2 * IPv6 tunneling device
3 * Linux INET6 implementation
6 * Ville Nuorvala <vnuorval@tcs.hut.fi>
7 * Yasuyuki Kozakai <kozakai@linux-ipv6.org>
10 * linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
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>
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>
44 #include <linux/etherdevice.h>
46 #include <asm/uaccess.h>
47 #include <linux/atomic.h>
51 #include <net/ip_tunnels.h>
53 #include <net/ip6_route.h>
54 #include <net/addrconf.h>
55 #include <net/ip6_tunnel.h>
57 #include <net/dsfield.h>
58 #include <net/inet_ecn.h>
59 #include <net/net_namespace.h>
60 #include <net/netns/generic.h>
62 MODULE_AUTHOR("Ville Nuorvala");
63 MODULE_DESCRIPTION("IPv6 tunneling device");
64 MODULE_LICENSE("GPL");
65 MODULE_ALIAS_NETDEV("ip6tnl0");
68 #define IP6_TNL_TRACE(x...) pr_debug("%s:" x "\n", __func__)
70 #define IP6_TNL_TRACE(x...) do {;} while(0)
73 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
74 #define IPV6_TCLASS_SHIFT 20
76 #define HASH_SIZE_SHIFT 5
77 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
79 static bool log_ecn_error = true;
80 module_param(log_ecn_error, bool, 0644);
81 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
83 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
85 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
87 return hash_32(hash, HASH_SIZE_SHIFT);
90 static int ip6_tnl_dev_init(struct net_device *dev);
91 static void ip6_tnl_dev_setup(struct net_device *dev);
92 static struct rtnl_link_ops ip6_link_ops __read_mostly;
94 static int ip6_tnl_net_id __read_mostly;
96 /* the IPv6 tunnel fallback device */
97 struct net_device *fb_tnl_dev;
98 /* lists for storing tunnels in use */
99 struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
100 struct ip6_tnl __rcu *tnls_wc[1];
101 struct ip6_tnl __rcu **tnls[2];
104 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
106 struct pcpu_tstats sum = { 0 };
109 for_each_possible_cpu(i) {
110 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
112 sum.rx_packets += tstats->rx_packets;
113 sum.rx_bytes += tstats->rx_bytes;
114 sum.tx_packets += tstats->tx_packets;
115 sum.tx_bytes += tstats->tx_bytes;
117 dev->stats.rx_packets = sum.rx_packets;
118 dev->stats.rx_bytes = sum.rx_bytes;
119 dev->stats.tx_packets = sum.tx_packets;
120 dev->stats.tx_bytes = sum.tx_bytes;
125 * Locking : hash tables are protected by RCU and RTNL
128 struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
130 struct dst_entry *dst = t->dst_cache;
132 if (dst && dst->obsolete &&
133 dst->ops->check(dst, t->dst_cookie) == NULL) {
141 EXPORT_SYMBOL_GPL(ip6_tnl_dst_check);
143 void ip6_tnl_dst_reset(struct ip6_tnl *t)
145 dst_release(t->dst_cache);
148 EXPORT_SYMBOL_GPL(ip6_tnl_dst_reset);
150 void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
152 struct rt6_info *rt = (struct rt6_info *) dst;
153 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
154 dst_release(t->dst_cache);
157 EXPORT_SYMBOL_GPL(ip6_tnl_dst_store);
160 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
161 * @remote: the address of the tunnel exit-point
162 * @local: the address of the tunnel entry-point
165 * tunnel matching given end-points if found,
166 * else fallback tunnel if its device is up,
170 #define for_each_ip6_tunnel_rcu(start) \
171 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
173 static struct ip6_tnl *
174 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
176 unsigned int hash = HASH(remote, local);
178 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
180 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
181 if (ipv6_addr_equal(local, &t->parms.laddr) &&
182 ipv6_addr_equal(remote, &t->parms.raddr) &&
183 (t->dev->flags & IFF_UP))
186 t = rcu_dereference(ip6n->tnls_wc[0]);
187 if (t && (t->dev->flags & IFF_UP))
194 * ip6_tnl_bucket - get head of list matching given tunnel parameters
195 * @p: parameters containing tunnel end-points
198 * ip6_tnl_bucket() returns the head of the list matching the
199 * &struct in6_addr entries laddr and raddr in @p.
201 * Return: head of IPv6 tunnel list
204 static struct ip6_tnl __rcu **
205 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
207 const struct in6_addr *remote = &p->raddr;
208 const struct in6_addr *local = &p->laddr;
212 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
214 h = HASH(remote, local);
216 return &ip6n->tnls[prio][h];
220 * ip6_tnl_link - add tunnel to hash table
221 * @t: tunnel to be added
225 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
227 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
229 rcu_assign_pointer(t->next , rtnl_dereference(*tp));
230 rcu_assign_pointer(*tp, t);
234 * ip6_tnl_unlink - remove tunnel from hash table
235 * @t: tunnel to be removed
239 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
241 struct ip6_tnl __rcu **tp;
242 struct ip6_tnl *iter;
244 for (tp = ip6_tnl_bucket(ip6n, &t->parms);
245 (iter = rtnl_dereference(*tp)) != NULL;
248 rcu_assign_pointer(*tp, t->next);
254 static void ip6_dev_free(struct net_device *dev)
256 free_percpu(dev->tstats);
260 static int ip6_tnl_create2(struct net_device *dev)
262 struct ip6_tnl *t = netdev_priv(dev);
263 struct net *net = dev_net(dev);
264 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
267 t = netdev_priv(dev);
268 err = ip6_tnl_dev_init(dev);
272 err = register_netdevice(dev);
276 strcpy(t->parms.name, dev->name);
277 dev->rtnl_link_ops = &ip6_link_ops;
280 ip6_tnl_link(ip6n, t);
288 * ip6_tnl_create - create a new tunnel
289 * @p: tunnel parameters
290 * @pt: pointer to new tunnel
293 * Create tunnel matching given parameters.
296 * created tunnel or NULL
299 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
301 struct net_device *dev;
307 strlcpy(name, p->name, IFNAMSIZ);
309 sprintf(name, "ip6tnl%%d");
311 dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
315 dev_net_set(dev, net);
317 t = netdev_priv(dev);
319 t->net = dev_net(dev);
320 err = ip6_tnl_create2(dev);
333 * ip6_tnl_locate - find or create tunnel matching given parameters
334 * @p: tunnel parameters
335 * @create: != 0 if allowed to create new tunnel if no match found
338 * ip6_tnl_locate() first tries to locate an existing tunnel
339 * based on @parms. If this is unsuccessful, but @create is set a new
340 * tunnel device is created and registered for use.
343 * matching tunnel or NULL
346 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
347 struct __ip6_tnl_parm *p, int create)
349 const struct in6_addr *remote = &p->raddr;
350 const struct in6_addr *local = &p->laddr;
351 struct ip6_tnl __rcu **tp;
353 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
355 for (tp = ip6_tnl_bucket(ip6n, p);
356 (t = rtnl_dereference(*tp)) != NULL;
358 if (ipv6_addr_equal(local, &t->parms.laddr) &&
359 ipv6_addr_equal(remote, &t->parms.raddr))
364 return ip6_tnl_create(net, p);
368 * ip6_tnl_dev_uninit - tunnel device uninitializer
369 * @dev: the device to be destroyed
372 * ip6_tnl_dev_uninit() removes tunnel from its list
376 ip6_tnl_dev_uninit(struct net_device *dev)
378 struct ip6_tnl *t = netdev_priv(dev);
379 struct net *net = t->net;
380 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
382 if (dev == ip6n->fb_tnl_dev)
383 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
385 ip6_tnl_unlink(ip6n, t);
386 ip6_tnl_dst_reset(t);
391 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
392 * @skb: received socket buffer
395 * 0 if none was found,
396 * else index to encapsulation limit
399 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
401 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
402 __u8 nexthdr = ipv6h->nexthdr;
403 __u16 off = sizeof (*ipv6h);
405 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
407 struct ipv6_opt_hdr *hdr;
408 if (raw + off + sizeof (*hdr) > skb->data &&
409 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
412 hdr = (struct ipv6_opt_hdr *) (raw + off);
413 if (nexthdr == NEXTHDR_FRAGMENT) {
414 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
415 if (frag_hdr->frag_off)
418 } else if (nexthdr == NEXTHDR_AUTH) {
419 optlen = (hdr->hdrlen + 2) << 2;
421 optlen = ipv6_optlen(hdr);
423 if (nexthdr == NEXTHDR_DEST) {
426 struct ipv6_tlv_tnl_enc_lim *tel;
428 /* No more room for encapsulation limit */
429 if (i + sizeof (*tel) > off + optlen)
432 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
433 /* return index of option if found and valid */
434 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
437 /* else jump to next option */
439 i += tel->length + 2;
444 nexthdr = hdr->nexthdr;
449 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
452 * ip6_tnl_err - tunnel error handler
455 * ip6_tnl_err() should handle errors in the tunnel according
456 * to the specifications in RFC 2473.
460 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
461 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
463 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
466 u8 rel_type = ICMPV6_DEST_UNREACH;
467 u8 rel_code = ICMPV6_ADDR_UNREACH;
472 /* If the packet doesn't contain the original IPv6 header we are
473 in trouble since we might need the source address for further
474 processing of the error. */
477 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
478 &ipv6h->saddr)) == NULL)
481 if (t->parms.proto != ipproto && t->parms.proto != 0)
488 struct ipv6_tlv_tnl_enc_lim *tel;
490 case ICMPV6_DEST_UNREACH:
491 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
495 case ICMPV6_TIME_EXCEED:
496 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
497 net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
502 case ICMPV6_PARAMPROB:
504 if ((*code) == ICMPV6_HDR_FIELD)
505 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
507 if (teli && teli == *info - 2) {
508 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
509 if (tel->encap_limit == 0) {
510 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
515 net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
519 case ICMPV6_PKT_TOOBIG:
520 mtu = *info - offset;
521 if (mtu < IPV6_MIN_MTU)
525 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
526 rel_type = ICMPV6_PKT_TOOBIG;
545 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
546 u8 type, u8 code, int offset, __be32 info)
551 __u32 rel_info = ntohl(info);
553 struct sk_buff *skb2;
554 const struct iphdr *eiph;
558 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
559 &rel_msg, &rel_info, offset);
567 case ICMPV6_DEST_UNREACH:
568 if (rel_code != ICMPV6_ADDR_UNREACH)
570 rel_type = ICMP_DEST_UNREACH;
571 rel_code = ICMP_HOST_UNREACH;
573 case ICMPV6_PKT_TOOBIG:
576 rel_type = ICMP_DEST_UNREACH;
577 rel_code = ICMP_FRAG_NEEDED;
580 rel_type = ICMP_REDIRECT;
581 rel_code = ICMP_REDIR_HOST;
586 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
589 skb2 = skb_clone(skb, GFP_ATOMIC);
595 skb_pull(skb2, offset);
596 skb_reset_network_header(skb2);
599 /* Try to guess incoming interface */
600 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
603 IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
607 skb2->dev = rt->dst.dev;
609 /* route "incoming" packet */
610 if (rt->rt_flags & RTCF_LOCAL) {
613 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
614 eiph->daddr, eiph->saddr,
617 RT_TOS(eiph->tos), 0);
619 rt->dst.dev->type != ARPHRD_TUNNEL) {
624 skb_dst_set(skb2, &rt->dst);
627 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
629 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
633 /* change mtu on this route */
634 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
635 if (rel_info > dst_mtu(skb_dst(skb2)))
638 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
640 if (rel_type == ICMP_REDIRECT)
641 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
643 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
651 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
652 u8 type, u8 code, int offset, __be32 info)
657 __u32 rel_info = ntohl(info);
660 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
661 &rel_msg, &rel_info, offset);
665 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
667 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
673 skb_pull(skb2, offset);
674 skb_reset_network_header(skb2);
676 /* Try to guess incoming interface */
677 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
680 if (rt && rt->dst.dev)
681 skb2->dev = rt->dst.dev;
683 icmpv6_send(skb2, rel_type, rel_code, rel_info);
693 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
694 const struct ipv6hdr *ipv6h,
697 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
699 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
700 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
702 return IP6_ECN_decapsulate(ipv6h, skb);
705 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
706 const struct ipv6hdr *ipv6h,
709 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
710 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
712 return IP6_ECN_decapsulate(ipv6h, skb);
715 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
716 const struct in6_addr *laddr,
717 const struct in6_addr *raddr)
719 struct __ip6_tnl_parm *p = &t->parms;
720 int ltype = ipv6_addr_type(laddr);
721 int rtype = ipv6_addr_type(raddr);
724 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
725 flags = IP6_TNL_F_CAP_PER_PACKET;
726 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
727 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
728 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
729 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
730 if (ltype&IPV6_ADDR_UNICAST)
731 flags |= IP6_TNL_F_CAP_XMIT;
732 if (rtype&IPV6_ADDR_UNICAST)
733 flags |= IP6_TNL_F_CAP_RCV;
737 EXPORT_SYMBOL(ip6_tnl_get_cap);
739 /* called with rcu_read_lock() */
740 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
741 const struct in6_addr *laddr,
742 const struct in6_addr *raddr)
744 struct __ip6_tnl_parm *p = &t->parms;
746 struct net *net = t->net;
748 if ((p->flags & IP6_TNL_F_CAP_RCV) ||
749 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
750 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
751 struct net_device *ldev = NULL;
754 ldev = dev_get_by_index_rcu(net, p->link);
756 if ((ipv6_addr_is_multicast(laddr) ||
757 likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
758 likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
763 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
766 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
767 * @skb: received socket buffer
768 * @protocol: ethernet protocol ID
769 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
774 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
776 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
777 const struct ipv6hdr *ipv6h,
778 struct sk_buff *skb))
781 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
786 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
787 &ipv6h->daddr)) != NULL) {
788 struct pcpu_tstats *tstats;
790 if (t->parms.proto != ipproto && t->parms.proto != 0) {
795 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
800 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
801 t->dev->stats.rx_dropped++;
805 skb->mac_header = skb->network_header;
806 skb_reset_network_header(skb);
807 skb->protocol = htons(protocol);
808 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
810 __skb_tunnel_rx(skb, t->dev, t->net);
812 err = dscp_ecn_decapsulate(t, ipv6h, skb);
815 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
817 ipv6_get_dsfield(ipv6h));
819 ++t->dev->stats.rx_frame_errors;
820 ++t->dev->stats.rx_errors;
826 tstats = this_cpu_ptr(t->dev->tstats);
827 tstats->rx_packets++;
828 tstats->rx_bytes += skb->len;
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;
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;
866 opt->dst_opt[4] = encap_limit;
867 opt->dst_opt[5] = IPV6_TLV_PADN;
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
880 * Avoid trivial tunneling loop by checking that tunnel exit-point
881 * doesn't match source of incoming packet.
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;
898 struct net *net = t->net;
900 if (p->flags & IP6_TNL_F_CAP_XMIT) {
901 struct net_device *ldev = NULL;
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",
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",
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
932 * Build new header and do some sanity checks on the packet before sending
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,
948 struct ip6_tnl *t = netdev_priv(dev);
949 struct net *net = t->net;
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;
956 unsigned int max_headroom = sizeof(struct ipv6hdr);
960 if (!fl6->flowi6_mark)
961 dst = ip6_tnl_dst_check(t);
963 ndst = ip6_route_output(net, NULL, fl6);
966 goto tx_err_link_failure;
967 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
971 goto tx_err_link_failure;
980 net_warn_ratelimited("%s: Local routing loop detected!\n",
982 goto tx_err_dst_release;
984 mtu = dst_mtu(dst) - sizeof (*ipv6h);
985 if (encap_limit >= 0) {
989 if (mtu < IPV6_MIN_MTU)
992 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
993 if (skb->len > mtu) {
996 goto tx_err_dst_release;
999 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1002 * Okay, now see if we can stuff it in the buffer as-is.
1004 max_headroom += LL_RESERVED_SPACE(tdev);
1006 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1007 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1008 struct sk_buff *new_skb;
1010 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
1011 goto tx_err_dst_release;
1014 skb_set_owner_w(new_skb, skb->sk);
1018 if (fl6->flowi6_mark) {
1019 skb_dst_set(skb, dst);
1022 skb_dst_set_noref(skb, dst);
1024 skb->transport_header = skb->network_header;
1026 proto = fl6->flowi6_proto;
1027 if (encap_limit >= 0) {
1028 init_tel_txopt(&opt, encap_limit);
1029 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1031 skb_push(skb, sizeof(struct ipv6hdr));
1032 skb_reset_network_header(skb);
1033 ipv6h = ipv6_hdr(skb);
1034 ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield), fl6->flowlabel);
1035 ipv6h->hop_limit = t->parms.hop_limit;
1036 ipv6h->nexthdr = proto;
1037 ipv6h->saddr = fl6->saddr;
1038 ipv6h->daddr = fl6->daddr;
1039 ip6tunnel_xmit(skb, dev);
1041 ip6_tnl_dst_store(t, ndst);
1043 tx_err_link_failure:
1044 stats->tx_carrier_errors++;
1045 dst_link_failure(skb);
1052 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1054 struct ip6_tnl *t = netdev_priv(dev);
1055 const struct iphdr *iph = ip_hdr(skb);
1056 int encap_limit = -1;
1062 if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
1063 !ip6_tnl_xmit_ctl(t))
1066 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1067 encap_limit = t->parms.encap_limit;
1069 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1070 fl6.flowi6_proto = IPPROTO_IPIP;
1072 dsfield = ipv4_get_dsfield(iph);
1074 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1075 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1077 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1078 fl6.flowi6_mark = skb->mark;
1080 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1082 /* XXX: send ICMP error even if DF is not set. */
1083 if (err == -EMSGSIZE)
1084 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1093 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1095 struct ip6_tnl *t = netdev_priv(dev);
1096 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1097 int encap_limit = -1;
1104 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1105 !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1108 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1110 struct ipv6_tlv_tnl_enc_lim *tel;
1111 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1112 if (tel->encap_limit == 0) {
1113 icmpv6_send(skb, ICMPV6_PARAMPROB,
1114 ICMPV6_HDR_FIELD, offset + 2);
1117 encap_limit = tel->encap_limit - 1;
1118 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1119 encap_limit = t->parms.encap_limit;
1121 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1122 fl6.flowi6_proto = IPPROTO_IPV6;
1124 dsfield = ipv6_get_dsfield(ipv6h);
1125 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1126 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1127 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1128 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1129 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1130 fl6.flowi6_mark = skb->mark;
1132 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1134 if (err == -EMSGSIZE)
1135 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1143 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1145 struct ip6_tnl *t = netdev_priv(dev);
1146 struct net_device_stats *stats = &t->dev->stats;
1149 switch (skb->protocol) {
1150 case htons(ETH_P_IP):
1151 ret = ip4ip6_tnl_xmit(skb, dev);
1153 case htons(ETH_P_IPV6):
1154 ret = ip6ip6_tnl_xmit(skb, dev);
1163 return NETDEV_TX_OK;
1167 stats->tx_dropped++;
1169 return NETDEV_TX_OK;
1172 static void ip6_tnl_link_config(struct ip6_tnl *t)
1174 struct net_device *dev = t->dev;
1175 struct __ip6_tnl_parm *p = &t->parms;
1176 struct flowi6 *fl6 = &t->fl.u.ip6;
1178 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1179 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1181 /* Set up flowi template */
1182 fl6->saddr = p->laddr;
1183 fl6->daddr = p->raddr;
1184 fl6->flowi6_oif = p->link;
1187 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1188 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1189 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1190 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1192 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1193 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1195 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1196 dev->flags |= IFF_POINTOPOINT;
1198 dev->flags &= ~IFF_POINTOPOINT;
1200 dev->iflink = p->link;
1202 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1203 int strict = (ipv6_addr_type(&p->raddr) &
1204 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1206 struct rt6_info *rt = rt6_lookup(t->net,
1207 &p->raddr, &p->laddr,
1214 dev->hard_header_len = rt->dst.dev->hard_header_len +
1215 sizeof (struct ipv6hdr);
1217 dev->mtu = rt->dst.dev->mtu - sizeof (struct ipv6hdr);
1218 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1221 if (dev->mtu < IPV6_MIN_MTU)
1222 dev->mtu = IPV6_MIN_MTU;
1229 * ip6_tnl_change - update the tunnel parameters
1230 * @t: tunnel to be changed
1231 * @p: tunnel configuration parameters
1234 * ip6_tnl_change() updates the tunnel parameters
1238 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1240 t->parms.laddr = p->laddr;
1241 t->parms.raddr = p->raddr;
1242 t->parms.flags = p->flags;
1243 t->parms.hop_limit = p->hop_limit;
1244 t->parms.encap_limit = p->encap_limit;
1245 t->parms.flowinfo = p->flowinfo;
1246 t->parms.link = p->link;
1247 t->parms.proto = p->proto;
1248 ip6_tnl_dst_reset(t);
1249 ip6_tnl_link_config(t);
1253 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1255 struct net *net = t->net;
1256 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1259 ip6_tnl_unlink(ip6n, t);
1261 err = ip6_tnl_change(t, p);
1262 ip6_tnl_link(ip6n, t);
1263 netdev_state_change(t->dev);
1268 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1270 p->laddr = u->laddr;
1271 p->raddr = u->raddr;
1272 p->flags = u->flags;
1273 p->hop_limit = u->hop_limit;
1274 p->encap_limit = u->encap_limit;
1275 p->flowinfo = u->flowinfo;
1277 p->proto = u->proto;
1278 memcpy(p->name, u->name, sizeof(u->name));
1282 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1284 u->laddr = p->laddr;
1285 u->raddr = p->raddr;
1286 u->flags = p->flags;
1287 u->hop_limit = p->hop_limit;
1288 u->encap_limit = p->encap_limit;
1289 u->flowinfo = p->flowinfo;
1291 u->proto = p->proto;
1292 memcpy(u->name, p->name, sizeof(u->name));
1296 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1297 * @dev: virtual device associated with tunnel
1298 * @ifr: parameters passed from userspace
1299 * @cmd: command to be performed
1302 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1305 * The possible commands are the following:
1306 * %SIOCGETTUNNEL: get tunnel parameters for device
1307 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1308 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1309 * %SIOCDELTUNNEL: delete tunnel
1311 * The fallback device "ip6tnl0", created during module
1312 * initialization, can be used for creating other tunnel devices.
1316 * %-EFAULT if unable to copy data to or from userspace,
1317 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1318 * %-EINVAL if passed tunnel parameters are invalid,
1319 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1320 * %-ENODEV if attempting to change or delete a nonexisting device
1324 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1327 struct ip6_tnl_parm p;
1328 struct __ip6_tnl_parm p1;
1329 struct ip6_tnl *t = NULL;
1330 struct net *net = dev_net(dev);
1331 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1335 if (dev == ip6n->fb_tnl_dev) {
1336 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1340 ip6_tnl_parm_from_user(&p1, &p);
1341 t = ip6_tnl_locate(net, &p1, 0);
1343 memset(&p, 0, sizeof(p));
1346 t = netdev_priv(dev);
1347 ip6_tnl_parm_to_user(&p, &t->parms);
1348 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1355 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1358 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1361 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1364 ip6_tnl_parm_from_user(&p1, &p);
1365 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1366 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1368 if (t->dev != dev) {
1373 t = netdev_priv(dev);
1375 err = ip6_tnl_update(t, &p1);
1379 ip6_tnl_parm_to_user(&p, &t->parms);
1380 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1384 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1388 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1391 if (dev == ip6n->fb_tnl_dev) {
1393 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1396 ip6_tnl_parm_from_user(&p1, &p);
1397 t = ip6_tnl_locate(net, &p1, 0);
1401 if (t->dev == ip6n->fb_tnl_dev)
1406 unregister_netdevice(dev);
1415 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1416 * @dev: virtual device associated with tunnel
1417 * @new_mtu: the new mtu
1421 * %-EINVAL if mtu too small
1425 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1427 if (new_mtu < IPV6_MIN_MTU) {
1435 static const struct net_device_ops ip6_tnl_netdev_ops = {
1436 .ndo_uninit = ip6_tnl_dev_uninit,
1437 .ndo_start_xmit = ip6_tnl_xmit,
1438 .ndo_do_ioctl = ip6_tnl_ioctl,
1439 .ndo_change_mtu = ip6_tnl_change_mtu,
1440 .ndo_get_stats = ip6_get_stats,
1445 * ip6_tnl_dev_setup - setup virtual tunnel device
1446 * @dev: virtual device associated with tunnel
1449 * Initialize function pointers and device parameters
1452 static void ip6_tnl_dev_setup(struct net_device *dev)
1456 dev->netdev_ops = &ip6_tnl_netdev_ops;
1457 dev->destructor = ip6_dev_free;
1459 dev->type = ARPHRD_TUNNEL6;
1460 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1461 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1462 t = netdev_priv(dev);
1463 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1465 dev->flags |= IFF_NOARP;
1466 dev->addr_len = sizeof(struct in6_addr);
1467 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1468 /* This perm addr will be used as interface identifier by IPv6 */
1469 dev->addr_assign_type = NET_ADDR_RANDOM;
1470 eth_random_addr(dev->perm_addr);
1475 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1476 * @dev: virtual device associated with tunnel
1480 ip6_tnl_dev_init_gen(struct net_device *dev)
1482 struct ip6_tnl *t = netdev_priv(dev);
1485 t->net = dev_net(dev);
1486 dev->tstats = alloc_percpu(struct pcpu_tstats);
1493 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1494 * @dev: virtual device associated with tunnel
1497 static int ip6_tnl_dev_init(struct net_device *dev)
1499 struct ip6_tnl *t = netdev_priv(dev);
1500 int err = ip6_tnl_dev_init_gen(dev);
1504 ip6_tnl_link_config(t);
1509 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1510 * @dev: fallback device
1515 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1517 struct ip6_tnl *t = netdev_priv(dev);
1518 struct net *net = dev_net(dev);
1519 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1520 int err = ip6_tnl_dev_init_gen(dev);
1525 t->parms.proto = IPPROTO_IPV6;
1528 ip6_tnl_link_config(t);
1530 rcu_assign_pointer(ip6n->tnls_wc[0], t);
1534 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1541 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1542 if (proto != IPPROTO_IPV6 &&
1543 proto != IPPROTO_IPIP &&
1550 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1551 struct __ip6_tnl_parm *parms)
1553 memset(parms, 0, sizeof(*parms));
1558 if (data[IFLA_IPTUN_LINK])
1559 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1561 if (data[IFLA_IPTUN_LOCAL])
1562 nla_memcpy(&parms->laddr, data[IFLA_IPTUN_LOCAL],
1563 sizeof(struct in6_addr));
1565 if (data[IFLA_IPTUN_REMOTE])
1566 nla_memcpy(&parms->raddr, data[IFLA_IPTUN_REMOTE],
1567 sizeof(struct in6_addr));
1569 if (data[IFLA_IPTUN_TTL])
1570 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1572 if (data[IFLA_IPTUN_ENCAP_LIMIT])
1573 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1575 if (data[IFLA_IPTUN_FLOWINFO])
1576 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1578 if (data[IFLA_IPTUN_FLAGS])
1579 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1581 if (data[IFLA_IPTUN_PROTO])
1582 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1585 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1586 struct nlattr *tb[], struct nlattr *data[])
1588 struct net *net = dev_net(dev);
1591 nt = netdev_priv(dev);
1592 ip6_tnl_netlink_parms(data, &nt->parms);
1594 if (ip6_tnl_locate(net, &nt->parms, 0))
1597 return ip6_tnl_create2(dev);
1600 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1601 struct nlattr *data[])
1603 struct ip6_tnl *t = netdev_priv(dev);
1604 struct __ip6_tnl_parm p;
1605 struct net *net = t->net;
1606 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1608 if (dev == ip6n->fb_tnl_dev)
1611 ip6_tnl_netlink_parms(data, &p);
1613 t = ip6_tnl_locate(net, &p, 0);
1619 t = netdev_priv(dev);
1621 return ip6_tnl_update(t, &p);
1624 static size_t ip6_tnl_get_size(const struct net_device *dev)
1627 /* IFLA_IPTUN_LINK */
1629 /* IFLA_IPTUN_LOCAL */
1630 nla_total_size(sizeof(struct in6_addr)) +
1631 /* IFLA_IPTUN_REMOTE */
1632 nla_total_size(sizeof(struct in6_addr)) +
1633 /* IFLA_IPTUN_TTL */
1635 /* IFLA_IPTUN_ENCAP_LIMIT */
1637 /* IFLA_IPTUN_FLOWINFO */
1639 /* IFLA_IPTUN_FLAGS */
1641 /* IFLA_IPTUN_PROTO */
1646 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
1648 struct ip6_tnl *tunnel = netdev_priv(dev);
1649 struct __ip6_tnl_parm *parm = &tunnel->parms;
1651 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1652 nla_put(skb, IFLA_IPTUN_LOCAL, sizeof(struct in6_addr),
1654 nla_put(skb, IFLA_IPTUN_REMOTE, sizeof(struct in6_addr),
1656 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
1657 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
1658 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
1659 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
1660 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
1661 goto nla_put_failure;
1668 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
1669 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1670 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) },
1671 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) },
1672 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1673 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 },
1674 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 },
1675 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 },
1676 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1679 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
1681 .maxtype = IFLA_IPTUN_MAX,
1682 .policy = ip6_tnl_policy,
1683 .priv_size = sizeof(struct ip6_tnl),
1684 .setup = ip6_tnl_dev_setup,
1685 .validate = ip6_tnl_validate,
1686 .newlink = ip6_tnl_newlink,
1687 .changelink = ip6_tnl_changelink,
1688 .get_size = ip6_tnl_get_size,
1689 .fill_info = ip6_tnl_fill_info,
1692 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1693 .handler = ip4ip6_rcv,
1694 .err_handler = ip4ip6_err,
1698 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1699 .handler = ip6ip6_rcv,
1700 .err_handler = ip6ip6_err,
1704 static void __net_exit ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n)
1706 struct net *net = dev_net(ip6n->fb_tnl_dev);
1707 struct net_device *dev, *aux;
1712 for_each_netdev_safe(net, dev, aux)
1713 if (dev->rtnl_link_ops == &ip6_link_ops)
1714 unregister_netdevice_queue(dev, &list);
1716 for (h = 0; h < HASH_SIZE; h++) {
1717 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1719 /* If dev is in the same netns, it has already
1720 * been added to the list by the previous loop.
1722 if (!net_eq(dev_net(t->dev), net))
1723 unregister_netdevice_queue(t->dev, &list);
1724 t = rtnl_dereference(t->next);
1728 t = rtnl_dereference(ip6n->tnls_wc[0]);
1729 unregister_netdevice_queue(t->dev, &list);
1730 unregister_netdevice_many(&list);
1733 static int __net_init ip6_tnl_init_net(struct net *net)
1735 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1736 struct ip6_tnl *t = NULL;
1739 ip6n->tnls[0] = ip6n->tnls_wc;
1740 ip6n->tnls[1] = ip6n->tnls_r_l;
1743 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1746 if (!ip6n->fb_tnl_dev)
1748 dev_net_set(ip6n->fb_tnl_dev, net);
1749 /* FB netdevice is special: we have one, and only one per netns.
1750 * Allowing to move it to another netns is clearly unsafe.
1752 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
1754 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1758 err = register_netdev(ip6n->fb_tnl_dev);
1762 t = netdev_priv(ip6n->fb_tnl_dev);
1764 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1768 ip6_dev_free(ip6n->fb_tnl_dev);
1773 static void __net_exit ip6_tnl_exit_net(struct net *net)
1775 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1778 ip6_tnl_destroy_tunnels(ip6n);
1782 static struct pernet_operations ip6_tnl_net_ops = {
1783 .init = ip6_tnl_init_net,
1784 .exit = ip6_tnl_exit_net,
1785 .id = &ip6_tnl_net_id,
1786 .size = sizeof(struct ip6_tnl_net),
1790 * ip6_tunnel_init - register protocol and reserve needed resources
1792 * Return: 0 on success
1795 static int __init ip6_tunnel_init(void)
1799 err = register_pernet_device(&ip6_tnl_net_ops);
1803 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1805 pr_err("%s: can't register ip4ip6\n", __func__);
1809 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1811 pr_err("%s: can't register ip6ip6\n", __func__);
1814 err = rtnl_link_register(&ip6_link_ops);
1816 goto rtnl_link_failed;
1821 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1823 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1825 unregister_pernet_device(&ip6_tnl_net_ops);
1831 * ip6_tunnel_cleanup - free resources and unregister protocol
1834 static void __exit ip6_tunnel_cleanup(void)
1836 rtnl_link_unregister(&ip6_link_ops);
1837 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1838 pr_info("%s: can't deregister ip4ip6\n", __func__);
1840 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1841 pr_info("%s: can't deregister ip6ip6\n", __func__);
1843 unregister_pernet_device(&ip6_tnl_net_ops);
1846 module_init(ip6_tunnel_init);
1847 module_exit(ip6_tunnel_cleanup);