1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * Linux NET3: IP/IP protocol decoder modified to support
4 * virtual tunnel interface
11 This version of net/ipv4/ip_vti.c is cloned of net/ipv4/ipip.c
13 For comments look at net/ipv4/ip_gre.c --ANK
17 #include <linux/capability.h>
18 #include <linux/module.h>
19 #include <linux/types.h>
20 #include <linux/kernel.h>
21 #include <linux/uaccess.h>
22 #include <linux/skbuff.h>
23 #include <linux/netdevice.h>
25 #include <linux/tcp.h>
26 #include <linux/udp.h>
27 #include <linux/if_arp.h>
28 #include <linux/init.h>
29 #include <linux/netfilter_ipv4.h>
30 #include <linux/if_ether.h>
31 #include <linux/icmpv6.h>
36 #include <net/ip_tunnels.h>
37 #include <net/inet_ecn.h>
39 #include <net/net_namespace.h>
40 #include <net/netns/generic.h>
42 static struct rtnl_link_ops vti_link_ops __read_mostly;
44 static unsigned int vti_net_id __read_mostly;
45 static int vti_tunnel_init(struct net_device *dev);
47 static int vti_input(struct sk_buff *skb, int nexthdr, __be32 spi,
48 int encap_type, bool update_skb_dev)
50 struct ip_tunnel *tunnel;
51 const struct iphdr *iph = ip_hdr(skb);
52 struct net *net = dev_net(skb->dev);
53 struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
55 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
56 iph->saddr, iph->daddr, 0);
58 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
61 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4 = tunnel;
64 skb->dev = tunnel->dev;
66 return xfrm_input(skb, nexthdr, spi, encap_type);
75 static int vti_input_proto(struct sk_buff *skb, int nexthdr, __be32 spi,
78 return vti_input(skb, nexthdr, spi, encap_type, false);
81 static int vti_rcv(struct sk_buff *skb, __be32 spi, bool update_skb_dev)
83 XFRM_SPI_SKB_CB(skb)->family = AF_INET;
84 XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct iphdr, daddr);
86 return vti_input(skb, ip_hdr(skb)->protocol, spi, 0, update_skb_dev);
89 static int vti_rcv_proto(struct sk_buff *skb)
91 return vti_rcv(skb, 0, false);
94 static int vti_rcv_tunnel(struct sk_buff *skb)
96 return vti_rcv(skb, ip_hdr(skb)->saddr, true);
99 static int vti_rcv_cb(struct sk_buff *skb, int err)
101 unsigned short family;
102 struct net_device *dev;
103 struct pcpu_sw_netstats *tstats;
104 struct xfrm_state *x;
105 const struct xfrm_mode *inner_mode;
106 struct ip_tunnel *tunnel = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4;
107 u32 orig_mark = skb->mark;
116 dev->stats.rx_errors++;
117 dev->stats.rx_dropped++;
122 x = xfrm_input_state(skb);
124 inner_mode = &x->inner_mode;
126 if (x->sel.family == AF_UNSPEC) {
127 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
128 if (inner_mode == NULL) {
129 XFRM_INC_STATS(dev_net(skb->dev),
130 LINUX_MIB_XFRMINSTATEMODEERROR);
135 family = inner_mode->family;
137 skb->mark = be32_to_cpu(tunnel->parms.i_key);
138 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
139 skb->mark = orig_mark;
144 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(skb->dev)));
147 tstats = this_cpu_ptr(dev->tstats);
149 u64_stats_update_begin(&tstats->syncp);
150 tstats->rx_packets++;
151 tstats->rx_bytes += skb->len;
152 u64_stats_update_end(&tstats->syncp);
157 static bool vti_state_check(const struct xfrm_state *x, __be32 dst, __be32 src)
159 xfrm_address_t *daddr = (xfrm_address_t *)&dst;
160 xfrm_address_t *saddr = (xfrm_address_t *)&src;
162 /* if there is no transform then this tunnel is not functional.
163 * Or if the xfrm is not mode tunnel.
165 if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
166 x->props.family != AF_INET)
170 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET);
172 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET))
178 static netdev_tx_t vti_xmit(struct sk_buff *skb, struct net_device *dev,
181 struct ip_tunnel *tunnel = netdev_priv(dev);
182 struct ip_tunnel_parm *parms = &tunnel->parms;
183 struct dst_entry *dst = skb_dst(skb);
184 struct net_device *tdev; /* Device to other host */
185 int pkt_len = skb->len;
190 switch (skb->protocol) {
191 case htons(ETH_P_IP): {
194 fl->u.ip4.flowi4_oif = dev->ifindex;
195 fl->u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC;
196 rt = __ip_route_output_key(dev_net(dev), &fl->u.ip4);
198 dev->stats.tx_carrier_errors++;
202 skb_dst_set(skb, dst);
205 #if IS_ENABLED(CONFIG_IPV6)
206 case htons(ETH_P_IPV6):
207 fl->u.ip6.flowi6_oif = dev->ifindex;
208 fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
209 dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6);
213 dev->stats.tx_carrier_errors++;
216 skb_dst_set(skb, dst);
220 dev->stats.tx_carrier_errors++;
226 dst = xfrm_lookup(tunnel->net, dst, fl, NULL, 0);
228 dev->stats.tx_carrier_errors++;
232 if (!vti_state_check(dst->xfrm, parms->iph.daddr, parms->iph.saddr)) {
233 dev->stats.tx_carrier_errors++;
242 dev->stats.collisions++;
247 if (skb->len > mtu) {
248 skb_dst_update_pmtu_no_confirm(skb, mtu);
249 if (skb->protocol == htons(ETH_P_IP)) {
250 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
253 if (mtu < IPV6_MIN_MTU)
256 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
263 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(dev)));
264 skb_dst_set(skb, dst);
265 skb->dev = skb_dst(skb)->dev;
267 err = dst_output(tunnel->net, skb->sk, skb);
268 if (net_xmit_eval(err) == 0)
270 iptunnel_xmit_stats(dev, err);
274 dst_link_failure(skb);
276 dev->stats.tx_errors++;
281 /* This function assumes it is being called from dev_queue_xmit()
282 * and that skb is filled properly by that function.
284 static netdev_tx_t vti_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
286 struct ip_tunnel *tunnel = netdev_priv(dev);
289 if (!pskb_inet_may_pull(skb))
292 memset(&fl, 0, sizeof(fl));
294 switch (skb->protocol) {
295 case htons(ETH_P_IP):
296 xfrm_decode_session(skb, &fl, AF_INET);
297 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
299 case htons(ETH_P_IPV6):
300 xfrm_decode_session(skb, &fl, AF_INET6);
301 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
307 /* override mark with tunnel output key */
308 fl.flowi_mark = be32_to_cpu(tunnel->parms.o_key);
310 return vti_xmit(skb, dev, &fl);
313 dev->stats.tx_errors++;
318 static int vti4_err(struct sk_buff *skb, u32 info)
322 struct xfrm_state *x;
323 struct ip_tunnel *tunnel;
324 struct ip_esp_hdr *esph;
325 struct ip_auth_hdr *ah ;
326 struct ip_comp_hdr *ipch;
327 struct net *net = dev_net(skb->dev);
328 const struct iphdr *iph = (const struct iphdr *)skb->data;
329 int protocol = iph->protocol;
330 struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
332 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
333 iph->daddr, iph->saddr, 0);
337 mark = be32_to_cpu(tunnel->parms.o_key);
341 esph = (struct ip_esp_hdr *)(skb->data+(iph->ihl<<2));
345 ah = (struct ip_auth_hdr *)(skb->data+(iph->ihl<<2));
349 ipch = (struct ip_comp_hdr *)(skb->data+(iph->ihl<<2));
350 spi = htonl(ntohs(ipch->cpi));
356 switch (icmp_hdr(skb)->type) {
357 case ICMP_DEST_UNREACH:
358 if (icmp_hdr(skb)->code != ICMP_FRAG_NEEDED)
366 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
367 spi, protocol, AF_INET);
371 if (icmp_hdr(skb)->type == ICMP_DEST_UNREACH)
372 ipv4_update_pmtu(skb, net, info, 0, protocol);
374 ipv4_redirect(skb, net, 0, protocol);
381 vti_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
384 struct ip_tunnel_parm p;
386 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
389 if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
390 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPIP ||
395 if (!(p.i_flags & GRE_KEY))
397 if (!(p.o_flags & GRE_KEY))
400 p.i_flags = VTI_ISVTI;
402 err = ip_tunnel_ioctl(dev, &p, cmd);
406 if (cmd != SIOCDELTUNNEL) {
407 p.i_flags |= GRE_KEY;
408 p.o_flags |= GRE_KEY;
411 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
416 static const struct net_device_ops vti_netdev_ops = {
417 .ndo_init = vti_tunnel_init,
418 .ndo_uninit = ip_tunnel_uninit,
419 .ndo_start_xmit = vti_tunnel_xmit,
420 .ndo_do_ioctl = vti_tunnel_ioctl,
421 .ndo_change_mtu = ip_tunnel_change_mtu,
422 .ndo_get_stats64 = ip_tunnel_get_stats64,
423 .ndo_get_iflink = ip_tunnel_get_iflink,
426 static void vti_tunnel_setup(struct net_device *dev)
428 dev->netdev_ops = &vti_netdev_ops;
429 dev->type = ARPHRD_TUNNEL;
430 ip_tunnel_setup(dev, vti_net_id);
433 static int vti_tunnel_init(struct net_device *dev)
435 struct ip_tunnel *tunnel = netdev_priv(dev);
436 struct iphdr *iph = &tunnel->parms.iph;
438 memcpy(dev->dev_addr, &iph->saddr, 4);
439 memcpy(dev->broadcast, &iph->daddr, 4);
441 dev->flags = IFF_NOARP;
443 dev->features |= NETIF_F_LLTX;
446 return ip_tunnel_init(dev);
449 static void __net_init vti_fb_tunnel_init(struct net_device *dev)
451 struct ip_tunnel *tunnel = netdev_priv(dev);
452 struct iphdr *iph = &tunnel->parms.iph;
455 iph->protocol = IPPROTO_IPIP;
459 static struct xfrm4_protocol vti_esp4_protocol __read_mostly = {
460 .handler = vti_rcv_proto,
461 .input_handler = vti_input_proto,
462 .cb_handler = vti_rcv_cb,
463 .err_handler = vti4_err,
467 static struct xfrm4_protocol vti_ah4_protocol __read_mostly = {
468 .handler = vti_rcv_proto,
469 .input_handler = vti_input_proto,
470 .cb_handler = vti_rcv_cb,
471 .err_handler = vti4_err,
475 static struct xfrm4_protocol vti_ipcomp4_protocol __read_mostly = {
476 .handler = vti_rcv_proto,
477 .input_handler = vti_input_proto,
478 .cb_handler = vti_rcv_cb,
479 .err_handler = vti4_err,
483 static struct xfrm_tunnel ipip_handler __read_mostly = {
484 .handler = vti_rcv_tunnel,
485 .err_handler = vti4_err,
489 static int __net_init vti_init_net(struct net *net)
492 struct ip_tunnel_net *itn;
494 err = ip_tunnel_init_net(net, vti_net_id, &vti_link_ops, "ip_vti0");
497 itn = net_generic(net, vti_net_id);
498 if (itn->fb_tunnel_dev)
499 vti_fb_tunnel_init(itn->fb_tunnel_dev);
503 static void __net_exit vti_exit_batch_net(struct list_head *list_net)
505 ip_tunnel_delete_nets(list_net, vti_net_id, &vti_link_ops);
508 static struct pernet_operations vti_net_ops = {
509 .init = vti_init_net,
510 .exit_batch = vti_exit_batch_net,
512 .size = sizeof(struct ip_tunnel_net),
515 static int vti_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
516 struct netlink_ext_ack *extack)
521 static void vti_netlink_parms(struct nlattr *data[],
522 struct ip_tunnel_parm *parms,
525 memset(parms, 0, sizeof(*parms));
527 parms->iph.protocol = IPPROTO_IPIP;
532 parms->i_flags = VTI_ISVTI;
534 if (data[IFLA_VTI_LINK])
535 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
537 if (data[IFLA_VTI_IKEY])
538 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
540 if (data[IFLA_VTI_OKEY])
541 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
543 if (data[IFLA_VTI_LOCAL])
544 parms->iph.saddr = nla_get_in_addr(data[IFLA_VTI_LOCAL]);
546 if (data[IFLA_VTI_REMOTE])
547 parms->iph.daddr = nla_get_in_addr(data[IFLA_VTI_REMOTE]);
549 if (data[IFLA_VTI_FWMARK])
550 *fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]);
553 static int vti_newlink(struct net *src_net, struct net_device *dev,
554 struct nlattr *tb[], struct nlattr *data[],
555 struct netlink_ext_ack *extack)
557 struct ip_tunnel_parm parms;
560 vti_netlink_parms(data, &parms, &fwmark);
561 return ip_tunnel_newlink(dev, tb, &parms, fwmark);
564 static int vti_changelink(struct net_device *dev, struct nlattr *tb[],
565 struct nlattr *data[],
566 struct netlink_ext_ack *extack)
568 struct ip_tunnel *t = netdev_priv(dev);
569 __u32 fwmark = t->fwmark;
570 struct ip_tunnel_parm p;
572 vti_netlink_parms(data, &p, &fwmark);
573 return ip_tunnel_changelink(dev, tb, &p, fwmark);
576 static size_t vti_get_size(const struct net_device *dev)
587 /* IFLA_VTI_REMOTE */
589 /* IFLA_VTI_FWMARK */
594 static int vti_fill_info(struct sk_buff *skb, const struct net_device *dev)
596 struct ip_tunnel *t = netdev_priv(dev);
597 struct ip_tunnel_parm *p = &t->parms;
599 if (nla_put_u32(skb, IFLA_VTI_LINK, p->link) ||
600 nla_put_be32(skb, IFLA_VTI_IKEY, p->i_key) ||
601 nla_put_be32(skb, IFLA_VTI_OKEY, p->o_key) ||
602 nla_put_in_addr(skb, IFLA_VTI_LOCAL, p->iph.saddr) ||
603 nla_put_in_addr(skb, IFLA_VTI_REMOTE, p->iph.daddr) ||
604 nla_put_u32(skb, IFLA_VTI_FWMARK, t->fwmark))
610 static const struct nla_policy vti_policy[IFLA_VTI_MAX + 1] = {
611 [IFLA_VTI_LINK] = { .type = NLA_U32 },
612 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
613 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
614 [IFLA_VTI_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) },
615 [IFLA_VTI_REMOTE] = { .len = sizeof_field(struct iphdr, daddr) },
616 [IFLA_VTI_FWMARK] = { .type = NLA_U32 },
619 static struct rtnl_link_ops vti_link_ops __read_mostly = {
621 .maxtype = IFLA_VTI_MAX,
622 .policy = vti_policy,
623 .priv_size = sizeof(struct ip_tunnel),
624 .setup = vti_tunnel_setup,
625 .validate = vti_tunnel_validate,
626 .newlink = vti_newlink,
627 .changelink = vti_changelink,
628 .dellink = ip_tunnel_dellink,
629 .get_size = vti_get_size,
630 .fill_info = vti_fill_info,
631 .get_link_net = ip_tunnel_get_link_net,
634 static int __init vti_init(void)
639 pr_info("IPv4 over IPsec tunneling driver\n");
641 msg = "tunnel device";
642 err = register_pernet_device(&vti_net_ops);
644 goto pernet_dev_failed;
646 msg = "tunnel protocols";
647 err = xfrm4_protocol_register(&vti_esp4_protocol, IPPROTO_ESP);
649 goto xfrm_proto_esp_failed;
650 err = xfrm4_protocol_register(&vti_ah4_protocol, IPPROTO_AH);
652 goto xfrm_proto_ah_failed;
653 err = xfrm4_protocol_register(&vti_ipcomp4_protocol, IPPROTO_COMP);
655 goto xfrm_proto_comp_failed;
658 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
660 goto xfrm_tunnel_failed;
662 msg = "netlink interface";
663 err = rtnl_link_register(&vti_link_ops);
665 goto rtnl_link_failed;
670 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
672 xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP);
673 xfrm_proto_comp_failed:
674 xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
675 xfrm_proto_ah_failed:
676 xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
677 xfrm_proto_esp_failed:
678 unregister_pernet_device(&vti_net_ops);
680 pr_err("vti init: failed to register %s\n", msg);
684 static void __exit vti_fini(void)
686 rtnl_link_unregister(&vti_link_ops);
687 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
688 xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP);
689 xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
690 xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
691 unregister_pernet_device(&vti_net_ops);
694 module_init(vti_init);
695 module_exit(vti_fini);
696 MODULE_LICENSE("GPL");
697 MODULE_ALIAS_RTNL_LINK("vti");
698 MODULE_ALIAS_NETDEV("ip_vti0");