blob: ff9040c92556d20fb453b5ff815e402dc6c6ff60 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * IPv6 over IPv6 tunnel device
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Ville Nuorvala <vnuorval@tcs.hut.fi>
7 *
8 * $Id$
9 *
10 * Based on:
11 * linux/net/ipv6/sit.c
12 *
13 * RFC 2473
14 *
15 * This program is free software; you can redistribute it and/or
16 * modify it under the terms of the GNU General Public License
17 * as published by the Free Software Foundation; either version
18 * 2 of the License, or (at your option) any later version.
19 *
20 */
21
22#include <linux/config.h>
23#include <linux/module.h>
Randy Dunlap4fc268d2006-01-11 12:17:47 -080024#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <linux/errno.h>
26#include <linux/types.h>
27#include <linux/sockios.h>
28#include <linux/if.h>
29#include <linux/in.h>
30#include <linux/ip.h>
31#include <linux/if_tunnel.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
42#include <asm/uaccess.h>
43#include <asm/atomic.h>
44
45#include <net/ip.h>
46#include <net/ipv6.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070047#include <net/ip6_route.h>
48#include <net/addrconf.h>
49#include <net/ip6_tunnel.h>
50#include <net/xfrm.h>
51#include <net/dsfield.h>
52#include <net/inet_ecn.h>
53
54MODULE_AUTHOR("Ville Nuorvala");
55MODULE_DESCRIPTION("IPv6-in-IPv6 tunnel");
56MODULE_LICENSE("GPL");
57
58#define IPV6_TLV_TEL_DST_SIZE 8
59
60#ifdef IP6_TNL_DEBUG
61#define IP6_TNL_TRACE(x...) printk(KERN_DEBUG "%s:" x "\n", __FUNCTION__)
62#else
63#define IP6_TNL_TRACE(x...) do {;} while(0)
64#endif
65
66#define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
67
68#define HASH_SIZE 32
69
70#define HASH(addr) (((addr)->s6_addr32[0] ^ (addr)->s6_addr32[1] ^ \
71 (addr)->s6_addr32[2] ^ (addr)->s6_addr32[3]) & \
72 (HASH_SIZE - 1))
73
74static int ip6ip6_fb_tnl_dev_init(struct net_device *dev);
75static int ip6ip6_tnl_dev_init(struct net_device *dev);
76static void ip6ip6_tnl_dev_setup(struct net_device *dev);
77
78/* the IPv6 tunnel fallback device */
79static struct net_device *ip6ip6_fb_tnl_dev;
80
81
82/* lists for storing tunnels in use */
83static struct ip6_tnl *tnls_r_l[HASH_SIZE];
84static struct ip6_tnl *tnls_wc[1];
85static struct ip6_tnl **tnls[2] = { tnls_wc, tnls_r_l };
86
87/* lock for the tunnel lists */
88static DEFINE_RWLOCK(ip6ip6_lock);
89
90static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
91{
92 struct dst_entry *dst = t->dst_cache;
93
94 if (dst && dst->obsolete &&
95 dst->ops->check(dst, t->dst_cookie) == NULL) {
96 t->dst_cache = NULL;
97 dst_release(dst);
98 return NULL;
99 }
100
101 return dst;
102}
103
104static inline void ip6_tnl_dst_reset(struct ip6_tnl *t)
105{
106 dst_release(t->dst_cache);
107 t->dst_cache = NULL;
108}
109
110static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
111{
112 struct rt6_info *rt = (struct rt6_info *) dst;
113 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
114 dst_release(t->dst_cache);
115 t->dst_cache = dst;
116}
117
118/**
119 * ip6ip6_tnl_lookup - fetch tunnel matching the end-point addresses
120 * @remote: the address of the tunnel exit-point
121 * @local: the address of the tunnel entry-point
122 *
123 * Return:
124 * tunnel matching given end-points if found,
125 * else fallback tunnel if its device is up,
126 * else %NULL
127 **/
128
129static struct ip6_tnl *
130ip6ip6_tnl_lookup(struct in6_addr *remote, struct in6_addr *local)
131{
132 unsigned h0 = HASH(remote);
133 unsigned h1 = HASH(local);
134 struct ip6_tnl *t;
135
136 for (t = tnls_r_l[h0 ^ h1]; t; t = t->next) {
137 if (ipv6_addr_equal(local, &t->parms.laddr) &&
138 ipv6_addr_equal(remote, &t->parms.raddr) &&
139 (t->dev->flags & IFF_UP))
140 return t;
141 }
142 if ((t = tnls_wc[0]) != NULL && (t->dev->flags & IFF_UP))
143 return t;
144
145 return NULL;
146}
147
148/**
149 * ip6ip6_bucket - get head of list matching given tunnel parameters
150 * @p: parameters containing tunnel end-points
151 *
152 * Description:
153 * ip6ip6_bucket() returns the head of the list matching the
154 * &struct in6_addr entries laddr and raddr in @p.
155 *
156 * Return: head of IPv6 tunnel list
157 **/
158
159static struct ip6_tnl **
160ip6ip6_bucket(struct ip6_tnl_parm *p)
161{
162 struct in6_addr *remote = &p->raddr;
163 struct in6_addr *local = &p->laddr;
164 unsigned h = 0;
165 int prio = 0;
166
167 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
168 prio = 1;
169 h = HASH(remote) ^ HASH(local);
170 }
171 return &tnls[prio][h];
172}
173
174/**
175 * ip6ip6_tnl_link - add tunnel to hash table
176 * @t: tunnel to be added
177 **/
178
179static void
180ip6ip6_tnl_link(struct ip6_tnl *t)
181{
182 struct ip6_tnl **tp = ip6ip6_bucket(&t->parms);
183
184 t->next = *tp;
185 write_lock_bh(&ip6ip6_lock);
186 *tp = t;
187 write_unlock_bh(&ip6ip6_lock);
188}
189
190/**
191 * ip6ip6_tnl_unlink - remove tunnel from hash table
192 * @t: tunnel to be removed
193 **/
194
195static void
196ip6ip6_tnl_unlink(struct ip6_tnl *t)
197{
198 struct ip6_tnl **tp;
199
200 for (tp = ip6ip6_bucket(&t->parms); *tp; tp = &(*tp)->next) {
201 if (t == *tp) {
202 write_lock_bh(&ip6ip6_lock);
203 *tp = t->next;
204 write_unlock_bh(&ip6ip6_lock);
205 break;
206 }
207 }
208}
209
210/**
211 * ip6_tnl_create() - create a new tunnel
212 * @p: tunnel parameters
213 * @pt: pointer to new tunnel
214 *
215 * Description:
216 * Create tunnel matching given parameters.
217 *
218 * Return:
219 * 0 on success
220 **/
221
222static int
223ip6_tnl_create(struct ip6_tnl_parm *p, struct ip6_tnl **pt)
224{
225 struct net_device *dev;
226 struct ip6_tnl *t;
227 char name[IFNAMSIZ];
228 int err;
229
230 if (p->name[0]) {
231 strlcpy(name, p->name, IFNAMSIZ);
232 } else {
233 int i;
234 for (i = 1; i < IP6_TNL_MAX; i++) {
235 sprintf(name, "ip6tnl%d", i);
236 if (__dev_get_by_name(name) == NULL)
237 break;
238 }
239 if (i == IP6_TNL_MAX)
240 return -ENOBUFS;
241 }
242 dev = alloc_netdev(sizeof (*t), name, ip6ip6_tnl_dev_setup);
243 if (dev == NULL)
244 return -ENOMEM;
245
Patrick McHardy2941a482006-01-08 22:05:26 -0800246 t = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247 dev->init = ip6ip6_tnl_dev_init;
248 t->parms = *p;
249
250 if ((err = register_netdevice(dev)) < 0) {
251 free_netdev(dev);
252 return err;
253 }
254 dev_hold(dev);
255
256 ip6ip6_tnl_link(t);
257 *pt = t;
258 return 0;
259}
260
261/**
262 * ip6ip6_tnl_locate - find or create tunnel matching given parameters
263 * @p: tunnel parameters
264 * @create: != 0 if allowed to create new tunnel if no match found
265 *
266 * Description:
267 * ip6ip6_tnl_locate() first tries to locate an existing tunnel
268 * based on @parms. If this is unsuccessful, but @create is set a new
269 * tunnel device is created and registered for use.
270 *
271 * Return:
272 * 0 if tunnel located or created,
273 * -EINVAL if parameters incorrect,
274 * -ENODEV if no matching tunnel available
275 **/
276
277static int
278ip6ip6_tnl_locate(struct ip6_tnl_parm *p, struct ip6_tnl **pt, int create)
279{
280 struct in6_addr *remote = &p->raddr;
281 struct in6_addr *local = &p->laddr;
282 struct ip6_tnl *t;
283
284 if (p->proto != IPPROTO_IPV6)
285 return -EINVAL;
286
287 for (t = *ip6ip6_bucket(p); t; t = t->next) {
288 if (ipv6_addr_equal(local, &t->parms.laddr) &&
289 ipv6_addr_equal(remote, &t->parms.raddr)) {
290 *pt = t;
291 return (create ? -EEXIST : 0);
292 }
293 }
294 if (!create)
295 return -ENODEV;
296
297 return ip6_tnl_create(p, pt);
298}
299
300/**
301 * ip6ip6_tnl_dev_uninit - tunnel device uninitializer
302 * @dev: the device to be destroyed
303 *
304 * Description:
305 * ip6ip6_tnl_dev_uninit() removes tunnel from its list
306 **/
307
308static void
309ip6ip6_tnl_dev_uninit(struct net_device *dev)
310{
Patrick McHardy2941a482006-01-08 22:05:26 -0800311 struct ip6_tnl *t = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312
313 if (dev == ip6ip6_fb_tnl_dev) {
314 write_lock_bh(&ip6ip6_lock);
315 tnls_wc[0] = NULL;
316 write_unlock_bh(&ip6ip6_lock);
317 } else {
318 ip6ip6_tnl_unlink(t);
319 }
320 ip6_tnl_dst_reset(t);
321 dev_put(dev);
322}
323
324/**
325 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
326 * @skb: received socket buffer
327 *
328 * Return:
329 * 0 if none was found,
330 * else index to encapsulation limit
331 **/
332
333static __u16
334parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
335{
336 struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw;
337 __u8 nexthdr = ipv6h->nexthdr;
338 __u16 off = sizeof (*ipv6h);
339
340 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
341 __u16 optlen = 0;
342 struct ipv6_opt_hdr *hdr;
343 if (raw + off + sizeof (*hdr) > skb->data &&
344 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
345 break;
346
347 hdr = (struct ipv6_opt_hdr *) (raw + off);
348 if (nexthdr == NEXTHDR_FRAGMENT) {
349 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
350 if (frag_hdr->frag_off)
351 break;
352 optlen = 8;
353 } else if (nexthdr == NEXTHDR_AUTH) {
354 optlen = (hdr->hdrlen + 2) << 2;
355 } else {
356 optlen = ipv6_optlen(hdr);
357 }
358 if (nexthdr == NEXTHDR_DEST) {
359 __u16 i = off + 2;
360 while (1) {
361 struct ipv6_tlv_tnl_enc_lim *tel;
362
363 /* No more room for encapsulation limit */
364 if (i + sizeof (*tel) > off + optlen)
365 break;
366
367 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
368 /* return index of option if found and valid */
369 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
370 tel->length == 1)
371 return i;
372 /* else jump to next option */
373 if (tel->type)
374 i += tel->length + 2;
375 else
376 i++;
377 }
378 }
379 nexthdr = hdr->nexthdr;
380 off += optlen;
381 }
382 return 0;
383}
384
385/**
386 * ip6ip6_err - tunnel error handler
387 *
388 * Description:
389 * ip6ip6_err() should handle errors in the tunnel according
390 * to the specifications in RFC 2473.
391 **/
392
Herbert Xud2acc342006-03-28 01:12:13 -0800393static int
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
395 int type, int code, int offset, __u32 info)
396{
397 struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
398 struct ip6_tnl *t;
399 int rel_msg = 0;
400 int rel_type = ICMPV6_DEST_UNREACH;
401 int rel_code = ICMPV6_ADDR_UNREACH;
402 __u32 rel_info = 0;
403 __u16 len;
Herbert Xud2acc342006-03-28 01:12:13 -0800404 int err = -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405
406 /* If the packet doesn't contain the original IPv6 header we are
407 in trouble since we might need the source address for further
408 processing of the error. */
409
410 read_lock(&ip6ip6_lock);
411 if ((t = ip6ip6_tnl_lookup(&ipv6h->daddr, &ipv6h->saddr)) == NULL)
412 goto out;
413
Herbert Xud2acc342006-03-28 01:12:13 -0800414 err = 0;
415
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 switch (type) {
417 __u32 teli;
418 struct ipv6_tlv_tnl_enc_lim *tel;
419 __u32 mtu;
420 case ICMPV6_DEST_UNREACH:
421 if (net_ratelimit())
422 printk(KERN_WARNING
423 "%s: Path to destination invalid "
424 "or inactive!\n", t->parms.name);
425 rel_msg = 1;
426 break;
427 case ICMPV6_TIME_EXCEED:
428 if (code == ICMPV6_EXC_HOPLIMIT) {
429 if (net_ratelimit())
430 printk(KERN_WARNING
431 "%s: Too small hop limit or "
432 "routing loop in tunnel!\n",
433 t->parms.name);
434 rel_msg = 1;
435 }
436 break;
437 case ICMPV6_PARAMPROB:
438 /* ignore if parameter problem not caused by a tunnel
439 encapsulation limit sub-option */
440 if (code != ICMPV6_HDR_FIELD) {
441 break;
442 }
443 teli = parse_tlv_tnl_enc_lim(skb, skb->data);
444
445 if (teli && teli == ntohl(info) - 2) {
446 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
447 if (tel->encap_limit == 0) {
448 if (net_ratelimit())
449 printk(KERN_WARNING
450 "%s: Too small encapsulation "
451 "limit or routing loop in "
452 "tunnel!\n", t->parms.name);
453 rel_msg = 1;
454 }
455 }
456 break;
457 case ICMPV6_PKT_TOOBIG:
458 mtu = ntohl(info) - offset;
459 if (mtu < IPV6_MIN_MTU)
460 mtu = IPV6_MIN_MTU;
461 t->dev->mtu = mtu;
462
Al Virocc6cdac2006-02-18 16:02:18 -0500463 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 rel_type = ICMPV6_PKT_TOOBIG;
465 rel_code = 0;
466 rel_info = mtu;
467 rel_msg = 1;
468 }
469 break;
470 }
471 if (rel_msg && pskb_may_pull(skb, offset + sizeof (*ipv6h))) {
472 struct rt6_info *rt;
473 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
474 if (!skb2)
475 goto out;
476
477 dst_release(skb2->dst);
478 skb2->dst = NULL;
479 skb_pull(skb2, offset);
480 skb2->nh.raw = skb2->data;
481
482 /* Try to guess incoming interface */
483 rt = rt6_lookup(&skb2->nh.ipv6h->saddr, NULL, 0, 0);
484
485 if (rt && rt->rt6i_dev)
486 skb2->dev = rt->rt6i_dev;
487
488 icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
489
490 if (rt)
491 dst_release(&rt->u.dst);
492
493 kfree_skb(skb2);
494 }
495out:
496 read_unlock(&ip6ip6_lock);
Herbert Xud2acc342006-03-28 01:12:13 -0800497 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498}
499
500static inline void ip6ip6_ecn_decapsulate(struct ipv6hdr *outer_iph,
501 struct sk_buff *skb)
502{
503 struct ipv6hdr *inner_iph = skb->nh.ipv6h;
504
505 if (INET_ECN_is_ce(ipv6_get_dsfield(outer_iph)))
506 IP6_ECN_set_ce(inner_iph);
507}
508
509/**
510 * ip6ip6_rcv - decapsulate IPv6 packet and retransmit it locally
511 * @skb: received socket buffer
512 *
513 * Return: 0
514 **/
515
516static int
Herbert Xud2acc342006-03-28 01:12:13 -0800517ip6ip6_rcv(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 struct ipv6hdr *ipv6h;
520 struct ip6_tnl *t;
521
522 if (!pskb_may_pull(skb, sizeof (*ipv6h)))
523 goto discard;
524
525 ipv6h = skb->nh.ipv6h;
526
527 read_lock(&ip6ip6_lock);
528
529 if ((t = ip6ip6_tnl_lookup(&ipv6h->saddr, &ipv6h->daddr)) != NULL) {
530 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
Ken-ichirou MATSUZAWA9f0ede52005-11-09 13:08:29 -0800531 read_unlock(&ip6ip6_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 kfree_skb(skb);
533 return 0;
534 }
535
536 if (!(t->parms.flags & IP6_TNL_F_CAP_RCV)) {
537 t->stat.rx_dropped++;
538 read_unlock(&ip6ip6_lock);
539 goto discard;
540 }
541 secpath_reset(skb);
542 skb->mac.raw = skb->nh.raw;
543 skb->nh.raw = skb->data;
544 skb->protocol = htons(ETH_P_IPV6);
545 skb->pkt_type = PACKET_HOST;
546 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
547 skb->dev = t->dev;
548 dst_release(skb->dst);
549 skb->dst = NULL;
550 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
551 ipv6_copy_dscp(ipv6h, skb->nh.ipv6h);
552 ip6ip6_ecn_decapsulate(ipv6h, skb);
553 t->stat.rx_packets++;
554 t->stat.rx_bytes += skb->len;
555 netif_rx(skb);
556 read_unlock(&ip6ip6_lock);
557 return 0;
558 }
559 read_unlock(&ip6ip6_lock);
560 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0, skb->dev);
561discard:
562 return 1;
563}
564
565static inline struct ipv6_txoptions *create_tel(__u8 encap_limit)
566{
567 struct ipv6_tlv_tnl_enc_lim *tel;
568 struct ipv6_txoptions *opt;
569 __u8 *raw;
570
571 int opt_len = sizeof(*opt) + 8;
572
573 if (!(opt = kmalloc(opt_len, GFP_ATOMIC))) {
574 return NULL;
575 }
576 memset(opt, 0, opt_len);
577 opt->tot_len = opt_len;
578 opt->dst0opt = (struct ipv6_opt_hdr *) (opt + 1);
579 opt->opt_nflen = 8;
580
581 tel = (struct ipv6_tlv_tnl_enc_lim *) (opt->dst0opt + 1);
582 tel->type = IPV6_TLV_TNL_ENCAP_LIMIT;
583 tel->length = 1;
584 tel->encap_limit = encap_limit;
585
586 raw = (__u8 *) opt->dst0opt;
587 raw[5] = IPV6_TLV_PADN;
588 raw[6] = 1;
589
590 return opt;
591}
592
593/**
594 * ip6ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
595 * @t: the outgoing tunnel device
596 * @hdr: IPv6 header from the incoming packet
597 *
598 * Description:
599 * Avoid trivial tunneling loop by checking that tunnel exit-point
600 * doesn't match source of incoming packet.
601 *
602 * Return:
603 * 1 if conflict,
604 * 0 else
605 **/
606
607static inline int
608ip6ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
609{
610 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
611}
612
613/**
614 * ip6ip6_tnl_xmit - encapsulate packet and send
615 * @skb: the outgoing socket buffer
616 * @dev: the outgoing tunnel device
617 *
618 * Description:
619 * Build new header and do some sanity checks on the packet before sending
620 * it.
621 *
622 * Return:
623 * 0
624 **/
625
626static int
627ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
628{
Patrick McHardy2941a482006-01-08 22:05:26 -0800629 struct ip6_tnl *t = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630 struct net_device_stats *stats = &t->stat;
631 struct ipv6hdr *ipv6h = skb->nh.ipv6h;
632 struct ipv6_txoptions *opt = NULL;
633 int encap_limit = -1;
634 __u16 offset;
635 struct flowi fl;
636 struct dst_entry *dst;
637 struct net_device *tdev;
638 int mtu;
639 int max_headroom = sizeof(struct ipv6hdr);
640 u8 proto;
641 int err;
642 int pkt_len;
643 int dsfield;
644
645 if (t->recursion++) {
646 stats->collisions++;
647 goto tx_err;
648 }
649 if (skb->protocol != htons(ETH_P_IPV6) ||
650 !(t->parms.flags & IP6_TNL_F_CAP_XMIT) ||
651 ip6ip6_tnl_addr_conflict(t, ipv6h)) {
652 goto tx_err;
653 }
654 if ((offset = parse_tlv_tnl_enc_lim(skb, skb->nh.raw)) > 0) {
655 struct ipv6_tlv_tnl_enc_lim *tel;
656 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->nh.raw[offset];
657 if (tel->encap_limit == 0) {
658 icmpv6_send(skb, ICMPV6_PARAMPROB,
659 ICMPV6_HDR_FIELD, offset + 2, skb->dev);
660 goto tx_err;
661 }
662 encap_limit = tel->encap_limit - 1;
663 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
664 encap_limit = t->parms.encap_limit;
665 }
666 memcpy(&fl, &t->fl, sizeof (fl));
667 proto = fl.proto;
668
669 dsfield = ipv6_get_dsfield(ipv6h);
670 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
671 fl.fl6_flowlabel |= (*(__u32 *) ipv6h & IPV6_TCLASS_MASK);
672 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
673 fl.fl6_flowlabel |= (*(__u32 *) ipv6h & IPV6_FLOWLABEL_MASK);
674
675 if (encap_limit >= 0 && (opt = create_tel(encap_limit)) == NULL)
676 goto tx_err;
677
678 if ((dst = ip6_tnl_dst_check(t)) != NULL)
679 dst_hold(dst);
Patrick McHardya57ebc92005-09-08 14:27:47 -0700680 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681 dst = ip6_route_output(NULL, &fl);
682
Patrick McHardya57ebc92005-09-08 14:27:47 -0700683 if (dst->error || xfrm_lookup(&dst, &fl, NULL, 0) < 0)
684 goto tx_err_link_failure;
685 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686
687 tdev = dst->dev;
688
689 if (tdev == dev) {
690 stats->collisions++;
691 if (net_ratelimit())
692 printk(KERN_WARNING
693 "%s: Local routing loop detected!\n",
694 t->parms.name);
695 goto tx_err_dst_release;
696 }
697 mtu = dst_mtu(dst) - sizeof (*ipv6h);
698 if (opt) {
699 max_headroom += 8;
700 mtu -= 8;
701 }
702 if (mtu < IPV6_MIN_MTU)
703 mtu = IPV6_MIN_MTU;
704 if (skb->dst && mtu < dst_mtu(skb->dst)) {
705 struct rt6_info *rt = (struct rt6_info *) skb->dst;
706 rt->rt6i_flags |= RTF_MODIFIED;
707 rt->u.dst.metrics[RTAX_MTU-1] = mtu;
708 }
709 if (skb->len > mtu) {
710 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
711 goto tx_err_dst_release;
712 }
713
714 /*
715 * Okay, now see if we can stuff it in the buffer as-is.
716 */
717 max_headroom += LL_RESERVED_SPACE(tdev);
718
719 if (skb_headroom(skb) < max_headroom ||
720 skb_cloned(skb) || skb_shared(skb)) {
721 struct sk_buff *new_skb;
722
723 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
724 goto tx_err_dst_release;
725
726 if (skb->sk)
727 skb_set_owner_w(new_skb, skb->sk);
728 kfree_skb(skb);
729 skb = new_skb;
730 }
731 dst_release(skb->dst);
732 skb->dst = dst_clone(dst);
733
734 skb->h.raw = skb->nh.raw;
735
736 if (opt)
737 ipv6_push_nfrag_opts(skb, opt, &proto, NULL);
738
739 skb->nh.raw = skb_push(skb, sizeof(struct ipv6hdr));
740 ipv6h = skb->nh.ipv6h;
741 *(u32*)ipv6h = fl.fl6_flowlabel | htonl(0x60000000);
742 dsfield = INET_ECN_encapsulate(0, dsfield);
743 ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
744 ipv6h->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
745 ipv6h->hop_limit = t->parms.hop_limit;
746 ipv6h->nexthdr = proto;
747 ipv6_addr_copy(&ipv6h->saddr, &fl.fl6_src);
748 ipv6_addr_copy(&ipv6h->daddr, &fl.fl6_dst);
749 nf_reset(skb);
750 pkt_len = skb->len;
751 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL,
752 skb->dst->dev, dst_output);
753
754 if (err == NET_XMIT_SUCCESS || err == NET_XMIT_CN) {
755 stats->tx_bytes += pkt_len;
756 stats->tx_packets++;
757 } else {
758 stats->tx_errors++;
759 stats->tx_aborted_errors++;
760 }
761 ip6_tnl_dst_store(t, dst);
762
Jesper Juhla51482b2005-11-08 09:41:34 -0800763 kfree(opt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764
765 t->recursion--;
766 return 0;
767tx_err_link_failure:
768 stats->tx_carrier_errors++;
769 dst_link_failure(skb);
770tx_err_dst_release:
771 dst_release(dst);
Jesper Juhla51482b2005-11-08 09:41:34 -0800772 kfree(opt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773tx_err:
774 stats->tx_errors++;
775 stats->tx_dropped++;
776 kfree_skb(skb);
777 t->recursion--;
778 return 0;
779}
780
781static void ip6_tnl_set_cap(struct ip6_tnl *t)
782{
783 struct ip6_tnl_parm *p = &t->parms;
784 struct in6_addr *laddr = &p->laddr;
785 struct in6_addr *raddr = &p->raddr;
786 int ltype = ipv6_addr_type(laddr);
787 int rtype = ipv6_addr_type(raddr);
788
789 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
790
791 if (ltype != IPV6_ADDR_ANY && rtype != IPV6_ADDR_ANY &&
792 ((ltype|rtype) &
793 (IPV6_ADDR_UNICAST|
794 IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL|
795 IPV6_ADDR_MAPPED|IPV6_ADDR_RESERVED)) == IPV6_ADDR_UNICAST) {
796 struct net_device *ldev = NULL;
797 int l_ok = 1;
798 int r_ok = 1;
799
800 if (p->link)
801 ldev = dev_get_by_index(p->link);
802
803 if (ltype&IPV6_ADDR_UNICAST && !ipv6_chk_addr(laddr, ldev, 0))
804 l_ok = 0;
805
806 if (rtype&IPV6_ADDR_UNICAST && ipv6_chk_addr(raddr, NULL, 0))
807 r_ok = 0;
808
809 if (l_ok && r_ok) {
810 if (ltype&IPV6_ADDR_UNICAST)
811 p->flags |= IP6_TNL_F_CAP_XMIT;
812 if (rtype&IPV6_ADDR_UNICAST)
813 p->flags |= IP6_TNL_F_CAP_RCV;
814 }
815 if (ldev)
816 dev_put(ldev);
817 }
818}
819
820static void ip6ip6_tnl_link_config(struct ip6_tnl *t)
821{
822 struct net_device *dev = t->dev;
823 struct ip6_tnl_parm *p = &t->parms;
824 struct flowi *fl = &t->fl;
825
826 memcpy(&dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
827 memcpy(&dev->broadcast, &p->raddr, sizeof(struct in6_addr));
828
829 /* Set up flowi template */
830 ipv6_addr_copy(&fl->fl6_src, &p->laddr);
831 ipv6_addr_copy(&fl->fl6_dst, &p->raddr);
832 fl->oif = p->link;
833 fl->fl6_flowlabel = 0;
834
835 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
836 fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
837 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
838 fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
839
840 ip6_tnl_set_cap(t);
841
842 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
843 dev->flags |= IFF_POINTOPOINT;
844 else
845 dev->flags &= ~IFF_POINTOPOINT;
846
847 dev->iflink = p->link;
848
849 if (p->flags & IP6_TNL_F_CAP_XMIT) {
850 struct rt6_info *rt = rt6_lookup(&p->raddr, &p->laddr,
851 p->link, 0);
852
853 if (rt == NULL)
854 return;
855
856 if (rt->rt6i_dev) {
857 dev->hard_header_len = rt->rt6i_dev->hard_header_len +
858 sizeof (struct ipv6hdr);
859
860 dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
861
862 if (dev->mtu < IPV6_MIN_MTU)
863 dev->mtu = IPV6_MIN_MTU;
864 }
865 dst_release(&rt->u.dst);
866 }
867}
868
869/**
870 * ip6ip6_tnl_change - update the tunnel parameters
871 * @t: tunnel to be changed
872 * @p: tunnel configuration parameters
873 * @active: != 0 if tunnel is ready for use
874 *
875 * Description:
876 * ip6ip6_tnl_change() updates the tunnel parameters
877 **/
878
879static int
880ip6ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
881{
882 ipv6_addr_copy(&t->parms.laddr, &p->laddr);
883 ipv6_addr_copy(&t->parms.raddr, &p->raddr);
884 t->parms.flags = p->flags;
885 t->parms.hop_limit = p->hop_limit;
886 t->parms.encap_limit = p->encap_limit;
887 t->parms.flowinfo = p->flowinfo;
Gabor Fekete8181b8c2005-06-08 14:54:38 -0700888 t->parms.link = p->link;
Hugo Santos0c088892006-02-24 13:16:25 -0800889 ip6_tnl_dst_reset(t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890 ip6ip6_tnl_link_config(t);
891 return 0;
892}
893
894/**
895 * ip6ip6_tnl_ioctl - configure ipv6 tunnels from userspace
896 * @dev: virtual device associated with tunnel
897 * @ifr: parameters passed from userspace
898 * @cmd: command to be performed
899 *
900 * Description:
901 * ip6ip6_tnl_ioctl() is used for managing IPv6 tunnels
902 * from userspace.
903 *
904 * The possible commands are the following:
905 * %SIOCGETTUNNEL: get tunnel parameters for device
906 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
907 * %SIOCCHGTUNNEL: change tunnel parameters to those given
908 * %SIOCDELTUNNEL: delete tunnel
909 *
910 * The fallback device "ip6tnl0", created during module
911 * initialization, can be used for creating other tunnel devices.
912 *
913 * Return:
914 * 0 on success,
915 * %-EFAULT if unable to copy data to or from userspace,
916 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
917 * %-EINVAL if passed tunnel parameters are invalid,
918 * %-EEXIST if changing a tunnel's parameters would cause a conflict
919 * %-ENODEV if attempting to change or delete a nonexisting device
920 **/
921
922static int
923ip6ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
924{
925 int err = 0;
926 int create;
927 struct ip6_tnl_parm p;
928 struct ip6_tnl *t = NULL;
929
930 switch (cmd) {
931 case SIOCGETTUNNEL:
932 if (dev == ip6ip6_fb_tnl_dev) {
933 if (copy_from_user(&p,
934 ifr->ifr_ifru.ifru_data,
935 sizeof (p))) {
936 err = -EFAULT;
937 break;
938 }
939 if ((err = ip6ip6_tnl_locate(&p, &t, 0)) == -ENODEV)
Patrick McHardy2941a482006-01-08 22:05:26 -0800940 t = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941 else if (err)
942 break;
943 } else
Patrick McHardy2941a482006-01-08 22:05:26 -0800944 t = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945
946 memcpy(&p, &t->parms, sizeof (p));
947 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
948 err = -EFAULT;
949 }
950 break;
951 case SIOCADDTUNNEL:
952 case SIOCCHGTUNNEL:
953 err = -EPERM;
954 create = (cmd == SIOCADDTUNNEL);
955 if (!capable(CAP_NET_ADMIN))
956 break;
957 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
958 err = -EFAULT;
959 break;
960 }
961 if (!create && dev != ip6ip6_fb_tnl_dev) {
Patrick McHardy2941a482006-01-08 22:05:26 -0800962 t = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 }
964 if (!t && (err = ip6ip6_tnl_locate(&p, &t, create))) {
965 break;
966 }
967 if (cmd == SIOCCHGTUNNEL) {
968 if (t->dev != dev) {
969 err = -EEXIST;
970 break;
971 }
972 ip6ip6_tnl_unlink(t);
973 err = ip6ip6_tnl_change(t, &p);
974 ip6ip6_tnl_link(t);
975 netdev_state_change(dev);
976 }
977 if (copy_to_user(ifr->ifr_ifru.ifru_data,
978 &t->parms, sizeof (p))) {
979 err = -EFAULT;
980 } else {
981 err = 0;
982 }
983 break;
984 case SIOCDELTUNNEL:
985 err = -EPERM;
986 if (!capable(CAP_NET_ADMIN))
987 break;
988
989 if (dev == ip6ip6_fb_tnl_dev) {
990 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data,
991 sizeof (p))) {
992 err = -EFAULT;
993 break;
994 }
995 err = ip6ip6_tnl_locate(&p, &t, 0);
996 if (err)
997 break;
Patrick McHardy2941a482006-01-08 22:05:26 -0800998 if (t == netdev_priv(ip6ip6_fb_tnl_dev)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999 err = -EPERM;
1000 break;
1001 }
1002 } else {
Patrick McHardy2941a482006-01-08 22:05:26 -08001003 t = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 }
1005 err = unregister_netdevice(t->dev);
1006 break;
1007 default:
1008 err = -EINVAL;
1009 }
1010 return err;
1011}
1012
1013/**
1014 * ip6ip6_tnl_get_stats - return the stats for tunnel device
1015 * @dev: virtual device associated with tunnel
1016 *
1017 * Return: stats for device
1018 **/
1019
1020static struct net_device_stats *
1021ip6ip6_tnl_get_stats(struct net_device *dev)
1022{
Patrick McHardy2941a482006-01-08 22:05:26 -08001023 return &(((struct ip6_tnl *)netdev_priv(dev))->stat);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024}
1025
1026/**
1027 * ip6ip6_tnl_change_mtu - change mtu manually for tunnel device
1028 * @dev: virtual device associated with tunnel
1029 * @new_mtu: the new mtu
1030 *
1031 * Return:
1032 * 0 on success,
1033 * %-EINVAL if mtu too small
1034 **/
1035
1036static int
1037ip6ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1038{
1039 if (new_mtu < IPV6_MIN_MTU) {
1040 return -EINVAL;
1041 }
1042 dev->mtu = new_mtu;
1043 return 0;
1044}
1045
1046/**
1047 * ip6ip6_tnl_dev_setup - setup virtual tunnel device
1048 * @dev: virtual device associated with tunnel
1049 *
1050 * Description:
1051 * Initialize function pointers and device parameters
1052 **/
1053
1054static void ip6ip6_tnl_dev_setup(struct net_device *dev)
1055{
1056 SET_MODULE_OWNER(dev);
1057 dev->uninit = ip6ip6_tnl_dev_uninit;
1058 dev->destructor = free_netdev;
1059 dev->hard_start_xmit = ip6ip6_tnl_xmit;
1060 dev->get_stats = ip6ip6_tnl_get_stats;
1061 dev->do_ioctl = ip6ip6_tnl_ioctl;
1062 dev->change_mtu = ip6ip6_tnl_change_mtu;
1063
1064 dev->type = ARPHRD_TUNNEL6;
1065 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1066 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1067 dev->flags |= IFF_NOARP;
1068 dev->addr_len = sizeof(struct in6_addr);
1069}
1070
1071
1072/**
1073 * ip6ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1074 * @dev: virtual device associated with tunnel
1075 **/
1076
1077static inline void
1078ip6ip6_tnl_dev_init_gen(struct net_device *dev)
1079{
Patrick McHardy2941a482006-01-08 22:05:26 -08001080 struct ip6_tnl *t = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 t->fl.proto = IPPROTO_IPV6;
1082 t->dev = dev;
1083 strcpy(t->parms.name, dev->name);
1084}
1085
1086/**
1087 * ip6ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1088 * @dev: virtual device associated with tunnel
1089 **/
1090
1091static int
1092ip6ip6_tnl_dev_init(struct net_device *dev)
1093{
Patrick McHardy2941a482006-01-08 22:05:26 -08001094 struct ip6_tnl *t = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095 ip6ip6_tnl_dev_init_gen(dev);
1096 ip6ip6_tnl_link_config(t);
1097 return 0;
1098}
1099
1100/**
1101 * ip6ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1102 * @dev: fallback device
1103 *
1104 * Return: 0
1105 **/
1106
1107static int
1108ip6ip6_fb_tnl_dev_init(struct net_device *dev)
1109{
Patrick McHardy2941a482006-01-08 22:05:26 -08001110 struct ip6_tnl *t = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111 ip6ip6_tnl_dev_init_gen(dev);
1112 dev_hold(dev);
1113 tnls_wc[0] = t;
1114 return 0;
1115}
1116
1117static struct xfrm6_tunnel ip6ip6_handler = {
Patrick McHardy03037702005-07-19 14:03:34 -07001118 .handler = ip6ip6_rcv,
1119 .err_handler = ip6ip6_err,
Herbert Xud2acc342006-03-28 01:12:13 -08001120 .priority = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121};
1122
1123/**
1124 * ip6_tunnel_init - register protocol and reserve needed resources
1125 *
1126 * Return: 0 on success
1127 **/
1128
1129static int __init ip6_tunnel_init(void)
1130{
1131 int err;
1132
Herbert Xud2acc342006-03-28 01:12:13 -08001133 if (xfrm6_tunnel_register(&ip6ip6_handler)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134 printk(KERN_ERR "ip6ip6 init: can't register tunnel\n");
1135 return -EAGAIN;
1136 }
1137 ip6ip6_fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1138 ip6ip6_tnl_dev_setup);
1139
1140 if (!ip6ip6_fb_tnl_dev) {
1141 err = -ENOMEM;
1142 goto fail;
1143 }
1144 ip6ip6_fb_tnl_dev->init = ip6ip6_fb_tnl_dev_init;
1145
1146 if ((err = register_netdev(ip6ip6_fb_tnl_dev))) {
1147 free_netdev(ip6ip6_fb_tnl_dev);
1148 goto fail;
1149 }
1150 return 0;
1151fail:
Herbert Xud2acc342006-03-28 01:12:13 -08001152 xfrm6_tunnel_deregister(&ip6ip6_handler);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153 return err;
1154}
1155
1156/**
1157 * ip6_tunnel_cleanup - free resources and unregister protocol
1158 **/
1159
1160static void __exit ip6_tunnel_cleanup(void)
1161{
Herbert Xud2acc342006-03-28 01:12:13 -08001162 if (xfrm6_tunnel_deregister(&ip6ip6_handler))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163 printk(KERN_INFO "ip6ip6 close: can't deregister tunnel\n");
1164
1165 unregister_netdev(ip6ip6_fb_tnl_dev);
1166}
1167
1168module_init(ip6_tunnel_init);
1169module_exit(ip6_tunnel_cleanup);