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 void vti6_dev_free(struct net_device *dev)
178{
179 free_percpu(dev->tstats);
180}
181
182static int vti6_tnl_create2(struct net_device *dev)
183{
184 struct ip6_tnl *t = netdev_priv(dev);
185 struct net *net = dev_net(dev);
186 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
187 int err;
188
189 dev->rtnl_link_ops = &vti6_link_ops;
190 err = register_netdevice(dev);
191 if (err < 0)
192 goto out;
193
194 strcpy(t->parms.name, dev->name);
195
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 strscpy(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 netdev_put(dev, &t->dev_tracker);
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 DEV_STATS_INC(t->dev, 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 xfrm_state *x;
351 const struct xfrm_mode *inner_mode;
352 struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6;
353 u32 orig_mark = skb->mark;
354 int ret;
355
356 if (!t)
357 return 1;
358
359 dev = t->dev;
360
361 if (err) {
362 DEV_STATS_INC(dev, rx_errors);
363 DEV_STATS_INC(dev, rx_dropped);
364
365 return 0;
366 }
367
368 x = xfrm_input_state(skb);
369
370 inner_mode = &x->inner_mode;
371
372 if (x->sel.family == AF_UNSPEC) {
373 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
374 if (inner_mode == NULL) {
375 XFRM_INC_STATS(dev_net(skb->dev),
376 LINUX_MIB_XFRMINSTATEMODEERROR);
377 return -EINVAL;
378 }
379 }
380
381 family = inner_mode->family;
382
383 skb->mark = be32_to_cpu(t->parms.i_key);
384 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
385 skb->mark = orig_mark;
386
387 if (!ret)
388 return -EPERM;
389
390 skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev)));
391 skb->dev = dev;
392 dev_sw_netstats_rx_add(dev, skb->len);
393
394 return 0;
395}
396
397/**
398 * vti6_addr_conflict - compare packet addresses to tunnel's own
399 * @t: the outgoing tunnel device
400 * @hdr: IPv6 header from the incoming packet
401 *
402 * Description:
403 * Avoid trivial tunneling loop by checking that tunnel exit-point
404 * doesn't match source of incoming packet.
405 *
406 * Return:
407 * 1 if conflict,
408 * 0 else
409 **/
410static inline bool
411vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
412{
413 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
414}
415
416static bool vti6_state_check(const struct xfrm_state *x,
417 const struct in6_addr *dst,
418 const struct in6_addr *src)
419{
420 xfrm_address_t *daddr = (xfrm_address_t *)dst;
421 xfrm_address_t *saddr = (xfrm_address_t *)src;
422
423 /* if there is no transform then this tunnel is not functional.
424 * Or if the xfrm is not mode tunnel.
425 */
426 if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
427 x->props.family != AF_INET6)
428 return false;
429
430 if (ipv6_addr_any(dst))
431 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6);
432
433 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6))
434 return false;
435
436 return true;
437}
438
439/**
440 * vti6_xmit - send a packet
441 * @skb: the outgoing socket buffer
442 * @dev: the outgoing tunnel device
443 * @fl: the flow informations for the xfrm_lookup
444 **/
445static int
446vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
447{
448 struct ip6_tnl *t = netdev_priv(dev);
449 struct dst_entry *dst = skb_dst(skb);
450 struct net_device *tdev;
451 struct xfrm_state *x;
452 int pkt_len = skb->len;
453 int err = -1;
454 int mtu;
455
456 if (!dst) {
457 switch (skb->protocol) {
458 case htons(ETH_P_IP): {
459 struct rtable *rt;
460
461 fl->u.ip4.flowi4_oif = dev->ifindex;
462 fl->u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC;
463 rt = __ip_route_output_key(dev_net(dev), &fl->u.ip4);
464 if (IS_ERR(rt))
465 goto tx_err_link_failure;
466 dst = &rt->dst;
467 skb_dst_set(skb, dst);
468 break;
469 }
470 case htons(ETH_P_IPV6):
471 fl->u.ip6.flowi6_oif = dev->ifindex;
472 fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
473 dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6);
474 if (dst->error) {
475 dst_release(dst);
476 dst = NULL;
477 goto tx_err_link_failure;
478 }
479 skb_dst_set(skb, dst);
480 break;
481 default:
482 goto tx_err_link_failure;
483 }
484 }
485
486 dst_hold(dst);
487 dst = xfrm_lookup_route(t->net, dst, fl, NULL, 0);
488 if (IS_ERR(dst)) {
489 err = PTR_ERR(dst);
490 dst = NULL;
491 goto tx_err_link_failure;
492 }
493
494 if (dst->flags & DST_XFRM_QUEUE)
495 goto xmit;
496
497 x = dst->xfrm;
498 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr))
499 goto tx_err_link_failure;
500
501 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr,
502 (const struct in6_addr *)&x->id.daddr))
503 goto tx_err_link_failure;
504
505 tdev = dst->dev;
506
507 if (tdev == dev) {
508 DEV_STATS_INC(dev, collisions);
509 net_warn_ratelimited("%s: Local routing loop detected!\n",
510 t->parms.name);
511 goto tx_err_dst_release;
512 }
513
514 mtu = dst_mtu(dst);
515 if (skb->len > mtu) {
516 skb_dst_update_pmtu_no_confirm(skb, mtu);
517
518 if (skb->protocol == htons(ETH_P_IPV6)) {
519 if (mtu < IPV6_MIN_MTU)
520 mtu = IPV6_MIN_MTU;
521
522 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
523 } else {
524 if (!(ip_hdr(skb)->frag_off & htons(IP_DF)))
525 goto xmit;
526 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
527 htonl(mtu));
528 }
529
530 err = -EMSGSIZE;
531 goto tx_err_dst_release;
532 }
533
534xmit:
535 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
536 skb_dst_set(skb, dst);
537 skb->dev = skb_dst(skb)->dev;
538
539 err = dst_output(t->net, skb->sk, skb);
540 if (net_xmit_eval(err) == 0)
541 err = pkt_len;
542 iptunnel_xmit_stats(dev, err);
543
544 return 0;
545tx_err_link_failure:
546 DEV_STATS_INC(dev, tx_carrier_errors);
547 dst_link_failure(skb);
548tx_err_dst_release:
549 dst_release(dst);
550 return err;
551}
552
553static netdev_tx_t
554vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
555{
556 struct ip6_tnl *t = netdev_priv(dev);
557 struct flowi fl;
558 int ret;
559
560 if (!pskb_inet_may_pull(skb))
561 goto tx_err;
562
563 memset(&fl, 0, sizeof(fl));
564
565 switch (skb->protocol) {
566 case htons(ETH_P_IPV6):
567 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
568 vti6_addr_conflict(t, ipv6_hdr(skb)))
569 goto tx_err;
570
571 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
572 xfrm_decode_session(dev_net(dev), skb, &fl, AF_INET6);
573 break;
574 case htons(ETH_P_IP):
575 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
576 xfrm_decode_session(dev_net(dev), skb, &fl, AF_INET);
577 break;
578 default:
579 goto tx_err;
580 }
581
582 /* override mark with tunnel output key */
583 fl.flowi_mark = be32_to_cpu(t->parms.o_key);
584
585 ret = vti6_xmit(skb, dev, &fl);
586 if (ret < 0)
587 goto tx_err;
588
589 return NETDEV_TX_OK;
590
591tx_err:
592 DEV_STATS_INC(dev, tx_errors);
593 DEV_STATS_INC(dev, tx_dropped);
594 kfree_skb(skb);
595 return NETDEV_TX_OK;
596}
597
598static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
599 u8 type, u8 code, int offset, __be32 info)
600{
601 __be32 spi;
602 __u32 mark;
603 struct xfrm_state *x;
604 struct ip6_tnl *t;
605 struct ip_esp_hdr *esph;
606 struct ip_auth_hdr *ah;
607 struct ip_comp_hdr *ipch;
608 struct net *net = dev_net(skb->dev);
609 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
610 int protocol = iph->nexthdr;
611
612 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
613 if (!t)
614 return -1;
615
616 mark = be32_to_cpu(t->parms.o_key);
617
618 switch (protocol) {
619 case IPPROTO_ESP:
620 esph = (struct ip_esp_hdr *)(skb->data + offset);
621 spi = esph->spi;
622 break;
623 case IPPROTO_AH:
624 ah = (struct ip_auth_hdr *)(skb->data + offset);
625 spi = ah->spi;
626 break;
627 case IPPROTO_COMP:
628 ipch = (struct ip_comp_hdr *)(skb->data + offset);
629 spi = htonl(ntohs(ipch->cpi));
630 break;
631 default:
632 return 0;
633 }
634
635 if (type != ICMPV6_PKT_TOOBIG &&
636 type != NDISC_REDIRECT)
637 return 0;
638
639 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
640 spi, protocol, AF_INET6);
641 if (!x)
642 return 0;
643
644 if (type == NDISC_REDIRECT)
645 ip6_redirect(skb, net, skb->dev->ifindex, 0,
646 sock_net_uid(net, NULL));
647 else
648 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
649 xfrm_state_put(x);
650
651 return 0;
652}
653
654static void vti6_link_config(struct ip6_tnl *t, bool keep_mtu)
655{
656 struct net_device *dev = t->dev;
657 struct __ip6_tnl_parm *p = &t->parms;
658 struct net_device *tdev = NULL;
659 int mtu;
660
661 __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr));
662 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
663
664 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
665 IP6_TNL_F_CAP_PER_PACKET);
666 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
667
668 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
669 dev->flags |= IFF_POINTOPOINT;
670 else
671 dev->flags &= ~IFF_POINTOPOINT;
672
673 if (keep_mtu && dev->mtu) {
674 dev->mtu = clamp(dev->mtu, dev->min_mtu, dev->max_mtu);
675 return;
676 }
677
678 if (p->flags & IP6_TNL_F_CAP_XMIT) {
679 int strict = (ipv6_addr_type(&p->raddr) &
680 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
681 struct rt6_info *rt = rt6_lookup(t->net,
682 &p->raddr, &p->laddr,
683 p->link, NULL, strict);
684
685 if (rt)
686 tdev = rt->dst.dev;
687 ip6_rt_put(rt);
688 }
689
690 if (!tdev && p->link)
691 tdev = __dev_get_by_index(t->net, p->link);
692
693 if (tdev)
694 mtu = tdev->mtu - sizeof(struct ipv6hdr);
695 else
696 mtu = ETH_DATA_LEN - LL_MAX_HEADER - sizeof(struct ipv6hdr);
697
698 dev->mtu = max_t(int, mtu, IPV4_MIN_MTU);
699}
700
701/**
702 * vti6_tnl_change - update the tunnel parameters
703 * @t: tunnel to be changed
704 * @p: tunnel configuration parameters
705 * @keep_mtu: MTU was set from userspace, don't re-compute it
706 *
707 * Description:
708 * vti6_tnl_change() updates the tunnel parameters
709 **/
710static int
711vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
712 bool keep_mtu)
713{
714 t->parms.laddr = p->laddr;
715 t->parms.raddr = p->raddr;
716 t->parms.link = p->link;
717 t->parms.i_key = p->i_key;
718 t->parms.o_key = p->o_key;
719 t->parms.proto = p->proto;
720 t->parms.fwmark = p->fwmark;
721 dst_cache_reset(&t->dst_cache);
722 vti6_link_config(t, keep_mtu);
723 return 0;
724}
725
726static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p,
727 bool keep_mtu)
728{
729 struct net *net = dev_net(t->dev);
730 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
731 int err;
732
733 vti6_tnl_unlink(ip6n, t);
734 synchronize_net();
735 err = vti6_tnl_change(t, p, keep_mtu);
736 vti6_tnl_link(ip6n, t);
737 netdev_state_change(t->dev);
738 return err;
739}
740
741static void
742vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
743{
744 p->laddr = u->laddr;
745 p->raddr = u->raddr;
746 p->link = u->link;
747 p->i_key = u->i_key;
748 p->o_key = u->o_key;
749 p->proto = u->proto;
750
751 memcpy(p->name, u->name, sizeof(u->name));
752}
753
754static void
755vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
756{
757 u->laddr = p->laddr;
758 u->raddr = p->raddr;
759 u->link = p->link;
760 u->i_key = p->i_key;
761 u->o_key = p->o_key;
762 if (u->i_key)
763 u->i_flags |= GRE_KEY;
764 if (u->o_key)
765 u->o_flags |= GRE_KEY;
766 u->proto = p->proto;
767
768 memcpy(u->name, p->name, sizeof(u->name));
769}
770
771/**
772 * vti6_siocdevprivate - configure vti6 tunnels from userspace
773 * @dev: virtual device associated with tunnel
774 * @ifr: unused
775 * @data: parameters passed from userspace
776 * @cmd: command to be performed
777 *
778 * Description:
779 * vti6_siocdevprivate() is used for managing vti6 tunnels
780 * from userspace.
781 *
782 * The possible commands are the following:
783 * %SIOCGETTUNNEL: get tunnel parameters for device
784 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
785 * %SIOCCHGTUNNEL: change tunnel parameters to those given
786 * %SIOCDELTUNNEL: delete tunnel
787 *
788 * The fallback device "ip6_vti0", created during module
789 * initialization, can be used for creating other tunnel devices.
790 *
791 * Return:
792 * 0 on success,
793 * %-EFAULT if unable to copy data to or from userspace,
794 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
795 * %-EINVAL if passed tunnel parameters are invalid,
796 * %-EEXIST if changing a tunnel's parameters would cause a conflict
797 * %-ENODEV if attempting to change or delete a nonexisting device
798 **/
799static int
800vti6_siocdevprivate(struct net_device *dev, struct ifreq *ifr, void __user *data, int cmd)
801{
802 int err = 0;
803 struct ip6_tnl_parm2 p;
804 struct __ip6_tnl_parm p1;
805 struct ip6_tnl *t = NULL;
806 struct net *net = dev_net(dev);
807 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
808
809 memset(&p1, 0, sizeof(p1));
810
811 switch (cmd) {
812 case SIOCGETTUNNEL:
813 if (dev == ip6n->fb_tnl_dev) {
814 if (copy_from_user(&p, data, sizeof(p))) {
815 err = -EFAULT;
816 break;
817 }
818 vti6_parm_from_user(&p1, &p);
819 t = vti6_locate(net, &p1, 0);
820 } else {
821 memset(&p, 0, sizeof(p));
822 }
823 if (!t)
824 t = netdev_priv(dev);
825 vti6_parm_to_user(&p, &t->parms);
826 if (copy_to_user(data, &p, sizeof(p)))
827 err = -EFAULT;
828 break;
829 case SIOCADDTUNNEL:
830 case SIOCCHGTUNNEL:
831 err = -EPERM;
832 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
833 break;
834 err = -EFAULT;
835 if (copy_from_user(&p, data, sizeof(p)))
836 break;
837 err = -EINVAL;
838 if (p.proto != IPPROTO_IPV6 && p.proto != 0)
839 break;
840 vti6_parm_from_user(&p1, &p);
841 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
842 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
843 if (t) {
844 if (t->dev != dev) {
845 err = -EEXIST;
846 break;
847 }
848 } else
849 t = netdev_priv(dev);
850
851 err = vti6_update(t, &p1, false);
852 }
853 if (t) {
854 err = 0;
855 vti6_parm_to_user(&p, &t->parms);
856 if (copy_to_user(data, &p, sizeof(p)))
857 err = -EFAULT;
858
859 } else
860 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
861 break;
862 case SIOCDELTUNNEL:
863 err = -EPERM;
864 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
865 break;
866
867 if (dev == ip6n->fb_tnl_dev) {
868 err = -EFAULT;
869 if (copy_from_user(&p, data, sizeof(p)))
870 break;
871 err = -ENOENT;
872 vti6_parm_from_user(&p1, &p);
873 t = vti6_locate(net, &p1, 0);
874 if (!t)
875 break;
876 err = -EPERM;
877 if (t->dev == ip6n->fb_tnl_dev)
878 break;
879 dev = t->dev;
880 }
881 err = 0;
882 unregister_netdevice(dev);
883 break;
884 default:
885 err = -EINVAL;
886 }
887 return err;
888}
889
890static const struct net_device_ops vti6_netdev_ops = {
891 .ndo_init = vti6_dev_init,
892 .ndo_uninit = vti6_dev_uninit,
893 .ndo_start_xmit = vti6_tnl_xmit,
894 .ndo_siocdevprivate = vti6_siocdevprivate,
895 .ndo_get_stats64 = dev_get_tstats64,
896 .ndo_get_iflink = ip6_tnl_get_iflink,
897};
898
899/**
900 * vti6_dev_setup - setup virtual tunnel device
901 * @dev: virtual device associated with tunnel
902 *
903 * Description:
904 * Initialize function pointers and device parameters
905 **/
906static void vti6_dev_setup(struct net_device *dev)
907{
908 dev->netdev_ops = &vti6_netdev_ops;
909 dev->header_ops = &ip_tunnel_header_ops;
910 dev->needs_free_netdev = true;
911 dev->priv_destructor = vti6_dev_free;
912
913 dev->type = ARPHRD_TUNNEL6;
914 dev->min_mtu = IPV4_MIN_MTU;
915 dev->max_mtu = IP_MAX_MTU - sizeof(struct ipv6hdr);
916 dev->flags |= IFF_NOARP;
917 dev->addr_len = sizeof(struct in6_addr);
918 netif_keep_dst(dev);
919 /* This perm addr will be used as interface identifier by IPv6 */
920 dev->addr_assign_type = NET_ADDR_RANDOM;
921 eth_random_addr(dev->perm_addr);
922}
923
924/**
925 * vti6_dev_init_gen - general initializer for all tunnel devices
926 * @dev: virtual device associated with tunnel
927 **/
928static inline int vti6_dev_init_gen(struct net_device *dev)
929{
930 struct ip6_tnl *t = netdev_priv(dev);
931
932 t->dev = dev;
933 t->net = dev_net(dev);
934 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
935 if (!dev->tstats)
936 return -ENOMEM;
937 netdev_hold(dev, &t->dev_tracker, GFP_KERNEL);
938 return 0;
939}
940
941/**
942 * vti6_dev_init - initializer for all non fallback tunnel devices
943 * @dev: virtual device associated with tunnel
944 **/
945static int vti6_dev_init(struct net_device *dev)
946{
947 struct ip6_tnl *t = netdev_priv(dev);
948 int err = vti6_dev_init_gen(dev);
949
950 if (err)
951 return err;
952 vti6_link_config(t, true);
953 return 0;
954}
955
956/**
957 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
958 * @dev: fallback device
959 *
960 * Return: 0
961 **/
962static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
963{
964 struct ip6_tnl *t = netdev_priv(dev);
965 struct net *net = dev_net(dev);
966 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
967
968 t->parms.proto = IPPROTO_IPV6;
969
970 rcu_assign_pointer(ip6n->tnls_wc[0], t);
971 return 0;
972}
973
974static int vti6_validate(struct nlattr *tb[], struct nlattr *data[],
975 struct netlink_ext_ack *extack)
976{
977 return 0;
978}
979
980static void vti6_netlink_parms(struct nlattr *data[],
981 struct __ip6_tnl_parm *parms)
982{
983 memset(parms, 0, sizeof(*parms));
984
985 if (!data)
986 return;
987
988 if (data[IFLA_VTI_LINK])
989 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
990
991 if (data[IFLA_VTI_LOCAL])
992 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]);
993
994 if (data[IFLA_VTI_REMOTE])
995 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]);
996
997 if (data[IFLA_VTI_IKEY])
998 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
999
1000 if (data[IFLA_VTI_OKEY])
1001 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
1002
1003 if (data[IFLA_VTI_FWMARK])
1004 parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]);
1005}
1006
1007static int vti6_newlink(struct net *src_net, struct net_device *dev,
1008 struct nlattr *tb[], struct nlattr *data[],
1009 struct netlink_ext_ack *extack)
1010{
1011 struct net *net = dev_net(dev);
1012 struct ip6_tnl *nt;
1013
1014 nt = netdev_priv(dev);
1015 vti6_netlink_parms(data, &nt->parms);
1016
1017 nt->parms.proto = IPPROTO_IPV6;
1018
1019 if (vti6_locate(net, &nt->parms, 0))
1020 return -EEXIST;
1021
1022 return vti6_tnl_create2(dev);
1023}
1024
1025static void vti6_dellink(struct net_device *dev, struct list_head *head)
1026{
1027 struct net *net = dev_net(dev);
1028 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1029
1030 if (dev != ip6n->fb_tnl_dev)
1031 unregister_netdevice_queue(dev, head);
1032}
1033
1034static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
1035 struct nlattr *data[],
1036 struct netlink_ext_ack *extack)
1037{
1038 struct ip6_tnl *t;
1039 struct __ip6_tnl_parm p;
1040 struct net *net = dev_net(dev);
1041 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1042
1043 if (dev == ip6n->fb_tnl_dev)
1044 return -EINVAL;
1045
1046 vti6_netlink_parms(data, &p);
1047
1048 t = vti6_locate(net, &p, 0);
1049
1050 if (t) {
1051 if (t->dev != dev)
1052 return -EEXIST;
1053 } else
1054 t = netdev_priv(dev);
1055
1056 return vti6_update(t, &p, tb && tb[IFLA_MTU]);
1057}
1058
1059static size_t vti6_get_size(const struct net_device *dev)
1060{
1061 return
1062 /* IFLA_VTI_LINK */
1063 nla_total_size(4) +
1064 /* IFLA_VTI_LOCAL */
1065 nla_total_size(sizeof(struct in6_addr)) +
1066 /* IFLA_VTI_REMOTE */
1067 nla_total_size(sizeof(struct in6_addr)) +
1068 /* IFLA_VTI_IKEY */
1069 nla_total_size(4) +
1070 /* IFLA_VTI_OKEY */
1071 nla_total_size(4) +
1072 /* IFLA_VTI_FWMARK */
1073 nla_total_size(4) +
1074 0;
1075}
1076
1077static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1078{
1079 struct ip6_tnl *tunnel = netdev_priv(dev);
1080 struct __ip6_tnl_parm *parm = &tunnel->parms;
1081
1082 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
1083 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) ||
1084 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) ||
1085 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
1086 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) ||
1087 nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark))
1088 goto nla_put_failure;
1089 return 0;
1090
1091nla_put_failure:
1092 return -EMSGSIZE;
1093}
1094
1095static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1096 [IFLA_VTI_LINK] = { .type = NLA_U32 },
1097 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) },
1098 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) },
1099 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
1100 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
1101 [IFLA_VTI_FWMARK] = { .type = NLA_U32 },
1102};
1103
1104static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1105 .kind = "vti6",
1106 .maxtype = IFLA_VTI_MAX,
1107 .policy = vti6_policy,
1108 .priv_size = sizeof(struct ip6_tnl),
1109 .setup = vti6_dev_setup,
1110 .validate = vti6_validate,
1111 .newlink = vti6_newlink,
1112 .dellink = vti6_dellink,
1113 .changelink = vti6_changelink,
1114 .get_size = vti6_get_size,
1115 .fill_info = vti6_fill_info,
1116 .get_link_net = ip6_tnl_get_link_net,
1117};
1118
1119static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n,
1120 struct list_head *list)
1121{
1122 int h;
1123 struct ip6_tnl *t;
1124
1125 for (h = 0; h < IP6_VTI_HASH_SIZE; h++) {
1126 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1127 while (t) {
1128 unregister_netdevice_queue(t->dev, list);
1129 t = rtnl_dereference(t->next);
1130 }
1131 }
1132
1133 t = rtnl_dereference(ip6n->tnls_wc[0]);
1134 if (t)
1135 unregister_netdevice_queue(t->dev, list);
1136}
1137
1138static int __net_init vti6_init_net(struct net *net)
1139{
1140 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1141 struct ip6_tnl *t = NULL;
1142 int err;
1143
1144 ip6n->tnls[0] = ip6n->tnls_wc;
1145 ip6n->tnls[1] = ip6n->tnls_r_l;
1146
1147 if (!net_has_fallback_tunnels(net))
1148 return 0;
1149 err = -ENOMEM;
1150 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1151 NET_NAME_UNKNOWN, vti6_dev_setup);
1152
1153 if (!ip6n->fb_tnl_dev)
1154 goto err_alloc_dev;
1155 dev_net_set(ip6n->fb_tnl_dev, net);
1156 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1157
1158 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1159 if (err < 0)
1160 goto err_register;
1161
1162 err = register_netdev(ip6n->fb_tnl_dev);
1163 if (err < 0)
1164 goto err_register;
1165
1166 t = netdev_priv(ip6n->fb_tnl_dev);
1167
1168 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1169 return 0;
1170
1171err_register:
1172 free_netdev(ip6n->fb_tnl_dev);
1173err_alloc_dev:
1174 return err;
1175}
1176
1177static void __net_exit vti6_exit_batch_net(struct list_head *net_list)
1178{
1179 struct vti6_net *ip6n;
1180 struct net *net;
1181 LIST_HEAD(list);
1182
1183 rtnl_lock();
1184 list_for_each_entry(net, net_list, exit_list) {
1185 ip6n = net_generic(net, vti6_net_id);
1186 vti6_destroy_tunnels(ip6n, &list);
1187 }
1188 unregister_netdevice_many(&list);
1189 rtnl_unlock();
1190}
1191
1192static struct pernet_operations vti6_net_ops = {
1193 .init = vti6_init_net,
1194 .exit_batch = vti6_exit_batch_net,
1195 .id = &vti6_net_id,
1196 .size = sizeof(struct vti6_net),
1197};
1198
1199static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1200 .handler = vti6_rcv,
1201 .input_handler = vti6_input_proto,
1202 .cb_handler = vti6_rcv_cb,
1203 .err_handler = vti6_err,
1204 .priority = 100,
1205};
1206
1207static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1208 .handler = vti6_rcv,
1209 .input_handler = vti6_input_proto,
1210 .cb_handler = vti6_rcv_cb,
1211 .err_handler = vti6_err,
1212 .priority = 100,
1213};
1214
1215static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1216 .handler = vti6_rcv,
1217 .input_handler = vti6_input_proto,
1218 .cb_handler = vti6_rcv_cb,
1219 .err_handler = vti6_err,
1220 .priority = 100,
1221};
1222
1223#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1224static int vti6_rcv_tunnel(struct sk_buff *skb)
1225{
1226 const xfrm_address_t *saddr;
1227 __be32 spi;
1228
1229 saddr = (const xfrm_address_t *)&ipv6_hdr(skb)->saddr;
1230 spi = xfrm6_tunnel_spi_lookup(dev_net(skb->dev), saddr);
1231
1232 return vti6_input_proto(skb, IPPROTO_IPV6, spi, 0);
1233}
1234
1235static struct xfrm6_tunnel vti_ipv6_handler __read_mostly = {
1236 .handler = vti6_rcv_tunnel,
1237 .cb_handler = vti6_rcv_cb,
1238 .err_handler = vti6_err,
1239 .priority = 0,
1240};
1241
1242static struct xfrm6_tunnel vti_ip6ip_handler __read_mostly = {
1243 .handler = vti6_rcv_tunnel,
1244 .cb_handler = vti6_rcv_cb,
1245 .err_handler = vti6_err,
1246 .priority = 0,
1247};
1248#endif
1249
1250/**
1251 * vti6_tunnel_init - register protocol and reserve needed resources
1252 *
1253 * Return: 0 on success
1254 **/
1255static int __init vti6_tunnel_init(void)
1256{
1257 const char *msg;
1258 int err;
1259
1260 msg = "tunnel device";
1261 err = register_pernet_device(&vti6_net_ops);
1262 if (err < 0)
1263 goto pernet_dev_failed;
1264
1265 msg = "tunnel protocols";
1266 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1267 if (err < 0)
1268 goto xfrm_proto_esp_failed;
1269 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1270 if (err < 0)
1271 goto xfrm_proto_ah_failed;
1272 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1273 if (err < 0)
1274 goto xfrm_proto_comp_failed;
1275#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1276 msg = "ipv6 tunnel";
1277 err = xfrm6_tunnel_register(&vti_ipv6_handler, AF_INET6);
1278 if (err < 0)
1279 goto vti_tunnel_ipv6_failed;
1280 err = xfrm6_tunnel_register(&vti_ip6ip_handler, AF_INET);
1281 if (err < 0)
1282 goto vti_tunnel_ip6ip_failed;
1283#endif
1284
1285 msg = "netlink interface";
1286 err = rtnl_link_register(&vti6_link_ops);
1287 if (err < 0)
1288 goto rtnl_link_failed;
1289
1290 return 0;
1291
1292rtnl_link_failed:
1293#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1294 err = xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET);
1295vti_tunnel_ip6ip_failed:
1296 err = xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6);
1297vti_tunnel_ipv6_failed:
1298#endif
1299 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1300xfrm_proto_comp_failed:
1301 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1302xfrm_proto_ah_failed:
1303 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1304xfrm_proto_esp_failed:
1305 unregister_pernet_device(&vti6_net_ops);
1306pernet_dev_failed:
1307 pr_err("vti6 init: failed to register %s\n", msg);
1308 return err;
1309}
1310
1311/**
1312 * vti6_tunnel_cleanup - free resources and unregister protocol
1313 **/
1314static void __exit vti6_tunnel_cleanup(void)
1315{
1316 rtnl_link_unregister(&vti6_link_ops);
1317#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1318 xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET);
1319 xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6);
1320#endif
1321 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1322 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1323 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1324 unregister_pernet_device(&vti6_net_ops);
1325}
1326
1327module_init(vti6_tunnel_init);
1328module_exit(vti6_tunnel_cleanup);
1329MODULE_LICENSE("GPL");
1330MODULE_ALIAS_RTNL_LINK("vti6");
1331MODULE_ALIAS_NETDEV("ip6_vti0");
1332MODULE_AUTHOR("Steffen Klassert");
1333MODULE_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 * @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_rcv(struct sk_buff *skb)
300{
301 struct ip6_tnl *t;
302 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
303
304 rcu_read_lock();
305 t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
306 if (t) {
307 if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) {
308 rcu_read_unlock();
309 goto discard;
310 }
311
312 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
313 rcu_read_unlock();
314 return 0;
315 }
316
317 ipv6h = ipv6_hdr(skb);
318 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
319 t->dev->stats.rx_dropped++;
320 rcu_read_unlock();
321 goto discard;
322 }
323
324 rcu_read_unlock();
325
326 return xfrm6_rcv_tnl(skb, t);
327 }
328 rcu_read_unlock();
329 return -EINVAL;
330discard:
331 kfree_skb(skb);
332 return 0;
333}
334
335static int vti6_rcv_cb(struct sk_buff *skb, int err)
336{
337 unsigned short family;
338 struct net_device *dev;
339 struct pcpu_sw_netstats *tstats;
340 struct xfrm_state *x;
341 const struct xfrm_mode *inner_mode;
342 struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6;
343 u32 orig_mark = skb->mark;
344 int ret;
345
346 if (!t)
347 return 1;
348
349 dev = t->dev;
350
351 if (err) {
352 dev->stats.rx_errors++;
353 dev->stats.rx_dropped++;
354
355 return 0;
356 }
357
358 x = xfrm_input_state(skb);
359
360 inner_mode = &x->inner_mode;
361
362 if (x->sel.family == AF_UNSPEC) {
363 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
364 if (inner_mode == NULL) {
365 XFRM_INC_STATS(dev_net(skb->dev),
366 LINUX_MIB_XFRMINSTATEMODEERROR);
367 return -EINVAL;
368 }
369 }
370
371 family = inner_mode->family;
372
373 skb->mark = be32_to_cpu(t->parms.i_key);
374 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
375 skb->mark = orig_mark;
376
377 if (!ret)
378 return -EPERM;
379
380 skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev)));
381 skb->dev = dev;
382
383 tstats = this_cpu_ptr(dev->tstats);
384 u64_stats_update_begin(&tstats->syncp);
385 tstats->rx_packets++;
386 tstats->rx_bytes += skb->len;
387 u64_stats_update_end(&tstats->syncp);
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 net_device_stats *stats = &t->dev->stats;
445 struct dst_entry *dst = skb_dst(skb);
446 struct net_device *tdev;
447 struct xfrm_state *x;
448 int pkt_len = skb->len;
449 int err = -1;
450 int mtu;
451
452 if (!dst)
453 goto tx_err_link_failure;
454
455 dst_hold(dst);
456 dst = xfrm_lookup(t->net, dst, fl, NULL, 0);
457 if (IS_ERR(dst)) {
458 err = PTR_ERR(dst);
459 dst = NULL;
460 goto tx_err_link_failure;
461 }
462
463 x = dst->xfrm;
464 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr))
465 goto tx_err_link_failure;
466
467 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr,
468 (const struct in6_addr *)&x->id.daddr))
469 goto tx_err_link_failure;
470
471 tdev = dst->dev;
472
473 if (tdev == dev) {
474 stats->collisions++;
475 net_warn_ratelimited("%s: Local routing loop detected!\n",
476 t->parms.name);
477 goto tx_err_dst_release;
478 }
479
480 mtu = dst_mtu(dst);
481 if (skb->len > mtu) {
482 skb_dst_update_pmtu(skb, mtu);
483
484 if (skb->protocol == htons(ETH_P_IPV6)) {
485 if (mtu < IPV6_MIN_MTU)
486 mtu = IPV6_MIN_MTU;
487
488 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
489 } else {
490 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
491 htonl(mtu));
492 }
493
494 err = -EMSGSIZE;
495 goto tx_err_dst_release;
496 }
497
498 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
499 skb_dst_set(skb, dst);
500 skb->dev = skb_dst(skb)->dev;
501
502 err = dst_output(t->net, skb->sk, skb);
503 if (net_xmit_eval(err) == 0)
504 err = pkt_len;
505 iptunnel_xmit_stats(dev, err);
506
507 return 0;
508tx_err_link_failure:
509 stats->tx_carrier_errors++;
510 dst_link_failure(skb);
511tx_err_dst_release:
512 dst_release(dst);
513 return err;
514}
515
516static netdev_tx_t
517vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
518{
519 struct ip6_tnl *t = netdev_priv(dev);
520 struct net_device_stats *stats = &t->dev->stats;
521 struct flowi fl;
522 int ret;
523
524 if (!pskb_inet_may_pull(skb))
525 goto tx_err;
526
527 memset(&fl, 0, sizeof(fl));
528
529 switch (skb->protocol) {
530 case htons(ETH_P_IPV6):
531 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
532 vti6_addr_conflict(t, ipv6_hdr(skb)))
533 goto tx_err;
534
535 xfrm_decode_session(skb, &fl, AF_INET6);
536 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
537 break;
538 case htons(ETH_P_IP):
539 xfrm_decode_session(skb, &fl, AF_INET);
540 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
541 break;
542 default:
543 goto tx_err;
544 }
545
546 /* override mark with tunnel output key */
547 fl.flowi_mark = be32_to_cpu(t->parms.o_key);
548
549 ret = vti6_xmit(skb, dev, &fl);
550 if (ret < 0)
551 goto tx_err;
552
553 return NETDEV_TX_OK;
554
555tx_err:
556 stats->tx_errors++;
557 stats->tx_dropped++;
558 kfree_skb(skb);
559 return NETDEV_TX_OK;
560}
561
562static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
563 u8 type, u8 code, int offset, __be32 info)
564{
565 __be32 spi;
566 __u32 mark;
567 struct xfrm_state *x;
568 struct ip6_tnl *t;
569 struct ip_esp_hdr *esph;
570 struct ip_auth_hdr *ah;
571 struct ip_comp_hdr *ipch;
572 struct net *net = dev_net(skb->dev);
573 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
574 int protocol = iph->nexthdr;
575
576 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
577 if (!t)
578 return -1;
579
580 mark = be32_to_cpu(t->parms.o_key);
581
582 switch (protocol) {
583 case IPPROTO_ESP:
584 esph = (struct ip_esp_hdr *)(skb->data + offset);
585 spi = esph->spi;
586 break;
587 case IPPROTO_AH:
588 ah = (struct ip_auth_hdr *)(skb->data + offset);
589 spi = ah->spi;
590 break;
591 case IPPROTO_COMP:
592 ipch = (struct ip_comp_hdr *)(skb->data + offset);
593 spi = htonl(ntohs(ipch->cpi));
594 break;
595 default:
596 return 0;
597 }
598
599 if (type != ICMPV6_PKT_TOOBIG &&
600 type != NDISC_REDIRECT)
601 return 0;
602
603 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
604 spi, protocol, AF_INET6);
605 if (!x)
606 return 0;
607
608 if (type == NDISC_REDIRECT)
609 ip6_redirect(skb, net, skb->dev->ifindex, 0,
610 sock_net_uid(net, NULL));
611 else
612 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
613 xfrm_state_put(x);
614
615 return 0;
616}
617
618static void vti6_link_config(struct ip6_tnl *t, bool keep_mtu)
619{
620 struct net_device *dev = t->dev;
621 struct __ip6_tnl_parm *p = &t->parms;
622 struct net_device *tdev = NULL;
623 int mtu;
624
625 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
626 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
627
628 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
629 IP6_TNL_F_CAP_PER_PACKET);
630 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
631
632 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
633 dev->flags |= IFF_POINTOPOINT;
634 else
635 dev->flags &= ~IFF_POINTOPOINT;
636
637 if (keep_mtu && dev->mtu) {
638 dev->mtu = clamp(dev->mtu, dev->min_mtu, dev->max_mtu);
639 return;
640 }
641
642 if (p->flags & IP6_TNL_F_CAP_XMIT) {
643 int strict = (ipv6_addr_type(&p->raddr) &
644 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
645 struct rt6_info *rt = rt6_lookup(t->net,
646 &p->raddr, &p->laddr,
647 p->link, NULL, strict);
648
649 if (rt)
650 tdev = rt->dst.dev;
651 ip6_rt_put(rt);
652 }
653
654 if (!tdev && p->link)
655 tdev = __dev_get_by_index(t->net, p->link);
656
657 if (tdev)
658 mtu = tdev->mtu - sizeof(struct ipv6hdr);
659 else
660 mtu = ETH_DATA_LEN - LL_MAX_HEADER - sizeof(struct ipv6hdr);
661
662 dev->mtu = max_t(int, mtu, IPV4_MIN_MTU);
663}
664
665/**
666 * vti6_tnl_change - update the tunnel parameters
667 * @t: tunnel to be changed
668 * @p: tunnel configuration parameters
669 * @keep_mtu: MTU was set from userspace, don't re-compute it
670 *
671 * Description:
672 * vti6_tnl_change() updates the tunnel parameters
673 **/
674static int
675vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
676 bool keep_mtu)
677{
678 t->parms.laddr = p->laddr;
679 t->parms.raddr = p->raddr;
680 t->parms.link = p->link;
681 t->parms.i_key = p->i_key;
682 t->parms.o_key = p->o_key;
683 t->parms.proto = p->proto;
684 t->parms.fwmark = p->fwmark;
685 dst_cache_reset(&t->dst_cache);
686 vti6_link_config(t, keep_mtu);
687 return 0;
688}
689
690static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p,
691 bool keep_mtu)
692{
693 struct net *net = dev_net(t->dev);
694 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
695 int err;
696
697 vti6_tnl_unlink(ip6n, t);
698 synchronize_net();
699 err = vti6_tnl_change(t, p, keep_mtu);
700 vti6_tnl_link(ip6n, t);
701 netdev_state_change(t->dev);
702 return err;
703}
704
705static void
706vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
707{
708 p->laddr = u->laddr;
709 p->raddr = u->raddr;
710 p->link = u->link;
711 p->i_key = u->i_key;
712 p->o_key = u->o_key;
713 p->proto = u->proto;
714
715 memcpy(p->name, u->name, sizeof(u->name));
716}
717
718static void
719vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
720{
721 u->laddr = p->laddr;
722 u->raddr = p->raddr;
723 u->link = p->link;
724 u->i_key = p->i_key;
725 u->o_key = p->o_key;
726 if (u->i_key)
727 u->i_flags |= GRE_KEY;
728 if (u->o_key)
729 u->o_flags |= GRE_KEY;
730 u->proto = p->proto;
731
732 memcpy(u->name, p->name, sizeof(u->name));
733}
734
735/**
736 * vti6_ioctl - configure vti6 tunnels from userspace
737 * @dev: virtual device associated with tunnel
738 * @ifr: parameters passed from userspace
739 * @cmd: command to be performed
740 *
741 * Description:
742 * vti6_ioctl() is used for managing vti6 tunnels
743 * from userspace.
744 *
745 * The possible commands are the following:
746 * %SIOCGETTUNNEL: get tunnel parameters for device
747 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
748 * %SIOCCHGTUNNEL: change tunnel parameters to those given
749 * %SIOCDELTUNNEL: delete tunnel
750 *
751 * The fallback device "ip6_vti0", created during module
752 * initialization, can be used for creating other tunnel devices.
753 *
754 * Return:
755 * 0 on success,
756 * %-EFAULT if unable to copy data to or from userspace,
757 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
758 * %-EINVAL if passed tunnel parameters are invalid,
759 * %-EEXIST if changing a tunnel's parameters would cause a conflict
760 * %-ENODEV if attempting to change or delete a nonexisting device
761 **/
762static int
763vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
764{
765 int err = 0;
766 struct ip6_tnl_parm2 p;
767 struct __ip6_tnl_parm p1;
768 struct ip6_tnl *t = NULL;
769 struct net *net = dev_net(dev);
770 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
771
772 switch (cmd) {
773 case SIOCGETTUNNEL:
774 if (dev == ip6n->fb_tnl_dev) {
775 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
776 err = -EFAULT;
777 break;
778 }
779 vti6_parm_from_user(&p1, &p);
780 t = vti6_locate(net, &p1, 0);
781 } else {
782 memset(&p, 0, sizeof(p));
783 }
784 if (!t)
785 t = netdev_priv(dev);
786 vti6_parm_to_user(&p, &t->parms);
787 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
788 err = -EFAULT;
789 break;
790 case SIOCADDTUNNEL:
791 case SIOCCHGTUNNEL:
792 err = -EPERM;
793 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
794 break;
795 err = -EFAULT;
796 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
797 break;
798 err = -EINVAL;
799 if (p.proto != IPPROTO_IPV6 && p.proto != 0)
800 break;
801 vti6_parm_from_user(&p1, &p);
802 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
803 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
804 if (t) {
805 if (t->dev != dev) {
806 err = -EEXIST;
807 break;
808 }
809 } else
810 t = netdev_priv(dev);
811
812 err = vti6_update(t, &p1, false);
813 }
814 if (t) {
815 err = 0;
816 vti6_parm_to_user(&p, &t->parms);
817 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
818 err = -EFAULT;
819
820 } else
821 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
822 break;
823 case SIOCDELTUNNEL:
824 err = -EPERM;
825 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
826 break;
827
828 if (dev == ip6n->fb_tnl_dev) {
829 err = -EFAULT;
830 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
831 break;
832 err = -ENOENT;
833 vti6_parm_from_user(&p1, &p);
834 t = vti6_locate(net, &p1, 0);
835 if (!t)
836 break;
837 err = -EPERM;
838 if (t->dev == ip6n->fb_tnl_dev)
839 break;
840 dev = t->dev;
841 }
842 err = 0;
843 unregister_netdevice(dev);
844 break;
845 default:
846 err = -EINVAL;
847 }
848 return err;
849}
850
851static const struct net_device_ops vti6_netdev_ops = {
852 .ndo_init = vti6_dev_init,
853 .ndo_uninit = vti6_dev_uninit,
854 .ndo_start_xmit = vti6_tnl_xmit,
855 .ndo_do_ioctl = vti6_ioctl,
856 .ndo_get_stats64 = ip_tunnel_get_stats64,
857 .ndo_get_iflink = ip6_tnl_get_iflink,
858};
859
860/**
861 * vti6_dev_setup - setup virtual tunnel device
862 * @dev: virtual device associated with tunnel
863 *
864 * Description:
865 * Initialize function pointers and device parameters
866 **/
867static void vti6_dev_setup(struct net_device *dev)
868{
869 dev->netdev_ops = &vti6_netdev_ops;
870 dev->needs_free_netdev = true;
871 dev->priv_destructor = vti6_dev_free;
872
873 dev->type = ARPHRD_TUNNEL6;
874 dev->min_mtu = IPV4_MIN_MTU;
875 dev->max_mtu = IP_MAX_MTU - sizeof(struct ipv6hdr);
876 dev->flags |= IFF_NOARP;
877 dev->addr_len = sizeof(struct in6_addr);
878 netif_keep_dst(dev);
879 /* This perm addr will be used as interface identifier by IPv6 */
880 dev->addr_assign_type = NET_ADDR_RANDOM;
881 eth_random_addr(dev->perm_addr);
882}
883
884/**
885 * vti6_dev_init_gen - general initializer for all tunnel devices
886 * @dev: virtual device associated with tunnel
887 **/
888static inline int vti6_dev_init_gen(struct net_device *dev)
889{
890 struct ip6_tnl *t = netdev_priv(dev);
891
892 t->dev = dev;
893 t->net = dev_net(dev);
894 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
895 if (!dev->tstats)
896 return -ENOMEM;
897 return 0;
898}
899
900/**
901 * vti6_dev_init - initializer for all non fallback tunnel devices
902 * @dev: virtual device associated with tunnel
903 **/
904static int vti6_dev_init(struct net_device *dev)
905{
906 struct ip6_tnl *t = netdev_priv(dev);
907 int err = vti6_dev_init_gen(dev);
908
909 if (err)
910 return err;
911 vti6_link_config(t, true);
912 return 0;
913}
914
915/**
916 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
917 * @dev: fallback device
918 *
919 * Return: 0
920 **/
921static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
922{
923 struct ip6_tnl *t = netdev_priv(dev);
924 struct net *net = dev_net(dev);
925 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
926
927 t->parms.proto = IPPROTO_IPV6;
928 dev_hold(dev);
929
930 rcu_assign_pointer(ip6n->tnls_wc[0], t);
931 return 0;
932}
933
934static int vti6_validate(struct nlattr *tb[], struct nlattr *data[],
935 struct netlink_ext_ack *extack)
936{
937 return 0;
938}
939
940static void vti6_netlink_parms(struct nlattr *data[],
941 struct __ip6_tnl_parm *parms)
942{
943 memset(parms, 0, sizeof(*parms));
944
945 if (!data)
946 return;
947
948 if (data[IFLA_VTI_LINK])
949 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
950
951 if (data[IFLA_VTI_LOCAL])
952 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]);
953
954 if (data[IFLA_VTI_REMOTE])
955 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]);
956
957 if (data[IFLA_VTI_IKEY])
958 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
959
960 if (data[IFLA_VTI_OKEY])
961 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
962
963 if (data[IFLA_VTI_FWMARK])
964 parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]);
965}
966
967static int vti6_newlink(struct net *src_net, struct net_device *dev,
968 struct nlattr *tb[], struct nlattr *data[],
969 struct netlink_ext_ack *extack)
970{
971 struct net *net = dev_net(dev);
972 struct ip6_tnl *nt;
973
974 nt = netdev_priv(dev);
975 vti6_netlink_parms(data, &nt->parms);
976
977 nt->parms.proto = IPPROTO_IPV6;
978
979 if (vti6_locate(net, &nt->parms, 0))
980 return -EEXIST;
981
982 return vti6_tnl_create2(dev);
983}
984
985static void vti6_dellink(struct net_device *dev, struct list_head *head)
986{
987 struct net *net = dev_net(dev);
988 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
989
990 if (dev != ip6n->fb_tnl_dev)
991 unregister_netdevice_queue(dev, head);
992}
993
994static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
995 struct nlattr *data[],
996 struct netlink_ext_ack *extack)
997{
998 struct ip6_tnl *t;
999 struct __ip6_tnl_parm p;
1000 struct net *net = dev_net(dev);
1001 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1002
1003 if (dev == ip6n->fb_tnl_dev)
1004 return -EINVAL;
1005
1006 vti6_netlink_parms(data, &p);
1007
1008 t = vti6_locate(net, &p, 0);
1009
1010 if (t) {
1011 if (t->dev != dev)
1012 return -EEXIST;
1013 } else
1014 t = netdev_priv(dev);
1015
1016 return vti6_update(t, &p, tb && tb[IFLA_MTU]);
1017}
1018
1019static size_t vti6_get_size(const struct net_device *dev)
1020{
1021 return
1022 /* IFLA_VTI_LINK */
1023 nla_total_size(4) +
1024 /* IFLA_VTI_LOCAL */
1025 nla_total_size(sizeof(struct in6_addr)) +
1026 /* IFLA_VTI_REMOTE */
1027 nla_total_size(sizeof(struct in6_addr)) +
1028 /* IFLA_VTI_IKEY */
1029 nla_total_size(4) +
1030 /* IFLA_VTI_OKEY */
1031 nla_total_size(4) +
1032 /* IFLA_VTI_FWMARK */
1033 nla_total_size(4) +
1034 0;
1035}
1036
1037static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1038{
1039 struct ip6_tnl *tunnel = netdev_priv(dev);
1040 struct __ip6_tnl_parm *parm = &tunnel->parms;
1041
1042 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
1043 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) ||
1044 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) ||
1045 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
1046 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) ||
1047 nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark))
1048 goto nla_put_failure;
1049 return 0;
1050
1051nla_put_failure:
1052 return -EMSGSIZE;
1053}
1054
1055static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1056 [IFLA_VTI_LINK] = { .type = NLA_U32 },
1057 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) },
1058 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) },
1059 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
1060 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
1061 [IFLA_VTI_FWMARK] = { .type = NLA_U32 },
1062};
1063
1064static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1065 .kind = "vti6",
1066 .maxtype = IFLA_VTI_MAX,
1067 .policy = vti6_policy,
1068 .priv_size = sizeof(struct ip6_tnl),
1069 .setup = vti6_dev_setup,
1070 .validate = vti6_validate,
1071 .newlink = vti6_newlink,
1072 .dellink = vti6_dellink,
1073 .changelink = vti6_changelink,
1074 .get_size = vti6_get_size,
1075 .fill_info = vti6_fill_info,
1076 .get_link_net = ip6_tnl_get_link_net,
1077};
1078
1079static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n,
1080 struct list_head *list)
1081{
1082 int h;
1083 struct ip6_tnl *t;
1084
1085 for (h = 0; h < IP6_VTI_HASH_SIZE; h++) {
1086 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1087 while (t) {
1088 unregister_netdevice_queue(t->dev, list);
1089 t = rtnl_dereference(t->next);
1090 }
1091 }
1092
1093 t = rtnl_dereference(ip6n->tnls_wc[0]);
1094 if (t)
1095 unregister_netdevice_queue(t->dev, list);
1096}
1097
1098static int __net_init vti6_init_net(struct net *net)
1099{
1100 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1101 struct ip6_tnl *t = NULL;
1102 int err;
1103
1104 ip6n->tnls[0] = ip6n->tnls_wc;
1105 ip6n->tnls[1] = ip6n->tnls_r_l;
1106
1107 if (!net_has_fallback_tunnels(net))
1108 return 0;
1109 err = -ENOMEM;
1110 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1111 NET_NAME_UNKNOWN, vti6_dev_setup);
1112
1113 if (!ip6n->fb_tnl_dev)
1114 goto err_alloc_dev;
1115 dev_net_set(ip6n->fb_tnl_dev, net);
1116 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1117
1118 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1119 if (err < 0)
1120 goto err_register;
1121
1122 err = register_netdev(ip6n->fb_tnl_dev);
1123 if (err < 0)
1124 goto err_register;
1125
1126 t = netdev_priv(ip6n->fb_tnl_dev);
1127
1128 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1129 return 0;
1130
1131err_register:
1132 free_netdev(ip6n->fb_tnl_dev);
1133err_alloc_dev:
1134 return err;
1135}
1136
1137static void __net_exit vti6_exit_batch_net(struct list_head *net_list)
1138{
1139 struct vti6_net *ip6n;
1140 struct net *net;
1141 LIST_HEAD(list);
1142
1143 rtnl_lock();
1144 list_for_each_entry(net, net_list, exit_list) {
1145 ip6n = net_generic(net, vti6_net_id);
1146 vti6_destroy_tunnels(ip6n, &list);
1147 }
1148 unregister_netdevice_many(&list);
1149 rtnl_unlock();
1150}
1151
1152static struct pernet_operations vti6_net_ops = {
1153 .init = vti6_init_net,
1154 .exit_batch = vti6_exit_batch_net,
1155 .id = &vti6_net_id,
1156 .size = sizeof(struct vti6_net),
1157};
1158
1159static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1160 .handler = vti6_rcv,
1161 .cb_handler = vti6_rcv_cb,
1162 .err_handler = vti6_err,
1163 .priority = 100,
1164};
1165
1166static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1167 .handler = vti6_rcv,
1168 .cb_handler = vti6_rcv_cb,
1169 .err_handler = vti6_err,
1170 .priority = 100,
1171};
1172
1173static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1174 .handler = vti6_rcv,
1175 .cb_handler = vti6_rcv_cb,
1176 .err_handler = vti6_err,
1177 .priority = 100,
1178};
1179
1180/**
1181 * vti6_tunnel_init - register protocol and reserve needed resources
1182 *
1183 * Return: 0 on success
1184 **/
1185static int __init vti6_tunnel_init(void)
1186{
1187 const char *msg;
1188 int err;
1189
1190 msg = "tunnel device";
1191 err = register_pernet_device(&vti6_net_ops);
1192 if (err < 0)
1193 goto pernet_dev_failed;
1194
1195 msg = "tunnel protocols";
1196 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1197 if (err < 0)
1198 goto xfrm_proto_esp_failed;
1199 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1200 if (err < 0)
1201 goto xfrm_proto_ah_failed;
1202 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1203 if (err < 0)
1204 goto xfrm_proto_comp_failed;
1205
1206 msg = "netlink interface";
1207 err = rtnl_link_register(&vti6_link_ops);
1208 if (err < 0)
1209 goto rtnl_link_failed;
1210
1211 return 0;
1212
1213rtnl_link_failed:
1214 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1215xfrm_proto_comp_failed:
1216 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1217xfrm_proto_ah_failed:
1218 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1219xfrm_proto_esp_failed:
1220 unregister_pernet_device(&vti6_net_ops);
1221pernet_dev_failed:
1222 pr_err("vti6 init: failed to register %s\n", msg);
1223 return err;
1224}
1225
1226/**
1227 * vti6_tunnel_cleanup - free resources and unregister protocol
1228 **/
1229static void __exit vti6_tunnel_cleanup(void)
1230{
1231 rtnl_link_unregister(&vti6_link_ops);
1232 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1233 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1234 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1235 unregister_pernet_device(&vti6_net_ops);
1236}
1237
1238module_init(vti6_tunnel_init);
1239module_exit(vti6_tunnel_cleanup);
1240MODULE_LICENSE("GPL");
1241MODULE_ALIAS_RTNL_LINK("vti6");
1242MODULE_ALIAS_NETDEV("ip6_vti0");
1243MODULE_AUTHOR("Steffen Klassert");
1244MODULE_DESCRIPTION("IPv6 virtual tunnel interface");