1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * Linux NET3: IP/IP protocol decoder.
6 * Sam Lantinga (slouken@cs.ucdavis.edu) 02/01/95
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!
30 -Sam Lantinga (slouken@cs.ucdavis.edu) 02/01/95
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.
39 -Alan Cox (alan@lxorguk.ukuu.org.uk) 21 March 95
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.
47 -Sam Lantinga (slouken@cs.ucdavis.edu) 02/13/96
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 skb_reset_mac_header(skb);
247 return ip_tunnel_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
257 static int ipip_rcv(struct sk_buff *skb)
259 return ipip_tunnel_rcv(skb, IPPROTO_IPIP);
262 #if IS_ENABLED(CONFIG_MPLS)
263 static int mplsip_rcv(struct sk_buff *skb)
265 return ipip_tunnel_rcv(skb, IPPROTO_MPLS);
270 * This function assumes it is being called from dev_queue_xmit()
271 * and that skb is filled properly by that function.
273 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb,
274 struct net_device *dev)
276 struct ip_tunnel *tunnel = netdev_priv(dev);
277 const struct iphdr *tiph = &tunnel->parms.iph;
280 if (!pskb_inet_may_pull(skb))
283 switch (skb->protocol) {
284 case htons(ETH_P_IP):
285 ipproto = IPPROTO_IPIP;
287 #if IS_ENABLED(CONFIG_MPLS)
288 case htons(ETH_P_MPLS_UC):
289 ipproto = IPPROTO_MPLS;
296 if (tiph->protocol != ipproto && tiph->protocol != 0)
299 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
302 skb_set_inner_ipproto(skb, ipproto);
304 if (tunnel->collect_md)
305 ip_md_tunnel_xmit(skb, dev, ipproto, 0);
307 ip_tunnel_xmit(skb, dev, tiph, ipproto);
313 DEV_STATS_INC(dev, tx_errors);
317 static bool ipip_tunnel_ioctl_verify_protocol(u8 ipproto)
322 #if IS_ENABLED(CONFIG_MPLS)
332 ipip_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd)
334 if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
335 if (p->iph.version != 4 ||
336 !ipip_tunnel_ioctl_verify_protocol(p->iph.protocol) ||
337 p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)))
341 p->i_key = p->o_key = 0;
342 p->i_flags = p->o_flags = 0;
343 return ip_tunnel_ctl(dev, p, cmd);
346 static const struct net_device_ops ipip_netdev_ops = {
347 .ndo_init = ipip_tunnel_init,
348 .ndo_uninit = ip_tunnel_uninit,
349 .ndo_start_xmit = ipip_tunnel_xmit,
350 .ndo_siocdevprivate = ip_tunnel_siocdevprivate,
351 .ndo_change_mtu = ip_tunnel_change_mtu,
352 .ndo_get_stats64 = dev_get_tstats64,
353 .ndo_get_iflink = ip_tunnel_get_iflink,
354 .ndo_tunnel_ctl = ipip_tunnel_ctl,
357 #define IPIP_FEATURES (NETIF_F_SG | \
360 NETIF_F_GSO_SOFTWARE | \
363 static void ipip_tunnel_setup(struct net_device *dev)
365 dev->netdev_ops = &ipip_netdev_ops;
366 dev->header_ops = &ip_tunnel_header_ops;
368 dev->type = ARPHRD_TUNNEL;
369 dev->flags = IFF_NOARP;
371 dev->features |= NETIF_F_LLTX;
374 dev->features |= IPIP_FEATURES;
375 dev->hw_features |= IPIP_FEATURES;
376 ip_tunnel_setup(dev, ipip_net_id);
379 static int ipip_tunnel_init(struct net_device *dev)
381 struct ip_tunnel *tunnel = netdev_priv(dev);
383 __dev_addr_set(dev, &tunnel->parms.iph.saddr, 4);
384 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
386 tunnel->tun_hlen = 0;
387 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
388 return ip_tunnel_init(dev);
391 static int ipip_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
392 struct netlink_ext_ack *extack)
396 if (!data || !data[IFLA_IPTUN_PROTO])
399 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
400 if (proto != IPPROTO_IPIP && proto != IPPROTO_MPLS && proto != 0)
406 static void ipip_netlink_parms(struct nlattr *data[],
407 struct ip_tunnel_parm *parms, bool *collect_md,
410 memset(parms, 0, sizeof(*parms));
412 parms->iph.version = 4;
413 parms->iph.protocol = IPPROTO_IPIP;
420 ip_tunnel_netlink_parms(data, parms);
422 if (data[IFLA_IPTUN_COLLECT_METADATA])
425 if (data[IFLA_IPTUN_FWMARK])
426 *fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
429 static int ipip_newlink(struct net *src_net, struct net_device *dev,
430 struct nlattr *tb[], struct nlattr *data[],
431 struct netlink_ext_ack *extack)
433 struct ip_tunnel *t = netdev_priv(dev);
434 struct ip_tunnel_parm p;
435 struct ip_tunnel_encap ipencap;
438 if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
439 int err = ip_tunnel_encap_setup(t, &ipencap);
445 ipip_netlink_parms(data, &p, &t->collect_md, &fwmark);
446 return ip_tunnel_newlink(dev, tb, &p, fwmark);
449 static int ipip_changelink(struct net_device *dev, struct nlattr *tb[],
450 struct nlattr *data[],
451 struct netlink_ext_ack *extack)
453 struct ip_tunnel *t = netdev_priv(dev);
454 struct ip_tunnel_parm p;
455 struct ip_tunnel_encap ipencap;
457 __u32 fwmark = t->fwmark;
459 if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
460 int err = ip_tunnel_encap_setup(t, &ipencap);
466 ipip_netlink_parms(data, &p, &collect_md, &fwmark);
470 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
471 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
474 return ip_tunnel_changelink(dev, tb, &p, fwmark);
477 static size_t ipip_get_size(const struct net_device *dev)
480 /* IFLA_IPTUN_LINK */
482 /* IFLA_IPTUN_LOCAL */
484 /* IFLA_IPTUN_REMOTE */
490 /* IFLA_IPTUN_PROTO */
492 /* IFLA_IPTUN_PMTUDISC */
494 /* IFLA_IPTUN_ENCAP_TYPE */
496 /* IFLA_IPTUN_ENCAP_FLAGS */
498 /* IFLA_IPTUN_ENCAP_SPORT */
500 /* IFLA_IPTUN_ENCAP_DPORT */
502 /* IFLA_IPTUN_COLLECT_METADATA */
504 /* IFLA_IPTUN_FWMARK */
509 static int ipip_fill_info(struct sk_buff *skb, const struct net_device *dev)
511 struct ip_tunnel *tunnel = netdev_priv(dev);
512 struct ip_tunnel_parm *parm = &tunnel->parms;
514 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
515 nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
516 nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
517 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
518 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
519 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
520 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
521 !!(parm->iph.frag_off & htons(IP_DF))) ||
522 nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
523 goto nla_put_failure;
525 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
526 tunnel->encap.type) ||
527 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
528 tunnel->encap.sport) ||
529 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
530 tunnel->encap.dport) ||
531 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
532 tunnel->encap.flags))
533 goto nla_put_failure;
535 if (tunnel->collect_md)
536 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
537 goto nla_put_failure;
544 static const struct nla_policy ipip_policy[IFLA_IPTUN_MAX + 1] = {
545 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
546 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 },
547 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 },
548 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
549 [IFLA_IPTUN_TOS] = { .type = NLA_U8 },
550 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
551 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 },
552 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
553 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
554 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
555 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
556 [IFLA_IPTUN_COLLECT_METADATA] = { .type = NLA_FLAG },
557 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 },
560 static struct rtnl_link_ops ipip_link_ops __read_mostly = {
562 .maxtype = IFLA_IPTUN_MAX,
563 .policy = ipip_policy,
564 .priv_size = sizeof(struct ip_tunnel),
565 .setup = ipip_tunnel_setup,
566 .validate = ipip_tunnel_validate,
567 .newlink = ipip_newlink,
568 .changelink = ipip_changelink,
569 .dellink = ip_tunnel_dellink,
570 .get_size = ipip_get_size,
571 .fill_info = ipip_fill_info,
572 .get_link_net = ip_tunnel_get_link_net,
575 static struct xfrm_tunnel ipip_handler __read_mostly = {
577 .err_handler = ipip_err,
581 #if IS_ENABLED(CONFIG_MPLS)
582 static struct xfrm_tunnel mplsip_handler __read_mostly = {
583 .handler = mplsip_rcv,
584 .err_handler = ipip_err,
589 static int __net_init ipip_init_net(struct net *net)
591 return ip_tunnel_init_net(net, ipip_net_id, &ipip_link_ops, "tunl0");
594 static void __net_exit ipip_exit_batch_net(struct list_head *list_net)
596 ip_tunnel_delete_nets(list_net, ipip_net_id, &ipip_link_ops);
599 static struct pernet_operations ipip_net_ops = {
600 .init = ipip_init_net,
601 .exit_batch = ipip_exit_batch_net,
603 .size = sizeof(struct ip_tunnel_net),
606 static int __init ipip_init(void)
610 pr_info("ipip: IPv4 and MPLS over IPv4 tunneling driver\n");
612 err = register_pernet_device(&ipip_net_ops);
615 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
617 pr_info("%s: can't register tunnel\n", __func__);
618 goto xfrm_tunnel_ipip_failed;
620 #if IS_ENABLED(CONFIG_MPLS)
621 err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
623 pr_info("%s: can't register tunnel\n", __func__);
624 goto xfrm_tunnel_mplsip_failed;
627 err = rtnl_link_register(&ipip_link_ops);
629 goto rtnl_link_failed;
635 #if IS_ENABLED(CONFIG_MPLS)
636 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
637 xfrm_tunnel_mplsip_failed:
640 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
641 xfrm_tunnel_ipip_failed:
642 unregister_pernet_device(&ipip_net_ops);
646 static void __exit ipip_fini(void)
648 rtnl_link_unregister(&ipip_link_ops);
649 if (xfrm4_tunnel_deregister(&ipip_handler, AF_INET))
650 pr_info("%s: can't deregister tunnel\n", __func__);
651 #if IS_ENABLED(CONFIG_MPLS)
652 if (xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS))
653 pr_info("%s: can't deregister tunnel\n", __func__);
655 unregister_pernet_device(&ipip_net_ops);
658 module_init(ipip_init);
659 module_exit(ipip_fini);
660 MODULE_LICENSE("GPL");
661 MODULE_ALIAS_RTNL_LINK("ipip");
662 MODULE_ALIAS_NETDEV("tunl0");