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>
47 #include <net/protocol.h>
48 #include <net/ip_tunnels.h>
50 #include <net/checksum.h>
51 #include <net/dsfield.h>
52 #include <net/inet_ecn.h>
54 #include <net/net_namespace.h>
55 #include <net/netns/generic.h>
56 #include <net/rtnetlink.h>
58 #if IS_ENABLED(CONFIG_IPV6)
60 #include <net/ip6_fib.h>
61 #include <net/ip6_route.h>
64 static unsigned int ip_tunnel_hash(struct ip_tunnel_net *itn,
65 __be32 key, __be32 remote)
67 return hash_32((__force u32)key ^ (__force u32)remote,
71 static inline void __tunnel_dst_set(struct ip_tunnel_dst *idst,
72 struct dst_entry *dst)
74 struct dst_entry *old_dst;
76 if (dst && (dst->flags & DST_NOCACHE))
79 spin_lock_bh(&idst->lock);
80 old_dst = rcu_dereference(idst->dst);
81 rcu_assign_pointer(idst->dst, dst);
83 spin_unlock_bh(&idst->lock);
86 static inline 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 inline void tunnel_dst_reset(struct ip_tunnel *t)
93 tunnel_dst_set(t, NULL);
96 static void 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);
104 static inline struct dst_entry *tunnel_dst_get(struct ip_tunnel *t)
106 struct dst_entry *dst;
109 dst = rcu_dereference(this_cpu_ptr(t->dst_cache)->dst);
116 struct dst_entry *tunnel_dst_check(struct ip_tunnel *t, u32 cookie)
118 struct dst_entry *dst = tunnel_dst_get(t);
120 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
128 /* Often modified stats are per cpu, other are shared (netdev->stats) */
129 struct rtnl_link_stats64 *ip_tunnel_get_stats64(struct net_device *dev,
130 struct rtnl_link_stats64 *tot)
134 for_each_possible_cpu(i) {
135 const struct pcpu_sw_netstats *tstats =
136 per_cpu_ptr(dev->tstats, i);
137 u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
141 start = u64_stats_fetch_begin_bh(&tstats->syncp);
142 rx_packets = tstats->rx_packets;
143 tx_packets = tstats->tx_packets;
144 rx_bytes = tstats->rx_bytes;
145 tx_bytes = tstats->tx_bytes;
146 } while (u64_stats_fetch_retry_bh(&tstats->syncp, start));
148 tot->rx_packets += rx_packets;
149 tot->tx_packets += tx_packets;
150 tot->rx_bytes += rx_bytes;
151 tot->tx_bytes += tx_bytes;
154 tot->multicast = dev->stats.multicast;
156 tot->rx_crc_errors = dev->stats.rx_crc_errors;
157 tot->rx_fifo_errors = dev->stats.rx_fifo_errors;
158 tot->rx_length_errors = dev->stats.rx_length_errors;
159 tot->rx_frame_errors = dev->stats.rx_frame_errors;
160 tot->rx_errors = dev->stats.rx_errors;
162 tot->tx_fifo_errors = dev->stats.tx_fifo_errors;
163 tot->tx_carrier_errors = dev->stats.tx_carrier_errors;
164 tot->tx_dropped = dev->stats.tx_dropped;
165 tot->tx_aborted_errors = dev->stats.tx_aborted_errors;
166 tot->tx_errors = dev->stats.tx_errors;
168 tot->collisions = dev->stats.collisions;
172 EXPORT_SYMBOL_GPL(ip_tunnel_get_stats64);
174 static bool ip_tunnel_key_match(const struct ip_tunnel_parm *p,
175 __be16 flags, __be32 key)
177 if (p->i_flags & TUNNEL_KEY) {
178 if (flags & TUNNEL_KEY)
179 return key == p->i_key;
181 /* key expected, none present */
184 return !(flags & TUNNEL_KEY);
187 /* Fallback tunnel: no source, no destination, no key, no options
190 We require exact key match i.e. if a key is present in packet
191 it will match only tunnel with the same key; if it is not present,
192 it will match only keyless tunnel.
194 All keysless packets, if not matched configured keyless tunnels
195 will match fallback tunnel.
196 Given src, dst and key, find appropriate for input tunnel.
198 struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
199 int link, __be16 flags,
200 __be32 remote, __be32 local,
204 struct ip_tunnel *t, *cand = NULL;
205 struct hlist_head *head;
207 hash = ip_tunnel_hash(itn, key, remote);
208 head = &itn->tunnels[hash];
210 hlist_for_each_entry_rcu(t, head, hash_node) {
211 if (local != t->parms.iph.saddr ||
212 remote != t->parms.iph.daddr ||
213 !(t->dev->flags & IFF_UP))
216 if (!ip_tunnel_key_match(&t->parms, flags, key))
219 if (t->parms.link == link)
225 hlist_for_each_entry_rcu(t, head, hash_node) {
226 if (remote != t->parms.iph.daddr ||
227 !(t->dev->flags & IFF_UP))
230 if (!ip_tunnel_key_match(&t->parms, flags, key))
233 if (t->parms.link == link)
239 hash = ip_tunnel_hash(itn, key, 0);
240 head = &itn->tunnels[hash];
242 hlist_for_each_entry_rcu(t, head, hash_node) {
243 if ((local != t->parms.iph.saddr &&
244 (local != t->parms.iph.daddr ||
245 !ipv4_is_multicast(local))) ||
246 !(t->dev->flags & IFF_UP))
249 if (!ip_tunnel_key_match(&t->parms, flags, key))
252 if (t->parms.link == link)
258 if (flags & TUNNEL_NO_KEY)
259 goto skip_key_lookup;
261 hlist_for_each_entry_rcu(t, head, hash_node) {
262 if (t->parms.i_key != key ||
263 !(t->dev->flags & IFF_UP))
266 if (t->parms.link == link)
276 if (itn->fb_tunnel_dev && itn->fb_tunnel_dev->flags & IFF_UP)
277 return netdev_priv(itn->fb_tunnel_dev);
282 EXPORT_SYMBOL_GPL(ip_tunnel_lookup);
284 static struct hlist_head *ip_bucket(struct ip_tunnel_net *itn,
285 struct ip_tunnel_parm *parms)
290 if (parms->iph.daddr && !ipv4_is_multicast(parms->iph.daddr))
291 remote = parms->iph.daddr;
295 h = ip_tunnel_hash(itn, parms->i_key, remote);
296 return &itn->tunnels[h];
299 static void ip_tunnel_add(struct ip_tunnel_net *itn, struct ip_tunnel *t)
301 struct hlist_head *head = ip_bucket(itn, &t->parms);
303 hlist_add_head_rcu(&t->hash_node, head);
306 static void ip_tunnel_del(struct ip_tunnel *t)
308 hlist_del_init_rcu(&t->hash_node);
311 static struct ip_tunnel *ip_tunnel_find(struct ip_tunnel_net *itn,
312 struct ip_tunnel_parm *parms,
315 __be32 remote = parms->iph.daddr;
316 __be32 local = parms->iph.saddr;
317 __be32 key = parms->i_key;
318 int link = parms->link;
319 struct ip_tunnel *t = NULL;
320 struct hlist_head *head = ip_bucket(itn, parms);
322 hlist_for_each_entry_rcu(t, head, hash_node) {
323 if (local == t->parms.iph.saddr &&
324 remote == t->parms.iph.daddr &&
325 key == t->parms.i_key &&
326 link == t->parms.link &&
327 type == t->dev->type)
333 static struct net_device *__ip_tunnel_create(struct net *net,
334 const struct rtnl_link_ops *ops,
335 struct ip_tunnel_parm *parms)
338 struct ip_tunnel *tunnel;
339 struct net_device *dev;
343 strlcpy(name, parms->name, IFNAMSIZ);
345 if (strlen(ops->kind) > (IFNAMSIZ - 3)) {
349 strlcpy(name, ops->kind, IFNAMSIZ);
350 strncat(name, "%d", 2);
354 dev = alloc_netdev(ops->priv_size, name, ops->setup);
359 dev_net_set(dev, net);
361 dev->rtnl_link_ops = ops;
363 tunnel = netdev_priv(dev);
364 tunnel->parms = *parms;
367 err = register_netdevice(dev);
379 static inline void init_tunnel_flow(struct flowi4 *fl4,
381 __be32 daddr, __be32 saddr,
382 __be32 key, __u8 tos, int oif)
384 memset(fl4, 0, sizeof(*fl4));
385 fl4->flowi4_oif = oif;
388 fl4->flowi4_tos = tos;
389 fl4->flowi4_proto = proto;
390 fl4->fl4_gre_key = key;
393 static int ip_tunnel_bind_dev(struct net_device *dev)
395 struct net_device *tdev = NULL;
396 struct ip_tunnel *tunnel = netdev_priv(dev);
397 const struct iphdr *iph;
398 int hlen = LL_MAX_HEADER;
399 int mtu = ETH_DATA_LEN;
400 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
402 iph = &tunnel->parms.iph;
404 /* Guess output device to choose reasonable mtu and needed_headroom */
409 init_tunnel_flow(&fl4, iph->protocol, iph->daddr,
410 iph->saddr, tunnel->parms.o_key,
411 RT_TOS(iph->tos), tunnel->parms.link);
412 rt = ip_route_output_key(tunnel->net, &fl4);
416 tunnel_dst_set(tunnel, dst_clone(&rt->dst));
419 if (dev->type != ARPHRD_ETHER)
420 dev->flags |= IFF_POINTOPOINT;
423 if (!tdev && tunnel->parms.link)
424 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
427 hlen = tdev->hard_header_len + tdev->needed_headroom;
430 dev->iflink = tunnel->parms.link;
432 dev->needed_headroom = t_hlen + hlen;
433 mtu -= (dev->hard_header_len + t_hlen);
441 static struct ip_tunnel *ip_tunnel_create(struct net *net,
442 struct ip_tunnel_net *itn,
443 struct ip_tunnel_parm *parms)
445 struct ip_tunnel *nt, *fbt;
446 struct net_device *dev;
448 BUG_ON(!itn->fb_tunnel_dev);
449 fbt = netdev_priv(itn->fb_tunnel_dev);
450 dev = __ip_tunnel_create(net, itn->fb_tunnel_dev->rtnl_link_ops, parms);
454 dev->mtu = ip_tunnel_bind_dev(dev);
456 nt = netdev_priv(dev);
457 ip_tunnel_add(itn, nt);
461 int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
462 const struct tnl_ptk_info *tpi, bool log_ecn_error)
464 struct pcpu_sw_netstats *tstats;
465 const struct iphdr *iph = ip_hdr(skb);
468 #ifdef CONFIG_NET_IPGRE_BROADCAST
469 if (ipv4_is_multicast(iph->daddr)) {
470 /* Looped back packet, drop it! */
471 if (rt_is_output_route(skb_rtable(skb)))
473 tunnel->dev->stats.multicast++;
474 skb->pkt_type = PACKET_BROADCAST;
478 if ((!(tpi->flags&TUNNEL_CSUM) && (tunnel->parms.i_flags&TUNNEL_CSUM)) ||
479 ((tpi->flags&TUNNEL_CSUM) && !(tunnel->parms.i_flags&TUNNEL_CSUM))) {
480 tunnel->dev->stats.rx_crc_errors++;
481 tunnel->dev->stats.rx_errors++;
485 if (tunnel->parms.i_flags&TUNNEL_SEQ) {
486 if (!(tpi->flags&TUNNEL_SEQ) ||
487 (tunnel->i_seqno && (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
488 tunnel->dev->stats.rx_fifo_errors++;
489 tunnel->dev->stats.rx_errors++;
492 tunnel->i_seqno = ntohl(tpi->seq) + 1;
495 err = IP_ECN_decapsulate(iph, skb);
498 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
499 &iph->saddr, iph->tos);
501 ++tunnel->dev->stats.rx_frame_errors;
502 ++tunnel->dev->stats.rx_errors;
507 tstats = this_cpu_ptr(tunnel->dev->tstats);
508 u64_stats_update_begin(&tstats->syncp);
509 tstats->rx_packets++;
510 tstats->rx_bytes += skb->len;
511 u64_stats_update_end(&tstats->syncp);
513 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
515 if (tunnel->dev->type == ARPHRD_ETHER) {
516 skb->protocol = eth_type_trans(skb, tunnel->dev);
517 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
519 skb->dev = tunnel->dev;
522 gro_cells_receive(&tunnel->gro_cells, skb);
529 EXPORT_SYMBOL_GPL(ip_tunnel_rcv);
531 static int tnl_update_pmtu(struct net_device *dev, struct sk_buff *skb,
532 struct rtable *rt, __be16 df)
534 struct ip_tunnel *tunnel = netdev_priv(dev);
535 int pkt_size = skb->len - tunnel->hlen - dev->hard_header_len;
539 mtu = dst_mtu(&rt->dst) - dev->hard_header_len
540 - sizeof(struct iphdr) - tunnel->hlen;
542 mtu = skb_dst(skb) ? dst_mtu(skb_dst(skb)) : dev->mtu;
545 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
547 if (skb->protocol == htons(ETH_P_IP)) {
548 if (!skb_is_gso(skb) &&
549 (df & htons(IP_DF)) && mtu < pkt_size) {
550 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
551 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, htonl(mtu));
555 #if IS_ENABLED(CONFIG_IPV6)
556 else if (skb->protocol == htons(ETH_P_IPV6)) {
557 struct rt6_info *rt6 = (struct rt6_info *)skb_dst(skb);
559 if (rt6 && mtu < dst_mtu(skb_dst(skb)) &&
560 mtu >= IPV6_MIN_MTU) {
561 if ((tunnel->parms.iph.daddr &&
562 !ipv4_is_multicast(tunnel->parms.iph.daddr)) ||
563 rt6->rt6i_dst.plen == 128) {
564 rt6->rt6i_flags |= RTF_MODIFIED;
565 dst_metric_set(skb_dst(skb), RTAX_MTU, mtu);
569 if (!skb_is_gso(skb) && mtu >= IPV6_MIN_MTU &&
571 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
579 void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
580 const struct iphdr *tnl_params, const u8 protocol)
582 struct ip_tunnel *tunnel = netdev_priv(dev);
583 const struct iphdr *inner_iph;
587 struct rtable *rt = NULL; /* Route to the other host */
588 unsigned int max_headroom; /* The extra header space needed */
591 bool connected = true;
593 inner_iph = (const struct iphdr *)skb_inner_network_header(skb);
595 dst = tnl_params->daddr;
599 if (skb_dst(skb) == NULL) {
600 dev->stats.tx_fifo_errors++;
604 if (skb->protocol == htons(ETH_P_IP)) {
605 rt = skb_rtable(skb);
606 dst = rt_nexthop(rt, inner_iph->daddr);
608 #if IS_ENABLED(CONFIG_IPV6)
609 else if (skb->protocol == htons(ETH_P_IPV6)) {
610 const struct in6_addr *addr6;
611 struct neighbour *neigh;
612 bool do_tx_error_icmp;
615 neigh = dst_neigh_lookup(skb_dst(skb),
616 &ipv6_hdr(skb)->daddr);
620 addr6 = (const struct in6_addr *)&neigh->primary_key;
621 addr_type = ipv6_addr_type(addr6);
623 if (addr_type == IPV6_ADDR_ANY) {
624 addr6 = &ipv6_hdr(skb)->daddr;
625 addr_type = ipv6_addr_type(addr6);
628 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
629 do_tx_error_icmp = true;
631 do_tx_error_icmp = false;
632 dst = addr6->s6_addr32[3];
634 neigh_release(neigh);
635 if (do_tx_error_icmp)
645 tos = tnl_params->tos;
648 if (skb->protocol == htons(ETH_P_IP)) {
649 tos = inner_iph->tos;
651 } else if (skb->protocol == htons(ETH_P_IPV6)) {
652 tos = ipv6_get_dsfield((const struct ipv6hdr *)inner_iph);
657 init_tunnel_flow(&fl4, protocol, dst, tnl_params->saddr,
658 tunnel->parms.o_key, RT_TOS(tos), tunnel->parms.link);
661 rt = (struct rtable *)tunnel_dst_check(tunnel, 0);
664 rt = ip_route_output_key(tunnel->net, &fl4);
667 dev->stats.tx_carrier_errors++;
671 tunnel_dst_set(tunnel, dst_clone(&rt->dst));
674 if (rt->dst.dev == dev) {
676 dev->stats.collisions++;
680 if (tnl_update_pmtu(dev, skb, rt, tnl_params->frag_off)) {
685 if (tunnel->err_count > 0) {
686 if (time_before(jiffies,
687 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
690 dst_link_failure(skb);
692 tunnel->err_count = 0;
695 tos = ip_tunnel_ecn_encap(tos, inner_iph, skb);
696 ttl = tnl_params->ttl;
698 if (skb->protocol == htons(ETH_P_IP))
699 ttl = inner_iph->ttl;
700 #if IS_ENABLED(CONFIG_IPV6)
701 else if (skb->protocol == htons(ETH_P_IPV6))
702 ttl = ((const struct ipv6hdr *)inner_iph)->hop_limit;
705 ttl = ip4_dst_hoplimit(&rt->dst);
708 df = tnl_params->frag_off;
709 if (skb->protocol == htons(ETH_P_IP))
710 df |= (inner_iph->frag_off&htons(IP_DF));
712 max_headroom = LL_RESERVED_SPACE(rt->dst.dev) + sizeof(struct iphdr)
713 + rt->dst.header_len;
714 if (max_headroom > dev->needed_headroom)
715 dev->needed_headroom = max_headroom;
717 if (skb_cow_head(skb, dev->needed_headroom)) {
718 dev->stats.tx_dropped++;
723 err = iptunnel_xmit(rt, skb, fl4.saddr, fl4.daddr, protocol,
724 tos, ttl, df, !net_eq(tunnel->net, dev_net(dev)));
725 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
729 #if IS_ENABLED(CONFIG_IPV6)
731 dst_link_failure(skb);
734 dev->stats.tx_errors++;
737 EXPORT_SYMBOL_GPL(ip_tunnel_xmit);
739 static void ip_tunnel_update(struct ip_tunnel_net *itn,
741 struct net_device *dev,
742 struct ip_tunnel_parm *p,
746 t->parms.iph.saddr = p->iph.saddr;
747 t->parms.iph.daddr = p->iph.daddr;
748 t->parms.i_key = p->i_key;
749 t->parms.o_key = p->o_key;
750 if (dev->type != ARPHRD_ETHER) {
751 memcpy(dev->dev_addr, &p->iph.saddr, 4);
752 memcpy(dev->broadcast, &p->iph.daddr, 4);
754 ip_tunnel_add(itn, t);
756 t->parms.iph.ttl = p->iph.ttl;
757 t->parms.iph.tos = p->iph.tos;
758 t->parms.iph.frag_off = p->iph.frag_off;
760 if (t->parms.link != p->link) {
763 t->parms.link = p->link;
764 mtu = ip_tunnel_bind_dev(dev);
768 tunnel_dst_reset_all(t);
769 netdev_state_change(dev);
772 int ip_tunnel_ioctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd)
776 struct net *net = dev_net(dev);
777 struct ip_tunnel *tunnel = netdev_priv(dev);
778 struct ip_tunnel_net *itn = net_generic(net, tunnel->ip_tnl_net_id);
780 BUG_ON(!itn->fb_tunnel_dev);
784 if (dev == itn->fb_tunnel_dev)
785 t = ip_tunnel_find(itn, p, itn->fb_tunnel_dev->type);
787 t = netdev_priv(dev);
788 memcpy(p, &t->parms, sizeof(*p));
794 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
797 p->iph.frag_off |= htons(IP_DF);
798 if (!(p->i_flags&TUNNEL_KEY))
800 if (!(p->o_flags&TUNNEL_KEY))
803 t = ip_tunnel_find(itn, p, itn->fb_tunnel_dev->type);
805 if (!t && (cmd == SIOCADDTUNNEL))
806 t = ip_tunnel_create(net, itn, p);
808 if (dev != itn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
815 unsigned int nflags = 0;
817 if (ipv4_is_multicast(p->iph.daddr))
818 nflags = IFF_BROADCAST;
819 else if (p->iph.daddr)
820 nflags = IFF_POINTOPOINT;
822 if ((dev->flags^nflags)&(IFF_POINTOPOINT|IFF_BROADCAST)) {
827 t = netdev_priv(dev);
833 ip_tunnel_update(itn, t, dev, p, true);
835 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
840 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
843 if (dev == itn->fb_tunnel_dev) {
845 t = ip_tunnel_find(itn, p, itn->fb_tunnel_dev->type);
849 if (t == netdev_priv(itn->fb_tunnel_dev))
853 unregister_netdevice(dev);
864 EXPORT_SYMBOL_GPL(ip_tunnel_ioctl);
866 int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu)
868 struct ip_tunnel *tunnel = netdev_priv(dev);
869 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
872 new_mtu > 0xFFF8 - dev->hard_header_len - t_hlen)
877 EXPORT_SYMBOL_GPL(ip_tunnel_change_mtu);
879 static void ip_tunnel_dev_free(struct net_device *dev)
881 struct ip_tunnel *tunnel = netdev_priv(dev);
883 gro_cells_destroy(&tunnel->gro_cells);
884 free_percpu(tunnel->dst_cache);
885 free_percpu(dev->tstats);
889 void ip_tunnel_dellink(struct net_device *dev, struct list_head *head)
891 struct ip_tunnel *tunnel = netdev_priv(dev);
892 struct ip_tunnel_net *itn;
894 itn = net_generic(tunnel->net, tunnel->ip_tnl_net_id);
896 if (itn->fb_tunnel_dev != dev) {
897 ip_tunnel_del(netdev_priv(dev));
898 unregister_netdevice_queue(dev, head);
901 EXPORT_SYMBOL_GPL(ip_tunnel_dellink);
903 int ip_tunnel_init_net(struct net *net, int ip_tnl_net_id,
904 struct rtnl_link_ops *ops, char *devname)
906 struct ip_tunnel_net *itn = net_generic(net, ip_tnl_net_id);
907 struct ip_tunnel_parm parms;
910 for (i = 0; i < IP_TNL_HASH_SIZE; i++)
911 INIT_HLIST_HEAD(&itn->tunnels[i]);
914 itn->fb_tunnel_dev = NULL;
918 memset(&parms, 0, sizeof(parms));
920 strlcpy(parms.name, devname, IFNAMSIZ);
923 itn->fb_tunnel_dev = __ip_tunnel_create(net, ops, &parms);
924 /* FB netdevice is special: we have one, and only one per netns.
925 * Allowing to move it to another netns is clearly unsafe.
927 if (!IS_ERR(itn->fb_tunnel_dev)) {
928 itn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
929 ip_tunnel_add(itn, netdev_priv(itn->fb_tunnel_dev));
933 return PTR_RET(itn->fb_tunnel_dev);
935 EXPORT_SYMBOL_GPL(ip_tunnel_init_net);
937 static void ip_tunnel_destroy(struct ip_tunnel_net *itn, struct list_head *head,
938 struct rtnl_link_ops *ops)
940 struct net *net = dev_net(itn->fb_tunnel_dev);
941 struct net_device *dev, *aux;
944 for_each_netdev_safe(net, dev, aux)
945 if (dev->rtnl_link_ops == ops)
946 unregister_netdevice_queue(dev, head);
948 for (h = 0; h < IP_TNL_HASH_SIZE; h++) {
950 struct hlist_node *n;
951 struct hlist_head *thead = &itn->tunnels[h];
953 hlist_for_each_entry_safe(t, n, thead, hash_node)
954 /* If dev is in the same netns, it has already
955 * been added to the list by the previous loop.
957 if (!net_eq(dev_net(t->dev), net))
958 unregister_netdevice_queue(t->dev, head);
962 void ip_tunnel_delete_net(struct ip_tunnel_net *itn, struct rtnl_link_ops *ops)
967 ip_tunnel_destroy(itn, &list, ops);
968 unregister_netdevice_many(&list);
971 EXPORT_SYMBOL_GPL(ip_tunnel_delete_net);
973 int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
974 struct ip_tunnel_parm *p)
976 struct ip_tunnel *nt;
977 struct net *net = dev_net(dev);
978 struct ip_tunnel_net *itn;
982 nt = netdev_priv(dev);
983 itn = net_generic(net, nt->ip_tnl_net_id);
985 if (ip_tunnel_find(itn, p, dev->type))
990 err = register_netdevice(dev);
994 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
995 eth_hw_addr_random(dev);
997 mtu = ip_tunnel_bind_dev(dev);
1001 ip_tunnel_add(itn, nt);
1006 EXPORT_SYMBOL_GPL(ip_tunnel_newlink);
1008 int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
1009 struct ip_tunnel_parm *p)
1011 struct ip_tunnel *t;
1012 struct ip_tunnel *tunnel = netdev_priv(dev);
1013 struct net *net = tunnel->net;
1014 struct ip_tunnel_net *itn = net_generic(net, tunnel->ip_tnl_net_id);
1016 if (dev == itn->fb_tunnel_dev)
1019 t = ip_tunnel_find(itn, p, dev->type);
1027 if (dev->type != ARPHRD_ETHER) {
1028 unsigned int nflags = 0;
1030 if (ipv4_is_multicast(p->iph.daddr))
1031 nflags = IFF_BROADCAST;
1032 else if (p->iph.daddr)
1033 nflags = IFF_POINTOPOINT;
1035 if ((dev->flags ^ nflags) &
1036 (IFF_POINTOPOINT | IFF_BROADCAST))
1041 ip_tunnel_update(itn, t, dev, p, !tb[IFLA_MTU]);
1044 EXPORT_SYMBOL_GPL(ip_tunnel_changelink);
1046 int ip_tunnel_init(struct net_device *dev)
1048 struct ip_tunnel *tunnel = netdev_priv(dev);
1049 struct iphdr *iph = &tunnel->parms.iph;
1052 dev->destructor = ip_tunnel_dev_free;
1053 dev->tstats = alloc_percpu(struct pcpu_sw_netstats);
1057 for_each_possible_cpu(i) {
1058 struct pcpu_sw_netstats *ipt_stats;
1059 ipt_stats = per_cpu_ptr(dev->tstats, i);
1060 u64_stats_init(&ipt_stats->syncp);
1063 tunnel->dst_cache = alloc_percpu(struct ip_tunnel_dst);
1064 if (!tunnel->dst_cache) {
1065 free_percpu(dev->tstats);
1069 for_each_possible_cpu(i) {
1070 struct ip_tunnel_dst *idst = per_cpu_ptr(tunnel->dst_cache, i);
1072 spin_lock_init(&idst->lock);
1075 err = gro_cells_init(&tunnel->gro_cells, dev);
1077 free_percpu(tunnel->dst_cache);
1078 free_percpu(dev->tstats);
1083 tunnel->net = dev_net(dev);
1084 strcpy(tunnel->parms.name, dev->name);
1090 EXPORT_SYMBOL_GPL(ip_tunnel_init);
1092 void ip_tunnel_uninit(struct net_device *dev)
1094 struct ip_tunnel *tunnel = netdev_priv(dev);
1095 struct net *net = tunnel->net;
1096 struct ip_tunnel_net *itn;
1098 itn = net_generic(net, tunnel->ip_tnl_net_id);
1099 /* fb_tunnel_dev will be unregisted in net-exit call. */
1100 if (itn->fb_tunnel_dev != dev)
1101 ip_tunnel_del(netdev_priv(dev));
1103 tunnel_dst_reset_all(tunnel);
1105 EXPORT_SYMBOL_GPL(ip_tunnel_uninit);
1107 /* Do least required initialization, rest of init is done in tunnel_init call */
1108 void ip_tunnel_setup(struct net_device *dev, int net_id)
1110 struct ip_tunnel *tunnel = netdev_priv(dev);
1111 tunnel->ip_tnl_net_id = net_id;
1113 EXPORT_SYMBOL_GPL(ip_tunnel_setup);
1115 MODULE_LICENSE("GPL");