2 * IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
15 * Roger Venning <r.venning@telstra.com>: 6to4 support
16 * Nate Thompson <nate@thebog.net>: 6to4 support
17 * Fred Templin <fred.l.templin@boeing.com>: isatap support
20 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
22 #include <linux/module.h>
23 #include <linux/capability.h>
24 #include <linux/errno.h>
25 #include <linux/types.h>
26 #include <linux/socket.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/in6.h>
30 #include <linux/netdevice.h>
31 #include <linux/if_arp.h>
32 #include <linux/icmp.h>
33 #include <linux/slab.h>
34 #include <asm/uaccess.h>
35 #include <linux/init.h>
36 #include <linux/netfilter_ipv4.h>
37 #include <linux/if_ether.h>
43 #include <net/protocol.h>
44 #include <net/transp_v6.h>
45 #include <net/ip6_fib.h>
46 #include <net/ip6_route.h>
47 #include <net/ndisc.h>
48 #include <net/addrconf.h>
52 #include <net/ip_tunnels.h>
53 #include <net/inet_ecn.h>
55 #include <net/dsfield.h>
56 #include <net/net_namespace.h>
57 #include <net/netns/generic.h>
60 This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
62 For comments look at net/ipv4/ip_gre.c --ANK
66 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
68 static bool log_ecn_error = true;
69 module_param(log_ecn_error, bool, 0644);
70 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
72 static int ipip6_tunnel_init(struct net_device *dev);
73 static void ipip6_tunnel_setup(struct net_device *dev);
74 static void ipip6_dev_free(struct net_device *dev);
75 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
77 static struct rtnl_link_ops sit_link_ops __read_mostly;
79 static int sit_net_id __read_mostly;
81 struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
82 struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
83 struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
84 struct ip_tunnel __rcu *tunnels_wc[1];
85 struct ip_tunnel __rcu **tunnels[4];
87 struct net_device *fb_tunnel_dev;
91 * Must be invoked with rcu_read_lock
93 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
94 struct net_device *dev, __be32 remote, __be32 local)
96 unsigned int h0 = HASH(remote);
97 unsigned int h1 = HASH(local);
99 struct sit_net *sitn = net_generic(net, sit_net_id);
101 for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
102 if (local == t->parms.iph.saddr &&
103 remote == t->parms.iph.daddr &&
104 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
105 (t->dev->flags & IFF_UP))
108 for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
109 if (remote == t->parms.iph.daddr &&
110 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
111 (t->dev->flags & IFF_UP))
114 for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
115 if (local == t->parms.iph.saddr &&
116 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
117 (t->dev->flags & IFF_UP))
120 t = rcu_dereference(sitn->tunnels_wc[0]);
121 if ((t != NULL) && (t->dev->flags & IFF_UP))
126 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
127 struct ip_tunnel_parm *parms)
129 __be32 remote = parms->iph.daddr;
130 __be32 local = parms->iph.saddr;
142 return &sitn->tunnels[prio][h];
145 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
148 return __ipip6_bucket(sitn, &t->parms);
151 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
153 struct ip_tunnel __rcu **tp;
154 struct ip_tunnel *iter;
156 for (tp = ipip6_bucket(sitn, t);
157 (iter = rtnl_dereference(*tp)) != NULL;
160 rcu_assign_pointer(*tp, t->next);
166 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
168 struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
170 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
171 rcu_assign_pointer(*tp, t);
174 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
176 #ifdef CONFIG_IPV6_SIT_6RD
177 struct ip_tunnel *t = netdev_priv(dev);
179 if (t->dev == sitn->fb_tunnel_dev) {
180 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
181 t->ip6rd.relay_prefix = 0;
182 t->ip6rd.prefixlen = 16;
183 t->ip6rd.relay_prefixlen = 0;
185 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
186 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
191 static int ipip6_tunnel_create(struct net_device *dev)
193 struct ip_tunnel *t = netdev_priv(dev);
194 struct net *net = dev_net(dev);
195 struct sit_net *sitn = net_generic(net, sit_net_id);
198 err = ipip6_tunnel_init(dev);
201 ipip6_tunnel_clone_6rd(dev, sitn);
203 if ((__force u16)t->parms.i_flags & SIT_ISATAP)
204 dev->priv_flags |= IFF_ISATAP;
206 err = register_netdevice(dev);
210 strcpy(t->parms.name, dev->name);
211 dev->rtnl_link_ops = &sit_link_ops;
215 ipip6_tunnel_link(sitn, t);
222 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
223 struct ip_tunnel_parm *parms, int create)
225 __be32 remote = parms->iph.daddr;
226 __be32 local = parms->iph.saddr;
227 struct ip_tunnel *t, *nt;
228 struct ip_tunnel __rcu **tp;
229 struct net_device *dev;
231 struct sit_net *sitn = net_generic(net, sit_net_id);
233 for (tp = __ipip6_bucket(sitn, parms);
234 (t = rtnl_dereference(*tp)) != NULL;
236 if (local == t->parms.iph.saddr &&
237 remote == t->parms.iph.daddr &&
238 parms->link == t->parms.link) {
249 strlcpy(name, parms->name, IFNAMSIZ);
251 strcpy(name, "sit%d");
253 dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
257 dev_net_set(dev, net);
259 nt = netdev_priv(dev);
262 if (ipip6_tunnel_create(dev) < 0)
273 #define for_each_prl_rcu(start) \
274 for (prl = rcu_dereference(start); \
276 prl = rcu_dereference(prl->next))
278 static struct ip_tunnel_prl_entry *
279 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
281 struct ip_tunnel_prl_entry *prl;
283 for_each_prl_rcu(t->prl)
284 if (prl->addr == addr)
290 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
291 struct ip_tunnel_prl __user *a)
293 struct ip_tunnel_prl kprl, *kp;
294 struct ip_tunnel_prl_entry *prl;
295 unsigned int cmax, c = 0, ca, len;
298 if (copy_from_user(&kprl, a, sizeof(kprl)))
300 cmax = kprl.datalen / sizeof(kprl);
301 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
304 /* For simple GET or for root users,
305 * we try harder to allocate.
307 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
308 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
313 ca = t->prl_count < cmax ? t->prl_count : cmax;
316 /* We don't try hard to allocate much memory for
318 * For root users, retry allocating enough memory for
321 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
329 for_each_prl_rcu(t->prl) {
332 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
334 kp[c].addr = prl->addr;
335 kp[c].flags = prl->flags;
337 if (kprl.addr != htonl(INADDR_ANY))
343 len = sizeof(*kp) * c;
345 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
354 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
356 struct ip_tunnel_prl_entry *p;
359 if (a->addr == htonl(INADDR_ANY))
364 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
365 if (p->addr == a->addr) {
380 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
390 rcu_assign_pointer(t->prl, p);
395 static void prl_list_destroy_rcu(struct rcu_head *head)
397 struct ip_tunnel_prl_entry *p, *n;
399 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
401 n = rcu_dereference_protected(p->next, 1);
408 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
410 struct ip_tunnel_prl_entry *x;
411 struct ip_tunnel_prl_entry __rcu **p;
416 if (a && a->addr != htonl(INADDR_ANY)) {
418 (x = rtnl_dereference(*p)) != NULL;
420 if (x->addr == a->addr) {
422 kfree_rcu(x, rcu_head);
429 x = rtnl_dereference(t->prl);
432 call_rcu(&x->rcu_head, prl_list_destroy_rcu);
441 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
443 struct ip_tunnel_prl_entry *p;
447 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
449 if (p->flags & PRL_DEFAULT)
450 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
452 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
454 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
456 if (ipv6_addr_is_isatap(addr6) &&
457 (addr6->s6_addr32[3] == iph->saddr) &&
458 ipv6_chk_prefix(addr6, t->dev))
459 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
467 static void ipip6_tunnel_uninit(struct net_device *dev)
469 struct net *net = dev_net(dev);
470 struct sit_net *sitn = net_generic(net, sit_net_id);
472 if (dev == sitn->fb_tunnel_dev) {
473 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
475 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
476 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
482 static int ipip6_err(struct sk_buff *skb, u32 info)
485 /* All the routers (except for Linux) return only
486 8 bytes of packet payload. It means, that precise relaying of
487 ICMP in the real Internet is absolutely infeasible.
489 const struct iphdr *iph = (const struct iphdr *)skb->data;
490 const int type = icmp_hdr(skb)->type;
491 const int code = icmp_hdr(skb)->code;
497 case ICMP_PARAMETERPROB:
500 case ICMP_DEST_UNREACH:
503 case ICMP_PORT_UNREACH:
504 /* Impossible event. */
507 /* All others are translated to HOST_UNREACH.
508 rfc2003 contains "deep thoughts" about NET_UNREACH,
509 I believe they are just ether pollution. --ANK
514 case ICMP_TIME_EXCEEDED:
515 if (code != ICMP_EXC_TTL)
524 t = ipip6_tunnel_lookup(dev_net(skb->dev),
531 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
532 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
533 t->dev->ifindex, 0, IPPROTO_IPV6, 0);
537 if (type == ICMP_REDIRECT) {
538 ipv4_redirect(skb, dev_net(skb->dev), t->dev->ifindex, 0,
544 if (t->parms.iph.daddr == 0)
548 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
551 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
555 t->err_time = jiffies;
560 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
561 const struct in6_addr *v6addr)
564 if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
569 static int ipip6_rcv(struct sk_buff *skb)
571 const struct iphdr *iph = ip_hdr(skb);
572 struct ip_tunnel *tunnel;
575 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
576 iph->saddr, iph->daddr);
577 if (tunnel != NULL) {
578 struct pcpu_tstats *tstats;
580 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
581 tunnel->parms.iph.protocol != 0)
585 skb->mac_header = skb->network_header;
586 skb_reset_network_header(skb);
587 IPCB(skb)->flags = 0;
588 skb->protocol = htons(ETH_P_IPV6);
589 skb->pkt_type = PACKET_HOST;
591 if (tunnel->dev->priv_flags & IFF_ISATAP) {
592 if (!isatap_chksrc(skb, iph, tunnel)) {
593 tunnel->dev->stats.rx_errors++;
597 if (is_spoofed_6rd(tunnel, iph->saddr,
598 &ipv6_hdr(skb)->saddr) ||
599 is_spoofed_6rd(tunnel, iph->daddr,
600 &ipv6_hdr(skb)->daddr)) {
601 tunnel->dev->stats.rx_errors++;
606 __skb_tunnel_rx(skb, tunnel->dev);
608 err = IP_ECN_decapsulate(iph, skb);
611 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
612 &iph->saddr, iph->tos);
614 ++tunnel->dev->stats.rx_frame_errors;
615 ++tunnel->dev->stats.rx_errors;
620 tstats = this_cpu_ptr(tunnel->dev->tstats);
621 tstats->rx_packets++;
622 tstats->rx_bytes += skb->len;
629 /* no tunnel matched, let upstream know, ipsec may handle it */
636 static const struct tnl_ptk_info tpi = {
637 /* no tunnel info required for ipip. */
638 .proto = htons(ETH_P_IP),
641 static int ipip_rcv(struct sk_buff *skb)
643 const struct iphdr *iph;
644 struct ip_tunnel *tunnel;
646 if (iptunnel_pull_header(skb, 0, tpi.proto))
651 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
652 iph->saddr, iph->daddr);
653 if (tunnel != NULL) {
654 if (tunnel->parms.iph.protocol != IPPROTO_IPIP &&
655 tunnel->parms.iph.protocol != 0)
658 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
660 return ip_tunnel_rcv(tunnel, skb, &tpi, log_ecn_error);
671 * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
672 * stores the embedded IPv4 address in v4dst and returns true.
674 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
677 #ifdef CONFIG_IPV6_SIT_6RD
678 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
679 tunnel->ip6rd.prefixlen)) {
680 unsigned int pbw0, pbi0;
684 pbw0 = tunnel->ip6rd.prefixlen >> 5;
685 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
687 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
688 tunnel->ip6rd.relay_prefixlen;
690 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
692 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
695 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
699 if (v6dst->s6_addr16[0] == htons(0x2002)) {
700 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
701 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
708 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
709 const struct in6_addr *v6dst)
712 check_6rd(tunnel, v6dst, &dst);
717 * This function assumes it is being called from dev_queue_xmit()
718 * and that skb is filled properly by that function.
721 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
722 struct net_device *dev)
724 struct ip_tunnel *tunnel = netdev_priv(dev);
725 const struct iphdr *tiph = &tunnel->parms.iph;
726 const struct ipv6hdr *iph6 = ipv6_hdr(skb);
727 u8 tos = tunnel->parms.iph.tos;
728 __be16 df = tiph->frag_off;
729 struct rtable *rt; /* Route to the other host */
730 struct net_device *tdev; /* Device to other host */
731 unsigned int max_headroom; /* The extra header space needed */
732 __be32 dst = tiph->daddr;
735 const struct in6_addr *addr6;
740 if (skb->protocol != htons(ETH_P_IPV6))
744 tos = ipv6_get_dsfield(iph6);
746 /* ISATAP (RFC4214) - must come before 6to4 */
747 if (dev->priv_flags & IFF_ISATAP) {
748 struct neighbour *neigh = NULL;
749 bool do_tx_error = false;
752 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
755 net_dbg_ratelimited("sit: nexthop == NULL\n");
759 addr6 = (const struct in6_addr *)&neigh->primary_key;
760 addr_type = ipv6_addr_type(addr6);
762 if ((addr_type & IPV6_ADDR_UNICAST) &&
763 ipv6_addr_is_isatap(addr6))
764 dst = addr6->s6_addr32[3];
768 neigh_release(neigh);
774 dst = try_6rd(tunnel, &iph6->daddr);
777 struct neighbour *neigh = NULL;
778 bool do_tx_error = false;
781 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
784 net_dbg_ratelimited("sit: nexthop == NULL\n");
788 addr6 = (const struct in6_addr *)&neigh->primary_key;
789 addr_type = ipv6_addr_type(addr6);
791 if (addr_type == IPV6_ADDR_ANY) {
792 addr6 = &ipv6_hdr(skb)->daddr;
793 addr_type = ipv6_addr_type(addr6);
796 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
797 dst = addr6->s6_addr32[3];
801 neigh_release(neigh);
806 rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
809 IPPROTO_IPV6, RT_TOS(tos),
812 dev->stats.tx_carrier_errors++;
815 if (rt->rt_type != RTN_UNICAST) {
817 dev->stats.tx_carrier_errors++;
824 dev->stats.collisions++;
829 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
832 dev->stats.collisions++;
837 if (mtu < IPV6_MIN_MTU) {
842 if (tunnel->parms.iph.daddr && skb_dst(skb))
843 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
845 if (skb->len > mtu) {
846 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
852 if (tunnel->err_count > 0) {
853 if (time_before(jiffies,
854 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
856 dst_link_failure(skb);
858 tunnel->err_count = 0;
862 * Okay, now see if we can stuff it in the buffer as-is.
864 max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
866 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
867 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
868 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
871 dev->stats.tx_dropped++;
876 skb_set_owner_w(new_skb, skb->sk);
879 iph6 = ipv6_hdr(skb);
883 ttl = iph6->hop_limit;
884 tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
886 err = iptunnel_xmit(dev_net(dev), rt, skb, fl4.saddr, fl4.daddr,
887 IPPROTO_IPV6, tos, ttl, df);
888 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
892 dst_link_failure(skb);
894 dev->stats.tx_errors++;
899 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
901 struct ip_tunnel *tunnel = netdev_priv(dev);
902 const struct iphdr *tiph = &tunnel->parms.iph;
904 if (likely(!skb->encapsulation)) {
905 skb_reset_inner_headers(skb);
906 skb->encapsulation = 1;
909 ip_tunnel_xmit(skb, dev, tiph, IPPROTO_IPIP);
913 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
914 struct net_device *dev)
916 switch (skb->protocol) {
917 case htons(ETH_P_IP):
918 ipip_tunnel_xmit(skb, dev);
920 case htons(ETH_P_IPV6):
921 ipip6_tunnel_xmit(skb, dev);
930 dev->stats.tx_errors++;
936 static void ipip6_tunnel_bind_dev(struct net_device *dev)
938 struct net_device *tdev = NULL;
939 struct ip_tunnel *tunnel;
940 const struct iphdr *iph;
943 tunnel = netdev_priv(dev);
944 iph = &tunnel->parms.iph;
947 struct rtable *rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
948 iph->daddr, iph->saddr,
958 dev->flags |= IFF_POINTOPOINT;
961 if (!tdev && tunnel->parms.link)
962 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
965 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
966 dev->mtu = tdev->mtu - sizeof(struct iphdr);
967 if (dev->mtu < IPV6_MIN_MTU)
968 dev->mtu = IPV6_MIN_MTU;
970 dev->iflink = tunnel->parms.link;
973 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
975 struct net *net = dev_net(t->dev);
976 struct sit_net *sitn = net_generic(net, sit_net_id);
978 ipip6_tunnel_unlink(sitn, t);
980 t->parms.iph.saddr = p->iph.saddr;
981 t->parms.iph.daddr = p->iph.daddr;
982 memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
983 memcpy(t->dev->broadcast, &p->iph.daddr, 4);
984 ipip6_tunnel_link(sitn, t);
985 t->parms.iph.ttl = p->iph.ttl;
986 t->parms.iph.tos = p->iph.tos;
987 if (t->parms.link != p->link) {
988 t->parms.link = p->link;
989 ipip6_tunnel_bind_dev(t->dev);
991 netdev_state_change(t->dev);
994 #ifdef CONFIG_IPV6_SIT_6RD
995 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
996 struct ip_tunnel_6rd *ip6rd)
998 struct in6_addr prefix;
1001 if (ip6rd->relay_prefixlen > 32 ||
1002 ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1005 ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1006 if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1008 if (ip6rd->relay_prefixlen)
1009 relay_prefix = ip6rd->relay_prefix &
1010 htonl(0xffffffffUL <<
1011 (32 - ip6rd->relay_prefixlen));
1014 if (relay_prefix != ip6rd->relay_prefix)
1017 t->ip6rd.prefix = prefix;
1018 t->ip6rd.relay_prefix = relay_prefix;
1019 t->ip6rd.prefixlen = ip6rd->prefixlen;
1020 t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1021 netdev_state_change(t->dev);
1027 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
1030 struct ip_tunnel_parm p;
1031 struct ip_tunnel_prl prl;
1032 struct ip_tunnel *t;
1033 struct net *net = dev_net(dev);
1034 struct sit_net *sitn = net_generic(net, sit_net_id);
1035 #ifdef CONFIG_IPV6_SIT_6RD
1036 struct ip_tunnel_6rd ip6rd;
1041 #ifdef CONFIG_IPV6_SIT_6RD
1045 if (dev == sitn->fb_tunnel_dev) {
1046 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1050 t = ipip6_tunnel_locate(net, &p, 0);
1053 t = netdev_priv(dev);
1056 if (cmd == SIOCGETTUNNEL) {
1057 memcpy(&p, &t->parms, sizeof(p));
1058 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1061 #ifdef CONFIG_IPV6_SIT_6RD
1063 ip6rd.prefix = t->ip6rd.prefix;
1064 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1065 ip6rd.prefixlen = t->ip6rd.prefixlen;
1066 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1067 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1078 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1082 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1086 if (p.iph.protocol != IPPROTO_IPV6 &&
1087 p.iph.protocol != IPPROTO_IPIP &&
1088 p.iph.protocol != 0)
1090 if (p.iph.version != 4 ||
1091 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1094 p.iph.frag_off |= htons(IP_DF);
1096 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1098 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1100 if (t->dev != dev) {
1105 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1106 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1110 t = netdev_priv(dev);
1113 ipip6_tunnel_update(t, &p);
1118 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1121 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1126 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1129 if (dev == sitn->fb_tunnel_dev) {
1131 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1134 if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1137 if (t == netdev_priv(sitn->fb_tunnel_dev))
1141 unregister_netdevice(dev);
1147 if (dev == sitn->fb_tunnel_dev)
1150 if (!(t = netdev_priv(dev)))
1152 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1159 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1162 if (dev == sitn->fb_tunnel_dev)
1165 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1168 if (!(t = netdev_priv(dev)))
1173 err = ipip6_tunnel_del_prl(t, &prl);
1177 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1180 netdev_state_change(dev);
1183 #ifdef CONFIG_IPV6_SIT_6RD
1188 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1192 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1196 t = netdev_priv(dev);
1198 if (cmd != SIOCDEL6RD) {
1199 err = ipip6_tunnel_update_6rd(t, &ip6rd);
1203 ipip6_tunnel_clone_6rd(dev, sitn);
1217 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1219 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1225 static const struct net_device_ops ipip6_netdev_ops = {
1226 .ndo_uninit = ipip6_tunnel_uninit,
1227 .ndo_start_xmit = sit_tunnel_xmit,
1228 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1229 .ndo_change_mtu = ipip6_tunnel_change_mtu,
1230 .ndo_get_stats64 = ip_tunnel_get_stats64,
1233 static void ipip6_dev_free(struct net_device *dev)
1235 free_percpu(dev->tstats);
1239 static void ipip6_tunnel_setup(struct net_device *dev)
1241 dev->netdev_ops = &ipip6_netdev_ops;
1242 dev->destructor = ipip6_dev_free;
1244 dev->type = ARPHRD_SIT;
1245 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
1246 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr);
1247 dev->flags = IFF_NOARP;
1248 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1251 dev->features |= NETIF_F_NETNS_LOCAL;
1252 dev->features |= NETIF_F_LLTX;
1255 static int ipip6_tunnel_init(struct net_device *dev)
1257 struct ip_tunnel *tunnel = netdev_priv(dev);
1261 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1262 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1264 ipip6_tunnel_bind_dev(dev);
1265 dev->tstats = alloc_percpu(struct pcpu_tstats);
1272 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1274 struct ip_tunnel *tunnel = netdev_priv(dev);
1275 struct iphdr *iph = &tunnel->parms.iph;
1276 struct net *net = dev_net(dev);
1277 struct sit_net *sitn = net_generic(net, sit_net_id);
1280 strcpy(tunnel->parms.name, dev->name);
1283 iph->protocol = IPPROTO_IPV6;
1287 dev->tstats = alloc_percpu(struct pcpu_tstats);
1291 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1295 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[])
1299 if (!data || !data[IFLA_IPTUN_PROTO])
1302 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1303 if (proto != IPPROTO_IPV6 &&
1304 proto != IPPROTO_IPIP &&
1311 static void ipip6_netlink_parms(struct nlattr *data[],
1312 struct ip_tunnel_parm *parms)
1314 memset(parms, 0, sizeof(*parms));
1316 parms->iph.version = 4;
1317 parms->iph.protocol = IPPROTO_IPV6;
1319 parms->iph.ttl = 64;
1324 if (data[IFLA_IPTUN_LINK])
1325 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1327 if (data[IFLA_IPTUN_LOCAL])
1328 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1330 if (data[IFLA_IPTUN_REMOTE])
1331 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1333 if (data[IFLA_IPTUN_TTL]) {
1334 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1336 parms->iph.frag_off = htons(IP_DF);
1339 if (data[IFLA_IPTUN_TOS])
1340 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1342 if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1343 parms->iph.frag_off = htons(IP_DF);
1345 if (data[IFLA_IPTUN_FLAGS])
1346 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1348 if (data[IFLA_IPTUN_PROTO])
1349 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1353 #ifdef CONFIG_IPV6_SIT_6RD
1354 /* This function returns true when 6RD attributes are present in the nl msg */
1355 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1356 struct ip_tunnel_6rd *ip6rd)
1359 memset(ip6rd, 0, sizeof(*ip6rd));
1364 if (data[IFLA_IPTUN_6RD_PREFIX]) {
1366 nla_memcpy(&ip6rd->prefix, data[IFLA_IPTUN_6RD_PREFIX],
1367 sizeof(struct in6_addr));
1370 if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1372 ip6rd->relay_prefix =
1373 nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1376 if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1378 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1381 if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1383 ip6rd->relay_prefixlen =
1384 nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1391 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1392 struct nlattr *tb[], struct nlattr *data[])
1394 struct net *net = dev_net(dev);
1395 struct ip_tunnel *nt;
1396 #ifdef CONFIG_IPV6_SIT_6RD
1397 struct ip_tunnel_6rd ip6rd;
1401 nt = netdev_priv(dev);
1402 ipip6_netlink_parms(data, &nt->parms);
1404 if (ipip6_tunnel_locate(net, &nt->parms, 0))
1407 err = ipip6_tunnel_create(dev);
1411 #ifdef CONFIG_IPV6_SIT_6RD
1412 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1413 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1419 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1420 struct nlattr *data[])
1422 struct ip_tunnel *t;
1423 struct ip_tunnel_parm p;
1424 struct net *net = dev_net(dev);
1425 struct sit_net *sitn = net_generic(net, sit_net_id);
1426 #ifdef CONFIG_IPV6_SIT_6RD
1427 struct ip_tunnel_6rd ip6rd;
1430 if (dev == sitn->fb_tunnel_dev)
1433 ipip6_netlink_parms(data, &p);
1435 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1436 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1439 t = ipip6_tunnel_locate(net, &p, 0);
1445 t = netdev_priv(dev);
1447 ipip6_tunnel_update(t, &p);
1449 #ifdef CONFIG_IPV6_SIT_6RD
1450 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1451 return ipip6_tunnel_update_6rd(t, &ip6rd);
1457 static size_t ipip6_get_size(const struct net_device *dev)
1460 /* IFLA_IPTUN_LINK */
1462 /* IFLA_IPTUN_LOCAL */
1464 /* IFLA_IPTUN_REMOTE */
1466 /* IFLA_IPTUN_TTL */
1468 /* IFLA_IPTUN_TOS */
1470 /* IFLA_IPTUN_PMTUDISC */
1472 /* IFLA_IPTUN_FLAGS */
1474 /* IFLA_IPTUN_PROTO */
1476 #ifdef CONFIG_IPV6_SIT_6RD
1477 /* IFLA_IPTUN_6RD_PREFIX */
1478 nla_total_size(sizeof(struct in6_addr)) +
1479 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1481 /* IFLA_IPTUN_6RD_PREFIXLEN */
1483 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1489 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1491 struct ip_tunnel *tunnel = netdev_priv(dev);
1492 struct ip_tunnel_parm *parm = &tunnel->parms;
1494 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1495 nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1496 nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1497 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1498 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1499 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1500 !!(parm->iph.frag_off & htons(IP_DF))) ||
1501 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1502 nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1503 goto nla_put_failure;
1505 #ifdef CONFIG_IPV6_SIT_6RD
1506 if (nla_put(skb, IFLA_IPTUN_6RD_PREFIX, sizeof(struct in6_addr),
1507 &tunnel->ip6rd.prefix) ||
1508 nla_put_be32(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1509 tunnel->ip6rd.relay_prefix) ||
1510 nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1511 tunnel->ip6rd.prefixlen) ||
1512 nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1513 tunnel->ip6rd.relay_prefixlen))
1514 goto nla_put_failure;
1523 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1524 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1525 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 },
1526 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 },
1527 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1528 [IFLA_IPTUN_TOS] = { .type = NLA_U8 },
1529 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 },
1530 [IFLA_IPTUN_FLAGS] = { .type = NLA_U16 },
1531 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1532 #ifdef CONFIG_IPV6_SIT_6RD
1533 [IFLA_IPTUN_6RD_PREFIX] = { .len = sizeof(struct in6_addr) },
1534 [IFLA_IPTUN_6RD_RELAY_PREFIX] = { .type = NLA_U32 },
1535 [IFLA_IPTUN_6RD_PREFIXLEN] = { .type = NLA_U16 },
1536 [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1540 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1542 .maxtype = IFLA_IPTUN_MAX,
1543 .policy = ipip6_policy,
1544 .priv_size = sizeof(struct ip_tunnel),
1545 .setup = ipip6_tunnel_setup,
1546 .validate = ipip6_validate,
1547 .newlink = ipip6_newlink,
1548 .changelink = ipip6_changelink,
1549 .get_size = ipip6_get_size,
1550 .fill_info = ipip6_fill_info,
1553 static struct xfrm_tunnel sit_handler __read_mostly = {
1554 .handler = ipip6_rcv,
1555 .err_handler = ipip6_err,
1559 static struct xfrm_tunnel ipip_handler __read_mostly = {
1560 .handler = ipip_rcv,
1561 .err_handler = ipip6_err,
1565 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1569 for (prio = 1; prio < 4; prio++) {
1571 for (h = 0; h < HASH_SIZE; h++) {
1572 struct ip_tunnel *t;
1574 t = rtnl_dereference(sitn->tunnels[prio][h]);
1576 unregister_netdevice_queue(t->dev, head);
1577 t = rtnl_dereference(t->next);
1583 static int __net_init sit_init_net(struct net *net)
1585 struct sit_net *sitn = net_generic(net, sit_net_id);
1586 struct ip_tunnel *t;
1589 sitn->tunnels[0] = sitn->tunnels_wc;
1590 sitn->tunnels[1] = sitn->tunnels_l;
1591 sitn->tunnels[2] = sitn->tunnels_r;
1592 sitn->tunnels[3] = sitn->tunnels_r_l;
1594 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1595 ipip6_tunnel_setup);
1596 if (!sitn->fb_tunnel_dev) {
1600 dev_net_set(sitn->fb_tunnel_dev, net);
1602 err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1606 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1608 if ((err = register_netdev(sitn->fb_tunnel_dev)))
1611 t = netdev_priv(sitn->fb_tunnel_dev);
1613 strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1617 dev_put(sitn->fb_tunnel_dev);
1619 ipip6_dev_free(sitn->fb_tunnel_dev);
1624 static void __net_exit sit_exit_net(struct net *net)
1626 struct sit_net *sitn = net_generic(net, sit_net_id);
1630 sit_destroy_tunnels(sitn, &list);
1631 unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1632 unregister_netdevice_many(&list);
1636 static struct pernet_operations sit_net_ops = {
1637 .init = sit_init_net,
1638 .exit = sit_exit_net,
1640 .size = sizeof(struct sit_net),
1643 static void __exit sit_cleanup(void)
1645 rtnl_link_unregister(&sit_link_ops);
1646 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1647 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1649 unregister_pernet_device(&sit_net_ops);
1650 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1653 static int __init sit_init(void)
1657 pr_info("IPv6 over IPv4 tunneling driver\n");
1659 err = register_pernet_device(&sit_net_ops);
1662 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1664 pr_info("%s: can't register ip6ip4\n", __func__);
1665 goto xfrm_tunnel_failed;
1667 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1669 pr_info("%s: can't register ip4ip4\n", __func__);
1670 goto xfrm_tunnel4_failed;
1672 err = rtnl_link_register(&sit_link_ops);
1674 goto rtnl_link_failed;
1680 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1681 xfrm_tunnel4_failed:
1682 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1684 unregister_pernet_device(&sit_net_ops);
1688 module_init(sit_init);
1689 module_exit(sit_cleanup);
1690 MODULE_LICENSE("GPL");
1691 MODULE_ALIAS_NETDEV("sit0");