2 * Linux NET3: IP/IP protocol decoder modified to support
3 * virtual tunnel interface
6 * Saurabh Mohan (saurabh.mohan@vyatta.com) 05/07/2012
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
16 This version of net/ipv4/ip_vti.c is cloned of net/ipv4/ipip.c
18 For comments look at net/ipv4/ip_gre.c --ANK
22 #include <linux/capability.h>
23 #include <linux/module.h>
24 #include <linux/types.h>
25 #include <linux/kernel.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/netfilter_ipv4.h>
35 #include <linux/if_ether.h>
36 #include <linux/icmpv6.h>
41 #include <net/ip_tunnels.h>
42 #include <net/inet_ecn.h>
44 #include <net/net_namespace.h>
45 #include <net/netns/generic.h>
47 static struct rtnl_link_ops vti_link_ops __read_mostly;
49 static int vti_net_id __read_mostly;
50 static int vti_tunnel_init(struct net_device *dev);
52 static int vti_input(struct sk_buff *skb, int nexthdr, __be32 spi,
53 int encap_type, bool update_skb_dev)
55 struct ip_tunnel *tunnel;
56 const struct iphdr *iph = ip_hdr(skb);
57 struct net *net = dev_net(skb->dev);
58 struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
60 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
61 iph->saddr, iph->daddr, 0);
63 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
66 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4 = tunnel;
69 skb->dev = tunnel->dev;
71 return xfrm_input(skb, nexthdr, spi, encap_type);
80 static int vti_input_proto(struct sk_buff *skb, int nexthdr, __be32 spi,
83 return vti_input(skb, nexthdr, spi, encap_type, false);
86 static int vti_rcv(struct sk_buff *skb, __be32 spi, bool update_skb_dev)
88 XFRM_SPI_SKB_CB(skb)->family = AF_INET;
89 XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct iphdr, daddr);
91 return vti_input(skb, ip_hdr(skb)->protocol, spi, 0, update_skb_dev);
94 static int vti_rcv_proto(struct sk_buff *skb)
96 return vti_rcv(skb, 0, false);
99 static int vti_rcv_tunnel(struct sk_buff *skb)
101 struct ip_tunnel_net *itn = net_generic(dev_net(skb->dev), vti_net_id);
102 const struct iphdr *iph = ip_hdr(skb);
103 struct ip_tunnel *tunnel;
105 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
106 iph->saddr, iph->daddr, 0);
108 struct tnl_ptk_info tpi = {
109 .proto = htons(ETH_P_IP),
112 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
114 if (iptunnel_pull_header(skb, 0, tpi.proto, false))
116 return ip_tunnel_rcv(tunnel, skb, &tpi, NULL, false);
125 static int vti_rcv_cb(struct sk_buff *skb, int err)
127 unsigned short family;
128 struct net_device *dev;
129 struct pcpu_sw_netstats *tstats;
130 struct xfrm_state *x;
131 struct xfrm_mode *inner_mode;
132 struct ip_tunnel *tunnel = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4;
133 u32 orig_mark = skb->mark;
142 dev->stats.rx_errors++;
143 dev->stats.rx_dropped++;
148 x = xfrm_input_state(skb);
150 inner_mode = x->inner_mode;
152 if (x->sel.family == AF_UNSPEC) {
153 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
154 if (inner_mode == NULL) {
155 XFRM_INC_STATS(dev_net(skb->dev),
156 LINUX_MIB_XFRMINSTATEMODEERROR);
161 family = inner_mode->afinfo->family;
163 skb->mark = be32_to_cpu(tunnel->parms.i_key);
164 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
165 skb->mark = orig_mark;
170 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(skb->dev)));
173 tstats = this_cpu_ptr(dev->tstats);
175 u64_stats_update_begin(&tstats->syncp);
176 tstats->rx_packets++;
177 tstats->rx_bytes += skb->len;
178 u64_stats_update_end(&tstats->syncp);
183 static bool vti_state_check(const struct xfrm_state *x, __be32 dst, __be32 src)
185 xfrm_address_t *daddr = (xfrm_address_t *)&dst;
186 xfrm_address_t *saddr = (xfrm_address_t *)&src;
188 /* if there is no transform then this tunnel is not functional.
189 * Or if the xfrm is not mode tunnel.
191 if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
192 x->props.family != AF_INET)
196 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET);
198 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET))
204 static netdev_tx_t vti_xmit(struct sk_buff *skb, struct net_device *dev,
207 struct ip_tunnel *tunnel = netdev_priv(dev);
208 struct ip_tunnel_parm *parms = &tunnel->parms;
209 struct dst_entry *dst = skb_dst(skb);
210 struct net_device *tdev; /* Device to other host */
211 int pkt_len = skb->len;
216 switch (skb->protocol) {
217 case htons(ETH_P_IP): {
220 fl->u.ip4.flowi4_oif = dev->ifindex;
221 fl->u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC;
222 rt = __ip_route_output_key(dev_net(dev), &fl->u.ip4);
224 dev->stats.tx_carrier_errors++;
228 skb_dst_set(skb, dst);
231 #if IS_ENABLED(CONFIG_IPV6)
232 case htons(ETH_P_IPV6):
233 fl->u.ip6.flowi6_oif = dev->ifindex;
234 fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
235 dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6);
239 dev->stats.tx_carrier_errors++;
242 skb_dst_set(skb, dst);
246 dev->stats.tx_carrier_errors++;
252 dst = xfrm_lookup(tunnel->net, dst, fl, NULL, 0);
254 dev->stats.tx_carrier_errors++;
258 if (!vti_state_check(dst->xfrm, parms->iph.daddr, parms->iph.saddr)) {
259 dev->stats.tx_carrier_errors++;
268 dev->stats.collisions++;
272 if (tunnel->err_count > 0) {
273 if (time_before(jiffies,
274 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
276 dst_link_failure(skb);
278 tunnel->err_count = 0;
282 if (skb->len > mtu) {
283 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
284 if (skb->protocol == htons(ETH_P_IP)) {
285 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
288 if (mtu < IPV6_MIN_MTU)
291 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
298 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(dev)));
299 skb_dst_set(skb, dst);
300 skb->dev = skb_dst(skb)->dev;
302 err = dst_output(tunnel->net, skb->sk, skb);
303 if (net_xmit_eval(err) == 0)
305 iptunnel_xmit_stats(dev, err);
309 dst_link_failure(skb);
311 dev->stats.tx_errors++;
316 /* This function assumes it is being called from dev_queue_xmit()
317 * and that skb is filled properly by that function.
319 static netdev_tx_t vti_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
321 struct ip_tunnel *tunnel = netdev_priv(dev);
324 memset(&fl, 0, sizeof(fl));
326 switch (skb->protocol) {
327 case htons(ETH_P_IP):
328 xfrm_decode_session(skb, &fl, AF_INET);
329 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
331 case htons(ETH_P_IPV6):
332 xfrm_decode_session(skb, &fl, AF_INET6);
333 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
336 dev->stats.tx_errors++;
341 /* override mark with tunnel output key */
342 fl.flowi_mark = be32_to_cpu(tunnel->parms.o_key);
344 return vti_xmit(skb, dev, &fl);
347 static int vti4_err(struct sk_buff *skb, u32 info)
351 struct xfrm_state *x;
352 struct ip_tunnel *tunnel;
353 struct ip_esp_hdr *esph;
354 struct ip_auth_hdr *ah ;
355 struct ip_comp_hdr *ipch;
356 struct net *net = dev_net(skb->dev);
357 const struct iphdr *iph = (const struct iphdr *)skb->data;
358 int protocol = iph->protocol;
359 struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
361 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
362 iph->daddr, iph->saddr, 0);
366 mark = be32_to_cpu(tunnel->parms.o_key);
370 esph = (struct ip_esp_hdr *)(skb->data+(iph->ihl<<2));
374 ah = (struct ip_auth_hdr *)(skb->data+(iph->ihl<<2));
378 ipch = (struct ip_comp_hdr *)(skb->data+(iph->ihl<<2));
379 spi = htonl(ntohs(ipch->cpi));
385 switch (icmp_hdr(skb)->type) {
386 case ICMP_DEST_UNREACH:
387 if (icmp_hdr(skb)->code != ICMP_FRAG_NEEDED)
395 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
396 spi, protocol, AF_INET);
400 if (icmp_hdr(skb)->type == ICMP_DEST_UNREACH)
401 ipv4_update_pmtu(skb, net, info, 0, 0, protocol, 0);
403 ipv4_redirect(skb, net, 0, 0, protocol, 0);
410 vti_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
413 struct ip_tunnel_parm p;
415 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
418 if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
419 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPIP ||
424 if (!(p.i_flags & GRE_KEY))
426 if (!(p.o_flags & GRE_KEY))
429 p.i_flags = VTI_ISVTI;
431 err = ip_tunnel_ioctl(dev, &p, cmd);
435 if (cmd != SIOCDELTUNNEL) {
436 p.i_flags |= GRE_KEY;
437 p.o_flags |= GRE_KEY;
440 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
445 static const struct net_device_ops vti_netdev_ops = {
446 .ndo_init = vti_tunnel_init,
447 .ndo_uninit = ip_tunnel_uninit,
448 .ndo_start_xmit = vti_tunnel_xmit,
449 .ndo_do_ioctl = vti_tunnel_ioctl,
450 .ndo_change_mtu = ip_tunnel_change_mtu,
451 .ndo_get_stats64 = ip_tunnel_get_stats64,
452 .ndo_get_iflink = ip_tunnel_get_iflink,
455 static void vti_tunnel_setup(struct net_device *dev)
457 dev->netdev_ops = &vti_netdev_ops;
458 dev->type = ARPHRD_TUNNEL;
459 ip_tunnel_setup(dev, vti_net_id);
462 static int vti_tunnel_init(struct net_device *dev)
464 struct ip_tunnel *tunnel = netdev_priv(dev);
465 struct iphdr *iph = &tunnel->parms.iph;
467 memcpy(dev->dev_addr, &iph->saddr, 4);
468 memcpy(dev->broadcast, &iph->daddr, 4);
470 dev->mtu = ETH_DATA_LEN;
471 dev->flags = IFF_NOARP;
473 dev->features |= NETIF_F_LLTX;
476 return ip_tunnel_init(dev);
479 static void __net_init vti_fb_tunnel_init(struct net_device *dev)
481 struct ip_tunnel *tunnel = netdev_priv(dev);
482 struct iphdr *iph = &tunnel->parms.iph;
485 iph->protocol = IPPROTO_IPIP;
489 static struct xfrm4_protocol vti_esp4_protocol __read_mostly = {
490 .handler = vti_rcv_proto,
491 .input_handler = vti_input_proto,
492 .cb_handler = vti_rcv_cb,
493 .err_handler = vti4_err,
497 static struct xfrm4_protocol vti_ah4_protocol __read_mostly = {
498 .handler = vti_rcv_proto,
499 .input_handler = vti_input_proto,
500 .cb_handler = vti_rcv_cb,
501 .err_handler = vti4_err,
505 static struct xfrm4_protocol vti_ipcomp4_protocol __read_mostly = {
506 .handler = vti_rcv_proto,
507 .input_handler = vti_input_proto,
508 .cb_handler = vti_rcv_cb,
509 .err_handler = vti4_err,
513 static struct xfrm_tunnel ipip_handler __read_mostly = {
514 .handler = vti_rcv_tunnel,
515 .err_handler = vti4_err,
519 static int __net_init vti_init_net(struct net *net)
522 struct ip_tunnel_net *itn;
524 err = ip_tunnel_init_net(net, vti_net_id, &vti_link_ops, "ip_vti0");
527 itn = net_generic(net, vti_net_id);
528 vti_fb_tunnel_init(itn->fb_tunnel_dev);
532 static void __net_exit vti_exit_net(struct net *net)
534 struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
535 ip_tunnel_delete_net(itn, &vti_link_ops);
538 static struct pernet_operations vti_net_ops = {
539 .init = vti_init_net,
540 .exit = vti_exit_net,
542 .size = sizeof(struct ip_tunnel_net),
545 static int vti_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
550 static void vti_netlink_parms(struct nlattr *data[],
551 struct ip_tunnel_parm *parms)
553 memset(parms, 0, sizeof(*parms));
555 parms->iph.protocol = IPPROTO_IPIP;
560 parms->i_flags = VTI_ISVTI;
562 if (data[IFLA_VTI_LINK])
563 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
565 if (data[IFLA_VTI_IKEY])
566 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
568 if (data[IFLA_VTI_OKEY])
569 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
571 if (data[IFLA_VTI_LOCAL])
572 parms->iph.saddr = nla_get_in_addr(data[IFLA_VTI_LOCAL]);
574 if (data[IFLA_VTI_REMOTE])
575 parms->iph.daddr = nla_get_in_addr(data[IFLA_VTI_REMOTE]);
579 static int vti_newlink(struct net *src_net, struct net_device *dev,
580 struct nlattr *tb[], struct nlattr *data[])
582 struct ip_tunnel_parm parms;
584 vti_netlink_parms(data, &parms);
585 return ip_tunnel_newlink(dev, tb, &parms);
588 static int vti_changelink(struct net_device *dev, struct nlattr *tb[],
589 struct nlattr *data[])
591 struct ip_tunnel_parm p;
593 vti_netlink_parms(data, &p);
594 return ip_tunnel_changelink(dev, tb, &p);
597 static size_t vti_get_size(const struct net_device *dev)
608 /* IFLA_VTI_REMOTE */
613 static int vti_fill_info(struct sk_buff *skb, const struct net_device *dev)
615 struct ip_tunnel *t = netdev_priv(dev);
616 struct ip_tunnel_parm *p = &t->parms;
618 nla_put_u32(skb, IFLA_VTI_LINK, p->link);
619 nla_put_be32(skb, IFLA_VTI_IKEY, p->i_key);
620 nla_put_be32(skb, IFLA_VTI_OKEY, p->o_key);
621 nla_put_in_addr(skb, IFLA_VTI_LOCAL, p->iph.saddr);
622 nla_put_in_addr(skb, IFLA_VTI_REMOTE, p->iph.daddr);
627 static const struct nla_policy vti_policy[IFLA_VTI_MAX + 1] = {
628 [IFLA_VTI_LINK] = { .type = NLA_U32 },
629 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
630 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
631 [IFLA_VTI_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
632 [IFLA_VTI_REMOTE] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
635 static struct rtnl_link_ops vti_link_ops __read_mostly = {
637 .maxtype = IFLA_VTI_MAX,
638 .policy = vti_policy,
639 .priv_size = sizeof(struct ip_tunnel),
640 .setup = vti_tunnel_setup,
641 .validate = vti_tunnel_validate,
642 .newlink = vti_newlink,
643 .changelink = vti_changelink,
644 .dellink = ip_tunnel_dellink,
645 .get_size = vti_get_size,
646 .fill_info = vti_fill_info,
647 .get_link_net = ip_tunnel_get_link_net,
650 static bool is_vti_tunnel(const struct net_device *dev)
652 return dev->netdev_ops == &vti_netdev_ops;
655 static int vti_device_event(struct notifier_block *unused,
656 unsigned long event, void *ptr)
658 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
659 struct ip_tunnel *tunnel = netdev_priv(dev);
661 if (!is_vti_tunnel(dev))
666 if (!net_eq(tunnel->net, dev_net(dev)))
667 xfrm_garbage_collect(tunnel->net);
673 static struct notifier_block vti_notifier_block __read_mostly = {
674 .notifier_call = vti_device_event,
677 static int __init vti_init(void)
682 pr_info("IPv4 over IPsec tunneling driver\n");
684 register_netdevice_notifier(&vti_notifier_block);
686 msg = "tunnel device";
687 err = register_pernet_device(&vti_net_ops);
689 goto pernet_dev_failed;
691 msg = "tunnel protocols";
692 err = xfrm4_protocol_register(&vti_esp4_protocol, IPPROTO_ESP);
694 goto xfrm_proto_esp_failed;
695 err = xfrm4_protocol_register(&vti_ah4_protocol, IPPROTO_AH);
697 goto xfrm_proto_ah_failed;
698 err = xfrm4_protocol_register(&vti_ipcomp4_protocol, IPPROTO_COMP);
700 goto xfrm_proto_comp_failed;
703 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
705 goto xfrm_tunnel_failed;
707 msg = "netlink interface";
708 err = rtnl_link_register(&vti_link_ops);
710 goto rtnl_link_failed;
715 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
717 xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP);
718 xfrm_proto_comp_failed:
719 xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
720 xfrm_proto_ah_failed:
721 xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
722 xfrm_proto_esp_failed:
723 unregister_pernet_device(&vti_net_ops);
725 unregister_netdevice_notifier(&vti_notifier_block);
726 pr_err("vti init: failed to register %s\n", msg);
730 static void __exit vti_fini(void)
732 rtnl_link_unregister(&vti_link_ops);
733 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
734 xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP);
735 xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
736 xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
737 unregister_pernet_device(&vti_net_ops);
738 unregister_netdevice_notifier(&vti_notifier_block);
741 module_init(vti_init);
742 module_exit(vti_fini);
743 MODULE_LICENSE("GPL");
744 MODULE_ALIAS_RTNL_LINK("vti");
745 MODULE_ALIAS_NETDEV("ip_vti0");