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/init.h>
34 #include <linux/in6.h>
35 #include <linux/inetdevice.h>
36 #include <linux/igmp.h>
37 #include <linux/netfilter_ipv4.h>
38 #include <linux/etherdevice.h>
39 #include <linux/if_ether.h>
40 #include <linux/if_vlan.h>
41 #include <linux/rculist.h>
42 #include <linux/err.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>
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, __be32 saddr)
74 struct dst_entry *old_dst;
77 old_dst = xchg((__force struct dst_entry **)&idst->dst, dst);
82 static noinline void tunnel_dst_set(struct ip_tunnel *t,
83 struct dst_entry *dst, __be32 saddr)
85 __tunnel_dst_set(raw_cpu_ptr(t->dst_cache), dst, saddr);
88 static void tunnel_dst_reset(struct ip_tunnel *t)
90 tunnel_dst_set(t, NULL, 0);
93 void ip_tunnel_dst_reset_all(struct ip_tunnel *t)
97 for_each_possible_cpu(i)
98 __tunnel_dst_set(per_cpu_ptr(t->dst_cache, i), NULL, 0);
100 EXPORT_SYMBOL(ip_tunnel_dst_reset_all);
102 static struct rtable *tunnel_rtable_get(struct ip_tunnel *t,
103 u32 cookie, __be32 *saddr)
105 struct ip_tunnel_dst *idst;
106 struct dst_entry *dst;
109 idst = raw_cpu_ptr(t->dst_cache);
110 dst = rcu_dereference(idst->dst);
111 if (dst && !atomic_inc_not_zero(&dst->__refcnt))
114 if (!dst->obsolete || dst->ops->check(dst, cookie)) {
115 *saddr = idst->saddr;
123 return (struct rtable *)dst;
126 static bool ip_tunnel_key_match(const struct ip_tunnel_parm *p,
127 __be16 flags, __be32 key)
129 if (p->i_flags & TUNNEL_KEY) {
130 if (flags & TUNNEL_KEY)
131 return key == p->i_key;
133 /* key expected, none present */
136 return !(flags & TUNNEL_KEY);
139 /* Fallback tunnel: no source, no destination, no key, no options
142 We require exact key match i.e. if a key is present in packet
143 it will match only tunnel with the same key; if it is not present,
144 it will match only keyless tunnel.
146 All keysless packets, if not matched configured keyless tunnels
147 will match fallback tunnel.
148 Given src, dst and key, find appropriate for input tunnel.
150 struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
151 int link, __be16 flags,
152 __be32 remote, __be32 local,
156 struct ip_tunnel *t, *cand = NULL;
157 struct hlist_head *head;
159 hash = ip_tunnel_hash(key, remote);
160 head = &itn->tunnels[hash];
162 hlist_for_each_entry_rcu(t, head, hash_node) {
163 if (local != t->parms.iph.saddr ||
164 remote != t->parms.iph.daddr ||
165 !(t->dev->flags & IFF_UP))
168 if (!ip_tunnel_key_match(&t->parms, flags, key))
171 if (t->parms.link == link)
177 hlist_for_each_entry_rcu(t, head, hash_node) {
178 if (remote != t->parms.iph.daddr ||
179 t->parms.iph.saddr != 0 ||
180 !(t->dev->flags & IFF_UP))
183 if (!ip_tunnel_key_match(&t->parms, flags, key))
186 if (t->parms.link == link)
192 hash = ip_tunnel_hash(key, 0);
193 head = &itn->tunnels[hash];
195 hlist_for_each_entry_rcu(t, head, hash_node) {
196 if ((local != t->parms.iph.saddr || t->parms.iph.daddr != 0) &&
197 (local != t->parms.iph.daddr || !ipv4_is_multicast(local)))
200 if (!(t->dev->flags & IFF_UP))
203 if (!ip_tunnel_key_match(&t->parms, flags, key))
206 if (t->parms.link == link)
212 if (flags & TUNNEL_NO_KEY)
213 goto skip_key_lookup;
215 hlist_for_each_entry_rcu(t, head, hash_node) {
216 if (t->parms.i_key != key ||
217 t->parms.iph.saddr != 0 ||
218 t->parms.iph.daddr != 0 ||
219 !(t->dev->flags & IFF_UP))
222 if (t->parms.link == link)
232 t = rcu_dereference(itn->collect_md_tun);
236 if (itn->fb_tunnel_dev && itn->fb_tunnel_dev->flags & IFF_UP)
237 return netdev_priv(itn->fb_tunnel_dev);
241 EXPORT_SYMBOL_GPL(ip_tunnel_lookup);
243 static struct hlist_head *ip_bucket(struct ip_tunnel_net *itn,
244 struct ip_tunnel_parm *parms)
248 __be32 i_key = parms->i_key;
250 if (parms->iph.daddr && !ipv4_is_multicast(parms->iph.daddr))
251 remote = parms->iph.daddr;
255 if (!(parms->i_flags & TUNNEL_KEY) && (parms->i_flags & VTI_ISVTI))
258 h = ip_tunnel_hash(i_key, remote);
259 return &itn->tunnels[h];
262 static void ip_tunnel_add(struct ip_tunnel_net *itn, struct ip_tunnel *t)
264 struct hlist_head *head = ip_bucket(itn, &t->parms);
267 rcu_assign_pointer(itn->collect_md_tun, t);
268 hlist_add_head_rcu(&t->hash_node, head);
271 static void ip_tunnel_del(struct ip_tunnel_net *itn, struct ip_tunnel *t)
274 rcu_assign_pointer(itn->collect_md_tun, NULL);
275 hlist_del_init_rcu(&t->hash_node);
278 static struct ip_tunnel *ip_tunnel_find(struct ip_tunnel_net *itn,
279 struct ip_tunnel_parm *parms,
282 __be32 remote = parms->iph.daddr;
283 __be32 local = parms->iph.saddr;
284 __be32 key = parms->i_key;
285 __be16 flags = parms->i_flags;
286 int link = parms->link;
287 struct ip_tunnel *t = NULL;
288 struct hlist_head *head = ip_bucket(itn, parms);
290 hlist_for_each_entry_rcu(t, head, hash_node) {
291 if (local == t->parms.iph.saddr &&
292 remote == t->parms.iph.daddr &&
293 link == t->parms.link &&
294 type == t->dev->type &&
295 ip_tunnel_key_match(&t->parms, flags, key))
301 static struct net_device *__ip_tunnel_create(struct net *net,
302 const struct rtnl_link_ops *ops,
303 struct ip_tunnel_parm *parms)
306 struct ip_tunnel *tunnel;
307 struct net_device *dev;
311 strlcpy(name, parms->name, IFNAMSIZ);
313 if (strlen(ops->kind) > (IFNAMSIZ - 3)) {
317 strlcpy(name, ops->kind, IFNAMSIZ);
318 strncat(name, "%d", 2);
322 dev = alloc_netdev(ops->priv_size, name, NET_NAME_UNKNOWN, ops->setup);
327 dev_net_set(dev, net);
329 dev->rtnl_link_ops = ops;
331 tunnel = netdev_priv(dev);
332 tunnel->parms = *parms;
335 err = register_netdevice(dev);
347 static inline void init_tunnel_flow(struct flowi4 *fl4,
349 __be32 daddr, __be32 saddr,
350 __be32 key, __u8 tos, int oif)
352 memset(fl4, 0, sizeof(*fl4));
353 fl4->flowi4_oif = oif;
356 fl4->flowi4_tos = tos;
357 fl4->flowi4_proto = proto;
358 fl4->fl4_gre_key = key;
361 static int ip_tunnel_bind_dev(struct net_device *dev)
363 struct net_device *tdev = NULL;
364 struct ip_tunnel *tunnel = netdev_priv(dev);
365 const struct iphdr *iph;
366 int hlen = LL_MAX_HEADER;
367 int mtu = ETH_DATA_LEN;
368 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
370 iph = &tunnel->parms.iph;
372 /* Guess output device to choose reasonable mtu and needed_headroom */
377 init_tunnel_flow(&fl4, iph->protocol, iph->daddr,
378 iph->saddr, tunnel->parms.o_key,
379 RT_TOS(iph->tos), tunnel->parms.link);
380 rt = ip_route_output_key(tunnel->net, &fl4);
384 tunnel_dst_set(tunnel, &rt->dst, fl4.saddr);
387 if (dev->type != ARPHRD_ETHER)
388 dev->flags |= IFF_POINTOPOINT;
391 if (!tdev && tunnel->parms.link)
392 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
395 hlen = tdev->hard_header_len + tdev->needed_headroom;
399 dev->needed_headroom = t_hlen + hlen;
400 mtu -= (dev->hard_header_len + t_hlen);
408 static struct ip_tunnel *ip_tunnel_create(struct net *net,
409 struct ip_tunnel_net *itn,
410 struct ip_tunnel_parm *parms)
412 struct ip_tunnel *nt;
413 struct net_device *dev;
415 BUG_ON(!itn->fb_tunnel_dev);
416 dev = __ip_tunnel_create(net, itn->fb_tunnel_dev->rtnl_link_ops, parms);
418 return ERR_CAST(dev);
420 dev->mtu = ip_tunnel_bind_dev(dev);
422 nt = netdev_priv(dev);
423 ip_tunnel_add(itn, nt);
427 int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
428 const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
431 struct pcpu_sw_netstats *tstats;
432 const struct iphdr *iph = ip_hdr(skb);
435 #ifdef CONFIG_NET_IPGRE_BROADCAST
436 if (ipv4_is_multicast(iph->daddr)) {
437 tunnel->dev->stats.multicast++;
438 skb->pkt_type = PACKET_BROADCAST;
442 if ((!(tpi->flags&TUNNEL_CSUM) && (tunnel->parms.i_flags&TUNNEL_CSUM)) ||
443 ((tpi->flags&TUNNEL_CSUM) && !(tunnel->parms.i_flags&TUNNEL_CSUM))) {
444 tunnel->dev->stats.rx_crc_errors++;
445 tunnel->dev->stats.rx_errors++;
449 if (tunnel->parms.i_flags&TUNNEL_SEQ) {
450 if (!(tpi->flags&TUNNEL_SEQ) ||
451 (tunnel->i_seqno && (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
452 tunnel->dev->stats.rx_fifo_errors++;
453 tunnel->dev->stats.rx_errors++;
456 tunnel->i_seqno = ntohl(tpi->seq) + 1;
459 skb_reset_network_header(skb);
461 err = IP_ECN_decapsulate(iph, skb);
464 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
465 &iph->saddr, iph->tos);
467 ++tunnel->dev->stats.rx_frame_errors;
468 ++tunnel->dev->stats.rx_errors;
473 tstats = this_cpu_ptr(tunnel->dev->tstats);
474 u64_stats_update_begin(&tstats->syncp);
475 tstats->rx_packets++;
476 tstats->rx_bytes += skb->len;
477 u64_stats_update_end(&tstats->syncp);
479 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
481 if (tunnel->dev->type == ARPHRD_ETHER) {
482 skb->protocol = eth_type_trans(skb, tunnel->dev);
483 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
485 skb->dev = tunnel->dev;
489 skb_dst_set(skb, (struct dst_entry *)tun_dst);
491 gro_cells_receive(&tunnel->gro_cells, skb);
498 EXPORT_SYMBOL_GPL(ip_tunnel_rcv);
500 static int ip_encap_hlen(struct ip_tunnel_encap *e)
502 const struct ip_tunnel_encap_ops *ops;
505 if (e->type == TUNNEL_ENCAP_NONE)
508 if (e->type >= MAX_IPTUN_ENCAP_OPS)
512 ops = rcu_dereference(iptun_encaps[e->type]);
513 if (likely(ops && ops->encap_hlen))
514 hlen = ops->encap_hlen(e);
520 const struct ip_tunnel_encap_ops __rcu *
521 iptun_encaps[MAX_IPTUN_ENCAP_OPS] __read_mostly;
523 int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *ops,
526 if (num >= MAX_IPTUN_ENCAP_OPS)
529 return !cmpxchg((const struct ip_tunnel_encap_ops **)
533 EXPORT_SYMBOL(ip_tunnel_encap_add_ops);
535 int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *ops,
540 if (num >= MAX_IPTUN_ENCAP_OPS)
543 ret = (cmpxchg((const struct ip_tunnel_encap_ops **)
545 ops, NULL) == ops) ? 0 : -1;
551 EXPORT_SYMBOL(ip_tunnel_encap_del_ops);
553 int ip_tunnel_encap_setup(struct ip_tunnel *t,
554 struct ip_tunnel_encap *ipencap)
558 memset(&t->encap, 0, sizeof(t->encap));
560 hlen = ip_encap_hlen(ipencap);
564 t->encap.type = ipencap->type;
565 t->encap.sport = ipencap->sport;
566 t->encap.dport = ipencap->dport;
567 t->encap.flags = ipencap->flags;
569 t->encap_hlen = hlen;
570 t->hlen = t->encap_hlen + t->tun_hlen;
574 EXPORT_SYMBOL_GPL(ip_tunnel_encap_setup);
576 int ip_tunnel_encap(struct sk_buff *skb, struct ip_tunnel *t,
577 u8 *protocol, struct flowi4 *fl4)
579 const struct ip_tunnel_encap_ops *ops;
582 if (t->encap.type == TUNNEL_ENCAP_NONE)
585 if (t->encap.type >= MAX_IPTUN_ENCAP_OPS)
589 ops = rcu_dereference(iptun_encaps[t->encap.type]);
590 if (likely(ops && ops->build_header))
591 ret = ops->build_header(skb, &t->encap, protocol, fl4);
596 EXPORT_SYMBOL(ip_tunnel_encap);
598 static int tnl_update_pmtu(struct net_device *dev, struct sk_buff *skb,
599 struct rtable *rt, __be16 df,
600 const struct iphdr *inner_iph)
602 struct ip_tunnel *tunnel = netdev_priv(dev);
603 int pkt_size = skb->len - tunnel->hlen - dev->hard_header_len;
607 mtu = dst_mtu(&rt->dst) - dev->hard_header_len
608 - sizeof(struct iphdr) - tunnel->hlen;
610 mtu = skb_dst(skb) ? dst_mtu(skb_dst(skb)) : dev->mtu;
613 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
615 if (skb->protocol == htons(ETH_P_IP)) {
616 if (!skb_is_gso(skb) &&
617 (inner_iph->frag_off & htons(IP_DF)) &&
619 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
620 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, htonl(mtu));
624 #if IS_ENABLED(CONFIG_IPV6)
625 else if (skb->protocol == htons(ETH_P_IPV6)) {
626 struct rt6_info *rt6 = (struct rt6_info *)skb_dst(skb);
628 if (rt6 && mtu < dst_mtu(skb_dst(skb)) &&
629 mtu >= IPV6_MIN_MTU) {
630 if ((tunnel->parms.iph.daddr &&
631 !ipv4_is_multicast(tunnel->parms.iph.daddr)) ||
632 rt6->rt6i_dst.plen == 128) {
633 rt6->rt6i_flags |= RTF_MODIFIED;
634 dst_metric_set(skb_dst(skb), RTAX_MTU, mtu);
638 if (!skb_is_gso(skb) && mtu >= IPV6_MIN_MTU &&
640 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
648 void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
649 const struct iphdr *tnl_params, u8 protocol)
651 struct ip_tunnel *tunnel = netdev_priv(dev);
652 const struct iphdr *inner_iph;
656 struct rtable *rt; /* Route to the other host */
657 unsigned int max_headroom; /* The extra header space needed */
661 inner_iph = (const struct iphdr *)skb_inner_network_header(skb);
662 connected = (tunnel->parms.iph.daddr != 0);
664 dst = tnl_params->daddr;
669 dev->stats.tx_fifo_errors++;
673 if (skb->protocol == htons(ETH_P_IP)) {
674 rt = skb_rtable(skb);
675 dst = rt_nexthop(rt, inner_iph->daddr);
677 #if IS_ENABLED(CONFIG_IPV6)
678 else if (skb->protocol == htons(ETH_P_IPV6)) {
679 const struct in6_addr *addr6;
680 struct neighbour *neigh;
681 bool do_tx_error_icmp;
684 neigh = dst_neigh_lookup(skb_dst(skb),
685 &ipv6_hdr(skb)->daddr);
689 addr6 = (const struct in6_addr *)&neigh->primary_key;
690 addr_type = ipv6_addr_type(addr6);
692 if (addr_type == IPV6_ADDR_ANY) {
693 addr6 = &ipv6_hdr(skb)->daddr;
694 addr_type = ipv6_addr_type(addr6);
697 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
698 do_tx_error_icmp = true;
700 do_tx_error_icmp = false;
701 dst = addr6->s6_addr32[3];
703 neigh_release(neigh);
704 if (do_tx_error_icmp)
714 tos = tnl_params->tos;
717 if (skb->protocol == htons(ETH_P_IP)) {
718 tos = inner_iph->tos;
720 } else if (skb->protocol == htons(ETH_P_IPV6)) {
721 tos = ipv6_get_dsfield((const struct ipv6hdr *)inner_iph);
726 init_tunnel_flow(&fl4, protocol, dst, tnl_params->saddr,
727 tunnel->parms.o_key, RT_TOS(tos), tunnel->parms.link);
729 if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0)
732 rt = connected ? tunnel_rtable_get(tunnel, 0, &fl4.saddr) : NULL;
735 rt = ip_route_output_key(tunnel->net, &fl4);
738 dev->stats.tx_carrier_errors++;
742 tunnel_dst_set(tunnel, &rt->dst, fl4.saddr);
745 if (rt->dst.dev == dev) {
747 dev->stats.collisions++;
751 if (tnl_update_pmtu(dev, skb, rt, tnl_params->frag_off, inner_iph)) {
756 if (tunnel->err_count > 0) {
757 if (time_before(jiffies,
758 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
761 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
762 dst_link_failure(skb);
764 tunnel->err_count = 0;
767 tos = ip_tunnel_ecn_encap(tos, inner_iph, skb);
768 ttl = tnl_params->ttl;
770 if (skb->protocol == htons(ETH_P_IP))
771 ttl = inner_iph->ttl;
772 #if IS_ENABLED(CONFIG_IPV6)
773 else if (skb->protocol == htons(ETH_P_IPV6))
774 ttl = ((const struct ipv6hdr *)inner_iph)->hop_limit;
777 ttl = ip4_dst_hoplimit(&rt->dst);
780 df = tnl_params->frag_off;
781 if (skb->protocol == htons(ETH_P_IP))
782 df |= (inner_iph->frag_off&htons(IP_DF));
784 max_headroom = LL_RESERVED_SPACE(rt->dst.dev) + sizeof(struct iphdr)
785 + rt->dst.header_len + ip_encap_hlen(&tunnel->encap);
786 if (max_headroom > dev->needed_headroom)
787 dev->needed_headroom = max_headroom;
789 if (skb_cow_head(skb, dev->needed_headroom)) {
791 dev->stats.tx_dropped++;
796 iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
797 df, !net_eq(tunnel->net, dev_net(dev)));
800 #if IS_ENABLED(CONFIG_IPV6)
802 dst_link_failure(skb);
805 dev->stats.tx_errors++;
808 EXPORT_SYMBOL_GPL(ip_tunnel_xmit);
810 static void ip_tunnel_update(struct ip_tunnel_net *itn,
812 struct net_device *dev,
813 struct ip_tunnel_parm *p,
816 ip_tunnel_del(itn, t);
817 t->parms.iph.saddr = p->iph.saddr;
818 t->parms.iph.daddr = p->iph.daddr;
819 t->parms.i_key = p->i_key;
820 t->parms.o_key = p->o_key;
821 if (dev->type != ARPHRD_ETHER) {
822 memcpy(dev->dev_addr, &p->iph.saddr, 4);
823 memcpy(dev->broadcast, &p->iph.daddr, 4);
825 ip_tunnel_add(itn, t);
827 t->parms.iph.ttl = p->iph.ttl;
828 t->parms.iph.tos = p->iph.tos;
829 t->parms.iph.frag_off = p->iph.frag_off;
831 if (t->parms.link != p->link) {
834 t->parms.link = p->link;
835 mtu = ip_tunnel_bind_dev(dev);
839 ip_tunnel_dst_reset_all(t);
840 netdev_state_change(dev);
843 int ip_tunnel_ioctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd)
846 struct ip_tunnel *t = netdev_priv(dev);
847 struct net *net = t->net;
848 struct ip_tunnel_net *itn = net_generic(net, t->ip_tnl_net_id);
850 BUG_ON(!itn->fb_tunnel_dev);
853 if (dev == itn->fb_tunnel_dev) {
854 t = ip_tunnel_find(itn, p, itn->fb_tunnel_dev->type);
856 t = netdev_priv(dev);
858 memcpy(p, &t->parms, sizeof(*p));
864 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
867 p->iph.frag_off |= htons(IP_DF);
868 if (!(p->i_flags & VTI_ISVTI)) {
869 if (!(p->i_flags & TUNNEL_KEY))
871 if (!(p->o_flags & TUNNEL_KEY))
875 t = ip_tunnel_find(itn, p, itn->fb_tunnel_dev->type);
877 if (cmd == SIOCADDTUNNEL) {
879 t = ip_tunnel_create(net, itn, p);
880 err = PTR_ERR_OR_ZERO(t);
887 if (dev != itn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
894 unsigned int nflags = 0;
896 if (ipv4_is_multicast(p->iph.daddr))
897 nflags = IFF_BROADCAST;
898 else if (p->iph.daddr)
899 nflags = IFF_POINTOPOINT;
901 if ((dev->flags^nflags)&(IFF_POINTOPOINT|IFF_BROADCAST)) {
906 t = netdev_priv(dev);
912 ip_tunnel_update(itn, t, dev, p, true);
920 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
923 if (dev == itn->fb_tunnel_dev) {
925 t = ip_tunnel_find(itn, p, itn->fb_tunnel_dev->type);
929 if (t == netdev_priv(itn->fb_tunnel_dev))
933 unregister_netdevice(dev);
944 EXPORT_SYMBOL_GPL(ip_tunnel_ioctl);
946 int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu)
948 struct ip_tunnel *tunnel = netdev_priv(dev);
949 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
952 new_mtu > 0xFFF8 - dev->hard_header_len - t_hlen)
957 EXPORT_SYMBOL_GPL(ip_tunnel_change_mtu);
959 static void ip_tunnel_dev_free(struct net_device *dev)
961 struct ip_tunnel *tunnel = netdev_priv(dev);
963 gro_cells_destroy(&tunnel->gro_cells);
964 free_percpu(tunnel->dst_cache);
965 free_percpu(dev->tstats);
969 void ip_tunnel_dellink(struct net_device *dev, struct list_head *head)
971 struct ip_tunnel *tunnel = netdev_priv(dev);
972 struct ip_tunnel_net *itn;
974 itn = net_generic(tunnel->net, tunnel->ip_tnl_net_id);
976 if (itn->fb_tunnel_dev != dev) {
977 ip_tunnel_del(itn, netdev_priv(dev));
978 unregister_netdevice_queue(dev, head);
981 EXPORT_SYMBOL_GPL(ip_tunnel_dellink);
983 struct net *ip_tunnel_get_link_net(const struct net_device *dev)
985 struct ip_tunnel *tunnel = netdev_priv(dev);
989 EXPORT_SYMBOL(ip_tunnel_get_link_net);
991 int ip_tunnel_get_iflink(const struct net_device *dev)
993 struct ip_tunnel *tunnel = netdev_priv(dev);
995 return tunnel->parms.link;
997 EXPORT_SYMBOL(ip_tunnel_get_iflink);
999 int ip_tunnel_init_net(struct net *net, int ip_tnl_net_id,
1000 struct rtnl_link_ops *ops, char *devname)
1002 struct ip_tunnel_net *itn = net_generic(net, ip_tnl_net_id);
1003 struct ip_tunnel_parm parms;
1006 for (i = 0; i < IP_TNL_HASH_SIZE; i++)
1007 INIT_HLIST_HEAD(&itn->tunnels[i]);
1010 itn->fb_tunnel_dev = NULL;
1014 memset(&parms, 0, sizeof(parms));
1016 strlcpy(parms.name, devname, IFNAMSIZ);
1019 itn->fb_tunnel_dev = __ip_tunnel_create(net, ops, &parms);
1020 /* FB netdevice is special: we have one, and only one per netns.
1021 * Allowing to move it to another netns is clearly unsafe.
1023 if (!IS_ERR(itn->fb_tunnel_dev)) {
1024 itn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1025 itn->fb_tunnel_dev->mtu = ip_tunnel_bind_dev(itn->fb_tunnel_dev);
1026 ip_tunnel_add(itn, netdev_priv(itn->fb_tunnel_dev));
1030 return PTR_ERR_OR_ZERO(itn->fb_tunnel_dev);
1032 EXPORT_SYMBOL_GPL(ip_tunnel_init_net);
1034 static void ip_tunnel_destroy(struct ip_tunnel_net *itn, struct list_head *head,
1035 struct rtnl_link_ops *ops)
1037 struct net *net = dev_net(itn->fb_tunnel_dev);
1038 struct net_device *dev, *aux;
1041 for_each_netdev_safe(net, dev, aux)
1042 if (dev->rtnl_link_ops == ops)
1043 unregister_netdevice_queue(dev, head);
1045 for (h = 0; h < IP_TNL_HASH_SIZE; h++) {
1046 struct ip_tunnel *t;
1047 struct hlist_node *n;
1048 struct hlist_head *thead = &itn->tunnels[h];
1050 hlist_for_each_entry_safe(t, n, thead, hash_node)
1051 /* If dev is in the same netns, it has already
1052 * been added to the list by the previous loop.
1054 if (!net_eq(dev_net(t->dev), net))
1055 unregister_netdevice_queue(t->dev, head);
1059 void ip_tunnel_delete_net(struct ip_tunnel_net *itn, struct rtnl_link_ops *ops)
1064 ip_tunnel_destroy(itn, &list, ops);
1065 unregister_netdevice_many(&list);
1068 EXPORT_SYMBOL_GPL(ip_tunnel_delete_net);
1070 int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
1071 struct ip_tunnel_parm *p)
1073 struct ip_tunnel *nt;
1074 struct net *net = dev_net(dev);
1075 struct ip_tunnel_net *itn;
1079 nt = netdev_priv(dev);
1080 itn = net_generic(net, nt->ip_tnl_net_id);
1082 if (nt->collect_md) {
1083 if (rtnl_dereference(itn->collect_md_tun))
1086 if (ip_tunnel_find(itn, p, dev->type))
1092 err = register_netdevice(dev);
1096 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1097 eth_hw_addr_random(dev);
1099 mtu = ip_tunnel_bind_dev(dev);
1103 ip_tunnel_add(itn, nt);
1107 EXPORT_SYMBOL_GPL(ip_tunnel_newlink);
1109 int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
1110 struct ip_tunnel_parm *p)
1112 struct ip_tunnel *t;
1113 struct ip_tunnel *tunnel = netdev_priv(dev);
1114 struct net *net = tunnel->net;
1115 struct ip_tunnel_net *itn = net_generic(net, tunnel->ip_tnl_net_id);
1117 if (dev == itn->fb_tunnel_dev)
1120 t = ip_tunnel_find(itn, p, dev->type);
1128 if (dev->type != ARPHRD_ETHER) {
1129 unsigned int nflags = 0;
1131 if (ipv4_is_multicast(p->iph.daddr))
1132 nflags = IFF_BROADCAST;
1133 else if (p->iph.daddr)
1134 nflags = IFF_POINTOPOINT;
1136 if ((dev->flags ^ nflags) &
1137 (IFF_POINTOPOINT | IFF_BROADCAST))
1142 ip_tunnel_update(itn, t, dev, p, !tb[IFLA_MTU]);
1145 EXPORT_SYMBOL_GPL(ip_tunnel_changelink);
1147 int ip_tunnel_init(struct net_device *dev)
1149 struct ip_tunnel *tunnel = netdev_priv(dev);
1150 struct iphdr *iph = &tunnel->parms.iph;
1153 dev->destructor = ip_tunnel_dev_free;
1154 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1158 tunnel->dst_cache = alloc_percpu(struct ip_tunnel_dst);
1159 if (!tunnel->dst_cache) {
1160 free_percpu(dev->tstats);
1164 err = gro_cells_init(&tunnel->gro_cells, dev);
1166 free_percpu(tunnel->dst_cache);
1167 free_percpu(dev->tstats);
1172 tunnel->net = dev_net(dev);
1173 strcpy(tunnel->parms.name, dev->name);
1177 if (tunnel->collect_md) {
1178 dev->features |= NETIF_F_NETNS_LOCAL;
1179 netif_keep_dst(dev);
1183 EXPORT_SYMBOL_GPL(ip_tunnel_init);
1185 void ip_tunnel_uninit(struct net_device *dev)
1187 struct ip_tunnel *tunnel = netdev_priv(dev);
1188 struct net *net = tunnel->net;
1189 struct ip_tunnel_net *itn;
1191 itn = net_generic(net, tunnel->ip_tnl_net_id);
1192 /* fb_tunnel_dev will be unregisted in net-exit call. */
1193 if (itn->fb_tunnel_dev != dev)
1194 ip_tunnel_del(itn, netdev_priv(dev));
1196 ip_tunnel_dst_reset_all(tunnel);
1198 EXPORT_SYMBOL_GPL(ip_tunnel_uninit);
1200 /* Do least required initialization, rest of init is done in tunnel_init call */
1201 void ip_tunnel_setup(struct net_device *dev, int net_id)
1203 struct ip_tunnel *tunnel = netdev_priv(dev);
1204 tunnel->ip_tnl_net_id = net_id;
1206 EXPORT_SYMBOL_GPL(ip_tunnel_setup);
1208 MODULE_LICENSE("GPL");