2 * Linux NET3: IP/IP protocol decoder.
5 * Sam Lantinga (slouken@cs.ucdavis.edu) 02/01/95
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!
35 -Sam Lantinga (slouken@cs.ucdavis.edu) 02/01/95
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.
44 -Alan Cox (alan@lxorguk.ukuu.org.uk) 21 March 95
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.
52 -Sam Lantinga (slouken@cs.ucdavis.edu) 02/13/96
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 <asm/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/mroute.h>
107 #include <linux/init.h>
108 #include <linux/netfilter_ipv4.h>
109 #include <linux/if_ether.h>
111 #include <net/sock.h>
113 #include <net/icmp.h>
114 #include <net/ip_tunnels.h>
115 #include <net/inet_ecn.h>
116 #include <net/xfrm.h>
117 #include <net/net_namespace.h>
118 #include <net/netns/generic.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 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
;
139 const int type
= icmp_hdr(skb
)->type
;
140 const int code
= icmp_hdr(skb
)->code
;
145 case ICMP_DEST_UNREACH
:
148 /* Impossible event. */
151 /* All others are translated to HOST_UNREACH.
152 * rfc2003 contains "deep thoughts" about NET_UNREACH,
153 * I believe they are just ether pollution. --ANK
159 case ICMP_TIME_EXCEEDED
:
160 if (code
!= ICMP_EXC_TTL
)
171 t
= ip_tunnel_lookup(itn
, skb
->dev
->ifindex
, TUNNEL_NO_KEY
,
172 iph
->daddr
, iph
->saddr
, 0);
178 if (type
== ICMP_DEST_UNREACH
&& code
== ICMP_FRAG_NEEDED
) {
179 ipv4_update_pmtu(skb
, net
, info
, t
->parms
.link
, 0,
184 if (type
== ICMP_REDIRECT
) {
185 ipv4_redirect(skb
, net
, t
->parms
.link
, 0, iph
->protocol
, 0);
189 if (t
->parms
.iph
.daddr
== 0) {
194 if (t
->parms
.iph
.ttl
== 0 && type
== ICMP_TIME_EXCEEDED
)
197 if (time_before(jiffies
, t
->err_time
+ IPTUNNEL_ERR_TIMEO
))
201 t
->err_time
= jiffies
;
207 static const struct tnl_ptk_info tpi
= {
208 /* no tunnel info required for ipip. */
209 .proto
= htons(ETH_P_IP
),
212 static int ipip_rcv(struct sk_buff
*skb
)
214 struct net
*net
= dev_net(skb
->dev
);
215 struct ip_tunnel_net
*itn
= net_generic(net
, ipip_net_id
);
216 struct ip_tunnel
*tunnel
;
217 const struct iphdr
*iph
;
220 tunnel
= ip_tunnel_lookup(itn
, skb
->dev
->ifindex
, TUNNEL_NO_KEY
,
221 iph
->saddr
, iph
->daddr
, 0);
223 if (!xfrm4_policy_check(NULL
, XFRM_POLICY_IN
, skb
))
225 if (iptunnel_pull_header(skb
, 0, tpi
.proto
))
227 return ip_tunnel_rcv(tunnel
, skb
, &tpi
, NULL
, log_ecn_error
);
238 * This function assumes it is being called from dev_queue_xmit()
239 * and that skb is filled properly by that function.
241 static netdev_tx_t
ipip_tunnel_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
243 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
244 const struct iphdr
*tiph
= &tunnel
->parms
.iph
;
246 if (unlikely(skb
->protocol
!= htons(ETH_P_IP
)))
249 skb
= iptunnel_handle_offloads(skb
, false, SKB_GSO_IPIP
);
253 skb_set_inner_ipproto(skb
, IPPROTO_IPIP
);
255 ip_tunnel_xmit(skb
, dev
, tiph
, tiph
->protocol
);
261 dev
->stats
.tx_errors
++;
266 ipip_tunnel_ioctl(struct net_device
*dev
, struct ifreq
*ifr
, int cmd
)
269 struct ip_tunnel_parm p
;
271 if (copy_from_user(&p
, ifr
->ifr_ifru
.ifru_data
, sizeof(p
)))
274 if (cmd
== SIOCADDTUNNEL
|| cmd
== SIOCCHGTUNNEL
) {
275 if (p
.iph
.version
!= 4 || p
.iph
.protocol
!= IPPROTO_IPIP
||
276 p
.iph
.ihl
!= 5 || (p
.iph
.frag_off
&htons(~IP_DF
)))
280 p
.i_key
= p
.o_key
= 0;
281 p
.i_flags
= p
.o_flags
= 0;
282 err
= ip_tunnel_ioctl(dev
, &p
, cmd
);
286 if (copy_to_user(ifr
->ifr_ifru
.ifru_data
, &p
, sizeof(p
)))
292 static const struct net_device_ops ipip_netdev_ops
= {
293 .ndo_init
= ipip_tunnel_init
,
294 .ndo_uninit
= ip_tunnel_uninit
,
295 .ndo_start_xmit
= ipip_tunnel_xmit
,
296 .ndo_do_ioctl
= ipip_tunnel_ioctl
,
297 .ndo_change_mtu
= ip_tunnel_change_mtu
,
298 .ndo_get_stats64
= ip_tunnel_get_stats64
,
299 .ndo_get_iflink
= ip_tunnel_get_iflink
,
302 #define IPIP_FEATURES (NETIF_F_SG | \
305 NETIF_F_GSO_SOFTWARE | \
308 static void ipip_tunnel_setup(struct net_device
*dev
)
310 dev
->netdev_ops
= &ipip_netdev_ops
;
312 dev
->type
= ARPHRD_TUNNEL
;
313 dev
->flags
= IFF_NOARP
;
315 dev
->features
|= NETIF_F_LLTX
;
318 dev
->features
|= IPIP_FEATURES
;
319 dev
->hw_features
|= IPIP_FEATURES
;
320 ip_tunnel_setup(dev
, ipip_net_id
);
323 static int ipip_tunnel_init(struct net_device
*dev
)
325 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
327 memcpy(dev
->dev_addr
, &tunnel
->parms
.iph
.saddr
, 4);
328 memcpy(dev
->broadcast
, &tunnel
->parms
.iph
.daddr
, 4);
330 tunnel
->tun_hlen
= 0;
331 tunnel
->hlen
= tunnel
->tun_hlen
+ tunnel
->encap_hlen
;
332 tunnel
->parms
.iph
.protocol
= IPPROTO_IPIP
;
333 return ip_tunnel_init(dev
);
336 static void ipip_netlink_parms(struct nlattr
*data
[],
337 struct ip_tunnel_parm
*parms
)
339 memset(parms
, 0, sizeof(*parms
));
341 parms
->iph
.version
= 4;
342 parms
->iph
.protocol
= IPPROTO_IPIP
;
348 if (data
[IFLA_IPTUN_LINK
])
349 parms
->link
= nla_get_u32(data
[IFLA_IPTUN_LINK
]);
351 if (data
[IFLA_IPTUN_LOCAL
])
352 parms
->iph
.saddr
= nla_get_in_addr(data
[IFLA_IPTUN_LOCAL
]);
354 if (data
[IFLA_IPTUN_REMOTE
])
355 parms
->iph
.daddr
= nla_get_in_addr(data
[IFLA_IPTUN_REMOTE
]);
357 if (data
[IFLA_IPTUN_TTL
]) {
358 parms
->iph
.ttl
= nla_get_u8(data
[IFLA_IPTUN_TTL
]);
360 parms
->iph
.frag_off
= htons(IP_DF
);
363 if (data
[IFLA_IPTUN_TOS
])
364 parms
->iph
.tos
= nla_get_u8(data
[IFLA_IPTUN_TOS
]);
366 if (!data
[IFLA_IPTUN_PMTUDISC
] || nla_get_u8(data
[IFLA_IPTUN_PMTUDISC
]))
367 parms
->iph
.frag_off
= htons(IP_DF
);
370 /* This function returns true when ENCAP attributes are present in the nl msg */
371 static bool ipip_netlink_encap_parms(struct nlattr
*data
[],
372 struct ip_tunnel_encap
*ipencap
)
376 memset(ipencap
, 0, sizeof(*ipencap
));
381 if (data
[IFLA_IPTUN_ENCAP_TYPE
]) {
383 ipencap
->type
= nla_get_u16(data
[IFLA_IPTUN_ENCAP_TYPE
]);
386 if (data
[IFLA_IPTUN_ENCAP_FLAGS
]) {
388 ipencap
->flags
= nla_get_u16(data
[IFLA_IPTUN_ENCAP_FLAGS
]);
391 if (data
[IFLA_IPTUN_ENCAP_SPORT
]) {
393 ipencap
->sport
= nla_get_be16(data
[IFLA_IPTUN_ENCAP_SPORT
]);
396 if (data
[IFLA_IPTUN_ENCAP_DPORT
]) {
398 ipencap
->dport
= nla_get_be16(data
[IFLA_IPTUN_ENCAP_DPORT
]);
404 static int ipip_newlink(struct net
*src_net
, struct net_device
*dev
,
405 struct nlattr
*tb
[], struct nlattr
*data
[])
407 struct ip_tunnel_parm p
;
408 struct ip_tunnel_encap ipencap
;
410 if (ipip_netlink_encap_parms(data
, &ipencap
)) {
411 struct ip_tunnel
*t
= netdev_priv(dev
);
412 int err
= ip_tunnel_encap_setup(t
, &ipencap
);
418 ipip_netlink_parms(data
, &p
);
419 return ip_tunnel_newlink(dev
, tb
, &p
);
422 static int ipip_changelink(struct net_device
*dev
, struct nlattr
*tb
[],
423 struct nlattr
*data
[])
425 struct ip_tunnel_parm p
;
426 struct ip_tunnel_encap ipencap
;
428 if (ipip_netlink_encap_parms(data
, &ipencap
)) {
429 struct ip_tunnel
*t
= netdev_priv(dev
);
430 int err
= ip_tunnel_encap_setup(t
, &ipencap
);
436 ipip_netlink_parms(data
, &p
);
438 if (((dev
->flags
& IFF_POINTOPOINT
) && !p
.iph
.daddr
) ||
439 (!(dev
->flags
& IFF_POINTOPOINT
) && p
.iph
.daddr
))
442 return ip_tunnel_changelink(dev
, tb
, &p
);
445 static size_t ipip_get_size(const struct net_device
*dev
)
448 /* IFLA_IPTUN_LINK */
450 /* IFLA_IPTUN_LOCAL */
452 /* IFLA_IPTUN_REMOTE */
458 /* IFLA_IPTUN_PMTUDISC */
460 /* IFLA_IPTUN_ENCAP_TYPE */
462 /* IFLA_IPTUN_ENCAP_FLAGS */
464 /* IFLA_IPTUN_ENCAP_SPORT */
466 /* IFLA_IPTUN_ENCAP_DPORT */
471 static int ipip_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
473 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
474 struct ip_tunnel_parm
*parm
= &tunnel
->parms
;
476 if (nla_put_u32(skb
, IFLA_IPTUN_LINK
, parm
->link
) ||
477 nla_put_in_addr(skb
, IFLA_IPTUN_LOCAL
, parm
->iph
.saddr
) ||
478 nla_put_in_addr(skb
, IFLA_IPTUN_REMOTE
, parm
->iph
.daddr
) ||
479 nla_put_u8(skb
, IFLA_IPTUN_TTL
, parm
->iph
.ttl
) ||
480 nla_put_u8(skb
, IFLA_IPTUN_TOS
, parm
->iph
.tos
) ||
481 nla_put_u8(skb
, IFLA_IPTUN_PMTUDISC
,
482 !!(parm
->iph
.frag_off
& htons(IP_DF
))))
483 goto nla_put_failure
;
485 if (nla_put_u16(skb
, IFLA_IPTUN_ENCAP_TYPE
,
486 tunnel
->encap
.type
) ||
487 nla_put_be16(skb
, IFLA_IPTUN_ENCAP_SPORT
,
488 tunnel
->encap
.sport
) ||
489 nla_put_be16(skb
, IFLA_IPTUN_ENCAP_DPORT
,
490 tunnel
->encap
.dport
) ||
491 nla_put_u16(skb
, IFLA_IPTUN_ENCAP_FLAGS
,
492 tunnel
->encap
.flags
))
493 goto nla_put_failure
;
501 static const struct nla_policy ipip_policy
[IFLA_IPTUN_MAX
+ 1] = {
502 [IFLA_IPTUN_LINK
] = { .type
= NLA_U32
},
503 [IFLA_IPTUN_LOCAL
] = { .type
= NLA_U32
},
504 [IFLA_IPTUN_REMOTE
] = { .type
= NLA_U32
},
505 [IFLA_IPTUN_TTL
] = { .type
= NLA_U8
},
506 [IFLA_IPTUN_TOS
] = { .type
= NLA_U8
},
507 [IFLA_IPTUN_PMTUDISC
] = { .type
= NLA_U8
},
508 [IFLA_IPTUN_ENCAP_TYPE
] = { .type
= NLA_U16
},
509 [IFLA_IPTUN_ENCAP_FLAGS
] = { .type
= NLA_U16
},
510 [IFLA_IPTUN_ENCAP_SPORT
] = { .type
= NLA_U16
},
511 [IFLA_IPTUN_ENCAP_DPORT
] = { .type
= NLA_U16
},
514 static struct rtnl_link_ops ipip_link_ops __read_mostly
= {
516 .maxtype
= IFLA_IPTUN_MAX
,
517 .policy
= ipip_policy
,
518 .priv_size
= sizeof(struct ip_tunnel
),
519 .setup
= ipip_tunnel_setup
,
520 .newlink
= ipip_newlink
,
521 .changelink
= ipip_changelink
,
522 .dellink
= ip_tunnel_dellink
,
523 .get_size
= ipip_get_size
,
524 .fill_info
= ipip_fill_info
,
525 .get_link_net
= ip_tunnel_get_link_net
,
528 static struct xfrm_tunnel ipip_handler __read_mostly
= {
530 .err_handler
= ipip_err
,
534 static int __net_init
ipip_init_net(struct net
*net
)
536 return ip_tunnel_init_net(net
, ipip_net_id
, &ipip_link_ops
, "tunl0");
539 static void __net_exit
ipip_exit_net(struct net
*net
)
541 struct ip_tunnel_net
*itn
= net_generic(net
, ipip_net_id
);
542 ip_tunnel_delete_net(itn
, &ipip_link_ops
);
545 static struct pernet_operations ipip_net_ops
= {
546 .init
= ipip_init_net
,
547 .exit
= ipip_exit_net
,
549 .size
= sizeof(struct ip_tunnel_net
),
552 static int __init
ipip_init(void)
556 pr_info("ipip: IPv4 over IPv4 tunneling driver\n");
558 err
= register_pernet_device(&ipip_net_ops
);
561 err
= xfrm4_tunnel_register(&ipip_handler
, AF_INET
);
563 pr_info("%s: can't register tunnel\n", __func__
);
564 goto xfrm_tunnel_failed
;
566 err
= rtnl_link_register(&ipip_link_ops
);
568 goto rtnl_link_failed
;
574 xfrm4_tunnel_deregister(&ipip_handler
, AF_INET
);
576 unregister_pernet_device(&ipip_net_ops
);
580 static void __exit
ipip_fini(void)
582 rtnl_link_unregister(&ipip_link_ops
);
583 if (xfrm4_tunnel_deregister(&ipip_handler
, AF_INET
))
584 pr_info("%s: can't deregister tunnel\n", __func__
);
586 unregister_pernet_device(&ipip_net_ops
);
589 module_init(ipip_init
);
590 module_exit(ipip_fini
);
591 MODULE_LICENSE("GPL");
592 MODULE_ALIAS_RTNL_LINK("ipip");
593 MODULE_ALIAS_NETDEV("tunl0");