2 * Linux NET3: IP/IP protocol decoder.
8 * Alan Cox : Merged and made usable non modular (its so tiny its silly as
9 * a module taking up 2 pages).
10 * Alan Cox : Fixed bug with 1.3.18 and IPIP not working (now needs to set skb->h.iph)
11 * to keep ip_forward happy.
12 * Alan Cox : More fixes for 1.3.21, and firewall fix. Maybe this will work soon 8).
13 * Kai Schulte : Fixed #defines for IP_FIREWALL->FIREWALL
14 * David Woodhouse : Perform some basic ICMP handling.
15 * IPIP Routing without decapsulation.
16 * Carlos Picoto : GRE over IP support
17 * Alexey Kuznetsov: Reworked. Really, now it is truncated version of ipv4/ip_gre.c.
18 * I do not want to merge them together.
20 * This program is free software; you can redistribute it and/or
21 * modify it under the terms of the GNU General Public License
22 * as published by the Free Software Foundation; either version
23 * 2 of the License, or (at your option) any later version.
27 /* tunnel.c: an IP tunnel driver
29 The purpose of this driver is to provide an IP tunnel through
30 which you can tunnel network traffic transparently across subnets.
32 This was written by looking at Nick Holloway's dummy driver
33 Thanks for the great code!
38 Cleaned up the code a little and added some pre-1.3.0 tweaks.
39 dev->hard_header/hard_header_len changed to use no headers.
40 Comments/bracketing tweaked.
41 Made the tunnels use dev->name not tunnel: when error reporting.
47 Changed to tunnel to destination gateway in addition to the
48 tunnel's pointopoint address
49 Almost completely rewritten
50 Note: There is currently no firewall or ICMP handling done.
56 /* Things I wish I had known when writing the tunnel driver:
58 When the tunnel_xmit() function is called, the skb contains the
59 packet to be sent (plus a great deal of extra info), and dev
60 contains the tunnel device that _we_ are.
62 When we are passed a packet, we are expected to fill in the
63 source address with our source IP address.
65 What is the proper way to allocate, copy and free a buffer?
66 After you allocate it, it is a "0 length" chunk of memory
67 starting at zero. If you want to add headers to the buffer
68 later, you'll have to call "skb_reserve(skb, amount)" with
69 the amount of memory you want reserved. Then, you call
70 "skb_put(skb, amount)" with the amount of space you want in
71 the buffer. skb_put() returns a pointer to the top (#0) of
72 that buffer. skb->len is set to the amount of space you have
73 "allocated" with skb_put(). You can then write up to skb->len
74 bytes to that buffer. If you need more, you can call skb_put()
75 again with the additional amount of space you need. You can
76 find out how much more space you can allocate by calling
78 Now, to add header space, call "skb_push(skb, header_len)".
79 This creates space at the beginning of the buffer and returns
80 a pointer to this new space. If later you need to strip a
81 header from a buffer, call "skb_pull(skb, header_len)".
82 skb_headroom() will return how much space is left at the top
83 of the buffer (before the main data). Remember, this headroom
84 space must be reserved before the skb_put() function is called.
88 This version of net/ipv4/ipip.c is cloned of net/ipv4/ip_gre.c
90 For comments look at net/ipv4/ip_gre.c --ANK
94 #include <linux/capability.h>
95 #include <linux/module.h>
96 #include <linux/types.h>
97 #include <linux/kernel.h>
98 #include <linux/slab.h>
99 #include <linux/uaccess.h>
100 #include <linux/skbuff.h>
101 #include <linux/netdevice.h>
102 #include <linux/in.h>
103 #include <linux/tcp.h>
104 #include <linux/udp.h>
105 #include <linux/if_arp.h>
106 #include <linux/init.h>
107 #include <linux/netfilter_ipv4.h>
108 #include <linux/if_ether.h>
110 #include <net/sock.h>
112 #include <net/icmp.h>
113 #include <net/ip_tunnels.h>
114 #include <net/inet_ecn.h>
115 #include <net/xfrm.h>
116 #include <net/net_namespace.h>
117 #include <net/netns/generic.h>
118 #include <net/dst_metadata.h>
120 static bool log_ecn_error = true;
121 module_param(log_ecn_error, bool, 0644);
122 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
124 static unsigned int ipip_net_id __read_mostly;
126 static int ipip_tunnel_init(struct net_device *dev);
127 static struct rtnl_link_ops ipip_link_ops __read_mostly;
129 static int ipip_err(struct sk_buff *skb, u32 info)
132 /* All the routers (except for Linux) return only
133 8 bytes of packet payload. It means, that precise relaying of
134 ICMP in the real Internet is absolutely infeasible.
136 struct net *net = dev_net(skb->dev);
137 struct ip_tunnel_net *itn = net_generic(net, ipip_net_id);
138 const struct iphdr *iph = (const struct iphdr *)skb->data;
141 const int type = icmp_hdr(skb)->type;
142 const int code = icmp_hdr(skb)->code;
145 t = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
146 iph->daddr, iph->saddr, 0);
150 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
151 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
152 t->parms.link, 0, iph->protocol, 0);
157 if (type == ICMP_REDIRECT) {
158 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 0,
164 if (t->parms.iph.daddr == 0)
168 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
171 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
175 t->err_time = jiffies;
181 static const struct tnl_ptk_info ipip_tpi = {
182 /* no tunnel info required for ipip. */
183 .proto = htons(ETH_P_IP),
186 #if IS_ENABLED(CONFIG_MPLS)
187 static const struct tnl_ptk_info mplsip_tpi = {
188 /* no tunnel info required for mplsip. */
189 .proto = htons(ETH_P_MPLS_UC),
193 static int ipip_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
195 struct net *net = dev_net(skb->dev);
196 struct ip_tunnel_net *itn = net_generic(net, ipip_net_id);
197 struct metadata_dst *tun_dst = NULL;
198 struct ip_tunnel *tunnel;
199 const struct iphdr *iph;
202 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
203 iph->saddr, iph->daddr, 0);
205 const struct tnl_ptk_info *tpi;
207 if (tunnel->parms.iph.protocol != ipproto &&
208 tunnel->parms.iph.protocol != 0)
211 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
213 #if IS_ENABLED(CONFIG_MPLS)
214 if (ipproto == IPPROTO_MPLS)
219 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
221 if (tunnel->collect_md) {
222 tun_dst = ip_tun_rx_dst(skb, 0, 0, 0);
226 return ip_tunnel_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
236 static int ipip_rcv(struct sk_buff *skb)
238 return ipip_tunnel_rcv(skb, IPPROTO_IPIP);
241 #if IS_ENABLED(CONFIG_MPLS)
242 static int mplsip_rcv(struct sk_buff *skb)
244 return ipip_tunnel_rcv(skb, IPPROTO_MPLS);
249 * This function assumes it is being called from dev_queue_xmit()
250 * and that skb is filled properly by that function.
252 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb,
253 struct net_device *dev)
255 struct ip_tunnel *tunnel = netdev_priv(dev);
256 const struct iphdr *tiph = &tunnel->parms.iph;
259 switch (skb->protocol) {
260 case htons(ETH_P_IP):
261 ipproto = IPPROTO_IPIP;
263 #if IS_ENABLED(CONFIG_MPLS)
264 case htons(ETH_P_MPLS_UC):
265 ipproto = IPPROTO_MPLS;
272 if (tiph->protocol != ipproto && tiph->protocol != 0)
275 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
278 skb_set_inner_ipproto(skb, ipproto);
280 if (tunnel->collect_md)
281 ip_md_tunnel_xmit(skb, dev, ipproto);
283 ip_tunnel_xmit(skb, dev, tiph, ipproto);
289 dev->stats.tx_errors++;
293 static bool ipip_tunnel_ioctl_verify_protocol(u8 ipproto)
298 #if IS_ENABLED(CONFIG_MPLS)
308 ipip_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
311 struct ip_tunnel_parm p;
313 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
316 if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
317 if (p.iph.version != 4 ||
318 !ipip_tunnel_ioctl_verify_protocol(p.iph.protocol) ||
319 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
323 p.i_key = p.o_key = 0;
324 p.i_flags = p.o_flags = 0;
325 err = ip_tunnel_ioctl(dev, &p, cmd);
329 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
335 static const struct net_device_ops ipip_netdev_ops = {
336 .ndo_init = ipip_tunnel_init,
337 .ndo_uninit = ip_tunnel_uninit,
338 .ndo_start_xmit = ipip_tunnel_xmit,
339 .ndo_do_ioctl = ipip_tunnel_ioctl,
340 .ndo_change_mtu = ip_tunnel_change_mtu,
341 .ndo_get_stats64 = ip_tunnel_get_stats64,
342 .ndo_get_iflink = ip_tunnel_get_iflink,
345 #define IPIP_FEATURES (NETIF_F_SG | \
348 NETIF_F_GSO_SOFTWARE | \
351 static void ipip_tunnel_setup(struct net_device *dev)
353 dev->netdev_ops = &ipip_netdev_ops;
355 dev->type = ARPHRD_TUNNEL;
356 dev->flags = IFF_NOARP;
358 dev->features |= NETIF_F_LLTX;
361 dev->features |= IPIP_FEATURES;
362 dev->hw_features |= IPIP_FEATURES;
363 ip_tunnel_setup(dev, ipip_net_id);
366 static int ipip_tunnel_init(struct net_device *dev)
368 struct ip_tunnel *tunnel = netdev_priv(dev);
370 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
371 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
373 tunnel->tun_hlen = 0;
374 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
375 return ip_tunnel_init(dev);
378 static int ipip_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
382 if (!data || !data[IFLA_IPTUN_PROTO])
385 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
386 if (proto != IPPROTO_IPIP && proto != IPPROTO_MPLS && proto != 0)
392 static void ipip_netlink_parms(struct nlattr *data[],
393 struct ip_tunnel_parm *parms, bool *collect_md,
396 memset(parms, 0, sizeof(*parms));
398 parms->iph.version = 4;
399 parms->iph.protocol = IPPROTO_IPIP;
406 if (data[IFLA_IPTUN_LINK])
407 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
409 if (data[IFLA_IPTUN_LOCAL])
410 parms->iph.saddr = nla_get_in_addr(data[IFLA_IPTUN_LOCAL]);
412 if (data[IFLA_IPTUN_REMOTE])
413 parms->iph.daddr = nla_get_in_addr(data[IFLA_IPTUN_REMOTE]);
415 if (data[IFLA_IPTUN_TTL]) {
416 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
418 parms->iph.frag_off = htons(IP_DF);
421 if (data[IFLA_IPTUN_TOS])
422 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
424 if (data[IFLA_IPTUN_PROTO])
425 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
427 if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
428 parms->iph.frag_off = htons(IP_DF);
430 if (data[IFLA_IPTUN_COLLECT_METADATA])
433 if (data[IFLA_IPTUN_FWMARK])
434 *fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
437 /* This function returns true when ENCAP attributes are present in the nl msg */
438 static bool ipip_netlink_encap_parms(struct nlattr *data[],
439 struct ip_tunnel_encap *ipencap)
443 memset(ipencap, 0, sizeof(*ipencap));
448 if (data[IFLA_IPTUN_ENCAP_TYPE]) {
450 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
453 if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
455 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
458 if (data[IFLA_IPTUN_ENCAP_SPORT]) {
460 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
463 if (data[IFLA_IPTUN_ENCAP_DPORT]) {
465 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
471 static int ipip_newlink(struct net *src_net, struct net_device *dev,
472 struct nlattr *tb[], struct nlattr *data[])
474 struct ip_tunnel *t = netdev_priv(dev);
475 struct ip_tunnel_parm p;
476 struct ip_tunnel_encap ipencap;
479 if (ipip_netlink_encap_parms(data, &ipencap)) {
480 int err = ip_tunnel_encap_setup(t, &ipencap);
486 ipip_netlink_parms(data, &p, &t->collect_md, &fwmark);
487 return ip_tunnel_newlink(dev, tb, &p, fwmark);
490 static int ipip_changelink(struct net_device *dev, struct nlattr *tb[],
491 struct nlattr *data[])
493 struct ip_tunnel *t = netdev_priv(dev);
494 struct ip_tunnel_parm p;
495 struct ip_tunnel_encap ipencap;
497 __u32 fwmark = t->fwmark;
499 if (ipip_netlink_encap_parms(data, &ipencap)) {
500 int err = ip_tunnel_encap_setup(t, &ipencap);
506 ipip_netlink_parms(data, &p, &collect_md, &fwmark);
510 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
511 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
514 return ip_tunnel_changelink(dev, tb, &p, fwmark);
517 static size_t ipip_get_size(const struct net_device *dev)
520 /* IFLA_IPTUN_LINK */
522 /* IFLA_IPTUN_LOCAL */
524 /* IFLA_IPTUN_REMOTE */
530 /* IFLA_IPTUN_PROTO */
532 /* IFLA_IPTUN_PMTUDISC */
534 /* IFLA_IPTUN_ENCAP_TYPE */
536 /* IFLA_IPTUN_ENCAP_FLAGS */
538 /* IFLA_IPTUN_ENCAP_SPORT */
540 /* IFLA_IPTUN_ENCAP_DPORT */
542 /* IFLA_IPTUN_COLLECT_METADATA */
544 /* IFLA_IPTUN_FWMARK */
549 static int ipip_fill_info(struct sk_buff *skb, const struct net_device *dev)
551 struct ip_tunnel *tunnel = netdev_priv(dev);
552 struct ip_tunnel_parm *parm = &tunnel->parms;
554 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
555 nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
556 nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
557 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
558 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
559 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
560 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
561 !!(parm->iph.frag_off & htons(IP_DF))) ||
562 nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
563 goto nla_put_failure;
565 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
566 tunnel->encap.type) ||
567 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
568 tunnel->encap.sport) ||
569 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
570 tunnel->encap.dport) ||
571 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
572 tunnel->encap.flags))
573 goto nla_put_failure;
575 if (tunnel->collect_md)
576 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
577 goto nla_put_failure;
584 static const struct nla_policy ipip_policy[IFLA_IPTUN_MAX + 1] = {
585 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
586 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 },
587 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 },
588 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
589 [IFLA_IPTUN_TOS] = { .type = NLA_U8 },
590 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
591 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 },
592 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
593 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
594 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
595 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
596 [IFLA_IPTUN_COLLECT_METADATA] = { .type = NLA_FLAG },
597 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 },
600 static struct rtnl_link_ops ipip_link_ops __read_mostly = {
602 .maxtype = IFLA_IPTUN_MAX,
603 .policy = ipip_policy,
604 .priv_size = sizeof(struct ip_tunnel),
605 .setup = ipip_tunnel_setup,
606 .validate = ipip_tunnel_validate,
607 .newlink = ipip_newlink,
608 .changelink = ipip_changelink,
609 .dellink = ip_tunnel_dellink,
610 .get_size = ipip_get_size,
611 .fill_info = ipip_fill_info,
612 .get_link_net = ip_tunnel_get_link_net,
615 static struct xfrm_tunnel ipip_handler __read_mostly = {
617 .err_handler = ipip_err,
621 #if IS_ENABLED(CONFIG_MPLS)
622 static struct xfrm_tunnel mplsip_handler __read_mostly = {
623 .handler = mplsip_rcv,
624 .err_handler = ipip_err,
629 static int __net_init ipip_init_net(struct net *net)
631 return ip_tunnel_init_net(net, ipip_net_id, &ipip_link_ops, "tunl0");
634 static void __net_exit ipip_exit_net(struct net *net)
636 struct ip_tunnel_net *itn = net_generic(net, ipip_net_id);
637 ip_tunnel_delete_net(itn, &ipip_link_ops);
640 static struct pernet_operations ipip_net_ops = {
641 .init = ipip_init_net,
642 .exit = ipip_exit_net,
644 .size = sizeof(struct ip_tunnel_net),
647 static int __init ipip_init(void)
651 pr_info("ipip: IPv4 and MPLS over IPv4 tunneling driver\n");
653 err = register_pernet_device(&ipip_net_ops);
656 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
658 pr_info("%s: can't register tunnel\n", __func__);
659 goto xfrm_tunnel_ipip_failed;
661 #if IS_ENABLED(CONFIG_MPLS)
662 err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
664 pr_info("%s: can't register tunnel\n", __func__);
665 goto xfrm_tunnel_mplsip_failed;
668 err = rtnl_link_register(&ipip_link_ops);
670 goto rtnl_link_failed;
676 #if IS_ENABLED(CONFIG_MPLS)
677 xfrm4_tunnel_deregister(&mplsip_handler, AF_INET);
678 xfrm_tunnel_mplsip_failed:
681 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
682 xfrm_tunnel_ipip_failed:
683 unregister_pernet_device(&ipip_net_ops);
687 static void __exit ipip_fini(void)
689 rtnl_link_unregister(&ipip_link_ops);
690 if (xfrm4_tunnel_deregister(&ipip_handler, AF_INET))
691 pr_info("%s: can't deregister tunnel\n", __func__);
692 #if IS_ENABLED(CONFIG_MPLS)
693 if (xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS))
694 pr_info("%s: can't deregister tunnel\n", __func__);
696 unregister_pernet_device(&ipip_net_ops);
699 module_init(ipip_init);
700 module_exit(ipip_fini);
701 MODULE_LICENSE("GPL");
702 MODULE_ALIAS_RTNL_LINK("ipip");
703 MODULE_ALIAS_NETDEV("tunl0");