pktgen: add needed include file
[linux/fpc-iii.git] / net / ipv6 / ip6_tunnel.c
blob1e55866cead7425eefdff36c1ddca1aab9504286
1 /*
2 * IPv6 tunneling device
3 * Linux INET6 implementation
5 * Authors:
6 * Ville Nuorvala <vnuorval@tcs.hut.fi>
7 * Yasuyuki Kozakai <kozakai@linux-ipv6.org>
9 * Based on:
10 * linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
12 * RFC 2473
14 * This program is free software; you can redistribute it and/or
15 * modify it under the terms of the GNU General Public License
16 * as published by the Free Software Foundation; either version
17 * 2 of the License, or (at your option) any later version.
21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
23 #include <linux/module.h>
24 #include <linux/capability.h>
25 #include <linux/errno.h>
26 #include <linux/types.h>
27 #include <linux/sockios.h>
28 #include <linux/icmp.h>
29 #include <linux/if.h>
30 #include <linux/in.h>
31 #include <linux/ip.h>
32 #include <linux/if_tunnel.h>
33 #include <linux/net.h>
34 #include <linux/in6.h>
35 #include <linux/netdevice.h>
36 #include <linux/if_arp.h>
37 #include <linux/icmpv6.h>
38 #include <linux/init.h>
39 #include <linux/route.h>
40 #include <linux/rtnetlink.h>
41 #include <linux/netfilter_ipv6.h>
42 #include <linux/slab.h>
43 #include <linux/hash.h>
45 #include <asm/uaccess.h>
46 #include <linux/atomic.h>
48 #include <net/icmp.h>
49 #include <net/ip.h>
50 #include <net/ip_tunnels.h>
51 #include <net/ipv6.h>
52 #include <net/ip6_route.h>
53 #include <net/addrconf.h>
54 #include <net/ip6_tunnel.h>
55 #include <net/xfrm.h>
56 #include <net/dsfield.h>
57 #include <net/inet_ecn.h>
58 #include <net/net_namespace.h>
59 #include <net/netns/generic.h>
61 MODULE_AUTHOR("Ville Nuorvala");
62 MODULE_DESCRIPTION("IPv6 tunneling device");
63 MODULE_LICENSE("GPL");
64 MODULE_ALIAS_NETDEV("ip6tnl0");
66 #ifdef IP6_TNL_DEBUG
67 #define IP6_TNL_TRACE(x...) pr_debug("%s:" x "\n", __func__)
68 #else
69 #define IP6_TNL_TRACE(x...) do {;} while(0)
70 #endif
72 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
73 #define IPV6_TCLASS_SHIFT 20
75 #define HASH_SIZE_SHIFT 5
76 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
78 static bool log_ecn_error = true;
79 module_param(log_ecn_error, bool, 0644);
80 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
82 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
84 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
86 return hash_32(hash, HASH_SIZE_SHIFT);
89 static int ip6_tnl_dev_init(struct net_device *dev);
90 static void ip6_tnl_dev_setup(struct net_device *dev);
91 static struct rtnl_link_ops ip6_link_ops __read_mostly;
93 static int ip6_tnl_net_id __read_mostly;
94 struct ip6_tnl_net {
95 /* the IPv6 tunnel fallback device */
96 struct net_device *fb_tnl_dev;
97 /* lists for storing tunnels in use */
98 struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
99 struct ip6_tnl __rcu *tnls_wc[1];
100 struct ip6_tnl __rcu **tnls[2];
103 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
105 struct pcpu_tstats sum = { 0 };
106 int i;
108 for_each_possible_cpu(i) {
109 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
111 sum.rx_packets += tstats->rx_packets;
112 sum.rx_bytes += tstats->rx_bytes;
113 sum.tx_packets += tstats->tx_packets;
114 sum.tx_bytes += tstats->tx_bytes;
116 dev->stats.rx_packets = sum.rx_packets;
117 dev->stats.rx_bytes = sum.rx_bytes;
118 dev->stats.tx_packets = sum.tx_packets;
119 dev->stats.tx_bytes = sum.tx_bytes;
120 return &dev->stats;
124 * Locking : hash tables are protected by RCU and RTNL
127 struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
129 struct dst_entry *dst = t->dst_cache;
131 if (dst && dst->obsolete &&
132 dst->ops->check(dst, t->dst_cookie) == NULL) {
133 t->dst_cache = NULL;
134 dst_release(dst);
135 return NULL;
138 return dst;
140 EXPORT_SYMBOL_GPL(ip6_tnl_dst_check);
142 void ip6_tnl_dst_reset(struct ip6_tnl *t)
144 dst_release(t->dst_cache);
145 t->dst_cache = NULL;
147 EXPORT_SYMBOL_GPL(ip6_tnl_dst_reset);
149 void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
151 struct rt6_info *rt = (struct rt6_info *) dst;
152 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
153 dst_release(t->dst_cache);
154 t->dst_cache = dst;
156 EXPORT_SYMBOL_GPL(ip6_tnl_dst_store);
159 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
160 * @remote: the address of the tunnel exit-point
161 * @local: the address of the tunnel entry-point
163 * Return:
164 * tunnel matching given end-points if found,
165 * else fallback tunnel if its device is up,
166 * else %NULL
169 #define for_each_ip6_tunnel_rcu(start) \
170 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
172 static struct ip6_tnl *
173 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
175 unsigned int hash = HASH(remote, local);
176 struct ip6_tnl *t;
177 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
179 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
180 if (ipv6_addr_equal(local, &t->parms.laddr) &&
181 ipv6_addr_equal(remote, &t->parms.raddr) &&
182 (t->dev->flags & IFF_UP))
183 return t;
185 t = rcu_dereference(ip6n->tnls_wc[0]);
186 if (t && (t->dev->flags & IFF_UP))
187 return t;
189 return NULL;
193 * ip6_tnl_bucket - get head of list matching given tunnel parameters
194 * @p: parameters containing tunnel end-points
196 * Description:
197 * ip6_tnl_bucket() returns the head of the list matching the
198 * &struct in6_addr entries laddr and raddr in @p.
200 * Return: head of IPv6 tunnel list
203 static struct ip6_tnl __rcu **
204 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
206 const struct in6_addr *remote = &p->raddr;
207 const struct in6_addr *local = &p->laddr;
208 unsigned int h = 0;
209 int prio = 0;
211 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
212 prio = 1;
213 h = HASH(remote, local);
215 return &ip6n->tnls[prio][h];
219 * ip6_tnl_link - add tunnel to hash table
220 * @t: tunnel to be added
223 static void
224 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
226 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
228 rcu_assign_pointer(t->next , rtnl_dereference(*tp));
229 rcu_assign_pointer(*tp, t);
233 * ip6_tnl_unlink - remove tunnel from hash table
234 * @t: tunnel to be removed
237 static void
238 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
240 struct ip6_tnl __rcu **tp;
241 struct ip6_tnl *iter;
243 for (tp = ip6_tnl_bucket(ip6n, &t->parms);
244 (iter = rtnl_dereference(*tp)) != NULL;
245 tp = &iter->next) {
246 if (t == iter) {
247 rcu_assign_pointer(*tp, t->next);
248 break;
253 static void ip6_dev_free(struct net_device *dev)
255 free_percpu(dev->tstats);
256 free_netdev(dev);
259 static int ip6_tnl_create2(struct net_device *dev)
261 struct ip6_tnl *t = netdev_priv(dev);
262 struct net *net = dev_net(dev);
263 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
264 int err;
266 t = netdev_priv(dev);
267 err = ip6_tnl_dev_init(dev);
268 if (err < 0)
269 goto out;
271 err = register_netdevice(dev);
272 if (err < 0)
273 goto out;
275 strcpy(t->parms.name, dev->name);
276 dev->rtnl_link_ops = &ip6_link_ops;
278 dev_hold(dev);
279 ip6_tnl_link(ip6n, t);
280 return 0;
282 out:
283 return err;
287 * ip6_tnl_create - create a new tunnel
288 * @p: tunnel parameters
289 * @pt: pointer to new tunnel
291 * Description:
292 * Create tunnel matching given parameters.
294 * Return:
295 * created tunnel or NULL
298 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
300 struct net_device *dev;
301 struct ip6_tnl *t;
302 char name[IFNAMSIZ];
303 int err;
305 if (p->name[0])
306 strlcpy(name, p->name, IFNAMSIZ);
307 else
308 sprintf(name, "ip6tnl%%d");
310 dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
311 if (dev == NULL)
312 goto failed;
314 dev_net_set(dev, net);
316 t = netdev_priv(dev);
317 t->parms = *p;
318 err = ip6_tnl_create2(dev);
319 if (err < 0)
320 goto failed_free;
322 return t;
324 failed_free:
325 ip6_dev_free(dev);
326 failed:
327 return NULL;
331 * ip6_tnl_locate - find or create tunnel matching given parameters
332 * @p: tunnel parameters
333 * @create: != 0 if allowed to create new tunnel if no match found
335 * Description:
336 * ip6_tnl_locate() first tries to locate an existing tunnel
337 * based on @parms. If this is unsuccessful, but @create is set a new
338 * tunnel device is created and registered for use.
340 * Return:
341 * matching tunnel or NULL
344 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
345 struct __ip6_tnl_parm *p, int create)
347 const struct in6_addr *remote = &p->raddr;
348 const struct in6_addr *local = &p->laddr;
349 struct ip6_tnl __rcu **tp;
350 struct ip6_tnl *t;
351 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
353 for (tp = ip6_tnl_bucket(ip6n, p);
354 (t = rtnl_dereference(*tp)) != NULL;
355 tp = &t->next) {
356 if (ipv6_addr_equal(local, &t->parms.laddr) &&
357 ipv6_addr_equal(remote, &t->parms.raddr))
358 return t;
360 if (!create)
361 return NULL;
362 return ip6_tnl_create(net, p);
366 * ip6_tnl_dev_uninit - tunnel device uninitializer
367 * @dev: the device to be destroyed
369 * Description:
370 * ip6_tnl_dev_uninit() removes tunnel from its list
373 static void
374 ip6_tnl_dev_uninit(struct net_device *dev)
376 struct ip6_tnl *t = netdev_priv(dev);
377 struct net *net = dev_net(dev);
378 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
380 if (dev == ip6n->fb_tnl_dev)
381 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
382 else
383 ip6_tnl_unlink(ip6n, t);
384 ip6_tnl_dst_reset(t);
385 dev_put(dev);
389 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
390 * @skb: received socket buffer
392 * Return:
393 * 0 if none was found,
394 * else index to encapsulation limit
397 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
399 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
400 __u8 nexthdr = ipv6h->nexthdr;
401 __u16 off = sizeof (*ipv6h);
403 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
404 __u16 optlen = 0;
405 struct ipv6_opt_hdr *hdr;
406 if (raw + off + sizeof (*hdr) > skb->data &&
407 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
408 break;
410 hdr = (struct ipv6_opt_hdr *) (raw + off);
411 if (nexthdr == NEXTHDR_FRAGMENT) {
412 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
413 if (frag_hdr->frag_off)
414 break;
415 optlen = 8;
416 } else if (nexthdr == NEXTHDR_AUTH) {
417 optlen = (hdr->hdrlen + 2) << 2;
418 } else {
419 optlen = ipv6_optlen(hdr);
421 if (nexthdr == NEXTHDR_DEST) {
422 __u16 i = off + 2;
423 while (1) {
424 struct ipv6_tlv_tnl_enc_lim *tel;
426 /* No more room for encapsulation limit */
427 if (i + sizeof (*tel) > off + optlen)
428 break;
430 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
431 /* return index of option if found and valid */
432 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
433 tel->length == 1)
434 return i;
435 /* else jump to next option */
436 if (tel->type)
437 i += tel->length + 2;
438 else
439 i++;
442 nexthdr = hdr->nexthdr;
443 off += optlen;
445 return 0;
447 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
450 * ip6_tnl_err - tunnel error handler
452 * Description:
453 * ip6_tnl_err() should handle errors in the tunnel according
454 * to the specifications in RFC 2473.
457 static int
458 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
459 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
461 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
462 struct ip6_tnl *t;
463 int rel_msg = 0;
464 u8 rel_type = ICMPV6_DEST_UNREACH;
465 u8 rel_code = ICMPV6_ADDR_UNREACH;
466 __u32 rel_info = 0;
467 __u16 len;
468 int err = -ENOENT;
470 /* If the packet doesn't contain the original IPv6 header we are
471 in trouble since we might need the source address for further
472 processing of the error. */
474 rcu_read_lock();
475 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
476 &ipv6h->saddr)) == NULL)
477 goto out;
479 if (t->parms.proto != ipproto && t->parms.proto != 0)
480 goto out;
482 err = 0;
484 switch (*type) {
485 __u32 teli;
486 struct ipv6_tlv_tnl_enc_lim *tel;
487 __u32 mtu;
488 case ICMPV6_DEST_UNREACH:
489 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
490 t->parms.name);
491 rel_msg = 1;
492 break;
493 case ICMPV6_TIME_EXCEED:
494 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
495 net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
496 t->parms.name);
497 rel_msg = 1;
499 break;
500 case ICMPV6_PARAMPROB:
501 teli = 0;
502 if ((*code) == ICMPV6_HDR_FIELD)
503 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
505 if (teli && teli == *info - 2) {
506 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
507 if (tel->encap_limit == 0) {
508 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
509 t->parms.name);
510 rel_msg = 1;
512 } else {
513 net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
514 t->parms.name);
516 break;
517 case ICMPV6_PKT_TOOBIG:
518 mtu = *info - offset;
519 if (mtu < IPV6_MIN_MTU)
520 mtu = IPV6_MIN_MTU;
521 t->dev->mtu = mtu;
523 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
524 rel_type = ICMPV6_PKT_TOOBIG;
525 rel_code = 0;
526 rel_info = mtu;
527 rel_msg = 1;
529 break;
532 *type = rel_type;
533 *code = rel_code;
534 *info = rel_info;
535 *msg = rel_msg;
537 out:
538 rcu_read_unlock();
539 return err;
542 static int
543 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
544 u8 type, u8 code, int offset, __be32 info)
546 int rel_msg = 0;
547 u8 rel_type = type;
548 u8 rel_code = code;
549 __u32 rel_info = ntohl(info);
550 int err;
551 struct sk_buff *skb2;
552 const struct iphdr *eiph;
553 struct rtable *rt;
554 struct flowi4 fl4;
556 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
557 &rel_msg, &rel_info, offset);
558 if (err < 0)
559 return err;
561 if (rel_msg == 0)
562 return 0;
564 switch (rel_type) {
565 case ICMPV6_DEST_UNREACH:
566 if (rel_code != ICMPV6_ADDR_UNREACH)
567 return 0;
568 rel_type = ICMP_DEST_UNREACH;
569 rel_code = ICMP_HOST_UNREACH;
570 break;
571 case ICMPV6_PKT_TOOBIG:
572 if (rel_code != 0)
573 return 0;
574 rel_type = ICMP_DEST_UNREACH;
575 rel_code = ICMP_FRAG_NEEDED;
576 break;
577 case NDISC_REDIRECT:
578 rel_type = ICMP_REDIRECT;
579 rel_code = ICMP_REDIR_HOST;
580 default:
581 return 0;
584 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
585 return 0;
587 skb2 = skb_clone(skb, GFP_ATOMIC);
588 if (!skb2)
589 return 0;
591 skb_dst_drop(skb2);
593 skb_pull(skb2, offset);
594 skb_reset_network_header(skb2);
595 eiph = ip_hdr(skb2);
597 /* Try to guess incoming interface */
598 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
599 eiph->saddr, 0,
600 0, 0,
601 IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
602 if (IS_ERR(rt))
603 goto out;
605 skb2->dev = rt->dst.dev;
607 /* route "incoming" packet */
608 if (rt->rt_flags & RTCF_LOCAL) {
609 ip_rt_put(rt);
610 rt = NULL;
611 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
612 eiph->daddr, eiph->saddr,
613 0, 0,
614 IPPROTO_IPIP,
615 RT_TOS(eiph->tos), 0);
616 if (IS_ERR(rt) ||
617 rt->dst.dev->type != ARPHRD_TUNNEL) {
618 if (!IS_ERR(rt))
619 ip_rt_put(rt);
620 goto out;
622 skb_dst_set(skb2, &rt->dst);
623 } else {
624 ip_rt_put(rt);
625 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
626 skb2->dev) ||
627 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
628 goto out;
631 /* change mtu on this route */
632 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
633 if (rel_info > dst_mtu(skb_dst(skb2)))
634 goto out;
636 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
638 if (rel_type == ICMP_REDIRECT)
639 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
641 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
643 out:
644 kfree_skb(skb2);
645 return 0;
648 static int
649 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
650 u8 type, u8 code, int offset, __be32 info)
652 int rel_msg = 0;
653 u8 rel_type = type;
654 u8 rel_code = code;
655 __u32 rel_info = ntohl(info);
656 int err;
658 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
659 &rel_msg, &rel_info, offset);
660 if (err < 0)
661 return err;
663 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
664 struct rt6_info *rt;
665 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
667 if (!skb2)
668 return 0;
670 skb_dst_drop(skb2);
671 skb_pull(skb2, offset);
672 skb_reset_network_header(skb2);
674 /* Try to guess incoming interface */
675 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
676 NULL, 0, 0);
678 if (rt && rt->dst.dev)
679 skb2->dev = rt->dst.dev;
681 icmpv6_send(skb2, rel_type, rel_code, rel_info);
683 ip6_rt_put(rt);
685 kfree_skb(skb2);
688 return 0;
691 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
692 const struct ipv6hdr *ipv6h,
693 struct sk_buff *skb)
695 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
697 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
698 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
700 return IP6_ECN_decapsulate(ipv6h, skb);
703 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
704 const struct ipv6hdr *ipv6h,
705 struct sk_buff *skb)
707 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
708 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
710 return IP6_ECN_decapsulate(ipv6h, skb);
713 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
714 const struct in6_addr *laddr,
715 const struct in6_addr *raddr)
717 struct __ip6_tnl_parm *p = &t->parms;
718 int ltype = ipv6_addr_type(laddr);
719 int rtype = ipv6_addr_type(raddr);
720 __u32 flags = 0;
722 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
723 flags = IP6_TNL_F_CAP_PER_PACKET;
724 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
725 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
726 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
727 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
728 if (ltype&IPV6_ADDR_UNICAST)
729 flags |= IP6_TNL_F_CAP_XMIT;
730 if (rtype&IPV6_ADDR_UNICAST)
731 flags |= IP6_TNL_F_CAP_RCV;
733 return flags;
735 EXPORT_SYMBOL(ip6_tnl_get_cap);
737 /* called with rcu_read_lock() */
738 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
739 const struct in6_addr *laddr,
740 const struct in6_addr *raddr)
742 struct __ip6_tnl_parm *p = &t->parms;
743 int ret = 0;
744 struct net *net = dev_net(t->dev);
746 if ((p->flags & IP6_TNL_F_CAP_RCV) ||
747 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
748 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
749 struct net_device *ldev = NULL;
751 if (p->link)
752 ldev = dev_get_by_index_rcu(net, p->link);
754 if ((ipv6_addr_is_multicast(laddr) ||
755 likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
756 likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
757 ret = 1;
759 return ret;
761 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
764 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
765 * @skb: received socket buffer
766 * @protocol: ethernet protocol ID
767 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
769 * Return: 0
772 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
773 __u8 ipproto,
774 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
775 const struct ipv6hdr *ipv6h,
776 struct sk_buff *skb))
778 struct ip6_tnl *t;
779 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
780 int err;
782 rcu_read_lock();
784 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
785 &ipv6h->daddr)) != NULL) {
786 struct pcpu_tstats *tstats;
788 if (t->parms.proto != ipproto && t->parms.proto != 0) {
789 rcu_read_unlock();
790 goto discard;
793 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
794 rcu_read_unlock();
795 goto discard;
798 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
799 t->dev->stats.rx_dropped++;
800 rcu_read_unlock();
801 goto discard;
803 secpath_reset(skb);
804 skb->mac_header = skb->network_header;
805 skb_reset_network_header(skb);
806 skb->protocol = htons(protocol);
807 skb->pkt_type = PACKET_HOST;
808 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
810 __skb_tunnel_rx(skb, t->dev);
812 err = dscp_ecn_decapsulate(t, ipv6h, skb);
813 if (unlikely(err)) {
814 if (log_ecn_error)
815 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
816 &ipv6h->saddr,
817 ipv6_get_dsfield(ipv6h));
818 if (err > 1) {
819 ++t->dev->stats.rx_frame_errors;
820 ++t->dev->stats.rx_errors;
821 rcu_read_unlock();
822 goto discard;
826 tstats = this_cpu_ptr(t->dev->tstats);
827 tstats->rx_packets++;
828 tstats->rx_bytes += skb->len;
830 netif_rx(skb);
832 rcu_read_unlock();
833 return 0;
835 rcu_read_unlock();
836 return 1;
838 discard:
839 kfree_skb(skb);
840 return 0;
843 static int ip4ip6_rcv(struct sk_buff *skb)
845 return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
846 ip4ip6_dscp_ecn_decapsulate);
849 static int ip6ip6_rcv(struct sk_buff *skb)
851 return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
852 ip6ip6_dscp_ecn_decapsulate);
855 struct ipv6_tel_txoption {
856 struct ipv6_txoptions ops;
857 __u8 dst_opt[8];
860 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
862 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
864 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
865 opt->dst_opt[3] = 1;
866 opt->dst_opt[4] = encap_limit;
867 opt->dst_opt[5] = IPV6_TLV_PADN;
868 opt->dst_opt[6] = 1;
870 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
871 opt->ops.opt_nflen = 8;
875 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
876 * @t: the outgoing tunnel device
877 * @hdr: IPv6 header from the incoming packet
879 * Description:
880 * Avoid trivial tunneling loop by checking that tunnel exit-point
881 * doesn't match source of incoming packet.
883 * Return:
884 * 1 if conflict,
885 * 0 else
888 static inline bool
889 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
891 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
894 int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
896 struct __ip6_tnl_parm *p = &t->parms;
897 int ret = 0;
898 struct net *net = dev_net(t->dev);
900 if (p->flags & IP6_TNL_F_CAP_XMIT) {
901 struct net_device *ldev = NULL;
903 rcu_read_lock();
904 if (p->link)
905 ldev = dev_get_by_index_rcu(net, p->link);
907 if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
908 pr_warn("%s xmit: Local address not yet configured!\n",
909 p->name);
910 else if (!ipv6_addr_is_multicast(&p->raddr) &&
911 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
912 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
913 p->name);
914 else
915 ret = 1;
916 rcu_read_unlock();
918 return ret;
920 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
923 * ip6_tnl_xmit2 - encapsulate packet and send
924 * @skb: the outgoing socket buffer
925 * @dev: the outgoing tunnel device
926 * @dsfield: dscp code for outer header
927 * @fl: flow of tunneled packet
928 * @encap_limit: encapsulation limit
929 * @pmtu: Path MTU is stored if packet is too big
931 * Description:
932 * Build new header and do some sanity checks on the packet before sending
933 * it.
935 * Return:
936 * 0 on success
937 * -1 fail
938 * %-EMSGSIZE message too big. return mtu in this case.
941 static int ip6_tnl_xmit2(struct sk_buff *skb,
942 struct net_device *dev,
943 __u8 dsfield,
944 struct flowi6 *fl6,
945 int encap_limit,
946 __u32 *pmtu)
948 struct net *net = dev_net(dev);
949 struct ip6_tnl *t = netdev_priv(dev);
950 struct net_device_stats *stats = &t->dev->stats;
951 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
952 struct ipv6_tel_txoption opt;
953 struct dst_entry *dst = NULL, *ndst = NULL;
954 struct net_device *tdev;
955 int mtu;
956 unsigned int max_headroom = sizeof(struct ipv6hdr);
957 u8 proto;
958 int err = -1;
960 if (!fl6->flowi6_mark)
961 dst = ip6_tnl_dst_check(t);
962 if (!dst) {
963 ndst = ip6_route_output(net, NULL, fl6);
965 if (ndst->error)
966 goto tx_err_link_failure;
967 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
968 if (IS_ERR(ndst)) {
969 err = PTR_ERR(ndst);
970 ndst = NULL;
971 goto tx_err_link_failure;
973 dst = ndst;
976 tdev = dst->dev;
978 if (tdev == dev) {
979 stats->collisions++;
980 net_warn_ratelimited("%s: Local routing loop detected!\n",
981 t->parms.name);
982 goto tx_err_dst_release;
984 mtu = dst_mtu(dst) - sizeof (*ipv6h);
985 if (encap_limit >= 0) {
986 max_headroom += 8;
987 mtu -= 8;
989 if (mtu < IPV6_MIN_MTU)
990 mtu = IPV6_MIN_MTU;
991 if (skb_dst(skb))
992 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
993 if (skb->len > mtu) {
994 *pmtu = mtu;
995 err = -EMSGSIZE;
996 goto tx_err_dst_release;
1000 * Okay, now see if we can stuff it in the buffer as-is.
1002 max_headroom += LL_RESERVED_SPACE(tdev);
1004 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1005 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1006 struct sk_buff *new_skb;
1008 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
1009 goto tx_err_dst_release;
1011 if (skb->sk)
1012 skb_set_owner_w(new_skb, skb->sk);
1013 consume_skb(skb);
1014 skb = new_skb;
1016 skb_dst_drop(skb);
1017 if (fl6->flowi6_mark) {
1018 skb_dst_set(skb, dst);
1019 ndst = NULL;
1020 } else {
1021 skb_dst_set_noref(skb, dst);
1023 skb->transport_header = skb->network_header;
1025 proto = fl6->flowi6_proto;
1026 if (encap_limit >= 0) {
1027 init_tel_txopt(&opt, encap_limit);
1028 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1030 skb_push(skb, sizeof(struct ipv6hdr));
1031 skb_reset_network_header(skb);
1032 ipv6h = ipv6_hdr(skb);
1033 ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield), fl6->flowlabel);
1034 ipv6h->hop_limit = t->parms.hop_limit;
1035 ipv6h->nexthdr = proto;
1036 ipv6h->saddr = fl6->saddr;
1037 ipv6h->daddr = fl6->daddr;
1038 ip6tunnel_xmit(skb, dev);
1039 if (ndst)
1040 ip6_tnl_dst_store(t, ndst);
1041 return 0;
1042 tx_err_link_failure:
1043 stats->tx_carrier_errors++;
1044 dst_link_failure(skb);
1045 tx_err_dst_release:
1046 dst_release(ndst);
1047 return err;
1050 static inline int
1051 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1053 struct ip6_tnl *t = netdev_priv(dev);
1054 const struct iphdr *iph = ip_hdr(skb);
1055 int encap_limit = -1;
1056 struct flowi6 fl6;
1057 __u8 dsfield;
1058 __u32 mtu;
1059 int err;
1061 if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
1062 !ip6_tnl_xmit_ctl(t))
1063 return -1;
1065 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1066 encap_limit = t->parms.encap_limit;
1068 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1069 fl6.flowi6_proto = IPPROTO_IPIP;
1071 dsfield = ipv4_get_dsfield(iph);
1073 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1074 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1075 & IPV6_TCLASS_MASK;
1076 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1077 fl6.flowi6_mark = skb->mark;
1079 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1080 if (err != 0) {
1081 /* XXX: send ICMP error even if DF is not set. */
1082 if (err == -EMSGSIZE)
1083 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1084 htonl(mtu));
1085 return -1;
1088 return 0;
1091 static inline int
1092 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1094 struct ip6_tnl *t = netdev_priv(dev);
1095 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1096 int encap_limit = -1;
1097 __u16 offset;
1098 struct flowi6 fl6;
1099 __u8 dsfield;
1100 __u32 mtu;
1101 int err;
1103 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1104 !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1105 return -1;
1107 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1108 if (offset > 0) {
1109 struct ipv6_tlv_tnl_enc_lim *tel;
1110 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1111 if (tel->encap_limit == 0) {
1112 icmpv6_send(skb, ICMPV6_PARAMPROB,
1113 ICMPV6_HDR_FIELD, offset + 2);
1114 return -1;
1116 encap_limit = tel->encap_limit - 1;
1117 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1118 encap_limit = t->parms.encap_limit;
1120 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1121 fl6.flowi6_proto = IPPROTO_IPV6;
1123 dsfield = ipv6_get_dsfield(ipv6h);
1124 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1125 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1126 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1127 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1128 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1129 fl6.flowi6_mark = skb->mark;
1131 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1132 if (err != 0) {
1133 if (err == -EMSGSIZE)
1134 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1135 return -1;
1138 return 0;
1141 static netdev_tx_t
1142 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1144 struct ip6_tnl *t = netdev_priv(dev);
1145 struct net_device_stats *stats = &t->dev->stats;
1146 int ret;
1148 switch (skb->protocol) {
1149 case htons(ETH_P_IP):
1150 ret = ip4ip6_tnl_xmit(skb, dev);
1151 break;
1152 case htons(ETH_P_IPV6):
1153 ret = ip6ip6_tnl_xmit(skb, dev);
1154 break;
1155 default:
1156 goto tx_err;
1159 if (ret < 0)
1160 goto tx_err;
1162 return NETDEV_TX_OK;
1164 tx_err:
1165 stats->tx_errors++;
1166 stats->tx_dropped++;
1167 kfree_skb(skb);
1168 return NETDEV_TX_OK;
1171 static void ip6_tnl_link_config(struct ip6_tnl *t)
1173 struct net_device *dev = t->dev;
1174 struct __ip6_tnl_parm *p = &t->parms;
1175 struct flowi6 *fl6 = &t->fl.u.ip6;
1177 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1178 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1180 /* Set up flowi template */
1181 fl6->saddr = p->laddr;
1182 fl6->daddr = p->raddr;
1183 fl6->flowi6_oif = p->link;
1184 fl6->flowlabel = 0;
1186 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1187 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1188 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1189 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1191 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1192 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1194 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1195 dev->flags |= IFF_POINTOPOINT;
1196 else
1197 dev->flags &= ~IFF_POINTOPOINT;
1199 dev->iflink = p->link;
1201 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1202 int strict = (ipv6_addr_type(&p->raddr) &
1203 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1205 struct rt6_info *rt = rt6_lookup(dev_net(dev),
1206 &p->raddr, &p->laddr,
1207 p->link, strict);
1209 if (rt == NULL)
1210 return;
1212 if (rt->dst.dev) {
1213 dev->hard_header_len = rt->dst.dev->hard_header_len +
1214 sizeof (struct ipv6hdr);
1216 dev->mtu = rt->dst.dev->mtu - sizeof (struct ipv6hdr);
1217 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1218 dev->mtu-=8;
1220 if (dev->mtu < IPV6_MIN_MTU)
1221 dev->mtu = IPV6_MIN_MTU;
1223 ip6_rt_put(rt);
1228 * ip6_tnl_change - update the tunnel parameters
1229 * @t: tunnel to be changed
1230 * @p: tunnel configuration parameters
1232 * Description:
1233 * ip6_tnl_change() updates the tunnel parameters
1236 static int
1237 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1239 t->parms.laddr = p->laddr;
1240 t->parms.raddr = p->raddr;
1241 t->parms.flags = p->flags;
1242 t->parms.hop_limit = p->hop_limit;
1243 t->parms.encap_limit = p->encap_limit;
1244 t->parms.flowinfo = p->flowinfo;
1245 t->parms.link = p->link;
1246 t->parms.proto = p->proto;
1247 ip6_tnl_dst_reset(t);
1248 ip6_tnl_link_config(t);
1249 return 0;
1252 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1254 struct net *net = dev_net(t->dev);
1255 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1256 int err;
1258 ip6_tnl_unlink(ip6n, t);
1259 synchronize_net();
1260 err = ip6_tnl_change(t, p);
1261 ip6_tnl_link(ip6n, t);
1262 netdev_state_change(t->dev);
1263 return err;
1266 static void
1267 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1269 p->laddr = u->laddr;
1270 p->raddr = u->raddr;
1271 p->flags = u->flags;
1272 p->hop_limit = u->hop_limit;
1273 p->encap_limit = u->encap_limit;
1274 p->flowinfo = u->flowinfo;
1275 p->link = u->link;
1276 p->proto = u->proto;
1277 memcpy(p->name, u->name, sizeof(u->name));
1280 static void
1281 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1283 u->laddr = p->laddr;
1284 u->raddr = p->raddr;
1285 u->flags = p->flags;
1286 u->hop_limit = p->hop_limit;
1287 u->encap_limit = p->encap_limit;
1288 u->flowinfo = p->flowinfo;
1289 u->link = p->link;
1290 u->proto = p->proto;
1291 memcpy(u->name, p->name, sizeof(u->name));
1295 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1296 * @dev: virtual device associated with tunnel
1297 * @ifr: parameters passed from userspace
1298 * @cmd: command to be performed
1300 * Description:
1301 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1302 * from userspace.
1304 * The possible commands are the following:
1305 * %SIOCGETTUNNEL: get tunnel parameters for device
1306 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1307 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1308 * %SIOCDELTUNNEL: delete tunnel
1310 * The fallback device "ip6tnl0", created during module
1311 * initialization, can be used for creating other tunnel devices.
1313 * Return:
1314 * 0 on success,
1315 * %-EFAULT if unable to copy data to or from userspace,
1316 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1317 * %-EINVAL if passed tunnel parameters are invalid,
1318 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1319 * %-ENODEV if attempting to change or delete a nonexisting device
1322 static int
1323 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1325 int err = 0;
1326 struct ip6_tnl_parm p;
1327 struct __ip6_tnl_parm p1;
1328 struct ip6_tnl *t = NULL;
1329 struct net *net = dev_net(dev);
1330 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1332 switch (cmd) {
1333 case SIOCGETTUNNEL:
1334 if (dev == ip6n->fb_tnl_dev) {
1335 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1336 err = -EFAULT;
1337 break;
1339 ip6_tnl_parm_from_user(&p1, &p);
1340 t = ip6_tnl_locate(net, &p1, 0);
1341 } else {
1342 memset(&p, 0, sizeof(p));
1344 if (t == NULL)
1345 t = netdev_priv(dev);
1346 ip6_tnl_parm_to_user(&p, &t->parms);
1347 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1348 err = -EFAULT;
1350 break;
1351 case SIOCADDTUNNEL:
1352 case SIOCCHGTUNNEL:
1353 err = -EPERM;
1354 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1355 break;
1356 err = -EFAULT;
1357 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1358 break;
1359 err = -EINVAL;
1360 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1361 p.proto != 0)
1362 break;
1363 ip6_tnl_parm_from_user(&p1, &p);
1364 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1365 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1366 if (t != NULL) {
1367 if (t->dev != dev) {
1368 err = -EEXIST;
1369 break;
1371 } else
1372 t = netdev_priv(dev);
1374 err = ip6_tnl_update(t, &p1);
1376 if (t) {
1377 err = 0;
1378 ip6_tnl_parm_to_user(&p, &t->parms);
1379 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1380 err = -EFAULT;
1382 } else
1383 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1384 break;
1385 case SIOCDELTUNNEL:
1386 err = -EPERM;
1387 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1388 break;
1390 if (dev == ip6n->fb_tnl_dev) {
1391 err = -EFAULT;
1392 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1393 break;
1394 err = -ENOENT;
1395 ip6_tnl_parm_from_user(&p1, &p);
1396 t = ip6_tnl_locate(net, &p1, 0);
1397 if (t == NULL)
1398 break;
1399 err = -EPERM;
1400 if (t->dev == ip6n->fb_tnl_dev)
1401 break;
1402 dev = t->dev;
1404 err = 0;
1405 unregister_netdevice(dev);
1406 break;
1407 default:
1408 err = -EINVAL;
1410 return err;
1414 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1415 * @dev: virtual device associated with tunnel
1416 * @new_mtu: the new mtu
1418 * Return:
1419 * 0 on success,
1420 * %-EINVAL if mtu too small
1423 static int
1424 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1426 if (new_mtu < IPV6_MIN_MTU) {
1427 return -EINVAL;
1429 dev->mtu = new_mtu;
1430 return 0;
1434 static const struct net_device_ops ip6_tnl_netdev_ops = {
1435 .ndo_uninit = ip6_tnl_dev_uninit,
1436 .ndo_start_xmit = ip6_tnl_xmit,
1437 .ndo_do_ioctl = ip6_tnl_ioctl,
1438 .ndo_change_mtu = ip6_tnl_change_mtu,
1439 .ndo_get_stats = ip6_get_stats,
1444 * ip6_tnl_dev_setup - setup virtual tunnel device
1445 * @dev: virtual device associated with tunnel
1447 * Description:
1448 * Initialize function pointers and device parameters
1451 static void ip6_tnl_dev_setup(struct net_device *dev)
1453 struct ip6_tnl *t;
1455 dev->netdev_ops = &ip6_tnl_netdev_ops;
1456 dev->destructor = ip6_dev_free;
1458 dev->type = ARPHRD_TUNNEL6;
1459 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1460 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1461 t = netdev_priv(dev);
1462 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1463 dev->mtu-=8;
1464 dev->flags |= IFF_NOARP;
1465 dev->addr_len = sizeof(struct in6_addr);
1466 dev->features |= NETIF_F_NETNS_LOCAL;
1467 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1472 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1473 * @dev: virtual device associated with tunnel
1476 static inline int
1477 ip6_tnl_dev_init_gen(struct net_device *dev)
1479 struct ip6_tnl *t = netdev_priv(dev);
1481 t->dev = dev;
1482 dev->tstats = alloc_percpu(struct pcpu_tstats);
1483 if (!dev->tstats)
1484 return -ENOMEM;
1485 return 0;
1489 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1490 * @dev: virtual device associated with tunnel
1493 static int ip6_tnl_dev_init(struct net_device *dev)
1495 struct ip6_tnl *t = netdev_priv(dev);
1496 int err = ip6_tnl_dev_init_gen(dev);
1498 if (err)
1499 return err;
1500 ip6_tnl_link_config(t);
1501 return 0;
1505 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1506 * @dev: fallback device
1508 * Return: 0
1511 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1513 struct ip6_tnl *t = netdev_priv(dev);
1514 struct net *net = dev_net(dev);
1515 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1516 int err = ip6_tnl_dev_init_gen(dev);
1518 if (err)
1519 return err;
1521 t->parms.proto = IPPROTO_IPV6;
1522 dev_hold(dev);
1524 ip6_tnl_link_config(t);
1526 rcu_assign_pointer(ip6n->tnls_wc[0], t);
1527 return 0;
1530 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1532 u8 proto;
1534 if (!data)
1535 return 0;
1537 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1538 if (proto != IPPROTO_IPV6 &&
1539 proto != IPPROTO_IPIP &&
1540 proto != 0)
1541 return -EINVAL;
1543 return 0;
1546 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1547 struct __ip6_tnl_parm *parms)
1549 memset(parms, 0, sizeof(*parms));
1551 if (!data)
1552 return;
1554 if (data[IFLA_IPTUN_LINK])
1555 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1557 if (data[IFLA_IPTUN_LOCAL])
1558 nla_memcpy(&parms->laddr, data[IFLA_IPTUN_LOCAL],
1559 sizeof(struct in6_addr));
1561 if (data[IFLA_IPTUN_REMOTE])
1562 nla_memcpy(&parms->raddr, data[IFLA_IPTUN_REMOTE],
1563 sizeof(struct in6_addr));
1565 if (data[IFLA_IPTUN_TTL])
1566 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1568 if (data[IFLA_IPTUN_ENCAP_LIMIT])
1569 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1571 if (data[IFLA_IPTUN_FLOWINFO])
1572 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1574 if (data[IFLA_IPTUN_FLAGS])
1575 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1577 if (data[IFLA_IPTUN_PROTO])
1578 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1581 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1582 struct nlattr *tb[], struct nlattr *data[])
1584 struct net *net = dev_net(dev);
1585 struct ip6_tnl *nt;
1587 nt = netdev_priv(dev);
1588 ip6_tnl_netlink_parms(data, &nt->parms);
1590 if (ip6_tnl_locate(net, &nt->parms, 0))
1591 return -EEXIST;
1593 return ip6_tnl_create2(dev);
1596 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1597 struct nlattr *data[])
1599 struct ip6_tnl *t;
1600 struct __ip6_tnl_parm p;
1601 struct net *net = dev_net(dev);
1602 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1604 if (dev == ip6n->fb_tnl_dev)
1605 return -EINVAL;
1607 ip6_tnl_netlink_parms(data, &p);
1609 t = ip6_tnl_locate(net, &p, 0);
1611 if (t) {
1612 if (t->dev != dev)
1613 return -EEXIST;
1614 } else
1615 t = netdev_priv(dev);
1617 return ip6_tnl_update(t, &p);
1620 static size_t ip6_tnl_get_size(const struct net_device *dev)
1622 return
1623 /* IFLA_IPTUN_LINK */
1624 nla_total_size(4) +
1625 /* IFLA_IPTUN_LOCAL */
1626 nla_total_size(sizeof(struct in6_addr)) +
1627 /* IFLA_IPTUN_REMOTE */
1628 nla_total_size(sizeof(struct in6_addr)) +
1629 /* IFLA_IPTUN_TTL */
1630 nla_total_size(1) +
1631 /* IFLA_IPTUN_ENCAP_LIMIT */
1632 nla_total_size(1) +
1633 /* IFLA_IPTUN_FLOWINFO */
1634 nla_total_size(4) +
1635 /* IFLA_IPTUN_FLAGS */
1636 nla_total_size(4) +
1637 /* IFLA_IPTUN_PROTO */
1638 nla_total_size(1) +
1642 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
1644 struct ip6_tnl *tunnel = netdev_priv(dev);
1645 struct __ip6_tnl_parm *parm = &tunnel->parms;
1647 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1648 nla_put(skb, IFLA_IPTUN_LOCAL, sizeof(struct in6_addr),
1649 &parm->raddr) ||
1650 nla_put(skb, IFLA_IPTUN_REMOTE, sizeof(struct in6_addr),
1651 &parm->laddr) ||
1652 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
1653 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
1654 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
1655 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
1656 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
1657 goto nla_put_failure;
1658 return 0;
1660 nla_put_failure:
1661 return -EMSGSIZE;
1664 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
1665 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1666 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) },
1667 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) },
1668 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1669 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 },
1670 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 },
1671 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 },
1672 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1675 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
1676 .kind = "ip6tnl",
1677 .maxtype = IFLA_IPTUN_MAX,
1678 .policy = ip6_tnl_policy,
1679 .priv_size = sizeof(struct ip6_tnl),
1680 .setup = ip6_tnl_dev_setup,
1681 .validate = ip6_tnl_validate,
1682 .newlink = ip6_tnl_newlink,
1683 .changelink = ip6_tnl_changelink,
1684 .get_size = ip6_tnl_get_size,
1685 .fill_info = ip6_tnl_fill_info,
1688 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1689 .handler = ip4ip6_rcv,
1690 .err_handler = ip4ip6_err,
1691 .priority = 1,
1694 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1695 .handler = ip6ip6_rcv,
1696 .err_handler = ip6ip6_err,
1697 .priority = 1,
1700 static void __net_exit ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n)
1702 int h;
1703 struct ip6_tnl *t;
1704 LIST_HEAD(list);
1706 for (h = 0; h < HASH_SIZE; h++) {
1707 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1708 while (t != NULL) {
1709 unregister_netdevice_queue(t->dev, &list);
1710 t = rtnl_dereference(t->next);
1714 t = rtnl_dereference(ip6n->tnls_wc[0]);
1715 unregister_netdevice_queue(t->dev, &list);
1716 unregister_netdevice_many(&list);
1719 static int __net_init ip6_tnl_init_net(struct net *net)
1721 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1722 struct ip6_tnl *t = NULL;
1723 int err;
1725 ip6n->tnls[0] = ip6n->tnls_wc;
1726 ip6n->tnls[1] = ip6n->tnls_r_l;
1728 err = -ENOMEM;
1729 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1730 ip6_tnl_dev_setup);
1732 if (!ip6n->fb_tnl_dev)
1733 goto err_alloc_dev;
1734 dev_net_set(ip6n->fb_tnl_dev, net);
1736 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1737 if (err < 0)
1738 goto err_register;
1740 err = register_netdev(ip6n->fb_tnl_dev);
1741 if (err < 0)
1742 goto err_register;
1744 t = netdev_priv(ip6n->fb_tnl_dev);
1746 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1747 return 0;
1749 err_register:
1750 ip6_dev_free(ip6n->fb_tnl_dev);
1751 err_alloc_dev:
1752 return err;
1755 static void __net_exit ip6_tnl_exit_net(struct net *net)
1757 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1759 rtnl_lock();
1760 ip6_tnl_destroy_tunnels(ip6n);
1761 rtnl_unlock();
1764 static struct pernet_operations ip6_tnl_net_ops = {
1765 .init = ip6_tnl_init_net,
1766 .exit = ip6_tnl_exit_net,
1767 .id = &ip6_tnl_net_id,
1768 .size = sizeof(struct ip6_tnl_net),
1772 * ip6_tunnel_init - register protocol and reserve needed resources
1774 * Return: 0 on success
1777 static int __init ip6_tunnel_init(void)
1779 int err;
1781 err = register_pernet_device(&ip6_tnl_net_ops);
1782 if (err < 0)
1783 goto out_pernet;
1785 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1786 if (err < 0) {
1787 pr_err("%s: can't register ip4ip6\n", __func__);
1788 goto out_ip4ip6;
1791 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1792 if (err < 0) {
1793 pr_err("%s: can't register ip6ip6\n", __func__);
1794 goto out_ip6ip6;
1796 err = rtnl_link_register(&ip6_link_ops);
1797 if (err < 0)
1798 goto rtnl_link_failed;
1800 return 0;
1802 rtnl_link_failed:
1803 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1804 out_ip6ip6:
1805 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1806 out_ip4ip6:
1807 unregister_pernet_device(&ip6_tnl_net_ops);
1808 out_pernet:
1809 return err;
1813 * ip6_tunnel_cleanup - free resources and unregister protocol
1816 static void __exit ip6_tunnel_cleanup(void)
1818 rtnl_link_unregister(&ip6_link_ops);
1819 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1820 pr_info("%s: can't deregister ip4ip6\n", __func__);
1822 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1823 pr_info("%s: can't deregister ip6ip6\n", __func__);
1825 unregister_pernet_device(&ip6_tnl_net_ops);
1828 module_init(ip6_tunnel_init);
1829 module_exit(ip6_tunnel_cleanup);