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 int ipip6_tunnel_init(struct net_device *dev);
67 static void ipip6_tunnel_setup(struct net_device *dev);
68 static void ipip6_dev_free(struct net_device *dev);
70 static int sit_net_id __read_mostly;
72 struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
73 struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
74 struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
75 struct ip_tunnel __rcu *tunnels_wc[1];
76 struct ip_tunnel __rcu **tunnels[4];
78 struct net_device *fb_tunnel_dev;
82 * Locking : hash tables are protected by RCU and RTNL
85 #define for_each_ip_tunnel_rcu(start) \
86 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
88 /* often modified stats are per cpu, other are shared (netdev->stats) */
90 unsigned long rx_packets;
91 unsigned long rx_bytes;
92 unsigned long tx_packets;
93 unsigned long tx_bytes;
96 static struct net_device_stats *ipip6_get_stats(struct net_device *dev)
98 struct pcpu_tstats sum = { 0 };
101 for_each_possible_cpu(i) {
102 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
104 sum.rx_packets += tstats->rx_packets;
105 sum.rx_bytes += tstats->rx_bytes;
106 sum.tx_packets += tstats->tx_packets;
107 sum.tx_bytes += tstats->tx_bytes;
109 dev->stats.rx_packets = sum.rx_packets;
110 dev->stats.rx_bytes = sum.rx_bytes;
111 dev->stats.tx_packets = sum.tx_packets;
112 dev->stats.tx_bytes = sum.tx_bytes;
116 * Must be invoked with rcu_read_lock
118 static struct ip_tunnel * ipip6_tunnel_lookup(struct net *net,
119 struct net_device *dev, __be32 remote, __be32 local)
121 unsigned int h0 = HASH(remote);
122 unsigned int h1 = HASH(local);
124 struct sit_net *sitn = net_generic(net, sit_net_id);
126 for_each_ip_tunnel_rcu(sitn->tunnels_r_l[h0 ^ h1]) {
127 if (local == t->parms.iph.saddr &&
128 remote == t->parms.iph.daddr &&
129 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
130 (t->dev->flags & IFF_UP))
133 for_each_ip_tunnel_rcu(sitn->tunnels_r[h0]) {
134 if (remote == t->parms.iph.daddr &&
135 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
136 (t->dev->flags & IFF_UP))
139 for_each_ip_tunnel_rcu(sitn->tunnels_l[h1]) {
140 if (local == t->parms.iph.saddr &&
141 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
142 (t->dev->flags & IFF_UP))
145 t = rcu_dereference(sitn->tunnels_wc[0]);
146 if ((t != NULL) && (t->dev->flags & IFF_UP))
151 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
152 struct ip_tunnel_parm *parms)
154 __be32 remote = parms->iph.daddr;
155 __be32 local = parms->iph.saddr;
167 return &sitn->tunnels[prio][h];
170 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
173 return __ipip6_bucket(sitn, &t->parms);
176 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
178 struct ip_tunnel __rcu **tp;
179 struct ip_tunnel *iter;
181 for (tp = ipip6_bucket(sitn, t);
182 (iter = rtnl_dereference(*tp)) != NULL;
185 rcu_assign_pointer(*tp, t->next);
191 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
193 struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
195 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
196 rcu_assign_pointer(*tp, t);
199 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
201 #ifdef CONFIG_IPV6_SIT_6RD
202 struct ip_tunnel *t = netdev_priv(dev);
204 if (t->dev == sitn->fb_tunnel_dev) {
205 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
206 t->ip6rd.relay_prefix = 0;
207 t->ip6rd.prefixlen = 16;
208 t->ip6rd.relay_prefixlen = 0;
210 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
211 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
216 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
217 struct ip_tunnel_parm *parms, int create)
219 __be32 remote = parms->iph.daddr;
220 __be32 local = parms->iph.saddr;
221 struct ip_tunnel *t, *nt;
222 struct ip_tunnel __rcu **tp;
223 struct net_device *dev;
225 struct sit_net *sitn = net_generic(net, sit_net_id);
227 for (tp = __ipip6_bucket(sitn, parms);
228 (t = rtnl_dereference(*tp)) != NULL;
230 if (local == t->parms.iph.saddr &&
231 remote == t->parms.iph.daddr &&
232 parms->link == t->parms.link) {
243 strlcpy(name, parms->name, IFNAMSIZ);
245 strcpy(name, "sit%d");
247 dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
251 dev_net_set(dev, net);
253 if (strchr(name, '%')) {
254 if (dev_alloc_name(dev, name) < 0)
258 nt = netdev_priv(dev);
261 if (ipip6_tunnel_init(dev) < 0)
263 ipip6_tunnel_clone_6rd(dev, sitn);
265 if (parms->i_flags & SIT_ISATAP)
266 dev->priv_flags |= IFF_ISATAP;
268 if (register_netdevice(dev) < 0)
273 ipip6_tunnel_link(sitn, nt);
282 #define for_each_prl_rcu(start) \
283 for (prl = rcu_dereference(start); \
285 prl = rcu_dereference(prl->next))
287 static struct ip_tunnel_prl_entry *
288 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
290 struct ip_tunnel_prl_entry *prl;
292 for_each_prl_rcu(t->prl)
293 if (prl->addr == addr)
299 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
300 struct ip_tunnel_prl __user *a)
302 struct ip_tunnel_prl kprl, *kp;
303 struct ip_tunnel_prl_entry *prl;
304 unsigned int cmax, c = 0, ca, len;
307 if (copy_from_user(&kprl, a, sizeof(kprl)))
309 cmax = kprl.datalen / sizeof(kprl);
310 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
313 /* For simple GET or for root users,
314 * we try harder to allocate.
316 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
317 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
322 ca = t->prl_count < cmax ? t->prl_count : cmax;
325 /* We don't try hard to allocate much memory for
327 * For root users, retry allocating enough memory for
330 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
338 for_each_prl_rcu(t->prl) {
341 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
343 kp[c].addr = prl->addr;
344 kp[c].flags = prl->flags;
346 if (kprl.addr != htonl(INADDR_ANY))
352 len = sizeof(*kp) * c;
354 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
363 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
365 struct ip_tunnel_prl_entry *p;
368 if (a->addr == htonl(INADDR_ANY))
373 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
374 if (p->addr == a->addr) {
389 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
399 rcu_assign_pointer(t->prl, p);
404 static void prl_entry_destroy_rcu(struct rcu_head *head)
406 kfree(container_of(head, struct ip_tunnel_prl_entry, rcu_head));
409 static void prl_list_destroy_rcu(struct rcu_head *head)
411 struct ip_tunnel_prl_entry *p, *n;
413 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
422 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
424 struct ip_tunnel_prl_entry *x, **p;
429 if (a && a->addr != htonl(INADDR_ANY)) {
430 for (p = &t->prl; *p; p = &(*p)->next) {
431 if ((*p)->addr == a->addr) {
434 call_rcu(&x->rcu_head, prl_entry_destroy_rcu);
444 call_rcu(&x->rcu_head, prl_list_destroy_rcu);
453 isatap_chksrc(struct sk_buff *skb, struct iphdr *iph, struct ip_tunnel *t)
455 struct ip_tunnel_prl_entry *p;
459 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
461 if (p->flags & PRL_DEFAULT)
462 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
464 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
466 struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
467 if (ipv6_addr_is_isatap(addr6) &&
468 (addr6->s6_addr32[3] == iph->saddr) &&
469 ipv6_chk_prefix(addr6, t->dev))
470 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
478 static void ipip6_tunnel_uninit(struct net_device *dev)
480 struct net *net = dev_net(dev);
481 struct sit_net *sitn = net_generic(net, sit_net_id);
483 if (dev == sitn->fb_tunnel_dev) {
484 rcu_assign_pointer(sitn->tunnels_wc[0], NULL);
486 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
487 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
493 static int ipip6_err(struct sk_buff *skb, u32 info)
496 /* All the routers (except for Linux) return only
497 8 bytes of packet payload. It means, that precise relaying of
498 ICMP in the real Internet is absolutely infeasible.
500 struct iphdr *iph = (struct iphdr*)skb->data;
501 const int type = icmp_hdr(skb)->type;
502 const int code = icmp_hdr(skb)->code;
508 case ICMP_PARAMETERPROB:
511 case ICMP_DEST_UNREACH:
514 case ICMP_PORT_UNREACH:
515 /* Impossible event. */
517 case ICMP_FRAG_NEEDED:
518 /* Soft state for pmtu is maintained by IP core. */
521 /* All others are translated to HOST_UNREACH.
522 rfc2003 contains "deep thoughts" about NET_UNREACH,
523 I believe they are just ether pollution. --ANK
528 case ICMP_TIME_EXCEEDED:
529 if (code != ICMP_EXC_TTL)
537 t = ipip6_tunnel_lookup(dev_net(skb->dev),
541 if (t == NULL || t->parms.iph.daddr == 0)
545 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
548 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
552 t->err_time = jiffies;
558 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
560 if (INET_ECN_is_ce(iph->tos))
561 IP6_ECN_set_ce(ipv6_hdr(skb));
564 static int ipip6_rcv(struct sk_buff *skb)
567 struct ip_tunnel *tunnel;
569 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
575 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
576 iph->saddr, iph->daddr);
577 if (tunnel != NULL) {
578 struct pcpu_tstats *tstats;
581 skb->mac_header = skb->network_header;
582 skb_reset_network_header(skb);
583 IPCB(skb)->flags = 0;
584 skb->protocol = htons(ETH_P_IPV6);
585 skb->pkt_type = PACKET_HOST;
587 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
588 !isatap_chksrc(skb, iph, tunnel)) {
589 tunnel->dev->stats.rx_errors++;
595 tstats = this_cpu_ptr(tunnel->dev->tstats);
596 tstats->rx_packets++;
597 tstats->rx_bytes += skb->len;
599 __skb_tunnel_rx(skb, tunnel->dev);
601 ipip6_ecn_decapsulate(iph, skb);
609 /* no tunnel matched, let upstream know, ipsec may handle it */
618 * Returns the embedded IPv4 address if the IPv6 address
619 * comes from 6rd / 6to4 (RFC 3056) addr space.
622 __be32 try_6rd(struct in6_addr *v6dst, struct ip_tunnel *tunnel)
626 #ifdef CONFIG_IPV6_SIT_6RD
627 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
628 tunnel->ip6rd.prefixlen)) {
629 unsigned int pbw0, pbi0;
633 pbw0 = tunnel->ip6rd.prefixlen >> 5;
634 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
636 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
637 tunnel->ip6rd.relay_prefixlen;
639 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
641 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
644 dst = tunnel->ip6rd.relay_prefix | htonl(d);
647 if (v6dst->s6_addr16[0] == htons(0x2002)) {
648 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
649 memcpy(&dst, &v6dst->s6_addr16[1], 4);
656 * This function assumes it is being called from dev_queue_xmit()
657 * and that skb is filled properly by that function.
660 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
661 struct net_device *dev)
663 struct ip_tunnel *tunnel = netdev_priv(dev);
664 struct pcpu_tstats *tstats;
665 struct iphdr *tiph = &tunnel->parms.iph;
666 struct ipv6hdr *iph6 = ipv6_hdr(skb);
667 u8 tos = tunnel->parms.iph.tos;
668 __be16 df = tiph->frag_off;
669 struct rtable *rt; /* Route to the other host */
670 struct net_device *tdev; /* Device to other host */
671 struct iphdr *iph; /* Our new IP header */
672 unsigned int max_headroom; /* The extra header space needed */
673 __be32 dst = tiph->daddr;
675 struct in6_addr *addr6;
678 if (skb->protocol != htons(ETH_P_IPV6))
681 /* ISATAP (RFC4214) - must come before 6to4 */
682 if (dev->priv_flags & IFF_ISATAP) {
683 struct neighbour *neigh = NULL;
686 neigh = skb_dst(skb)->neighbour;
690 printk(KERN_DEBUG "sit: nexthop == NULL\n");
694 addr6 = (struct in6_addr*)&neigh->primary_key;
695 addr_type = ipv6_addr_type(addr6);
697 if ((addr_type & IPV6_ADDR_UNICAST) &&
698 ipv6_addr_is_isatap(addr6))
699 dst = addr6->s6_addr32[3];
705 dst = try_6rd(&iph6->daddr, tunnel);
708 struct neighbour *neigh = NULL;
711 neigh = skb_dst(skb)->neighbour;
715 printk(KERN_DEBUG "sit: nexthop == NULL\n");
719 addr6 = (struct in6_addr*)&neigh->primary_key;
720 addr_type = ipv6_addr_type(addr6);
722 if (addr_type == IPV6_ADDR_ANY) {
723 addr6 = &ipv6_hdr(skb)->daddr;
724 addr_type = ipv6_addr_type(addr6);
727 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
730 dst = addr6->s6_addr32[3];
734 struct flowi fl = { .fl4_dst = dst,
735 .fl4_src = tiph->saddr,
736 .fl4_tos = RT_TOS(tos),
737 .oif = tunnel->parms.link,
738 .proto = IPPROTO_IPV6 };
739 if (ip_route_output_key(dev_net(dev), &rt, &fl)) {
740 dev->stats.tx_carrier_errors++;
744 if (rt->rt_type != RTN_UNICAST) {
746 dev->stats.tx_carrier_errors++;
753 dev->stats.collisions++;
758 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
761 dev->stats.collisions++;
766 if (mtu < IPV6_MIN_MTU) {
771 if (tunnel->parms.iph.daddr && skb_dst(skb))
772 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
774 if (skb->len > mtu) {
775 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
781 if (tunnel->err_count > 0) {
782 if (time_before(jiffies,
783 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
785 dst_link_failure(skb);
787 tunnel->err_count = 0;
791 * Okay, now see if we can stuff it in the buffer as-is.
793 max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
795 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
796 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
797 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
800 dev->stats.tx_dropped++;
805 skb_set_owner_w(new_skb, skb->sk);
808 iph6 = ipv6_hdr(skb);
811 skb->transport_header = skb->network_header;
812 skb_push(skb, sizeof(struct iphdr));
813 skb_reset_network_header(skb);
814 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
815 IPCB(skb)->flags = 0;
817 skb_dst_set(skb, &rt->dst);
820 * Push down and install the IPIP header.
825 iph->ihl = sizeof(struct iphdr)>>2;
827 iph->protocol = IPPROTO_IPV6;
828 iph->tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
829 iph->daddr = rt->rt_dst;
830 iph->saddr = rt->rt_src;
832 if ((iph->ttl = tiph->ttl) == 0)
833 iph->ttl = iph6->hop_limit;
836 tstats = this_cpu_ptr(dev->tstats);
837 __IPTUNNEL_XMIT(tstats, &dev->stats);
841 dst_link_failure(skb);
843 dev->stats.tx_errors++;
848 static void ipip6_tunnel_bind_dev(struct net_device *dev)
850 struct net_device *tdev = NULL;
851 struct ip_tunnel *tunnel;
854 tunnel = netdev_priv(dev);
855 iph = &tunnel->parms.iph;
858 struct flowi fl = { .fl4_dst = iph->daddr,
859 .fl4_src = iph->saddr,
860 .fl4_tos = RT_TOS(iph->tos),
861 .oif = tunnel->parms.link,
862 .proto = IPPROTO_IPV6 };
864 if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
868 dev->flags |= IFF_POINTOPOINT;
871 if (!tdev && tunnel->parms.link)
872 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
875 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
876 dev->mtu = tdev->mtu - sizeof(struct iphdr);
877 if (dev->mtu < IPV6_MIN_MTU)
878 dev->mtu = IPV6_MIN_MTU;
880 dev->iflink = tunnel->parms.link;
884 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
887 struct ip_tunnel_parm p;
888 struct ip_tunnel_prl prl;
890 struct net *net = dev_net(dev);
891 struct sit_net *sitn = net_generic(net, sit_net_id);
892 #ifdef CONFIG_IPV6_SIT_6RD
893 struct ip_tunnel_6rd ip6rd;
898 #ifdef CONFIG_IPV6_SIT_6RD
902 if (dev == sitn->fb_tunnel_dev) {
903 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
907 t = ipip6_tunnel_locate(net, &p, 0);
910 t = netdev_priv(dev);
913 if (cmd == SIOCGETTUNNEL) {
914 memcpy(&p, &t->parms, sizeof(p));
915 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
918 #ifdef CONFIG_IPV6_SIT_6RD
920 ipv6_addr_copy(&ip6rd.prefix, &t->ip6rd.prefix);
921 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
922 ip6rd.prefixlen = t->ip6rd.prefixlen;
923 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
924 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
935 if (!capable(CAP_NET_ADMIN))
939 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
943 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
944 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
947 p.iph.frag_off |= htons(IP_DF);
949 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
951 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
958 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
959 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
963 t = netdev_priv(dev);
964 ipip6_tunnel_unlink(sitn, t);
966 t->parms.iph.saddr = p.iph.saddr;
967 t->parms.iph.daddr = p.iph.daddr;
968 memcpy(dev->dev_addr, &p.iph.saddr, 4);
969 memcpy(dev->broadcast, &p.iph.daddr, 4);
970 ipip6_tunnel_link(sitn, t);
971 netdev_state_change(dev);
977 if (cmd == SIOCCHGTUNNEL) {
978 t->parms.iph.ttl = p.iph.ttl;
979 t->parms.iph.tos = p.iph.tos;
980 if (t->parms.link != p.link) {
981 t->parms.link = p.link;
982 ipip6_tunnel_bind_dev(dev);
983 netdev_state_change(dev);
986 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
989 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
994 if (!capable(CAP_NET_ADMIN))
997 if (dev == sitn->fb_tunnel_dev) {
999 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1002 if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1005 if (t == netdev_priv(sitn->fb_tunnel_dev))
1009 unregister_netdevice(dev);
1015 if (dev == sitn->fb_tunnel_dev)
1018 if (!(t = netdev_priv(dev)))
1020 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1027 if (!capable(CAP_NET_ADMIN))
1030 if (dev == sitn->fb_tunnel_dev)
1033 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1036 if (!(t = netdev_priv(dev)))
1041 err = ipip6_tunnel_del_prl(t, &prl);
1045 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1048 netdev_state_change(dev);
1051 #ifdef CONFIG_IPV6_SIT_6RD
1056 if (!capable(CAP_NET_ADMIN))
1060 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1064 t = netdev_priv(dev);
1066 if (cmd != SIOCDEL6RD) {
1067 struct in6_addr prefix;
1068 __be32 relay_prefix;
1071 if (ip6rd.relay_prefixlen > 32 ||
1072 ip6rd.prefixlen + (32 - ip6rd.relay_prefixlen) > 64)
1075 ipv6_addr_prefix(&prefix, &ip6rd.prefix,
1077 if (!ipv6_addr_equal(&prefix, &ip6rd.prefix))
1079 if (ip6rd.relay_prefixlen)
1080 relay_prefix = ip6rd.relay_prefix &
1081 htonl(0xffffffffUL <<
1082 (32 - ip6rd.relay_prefixlen));
1085 if (relay_prefix != ip6rd.relay_prefix)
1088 ipv6_addr_copy(&t->ip6rd.prefix, &prefix);
1089 t->ip6rd.relay_prefix = relay_prefix;
1090 t->ip6rd.prefixlen = ip6rd.prefixlen;
1091 t->ip6rd.relay_prefixlen = ip6rd.relay_prefixlen;
1093 ipip6_tunnel_clone_6rd(dev, sitn);
1107 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1109 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1115 static const struct net_device_ops ipip6_netdev_ops = {
1116 .ndo_uninit = ipip6_tunnel_uninit,
1117 .ndo_start_xmit = ipip6_tunnel_xmit,
1118 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1119 .ndo_change_mtu = ipip6_tunnel_change_mtu,
1120 .ndo_get_stats = ipip6_get_stats,
1123 static void ipip6_dev_free(struct net_device *dev)
1125 free_percpu(dev->tstats);
1129 static void ipip6_tunnel_setup(struct net_device *dev)
1131 dev->netdev_ops = &ipip6_netdev_ops;
1132 dev->destructor = ipip6_dev_free;
1134 dev->type = ARPHRD_SIT;
1135 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
1136 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr);
1137 dev->flags = IFF_NOARP;
1138 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1141 dev->features |= NETIF_F_NETNS_LOCAL;
1142 dev->features |= NETIF_F_LLTX;
1145 static int ipip6_tunnel_init(struct net_device *dev)
1147 struct ip_tunnel *tunnel = netdev_priv(dev);
1150 strcpy(tunnel->parms.name, dev->name);
1152 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1153 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1155 ipip6_tunnel_bind_dev(dev);
1156 dev->tstats = alloc_percpu(struct pcpu_tstats);
1163 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1165 struct ip_tunnel *tunnel = netdev_priv(dev);
1166 struct iphdr *iph = &tunnel->parms.iph;
1167 struct net *net = dev_net(dev);
1168 struct sit_net *sitn = net_generic(net, sit_net_id);
1171 strcpy(tunnel->parms.name, dev->name);
1174 iph->protocol = IPPROTO_IPV6;
1178 dev->tstats = alloc_percpu(struct pcpu_tstats);
1182 sitn->tunnels_wc[0] = tunnel;
1186 static struct xfrm_tunnel sit_handler __read_mostly = {
1187 .handler = ipip6_rcv,
1188 .err_handler = ipip6_err,
1192 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1196 for (prio = 1; prio < 4; prio++) {
1198 for (h = 0; h < HASH_SIZE; h++) {
1199 struct ip_tunnel *t = sitn->tunnels[prio][h];
1202 unregister_netdevice_queue(t->dev, head);
1209 static int __net_init sit_init_net(struct net *net)
1211 struct sit_net *sitn = net_generic(net, sit_net_id);
1214 sitn->tunnels[0] = sitn->tunnels_wc;
1215 sitn->tunnels[1] = sitn->tunnels_l;
1216 sitn->tunnels[2] = sitn->tunnels_r;
1217 sitn->tunnels[3] = sitn->tunnels_r_l;
1219 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1220 ipip6_tunnel_setup);
1221 if (!sitn->fb_tunnel_dev) {
1225 dev_net_set(sitn->fb_tunnel_dev, net);
1227 err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1231 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1233 if ((err = register_netdev(sitn->fb_tunnel_dev)))
1239 dev_put(sitn->fb_tunnel_dev);
1241 ipip6_dev_free(sitn->fb_tunnel_dev);
1246 static void __net_exit sit_exit_net(struct net *net)
1248 struct sit_net *sitn = net_generic(net, sit_net_id);
1252 sit_destroy_tunnels(sitn, &list);
1253 unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1254 unregister_netdevice_many(&list);
1258 static struct pernet_operations sit_net_ops = {
1259 .init = sit_init_net,
1260 .exit = sit_exit_net,
1262 .size = sizeof(struct sit_net),
1265 static void __exit sit_cleanup(void)
1267 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1269 unregister_pernet_device(&sit_net_ops);
1270 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1273 static int __init sit_init(void)
1277 printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1279 err = register_pernet_device(&sit_net_ops);
1282 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1284 unregister_pernet_device(&sit_net_ops);
1285 printk(KERN_INFO "sit init: Can't add protocol\n");
1290 module_init(sit_init);
1291 module_exit(sit_cleanup);
1292 MODULE_LICENSE("GPL");
1293 MODULE_ALIAS_NETDEV("sit0");