blob: ba3b0c267f759565c3eab810e3ed8bbbaaa91bfd [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>
24#include <linux/errno.h>
25#include <linux/types.h>
26#include <linux/sockios.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
41#include <asm/uaccess.h>
42#include <asm/atomic.h>
43
44#include <net/ip.h>
45#include <net/ipv6.h>
46#include <net/protocol.h>
47#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
246 t = dev->priv;
247 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{
311 struct ip6_tnl *t = dev->priv;
312
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
393static void
394ip6ip6_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;
404
405 /* If the packet doesn't contain the original IPv6 header we are
406 in trouble since we might need the source address for further
407 processing of the error. */
408
409 read_lock(&ip6ip6_lock);
410 if ((t = ip6ip6_tnl_lookup(&ipv6h->daddr, &ipv6h->saddr)) == NULL)
411 goto out;
412
413 switch (type) {
414 __u32 teli;
415 struct ipv6_tlv_tnl_enc_lim *tel;
416 __u32 mtu;
417 case ICMPV6_DEST_UNREACH:
418 if (net_ratelimit())
419 printk(KERN_WARNING
420 "%s: Path to destination invalid "
421 "or inactive!\n", t->parms.name);
422 rel_msg = 1;
423 break;
424 case ICMPV6_TIME_EXCEED:
425 if (code == ICMPV6_EXC_HOPLIMIT) {
426 if (net_ratelimit())
427 printk(KERN_WARNING
428 "%s: Too small hop limit or "
429 "routing loop in tunnel!\n",
430 t->parms.name);
431 rel_msg = 1;
432 }
433 break;
434 case ICMPV6_PARAMPROB:
435 /* ignore if parameter problem not caused by a tunnel
436 encapsulation limit sub-option */
437 if (code != ICMPV6_HDR_FIELD) {
438 break;
439 }
440 teli = parse_tlv_tnl_enc_lim(skb, skb->data);
441
442 if (teli && teli == ntohl(info) - 2) {
443 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
444 if (tel->encap_limit == 0) {
445 if (net_ratelimit())
446 printk(KERN_WARNING
447 "%s: Too small encapsulation "
448 "limit or routing loop in "
449 "tunnel!\n", t->parms.name);
450 rel_msg = 1;
451 }
452 }
453 break;
454 case ICMPV6_PKT_TOOBIG:
455 mtu = ntohl(info) - offset;
456 if (mtu < IPV6_MIN_MTU)
457 mtu = IPV6_MIN_MTU;
458 t->dev->mtu = mtu;
459
460 if ((len = sizeof (*ipv6h) + ipv6h->payload_len) > mtu) {
461 rel_type = ICMPV6_PKT_TOOBIG;
462 rel_code = 0;
463 rel_info = mtu;
464 rel_msg = 1;
465 }
466 break;
467 }
468 if (rel_msg && pskb_may_pull(skb, offset + sizeof (*ipv6h))) {
469 struct rt6_info *rt;
470 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
471 if (!skb2)
472 goto out;
473
474 dst_release(skb2->dst);
475 skb2->dst = NULL;
476 skb_pull(skb2, offset);
477 skb2->nh.raw = skb2->data;
478
479 /* Try to guess incoming interface */
480 rt = rt6_lookup(&skb2->nh.ipv6h->saddr, NULL, 0, 0);
481
482 if (rt && rt->rt6i_dev)
483 skb2->dev = rt->rt6i_dev;
484
485 icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
486
487 if (rt)
488 dst_release(&rt->u.dst);
489
490 kfree_skb(skb2);
491 }
492out:
493 read_unlock(&ip6ip6_lock);
494}
495
496static inline void ip6ip6_ecn_decapsulate(struct ipv6hdr *outer_iph,
497 struct sk_buff *skb)
498{
499 struct ipv6hdr *inner_iph = skb->nh.ipv6h;
500
501 if (INET_ECN_is_ce(ipv6_get_dsfield(outer_iph)))
502 IP6_ECN_set_ce(inner_iph);
503}
504
505/**
506 * ip6ip6_rcv - decapsulate IPv6 packet and retransmit it locally
507 * @skb: received socket buffer
508 *
509 * Return: 0
510 **/
511
512static int
513ip6ip6_rcv(struct sk_buff **pskb, unsigned int *nhoffp)
514{
515 struct sk_buff *skb = *pskb;
516 struct ipv6hdr *ipv6h;
517 struct ip6_tnl *t;
518
519 if (!pskb_may_pull(skb, sizeof (*ipv6h)))
520 goto discard;
521
522 ipv6h = skb->nh.ipv6h;
523
524 read_lock(&ip6ip6_lock);
525
526 if ((t = ip6ip6_tnl_lookup(&ipv6h->saddr, &ipv6h->daddr)) != NULL) {
527 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
528 kfree_skb(skb);
529 return 0;
530 }
531
532 if (!(t->parms.flags & IP6_TNL_F_CAP_RCV)) {
533 t->stat.rx_dropped++;
534 read_unlock(&ip6ip6_lock);
535 goto discard;
536 }
537 secpath_reset(skb);
538 skb->mac.raw = skb->nh.raw;
539 skb->nh.raw = skb->data;
540 skb->protocol = htons(ETH_P_IPV6);
541 skb->pkt_type = PACKET_HOST;
542 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
543 skb->dev = t->dev;
544 dst_release(skb->dst);
545 skb->dst = NULL;
546 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
547 ipv6_copy_dscp(ipv6h, skb->nh.ipv6h);
548 ip6ip6_ecn_decapsulate(ipv6h, skb);
549 t->stat.rx_packets++;
550 t->stat.rx_bytes += skb->len;
551 netif_rx(skb);
552 read_unlock(&ip6ip6_lock);
553 return 0;
554 }
555 read_unlock(&ip6ip6_lock);
556 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0, skb->dev);
557discard:
558 return 1;
559}
560
561static inline struct ipv6_txoptions *create_tel(__u8 encap_limit)
562{
563 struct ipv6_tlv_tnl_enc_lim *tel;
564 struct ipv6_txoptions *opt;
565 __u8 *raw;
566
567 int opt_len = sizeof(*opt) + 8;
568
569 if (!(opt = kmalloc(opt_len, GFP_ATOMIC))) {
570 return NULL;
571 }
572 memset(opt, 0, opt_len);
573 opt->tot_len = opt_len;
574 opt->dst0opt = (struct ipv6_opt_hdr *) (opt + 1);
575 opt->opt_nflen = 8;
576
577 tel = (struct ipv6_tlv_tnl_enc_lim *) (opt->dst0opt + 1);
578 tel->type = IPV6_TLV_TNL_ENCAP_LIMIT;
579 tel->length = 1;
580 tel->encap_limit = encap_limit;
581
582 raw = (__u8 *) opt->dst0opt;
583 raw[5] = IPV6_TLV_PADN;
584 raw[6] = 1;
585
586 return opt;
587}
588
589/**
590 * ip6ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
591 * @t: the outgoing tunnel device
592 * @hdr: IPv6 header from the incoming packet
593 *
594 * Description:
595 * Avoid trivial tunneling loop by checking that tunnel exit-point
596 * doesn't match source of incoming packet.
597 *
598 * Return:
599 * 1 if conflict,
600 * 0 else
601 **/
602
603static inline int
604ip6ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
605{
606 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
607}
608
609/**
610 * ip6ip6_tnl_xmit - encapsulate packet and send
611 * @skb: the outgoing socket buffer
612 * @dev: the outgoing tunnel device
613 *
614 * Description:
615 * Build new header and do some sanity checks on the packet before sending
616 * it.
617 *
618 * Return:
619 * 0
620 **/
621
622static int
623ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
624{
625 struct ip6_tnl *t = (struct ip6_tnl *) dev->priv;
626 struct net_device_stats *stats = &t->stat;
627 struct ipv6hdr *ipv6h = skb->nh.ipv6h;
628 struct ipv6_txoptions *opt = NULL;
629 int encap_limit = -1;
630 __u16 offset;
631 struct flowi fl;
632 struct dst_entry *dst;
633 struct net_device *tdev;
634 int mtu;
635 int max_headroom = sizeof(struct ipv6hdr);
636 u8 proto;
637 int err;
638 int pkt_len;
639 int dsfield;
640
641 if (t->recursion++) {
642 stats->collisions++;
643 goto tx_err;
644 }
645 if (skb->protocol != htons(ETH_P_IPV6) ||
646 !(t->parms.flags & IP6_TNL_F_CAP_XMIT) ||
647 ip6ip6_tnl_addr_conflict(t, ipv6h)) {
648 goto tx_err;
649 }
650 if ((offset = parse_tlv_tnl_enc_lim(skb, skb->nh.raw)) > 0) {
651 struct ipv6_tlv_tnl_enc_lim *tel;
652 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->nh.raw[offset];
653 if (tel->encap_limit == 0) {
654 icmpv6_send(skb, ICMPV6_PARAMPROB,
655 ICMPV6_HDR_FIELD, offset + 2, skb->dev);
656 goto tx_err;
657 }
658 encap_limit = tel->encap_limit - 1;
659 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
660 encap_limit = t->parms.encap_limit;
661 }
662 memcpy(&fl, &t->fl, sizeof (fl));
663 proto = fl.proto;
664
665 dsfield = ipv6_get_dsfield(ipv6h);
666 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
667 fl.fl6_flowlabel |= (*(__u32 *) ipv6h & IPV6_TCLASS_MASK);
668 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
669 fl.fl6_flowlabel |= (*(__u32 *) ipv6h & IPV6_FLOWLABEL_MASK);
670
671 if (encap_limit >= 0 && (opt = create_tel(encap_limit)) == NULL)
672 goto tx_err;
673
674 if ((dst = ip6_tnl_dst_check(t)) != NULL)
675 dst_hold(dst);
676 else
677 dst = ip6_route_output(NULL, &fl);
678
679 if (dst->error || xfrm_lookup(&dst, &fl, NULL, 0) < 0)
680 goto tx_err_link_failure;
681
682 tdev = dst->dev;
683
684 if (tdev == dev) {
685 stats->collisions++;
686 if (net_ratelimit())
687 printk(KERN_WARNING
688 "%s: Local routing loop detected!\n",
689 t->parms.name);
690 goto tx_err_dst_release;
691 }
692 mtu = dst_mtu(dst) - sizeof (*ipv6h);
693 if (opt) {
694 max_headroom += 8;
695 mtu -= 8;
696 }
697 if (mtu < IPV6_MIN_MTU)
698 mtu = IPV6_MIN_MTU;
699 if (skb->dst && mtu < dst_mtu(skb->dst)) {
700 struct rt6_info *rt = (struct rt6_info *) skb->dst;
701 rt->rt6i_flags |= RTF_MODIFIED;
702 rt->u.dst.metrics[RTAX_MTU-1] = mtu;
703 }
704 if (skb->len > mtu) {
705 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
706 goto tx_err_dst_release;
707 }
708
709 /*
710 * Okay, now see if we can stuff it in the buffer as-is.
711 */
712 max_headroom += LL_RESERVED_SPACE(tdev);
713
714 if (skb_headroom(skb) < max_headroom ||
715 skb_cloned(skb) || skb_shared(skb)) {
716 struct sk_buff *new_skb;
717
718 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
719 goto tx_err_dst_release;
720
721 if (skb->sk)
722 skb_set_owner_w(new_skb, skb->sk);
723 kfree_skb(skb);
724 skb = new_skb;
725 }
726 dst_release(skb->dst);
727 skb->dst = dst_clone(dst);
728
729 skb->h.raw = skb->nh.raw;
730
731 if (opt)
732 ipv6_push_nfrag_opts(skb, opt, &proto, NULL);
733
734 skb->nh.raw = skb_push(skb, sizeof(struct ipv6hdr));
735 ipv6h = skb->nh.ipv6h;
736 *(u32*)ipv6h = fl.fl6_flowlabel | htonl(0x60000000);
737 dsfield = INET_ECN_encapsulate(0, dsfield);
738 ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
739 ipv6h->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
740 ipv6h->hop_limit = t->parms.hop_limit;
741 ipv6h->nexthdr = proto;
742 ipv6_addr_copy(&ipv6h->saddr, &fl.fl6_src);
743 ipv6_addr_copy(&ipv6h->daddr, &fl.fl6_dst);
744 nf_reset(skb);
745 pkt_len = skb->len;
746 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL,
747 skb->dst->dev, dst_output);
748
749 if (err == NET_XMIT_SUCCESS || err == NET_XMIT_CN) {
750 stats->tx_bytes += pkt_len;
751 stats->tx_packets++;
752 } else {
753 stats->tx_errors++;
754 stats->tx_aborted_errors++;
755 }
756 ip6_tnl_dst_store(t, dst);
757
758 if (opt)
759 kfree(opt);
760
761 t->recursion--;
762 return 0;
763tx_err_link_failure:
764 stats->tx_carrier_errors++;
765 dst_link_failure(skb);
766tx_err_dst_release:
767 dst_release(dst);
768 if (opt)
769 kfree(opt);
770tx_err:
771 stats->tx_errors++;
772 stats->tx_dropped++;
773 kfree_skb(skb);
774 t->recursion--;
775 return 0;
776}
777
778static void ip6_tnl_set_cap(struct ip6_tnl *t)
779{
780 struct ip6_tnl_parm *p = &t->parms;
781 struct in6_addr *laddr = &p->laddr;
782 struct in6_addr *raddr = &p->raddr;
783 int ltype = ipv6_addr_type(laddr);
784 int rtype = ipv6_addr_type(raddr);
785
786 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
787
788 if (ltype != IPV6_ADDR_ANY && rtype != IPV6_ADDR_ANY &&
789 ((ltype|rtype) &
790 (IPV6_ADDR_UNICAST|
791 IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL|
792 IPV6_ADDR_MAPPED|IPV6_ADDR_RESERVED)) == IPV6_ADDR_UNICAST) {
793 struct net_device *ldev = NULL;
794 int l_ok = 1;
795 int r_ok = 1;
796
797 if (p->link)
798 ldev = dev_get_by_index(p->link);
799
800 if (ltype&IPV6_ADDR_UNICAST && !ipv6_chk_addr(laddr, ldev, 0))
801 l_ok = 0;
802
803 if (rtype&IPV6_ADDR_UNICAST && ipv6_chk_addr(raddr, NULL, 0))
804 r_ok = 0;
805
806 if (l_ok && r_ok) {
807 if (ltype&IPV6_ADDR_UNICAST)
808 p->flags |= IP6_TNL_F_CAP_XMIT;
809 if (rtype&IPV6_ADDR_UNICAST)
810 p->flags |= IP6_TNL_F_CAP_RCV;
811 }
812 if (ldev)
813 dev_put(ldev);
814 }
815}
816
817static void ip6ip6_tnl_link_config(struct ip6_tnl *t)
818{
819 struct net_device *dev = t->dev;
820 struct ip6_tnl_parm *p = &t->parms;
821 struct flowi *fl = &t->fl;
822
823 memcpy(&dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
824 memcpy(&dev->broadcast, &p->raddr, sizeof(struct in6_addr));
825
826 /* Set up flowi template */
827 ipv6_addr_copy(&fl->fl6_src, &p->laddr);
828 ipv6_addr_copy(&fl->fl6_dst, &p->raddr);
829 fl->oif = p->link;
830 fl->fl6_flowlabel = 0;
831
832 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
833 fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
834 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
835 fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
836
837 ip6_tnl_set_cap(t);
838
839 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
840 dev->flags |= IFF_POINTOPOINT;
841 else
842 dev->flags &= ~IFF_POINTOPOINT;
843
844 dev->iflink = p->link;
845
846 if (p->flags & IP6_TNL_F_CAP_XMIT) {
847 struct rt6_info *rt = rt6_lookup(&p->raddr, &p->laddr,
848 p->link, 0);
849
850 if (rt == NULL)
851 return;
852
853 if (rt->rt6i_dev) {
854 dev->hard_header_len = rt->rt6i_dev->hard_header_len +
855 sizeof (struct ipv6hdr);
856
857 dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
858
859 if (dev->mtu < IPV6_MIN_MTU)
860 dev->mtu = IPV6_MIN_MTU;
861 }
862 dst_release(&rt->u.dst);
863 }
864}
865
866/**
867 * ip6ip6_tnl_change - update the tunnel parameters
868 * @t: tunnel to be changed
869 * @p: tunnel configuration parameters
870 * @active: != 0 if tunnel is ready for use
871 *
872 * Description:
873 * ip6ip6_tnl_change() updates the tunnel parameters
874 **/
875
876static int
877ip6ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
878{
879 ipv6_addr_copy(&t->parms.laddr, &p->laddr);
880 ipv6_addr_copy(&t->parms.raddr, &p->raddr);
881 t->parms.flags = p->flags;
882 t->parms.hop_limit = p->hop_limit;
883 t->parms.encap_limit = p->encap_limit;
884 t->parms.flowinfo = p->flowinfo;
Gabor Fekete8181b8c2005-06-08 14:54:38 -0700885 t->parms.link = p->link;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 ip6ip6_tnl_link_config(t);
887 return 0;
888}
889
890/**
891 * ip6ip6_tnl_ioctl - configure ipv6 tunnels from userspace
892 * @dev: virtual device associated with tunnel
893 * @ifr: parameters passed from userspace
894 * @cmd: command to be performed
895 *
896 * Description:
897 * ip6ip6_tnl_ioctl() is used for managing IPv6 tunnels
898 * from userspace.
899 *
900 * The possible commands are the following:
901 * %SIOCGETTUNNEL: get tunnel parameters for device
902 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
903 * %SIOCCHGTUNNEL: change tunnel parameters to those given
904 * %SIOCDELTUNNEL: delete tunnel
905 *
906 * The fallback device "ip6tnl0", created during module
907 * initialization, can be used for creating other tunnel devices.
908 *
909 * Return:
910 * 0 on success,
911 * %-EFAULT if unable to copy data to or from userspace,
912 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
913 * %-EINVAL if passed tunnel parameters are invalid,
914 * %-EEXIST if changing a tunnel's parameters would cause a conflict
915 * %-ENODEV if attempting to change or delete a nonexisting device
916 **/
917
918static int
919ip6ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
920{
921 int err = 0;
922 int create;
923 struct ip6_tnl_parm p;
924 struct ip6_tnl *t = NULL;
925
926 switch (cmd) {
927 case SIOCGETTUNNEL:
928 if (dev == ip6ip6_fb_tnl_dev) {
929 if (copy_from_user(&p,
930 ifr->ifr_ifru.ifru_data,
931 sizeof (p))) {
932 err = -EFAULT;
933 break;
934 }
935 if ((err = ip6ip6_tnl_locate(&p, &t, 0)) == -ENODEV)
936 t = (struct ip6_tnl *) dev->priv;
937 else if (err)
938 break;
939 } else
940 t = (struct ip6_tnl *) dev->priv;
941
942 memcpy(&p, &t->parms, sizeof (p));
943 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
944 err = -EFAULT;
945 }
946 break;
947 case SIOCADDTUNNEL:
948 case SIOCCHGTUNNEL:
949 err = -EPERM;
950 create = (cmd == SIOCADDTUNNEL);
951 if (!capable(CAP_NET_ADMIN))
952 break;
953 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
954 err = -EFAULT;
955 break;
956 }
957 if (!create && dev != ip6ip6_fb_tnl_dev) {
958 t = (struct ip6_tnl *) dev->priv;
959 }
960 if (!t && (err = ip6ip6_tnl_locate(&p, &t, create))) {
961 break;
962 }
963 if (cmd == SIOCCHGTUNNEL) {
964 if (t->dev != dev) {
965 err = -EEXIST;
966 break;
967 }
968 ip6ip6_tnl_unlink(t);
969 err = ip6ip6_tnl_change(t, &p);
970 ip6ip6_tnl_link(t);
971 netdev_state_change(dev);
972 }
973 if (copy_to_user(ifr->ifr_ifru.ifru_data,
974 &t->parms, sizeof (p))) {
975 err = -EFAULT;
976 } else {
977 err = 0;
978 }
979 break;
980 case SIOCDELTUNNEL:
981 err = -EPERM;
982 if (!capable(CAP_NET_ADMIN))
983 break;
984
985 if (dev == ip6ip6_fb_tnl_dev) {
986 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data,
987 sizeof (p))) {
988 err = -EFAULT;
989 break;
990 }
991 err = ip6ip6_tnl_locate(&p, &t, 0);
992 if (err)
993 break;
994 if (t == ip6ip6_fb_tnl_dev->priv) {
995 err = -EPERM;
996 break;
997 }
998 } else {
999 t = (struct ip6_tnl *) dev->priv;
1000 }
1001 err = unregister_netdevice(t->dev);
1002 break;
1003 default:
1004 err = -EINVAL;
1005 }
1006 return err;
1007}
1008
1009/**
1010 * ip6ip6_tnl_get_stats - return the stats for tunnel device
1011 * @dev: virtual device associated with tunnel
1012 *
1013 * Return: stats for device
1014 **/
1015
1016static struct net_device_stats *
1017ip6ip6_tnl_get_stats(struct net_device *dev)
1018{
1019 return &(((struct ip6_tnl *) dev->priv)->stat);
1020}
1021
1022/**
1023 * ip6ip6_tnl_change_mtu - change mtu manually for tunnel device
1024 * @dev: virtual device associated with tunnel
1025 * @new_mtu: the new mtu
1026 *
1027 * Return:
1028 * 0 on success,
1029 * %-EINVAL if mtu too small
1030 **/
1031
1032static int
1033ip6ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1034{
1035 if (new_mtu < IPV6_MIN_MTU) {
1036 return -EINVAL;
1037 }
1038 dev->mtu = new_mtu;
1039 return 0;
1040}
1041
1042/**
1043 * ip6ip6_tnl_dev_setup - setup virtual tunnel device
1044 * @dev: virtual device associated with tunnel
1045 *
1046 * Description:
1047 * Initialize function pointers and device parameters
1048 **/
1049
1050static void ip6ip6_tnl_dev_setup(struct net_device *dev)
1051{
1052 SET_MODULE_OWNER(dev);
1053 dev->uninit = ip6ip6_tnl_dev_uninit;
1054 dev->destructor = free_netdev;
1055 dev->hard_start_xmit = ip6ip6_tnl_xmit;
1056 dev->get_stats = ip6ip6_tnl_get_stats;
1057 dev->do_ioctl = ip6ip6_tnl_ioctl;
1058 dev->change_mtu = ip6ip6_tnl_change_mtu;
1059
1060 dev->type = ARPHRD_TUNNEL6;
1061 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1062 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1063 dev->flags |= IFF_NOARP;
1064 dev->addr_len = sizeof(struct in6_addr);
1065}
1066
1067
1068/**
1069 * ip6ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1070 * @dev: virtual device associated with tunnel
1071 **/
1072
1073static inline void
1074ip6ip6_tnl_dev_init_gen(struct net_device *dev)
1075{
1076 struct ip6_tnl *t = (struct ip6_tnl *) dev->priv;
1077 t->fl.proto = IPPROTO_IPV6;
1078 t->dev = dev;
1079 strcpy(t->parms.name, dev->name);
1080}
1081
1082/**
1083 * ip6ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1084 * @dev: virtual device associated with tunnel
1085 **/
1086
1087static int
1088ip6ip6_tnl_dev_init(struct net_device *dev)
1089{
1090 struct ip6_tnl *t = (struct ip6_tnl *) dev->priv;
1091 ip6ip6_tnl_dev_init_gen(dev);
1092 ip6ip6_tnl_link_config(t);
1093 return 0;
1094}
1095
1096/**
1097 * ip6ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1098 * @dev: fallback device
1099 *
1100 * Return: 0
1101 **/
1102
1103static int
1104ip6ip6_fb_tnl_dev_init(struct net_device *dev)
1105{
1106 struct ip6_tnl *t = dev->priv;
1107 ip6ip6_tnl_dev_init_gen(dev);
1108 dev_hold(dev);
1109 tnls_wc[0] = t;
1110 return 0;
1111}
1112
1113static struct xfrm6_tunnel ip6ip6_handler = {
1114 .handler = ip6ip6_rcv,
1115 .err_handler = ip6ip6_err,
1116};
1117
1118/**
1119 * ip6_tunnel_init - register protocol and reserve needed resources
1120 *
1121 * Return: 0 on success
1122 **/
1123
1124static int __init ip6_tunnel_init(void)
1125{
1126 int err;
1127
1128 if (xfrm6_tunnel_register(&ip6ip6_handler) < 0) {
1129 printk(KERN_ERR "ip6ip6 init: can't register tunnel\n");
1130 return -EAGAIN;
1131 }
1132 ip6ip6_fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1133 ip6ip6_tnl_dev_setup);
1134
1135 if (!ip6ip6_fb_tnl_dev) {
1136 err = -ENOMEM;
1137 goto fail;
1138 }
1139 ip6ip6_fb_tnl_dev->init = ip6ip6_fb_tnl_dev_init;
1140
1141 if ((err = register_netdev(ip6ip6_fb_tnl_dev))) {
1142 free_netdev(ip6ip6_fb_tnl_dev);
1143 goto fail;
1144 }
1145 return 0;
1146fail:
1147 xfrm6_tunnel_deregister(&ip6ip6_handler);
1148 return err;
1149}
1150
1151/**
1152 * ip6_tunnel_cleanup - free resources and unregister protocol
1153 **/
1154
1155static void __exit ip6_tunnel_cleanup(void)
1156{
1157 if (xfrm6_tunnel_deregister(&ip6ip6_handler) < 0)
1158 printk(KERN_INFO "ip6ip6 close: can't deregister tunnel\n");
1159
1160 unregister_netdev(ip6ip6_fb_tnl_dev);
1161}
1162
1163module_init(ip6_tunnel_init);
1164module_exit(ip6_tunnel_cleanup);