1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * Linux NET3: IP/IP protocol decoder.
9 * Alan Cox : Merged and made usable non modular (its so tiny its silly as
10 * a module taking up 2 pages).
11 * Alan Cox : Fixed bug with 1.3.18 and IPIP not working (now needs to set skb->h.iph)
12 * to keep ip_forward happy.
13 * Alan Cox : More fixes for 1.3.21, and firewall fix. Maybe this will work soon 8).
14 * Kai Schulte : Fixed #defines for IP_FIREWALL->FIREWALL
15 * David Woodhouse : Perform some basic ICMP handling.
16 * IPIP Routing without decapsulation.
17 * Carlos Picoto : GRE over IP support
18 * Alexey Kuznetsov: Reworked. Really, now it is truncated version of ipv4/ip_gre.c.
19 * I do not want to merge them together.
22 /* tunnel.c: an IP tunnel driver
24 The purpose of this driver is to provide an IP tunnel through
25 which you can tunnel network traffic transparently across subnets.
27 This was written by looking at Nick Holloway's dummy driver
28 Thanks for the great code!
33 Cleaned up the code a little and added some pre-1.3.0 tweaks.
34 dev->hard_header/hard_header_len changed to use no headers.
35 Comments/bracketing tweaked.
36 Made the tunnels use dev->name not tunnel: when error reporting.
42 Changed to tunnel to destination gateway in addition to the
43 tunnel's pointopoint address
44 Almost completely rewritten
45 Note: There is currently no firewall or ICMP handling done.
51 /* Things I wish I had known when writing the tunnel driver:
53 When the tunnel_xmit() function is called, the skb contains the
54 packet to be sent (plus a great deal of extra info), and dev
55 contains the tunnel device that _we_ are.
57 When we are passed a packet, we are expected to fill in the
58 source address with our source IP address.
60 What is the proper way to allocate, copy and free a buffer?
61 After you allocate it, it is a "0 length" chunk of memory
62 starting at zero. If you want to add headers to the buffer
63 later, you'll have to call "skb_reserve(skb, amount)" with
64 the amount of memory you want reserved. Then, you call
65 "skb_put(skb, amount)" with the amount of space you want in
66 the buffer. skb_put() returns a pointer to the top (#0) of
67 that buffer. skb->len is set to the amount of space you have
68 "allocated" with skb_put(). You can then write up to skb->len
69 bytes to that buffer. If you need more, you can call skb_put()
70 again with the additional amount of space you need. You can
71 find out how much more space you can allocate by calling
73 Now, to add header space, call "skb_push(skb, header_len)".
74 This creates space at the beginning of the buffer and returns
75 a pointer to this new space. If later you need to strip a
76 header from a buffer, call "skb_pull(skb, header_len)".
77 skb_headroom() will return how much space is left at the top
78 of the buffer (before the main data). Remember, this headroom
79 space must be reserved before the skb_put() function is called.
83 This version of net/ipv4/ipip.c is cloned of net/ipv4/ip_gre.c
85 For comments look at net/ipv4/ip_gre.c --ANK
89 #include <linux/capability.h>
90 #include <linux/module.h>
91 #include <linux/types.h>
92 #include <linux/kernel.h>
93 #include <linux/slab.h>
94 #include <linux/uaccess.h>
95 #include <linux/skbuff.h>
96 #include <linux/netdevice.h>
98 #include <linux/tcp.h>
99 #include <linux/udp.h>
100 #include <linux/if_arp.h>
101 #include <linux/init.h>
102 #include <linux/netfilter_ipv4.h>
103 #include <linux/if_ether.h>
105 #include <net/sock.h>
107 #include <net/icmp.h>
108 #include <net/ip_tunnels.h>
109 #include <net/inet_ecn.h>
110 #include <net/xfrm.h>
111 #include <net/net_namespace.h>
112 #include <net/netns/generic.h>
113 #include <net/dst_metadata.h>
115 static bool log_ecn_error = true;
116 module_param(log_ecn_error, bool, 0644);
117 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
119 static unsigned int ipip_net_id __read_mostly;
121 static int ipip_tunnel_init(struct net_device *dev);
122 static struct rtnl_link_ops ipip_link_ops __read_mostly;
124 static int ipip_err(struct sk_buff *skb, u32 info)
126 /* All the routers (except for Linux) return only
127 * 8 bytes of packet payload. It means, that precise relaying of
128 * ICMP in the real Internet is absolutely infeasible.
130 struct net *net = dev_net(skb->dev);
131 struct ip_tunnel_net *itn = net_generic(net, ipip_net_id);
132 const struct iphdr *iph = (const struct iphdr *)skb->data;
133 const int type = icmp_hdr(skb)->type;
134 const int code = icmp_hdr(skb)->code;
138 t = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
139 iph->daddr, iph->saddr, 0);
146 case ICMP_DEST_UNREACH:
149 /* Impossible event. */
152 /* All others are translated to HOST_UNREACH.
153 * rfc2003 contains "deep thoughts" about NET_UNREACH,
154 * I believe they are just ether pollution. --ANK
160 case ICMP_TIME_EXCEEDED:
161 if (code != ICMP_EXC_TTL)
172 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
173 ipv4_update_pmtu(skb, net, info, t->parms.link, iph->protocol);
177 if (type == ICMP_REDIRECT) {
178 ipv4_redirect(skb, net, t->parms.link, iph->protocol);
182 if (t->parms.iph.daddr == 0) {
187 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
190 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
194 t->err_time = jiffies;
200 static const struct tnl_ptk_info ipip_tpi = {
201 /* no tunnel info required for ipip. */
202 .proto = htons(ETH_P_IP),
205 #if IS_ENABLED(CONFIG_MPLS)
206 static const struct tnl_ptk_info mplsip_tpi = {
207 /* no tunnel info required for mplsip. */
208 .proto = htons(ETH_P_MPLS_UC),
212 static int ipip_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
214 struct net *net = dev_net(skb->dev);
215 struct ip_tunnel_net *itn = net_generic(net, ipip_net_id);
216 struct metadata_dst *tun_dst = NULL;
217 struct ip_tunnel *tunnel;
218 const struct iphdr *iph;
221 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
222 iph->saddr, iph->daddr, 0);
224 const struct tnl_ptk_info *tpi;
226 if (tunnel->parms.iph.protocol != ipproto &&
227 tunnel->parms.iph.protocol != 0)
230 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
232 #if IS_ENABLED(CONFIG_MPLS)
233 if (ipproto == IPPROTO_MPLS)
238 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
240 if (tunnel->collect_md) {
241 tun_dst = ip_tun_rx_dst(skb, 0, 0, 0);
245 return ip_tunnel_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
255 static int ipip_rcv(struct sk_buff *skb)
257 return ipip_tunnel_rcv(skb, IPPROTO_IPIP);
260 #if IS_ENABLED(CONFIG_MPLS)
261 static int mplsip_rcv(struct sk_buff *skb)
263 return ipip_tunnel_rcv(skb, IPPROTO_MPLS);
268 * This function assumes it is being called from dev_queue_xmit()
269 * and that skb is filled properly by that function.
271 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb,
272 struct net_device *dev)
274 struct ip_tunnel *tunnel = netdev_priv(dev);
275 const struct iphdr *tiph = &tunnel->parms.iph;
278 if (!pskb_inet_may_pull(skb))
281 switch (skb->protocol) {
282 case htons(ETH_P_IP):
283 ipproto = IPPROTO_IPIP;
285 #if IS_ENABLED(CONFIG_MPLS)
286 case htons(ETH_P_MPLS_UC):
287 ipproto = IPPROTO_MPLS;
294 if (tiph->protocol != ipproto && tiph->protocol != 0)
297 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
300 skb_set_inner_ipproto(skb, ipproto);
302 if (tunnel->collect_md)
303 ip_md_tunnel_xmit(skb, dev, ipproto, 0);
305 ip_tunnel_xmit(skb, dev, tiph, ipproto);
311 dev->stats.tx_errors++;
315 static bool ipip_tunnel_ioctl_verify_protocol(u8 ipproto)
320 #if IS_ENABLED(CONFIG_MPLS)
330 ipip_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd)
332 if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
333 if (p->iph.version != 4 ||
334 !ipip_tunnel_ioctl_verify_protocol(p->iph.protocol) ||
335 p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)))
339 p->i_key = p->o_key = 0;
340 p->i_flags = p->o_flags = 0;
341 return ip_tunnel_ctl(dev, p, cmd);
344 static const struct net_device_ops ipip_netdev_ops = {
345 .ndo_init = ipip_tunnel_init,
346 .ndo_uninit = ip_tunnel_uninit,
347 .ndo_start_xmit = ipip_tunnel_xmit,
348 .ndo_do_ioctl = ip_tunnel_ioctl,
349 .ndo_change_mtu = ip_tunnel_change_mtu,
350 .ndo_get_stats64 = ip_tunnel_get_stats64,
351 .ndo_get_iflink = ip_tunnel_get_iflink,
352 .ndo_tunnel_ctl = ipip_tunnel_ctl,
355 #define IPIP_FEATURES (NETIF_F_SG | \
358 NETIF_F_GSO_SOFTWARE | \
361 static void ipip_tunnel_setup(struct net_device *dev)
363 dev->netdev_ops = &ipip_netdev_ops;
364 dev->header_ops = &ip_tunnel_header_ops;
366 dev->type = ARPHRD_TUNNEL;
367 dev->flags = IFF_NOARP;
369 dev->features |= NETIF_F_LLTX;
372 dev->features |= IPIP_FEATURES;
373 dev->hw_features |= IPIP_FEATURES;
374 ip_tunnel_setup(dev, ipip_net_id);
377 static int ipip_tunnel_init(struct net_device *dev)
379 struct ip_tunnel *tunnel = netdev_priv(dev);
381 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
382 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
384 tunnel->tun_hlen = 0;
385 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
386 return ip_tunnel_init(dev);
389 static int ipip_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
390 struct netlink_ext_ack *extack)
394 if (!data || !data[IFLA_IPTUN_PROTO])
397 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
398 if (proto != IPPROTO_IPIP && proto != IPPROTO_MPLS && proto != 0)
404 static void ipip_netlink_parms(struct nlattr *data[],
405 struct ip_tunnel_parm *parms, bool *collect_md,
408 memset(parms, 0, sizeof(*parms));
410 parms->iph.version = 4;
411 parms->iph.protocol = IPPROTO_IPIP;
418 if (data[IFLA_IPTUN_LINK])
419 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
421 if (data[IFLA_IPTUN_LOCAL])
422 parms->iph.saddr = nla_get_in_addr(data[IFLA_IPTUN_LOCAL]);
424 if (data[IFLA_IPTUN_REMOTE])
425 parms->iph.daddr = nla_get_in_addr(data[IFLA_IPTUN_REMOTE]);
427 if (data[IFLA_IPTUN_TTL]) {
428 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
430 parms->iph.frag_off = htons(IP_DF);
433 if (data[IFLA_IPTUN_TOS])
434 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
436 if (data[IFLA_IPTUN_PROTO])
437 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
439 if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
440 parms->iph.frag_off = htons(IP_DF);
442 if (data[IFLA_IPTUN_COLLECT_METADATA])
445 if (data[IFLA_IPTUN_FWMARK])
446 *fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
449 /* This function returns true when ENCAP attributes are present in the nl msg */
450 static bool ipip_netlink_encap_parms(struct nlattr *data[],
451 struct ip_tunnel_encap *ipencap)
455 memset(ipencap, 0, sizeof(*ipencap));
460 if (data[IFLA_IPTUN_ENCAP_TYPE]) {
462 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
465 if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
467 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
470 if (data[IFLA_IPTUN_ENCAP_SPORT]) {
472 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
475 if (data[IFLA_IPTUN_ENCAP_DPORT]) {
477 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
483 static int ipip_newlink(struct net *src_net, struct net_device *dev,
484 struct nlattr *tb[], struct nlattr *data[],
485 struct netlink_ext_ack *extack)
487 struct ip_tunnel *t = netdev_priv(dev);
488 struct ip_tunnel_parm p;
489 struct ip_tunnel_encap ipencap;
492 if (ipip_netlink_encap_parms(data, &ipencap)) {
493 int err = ip_tunnel_encap_setup(t, &ipencap);
499 ipip_netlink_parms(data, &p, &t->collect_md, &fwmark);
500 return ip_tunnel_newlink(dev, tb, &p, fwmark);
503 static int ipip_changelink(struct net_device *dev, struct nlattr *tb[],
504 struct nlattr *data[],
505 struct netlink_ext_ack *extack)
507 struct ip_tunnel *t = netdev_priv(dev);
508 struct ip_tunnel_parm p;
509 struct ip_tunnel_encap ipencap;
511 __u32 fwmark = t->fwmark;
513 if (ipip_netlink_encap_parms(data, &ipencap)) {
514 int err = ip_tunnel_encap_setup(t, &ipencap);
520 ipip_netlink_parms(data, &p, &collect_md, &fwmark);
524 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
525 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
528 return ip_tunnel_changelink(dev, tb, &p, fwmark);
531 static size_t ipip_get_size(const struct net_device *dev)
534 /* IFLA_IPTUN_LINK */
536 /* IFLA_IPTUN_LOCAL */
538 /* IFLA_IPTUN_REMOTE */
544 /* IFLA_IPTUN_PROTO */
546 /* IFLA_IPTUN_PMTUDISC */
548 /* IFLA_IPTUN_ENCAP_TYPE */
550 /* IFLA_IPTUN_ENCAP_FLAGS */
552 /* IFLA_IPTUN_ENCAP_SPORT */
554 /* IFLA_IPTUN_ENCAP_DPORT */
556 /* IFLA_IPTUN_COLLECT_METADATA */
558 /* IFLA_IPTUN_FWMARK */
563 static int ipip_fill_info(struct sk_buff *skb, const struct net_device *dev)
565 struct ip_tunnel *tunnel = netdev_priv(dev);
566 struct ip_tunnel_parm *parm = &tunnel->parms;
568 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
569 nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
570 nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
571 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
572 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
573 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
574 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
575 !!(parm->iph.frag_off & htons(IP_DF))) ||
576 nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
577 goto nla_put_failure;
579 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
580 tunnel->encap.type) ||
581 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
582 tunnel->encap.sport) ||
583 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
584 tunnel->encap.dport) ||
585 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
586 tunnel->encap.flags))
587 goto nla_put_failure;
589 if (tunnel->collect_md)
590 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
591 goto nla_put_failure;
598 static const struct nla_policy ipip_policy[IFLA_IPTUN_MAX + 1] = {
599 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
600 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 },
601 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 },
602 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
603 [IFLA_IPTUN_TOS] = { .type = NLA_U8 },
604 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
605 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 },
606 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
607 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
608 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
609 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
610 [IFLA_IPTUN_COLLECT_METADATA] = { .type = NLA_FLAG },
611 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 },
614 static struct rtnl_link_ops ipip_link_ops __read_mostly = {
616 .maxtype = IFLA_IPTUN_MAX,
617 .policy = ipip_policy,
618 .priv_size = sizeof(struct ip_tunnel),
619 .setup = ipip_tunnel_setup,
620 .validate = ipip_tunnel_validate,
621 .newlink = ipip_newlink,
622 .changelink = ipip_changelink,
623 .dellink = ip_tunnel_dellink,
624 .get_size = ipip_get_size,
625 .fill_info = ipip_fill_info,
626 .get_link_net = ip_tunnel_get_link_net,
629 static struct xfrm_tunnel ipip_handler __read_mostly = {
631 .err_handler = ipip_err,
635 #if IS_ENABLED(CONFIG_MPLS)
636 static struct xfrm_tunnel mplsip_handler __read_mostly = {
637 .handler = mplsip_rcv,
638 .err_handler = ipip_err,
643 static int __net_init ipip_init_net(struct net *net)
645 return ip_tunnel_init_net(net, ipip_net_id, &ipip_link_ops, "tunl0");
648 static void __net_exit ipip_exit_batch_net(struct list_head *list_net)
650 ip_tunnel_delete_nets(list_net, ipip_net_id, &ipip_link_ops);
653 static struct pernet_operations ipip_net_ops = {
654 .init = ipip_init_net,
655 .exit_batch = ipip_exit_batch_net,
657 .size = sizeof(struct ip_tunnel_net),
660 static int __init ipip_init(void)
664 pr_info("ipip: IPv4 and MPLS over IPv4 tunneling driver\n");
666 err = register_pernet_device(&ipip_net_ops);
669 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
671 pr_info("%s: can't register tunnel\n", __func__);
672 goto xfrm_tunnel_ipip_failed;
674 #if IS_ENABLED(CONFIG_MPLS)
675 err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
677 pr_info("%s: can't register tunnel\n", __func__);
678 goto xfrm_tunnel_mplsip_failed;
681 err = rtnl_link_register(&ipip_link_ops);
683 goto rtnl_link_failed;
689 #if IS_ENABLED(CONFIG_MPLS)
690 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
691 xfrm_tunnel_mplsip_failed:
694 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
695 xfrm_tunnel_ipip_failed:
696 unregister_pernet_device(&ipip_net_ops);
700 static void __exit ipip_fini(void)
702 rtnl_link_unregister(&ipip_link_ops);
703 if (xfrm4_tunnel_deregister(&ipip_handler, AF_INET))
704 pr_info("%s: can't deregister tunnel\n", __func__);
705 #if IS_ENABLED(CONFIG_MPLS)
706 if (xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS))
707 pr_info("%s: can't deregister tunnel\n", __func__);
709 unregister_pernet_device(&ipip_net_ops);
712 module_init(ipip_init);
713 module_exit(ipip_fini);
714 MODULE_LICENSE("GPL");
715 MODULE_ALIAS_RTNL_LINK("ipip");
716 MODULE_ALIAS_NETDEV("tunl0");