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 <linux/slab.h>
32 #include <asm/uaccess.h>
33 #include <linux/init.h>
34 #include <linux/netfilter_ipv4.h>
35 #include <linux/if_ether.h>
41 #include <net/protocol.h>
42 #include <net/transp_v6.h>
43 #include <net/ip6_fib.h>
44 #include <net/ip6_route.h>
45 #include <net/ndisc.h>
46 #include <net/addrconf.h>
51 #include <net/inet_ecn.h>
53 #include <net/dsfield.h>
54 #include <net/net_namespace.h>
55 #include <net/netns/generic.h>
58 This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
60 For comments look at net/ipv4/ip_gre.c --ANK
64 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
66 static void ipip6_tunnel_init(struct net_device *dev);
67 static void ipip6_tunnel_setup(struct net_device *dev);
69 static int sit_net_id __read_mostly;
71 struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
72 struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
73 struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
74 struct ip_tunnel __rcu *tunnels_wc[1];
75 struct ip_tunnel __rcu **tunnels[4];
77 struct net_device *fb_tunnel_dev;
81 * Locking : hash tables are protected by RCU and RTNL
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 int h0 = HASH(remote);
94 unsigned int 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 __rcu **__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 __rcu **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 __rcu **tp;
151 struct ip_tunnel *iter;
153 for (tp = ipip6_bucket(sitn, t);
154 (iter = rtnl_dereference(*tp)) != NULL;
157 rcu_assign_pointer(*tp, t->next);
163 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
165 struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
167 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
168 rcu_assign_pointer(*tp, t);
171 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
173 #ifdef CONFIG_IPV6_SIT_6RD
174 struct ip_tunnel *t = netdev_priv(dev);
176 if (t->dev == sitn->fb_tunnel_dev) {
177 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
178 t->ip6rd.relay_prefix = 0;
179 t->ip6rd.prefixlen = 16;
180 t->ip6rd.relay_prefixlen = 0;
182 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
183 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
188 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
189 struct ip_tunnel_parm *parms, int create)
191 __be32 remote = parms->iph.daddr;
192 __be32 local = parms->iph.saddr;
193 struct ip_tunnel *t, *nt;
194 struct ip_tunnel __rcu **tp;
195 struct net_device *dev;
197 struct sit_net *sitn = net_generic(net, sit_net_id);
199 for (tp = __ipip6_bucket(sitn, parms);
200 (t = rtnl_dereference(*tp)) != NULL;
202 if (local == t->parms.iph.saddr &&
203 remote == t->parms.iph.daddr &&
204 parms->link == t->parms.link) {
215 strlcpy(name, parms->name, IFNAMSIZ);
217 sprintf(name, "sit%%d");
219 dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
223 dev_net_set(dev, net);
225 if (strchr(name, '%')) {
226 if (dev_alloc_name(dev, name) < 0)
230 nt = netdev_priv(dev);
233 ipip6_tunnel_init(dev);
234 ipip6_tunnel_clone_6rd(dev, sitn);
236 if (parms->i_flags & SIT_ISATAP)
237 dev->priv_flags |= IFF_ISATAP;
239 if (register_netdevice(dev) < 0)
244 ipip6_tunnel_link(sitn, nt);
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))
344 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(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);
375 static void prl_entry_destroy_rcu(struct rcu_head *head)
377 kfree(container_of(head, struct ip_tunnel_prl_entry, rcu_head));
380 static void prl_list_destroy_rcu(struct rcu_head *head)
382 struct ip_tunnel_prl_entry *p, *n;
384 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
393 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
395 struct ip_tunnel_prl_entry *x, **p;
400 if (a && a->addr != htonl(INADDR_ANY)) {
401 for (p = &t->prl; *p; p = &(*p)->next) {
402 if ((*p)->addr == a->addr) {
405 call_rcu(&x->rcu_head, prl_entry_destroy_rcu);
415 call_rcu(&x->rcu_head, prl_list_destroy_rcu);
424 isatap_chksrc(struct sk_buff *skb, struct iphdr *iph, struct ip_tunnel *t)
426 struct ip_tunnel_prl_entry *p;
430 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
432 if (p->flags & PRL_DEFAULT)
433 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
435 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
437 struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
438 if (ipv6_addr_is_isatap(addr6) &&
439 (addr6->s6_addr32[3] == iph->saddr) &&
440 ipv6_chk_prefix(addr6, t->dev))
441 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
449 static void ipip6_tunnel_uninit(struct net_device *dev)
451 struct net *net = dev_net(dev);
452 struct sit_net *sitn = net_generic(net, sit_net_id);
454 if (dev == sitn->fb_tunnel_dev) {
455 rcu_assign_pointer(sitn->tunnels_wc[0], NULL);
457 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
458 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
464 static int ipip6_err(struct sk_buff *skb, u32 info)
467 /* All the routers (except for Linux) return only
468 8 bytes of packet payload. It means, that precise relaying of
469 ICMP in the real Internet is absolutely infeasible.
471 struct iphdr *iph = (struct iphdr*)skb->data;
472 const int type = icmp_hdr(skb)->type;
473 const int code = icmp_hdr(skb)->code;
479 case ICMP_PARAMETERPROB:
482 case ICMP_DEST_UNREACH:
485 case ICMP_PORT_UNREACH:
486 /* Impossible event. */
488 case ICMP_FRAG_NEEDED:
489 /* Soft state for pmtu is maintained by IP core. */
492 /* All others are translated to HOST_UNREACH.
493 rfc2003 contains "deep thoughts" about NET_UNREACH,
494 I believe they are just ether pollution. --ANK
499 case ICMP_TIME_EXCEEDED:
500 if (code != ICMP_EXC_TTL)
508 t = ipip6_tunnel_lookup(dev_net(skb->dev),
512 if (t == NULL || t->parms.iph.daddr == 0)
516 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
519 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
523 t->err_time = jiffies;
529 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
531 if (INET_ECN_is_ce(iph->tos))
532 IP6_ECN_set_ce(ipv6_hdr(skb));
535 static int ipip6_rcv(struct sk_buff *skb)
538 struct ip_tunnel *tunnel;
540 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
546 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
547 iph->saddr, iph->daddr);
548 if (tunnel != NULL) {
550 skb->mac_header = skb->network_header;
551 skb_reset_network_header(skb);
552 IPCB(skb)->flags = 0;
553 skb->protocol = htons(ETH_P_IPV6);
554 skb->pkt_type = PACKET_HOST;
556 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
557 !isatap_chksrc(skb, iph, tunnel)) {
558 tunnel->dev->stats.rx_errors++;
564 skb_tunnel_rx(skb, tunnel->dev);
566 ipip6_ecn_decapsulate(iph, skb);
568 if (netif_rx(skb) == NET_RX_DROP)
569 tunnel->dev->stats.rx_dropped++;
575 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
583 * Returns the embedded IPv4 address if the IPv6 address
584 * comes from 6rd / 6to4 (RFC 3056) addr space.
587 __be32 try_6rd(struct in6_addr *v6dst, struct ip_tunnel *tunnel)
591 #ifdef CONFIG_IPV6_SIT_6RD
592 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
593 tunnel->ip6rd.prefixlen)) {
594 unsigned int pbw0, pbi0;
598 pbw0 = tunnel->ip6rd.prefixlen >> 5;
599 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
601 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
602 tunnel->ip6rd.relay_prefixlen;
604 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
606 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
609 dst = tunnel->ip6rd.relay_prefix | htonl(d);
612 if (v6dst->s6_addr16[0] == htons(0x2002)) {
613 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
614 memcpy(&dst, &v6dst->s6_addr16[1], 4);
621 * This function assumes it is being called from dev_queue_xmit()
622 * and that skb is filled properly by that function.
625 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
626 struct net_device *dev)
628 struct ip_tunnel *tunnel = netdev_priv(dev);
629 struct net_device_stats *stats = &dev->stats;
630 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
631 struct iphdr *tiph = &tunnel->parms.iph;
632 struct ipv6hdr *iph6 = ipv6_hdr(skb);
633 u8 tos = tunnel->parms.iph.tos;
634 __be16 df = tiph->frag_off;
635 struct rtable *rt; /* Route to the other host */
636 struct net_device *tdev; /* Device to other host */
637 struct iphdr *iph; /* Our new IP header */
638 unsigned int max_headroom; /* The extra header space needed */
639 __be32 dst = tiph->daddr;
641 struct in6_addr *addr6;
644 if (skb->protocol != htons(ETH_P_IPV6))
647 /* ISATAP (RFC4214) - must come before 6to4 */
648 if (dev->priv_flags & IFF_ISATAP) {
649 struct neighbour *neigh = NULL;
652 neigh = skb_dst(skb)->neighbour;
656 printk(KERN_DEBUG "sit: nexthop == NULL\n");
660 addr6 = (struct in6_addr*)&neigh->primary_key;
661 addr_type = ipv6_addr_type(addr6);
663 if ((addr_type & IPV6_ADDR_UNICAST) &&
664 ipv6_addr_is_isatap(addr6))
665 dst = addr6->s6_addr32[3];
671 dst = try_6rd(&iph6->daddr, tunnel);
674 struct neighbour *neigh = NULL;
677 neigh = skb_dst(skb)->neighbour;
681 printk(KERN_DEBUG "sit: nexthop == NULL\n");
685 addr6 = (struct in6_addr*)&neigh->primary_key;
686 addr_type = ipv6_addr_type(addr6);
688 if (addr_type == IPV6_ADDR_ANY) {
689 addr6 = &ipv6_hdr(skb)->daddr;
690 addr_type = ipv6_addr_type(addr6);
693 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
696 dst = addr6->s6_addr32[3];
700 struct flowi fl = { .nl_u = { .ip4_u =
702 .saddr = tiph->saddr,
703 .tos = RT_TOS(tos) } },
704 .oif = tunnel->parms.link,
705 .proto = IPPROTO_IPV6 };
706 if (ip_route_output_key(dev_net(dev), &rt, &fl)) {
707 stats->tx_carrier_errors++;
711 if (rt->rt_type != RTN_UNICAST) {
713 stats->tx_carrier_errors++;
725 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
733 if (mtu < IPV6_MIN_MTU) {
738 if (tunnel->parms.iph.daddr && skb_dst(skb))
739 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
741 if (skb->len > mtu) {
742 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
748 if (tunnel->err_count > 0) {
749 if (time_before(jiffies,
750 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
752 dst_link_failure(skb);
754 tunnel->err_count = 0;
758 * Okay, now see if we can stuff it in the buffer as-is.
760 max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
762 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
763 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
764 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
772 skb_set_owner_w(new_skb, skb->sk);
775 iph6 = ipv6_hdr(skb);
778 skb->transport_header = skb->network_header;
779 skb_push(skb, sizeof(struct iphdr));
780 skb_reset_network_header(skb);
781 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
782 IPCB(skb)->flags = 0;
784 skb_dst_set(skb, &rt->dst);
787 * Push down and install the IPIP header.
792 iph->ihl = sizeof(struct iphdr)>>2;
794 iph->protocol = IPPROTO_IPV6;
795 iph->tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
796 iph->daddr = rt->rt_dst;
797 iph->saddr = rt->rt_src;
799 if ((iph->ttl = tiph->ttl) == 0)
800 iph->ttl = iph6->hop_limit;
808 dst_link_failure(skb);
815 static void ipip6_tunnel_bind_dev(struct net_device *dev)
817 struct net_device *tdev = NULL;
818 struct ip_tunnel *tunnel;
821 tunnel = netdev_priv(dev);
822 iph = &tunnel->parms.iph;
825 struct flowi fl = { .nl_u = { .ip4_u =
826 { .daddr = iph->daddr,
828 .tos = RT_TOS(iph->tos) } },
829 .oif = tunnel->parms.link,
830 .proto = IPPROTO_IPV6 };
832 if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
836 dev->flags |= IFF_POINTOPOINT;
839 if (!tdev && tunnel->parms.link)
840 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
843 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
844 dev->mtu = tdev->mtu - sizeof(struct iphdr);
845 if (dev->mtu < IPV6_MIN_MTU)
846 dev->mtu = IPV6_MIN_MTU;
848 dev->iflink = tunnel->parms.link;
852 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
855 struct ip_tunnel_parm p;
856 struct ip_tunnel_prl prl;
858 struct net *net = dev_net(dev);
859 struct sit_net *sitn = net_generic(net, sit_net_id);
860 #ifdef CONFIG_IPV6_SIT_6RD
861 struct ip_tunnel_6rd ip6rd;
866 #ifdef CONFIG_IPV6_SIT_6RD
870 if (dev == sitn->fb_tunnel_dev) {
871 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
875 t = ipip6_tunnel_locate(net, &p, 0);
878 t = netdev_priv(dev);
881 if (cmd == SIOCGETTUNNEL) {
882 memcpy(&p, &t->parms, sizeof(p));
883 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
886 #ifdef CONFIG_IPV6_SIT_6RD
888 ipv6_addr_copy(&ip6rd.prefix, &t->ip6rd.prefix);
889 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
890 ip6rd.prefixlen = t->ip6rd.prefixlen;
891 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
892 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
903 if (!capable(CAP_NET_ADMIN))
907 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
911 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
912 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
915 p.iph.frag_off |= htons(IP_DF);
917 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
919 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
926 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
927 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
931 t = netdev_priv(dev);
932 ipip6_tunnel_unlink(sitn, t);
933 t->parms.iph.saddr = p.iph.saddr;
934 t->parms.iph.daddr = p.iph.daddr;
935 memcpy(dev->dev_addr, &p.iph.saddr, 4);
936 memcpy(dev->broadcast, &p.iph.daddr, 4);
937 ipip6_tunnel_link(sitn, t);
938 netdev_state_change(dev);
944 if (cmd == SIOCCHGTUNNEL) {
945 t->parms.iph.ttl = p.iph.ttl;
946 t->parms.iph.tos = p.iph.tos;
947 if (t->parms.link != p.link) {
948 t->parms.link = p.link;
949 ipip6_tunnel_bind_dev(dev);
950 netdev_state_change(dev);
953 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
956 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
961 if (!capable(CAP_NET_ADMIN))
964 if (dev == sitn->fb_tunnel_dev) {
966 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
969 if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
972 if (t == netdev_priv(sitn->fb_tunnel_dev))
976 unregister_netdevice(dev);
982 if (dev == sitn->fb_tunnel_dev)
985 if (!(t = netdev_priv(dev)))
987 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
994 if (!capable(CAP_NET_ADMIN))
997 if (dev == sitn->fb_tunnel_dev)
1000 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1003 if (!(t = netdev_priv(dev)))
1008 err = ipip6_tunnel_del_prl(t, &prl);
1012 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1015 netdev_state_change(dev);
1018 #ifdef CONFIG_IPV6_SIT_6RD
1023 if (!capable(CAP_NET_ADMIN))
1027 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1031 t = netdev_priv(dev);
1033 if (cmd != SIOCDEL6RD) {
1034 struct in6_addr prefix;
1035 __be32 relay_prefix;
1038 if (ip6rd.relay_prefixlen > 32 ||
1039 ip6rd.prefixlen + (32 - ip6rd.relay_prefixlen) > 64)
1042 ipv6_addr_prefix(&prefix, &ip6rd.prefix,
1044 if (!ipv6_addr_equal(&prefix, &ip6rd.prefix))
1046 if (ip6rd.relay_prefixlen)
1047 relay_prefix = ip6rd.relay_prefix &
1048 htonl(0xffffffffUL <<
1049 (32 - ip6rd.relay_prefixlen));
1052 if (relay_prefix != ip6rd.relay_prefix)
1055 ipv6_addr_copy(&t->ip6rd.prefix, &prefix);
1056 t->ip6rd.relay_prefix = relay_prefix;
1057 t->ip6rd.prefixlen = ip6rd.prefixlen;
1058 t->ip6rd.relay_prefixlen = ip6rd.relay_prefixlen;
1060 ipip6_tunnel_clone_6rd(dev, sitn);
1074 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1076 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1082 static const struct net_device_ops ipip6_netdev_ops = {
1083 .ndo_uninit = ipip6_tunnel_uninit,
1084 .ndo_start_xmit = ipip6_tunnel_xmit,
1085 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1086 .ndo_change_mtu = ipip6_tunnel_change_mtu,
1089 static void ipip6_tunnel_setup(struct net_device *dev)
1091 dev->netdev_ops = &ipip6_netdev_ops;
1092 dev->destructor = free_netdev;
1094 dev->type = ARPHRD_SIT;
1095 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
1096 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr);
1097 dev->flags = IFF_NOARP;
1098 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1101 dev->features |= NETIF_F_NETNS_LOCAL;
1104 static void ipip6_tunnel_init(struct net_device *dev)
1106 struct ip_tunnel *tunnel = netdev_priv(dev);
1109 strcpy(tunnel->parms.name, dev->name);
1111 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1112 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1114 ipip6_tunnel_bind_dev(dev);
1117 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1119 struct ip_tunnel *tunnel = netdev_priv(dev);
1120 struct iphdr *iph = &tunnel->parms.iph;
1121 struct net *net = dev_net(dev);
1122 struct sit_net *sitn = net_generic(net, sit_net_id);
1125 strcpy(tunnel->parms.name, dev->name);
1128 iph->protocol = IPPROTO_IPV6;
1133 sitn->tunnels_wc[0] = tunnel;
1136 static struct xfrm_tunnel sit_handler __read_mostly = {
1137 .handler = ipip6_rcv,
1138 .err_handler = ipip6_err,
1142 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1146 for (prio = 1; prio < 4; prio++) {
1148 for (h = 0; h < HASH_SIZE; h++) {
1149 struct ip_tunnel *t = sitn->tunnels[prio][h];
1152 unregister_netdevice_queue(t->dev, head);
1159 static int __net_init sit_init_net(struct net *net)
1161 struct sit_net *sitn = net_generic(net, sit_net_id);
1164 sitn->tunnels[0] = sitn->tunnels_wc;
1165 sitn->tunnels[1] = sitn->tunnels_l;
1166 sitn->tunnels[2] = sitn->tunnels_r;
1167 sitn->tunnels[3] = sitn->tunnels_r_l;
1169 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1170 ipip6_tunnel_setup);
1171 if (!sitn->fb_tunnel_dev) {
1175 dev_net_set(sitn->fb_tunnel_dev, net);
1177 ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1178 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1180 if ((err = register_netdev(sitn->fb_tunnel_dev)))
1186 dev_put(sitn->fb_tunnel_dev);
1187 free_netdev(sitn->fb_tunnel_dev);
1192 static void __net_exit sit_exit_net(struct net *net)
1194 struct sit_net *sitn = net_generic(net, sit_net_id);
1198 sit_destroy_tunnels(sitn, &list);
1199 unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1200 unregister_netdevice_many(&list);
1204 static struct pernet_operations sit_net_ops = {
1205 .init = sit_init_net,
1206 .exit = sit_exit_net,
1208 .size = sizeof(struct sit_net),
1211 static void __exit sit_cleanup(void)
1213 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1215 unregister_pernet_device(&sit_net_ops);
1216 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1219 static int __init sit_init(void)
1223 printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1225 err = register_pernet_device(&sit_net_ops);
1228 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1230 unregister_pernet_device(&sit_net_ops);
1231 printk(KERN_INFO "sit init: Can't add protocol\n");
1236 module_init(sit_init);
1237 module_exit(sit_cleanup);
1238 MODULE_LICENSE("GPL");
1239 MODULE_ALIAS("sit0");