usb: gadget: f_mtp: Avoid race between mtp_read and mtp_function_disable
[GitHub/exynos8895/android_kernel_samsung_universal8895.git] / net / ipv6 / ip6_vti.c
CommitLineData
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
56static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
57{
58 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
59
60 return hash_32(hash, HASH_SIZE_SHIFT);
61}
62
63static int vti6_dev_init(struct net_device *dev);
64static void vti6_dev_setup(struct net_device *dev);
65static struct rtnl_link_ops vti6_link_ops __read_mostly;
66
67static int vti6_net_id __read_mostly;
68struct vti6_net {
69 /* the vti6 tunnel fallback device */
70 struct net_device *fb_tnl_dev;
71 /* lists for storing tunnels in use */
72 struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
73 struct ip6_tnl __rcu *tnls_wc[1];
74 struct ip6_tnl __rcu **tnls[2];
75};
76
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 **/
91static struct ip6_tnl *
92vti6_tnl_lookup(struct net *net, const struct in6_addr *remote,
93 const struct in6_addr *local)
94{
95 unsigned int hash = HASH(remote, local);
96 struct ip6_tnl *t;
97 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
fbe68ee8 98 struct in6_addr any;
ed1efb2a
SK
99
100 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
101 if (ipv6_addr_equal(local, &t->parms.laddr) &&
102 ipv6_addr_equal(remote, &t->parms.raddr) &&
103 (t->dev->flags & IFF_UP))
104 return t;
105 }
fbe68ee8
SK
106
107 memset(&any, 0, sizeof(any));
108 hash = HASH(&any, local);
109 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
110 if (ipv6_addr_equal(local, &t->parms.laddr) &&
111 (t->dev->flags & IFF_UP))
112 return t;
113 }
114
115 hash = HASH(remote, &any);
116 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
117 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
118 (t->dev->flags & IFF_UP))
119 return t;
120 }
121
ed1efb2a
SK
122 t = rcu_dereference(ip6n->tnls_wc[0]);
123 if (t && (t->dev->flags & IFF_UP))
124 return t;
125
126 return NULL;
127}
128
129/**
130 * vti6_tnl_bucket - get head of list matching given tunnel parameters
131 * @p: parameters containing tunnel end-points
132 *
133 * Description:
134 * vti6_tnl_bucket() returns the head of the list matching the
135 * &struct in6_addr entries laddr and raddr in @p.
136 *
137 * Return: head of IPv6 tunnel list
138 **/
139static struct ip6_tnl __rcu **
140vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p)
141{
142 const struct in6_addr *remote = &p->raddr;
143 const struct in6_addr *local = &p->laddr;
144 unsigned int h = 0;
145 int prio = 0;
146
147 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
148 prio = 1;
149 h = HASH(remote, local);
150 }
151 return &ip6n->tnls[prio][h];
152}
153
154static void
155vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t)
156{
157 struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms);
158
159 rcu_assign_pointer(t->next , rtnl_dereference(*tp));
160 rcu_assign_pointer(*tp, t);
161}
162
163static void
164vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t)
165{
166 struct ip6_tnl __rcu **tp;
167 struct ip6_tnl *iter;
168
169 for (tp = vti6_tnl_bucket(ip6n, &t->parms);
170 (iter = rtnl_dereference(*tp)) != NULL;
171 tp = &iter->next) {
172 if (t == iter) {
173 rcu_assign_pointer(*tp, t->next);
174 break;
175 }
176 }
177}
178
179static void vti6_dev_free(struct net_device *dev)
180{
181 free_percpu(dev->tstats);
182 free_netdev(dev);
183}
184
185static int vti6_tnl_create2(struct net_device *dev)
186{
187 struct ip6_tnl *t = netdev_priv(dev);
188 struct net *net = dev_net(dev);
189 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
190 int err;
191
c73b58cf 192 dev->rtnl_link_ops = &vti6_link_ops;
ed1efb2a
SK
193 err = register_netdevice(dev);
194 if (err < 0)
195 goto out;
196
197 strcpy(t->parms.name, dev->name);
ed1efb2a
SK
198
199 dev_hold(dev);
200 vti6_tnl_link(ip6n, t);
201
202 return 0;
203
204out:
205 return err;
206}
207
208static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
209{
210 struct net_device *dev;
211 struct ip6_tnl *t;
212 char name[IFNAMSIZ];
213 int err;
214
215 if (p->name[0])
216 strlcpy(name, p->name, IFNAMSIZ);
217 else
218 sprintf(name, "ip6_vti%%d");
219
c835a677 220 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup);
63159f29 221 if (!dev)
ed1efb2a
SK
222 goto failed;
223
224 dev_net_set(dev, net);
225
226 t = netdev_priv(dev);
227 t->parms = *p;
228 t->net = dev_net(dev);
229
230 err = vti6_tnl_create2(dev);
231 if (err < 0)
232 goto failed_free;
233
234 return t;
235
236failed_free:
237 vti6_dev_free(dev);
238failed:
239 return NULL;
240}
241
242/**
243 * vti6_locate - find or create tunnel matching given parameters
244 * @net: network namespace
245 * @p: tunnel parameters
246 * @create: != 0 if allowed to create new tunnel if no match found
247 *
248 * Description:
249 * vti6_locate() first tries to locate an existing tunnel
250 * based on @parms. If this is unsuccessful, but @create is set a new
251 * tunnel device is created and registered for use.
252 *
253 * Return:
254 * matching tunnel or NULL
255 **/
256static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p,
257 int create)
258{
259 const struct in6_addr *remote = &p->raddr;
260 const struct in6_addr *local = &p->laddr;
261 struct ip6_tnl __rcu **tp;
262 struct ip6_tnl *t;
263 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
264
265 for (tp = vti6_tnl_bucket(ip6n, p);
266 (t = rtnl_dereference(*tp)) != NULL;
267 tp = &t->next) {
268 if (ipv6_addr_equal(local, &t->parms.laddr) &&
d814b847
SK
269 ipv6_addr_equal(remote, &t->parms.raddr)) {
270 if (create)
271 return NULL;
272
ed1efb2a 273 return t;
d814b847 274 }
ed1efb2a
SK
275 }
276 if (!create)
277 return NULL;
278 return vti6_tnl_create(net, p);
279}
280
281/**
282 * vti6_dev_uninit - tunnel device uninitializer
283 * @dev: the device to be destroyed
284 *
285 * Description:
286 * vti6_dev_uninit() removes tunnel from its list
287 **/
288static void vti6_dev_uninit(struct net_device *dev)
289{
290 struct ip6_tnl *t = netdev_priv(dev);
092a29a4 291 struct vti6_net *ip6n = net_generic(t->net, vti6_net_id);
ed1efb2a
SK
292
293 if (dev == ip6n->fb_tnl_dev)
294 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
295 else
296 vti6_tnl_unlink(ip6n, t);
ed1efb2a
SK
297 dev_put(dev);
298}
299
300static int vti6_rcv(struct sk_buff *skb)
301{
302 struct ip6_tnl *t;
303 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
304
305 rcu_read_lock();
e5d08d71 306 t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
53b24b8f 307 if (t) {
ed1efb2a
SK
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 return 0;
316 }
317
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
fa9ad96d 324 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6 = t;
ed1efb2a
SK
325
326 rcu_read_unlock();
fa9ad96d
SK
327
328 return xfrm6_rcv(skb);
ed1efb2a
SK
329 }
330 rcu_read_unlock();
fa9ad96d 331 return -EINVAL;
ed1efb2a
SK
332discard:
333 kfree_skb(skb);
334 return 0;
335}
336
fa9ad96d
SK
337static int vti6_rcv_cb(struct sk_buff *skb, int err)
338{
339 unsigned short family;
340 struct net_device *dev;
341 struct pcpu_sw_netstats *tstats;
342 struct xfrm_state *x;
343 struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6;
d55c670c
AD
344 u32 orig_mark = skb->mark;
345 int ret;
fa9ad96d
SK
346
347 if (!t)
348 return 1;
349
350 dev = t->dev;
351
352 if (err) {
353 dev->stats.rx_errors++;
354 dev->stats.rx_dropped++;
355
356 return 0;
357 }
358
359 x = xfrm_input_state(skb);
360 family = x->inner_mode->afinfo->family;
361
d55c670c
AD
362 skb->mark = be32_to_cpu(t->parms.i_key);
363 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
364 skb->mark = orig_mark;
365
366 if (!ret)
fa9ad96d
SK
367 return -EPERM;
368
369 skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev)));
370 skb->dev = dev;
371
372 tstats = this_cpu_ptr(dev->tstats);
373 u64_stats_update_begin(&tstats->syncp);
374 tstats->rx_packets++;
375 tstats->rx_bytes += skb->len;
376 u64_stats_update_end(&tstats->syncp);
377
378 return 0;
379}
380
ed1efb2a
SK
381/**
382 * vti6_addr_conflict - compare packet addresses to tunnel's own
383 * @t: the outgoing tunnel device
384 * @hdr: IPv6 header from the incoming packet
385 *
386 * Description:
387 * Avoid trivial tunneling loop by checking that tunnel exit-point
388 * doesn't match source of incoming packet.
389 *
390 * Return:
391 * 1 if conflict,
392 * 0 else
393 **/
394static inline bool
395vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
396{
397 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
398}
399
26be8e2d
SK
400static bool vti6_state_check(const struct xfrm_state *x,
401 const struct in6_addr *dst,
402 const struct in6_addr *src)
403{
404 xfrm_address_t *daddr = (xfrm_address_t *)dst;
405 xfrm_address_t *saddr = (xfrm_address_t *)src;
406
407 /* if there is no transform then this tunnel is not functional.
408 * Or if the xfrm is not mode tunnel.
409 */
410 if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
411 x->props.family != AF_INET6)
412 return false;
413
414 if (ipv6_addr_any(dst))
415 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6);
416
417 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6))
418 return false;
419
420 return true;
421}
422
ed1efb2a
SK
423/**
424 * vti6_xmit - send a packet
425 * @skb: the outgoing socket buffer
426 * @dev: the outgoing tunnel device
22e1b23d 427 * @fl: the flow informations for the xfrm_lookup
ed1efb2a 428 **/
22e1b23d
SK
429static int
430vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
ed1efb2a 431{
ed1efb2a
SK
432 struct ip6_tnl *t = netdev_priv(dev);
433 struct net_device_stats *stats = &t->dev->stats;
7c852581 434 struct dst_entry *dst = skb_dst(skb);
ed1efb2a 435 struct net_device *tdev;
d5005140 436 struct xfrm_state *x;
93040aa1 437 int pkt_len = skb->len;
ed1efb2a 438 int err = -1;
ccd740cb 439 int mtu;
ed1efb2a 440
7c852581
SK
441 if (!dst)
442 goto tx_err_link_failure;
ed1efb2a 443
7c852581 444 dst_hold(dst);
22e1b23d 445 dst = xfrm_lookup(t->net, dst, fl, NULL, 0);
7c852581
SK
446 if (IS_ERR(dst)) {
447 err = PTR_ERR(dst);
448 dst = NULL;
449 goto tx_err_link_failure;
ed1efb2a
SK
450 }
451
d5005140
SK
452 x = dst->xfrm;
453 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr))
454 goto tx_err_link_failure;
455
456 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr,
457 (const struct in6_addr *)&x->id.daddr))
ed1efb2a
SK
458 goto tx_err_link_failure;
459
460 tdev = dst->dev;
461
462 if (tdev == dev) {
463 stats->collisions++;
464 net_warn_ratelimited("%s: Local routing loop detected!\n",
465 t->parms.name);
466 goto tx_err_dst_release;
467 }
468
7c852581
SK
469 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
470 skb_dst_set(skb, dst);
471 skb->dev = skb_dst(skb)->dev;
ed1efb2a 472
ccd740cb
SK
473 mtu = dst_mtu(dst);
474 if (!skb->ignore_df && skb->len > mtu) {
475 skb_dst(skb)->ops->update_pmtu(dst, NULL, skb, mtu);
476
572b639b
SK
477 if (skb->protocol == htons(ETH_P_IPV6)) {
478 if (mtu < IPV6_MIN_MTU)
479 mtu = IPV6_MIN_MTU;
480
ccd740cb 481 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
572b639b 482 } else {
ccd740cb
SK
483 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
484 htonl(mtu));
572b639b 485 }
ccd740cb
SK
486
487 return -EMSGSIZE;
488 }
489
13206b6b 490 err = dst_output(t->net, skb->sk, skb);
22e1b23d
SK
491 if (net_xmit_eval(err) == 0) {
492 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
493
494 u64_stats_update_begin(&tstats->syncp);
93040aa1 495 tstats->tx_bytes += pkt_len;
22e1b23d
SK
496 tstats->tx_packets++;
497 u64_stats_update_end(&tstats->syncp);
498 } else {
499 stats->tx_errors++;
500 stats->tx_aborted_errors++;
501 }
ed1efb2a
SK
502
503 return 0;
504tx_err_link_failure:
505 stats->tx_carrier_errors++;
506 dst_link_failure(skb);
507tx_err_dst_release:
7c852581 508 dst_release(dst);
ed1efb2a
SK
509 return err;
510}
511
512static netdev_tx_t
513vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
514{
515 struct ip6_tnl *t = netdev_priv(dev);
516 struct net_device_stats *stats = &t->dev->stats;
22e1b23d
SK
517 struct ipv6hdr *ipv6h;
518 struct flowi fl;
ed1efb2a
SK
519 int ret;
520
22e1b23d 521 memset(&fl, 0, sizeof(fl));
22e1b23d 522
ed1efb2a
SK
523 switch (skb->protocol) {
524 case htons(ETH_P_IPV6):
22e1b23d
SK
525 ipv6h = ipv6_hdr(skb);
526
527 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
d5005140 528 vti6_addr_conflict(t, ipv6h))
22e1b23d
SK
529 goto tx_err;
530
531 xfrm_decode_session(skb, &fl, AF_INET6);
532 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
533 break;
534 case htons(ETH_P_IP):
535 xfrm_decode_session(skb, &fl, AF_INET);
536 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
ed1efb2a
SK
537 break;
538 default:
539 goto tx_err;
540 }
541
cd5279c1
AD
542 /* override mark with tunnel output key */
543 fl.flowi_mark = be32_to_cpu(t->parms.o_key);
544
22e1b23d 545 ret = vti6_xmit(skb, dev, &fl);
ed1efb2a
SK
546 if (ret < 0)
547 goto tx_err;
548
549 return NETDEV_TX_OK;
550
551tx_err:
552 stats->tx_errors++;
553 stats->tx_dropped++;
554 kfree_skb(skb);
555 return NETDEV_TX_OK;
556}
557
fa9ad96d
SK
558static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
559 u8 type, u8 code, int offset, __be32 info)
560{
561 __be32 spi;
6d004d6c 562 __u32 mark;
fa9ad96d
SK
563 struct xfrm_state *x;
564 struct ip6_tnl *t;
565 struct ip_esp_hdr *esph;
566 struct ip_auth_hdr *ah;
567 struct ip_comp_hdr *ipch;
568 struct net *net = dev_net(skb->dev);
569 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
570 int protocol = iph->nexthdr;
571
572 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
573 if (!t)
574 return -1;
575
6d004d6c
SK
576 mark = be32_to_cpu(t->parms.o_key);
577
fa9ad96d
SK
578 switch (protocol) {
579 case IPPROTO_ESP:
580 esph = (struct ip_esp_hdr *)(skb->data + offset);
581 spi = esph->spi;
582 break;
583 case IPPROTO_AH:
584 ah = (struct ip_auth_hdr *)(skb->data + offset);
585 spi = ah->spi;
586 break;
587 case IPPROTO_COMP:
588 ipch = (struct ip_comp_hdr *)(skb->data + offset);
589 spi = htonl(ntohs(ipch->cpi));
590 break;
591 default:
592 return 0;
593 }
594
595 if (type != ICMPV6_PKT_TOOBIG &&
596 type != NDISC_REDIRECT)
597 return 0;
598
6d004d6c 599 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
fa9ad96d
SK
600 spi, protocol, AF_INET6);
601 if (!x)
602 return 0;
603
604 if (type == NDISC_REDIRECT)
344afd62
LC
605 ip6_redirect(skb, net, skb->dev->ifindex, 0,
606 sock_net_uid(net, NULL));
fa9ad96d 607 else
344afd62 608 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
fa9ad96d
SK
609 xfrm_state_put(x);
610
611 return 0;
612}
613
ed1efb2a
SK
614static void vti6_link_config(struct ip6_tnl *t)
615{
ed1efb2a
SK
616 struct net_device *dev = t->dev;
617 struct __ip6_tnl_parm *p = &t->parms;
ed1efb2a
SK
618
619 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
620 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
621
ed1efb2a
SK
622 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
623 IP6_TNL_F_CAP_PER_PACKET);
624 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
625
626 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
627 dev->flags |= IFF_POINTOPOINT;
628 else
629 dev->flags &= ~IFF_POINTOPOINT;
ed1efb2a
SK
630}
631
632/**
633 * vti6_tnl_change - update the tunnel parameters
634 * @t: tunnel to be changed
635 * @p: tunnel configuration parameters
636 *
637 * Description:
638 * vti6_tnl_change() updates the tunnel parameters
639 **/
640static int
641vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
642{
643 t->parms.laddr = p->laddr;
644 t->parms.raddr = p->raddr;
645 t->parms.link = p->link;
646 t->parms.i_key = p->i_key;
647 t->parms.o_key = p->o_key;
648 t->parms.proto = p->proto;
649 ip6_tnl_dst_reset(t);
650 vti6_link_config(t);
651 return 0;
652}
653
654static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
655{
656 struct net *net = dev_net(t->dev);
657 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
658 int err;
659
660 vti6_tnl_unlink(ip6n, t);
661 synchronize_net();
662 err = vti6_tnl_change(t, p);
663 vti6_tnl_link(ip6n, t);
664 netdev_state_change(t->dev);
665 return err;
666}
667
668static void
669vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
670{
671 p->laddr = u->laddr;
672 p->raddr = u->raddr;
673 p->link = u->link;
674 p->i_key = u->i_key;
675 p->o_key = u->o_key;
676 p->proto = u->proto;
677
678 memcpy(p->name, u->name, sizeof(u->name));
679}
680
681static void
682vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
683{
684 u->laddr = p->laddr;
685 u->raddr = p->raddr;
686 u->link = p->link;
687 u->i_key = p->i_key;
688 u->o_key = p->o_key;
f1b3aae1
DF
689 if (u->i_key)
690 u->i_flags |= GRE_KEY;
691 if (u->o_key)
692 u->o_flags |= GRE_KEY;
ed1efb2a
SK
693 u->proto = p->proto;
694
695 memcpy(u->name, p->name, sizeof(u->name));
696}
697
698/**
699 * vti6_tnl_ioctl - configure vti6 tunnels from userspace
700 * @dev: virtual device associated with tunnel
701 * @ifr: parameters passed from userspace
702 * @cmd: command to be performed
703 *
704 * Description:
705 * vti6_ioctl() is used for managing vti6 tunnels
706 * from userspace.
707 *
708 * The possible commands are the following:
709 * %SIOCGETTUNNEL: get tunnel parameters for device
710 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
711 * %SIOCCHGTUNNEL: change tunnel parameters to those given
712 * %SIOCDELTUNNEL: delete tunnel
713 *
714 * The fallback device "ip6_vti0", created during module
715 * initialization, can be used for creating other tunnel devices.
716 *
717 * Return:
718 * 0 on success,
719 * %-EFAULT if unable to copy data to or from userspace,
720 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
721 * %-EINVAL if passed tunnel parameters are invalid,
722 * %-EEXIST if changing a tunnel's parameters would cause a conflict
723 * %-ENODEV if attempting to change or delete a nonexisting device
724 **/
725static int
726vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
727{
728 int err = 0;
729 struct ip6_tnl_parm2 p;
730 struct __ip6_tnl_parm p1;
731 struct ip6_tnl *t = NULL;
732 struct net *net = dev_net(dev);
733 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
734
735 switch (cmd) {
736 case SIOCGETTUNNEL:
737 if (dev == ip6n->fb_tnl_dev) {
738 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
739 err = -EFAULT;
740 break;
741 }
742 vti6_parm_from_user(&p1, &p);
743 t = vti6_locate(net, &p1, 0);
744 } else {
745 memset(&p, 0, sizeof(p));
746 }
63159f29 747 if (!t)
ed1efb2a
SK
748 t = netdev_priv(dev);
749 vti6_parm_to_user(&p, &t->parms);
750 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
751 err = -EFAULT;
752 break;
753 case SIOCADDTUNNEL:
754 case SIOCCHGTUNNEL:
755 err = -EPERM;
756 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
757 break;
758 err = -EFAULT;
759 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
760 break;
761 err = -EINVAL;
762 if (p.proto != IPPROTO_IPV6 && p.proto != 0)
763 break;
764 vti6_parm_from_user(&p1, &p);
765 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
766 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
53b24b8f 767 if (t) {
ed1efb2a
SK
768 if (t->dev != dev) {
769 err = -EEXIST;
770 break;
771 }
772 } else
773 t = netdev_priv(dev);
774
775 err = vti6_update(t, &p1);
776 }
777 if (t) {
778 err = 0;
779 vti6_parm_to_user(&p, &t->parms);
780 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
781 err = -EFAULT;
782
783 } else
784 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
785 break;
786 case SIOCDELTUNNEL:
787 err = -EPERM;
788 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
789 break;
790
791 if (dev == ip6n->fb_tnl_dev) {
792 err = -EFAULT;
793 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
794 break;
795 err = -ENOENT;
796 vti6_parm_from_user(&p1, &p);
797 t = vti6_locate(net, &p1, 0);
63159f29 798 if (!t)
ed1efb2a
SK
799 break;
800 err = -EPERM;
801 if (t->dev == ip6n->fb_tnl_dev)
802 break;
803 dev = t->dev;
804 }
805 err = 0;
806 unregister_netdevice(dev);
807 break;
808 default:
809 err = -EINVAL;
810 }
811 return err;
812}
813
814/**
815 * vti6_tnl_change_mtu - change mtu manually for tunnel device
816 * @dev: virtual device associated with tunnel
817 * @new_mtu: the new mtu
818 *
819 * Return:
820 * 0 on success,
821 * %-EINVAL if mtu too small
822 **/
823static int vti6_change_mtu(struct net_device *dev, int new_mtu)
824{
825 if (new_mtu < IPV6_MIN_MTU)
826 return -EINVAL;
827
828 dev->mtu = new_mtu;
829 return 0;
830}
831
832static const struct net_device_ops vti6_netdev_ops = {
16a0231b 833 .ndo_init = vti6_dev_init,
ed1efb2a
SK
834 .ndo_uninit = vti6_dev_uninit,
835 .ndo_start_xmit = vti6_tnl_xmit,
836 .ndo_do_ioctl = vti6_ioctl,
837 .ndo_change_mtu = vti6_change_mtu,
469bdcef 838 .ndo_get_stats64 = ip_tunnel_get_stats64,
ecf2c06a 839 .ndo_get_iflink = ip6_tnl_get_iflink,
ed1efb2a
SK
840};
841
842/**
843 * vti6_dev_setup - setup virtual tunnel device
844 * @dev: virtual device associated with tunnel
845 *
846 * Description:
847 * Initialize function pointers and device parameters
848 **/
849static void vti6_dev_setup(struct net_device *dev)
850{
ed1efb2a
SK
851 dev->netdev_ops = &vti6_netdev_ops;
852 dev->destructor = vti6_dev_free;
853
854 dev->type = ARPHRD_TUNNEL6;
855 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
856 dev->mtu = ETH_DATA_LEN;
ed1efb2a
SK
857 dev->flags |= IFF_NOARP;
858 dev->addr_len = sizeof(struct in6_addr);
02875878 859 netif_keep_dst(dev);
ed1efb2a
SK
860}
861
862/**
863 * vti6_dev_init_gen - general initializer for all tunnel devices
864 * @dev: virtual device associated with tunnel
865 **/
866static inline int vti6_dev_init_gen(struct net_device *dev)
867{
868 struct ip6_tnl *t = netdev_priv(dev);
869
870 t->dev = dev;
871 t->net = dev_net(dev);
1c213bd2 872 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
ed1efb2a
SK
873 if (!dev->tstats)
874 return -ENOMEM;
875 return 0;
876}
877
878/**
879 * vti6_dev_init - initializer for all non fallback tunnel devices
880 * @dev: virtual device associated with tunnel
881 **/
882static int vti6_dev_init(struct net_device *dev)
883{
884 struct ip6_tnl *t = netdev_priv(dev);
885 int err = vti6_dev_init_gen(dev);
886
887 if (err)
888 return err;
889 vti6_link_config(t);
890 return 0;
891}
892
893/**
894 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
895 * @dev: fallback device
896 *
897 * Return: 0
898 **/
899static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
900{
901 struct ip6_tnl *t = netdev_priv(dev);
902 struct net *net = dev_net(dev);
903 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
ed1efb2a
SK
904
905 t->parms.proto = IPPROTO_IPV6;
906 dev_hold(dev);
907
ed1efb2a
SK
908 rcu_assign_pointer(ip6n->tnls_wc[0], t);
909 return 0;
910}
911
912static int vti6_validate(struct nlattr *tb[], struct nlattr *data[])
913{
914 return 0;
915}
916
917static void vti6_netlink_parms(struct nlattr *data[],
918 struct __ip6_tnl_parm *parms)
919{
920 memset(parms, 0, sizeof(*parms));
921
922 if (!data)
923 return;
924
925 if (data[IFLA_VTI_LINK])
926 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
927
928 if (data[IFLA_VTI_LOCAL])
67b61f6c 929 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]);
ed1efb2a
SK
930
931 if (data[IFLA_VTI_REMOTE])
67b61f6c 932 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]);
ed1efb2a
SK
933
934 if (data[IFLA_VTI_IKEY])
935 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
936
937 if (data[IFLA_VTI_OKEY])
938 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
939}
940
941static int vti6_newlink(struct net *src_net, struct net_device *dev,
942 struct nlattr *tb[], struct nlattr *data[])
943{
944 struct net *net = dev_net(dev);
945 struct ip6_tnl *nt;
946
947 nt = netdev_priv(dev);
948 vti6_netlink_parms(data, &nt->parms);
949
950 nt->parms.proto = IPPROTO_IPV6;
951
952 if (vti6_locate(net, &nt->parms, 0))
953 return -EEXIST;
954
955 return vti6_tnl_create2(dev);
956}
957
20ea60ca 958static void vti6_dellink(struct net_device *dev, struct list_head *head)
959{
960 struct net *net = dev_net(dev);
961 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
962
963 if (dev != ip6n->fb_tnl_dev)
964 unregister_netdevice_queue(dev, head);
965}
966
ed1efb2a
SK
967static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
968 struct nlattr *data[])
969{
970 struct ip6_tnl *t;
971 struct __ip6_tnl_parm p;
972 struct net *net = dev_net(dev);
973 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
974
975 if (dev == ip6n->fb_tnl_dev)
976 return -EINVAL;
977
978 vti6_netlink_parms(data, &p);
979
980 t = vti6_locate(net, &p, 0);
981
982 if (t) {
983 if (t->dev != dev)
984 return -EEXIST;
985 } else
986 t = netdev_priv(dev);
987
988 return vti6_update(t, &p);
989}
990
991static size_t vti6_get_size(const struct net_device *dev)
992{
993 return
994 /* IFLA_VTI_LINK */
995 nla_total_size(4) +
996 /* IFLA_VTI_LOCAL */
997 nla_total_size(sizeof(struct in6_addr)) +
998 /* IFLA_VTI_REMOTE */
999 nla_total_size(sizeof(struct in6_addr)) +
1000 /* IFLA_VTI_IKEY */
1001 nla_total_size(4) +
1002 /* IFLA_VTI_OKEY */
1003 nla_total_size(4) +
1004 0;
1005}
1006
1007static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1008{
1009 struct ip6_tnl *tunnel = netdev_priv(dev);
1010 struct __ip6_tnl_parm *parm = &tunnel->parms;
1011
1012 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
930345ea
JB
1013 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) ||
1014 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) ||
ed1efb2a
SK
1015 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
1016 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key))
1017 goto nla_put_failure;
1018 return 0;
1019
1020nla_put_failure:
1021 return -EMSGSIZE;
1022}
1023
1024static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1025 [IFLA_VTI_LINK] = { .type = NLA_U32 },
1026 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) },
1027 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) },
1028 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
1029 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
1030};
1031
1032static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1033 .kind = "vti6",
1034 .maxtype = IFLA_VTI_MAX,
1035 .policy = vti6_policy,
1036 .priv_size = sizeof(struct ip6_tnl),
1037 .setup = vti6_dev_setup,
1038 .validate = vti6_validate,
1039 .newlink = vti6_newlink,
20ea60ca 1040 .dellink = vti6_dellink,
ed1efb2a
SK
1041 .changelink = vti6_changelink,
1042 .get_size = vti6_get_size,
1043 .fill_info = vti6_fill_info,
1728d4fa 1044 .get_link_net = ip6_tnl_get_link_net,
ed1efb2a
SK
1045};
1046
ed1efb2a
SK
1047static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n)
1048{
1049 int h;
1050 struct ip6_tnl *t;
1051 LIST_HEAD(list);
1052
1053 for (h = 0; h < HASH_SIZE; h++) {
1054 t = rtnl_dereference(ip6n->tnls_r_l[h]);
53b24b8f 1055 while (t) {
ed1efb2a
SK
1056 unregister_netdevice_queue(t->dev, &list);
1057 t = rtnl_dereference(t->next);
1058 }
1059 }
1060
1061 t = rtnl_dereference(ip6n->tnls_wc[0]);
1062 unregister_netdevice_queue(t->dev, &list);
1063 unregister_netdevice_many(&list);
1064}
1065
1066static int __net_init vti6_init_net(struct net *net)
1067{
1068 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1069 struct ip6_tnl *t = NULL;
1070 int err;
1071
1072 ip6n->tnls[0] = ip6n->tnls_wc;
1073 ip6n->tnls[1] = ip6n->tnls_r_l;
1074
1075 err = -ENOMEM;
1076 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
c835a677 1077 NET_NAME_UNKNOWN, vti6_dev_setup);
ed1efb2a
SK
1078
1079 if (!ip6n->fb_tnl_dev)
1080 goto err_alloc_dev;
1081 dev_net_set(ip6n->fb_tnl_dev, net);
20ea60ca 1082 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
ed1efb2a
SK
1083
1084 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1085 if (err < 0)
1086 goto err_register;
1087
1088 err = register_netdev(ip6n->fb_tnl_dev);
1089 if (err < 0)
1090 goto err_register;
1091
1092 t = netdev_priv(ip6n->fb_tnl_dev);
1093
1094 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1095 return 0;
1096
1097err_register:
1098 vti6_dev_free(ip6n->fb_tnl_dev);
1099err_alloc_dev:
1100 return err;
1101}
1102
1103static void __net_exit vti6_exit_net(struct net *net)
1104{
1105 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1106
1107 rtnl_lock();
1108 vti6_destroy_tunnels(ip6n);
1109 rtnl_unlock();
1110}
1111
1112static struct pernet_operations vti6_net_ops = {
1113 .init = vti6_init_net,
1114 .exit = vti6_exit_net,
1115 .id = &vti6_net_id,
1116 .size = sizeof(struct vti6_net),
1117};
1118
fa9ad96d
SK
1119static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1120 .handler = vti6_rcv,
1121 .cb_handler = vti6_rcv_cb,
1122 .err_handler = vti6_err,
1123 .priority = 100,
1124};
1125
1126static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1127 .handler = vti6_rcv,
1128 .cb_handler = vti6_rcv_cb,
1129 .err_handler = vti6_err,
1130 .priority = 100,
1131};
1132
1133static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1134 .handler = vti6_rcv,
1135 .cb_handler = vti6_rcv_cb,
1136 .err_handler = vti6_err,
1137 .priority = 100,
1138};
1139
ed1efb2a
SK
1140/**
1141 * vti6_tunnel_init - register protocol and reserve needed resources
1142 *
1143 * Return: 0 on success
1144 **/
1145static int __init vti6_tunnel_init(void)
1146{
e59d82fd
MK
1147 const char *msg;
1148 int err;
ed1efb2a 1149
e59d82fd 1150 msg = "tunnel device";
ed1efb2a
SK
1151 err = register_pernet_device(&vti6_net_ops);
1152 if (err < 0)
e59d82fd 1153 goto pernet_dev_failed;
ed1efb2a 1154
e59d82fd 1155 msg = "tunnel protocols";
fa9ad96d 1156 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
e59d82fd
MK
1157 if (err < 0)
1158 goto xfrm_proto_esp_failed;
fa9ad96d 1159 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
e59d82fd
MK
1160 if (err < 0)
1161 goto xfrm_proto_ah_failed;
fa9ad96d 1162 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
e59d82fd
MK
1163 if (err < 0)
1164 goto xfrm_proto_comp_failed;
fa9ad96d 1165
e59d82fd 1166 msg = "netlink interface";
ed1efb2a
SK
1167 err = rtnl_link_register(&vti6_link_ops);
1168 if (err < 0)
1169 goto rtnl_link_failed;
1170
1171 return 0;
1172
1173rtnl_link_failed:
fa9ad96d 1174 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
e59d82fd 1175xfrm_proto_comp_failed:
fa9ad96d 1176 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
e59d82fd 1177xfrm_proto_ah_failed:
fa9ad96d 1178 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
e59d82fd 1179xfrm_proto_esp_failed:
ed1efb2a 1180 unregister_pernet_device(&vti6_net_ops);
e59d82fd
MK
1181pernet_dev_failed:
1182 pr_err("vti6 init: failed to register %s\n", msg);
ed1efb2a
SK
1183 return err;
1184}
1185
1186/**
1187 * vti6_tunnel_cleanup - free resources and unregister protocol
1188 **/
1189static void __exit vti6_tunnel_cleanup(void)
1190{
1191 rtnl_link_unregister(&vti6_link_ops);
e59d82fd
MK
1192 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1193 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1194 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
ed1efb2a
SK
1195 unregister_pernet_device(&vti6_net_ops);
1196}
1197
1198module_init(vti6_tunnel_init);
1199module_exit(vti6_tunnel_cleanup);
1200MODULE_LICENSE("GPL");
1201MODULE_ALIAS_RTNL_LINK("vti6");
1202MODULE_ALIAS_NETDEV("ip6_vti0");
1203MODULE_AUTHOR("Steffen Klassert");
1204MODULE_DESCRIPTION("IPv6 virtual tunnel interface");