2 * Copyright (c) 2013 Nicira, Inc.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public License
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
19 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
21 #include <linux/capability.h>
22 #include <linux/module.h>
23 #include <linux/types.h>
24 #include <linux/kernel.h>
25 #include <linux/slab.h>
26 #include <linux/uaccess.h>
27 #include <linux/skbuff.h>
28 #include <linux/netdevice.h>
30 #include <linux/tcp.h>
31 #include <linux/udp.h>
32 #include <linux/if_arp.h>
33 #include <linux/mroute.h>
34 #include <linux/init.h>
35 #include <linux/in6.h>
36 #include <linux/inetdevice.h>
37 #include <linux/igmp.h>
38 #include <linux/netfilter_ipv4.h>
39 #include <linux/etherdevice.h>
40 #include <linux/if_ether.h>
41 #include <linux/if_vlan.h>
42 #include <linux/rculist.h>
43 #include <linux/err.h>
48 #include <net/protocol.h>
49 #include <net/ip_tunnels.h>
51 #include <net/checksum.h>
52 #include <net/dsfield.h>
53 #include <net/inet_ecn.h>
55 #include <net/net_namespace.h>
56 #include <net/netns/generic.h>
57 #include <net/rtnetlink.h>
59 #if IS_ENABLED(CONFIG_IPV6)
61 #include <net/ip6_fib.h>
62 #include <net/ip6_route.h>
65 static unsigned int ip_tunnel_hash(__be32 key, __be32 remote)
67 return hash_32((__force u32)key ^ (__force u32)remote,
71 static void __tunnel_dst_set(struct ip_tunnel_dst *idst,
72 struct dst_entry *dst)
74 struct dst_entry *old_dst;
77 if (dst->flags & DST_NOCACHE)
82 old_dst = xchg((__force struct dst_entry **)&idst->dst, dst);
86 static void tunnel_dst_set(struct ip_tunnel *t, struct dst_entry *dst)
88 __tunnel_dst_set(this_cpu_ptr(t->dst_cache), dst);
91 static void tunnel_dst_reset(struct ip_tunnel *t)
93 tunnel_dst_set(t, NULL);
96 void ip_tunnel_dst_reset_all(struct ip_tunnel *t)
100 for_each_possible_cpu(i)
101 __tunnel_dst_set(per_cpu_ptr(t->dst_cache, i), NULL);
103 EXPORT_SYMBOL(ip_tunnel_dst_reset_all);
105 static struct rtable *tunnel_rtable_get(struct ip_tunnel *t, u32 cookie)
107 struct dst_entry *dst;
110 dst = rcu_dereference(this_cpu_ptr(t->dst_cache)->dst);
112 if (dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
120 return (struct rtable *)dst;
123 static bool ip_tunnel_key_match(const struct ip_tunnel_parm *p,
124 __be16 flags, __be32 key)
126 if (p->i_flags & TUNNEL_KEY) {
127 if (flags & TUNNEL_KEY)
128 return key == p->i_key;
130 /* key expected, none present */
133 return !(flags & TUNNEL_KEY);
136 /* Fallback tunnel: no source, no destination, no key, no options
139 We require exact key match i.e. if a key is present in packet
140 it will match only tunnel with the same key; if it is not present,
141 it will match only keyless tunnel.
143 All keysless packets, if not matched configured keyless tunnels
144 will match fallback tunnel.
145 Given src, dst and key, find appropriate for input tunnel.
147 struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
148 int link, __be16 flags,
149 __be32 remote, __be32 local,
153 struct ip_tunnel *t, *cand = NULL;
154 struct hlist_head *head;
156 hash = ip_tunnel_hash(key, remote);
157 head = &itn->tunnels[hash];
159 hlist_for_each_entry_rcu(t, head, hash_node) {
160 if (local != t->parms.iph.saddr ||
161 remote != t->parms.iph.daddr ||
162 !(t->dev->flags & IFF_UP))
165 if (!ip_tunnel_key_match(&t->parms, flags, key))
168 if (t->parms.link == link)
174 hlist_for_each_entry_rcu(t, head, hash_node) {
175 if (remote != t->parms.iph.daddr ||
176 !(t->dev->flags & IFF_UP))
179 if (!ip_tunnel_key_match(&t->parms, flags, key))
182 if (t->parms.link == link)
188 hash = ip_tunnel_hash(key, 0);
189 head = &itn->tunnels[hash];
191 hlist_for_each_entry_rcu(t, head, hash_node) {
192 if ((local != t->parms.iph.saddr &&
193 (local != t->parms.iph.daddr ||
194 !ipv4_is_multicast(local))) ||
195 !(t->dev->flags & IFF_UP))
198 if (!ip_tunnel_key_match(&t->parms, flags, key))
201 if (t->parms.link == link)
207 if (flags & TUNNEL_NO_KEY)
208 goto skip_key_lookup;
210 hlist_for_each_entry_rcu(t, head, hash_node) {
211 if (t->parms.i_key != key ||
212 !(t->dev->flags & IFF_UP))
215 if (t->parms.link == link)
225 if (itn->fb_tunnel_dev && itn->fb_tunnel_dev->flags & IFF_UP)
226 return netdev_priv(itn->fb_tunnel_dev);
231 EXPORT_SYMBOL_GPL(ip_tunnel_lookup);
233 static struct hlist_head *ip_bucket(struct ip_tunnel_net *itn,
234 struct ip_tunnel_parm *parms)
238 __be32 i_key = parms->i_key;
240 if (parms->iph.daddr && !ipv4_is_multicast(parms->iph.daddr))
241 remote = parms->iph.daddr;
245 if (!(parms->i_flags & TUNNEL_KEY) && (parms->i_flags & VTI_ISVTI))
248 h = ip_tunnel_hash(i_key, remote);
249 return &itn->tunnels[h];
252 static void ip_tunnel_add(struct ip_tunnel_net *itn, struct ip_tunnel *t)
254 struct hlist_head *head = ip_bucket(itn, &t->parms);
256 hlist_add_head_rcu(&t->hash_node, head);
259 static void ip_tunnel_del(struct ip_tunnel *t)
261 hlist_del_init_rcu(&t->hash_node);
264 static struct ip_tunnel *ip_tunnel_find(struct ip_tunnel_net *itn,
265 struct ip_tunnel_parm *parms,
268 __be32 remote = parms->iph.daddr;
269 __be32 local = parms->iph.saddr;
270 __be32 key = parms->i_key;
271 int link = parms->link;
272 struct ip_tunnel *t = NULL;
273 struct hlist_head *head = ip_bucket(itn, parms);
275 hlist_for_each_entry_rcu(t, head, hash_node) {
276 if (local == t->parms.iph.saddr &&
277 remote == t->parms.iph.daddr &&
278 key == t->parms.i_key &&
279 link == t->parms.link &&
280 type == t->dev->type)
286 static struct net_device *__ip_tunnel_create(struct net *net,
287 const struct rtnl_link_ops *ops,
288 struct ip_tunnel_parm *parms)
291 struct ip_tunnel *tunnel;
292 struct net_device *dev;
296 strlcpy(name, parms->name, IFNAMSIZ);
298 if (strlen(ops->kind) > (IFNAMSIZ - 3)) {
302 strlcpy(name, ops->kind, IFNAMSIZ);
303 strncat(name, "%d", 2);
307 dev = alloc_netdev(ops->priv_size, name, ops->setup);
312 dev_net_set(dev, net);
314 dev->rtnl_link_ops = ops;
316 tunnel = netdev_priv(dev);
317 tunnel->parms = *parms;
320 err = register_netdevice(dev);
332 static inline void init_tunnel_flow(struct flowi4 *fl4,
334 __be32 daddr, __be32 saddr,
335 __be32 key, __u8 tos, int oif)
337 memset(fl4, 0, sizeof(*fl4));
338 fl4->flowi4_oif = oif;
341 fl4->flowi4_tos = tos;
342 fl4->flowi4_proto = proto;
343 fl4->fl4_gre_key = key;
346 static int ip_tunnel_bind_dev(struct net_device *dev)
348 struct net_device *tdev = NULL;
349 struct ip_tunnel *tunnel = netdev_priv(dev);
350 const struct iphdr *iph;
351 int hlen = LL_MAX_HEADER;
352 int mtu = ETH_DATA_LEN;
353 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
355 iph = &tunnel->parms.iph;
357 /* Guess output device to choose reasonable mtu and needed_headroom */
362 init_tunnel_flow(&fl4, iph->protocol, iph->daddr,
363 iph->saddr, tunnel->parms.o_key,
364 RT_TOS(iph->tos), tunnel->parms.link);
365 rt = ip_route_output_key(tunnel->net, &fl4);
369 tunnel_dst_set(tunnel, &rt->dst);
372 if (dev->type != ARPHRD_ETHER)
373 dev->flags |= IFF_POINTOPOINT;
376 if (!tdev && tunnel->parms.link)
377 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
380 hlen = tdev->hard_header_len + tdev->needed_headroom;
383 dev->iflink = tunnel->parms.link;
385 dev->needed_headroom = t_hlen + hlen;
386 mtu -= (dev->hard_header_len + t_hlen);
394 static struct ip_tunnel *ip_tunnel_create(struct net *net,
395 struct ip_tunnel_net *itn,
396 struct ip_tunnel_parm *parms)
398 struct ip_tunnel *nt, *fbt;
399 struct net_device *dev;
401 BUG_ON(!itn->fb_tunnel_dev);
402 fbt = netdev_priv(itn->fb_tunnel_dev);
403 dev = __ip_tunnel_create(net, itn->fb_tunnel_dev->rtnl_link_ops, parms);
405 return ERR_CAST(dev);
407 dev->mtu = ip_tunnel_bind_dev(dev);
409 nt = netdev_priv(dev);
410 ip_tunnel_add(itn, nt);
414 int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
415 const struct tnl_ptk_info *tpi, bool log_ecn_error)
417 struct pcpu_sw_netstats *tstats;
418 const struct iphdr *iph = ip_hdr(skb);
421 #ifdef CONFIG_NET_IPGRE_BROADCAST
422 if (ipv4_is_multicast(iph->daddr)) {
423 tunnel->dev->stats.multicast++;
424 skb->pkt_type = PACKET_BROADCAST;
428 if ((!(tpi->flags&TUNNEL_CSUM) && (tunnel->parms.i_flags&TUNNEL_CSUM)) ||
429 ((tpi->flags&TUNNEL_CSUM) && !(tunnel->parms.i_flags&TUNNEL_CSUM))) {
430 tunnel->dev->stats.rx_crc_errors++;
431 tunnel->dev->stats.rx_errors++;
435 if (tunnel->parms.i_flags&TUNNEL_SEQ) {
436 if (!(tpi->flags&TUNNEL_SEQ) ||
437 (tunnel->i_seqno && (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
438 tunnel->dev->stats.rx_fifo_errors++;
439 tunnel->dev->stats.rx_errors++;
442 tunnel->i_seqno = ntohl(tpi->seq) + 1;
445 err = IP_ECN_decapsulate(iph, skb);
448 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
449 &iph->saddr, iph->tos);
451 ++tunnel->dev->stats.rx_frame_errors;
452 ++tunnel->dev->stats.rx_errors;
457 tstats = this_cpu_ptr(tunnel->dev->tstats);
458 u64_stats_update_begin(&tstats->syncp);
459 tstats->rx_packets++;
460 tstats->rx_bytes += skb->len;
461 u64_stats_update_end(&tstats->syncp);
463 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
465 if (tunnel->dev->type == ARPHRD_ETHER) {
466 skb->protocol = eth_type_trans(skb, tunnel->dev);
467 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
469 skb->dev = tunnel->dev;
472 gro_cells_receive(&tunnel->gro_cells, skb);
479 EXPORT_SYMBOL_GPL(ip_tunnel_rcv);
481 static int tnl_update_pmtu(struct net_device *dev, struct sk_buff *skb,
482 struct rtable *rt, __be16 df)
484 struct ip_tunnel *tunnel = netdev_priv(dev);
485 int pkt_size = skb->len - tunnel->hlen - dev->hard_header_len;
489 mtu = dst_mtu(&rt->dst) - dev->hard_header_len
490 - sizeof(struct iphdr) - tunnel->hlen;
492 mtu = skb_dst(skb) ? dst_mtu(skb_dst(skb)) : dev->mtu;
495 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
497 if (skb->protocol == htons(ETH_P_IP)) {
498 if (!skb_is_gso(skb) &&
499 (df & htons(IP_DF)) && mtu < pkt_size) {
500 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
501 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, htonl(mtu));
505 #if IS_ENABLED(CONFIG_IPV6)
506 else if (skb->protocol == htons(ETH_P_IPV6)) {
507 struct rt6_info *rt6 = (struct rt6_info *)skb_dst(skb);
509 if (rt6 && mtu < dst_mtu(skb_dst(skb)) &&
510 mtu >= IPV6_MIN_MTU) {
511 if ((tunnel->parms.iph.daddr &&
512 !ipv4_is_multicast(tunnel->parms.iph.daddr)) ||
513 rt6->rt6i_dst.plen == 128) {
514 rt6->rt6i_flags |= RTF_MODIFIED;
515 dst_metric_set(skb_dst(skb), RTAX_MTU, mtu);
519 if (!skb_is_gso(skb) && mtu >= IPV6_MIN_MTU &&
521 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
529 void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
530 const struct iphdr *tnl_params, const u8 protocol)
532 struct ip_tunnel *tunnel = netdev_priv(dev);
533 const struct iphdr *inner_iph;
537 struct rtable *rt; /* Route to the other host */
538 unsigned int max_headroom; /* The extra header space needed */
541 bool connected = true;
543 inner_iph = (const struct iphdr *)skb_inner_network_header(skb);
545 dst = tnl_params->daddr;
549 if (skb_dst(skb) == NULL) {
550 dev->stats.tx_fifo_errors++;
554 if (skb->protocol == htons(ETH_P_IP)) {
555 rt = skb_rtable(skb);
556 dst = rt_nexthop(rt, inner_iph->daddr);
558 #if IS_ENABLED(CONFIG_IPV6)
559 else if (skb->protocol == htons(ETH_P_IPV6)) {
560 const struct in6_addr *addr6;
561 struct neighbour *neigh;
562 bool do_tx_error_icmp;
565 neigh = dst_neigh_lookup(skb_dst(skb),
566 &ipv6_hdr(skb)->daddr);
570 addr6 = (const struct in6_addr *)&neigh->primary_key;
571 addr_type = ipv6_addr_type(addr6);
573 if (addr_type == IPV6_ADDR_ANY) {
574 addr6 = &ipv6_hdr(skb)->daddr;
575 addr_type = ipv6_addr_type(addr6);
578 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
579 do_tx_error_icmp = true;
581 do_tx_error_icmp = false;
582 dst = addr6->s6_addr32[3];
584 neigh_release(neigh);
585 if (do_tx_error_icmp)
595 tos = tnl_params->tos;
598 if (skb->protocol == htons(ETH_P_IP)) {
599 tos = inner_iph->tos;
601 } else if (skb->protocol == htons(ETH_P_IPV6)) {
602 tos = ipv6_get_dsfield((const struct ipv6hdr *)inner_iph);
607 init_tunnel_flow(&fl4, protocol, dst, tnl_params->saddr,
608 tunnel->parms.o_key, RT_TOS(tos), tunnel->parms.link);
610 rt = connected ? tunnel_rtable_get(tunnel, 0) : NULL;
613 rt = ip_route_output_key(tunnel->net, &fl4);
616 dev->stats.tx_carrier_errors++;
620 tunnel_dst_set(tunnel, &rt->dst);
623 if (rt->dst.dev == dev) {
625 dev->stats.collisions++;
629 if (tnl_update_pmtu(dev, skb, rt, tnl_params->frag_off)) {
634 if (tunnel->err_count > 0) {
635 if (time_before(jiffies,
636 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
639 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
640 dst_link_failure(skb);
642 tunnel->err_count = 0;
645 tos = ip_tunnel_ecn_encap(tos, inner_iph, skb);
646 ttl = tnl_params->ttl;
648 if (skb->protocol == htons(ETH_P_IP))
649 ttl = inner_iph->ttl;
650 #if IS_ENABLED(CONFIG_IPV6)
651 else if (skb->protocol == htons(ETH_P_IPV6))
652 ttl = ((const struct ipv6hdr *)inner_iph)->hop_limit;
655 ttl = ip4_dst_hoplimit(&rt->dst);
658 df = tnl_params->frag_off;
659 if (skb->protocol == htons(ETH_P_IP))
660 df |= (inner_iph->frag_off&htons(IP_DF));
662 max_headroom = LL_RESERVED_SPACE(rt->dst.dev) + sizeof(struct iphdr)
663 + rt->dst.header_len;
664 if (max_headroom > dev->needed_headroom)
665 dev->needed_headroom = max_headroom;
667 if (skb_cow_head(skb, dev->needed_headroom)) {
668 dev->stats.tx_dropped++;
673 err = iptunnel_xmit(rt, skb, fl4.saddr, fl4.daddr, protocol,
674 tos, ttl, df, !net_eq(tunnel->net, dev_net(dev)));
675 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
679 #if IS_ENABLED(CONFIG_IPV6)
681 dst_link_failure(skb);
684 dev->stats.tx_errors++;
687 EXPORT_SYMBOL_GPL(ip_tunnel_xmit);
689 static void ip_tunnel_update(struct ip_tunnel_net *itn,
691 struct net_device *dev,
692 struct ip_tunnel_parm *p,
696 t->parms.iph.saddr = p->iph.saddr;
697 t->parms.iph.daddr = p->iph.daddr;
698 t->parms.i_key = p->i_key;
699 t->parms.o_key = p->o_key;
700 if (dev->type != ARPHRD_ETHER) {
701 memcpy(dev->dev_addr, &p->iph.saddr, 4);
702 memcpy(dev->broadcast, &p->iph.daddr, 4);
704 ip_tunnel_add(itn, t);
706 t->parms.iph.ttl = p->iph.ttl;
707 t->parms.iph.tos = p->iph.tos;
708 t->parms.iph.frag_off = p->iph.frag_off;
710 if (t->parms.link != p->link) {
713 t->parms.link = p->link;
714 mtu = ip_tunnel_bind_dev(dev);
718 ip_tunnel_dst_reset_all(t);
719 netdev_state_change(dev);
722 int ip_tunnel_ioctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd)
726 struct net *net = dev_net(dev);
727 struct ip_tunnel *tunnel = netdev_priv(dev);
728 struct ip_tunnel_net *itn = net_generic(net, tunnel->ip_tnl_net_id);
730 BUG_ON(!itn->fb_tunnel_dev);
734 if (dev == itn->fb_tunnel_dev)
735 t = ip_tunnel_find(itn, p, itn->fb_tunnel_dev->type);
737 t = netdev_priv(dev);
738 memcpy(p, &t->parms, sizeof(*p));
744 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
747 p->iph.frag_off |= htons(IP_DF);
748 if (!(p->i_flags&TUNNEL_KEY))
750 if (!(p->o_flags&TUNNEL_KEY))
753 t = ip_tunnel_find(itn, p, itn->fb_tunnel_dev->type);
755 if (!t && (cmd == SIOCADDTUNNEL)) {
756 t = ip_tunnel_create(net, itn, p);
762 if (dev != itn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
769 unsigned int nflags = 0;
771 if (ipv4_is_multicast(p->iph.daddr))
772 nflags = IFF_BROADCAST;
773 else if (p->iph.daddr)
774 nflags = IFF_POINTOPOINT;
776 if ((dev->flags^nflags)&(IFF_POINTOPOINT|IFF_BROADCAST)) {
781 t = netdev_priv(dev);
787 ip_tunnel_update(itn, t, dev, p, true);
795 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
798 if (dev == itn->fb_tunnel_dev) {
800 t = ip_tunnel_find(itn, p, itn->fb_tunnel_dev->type);
804 if (t == netdev_priv(itn->fb_tunnel_dev))
808 unregister_netdevice(dev);
819 EXPORT_SYMBOL_GPL(ip_tunnel_ioctl);
821 int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu)
823 struct ip_tunnel *tunnel = netdev_priv(dev);
824 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
827 new_mtu > 0xFFF8 - dev->hard_header_len - t_hlen)
832 EXPORT_SYMBOL_GPL(ip_tunnel_change_mtu);
834 static void ip_tunnel_dev_free(struct net_device *dev)
836 struct ip_tunnel *tunnel = netdev_priv(dev);
838 gro_cells_destroy(&tunnel->gro_cells);
839 free_percpu(tunnel->dst_cache);
840 free_percpu(dev->tstats);
844 void ip_tunnel_dellink(struct net_device *dev, struct list_head *head)
846 struct ip_tunnel *tunnel = netdev_priv(dev);
847 struct ip_tunnel_net *itn;
849 itn = net_generic(tunnel->net, tunnel->ip_tnl_net_id);
851 if (itn->fb_tunnel_dev != dev) {
852 ip_tunnel_del(netdev_priv(dev));
853 unregister_netdevice_queue(dev, head);
856 EXPORT_SYMBOL_GPL(ip_tunnel_dellink);
858 int ip_tunnel_init_net(struct net *net, int ip_tnl_net_id,
859 struct rtnl_link_ops *ops, char *devname)
861 struct ip_tunnel_net *itn = net_generic(net, ip_tnl_net_id);
862 struct ip_tunnel_parm parms;
865 for (i = 0; i < IP_TNL_HASH_SIZE; i++)
866 INIT_HLIST_HEAD(&itn->tunnels[i]);
869 itn->fb_tunnel_dev = NULL;
873 memset(&parms, 0, sizeof(parms));
875 strlcpy(parms.name, devname, IFNAMSIZ);
878 itn->fb_tunnel_dev = __ip_tunnel_create(net, ops, &parms);
879 /* FB netdevice is special: we have one, and only one per netns.
880 * Allowing to move it to another netns is clearly unsafe.
882 if (!IS_ERR(itn->fb_tunnel_dev)) {
883 itn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
884 ip_tunnel_add(itn, netdev_priv(itn->fb_tunnel_dev));
888 return PTR_ERR_OR_ZERO(itn->fb_tunnel_dev);
890 EXPORT_SYMBOL_GPL(ip_tunnel_init_net);
892 static void ip_tunnel_destroy(struct ip_tunnel_net *itn, struct list_head *head,
893 struct rtnl_link_ops *ops)
895 struct net *net = dev_net(itn->fb_tunnel_dev);
896 struct net_device *dev, *aux;
899 for_each_netdev_safe(net, dev, aux)
900 if (dev->rtnl_link_ops == ops)
901 unregister_netdevice_queue(dev, head);
903 for (h = 0; h < IP_TNL_HASH_SIZE; h++) {
905 struct hlist_node *n;
906 struct hlist_head *thead = &itn->tunnels[h];
908 hlist_for_each_entry_safe(t, n, thead, hash_node)
909 /* If dev is in the same netns, it has already
910 * been added to the list by the previous loop.
912 if (!net_eq(dev_net(t->dev), net))
913 unregister_netdevice_queue(t->dev, head);
917 void ip_tunnel_delete_net(struct ip_tunnel_net *itn, struct rtnl_link_ops *ops)
922 ip_tunnel_destroy(itn, &list, ops);
923 unregister_netdevice_many(&list);
926 EXPORT_SYMBOL_GPL(ip_tunnel_delete_net);
928 int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
929 struct ip_tunnel_parm *p)
931 struct ip_tunnel *nt;
932 struct net *net = dev_net(dev);
933 struct ip_tunnel_net *itn;
937 nt = netdev_priv(dev);
938 itn = net_generic(net, nt->ip_tnl_net_id);
940 if (ip_tunnel_find(itn, p, dev->type))
945 err = register_netdevice(dev);
949 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
950 eth_hw_addr_random(dev);
952 mtu = ip_tunnel_bind_dev(dev);
956 ip_tunnel_add(itn, nt);
961 EXPORT_SYMBOL_GPL(ip_tunnel_newlink);
963 int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
964 struct ip_tunnel_parm *p)
967 struct ip_tunnel *tunnel = netdev_priv(dev);
968 struct net *net = tunnel->net;
969 struct ip_tunnel_net *itn = net_generic(net, tunnel->ip_tnl_net_id);
971 if (dev == itn->fb_tunnel_dev)
974 t = ip_tunnel_find(itn, p, dev->type);
982 if (dev->type != ARPHRD_ETHER) {
983 unsigned int nflags = 0;
985 if (ipv4_is_multicast(p->iph.daddr))
986 nflags = IFF_BROADCAST;
987 else if (p->iph.daddr)
988 nflags = IFF_POINTOPOINT;
990 if ((dev->flags ^ nflags) &
991 (IFF_POINTOPOINT | IFF_BROADCAST))
996 ip_tunnel_update(itn, t, dev, p, !tb[IFLA_MTU]);
999 EXPORT_SYMBOL_GPL(ip_tunnel_changelink);
1001 int ip_tunnel_init(struct net_device *dev)
1003 struct ip_tunnel *tunnel = netdev_priv(dev);
1004 struct iphdr *iph = &tunnel->parms.iph;
1007 dev->destructor = ip_tunnel_dev_free;
1008 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1012 tunnel->dst_cache = alloc_percpu(struct ip_tunnel_dst);
1013 if (!tunnel->dst_cache) {
1014 free_percpu(dev->tstats);
1018 err = gro_cells_init(&tunnel->gro_cells, dev);
1020 free_percpu(tunnel->dst_cache);
1021 free_percpu(dev->tstats);
1026 tunnel->net = dev_net(dev);
1027 strcpy(tunnel->parms.name, dev->name);
1033 EXPORT_SYMBOL_GPL(ip_tunnel_init);
1035 void ip_tunnel_uninit(struct net_device *dev)
1037 struct ip_tunnel *tunnel = netdev_priv(dev);
1038 struct net *net = tunnel->net;
1039 struct ip_tunnel_net *itn;
1041 itn = net_generic(net, tunnel->ip_tnl_net_id);
1042 /* fb_tunnel_dev will be unregisted in net-exit call. */
1043 if (itn->fb_tunnel_dev != dev)
1044 ip_tunnel_del(netdev_priv(dev));
1046 ip_tunnel_dst_reset_all(tunnel);
1048 EXPORT_SYMBOL_GPL(ip_tunnel_uninit);
1050 /* Do least required initialization, rest of init is done in tunnel_init call */
1051 void ip_tunnel_setup(struct net_device *dev, int net_id)
1053 struct ip_tunnel *tunnel = netdev_priv(dev);
1054 tunnel->ip_tnl_net_id = net_id;
1056 EXPORT_SYMBOL_GPL(ip_tunnel_setup);
1058 MODULE_LICENSE("GPL");