x86/paravirt: Remove paravirt_enabled()
[linux/fpc-iii.git] / net / ipv6 / ip6_tunnel.c
blob1f20345cbc970e88dd3b1ace2cc111cb56e5f73b
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/net.h>
33 #include <linux/in6.h>
34 #include <linux/netdevice.h>
35 #include <linux/if_arp.h>
36 #include <linux/icmpv6.h>
37 #include <linux/init.h>
38 #include <linux/route.h>
39 #include <linux/rtnetlink.h>
40 #include <linux/netfilter_ipv6.h>
41 #include <linux/slab.h>
42 #include <linux/hash.h>
43 #include <linux/etherdevice.h>
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_RTNL_LINK("ip6tnl");
65 MODULE_ALIAS_NETDEV("ip6tnl0");
67 #define HASH_SIZE_SHIFT 5
68 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
70 static bool log_ecn_error = true;
71 module_param(log_ecn_error, bool, 0644);
72 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
74 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
76 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
78 return hash_32(hash, HASH_SIZE_SHIFT);
81 static int ip6_tnl_dev_init(struct net_device *dev);
82 static void ip6_tnl_dev_setup(struct net_device *dev);
83 static struct rtnl_link_ops ip6_link_ops __read_mostly;
85 static int ip6_tnl_net_id __read_mostly;
86 struct ip6_tnl_net {
87 /* the IPv6 tunnel fallback device */
88 struct net_device *fb_tnl_dev;
89 /* lists for storing tunnels in use */
90 struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
91 struct ip6_tnl __rcu *tnls_wc[1];
92 struct ip6_tnl __rcu **tnls[2];
95 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
97 struct pcpu_sw_netstats tmp, sum = { 0 };
98 int i;
100 for_each_possible_cpu(i) {
101 unsigned int start;
102 const struct pcpu_sw_netstats *tstats =
103 per_cpu_ptr(dev->tstats, i);
105 do {
106 start = u64_stats_fetch_begin_irq(&tstats->syncp);
107 tmp.rx_packets = tstats->rx_packets;
108 tmp.rx_bytes = tstats->rx_bytes;
109 tmp.tx_packets = tstats->tx_packets;
110 tmp.tx_bytes = tstats->tx_bytes;
111 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
113 sum.rx_packets += tmp.rx_packets;
114 sum.rx_bytes += tmp.rx_bytes;
115 sum.tx_packets += tmp.tx_packets;
116 sum.tx_bytes += tmp.tx_bytes;
118 dev->stats.rx_packets = sum.rx_packets;
119 dev->stats.rx_bytes = sum.rx_bytes;
120 dev->stats.tx_packets = sum.tx_packets;
121 dev->stats.tx_bytes = sum.tx_bytes;
122 return &dev->stats;
126 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
127 * @remote: the address of the tunnel exit-point
128 * @local: the address of the tunnel entry-point
130 * Return:
131 * tunnel matching given end-points if found,
132 * else fallback tunnel if its device is up,
133 * else %NULL
136 #define for_each_ip6_tunnel_rcu(start) \
137 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
139 static struct ip6_tnl *
140 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
142 unsigned int hash = HASH(remote, local);
143 struct ip6_tnl *t;
144 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
145 struct in6_addr any;
147 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
148 if (ipv6_addr_equal(local, &t->parms.laddr) &&
149 ipv6_addr_equal(remote, &t->parms.raddr) &&
150 (t->dev->flags & IFF_UP))
151 return t;
154 memset(&any, 0, sizeof(any));
155 hash = HASH(&any, local);
156 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
157 if (ipv6_addr_equal(local, &t->parms.laddr) &&
158 (t->dev->flags & IFF_UP))
159 return t;
162 hash = HASH(remote, &any);
163 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
164 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
165 (t->dev->flags & IFF_UP))
166 return t;
169 t = rcu_dereference(ip6n->tnls_wc[0]);
170 if (t && (t->dev->flags & IFF_UP))
171 return t;
173 return NULL;
177 * ip6_tnl_bucket - get head of list matching given tunnel parameters
178 * @p: parameters containing tunnel end-points
180 * Description:
181 * ip6_tnl_bucket() returns the head of the list matching the
182 * &struct in6_addr entries laddr and raddr in @p.
184 * Return: head of IPv6 tunnel list
187 static struct ip6_tnl __rcu **
188 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
190 const struct in6_addr *remote = &p->raddr;
191 const struct in6_addr *local = &p->laddr;
192 unsigned int h = 0;
193 int prio = 0;
195 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
196 prio = 1;
197 h = HASH(remote, local);
199 return &ip6n->tnls[prio][h];
203 * ip6_tnl_link - add tunnel to hash table
204 * @t: tunnel to be added
207 static void
208 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
210 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
212 rcu_assign_pointer(t->next , rtnl_dereference(*tp));
213 rcu_assign_pointer(*tp, t);
217 * ip6_tnl_unlink - remove tunnel from hash table
218 * @t: tunnel to be removed
221 static void
222 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
224 struct ip6_tnl __rcu **tp;
225 struct ip6_tnl *iter;
227 for (tp = ip6_tnl_bucket(ip6n, &t->parms);
228 (iter = rtnl_dereference(*tp)) != NULL;
229 tp = &iter->next) {
230 if (t == iter) {
231 rcu_assign_pointer(*tp, t->next);
232 break;
237 static void ip6_dev_free(struct net_device *dev)
239 struct ip6_tnl *t = netdev_priv(dev);
241 dst_cache_destroy(&t->dst_cache);
242 free_percpu(dev->tstats);
243 free_netdev(dev);
246 static int ip6_tnl_create2(struct net_device *dev)
248 struct ip6_tnl *t = netdev_priv(dev);
249 struct net *net = dev_net(dev);
250 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
251 int err;
253 t = netdev_priv(dev);
255 dev->rtnl_link_ops = &ip6_link_ops;
256 err = register_netdevice(dev);
257 if (err < 0)
258 goto out;
260 strcpy(t->parms.name, dev->name);
262 dev_hold(dev);
263 ip6_tnl_link(ip6n, t);
264 return 0;
266 out:
267 return err;
271 * ip6_tnl_create - create a new tunnel
272 * @p: tunnel parameters
273 * @pt: pointer to new tunnel
275 * Description:
276 * Create tunnel matching given parameters.
278 * Return:
279 * created tunnel or error pointer
282 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
284 struct net_device *dev;
285 struct ip6_tnl *t;
286 char name[IFNAMSIZ];
287 int err = -ENOMEM;
289 if (p->name[0])
290 strlcpy(name, p->name, IFNAMSIZ);
291 else
292 sprintf(name, "ip6tnl%%d");
294 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
295 ip6_tnl_dev_setup);
296 if (!dev)
297 goto failed;
299 dev_net_set(dev, net);
301 t = netdev_priv(dev);
302 t->parms = *p;
303 t->net = dev_net(dev);
304 err = ip6_tnl_create2(dev);
305 if (err < 0)
306 goto failed_free;
308 return t;
310 failed_free:
311 ip6_dev_free(dev);
312 failed:
313 return ERR_PTR(err);
317 * ip6_tnl_locate - find or create tunnel matching given parameters
318 * @p: tunnel parameters
319 * @create: != 0 if allowed to create new tunnel if no match found
321 * Description:
322 * ip6_tnl_locate() first tries to locate an existing tunnel
323 * based on @parms. If this is unsuccessful, but @create is set a new
324 * tunnel device is created and registered for use.
326 * Return:
327 * matching tunnel or error pointer
330 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
331 struct __ip6_tnl_parm *p, int create)
333 const struct in6_addr *remote = &p->raddr;
334 const struct in6_addr *local = &p->laddr;
335 struct ip6_tnl __rcu **tp;
336 struct ip6_tnl *t;
337 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
339 for (tp = ip6_tnl_bucket(ip6n, p);
340 (t = rtnl_dereference(*tp)) != NULL;
341 tp = &t->next) {
342 if (ipv6_addr_equal(local, &t->parms.laddr) &&
343 ipv6_addr_equal(remote, &t->parms.raddr)) {
344 if (create)
345 return ERR_PTR(-EEXIST);
347 return t;
350 if (!create)
351 return ERR_PTR(-ENODEV);
352 return ip6_tnl_create(net, p);
356 * ip6_tnl_dev_uninit - tunnel device uninitializer
357 * @dev: the device to be destroyed
359 * Description:
360 * ip6_tnl_dev_uninit() removes tunnel from its list
363 static void
364 ip6_tnl_dev_uninit(struct net_device *dev)
366 struct ip6_tnl *t = netdev_priv(dev);
367 struct net *net = t->net;
368 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
370 if (dev == ip6n->fb_tnl_dev)
371 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
372 else
373 ip6_tnl_unlink(ip6n, t);
374 dst_cache_reset(&t->dst_cache);
375 dev_put(dev);
379 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
380 * @skb: received socket buffer
382 * Return:
383 * 0 if none was found,
384 * else index to encapsulation limit
387 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
389 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
390 __u8 nexthdr = ipv6h->nexthdr;
391 __u16 off = sizeof(*ipv6h);
393 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
394 __u16 optlen = 0;
395 struct ipv6_opt_hdr *hdr;
396 if (raw + off + sizeof(*hdr) > skb->data &&
397 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
398 break;
400 hdr = (struct ipv6_opt_hdr *) (raw + off);
401 if (nexthdr == NEXTHDR_FRAGMENT) {
402 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
403 if (frag_hdr->frag_off)
404 break;
405 optlen = 8;
406 } else if (nexthdr == NEXTHDR_AUTH) {
407 optlen = (hdr->hdrlen + 2) << 2;
408 } else {
409 optlen = ipv6_optlen(hdr);
411 if (nexthdr == NEXTHDR_DEST) {
412 __u16 i = off + 2;
413 while (1) {
414 struct ipv6_tlv_tnl_enc_lim *tel;
416 /* No more room for encapsulation limit */
417 if (i + sizeof (*tel) > off + optlen)
418 break;
420 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
421 /* return index of option if found and valid */
422 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
423 tel->length == 1)
424 return i;
425 /* else jump to next option */
426 if (tel->type)
427 i += tel->length + 2;
428 else
429 i++;
432 nexthdr = hdr->nexthdr;
433 off += optlen;
435 return 0;
437 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
440 * ip6_tnl_err - tunnel error handler
442 * Description:
443 * ip6_tnl_err() should handle errors in the tunnel according
444 * to the specifications in RFC 2473.
447 static int
448 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
449 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
451 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
452 struct ip6_tnl *t;
453 int rel_msg = 0;
454 u8 rel_type = ICMPV6_DEST_UNREACH;
455 u8 rel_code = ICMPV6_ADDR_UNREACH;
456 u8 tproto;
457 __u32 rel_info = 0;
458 __u16 len;
459 int err = -ENOENT;
461 /* If the packet doesn't contain the original IPv6 header we are
462 in trouble since we might need the source address for further
463 processing of the error. */
465 rcu_read_lock();
466 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
467 if (!t)
468 goto out;
470 tproto = ACCESS_ONCE(t->parms.proto);
471 if (tproto != ipproto && tproto != 0)
472 goto out;
474 err = 0;
476 switch (*type) {
477 __u32 teli;
478 struct ipv6_tlv_tnl_enc_lim *tel;
479 __u32 mtu;
480 case ICMPV6_DEST_UNREACH:
481 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
482 t->parms.name);
483 rel_msg = 1;
484 break;
485 case ICMPV6_TIME_EXCEED:
486 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
487 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
488 t->parms.name);
489 rel_msg = 1;
491 break;
492 case ICMPV6_PARAMPROB:
493 teli = 0;
494 if ((*code) == ICMPV6_HDR_FIELD)
495 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
497 if (teli && teli == *info - 2) {
498 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
499 if (tel->encap_limit == 0) {
500 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
501 t->parms.name);
502 rel_msg = 1;
504 } else {
505 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
506 t->parms.name);
508 break;
509 case ICMPV6_PKT_TOOBIG:
510 mtu = *info - offset;
511 if (mtu < IPV6_MIN_MTU)
512 mtu = IPV6_MIN_MTU;
513 t->dev->mtu = mtu;
515 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
516 if (len > mtu) {
517 rel_type = ICMPV6_PKT_TOOBIG;
518 rel_code = 0;
519 rel_info = mtu;
520 rel_msg = 1;
522 break;
525 *type = rel_type;
526 *code = rel_code;
527 *info = rel_info;
528 *msg = rel_msg;
530 out:
531 rcu_read_unlock();
532 return err;
535 static int
536 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
537 u8 type, u8 code, int offset, __be32 info)
539 int rel_msg = 0;
540 u8 rel_type = type;
541 u8 rel_code = code;
542 __u32 rel_info = ntohl(info);
543 int err;
544 struct sk_buff *skb2;
545 const struct iphdr *eiph;
546 struct rtable *rt;
547 struct flowi4 fl4;
549 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
550 &rel_msg, &rel_info, offset);
551 if (err < 0)
552 return err;
554 if (rel_msg == 0)
555 return 0;
557 switch (rel_type) {
558 case ICMPV6_DEST_UNREACH:
559 if (rel_code != ICMPV6_ADDR_UNREACH)
560 return 0;
561 rel_type = ICMP_DEST_UNREACH;
562 rel_code = ICMP_HOST_UNREACH;
563 break;
564 case ICMPV6_PKT_TOOBIG:
565 if (rel_code != 0)
566 return 0;
567 rel_type = ICMP_DEST_UNREACH;
568 rel_code = ICMP_FRAG_NEEDED;
569 break;
570 case NDISC_REDIRECT:
571 rel_type = ICMP_REDIRECT;
572 rel_code = ICMP_REDIR_HOST;
573 default:
574 return 0;
577 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
578 return 0;
580 skb2 = skb_clone(skb, GFP_ATOMIC);
581 if (!skb2)
582 return 0;
584 skb_dst_drop(skb2);
586 skb_pull(skb2, offset);
587 skb_reset_network_header(skb2);
588 eiph = ip_hdr(skb2);
590 /* Try to guess incoming interface */
591 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
592 eiph->saddr, 0,
593 0, 0,
594 IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
595 if (IS_ERR(rt))
596 goto out;
598 skb2->dev = rt->dst.dev;
600 /* route "incoming" packet */
601 if (rt->rt_flags & RTCF_LOCAL) {
602 ip_rt_put(rt);
603 rt = NULL;
604 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
605 eiph->daddr, eiph->saddr,
606 0, 0,
607 IPPROTO_IPIP,
608 RT_TOS(eiph->tos), 0);
609 if (IS_ERR(rt) ||
610 rt->dst.dev->type != ARPHRD_TUNNEL) {
611 if (!IS_ERR(rt))
612 ip_rt_put(rt);
613 goto out;
615 skb_dst_set(skb2, &rt->dst);
616 } else {
617 ip_rt_put(rt);
618 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
619 skb2->dev) ||
620 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
621 goto out;
624 /* change mtu on this route */
625 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
626 if (rel_info > dst_mtu(skb_dst(skb2)))
627 goto out;
629 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
631 if (rel_type == ICMP_REDIRECT)
632 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
634 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
636 out:
637 kfree_skb(skb2);
638 return 0;
641 static int
642 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
643 u8 type, u8 code, int offset, __be32 info)
645 int rel_msg = 0;
646 u8 rel_type = type;
647 u8 rel_code = code;
648 __u32 rel_info = ntohl(info);
649 int err;
651 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
652 &rel_msg, &rel_info, offset);
653 if (err < 0)
654 return err;
656 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
657 struct rt6_info *rt;
658 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
660 if (!skb2)
661 return 0;
663 skb_dst_drop(skb2);
664 skb_pull(skb2, offset);
665 skb_reset_network_header(skb2);
667 /* Try to guess incoming interface */
668 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
669 NULL, 0, 0);
671 if (rt && rt->dst.dev)
672 skb2->dev = rt->dst.dev;
674 icmpv6_send(skb2, rel_type, rel_code, rel_info);
676 ip6_rt_put(rt);
678 kfree_skb(skb2);
681 return 0;
684 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
685 const struct ipv6hdr *ipv6h,
686 struct sk_buff *skb)
688 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
690 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
691 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
693 return IP6_ECN_decapsulate(ipv6h, skb);
696 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
697 const struct ipv6hdr *ipv6h,
698 struct sk_buff *skb)
700 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
701 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
703 return IP6_ECN_decapsulate(ipv6h, skb);
706 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
707 const struct in6_addr *laddr,
708 const struct in6_addr *raddr)
710 struct __ip6_tnl_parm *p = &t->parms;
711 int ltype = ipv6_addr_type(laddr);
712 int rtype = ipv6_addr_type(raddr);
713 __u32 flags = 0;
715 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
716 flags = IP6_TNL_F_CAP_PER_PACKET;
717 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
718 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
719 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
720 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
721 if (ltype&IPV6_ADDR_UNICAST)
722 flags |= IP6_TNL_F_CAP_XMIT;
723 if (rtype&IPV6_ADDR_UNICAST)
724 flags |= IP6_TNL_F_CAP_RCV;
726 return flags;
728 EXPORT_SYMBOL(ip6_tnl_get_cap);
730 /* called with rcu_read_lock() */
731 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
732 const struct in6_addr *laddr,
733 const struct in6_addr *raddr)
735 struct __ip6_tnl_parm *p = &t->parms;
736 int ret = 0;
737 struct net *net = t->net;
739 if ((p->flags & IP6_TNL_F_CAP_RCV) ||
740 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
741 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
742 struct net_device *ldev = NULL;
744 if (p->link)
745 ldev = dev_get_by_index_rcu(net, p->link);
747 if ((ipv6_addr_is_multicast(laddr) ||
748 likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
749 likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
750 ret = 1;
752 return ret;
754 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
757 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
758 * @skb: received socket buffer
759 * @protocol: ethernet protocol ID
760 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
762 * Return: 0
765 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
766 __u8 ipproto,
767 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
768 const struct ipv6hdr *ipv6h,
769 struct sk_buff *skb))
771 struct ip6_tnl *t;
772 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
773 u8 tproto;
774 int err;
776 rcu_read_lock();
777 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
778 if (t) {
779 struct pcpu_sw_netstats *tstats;
781 tproto = ACCESS_ONCE(t->parms.proto);
782 if (tproto != ipproto && tproto != 0) {
783 rcu_read_unlock();
784 goto discard;
787 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
788 rcu_read_unlock();
789 goto discard;
792 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
793 t->dev->stats.rx_dropped++;
794 rcu_read_unlock();
795 goto discard;
797 skb->mac_header = skb->network_header;
798 skb_reset_network_header(skb);
799 skb->protocol = htons(protocol);
800 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
802 __skb_tunnel_rx(skb, t->dev, t->net);
804 err = dscp_ecn_decapsulate(t, ipv6h, skb);
805 if (unlikely(err)) {
806 if (log_ecn_error)
807 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
808 &ipv6h->saddr,
809 ipv6_get_dsfield(ipv6h));
810 if (err > 1) {
811 ++t->dev->stats.rx_frame_errors;
812 ++t->dev->stats.rx_errors;
813 rcu_read_unlock();
814 goto discard;
818 tstats = this_cpu_ptr(t->dev->tstats);
819 u64_stats_update_begin(&tstats->syncp);
820 tstats->rx_packets++;
821 tstats->rx_bytes += skb->len;
822 u64_stats_update_end(&tstats->syncp);
824 netif_rx(skb);
826 rcu_read_unlock();
827 return 0;
829 rcu_read_unlock();
830 return 1;
832 discard:
833 kfree_skb(skb);
834 return 0;
837 static int ip4ip6_rcv(struct sk_buff *skb)
839 return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
840 ip4ip6_dscp_ecn_decapsulate);
843 static int ip6ip6_rcv(struct sk_buff *skb)
845 return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
846 ip6ip6_dscp_ecn_decapsulate);
849 struct ipv6_tel_txoption {
850 struct ipv6_txoptions ops;
851 __u8 dst_opt[8];
854 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
856 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
858 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
859 opt->dst_opt[3] = 1;
860 opt->dst_opt[4] = encap_limit;
861 opt->dst_opt[5] = IPV6_TLV_PADN;
862 opt->dst_opt[6] = 1;
864 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
865 opt->ops.opt_nflen = 8;
869 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
870 * @t: the outgoing tunnel device
871 * @hdr: IPv6 header from the incoming packet
873 * Description:
874 * Avoid trivial tunneling loop by checking that tunnel exit-point
875 * doesn't match source of incoming packet.
877 * Return:
878 * 1 if conflict,
879 * 0 else
882 static inline bool
883 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
885 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
888 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
889 const struct in6_addr *laddr,
890 const struct in6_addr *raddr)
892 struct __ip6_tnl_parm *p = &t->parms;
893 int ret = 0;
894 struct net *net = t->net;
896 if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
897 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
898 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
899 struct net_device *ldev = NULL;
901 rcu_read_lock();
902 if (p->link)
903 ldev = dev_get_by_index_rcu(net, p->link);
905 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
906 pr_warn("%s xmit: Local address not yet configured!\n",
907 p->name);
908 else if (!ipv6_addr_is_multicast(raddr) &&
909 unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
910 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
911 p->name);
912 else
913 ret = 1;
914 rcu_read_unlock();
916 return ret;
918 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
921 * ip6_tnl_xmit2 - encapsulate packet and send
922 * @skb: the outgoing socket buffer
923 * @dev: the outgoing tunnel device
924 * @dsfield: dscp code for outer header
925 * @fl: flow of tunneled packet
926 * @encap_limit: encapsulation limit
927 * @pmtu: Path MTU is stored if packet is too big
929 * Description:
930 * Build new header and do some sanity checks on the packet before sending
931 * it.
933 * Return:
934 * 0 on success
935 * -1 fail
936 * %-EMSGSIZE message too big. return mtu in this case.
939 static int ip6_tnl_xmit2(struct sk_buff *skb,
940 struct net_device *dev,
941 __u8 dsfield,
942 struct flowi6 *fl6,
943 int encap_limit,
944 __u32 *pmtu)
946 struct ip6_tnl *t = netdev_priv(dev);
947 struct net *net = t->net;
948 struct net_device_stats *stats = &t->dev->stats;
949 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
950 struct ipv6_tel_txoption opt;
951 struct dst_entry *dst = NULL, *ndst = NULL;
952 struct net_device *tdev;
953 int mtu;
954 unsigned int max_headroom = sizeof(struct ipv6hdr);
955 u8 proto;
956 int err = -1;
958 /* NBMA tunnel */
959 if (ipv6_addr_any(&t->parms.raddr)) {
960 struct in6_addr *addr6;
961 struct neighbour *neigh;
962 int addr_type;
964 if (!skb_dst(skb))
965 goto tx_err_link_failure;
967 neigh = dst_neigh_lookup(skb_dst(skb),
968 &ipv6_hdr(skb)->daddr);
969 if (!neigh)
970 goto tx_err_link_failure;
972 addr6 = (struct in6_addr *)&neigh->primary_key;
973 addr_type = ipv6_addr_type(addr6);
975 if (addr_type == IPV6_ADDR_ANY)
976 addr6 = &ipv6_hdr(skb)->daddr;
978 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
979 neigh_release(neigh);
980 } else if (!fl6->flowi6_mark)
981 dst = dst_cache_get(&t->dst_cache);
983 if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
984 goto tx_err_link_failure;
986 if (!dst) {
987 dst = ip6_route_output(net, NULL, fl6);
989 if (dst->error)
990 goto tx_err_link_failure;
991 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
992 if (IS_ERR(dst)) {
993 err = PTR_ERR(dst);
994 dst = NULL;
995 goto tx_err_link_failure;
997 ndst = dst;
1000 tdev = dst->dev;
1002 if (tdev == dev) {
1003 stats->collisions++;
1004 net_warn_ratelimited("%s: Local routing loop detected!\n",
1005 t->parms.name);
1006 goto tx_err_dst_release;
1008 mtu = dst_mtu(dst) - sizeof(*ipv6h);
1009 if (encap_limit >= 0) {
1010 max_headroom += 8;
1011 mtu -= 8;
1013 if (mtu < IPV6_MIN_MTU)
1014 mtu = IPV6_MIN_MTU;
1015 if (skb_dst(skb))
1016 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1017 if (skb->len > mtu) {
1018 *pmtu = mtu;
1019 err = -EMSGSIZE;
1020 goto tx_err_dst_release;
1023 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1026 * Okay, now see if we can stuff it in the buffer as-is.
1028 max_headroom += LL_RESERVED_SPACE(tdev);
1030 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1031 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1032 struct sk_buff *new_skb;
1034 new_skb = skb_realloc_headroom(skb, max_headroom);
1035 if (!new_skb)
1036 goto tx_err_dst_release;
1038 if (skb->sk)
1039 skb_set_owner_w(new_skb, skb->sk);
1040 consume_skb(skb);
1041 skb = new_skb;
1044 if (!fl6->flowi6_mark && ndst)
1045 dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1046 skb_dst_set(skb, dst);
1048 skb->transport_header = skb->network_header;
1050 proto = fl6->flowi6_proto;
1051 if (encap_limit >= 0) {
1052 init_tel_txopt(&opt, encap_limit);
1053 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1056 if (likely(!skb->encapsulation)) {
1057 skb_reset_inner_headers(skb);
1058 skb->encapsulation = 1;
1061 skb_push(skb, sizeof(struct ipv6hdr));
1062 skb_reset_network_header(skb);
1063 ipv6h = ipv6_hdr(skb);
1064 ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
1065 ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1066 ipv6h->hop_limit = t->parms.hop_limit;
1067 ipv6h->nexthdr = proto;
1068 ipv6h->saddr = fl6->saddr;
1069 ipv6h->daddr = fl6->daddr;
1070 ip6tunnel_xmit(NULL, skb, dev);
1071 return 0;
1072 tx_err_link_failure:
1073 stats->tx_carrier_errors++;
1074 dst_link_failure(skb);
1075 tx_err_dst_release:
1076 dst_release(dst);
1077 return err;
1080 static inline int
1081 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1083 struct ip6_tnl *t = netdev_priv(dev);
1084 const struct iphdr *iph = ip_hdr(skb);
1085 int encap_limit = -1;
1086 struct flowi6 fl6;
1087 __u8 dsfield;
1088 __u32 mtu;
1089 u8 tproto;
1090 int err;
1092 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1094 tproto = ACCESS_ONCE(t->parms.proto);
1095 if (tproto != IPPROTO_IPIP && tproto != 0)
1096 return -1;
1098 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1099 encap_limit = t->parms.encap_limit;
1101 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1102 fl6.flowi6_proto = IPPROTO_IPIP;
1104 dsfield = ipv4_get_dsfield(iph);
1106 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1107 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1108 & IPV6_TCLASS_MASK;
1109 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1110 fl6.flowi6_mark = skb->mark;
1112 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1113 if (err != 0) {
1114 /* XXX: send ICMP error even if DF is not set. */
1115 if (err == -EMSGSIZE)
1116 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1117 htonl(mtu));
1118 return -1;
1121 return 0;
1124 static inline int
1125 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1127 struct ip6_tnl *t = netdev_priv(dev);
1128 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1129 int encap_limit = -1;
1130 __u16 offset;
1131 struct flowi6 fl6;
1132 __u8 dsfield;
1133 __u32 mtu;
1134 u8 tproto;
1135 int err;
1137 tproto = ACCESS_ONCE(t->parms.proto);
1138 if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1139 ip6_tnl_addr_conflict(t, ipv6h))
1140 return -1;
1142 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1143 if (offset > 0) {
1144 struct ipv6_tlv_tnl_enc_lim *tel;
1145 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1146 if (tel->encap_limit == 0) {
1147 icmpv6_send(skb, ICMPV6_PARAMPROB,
1148 ICMPV6_HDR_FIELD, offset + 2);
1149 return -1;
1151 encap_limit = tel->encap_limit - 1;
1152 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1153 encap_limit = t->parms.encap_limit;
1155 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1156 fl6.flowi6_proto = IPPROTO_IPV6;
1158 dsfield = ipv6_get_dsfield(ipv6h);
1159 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1160 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1161 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1162 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1163 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1164 fl6.flowi6_mark = skb->mark;
1166 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1167 if (err != 0) {
1168 if (err == -EMSGSIZE)
1169 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1170 return -1;
1173 return 0;
1176 static netdev_tx_t
1177 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1179 struct ip6_tnl *t = netdev_priv(dev);
1180 struct net_device_stats *stats = &t->dev->stats;
1181 int ret;
1183 switch (skb->protocol) {
1184 case htons(ETH_P_IP):
1185 ret = ip4ip6_tnl_xmit(skb, dev);
1186 break;
1187 case htons(ETH_P_IPV6):
1188 ret = ip6ip6_tnl_xmit(skb, dev);
1189 break;
1190 default:
1191 goto tx_err;
1194 if (ret < 0)
1195 goto tx_err;
1197 return NETDEV_TX_OK;
1199 tx_err:
1200 stats->tx_errors++;
1201 stats->tx_dropped++;
1202 kfree_skb(skb);
1203 return NETDEV_TX_OK;
1206 static void ip6_tnl_link_config(struct ip6_tnl *t)
1208 struct net_device *dev = t->dev;
1209 struct __ip6_tnl_parm *p = &t->parms;
1210 struct flowi6 *fl6 = &t->fl.u.ip6;
1212 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1213 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1215 /* Set up flowi template */
1216 fl6->saddr = p->laddr;
1217 fl6->daddr = p->raddr;
1218 fl6->flowi6_oif = p->link;
1219 fl6->flowlabel = 0;
1221 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1222 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1223 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1224 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1226 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1227 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1229 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1230 dev->flags |= IFF_POINTOPOINT;
1231 else
1232 dev->flags &= ~IFF_POINTOPOINT;
1234 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1235 int strict = (ipv6_addr_type(&p->raddr) &
1236 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1238 struct rt6_info *rt = rt6_lookup(t->net,
1239 &p->raddr, &p->laddr,
1240 p->link, strict);
1242 if (!rt)
1243 return;
1245 if (rt->dst.dev) {
1246 dev->hard_header_len = rt->dst.dev->hard_header_len +
1247 sizeof(struct ipv6hdr);
1249 dev->mtu = rt->dst.dev->mtu - sizeof(struct ipv6hdr);
1250 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1251 dev->mtu -= 8;
1253 if (dev->mtu < IPV6_MIN_MTU)
1254 dev->mtu = IPV6_MIN_MTU;
1256 ip6_rt_put(rt);
1261 * ip6_tnl_change - update the tunnel parameters
1262 * @t: tunnel to be changed
1263 * @p: tunnel configuration parameters
1265 * Description:
1266 * ip6_tnl_change() updates the tunnel parameters
1269 static int
1270 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1272 t->parms.laddr = p->laddr;
1273 t->parms.raddr = p->raddr;
1274 t->parms.flags = p->flags;
1275 t->parms.hop_limit = p->hop_limit;
1276 t->parms.encap_limit = p->encap_limit;
1277 t->parms.flowinfo = p->flowinfo;
1278 t->parms.link = p->link;
1279 t->parms.proto = p->proto;
1280 dst_cache_reset(&t->dst_cache);
1281 ip6_tnl_link_config(t);
1282 return 0;
1285 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1287 struct net *net = t->net;
1288 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1289 int err;
1291 ip6_tnl_unlink(ip6n, t);
1292 synchronize_net();
1293 err = ip6_tnl_change(t, p);
1294 ip6_tnl_link(ip6n, t);
1295 netdev_state_change(t->dev);
1296 return err;
1299 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1301 /* for default tnl0 device allow to change only the proto */
1302 t->parms.proto = p->proto;
1303 netdev_state_change(t->dev);
1304 return 0;
1307 static void
1308 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1310 p->laddr = u->laddr;
1311 p->raddr = u->raddr;
1312 p->flags = u->flags;
1313 p->hop_limit = u->hop_limit;
1314 p->encap_limit = u->encap_limit;
1315 p->flowinfo = u->flowinfo;
1316 p->link = u->link;
1317 p->proto = u->proto;
1318 memcpy(p->name, u->name, sizeof(u->name));
1321 static void
1322 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1324 u->laddr = p->laddr;
1325 u->raddr = p->raddr;
1326 u->flags = p->flags;
1327 u->hop_limit = p->hop_limit;
1328 u->encap_limit = p->encap_limit;
1329 u->flowinfo = p->flowinfo;
1330 u->link = p->link;
1331 u->proto = p->proto;
1332 memcpy(u->name, p->name, sizeof(u->name));
1336 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1337 * @dev: virtual device associated with tunnel
1338 * @ifr: parameters passed from userspace
1339 * @cmd: command to be performed
1341 * Description:
1342 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1343 * from userspace.
1345 * The possible commands are the following:
1346 * %SIOCGETTUNNEL: get tunnel parameters for device
1347 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1348 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1349 * %SIOCDELTUNNEL: delete tunnel
1351 * The fallback device "ip6tnl0", created during module
1352 * initialization, can be used for creating other tunnel devices.
1354 * Return:
1355 * 0 on success,
1356 * %-EFAULT if unable to copy data to or from userspace,
1357 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1358 * %-EINVAL if passed tunnel parameters are invalid,
1359 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1360 * %-ENODEV if attempting to change or delete a nonexisting device
1363 static int
1364 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1366 int err = 0;
1367 struct ip6_tnl_parm p;
1368 struct __ip6_tnl_parm p1;
1369 struct ip6_tnl *t = netdev_priv(dev);
1370 struct net *net = t->net;
1371 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1373 switch (cmd) {
1374 case SIOCGETTUNNEL:
1375 if (dev == ip6n->fb_tnl_dev) {
1376 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1377 err = -EFAULT;
1378 break;
1380 ip6_tnl_parm_from_user(&p1, &p);
1381 t = ip6_tnl_locate(net, &p1, 0);
1382 if (IS_ERR(t))
1383 t = netdev_priv(dev);
1384 } else {
1385 memset(&p, 0, sizeof(p));
1387 ip6_tnl_parm_to_user(&p, &t->parms);
1388 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1389 err = -EFAULT;
1391 break;
1392 case SIOCADDTUNNEL:
1393 case SIOCCHGTUNNEL:
1394 err = -EPERM;
1395 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1396 break;
1397 err = -EFAULT;
1398 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1399 break;
1400 err = -EINVAL;
1401 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1402 p.proto != 0)
1403 break;
1404 ip6_tnl_parm_from_user(&p1, &p);
1405 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1406 if (cmd == SIOCCHGTUNNEL) {
1407 if (!IS_ERR(t)) {
1408 if (t->dev != dev) {
1409 err = -EEXIST;
1410 break;
1412 } else
1413 t = netdev_priv(dev);
1414 if (dev == ip6n->fb_tnl_dev)
1415 err = ip6_tnl0_update(t, &p1);
1416 else
1417 err = ip6_tnl_update(t, &p1);
1419 if (!IS_ERR(t)) {
1420 err = 0;
1421 ip6_tnl_parm_to_user(&p, &t->parms);
1422 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1423 err = -EFAULT;
1425 } else {
1426 err = PTR_ERR(t);
1428 break;
1429 case SIOCDELTUNNEL:
1430 err = -EPERM;
1431 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1432 break;
1434 if (dev == ip6n->fb_tnl_dev) {
1435 err = -EFAULT;
1436 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1437 break;
1438 err = -ENOENT;
1439 ip6_tnl_parm_from_user(&p1, &p);
1440 t = ip6_tnl_locate(net, &p1, 0);
1441 if (IS_ERR(t))
1442 break;
1443 err = -EPERM;
1444 if (t->dev == ip6n->fb_tnl_dev)
1445 break;
1446 dev = t->dev;
1448 err = 0;
1449 unregister_netdevice(dev);
1450 break;
1451 default:
1452 err = -EINVAL;
1454 return err;
1458 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1459 * @dev: virtual device associated with tunnel
1460 * @new_mtu: the new mtu
1462 * Return:
1463 * 0 on success,
1464 * %-EINVAL if mtu too small
1467 static int
1468 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1470 struct ip6_tnl *tnl = netdev_priv(dev);
1472 if (tnl->parms.proto == IPPROTO_IPIP) {
1473 if (new_mtu < 68)
1474 return -EINVAL;
1475 } else {
1476 if (new_mtu < IPV6_MIN_MTU)
1477 return -EINVAL;
1479 if (new_mtu > 0xFFF8 - dev->hard_header_len)
1480 return -EINVAL;
1481 dev->mtu = new_mtu;
1482 return 0;
1485 int ip6_tnl_get_iflink(const struct net_device *dev)
1487 struct ip6_tnl *t = netdev_priv(dev);
1489 return t->parms.link;
1491 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1493 static const struct net_device_ops ip6_tnl_netdev_ops = {
1494 .ndo_init = ip6_tnl_dev_init,
1495 .ndo_uninit = ip6_tnl_dev_uninit,
1496 .ndo_start_xmit = ip6_tnl_xmit,
1497 .ndo_do_ioctl = ip6_tnl_ioctl,
1498 .ndo_change_mtu = ip6_tnl_change_mtu,
1499 .ndo_get_stats = ip6_get_stats,
1500 .ndo_get_iflink = ip6_tnl_get_iflink,
1505 * ip6_tnl_dev_setup - setup virtual tunnel device
1506 * @dev: virtual device associated with tunnel
1508 * Description:
1509 * Initialize function pointers and device parameters
1512 static void ip6_tnl_dev_setup(struct net_device *dev)
1514 struct ip6_tnl *t;
1516 dev->netdev_ops = &ip6_tnl_netdev_ops;
1517 dev->destructor = ip6_dev_free;
1519 dev->type = ARPHRD_TUNNEL6;
1520 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
1521 dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr);
1522 t = netdev_priv(dev);
1523 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1524 dev->mtu -= 8;
1525 dev->flags |= IFF_NOARP;
1526 dev->addr_len = sizeof(struct in6_addr);
1527 netif_keep_dst(dev);
1528 /* This perm addr will be used as interface identifier by IPv6 */
1529 dev->addr_assign_type = NET_ADDR_RANDOM;
1530 eth_random_addr(dev->perm_addr);
1535 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1536 * @dev: virtual device associated with tunnel
1539 static inline int
1540 ip6_tnl_dev_init_gen(struct net_device *dev)
1542 struct ip6_tnl *t = netdev_priv(dev);
1543 int ret;
1545 t->dev = dev;
1546 t->net = dev_net(dev);
1547 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1548 if (!dev->tstats)
1549 return -ENOMEM;
1551 ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1552 if (ret) {
1553 free_percpu(dev->tstats);
1554 dev->tstats = NULL;
1555 return ret;
1558 return 0;
1562 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1563 * @dev: virtual device associated with tunnel
1566 static int ip6_tnl_dev_init(struct net_device *dev)
1568 struct ip6_tnl *t = netdev_priv(dev);
1569 int err = ip6_tnl_dev_init_gen(dev);
1571 if (err)
1572 return err;
1573 ip6_tnl_link_config(t);
1574 return 0;
1578 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1579 * @dev: fallback device
1581 * Return: 0
1584 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1586 struct ip6_tnl *t = netdev_priv(dev);
1587 struct net *net = dev_net(dev);
1588 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1590 t->parms.proto = IPPROTO_IPV6;
1591 dev_hold(dev);
1593 rcu_assign_pointer(ip6n->tnls_wc[0], t);
1594 return 0;
1597 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1599 u8 proto;
1601 if (!data || !data[IFLA_IPTUN_PROTO])
1602 return 0;
1604 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1605 if (proto != IPPROTO_IPV6 &&
1606 proto != IPPROTO_IPIP &&
1607 proto != 0)
1608 return -EINVAL;
1610 return 0;
1613 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1614 struct __ip6_tnl_parm *parms)
1616 memset(parms, 0, sizeof(*parms));
1618 if (!data)
1619 return;
1621 if (data[IFLA_IPTUN_LINK])
1622 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1624 if (data[IFLA_IPTUN_LOCAL])
1625 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1627 if (data[IFLA_IPTUN_REMOTE])
1628 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1630 if (data[IFLA_IPTUN_TTL])
1631 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1633 if (data[IFLA_IPTUN_ENCAP_LIMIT])
1634 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1636 if (data[IFLA_IPTUN_FLOWINFO])
1637 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1639 if (data[IFLA_IPTUN_FLAGS])
1640 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1642 if (data[IFLA_IPTUN_PROTO])
1643 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1646 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1647 struct nlattr *tb[], struct nlattr *data[])
1649 struct net *net = dev_net(dev);
1650 struct ip6_tnl *nt, *t;
1652 nt = netdev_priv(dev);
1653 ip6_tnl_netlink_parms(data, &nt->parms);
1655 t = ip6_tnl_locate(net, &nt->parms, 0);
1656 if (!IS_ERR(t))
1657 return -EEXIST;
1659 return ip6_tnl_create2(dev);
1662 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1663 struct nlattr *data[])
1665 struct ip6_tnl *t = netdev_priv(dev);
1666 struct __ip6_tnl_parm p;
1667 struct net *net = t->net;
1668 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1670 if (dev == ip6n->fb_tnl_dev)
1671 return -EINVAL;
1673 ip6_tnl_netlink_parms(data, &p);
1675 t = ip6_tnl_locate(net, &p, 0);
1676 if (!IS_ERR(t)) {
1677 if (t->dev != dev)
1678 return -EEXIST;
1679 } else
1680 t = netdev_priv(dev);
1682 return ip6_tnl_update(t, &p);
1685 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
1687 struct net *net = dev_net(dev);
1688 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1690 if (dev != ip6n->fb_tnl_dev)
1691 unregister_netdevice_queue(dev, head);
1694 static size_t ip6_tnl_get_size(const struct net_device *dev)
1696 return
1697 /* IFLA_IPTUN_LINK */
1698 nla_total_size(4) +
1699 /* IFLA_IPTUN_LOCAL */
1700 nla_total_size(sizeof(struct in6_addr)) +
1701 /* IFLA_IPTUN_REMOTE */
1702 nla_total_size(sizeof(struct in6_addr)) +
1703 /* IFLA_IPTUN_TTL */
1704 nla_total_size(1) +
1705 /* IFLA_IPTUN_ENCAP_LIMIT */
1706 nla_total_size(1) +
1707 /* IFLA_IPTUN_FLOWINFO */
1708 nla_total_size(4) +
1709 /* IFLA_IPTUN_FLAGS */
1710 nla_total_size(4) +
1711 /* IFLA_IPTUN_PROTO */
1712 nla_total_size(1) +
1716 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
1718 struct ip6_tnl *tunnel = netdev_priv(dev);
1719 struct __ip6_tnl_parm *parm = &tunnel->parms;
1721 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1722 nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
1723 nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
1724 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
1725 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
1726 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
1727 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
1728 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
1729 goto nla_put_failure;
1730 return 0;
1732 nla_put_failure:
1733 return -EMSGSIZE;
1736 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
1738 struct ip6_tnl *tunnel = netdev_priv(dev);
1740 return tunnel->net;
1742 EXPORT_SYMBOL(ip6_tnl_get_link_net);
1744 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
1745 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1746 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) },
1747 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) },
1748 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1749 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 },
1750 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 },
1751 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 },
1752 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1755 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
1756 .kind = "ip6tnl",
1757 .maxtype = IFLA_IPTUN_MAX,
1758 .policy = ip6_tnl_policy,
1759 .priv_size = sizeof(struct ip6_tnl),
1760 .setup = ip6_tnl_dev_setup,
1761 .validate = ip6_tnl_validate,
1762 .newlink = ip6_tnl_newlink,
1763 .changelink = ip6_tnl_changelink,
1764 .dellink = ip6_tnl_dellink,
1765 .get_size = ip6_tnl_get_size,
1766 .fill_info = ip6_tnl_fill_info,
1767 .get_link_net = ip6_tnl_get_link_net,
1770 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1771 .handler = ip4ip6_rcv,
1772 .err_handler = ip4ip6_err,
1773 .priority = 1,
1776 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1777 .handler = ip6ip6_rcv,
1778 .err_handler = ip6ip6_err,
1779 .priority = 1,
1782 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
1784 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1785 struct net_device *dev, *aux;
1786 int h;
1787 struct ip6_tnl *t;
1788 LIST_HEAD(list);
1790 for_each_netdev_safe(net, dev, aux)
1791 if (dev->rtnl_link_ops == &ip6_link_ops)
1792 unregister_netdevice_queue(dev, &list);
1794 for (h = 0; h < HASH_SIZE; h++) {
1795 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1796 while (t) {
1797 /* If dev is in the same netns, it has already
1798 * been added to the list by the previous loop.
1800 if (!net_eq(dev_net(t->dev), net))
1801 unregister_netdevice_queue(t->dev, &list);
1802 t = rtnl_dereference(t->next);
1806 unregister_netdevice_many(&list);
1809 static int __net_init ip6_tnl_init_net(struct net *net)
1811 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1812 struct ip6_tnl *t = NULL;
1813 int err;
1815 ip6n->tnls[0] = ip6n->tnls_wc;
1816 ip6n->tnls[1] = ip6n->tnls_r_l;
1818 err = -ENOMEM;
1819 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1820 NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
1822 if (!ip6n->fb_tnl_dev)
1823 goto err_alloc_dev;
1824 dev_net_set(ip6n->fb_tnl_dev, net);
1825 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
1826 /* FB netdevice is special: we have one, and only one per netns.
1827 * Allowing to move it to another netns is clearly unsafe.
1829 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
1831 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1832 if (err < 0)
1833 goto err_register;
1835 err = register_netdev(ip6n->fb_tnl_dev);
1836 if (err < 0)
1837 goto err_register;
1839 t = netdev_priv(ip6n->fb_tnl_dev);
1841 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1842 return 0;
1844 err_register:
1845 ip6_dev_free(ip6n->fb_tnl_dev);
1846 err_alloc_dev:
1847 return err;
1850 static void __net_exit ip6_tnl_exit_net(struct net *net)
1852 rtnl_lock();
1853 ip6_tnl_destroy_tunnels(net);
1854 rtnl_unlock();
1857 static struct pernet_operations ip6_tnl_net_ops = {
1858 .init = ip6_tnl_init_net,
1859 .exit = ip6_tnl_exit_net,
1860 .id = &ip6_tnl_net_id,
1861 .size = sizeof(struct ip6_tnl_net),
1865 * ip6_tunnel_init - register protocol and reserve needed resources
1867 * Return: 0 on success
1870 static int __init ip6_tunnel_init(void)
1872 int err;
1874 err = register_pernet_device(&ip6_tnl_net_ops);
1875 if (err < 0)
1876 goto out_pernet;
1878 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1879 if (err < 0) {
1880 pr_err("%s: can't register ip4ip6\n", __func__);
1881 goto out_ip4ip6;
1884 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1885 if (err < 0) {
1886 pr_err("%s: can't register ip6ip6\n", __func__);
1887 goto out_ip6ip6;
1889 err = rtnl_link_register(&ip6_link_ops);
1890 if (err < 0)
1891 goto rtnl_link_failed;
1893 return 0;
1895 rtnl_link_failed:
1896 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1897 out_ip6ip6:
1898 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1899 out_ip4ip6:
1900 unregister_pernet_device(&ip6_tnl_net_ops);
1901 out_pernet:
1902 return err;
1906 * ip6_tunnel_cleanup - free resources and unregister protocol
1909 static void __exit ip6_tunnel_cleanup(void)
1911 rtnl_link_unregister(&ip6_link_ops);
1912 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1913 pr_info("%s: can't deregister ip4ip6\n", __func__);
1915 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1916 pr_info("%s: can't deregister ip6ip6\n", __func__);
1918 unregister_pernet_device(&ip6_tnl_net_ops);
1921 module_init(ip6_tunnel_init);
1922 module_exit(ip6_tunnel_cleanup);