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);
71 static struct rtnl_link_ops sit_link_ops __read_mostly;
73 static int sit_net_id __read_mostly;
75 struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
76 struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
77 struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
78 struct ip_tunnel __rcu *tunnels_wc[1];
79 struct ip_tunnel __rcu **tunnels[4];
81 struct net_device *fb_tunnel_dev;
85 * Locking : hash tables are protected by RCU and RTNL
88 #define for_each_ip_tunnel_rcu(start) \
89 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
91 static struct rtnl_link_stats64 *ipip6_get_stats64(struct net_device *dev,
92 struct rtnl_link_stats64 *tot)
96 for_each_possible_cpu(i) {
97 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
98 u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
102 start = u64_stats_fetch_begin_bh(&tstats->syncp);
103 rx_packets = tstats->rx_packets;
104 tx_packets = tstats->tx_packets;
105 rx_bytes = tstats->rx_bytes;
106 tx_bytes = tstats->tx_bytes;
107 } while (u64_stats_fetch_retry_bh(&tstats->syncp, start));
109 tot->rx_packets += rx_packets;
110 tot->tx_packets += tx_packets;
111 tot->rx_bytes += rx_bytes;
112 tot->tx_bytes += tx_bytes;
115 tot->rx_errors = dev->stats.rx_errors;
116 tot->tx_fifo_errors = dev->stats.tx_fifo_errors;
117 tot->tx_carrier_errors = dev->stats.tx_carrier_errors;
118 tot->tx_dropped = dev->stats.tx_dropped;
119 tot->tx_aborted_errors = dev->stats.tx_aborted_errors;
120 tot->tx_errors = dev->stats.tx_errors;
126 * Must be invoked with rcu_read_lock
128 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
129 struct net_device *dev, __be32 remote, __be32 local)
131 unsigned int h0 = HASH(remote);
132 unsigned int h1 = HASH(local);
134 struct sit_net *sitn = net_generic(net, sit_net_id);
136 for_each_ip_tunnel_rcu(sitn->tunnels_r_l[h0 ^ h1]) {
137 if (local == t->parms.iph.saddr &&
138 remote == t->parms.iph.daddr &&
139 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
140 (t->dev->flags & IFF_UP))
143 for_each_ip_tunnel_rcu(sitn->tunnels_r[h0]) {
144 if (remote == t->parms.iph.daddr &&
145 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
146 (t->dev->flags & IFF_UP))
149 for_each_ip_tunnel_rcu(sitn->tunnels_l[h1]) {
150 if (local == t->parms.iph.saddr &&
151 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
152 (t->dev->flags & IFF_UP))
155 t = rcu_dereference(sitn->tunnels_wc[0]);
156 if ((t != NULL) && (t->dev->flags & IFF_UP))
161 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
162 struct ip_tunnel_parm *parms)
164 __be32 remote = parms->iph.daddr;
165 __be32 local = parms->iph.saddr;
177 return &sitn->tunnels[prio][h];
180 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
183 return __ipip6_bucket(sitn, &t->parms);
186 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
188 struct ip_tunnel __rcu **tp;
189 struct ip_tunnel *iter;
191 for (tp = ipip6_bucket(sitn, t);
192 (iter = rtnl_dereference(*tp)) != NULL;
195 rcu_assign_pointer(*tp, t->next);
201 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
203 struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
205 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
206 rcu_assign_pointer(*tp, t);
209 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
211 #ifdef CONFIG_IPV6_SIT_6RD
212 struct ip_tunnel *t = netdev_priv(dev);
214 if (t->dev == sitn->fb_tunnel_dev) {
215 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
216 t->ip6rd.relay_prefix = 0;
217 t->ip6rd.prefixlen = 16;
218 t->ip6rd.relay_prefixlen = 0;
220 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
221 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
226 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
227 struct ip_tunnel_parm *parms, int create)
229 __be32 remote = parms->iph.daddr;
230 __be32 local = parms->iph.saddr;
231 struct ip_tunnel *t, *nt;
232 struct ip_tunnel __rcu **tp;
233 struct net_device *dev;
235 struct sit_net *sitn = net_generic(net, sit_net_id);
237 for (tp = __ipip6_bucket(sitn, parms);
238 (t = rtnl_dereference(*tp)) != NULL;
240 if (local == t->parms.iph.saddr &&
241 remote == t->parms.iph.daddr &&
242 parms->link == t->parms.link) {
253 strlcpy(name, parms->name, IFNAMSIZ);
255 strcpy(name, "sit%d");
257 dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
261 dev_net_set(dev, net);
263 nt = netdev_priv(dev);
266 if (ipip6_tunnel_init(dev) < 0)
268 ipip6_tunnel_clone_6rd(dev, sitn);
270 if (parms->i_flags & SIT_ISATAP)
271 dev->priv_flags |= IFF_ISATAP;
273 if (register_netdevice(dev) < 0)
276 strcpy(nt->parms.name, dev->name);
277 dev->rtnl_link_ops = &sit_link_ops;
281 ipip6_tunnel_link(sitn, nt);
290 #define for_each_prl_rcu(start) \
291 for (prl = rcu_dereference(start); \
293 prl = rcu_dereference(prl->next))
295 static struct ip_tunnel_prl_entry *
296 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
298 struct ip_tunnel_prl_entry *prl;
300 for_each_prl_rcu(t->prl)
301 if (prl->addr == addr)
307 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
308 struct ip_tunnel_prl __user *a)
310 struct ip_tunnel_prl kprl, *kp;
311 struct ip_tunnel_prl_entry *prl;
312 unsigned int cmax, c = 0, ca, len;
315 if (copy_from_user(&kprl, a, sizeof(kprl)))
317 cmax = kprl.datalen / sizeof(kprl);
318 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
321 /* For simple GET or for root users,
322 * we try harder to allocate.
324 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
325 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
330 ca = t->prl_count < cmax ? t->prl_count : cmax;
333 /* We don't try hard to allocate much memory for
335 * For root users, retry allocating enough memory for
338 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
346 for_each_prl_rcu(t->prl) {
349 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
351 kp[c].addr = prl->addr;
352 kp[c].flags = prl->flags;
354 if (kprl.addr != htonl(INADDR_ANY))
360 len = sizeof(*kp) * c;
362 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
371 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
373 struct ip_tunnel_prl_entry *p;
376 if (a->addr == htonl(INADDR_ANY))
381 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
382 if (p->addr == a->addr) {
397 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
407 rcu_assign_pointer(t->prl, p);
412 static void prl_list_destroy_rcu(struct rcu_head *head)
414 struct ip_tunnel_prl_entry *p, *n;
416 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
418 n = rcu_dereference_protected(p->next, 1);
425 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
427 struct ip_tunnel_prl_entry *x;
428 struct ip_tunnel_prl_entry __rcu **p;
433 if (a && a->addr != htonl(INADDR_ANY)) {
435 (x = rtnl_dereference(*p)) != NULL;
437 if (x->addr == a->addr) {
439 kfree_rcu(x, rcu_head);
446 x = rtnl_dereference(t->prl);
449 call_rcu(&x->rcu_head, prl_list_destroy_rcu);
458 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
460 struct ip_tunnel_prl_entry *p;
464 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
466 if (p->flags & PRL_DEFAULT)
467 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
469 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
471 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
473 if (ipv6_addr_is_isatap(addr6) &&
474 (addr6->s6_addr32[3] == iph->saddr) &&
475 ipv6_chk_prefix(addr6, t->dev))
476 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
484 static void ipip6_tunnel_uninit(struct net_device *dev)
486 struct net *net = dev_net(dev);
487 struct sit_net *sitn = net_generic(net, sit_net_id);
489 if (dev == sitn->fb_tunnel_dev) {
490 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
492 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
493 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
499 static int ipip6_err(struct sk_buff *skb, u32 info)
502 /* All the routers (except for Linux) return only
503 8 bytes of packet payload. It means, that precise relaying of
504 ICMP in the real Internet is absolutely infeasible.
506 const struct iphdr *iph = (const struct iphdr *)skb->data;
507 const int type = icmp_hdr(skb)->type;
508 const int code = icmp_hdr(skb)->code;
514 case ICMP_PARAMETERPROB:
517 case ICMP_DEST_UNREACH:
520 case ICMP_PORT_UNREACH:
521 /* Impossible event. */
524 /* All others are translated to HOST_UNREACH.
525 rfc2003 contains "deep thoughts" about NET_UNREACH,
526 I believe they are just ether pollution. --ANK
531 case ICMP_TIME_EXCEEDED:
532 if (code != ICMP_EXC_TTL)
541 t = ipip6_tunnel_lookup(dev_net(skb->dev),
548 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
549 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
550 t->dev->ifindex, 0, IPPROTO_IPV6, 0);
554 if (type == ICMP_REDIRECT) {
555 ipv4_redirect(skb, dev_net(skb->dev), t->dev->ifindex, 0,
561 if (t->parms.iph.daddr == 0)
565 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
568 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
572 t->err_time = jiffies;
577 static inline void ipip6_ecn_decapsulate(const struct iphdr *iph, struct sk_buff *skb)
579 if (INET_ECN_is_ce(iph->tos))
580 IP6_ECN_set_ce(ipv6_hdr(skb));
583 static int ipip6_rcv(struct sk_buff *skb)
585 const struct iphdr *iph;
586 struct ip_tunnel *tunnel;
588 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
593 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
594 iph->saddr, iph->daddr);
595 if (tunnel != NULL) {
596 struct pcpu_tstats *tstats;
599 skb->mac_header = skb->network_header;
600 skb_reset_network_header(skb);
601 IPCB(skb)->flags = 0;
602 skb->protocol = htons(ETH_P_IPV6);
603 skb->pkt_type = PACKET_HOST;
605 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
606 !isatap_chksrc(skb, iph, tunnel)) {
607 tunnel->dev->stats.rx_errors++;
612 tstats = this_cpu_ptr(tunnel->dev->tstats);
613 tstats->rx_packets++;
614 tstats->rx_bytes += skb->len;
616 __skb_tunnel_rx(skb, tunnel->dev);
618 ipip6_ecn_decapsulate(iph, skb);
625 /* no tunnel matched, let upstream know, ipsec may handle it */
633 * Returns the embedded IPv4 address if the IPv6 address
634 * comes from 6rd / 6to4 (RFC 3056) addr space.
637 __be32 try_6rd(const struct in6_addr *v6dst, struct ip_tunnel *tunnel)
641 #ifdef CONFIG_IPV6_SIT_6RD
642 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
643 tunnel->ip6rd.prefixlen)) {
644 unsigned int pbw0, pbi0;
648 pbw0 = tunnel->ip6rd.prefixlen >> 5;
649 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
651 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
652 tunnel->ip6rd.relay_prefixlen;
654 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
656 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
659 dst = tunnel->ip6rd.relay_prefix | htonl(d);
662 if (v6dst->s6_addr16[0] == htons(0x2002)) {
663 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
664 memcpy(&dst, &v6dst->s6_addr16[1], 4);
671 * This function assumes it is being called from dev_queue_xmit()
672 * and that skb is filled properly by that function.
675 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
676 struct net_device *dev)
678 struct ip_tunnel *tunnel = netdev_priv(dev);
679 const struct iphdr *tiph = &tunnel->parms.iph;
680 const struct ipv6hdr *iph6 = ipv6_hdr(skb);
681 u8 tos = tunnel->parms.iph.tos;
682 __be16 df = tiph->frag_off;
683 struct rtable *rt; /* Route to the other host */
684 struct net_device *tdev; /* Device to other host */
685 struct iphdr *iph; /* Our new IP header */
686 unsigned int max_headroom; /* The extra header space needed */
687 __be32 dst = tiph->daddr;
690 const struct in6_addr *addr6;
693 if (skb->protocol != htons(ETH_P_IPV6))
697 tos = ipv6_get_dsfield(iph6);
699 /* ISATAP (RFC4214) - must come before 6to4 */
700 if (dev->priv_flags & IFF_ISATAP) {
701 struct neighbour *neigh = NULL;
702 bool do_tx_error = false;
705 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
708 net_dbg_ratelimited("sit: nexthop == NULL\n");
712 addr6 = (const struct in6_addr *)&neigh->primary_key;
713 addr_type = ipv6_addr_type(addr6);
715 if ((addr_type & IPV6_ADDR_UNICAST) &&
716 ipv6_addr_is_isatap(addr6))
717 dst = addr6->s6_addr32[3];
721 neigh_release(neigh);
727 dst = try_6rd(&iph6->daddr, tunnel);
730 struct neighbour *neigh = NULL;
731 bool do_tx_error = false;
734 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
737 net_dbg_ratelimited("sit: nexthop == NULL\n");
741 addr6 = (const struct in6_addr *)&neigh->primary_key;
742 addr_type = ipv6_addr_type(addr6);
744 if (addr_type == IPV6_ADDR_ANY) {
745 addr6 = &ipv6_hdr(skb)->daddr;
746 addr_type = ipv6_addr_type(addr6);
749 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
750 dst = addr6->s6_addr32[3];
754 neigh_release(neigh);
759 rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
762 IPPROTO_IPV6, RT_TOS(tos),
765 dev->stats.tx_carrier_errors++;
768 if (rt->rt_type != RTN_UNICAST) {
770 dev->stats.tx_carrier_errors++;
777 dev->stats.collisions++;
782 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
785 dev->stats.collisions++;
790 if (mtu < IPV6_MIN_MTU) {
795 if (tunnel->parms.iph.daddr && skb_dst(skb))
796 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
798 if (skb->len > mtu) {
799 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
805 if (tunnel->err_count > 0) {
806 if (time_before(jiffies,
807 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
809 dst_link_failure(skb);
811 tunnel->err_count = 0;
815 * Okay, now see if we can stuff it in the buffer as-is.
817 max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
819 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
820 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
821 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
824 dev->stats.tx_dropped++;
829 skb_set_owner_w(new_skb, skb->sk);
832 iph6 = ipv6_hdr(skb);
835 skb->transport_header = skb->network_header;
836 skb_push(skb, sizeof(struct iphdr));
837 skb_reset_network_header(skb);
838 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
839 IPCB(skb)->flags = 0;
841 skb_dst_set(skb, &rt->dst);
844 * Push down and install the IPIP header.
849 iph->ihl = sizeof(struct iphdr)>>2;
851 iph->protocol = IPPROTO_IPV6;
852 iph->tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
853 iph->daddr = fl4.daddr;
854 iph->saddr = fl4.saddr;
856 if ((iph->ttl = tiph->ttl) == 0)
857 iph->ttl = iph6->hop_limit;
859 iptunnel_xmit(skb, dev);
863 dst_link_failure(skb);
865 dev->stats.tx_errors++;
870 static void ipip6_tunnel_bind_dev(struct net_device *dev)
872 struct net_device *tdev = NULL;
873 struct ip_tunnel *tunnel;
874 const struct iphdr *iph;
877 tunnel = netdev_priv(dev);
878 iph = &tunnel->parms.iph;
881 struct rtable *rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
882 iph->daddr, iph->saddr,
892 dev->flags |= IFF_POINTOPOINT;
895 if (!tdev && tunnel->parms.link)
896 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
899 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
900 dev->mtu = tdev->mtu - sizeof(struct iphdr);
901 if (dev->mtu < IPV6_MIN_MTU)
902 dev->mtu = IPV6_MIN_MTU;
904 dev->iflink = tunnel->parms.link;
908 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
911 struct ip_tunnel_parm p;
912 struct ip_tunnel_prl prl;
914 struct net *net = dev_net(dev);
915 struct sit_net *sitn = net_generic(net, sit_net_id);
916 #ifdef CONFIG_IPV6_SIT_6RD
917 struct ip_tunnel_6rd ip6rd;
922 #ifdef CONFIG_IPV6_SIT_6RD
926 if (dev == sitn->fb_tunnel_dev) {
927 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
931 t = ipip6_tunnel_locate(net, &p, 0);
934 t = netdev_priv(dev);
937 if (cmd == SIOCGETTUNNEL) {
938 memcpy(&p, &t->parms, sizeof(p));
939 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
942 #ifdef CONFIG_IPV6_SIT_6RD
944 ip6rd.prefix = t->ip6rd.prefix;
945 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
946 ip6rd.prefixlen = t->ip6rd.prefixlen;
947 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
948 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
959 if (!capable(CAP_NET_ADMIN))
963 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
967 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
968 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
971 p.iph.frag_off |= htons(IP_DF);
973 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
975 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
982 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
983 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
987 t = netdev_priv(dev);
988 ipip6_tunnel_unlink(sitn, t);
990 t->parms.iph.saddr = p.iph.saddr;
991 t->parms.iph.daddr = p.iph.daddr;
992 memcpy(dev->dev_addr, &p.iph.saddr, 4);
993 memcpy(dev->broadcast, &p.iph.daddr, 4);
994 ipip6_tunnel_link(sitn, t);
995 netdev_state_change(dev);
1001 if (cmd == SIOCCHGTUNNEL) {
1002 t->parms.iph.ttl = p.iph.ttl;
1003 t->parms.iph.tos = p.iph.tos;
1004 if (t->parms.link != p.link) {
1005 t->parms.link = p.link;
1006 ipip6_tunnel_bind_dev(dev);
1007 netdev_state_change(dev);
1010 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1013 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1018 if (!capable(CAP_NET_ADMIN))
1021 if (dev == sitn->fb_tunnel_dev) {
1023 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1026 if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1029 if (t == netdev_priv(sitn->fb_tunnel_dev))
1033 unregister_netdevice(dev);
1039 if (dev == sitn->fb_tunnel_dev)
1042 if (!(t = netdev_priv(dev)))
1044 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1051 if (!capable(CAP_NET_ADMIN))
1054 if (dev == sitn->fb_tunnel_dev)
1057 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1060 if (!(t = netdev_priv(dev)))
1065 err = ipip6_tunnel_del_prl(t, &prl);
1069 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1072 netdev_state_change(dev);
1075 #ifdef CONFIG_IPV6_SIT_6RD
1080 if (!capable(CAP_NET_ADMIN))
1084 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1088 t = netdev_priv(dev);
1090 if (cmd != SIOCDEL6RD) {
1091 struct in6_addr prefix;
1092 __be32 relay_prefix;
1095 if (ip6rd.relay_prefixlen > 32 ||
1096 ip6rd.prefixlen + (32 - ip6rd.relay_prefixlen) > 64)
1099 ipv6_addr_prefix(&prefix, &ip6rd.prefix,
1101 if (!ipv6_addr_equal(&prefix, &ip6rd.prefix))
1103 if (ip6rd.relay_prefixlen)
1104 relay_prefix = ip6rd.relay_prefix &
1105 htonl(0xffffffffUL <<
1106 (32 - ip6rd.relay_prefixlen));
1109 if (relay_prefix != ip6rd.relay_prefix)
1112 t->ip6rd.prefix = prefix;
1113 t->ip6rd.relay_prefix = relay_prefix;
1114 t->ip6rd.prefixlen = ip6rd.prefixlen;
1115 t->ip6rd.relay_prefixlen = ip6rd.relay_prefixlen;
1117 ipip6_tunnel_clone_6rd(dev, sitn);
1131 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1133 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1139 static const struct net_device_ops ipip6_netdev_ops = {
1140 .ndo_uninit = ipip6_tunnel_uninit,
1141 .ndo_start_xmit = ipip6_tunnel_xmit,
1142 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1143 .ndo_change_mtu = ipip6_tunnel_change_mtu,
1144 .ndo_get_stats64= ipip6_get_stats64,
1147 static void ipip6_dev_free(struct net_device *dev)
1149 free_percpu(dev->tstats);
1153 static void ipip6_tunnel_setup(struct net_device *dev)
1155 dev->netdev_ops = &ipip6_netdev_ops;
1156 dev->destructor = ipip6_dev_free;
1158 dev->type = ARPHRD_SIT;
1159 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
1160 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr);
1161 dev->flags = IFF_NOARP;
1162 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1165 dev->features |= NETIF_F_NETNS_LOCAL;
1166 dev->features |= NETIF_F_LLTX;
1169 static int ipip6_tunnel_init(struct net_device *dev)
1171 struct ip_tunnel *tunnel = netdev_priv(dev);
1175 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1176 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1178 ipip6_tunnel_bind_dev(dev);
1179 dev->tstats = alloc_percpu(struct pcpu_tstats);
1186 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1188 struct ip_tunnel *tunnel = netdev_priv(dev);
1189 struct iphdr *iph = &tunnel->parms.iph;
1190 struct net *net = dev_net(dev);
1191 struct sit_net *sitn = net_generic(net, sit_net_id);
1194 strcpy(tunnel->parms.name, dev->name);
1197 iph->protocol = IPPROTO_IPV6;
1201 dev->tstats = alloc_percpu(struct pcpu_tstats);
1205 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1209 static size_t sit_get_size(const struct net_device *dev)
1212 /* IFLA_IPTUN_LINK */
1214 /* IFLA_IPTUN_LOCAL */
1216 /* IFLA_IPTUN_REMOTE */
1218 /* IFLA_IPTUN_TTL */
1220 /* IFLA_IPTUN_TOS */
1225 static int sit_fill_info(struct sk_buff *skb, const struct net_device *dev)
1227 struct ip_tunnel *tunnel = netdev_priv(dev);
1228 struct ip_tunnel_parm *parm = &tunnel->parms;
1230 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1231 nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1232 nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1233 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1234 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos))
1235 goto nla_put_failure;
1242 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1244 .maxtype = IFLA_IPTUN_MAX,
1245 .priv_size = sizeof(struct ip_tunnel),
1246 .get_size = sit_get_size,
1247 .fill_info = sit_fill_info,
1250 static struct xfrm_tunnel sit_handler __read_mostly = {
1251 .handler = ipip6_rcv,
1252 .err_handler = ipip6_err,
1256 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1260 for (prio = 1; prio < 4; prio++) {
1262 for (h = 0; h < HASH_SIZE; h++) {
1263 struct ip_tunnel *t;
1265 t = rtnl_dereference(sitn->tunnels[prio][h]);
1267 unregister_netdevice_queue(t->dev, head);
1268 t = rtnl_dereference(t->next);
1274 static int __net_init sit_init_net(struct net *net)
1276 struct sit_net *sitn = net_generic(net, sit_net_id);
1277 struct ip_tunnel *t;
1280 sitn->tunnels[0] = sitn->tunnels_wc;
1281 sitn->tunnels[1] = sitn->tunnels_l;
1282 sitn->tunnels[2] = sitn->tunnels_r;
1283 sitn->tunnels[3] = sitn->tunnels_r_l;
1285 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1286 ipip6_tunnel_setup);
1287 if (!sitn->fb_tunnel_dev) {
1291 dev_net_set(sitn->fb_tunnel_dev, net);
1293 err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1297 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1299 if ((err = register_netdev(sitn->fb_tunnel_dev)))
1302 t = netdev_priv(sitn->fb_tunnel_dev);
1304 strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1308 dev_put(sitn->fb_tunnel_dev);
1310 ipip6_dev_free(sitn->fb_tunnel_dev);
1315 static void __net_exit sit_exit_net(struct net *net)
1317 struct sit_net *sitn = net_generic(net, sit_net_id);
1321 sit_destroy_tunnels(sitn, &list);
1322 unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1323 unregister_netdevice_many(&list);
1327 static struct pernet_operations sit_net_ops = {
1328 .init = sit_init_net,
1329 .exit = sit_exit_net,
1331 .size = sizeof(struct sit_net),
1334 static void __exit sit_cleanup(void)
1336 rtnl_link_unregister(&sit_link_ops);
1337 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1339 unregister_pernet_device(&sit_net_ops);
1340 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1343 static int __init sit_init(void)
1347 pr_info("IPv6 over IPv4 tunneling driver\n");
1349 err = register_pernet_device(&sit_net_ops);
1352 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1354 pr_info("%s: can't add protocol\n", __func__);
1355 goto xfrm_tunnel_failed;
1357 err = rtnl_link_register(&sit_link_ops);
1359 goto rtnl_link_failed;
1365 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1367 unregister_pernet_device(&sit_net_ops);
1371 module_init(sit_init);
1372 module_exit(sit_cleanup);
1373 MODULE_LICENSE("GPL");
1374 MODULE_ALIAS_NETDEV("sit0");