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>
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 int ipip6_tunnel_init(struct net_device *dev);
69 static void ipip6_tunnel_setup(struct net_device *dev);
70 static void ipip6_dev_free(struct net_device *dev);
72 static int sit_net_id __read_mostly;
74 struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
75 struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
76 struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
77 struct ip_tunnel __rcu *tunnels_wc[1];
78 struct ip_tunnel __rcu **tunnels[4];
80 struct net_device *fb_tunnel_dev;
84 * Locking : hash tables are protected by RCU and RTNL
87 #define for_each_ip_tunnel_rcu(start) \
88 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
90 /* often modified stats are per cpu, other are shared (netdev->stats) */
96 struct u64_stats_sync syncp;
99 static struct rtnl_link_stats64 *ipip6_get_stats64(struct net_device *dev,
100 struct rtnl_link_stats64 *tot)
104 for_each_possible_cpu(i) {
105 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
106 u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
110 start = u64_stats_fetch_begin_bh(&tstats->syncp);
111 rx_packets = tstats->rx_packets;
112 tx_packets = tstats->tx_packets;
113 rx_bytes = tstats->rx_bytes;
114 tx_bytes = tstats->tx_bytes;
115 } while (u64_stats_fetch_retry_bh(&tstats->syncp, start));
117 tot->rx_packets += rx_packets;
118 tot->tx_packets += tx_packets;
119 tot->rx_bytes += rx_bytes;
120 tot->tx_bytes += tx_bytes;
123 tot->rx_errors = dev->stats.rx_errors;
124 tot->tx_fifo_errors = dev->stats.tx_fifo_errors;
125 tot->tx_carrier_errors = dev->stats.tx_carrier_errors;
126 tot->tx_dropped = dev->stats.tx_dropped;
127 tot->tx_aborted_errors = dev->stats.tx_aborted_errors;
128 tot->tx_errors = dev->stats.tx_errors;
134 * Must be invoked with rcu_read_lock
136 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
137 struct net_device *dev, __be32 remote, __be32 local)
139 unsigned int h0 = HASH(remote);
140 unsigned int h1 = HASH(local);
142 struct sit_net *sitn = net_generic(net, sit_net_id);
144 for_each_ip_tunnel_rcu(sitn->tunnels_r_l[h0 ^ h1]) {
145 if (local == t->parms.iph.saddr &&
146 remote == t->parms.iph.daddr &&
147 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
148 (t->dev->flags & IFF_UP))
151 for_each_ip_tunnel_rcu(sitn->tunnels_r[h0]) {
152 if (remote == t->parms.iph.daddr &&
153 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
154 (t->dev->flags & IFF_UP))
157 for_each_ip_tunnel_rcu(sitn->tunnels_l[h1]) {
158 if (local == t->parms.iph.saddr &&
159 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
160 (t->dev->flags & IFF_UP))
163 t = rcu_dereference(sitn->tunnels_wc[0]);
164 if ((t != NULL) && (t->dev->flags & IFF_UP))
169 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
170 struct ip_tunnel_parm *parms)
172 __be32 remote = parms->iph.daddr;
173 __be32 local = parms->iph.saddr;
185 return &sitn->tunnels[prio][h];
188 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
191 return __ipip6_bucket(sitn, &t->parms);
194 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
196 struct ip_tunnel __rcu **tp;
197 struct ip_tunnel *iter;
199 for (tp = ipip6_bucket(sitn, t);
200 (iter = rtnl_dereference(*tp)) != NULL;
203 rcu_assign_pointer(*tp, t->next);
209 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
211 struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
213 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
214 rcu_assign_pointer(*tp, t);
217 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
219 #ifdef CONFIG_IPV6_SIT_6RD
220 struct ip_tunnel *t = netdev_priv(dev);
222 if (t->dev == sitn->fb_tunnel_dev) {
223 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
224 t->ip6rd.relay_prefix = 0;
225 t->ip6rd.prefixlen = 16;
226 t->ip6rd.relay_prefixlen = 0;
228 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
229 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
234 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
235 struct ip_tunnel_parm *parms, int create)
237 __be32 remote = parms->iph.daddr;
238 __be32 local = parms->iph.saddr;
239 struct ip_tunnel *t, *nt;
240 struct ip_tunnel __rcu **tp;
241 struct net_device *dev;
243 struct sit_net *sitn = net_generic(net, sit_net_id);
245 for (tp = __ipip6_bucket(sitn, parms);
246 (t = rtnl_dereference(*tp)) != NULL;
248 if (local == t->parms.iph.saddr &&
249 remote == t->parms.iph.daddr &&
250 parms->link == t->parms.link) {
261 strlcpy(name, parms->name, IFNAMSIZ);
263 strcpy(name, "sit%d");
265 dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
269 dev_net_set(dev, net);
271 nt = netdev_priv(dev);
274 if (ipip6_tunnel_init(dev) < 0)
276 ipip6_tunnel_clone_6rd(dev, sitn);
278 if (parms->i_flags & SIT_ISATAP)
279 dev->priv_flags |= IFF_ISATAP;
281 if (register_netdevice(dev) < 0)
284 strcpy(nt->parms.name, dev->name);
288 ipip6_tunnel_link(sitn, nt);
297 #define for_each_prl_rcu(start) \
298 for (prl = rcu_dereference(start); \
300 prl = rcu_dereference(prl->next))
302 static struct ip_tunnel_prl_entry *
303 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
305 struct ip_tunnel_prl_entry *prl;
307 for_each_prl_rcu(t->prl)
308 if (prl->addr == addr)
314 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
315 struct ip_tunnel_prl __user *a)
317 struct ip_tunnel_prl kprl, *kp;
318 struct ip_tunnel_prl_entry *prl;
319 unsigned int cmax, c = 0, ca, len;
322 if (copy_from_user(&kprl, a, sizeof(kprl)))
324 cmax = kprl.datalen / sizeof(kprl);
325 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
328 /* For simple GET or for root users,
329 * we try harder to allocate.
331 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
332 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
337 ca = t->prl_count < cmax ? t->prl_count : cmax;
340 /* We don't try hard to allocate much memory for
342 * For root users, retry allocating enough memory for
345 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
353 for_each_prl_rcu(t->prl) {
356 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
358 kp[c].addr = prl->addr;
359 kp[c].flags = prl->flags;
361 if (kprl.addr != htonl(INADDR_ANY))
367 len = sizeof(*kp) * c;
369 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
378 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
380 struct ip_tunnel_prl_entry *p;
383 if (a->addr == htonl(INADDR_ANY))
388 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
389 if (p->addr == a->addr) {
404 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
414 rcu_assign_pointer(t->prl, p);
419 static void prl_list_destroy_rcu(struct rcu_head *head)
421 struct ip_tunnel_prl_entry *p, *n;
423 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
425 n = rcu_dereference_protected(p->next, 1);
432 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
434 struct ip_tunnel_prl_entry *x;
435 struct ip_tunnel_prl_entry __rcu **p;
440 if (a && a->addr != htonl(INADDR_ANY)) {
442 (x = rtnl_dereference(*p)) != NULL;
444 if (x->addr == a->addr) {
446 kfree_rcu(x, rcu_head);
453 x = rtnl_dereference(t->prl);
456 call_rcu(&x->rcu_head, prl_list_destroy_rcu);
465 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
467 struct ip_tunnel_prl_entry *p;
471 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
473 if (p->flags & PRL_DEFAULT)
474 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
476 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
478 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
480 if (ipv6_addr_is_isatap(addr6) &&
481 (addr6->s6_addr32[3] == iph->saddr) &&
482 ipv6_chk_prefix(addr6, t->dev))
483 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
491 static void ipip6_tunnel_uninit(struct net_device *dev)
493 struct net *net = dev_net(dev);
494 struct sit_net *sitn = net_generic(net, sit_net_id);
496 if (dev == sitn->fb_tunnel_dev) {
497 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
499 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
500 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
506 static int ipip6_err(struct sk_buff *skb, u32 info)
509 /* All the routers (except for Linux) return only
510 8 bytes of packet payload. It means, that precise relaying of
511 ICMP in the real Internet is absolutely infeasible.
513 const struct iphdr *iph = (const struct iphdr *)skb->data;
514 const int type = icmp_hdr(skb)->type;
515 const int code = icmp_hdr(skb)->code;
521 case ICMP_PARAMETERPROB:
524 case ICMP_DEST_UNREACH:
527 case ICMP_PORT_UNREACH:
528 /* Impossible event. */
531 /* All others are translated to HOST_UNREACH.
532 rfc2003 contains "deep thoughts" about NET_UNREACH,
533 I believe they are just ether pollution. --ANK
538 case ICMP_TIME_EXCEEDED:
539 if (code != ICMP_EXC_TTL)
549 t = ipip6_tunnel_lookup(dev_net(skb->dev),
556 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
557 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
558 t->dev->ifindex, 0, IPPROTO_IPV6, 0);
562 if (type == ICMP_REDIRECT) {
563 ipv4_redirect(skb, dev_net(skb->dev), t->dev->ifindex, 0,
569 if (t->parms.iph.daddr == 0)
573 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
576 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
580 t->err_time = jiffies;
586 static inline void ipip6_ecn_decapsulate(const struct iphdr *iph, struct sk_buff *skb)
588 if (INET_ECN_is_ce(iph->tos))
589 IP6_ECN_set_ce(ipv6_hdr(skb));
592 static int ipip6_rcv(struct sk_buff *skb)
594 const struct iphdr *iph;
595 struct ip_tunnel *tunnel;
597 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
603 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
604 iph->saddr, iph->daddr);
605 if (tunnel != NULL) {
606 struct pcpu_tstats *tstats;
609 skb->mac_header = skb->network_header;
610 skb_reset_network_header(skb);
611 IPCB(skb)->flags = 0;
612 skb->protocol = htons(ETH_P_IPV6);
613 skb->pkt_type = PACKET_HOST;
615 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
616 !isatap_chksrc(skb, iph, tunnel)) {
617 tunnel->dev->stats.rx_errors++;
623 tstats = this_cpu_ptr(tunnel->dev->tstats);
624 tstats->rx_packets++;
625 tstats->rx_bytes += skb->len;
627 __skb_tunnel_rx(skb, tunnel->dev);
629 ipip6_ecn_decapsulate(iph, skb);
637 /* no tunnel matched, let upstream know, ipsec may handle it */
646 * Returns the embedded IPv4 address if the IPv6 address
647 * comes from 6rd / 6to4 (RFC 3056) addr space.
650 __be32 try_6rd(const struct in6_addr *v6dst, struct ip_tunnel *tunnel)
654 #ifdef CONFIG_IPV6_SIT_6RD
655 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
656 tunnel->ip6rd.prefixlen)) {
657 unsigned int pbw0, pbi0;
661 pbw0 = tunnel->ip6rd.prefixlen >> 5;
662 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
664 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
665 tunnel->ip6rd.relay_prefixlen;
667 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
669 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
672 dst = tunnel->ip6rd.relay_prefix | htonl(d);
675 if (v6dst->s6_addr16[0] == htons(0x2002)) {
676 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
677 memcpy(&dst, &v6dst->s6_addr16[1], 4);
684 * This function assumes it is being called from dev_queue_xmit()
685 * and that skb is filled properly by that function.
688 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
689 struct net_device *dev)
691 struct ip_tunnel *tunnel = netdev_priv(dev);
692 struct pcpu_tstats *tstats;
693 const struct iphdr *tiph = &tunnel->parms.iph;
694 const struct ipv6hdr *iph6 = ipv6_hdr(skb);
695 u8 tos = tunnel->parms.iph.tos;
696 __be16 df = tiph->frag_off;
697 struct rtable *rt; /* Route to the other host */
698 struct net_device *tdev; /* Device to other host */
699 struct iphdr *iph; /* Our new IP header */
700 unsigned int max_headroom; /* The extra header space needed */
701 __be32 dst = tiph->daddr;
704 const struct in6_addr *addr6;
707 if (skb->protocol != htons(ETH_P_IPV6))
711 tos = ipv6_get_dsfield(iph6);
713 /* ISATAP (RFC4214) - must come before 6to4 */
714 if (dev->priv_flags & IFF_ISATAP) {
715 struct neighbour *neigh = NULL;
716 bool do_tx_error = false;
719 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
722 net_dbg_ratelimited("sit: nexthop == NULL\n");
726 addr6 = (const struct in6_addr *)&neigh->primary_key;
727 addr_type = ipv6_addr_type(addr6);
729 if ((addr_type & IPV6_ADDR_UNICAST) &&
730 ipv6_addr_is_isatap(addr6))
731 dst = addr6->s6_addr32[3];
735 neigh_release(neigh);
741 dst = try_6rd(&iph6->daddr, tunnel);
744 struct neighbour *neigh = NULL;
745 bool do_tx_error = false;
748 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
751 net_dbg_ratelimited("sit: nexthop == NULL\n");
755 addr6 = (const struct in6_addr *)&neigh->primary_key;
756 addr_type = ipv6_addr_type(addr6);
758 if (addr_type == IPV6_ADDR_ANY) {
759 addr6 = &ipv6_hdr(skb)->daddr;
760 addr_type = ipv6_addr_type(addr6);
763 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
764 dst = addr6->s6_addr32[3];
768 neigh_release(neigh);
773 rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
776 IPPROTO_IPV6, RT_TOS(tos),
779 dev->stats.tx_carrier_errors++;
782 if (rt->rt_type != RTN_UNICAST) {
784 dev->stats.tx_carrier_errors++;
791 dev->stats.collisions++;
796 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
799 dev->stats.collisions++;
804 if (mtu < IPV6_MIN_MTU) {
809 if (tunnel->parms.iph.daddr && skb_dst(skb))
810 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
812 if (skb->len > mtu) {
813 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
819 if (tunnel->err_count > 0) {
820 if (time_before(jiffies,
821 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
823 dst_link_failure(skb);
825 tunnel->err_count = 0;
829 * Okay, now see if we can stuff it in the buffer as-is.
831 max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
833 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
834 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
835 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
838 dev->stats.tx_dropped++;
843 skb_set_owner_w(new_skb, skb->sk);
846 iph6 = ipv6_hdr(skb);
849 skb->transport_header = skb->network_header;
850 skb_push(skb, sizeof(struct iphdr));
851 skb_reset_network_header(skb);
852 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
853 IPCB(skb)->flags = 0;
855 skb_dst_set(skb, &rt->dst);
858 * Push down and install the IPIP header.
863 iph->ihl = sizeof(struct iphdr)>>2;
865 iph->protocol = IPPROTO_IPV6;
866 iph->tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
867 iph->daddr = fl4.daddr;
868 iph->saddr = fl4.saddr;
870 if ((iph->ttl = tiph->ttl) == 0)
871 iph->ttl = iph6->hop_limit;
874 tstats = this_cpu_ptr(dev->tstats);
875 __IPTUNNEL_XMIT(tstats, &dev->stats);
879 dst_link_failure(skb);
881 dev->stats.tx_errors++;
886 static void ipip6_tunnel_bind_dev(struct net_device *dev)
888 struct net_device *tdev = NULL;
889 struct ip_tunnel *tunnel;
890 const struct iphdr *iph;
893 tunnel = netdev_priv(dev);
894 iph = &tunnel->parms.iph;
897 struct rtable *rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
898 iph->daddr, iph->saddr,
908 dev->flags |= IFF_POINTOPOINT;
911 if (!tdev && tunnel->parms.link)
912 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
915 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
916 dev->mtu = tdev->mtu - sizeof(struct iphdr);
917 if (dev->mtu < IPV6_MIN_MTU)
918 dev->mtu = IPV6_MIN_MTU;
920 dev->iflink = tunnel->parms.link;
924 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
927 struct ip_tunnel_parm p;
928 struct ip_tunnel_prl prl;
930 struct net *net = dev_net(dev);
931 struct sit_net *sitn = net_generic(net, sit_net_id);
932 #ifdef CONFIG_IPV6_SIT_6RD
933 struct ip_tunnel_6rd ip6rd;
938 #ifdef CONFIG_IPV6_SIT_6RD
942 if (dev == sitn->fb_tunnel_dev) {
943 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
947 t = ipip6_tunnel_locate(net, &p, 0);
950 t = netdev_priv(dev);
953 if (cmd == SIOCGETTUNNEL) {
954 memcpy(&p, &t->parms, sizeof(p));
955 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
958 #ifdef CONFIG_IPV6_SIT_6RD
960 ip6rd.prefix = t->ip6rd.prefix;
961 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
962 ip6rd.prefixlen = t->ip6rd.prefixlen;
963 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
964 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
975 if (!capable(CAP_NET_ADMIN))
979 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
983 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
984 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
987 p.iph.frag_off |= htons(IP_DF);
989 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
991 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
998 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
999 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1003 t = netdev_priv(dev);
1004 ipip6_tunnel_unlink(sitn, t);
1006 t->parms.iph.saddr = p.iph.saddr;
1007 t->parms.iph.daddr = p.iph.daddr;
1008 memcpy(dev->dev_addr, &p.iph.saddr, 4);
1009 memcpy(dev->broadcast, &p.iph.daddr, 4);
1010 ipip6_tunnel_link(sitn, t);
1011 netdev_state_change(dev);
1017 if (cmd == SIOCCHGTUNNEL) {
1018 t->parms.iph.ttl = p.iph.ttl;
1019 t->parms.iph.tos = p.iph.tos;
1020 if (t->parms.link != p.link) {
1021 t->parms.link = p.link;
1022 ipip6_tunnel_bind_dev(dev);
1023 netdev_state_change(dev);
1026 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1029 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1034 if (!capable(CAP_NET_ADMIN))
1037 if (dev == sitn->fb_tunnel_dev) {
1039 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1042 if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1045 if (t == netdev_priv(sitn->fb_tunnel_dev))
1049 unregister_netdevice(dev);
1055 if (dev == sitn->fb_tunnel_dev)
1058 if (!(t = netdev_priv(dev)))
1060 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1067 if (!capable(CAP_NET_ADMIN))
1070 if (dev == sitn->fb_tunnel_dev)
1073 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1076 if (!(t = netdev_priv(dev)))
1081 err = ipip6_tunnel_del_prl(t, &prl);
1085 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1088 netdev_state_change(dev);
1091 #ifdef CONFIG_IPV6_SIT_6RD
1096 if (!capable(CAP_NET_ADMIN))
1100 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1104 t = netdev_priv(dev);
1106 if (cmd != SIOCDEL6RD) {
1107 struct in6_addr prefix;
1108 __be32 relay_prefix;
1111 if (ip6rd.relay_prefixlen > 32 ||
1112 ip6rd.prefixlen + (32 - ip6rd.relay_prefixlen) > 64)
1115 ipv6_addr_prefix(&prefix, &ip6rd.prefix,
1117 if (!ipv6_addr_equal(&prefix, &ip6rd.prefix))
1119 if (ip6rd.relay_prefixlen)
1120 relay_prefix = ip6rd.relay_prefix &
1121 htonl(0xffffffffUL <<
1122 (32 - ip6rd.relay_prefixlen));
1125 if (relay_prefix != ip6rd.relay_prefix)
1128 t->ip6rd.prefix = prefix;
1129 t->ip6rd.relay_prefix = relay_prefix;
1130 t->ip6rd.prefixlen = ip6rd.prefixlen;
1131 t->ip6rd.relay_prefixlen = ip6rd.relay_prefixlen;
1133 ipip6_tunnel_clone_6rd(dev, sitn);
1147 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1149 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1155 static const struct net_device_ops ipip6_netdev_ops = {
1156 .ndo_uninit = ipip6_tunnel_uninit,
1157 .ndo_start_xmit = ipip6_tunnel_xmit,
1158 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1159 .ndo_change_mtu = ipip6_tunnel_change_mtu,
1160 .ndo_get_stats64= ipip6_get_stats64,
1163 static void ipip6_dev_free(struct net_device *dev)
1165 free_percpu(dev->tstats);
1169 static void ipip6_tunnel_setup(struct net_device *dev)
1171 dev->netdev_ops = &ipip6_netdev_ops;
1172 dev->destructor = ipip6_dev_free;
1174 dev->type = ARPHRD_SIT;
1175 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
1176 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr);
1177 dev->flags = IFF_NOARP;
1178 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1181 dev->features |= NETIF_F_NETNS_LOCAL;
1182 dev->features |= NETIF_F_LLTX;
1185 static int ipip6_tunnel_init(struct net_device *dev)
1187 struct ip_tunnel *tunnel = netdev_priv(dev);
1191 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1192 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1194 ipip6_tunnel_bind_dev(dev);
1195 dev->tstats = alloc_percpu(struct pcpu_tstats);
1202 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1204 struct ip_tunnel *tunnel = netdev_priv(dev);
1205 struct iphdr *iph = &tunnel->parms.iph;
1206 struct net *net = dev_net(dev);
1207 struct sit_net *sitn = net_generic(net, sit_net_id);
1210 strcpy(tunnel->parms.name, dev->name);
1213 iph->protocol = IPPROTO_IPV6;
1217 dev->tstats = alloc_percpu(struct pcpu_tstats);
1221 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1225 static struct xfrm_tunnel sit_handler __read_mostly = {
1226 .handler = ipip6_rcv,
1227 .err_handler = ipip6_err,
1231 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1235 for (prio = 1; prio < 4; prio++) {
1237 for (h = 0; h < HASH_SIZE; h++) {
1238 struct ip_tunnel *t;
1240 t = rtnl_dereference(sitn->tunnels[prio][h]);
1242 unregister_netdevice_queue(t->dev, head);
1243 t = rtnl_dereference(t->next);
1249 static int __net_init sit_init_net(struct net *net)
1251 struct sit_net *sitn = net_generic(net, sit_net_id);
1252 struct ip_tunnel *t;
1255 sitn->tunnels[0] = sitn->tunnels_wc;
1256 sitn->tunnels[1] = sitn->tunnels_l;
1257 sitn->tunnels[2] = sitn->tunnels_r;
1258 sitn->tunnels[3] = sitn->tunnels_r_l;
1260 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1261 ipip6_tunnel_setup);
1262 if (!sitn->fb_tunnel_dev) {
1266 dev_net_set(sitn->fb_tunnel_dev, net);
1268 err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1272 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1274 if ((err = register_netdev(sitn->fb_tunnel_dev)))
1277 t = netdev_priv(sitn->fb_tunnel_dev);
1279 strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1283 dev_put(sitn->fb_tunnel_dev);
1285 ipip6_dev_free(sitn->fb_tunnel_dev);
1290 static void __net_exit sit_exit_net(struct net *net)
1292 struct sit_net *sitn = net_generic(net, sit_net_id);
1296 sit_destroy_tunnels(sitn, &list);
1297 unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1298 unregister_netdevice_many(&list);
1302 static struct pernet_operations sit_net_ops = {
1303 .init = sit_init_net,
1304 .exit = sit_exit_net,
1306 .size = sizeof(struct sit_net),
1309 static void __exit sit_cleanup(void)
1311 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1313 unregister_pernet_device(&sit_net_ops);
1314 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1317 static int __init sit_init(void)
1321 pr_info("IPv6 over IPv4 tunneling driver\n");
1323 err = register_pernet_device(&sit_net_ops);
1326 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1328 unregister_pernet_device(&sit_net_ops);
1329 pr_info("%s: can't add protocol\n", __func__);
1334 module_init(sit_init);
1335 module_exit(sit_cleanup);
1336 MODULE_LICENSE("GPL");
1337 MODULE_ALIAS_NETDEV("sit0");