Loading...
1/*
2 * IPv6 tunneling device
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Ville Nuorvala <vnuorval@tcs.hut.fi>
7 * Yasuyuki Kozakai <kozakai@linux-ipv6.org>
8 *
9 * Based on:
10 * linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
11 *
12 * RFC 2473
13 *
14 * This program is free software; you can redistribute it and/or
15 * modify it under the terms of the GNU General Public License
16 * as published by the Free Software Foundation; either version
17 * 2 of the License, or (at your option) any later version.
18 *
19 */
20
21#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
22
23#include <linux/module.h>
24#include <linux/capability.h>
25#include <linux/errno.h>
26#include <linux/types.h>
27#include <linux/sockios.h>
28#include <linux/icmp.h>
29#include <linux/if.h>
30#include <linux/in.h>
31#include <linux/ip.h>
32#include <linux/net.h>
33#include <linux/in6.h>
34#include <linux/netdevice.h>
35#include <linux/if_arp.h>
36#include <linux/icmpv6.h>
37#include <linux/init.h>
38#include <linux/route.h>
39#include <linux/rtnetlink.h>
40#include <linux/netfilter_ipv6.h>
41#include <linux/slab.h>
42#include <linux/hash.h>
43#include <linux/etherdevice.h>
44
45#include <asm/uaccess.h>
46#include <linux/atomic.h>
47
48#include <net/icmp.h>
49#include <net/ip.h>
50#include <net/ip_tunnels.h>
51#include <net/ipv6.h>
52#include <net/ip6_route.h>
53#include <net/addrconf.h>
54#include <net/ip6_tunnel.h>
55#include <net/xfrm.h>
56#include <net/dsfield.h>
57#include <net/inet_ecn.h>
58#include <net/net_namespace.h>
59#include <net/netns/generic.h>
60
61MODULE_AUTHOR("Ville Nuorvala");
62MODULE_DESCRIPTION("IPv6 tunneling device");
63MODULE_LICENSE("GPL");
64MODULE_ALIAS_RTNL_LINK("ip6tnl");
65MODULE_ALIAS_NETDEV("ip6tnl0");
66
67#define HASH_SIZE_SHIFT 5
68#define HASH_SIZE (1 << HASH_SIZE_SHIFT)
69
70static bool log_ecn_error = true;
71module_param(log_ecn_error, bool, 0644);
72MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
73
74static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
75{
76 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
77
78 return hash_32(hash, HASH_SIZE_SHIFT);
79}
80
81static int ip6_tnl_dev_init(struct net_device *dev);
82static void ip6_tnl_dev_setup(struct net_device *dev);
83static struct rtnl_link_ops ip6_link_ops __read_mostly;
84
85static int ip6_tnl_net_id __read_mostly;
86struct ip6_tnl_net {
87 /* the IPv6 tunnel fallback device */
88 struct net_device *fb_tnl_dev;
89 /* lists for storing tunnels in use */
90 struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
91 struct ip6_tnl __rcu *tnls_wc[1];
92 struct ip6_tnl __rcu **tnls[2];
93};
94
95static struct net_device_stats *ip6_get_stats(struct net_device *dev)
96{
97 struct pcpu_sw_netstats tmp, sum = { 0 };
98 int i;
99
100 for_each_possible_cpu(i) {
101 unsigned int start;
102 const struct pcpu_sw_netstats *tstats =
103 per_cpu_ptr(dev->tstats, i);
104
105 do {
106 start = u64_stats_fetch_begin_irq(&tstats->syncp);
107 tmp.rx_packets = tstats->rx_packets;
108 tmp.rx_bytes = tstats->rx_bytes;
109 tmp.tx_packets = tstats->tx_packets;
110 tmp.tx_bytes = tstats->tx_bytes;
111 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
112
113 sum.rx_packets += tmp.rx_packets;
114 sum.rx_bytes += tmp.rx_bytes;
115 sum.tx_packets += tmp.tx_packets;
116 sum.tx_bytes += tmp.tx_bytes;
117 }
118 dev->stats.rx_packets = sum.rx_packets;
119 dev->stats.rx_bytes = sum.rx_bytes;
120 dev->stats.tx_packets = sum.tx_packets;
121 dev->stats.tx_bytes = sum.tx_bytes;
122 return &dev->stats;
123}
124
125/**
126 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
127 * @remote: the address of the tunnel exit-point
128 * @local: the address of the tunnel entry-point
129 *
130 * Return:
131 * tunnel matching given end-points if found,
132 * else fallback tunnel if its device is up,
133 * else %NULL
134 **/
135
136#define for_each_ip6_tunnel_rcu(start) \
137 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
138
139static struct ip6_tnl *
140ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
141{
142 unsigned int hash = HASH(remote, local);
143 struct ip6_tnl *t;
144 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
145 struct in6_addr any;
146
147 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
148 if (ipv6_addr_equal(local, &t->parms.laddr) &&
149 ipv6_addr_equal(remote, &t->parms.raddr) &&
150 (t->dev->flags & IFF_UP))
151 return t;
152 }
153
154 memset(&any, 0, sizeof(any));
155 hash = HASH(&any, local);
156 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
157 if (ipv6_addr_equal(local, &t->parms.laddr) &&
158 (t->dev->flags & IFF_UP))
159 return t;
160 }
161
162 hash = HASH(remote, &any);
163 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
164 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
165 (t->dev->flags & IFF_UP))
166 return t;
167 }
168
169 t = rcu_dereference(ip6n->tnls_wc[0]);
170 if (t && (t->dev->flags & IFF_UP))
171 return t;
172
173 return NULL;
174}
175
176/**
177 * ip6_tnl_bucket - get head of list matching given tunnel parameters
178 * @p: parameters containing tunnel end-points
179 *
180 * Description:
181 * ip6_tnl_bucket() returns the head of the list matching the
182 * &struct in6_addr entries laddr and raddr in @p.
183 *
184 * Return: head of IPv6 tunnel list
185 **/
186
187static struct ip6_tnl __rcu **
188ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
189{
190 const struct in6_addr *remote = &p->raddr;
191 const struct in6_addr *local = &p->laddr;
192 unsigned int h = 0;
193 int prio = 0;
194
195 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
196 prio = 1;
197 h = HASH(remote, local);
198 }
199 return &ip6n->tnls[prio][h];
200}
201
202/**
203 * ip6_tnl_link - add tunnel to hash table
204 * @t: tunnel to be added
205 **/
206
207static void
208ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
209{
210 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
211
212 rcu_assign_pointer(t->next , rtnl_dereference(*tp));
213 rcu_assign_pointer(*tp, t);
214}
215
216/**
217 * ip6_tnl_unlink - remove tunnel from hash table
218 * @t: tunnel to be removed
219 **/
220
221static void
222ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
223{
224 struct ip6_tnl __rcu **tp;
225 struct ip6_tnl *iter;
226
227 for (tp = ip6_tnl_bucket(ip6n, &t->parms);
228 (iter = rtnl_dereference(*tp)) != NULL;
229 tp = &iter->next) {
230 if (t == iter) {
231 rcu_assign_pointer(*tp, t->next);
232 break;
233 }
234 }
235}
236
237static void ip6_dev_free(struct net_device *dev)
238{
239 struct ip6_tnl *t = netdev_priv(dev);
240
241 dst_cache_destroy(&t->dst_cache);
242 free_percpu(dev->tstats);
243 free_netdev(dev);
244}
245
246static int ip6_tnl_create2(struct net_device *dev)
247{
248 struct ip6_tnl *t = netdev_priv(dev);
249 struct net *net = dev_net(dev);
250 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
251 int err;
252
253 t = netdev_priv(dev);
254
255 dev->rtnl_link_ops = &ip6_link_ops;
256 err = register_netdevice(dev);
257 if (err < 0)
258 goto out;
259
260 strcpy(t->parms.name, dev->name);
261
262 dev_hold(dev);
263 ip6_tnl_link(ip6n, t);
264 return 0;
265
266out:
267 return err;
268}
269
270/**
271 * ip6_tnl_create - create a new tunnel
272 * @p: tunnel parameters
273 * @pt: pointer to new tunnel
274 *
275 * Description:
276 * Create tunnel matching given parameters.
277 *
278 * Return:
279 * created tunnel or error pointer
280 **/
281
282static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
283{
284 struct net_device *dev;
285 struct ip6_tnl *t;
286 char name[IFNAMSIZ];
287 int err = -ENOMEM;
288
289 if (p->name[0])
290 strlcpy(name, p->name, IFNAMSIZ);
291 else
292 sprintf(name, "ip6tnl%%d");
293
294 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
295 ip6_tnl_dev_setup);
296 if (!dev)
297 goto failed;
298
299 dev_net_set(dev, net);
300
301 t = netdev_priv(dev);
302 t->parms = *p;
303 t->net = dev_net(dev);
304 err = ip6_tnl_create2(dev);
305 if (err < 0)
306 goto failed_free;
307
308 return t;
309
310failed_free:
311 ip6_dev_free(dev);
312failed:
313 return ERR_PTR(err);
314}
315
316/**
317 * ip6_tnl_locate - find or create tunnel matching given parameters
318 * @p: tunnel parameters
319 * @create: != 0 if allowed to create new tunnel if no match found
320 *
321 * Description:
322 * ip6_tnl_locate() first tries to locate an existing tunnel
323 * based on @parms. If this is unsuccessful, but @create is set a new
324 * tunnel device is created and registered for use.
325 *
326 * Return:
327 * matching tunnel or error pointer
328 **/
329
330static struct ip6_tnl *ip6_tnl_locate(struct net *net,
331 struct __ip6_tnl_parm *p, int create)
332{
333 const struct in6_addr *remote = &p->raddr;
334 const struct in6_addr *local = &p->laddr;
335 struct ip6_tnl __rcu **tp;
336 struct ip6_tnl *t;
337 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
338
339 for (tp = ip6_tnl_bucket(ip6n, p);
340 (t = rtnl_dereference(*tp)) != NULL;
341 tp = &t->next) {
342 if (ipv6_addr_equal(local, &t->parms.laddr) &&
343 ipv6_addr_equal(remote, &t->parms.raddr)) {
344 if (create)
345 return ERR_PTR(-EEXIST);
346
347 return t;
348 }
349 }
350 if (!create)
351 return ERR_PTR(-ENODEV);
352 return ip6_tnl_create(net, p);
353}
354
355/**
356 * ip6_tnl_dev_uninit - tunnel device uninitializer
357 * @dev: the device to be destroyed
358 *
359 * Description:
360 * ip6_tnl_dev_uninit() removes tunnel from its list
361 **/
362
363static void
364ip6_tnl_dev_uninit(struct net_device *dev)
365{
366 struct ip6_tnl *t = netdev_priv(dev);
367 struct net *net = t->net;
368 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
369
370 if (dev == ip6n->fb_tnl_dev)
371 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
372 else
373 ip6_tnl_unlink(ip6n, t);
374 dst_cache_reset(&t->dst_cache);
375 dev_put(dev);
376}
377
378/**
379 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
380 * @skb: received socket buffer
381 *
382 * Return:
383 * 0 if none was found,
384 * else index to encapsulation limit
385 **/
386
387__u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
388{
389 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
390 __u8 nexthdr = ipv6h->nexthdr;
391 __u16 off = sizeof(*ipv6h);
392
393 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
394 __u16 optlen = 0;
395 struct ipv6_opt_hdr *hdr;
396 if (raw + off + sizeof(*hdr) > skb->data &&
397 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
398 break;
399
400 hdr = (struct ipv6_opt_hdr *) (raw + off);
401 if (nexthdr == NEXTHDR_FRAGMENT) {
402 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
403 if (frag_hdr->frag_off)
404 break;
405 optlen = 8;
406 } else if (nexthdr == NEXTHDR_AUTH) {
407 optlen = (hdr->hdrlen + 2) << 2;
408 } else {
409 optlen = ipv6_optlen(hdr);
410 }
411 if (nexthdr == NEXTHDR_DEST) {
412 __u16 i = off + 2;
413 while (1) {
414 struct ipv6_tlv_tnl_enc_lim *tel;
415
416 /* No more room for encapsulation limit */
417 if (i + sizeof (*tel) > off + optlen)
418 break;
419
420 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
421 /* return index of option if found and valid */
422 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
423 tel->length == 1)
424 return i;
425 /* else jump to next option */
426 if (tel->type)
427 i += tel->length + 2;
428 else
429 i++;
430 }
431 }
432 nexthdr = hdr->nexthdr;
433 off += optlen;
434 }
435 return 0;
436}
437EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
438
439/**
440 * ip6_tnl_err - tunnel error handler
441 *
442 * Description:
443 * ip6_tnl_err() should handle errors in the tunnel according
444 * to the specifications in RFC 2473.
445 **/
446
447static int
448ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
449 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
450{
451 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
452 struct ip6_tnl *t;
453 int rel_msg = 0;
454 u8 rel_type = ICMPV6_DEST_UNREACH;
455 u8 rel_code = ICMPV6_ADDR_UNREACH;
456 u8 tproto;
457 __u32 rel_info = 0;
458 __u16 len;
459 int err = -ENOENT;
460
461 /* If the packet doesn't contain the original IPv6 header we are
462 in trouble since we might need the source address for further
463 processing of the error. */
464
465 rcu_read_lock();
466 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
467 if (!t)
468 goto out;
469
470 tproto = ACCESS_ONCE(t->parms.proto);
471 if (tproto != ipproto && tproto != 0)
472 goto out;
473
474 err = 0;
475
476 switch (*type) {
477 __u32 teli;
478 struct ipv6_tlv_tnl_enc_lim *tel;
479 __u32 mtu;
480 case ICMPV6_DEST_UNREACH:
481 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
482 t->parms.name);
483 rel_msg = 1;
484 break;
485 case ICMPV6_TIME_EXCEED:
486 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
487 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
488 t->parms.name);
489 rel_msg = 1;
490 }
491 break;
492 case ICMPV6_PARAMPROB:
493 teli = 0;
494 if ((*code) == ICMPV6_HDR_FIELD)
495 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
496
497 if (teli && teli == *info - 2) {
498 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
499 if (tel->encap_limit == 0) {
500 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
501 t->parms.name);
502 rel_msg = 1;
503 }
504 } else {
505 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
506 t->parms.name);
507 }
508 break;
509 case ICMPV6_PKT_TOOBIG:
510 mtu = *info - offset;
511 if (mtu < IPV6_MIN_MTU)
512 mtu = IPV6_MIN_MTU;
513 t->dev->mtu = mtu;
514
515 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
516 if (len > mtu) {
517 rel_type = ICMPV6_PKT_TOOBIG;
518 rel_code = 0;
519 rel_info = mtu;
520 rel_msg = 1;
521 }
522 break;
523 }
524
525 *type = rel_type;
526 *code = rel_code;
527 *info = rel_info;
528 *msg = rel_msg;
529
530out:
531 rcu_read_unlock();
532 return err;
533}
534
535static int
536ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
537 u8 type, u8 code, int offset, __be32 info)
538{
539 int rel_msg = 0;
540 u8 rel_type = type;
541 u8 rel_code = code;
542 __u32 rel_info = ntohl(info);
543 int err;
544 struct sk_buff *skb2;
545 const struct iphdr *eiph;
546 struct rtable *rt;
547 struct flowi4 fl4;
548
549 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
550 &rel_msg, &rel_info, offset);
551 if (err < 0)
552 return err;
553
554 if (rel_msg == 0)
555 return 0;
556
557 switch (rel_type) {
558 case ICMPV6_DEST_UNREACH:
559 if (rel_code != ICMPV6_ADDR_UNREACH)
560 return 0;
561 rel_type = ICMP_DEST_UNREACH;
562 rel_code = ICMP_HOST_UNREACH;
563 break;
564 case ICMPV6_PKT_TOOBIG:
565 if (rel_code != 0)
566 return 0;
567 rel_type = ICMP_DEST_UNREACH;
568 rel_code = ICMP_FRAG_NEEDED;
569 break;
570 case NDISC_REDIRECT:
571 rel_type = ICMP_REDIRECT;
572 rel_code = ICMP_REDIR_HOST;
573 default:
574 return 0;
575 }
576
577 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
578 return 0;
579
580 skb2 = skb_clone(skb, GFP_ATOMIC);
581 if (!skb2)
582 return 0;
583
584 skb_dst_drop(skb2);
585
586 skb_pull(skb2, offset);
587 skb_reset_network_header(skb2);
588 eiph = ip_hdr(skb2);
589
590 /* Try to guess incoming interface */
591 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
592 eiph->saddr, 0,
593 0, 0,
594 IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
595 if (IS_ERR(rt))
596 goto out;
597
598 skb2->dev = rt->dst.dev;
599
600 /* route "incoming" packet */
601 if (rt->rt_flags & RTCF_LOCAL) {
602 ip_rt_put(rt);
603 rt = NULL;
604 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
605 eiph->daddr, eiph->saddr,
606 0, 0,
607 IPPROTO_IPIP,
608 RT_TOS(eiph->tos), 0);
609 if (IS_ERR(rt) ||
610 rt->dst.dev->type != ARPHRD_TUNNEL) {
611 if (!IS_ERR(rt))
612 ip_rt_put(rt);
613 goto out;
614 }
615 skb_dst_set(skb2, &rt->dst);
616 } else {
617 ip_rt_put(rt);
618 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
619 skb2->dev) ||
620 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
621 goto out;
622 }
623
624 /* change mtu on this route */
625 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
626 if (rel_info > dst_mtu(skb_dst(skb2)))
627 goto out;
628
629 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
630 }
631 if (rel_type == ICMP_REDIRECT)
632 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
633
634 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
635
636out:
637 kfree_skb(skb2);
638 return 0;
639}
640
641static int
642ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
643 u8 type, u8 code, int offset, __be32 info)
644{
645 int rel_msg = 0;
646 u8 rel_type = type;
647 u8 rel_code = code;
648 __u32 rel_info = ntohl(info);
649 int err;
650
651 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
652 &rel_msg, &rel_info, offset);
653 if (err < 0)
654 return err;
655
656 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
657 struct rt6_info *rt;
658 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
659
660 if (!skb2)
661 return 0;
662
663 skb_dst_drop(skb2);
664 skb_pull(skb2, offset);
665 skb_reset_network_header(skb2);
666
667 /* Try to guess incoming interface */
668 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
669 NULL, 0, 0);
670
671 if (rt && rt->dst.dev)
672 skb2->dev = rt->dst.dev;
673
674 icmpv6_send(skb2, rel_type, rel_code, rel_info);
675
676 ip6_rt_put(rt);
677
678 kfree_skb(skb2);
679 }
680
681 return 0;
682}
683
684static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
685 const struct ipv6hdr *ipv6h,
686 struct sk_buff *skb)
687{
688 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
689
690 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
691 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
692
693 return IP6_ECN_decapsulate(ipv6h, skb);
694}
695
696static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
697 const struct ipv6hdr *ipv6h,
698 struct sk_buff *skb)
699{
700 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
701 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
702
703 return IP6_ECN_decapsulate(ipv6h, skb);
704}
705
706__u32 ip6_tnl_get_cap(struct ip6_tnl *t,
707 const struct in6_addr *laddr,
708 const struct in6_addr *raddr)
709{
710 struct __ip6_tnl_parm *p = &t->parms;
711 int ltype = ipv6_addr_type(laddr);
712 int rtype = ipv6_addr_type(raddr);
713 __u32 flags = 0;
714
715 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
716 flags = IP6_TNL_F_CAP_PER_PACKET;
717 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
718 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
719 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
720 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
721 if (ltype&IPV6_ADDR_UNICAST)
722 flags |= IP6_TNL_F_CAP_XMIT;
723 if (rtype&IPV6_ADDR_UNICAST)
724 flags |= IP6_TNL_F_CAP_RCV;
725 }
726 return flags;
727}
728EXPORT_SYMBOL(ip6_tnl_get_cap);
729
730/* called with rcu_read_lock() */
731int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
732 const struct in6_addr *laddr,
733 const struct in6_addr *raddr)
734{
735 struct __ip6_tnl_parm *p = &t->parms;
736 int ret = 0;
737 struct net *net = t->net;
738
739 if ((p->flags & IP6_TNL_F_CAP_RCV) ||
740 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
741 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
742 struct net_device *ldev = NULL;
743
744 if (p->link)
745 ldev = dev_get_by_index_rcu(net, p->link);
746
747 if ((ipv6_addr_is_multicast(laddr) ||
748 likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
749 likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
750 ret = 1;
751 }
752 return ret;
753}
754EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
755
756/**
757 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
758 * @skb: received socket buffer
759 * @protocol: ethernet protocol ID
760 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
761 *
762 * Return: 0
763 **/
764
765static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
766 __u8 ipproto,
767 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
768 const struct ipv6hdr *ipv6h,
769 struct sk_buff *skb))
770{
771 struct ip6_tnl *t;
772 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
773 u8 tproto;
774 int err;
775
776 rcu_read_lock();
777 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
778 if (t) {
779 struct pcpu_sw_netstats *tstats;
780
781 tproto = ACCESS_ONCE(t->parms.proto);
782 if (tproto != ipproto && tproto != 0) {
783 rcu_read_unlock();
784 goto discard;
785 }
786
787 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
788 rcu_read_unlock();
789 goto discard;
790 }
791
792 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
793 t->dev->stats.rx_dropped++;
794 rcu_read_unlock();
795 goto discard;
796 }
797 skb->mac_header = skb->network_header;
798 skb_reset_network_header(skb);
799 skb->protocol = htons(protocol);
800 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
801
802 __skb_tunnel_rx(skb, t->dev, t->net);
803
804 err = dscp_ecn_decapsulate(t, ipv6h, skb);
805 if (unlikely(err)) {
806 if (log_ecn_error)
807 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
808 &ipv6h->saddr,
809 ipv6_get_dsfield(ipv6h));
810 if (err > 1) {
811 ++t->dev->stats.rx_frame_errors;
812 ++t->dev->stats.rx_errors;
813 rcu_read_unlock();
814 goto discard;
815 }
816 }
817
818 tstats = this_cpu_ptr(t->dev->tstats);
819 u64_stats_update_begin(&tstats->syncp);
820 tstats->rx_packets++;
821 tstats->rx_bytes += skb->len;
822 u64_stats_update_end(&tstats->syncp);
823
824 netif_rx(skb);
825
826 rcu_read_unlock();
827 return 0;
828 }
829 rcu_read_unlock();
830 return 1;
831
832discard:
833 kfree_skb(skb);
834 return 0;
835}
836
837static int ip4ip6_rcv(struct sk_buff *skb)
838{
839 return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
840 ip4ip6_dscp_ecn_decapsulate);
841}
842
843static int ip6ip6_rcv(struct sk_buff *skb)
844{
845 return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
846 ip6ip6_dscp_ecn_decapsulate);
847}
848
849struct ipv6_tel_txoption {
850 struct ipv6_txoptions ops;
851 __u8 dst_opt[8];
852};
853
854static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
855{
856 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
857
858 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
859 opt->dst_opt[3] = 1;
860 opt->dst_opt[4] = encap_limit;
861 opt->dst_opt[5] = IPV6_TLV_PADN;
862 opt->dst_opt[6] = 1;
863
864 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
865 opt->ops.opt_nflen = 8;
866}
867
868/**
869 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
870 * @t: the outgoing tunnel device
871 * @hdr: IPv6 header from the incoming packet
872 *
873 * Description:
874 * Avoid trivial tunneling loop by checking that tunnel exit-point
875 * doesn't match source of incoming packet.
876 *
877 * Return:
878 * 1 if conflict,
879 * 0 else
880 **/
881
882static inline bool
883ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
884{
885 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
886}
887
888int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
889 const struct in6_addr *laddr,
890 const struct in6_addr *raddr)
891{
892 struct __ip6_tnl_parm *p = &t->parms;
893 int ret = 0;
894 struct net *net = t->net;
895
896 if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
897 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
898 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
899 struct net_device *ldev = NULL;
900
901 rcu_read_lock();
902 if (p->link)
903 ldev = dev_get_by_index_rcu(net, p->link);
904
905 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
906 pr_warn("%s xmit: Local address not yet configured!\n",
907 p->name);
908 else if (!ipv6_addr_is_multicast(raddr) &&
909 unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
910 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
911 p->name);
912 else
913 ret = 1;
914 rcu_read_unlock();
915 }
916 return ret;
917}
918EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
919
920/**
921 * ip6_tnl_xmit2 - encapsulate packet and send
922 * @skb: the outgoing socket buffer
923 * @dev: the outgoing tunnel device
924 * @dsfield: dscp code for outer header
925 * @fl: flow of tunneled packet
926 * @encap_limit: encapsulation limit
927 * @pmtu: Path MTU is stored if packet is too big
928 *
929 * Description:
930 * Build new header and do some sanity checks on the packet before sending
931 * it.
932 *
933 * Return:
934 * 0 on success
935 * -1 fail
936 * %-EMSGSIZE message too big. return mtu in this case.
937 **/
938
939static int ip6_tnl_xmit2(struct sk_buff *skb,
940 struct net_device *dev,
941 __u8 dsfield,
942 struct flowi6 *fl6,
943 int encap_limit,
944 __u32 *pmtu)
945{
946 struct ip6_tnl *t = netdev_priv(dev);
947 struct net *net = t->net;
948 struct net_device_stats *stats = &t->dev->stats;
949 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
950 struct ipv6_tel_txoption opt;
951 struct dst_entry *dst = NULL, *ndst = NULL;
952 struct net_device *tdev;
953 int mtu;
954 unsigned int max_headroom = sizeof(struct ipv6hdr);
955 u8 proto;
956 int err = -1;
957
958 /* NBMA tunnel */
959 if (ipv6_addr_any(&t->parms.raddr)) {
960 struct in6_addr *addr6;
961 struct neighbour *neigh;
962 int addr_type;
963
964 if (!skb_dst(skb))
965 goto tx_err_link_failure;
966
967 neigh = dst_neigh_lookup(skb_dst(skb),
968 &ipv6_hdr(skb)->daddr);
969 if (!neigh)
970 goto tx_err_link_failure;
971
972 addr6 = (struct in6_addr *)&neigh->primary_key;
973 addr_type = ipv6_addr_type(addr6);
974
975 if (addr_type == IPV6_ADDR_ANY)
976 addr6 = &ipv6_hdr(skb)->daddr;
977
978 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
979 neigh_release(neigh);
980 } else if (!fl6->flowi6_mark)
981 dst = dst_cache_get(&t->dst_cache);
982
983 if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
984 goto tx_err_link_failure;
985
986 if (!dst) {
987 dst = ip6_route_output(net, NULL, fl6);
988
989 if (dst->error)
990 goto tx_err_link_failure;
991 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
992 if (IS_ERR(dst)) {
993 err = PTR_ERR(dst);
994 dst = NULL;
995 goto tx_err_link_failure;
996 }
997 ndst = dst;
998 }
999
1000 tdev = dst->dev;
1001
1002 if (tdev == dev) {
1003 stats->collisions++;
1004 net_warn_ratelimited("%s: Local routing loop detected!\n",
1005 t->parms.name);
1006 goto tx_err_dst_release;
1007 }
1008 mtu = dst_mtu(dst) - sizeof(*ipv6h);
1009 if (encap_limit >= 0) {
1010 max_headroom += 8;
1011 mtu -= 8;
1012 }
1013 if (mtu < IPV6_MIN_MTU)
1014 mtu = IPV6_MIN_MTU;
1015 if (skb_dst(skb))
1016 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1017 if (skb->len > mtu) {
1018 *pmtu = mtu;
1019 err = -EMSGSIZE;
1020 goto tx_err_dst_release;
1021 }
1022
1023 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1024
1025 /*
1026 * Okay, now see if we can stuff it in the buffer as-is.
1027 */
1028 max_headroom += LL_RESERVED_SPACE(tdev);
1029
1030 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1031 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1032 struct sk_buff *new_skb;
1033
1034 new_skb = skb_realloc_headroom(skb, max_headroom);
1035 if (!new_skb)
1036 goto tx_err_dst_release;
1037
1038 if (skb->sk)
1039 skb_set_owner_w(new_skb, skb->sk);
1040 consume_skb(skb);
1041 skb = new_skb;
1042 }
1043
1044 if (!fl6->flowi6_mark && ndst)
1045 dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1046 skb_dst_set(skb, dst);
1047
1048 skb->transport_header = skb->network_header;
1049
1050 proto = fl6->flowi6_proto;
1051 if (encap_limit >= 0) {
1052 init_tel_txopt(&opt, encap_limit);
1053 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1054 }
1055
1056 if (likely(!skb->encapsulation)) {
1057 skb_reset_inner_headers(skb);
1058 skb->encapsulation = 1;
1059 }
1060
1061 skb_push(skb, sizeof(struct ipv6hdr));
1062 skb_reset_network_header(skb);
1063 ipv6h = ipv6_hdr(skb);
1064 ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
1065 ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1066 ipv6h->hop_limit = t->parms.hop_limit;
1067 ipv6h->nexthdr = proto;
1068 ipv6h->saddr = fl6->saddr;
1069 ipv6h->daddr = fl6->daddr;
1070 ip6tunnel_xmit(NULL, skb, dev);
1071 return 0;
1072tx_err_link_failure:
1073 stats->tx_carrier_errors++;
1074 dst_link_failure(skb);
1075tx_err_dst_release:
1076 dst_release(dst);
1077 return err;
1078}
1079
1080static inline int
1081ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1082{
1083 struct ip6_tnl *t = netdev_priv(dev);
1084 const struct iphdr *iph = ip_hdr(skb);
1085 int encap_limit = -1;
1086 struct flowi6 fl6;
1087 __u8 dsfield;
1088 __u32 mtu;
1089 u8 tproto;
1090 int err;
1091
1092 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1093
1094 tproto = ACCESS_ONCE(t->parms.proto);
1095 if (tproto != IPPROTO_IPIP && tproto != 0)
1096 return -1;
1097
1098 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1099 encap_limit = t->parms.encap_limit;
1100
1101 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1102 fl6.flowi6_proto = IPPROTO_IPIP;
1103
1104 dsfield = ipv4_get_dsfield(iph);
1105
1106 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1107 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1108 & IPV6_TCLASS_MASK;
1109 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1110 fl6.flowi6_mark = skb->mark;
1111
1112 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1113 if (err != 0) {
1114 /* XXX: send ICMP error even if DF is not set. */
1115 if (err == -EMSGSIZE)
1116 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1117 htonl(mtu));
1118 return -1;
1119 }
1120
1121 return 0;
1122}
1123
1124static inline int
1125ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1126{
1127 struct ip6_tnl *t = netdev_priv(dev);
1128 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1129 int encap_limit = -1;
1130 __u16 offset;
1131 struct flowi6 fl6;
1132 __u8 dsfield;
1133 __u32 mtu;
1134 u8 tproto;
1135 int err;
1136
1137 tproto = ACCESS_ONCE(t->parms.proto);
1138 if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1139 ip6_tnl_addr_conflict(t, ipv6h))
1140 return -1;
1141
1142 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1143 if (offset > 0) {
1144 struct ipv6_tlv_tnl_enc_lim *tel;
1145 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1146 if (tel->encap_limit == 0) {
1147 icmpv6_send(skb, ICMPV6_PARAMPROB,
1148 ICMPV6_HDR_FIELD, offset + 2);
1149 return -1;
1150 }
1151 encap_limit = tel->encap_limit - 1;
1152 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1153 encap_limit = t->parms.encap_limit;
1154
1155 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1156 fl6.flowi6_proto = IPPROTO_IPV6;
1157
1158 dsfield = ipv6_get_dsfield(ipv6h);
1159 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1160 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1161 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1162 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1163 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1164 fl6.flowi6_mark = skb->mark;
1165
1166 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1167 if (err != 0) {
1168 if (err == -EMSGSIZE)
1169 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1170 return -1;
1171 }
1172
1173 return 0;
1174}
1175
1176static netdev_tx_t
1177ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1178{
1179 struct ip6_tnl *t = netdev_priv(dev);
1180 struct net_device_stats *stats = &t->dev->stats;
1181 int ret;
1182
1183 switch (skb->protocol) {
1184 case htons(ETH_P_IP):
1185 ret = ip4ip6_tnl_xmit(skb, dev);
1186 break;
1187 case htons(ETH_P_IPV6):
1188 ret = ip6ip6_tnl_xmit(skb, dev);
1189 break;
1190 default:
1191 goto tx_err;
1192 }
1193
1194 if (ret < 0)
1195 goto tx_err;
1196
1197 return NETDEV_TX_OK;
1198
1199tx_err:
1200 stats->tx_errors++;
1201 stats->tx_dropped++;
1202 kfree_skb(skb);
1203 return NETDEV_TX_OK;
1204}
1205
1206static void ip6_tnl_link_config(struct ip6_tnl *t)
1207{
1208 struct net_device *dev = t->dev;
1209 struct __ip6_tnl_parm *p = &t->parms;
1210 struct flowi6 *fl6 = &t->fl.u.ip6;
1211
1212 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1213 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1214
1215 /* Set up flowi template */
1216 fl6->saddr = p->laddr;
1217 fl6->daddr = p->raddr;
1218 fl6->flowi6_oif = p->link;
1219 fl6->flowlabel = 0;
1220
1221 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1222 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1223 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1224 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1225
1226 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1227 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1228
1229 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1230 dev->flags |= IFF_POINTOPOINT;
1231 else
1232 dev->flags &= ~IFF_POINTOPOINT;
1233
1234 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1235 int strict = (ipv6_addr_type(&p->raddr) &
1236 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1237
1238 struct rt6_info *rt = rt6_lookup(t->net,
1239 &p->raddr, &p->laddr,
1240 p->link, strict);
1241
1242 if (!rt)
1243 return;
1244
1245 if (rt->dst.dev) {
1246 dev->hard_header_len = rt->dst.dev->hard_header_len +
1247 sizeof(struct ipv6hdr);
1248
1249 dev->mtu = rt->dst.dev->mtu - sizeof(struct ipv6hdr);
1250 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1251 dev->mtu -= 8;
1252
1253 if (dev->mtu < IPV6_MIN_MTU)
1254 dev->mtu = IPV6_MIN_MTU;
1255 }
1256 ip6_rt_put(rt);
1257 }
1258}
1259
1260/**
1261 * ip6_tnl_change - update the tunnel parameters
1262 * @t: tunnel to be changed
1263 * @p: tunnel configuration parameters
1264 *
1265 * Description:
1266 * ip6_tnl_change() updates the tunnel parameters
1267 **/
1268
1269static int
1270ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1271{
1272 t->parms.laddr = p->laddr;
1273 t->parms.raddr = p->raddr;
1274 t->parms.flags = p->flags;
1275 t->parms.hop_limit = p->hop_limit;
1276 t->parms.encap_limit = p->encap_limit;
1277 t->parms.flowinfo = p->flowinfo;
1278 t->parms.link = p->link;
1279 t->parms.proto = p->proto;
1280 dst_cache_reset(&t->dst_cache);
1281 ip6_tnl_link_config(t);
1282 return 0;
1283}
1284
1285static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1286{
1287 struct net *net = t->net;
1288 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1289 int err;
1290
1291 ip6_tnl_unlink(ip6n, t);
1292 synchronize_net();
1293 err = ip6_tnl_change(t, p);
1294 ip6_tnl_link(ip6n, t);
1295 netdev_state_change(t->dev);
1296 return err;
1297}
1298
1299static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1300{
1301 /* for default tnl0 device allow to change only the proto */
1302 t->parms.proto = p->proto;
1303 netdev_state_change(t->dev);
1304 return 0;
1305}
1306
1307static void
1308ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1309{
1310 p->laddr = u->laddr;
1311 p->raddr = u->raddr;
1312 p->flags = u->flags;
1313 p->hop_limit = u->hop_limit;
1314 p->encap_limit = u->encap_limit;
1315 p->flowinfo = u->flowinfo;
1316 p->link = u->link;
1317 p->proto = u->proto;
1318 memcpy(p->name, u->name, sizeof(u->name));
1319}
1320
1321static void
1322ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1323{
1324 u->laddr = p->laddr;
1325 u->raddr = p->raddr;
1326 u->flags = p->flags;
1327 u->hop_limit = p->hop_limit;
1328 u->encap_limit = p->encap_limit;
1329 u->flowinfo = p->flowinfo;
1330 u->link = p->link;
1331 u->proto = p->proto;
1332 memcpy(u->name, p->name, sizeof(u->name));
1333}
1334
1335/**
1336 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1337 * @dev: virtual device associated with tunnel
1338 * @ifr: parameters passed from userspace
1339 * @cmd: command to be performed
1340 *
1341 * Description:
1342 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1343 * from userspace.
1344 *
1345 * The possible commands are the following:
1346 * %SIOCGETTUNNEL: get tunnel parameters for device
1347 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1348 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1349 * %SIOCDELTUNNEL: delete tunnel
1350 *
1351 * The fallback device "ip6tnl0", created during module
1352 * initialization, can be used for creating other tunnel devices.
1353 *
1354 * Return:
1355 * 0 on success,
1356 * %-EFAULT if unable to copy data to or from userspace,
1357 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1358 * %-EINVAL if passed tunnel parameters are invalid,
1359 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1360 * %-ENODEV if attempting to change or delete a nonexisting device
1361 **/
1362
1363static int
1364ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1365{
1366 int err = 0;
1367 struct ip6_tnl_parm p;
1368 struct __ip6_tnl_parm p1;
1369 struct ip6_tnl *t = netdev_priv(dev);
1370 struct net *net = t->net;
1371 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1372
1373 switch (cmd) {
1374 case SIOCGETTUNNEL:
1375 if (dev == ip6n->fb_tnl_dev) {
1376 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1377 err = -EFAULT;
1378 break;
1379 }
1380 ip6_tnl_parm_from_user(&p1, &p);
1381 t = ip6_tnl_locate(net, &p1, 0);
1382 if (IS_ERR(t))
1383 t = netdev_priv(dev);
1384 } else {
1385 memset(&p, 0, sizeof(p));
1386 }
1387 ip6_tnl_parm_to_user(&p, &t->parms);
1388 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1389 err = -EFAULT;
1390 }
1391 break;
1392 case SIOCADDTUNNEL:
1393 case SIOCCHGTUNNEL:
1394 err = -EPERM;
1395 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1396 break;
1397 err = -EFAULT;
1398 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1399 break;
1400 err = -EINVAL;
1401 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1402 p.proto != 0)
1403 break;
1404 ip6_tnl_parm_from_user(&p1, &p);
1405 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1406 if (cmd == SIOCCHGTUNNEL) {
1407 if (!IS_ERR(t)) {
1408 if (t->dev != dev) {
1409 err = -EEXIST;
1410 break;
1411 }
1412 } else
1413 t = netdev_priv(dev);
1414 if (dev == ip6n->fb_tnl_dev)
1415 err = ip6_tnl0_update(t, &p1);
1416 else
1417 err = ip6_tnl_update(t, &p1);
1418 }
1419 if (!IS_ERR(t)) {
1420 err = 0;
1421 ip6_tnl_parm_to_user(&p, &t->parms);
1422 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1423 err = -EFAULT;
1424
1425 } else {
1426 err = PTR_ERR(t);
1427 }
1428 break;
1429 case SIOCDELTUNNEL:
1430 err = -EPERM;
1431 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1432 break;
1433
1434 if (dev == ip6n->fb_tnl_dev) {
1435 err = -EFAULT;
1436 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1437 break;
1438 err = -ENOENT;
1439 ip6_tnl_parm_from_user(&p1, &p);
1440 t = ip6_tnl_locate(net, &p1, 0);
1441 if (IS_ERR(t))
1442 break;
1443 err = -EPERM;
1444 if (t->dev == ip6n->fb_tnl_dev)
1445 break;
1446 dev = t->dev;
1447 }
1448 err = 0;
1449 unregister_netdevice(dev);
1450 break;
1451 default:
1452 err = -EINVAL;
1453 }
1454 return err;
1455}
1456
1457/**
1458 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1459 * @dev: virtual device associated with tunnel
1460 * @new_mtu: the new mtu
1461 *
1462 * Return:
1463 * 0 on success,
1464 * %-EINVAL if mtu too small
1465 **/
1466
1467static int
1468ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1469{
1470 struct ip6_tnl *tnl = netdev_priv(dev);
1471
1472 if (tnl->parms.proto == IPPROTO_IPIP) {
1473 if (new_mtu < 68)
1474 return -EINVAL;
1475 } else {
1476 if (new_mtu < IPV6_MIN_MTU)
1477 return -EINVAL;
1478 }
1479 if (new_mtu > 0xFFF8 - dev->hard_header_len)
1480 return -EINVAL;
1481 dev->mtu = new_mtu;
1482 return 0;
1483}
1484
1485int ip6_tnl_get_iflink(const struct net_device *dev)
1486{
1487 struct ip6_tnl *t = netdev_priv(dev);
1488
1489 return t->parms.link;
1490}
1491EXPORT_SYMBOL(ip6_tnl_get_iflink);
1492
1493static const struct net_device_ops ip6_tnl_netdev_ops = {
1494 .ndo_init = ip6_tnl_dev_init,
1495 .ndo_uninit = ip6_tnl_dev_uninit,
1496 .ndo_start_xmit = ip6_tnl_xmit,
1497 .ndo_do_ioctl = ip6_tnl_ioctl,
1498 .ndo_change_mtu = ip6_tnl_change_mtu,
1499 .ndo_get_stats = ip6_get_stats,
1500 .ndo_get_iflink = ip6_tnl_get_iflink,
1501};
1502
1503
1504/**
1505 * ip6_tnl_dev_setup - setup virtual tunnel device
1506 * @dev: virtual device associated with tunnel
1507 *
1508 * Description:
1509 * Initialize function pointers and device parameters
1510 **/
1511
1512static void ip6_tnl_dev_setup(struct net_device *dev)
1513{
1514 struct ip6_tnl *t;
1515
1516 dev->netdev_ops = &ip6_tnl_netdev_ops;
1517 dev->destructor = ip6_dev_free;
1518
1519 dev->type = ARPHRD_TUNNEL6;
1520 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
1521 dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr);
1522 t = netdev_priv(dev);
1523 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1524 dev->mtu -= 8;
1525 dev->flags |= IFF_NOARP;
1526 dev->addr_len = sizeof(struct in6_addr);
1527 netif_keep_dst(dev);
1528 /* This perm addr will be used as interface identifier by IPv6 */
1529 dev->addr_assign_type = NET_ADDR_RANDOM;
1530 eth_random_addr(dev->perm_addr);
1531}
1532
1533
1534/**
1535 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1536 * @dev: virtual device associated with tunnel
1537 **/
1538
1539static inline int
1540ip6_tnl_dev_init_gen(struct net_device *dev)
1541{
1542 struct ip6_tnl *t = netdev_priv(dev);
1543 int ret;
1544
1545 t->dev = dev;
1546 t->net = dev_net(dev);
1547 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1548 if (!dev->tstats)
1549 return -ENOMEM;
1550
1551 ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1552 if (ret) {
1553 free_percpu(dev->tstats);
1554 dev->tstats = NULL;
1555 return ret;
1556 }
1557
1558 return 0;
1559}
1560
1561/**
1562 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1563 * @dev: virtual device associated with tunnel
1564 **/
1565
1566static int ip6_tnl_dev_init(struct net_device *dev)
1567{
1568 struct ip6_tnl *t = netdev_priv(dev);
1569 int err = ip6_tnl_dev_init_gen(dev);
1570
1571 if (err)
1572 return err;
1573 ip6_tnl_link_config(t);
1574 return 0;
1575}
1576
1577/**
1578 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1579 * @dev: fallback device
1580 *
1581 * Return: 0
1582 **/
1583
1584static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1585{
1586 struct ip6_tnl *t = netdev_priv(dev);
1587 struct net *net = dev_net(dev);
1588 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1589
1590 t->parms.proto = IPPROTO_IPV6;
1591 dev_hold(dev);
1592
1593 rcu_assign_pointer(ip6n->tnls_wc[0], t);
1594 return 0;
1595}
1596
1597static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1598{
1599 u8 proto;
1600
1601 if (!data || !data[IFLA_IPTUN_PROTO])
1602 return 0;
1603
1604 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1605 if (proto != IPPROTO_IPV6 &&
1606 proto != IPPROTO_IPIP &&
1607 proto != 0)
1608 return -EINVAL;
1609
1610 return 0;
1611}
1612
1613static void ip6_tnl_netlink_parms(struct nlattr *data[],
1614 struct __ip6_tnl_parm *parms)
1615{
1616 memset(parms, 0, sizeof(*parms));
1617
1618 if (!data)
1619 return;
1620
1621 if (data[IFLA_IPTUN_LINK])
1622 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1623
1624 if (data[IFLA_IPTUN_LOCAL])
1625 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1626
1627 if (data[IFLA_IPTUN_REMOTE])
1628 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1629
1630 if (data[IFLA_IPTUN_TTL])
1631 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1632
1633 if (data[IFLA_IPTUN_ENCAP_LIMIT])
1634 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1635
1636 if (data[IFLA_IPTUN_FLOWINFO])
1637 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1638
1639 if (data[IFLA_IPTUN_FLAGS])
1640 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1641
1642 if (data[IFLA_IPTUN_PROTO])
1643 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1644}
1645
1646static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1647 struct nlattr *tb[], struct nlattr *data[])
1648{
1649 struct net *net = dev_net(dev);
1650 struct ip6_tnl *nt, *t;
1651
1652 nt = netdev_priv(dev);
1653 ip6_tnl_netlink_parms(data, &nt->parms);
1654
1655 t = ip6_tnl_locate(net, &nt->parms, 0);
1656 if (!IS_ERR(t))
1657 return -EEXIST;
1658
1659 return ip6_tnl_create2(dev);
1660}
1661
1662static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1663 struct nlattr *data[])
1664{
1665 struct ip6_tnl *t = netdev_priv(dev);
1666 struct __ip6_tnl_parm p;
1667 struct net *net = t->net;
1668 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1669
1670 if (dev == ip6n->fb_tnl_dev)
1671 return -EINVAL;
1672
1673 ip6_tnl_netlink_parms(data, &p);
1674
1675 t = ip6_tnl_locate(net, &p, 0);
1676 if (!IS_ERR(t)) {
1677 if (t->dev != dev)
1678 return -EEXIST;
1679 } else
1680 t = netdev_priv(dev);
1681
1682 return ip6_tnl_update(t, &p);
1683}
1684
1685static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
1686{
1687 struct net *net = dev_net(dev);
1688 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1689
1690 if (dev != ip6n->fb_tnl_dev)
1691 unregister_netdevice_queue(dev, head);
1692}
1693
1694static size_t ip6_tnl_get_size(const struct net_device *dev)
1695{
1696 return
1697 /* IFLA_IPTUN_LINK */
1698 nla_total_size(4) +
1699 /* IFLA_IPTUN_LOCAL */
1700 nla_total_size(sizeof(struct in6_addr)) +
1701 /* IFLA_IPTUN_REMOTE */
1702 nla_total_size(sizeof(struct in6_addr)) +
1703 /* IFLA_IPTUN_TTL */
1704 nla_total_size(1) +
1705 /* IFLA_IPTUN_ENCAP_LIMIT */
1706 nla_total_size(1) +
1707 /* IFLA_IPTUN_FLOWINFO */
1708 nla_total_size(4) +
1709 /* IFLA_IPTUN_FLAGS */
1710 nla_total_size(4) +
1711 /* IFLA_IPTUN_PROTO */
1712 nla_total_size(1) +
1713 0;
1714}
1715
1716static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
1717{
1718 struct ip6_tnl *tunnel = netdev_priv(dev);
1719 struct __ip6_tnl_parm *parm = &tunnel->parms;
1720
1721 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1722 nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
1723 nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
1724 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
1725 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
1726 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
1727 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
1728 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
1729 goto nla_put_failure;
1730 return 0;
1731
1732nla_put_failure:
1733 return -EMSGSIZE;
1734}
1735
1736struct net *ip6_tnl_get_link_net(const struct net_device *dev)
1737{
1738 struct ip6_tnl *tunnel = netdev_priv(dev);
1739
1740 return tunnel->net;
1741}
1742EXPORT_SYMBOL(ip6_tnl_get_link_net);
1743
1744static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
1745 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1746 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) },
1747 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) },
1748 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1749 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 },
1750 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 },
1751 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 },
1752 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1753};
1754
1755static struct rtnl_link_ops ip6_link_ops __read_mostly = {
1756 .kind = "ip6tnl",
1757 .maxtype = IFLA_IPTUN_MAX,
1758 .policy = ip6_tnl_policy,
1759 .priv_size = sizeof(struct ip6_tnl),
1760 .setup = ip6_tnl_dev_setup,
1761 .validate = ip6_tnl_validate,
1762 .newlink = ip6_tnl_newlink,
1763 .changelink = ip6_tnl_changelink,
1764 .dellink = ip6_tnl_dellink,
1765 .get_size = ip6_tnl_get_size,
1766 .fill_info = ip6_tnl_fill_info,
1767 .get_link_net = ip6_tnl_get_link_net,
1768};
1769
1770static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1771 .handler = ip4ip6_rcv,
1772 .err_handler = ip4ip6_err,
1773 .priority = 1,
1774};
1775
1776static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1777 .handler = ip6ip6_rcv,
1778 .err_handler = ip6ip6_err,
1779 .priority = 1,
1780};
1781
1782static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
1783{
1784 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1785 struct net_device *dev, *aux;
1786 int h;
1787 struct ip6_tnl *t;
1788 LIST_HEAD(list);
1789
1790 for_each_netdev_safe(net, dev, aux)
1791 if (dev->rtnl_link_ops == &ip6_link_ops)
1792 unregister_netdevice_queue(dev, &list);
1793
1794 for (h = 0; h < HASH_SIZE; h++) {
1795 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1796 while (t) {
1797 /* If dev is in the same netns, it has already
1798 * been added to the list by the previous loop.
1799 */
1800 if (!net_eq(dev_net(t->dev), net))
1801 unregister_netdevice_queue(t->dev, &list);
1802 t = rtnl_dereference(t->next);
1803 }
1804 }
1805
1806 unregister_netdevice_many(&list);
1807}
1808
1809static int __net_init ip6_tnl_init_net(struct net *net)
1810{
1811 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1812 struct ip6_tnl *t = NULL;
1813 int err;
1814
1815 ip6n->tnls[0] = ip6n->tnls_wc;
1816 ip6n->tnls[1] = ip6n->tnls_r_l;
1817
1818 err = -ENOMEM;
1819 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1820 NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
1821
1822 if (!ip6n->fb_tnl_dev)
1823 goto err_alloc_dev;
1824 dev_net_set(ip6n->fb_tnl_dev, net);
1825 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
1826 /* FB netdevice is special: we have one, and only one per netns.
1827 * Allowing to move it to another netns is clearly unsafe.
1828 */
1829 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
1830
1831 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1832 if (err < 0)
1833 goto err_register;
1834
1835 err = register_netdev(ip6n->fb_tnl_dev);
1836 if (err < 0)
1837 goto err_register;
1838
1839 t = netdev_priv(ip6n->fb_tnl_dev);
1840
1841 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1842 return 0;
1843
1844err_register:
1845 ip6_dev_free(ip6n->fb_tnl_dev);
1846err_alloc_dev:
1847 return err;
1848}
1849
1850static void __net_exit ip6_tnl_exit_net(struct net *net)
1851{
1852 rtnl_lock();
1853 ip6_tnl_destroy_tunnels(net);
1854 rtnl_unlock();
1855}
1856
1857static struct pernet_operations ip6_tnl_net_ops = {
1858 .init = ip6_tnl_init_net,
1859 .exit = ip6_tnl_exit_net,
1860 .id = &ip6_tnl_net_id,
1861 .size = sizeof(struct ip6_tnl_net),
1862};
1863
1864/**
1865 * ip6_tunnel_init - register protocol and reserve needed resources
1866 *
1867 * Return: 0 on success
1868 **/
1869
1870static int __init ip6_tunnel_init(void)
1871{
1872 int err;
1873
1874 err = register_pernet_device(&ip6_tnl_net_ops);
1875 if (err < 0)
1876 goto out_pernet;
1877
1878 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1879 if (err < 0) {
1880 pr_err("%s: can't register ip4ip6\n", __func__);
1881 goto out_ip4ip6;
1882 }
1883
1884 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1885 if (err < 0) {
1886 pr_err("%s: can't register ip6ip6\n", __func__);
1887 goto out_ip6ip6;
1888 }
1889 err = rtnl_link_register(&ip6_link_ops);
1890 if (err < 0)
1891 goto rtnl_link_failed;
1892
1893 return 0;
1894
1895rtnl_link_failed:
1896 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1897out_ip6ip6:
1898 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1899out_ip4ip6:
1900 unregister_pernet_device(&ip6_tnl_net_ops);
1901out_pernet:
1902 return err;
1903}
1904
1905/**
1906 * ip6_tunnel_cleanup - free resources and unregister protocol
1907 **/
1908
1909static void __exit ip6_tunnel_cleanup(void)
1910{
1911 rtnl_link_unregister(&ip6_link_ops);
1912 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1913 pr_info("%s: can't deregister ip4ip6\n", __func__);
1914
1915 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1916 pr_info("%s: can't deregister ip6ip6\n", __func__);
1917
1918 unregister_pernet_device(&ip6_tnl_net_ops);
1919}
1920
1921module_init(ip6_tunnel_init);
1922module_exit(ip6_tunnel_cleanup);
1/*
2 * IPv6 tunneling device
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Ville Nuorvala <vnuorval@tcs.hut.fi>
7 * Yasuyuki Kozakai <kozakai@linux-ipv6.org>
8 *
9 * Based on:
10 * linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
11 *
12 * RFC 2473
13 *
14 * This program is free software; you can redistribute it and/or
15 * modify it under the terms of the GNU General Public License
16 * as published by the Free Software Foundation; either version
17 * 2 of the License, or (at your option) any later version.
18 *
19 */
20
21#include <linux/module.h>
22#include <linux/capability.h>
23#include <linux/errno.h>
24#include <linux/types.h>
25#include <linux/sockios.h>
26#include <linux/icmp.h>
27#include <linux/if.h>
28#include <linux/in.h>
29#include <linux/ip.h>
30#include <linux/if_tunnel.h>
31#include <linux/net.h>
32#include <linux/in6.h>
33#include <linux/netdevice.h>
34#include <linux/if_arp.h>
35#include <linux/icmpv6.h>
36#include <linux/init.h>
37#include <linux/route.h>
38#include <linux/rtnetlink.h>
39#include <linux/netfilter_ipv6.h>
40#include <linux/slab.h>
41
42#include <asm/uaccess.h>
43#include <linux/atomic.h>
44
45#include <net/icmp.h>
46#include <net/ip.h>
47#include <net/ipv6.h>
48#include <net/ip6_route.h>
49#include <net/addrconf.h>
50#include <net/ip6_tunnel.h>
51#include <net/xfrm.h>
52#include <net/dsfield.h>
53#include <net/inet_ecn.h>
54#include <net/net_namespace.h>
55#include <net/netns/generic.h>
56
57MODULE_AUTHOR("Ville Nuorvala");
58MODULE_DESCRIPTION("IPv6 tunneling device");
59MODULE_LICENSE("GPL");
60MODULE_ALIAS_NETDEV("ip6tnl0");
61
62#ifdef IP6_TNL_DEBUG
63#define IP6_TNL_TRACE(x...) printk(KERN_DEBUG "%s:" x "\n", __func__)
64#else
65#define IP6_TNL_TRACE(x...) do {;} while(0)
66#endif
67
68#define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
69#define IPV6_TCLASS_SHIFT 20
70
71#define HASH_SIZE 32
72
73#define HASH(addr) ((__force u32)((addr)->s6_addr32[0] ^ (addr)->s6_addr32[1] ^ \
74 (addr)->s6_addr32[2] ^ (addr)->s6_addr32[3]) & \
75 (HASH_SIZE - 1))
76
77static int ip6_tnl_dev_init(struct net_device *dev);
78static void ip6_tnl_dev_setup(struct net_device *dev);
79
80static int ip6_tnl_net_id __read_mostly;
81struct ip6_tnl_net {
82 /* the IPv6 tunnel fallback device */
83 struct net_device *fb_tnl_dev;
84 /* lists for storing tunnels in use */
85 struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
86 struct ip6_tnl __rcu *tnls_wc[1];
87 struct ip6_tnl __rcu **tnls[2];
88};
89
90/* often modified stats are per cpu, other are shared (netdev->stats) */
91struct pcpu_tstats {
92 unsigned long rx_packets;
93 unsigned long rx_bytes;
94 unsigned long tx_packets;
95 unsigned long tx_bytes;
96};
97
98static struct net_device_stats *ip6_get_stats(struct net_device *dev)
99{
100 struct pcpu_tstats sum = { 0 };
101 int i;
102
103 for_each_possible_cpu(i) {
104 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
105
106 sum.rx_packets += tstats->rx_packets;
107 sum.rx_bytes += tstats->rx_bytes;
108 sum.tx_packets += tstats->tx_packets;
109 sum.tx_bytes += tstats->tx_bytes;
110 }
111 dev->stats.rx_packets = sum.rx_packets;
112 dev->stats.rx_bytes = sum.rx_bytes;
113 dev->stats.tx_packets = sum.tx_packets;
114 dev->stats.tx_bytes = sum.tx_bytes;
115 return &dev->stats;
116}
117
118/*
119 * Locking : hash tables are protected by RCU and RTNL
120 */
121
122static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
123{
124 struct dst_entry *dst = t->dst_cache;
125
126 if (dst && dst->obsolete &&
127 dst->ops->check(dst, t->dst_cookie) == NULL) {
128 t->dst_cache = NULL;
129 dst_release(dst);
130 return NULL;
131 }
132
133 return dst;
134}
135
136static inline void ip6_tnl_dst_reset(struct ip6_tnl *t)
137{
138 dst_release(t->dst_cache);
139 t->dst_cache = NULL;
140}
141
142static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
143{
144 struct rt6_info *rt = (struct rt6_info *) dst;
145 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
146 dst_release(t->dst_cache);
147 t->dst_cache = dst;
148}
149
150/**
151 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
152 * @remote: the address of the tunnel exit-point
153 * @local: the address of the tunnel entry-point
154 *
155 * Return:
156 * tunnel matching given end-points if found,
157 * else fallback tunnel if its device is up,
158 * else %NULL
159 **/
160
161#define for_each_ip6_tunnel_rcu(start) \
162 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
163
164static struct ip6_tnl *
165ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
166{
167 unsigned int h0 = HASH(remote);
168 unsigned int h1 = HASH(local);
169 struct ip6_tnl *t;
170 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
171
172 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[h0 ^ h1]) {
173 if (ipv6_addr_equal(local, &t->parms.laddr) &&
174 ipv6_addr_equal(remote, &t->parms.raddr) &&
175 (t->dev->flags & IFF_UP))
176 return t;
177 }
178 t = rcu_dereference(ip6n->tnls_wc[0]);
179 if (t && (t->dev->flags & IFF_UP))
180 return t;
181
182 return NULL;
183}
184
185/**
186 * ip6_tnl_bucket - get head of list matching given tunnel parameters
187 * @p: parameters containing tunnel end-points
188 *
189 * Description:
190 * ip6_tnl_bucket() returns the head of the list matching the
191 * &struct in6_addr entries laddr and raddr in @p.
192 *
193 * Return: head of IPv6 tunnel list
194 **/
195
196static struct ip6_tnl __rcu **
197ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct ip6_tnl_parm *p)
198{
199 const struct in6_addr *remote = &p->raddr;
200 const struct in6_addr *local = &p->laddr;
201 unsigned h = 0;
202 int prio = 0;
203
204 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
205 prio = 1;
206 h = HASH(remote) ^ HASH(local);
207 }
208 return &ip6n->tnls[prio][h];
209}
210
211/**
212 * ip6_tnl_link - add tunnel to hash table
213 * @t: tunnel to be added
214 **/
215
216static void
217ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
218{
219 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
220
221 rcu_assign_pointer(t->next , rtnl_dereference(*tp));
222 rcu_assign_pointer(*tp, t);
223}
224
225/**
226 * ip6_tnl_unlink - remove tunnel from hash table
227 * @t: tunnel to be removed
228 **/
229
230static void
231ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
232{
233 struct ip6_tnl __rcu **tp;
234 struct ip6_tnl *iter;
235
236 for (tp = ip6_tnl_bucket(ip6n, &t->parms);
237 (iter = rtnl_dereference(*tp)) != NULL;
238 tp = &iter->next) {
239 if (t == iter) {
240 rcu_assign_pointer(*tp, t->next);
241 break;
242 }
243 }
244}
245
246static void ip6_dev_free(struct net_device *dev)
247{
248 free_percpu(dev->tstats);
249 free_netdev(dev);
250}
251
252/**
253 * ip6_tnl_create() - create a new tunnel
254 * @p: tunnel parameters
255 * @pt: pointer to new tunnel
256 *
257 * Description:
258 * Create tunnel matching given parameters.
259 *
260 * Return:
261 * created tunnel or NULL
262 **/
263
264static struct ip6_tnl *ip6_tnl_create(struct net *net, struct ip6_tnl_parm *p)
265{
266 struct net_device *dev;
267 struct ip6_tnl *t;
268 char name[IFNAMSIZ];
269 int err;
270 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
271
272 if (p->name[0])
273 strlcpy(name, p->name, IFNAMSIZ);
274 else
275 sprintf(name, "ip6tnl%%d");
276
277 dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
278 if (dev == NULL)
279 goto failed;
280
281 dev_net_set(dev, net);
282
283 t = netdev_priv(dev);
284 t->parms = *p;
285 err = ip6_tnl_dev_init(dev);
286 if (err < 0)
287 goto failed_free;
288
289 if ((err = register_netdevice(dev)) < 0)
290 goto failed_free;
291
292 dev_hold(dev);
293 ip6_tnl_link(ip6n, t);
294 return t;
295
296failed_free:
297 ip6_dev_free(dev);
298failed:
299 return NULL;
300}
301
302/**
303 * ip6_tnl_locate - find or create tunnel matching given parameters
304 * @p: tunnel parameters
305 * @create: != 0 if allowed to create new tunnel if no match found
306 *
307 * Description:
308 * ip6_tnl_locate() first tries to locate an existing tunnel
309 * based on @parms. If this is unsuccessful, but @create is set a new
310 * tunnel device is created and registered for use.
311 *
312 * Return:
313 * matching tunnel or NULL
314 **/
315
316static struct ip6_tnl *ip6_tnl_locate(struct net *net,
317 struct ip6_tnl_parm *p, int create)
318{
319 const struct in6_addr *remote = &p->raddr;
320 const struct in6_addr *local = &p->laddr;
321 struct ip6_tnl __rcu **tp;
322 struct ip6_tnl *t;
323 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
324
325 for (tp = ip6_tnl_bucket(ip6n, p);
326 (t = rtnl_dereference(*tp)) != NULL;
327 tp = &t->next) {
328 if (ipv6_addr_equal(local, &t->parms.laddr) &&
329 ipv6_addr_equal(remote, &t->parms.raddr))
330 return t;
331 }
332 if (!create)
333 return NULL;
334 return ip6_tnl_create(net, p);
335}
336
337/**
338 * ip6_tnl_dev_uninit - tunnel device uninitializer
339 * @dev: the device to be destroyed
340 *
341 * Description:
342 * ip6_tnl_dev_uninit() removes tunnel from its list
343 **/
344
345static void
346ip6_tnl_dev_uninit(struct net_device *dev)
347{
348 struct ip6_tnl *t = netdev_priv(dev);
349 struct net *net = dev_net(dev);
350 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
351
352 if (dev == ip6n->fb_tnl_dev)
353 rcu_assign_pointer(ip6n->tnls_wc[0], NULL);
354 else
355 ip6_tnl_unlink(ip6n, t);
356 ip6_tnl_dst_reset(t);
357 dev_put(dev);
358}
359
360/**
361 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
362 * @skb: received socket buffer
363 *
364 * Return:
365 * 0 if none was found,
366 * else index to encapsulation limit
367 **/
368
369static __u16
370parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
371{
372 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
373 __u8 nexthdr = ipv6h->nexthdr;
374 __u16 off = sizeof (*ipv6h);
375
376 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
377 __u16 optlen = 0;
378 struct ipv6_opt_hdr *hdr;
379 if (raw + off + sizeof (*hdr) > skb->data &&
380 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
381 break;
382
383 hdr = (struct ipv6_opt_hdr *) (raw + off);
384 if (nexthdr == NEXTHDR_FRAGMENT) {
385 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
386 if (frag_hdr->frag_off)
387 break;
388 optlen = 8;
389 } else if (nexthdr == NEXTHDR_AUTH) {
390 optlen = (hdr->hdrlen + 2) << 2;
391 } else {
392 optlen = ipv6_optlen(hdr);
393 }
394 if (nexthdr == NEXTHDR_DEST) {
395 __u16 i = off + 2;
396 while (1) {
397 struct ipv6_tlv_tnl_enc_lim *tel;
398
399 /* No more room for encapsulation limit */
400 if (i + sizeof (*tel) > off + optlen)
401 break;
402
403 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
404 /* return index of option if found and valid */
405 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
406 tel->length == 1)
407 return i;
408 /* else jump to next option */
409 if (tel->type)
410 i += tel->length + 2;
411 else
412 i++;
413 }
414 }
415 nexthdr = hdr->nexthdr;
416 off += optlen;
417 }
418 return 0;
419}
420
421/**
422 * ip6_tnl_err - tunnel error handler
423 *
424 * Description:
425 * ip6_tnl_err() should handle errors in the tunnel according
426 * to the specifications in RFC 2473.
427 **/
428
429static int
430ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
431 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
432{
433 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
434 struct ip6_tnl *t;
435 int rel_msg = 0;
436 u8 rel_type = ICMPV6_DEST_UNREACH;
437 u8 rel_code = ICMPV6_ADDR_UNREACH;
438 __u32 rel_info = 0;
439 __u16 len;
440 int err = -ENOENT;
441
442 /* If the packet doesn't contain the original IPv6 header we are
443 in trouble since we might need the source address for further
444 processing of the error. */
445
446 rcu_read_lock();
447 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
448 &ipv6h->saddr)) == NULL)
449 goto out;
450
451 if (t->parms.proto != ipproto && t->parms.proto != 0)
452 goto out;
453
454 err = 0;
455
456 switch (*type) {
457 __u32 teli;
458 struct ipv6_tlv_tnl_enc_lim *tel;
459 __u32 mtu;
460 case ICMPV6_DEST_UNREACH:
461 if (net_ratelimit())
462 printk(KERN_WARNING
463 "%s: Path to destination invalid "
464 "or inactive!\n", t->parms.name);
465 rel_msg = 1;
466 break;
467 case ICMPV6_TIME_EXCEED:
468 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
469 if (net_ratelimit())
470 printk(KERN_WARNING
471 "%s: Too small hop limit or "
472 "routing loop in tunnel!\n",
473 t->parms.name);
474 rel_msg = 1;
475 }
476 break;
477 case ICMPV6_PARAMPROB:
478 teli = 0;
479 if ((*code) == ICMPV6_HDR_FIELD)
480 teli = parse_tlv_tnl_enc_lim(skb, skb->data);
481
482 if (teli && teli == *info - 2) {
483 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
484 if (tel->encap_limit == 0) {
485 if (net_ratelimit())
486 printk(KERN_WARNING
487 "%s: Too small encapsulation "
488 "limit or routing loop in "
489 "tunnel!\n", t->parms.name);
490 rel_msg = 1;
491 }
492 } else if (net_ratelimit()) {
493 printk(KERN_WARNING
494 "%s: Recipient unable to parse tunneled "
495 "packet!\n ", t->parms.name);
496 }
497 break;
498 case ICMPV6_PKT_TOOBIG:
499 mtu = *info - offset;
500 if (mtu < IPV6_MIN_MTU)
501 mtu = IPV6_MIN_MTU;
502 t->dev->mtu = mtu;
503
504 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
505 rel_type = ICMPV6_PKT_TOOBIG;
506 rel_code = 0;
507 rel_info = mtu;
508 rel_msg = 1;
509 }
510 break;
511 }
512
513 *type = rel_type;
514 *code = rel_code;
515 *info = rel_info;
516 *msg = rel_msg;
517
518out:
519 rcu_read_unlock();
520 return err;
521}
522
523static int
524ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
525 u8 type, u8 code, int offset, __be32 info)
526{
527 int rel_msg = 0;
528 u8 rel_type = type;
529 u8 rel_code = code;
530 __u32 rel_info = ntohl(info);
531 int err;
532 struct sk_buff *skb2;
533 const struct iphdr *eiph;
534 struct rtable *rt;
535 struct flowi4 fl4;
536
537 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
538 &rel_msg, &rel_info, offset);
539 if (err < 0)
540 return err;
541
542 if (rel_msg == 0)
543 return 0;
544
545 switch (rel_type) {
546 case ICMPV6_DEST_UNREACH:
547 if (rel_code != ICMPV6_ADDR_UNREACH)
548 return 0;
549 rel_type = ICMP_DEST_UNREACH;
550 rel_code = ICMP_HOST_UNREACH;
551 break;
552 case ICMPV6_PKT_TOOBIG:
553 if (rel_code != 0)
554 return 0;
555 rel_type = ICMP_DEST_UNREACH;
556 rel_code = ICMP_FRAG_NEEDED;
557 break;
558 default:
559 return 0;
560 }
561
562 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
563 return 0;
564
565 skb2 = skb_clone(skb, GFP_ATOMIC);
566 if (!skb2)
567 return 0;
568
569 skb_dst_drop(skb2);
570
571 skb_pull(skb2, offset);
572 skb_reset_network_header(skb2);
573 eiph = ip_hdr(skb2);
574
575 /* Try to guess incoming interface */
576 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
577 eiph->saddr, 0,
578 0, 0,
579 IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
580 if (IS_ERR(rt))
581 goto out;
582
583 skb2->dev = rt->dst.dev;
584
585 /* route "incoming" packet */
586 if (rt->rt_flags & RTCF_LOCAL) {
587 ip_rt_put(rt);
588 rt = NULL;
589 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
590 eiph->daddr, eiph->saddr,
591 0, 0,
592 IPPROTO_IPIP,
593 RT_TOS(eiph->tos), 0);
594 if (IS_ERR(rt) ||
595 rt->dst.dev->type != ARPHRD_TUNNEL) {
596 if (!IS_ERR(rt))
597 ip_rt_put(rt);
598 goto out;
599 }
600 skb_dst_set(skb2, &rt->dst);
601 } else {
602 ip_rt_put(rt);
603 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
604 skb2->dev) ||
605 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
606 goto out;
607 }
608
609 /* change mtu on this route */
610 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
611 if (rel_info > dst_mtu(skb_dst(skb2)))
612 goto out;
613
614 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), rel_info);
615 }
616
617 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
618
619out:
620 kfree_skb(skb2);
621 return 0;
622}
623
624static int
625ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
626 u8 type, u8 code, int offset, __be32 info)
627{
628 int rel_msg = 0;
629 u8 rel_type = type;
630 u8 rel_code = code;
631 __u32 rel_info = ntohl(info);
632 int err;
633
634 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
635 &rel_msg, &rel_info, offset);
636 if (err < 0)
637 return err;
638
639 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
640 struct rt6_info *rt;
641 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
642
643 if (!skb2)
644 return 0;
645
646 skb_dst_drop(skb2);
647 skb_pull(skb2, offset);
648 skb_reset_network_header(skb2);
649
650 /* Try to guess incoming interface */
651 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
652 NULL, 0, 0);
653
654 if (rt && rt->rt6i_dev)
655 skb2->dev = rt->rt6i_dev;
656
657 icmpv6_send(skb2, rel_type, rel_code, rel_info);
658
659 if (rt)
660 dst_release(&rt->dst);
661
662 kfree_skb(skb2);
663 }
664
665 return 0;
666}
667
668static void ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
669 const struct ipv6hdr *ipv6h,
670 struct sk_buff *skb)
671{
672 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
673
674 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
675 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
676
677 if (INET_ECN_is_ce(dsfield))
678 IP_ECN_set_ce(ip_hdr(skb));
679}
680
681static void ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
682 const struct ipv6hdr *ipv6h,
683 struct sk_buff *skb)
684{
685 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
686 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
687
688 if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
689 IP6_ECN_set_ce(ipv6_hdr(skb));
690}
691
692/* called with rcu_read_lock() */
693static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
694{
695 struct ip6_tnl_parm *p = &t->parms;
696 int ret = 0;
697 struct net *net = dev_net(t->dev);
698
699 if (p->flags & IP6_TNL_F_CAP_RCV) {
700 struct net_device *ldev = NULL;
701
702 if (p->link)
703 ldev = dev_get_by_index_rcu(net, p->link);
704
705 if ((ipv6_addr_is_multicast(&p->laddr) ||
706 likely(ipv6_chk_addr(net, &p->laddr, ldev, 0))) &&
707 likely(!ipv6_chk_addr(net, &p->raddr, NULL, 0)))
708 ret = 1;
709
710 }
711 return ret;
712}
713
714/**
715 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
716 * @skb: received socket buffer
717 * @protocol: ethernet protocol ID
718 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
719 *
720 * Return: 0
721 **/
722
723static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
724 __u8 ipproto,
725 void (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
726 const struct ipv6hdr *ipv6h,
727 struct sk_buff *skb))
728{
729 struct ip6_tnl *t;
730 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
731
732 rcu_read_lock();
733
734 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
735 &ipv6h->daddr)) != NULL) {
736 struct pcpu_tstats *tstats;
737
738 if (t->parms.proto != ipproto && t->parms.proto != 0) {
739 rcu_read_unlock();
740 goto discard;
741 }
742
743 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
744 rcu_read_unlock();
745 goto discard;
746 }
747
748 if (!ip6_tnl_rcv_ctl(t)) {
749 t->dev->stats.rx_dropped++;
750 rcu_read_unlock();
751 goto discard;
752 }
753 secpath_reset(skb);
754 skb->mac_header = skb->network_header;
755 skb_reset_network_header(skb);
756 skb->protocol = htons(protocol);
757 skb->pkt_type = PACKET_HOST;
758 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
759
760 tstats = this_cpu_ptr(t->dev->tstats);
761 tstats->rx_packets++;
762 tstats->rx_bytes += skb->len;
763
764 __skb_tunnel_rx(skb, t->dev);
765
766 dscp_ecn_decapsulate(t, ipv6h, skb);
767
768 netif_rx(skb);
769
770 rcu_read_unlock();
771 return 0;
772 }
773 rcu_read_unlock();
774 return 1;
775
776discard:
777 kfree_skb(skb);
778 return 0;
779}
780
781static int ip4ip6_rcv(struct sk_buff *skb)
782{
783 return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
784 ip4ip6_dscp_ecn_decapsulate);
785}
786
787static int ip6ip6_rcv(struct sk_buff *skb)
788{
789 return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
790 ip6ip6_dscp_ecn_decapsulate);
791}
792
793struct ipv6_tel_txoption {
794 struct ipv6_txoptions ops;
795 __u8 dst_opt[8];
796};
797
798static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
799{
800 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
801
802 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
803 opt->dst_opt[3] = 1;
804 opt->dst_opt[4] = encap_limit;
805 opt->dst_opt[5] = IPV6_TLV_PADN;
806 opt->dst_opt[6] = 1;
807
808 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
809 opt->ops.opt_nflen = 8;
810}
811
812/**
813 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
814 * @t: the outgoing tunnel device
815 * @hdr: IPv6 header from the incoming packet
816 *
817 * Description:
818 * Avoid trivial tunneling loop by checking that tunnel exit-point
819 * doesn't match source of incoming packet.
820 *
821 * Return:
822 * 1 if conflict,
823 * 0 else
824 **/
825
826static inline int
827ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
828{
829 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
830}
831
832static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
833{
834 struct ip6_tnl_parm *p = &t->parms;
835 int ret = 0;
836 struct net *net = dev_net(t->dev);
837
838 if (p->flags & IP6_TNL_F_CAP_XMIT) {
839 struct net_device *ldev = NULL;
840
841 rcu_read_lock();
842 if (p->link)
843 ldev = dev_get_by_index_rcu(net, p->link);
844
845 if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
846 printk(KERN_WARNING
847 "%s xmit: Local address not yet configured!\n",
848 p->name);
849 else if (!ipv6_addr_is_multicast(&p->raddr) &&
850 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
851 printk(KERN_WARNING
852 "%s xmit: Routing loop! "
853 "Remote address found on this node!\n",
854 p->name);
855 else
856 ret = 1;
857 rcu_read_unlock();
858 }
859 return ret;
860}
861/**
862 * ip6_tnl_xmit2 - encapsulate packet and send
863 * @skb: the outgoing socket buffer
864 * @dev: the outgoing tunnel device
865 * @dsfield: dscp code for outer header
866 * @fl: flow of tunneled packet
867 * @encap_limit: encapsulation limit
868 * @pmtu: Path MTU is stored if packet is too big
869 *
870 * Description:
871 * Build new header and do some sanity checks on the packet before sending
872 * it.
873 *
874 * Return:
875 * 0 on success
876 * -1 fail
877 * %-EMSGSIZE message too big. return mtu in this case.
878 **/
879
880static int ip6_tnl_xmit2(struct sk_buff *skb,
881 struct net_device *dev,
882 __u8 dsfield,
883 struct flowi6 *fl6,
884 int encap_limit,
885 __u32 *pmtu)
886{
887 struct net *net = dev_net(dev);
888 struct ip6_tnl *t = netdev_priv(dev);
889 struct net_device_stats *stats = &t->dev->stats;
890 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
891 struct ipv6_tel_txoption opt;
892 struct dst_entry *dst;
893 struct net_device *tdev;
894 int mtu;
895 unsigned int max_headroom = sizeof(struct ipv6hdr);
896 u8 proto;
897 int err = -1;
898 int pkt_len;
899
900 if ((dst = ip6_tnl_dst_check(t)) != NULL)
901 dst_hold(dst);
902 else {
903 dst = ip6_route_output(net, NULL, fl6);
904
905 if (dst->error)
906 goto tx_err_link_failure;
907 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
908 if (IS_ERR(dst)) {
909 err = PTR_ERR(dst);
910 dst = NULL;
911 goto tx_err_link_failure;
912 }
913 }
914
915 tdev = dst->dev;
916
917 if (tdev == dev) {
918 stats->collisions++;
919 if (net_ratelimit())
920 printk(KERN_WARNING
921 "%s: Local routing loop detected!\n",
922 t->parms.name);
923 goto tx_err_dst_release;
924 }
925 mtu = dst_mtu(dst) - sizeof (*ipv6h);
926 if (encap_limit >= 0) {
927 max_headroom += 8;
928 mtu -= 8;
929 }
930 if (mtu < IPV6_MIN_MTU)
931 mtu = IPV6_MIN_MTU;
932 if (skb_dst(skb))
933 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
934 if (skb->len > mtu) {
935 *pmtu = mtu;
936 err = -EMSGSIZE;
937 goto tx_err_dst_release;
938 }
939
940 /*
941 * Okay, now see if we can stuff it in the buffer as-is.
942 */
943 max_headroom += LL_RESERVED_SPACE(tdev);
944
945 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
946 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
947 struct sk_buff *new_skb;
948
949 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
950 goto tx_err_dst_release;
951
952 if (skb->sk)
953 skb_set_owner_w(new_skb, skb->sk);
954 kfree_skb(skb);
955 skb = new_skb;
956 }
957 skb_dst_drop(skb);
958 skb_dst_set(skb, dst_clone(dst));
959
960 skb->transport_header = skb->network_header;
961
962 proto = fl6->flowi6_proto;
963 if (encap_limit >= 0) {
964 init_tel_txopt(&opt, encap_limit);
965 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
966 }
967 skb_push(skb, sizeof(struct ipv6hdr));
968 skb_reset_network_header(skb);
969 ipv6h = ipv6_hdr(skb);
970 *(__be32*)ipv6h = fl6->flowlabel | htonl(0x60000000);
971 dsfield = INET_ECN_encapsulate(0, dsfield);
972 ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
973 ipv6h->hop_limit = t->parms.hop_limit;
974 ipv6h->nexthdr = proto;
975 ipv6_addr_copy(&ipv6h->saddr, &fl6->saddr);
976 ipv6_addr_copy(&ipv6h->daddr, &fl6->daddr);
977 nf_reset(skb);
978 pkt_len = skb->len;
979 err = ip6_local_out(skb);
980
981 if (net_xmit_eval(err) == 0) {
982 struct pcpu_tstats *tstats = this_cpu_ptr(t->dev->tstats);
983
984 tstats->tx_bytes += pkt_len;
985 tstats->tx_packets++;
986 } else {
987 stats->tx_errors++;
988 stats->tx_aborted_errors++;
989 }
990 ip6_tnl_dst_store(t, dst);
991 return 0;
992tx_err_link_failure:
993 stats->tx_carrier_errors++;
994 dst_link_failure(skb);
995tx_err_dst_release:
996 dst_release(dst);
997 return err;
998}
999
1000static inline int
1001ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1002{
1003 struct ip6_tnl *t = netdev_priv(dev);
1004 const struct iphdr *iph = ip_hdr(skb);
1005 int encap_limit = -1;
1006 struct flowi6 fl6;
1007 __u8 dsfield;
1008 __u32 mtu;
1009 int err;
1010
1011 if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
1012 !ip6_tnl_xmit_ctl(t))
1013 return -1;
1014
1015 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1016 encap_limit = t->parms.encap_limit;
1017
1018 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1019 fl6.flowi6_proto = IPPROTO_IPIP;
1020
1021 dsfield = ipv4_get_dsfield(iph);
1022
1023 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1024 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1025 & IPV6_TCLASS_MASK;
1026
1027 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1028 if (err != 0) {
1029 /* XXX: send ICMP error even if DF is not set. */
1030 if (err == -EMSGSIZE)
1031 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1032 htonl(mtu));
1033 return -1;
1034 }
1035
1036 return 0;
1037}
1038
1039static inline int
1040ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1041{
1042 struct ip6_tnl *t = netdev_priv(dev);
1043 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1044 int encap_limit = -1;
1045 __u16 offset;
1046 struct flowi6 fl6;
1047 __u8 dsfield;
1048 __u32 mtu;
1049 int err;
1050
1051 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1052 !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1053 return -1;
1054
1055 offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb));
1056 if (offset > 0) {
1057 struct ipv6_tlv_tnl_enc_lim *tel;
1058 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1059 if (tel->encap_limit == 0) {
1060 icmpv6_send(skb, ICMPV6_PARAMPROB,
1061 ICMPV6_HDR_FIELD, offset + 2);
1062 return -1;
1063 }
1064 encap_limit = tel->encap_limit - 1;
1065 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1066 encap_limit = t->parms.encap_limit;
1067
1068 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1069 fl6.flowi6_proto = IPPROTO_IPV6;
1070
1071 dsfield = ipv6_get_dsfield(ipv6h);
1072 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1073 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1074 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
1075 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1076
1077 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1078 if (err != 0) {
1079 if (err == -EMSGSIZE)
1080 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1081 return -1;
1082 }
1083
1084 return 0;
1085}
1086
1087static netdev_tx_t
1088ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1089{
1090 struct ip6_tnl *t = netdev_priv(dev);
1091 struct net_device_stats *stats = &t->dev->stats;
1092 int ret;
1093
1094 switch (skb->protocol) {
1095 case htons(ETH_P_IP):
1096 ret = ip4ip6_tnl_xmit(skb, dev);
1097 break;
1098 case htons(ETH_P_IPV6):
1099 ret = ip6ip6_tnl_xmit(skb, dev);
1100 break;
1101 default:
1102 goto tx_err;
1103 }
1104
1105 if (ret < 0)
1106 goto tx_err;
1107
1108 return NETDEV_TX_OK;
1109
1110tx_err:
1111 stats->tx_errors++;
1112 stats->tx_dropped++;
1113 kfree_skb(skb);
1114 return NETDEV_TX_OK;
1115}
1116
1117static void ip6_tnl_set_cap(struct ip6_tnl *t)
1118{
1119 struct ip6_tnl_parm *p = &t->parms;
1120 int ltype = ipv6_addr_type(&p->laddr);
1121 int rtype = ipv6_addr_type(&p->raddr);
1122
1123 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
1124
1125 if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1126 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1127 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
1128 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
1129 if (ltype&IPV6_ADDR_UNICAST)
1130 p->flags |= IP6_TNL_F_CAP_XMIT;
1131 if (rtype&IPV6_ADDR_UNICAST)
1132 p->flags |= IP6_TNL_F_CAP_RCV;
1133 }
1134}
1135
1136static void ip6_tnl_link_config(struct ip6_tnl *t)
1137{
1138 struct net_device *dev = t->dev;
1139 struct ip6_tnl_parm *p = &t->parms;
1140 struct flowi6 *fl6 = &t->fl.u.ip6;
1141
1142 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1143 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1144
1145 /* Set up flowi template */
1146 ipv6_addr_copy(&fl6->saddr, &p->laddr);
1147 ipv6_addr_copy(&fl6->daddr, &p->raddr);
1148 fl6->flowi6_oif = p->link;
1149 fl6->flowlabel = 0;
1150
1151 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1152 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1153 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1154 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1155
1156 ip6_tnl_set_cap(t);
1157
1158 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1159 dev->flags |= IFF_POINTOPOINT;
1160 else
1161 dev->flags &= ~IFF_POINTOPOINT;
1162
1163 dev->iflink = p->link;
1164
1165 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1166 int strict = (ipv6_addr_type(&p->raddr) &
1167 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1168
1169 struct rt6_info *rt = rt6_lookup(dev_net(dev),
1170 &p->raddr, &p->laddr,
1171 p->link, strict);
1172
1173 if (rt == NULL)
1174 return;
1175
1176 if (rt->rt6i_dev) {
1177 dev->hard_header_len = rt->rt6i_dev->hard_header_len +
1178 sizeof (struct ipv6hdr);
1179
1180 dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
1181 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1182 dev->mtu-=8;
1183
1184 if (dev->mtu < IPV6_MIN_MTU)
1185 dev->mtu = IPV6_MIN_MTU;
1186 }
1187 dst_release(&rt->dst);
1188 }
1189}
1190
1191/**
1192 * ip6_tnl_change - update the tunnel parameters
1193 * @t: tunnel to be changed
1194 * @p: tunnel configuration parameters
1195 *
1196 * Description:
1197 * ip6_tnl_change() updates the tunnel parameters
1198 **/
1199
1200static int
1201ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
1202{
1203 ipv6_addr_copy(&t->parms.laddr, &p->laddr);
1204 ipv6_addr_copy(&t->parms.raddr, &p->raddr);
1205 t->parms.flags = p->flags;
1206 t->parms.hop_limit = p->hop_limit;
1207 t->parms.encap_limit = p->encap_limit;
1208 t->parms.flowinfo = p->flowinfo;
1209 t->parms.link = p->link;
1210 t->parms.proto = p->proto;
1211 ip6_tnl_dst_reset(t);
1212 ip6_tnl_link_config(t);
1213 return 0;
1214}
1215
1216/**
1217 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1218 * @dev: virtual device associated with tunnel
1219 * @ifr: parameters passed from userspace
1220 * @cmd: command to be performed
1221 *
1222 * Description:
1223 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1224 * from userspace.
1225 *
1226 * The possible commands are the following:
1227 * %SIOCGETTUNNEL: get tunnel parameters for device
1228 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1229 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1230 * %SIOCDELTUNNEL: delete tunnel
1231 *
1232 * The fallback device "ip6tnl0", created during module
1233 * initialization, can be used for creating other tunnel devices.
1234 *
1235 * Return:
1236 * 0 on success,
1237 * %-EFAULT if unable to copy data to or from userspace,
1238 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1239 * %-EINVAL if passed tunnel parameters are invalid,
1240 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1241 * %-ENODEV if attempting to change or delete a nonexisting device
1242 **/
1243
1244static int
1245ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1246{
1247 int err = 0;
1248 struct ip6_tnl_parm p;
1249 struct ip6_tnl *t = NULL;
1250 struct net *net = dev_net(dev);
1251 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1252
1253 switch (cmd) {
1254 case SIOCGETTUNNEL:
1255 if (dev == ip6n->fb_tnl_dev) {
1256 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1257 err = -EFAULT;
1258 break;
1259 }
1260 t = ip6_tnl_locate(net, &p, 0);
1261 }
1262 if (t == NULL)
1263 t = netdev_priv(dev);
1264 memcpy(&p, &t->parms, sizeof (p));
1265 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1266 err = -EFAULT;
1267 }
1268 break;
1269 case SIOCADDTUNNEL:
1270 case SIOCCHGTUNNEL:
1271 err = -EPERM;
1272 if (!capable(CAP_NET_ADMIN))
1273 break;
1274 err = -EFAULT;
1275 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1276 break;
1277 err = -EINVAL;
1278 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1279 p.proto != 0)
1280 break;
1281 t = ip6_tnl_locate(net, &p, cmd == SIOCADDTUNNEL);
1282 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1283 if (t != NULL) {
1284 if (t->dev != dev) {
1285 err = -EEXIST;
1286 break;
1287 }
1288 } else
1289 t = netdev_priv(dev);
1290
1291 ip6_tnl_unlink(ip6n, t);
1292 synchronize_net();
1293 err = ip6_tnl_change(t, &p);
1294 ip6_tnl_link(ip6n, t);
1295 netdev_state_change(dev);
1296 }
1297 if (t) {
1298 err = 0;
1299 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
1300 err = -EFAULT;
1301
1302 } else
1303 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1304 break;
1305 case SIOCDELTUNNEL:
1306 err = -EPERM;
1307 if (!capable(CAP_NET_ADMIN))
1308 break;
1309
1310 if (dev == ip6n->fb_tnl_dev) {
1311 err = -EFAULT;
1312 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1313 break;
1314 err = -ENOENT;
1315 if ((t = ip6_tnl_locate(net, &p, 0)) == NULL)
1316 break;
1317 err = -EPERM;
1318 if (t->dev == ip6n->fb_tnl_dev)
1319 break;
1320 dev = t->dev;
1321 }
1322 err = 0;
1323 unregister_netdevice(dev);
1324 break;
1325 default:
1326 err = -EINVAL;
1327 }
1328 return err;
1329}
1330
1331/**
1332 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1333 * @dev: virtual device associated with tunnel
1334 * @new_mtu: the new mtu
1335 *
1336 * Return:
1337 * 0 on success,
1338 * %-EINVAL if mtu too small
1339 **/
1340
1341static int
1342ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1343{
1344 if (new_mtu < IPV6_MIN_MTU) {
1345 return -EINVAL;
1346 }
1347 dev->mtu = new_mtu;
1348 return 0;
1349}
1350
1351
1352static const struct net_device_ops ip6_tnl_netdev_ops = {
1353 .ndo_uninit = ip6_tnl_dev_uninit,
1354 .ndo_start_xmit = ip6_tnl_xmit,
1355 .ndo_do_ioctl = ip6_tnl_ioctl,
1356 .ndo_change_mtu = ip6_tnl_change_mtu,
1357 .ndo_get_stats = ip6_get_stats,
1358};
1359
1360
1361/**
1362 * ip6_tnl_dev_setup - setup virtual tunnel device
1363 * @dev: virtual device associated with tunnel
1364 *
1365 * Description:
1366 * Initialize function pointers and device parameters
1367 **/
1368
1369static void ip6_tnl_dev_setup(struct net_device *dev)
1370{
1371 struct ip6_tnl *t;
1372
1373 dev->netdev_ops = &ip6_tnl_netdev_ops;
1374 dev->destructor = ip6_dev_free;
1375
1376 dev->type = ARPHRD_TUNNEL6;
1377 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1378 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1379 t = netdev_priv(dev);
1380 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1381 dev->mtu-=8;
1382 dev->flags |= IFF_NOARP;
1383 dev->addr_len = sizeof(struct in6_addr);
1384 dev->features |= NETIF_F_NETNS_LOCAL;
1385 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1386}
1387
1388
1389/**
1390 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1391 * @dev: virtual device associated with tunnel
1392 **/
1393
1394static inline int
1395ip6_tnl_dev_init_gen(struct net_device *dev)
1396{
1397 struct ip6_tnl *t = netdev_priv(dev);
1398
1399 t->dev = dev;
1400 strcpy(t->parms.name, dev->name);
1401 dev->tstats = alloc_percpu(struct pcpu_tstats);
1402 if (!dev->tstats)
1403 return -ENOMEM;
1404 return 0;
1405}
1406
1407/**
1408 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1409 * @dev: virtual device associated with tunnel
1410 **/
1411
1412static int ip6_tnl_dev_init(struct net_device *dev)
1413{
1414 struct ip6_tnl *t = netdev_priv(dev);
1415 int err = ip6_tnl_dev_init_gen(dev);
1416
1417 if (err)
1418 return err;
1419 ip6_tnl_link_config(t);
1420 return 0;
1421}
1422
1423/**
1424 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1425 * @dev: fallback device
1426 *
1427 * Return: 0
1428 **/
1429
1430static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1431{
1432 struct ip6_tnl *t = netdev_priv(dev);
1433 struct net *net = dev_net(dev);
1434 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1435 int err = ip6_tnl_dev_init_gen(dev);
1436
1437 if (err)
1438 return err;
1439
1440 t->parms.proto = IPPROTO_IPV6;
1441 dev_hold(dev);
1442 rcu_assign_pointer(ip6n->tnls_wc[0], t);
1443 return 0;
1444}
1445
1446static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1447 .handler = ip4ip6_rcv,
1448 .err_handler = ip4ip6_err,
1449 .priority = 1,
1450};
1451
1452static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1453 .handler = ip6ip6_rcv,
1454 .err_handler = ip6ip6_err,
1455 .priority = 1,
1456};
1457
1458static void __net_exit ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n)
1459{
1460 int h;
1461 struct ip6_tnl *t;
1462 LIST_HEAD(list);
1463
1464 for (h = 0; h < HASH_SIZE; h++) {
1465 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1466 while (t != NULL) {
1467 unregister_netdevice_queue(t->dev, &list);
1468 t = rtnl_dereference(t->next);
1469 }
1470 }
1471
1472 t = rtnl_dereference(ip6n->tnls_wc[0]);
1473 unregister_netdevice_queue(t->dev, &list);
1474 unregister_netdevice_many(&list);
1475}
1476
1477static int __net_init ip6_tnl_init_net(struct net *net)
1478{
1479 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1480 int err;
1481
1482 ip6n->tnls[0] = ip6n->tnls_wc;
1483 ip6n->tnls[1] = ip6n->tnls_r_l;
1484
1485 err = -ENOMEM;
1486 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1487 ip6_tnl_dev_setup);
1488
1489 if (!ip6n->fb_tnl_dev)
1490 goto err_alloc_dev;
1491 dev_net_set(ip6n->fb_tnl_dev, net);
1492
1493 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1494 if (err < 0)
1495 goto err_register;
1496
1497 err = register_netdev(ip6n->fb_tnl_dev);
1498 if (err < 0)
1499 goto err_register;
1500 return 0;
1501
1502err_register:
1503 ip6_dev_free(ip6n->fb_tnl_dev);
1504err_alloc_dev:
1505 return err;
1506}
1507
1508static void __net_exit ip6_tnl_exit_net(struct net *net)
1509{
1510 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1511
1512 rtnl_lock();
1513 ip6_tnl_destroy_tunnels(ip6n);
1514 rtnl_unlock();
1515}
1516
1517static struct pernet_operations ip6_tnl_net_ops = {
1518 .init = ip6_tnl_init_net,
1519 .exit = ip6_tnl_exit_net,
1520 .id = &ip6_tnl_net_id,
1521 .size = sizeof(struct ip6_tnl_net),
1522};
1523
1524/**
1525 * ip6_tunnel_init - register protocol and reserve needed resources
1526 *
1527 * Return: 0 on success
1528 **/
1529
1530static int __init ip6_tunnel_init(void)
1531{
1532 int err;
1533
1534 err = register_pernet_device(&ip6_tnl_net_ops);
1535 if (err < 0)
1536 goto out_pernet;
1537
1538 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1539 if (err < 0) {
1540 printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n");
1541 goto out_ip4ip6;
1542 }
1543
1544 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1545 if (err < 0) {
1546 printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n");
1547 goto out_ip6ip6;
1548 }
1549
1550 return 0;
1551
1552out_ip6ip6:
1553 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1554out_ip4ip6:
1555 unregister_pernet_device(&ip6_tnl_net_ops);
1556out_pernet:
1557 return err;
1558}
1559
1560/**
1561 * ip6_tunnel_cleanup - free resources and unregister protocol
1562 **/
1563
1564static void __exit ip6_tunnel_cleanup(void)
1565{
1566 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1567 printk(KERN_INFO "ip6_tunnel close: can't deregister ip4ip6\n");
1568
1569 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1570 printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n");
1571
1572 unregister_pernet_device(&ip6_tnl_net_ops);
1573}
1574
1575module_init(ip6_tunnel_init);
1576module_exit(ip6_tunnel_cleanup);