1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
4 * Linux INET6 implementation
7 * Pedro Roque <roque@di.fc.ul.pt>
8 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
11 * Roger Venning <r.venning@telstra.com>: 6to4 support
12 * Nate Thompson <nate@thebog.net>: 6to4 support
13 * Fred Templin <fred.l.templin@boeing.com>: isatap support
16 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
18 #include <linux/module.h>
19 #include <linux/capability.h>
20 #include <linux/errno.h>
21 #include <linux/types.h>
22 #include <linux/socket.h>
23 #include <linux/sockios.h>
24 #include <linux/net.h>
25 #include <linux/in6.h>
26 #include <linux/netdevice.h>
27 #include <linux/if_arp.h>
28 #include <linux/icmp.h>
29 #include <linux/slab.h>
30 #include <linux/uaccess.h>
31 #include <linux/init.h>
32 #include <linux/netfilter_ipv4.h>
33 #include <linux/if_ether.h>
39 #include <net/protocol.h>
40 #include <net/transp_v6.h>
41 #include <net/ip6_fib.h>
42 #include <net/ip6_route.h>
43 #include <net/ndisc.h>
44 #include <net/addrconf.h>
48 #include <net/ip_tunnels.h>
49 #include <net/inet_ecn.h>
51 #include <net/dsfield.h>
52 #include <net/net_namespace.h>
53 #include <net/netns/generic.h>
56 This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
58 For comments look at net/ipv4/ip_gre.c --ANK
61 #define IP6_SIT_HASH_SIZE 16
62 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
64 static bool log_ecn_error = true;
65 module_param(log_ecn_error, bool, 0644);
66 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
68 static int ipip6_tunnel_init(struct net_device *dev);
69 static void ipip6_tunnel_setup(struct net_device *dev);
70 static void ipip6_dev_free(struct net_device *dev);
71 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
73 static struct rtnl_link_ops sit_link_ops __read_mostly;
75 static unsigned int sit_net_id __read_mostly;
77 struct ip_tunnel __rcu *tunnels_r_l[IP6_SIT_HASH_SIZE];
78 struct ip_tunnel __rcu *tunnels_r[IP6_SIT_HASH_SIZE];
79 struct ip_tunnel __rcu *tunnels_l[IP6_SIT_HASH_SIZE];
80 struct ip_tunnel __rcu *tunnels_wc[1];
81 struct ip_tunnel __rcu **tunnels[4];
83 struct net_device *fb_tunnel_dev;
87 * Must be invoked with rcu_read_lock
89 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
90 struct net_device *dev,
91 __be32 remote, __be32 local,
94 unsigned int h0 = HASH(remote);
95 unsigned int h1 = HASH(local);
97 struct sit_net *sitn = net_generic(net, sit_net_id);
98 int ifindex = dev ? dev->ifindex : 0;
100 for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
101 if (local == t->parms.iph.saddr &&
102 remote == t->parms.iph.daddr &&
103 (!dev || !t->parms.link || ifindex == t->parms.link ||
104 sifindex == 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 || ifindex == t->parms.link ||
111 sifindex == t->parms.link) &&
112 (t->dev->flags & IFF_UP))
115 for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
116 if (local == t->parms.iph.saddr &&
117 (!dev || !t->parms.link || ifindex == t->parms.link ||
118 sifindex == t->parms.link) &&
119 (t->dev->flags & IFF_UP))
122 t = rcu_dereference(sitn->tunnels_wc[0]);
123 if (t && (t->dev->flags & IFF_UP))
128 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
129 struct ip_tunnel_parm *parms)
131 __be32 remote = parms->iph.daddr;
132 __be32 local = parms->iph.saddr;
144 return &sitn->tunnels[prio][h];
147 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
150 return __ipip6_bucket(sitn, &t->parms);
153 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
155 struct ip_tunnel __rcu **tp;
156 struct ip_tunnel *iter;
158 for (tp = ipip6_bucket(sitn, t);
159 (iter = rtnl_dereference(*tp)) != NULL;
162 rcu_assign_pointer(*tp, t->next);
168 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
170 struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
172 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
173 rcu_assign_pointer(*tp, t);
176 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
178 #ifdef CONFIG_IPV6_SIT_6RD
179 struct ip_tunnel *t = netdev_priv(dev);
181 if (dev == sitn->fb_tunnel_dev || !sitn->fb_tunnel_dev) {
182 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
183 t->ip6rd.relay_prefix = 0;
184 t->ip6rd.prefixlen = 16;
185 t->ip6rd.relay_prefixlen = 0;
187 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
188 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
193 static int ipip6_tunnel_create(struct net_device *dev)
195 struct ip_tunnel *t = netdev_priv(dev);
196 struct net *net = dev_net(dev);
197 struct sit_net *sitn = net_generic(net, sit_net_id);
200 memcpy(dev->dev_addr, &t->parms.iph.saddr, 4);
201 memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
203 if ((__force u16)t->parms.i_flags & SIT_ISATAP)
204 dev->priv_flags |= IFF_ISATAP;
206 dev->rtnl_link_ops = &sit_link_ops;
208 err = register_netdevice(dev);
212 ipip6_tunnel_clone_6rd(dev, sitn);
214 ipip6_tunnel_link(sitn, t);
221 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
222 struct ip_tunnel_parm *parms, int create)
224 __be32 remote = parms->iph.daddr;
225 __be32 local = parms->iph.saddr;
226 struct ip_tunnel *t, *nt;
227 struct ip_tunnel __rcu **tp;
228 struct net_device *dev;
230 struct sit_net *sitn = net_generic(net, sit_net_id);
232 for (tp = __ipip6_bucket(sitn, parms);
233 (t = rtnl_dereference(*tp)) != NULL;
235 if (local == t->parms.iph.saddr &&
236 remote == t->parms.iph.daddr &&
237 parms->link == t->parms.link) {
247 if (parms->name[0]) {
248 if (!dev_valid_name(parms->name))
250 strlcpy(name, parms->name, IFNAMSIZ);
252 strcpy(name, "sit%d");
254 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
259 dev_net_set(dev, net);
261 nt = netdev_priv(dev);
264 if (ipip6_tunnel_create(dev) < 0)
275 #define for_each_prl_rcu(start) \
276 for (prl = rcu_dereference(start); \
278 prl = rcu_dereference(prl->next))
280 static struct ip_tunnel_prl_entry *
281 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
283 struct ip_tunnel_prl_entry *prl;
285 for_each_prl_rcu(t->prl)
286 if (prl->addr == addr)
292 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
293 struct ip_tunnel_prl __user *a)
295 struct ip_tunnel_prl kprl, *kp;
296 struct ip_tunnel_prl_entry *prl;
297 unsigned int cmax, c = 0, ca, len;
300 if (copy_from_user(&kprl, a, sizeof(kprl)))
302 cmax = kprl.datalen / sizeof(kprl);
303 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
306 /* For simple GET or for root users,
307 * we try harder to allocate.
309 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
310 kcalloc(cmax, sizeof(*kp), GFP_KERNEL | __GFP_NOWARN) :
313 ca = min(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;
494 case ICMP_PARAMETERPROB:
497 case ICMP_DEST_UNREACH:
500 /* Impossible event. */
503 /* All others are translated to HOST_UNREACH.
504 rfc2003 contains "deep thoughts" about NET_UNREACH,
505 I believe they are just ether pollution. --ANK
510 case ICMP_TIME_EXCEEDED:
511 if (code != ICMP_EXC_TTL)
513 data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
521 sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
522 t = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
523 iph->daddr, iph->saddr, sifindex);
527 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
528 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
529 t->parms.link, iph->protocol);
533 if (type == ICMP_REDIRECT) {
534 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link,
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;
640 sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
641 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
642 iph->saddr, iph->daddr, sifindex);
644 struct pcpu_sw_netstats *tstats;
646 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
647 tunnel->parms.iph.protocol != 0)
650 skb->mac_header = skb->network_header;
651 skb_reset_network_header(skb);
652 IPCB(skb)->flags = 0;
653 skb->dev = tunnel->dev;
655 if (packet_is_spoofed(skb, iph, tunnel)) {
656 tunnel->dev->stats.rx_errors++;
660 if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
661 !net_eq(tunnel->net, dev_net(tunnel->dev))))
664 /* skb can be uncloned in iptunnel_pull_header, so
665 * old iph is no longer valid
667 iph = (const struct iphdr *)skb_mac_header(skb);
668 err = IP_ECN_decapsulate(iph, skb);
671 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
672 &iph->saddr, iph->tos);
674 ++tunnel->dev->stats.rx_frame_errors;
675 ++tunnel->dev->stats.rx_errors;
680 tstats = this_cpu_ptr(tunnel->dev->tstats);
681 u64_stats_update_begin(&tstats->syncp);
682 tstats->rx_packets++;
683 tstats->rx_bytes += skb->len;
684 u64_stats_update_end(&tstats->syncp);
691 /* no tunnel matched, let upstream know, ipsec may handle it */
698 static const struct tnl_ptk_info ipip_tpi = {
699 /* no tunnel info required for ipip. */
700 .proto = htons(ETH_P_IP),
703 #if IS_ENABLED(CONFIG_MPLS)
704 static const struct tnl_ptk_info mplsip_tpi = {
705 /* no tunnel info required for mplsip. */
706 .proto = htons(ETH_P_MPLS_UC),
710 static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
712 const struct iphdr *iph;
713 struct ip_tunnel *tunnel;
716 sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
719 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
720 iph->saddr, iph->daddr, sifindex);
722 const struct tnl_ptk_info *tpi;
724 if (tunnel->parms.iph.protocol != ipproto &&
725 tunnel->parms.iph.protocol != 0)
728 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
730 #if IS_ENABLED(CONFIG_MPLS)
731 if (ipproto == IPPROTO_MPLS)
736 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
738 return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
748 static int ipip_rcv(struct sk_buff *skb)
750 return sit_tunnel_rcv(skb, IPPROTO_IPIP);
753 #if IS_ENABLED(CONFIG_MPLS)
754 static int mplsip_rcv(struct sk_buff *skb)
756 return sit_tunnel_rcv(skb, IPPROTO_MPLS);
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;
829 u8 protocol = IPPROTO_IPV6;
830 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
833 tos = ipv6_get_dsfield(iph6);
835 /* ISATAP (RFC4214) - must come before 6to4 */
836 if (dev->priv_flags & IFF_ISATAP) {
837 struct neighbour *neigh = NULL;
838 bool do_tx_error = false;
841 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
844 net_dbg_ratelimited("nexthop == NULL\n");
848 addr6 = (const struct in6_addr *)&neigh->primary_key;
849 addr_type = ipv6_addr_type(addr6);
851 if ((addr_type & IPV6_ADDR_UNICAST) &&
852 ipv6_addr_is_isatap(addr6))
853 dst = addr6->s6_addr32[3];
857 neigh_release(neigh);
863 dst = try_6rd(tunnel, &iph6->daddr);
866 struct neighbour *neigh = NULL;
867 bool do_tx_error = false;
870 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
873 net_dbg_ratelimited("nexthop == NULL\n");
877 addr6 = (const struct in6_addr *)&neigh->primary_key;
878 addr_type = ipv6_addr_type(addr6);
880 if (addr_type == IPV6_ADDR_ANY) {
881 addr6 = &ipv6_hdr(skb)->daddr;
882 addr_type = ipv6_addr_type(addr6);
885 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
886 dst = addr6->s6_addr32[3];
890 neigh_release(neigh);
895 flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
896 RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6,
897 0, dst, tiph->saddr, 0, 0,
898 sock_net_uid(tunnel->net, NULL));
900 rt = dst_cache_get_ip4(&tunnel->dst_cache, &fl4.saddr);
902 rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
904 dev->stats.tx_carrier_errors++;
907 dst_cache_set_ip4(&tunnel->dst_cache, &rt->dst, fl4.saddr);
910 if (rt->rt_type != RTN_UNICAST) {
912 dev->stats.tx_carrier_errors++;
919 dev->stats.collisions++;
923 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
929 mtu = dst_mtu(&rt->dst) - t_hlen;
932 dev->stats.collisions++;
937 if (mtu < IPV6_MIN_MTU) {
942 if (tunnel->parms.iph.daddr)
943 skb_dst_update_pmtu_no_confirm(skb, mtu);
945 if (skb->len > mtu && !skb_is_gso(skb)) {
946 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
952 if (tunnel->err_count > 0) {
953 if (time_before(jiffies,
954 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
956 dst_link_failure(skb);
958 tunnel->err_count = 0;
962 * Okay, now see if we can stuff it in the buffer as-is.
964 max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
966 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
967 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
968 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
971 dev->stats.tx_dropped++;
976 skb_set_owner_w(new_skb, skb->sk);
979 iph6 = ipv6_hdr(skb);
983 ttl = iph6->hop_limit;
984 tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
986 if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
991 skb_set_inner_ipproto(skb, IPPROTO_IPV6);
993 iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
994 df, !net_eq(tunnel->net, dev_net(dev)));
998 dst_link_failure(skb);
1001 dev->stats.tx_errors++;
1002 return NETDEV_TX_OK;
1005 static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1006 struct net_device *dev, u8 ipproto)
1008 struct ip_tunnel *tunnel = netdev_priv(dev);
1009 const struct iphdr *tiph = &tunnel->parms.iph;
1011 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1014 skb_set_inner_ipproto(skb, ipproto);
1016 ip_tunnel_xmit(skb, dev, tiph, ipproto);
1017 return NETDEV_TX_OK;
1020 dev->stats.tx_errors++;
1021 return NETDEV_TX_OK;
1024 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1025 struct net_device *dev)
1027 if (!pskb_inet_may_pull(skb))
1030 switch (skb->protocol) {
1031 case htons(ETH_P_IP):
1032 sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1034 case htons(ETH_P_IPV6):
1035 ipip6_tunnel_xmit(skb, dev);
1037 #if IS_ENABLED(CONFIG_MPLS)
1038 case htons(ETH_P_MPLS_UC):
1039 sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1046 return NETDEV_TX_OK;
1049 dev->stats.tx_errors++;
1051 return NETDEV_TX_OK;
1055 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1057 struct net_device *tdev = NULL;
1058 struct ip_tunnel *tunnel;
1059 const struct iphdr *iph;
1062 tunnel = netdev_priv(dev);
1063 iph = &tunnel->parms.iph;
1066 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1068 iph->daddr, iph->saddr,
1072 tunnel->parms.link);
1078 dev->flags |= IFF_POINTOPOINT;
1081 if (!tdev && tunnel->parms.link)
1082 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1084 if (tdev && !netif_is_l3_master(tdev)) {
1085 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1087 dev->mtu = tdev->mtu - t_hlen;
1088 if (dev->mtu < IPV6_MIN_MTU)
1089 dev->mtu = IPV6_MIN_MTU;
1093 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p,
1096 struct net *net = t->net;
1097 struct sit_net *sitn = net_generic(net, sit_net_id);
1099 ipip6_tunnel_unlink(sitn, t);
1101 t->parms.iph.saddr = p->iph.saddr;
1102 t->parms.iph.daddr = p->iph.daddr;
1103 memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1104 memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1105 ipip6_tunnel_link(sitn, t);
1106 t->parms.iph.ttl = p->iph.ttl;
1107 t->parms.iph.tos = p->iph.tos;
1108 t->parms.iph.frag_off = p->iph.frag_off;
1109 if (t->parms.link != p->link || t->fwmark != fwmark) {
1110 t->parms.link = p->link;
1112 ipip6_tunnel_bind_dev(t->dev);
1114 dst_cache_reset(&t->dst_cache);
1115 netdev_state_change(t->dev);
1118 #ifdef CONFIG_IPV6_SIT_6RD
1119 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1120 struct ip_tunnel_6rd *ip6rd)
1122 struct in6_addr prefix;
1123 __be32 relay_prefix;
1125 if (ip6rd->relay_prefixlen > 32 ||
1126 ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1129 ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1130 if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1132 if (ip6rd->relay_prefixlen)
1133 relay_prefix = ip6rd->relay_prefix &
1134 htonl(0xffffffffUL <<
1135 (32 - ip6rd->relay_prefixlen));
1138 if (relay_prefix != ip6rd->relay_prefix)
1141 t->ip6rd.prefix = prefix;
1142 t->ip6rd.relay_prefix = relay_prefix;
1143 t->ip6rd.prefixlen = ip6rd->prefixlen;
1144 t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1145 dst_cache_reset(&t->dst_cache);
1146 netdev_state_change(t->dev);
1151 static bool ipip6_valid_ip_proto(u8 ipproto)
1153 return ipproto == IPPROTO_IPV6 ||
1154 ipproto == IPPROTO_IPIP ||
1155 #if IS_ENABLED(CONFIG_MPLS)
1156 ipproto == IPPROTO_MPLS ||
1162 ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1165 struct ip_tunnel_parm p;
1166 struct ip_tunnel_prl prl;
1167 struct ip_tunnel *t = netdev_priv(dev);
1168 struct net *net = t->net;
1169 struct sit_net *sitn = net_generic(net, sit_net_id);
1170 #ifdef CONFIG_IPV6_SIT_6RD
1171 struct ip_tunnel_6rd ip6rd;
1176 #ifdef CONFIG_IPV6_SIT_6RD
1179 if (dev == sitn->fb_tunnel_dev) {
1180 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1184 t = ipip6_tunnel_locate(net, &p, 0);
1186 t = netdev_priv(dev);
1190 if (cmd == SIOCGETTUNNEL) {
1191 memcpy(&p, &t->parms, sizeof(p));
1192 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1195 #ifdef CONFIG_IPV6_SIT_6RD
1197 ip6rd.prefix = t->ip6rd.prefix;
1198 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1199 ip6rd.prefixlen = t->ip6rd.prefixlen;
1200 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1201 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1212 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1216 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1220 if (!ipip6_valid_ip_proto(p.iph.protocol))
1222 if (p.iph.version != 4 ||
1223 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1226 p.iph.frag_off |= htons(IP_DF);
1228 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1230 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1232 if (t->dev != dev) {
1237 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1238 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1242 t = netdev_priv(dev);
1245 ipip6_tunnel_update(t, &p, t->fwmark);
1250 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1253 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1258 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1261 if (dev == sitn->fb_tunnel_dev) {
1263 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1266 t = ipip6_tunnel_locate(net, &p, 0);
1270 if (t == netdev_priv(sitn->fb_tunnel_dev))
1274 unregister_netdevice(dev);
1280 if (dev == sitn->fb_tunnel_dev)
1282 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1289 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1292 if (dev == sitn->fb_tunnel_dev)
1295 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1300 err = ipip6_tunnel_del_prl(t, &prl);
1304 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1307 dst_cache_reset(&t->dst_cache);
1308 netdev_state_change(dev);
1311 #ifdef CONFIG_IPV6_SIT_6RD
1316 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1320 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1324 if (cmd != SIOCDEL6RD) {
1325 err = ipip6_tunnel_update_6rd(t, &ip6rd);
1329 ipip6_tunnel_clone_6rd(dev, sitn);
1343 static const struct net_device_ops ipip6_netdev_ops = {
1344 .ndo_init = ipip6_tunnel_init,
1345 .ndo_uninit = ipip6_tunnel_uninit,
1346 .ndo_start_xmit = sit_tunnel_xmit,
1347 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1348 .ndo_get_stats64 = ip_tunnel_get_stats64,
1349 .ndo_get_iflink = ip_tunnel_get_iflink,
1352 static void ipip6_dev_free(struct net_device *dev)
1354 struct ip_tunnel *tunnel = netdev_priv(dev);
1356 dst_cache_destroy(&tunnel->dst_cache);
1357 free_percpu(dev->tstats);
1360 #define SIT_FEATURES (NETIF_F_SG | \
1361 NETIF_F_FRAGLIST | \
1363 NETIF_F_GSO_SOFTWARE | \
1366 static void ipip6_tunnel_setup(struct net_device *dev)
1368 struct ip_tunnel *tunnel = netdev_priv(dev);
1369 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1371 dev->netdev_ops = &ipip6_netdev_ops;
1372 dev->needs_free_netdev = true;
1373 dev->priv_destructor = ipip6_dev_free;
1375 dev->type = ARPHRD_SIT;
1376 dev->mtu = ETH_DATA_LEN - t_hlen;
1377 dev->min_mtu = IPV6_MIN_MTU;
1378 dev->max_mtu = IP6_MAX_MTU - t_hlen;
1379 dev->flags = IFF_NOARP;
1380 netif_keep_dst(dev);
1382 dev->features |= NETIF_F_LLTX;
1383 dev->features |= SIT_FEATURES;
1384 dev->hw_features |= SIT_FEATURES;
1387 static int ipip6_tunnel_init(struct net_device *dev)
1389 struct ip_tunnel *tunnel = netdev_priv(dev);
1393 tunnel->net = dev_net(dev);
1394 strcpy(tunnel->parms.name, dev->name);
1396 ipip6_tunnel_bind_dev(dev);
1397 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1401 err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1403 free_percpu(dev->tstats);
1411 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1413 struct ip_tunnel *tunnel = netdev_priv(dev);
1414 struct iphdr *iph = &tunnel->parms.iph;
1415 struct net *net = dev_net(dev);
1416 struct sit_net *sitn = net_generic(net, sit_net_id);
1419 iph->protocol = IPPROTO_IPV6;
1423 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1426 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1427 struct netlink_ext_ack *extack)
1431 if (!data || !data[IFLA_IPTUN_PROTO])
1434 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1435 if (!ipip6_valid_ip_proto(proto))
1441 static void ipip6_netlink_parms(struct nlattr *data[],
1442 struct ip_tunnel_parm *parms,
1445 memset(parms, 0, sizeof(*parms));
1447 parms->iph.version = 4;
1448 parms->iph.protocol = IPPROTO_IPV6;
1450 parms->iph.ttl = 64;
1455 if (data[IFLA_IPTUN_LINK])
1456 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1458 if (data[IFLA_IPTUN_LOCAL])
1459 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1461 if (data[IFLA_IPTUN_REMOTE])
1462 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1464 if (data[IFLA_IPTUN_TTL]) {
1465 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1467 parms->iph.frag_off = htons(IP_DF);
1470 if (data[IFLA_IPTUN_TOS])
1471 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1473 if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1474 parms->iph.frag_off = htons(IP_DF);
1476 if (data[IFLA_IPTUN_FLAGS])
1477 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1479 if (data[IFLA_IPTUN_PROTO])
1480 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1482 if (data[IFLA_IPTUN_FWMARK])
1483 *fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1486 /* This function returns true when ENCAP attributes are present in the nl msg */
1487 static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1488 struct ip_tunnel_encap *ipencap)
1492 memset(ipencap, 0, sizeof(*ipencap));
1497 if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1499 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1502 if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1504 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1507 if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1509 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1512 if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1514 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1520 #ifdef CONFIG_IPV6_SIT_6RD
1521 /* This function returns true when 6RD attributes are present in the nl msg */
1522 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1523 struct ip_tunnel_6rd *ip6rd)
1526 memset(ip6rd, 0, sizeof(*ip6rd));
1531 if (data[IFLA_IPTUN_6RD_PREFIX]) {
1533 ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1536 if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1538 ip6rd->relay_prefix =
1539 nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1542 if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1544 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1547 if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1549 ip6rd->relay_prefixlen =
1550 nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1557 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1558 struct nlattr *tb[], struct nlattr *data[],
1559 struct netlink_ext_ack *extack)
1561 struct net *net = dev_net(dev);
1562 struct ip_tunnel *nt;
1563 struct ip_tunnel_encap ipencap;
1564 #ifdef CONFIG_IPV6_SIT_6RD
1565 struct ip_tunnel_6rd ip6rd;
1569 nt = netdev_priv(dev);
1571 if (ipip6_netlink_encap_parms(data, &ipencap)) {
1572 err = ip_tunnel_encap_setup(nt, &ipencap);
1577 ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1579 if (ipip6_tunnel_locate(net, &nt->parms, 0))
1582 err = ipip6_tunnel_create(dev);
1587 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1589 if (mtu >= IPV6_MIN_MTU &&
1590 mtu <= IP6_MAX_MTU - dev->hard_header_len)
1594 #ifdef CONFIG_IPV6_SIT_6RD
1595 if (ipip6_netlink_6rd_parms(data, &ip6rd)) {
1596 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1598 unregister_netdevice_queue(dev, NULL);
1605 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1606 struct nlattr *data[],
1607 struct netlink_ext_ack *extack)
1609 struct ip_tunnel *t = netdev_priv(dev);
1610 struct ip_tunnel_parm p;
1611 struct ip_tunnel_encap ipencap;
1612 struct net *net = t->net;
1613 struct sit_net *sitn = net_generic(net, sit_net_id);
1614 #ifdef CONFIG_IPV6_SIT_6RD
1615 struct ip_tunnel_6rd ip6rd;
1617 __u32 fwmark = t->fwmark;
1620 if (dev == sitn->fb_tunnel_dev)
1623 if (ipip6_netlink_encap_parms(data, &ipencap)) {
1624 err = ip_tunnel_encap_setup(t, &ipencap);
1629 ipip6_netlink_parms(data, &p, &fwmark);
1631 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1632 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1635 t = ipip6_tunnel_locate(net, &p, 0);
1641 t = netdev_priv(dev);
1643 ipip6_tunnel_update(t, &p, fwmark);
1645 #ifdef CONFIG_IPV6_SIT_6RD
1646 if (ipip6_netlink_6rd_parms(data, &ip6rd))
1647 return ipip6_tunnel_update_6rd(t, &ip6rd);
1653 static size_t ipip6_get_size(const struct net_device *dev)
1656 /* IFLA_IPTUN_LINK */
1658 /* IFLA_IPTUN_LOCAL */
1660 /* IFLA_IPTUN_REMOTE */
1662 /* IFLA_IPTUN_TTL */
1664 /* IFLA_IPTUN_TOS */
1666 /* IFLA_IPTUN_PMTUDISC */
1668 /* IFLA_IPTUN_FLAGS */
1670 /* IFLA_IPTUN_PROTO */
1672 #ifdef CONFIG_IPV6_SIT_6RD
1673 /* IFLA_IPTUN_6RD_PREFIX */
1674 nla_total_size(sizeof(struct in6_addr)) +
1675 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1677 /* IFLA_IPTUN_6RD_PREFIXLEN */
1679 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1682 /* IFLA_IPTUN_ENCAP_TYPE */
1684 /* IFLA_IPTUN_ENCAP_FLAGS */
1686 /* IFLA_IPTUN_ENCAP_SPORT */
1688 /* IFLA_IPTUN_ENCAP_DPORT */
1690 /* IFLA_IPTUN_FWMARK */
1695 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1697 struct ip_tunnel *tunnel = netdev_priv(dev);
1698 struct ip_tunnel_parm *parm = &tunnel->parms;
1700 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1701 nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1702 nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1703 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1704 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1705 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1706 !!(parm->iph.frag_off & htons(IP_DF))) ||
1707 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1708 nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
1709 nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1710 goto nla_put_failure;
1712 #ifdef CONFIG_IPV6_SIT_6RD
1713 if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1714 &tunnel->ip6rd.prefix) ||
1715 nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1716 tunnel->ip6rd.relay_prefix) ||
1717 nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1718 tunnel->ip6rd.prefixlen) ||
1719 nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1720 tunnel->ip6rd.relay_prefixlen))
1721 goto nla_put_failure;
1724 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1725 tunnel->encap.type) ||
1726 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1727 tunnel->encap.sport) ||
1728 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1729 tunnel->encap.dport) ||
1730 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1731 tunnel->encap.flags))
1732 goto nla_put_failure;
1740 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1741 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1742 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 },
1743 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 },
1744 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1745 [IFLA_IPTUN_TOS] = { .type = NLA_U8 },
1746 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 },
1747 [IFLA_IPTUN_FLAGS] = { .type = NLA_U16 },
1748 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1749 #ifdef CONFIG_IPV6_SIT_6RD
1750 [IFLA_IPTUN_6RD_PREFIX] = { .len = sizeof(struct in6_addr) },
1751 [IFLA_IPTUN_6RD_RELAY_PREFIX] = { .type = NLA_U32 },
1752 [IFLA_IPTUN_6RD_PREFIXLEN] = { .type = NLA_U16 },
1753 [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1755 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
1756 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
1757 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
1758 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
1759 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 },
1762 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1764 struct net *net = dev_net(dev);
1765 struct sit_net *sitn = net_generic(net, sit_net_id);
1767 if (dev != sitn->fb_tunnel_dev)
1768 unregister_netdevice_queue(dev, head);
1771 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1773 .maxtype = IFLA_IPTUN_MAX,
1774 .policy = ipip6_policy,
1775 .priv_size = sizeof(struct ip_tunnel),
1776 .setup = ipip6_tunnel_setup,
1777 .validate = ipip6_validate,
1778 .newlink = ipip6_newlink,
1779 .changelink = ipip6_changelink,
1780 .get_size = ipip6_get_size,
1781 .fill_info = ipip6_fill_info,
1782 .dellink = ipip6_dellink,
1783 .get_link_net = ip_tunnel_get_link_net,
1786 static struct xfrm_tunnel sit_handler __read_mostly = {
1787 .handler = ipip6_rcv,
1788 .err_handler = ipip6_err,
1792 static struct xfrm_tunnel ipip_handler __read_mostly = {
1793 .handler = ipip_rcv,
1794 .err_handler = ipip6_err,
1798 #if IS_ENABLED(CONFIG_MPLS)
1799 static struct xfrm_tunnel mplsip_handler __read_mostly = {
1800 .handler = mplsip_rcv,
1801 .err_handler = ipip6_err,
1806 static void __net_exit sit_destroy_tunnels(struct net *net,
1807 struct list_head *head)
1809 struct sit_net *sitn = net_generic(net, sit_net_id);
1810 struct net_device *dev, *aux;
1813 for_each_netdev_safe(net, dev, aux)
1814 if (dev->rtnl_link_ops == &sit_link_ops)
1815 unregister_netdevice_queue(dev, head);
1817 for (prio = 0; prio < 4; prio++) {
1819 for (h = 0; h < (prio ? IP6_SIT_HASH_SIZE : 1); h++) {
1820 struct ip_tunnel *t;
1822 t = rtnl_dereference(sitn->tunnels[prio][h]);
1824 /* If dev is in the same netns, it has already
1825 * been added to the list by the previous loop.
1827 if (!net_eq(dev_net(t->dev), net))
1828 unregister_netdevice_queue(t->dev,
1830 t = rtnl_dereference(t->next);
1836 static int __net_init sit_init_net(struct net *net)
1838 struct sit_net *sitn = net_generic(net, sit_net_id);
1839 struct ip_tunnel *t;
1842 sitn->tunnels[0] = sitn->tunnels_wc;
1843 sitn->tunnels[1] = sitn->tunnels_l;
1844 sitn->tunnels[2] = sitn->tunnels_r;
1845 sitn->tunnels[3] = sitn->tunnels_r_l;
1847 if (!net_has_fallback_tunnels(net))
1850 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1852 ipip6_tunnel_setup);
1853 if (!sitn->fb_tunnel_dev) {
1857 dev_net_set(sitn->fb_tunnel_dev, net);
1858 sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1859 /* FB netdevice is special: we have one, and only one per netns.
1860 * Allowing to move it to another netns is clearly unsafe.
1862 sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1864 err = register_netdev(sitn->fb_tunnel_dev);
1868 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1869 ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1871 t = netdev_priv(sitn->fb_tunnel_dev);
1873 strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1877 free_netdev(sitn->fb_tunnel_dev);
1882 static void __net_exit sit_exit_batch_net(struct list_head *net_list)
1888 list_for_each_entry(net, net_list, exit_list)
1889 sit_destroy_tunnels(net, &list);
1891 unregister_netdevice_many(&list);
1895 static struct pernet_operations sit_net_ops = {
1896 .init = sit_init_net,
1897 .exit_batch = sit_exit_batch_net,
1899 .size = sizeof(struct sit_net),
1902 static void __exit sit_cleanup(void)
1904 rtnl_link_unregister(&sit_link_ops);
1905 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1906 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1907 #if IS_ENABLED(CONFIG_MPLS)
1908 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1911 unregister_pernet_device(&sit_net_ops);
1912 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1915 static int __init sit_init(void)
1919 pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1921 err = register_pernet_device(&sit_net_ops);
1924 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1926 pr_info("%s: can't register ip6ip4\n", __func__);
1927 goto xfrm_tunnel_failed;
1929 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1931 pr_info("%s: can't register ip4ip4\n", __func__);
1932 goto xfrm_tunnel4_failed;
1934 #if IS_ENABLED(CONFIG_MPLS)
1935 err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1937 pr_info("%s: can't register mplsip\n", __func__);
1938 goto xfrm_tunnel_mpls_failed;
1941 err = rtnl_link_register(&sit_link_ops);
1943 goto rtnl_link_failed;
1949 #if IS_ENABLED(CONFIG_MPLS)
1950 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1951 xfrm_tunnel_mpls_failed:
1953 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1954 xfrm_tunnel4_failed:
1955 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1957 unregister_pernet_device(&sit_net_ops);
1961 module_init(sit_init);
1962 module_exit(sit_cleanup);
1963 MODULE_LICENSE("GPL");
1964 MODULE_ALIAS_RTNL_LINK("sit");
1965 MODULE_ALIAS_NETDEV("sit0");