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;
84 static struct rtnl_link_stats64 *ipip6_get_stats64(struct net_device *dev,
85 struct rtnl_link_stats64 *tot)
89 for_each_possible_cpu(i) {
90 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
91 u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
95 start = u64_stats_fetch_begin_bh(&tstats->syncp);
96 rx_packets = tstats->rx_packets;
97 tx_packets = tstats->tx_packets;
98 rx_bytes = tstats->rx_bytes;
99 tx_bytes = tstats->tx_bytes;
100 } while (u64_stats_fetch_retry_bh(&tstats->syncp, start));
102 tot->rx_packets += rx_packets;
103 tot->tx_packets += tx_packets;
104 tot->rx_bytes += rx_bytes;
105 tot->tx_bytes += tx_bytes;
108 tot->rx_errors = dev->stats.rx_errors;
109 tot->tx_fifo_errors = dev->stats.tx_fifo_errors;
110 tot->tx_carrier_errors = dev->stats.tx_carrier_errors;
111 tot->tx_dropped = dev->stats.tx_dropped;
112 tot->tx_aborted_errors = dev->stats.tx_aborted_errors;
113 tot->tx_errors = dev->stats.tx_errors;
119 * Must be invoked with rcu_read_lock
121 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
122 struct net_device *dev, __be32 remote, __be32 local)
124 unsigned int h0 = HASH(remote);
125 unsigned int h1 = HASH(local);
127 struct sit_net *sitn = net_generic(net, sit_net_id);
129 for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
130 if (local == t->parms.iph.saddr &&
131 remote == t->parms.iph.daddr &&
132 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
133 (t->dev->flags & IFF_UP))
136 for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
137 if (remote == t->parms.iph.daddr &&
138 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
139 (t->dev->flags & IFF_UP))
142 for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
143 if (local == t->parms.iph.saddr &&
144 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
145 (t->dev->flags & IFF_UP))
148 t = rcu_dereference(sitn->tunnels_wc[0]);
149 if ((t != NULL) && (t->dev->flags & IFF_UP))
154 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
155 struct ip_tunnel_parm *parms)
157 __be32 remote = parms->iph.daddr;
158 __be32 local = parms->iph.saddr;
170 return &sitn->tunnels[prio][h];
173 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
176 return __ipip6_bucket(sitn, &t->parms);
179 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
181 struct ip_tunnel __rcu **tp;
182 struct ip_tunnel *iter;
184 for (tp = ipip6_bucket(sitn, t);
185 (iter = rtnl_dereference(*tp)) != NULL;
188 rcu_assign_pointer(*tp, t->next);
194 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
196 struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
198 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
199 rcu_assign_pointer(*tp, t);
202 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
204 #ifdef CONFIG_IPV6_SIT_6RD
205 struct ip_tunnel *t = netdev_priv(dev);
207 if (t->dev == sitn->fb_tunnel_dev) {
208 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
209 t->ip6rd.relay_prefix = 0;
210 t->ip6rd.prefixlen = 16;
211 t->ip6rd.relay_prefixlen = 0;
213 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
214 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
219 static int ipip6_tunnel_create(struct net_device *dev)
221 struct ip_tunnel *t = netdev_priv(dev);
222 struct net *net = dev_net(dev);
223 struct sit_net *sitn = net_generic(net, sit_net_id);
226 err = ipip6_tunnel_init(dev);
229 ipip6_tunnel_clone_6rd(dev, sitn);
231 if ((__force u16)t->parms.i_flags & SIT_ISATAP)
232 dev->priv_flags |= IFF_ISATAP;
234 err = register_netdevice(dev);
238 strcpy(t->parms.name, dev->name);
239 dev->rtnl_link_ops = &sit_link_ops;
243 ipip6_tunnel_link(sitn, t);
250 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
251 struct ip_tunnel_parm *parms, int create)
253 __be32 remote = parms->iph.daddr;
254 __be32 local = parms->iph.saddr;
255 struct ip_tunnel *t, *nt;
256 struct ip_tunnel __rcu **tp;
257 struct net_device *dev;
259 struct sit_net *sitn = net_generic(net, sit_net_id);
261 for (tp = __ipip6_bucket(sitn, parms);
262 (t = rtnl_dereference(*tp)) != NULL;
264 if (local == t->parms.iph.saddr &&
265 remote == t->parms.iph.daddr &&
266 parms->link == t->parms.link) {
277 strlcpy(name, parms->name, IFNAMSIZ);
279 strcpy(name, "sit%d");
281 dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
285 dev_net_set(dev, net);
287 nt = netdev_priv(dev);
290 if (ipip6_tunnel_create(dev) < 0)
301 #define for_each_prl_rcu(start) \
302 for (prl = rcu_dereference(start); \
304 prl = rcu_dereference(prl->next))
306 static struct ip_tunnel_prl_entry *
307 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
309 struct ip_tunnel_prl_entry *prl;
311 for_each_prl_rcu(t->prl)
312 if (prl->addr == addr)
318 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
319 struct ip_tunnel_prl __user *a)
321 struct ip_tunnel_prl kprl, *kp;
322 struct ip_tunnel_prl_entry *prl;
323 unsigned int cmax, c = 0, ca, len;
326 if (copy_from_user(&kprl, a, sizeof(kprl)))
328 cmax = kprl.datalen / sizeof(kprl);
329 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
332 /* For simple GET or for root users,
333 * we try harder to allocate.
335 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
336 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
341 ca = t->prl_count < cmax ? t->prl_count : cmax;
344 /* We don't try hard to allocate much memory for
346 * For root users, retry allocating enough memory for
349 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
357 for_each_prl_rcu(t->prl) {
360 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
362 kp[c].addr = prl->addr;
363 kp[c].flags = prl->flags;
365 if (kprl.addr != htonl(INADDR_ANY))
371 len = sizeof(*kp) * c;
373 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
382 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
384 struct ip_tunnel_prl_entry *p;
387 if (a->addr == htonl(INADDR_ANY))
392 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
393 if (p->addr == a->addr) {
408 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
418 rcu_assign_pointer(t->prl, p);
423 static void prl_list_destroy_rcu(struct rcu_head *head)
425 struct ip_tunnel_prl_entry *p, *n;
427 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
429 n = rcu_dereference_protected(p->next, 1);
436 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
438 struct ip_tunnel_prl_entry *x;
439 struct ip_tunnel_prl_entry __rcu **p;
444 if (a && a->addr != htonl(INADDR_ANY)) {
446 (x = rtnl_dereference(*p)) != NULL;
448 if (x->addr == a->addr) {
450 kfree_rcu(x, rcu_head);
457 x = rtnl_dereference(t->prl);
460 call_rcu(&x->rcu_head, prl_list_destroy_rcu);
469 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
471 struct ip_tunnel_prl_entry *p;
475 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
477 if (p->flags & PRL_DEFAULT)
478 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
480 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
482 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
484 if (ipv6_addr_is_isatap(addr6) &&
485 (addr6->s6_addr32[3] == iph->saddr) &&
486 ipv6_chk_prefix(addr6, t->dev))
487 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
495 static void ipip6_tunnel_uninit(struct net_device *dev)
497 struct net *net = dev_net(dev);
498 struct sit_net *sitn = net_generic(net, sit_net_id);
500 if (dev == sitn->fb_tunnel_dev) {
501 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
503 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
504 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
510 static int ipip6_err(struct sk_buff *skb, u32 info)
513 /* All the routers (except for Linux) return only
514 8 bytes of packet payload. It means, that precise relaying of
515 ICMP in the real Internet is absolutely infeasible.
517 const struct iphdr *iph = (const struct iphdr *)skb->data;
518 const int type = icmp_hdr(skb)->type;
519 const int code = icmp_hdr(skb)->code;
525 case ICMP_PARAMETERPROB:
528 case ICMP_DEST_UNREACH:
531 case ICMP_PORT_UNREACH:
532 /* Impossible event. */
535 /* All others are translated to HOST_UNREACH.
536 rfc2003 contains "deep thoughts" about NET_UNREACH,
537 I believe they are just ether pollution. --ANK
542 case ICMP_TIME_EXCEEDED:
543 if (code != ICMP_EXC_TTL)
552 t = ipip6_tunnel_lookup(dev_net(skb->dev),
559 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
560 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
561 t->dev->ifindex, 0, IPPROTO_IPV6, 0);
565 if (type == ICMP_REDIRECT) {
566 ipv4_redirect(skb, dev_net(skb->dev), t->dev->ifindex, 0,
572 if (t->parms.iph.daddr == 0)
576 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
579 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
583 t->err_time = jiffies;
588 static inline void ipip6_ecn_decapsulate(const struct iphdr *iph, struct sk_buff *skb)
590 if (INET_ECN_is_ce(iph->tos))
591 IP6_ECN_set_ce(ipv6_hdr(skb));
594 static int ipip6_rcv(struct sk_buff *skb)
596 const struct iphdr *iph;
597 struct ip_tunnel *tunnel;
599 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
604 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
605 iph->saddr, iph->daddr);
606 if (tunnel != NULL) {
607 struct pcpu_tstats *tstats;
610 skb->mac_header = skb->network_header;
611 skb_reset_network_header(skb);
612 IPCB(skb)->flags = 0;
613 skb->protocol = htons(ETH_P_IPV6);
614 skb->pkt_type = PACKET_HOST;
616 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
617 !isatap_chksrc(skb, iph, tunnel)) {
618 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);
636 /* no tunnel matched, let upstream know, ipsec may handle it */
644 * Returns the embedded IPv4 address if the IPv6 address
645 * comes from 6rd / 6to4 (RFC 3056) addr space.
648 __be32 try_6rd(const struct in6_addr *v6dst, struct ip_tunnel *tunnel)
652 #ifdef CONFIG_IPV6_SIT_6RD
653 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
654 tunnel->ip6rd.prefixlen)) {
655 unsigned int pbw0, pbi0;
659 pbw0 = tunnel->ip6rd.prefixlen >> 5;
660 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
662 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
663 tunnel->ip6rd.relay_prefixlen;
665 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
667 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
670 dst = tunnel->ip6rd.relay_prefix | htonl(d);
673 if (v6dst->s6_addr16[0] == htons(0x2002)) {
674 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
675 memcpy(&dst, &v6dst->s6_addr16[1], 4);
682 * This function assumes it is being called from dev_queue_xmit()
683 * and that skb is filled properly by that function.
686 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
687 struct net_device *dev)
689 struct ip_tunnel *tunnel = netdev_priv(dev);
690 const struct iphdr *tiph = &tunnel->parms.iph;
691 const struct ipv6hdr *iph6 = ipv6_hdr(skb);
692 u8 tos = tunnel->parms.iph.tos;
693 __be16 df = tiph->frag_off;
694 struct rtable *rt; /* Route to the other host */
695 struct net_device *tdev; /* Device to other host */
696 struct iphdr *iph; /* Our new IP header */
697 unsigned int max_headroom; /* The extra header space needed */
698 __be32 dst = tiph->daddr;
701 const struct in6_addr *addr6;
704 if (skb->protocol != htons(ETH_P_IPV6))
708 tos = ipv6_get_dsfield(iph6);
710 /* ISATAP (RFC4214) - must come before 6to4 */
711 if (dev->priv_flags & IFF_ISATAP) {
712 struct neighbour *neigh = NULL;
713 bool do_tx_error = false;
716 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
719 net_dbg_ratelimited("sit: nexthop == NULL\n");
723 addr6 = (const struct in6_addr *)&neigh->primary_key;
724 addr_type = ipv6_addr_type(addr6);
726 if ((addr_type & IPV6_ADDR_UNICAST) &&
727 ipv6_addr_is_isatap(addr6))
728 dst = addr6->s6_addr32[3];
732 neigh_release(neigh);
738 dst = try_6rd(&iph6->daddr, tunnel);
741 struct neighbour *neigh = NULL;
742 bool do_tx_error = false;
745 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
748 net_dbg_ratelimited("sit: nexthop == NULL\n");
752 addr6 = (const struct in6_addr *)&neigh->primary_key;
753 addr_type = ipv6_addr_type(addr6);
755 if (addr_type == IPV6_ADDR_ANY) {
756 addr6 = &ipv6_hdr(skb)->daddr;
757 addr_type = ipv6_addr_type(addr6);
760 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
761 dst = addr6->s6_addr32[3];
765 neigh_release(neigh);
770 rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
773 IPPROTO_IPV6, RT_TOS(tos),
776 dev->stats.tx_carrier_errors++;
779 if (rt->rt_type != RTN_UNICAST) {
781 dev->stats.tx_carrier_errors++;
788 dev->stats.collisions++;
793 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
796 dev->stats.collisions++;
801 if (mtu < IPV6_MIN_MTU) {
806 if (tunnel->parms.iph.daddr && skb_dst(skb))
807 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
809 if (skb->len > mtu) {
810 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
816 if (tunnel->err_count > 0) {
817 if (time_before(jiffies,
818 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
820 dst_link_failure(skb);
822 tunnel->err_count = 0;
826 * Okay, now see if we can stuff it in the buffer as-is.
828 max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
830 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
831 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
832 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
835 dev->stats.tx_dropped++;
840 skb_set_owner_w(new_skb, skb->sk);
843 iph6 = ipv6_hdr(skb);
846 skb->transport_header = skb->network_header;
847 skb_push(skb, sizeof(struct iphdr));
848 skb_reset_network_header(skb);
849 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
850 IPCB(skb)->flags = 0;
852 skb_dst_set(skb, &rt->dst);
855 * Push down and install the IPIP header.
860 iph->ihl = sizeof(struct iphdr)>>2;
862 iph->protocol = IPPROTO_IPV6;
863 iph->tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
864 iph->daddr = fl4.daddr;
865 iph->saddr = fl4.saddr;
867 if ((iph->ttl = tiph->ttl) == 0)
868 iph->ttl = iph6->hop_limit;
870 iptunnel_xmit(skb, dev);
874 dst_link_failure(skb);
876 dev->stats.tx_errors++;
881 static void ipip6_tunnel_bind_dev(struct net_device *dev)
883 struct net_device *tdev = NULL;
884 struct ip_tunnel *tunnel;
885 const struct iphdr *iph;
888 tunnel = netdev_priv(dev);
889 iph = &tunnel->parms.iph;
892 struct rtable *rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
893 iph->daddr, iph->saddr,
903 dev->flags |= IFF_POINTOPOINT;
906 if (!tdev && tunnel->parms.link)
907 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
910 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
911 dev->mtu = tdev->mtu - sizeof(struct iphdr);
912 if (dev->mtu < IPV6_MIN_MTU)
913 dev->mtu = IPV6_MIN_MTU;
915 dev->iflink = tunnel->parms.link;
918 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
920 struct net *net = dev_net(t->dev);
921 struct sit_net *sitn = net_generic(net, sit_net_id);
923 ipip6_tunnel_unlink(sitn, t);
925 t->parms.iph.saddr = p->iph.saddr;
926 t->parms.iph.daddr = p->iph.daddr;
927 memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
928 memcpy(t->dev->broadcast, &p->iph.daddr, 4);
929 ipip6_tunnel_link(sitn, t);
930 t->parms.iph.ttl = p->iph.ttl;
931 t->parms.iph.tos = p->iph.tos;
932 if (t->parms.link != p->link) {
933 t->parms.link = p->link;
934 ipip6_tunnel_bind_dev(t->dev);
936 netdev_state_change(t->dev);
939 #ifdef CONFIG_IPV6_SIT_6RD
940 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
941 struct ip_tunnel_6rd *ip6rd)
943 struct in6_addr prefix;
946 if (ip6rd->relay_prefixlen > 32 ||
947 ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
950 ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
951 if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
953 if (ip6rd->relay_prefixlen)
954 relay_prefix = ip6rd->relay_prefix &
955 htonl(0xffffffffUL <<
956 (32 - ip6rd->relay_prefixlen));
959 if (relay_prefix != ip6rd->relay_prefix)
962 t->ip6rd.prefix = prefix;
963 t->ip6rd.relay_prefix = relay_prefix;
964 t->ip6rd.prefixlen = ip6rd->prefixlen;
965 t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
966 netdev_state_change(t->dev);
972 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
975 struct ip_tunnel_parm p;
976 struct ip_tunnel_prl prl;
978 struct net *net = dev_net(dev);
979 struct sit_net *sitn = net_generic(net, sit_net_id);
980 #ifdef CONFIG_IPV6_SIT_6RD
981 struct ip_tunnel_6rd ip6rd;
986 #ifdef CONFIG_IPV6_SIT_6RD
990 if (dev == sitn->fb_tunnel_dev) {
991 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
995 t = ipip6_tunnel_locate(net, &p, 0);
998 t = netdev_priv(dev);
1001 if (cmd == SIOCGETTUNNEL) {
1002 memcpy(&p, &t->parms, sizeof(p));
1003 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1006 #ifdef CONFIG_IPV6_SIT_6RD
1008 ip6rd.prefix = t->ip6rd.prefix;
1009 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1010 ip6rd.prefixlen = t->ip6rd.prefixlen;
1011 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1012 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1023 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1027 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1031 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
1032 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1035 p.iph.frag_off |= htons(IP_DF);
1037 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1039 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1041 if (t->dev != dev) {
1046 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1047 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1051 t = netdev_priv(dev);
1054 ipip6_tunnel_update(t, &p);
1059 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1062 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1067 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1070 if (dev == sitn->fb_tunnel_dev) {
1072 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1075 if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1078 if (t == netdev_priv(sitn->fb_tunnel_dev))
1082 unregister_netdevice(dev);
1088 if (dev == sitn->fb_tunnel_dev)
1091 if (!(t = netdev_priv(dev)))
1093 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1100 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1103 if (dev == sitn->fb_tunnel_dev)
1106 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1109 if (!(t = netdev_priv(dev)))
1114 err = ipip6_tunnel_del_prl(t, &prl);
1118 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1121 netdev_state_change(dev);
1124 #ifdef CONFIG_IPV6_SIT_6RD
1129 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1133 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1137 t = netdev_priv(dev);
1139 if (cmd != SIOCDEL6RD) {
1140 err = ipip6_tunnel_update_6rd(t, &ip6rd);
1144 ipip6_tunnel_clone_6rd(dev, sitn);
1158 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1160 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1166 static const struct net_device_ops ipip6_netdev_ops = {
1167 .ndo_uninit = ipip6_tunnel_uninit,
1168 .ndo_start_xmit = ipip6_tunnel_xmit,
1169 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1170 .ndo_change_mtu = ipip6_tunnel_change_mtu,
1171 .ndo_get_stats64= ipip6_get_stats64,
1174 static void ipip6_dev_free(struct net_device *dev)
1176 free_percpu(dev->tstats);
1180 static void ipip6_tunnel_setup(struct net_device *dev)
1182 dev->netdev_ops = &ipip6_netdev_ops;
1183 dev->destructor = ipip6_dev_free;
1185 dev->type = ARPHRD_SIT;
1186 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
1187 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr);
1188 dev->flags = IFF_NOARP;
1189 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1192 dev->features |= NETIF_F_NETNS_LOCAL;
1193 dev->features |= NETIF_F_LLTX;
1196 static int ipip6_tunnel_init(struct net_device *dev)
1198 struct ip_tunnel *tunnel = netdev_priv(dev);
1202 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1203 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1205 ipip6_tunnel_bind_dev(dev);
1206 dev->tstats = alloc_percpu(struct pcpu_tstats);
1213 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1215 struct ip_tunnel *tunnel = netdev_priv(dev);
1216 struct iphdr *iph = &tunnel->parms.iph;
1217 struct net *net = dev_net(dev);
1218 struct sit_net *sitn = net_generic(net, sit_net_id);
1221 strcpy(tunnel->parms.name, dev->name);
1224 iph->protocol = IPPROTO_IPV6;
1228 dev->tstats = alloc_percpu(struct pcpu_tstats);
1232 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1236 static void ipip6_netlink_parms(struct nlattr *data[],
1237 struct ip_tunnel_parm *parms)
1239 memset(parms, 0, sizeof(*parms));
1241 parms->iph.version = 4;
1242 parms->iph.protocol = IPPROTO_IPV6;
1244 parms->iph.ttl = 64;
1249 if (data[IFLA_IPTUN_LINK])
1250 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1252 if (data[IFLA_IPTUN_LOCAL])
1253 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1255 if (data[IFLA_IPTUN_REMOTE])
1256 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1258 if (data[IFLA_IPTUN_TTL]) {
1259 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1261 parms->iph.frag_off = htons(IP_DF);
1264 if (data[IFLA_IPTUN_TOS])
1265 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1267 if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1268 parms->iph.frag_off = htons(IP_DF);
1270 if (data[IFLA_IPTUN_FLAGS])
1271 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1274 #ifdef CONFIG_IPV6_SIT_6RD
1275 /* This function returns true when 6RD attributes are present in the nl msg */
1276 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1277 struct ip_tunnel_6rd *ip6rd)
1280 memset(ip6rd, 0, sizeof(*ip6rd));
1285 if (data[IFLA_IPTUN_6RD_PREFIX]) {
1287 nla_memcpy(&ip6rd->prefix, data[IFLA_IPTUN_6RD_PREFIX],
1288 sizeof(struct in6_addr));
1291 if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1293 ip6rd->relay_prefix =
1294 nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1297 if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1299 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1302 if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1304 ip6rd->relay_prefixlen =
1305 nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1312 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1313 struct nlattr *tb[], struct nlattr *data[])
1315 struct net *net = dev_net(dev);
1316 struct ip_tunnel *nt;
1317 #ifdef CONFIG_IPV6_SIT_6RD
1318 struct ip_tunnel_6rd ip6rd;
1322 nt = netdev_priv(dev);
1323 ipip6_netlink_parms(data, &nt->parms);
1325 if (ipip6_tunnel_locate(net, &nt->parms, 0))
1328 err = ipip6_tunnel_create(dev);
1332 #ifdef CONFIG_IPV6_SIT_6RD
1333 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1334 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1340 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1341 struct nlattr *data[])
1343 struct ip_tunnel *t;
1344 struct ip_tunnel_parm p;
1345 struct net *net = dev_net(dev);
1346 struct sit_net *sitn = net_generic(net, sit_net_id);
1347 #ifdef CONFIG_IPV6_SIT_6RD
1348 struct ip_tunnel_6rd ip6rd;
1351 if (dev == sitn->fb_tunnel_dev)
1354 ipip6_netlink_parms(data, &p);
1356 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1357 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1360 t = ipip6_tunnel_locate(net, &p, 0);
1366 t = netdev_priv(dev);
1368 ipip6_tunnel_update(t, &p);
1370 #ifdef CONFIG_IPV6_SIT_6RD
1371 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1372 return ipip6_tunnel_update_6rd(t, &ip6rd);
1378 static size_t ipip6_get_size(const struct net_device *dev)
1381 /* IFLA_IPTUN_LINK */
1383 /* IFLA_IPTUN_LOCAL */
1385 /* IFLA_IPTUN_REMOTE */
1387 /* IFLA_IPTUN_TTL */
1389 /* IFLA_IPTUN_TOS */
1391 /* IFLA_IPTUN_PMTUDISC */
1393 /* IFLA_IPTUN_FLAGS */
1395 #ifdef CONFIG_IPV6_SIT_6RD
1396 /* IFLA_IPTUN_6RD_PREFIX */
1397 nla_total_size(sizeof(struct in6_addr)) +
1398 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1400 /* IFLA_IPTUN_6RD_PREFIXLEN */
1402 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1408 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1410 struct ip_tunnel *tunnel = netdev_priv(dev);
1411 struct ip_tunnel_parm *parm = &tunnel->parms;
1413 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1414 nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1415 nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1416 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1417 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1418 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1419 !!(parm->iph.frag_off & htons(IP_DF))) ||
1420 nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1421 goto nla_put_failure;
1423 #ifdef CONFIG_IPV6_SIT_6RD
1424 if (nla_put(skb, IFLA_IPTUN_6RD_PREFIX, sizeof(struct in6_addr),
1425 &tunnel->ip6rd.prefix) ||
1426 nla_put_be32(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1427 tunnel->ip6rd.relay_prefix) ||
1428 nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1429 tunnel->ip6rd.prefixlen) ||
1430 nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1431 tunnel->ip6rd.relay_prefixlen))
1432 goto nla_put_failure;
1441 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1442 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1443 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 },
1444 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 },
1445 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1446 [IFLA_IPTUN_TOS] = { .type = NLA_U8 },
1447 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 },
1448 [IFLA_IPTUN_FLAGS] = { .type = NLA_U16 },
1449 #ifdef CONFIG_IPV6_SIT_6RD
1450 [IFLA_IPTUN_6RD_PREFIX] = { .len = sizeof(struct in6_addr) },
1451 [IFLA_IPTUN_6RD_RELAY_PREFIX] = { .type = NLA_U32 },
1452 [IFLA_IPTUN_6RD_PREFIXLEN] = { .type = NLA_U16 },
1453 [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1457 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1459 .maxtype = IFLA_IPTUN_MAX,
1460 .policy = ipip6_policy,
1461 .priv_size = sizeof(struct ip_tunnel),
1462 .setup = ipip6_tunnel_setup,
1463 .newlink = ipip6_newlink,
1464 .changelink = ipip6_changelink,
1465 .get_size = ipip6_get_size,
1466 .fill_info = ipip6_fill_info,
1469 static struct xfrm_tunnel sit_handler __read_mostly = {
1470 .handler = ipip6_rcv,
1471 .err_handler = ipip6_err,
1475 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1479 for (prio = 1; prio < 4; prio++) {
1481 for (h = 0; h < HASH_SIZE; h++) {
1482 struct ip_tunnel *t;
1484 t = rtnl_dereference(sitn->tunnels[prio][h]);
1486 unregister_netdevice_queue(t->dev, head);
1487 t = rtnl_dereference(t->next);
1493 static int __net_init sit_init_net(struct net *net)
1495 struct sit_net *sitn = net_generic(net, sit_net_id);
1496 struct ip_tunnel *t;
1499 sitn->tunnels[0] = sitn->tunnels_wc;
1500 sitn->tunnels[1] = sitn->tunnels_l;
1501 sitn->tunnels[2] = sitn->tunnels_r;
1502 sitn->tunnels[3] = sitn->tunnels_r_l;
1504 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1505 ipip6_tunnel_setup);
1506 if (!sitn->fb_tunnel_dev) {
1510 dev_net_set(sitn->fb_tunnel_dev, net);
1512 err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1516 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1518 if ((err = register_netdev(sitn->fb_tunnel_dev)))
1521 t = netdev_priv(sitn->fb_tunnel_dev);
1523 strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1527 dev_put(sitn->fb_tunnel_dev);
1529 ipip6_dev_free(sitn->fb_tunnel_dev);
1534 static void __net_exit sit_exit_net(struct net *net)
1536 struct sit_net *sitn = net_generic(net, sit_net_id);
1540 sit_destroy_tunnels(sitn, &list);
1541 unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1542 unregister_netdevice_many(&list);
1546 static struct pernet_operations sit_net_ops = {
1547 .init = sit_init_net,
1548 .exit = sit_exit_net,
1550 .size = sizeof(struct sit_net),
1553 static void __exit sit_cleanup(void)
1555 rtnl_link_unregister(&sit_link_ops);
1556 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1558 unregister_pernet_device(&sit_net_ops);
1559 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1562 static int __init sit_init(void)
1566 pr_info("IPv6 over IPv4 tunneling driver\n");
1568 err = register_pernet_device(&sit_net_ops);
1571 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1573 pr_info("%s: can't add protocol\n", __func__);
1574 goto xfrm_tunnel_failed;
1576 err = rtnl_link_register(&sit_link_ops);
1578 goto rtnl_link_failed;
1584 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1586 unregister_pernet_device(&sit_net_ops);
1590 module_init(sit_init);
1591 module_exit(sit_cleanup);
1592 MODULE_LICENSE("GPL");
1593 MODULE_ALIAS_NETDEV("sit0");