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 #include <linux/module.h>
21 #include <linux/capability.h>
22 #include <linux/errno.h>
23 #include <linux/types.h>
24 #include <linux/socket.h>
25 #include <linux/sockios.h>
26 #include <linux/net.h>
27 #include <linux/in6.h>
28 #include <linux/netdevice.h>
29 #include <linux/if_arp.h>
30 #include <linux/icmp.h>
31 #include <asm/uaccess.h>
32 #include <linux/init.h>
33 #include <linux/netfilter_ipv4.h>
34 #include <linux/if_ether.h>
40 #include <net/protocol.h>
41 #include <net/transp_v6.h>
42 #include <net/ip6_fib.h>
43 #include <net/ip6_route.h>
44 #include <net/ndisc.h>
45 #include <net/addrconf.h>
50 #include <net/inet_ecn.h>
52 #include <net/dsfield.h>
53 #include <net/net_namespace.h>
54 #include <net/netns/generic.h>
57 This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
59 For comments look at net/ipv4/ip_gre.c --ANK
63 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
65 static void ipip6_tunnel_init(struct net_device *dev);
66 static void ipip6_tunnel_setup(struct net_device *dev);
68 static int sit_net_id __read_mostly;
70 struct ip_tunnel *tunnels_r_l[HASH_SIZE];
71 struct ip_tunnel *tunnels_r[HASH_SIZE];
72 struct ip_tunnel *tunnels_l[HASH_SIZE];
73 struct ip_tunnel *tunnels_wc[1];
74 struct ip_tunnel **tunnels[4];
76 struct net_device *fb_tunnel_dev;
80 * Locking : hash tables are protected by RCU and a spinlock
82 static DEFINE_SPINLOCK(ipip6_lock);
84 #define for_each_ip_tunnel_rcu(start) \
85 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
88 * Must be invoked with rcu_read_lock
90 static struct ip_tunnel * ipip6_tunnel_lookup(struct net *net,
91 struct net_device *dev, __be32 remote, __be32 local)
93 unsigned h0 = HASH(remote);
94 unsigned h1 = HASH(local);
96 struct sit_net *sitn = net_generic(net, sit_net_id);
98 for_each_ip_tunnel_rcu(sitn->tunnels_r_l[h0 ^ h1]) {
99 if (local == t->parms.iph.saddr &&
100 remote == t->parms.iph.daddr &&
101 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
102 (t->dev->flags & IFF_UP))
105 for_each_ip_tunnel_rcu(sitn->tunnels_r[h0]) {
106 if (remote == t->parms.iph.daddr &&
107 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
108 (t->dev->flags & IFF_UP))
111 for_each_ip_tunnel_rcu(sitn->tunnels_l[h1]) {
112 if (local == t->parms.iph.saddr &&
113 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
114 (t->dev->flags & IFF_UP))
117 t = rcu_dereference(sitn->tunnels_wc[0]);
118 if ((t != NULL) && (t->dev->flags & IFF_UP))
123 static struct ip_tunnel **__ipip6_bucket(struct sit_net *sitn,
124 struct ip_tunnel_parm *parms)
126 __be32 remote = parms->iph.daddr;
127 __be32 local = parms->iph.saddr;
139 return &sitn->tunnels[prio][h];
142 static inline struct ip_tunnel **ipip6_bucket(struct sit_net *sitn,
145 return __ipip6_bucket(sitn, &t->parms);
148 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
150 struct ip_tunnel **tp;
152 for (tp = ipip6_bucket(sitn, t); *tp; tp = &(*tp)->next) {
154 spin_lock_bh(&ipip6_lock);
156 spin_unlock_bh(&ipip6_lock);
162 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
164 struct ip_tunnel **tp = ipip6_bucket(sitn, t);
166 spin_lock_bh(&ipip6_lock);
168 rcu_assign_pointer(*tp, t);
169 spin_unlock_bh(&ipip6_lock);
172 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
174 #ifdef CONFIG_IPV6_SIT_6RD
175 struct ip_tunnel *t = netdev_priv(dev);
177 if (t->dev == sitn->fb_tunnel_dev) {
178 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
179 t->ip6rd.relay_prefix = 0;
180 t->ip6rd.prefixlen = 16;
181 t->ip6rd.relay_prefixlen = 0;
183 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
184 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
189 static struct ip_tunnel * ipip6_tunnel_locate(struct net *net,
190 struct ip_tunnel_parm *parms, int create)
192 __be32 remote = parms->iph.daddr;
193 __be32 local = parms->iph.saddr;
194 struct ip_tunnel *t, **tp, *nt;
195 struct net_device *dev;
197 struct sit_net *sitn = net_generic(net, sit_net_id);
199 for (tp = __ipip6_bucket(sitn, parms); (t = *tp) != NULL; tp = &t->next) {
200 if (local == t->parms.iph.saddr &&
201 remote == t->parms.iph.daddr &&
202 parms->link == t->parms.link) {
213 strlcpy(name, parms->name, IFNAMSIZ);
215 sprintf(name, "sit%%d");
217 dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
221 dev_net_set(dev, net);
223 if (strchr(name, '%')) {
224 if (dev_alloc_name(dev, name) < 0)
228 nt = netdev_priv(dev);
231 ipip6_tunnel_init(dev);
232 ipip6_tunnel_clone_6rd(dev, sitn);
234 if (parms->i_flags & SIT_ISATAP)
235 dev->priv_flags |= IFF_ISATAP;
237 if (register_netdevice(dev) < 0)
242 ipip6_tunnel_link(sitn, nt);
251 static DEFINE_SPINLOCK(ipip6_prl_lock);
253 #define for_each_prl_rcu(start) \
254 for (prl = rcu_dereference(start); \
256 prl = rcu_dereference(prl->next))
258 static struct ip_tunnel_prl_entry *
259 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
261 struct ip_tunnel_prl_entry *prl;
263 for_each_prl_rcu(t->prl)
264 if (prl->addr == addr)
270 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
271 struct ip_tunnel_prl __user *a)
273 struct ip_tunnel_prl kprl, *kp;
274 struct ip_tunnel_prl_entry *prl;
275 unsigned int cmax, c = 0, ca, len;
278 if (copy_from_user(&kprl, a, sizeof(kprl)))
280 cmax = kprl.datalen / sizeof(kprl);
281 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
284 /* For simple GET or for root users,
285 * we try harder to allocate.
287 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
288 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
293 ca = t->prl_count < cmax ? t->prl_count : cmax;
296 /* We don't try hard to allocate much memory for
298 * For root users, retry allocating enough memory for
301 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
309 for_each_prl_rcu(t->prl) {
312 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
314 kp[c].addr = prl->addr;
315 kp[c].flags = prl->flags;
317 if (kprl.addr != htonl(INADDR_ANY))
323 len = sizeof(*kp) * c;
325 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
334 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
336 struct ip_tunnel_prl_entry *p;
339 if (a->addr == htonl(INADDR_ANY))
342 spin_lock(&ipip6_prl_lock);
344 for (p = t->prl; p; p = p->next) {
345 if (p->addr == a->addr) {
360 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
370 rcu_assign_pointer(t->prl, p);
372 spin_unlock(&ipip6_prl_lock);
376 static void prl_entry_destroy_rcu(struct rcu_head *head)
378 kfree(container_of(head, struct ip_tunnel_prl_entry, rcu_head));
381 static void prl_list_destroy_rcu(struct rcu_head *head)
383 struct ip_tunnel_prl_entry *p, *n;
385 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
394 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
396 struct ip_tunnel_prl_entry *x, **p;
399 spin_lock(&ipip6_prl_lock);
401 if (a && a->addr != htonl(INADDR_ANY)) {
402 for (p = &t->prl; *p; p = &(*p)->next) {
403 if ((*p)->addr == a->addr) {
406 call_rcu(&x->rcu_head, prl_entry_destroy_rcu);
416 call_rcu(&x->rcu_head, prl_list_destroy_rcu);
421 spin_unlock(&ipip6_prl_lock);
426 isatap_chksrc(struct sk_buff *skb, struct iphdr *iph, struct ip_tunnel *t)
428 struct ip_tunnel_prl_entry *p;
432 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
434 if (p->flags & PRL_DEFAULT)
435 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
437 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
439 struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
440 if (ipv6_addr_is_isatap(addr6) &&
441 (addr6->s6_addr32[3] == iph->saddr) &&
442 ipv6_chk_prefix(addr6, t->dev))
443 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
451 static void ipip6_tunnel_uninit(struct net_device *dev)
453 struct net *net = dev_net(dev);
454 struct sit_net *sitn = net_generic(net, sit_net_id);
456 if (dev == sitn->fb_tunnel_dev) {
457 spin_lock_bh(&ipip6_lock);
458 sitn->tunnels_wc[0] = NULL;
459 spin_unlock_bh(&ipip6_lock);
462 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
463 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
469 static int ipip6_err(struct sk_buff *skb, u32 info)
472 /* All the routers (except for Linux) return only
473 8 bytes of packet payload. It means, that precise relaying of
474 ICMP in the real Internet is absolutely infeasible.
476 struct iphdr *iph = (struct iphdr*)skb->data;
477 const int type = icmp_hdr(skb)->type;
478 const int code = icmp_hdr(skb)->code;
484 case ICMP_PARAMETERPROB:
487 case ICMP_DEST_UNREACH:
490 case ICMP_PORT_UNREACH:
491 /* Impossible event. */
493 case ICMP_FRAG_NEEDED:
494 /* Soft state for pmtu is maintained by IP core. */
497 /* All others are translated to HOST_UNREACH.
498 rfc2003 contains "deep thoughts" about NET_UNREACH,
499 I believe they are just ether pollution. --ANK
504 case ICMP_TIME_EXCEEDED:
505 if (code != ICMP_EXC_TTL)
513 t = ipip6_tunnel_lookup(dev_net(skb->dev),
517 if (t == NULL || t->parms.iph.daddr == 0)
521 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
524 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
528 t->err_time = jiffies;
534 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
536 if (INET_ECN_is_ce(iph->tos))
537 IP6_ECN_set_ce(ipv6_hdr(skb));
540 static int ipip6_rcv(struct sk_buff *skb)
543 struct ip_tunnel *tunnel;
545 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
551 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
552 iph->saddr, iph->daddr);
553 if (tunnel != NULL) {
555 skb->mac_header = skb->network_header;
556 skb_reset_network_header(skb);
557 IPCB(skb)->flags = 0;
558 skb->protocol = htons(ETH_P_IPV6);
559 skb->pkt_type = PACKET_HOST;
561 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
562 !isatap_chksrc(skb, iph, tunnel)) {
563 tunnel->dev->stats.rx_errors++;
568 tunnel->dev->stats.rx_packets++;
569 tunnel->dev->stats.rx_bytes += skb->len;
570 skb->dev = tunnel->dev;
573 ipip6_ecn_decapsulate(iph, skb);
579 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
587 * Returns the embedded IPv4 address if the IPv6 address
588 * comes from 6rd / 6to4 (RFC 3056) addr space.
591 __be32 try_6rd(struct in6_addr *v6dst, struct ip_tunnel *tunnel)
595 #ifdef CONFIG_IPV6_SIT_6RD
596 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
597 tunnel->ip6rd.prefixlen)) {
602 pbw0 = tunnel->ip6rd.prefixlen >> 5;
603 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
605 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
606 tunnel->ip6rd.relay_prefixlen;
608 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
610 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
613 dst = tunnel->ip6rd.relay_prefix | htonl(d);
616 if (v6dst->s6_addr16[0] == htons(0x2002)) {
617 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
618 memcpy(&dst, &v6dst->s6_addr16[1], 4);
625 * This function assumes it is being called from dev_queue_xmit()
626 * and that skb is filled properly by that function.
629 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
630 struct net_device *dev)
632 struct ip_tunnel *tunnel = netdev_priv(dev);
633 struct net_device_stats *stats = &dev->stats;
634 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
635 struct iphdr *tiph = &tunnel->parms.iph;
636 struct ipv6hdr *iph6 = ipv6_hdr(skb);
637 u8 tos = tunnel->parms.iph.tos;
638 __be16 df = tiph->frag_off;
639 struct rtable *rt; /* Route to the other host */
640 struct net_device *tdev; /* Device to other host */
641 struct iphdr *iph; /* Our new IP header */
642 unsigned int max_headroom; /* The extra header space needed */
643 __be32 dst = tiph->daddr;
645 struct in6_addr *addr6;
648 if (skb->protocol != htons(ETH_P_IPV6))
651 /* ISATAP (RFC4214) - must come before 6to4 */
652 if (dev->priv_flags & IFF_ISATAP) {
653 struct neighbour *neigh = NULL;
656 neigh = skb_dst(skb)->neighbour;
660 printk(KERN_DEBUG "sit: nexthop == NULL\n");
664 addr6 = (struct in6_addr*)&neigh->primary_key;
665 addr_type = ipv6_addr_type(addr6);
667 if ((addr_type & IPV6_ADDR_UNICAST) &&
668 ipv6_addr_is_isatap(addr6))
669 dst = addr6->s6_addr32[3];
675 dst = try_6rd(&iph6->daddr, tunnel);
678 struct neighbour *neigh = NULL;
681 neigh = skb_dst(skb)->neighbour;
685 printk(KERN_DEBUG "sit: nexthop == NULL\n");
689 addr6 = (struct in6_addr*)&neigh->primary_key;
690 addr_type = ipv6_addr_type(addr6);
692 if (addr_type == IPV6_ADDR_ANY) {
693 addr6 = &ipv6_hdr(skb)->daddr;
694 addr_type = ipv6_addr_type(addr6);
697 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
700 dst = addr6->s6_addr32[3];
704 struct flowi fl = { .nl_u = { .ip4_u =
706 .saddr = tiph->saddr,
707 .tos = RT_TOS(tos) } },
708 .oif = tunnel->parms.link,
709 .proto = IPPROTO_IPV6 };
710 if (ip_route_output_key(dev_net(dev), &rt, &fl)) {
711 stats->tx_carrier_errors++;
715 if (rt->rt_type != RTN_UNICAST) {
717 stats->tx_carrier_errors++;
720 tdev = rt->u.dst.dev;
729 mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
737 if (mtu < IPV6_MIN_MTU) {
742 if (tunnel->parms.iph.daddr && skb_dst(skb))
743 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
745 if (skb->len > mtu) {
746 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
752 if (tunnel->err_count > 0) {
753 if (time_before(jiffies,
754 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
756 dst_link_failure(skb);
758 tunnel->err_count = 0;
762 * Okay, now see if we can stuff it in the buffer as-is.
764 max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
766 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
767 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
768 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
776 skb_set_owner_w(new_skb, skb->sk);
779 iph6 = ipv6_hdr(skb);
782 skb->transport_header = skb->network_header;
783 skb_push(skb, sizeof(struct iphdr));
784 skb_reset_network_header(skb);
785 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
786 IPCB(skb)->flags = 0;
788 skb_dst_set(skb, &rt->u.dst);
791 * Push down and install the IPIP header.
796 iph->ihl = sizeof(struct iphdr)>>2;
798 iph->protocol = IPPROTO_IPV6;
799 iph->tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
800 iph->daddr = rt->rt_dst;
801 iph->saddr = rt->rt_src;
803 if ((iph->ttl = tiph->ttl) == 0)
804 iph->ttl = iph6->hop_limit;
812 dst_link_failure(skb);
819 static void ipip6_tunnel_bind_dev(struct net_device *dev)
821 struct net_device *tdev = NULL;
822 struct ip_tunnel *tunnel;
825 tunnel = netdev_priv(dev);
826 iph = &tunnel->parms.iph;
829 struct flowi fl = { .nl_u = { .ip4_u =
830 { .daddr = iph->daddr,
832 .tos = RT_TOS(iph->tos) } },
833 .oif = tunnel->parms.link,
834 .proto = IPPROTO_IPV6 };
836 if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
837 tdev = rt->u.dst.dev;
840 dev->flags |= IFF_POINTOPOINT;
843 if (!tdev && tunnel->parms.link)
844 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
847 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
848 dev->mtu = tdev->mtu - sizeof(struct iphdr);
849 if (dev->mtu < IPV6_MIN_MTU)
850 dev->mtu = IPV6_MIN_MTU;
852 dev->iflink = tunnel->parms.link;
856 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
859 struct ip_tunnel_parm p;
860 struct ip_tunnel_prl prl;
862 struct net *net = dev_net(dev);
863 struct sit_net *sitn = net_generic(net, sit_net_id);
864 #ifdef CONFIG_IPV6_SIT_6RD
865 struct ip_tunnel_6rd ip6rd;
870 #ifdef CONFIG_IPV6_SIT_6RD
874 if (dev == sitn->fb_tunnel_dev) {
875 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
879 t = ipip6_tunnel_locate(net, &p, 0);
882 t = netdev_priv(dev);
885 if (cmd == SIOCGETTUNNEL) {
886 memcpy(&p, &t->parms, sizeof(p));
887 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
890 #ifdef CONFIG_IPV6_SIT_6RD
892 ipv6_addr_copy(&ip6rd.prefix, &t->ip6rd.prefix);
893 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
894 ip6rd.prefixlen = t->ip6rd.prefixlen;
895 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
896 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
907 if (!capable(CAP_NET_ADMIN))
911 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
915 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
916 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
919 p.iph.frag_off |= htons(IP_DF);
921 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
923 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
930 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
931 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
935 t = netdev_priv(dev);
936 ipip6_tunnel_unlink(sitn, t);
937 t->parms.iph.saddr = p.iph.saddr;
938 t->parms.iph.daddr = p.iph.daddr;
939 memcpy(dev->dev_addr, &p.iph.saddr, 4);
940 memcpy(dev->broadcast, &p.iph.daddr, 4);
941 ipip6_tunnel_link(sitn, t);
942 netdev_state_change(dev);
948 if (cmd == SIOCCHGTUNNEL) {
949 t->parms.iph.ttl = p.iph.ttl;
950 t->parms.iph.tos = p.iph.tos;
951 if (t->parms.link != p.link) {
952 t->parms.link = p.link;
953 ipip6_tunnel_bind_dev(dev);
954 netdev_state_change(dev);
957 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
960 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
965 if (!capable(CAP_NET_ADMIN))
968 if (dev == sitn->fb_tunnel_dev) {
970 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
973 if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
976 if (t == netdev_priv(sitn->fb_tunnel_dev))
980 unregister_netdevice(dev);
986 if (dev == sitn->fb_tunnel_dev)
989 if (!(t = netdev_priv(dev)))
991 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
998 if (!capable(CAP_NET_ADMIN))
1001 if (dev == sitn->fb_tunnel_dev)
1004 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1007 if (!(t = netdev_priv(dev)))
1012 err = ipip6_tunnel_del_prl(t, &prl);
1016 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1019 netdev_state_change(dev);
1022 #ifdef CONFIG_IPV6_SIT_6RD
1027 if (!capable(CAP_NET_ADMIN))
1031 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1035 t = netdev_priv(dev);
1037 if (cmd != SIOCDEL6RD) {
1038 struct in6_addr prefix;
1039 __be32 relay_prefix;
1042 if (ip6rd.relay_prefixlen > 32 ||
1043 ip6rd.prefixlen + (32 - ip6rd.relay_prefixlen) > 64)
1046 ipv6_addr_prefix(&prefix, &ip6rd.prefix,
1048 if (!ipv6_addr_equal(&prefix, &ip6rd.prefix))
1050 if (ip6rd.relay_prefixlen)
1051 relay_prefix = ip6rd.relay_prefix &
1052 htonl(0xffffffffUL <<
1053 (32 - ip6rd.relay_prefixlen));
1056 if (relay_prefix != ip6rd.relay_prefix)
1059 ipv6_addr_copy(&t->ip6rd.prefix, &prefix);
1060 t->ip6rd.relay_prefix = relay_prefix;
1061 t->ip6rd.prefixlen = ip6rd.prefixlen;
1062 t->ip6rd.relay_prefixlen = ip6rd.relay_prefixlen;
1064 ipip6_tunnel_clone_6rd(dev, sitn);
1078 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1080 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1086 static const struct net_device_ops ipip6_netdev_ops = {
1087 .ndo_uninit = ipip6_tunnel_uninit,
1088 .ndo_start_xmit = ipip6_tunnel_xmit,
1089 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1090 .ndo_change_mtu = ipip6_tunnel_change_mtu,
1093 static void ipip6_tunnel_setup(struct net_device *dev)
1095 dev->netdev_ops = &ipip6_netdev_ops;
1096 dev->destructor = free_netdev;
1098 dev->type = ARPHRD_SIT;
1099 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
1100 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr);
1101 dev->flags = IFF_NOARP;
1102 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1105 dev->features |= NETIF_F_NETNS_LOCAL;
1108 static void ipip6_tunnel_init(struct net_device *dev)
1110 struct ip_tunnel *tunnel = netdev_priv(dev);
1113 strcpy(tunnel->parms.name, dev->name);
1115 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1116 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1118 ipip6_tunnel_bind_dev(dev);
1121 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1123 struct ip_tunnel *tunnel = netdev_priv(dev);
1124 struct iphdr *iph = &tunnel->parms.iph;
1125 struct net *net = dev_net(dev);
1126 struct sit_net *sitn = net_generic(net, sit_net_id);
1129 strcpy(tunnel->parms.name, dev->name);
1132 iph->protocol = IPPROTO_IPV6;
1137 sitn->tunnels_wc[0] = tunnel;
1140 static struct xfrm_tunnel sit_handler = {
1141 .handler = ipip6_rcv,
1142 .err_handler = ipip6_err,
1146 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1150 for (prio = 1; prio < 4; prio++) {
1152 for (h = 0; h < HASH_SIZE; h++) {
1153 struct ip_tunnel *t = sitn->tunnels[prio][h];
1156 unregister_netdevice_queue(t->dev, head);
1163 static int __net_init sit_init_net(struct net *net)
1165 struct sit_net *sitn = net_generic(net, sit_net_id);
1168 sitn->tunnels[0] = sitn->tunnels_wc;
1169 sitn->tunnels[1] = sitn->tunnels_l;
1170 sitn->tunnels[2] = sitn->tunnels_r;
1171 sitn->tunnels[3] = sitn->tunnels_r_l;
1173 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1174 ipip6_tunnel_setup);
1175 if (!sitn->fb_tunnel_dev) {
1179 dev_net_set(sitn->fb_tunnel_dev, net);
1181 ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1182 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1184 if ((err = register_netdev(sitn->fb_tunnel_dev)))
1190 dev_put(sitn->fb_tunnel_dev);
1191 free_netdev(sitn->fb_tunnel_dev);
1196 static void __net_exit sit_exit_net(struct net *net)
1198 struct sit_net *sitn = net_generic(net, sit_net_id);
1202 sit_destroy_tunnels(sitn, &list);
1203 unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1204 unregister_netdevice_many(&list);
1208 static struct pernet_operations sit_net_ops = {
1209 .init = sit_init_net,
1210 .exit = sit_exit_net,
1212 .size = sizeof(struct sit_net),
1215 static void __exit sit_cleanup(void)
1217 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1219 unregister_pernet_device(&sit_net_ops);
1220 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1223 static int __init sit_init(void)
1227 printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1229 err = register_pernet_device(&sit_net_ops);
1232 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1234 unregister_pernet_device(&sit_net_ops);
1235 printk(KERN_INFO "sit init: Can't add protocol\n");
1240 module_init(sit_init);
1241 module_exit(sit_cleanup);
1242 MODULE_LICENSE("GPL");
1243 MODULE_ALIAS("sit0");