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>
52 #include <net/ip_tunnels.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 bool log_ecn_error = true;
69 module_param(log_ecn_error, bool, 0644);
70 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
72 static int ipip6_tunnel_init(struct net_device *dev);
73 static void ipip6_tunnel_setup(struct net_device *dev);
74 static void ipip6_dev_free(struct net_device *dev);
75 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
77 static struct rtnl_link_ops sit_link_ops __read_mostly;
79 static int sit_net_id __read_mostly;
81 struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
82 struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
83 struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
84 struct ip_tunnel __rcu *tunnels_wc[1];
85 struct ip_tunnel __rcu **tunnels[4];
87 struct net_device *fb_tunnel_dev;
91 * Must be invoked with rcu_read_lock
93 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
94 struct net_device *dev, __be32 remote, __be32 local)
96 unsigned int h0 = HASH(remote);
97 unsigned int h1 = HASH(local);
99 struct sit_net *sitn = net_generic(net, sit_net_id);
101 for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
102 if (local == t->parms.iph.saddr &&
103 remote == t->parms.iph.daddr &&
104 (!dev || !t->parms.link || dev->ifindex == t->parms.link) &&
105 (t->dev->flags & IFF_UP))
108 for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
109 if (remote == t->parms.iph.daddr &&
110 (!dev || !t->parms.link || dev->ifindex == t->parms.link) &&
111 (t->dev->flags & IFF_UP))
114 for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
115 if (local == t->parms.iph.saddr &&
116 (!dev || !t->parms.link || dev->ifindex == t->parms.link) &&
117 (t->dev->flags & IFF_UP))
120 t = rcu_dereference(sitn->tunnels_wc[0]);
121 if (t && (t->dev->flags & IFF_UP))
126 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
127 struct ip_tunnel_parm *parms)
129 __be32 remote = parms->iph.daddr;
130 __be32 local = parms->iph.saddr;
142 return &sitn->tunnels[prio][h];
145 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
148 return __ipip6_bucket(sitn, &t->parms);
151 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
153 struct ip_tunnel __rcu **tp;
154 struct ip_tunnel *iter;
156 for (tp = ipip6_bucket(sitn, t);
157 (iter = rtnl_dereference(*tp)) != NULL;
160 rcu_assign_pointer(*tp, t->next);
166 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
168 struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
170 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
171 rcu_assign_pointer(*tp, t);
174 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
176 #ifdef CONFIG_IPV6_SIT_6RD
177 struct ip_tunnel *t = netdev_priv(dev);
179 if (dev == sitn->fb_tunnel_dev) {
180 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
181 t->ip6rd.relay_prefix = 0;
182 t->ip6rd.prefixlen = 16;
183 t->ip6rd.relay_prefixlen = 0;
185 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
186 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
191 static int ipip6_tunnel_create(struct net_device *dev)
193 struct ip_tunnel *t = netdev_priv(dev);
194 struct net *net = dev_net(dev);
195 struct sit_net *sitn = net_generic(net, sit_net_id);
198 memcpy(dev->dev_addr, &t->parms.iph.saddr, 4);
199 memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
201 if ((__force u16)t->parms.i_flags & SIT_ISATAP)
202 dev->priv_flags |= IFF_ISATAP;
204 err = register_netdevice(dev);
208 ipip6_tunnel_clone_6rd(dev, sitn);
210 dev->rtnl_link_ops = &sit_link_ops;
212 ipip6_tunnel_link(sitn, t);
219 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
220 struct ip_tunnel_parm *parms, int create)
222 __be32 remote = parms->iph.daddr;
223 __be32 local = parms->iph.saddr;
224 struct ip_tunnel *t, *nt;
225 struct ip_tunnel __rcu **tp;
226 struct net_device *dev;
228 struct sit_net *sitn = net_generic(net, sit_net_id);
230 for (tp = __ipip6_bucket(sitn, parms);
231 (t = rtnl_dereference(*tp)) != NULL;
233 if (local == t->parms.iph.saddr &&
234 remote == t->parms.iph.daddr &&
235 parms->link == t->parms.link) {
245 if (parms->name[0]) {
246 if (!dev_valid_name(parms->name))
248 strlcpy(name, parms->name, IFNAMSIZ);
250 strcpy(name, "sit%d");
252 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
257 dev_net_set(dev, net);
259 nt = netdev_priv(dev);
262 if (ipip6_tunnel_create(dev) < 0)
273 #define for_each_prl_rcu(start) \
274 for (prl = rcu_dereference(start); \
276 prl = rcu_dereference(prl->next))
278 static struct ip_tunnel_prl_entry *
279 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
281 struct ip_tunnel_prl_entry *prl;
283 for_each_prl_rcu(t->prl)
284 if (prl->addr == addr)
290 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
291 struct ip_tunnel_prl __user *a)
293 struct ip_tunnel_prl kprl, *kp;
294 struct ip_tunnel_prl_entry *prl;
295 unsigned int cmax, c = 0, ca, len;
298 if (copy_from_user(&kprl, a, sizeof(kprl)))
300 cmax = kprl.datalen / sizeof(kprl);
301 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
304 /* For simple GET or for root users,
305 * we try harder to allocate.
307 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
308 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
313 ca = t->prl_count < cmax ? t->prl_count : cmax;
316 /* We don't try hard to allocate much memory for
318 * For root users, retry allocating enough memory for
321 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
329 for_each_prl_rcu(t->prl) {
332 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
334 kp[c].addr = prl->addr;
335 kp[c].flags = prl->flags;
337 if (kprl.addr != htonl(INADDR_ANY))
343 len = sizeof(*kp) * c;
345 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
354 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
356 struct ip_tunnel_prl_entry *p;
359 if (a->addr == htonl(INADDR_ANY))
364 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
365 if (p->addr == a->addr) {
380 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
390 rcu_assign_pointer(t->prl, p);
395 static void prl_list_destroy_rcu(struct rcu_head *head)
397 struct ip_tunnel_prl_entry *p, *n;
399 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
401 n = rcu_dereference_protected(p->next, 1);
408 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
410 struct ip_tunnel_prl_entry *x;
411 struct ip_tunnel_prl_entry __rcu **p;
416 if (a && a->addr != htonl(INADDR_ANY)) {
418 (x = rtnl_dereference(*p)) != NULL;
420 if (x->addr == a->addr) {
422 kfree_rcu(x, rcu_head);
429 x = rtnl_dereference(t->prl);
432 call_rcu(&x->rcu_head, prl_list_destroy_rcu);
441 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
443 struct ip_tunnel_prl_entry *p;
447 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
449 if (p->flags & PRL_DEFAULT)
450 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
452 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
454 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
456 if (ipv6_addr_is_isatap(addr6) &&
457 (addr6->s6_addr32[3] == iph->saddr) &&
458 ipv6_chk_prefix(addr6, t->dev))
459 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
467 static void ipip6_tunnel_uninit(struct net_device *dev)
469 struct ip_tunnel *tunnel = netdev_priv(dev);
470 struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
472 if (dev == sitn->fb_tunnel_dev) {
473 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
475 ipip6_tunnel_unlink(sitn, tunnel);
476 ipip6_tunnel_del_prl(tunnel, NULL);
478 dst_cache_reset(&tunnel->dst_cache);
482 /* Generate icmpv6 with type/code ICMPV6_DEST_UNREACH/ICMPV6_ADDR_UNREACH
483 * if sufficient data bytes are available
485 static int ipip6_err_gen_icmpv6_unreach(struct sk_buff *skb)
487 int ihl = ((const struct iphdr *)skb->data)->ihl*4;
489 struct sk_buff *skb2;
491 if (!pskb_may_pull(skb, ihl + sizeof(struct ipv6hdr) + 8))
494 skb2 = skb_clone(skb, GFP_ATOMIC);
501 skb_reset_network_header(skb2);
503 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, NULL, 0, 0);
505 if (rt && rt->dst.dev)
506 skb2->dev = rt->dst.dev;
508 icmpv6_send(skb2, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
518 static int ipip6_err(struct sk_buff *skb, u32 info)
520 const struct iphdr *iph = (const struct iphdr *)skb->data;
521 const int type = icmp_hdr(skb)->type;
522 const int code = icmp_hdr(skb)->code;
528 case ICMP_PARAMETERPROB:
531 case ICMP_DEST_UNREACH:
534 /* Impossible event. */
537 /* All others are translated to HOST_UNREACH.
538 rfc2003 contains "deep thoughts" about NET_UNREACH,
539 I believe they are just ether pollution. --ANK
544 case ICMP_TIME_EXCEEDED:
545 if (code != ICMP_EXC_TTL)
554 t = ipip6_tunnel_lookup(dev_net(skb->dev),
561 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
562 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
563 t->parms.link, 0, iph->protocol, 0);
567 if (type == ICMP_REDIRECT) {
568 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 0,
574 if (t->parms.iph.daddr == 0)
578 if (__in6_dev_get(skb->dev) && !ipip6_err_gen_icmpv6_unreach(skb))
581 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
584 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
588 t->err_time = jiffies;
593 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
594 const struct in6_addr *v6addr)
597 if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
602 /* Checks if an address matches an address on the tunnel interface.
603 * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
605 * This function is called after we considered the packet as spoofed
607 * We may have a router that is doing NAT for proto 41 packets
608 * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
609 * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
610 * function will return true, dropping the packet.
611 * But, we can still check if is spoofed against the IP
612 * addresses associated with the interface.
614 static bool only_dnatted(const struct ip_tunnel *tunnel,
615 const struct in6_addr *v6dst)
619 #ifdef CONFIG_IPV6_SIT_6RD
620 prefix_len = tunnel->ip6rd.prefixlen + 32
621 - tunnel->ip6rd.relay_prefixlen;
625 return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
628 /* Returns true if a packet is spoofed */
629 static bool packet_is_spoofed(struct sk_buff *skb,
630 const struct iphdr *iph,
631 struct ip_tunnel *tunnel)
633 const struct ipv6hdr *ipv6h;
635 if (tunnel->dev->priv_flags & IFF_ISATAP) {
636 if (!isatap_chksrc(skb, iph, tunnel))
642 if (tunnel->dev->flags & IFF_POINTOPOINT)
645 ipv6h = ipv6_hdr(skb);
647 if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
648 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
649 &iph->saddr, &ipv6h->saddr,
650 &iph->daddr, &ipv6h->daddr);
654 if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
657 if (only_dnatted(tunnel, &ipv6h->daddr))
660 net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
661 &iph->saddr, &ipv6h->saddr,
662 &iph->daddr, &ipv6h->daddr);
666 static int ipip6_rcv(struct sk_buff *skb)
668 const struct iphdr *iph = ip_hdr(skb);
669 struct ip_tunnel *tunnel;
672 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
673 iph->saddr, iph->daddr);
675 struct pcpu_sw_netstats *tstats;
677 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
678 tunnel->parms.iph.protocol != 0)
681 skb->mac_header = skb->network_header;
682 skb_reset_network_header(skb);
683 IPCB(skb)->flags = 0;
684 skb->dev = tunnel->dev;
686 if (packet_is_spoofed(skb, iph, tunnel)) {
687 tunnel->dev->stats.rx_errors++;
691 if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6)))
694 /* skb can be uncloned in iptunnel_pull_header, so
695 * old iph is no longer valid
697 iph = (const struct iphdr *)skb_mac_header(skb);
698 err = IP_ECN_decapsulate(iph, skb);
701 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
702 &iph->saddr, iph->tos);
704 ++tunnel->dev->stats.rx_frame_errors;
705 ++tunnel->dev->stats.rx_errors;
710 tstats = this_cpu_ptr(tunnel->dev->tstats);
711 u64_stats_update_begin(&tstats->syncp);
712 tstats->rx_packets++;
713 tstats->rx_bytes += skb->len;
714 u64_stats_update_end(&tstats->syncp);
721 /* no tunnel matched, let upstream know, ipsec may handle it */
728 static const struct tnl_ptk_info tpi = {
729 /* no tunnel info required for ipip. */
730 .proto = htons(ETH_P_IP),
733 static int ipip_rcv(struct sk_buff *skb)
735 const struct iphdr *iph;
736 struct ip_tunnel *tunnel;
739 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
740 iph->saddr, iph->daddr);
742 if (tunnel->parms.iph.protocol != IPPROTO_IPIP &&
743 tunnel->parms.iph.protocol != 0)
746 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
748 if (iptunnel_pull_header(skb, 0, tpi.proto))
750 return ip_tunnel_rcv(tunnel, skb, &tpi, NULL, log_ecn_error);
761 * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
762 * stores the embedded IPv4 address in v4dst and returns true.
764 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
767 #ifdef CONFIG_IPV6_SIT_6RD
768 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
769 tunnel->ip6rd.prefixlen)) {
770 unsigned int pbw0, pbi0;
774 pbw0 = tunnel->ip6rd.prefixlen >> 5;
775 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
777 d = tunnel->ip6rd.relay_prefixlen < 32 ?
778 (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
779 tunnel->ip6rd.relay_prefixlen : 0;
781 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
783 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
786 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
790 if (v6dst->s6_addr16[0] == htons(0x2002)) {
791 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
792 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
799 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
800 const struct in6_addr *v6dst)
803 check_6rd(tunnel, v6dst, &dst);
808 * This function assumes it is being called from dev_queue_xmit()
809 * and that skb is filled properly by that function.
812 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
813 struct net_device *dev)
815 struct ip_tunnel *tunnel = netdev_priv(dev);
816 const struct iphdr *tiph = &tunnel->parms.iph;
817 const struct ipv6hdr *iph6 = ipv6_hdr(skb);
818 u8 tos = tunnel->parms.iph.tos;
819 __be16 df = tiph->frag_off;
820 struct rtable *rt; /* Route to the other host */
821 struct net_device *tdev; /* Device to other host */
822 unsigned int max_headroom; /* The extra header space needed */
823 __be32 dst = tiph->daddr;
826 const struct in6_addr *addr6;
830 u8 protocol = IPPROTO_IPV6;
831 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
833 if (skb->protocol != htons(ETH_P_IPV6))
837 tos = ipv6_get_dsfield(iph6);
839 /* ISATAP (RFC4214) - must come before 6to4 */
840 if (dev->priv_flags & IFF_ISATAP) {
841 struct neighbour *neigh = NULL;
842 bool do_tx_error = false;
845 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
848 net_dbg_ratelimited("nexthop == NULL\n");
852 addr6 = (const struct in6_addr *)&neigh->primary_key;
853 addr_type = ipv6_addr_type(addr6);
855 if ((addr_type & IPV6_ADDR_UNICAST) &&
856 ipv6_addr_is_isatap(addr6))
857 dst = addr6->s6_addr32[3];
861 neigh_release(neigh);
867 dst = try_6rd(tunnel, &iph6->daddr);
870 struct neighbour *neigh = NULL;
871 bool do_tx_error = false;
874 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
877 net_dbg_ratelimited("nexthop == NULL\n");
881 addr6 = (const struct in6_addr *)&neigh->primary_key;
882 addr_type = ipv6_addr_type(addr6);
884 if (addr_type == IPV6_ADDR_ANY) {
885 addr6 = &ipv6_hdr(skb)->daddr;
886 addr_type = ipv6_addr_type(addr6);
889 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
890 dst = addr6->s6_addr32[3];
894 neigh_release(neigh);
899 rt = ip_route_output_ports(tunnel->net, &fl4, NULL,
902 IPPROTO_IPV6, RT_TOS(tos),
905 dev->stats.tx_carrier_errors++;
908 if (rt->rt_type != RTN_UNICAST) {
910 dev->stats.tx_carrier_errors++;
917 dev->stats.collisions++;
921 skb = iptunnel_handle_offloads(skb, false, SKB_GSO_SIT);
928 mtu = dst_mtu(&rt->dst) - t_hlen;
931 dev->stats.collisions++;
936 if (mtu < IPV6_MIN_MTU) {
941 if (tunnel->parms.iph.daddr && skb_dst(skb))
942 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
944 if (skb->len > mtu && !skb_is_gso(skb)) {
945 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
951 if (tunnel->err_count > 0) {
952 if (time_before(jiffies,
953 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
955 dst_link_failure(skb);
957 tunnel->err_count = 0;
961 * Okay, now see if we can stuff it in the buffer as-is.
963 max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
965 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
966 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
967 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
970 dev->stats.tx_dropped++;
975 skb_set_owner_w(new_skb, skb->sk);
978 iph6 = ipv6_hdr(skb);
982 ttl = iph6->hop_limit;
983 tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
985 if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
990 skb_set_inner_ipproto(skb, IPPROTO_IPV6);
992 err = iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr,
993 protocol, tos, ttl, df,
994 !net_eq(tunnel->net, dev_net(dev)));
995 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
999 dst_link_failure(skb);
1003 dev->stats.tx_errors++;
1004 return NETDEV_TX_OK;
1007 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
1009 struct ip_tunnel *tunnel = netdev_priv(dev);
1010 const struct iphdr *tiph = &tunnel->parms.iph;
1012 skb = iptunnel_handle_offloads(skb, false, SKB_GSO_IPIP);
1016 skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1018 ip_tunnel_xmit(skb, dev, tiph, IPPROTO_IPIP);
1019 return NETDEV_TX_OK;
1021 dev->stats.tx_errors++;
1022 return NETDEV_TX_OK;
1025 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1026 struct net_device *dev)
1028 switch (skb->protocol) {
1029 case htons(ETH_P_IP):
1030 ipip_tunnel_xmit(skb, dev);
1032 case htons(ETH_P_IPV6):
1033 ipip6_tunnel_xmit(skb, dev);
1039 return NETDEV_TX_OK;
1042 dev->stats.tx_errors++;
1044 return NETDEV_TX_OK;
1048 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1050 struct net_device *tdev = NULL;
1051 struct ip_tunnel *tunnel;
1052 const struct iphdr *iph;
1055 tunnel = netdev_priv(dev);
1056 iph = &tunnel->parms.iph;
1059 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1061 iph->daddr, iph->saddr,
1065 tunnel->parms.link);
1071 dev->flags |= IFF_POINTOPOINT;
1074 if (!tdev && tunnel->parms.link)
1075 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1077 if (tdev && !netif_is_l3_master(tdev)) {
1078 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1080 dev->mtu = tdev->mtu - t_hlen;
1081 if (dev->mtu < IPV6_MIN_MTU)
1082 dev->mtu = IPV6_MIN_MTU;
1086 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
1088 struct net *net = t->net;
1089 struct sit_net *sitn = net_generic(net, sit_net_id);
1091 ipip6_tunnel_unlink(sitn, t);
1093 t->parms.iph.saddr = p->iph.saddr;
1094 t->parms.iph.daddr = p->iph.daddr;
1095 memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1096 memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1097 ipip6_tunnel_link(sitn, t);
1098 t->parms.iph.ttl = p->iph.ttl;
1099 t->parms.iph.tos = p->iph.tos;
1100 t->parms.iph.frag_off = p->iph.frag_off;
1101 if (t->parms.link != p->link) {
1102 t->parms.link = p->link;
1103 ipip6_tunnel_bind_dev(t->dev);
1105 dst_cache_reset(&t->dst_cache);
1106 netdev_state_change(t->dev);
1109 #ifdef CONFIG_IPV6_SIT_6RD
1110 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1111 struct ip_tunnel_6rd *ip6rd)
1113 struct in6_addr prefix;
1114 __be32 relay_prefix;
1116 if (ip6rd->relay_prefixlen > 32 ||
1117 ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1120 ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1121 if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1123 if (ip6rd->relay_prefixlen)
1124 relay_prefix = ip6rd->relay_prefix &
1125 htonl(0xffffffffUL <<
1126 (32 - ip6rd->relay_prefixlen));
1129 if (relay_prefix != ip6rd->relay_prefix)
1132 t->ip6rd.prefix = prefix;
1133 t->ip6rd.relay_prefix = relay_prefix;
1134 t->ip6rd.prefixlen = ip6rd->prefixlen;
1135 t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1136 dst_cache_reset(&t->dst_cache);
1137 netdev_state_change(t->dev);
1143 ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1146 struct ip_tunnel_parm p;
1147 struct ip_tunnel_prl prl;
1148 struct ip_tunnel *t = netdev_priv(dev);
1149 struct net *net = t->net;
1150 struct sit_net *sitn = net_generic(net, sit_net_id);
1151 #ifdef CONFIG_IPV6_SIT_6RD
1152 struct ip_tunnel_6rd ip6rd;
1157 #ifdef CONFIG_IPV6_SIT_6RD
1160 if (dev == sitn->fb_tunnel_dev) {
1161 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1165 t = ipip6_tunnel_locate(net, &p, 0);
1167 t = netdev_priv(dev);
1171 if (cmd == SIOCGETTUNNEL) {
1172 memcpy(&p, &t->parms, sizeof(p));
1173 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1176 #ifdef CONFIG_IPV6_SIT_6RD
1178 ip6rd.prefix = t->ip6rd.prefix;
1179 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1180 ip6rd.prefixlen = t->ip6rd.prefixlen;
1181 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1182 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1193 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1197 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1201 if (p.iph.protocol != IPPROTO_IPV6 &&
1202 p.iph.protocol != IPPROTO_IPIP &&
1203 p.iph.protocol != 0)
1205 if (p.iph.version != 4 ||
1206 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1209 p.iph.frag_off |= htons(IP_DF);
1211 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1213 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1215 if (t->dev != dev) {
1220 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1221 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1225 t = netdev_priv(dev);
1228 ipip6_tunnel_update(t, &p);
1233 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1236 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1241 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1244 if (dev == sitn->fb_tunnel_dev) {
1246 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1249 t = ipip6_tunnel_locate(net, &p, 0);
1253 if (t == netdev_priv(sitn->fb_tunnel_dev))
1257 unregister_netdevice(dev);
1263 if (dev == sitn->fb_tunnel_dev)
1265 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1272 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1275 if (dev == sitn->fb_tunnel_dev)
1278 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1283 err = ipip6_tunnel_del_prl(t, &prl);
1287 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1290 dst_cache_reset(&t->dst_cache);
1291 netdev_state_change(dev);
1294 #ifdef CONFIG_IPV6_SIT_6RD
1299 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1303 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1307 if (cmd != SIOCDEL6RD) {
1308 err = ipip6_tunnel_update_6rd(t, &ip6rd);
1312 ipip6_tunnel_clone_6rd(dev, sitn);
1326 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1328 struct ip_tunnel *tunnel = netdev_priv(dev);
1329 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1331 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - t_hlen)
1337 static const struct net_device_ops ipip6_netdev_ops = {
1338 .ndo_init = ipip6_tunnel_init,
1339 .ndo_uninit = ipip6_tunnel_uninit,
1340 .ndo_start_xmit = sit_tunnel_xmit,
1341 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1342 .ndo_change_mtu = ipip6_tunnel_change_mtu,
1343 .ndo_get_stats64 = ip_tunnel_get_stats64,
1344 .ndo_get_iflink = ip_tunnel_get_iflink,
1347 static void ipip6_dev_free(struct net_device *dev)
1349 struct ip_tunnel *tunnel = netdev_priv(dev);
1351 dst_cache_destroy(&tunnel->dst_cache);
1352 free_percpu(dev->tstats);
1356 #define SIT_FEATURES (NETIF_F_SG | \
1357 NETIF_F_FRAGLIST | \
1359 NETIF_F_GSO_SOFTWARE | \
1362 static void ipip6_tunnel_setup(struct net_device *dev)
1364 struct ip_tunnel *tunnel = netdev_priv(dev);
1365 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1367 dev->netdev_ops = &ipip6_netdev_ops;
1368 dev->destructor = ipip6_dev_free;
1370 dev->type = ARPHRD_SIT;
1371 dev->mtu = ETH_DATA_LEN - t_hlen;
1372 dev->flags = IFF_NOARP;
1373 netif_keep_dst(dev);
1375 dev->features |= NETIF_F_LLTX;
1376 dev->features |= SIT_FEATURES;
1377 dev->hw_features |= SIT_FEATURES;
1380 static int ipip6_tunnel_init(struct net_device *dev)
1382 struct ip_tunnel *tunnel = netdev_priv(dev);
1386 tunnel->net = dev_net(dev);
1387 strcpy(tunnel->parms.name, dev->name);
1389 ipip6_tunnel_bind_dev(dev);
1390 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1394 err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1396 free_percpu(dev->tstats);
1404 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1406 struct ip_tunnel *tunnel = netdev_priv(dev);
1407 struct iphdr *iph = &tunnel->parms.iph;
1408 struct net *net = dev_net(dev);
1409 struct sit_net *sitn = net_generic(net, sit_net_id);
1412 iph->protocol = IPPROTO_IPV6;
1416 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1419 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[])
1423 if (!data || !data[IFLA_IPTUN_PROTO])
1426 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1427 if (proto != IPPROTO_IPV6 &&
1428 proto != IPPROTO_IPIP &&
1435 static void ipip6_netlink_parms(struct nlattr *data[],
1436 struct ip_tunnel_parm *parms)
1438 memset(parms, 0, sizeof(*parms));
1440 parms->iph.version = 4;
1441 parms->iph.protocol = IPPROTO_IPV6;
1443 parms->iph.ttl = 64;
1448 if (data[IFLA_IPTUN_LINK])
1449 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1451 if (data[IFLA_IPTUN_LOCAL])
1452 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1454 if (data[IFLA_IPTUN_REMOTE])
1455 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1457 if (data[IFLA_IPTUN_TTL]) {
1458 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1460 parms->iph.frag_off = htons(IP_DF);
1463 if (data[IFLA_IPTUN_TOS])
1464 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1466 if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1467 parms->iph.frag_off = htons(IP_DF);
1469 if (data[IFLA_IPTUN_FLAGS])
1470 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1472 if (data[IFLA_IPTUN_PROTO])
1473 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1477 /* This function returns true when ENCAP attributes are present in the nl msg */
1478 static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1479 struct ip_tunnel_encap *ipencap)
1483 memset(ipencap, 0, sizeof(*ipencap));
1488 if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1490 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1493 if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1495 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1498 if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1500 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1503 if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1505 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1511 #ifdef CONFIG_IPV6_SIT_6RD
1512 /* This function returns true when 6RD attributes are present in the nl msg */
1513 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1514 struct ip_tunnel_6rd *ip6rd)
1517 memset(ip6rd, 0, sizeof(*ip6rd));
1522 if (data[IFLA_IPTUN_6RD_PREFIX]) {
1524 ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1527 if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1529 ip6rd->relay_prefix =
1530 nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1533 if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1535 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1538 if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1540 ip6rd->relay_prefixlen =
1541 nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1548 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1549 struct nlattr *tb[], struct nlattr *data[])
1551 struct net *net = dev_net(dev);
1552 struct ip_tunnel *nt;
1553 struct ip_tunnel_encap ipencap;
1554 #ifdef CONFIG_IPV6_SIT_6RD
1555 struct ip_tunnel_6rd ip6rd;
1559 nt = netdev_priv(dev);
1561 if (ipip6_netlink_encap_parms(data, &ipencap)) {
1562 err = ip_tunnel_encap_setup(nt, &ipencap);
1567 ipip6_netlink_parms(data, &nt->parms);
1569 if (ipip6_tunnel_locate(net, &nt->parms, 0))
1572 err = ipip6_tunnel_create(dev);
1577 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1579 if (mtu >= IPV6_MIN_MTU && mtu <= 0xFFF8 - dev->hard_header_len)
1583 #ifdef CONFIG_IPV6_SIT_6RD
1584 if (ipip6_netlink_6rd_parms(data, &ip6rd)) {
1585 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1587 unregister_netdevice_queue(dev, NULL);
1594 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1595 struct nlattr *data[])
1597 struct ip_tunnel *t = netdev_priv(dev);
1598 struct ip_tunnel_parm p;
1599 struct ip_tunnel_encap ipencap;
1600 struct net *net = t->net;
1601 struct sit_net *sitn = net_generic(net, sit_net_id);
1602 #ifdef CONFIG_IPV6_SIT_6RD
1603 struct ip_tunnel_6rd ip6rd;
1607 if (dev == sitn->fb_tunnel_dev)
1610 if (ipip6_netlink_encap_parms(data, &ipencap)) {
1611 err = ip_tunnel_encap_setup(t, &ipencap);
1616 ipip6_netlink_parms(data, &p);
1618 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1619 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1622 t = ipip6_tunnel_locate(net, &p, 0);
1628 t = netdev_priv(dev);
1630 ipip6_tunnel_update(t, &p);
1632 #ifdef CONFIG_IPV6_SIT_6RD
1633 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1634 return ipip6_tunnel_update_6rd(t, &ip6rd);
1640 static size_t ipip6_get_size(const struct net_device *dev)
1643 /* IFLA_IPTUN_LINK */
1645 /* IFLA_IPTUN_LOCAL */
1647 /* IFLA_IPTUN_REMOTE */
1649 /* IFLA_IPTUN_TTL */
1651 /* IFLA_IPTUN_TOS */
1653 /* IFLA_IPTUN_PMTUDISC */
1655 /* IFLA_IPTUN_FLAGS */
1657 /* IFLA_IPTUN_PROTO */
1659 #ifdef CONFIG_IPV6_SIT_6RD
1660 /* IFLA_IPTUN_6RD_PREFIX */
1661 nla_total_size(sizeof(struct in6_addr)) +
1662 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1664 /* IFLA_IPTUN_6RD_PREFIXLEN */
1666 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1669 /* IFLA_IPTUN_ENCAP_TYPE */
1671 /* IFLA_IPTUN_ENCAP_FLAGS */
1673 /* IFLA_IPTUN_ENCAP_SPORT */
1675 /* IFLA_IPTUN_ENCAP_DPORT */
1680 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1682 struct ip_tunnel *tunnel = netdev_priv(dev);
1683 struct ip_tunnel_parm *parm = &tunnel->parms;
1685 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1686 nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1687 nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1688 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1689 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1690 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1691 !!(parm->iph.frag_off & htons(IP_DF))) ||
1692 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1693 nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1694 goto nla_put_failure;
1696 #ifdef CONFIG_IPV6_SIT_6RD
1697 if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1698 &tunnel->ip6rd.prefix) ||
1699 nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1700 tunnel->ip6rd.relay_prefix) ||
1701 nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1702 tunnel->ip6rd.prefixlen) ||
1703 nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1704 tunnel->ip6rd.relay_prefixlen))
1705 goto nla_put_failure;
1708 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1709 tunnel->encap.type) ||
1710 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1711 tunnel->encap.sport) ||
1712 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1713 tunnel->encap.dport) ||
1714 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1715 tunnel->encap.flags))
1716 goto nla_put_failure;
1724 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1725 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1726 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 },
1727 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 },
1728 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1729 [IFLA_IPTUN_TOS] = { .type = NLA_U8 },
1730 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 },
1731 [IFLA_IPTUN_FLAGS] = { .type = NLA_U16 },
1732 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1733 #ifdef CONFIG_IPV6_SIT_6RD
1734 [IFLA_IPTUN_6RD_PREFIX] = { .len = sizeof(struct in6_addr) },
1735 [IFLA_IPTUN_6RD_RELAY_PREFIX] = { .type = NLA_U32 },
1736 [IFLA_IPTUN_6RD_PREFIXLEN] = { .type = NLA_U16 },
1737 [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1739 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
1740 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
1741 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
1742 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
1745 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1747 struct net *net = dev_net(dev);
1748 struct sit_net *sitn = net_generic(net, sit_net_id);
1750 if (dev != sitn->fb_tunnel_dev)
1751 unregister_netdevice_queue(dev, head);
1754 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1756 .maxtype = IFLA_IPTUN_MAX,
1757 .policy = ipip6_policy,
1758 .priv_size = sizeof(struct ip_tunnel),
1759 .setup = ipip6_tunnel_setup,
1760 .validate = ipip6_validate,
1761 .newlink = ipip6_newlink,
1762 .changelink = ipip6_changelink,
1763 .get_size = ipip6_get_size,
1764 .fill_info = ipip6_fill_info,
1765 .dellink = ipip6_dellink,
1766 .get_link_net = ip_tunnel_get_link_net,
1769 static struct xfrm_tunnel sit_handler __read_mostly = {
1770 .handler = ipip6_rcv,
1771 .err_handler = ipip6_err,
1775 static struct xfrm_tunnel ipip_handler __read_mostly = {
1776 .handler = ipip_rcv,
1777 .err_handler = ipip6_err,
1781 static void __net_exit sit_destroy_tunnels(struct net *net,
1782 struct list_head *head)
1784 struct sit_net *sitn = net_generic(net, sit_net_id);
1785 struct net_device *dev, *aux;
1788 for_each_netdev_safe(net, dev, aux)
1789 if (dev->rtnl_link_ops == &sit_link_ops)
1790 unregister_netdevice_queue(dev, head);
1792 for (prio = 1; prio < 4; prio++) {
1794 for (h = 0; h < HASH_SIZE; h++) {
1795 struct ip_tunnel *t;
1797 t = rtnl_dereference(sitn->tunnels[prio][h]);
1799 /* If dev is in the same netns, it has already
1800 * been added to the list by the previous loop.
1802 if (!net_eq(dev_net(t->dev), net))
1803 unregister_netdevice_queue(t->dev,
1805 t = rtnl_dereference(t->next);
1811 static int __net_init sit_init_net(struct net *net)
1813 struct sit_net *sitn = net_generic(net, sit_net_id);
1814 struct ip_tunnel *t;
1817 sitn->tunnels[0] = sitn->tunnels_wc;
1818 sitn->tunnels[1] = sitn->tunnels_l;
1819 sitn->tunnels[2] = sitn->tunnels_r;
1820 sitn->tunnels[3] = sitn->tunnels_r_l;
1822 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1824 ipip6_tunnel_setup);
1825 if (!sitn->fb_tunnel_dev) {
1829 dev_net_set(sitn->fb_tunnel_dev, net);
1830 sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1831 /* FB netdevice is special: we have one, and only one per netns.
1832 * Allowing to move it to another netns is clearly unsafe.
1834 sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1836 err = register_netdev(sitn->fb_tunnel_dev);
1840 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1841 ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1843 t = netdev_priv(sitn->fb_tunnel_dev);
1845 strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1849 ipip6_dev_free(sitn->fb_tunnel_dev);
1854 static void __net_exit sit_exit_net(struct net *net)
1859 sit_destroy_tunnels(net, &list);
1860 unregister_netdevice_many(&list);
1864 static struct pernet_operations sit_net_ops = {
1865 .init = sit_init_net,
1866 .exit = sit_exit_net,
1868 .size = sizeof(struct sit_net),
1871 static void __exit sit_cleanup(void)
1873 rtnl_link_unregister(&sit_link_ops);
1874 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1875 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1877 unregister_pernet_device(&sit_net_ops);
1878 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1881 static int __init sit_init(void)
1885 pr_info("IPv6 over IPv4 tunneling driver\n");
1887 err = register_pernet_device(&sit_net_ops);
1890 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1892 pr_info("%s: can't register ip6ip4\n", __func__);
1893 goto xfrm_tunnel_failed;
1895 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1897 pr_info("%s: can't register ip4ip4\n", __func__);
1898 goto xfrm_tunnel4_failed;
1900 err = rtnl_link_register(&sit_link_ops);
1902 goto rtnl_link_failed;
1908 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1909 xfrm_tunnel4_failed:
1910 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1912 unregister_pernet_device(&sit_net_ops);
1916 module_init(sit_init);
1917 module_exit(sit_cleanup);
1918 MODULE_LICENSE("GPL");
1919 MODULE_ALIAS_RTNL_LINK("sit");
1920 MODULE_ALIAS_NETDEV("sit0");