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 <linux/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
65 #define IP6_SIT_HASH_SIZE 16
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 unsigned int sit_net_id __read_mostly;
81 struct ip_tunnel __rcu *tunnels_r_l[IP6_SIT_HASH_SIZE];
82 struct ip_tunnel __rcu *tunnels_r[IP6_SIT_HASH_SIZE];
83 struct ip_tunnel __rcu *tunnels_l[IP6_SIT_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 dev->rtnl_link_ops = &sit_link_ops;
206 err = register_netdevice(dev);
210 ipip6_tunnel_clone_6rd(dev, sitn);
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 | __GFP_NOWARN) :
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 static int ipip6_err(struct sk_buff *skb, u32 info)
484 const struct iphdr *iph = (const struct iphdr *)skb->data;
485 const int type = icmp_hdr(skb)->type;
486 const int code = icmp_hdr(skb)->code;
487 unsigned int data_len = 0;
493 case ICMP_PARAMETERPROB:
496 case ICMP_DEST_UNREACH:
499 /* Impossible event. */
502 /* All others are translated to HOST_UNREACH.
503 rfc2003 contains "deep thoughts" about NET_UNREACH,
504 I believe they are just ether pollution. --ANK
509 case ICMP_TIME_EXCEEDED:
510 if (code != ICMP_EXC_TTL)
512 data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
520 t = ipip6_tunnel_lookup(dev_net(skb->dev),
527 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
528 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
529 t->parms.link, 0, iph->protocol, 0);
533 if (type == ICMP_REDIRECT) {
534 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 0,
541 if (__in6_dev_get(skb->dev) &&
542 !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
545 if (t->parms.iph.daddr == 0)
548 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
551 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
555 t->err_time = jiffies;
560 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
561 const struct in6_addr *v6addr)
564 if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
569 /* Checks if an address matches an address on the tunnel interface.
570 * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
572 * This function is called after we considered the packet as spoofed
574 * We may have a router that is doing NAT for proto 41 packets
575 * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
576 * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
577 * function will return true, dropping the packet.
578 * But, we can still check if is spoofed against the IP
579 * addresses associated with the interface.
581 static bool only_dnatted(const struct ip_tunnel *tunnel,
582 const struct in6_addr *v6dst)
586 #ifdef CONFIG_IPV6_SIT_6RD
587 prefix_len = tunnel->ip6rd.prefixlen + 32
588 - tunnel->ip6rd.relay_prefixlen;
592 return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
595 /* Returns true if a packet is spoofed */
596 static bool packet_is_spoofed(struct sk_buff *skb,
597 const struct iphdr *iph,
598 struct ip_tunnel *tunnel)
600 const struct ipv6hdr *ipv6h;
602 if (tunnel->dev->priv_flags & IFF_ISATAP) {
603 if (!isatap_chksrc(skb, iph, tunnel))
609 if (tunnel->dev->flags & IFF_POINTOPOINT)
612 ipv6h = ipv6_hdr(skb);
614 if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
615 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
616 &iph->saddr, &ipv6h->saddr,
617 &iph->daddr, &ipv6h->daddr);
621 if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
624 if (only_dnatted(tunnel, &ipv6h->daddr))
627 net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
628 &iph->saddr, &ipv6h->saddr,
629 &iph->daddr, &ipv6h->daddr);
633 static int ipip6_rcv(struct sk_buff *skb)
635 const struct iphdr *iph = ip_hdr(skb);
636 struct ip_tunnel *tunnel;
639 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
640 iph->saddr, iph->daddr);
642 struct pcpu_sw_netstats *tstats;
644 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
645 tunnel->parms.iph.protocol != 0)
648 skb->mac_header = skb->network_header;
649 skb_reset_network_header(skb);
650 IPCB(skb)->flags = 0;
651 skb->dev = tunnel->dev;
653 if (packet_is_spoofed(skb, iph, tunnel)) {
654 tunnel->dev->stats.rx_errors++;
658 if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
659 !net_eq(tunnel->net, dev_net(tunnel->dev))))
662 /* skb can be uncloned in iptunnel_pull_header, so
663 * old iph is no longer valid
665 iph = (const struct iphdr *)skb_mac_header(skb);
666 err = IP_ECN_decapsulate(iph, skb);
669 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
670 &iph->saddr, iph->tos);
672 ++tunnel->dev->stats.rx_frame_errors;
673 ++tunnel->dev->stats.rx_errors;
678 tstats = this_cpu_ptr(tunnel->dev->tstats);
679 u64_stats_update_begin(&tstats->syncp);
680 tstats->rx_packets++;
681 tstats->rx_bytes += skb->len;
682 u64_stats_update_end(&tstats->syncp);
689 /* no tunnel matched, let upstream know, ipsec may handle it */
696 static const struct tnl_ptk_info ipip_tpi = {
697 /* no tunnel info required for ipip. */
698 .proto = htons(ETH_P_IP),
701 #if IS_ENABLED(CONFIG_MPLS)
702 static const struct tnl_ptk_info mplsip_tpi = {
703 /* no tunnel info required for mplsip. */
704 .proto = htons(ETH_P_MPLS_UC),
708 static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
710 const struct iphdr *iph;
711 struct ip_tunnel *tunnel;
714 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
715 iph->saddr, iph->daddr);
717 const struct tnl_ptk_info *tpi;
719 if (tunnel->parms.iph.protocol != ipproto &&
720 tunnel->parms.iph.protocol != 0)
723 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
725 #if IS_ENABLED(CONFIG_MPLS)
726 if (ipproto == IPPROTO_MPLS)
731 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
733 return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
743 static int ipip_rcv(struct sk_buff *skb)
745 return sit_tunnel_rcv(skb, IPPROTO_IPIP);
748 #if IS_ENABLED(CONFIG_MPLS)
749 static int mplsip_rcv(struct sk_buff *skb)
751 return sit_tunnel_rcv(skb, IPPROTO_MPLS);
756 * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
757 * stores the embedded IPv4 address in v4dst and returns true.
759 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
762 #ifdef CONFIG_IPV6_SIT_6RD
763 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
764 tunnel->ip6rd.prefixlen)) {
765 unsigned int pbw0, pbi0;
769 pbw0 = tunnel->ip6rd.prefixlen >> 5;
770 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
772 d = tunnel->ip6rd.relay_prefixlen < 32 ?
773 (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
774 tunnel->ip6rd.relay_prefixlen : 0;
776 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
778 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
781 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
785 if (v6dst->s6_addr16[0] == htons(0x2002)) {
786 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
787 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
794 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
795 const struct in6_addr *v6dst)
798 check_6rd(tunnel, v6dst, &dst);
803 * This function assumes it is being called from dev_queue_xmit()
804 * and that skb is filled properly by that function.
807 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
808 struct net_device *dev)
810 struct ip_tunnel *tunnel = netdev_priv(dev);
811 const struct iphdr *tiph = &tunnel->parms.iph;
812 const struct ipv6hdr *iph6 = ipv6_hdr(skb);
813 u8 tos = tunnel->parms.iph.tos;
814 __be16 df = tiph->frag_off;
815 struct rtable *rt; /* Route to the other host */
816 struct net_device *tdev; /* Device to other host */
817 unsigned int max_headroom; /* The extra header space needed */
818 __be32 dst = tiph->daddr;
821 const struct in6_addr *addr6;
824 u8 protocol = IPPROTO_IPV6;
825 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
828 tos = ipv6_get_dsfield(iph6);
830 /* ISATAP (RFC4214) - must come before 6to4 */
831 if (dev->priv_flags & IFF_ISATAP) {
832 struct neighbour *neigh = NULL;
833 bool do_tx_error = false;
836 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
839 net_dbg_ratelimited("nexthop == NULL\n");
843 addr6 = (const struct in6_addr *)&neigh->primary_key;
844 addr_type = ipv6_addr_type(addr6);
846 if ((addr_type & IPV6_ADDR_UNICAST) &&
847 ipv6_addr_is_isatap(addr6))
848 dst = addr6->s6_addr32[3];
852 neigh_release(neigh);
858 dst = try_6rd(tunnel, &iph6->daddr);
861 struct neighbour *neigh = NULL;
862 bool do_tx_error = false;
865 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
868 net_dbg_ratelimited("nexthop == NULL\n");
872 addr6 = (const struct in6_addr *)&neigh->primary_key;
873 addr_type = ipv6_addr_type(addr6);
875 if (addr_type == IPV6_ADDR_ANY) {
876 addr6 = &ipv6_hdr(skb)->daddr;
877 addr_type = ipv6_addr_type(addr6);
880 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
881 dst = addr6->s6_addr32[3];
885 neigh_release(neigh);
890 flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
891 RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6,
892 0, dst, tiph->saddr, 0, 0,
893 sock_net_uid(tunnel->net, NULL));
894 rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
897 dev->stats.tx_carrier_errors++;
900 if (rt->rt_type != RTN_UNICAST) {
902 dev->stats.tx_carrier_errors++;
909 dev->stats.collisions++;
913 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
919 mtu = dst_mtu(&rt->dst) - t_hlen;
922 dev->stats.collisions++;
927 if (mtu < IPV6_MIN_MTU) {
932 if (tunnel->parms.iph.daddr)
933 skb_dst_update_pmtu_no_confirm(skb, mtu);
935 if (skb->len > mtu && !skb_is_gso(skb)) {
936 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
942 if (tunnel->err_count > 0) {
943 if (time_before(jiffies,
944 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
946 dst_link_failure(skb);
948 tunnel->err_count = 0;
952 * Okay, now see if we can stuff it in the buffer as-is.
954 max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
956 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
957 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
958 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
961 dev->stats.tx_dropped++;
966 skb_set_owner_w(new_skb, skb->sk);
969 iph6 = ipv6_hdr(skb);
973 ttl = iph6->hop_limit;
974 tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
976 if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
981 skb_set_inner_ipproto(skb, IPPROTO_IPV6);
983 iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
984 df, !net_eq(tunnel->net, dev_net(dev)));
988 dst_link_failure(skb);
991 dev->stats.tx_errors++;
995 static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
996 struct net_device *dev, u8 ipproto)
998 struct ip_tunnel *tunnel = netdev_priv(dev);
999 const struct iphdr *tiph = &tunnel->parms.iph;
1001 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1004 skb_set_inner_ipproto(skb, ipproto);
1006 ip_tunnel_xmit(skb, dev, tiph, ipproto);
1007 return NETDEV_TX_OK;
1010 dev->stats.tx_errors++;
1011 return NETDEV_TX_OK;
1014 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1015 struct net_device *dev)
1017 switch (skb->protocol) {
1018 case htons(ETH_P_IP):
1019 sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1021 case htons(ETH_P_IPV6):
1022 ipip6_tunnel_xmit(skb, dev);
1024 #if IS_ENABLED(CONFIG_MPLS)
1025 case htons(ETH_P_MPLS_UC):
1026 sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1033 return NETDEV_TX_OK;
1036 dev->stats.tx_errors++;
1038 return NETDEV_TX_OK;
1042 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1044 struct net_device *tdev = NULL;
1045 struct ip_tunnel *tunnel;
1046 const struct iphdr *iph;
1049 tunnel = netdev_priv(dev);
1050 iph = &tunnel->parms.iph;
1053 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1055 iph->daddr, iph->saddr,
1059 tunnel->parms.link);
1065 dev->flags |= IFF_POINTOPOINT;
1068 if (!tdev && tunnel->parms.link)
1069 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1071 if (tdev && !netif_is_l3_master(tdev)) {
1072 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1074 dev->mtu = tdev->mtu - t_hlen;
1075 if (dev->mtu < IPV6_MIN_MTU)
1076 dev->mtu = IPV6_MIN_MTU;
1080 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p,
1083 struct net *net = t->net;
1084 struct sit_net *sitn = net_generic(net, sit_net_id);
1086 ipip6_tunnel_unlink(sitn, t);
1088 t->parms.iph.saddr = p->iph.saddr;
1089 t->parms.iph.daddr = p->iph.daddr;
1090 memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1091 memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1092 ipip6_tunnel_link(sitn, t);
1093 t->parms.iph.ttl = p->iph.ttl;
1094 t->parms.iph.tos = p->iph.tos;
1095 t->parms.iph.frag_off = p->iph.frag_off;
1096 if (t->parms.link != p->link || t->fwmark != fwmark) {
1097 t->parms.link = p->link;
1099 ipip6_tunnel_bind_dev(t->dev);
1101 dst_cache_reset(&t->dst_cache);
1102 netdev_state_change(t->dev);
1105 #ifdef CONFIG_IPV6_SIT_6RD
1106 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1107 struct ip_tunnel_6rd *ip6rd)
1109 struct in6_addr prefix;
1110 __be32 relay_prefix;
1112 if (ip6rd->relay_prefixlen > 32 ||
1113 ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1116 ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1117 if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1119 if (ip6rd->relay_prefixlen)
1120 relay_prefix = ip6rd->relay_prefix &
1121 htonl(0xffffffffUL <<
1122 (32 - ip6rd->relay_prefixlen));
1125 if (relay_prefix != ip6rd->relay_prefix)
1128 t->ip6rd.prefix = prefix;
1129 t->ip6rd.relay_prefix = relay_prefix;
1130 t->ip6rd.prefixlen = ip6rd->prefixlen;
1131 t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1132 dst_cache_reset(&t->dst_cache);
1133 netdev_state_change(t->dev);
1138 static bool ipip6_valid_ip_proto(u8 ipproto)
1140 return ipproto == IPPROTO_IPV6 ||
1141 ipproto == IPPROTO_IPIP ||
1142 #if IS_ENABLED(CONFIG_MPLS)
1143 ipproto == IPPROTO_MPLS ||
1149 ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1152 struct ip_tunnel_parm p;
1153 struct ip_tunnel_prl prl;
1154 struct ip_tunnel *t = netdev_priv(dev);
1155 struct net *net = t->net;
1156 struct sit_net *sitn = net_generic(net, sit_net_id);
1157 #ifdef CONFIG_IPV6_SIT_6RD
1158 struct ip_tunnel_6rd ip6rd;
1163 #ifdef CONFIG_IPV6_SIT_6RD
1166 if (dev == sitn->fb_tunnel_dev) {
1167 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1171 t = ipip6_tunnel_locate(net, &p, 0);
1173 t = netdev_priv(dev);
1177 if (cmd == SIOCGETTUNNEL) {
1178 memcpy(&p, &t->parms, sizeof(p));
1179 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1182 #ifdef CONFIG_IPV6_SIT_6RD
1184 ip6rd.prefix = t->ip6rd.prefix;
1185 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1186 ip6rd.prefixlen = t->ip6rd.prefixlen;
1187 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1188 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1199 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1203 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1207 if (!ipip6_valid_ip_proto(p.iph.protocol))
1209 if (p.iph.version != 4 ||
1210 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1213 p.iph.frag_off |= htons(IP_DF);
1215 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1217 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1219 if (t->dev != dev) {
1224 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1225 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1229 t = netdev_priv(dev);
1232 ipip6_tunnel_update(t, &p, t->fwmark);
1237 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1240 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1245 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1248 if (dev == sitn->fb_tunnel_dev) {
1250 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1253 t = ipip6_tunnel_locate(net, &p, 0);
1257 if (t == netdev_priv(sitn->fb_tunnel_dev))
1261 unregister_netdevice(dev);
1267 if (dev == sitn->fb_tunnel_dev)
1269 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1276 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1279 if (dev == sitn->fb_tunnel_dev)
1282 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1287 err = ipip6_tunnel_del_prl(t, &prl);
1291 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1294 dst_cache_reset(&t->dst_cache);
1295 netdev_state_change(dev);
1298 #ifdef CONFIG_IPV6_SIT_6RD
1303 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1307 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1311 if (cmd != SIOCDEL6RD) {
1312 err = ipip6_tunnel_update_6rd(t, &ip6rd);
1316 ipip6_tunnel_clone_6rd(dev, sitn);
1330 static const struct net_device_ops ipip6_netdev_ops = {
1331 .ndo_init = ipip6_tunnel_init,
1332 .ndo_uninit = ipip6_tunnel_uninit,
1333 .ndo_start_xmit = sit_tunnel_xmit,
1334 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1335 .ndo_get_stats64 = ip_tunnel_get_stats64,
1336 .ndo_get_iflink = ip_tunnel_get_iflink,
1339 static void ipip6_dev_free(struct net_device *dev)
1341 struct ip_tunnel *tunnel = netdev_priv(dev);
1343 dst_cache_destroy(&tunnel->dst_cache);
1344 free_percpu(dev->tstats);
1347 #define SIT_FEATURES (NETIF_F_SG | \
1348 NETIF_F_FRAGLIST | \
1350 NETIF_F_GSO_SOFTWARE | \
1353 static void ipip6_tunnel_setup(struct net_device *dev)
1355 struct ip_tunnel *tunnel = netdev_priv(dev);
1356 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1358 dev->netdev_ops = &ipip6_netdev_ops;
1359 dev->needs_free_netdev = true;
1360 dev->priv_destructor = ipip6_dev_free;
1362 dev->type = ARPHRD_SIT;
1363 dev->mtu = ETH_DATA_LEN - t_hlen;
1364 dev->min_mtu = IPV6_MIN_MTU;
1365 dev->max_mtu = IP6_MAX_MTU - t_hlen;
1366 dev->flags = IFF_NOARP;
1367 netif_keep_dst(dev);
1369 dev->features |= NETIF_F_LLTX;
1370 dev->features |= SIT_FEATURES;
1371 dev->hw_features |= SIT_FEATURES;
1374 static int ipip6_tunnel_init(struct net_device *dev)
1376 struct ip_tunnel *tunnel = netdev_priv(dev);
1380 tunnel->net = dev_net(dev);
1381 strcpy(tunnel->parms.name, dev->name);
1383 ipip6_tunnel_bind_dev(dev);
1384 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1388 err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1390 free_percpu(dev->tstats);
1398 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1400 struct ip_tunnel *tunnel = netdev_priv(dev);
1401 struct iphdr *iph = &tunnel->parms.iph;
1402 struct net *net = dev_net(dev);
1403 struct sit_net *sitn = net_generic(net, sit_net_id);
1406 iph->protocol = IPPROTO_IPV6;
1410 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1413 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1414 struct netlink_ext_ack *extack)
1418 if (!data || !data[IFLA_IPTUN_PROTO])
1421 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1422 if (!ipip6_valid_ip_proto(proto))
1428 static void ipip6_netlink_parms(struct nlattr *data[],
1429 struct ip_tunnel_parm *parms,
1432 memset(parms, 0, sizeof(*parms));
1434 parms->iph.version = 4;
1435 parms->iph.protocol = IPPROTO_IPV6;
1437 parms->iph.ttl = 64;
1442 if (data[IFLA_IPTUN_LINK])
1443 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1445 if (data[IFLA_IPTUN_LOCAL])
1446 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1448 if (data[IFLA_IPTUN_REMOTE])
1449 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1451 if (data[IFLA_IPTUN_TTL]) {
1452 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1454 parms->iph.frag_off = htons(IP_DF);
1457 if (data[IFLA_IPTUN_TOS])
1458 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1460 if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1461 parms->iph.frag_off = htons(IP_DF);
1463 if (data[IFLA_IPTUN_FLAGS])
1464 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1466 if (data[IFLA_IPTUN_PROTO])
1467 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1469 if (data[IFLA_IPTUN_FWMARK])
1470 *fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1473 /* This function returns true when ENCAP attributes are present in the nl msg */
1474 static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1475 struct ip_tunnel_encap *ipencap)
1479 memset(ipencap, 0, sizeof(*ipencap));
1484 if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1486 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1489 if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1491 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1494 if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1496 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1499 if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1501 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1507 #ifdef CONFIG_IPV6_SIT_6RD
1508 /* This function returns true when 6RD attributes are present in the nl msg */
1509 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1510 struct ip_tunnel_6rd *ip6rd)
1513 memset(ip6rd, 0, sizeof(*ip6rd));
1518 if (data[IFLA_IPTUN_6RD_PREFIX]) {
1520 ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1523 if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1525 ip6rd->relay_prefix =
1526 nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1529 if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1531 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1534 if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1536 ip6rd->relay_prefixlen =
1537 nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1544 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1545 struct nlattr *tb[], struct nlattr *data[],
1546 struct netlink_ext_ack *extack)
1548 struct net *net = dev_net(dev);
1549 struct ip_tunnel *nt;
1550 struct ip_tunnel_encap ipencap;
1551 #ifdef CONFIG_IPV6_SIT_6RD
1552 struct ip_tunnel_6rd ip6rd;
1556 nt = netdev_priv(dev);
1558 if (ipip6_netlink_encap_parms(data, &ipencap)) {
1559 err = ip_tunnel_encap_setup(nt, &ipencap);
1564 ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1566 if (ipip6_tunnel_locate(net, &nt->parms, 0))
1569 err = ipip6_tunnel_create(dev);
1574 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1576 if (mtu >= IPV6_MIN_MTU &&
1577 mtu <= IP6_MAX_MTU - dev->hard_header_len)
1581 #ifdef CONFIG_IPV6_SIT_6RD
1582 if (ipip6_netlink_6rd_parms(data, &ip6rd)) {
1583 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1585 unregister_netdevice_queue(dev, NULL);
1592 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1593 struct nlattr *data[],
1594 struct netlink_ext_ack *extack)
1596 struct ip_tunnel *t = netdev_priv(dev);
1597 struct ip_tunnel_parm p;
1598 struct ip_tunnel_encap ipencap;
1599 struct net *net = t->net;
1600 struct sit_net *sitn = net_generic(net, sit_net_id);
1601 #ifdef CONFIG_IPV6_SIT_6RD
1602 struct ip_tunnel_6rd ip6rd;
1604 __u32 fwmark = t->fwmark;
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, &fwmark);
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, fwmark);
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 */
1677 /* IFLA_IPTUN_FWMARK */
1682 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1684 struct ip_tunnel *tunnel = netdev_priv(dev);
1685 struct ip_tunnel_parm *parm = &tunnel->parms;
1687 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1688 nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1689 nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1690 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1691 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1692 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1693 !!(parm->iph.frag_off & htons(IP_DF))) ||
1694 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1695 nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
1696 nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1697 goto nla_put_failure;
1699 #ifdef CONFIG_IPV6_SIT_6RD
1700 if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1701 &tunnel->ip6rd.prefix) ||
1702 nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1703 tunnel->ip6rd.relay_prefix) ||
1704 nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1705 tunnel->ip6rd.prefixlen) ||
1706 nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1707 tunnel->ip6rd.relay_prefixlen))
1708 goto nla_put_failure;
1711 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1712 tunnel->encap.type) ||
1713 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1714 tunnel->encap.sport) ||
1715 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1716 tunnel->encap.dport) ||
1717 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1718 tunnel->encap.flags))
1719 goto nla_put_failure;
1727 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1728 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1729 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 },
1730 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 },
1731 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1732 [IFLA_IPTUN_TOS] = { .type = NLA_U8 },
1733 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 },
1734 [IFLA_IPTUN_FLAGS] = { .type = NLA_U16 },
1735 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1736 #ifdef CONFIG_IPV6_SIT_6RD
1737 [IFLA_IPTUN_6RD_PREFIX] = { .len = sizeof(struct in6_addr) },
1738 [IFLA_IPTUN_6RD_RELAY_PREFIX] = { .type = NLA_U32 },
1739 [IFLA_IPTUN_6RD_PREFIXLEN] = { .type = NLA_U16 },
1740 [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1742 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
1743 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
1744 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
1745 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
1746 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 },
1749 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1751 struct net *net = dev_net(dev);
1752 struct sit_net *sitn = net_generic(net, sit_net_id);
1754 if (dev != sitn->fb_tunnel_dev)
1755 unregister_netdevice_queue(dev, head);
1758 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1760 .maxtype = IFLA_IPTUN_MAX,
1761 .policy = ipip6_policy,
1762 .priv_size = sizeof(struct ip_tunnel),
1763 .setup = ipip6_tunnel_setup,
1764 .validate = ipip6_validate,
1765 .newlink = ipip6_newlink,
1766 .changelink = ipip6_changelink,
1767 .get_size = ipip6_get_size,
1768 .fill_info = ipip6_fill_info,
1769 .dellink = ipip6_dellink,
1770 .get_link_net = ip_tunnel_get_link_net,
1773 static struct xfrm_tunnel sit_handler __read_mostly = {
1774 .handler = ipip6_rcv,
1775 .err_handler = ipip6_err,
1779 static struct xfrm_tunnel ipip_handler __read_mostly = {
1780 .handler = ipip_rcv,
1781 .err_handler = ipip6_err,
1785 #if IS_ENABLED(CONFIG_MPLS)
1786 static struct xfrm_tunnel mplsip_handler __read_mostly = {
1787 .handler = mplsip_rcv,
1788 .err_handler = ipip6_err,
1793 static void __net_exit sit_destroy_tunnels(struct net *net,
1794 struct list_head *head)
1796 struct sit_net *sitn = net_generic(net, sit_net_id);
1797 struct net_device *dev, *aux;
1800 for_each_netdev_safe(net, dev, aux)
1801 if (dev->rtnl_link_ops == &sit_link_ops)
1802 unregister_netdevice_queue(dev, head);
1804 for (prio = 0; prio < 4; prio++) {
1806 for (h = 0; h < (prio ? IP6_SIT_HASH_SIZE : 1); h++) {
1807 struct ip_tunnel *t;
1809 t = rtnl_dereference(sitn->tunnels[prio][h]);
1811 /* If dev is in the same netns, it has already
1812 * been added to the list by the previous loop.
1814 if (!net_eq(dev_net(t->dev), net))
1815 unregister_netdevice_queue(t->dev,
1817 t = rtnl_dereference(t->next);
1823 static int __net_init sit_init_net(struct net *net)
1825 struct sit_net *sitn = net_generic(net, sit_net_id);
1826 struct ip_tunnel *t;
1829 sitn->tunnels[0] = sitn->tunnels_wc;
1830 sitn->tunnels[1] = sitn->tunnels_l;
1831 sitn->tunnels[2] = sitn->tunnels_r;
1832 sitn->tunnels[3] = sitn->tunnels_r_l;
1834 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1836 ipip6_tunnel_setup);
1837 if (!sitn->fb_tunnel_dev) {
1841 dev_net_set(sitn->fb_tunnel_dev, net);
1842 sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1843 /* FB netdevice is special: we have one, and only one per netns.
1844 * Allowing to move it to another netns is clearly unsafe.
1846 sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1848 err = register_netdev(sitn->fb_tunnel_dev);
1852 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1853 ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1855 t = netdev_priv(sitn->fb_tunnel_dev);
1857 strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1861 free_netdev(sitn->fb_tunnel_dev);
1866 static void __net_exit sit_exit_net(struct net *net)
1871 sit_destroy_tunnels(net, &list);
1872 unregister_netdevice_many(&list);
1876 static struct pernet_operations sit_net_ops = {
1877 .init = sit_init_net,
1878 .exit = sit_exit_net,
1880 .size = sizeof(struct sit_net),
1883 static void __exit sit_cleanup(void)
1885 rtnl_link_unregister(&sit_link_ops);
1886 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1887 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1888 #if IS_ENABLED(CONFIG_MPLS)
1889 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1892 unregister_pernet_device(&sit_net_ops);
1893 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1896 static int __init sit_init(void)
1900 pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1902 err = register_pernet_device(&sit_net_ops);
1905 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1907 pr_info("%s: can't register ip6ip4\n", __func__);
1908 goto xfrm_tunnel_failed;
1910 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1912 pr_info("%s: can't register ip4ip4\n", __func__);
1913 goto xfrm_tunnel4_failed;
1915 #if IS_ENABLED(CONFIG_MPLS)
1916 err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1918 pr_info("%s: can't register mplsip\n", __func__);
1919 goto xfrm_tunnel_mpls_failed;
1922 err = rtnl_link_register(&sit_link_ops);
1924 goto rtnl_link_failed;
1930 #if IS_ENABLED(CONFIG_MPLS)
1931 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1932 xfrm_tunnel_mpls_failed:
1934 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1935 xfrm_tunnel4_failed:
1936 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1938 unregister_pernet_device(&sit_net_ops);
1942 module_init(sit_init);
1943 module_exit(sit_cleanup);
1944 MODULE_LICENSE("GPL");
1945 MODULE_ALIAS_RTNL_LINK("sit");
1946 MODULE_ALIAS_NETDEV("sit0");