Commit | Line | Data |
---|---|---|
ed1efb2a SK |
1 | /* |
2 | * IPv6 virtual tunneling interface | |
3 | * | |
4 | * Copyright (C) 2013 secunet Security Networks AG | |
5 | * | |
6 | * Author: | |
7 | * Steffen Klassert <steffen.klassert@secunet.com> | |
8 | * | |
9 | * Based on: | |
10 | * net/ipv6/ip6_tunnel.c | |
11 | * | |
12 | * This program is free software; you can redistribute it and/or | |
13 | * modify it under the terms of the GNU General Public License | |
14 | * as published by the Free Software Foundation; either version | |
15 | * 2 of the License, or (at your option) any later version. | |
16 | */ | |
17 | ||
18 | #include <linux/module.h> | |
19 | #include <linux/capability.h> | |
20 | #include <linux/errno.h> | |
21 | #include <linux/types.h> | |
22 | #include <linux/sockios.h> | |
23 | #include <linux/icmp.h> | |
24 | #include <linux/if.h> | |
25 | #include <linux/in.h> | |
26 | #include <linux/ip.h> | |
ed1efb2a SK |
27 | #include <linux/net.h> |
28 | #include <linux/in6.h> | |
29 | #include <linux/netdevice.h> | |
30 | #include <linux/if_arp.h> | |
31 | #include <linux/icmpv6.h> | |
32 | #include <linux/init.h> | |
33 | #include <linux/route.h> | |
34 | #include <linux/rtnetlink.h> | |
35 | #include <linux/netfilter_ipv6.h> | |
36 | #include <linux/slab.h> | |
37 | #include <linux/hash.h> | |
38 | ||
39 | #include <linux/uaccess.h> | |
40 | #include <linux/atomic.h> | |
41 | ||
42 | #include <net/icmp.h> | |
43 | #include <net/ip.h> | |
44 | #include <net/ip_tunnels.h> | |
45 | #include <net/ipv6.h> | |
46 | #include <net/ip6_route.h> | |
47 | #include <net/addrconf.h> | |
48 | #include <net/ip6_tunnel.h> | |
49 | #include <net/xfrm.h> | |
50 | #include <net/net_namespace.h> | |
51 | #include <net/netns/generic.h> | |
52 | ||
53 | #define HASH_SIZE_SHIFT 5 | |
54 | #define HASH_SIZE (1 << HASH_SIZE_SHIFT) | |
55 | ||
56 | static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2) | |
57 | { | |
58 | u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2); | |
59 | ||
60 | return hash_32(hash, HASH_SIZE_SHIFT); | |
61 | } | |
62 | ||
63 | static int vti6_dev_init(struct net_device *dev); | |
64 | static void vti6_dev_setup(struct net_device *dev); | |
65 | static struct rtnl_link_ops vti6_link_ops __read_mostly; | |
66 | ||
67 | static int vti6_net_id __read_mostly; | |
68 | struct vti6_net { | |
69 | /* the vti6 tunnel fallback device */ | |
70 | struct net_device *fb_tnl_dev; | |
71 | /* lists for storing tunnels in use */ | |
72 | struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE]; | |
73 | struct ip6_tnl __rcu *tnls_wc[1]; | |
74 | struct ip6_tnl __rcu **tnls[2]; | |
75 | }; | |
76 | ||
ed1efb2a SK |
77 | #define for_each_vti6_tunnel_rcu(start) \ |
78 | for (t = rcu_dereference(start); t; t = rcu_dereference(t->next)) | |
79 | ||
80 | /** | |
81 | * vti6_tnl_lookup - fetch tunnel matching the end-point addresses | |
82 | * @net: network namespace | |
83 | * @remote: the address of the tunnel exit-point | |
84 | * @local: the address of the tunnel entry-point | |
85 | * | |
86 | * Return: | |
87 | * tunnel matching given end-points if found, | |
88 | * else fallback tunnel if its device is up, | |
89 | * else %NULL | |
90 | **/ | |
91 | static struct ip6_tnl * | |
92 | vti6_tnl_lookup(struct net *net, const struct in6_addr *remote, | |
93 | const struct in6_addr *local) | |
94 | { | |
95 | unsigned int hash = HASH(remote, local); | |
96 | struct ip6_tnl *t; | |
97 | struct vti6_net *ip6n = net_generic(net, vti6_net_id); | |
98 | ||
99 | for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) { | |
100 | if (ipv6_addr_equal(local, &t->parms.laddr) && | |
101 | ipv6_addr_equal(remote, &t->parms.raddr) && | |
102 | (t->dev->flags & IFF_UP)) | |
103 | return t; | |
104 | } | |
105 | t = rcu_dereference(ip6n->tnls_wc[0]); | |
106 | if (t && (t->dev->flags & IFF_UP)) | |
107 | return t; | |
108 | ||
109 | return NULL; | |
110 | } | |
111 | ||
112 | /** | |
113 | * vti6_tnl_bucket - get head of list matching given tunnel parameters | |
114 | * @p: parameters containing tunnel end-points | |
115 | * | |
116 | * Description: | |
117 | * vti6_tnl_bucket() returns the head of the list matching the | |
118 | * &struct in6_addr entries laddr and raddr in @p. | |
119 | * | |
120 | * Return: head of IPv6 tunnel list | |
121 | **/ | |
122 | static struct ip6_tnl __rcu ** | |
123 | vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p) | |
124 | { | |
125 | const struct in6_addr *remote = &p->raddr; | |
126 | const struct in6_addr *local = &p->laddr; | |
127 | unsigned int h = 0; | |
128 | int prio = 0; | |
129 | ||
130 | if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) { | |
131 | prio = 1; | |
132 | h = HASH(remote, local); | |
133 | } | |
134 | return &ip6n->tnls[prio][h]; | |
135 | } | |
136 | ||
137 | static void | |
138 | vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t) | |
139 | { | |
140 | struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms); | |
141 | ||
142 | rcu_assign_pointer(t->next , rtnl_dereference(*tp)); | |
143 | rcu_assign_pointer(*tp, t); | |
144 | } | |
145 | ||
146 | static void | |
147 | vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t) | |
148 | { | |
149 | struct ip6_tnl __rcu **tp; | |
150 | struct ip6_tnl *iter; | |
151 | ||
152 | for (tp = vti6_tnl_bucket(ip6n, &t->parms); | |
153 | (iter = rtnl_dereference(*tp)) != NULL; | |
154 | tp = &iter->next) { | |
155 | if (t == iter) { | |
156 | rcu_assign_pointer(*tp, t->next); | |
157 | break; | |
158 | } | |
159 | } | |
160 | } | |
161 | ||
162 | static void vti6_dev_free(struct net_device *dev) | |
163 | { | |
164 | free_percpu(dev->tstats); | |
165 | free_netdev(dev); | |
166 | } | |
167 | ||
168 | static int vti6_tnl_create2(struct net_device *dev) | |
169 | { | |
170 | struct ip6_tnl *t = netdev_priv(dev); | |
171 | struct net *net = dev_net(dev); | |
172 | struct vti6_net *ip6n = net_generic(net, vti6_net_id); | |
173 | int err; | |
174 | ||
175 | err = vti6_dev_init(dev); | |
176 | if (err < 0) | |
177 | goto out; | |
178 | ||
179 | err = register_netdevice(dev); | |
180 | if (err < 0) | |
181 | goto out; | |
182 | ||
183 | strcpy(t->parms.name, dev->name); | |
184 | dev->rtnl_link_ops = &vti6_link_ops; | |
185 | ||
186 | dev_hold(dev); | |
187 | vti6_tnl_link(ip6n, t); | |
188 | ||
189 | return 0; | |
190 | ||
191 | out: | |
192 | return err; | |
193 | } | |
194 | ||
195 | static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p) | |
196 | { | |
197 | struct net_device *dev; | |
198 | struct ip6_tnl *t; | |
199 | char name[IFNAMSIZ]; | |
200 | int err; | |
201 | ||
202 | if (p->name[0]) | |
203 | strlcpy(name, p->name, IFNAMSIZ); | |
204 | else | |
205 | sprintf(name, "ip6_vti%%d"); | |
206 | ||
207 | dev = alloc_netdev(sizeof(*t), name, vti6_dev_setup); | |
208 | if (dev == NULL) | |
209 | goto failed; | |
210 | ||
211 | dev_net_set(dev, net); | |
212 | ||
213 | t = netdev_priv(dev); | |
214 | t->parms = *p; | |
215 | t->net = dev_net(dev); | |
216 | ||
217 | err = vti6_tnl_create2(dev); | |
218 | if (err < 0) | |
219 | goto failed_free; | |
220 | ||
221 | return t; | |
222 | ||
223 | failed_free: | |
224 | vti6_dev_free(dev); | |
225 | failed: | |
226 | return NULL; | |
227 | } | |
228 | ||
229 | /** | |
230 | * vti6_locate - find or create tunnel matching given parameters | |
231 | * @net: network namespace | |
232 | * @p: tunnel parameters | |
233 | * @create: != 0 if allowed to create new tunnel if no match found | |
234 | * | |
235 | * Description: | |
236 | * vti6_locate() first tries to locate an existing tunnel | |
237 | * based on @parms. If this is unsuccessful, but @create is set a new | |
238 | * tunnel device is created and registered for use. | |
239 | * | |
240 | * Return: | |
241 | * matching tunnel or NULL | |
242 | **/ | |
243 | static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p, | |
244 | int create) | |
245 | { | |
246 | const struct in6_addr *remote = &p->raddr; | |
247 | const struct in6_addr *local = &p->laddr; | |
248 | struct ip6_tnl __rcu **tp; | |
249 | struct ip6_tnl *t; | |
250 | struct vti6_net *ip6n = net_generic(net, vti6_net_id); | |
251 | ||
252 | for (tp = vti6_tnl_bucket(ip6n, p); | |
253 | (t = rtnl_dereference(*tp)) != NULL; | |
254 | tp = &t->next) { | |
255 | if (ipv6_addr_equal(local, &t->parms.laddr) && | |
256 | ipv6_addr_equal(remote, &t->parms.raddr)) | |
257 | return t; | |
258 | } | |
259 | if (!create) | |
260 | return NULL; | |
261 | return vti6_tnl_create(net, p); | |
262 | } | |
263 | ||
264 | /** | |
265 | * vti6_dev_uninit - tunnel device uninitializer | |
266 | * @dev: the device to be destroyed | |
267 | * | |
268 | * Description: | |
269 | * vti6_dev_uninit() removes tunnel from its list | |
270 | **/ | |
271 | static void vti6_dev_uninit(struct net_device *dev) | |
272 | { | |
273 | struct ip6_tnl *t = netdev_priv(dev); | |
274 | struct net *net = dev_net(dev); | |
275 | struct vti6_net *ip6n = net_generic(net, vti6_net_id); | |
276 | ||
277 | if (dev == ip6n->fb_tnl_dev) | |
278 | RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL); | |
279 | else | |
280 | vti6_tnl_unlink(ip6n, t); | |
281 | ip6_tnl_dst_reset(t); | |
282 | dev_put(dev); | |
283 | } | |
284 | ||
285 | static int vti6_rcv(struct sk_buff *skb) | |
286 | { | |
287 | struct ip6_tnl *t; | |
288 | const struct ipv6hdr *ipv6h = ipv6_hdr(skb); | |
289 | ||
290 | rcu_read_lock(); | |
291 | ||
292 | if ((t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, | |
293 | &ipv6h->daddr)) != NULL) { | |
8f84985f | 294 | struct pcpu_sw_netstats *tstats; |
ed1efb2a SK |
295 | |
296 | if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) { | |
297 | rcu_read_unlock(); | |
298 | goto discard; | |
299 | } | |
300 | ||
301 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { | |
302 | rcu_read_unlock(); | |
303 | return 0; | |
304 | } | |
305 | ||
306 | if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) { | |
307 | t->dev->stats.rx_dropped++; | |
308 | rcu_read_unlock(); | |
309 | goto discard; | |
310 | } | |
311 | ||
312 | tstats = this_cpu_ptr(t->dev->tstats); | |
469bdcef | 313 | u64_stats_update_begin(&tstats->syncp); |
ed1efb2a SK |
314 | tstats->rx_packets++; |
315 | tstats->rx_bytes += skb->len; | |
469bdcef | 316 | u64_stats_update_end(&tstats->syncp); |
ed1efb2a SK |
317 | |
318 | skb->mark = 0; | |
319 | secpath_reset(skb); | |
320 | skb->dev = t->dev; | |
321 | ||
322 | rcu_read_unlock(); | |
323 | return 0; | |
324 | } | |
325 | rcu_read_unlock(); | |
326 | return 1; | |
327 | ||
328 | discard: | |
329 | kfree_skb(skb); | |
330 | return 0; | |
331 | } | |
332 | ||
333 | /** | |
334 | * vti6_addr_conflict - compare packet addresses to tunnel's own | |
335 | * @t: the outgoing tunnel device | |
336 | * @hdr: IPv6 header from the incoming packet | |
337 | * | |
338 | * Description: | |
339 | * Avoid trivial tunneling loop by checking that tunnel exit-point | |
340 | * doesn't match source of incoming packet. | |
341 | * | |
342 | * Return: | |
343 | * 1 if conflict, | |
344 | * 0 else | |
345 | **/ | |
346 | static inline bool | |
347 | vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr) | |
348 | { | |
349 | return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); | |
350 | } | |
351 | ||
352 | /** | |
353 | * vti6_xmit - send a packet | |
354 | * @skb: the outgoing socket buffer | |
355 | * @dev: the outgoing tunnel device | |
356 | **/ | |
357 | static int vti6_xmit(struct sk_buff *skb, struct net_device *dev) | |
358 | { | |
359 | struct net *net = dev_net(dev); | |
360 | struct ip6_tnl *t = netdev_priv(dev); | |
361 | struct net_device_stats *stats = &t->dev->stats; | |
362 | struct dst_entry *dst = NULL, *ndst = NULL; | |
363 | struct flowi6 fl6; | |
364 | struct ipv6hdr *ipv6h = ipv6_hdr(skb); | |
365 | struct net_device *tdev; | |
366 | int err = -1; | |
367 | ||
368 | if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) || | |
369 | !ip6_tnl_xmit_ctl(t) || vti6_addr_conflict(t, ipv6h)) | |
370 | return err; | |
371 | ||
372 | dst = ip6_tnl_dst_check(t); | |
373 | if (!dst) { | |
374 | memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); | |
375 | ||
376 | ndst = ip6_route_output(net, NULL, &fl6); | |
377 | ||
378 | if (ndst->error) | |
379 | goto tx_err_link_failure; | |
380 | ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(&fl6), NULL, 0); | |
381 | if (IS_ERR(ndst)) { | |
382 | err = PTR_ERR(ndst); | |
383 | ndst = NULL; | |
384 | goto tx_err_link_failure; | |
385 | } | |
386 | dst = ndst; | |
387 | } | |
388 | ||
389 | if (!dst->xfrm || dst->xfrm->props.mode != XFRM_MODE_TUNNEL) | |
390 | goto tx_err_link_failure; | |
391 | ||
392 | tdev = dst->dev; | |
393 | ||
394 | if (tdev == dev) { | |
395 | stats->collisions++; | |
396 | net_warn_ratelimited("%s: Local routing loop detected!\n", | |
397 | t->parms.name); | |
398 | goto tx_err_dst_release; | |
399 | } | |
400 | ||
401 | ||
402 | skb_dst_drop(skb); | |
403 | skb_dst_set_noref(skb, dst); | |
404 | ||
405 | ip6tunnel_xmit(skb, dev); | |
406 | if (ndst) { | |
407 | dev->mtu = dst_mtu(ndst); | |
408 | ip6_tnl_dst_store(t, ndst); | |
409 | } | |
410 | ||
411 | return 0; | |
412 | tx_err_link_failure: | |
413 | stats->tx_carrier_errors++; | |
414 | dst_link_failure(skb); | |
415 | tx_err_dst_release: | |
416 | dst_release(ndst); | |
417 | return err; | |
418 | } | |
419 | ||
420 | static netdev_tx_t | |
421 | vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) | |
422 | { | |
423 | struct ip6_tnl *t = netdev_priv(dev); | |
424 | struct net_device_stats *stats = &t->dev->stats; | |
425 | int ret; | |
426 | ||
427 | switch (skb->protocol) { | |
428 | case htons(ETH_P_IPV6): | |
429 | ret = vti6_xmit(skb, dev); | |
430 | break; | |
431 | default: | |
432 | goto tx_err; | |
433 | } | |
434 | ||
435 | if (ret < 0) | |
436 | goto tx_err; | |
437 | ||
438 | return NETDEV_TX_OK; | |
439 | ||
440 | tx_err: | |
441 | stats->tx_errors++; | |
442 | stats->tx_dropped++; | |
443 | kfree_skb(skb); | |
444 | return NETDEV_TX_OK; | |
445 | } | |
446 | ||
447 | static void vti6_link_config(struct ip6_tnl *t) | |
448 | { | |
449 | struct dst_entry *dst; | |
450 | struct net_device *dev = t->dev; | |
451 | struct __ip6_tnl_parm *p = &t->parms; | |
452 | struct flowi6 *fl6 = &t->fl.u.ip6; | |
453 | ||
454 | memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); | |
455 | memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); | |
456 | ||
457 | /* Set up flowi template */ | |
458 | fl6->saddr = p->laddr; | |
459 | fl6->daddr = p->raddr; | |
460 | fl6->flowi6_oif = p->link; | |
461 | fl6->flowi6_mark = be32_to_cpu(p->i_key); | |
462 | fl6->flowi6_proto = p->proto; | |
463 | fl6->flowlabel = 0; | |
464 | ||
465 | p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV | | |
466 | IP6_TNL_F_CAP_PER_PACKET); | |
467 | p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); | |
468 | ||
469 | if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV) | |
470 | dev->flags |= IFF_POINTOPOINT; | |
471 | else | |
472 | dev->flags &= ~IFF_POINTOPOINT; | |
473 | ||
474 | dev->iflink = p->link; | |
475 | ||
476 | if (p->flags & IP6_TNL_F_CAP_XMIT) { | |
477 | ||
478 | dst = ip6_route_output(dev_net(dev), NULL, fl6); | |
479 | if (dst->error) | |
480 | return; | |
481 | ||
482 | dst = xfrm_lookup(dev_net(dev), dst, flowi6_to_flowi(fl6), | |
483 | NULL, 0); | |
484 | if (IS_ERR(dst)) | |
485 | return; | |
486 | ||
487 | if (dst->dev) { | |
488 | dev->hard_header_len = dst->dev->hard_header_len; | |
489 | ||
490 | dev->mtu = dst_mtu(dst); | |
491 | ||
492 | if (dev->mtu < IPV6_MIN_MTU) | |
493 | dev->mtu = IPV6_MIN_MTU; | |
494 | } | |
495 | dst_release(dst); | |
496 | } | |
497 | } | |
498 | ||
499 | /** | |
500 | * vti6_tnl_change - update the tunnel parameters | |
501 | * @t: tunnel to be changed | |
502 | * @p: tunnel configuration parameters | |
503 | * | |
504 | * Description: | |
505 | * vti6_tnl_change() updates the tunnel parameters | |
506 | **/ | |
507 | static int | |
508 | vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p) | |
509 | { | |
510 | t->parms.laddr = p->laddr; | |
511 | t->parms.raddr = p->raddr; | |
512 | t->parms.link = p->link; | |
513 | t->parms.i_key = p->i_key; | |
514 | t->parms.o_key = p->o_key; | |
515 | t->parms.proto = p->proto; | |
516 | ip6_tnl_dst_reset(t); | |
517 | vti6_link_config(t); | |
518 | return 0; | |
519 | } | |
520 | ||
521 | static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p) | |
522 | { | |
523 | struct net *net = dev_net(t->dev); | |
524 | struct vti6_net *ip6n = net_generic(net, vti6_net_id); | |
525 | int err; | |
526 | ||
527 | vti6_tnl_unlink(ip6n, t); | |
528 | synchronize_net(); | |
529 | err = vti6_tnl_change(t, p); | |
530 | vti6_tnl_link(ip6n, t); | |
531 | netdev_state_change(t->dev); | |
532 | return err; | |
533 | } | |
534 | ||
535 | static void | |
536 | vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u) | |
537 | { | |
538 | p->laddr = u->laddr; | |
539 | p->raddr = u->raddr; | |
540 | p->link = u->link; | |
541 | p->i_key = u->i_key; | |
542 | p->o_key = u->o_key; | |
543 | p->proto = u->proto; | |
544 | ||
545 | memcpy(p->name, u->name, sizeof(u->name)); | |
546 | } | |
547 | ||
548 | static void | |
549 | vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p) | |
550 | { | |
551 | u->laddr = p->laddr; | |
552 | u->raddr = p->raddr; | |
553 | u->link = p->link; | |
554 | u->i_key = p->i_key; | |
555 | u->o_key = p->o_key; | |
556 | u->proto = p->proto; | |
557 | ||
558 | memcpy(u->name, p->name, sizeof(u->name)); | |
559 | } | |
560 | ||
561 | /** | |
562 | * vti6_tnl_ioctl - configure vti6 tunnels from userspace | |
563 | * @dev: virtual device associated with tunnel | |
564 | * @ifr: parameters passed from userspace | |
565 | * @cmd: command to be performed | |
566 | * | |
567 | * Description: | |
568 | * vti6_ioctl() is used for managing vti6 tunnels | |
569 | * from userspace. | |
570 | * | |
571 | * The possible commands are the following: | |
572 | * %SIOCGETTUNNEL: get tunnel parameters for device | |
573 | * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters | |
574 | * %SIOCCHGTUNNEL: change tunnel parameters to those given | |
575 | * %SIOCDELTUNNEL: delete tunnel | |
576 | * | |
577 | * The fallback device "ip6_vti0", created during module | |
578 | * initialization, can be used for creating other tunnel devices. | |
579 | * | |
580 | * Return: | |
581 | * 0 on success, | |
582 | * %-EFAULT if unable to copy data to or from userspace, | |
583 | * %-EPERM if current process hasn't %CAP_NET_ADMIN set | |
584 | * %-EINVAL if passed tunnel parameters are invalid, | |
585 | * %-EEXIST if changing a tunnel's parameters would cause a conflict | |
586 | * %-ENODEV if attempting to change or delete a nonexisting device | |
587 | **/ | |
588 | static int | |
589 | vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) | |
590 | { | |
591 | int err = 0; | |
592 | struct ip6_tnl_parm2 p; | |
593 | struct __ip6_tnl_parm p1; | |
594 | struct ip6_tnl *t = NULL; | |
595 | struct net *net = dev_net(dev); | |
596 | struct vti6_net *ip6n = net_generic(net, vti6_net_id); | |
597 | ||
598 | switch (cmd) { | |
599 | case SIOCGETTUNNEL: | |
600 | if (dev == ip6n->fb_tnl_dev) { | |
601 | if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { | |
602 | err = -EFAULT; | |
603 | break; | |
604 | } | |
605 | vti6_parm_from_user(&p1, &p); | |
606 | t = vti6_locate(net, &p1, 0); | |
607 | } else { | |
608 | memset(&p, 0, sizeof(p)); | |
609 | } | |
610 | if (t == NULL) | |
611 | t = netdev_priv(dev); | |
612 | vti6_parm_to_user(&p, &t->parms); | |
613 | if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) | |
614 | err = -EFAULT; | |
615 | break; | |
616 | case SIOCADDTUNNEL: | |
617 | case SIOCCHGTUNNEL: | |
618 | err = -EPERM; | |
619 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
620 | break; | |
621 | err = -EFAULT; | |
622 | if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) | |
623 | break; | |
624 | err = -EINVAL; | |
625 | if (p.proto != IPPROTO_IPV6 && p.proto != 0) | |
626 | break; | |
627 | vti6_parm_from_user(&p1, &p); | |
628 | t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL); | |
629 | if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) { | |
630 | if (t != NULL) { | |
631 | if (t->dev != dev) { | |
632 | err = -EEXIST; | |
633 | break; | |
634 | } | |
635 | } else | |
636 | t = netdev_priv(dev); | |
637 | ||
638 | err = vti6_update(t, &p1); | |
639 | } | |
640 | if (t) { | |
641 | err = 0; | |
642 | vti6_parm_to_user(&p, &t->parms); | |
643 | if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) | |
644 | err = -EFAULT; | |
645 | ||
646 | } else | |
647 | err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); | |
648 | break; | |
649 | case SIOCDELTUNNEL: | |
650 | err = -EPERM; | |
651 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
652 | break; | |
653 | ||
654 | if (dev == ip6n->fb_tnl_dev) { | |
655 | err = -EFAULT; | |
656 | if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) | |
657 | break; | |
658 | err = -ENOENT; | |
659 | vti6_parm_from_user(&p1, &p); | |
660 | t = vti6_locate(net, &p1, 0); | |
661 | if (t == NULL) | |
662 | break; | |
663 | err = -EPERM; | |
664 | if (t->dev == ip6n->fb_tnl_dev) | |
665 | break; | |
666 | dev = t->dev; | |
667 | } | |
668 | err = 0; | |
669 | unregister_netdevice(dev); | |
670 | break; | |
671 | default: | |
672 | err = -EINVAL; | |
673 | } | |
674 | return err; | |
675 | } | |
676 | ||
677 | /** | |
678 | * vti6_tnl_change_mtu - change mtu manually for tunnel device | |
679 | * @dev: virtual device associated with tunnel | |
680 | * @new_mtu: the new mtu | |
681 | * | |
682 | * Return: | |
683 | * 0 on success, | |
684 | * %-EINVAL if mtu too small | |
685 | **/ | |
686 | static int vti6_change_mtu(struct net_device *dev, int new_mtu) | |
687 | { | |
688 | if (new_mtu < IPV6_MIN_MTU) | |
689 | return -EINVAL; | |
690 | ||
691 | dev->mtu = new_mtu; | |
692 | return 0; | |
693 | } | |
694 | ||
695 | static const struct net_device_ops vti6_netdev_ops = { | |
696 | .ndo_uninit = vti6_dev_uninit, | |
697 | .ndo_start_xmit = vti6_tnl_xmit, | |
698 | .ndo_do_ioctl = vti6_ioctl, | |
699 | .ndo_change_mtu = vti6_change_mtu, | |
469bdcef | 700 | .ndo_get_stats64 = ip_tunnel_get_stats64, |
ed1efb2a SK |
701 | }; |
702 | ||
703 | /** | |
704 | * vti6_dev_setup - setup virtual tunnel device | |
705 | * @dev: virtual device associated with tunnel | |
706 | * | |
707 | * Description: | |
708 | * Initialize function pointers and device parameters | |
709 | **/ | |
710 | static void vti6_dev_setup(struct net_device *dev) | |
711 | { | |
712 | struct ip6_tnl *t; | |
713 | ||
714 | dev->netdev_ops = &vti6_netdev_ops; | |
715 | dev->destructor = vti6_dev_free; | |
716 | ||
717 | dev->type = ARPHRD_TUNNEL6; | |
718 | dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr); | |
719 | dev->mtu = ETH_DATA_LEN; | |
720 | t = netdev_priv(dev); | |
721 | dev->flags |= IFF_NOARP; | |
722 | dev->addr_len = sizeof(struct in6_addr); | |
723 | dev->features |= NETIF_F_NETNS_LOCAL; | |
724 | dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; | |
725 | } | |
726 | ||
727 | /** | |
728 | * vti6_dev_init_gen - general initializer for all tunnel devices | |
729 | * @dev: virtual device associated with tunnel | |
730 | **/ | |
731 | static inline int vti6_dev_init_gen(struct net_device *dev) | |
732 | { | |
733 | struct ip6_tnl *t = netdev_priv(dev); | |
734 | ||
735 | t->dev = dev; | |
736 | t->net = dev_net(dev); | |
1c213bd2 | 737 | dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); |
ed1efb2a SK |
738 | if (!dev->tstats) |
739 | return -ENOMEM; | |
740 | return 0; | |
741 | } | |
742 | ||
743 | /** | |
744 | * vti6_dev_init - initializer for all non fallback tunnel devices | |
745 | * @dev: virtual device associated with tunnel | |
746 | **/ | |
747 | static int vti6_dev_init(struct net_device *dev) | |
748 | { | |
749 | struct ip6_tnl *t = netdev_priv(dev); | |
750 | int err = vti6_dev_init_gen(dev); | |
751 | ||
752 | if (err) | |
753 | return err; | |
754 | vti6_link_config(t); | |
755 | return 0; | |
756 | } | |
757 | ||
758 | /** | |
759 | * vti6_fb_tnl_dev_init - initializer for fallback tunnel device | |
760 | * @dev: fallback device | |
761 | * | |
762 | * Return: 0 | |
763 | **/ | |
764 | static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev) | |
765 | { | |
766 | struct ip6_tnl *t = netdev_priv(dev); | |
767 | struct net *net = dev_net(dev); | |
768 | struct vti6_net *ip6n = net_generic(net, vti6_net_id); | |
769 | int err = vti6_dev_init_gen(dev); | |
770 | ||
771 | if (err) | |
772 | return err; | |
773 | ||
774 | t->parms.proto = IPPROTO_IPV6; | |
775 | dev_hold(dev); | |
776 | ||
777 | vti6_link_config(t); | |
778 | ||
779 | rcu_assign_pointer(ip6n->tnls_wc[0], t); | |
780 | return 0; | |
781 | } | |
782 | ||
783 | static int vti6_validate(struct nlattr *tb[], struct nlattr *data[]) | |
784 | { | |
785 | return 0; | |
786 | } | |
787 | ||
788 | static void vti6_netlink_parms(struct nlattr *data[], | |
789 | struct __ip6_tnl_parm *parms) | |
790 | { | |
791 | memset(parms, 0, sizeof(*parms)); | |
792 | ||
793 | if (!data) | |
794 | return; | |
795 | ||
796 | if (data[IFLA_VTI_LINK]) | |
797 | parms->link = nla_get_u32(data[IFLA_VTI_LINK]); | |
798 | ||
799 | if (data[IFLA_VTI_LOCAL]) | |
800 | nla_memcpy(&parms->laddr, data[IFLA_VTI_LOCAL], | |
801 | sizeof(struct in6_addr)); | |
802 | ||
803 | if (data[IFLA_VTI_REMOTE]) | |
804 | nla_memcpy(&parms->raddr, data[IFLA_VTI_REMOTE], | |
805 | sizeof(struct in6_addr)); | |
806 | ||
807 | if (data[IFLA_VTI_IKEY]) | |
808 | parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]); | |
809 | ||
810 | if (data[IFLA_VTI_OKEY]) | |
811 | parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]); | |
812 | } | |
813 | ||
814 | static int vti6_newlink(struct net *src_net, struct net_device *dev, | |
815 | struct nlattr *tb[], struct nlattr *data[]) | |
816 | { | |
817 | struct net *net = dev_net(dev); | |
818 | struct ip6_tnl *nt; | |
819 | ||
820 | nt = netdev_priv(dev); | |
821 | vti6_netlink_parms(data, &nt->parms); | |
822 | ||
823 | nt->parms.proto = IPPROTO_IPV6; | |
824 | ||
825 | if (vti6_locate(net, &nt->parms, 0)) | |
826 | return -EEXIST; | |
827 | ||
828 | return vti6_tnl_create2(dev); | |
829 | } | |
830 | ||
831 | static int vti6_changelink(struct net_device *dev, struct nlattr *tb[], | |
832 | struct nlattr *data[]) | |
833 | { | |
834 | struct ip6_tnl *t; | |
835 | struct __ip6_tnl_parm p; | |
836 | struct net *net = dev_net(dev); | |
837 | struct vti6_net *ip6n = net_generic(net, vti6_net_id); | |
838 | ||
839 | if (dev == ip6n->fb_tnl_dev) | |
840 | return -EINVAL; | |
841 | ||
842 | vti6_netlink_parms(data, &p); | |
843 | ||
844 | t = vti6_locate(net, &p, 0); | |
845 | ||
846 | if (t) { | |
847 | if (t->dev != dev) | |
848 | return -EEXIST; | |
849 | } else | |
850 | t = netdev_priv(dev); | |
851 | ||
852 | return vti6_update(t, &p); | |
853 | } | |
854 | ||
855 | static size_t vti6_get_size(const struct net_device *dev) | |
856 | { | |
857 | return | |
858 | /* IFLA_VTI_LINK */ | |
859 | nla_total_size(4) + | |
860 | /* IFLA_VTI_LOCAL */ | |
861 | nla_total_size(sizeof(struct in6_addr)) + | |
862 | /* IFLA_VTI_REMOTE */ | |
863 | nla_total_size(sizeof(struct in6_addr)) + | |
864 | /* IFLA_VTI_IKEY */ | |
865 | nla_total_size(4) + | |
866 | /* IFLA_VTI_OKEY */ | |
867 | nla_total_size(4) + | |
868 | 0; | |
869 | } | |
870 | ||
871 | static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev) | |
872 | { | |
873 | struct ip6_tnl *tunnel = netdev_priv(dev); | |
874 | struct __ip6_tnl_parm *parm = &tunnel->parms; | |
875 | ||
876 | if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) || | |
877 | nla_put(skb, IFLA_VTI_LOCAL, sizeof(struct in6_addr), | |
878 | &parm->laddr) || | |
879 | nla_put(skb, IFLA_VTI_REMOTE, sizeof(struct in6_addr), | |
880 | &parm->raddr) || | |
881 | nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) || | |
882 | nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key)) | |
883 | goto nla_put_failure; | |
884 | return 0; | |
885 | ||
886 | nla_put_failure: | |
887 | return -EMSGSIZE; | |
888 | } | |
889 | ||
890 | static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = { | |
891 | [IFLA_VTI_LINK] = { .type = NLA_U32 }, | |
892 | [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) }, | |
893 | [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) }, | |
894 | [IFLA_VTI_IKEY] = { .type = NLA_U32 }, | |
895 | [IFLA_VTI_OKEY] = { .type = NLA_U32 }, | |
896 | }; | |
897 | ||
898 | static struct rtnl_link_ops vti6_link_ops __read_mostly = { | |
899 | .kind = "vti6", | |
900 | .maxtype = IFLA_VTI_MAX, | |
901 | .policy = vti6_policy, | |
902 | .priv_size = sizeof(struct ip6_tnl), | |
903 | .setup = vti6_dev_setup, | |
904 | .validate = vti6_validate, | |
905 | .newlink = vti6_newlink, | |
906 | .changelink = vti6_changelink, | |
907 | .get_size = vti6_get_size, | |
908 | .fill_info = vti6_fill_info, | |
909 | }; | |
910 | ||
911 | static struct xfrm_tunnel_notifier vti6_handler __read_mostly = { | |
912 | .handler = vti6_rcv, | |
913 | .priority = 1, | |
914 | }; | |
915 | ||
916 | static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n) | |
917 | { | |
918 | int h; | |
919 | struct ip6_tnl *t; | |
920 | LIST_HEAD(list); | |
921 | ||
922 | for (h = 0; h < HASH_SIZE; h++) { | |
923 | t = rtnl_dereference(ip6n->tnls_r_l[h]); | |
924 | while (t != NULL) { | |
925 | unregister_netdevice_queue(t->dev, &list); | |
926 | t = rtnl_dereference(t->next); | |
927 | } | |
928 | } | |
929 | ||
930 | t = rtnl_dereference(ip6n->tnls_wc[0]); | |
931 | unregister_netdevice_queue(t->dev, &list); | |
932 | unregister_netdevice_many(&list); | |
933 | } | |
934 | ||
935 | static int __net_init vti6_init_net(struct net *net) | |
936 | { | |
937 | struct vti6_net *ip6n = net_generic(net, vti6_net_id); | |
938 | struct ip6_tnl *t = NULL; | |
939 | int err; | |
940 | ||
941 | ip6n->tnls[0] = ip6n->tnls_wc; | |
942 | ip6n->tnls[1] = ip6n->tnls_r_l; | |
943 | ||
944 | err = -ENOMEM; | |
945 | ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0", | |
946 | vti6_dev_setup); | |
947 | ||
948 | if (!ip6n->fb_tnl_dev) | |
949 | goto err_alloc_dev; | |
950 | dev_net_set(ip6n->fb_tnl_dev, net); | |
951 | ||
952 | err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev); | |
953 | if (err < 0) | |
954 | goto err_register; | |
955 | ||
956 | err = register_netdev(ip6n->fb_tnl_dev); | |
957 | if (err < 0) | |
958 | goto err_register; | |
959 | ||
960 | t = netdev_priv(ip6n->fb_tnl_dev); | |
961 | ||
962 | strcpy(t->parms.name, ip6n->fb_tnl_dev->name); | |
963 | return 0; | |
964 | ||
965 | err_register: | |
966 | vti6_dev_free(ip6n->fb_tnl_dev); | |
967 | err_alloc_dev: | |
968 | return err; | |
969 | } | |
970 | ||
971 | static void __net_exit vti6_exit_net(struct net *net) | |
972 | { | |
973 | struct vti6_net *ip6n = net_generic(net, vti6_net_id); | |
974 | ||
975 | rtnl_lock(); | |
976 | vti6_destroy_tunnels(ip6n); | |
977 | rtnl_unlock(); | |
978 | } | |
979 | ||
980 | static struct pernet_operations vti6_net_ops = { | |
981 | .init = vti6_init_net, | |
982 | .exit = vti6_exit_net, | |
983 | .id = &vti6_net_id, | |
984 | .size = sizeof(struct vti6_net), | |
985 | }; | |
986 | ||
987 | /** | |
988 | * vti6_tunnel_init - register protocol and reserve needed resources | |
989 | * | |
990 | * Return: 0 on success | |
991 | **/ | |
992 | static int __init vti6_tunnel_init(void) | |
993 | { | |
994 | int err; | |
995 | ||
996 | err = register_pernet_device(&vti6_net_ops); | |
997 | if (err < 0) | |
998 | goto out_pernet; | |
999 | ||
1000 | err = xfrm6_mode_tunnel_input_register(&vti6_handler); | |
1001 | if (err < 0) { | |
1002 | pr_err("%s: can't register vti6\n", __func__); | |
1003 | goto out; | |
1004 | } | |
1005 | err = rtnl_link_register(&vti6_link_ops); | |
1006 | if (err < 0) | |
1007 | goto rtnl_link_failed; | |
1008 | ||
1009 | return 0; | |
1010 | ||
1011 | rtnl_link_failed: | |
1012 | xfrm6_mode_tunnel_input_deregister(&vti6_handler); | |
1013 | out: | |
1014 | unregister_pernet_device(&vti6_net_ops); | |
1015 | out_pernet: | |
1016 | return err; | |
1017 | } | |
1018 | ||
1019 | /** | |
1020 | * vti6_tunnel_cleanup - free resources and unregister protocol | |
1021 | **/ | |
1022 | static void __exit vti6_tunnel_cleanup(void) | |
1023 | { | |
1024 | rtnl_link_unregister(&vti6_link_ops); | |
1025 | if (xfrm6_mode_tunnel_input_deregister(&vti6_handler)) | |
1026 | pr_info("%s: can't deregister vti6\n", __func__); | |
1027 | ||
1028 | unregister_pernet_device(&vti6_net_ops); | |
1029 | } | |
1030 | ||
1031 | module_init(vti6_tunnel_init); | |
1032 | module_exit(vti6_tunnel_cleanup); | |
1033 | MODULE_LICENSE("GPL"); | |
1034 | MODULE_ALIAS_RTNL_LINK("vti6"); | |
1035 | MODULE_ALIAS_NETDEV("ip6_vti0"); | |
1036 | MODULE_AUTHOR("Steffen Klassert"); | |
1037 | MODULE_DESCRIPTION("IPv6 virtual tunnel interface"); |