ia64/linux-2.6.18-xen.hg

view net/ipv6/ip6_tunnel.c @ 673:3161879fdf22

[IA64] xencomm: support XENMEM_add_to_physmap and XENMEM_remove_from_phsymap

support XENMEM_add_to_physmap and XENMEM_remove_from_phsymap.

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
author Isaku Yamahata <yamahata@valinux.co.jp>
date Tue Sep 16 21:26:15 2008 +0900 (2008-09-16)
parents 831230e53067
children
line source
1 /*
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 */
22 #include <linux/module.h>
23 #include <linux/capability.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>
41 #include <asm/uaccess.h>
42 #include <asm/atomic.h>
44 #include <net/ip.h>
45 #include <net/ipv6.h>
46 #include <net/ip6_route.h>
47 #include <net/addrconf.h>
48 #include <net/ip6_tunnel.h>
49 #include <net/xfrm.h>
50 #include <net/dsfield.h>
51 #include <net/inet_ecn.h>
53 MODULE_AUTHOR("Ville Nuorvala");
54 MODULE_DESCRIPTION("IPv6-in-IPv6 tunnel");
55 MODULE_LICENSE("GPL");
57 #define IPV6_TLV_TEL_DST_SIZE 8
59 #ifdef IP6_TNL_DEBUG
60 #define IP6_TNL_TRACE(x...) printk(KERN_DEBUG "%s:" x "\n", __FUNCTION__)
61 #else
62 #define IP6_TNL_TRACE(x...) do {;} while(0)
63 #endif
65 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
67 #define HASH_SIZE 32
69 #define HASH(addr) (((addr)->s6_addr32[0] ^ (addr)->s6_addr32[1] ^ \
70 (addr)->s6_addr32[2] ^ (addr)->s6_addr32[3]) & \
71 (HASH_SIZE - 1))
73 static int ip6ip6_fb_tnl_dev_init(struct net_device *dev);
74 static int ip6ip6_tnl_dev_init(struct net_device *dev);
75 static void ip6ip6_tnl_dev_setup(struct net_device *dev);
77 /* the IPv6 tunnel fallback device */
78 static struct net_device *ip6ip6_fb_tnl_dev;
81 /* lists for storing tunnels in use */
82 static struct ip6_tnl *tnls_r_l[HASH_SIZE];
83 static struct ip6_tnl *tnls_wc[1];
84 static struct ip6_tnl **tnls[2] = { tnls_wc, tnls_r_l };
86 /* lock for the tunnel lists */
87 static DEFINE_RWLOCK(ip6ip6_lock);
89 static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
90 {
91 struct dst_entry *dst = t->dst_cache;
93 if (dst && dst->obsolete &&
94 dst->ops->check(dst, t->dst_cookie) == NULL) {
95 t->dst_cache = NULL;
96 dst_release(dst);
97 return NULL;
98 }
100 return dst;
101 }
103 static inline void ip6_tnl_dst_reset(struct ip6_tnl *t)
104 {
105 dst_release(t->dst_cache);
106 t->dst_cache = NULL;
107 }
109 static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
110 {
111 struct rt6_info *rt = (struct rt6_info *) dst;
112 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
113 dst_release(t->dst_cache);
114 t->dst_cache = dst;
115 }
117 /**
118 * ip6ip6_tnl_lookup - fetch tunnel matching the end-point addresses
119 * @remote: the address of the tunnel exit-point
120 * @local: the address of the tunnel entry-point
121 *
122 * Return:
123 * tunnel matching given end-points if found,
124 * else fallback tunnel if its device is up,
125 * else %NULL
126 **/
128 static struct ip6_tnl *
129 ip6ip6_tnl_lookup(struct in6_addr *remote, struct in6_addr *local)
130 {
131 unsigned h0 = HASH(remote);
132 unsigned h1 = HASH(local);
133 struct ip6_tnl *t;
135 for (t = tnls_r_l[h0 ^ h1]; t; t = t->next) {
136 if (ipv6_addr_equal(local, &t->parms.laddr) &&
137 ipv6_addr_equal(remote, &t->parms.raddr) &&
138 (t->dev->flags & IFF_UP))
139 return t;
140 }
141 if ((t = tnls_wc[0]) != NULL && (t->dev->flags & IFF_UP))
142 return t;
144 return NULL;
145 }
147 /**
148 * ip6ip6_bucket - get head of list matching given tunnel parameters
149 * @p: parameters containing tunnel end-points
150 *
151 * Description:
152 * ip6ip6_bucket() returns the head of the list matching the
153 * &struct in6_addr entries laddr and raddr in @p.
154 *
155 * Return: head of IPv6 tunnel list
156 **/
158 static struct ip6_tnl **
159 ip6ip6_bucket(struct ip6_tnl_parm *p)
160 {
161 struct in6_addr *remote = &p->raddr;
162 struct in6_addr *local = &p->laddr;
163 unsigned h = 0;
164 int prio = 0;
166 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
167 prio = 1;
168 h = HASH(remote) ^ HASH(local);
169 }
170 return &tnls[prio][h];
171 }
173 /**
174 * ip6ip6_tnl_link - add tunnel to hash table
175 * @t: tunnel to be added
176 **/
178 static void
179 ip6ip6_tnl_link(struct ip6_tnl *t)
180 {
181 struct ip6_tnl **tp = ip6ip6_bucket(&t->parms);
183 t->next = *tp;
184 write_lock_bh(&ip6ip6_lock);
185 *tp = t;
186 write_unlock_bh(&ip6ip6_lock);
187 }
189 /**
190 * ip6ip6_tnl_unlink - remove tunnel from hash table
191 * @t: tunnel to be removed
192 **/
194 static void
195 ip6ip6_tnl_unlink(struct ip6_tnl *t)
196 {
197 struct ip6_tnl **tp;
199 for (tp = ip6ip6_bucket(&t->parms); *tp; tp = &(*tp)->next) {
200 if (t == *tp) {
201 write_lock_bh(&ip6ip6_lock);
202 *tp = t->next;
203 write_unlock_bh(&ip6ip6_lock);
204 break;
205 }
206 }
207 }
209 /**
210 * ip6_tnl_create() - create a new tunnel
211 * @p: tunnel parameters
212 * @pt: pointer to new tunnel
213 *
214 * Description:
215 * Create tunnel matching given parameters.
216 *
217 * Return:
218 * 0 on success
219 **/
221 static int
222 ip6_tnl_create(struct ip6_tnl_parm *p, struct ip6_tnl **pt)
223 {
224 struct net_device *dev;
225 struct ip6_tnl *t;
226 char name[IFNAMSIZ];
227 int err;
229 if (p->name[0]) {
230 strlcpy(name, p->name, IFNAMSIZ);
231 } else {
232 int i;
233 for (i = 1; i < IP6_TNL_MAX; i++) {
234 sprintf(name, "ip6tnl%d", i);
235 if (__dev_get_by_name(name) == NULL)
236 break;
237 }
238 if (i == IP6_TNL_MAX)
239 return -ENOBUFS;
240 }
241 dev = alloc_netdev(sizeof (*t), name, ip6ip6_tnl_dev_setup);
242 if (dev == NULL)
243 return -ENOMEM;
245 t = netdev_priv(dev);
246 dev->init = ip6ip6_tnl_dev_init;
247 t->parms = *p;
249 if ((err = register_netdevice(dev)) < 0) {
250 free_netdev(dev);
251 return err;
252 }
253 dev_hold(dev);
255 ip6ip6_tnl_link(t);
256 *pt = t;
257 return 0;
258 }
260 /**
261 * ip6ip6_tnl_locate - find or create tunnel matching given parameters
262 * @p: tunnel parameters
263 * @create: != 0 if allowed to create new tunnel if no match found
264 *
265 * Description:
266 * ip6ip6_tnl_locate() first tries to locate an existing tunnel
267 * based on @parms. If this is unsuccessful, but @create is set a new
268 * tunnel device is created and registered for use.
269 *
270 * Return:
271 * 0 if tunnel located or created,
272 * -EINVAL if parameters incorrect,
273 * -ENODEV if no matching tunnel available
274 **/
276 static int
277 ip6ip6_tnl_locate(struct ip6_tnl_parm *p, struct ip6_tnl **pt, int create)
278 {
279 struct in6_addr *remote = &p->raddr;
280 struct in6_addr *local = &p->laddr;
281 struct ip6_tnl *t;
283 if (p->proto != IPPROTO_IPV6)
284 return -EINVAL;
286 for (t = *ip6ip6_bucket(p); t; t = t->next) {
287 if (ipv6_addr_equal(local, &t->parms.laddr) &&
288 ipv6_addr_equal(remote, &t->parms.raddr)) {
289 *pt = t;
290 return (create ? -EEXIST : 0);
291 }
292 }
293 if (!create)
294 return -ENODEV;
296 return ip6_tnl_create(p, pt);
297 }
299 /**
300 * ip6ip6_tnl_dev_uninit - tunnel device uninitializer
301 * @dev: the device to be destroyed
302 *
303 * Description:
304 * ip6ip6_tnl_dev_uninit() removes tunnel from its list
305 **/
307 static void
308 ip6ip6_tnl_dev_uninit(struct net_device *dev)
309 {
310 struct ip6_tnl *t = netdev_priv(dev);
312 if (dev == ip6ip6_fb_tnl_dev) {
313 write_lock_bh(&ip6ip6_lock);
314 tnls_wc[0] = NULL;
315 write_unlock_bh(&ip6ip6_lock);
316 } else {
317 ip6ip6_tnl_unlink(t);
318 }
319 ip6_tnl_dst_reset(t);
320 dev_put(dev);
321 }
323 /**
324 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
325 * @skb: received socket buffer
326 *
327 * Return:
328 * 0 if none was found,
329 * else index to encapsulation limit
330 **/
332 static __u16
333 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
334 {
335 struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw;
336 __u8 nexthdr = ipv6h->nexthdr;
337 __u16 off = sizeof (*ipv6h);
339 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
340 __u16 optlen = 0;
341 struct ipv6_opt_hdr *hdr;
342 if (raw + off + sizeof (*hdr) > skb->data &&
343 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
344 break;
346 hdr = (struct ipv6_opt_hdr *) (raw + off);
347 if (nexthdr == NEXTHDR_FRAGMENT) {
348 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
349 if (frag_hdr->frag_off)
350 break;
351 optlen = 8;
352 } else if (nexthdr == NEXTHDR_AUTH) {
353 optlen = (hdr->hdrlen + 2) << 2;
354 } else {
355 optlen = ipv6_optlen(hdr);
356 }
357 if (nexthdr == NEXTHDR_DEST) {
358 __u16 i = off + 2;
359 while (1) {
360 struct ipv6_tlv_tnl_enc_lim *tel;
362 /* No more room for encapsulation limit */
363 if (i + sizeof (*tel) > off + optlen)
364 break;
366 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
367 /* return index of option if found and valid */
368 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
369 tel->length == 1)
370 return i;
371 /* else jump to next option */
372 if (tel->type)
373 i += tel->length + 2;
374 else
375 i++;
376 }
377 }
378 nexthdr = hdr->nexthdr;
379 off += optlen;
380 }
381 return 0;
382 }
384 /**
385 * ip6ip6_err - tunnel error handler
386 *
387 * Description:
388 * ip6ip6_err() should handle errors in the tunnel according
389 * to the specifications in RFC 2473.
390 **/
392 static int
393 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
394 int type, int code, int offset, __u32 info)
395 {
396 struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
397 struct ip6_tnl *t;
398 int rel_msg = 0;
399 int rel_type = ICMPV6_DEST_UNREACH;
400 int rel_code = ICMPV6_ADDR_UNREACH;
401 __u32 rel_info = 0;
402 __u16 len;
403 int err = -ENOENT;
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. */
409 read_lock(&ip6ip6_lock);
410 if ((t = ip6ip6_tnl_lookup(&ipv6h->daddr, &ipv6h->saddr)) == NULL)
411 goto out;
413 err = 0;
415 switch (type) {
416 __u32 teli;
417 struct ipv6_tlv_tnl_enc_lim *tel;
418 __u32 mtu;
419 case ICMPV6_DEST_UNREACH:
420 if (net_ratelimit())
421 printk(KERN_WARNING
422 "%s: Path to destination invalid "
423 "or inactive!\n", t->parms.name);
424 rel_msg = 1;
425 break;
426 case ICMPV6_TIME_EXCEED:
427 if (code == ICMPV6_EXC_HOPLIMIT) {
428 if (net_ratelimit())
429 printk(KERN_WARNING
430 "%s: Too small hop limit or "
431 "routing loop in tunnel!\n",
432 t->parms.name);
433 rel_msg = 1;
434 }
435 break;
436 case ICMPV6_PARAMPROB:
437 /* ignore if parameter problem not caused by a tunnel
438 encapsulation limit sub-option */
439 if (code != ICMPV6_HDR_FIELD) {
440 break;
441 }
442 teli = parse_tlv_tnl_enc_lim(skb, skb->data);
444 if (teli && teli == ntohl(info) - 2) {
445 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
446 if (tel->encap_limit == 0) {
447 if (net_ratelimit())
448 printk(KERN_WARNING
449 "%s: Too small encapsulation "
450 "limit or routing loop in "
451 "tunnel!\n", t->parms.name);
452 rel_msg = 1;
453 }
454 }
455 break;
456 case ICMPV6_PKT_TOOBIG:
457 mtu = ntohl(info) - offset;
458 if (mtu < IPV6_MIN_MTU)
459 mtu = IPV6_MIN_MTU;
460 t->dev->mtu = mtu;
462 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
463 rel_type = ICMPV6_PKT_TOOBIG;
464 rel_code = 0;
465 rel_info = mtu;
466 rel_msg = 1;
467 }
468 break;
469 }
470 if (rel_msg && pskb_may_pull(skb, offset + sizeof (*ipv6h))) {
471 struct rt6_info *rt;
472 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
473 if (!skb2)
474 goto out;
476 dst_release(skb2->dst);
477 skb2->dst = NULL;
478 skb_pull(skb2, offset);
479 skb2->nh.raw = skb2->data;
481 /* Try to guess incoming interface */
482 rt = rt6_lookup(&skb2->nh.ipv6h->saddr, NULL, 0, 0);
484 if (rt && rt->rt6i_dev)
485 skb2->dev = rt->rt6i_dev;
487 icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
489 if (rt)
490 dst_release(&rt->u.dst);
492 kfree_skb(skb2);
493 }
494 out:
495 read_unlock(&ip6ip6_lock);
496 return err;
497 }
499 static inline void ip6ip6_ecn_decapsulate(struct ipv6hdr *outer_iph,
500 struct sk_buff *skb)
501 {
502 struct ipv6hdr *inner_iph = skb->nh.ipv6h;
504 if (INET_ECN_is_ce(ipv6_get_dsfield(outer_iph)))
505 IP6_ECN_set_ce(inner_iph);
506 }
508 /**
509 * ip6ip6_rcv - decapsulate IPv6 packet and retransmit it locally
510 * @skb: received socket buffer
511 *
512 * Return: 0
513 **/
515 static int
516 ip6ip6_rcv(struct sk_buff *skb)
517 {
518 struct ipv6hdr *ipv6h;
519 struct ip6_tnl *t;
521 ipv6h = skb->nh.ipv6h;
523 read_lock(&ip6ip6_lock);
525 if ((t = ip6ip6_tnl_lookup(&ipv6h->saddr, &ipv6h->daddr)) != NULL) {
526 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
527 read_unlock(&ip6ip6_lock);
528 goto discard;
529 }
531 if (!(t->parms.flags & IP6_TNL_F_CAP_RCV)) {
532 t->stat.rx_dropped++;
533 read_unlock(&ip6ip6_lock);
534 goto discard;
535 }
536 secpath_reset(skb);
537 skb->mac.raw = skb->nh.raw;
538 skb->nh.raw = skb->data;
539 skb->protocol = htons(ETH_P_IPV6);
540 skb->pkt_type = PACKET_HOST;
541 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
542 skb->dev = t->dev;
543 dst_release(skb->dst);
544 skb->dst = NULL;
545 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
546 ipv6_copy_dscp(ipv6h, skb->nh.ipv6h);
547 ip6ip6_ecn_decapsulate(ipv6h, skb);
548 t->stat.rx_packets++;
549 t->stat.rx_bytes += skb->len;
550 netif_rx(skb);
551 read_unlock(&ip6ip6_lock);
552 return 0;
553 }
554 read_unlock(&ip6ip6_lock);
555 return 1;
557 discard:
558 kfree_skb(skb);
559 return 0;
560 }
562 static inline struct ipv6_txoptions *create_tel(__u8 encap_limit)
563 {
564 struct ipv6_tlv_tnl_enc_lim *tel;
565 struct ipv6_txoptions *opt;
566 __u8 *raw;
568 int opt_len = sizeof(*opt) + 8;
570 if (!(opt = kzalloc(opt_len, GFP_ATOMIC))) {
571 return NULL;
572 }
573 opt->tot_len = opt_len;
574 opt->dst0opt = (struct ipv6_opt_hdr *) (opt + 1);
575 opt->opt_nflen = 8;
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;
582 raw = (__u8 *) opt->dst0opt;
583 raw[5] = IPV6_TLV_PADN;
584 raw[6] = 1;
586 return opt;
587 }
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 **/
603 static inline int
604 ip6ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
605 {
606 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
607 }
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 **/
622 static int
623 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
624 {
625 struct ip6_tnl *t = netdev_priv(dev);
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;
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;
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);
671 if (encap_limit >= 0 && (opt = create_tel(encap_limit)) == NULL)
672 goto tx_err;
674 if ((dst = ip6_tnl_dst_check(t)) != NULL)
675 dst_hold(dst);
676 else {
677 dst = ip6_route_output(NULL, &fl);
679 if (dst->error || xfrm_lookup(&dst, &fl, NULL, 0) < 0)
680 goto tx_err_link_failure;
681 }
683 tdev = dst->dev;
685 if (tdev == dev) {
686 stats->collisions++;
687 if (net_ratelimit())
688 printk(KERN_WARNING
689 "%s: Local routing loop detected!\n",
690 t->parms.name);
691 goto tx_err_dst_release;
692 }
693 mtu = dst_mtu(dst) - sizeof (*ipv6h);
694 if (opt) {
695 max_headroom += 8;
696 mtu -= 8;
697 }
698 if (mtu < IPV6_MIN_MTU)
699 mtu = IPV6_MIN_MTU;
700 if (skb->dst && mtu < dst_mtu(skb->dst)) {
701 struct rt6_info *rt = (struct rt6_info *) skb->dst;
702 rt->rt6i_flags |= RTF_MODIFIED;
703 rt->u.dst.metrics[RTAX_MTU-1] = mtu;
704 }
705 if (skb->len > mtu) {
706 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
707 goto tx_err_dst_release;
708 }
710 /*
711 * Okay, now see if we can stuff it in the buffer as-is.
712 */
713 max_headroom += LL_RESERVED_SPACE(tdev);
715 if (skb_headroom(skb) < max_headroom ||
716 skb_cloned(skb) || skb_shared(skb)) {
717 struct sk_buff *new_skb;
719 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
720 goto tx_err_dst_release;
722 if (skb->sk)
723 skb_set_owner_w(new_skb, skb->sk);
724 kfree_skb(skb);
725 skb = new_skb;
726 }
727 dst_release(skb->dst);
728 skb->dst = dst_clone(dst);
730 skb->h.raw = skb->nh.raw;
732 if (opt)
733 ipv6_push_nfrag_opts(skb, opt, &proto, NULL);
735 skb->nh.raw = skb_push(skb, sizeof(struct ipv6hdr));
736 ipv6h = skb->nh.ipv6h;
737 *(u32*)ipv6h = fl.fl6_flowlabel | htonl(0x60000000);
738 dsfield = INET_ECN_encapsulate(0, dsfield);
739 ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
740 ipv6h->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
741 ipv6h->hop_limit = t->parms.hop_limit;
742 ipv6h->nexthdr = proto;
743 ipv6_addr_copy(&ipv6h->saddr, &fl.fl6_src);
744 ipv6_addr_copy(&ipv6h->daddr, &fl.fl6_dst);
745 nf_reset(skb);
746 pkt_len = skb->len;
747 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL,
748 skb->dst->dev, dst_output);
750 if (err == NET_XMIT_SUCCESS || err == NET_XMIT_CN) {
751 stats->tx_bytes += pkt_len;
752 stats->tx_packets++;
753 } else {
754 stats->tx_errors++;
755 stats->tx_aborted_errors++;
756 }
757 ip6_tnl_dst_store(t, dst);
759 kfree(opt);
761 t->recursion--;
762 return 0;
763 tx_err_link_failure:
764 stats->tx_carrier_errors++;
765 dst_link_failure(skb);
766 tx_err_dst_release:
767 dst_release(dst);
768 kfree(opt);
769 tx_err:
770 stats->tx_errors++;
771 stats->tx_dropped++;
772 kfree_skb(skb);
773 t->recursion--;
774 return 0;
775 }
777 static void ip6_tnl_set_cap(struct ip6_tnl *t)
778 {
779 struct ip6_tnl_parm *p = &t->parms;
780 struct in6_addr *laddr = &p->laddr;
781 struct in6_addr *raddr = &p->raddr;
782 int ltype = ipv6_addr_type(laddr);
783 int rtype = ipv6_addr_type(raddr);
785 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
787 if (ltype != IPV6_ADDR_ANY && rtype != IPV6_ADDR_ANY &&
788 ((ltype|rtype) &
789 (IPV6_ADDR_UNICAST|
790 IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL|
791 IPV6_ADDR_MAPPED|IPV6_ADDR_RESERVED)) == IPV6_ADDR_UNICAST) {
792 struct net_device *ldev = NULL;
793 int l_ok = 1;
794 int r_ok = 1;
796 if (p->link)
797 ldev = dev_get_by_index(p->link);
799 if (ltype&IPV6_ADDR_UNICAST && !ipv6_chk_addr(laddr, ldev, 0))
800 l_ok = 0;
802 if (rtype&IPV6_ADDR_UNICAST && ipv6_chk_addr(raddr, NULL, 0))
803 r_ok = 0;
805 if (l_ok && r_ok) {
806 if (ltype&IPV6_ADDR_UNICAST)
807 p->flags |= IP6_TNL_F_CAP_XMIT;
808 if (rtype&IPV6_ADDR_UNICAST)
809 p->flags |= IP6_TNL_F_CAP_RCV;
810 }
811 if (ldev)
812 dev_put(ldev);
813 }
814 }
816 static void ip6ip6_tnl_link_config(struct ip6_tnl *t)
817 {
818 struct net_device *dev = t->dev;
819 struct ip6_tnl_parm *p = &t->parms;
820 struct flowi *fl = &t->fl;
822 memcpy(&dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
823 memcpy(&dev->broadcast, &p->raddr, sizeof(struct in6_addr));
825 /* Set up flowi template */
826 ipv6_addr_copy(&fl->fl6_src, &p->laddr);
827 ipv6_addr_copy(&fl->fl6_dst, &p->raddr);
828 fl->oif = p->link;
829 fl->fl6_flowlabel = 0;
831 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
832 fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
833 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
834 fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
836 ip6_tnl_set_cap(t);
838 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
839 dev->flags |= IFF_POINTOPOINT;
840 else
841 dev->flags &= ~IFF_POINTOPOINT;
843 dev->iflink = p->link;
845 if (p->flags & IP6_TNL_F_CAP_XMIT) {
846 struct rt6_info *rt = rt6_lookup(&p->raddr, &p->laddr,
847 p->link, 0);
849 if (rt == NULL)
850 return;
852 if (rt->rt6i_dev) {
853 dev->hard_header_len = rt->rt6i_dev->hard_header_len +
854 sizeof (struct ipv6hdr);
856 dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
858 if (dev->mtu < IPV6_MIN_MTU)
859 dev->mtu = IPV6_MIN_MTU;
860 }
861 dst_release(&rt->u.dst);
862 }
863 }
865 /**
866 * ip6ip6_tnl_change - update the tunnel parameters
867 * @t: tunnel to be changed
868 * @p: tunnel configuration parameters
869 * @active: != 0 if tunnel is ready for use
870 *
871 * Description:
872 * ip6ip6_tnl_change() updates the tunnel parameters
873 **/
875 static int
876 ip6ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
877 {
878 ipv6_addr_copy(&t->parms.laddr, &p->laddr);
879 ipv6_addr_copy(&t->parms.raddr, &p->raddr);
880 t->parms.flags = p->flags;
881 t->parms.hop_limit = p->hop_limit;
882 t->parms.encap_limit = p->encap_limit;
883 t->parms.flowinfo = p->flowinfo;
884 t->parms.link = p->link;
885 ip6_tnl_dst_reset(t);
886 ip6ip6_tnl_link_config(t);
887 return 0;
888 }
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 **/
918 static int
919 ip6ip6_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;
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 = netdev_priv(dev);
937 else if (err)
938 break;
939 } else
940 t = netdev_priv(dev);
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 = netdev_priv(dev);
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;
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 == netdev_priv(ip6ip6_fb_tnl_dev)) {
995 err = -EPERM;
996 break;
997 }
998 } else {
999 t = netdev_priv(dev);
1001 err = unregister_netdevice(t->dev);
1002 break;
1003 default:
1004 err = -EINVAL;
1006 return err;
1009 /**
1010 * ip6ip6_tnl_get_stats - return the stats for tunnel device
1011 * @dev: virtual device associated with tunnel
1013 * Return: stats for device
1014 **/
1016 static struct net_device_stats *
1017 ip6ip6_tnl_get_stats(struct net_device *dev)
1019 return &(((struct ip6_tnl *)netdev_priv(dev))->stat);
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
1027 * Return:
1028 * 0 on success,
1029 * %-EINVAL if mtu too small
1030 **/
1032 static int
1033 ip6ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1035 if (new_mtu < IPV6_MIN_MTU) {
1036 return -EINVAL;
1038 dev->mtu = new_mtu;
1039 return 0;
1042 /**
1043 * ip6ip6_tnl_dev_setup - setup virtual tunnel device
1044 * @dev: virtual device associated with tunnel
1046 * Description:
1047 * Initialize function pointers and device parameters
1048 **/
1050 static void ip6ip6_tnl_dev_setup(struct net_device *dev)
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;
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);
1068 /**
1069 * ip6ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1070 * @dev: virtual device associated with tunnel
1071 **/
1073 static inline void
1074 ip6ip6_tnl_dev_init_gen(struct net_device *dev)
1076 struct ip6_tnl *t = netdev_priv(dev);
1077 t->fl.proto = IPPROTO_IPV6;
1078 t->dev = dev;
1079 strcpy(t->parms.name, dev->name);
1082 /**
1083 * ip6ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1084 * @dev: virtual device associated with tunnel
1085 **/
1087 static int
1088 ip6ip6_tnl_dev_init(struct net_device *dev)
1090 struct ip6_tnl *t = netdev_priv(dev);
1091 ip6ip6_tnl_dev_init_gen(dev);
1092 ip6ip6_tnl_link_config(t);
1093 return 0;
1096 /**
1097 * ip6ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1098 * @dev: fallback device
1100 * Return: 0
1101 **/
1103 static int
1104 ip6ip6_fb_tnl_dev_init(struct net_device *dev)
1106 struct ip6_tnl *t = netdev_priv(dev);
1107 ip6ip6_tnl_dev_init_gen(dev);
1108 dev_hold(dev);
1109 tnls_wc[0] = t;
1110 return 0;
1113 static struct xfrm6_tunnel ip6ip6_handler = {
1114 .handler = ip6ip6_rcv,
1115 .err_handler = ip6ip6_err,
1116 .priority = 1,
1117 };
1119 /**
1120 * ip6_tunnel_init - register protocol and reserve needed resources
1122 * Return: 0 on success
1123 **/
1125 static int __init ip6_tunnel_init(void)
1127 int err;
1129 if (xfrm6_tunnel_register(&ip6ip6_handler)) {
1130 printk(KERN_ERR "ip6ip6 init: can't register tunnel\n");
1131 return -EAGAIN;
1133 ip6ip6_fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1134 ip6ip6_tnl_dev_setup);
1136 if (!ip6ip6_fb_tnl_dev) {
1137 err = -ENOMEM;
1138 goto fail;
1140 ip6ip6_fb_tnl_dev->init = ip6ip6_fb_tnl_dev_init;
1142 if ((err = register_netdev(ip6ip6_fb_tnl_dev))) {
1143 free_netdev(ip6ip6_fb_tnl_dev);
1144 goto fail;
1146 return 0;
1147 fail:
1148 xfrm6_tunnel_deregister(&ip6ip6_handler);
1149 return err;
1152 /**
1153 * ip6_tunnel_cleanup - free resources and unregister protocol
1154 **/
1156 static void __exit ip6_tunnel_cleanup(void)
1158 if (xfrm6_tunnel_deregister(&ip6ip6_handler))
1159 printk(KERN_INFO "ip6ip6 close: can't deregister tunnel\n");
1161 unregister_netdev(ip6ip6_fb_tnl_dev);
1164 module_init(ip6_tunnel_init);
1165 module_exit(ip6_tunnel_cleanup);