2 * IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 * $Id: sit.c,v 1.53 2001/09/25 05:09:53 davem Exp $
11 * This program is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU General Public License
13 * as published by the Free Software Foundation; either version
14 * 2 of the License, or (at your option) any later version.
17 * Roger Venning <r.venning@telstra.com>: 6to4 support
18 * Nate Thompson <nate@thebog.net>: 6to4 support
21 #include <linux/module.h>
22 #include <linux/capability.h>
23 #include <linux/errno.h>
24 #include <linux/types.h>
25 #include <linux/socket.h>
26 #include <linux/sockios.h>
27 #include <linux/net.h>
28 #include <linux/in6.h>
29 #include <linux/netdevice.h>
30 #include <linux/if_arp.h>
31 #include <linux/icmp.h>
32 #include <asm/uaccess.h>
33 #include <linux/init.h>
34 #include <linux/netfilter_ipv4.h>
35 #include <linux/if_ether.h>
41 #include <net/protocol.h>
42 #include <net/transp_v6.h>
43 #include <net/ip6_fib.h>
44 #include <net/ip6_route.h>
45 #include <net/ndisc.h>
46 #include <net/addrconf.h>
51 #include <net/inet_ecn.h>
53 #include <net/dsfield.h>
56 This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
58 For comments look at net/ipv4/ip_gre.c --ANK
62 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
64 static int ipip6_fb_tunnel_init(struct net_device
*dev
);
65 static int ipip6_tunnel_init(struct net_device
*dev
);
66 static void ipip6_tunnel_setup(struct net_device
*dev
);
68 static struct net_device
*ipip6_fb_tunnel_dev
;
70 static struct ip_tunnel
*tunnels_r_l
[HASH_SIZE
];
71 static struct ip_tunnel
*tunnels_r
[HASH_SIZE
];
72 static struct ip_tunnel
*tunnels_l
[HASH_SIZE
];
73 static struct ip_tunnel
*tunnels_wc
[1];
74 static struct ip_tunnel
**tunnels
[4] = { tunnels_wc
, tunnels_l
, tunnels_r
, tunnels_r_l
};
76 static DEFINE_RWLOCK(ipip6_lock
);
78 static struct ip_tunnel
* ipip6_tunnel_lookup(__be32 remote
, __be32 local
)
80 unsigned h0
= HASH(remote
);
81 unsigned h1
= HASH(local
);
84 for (t
= tunnels_r_l
[h0
^h1
]; t
; t
= t
->next
) {
85 if (local
== t
->parms
.iph
.saddr
&&
86 remote
== t
->parms
.iph
.daddr
&& (t
->dev
->flags
&IFF_UP
))
89 for (t
= tunnels_r
[h0
]; t
; t
= t
->next
) {
90 if (remote
== t
->parms
.iph
.daddr
&& (t
->dev
->flags
&IFF_UP
))
93 for (t
= tunnels_l
[h1
]; t
; t
= t
->next
) {
94 if (local
== t
->parms
.iph
.saddr
&& (t
->dev
->flags
&IFF_UP
))
97 if ((t
= tunnels_wc
[0]) != NULL
&& (t
->dev
->flags
&IFF_UP
))
102 static struct ip_tunnel
**__ipip6_bucket(struct ip_tunnel_parm
*parms
)
104 __be32 remote
= parms
->iph
.daddr
;
105 __be32 local
= parms
->iph
.saddr
;
117 return &tunnels
[prio
][h
];
120 static inline struct ip_tunnel
**ipip6_bucket(struct ip_tunnel
*t
)
122 return __ipip6_bucket(&t
->parms
);
125 static void ipip6_tunnel_unlink(struct ip_tunnel
*t
)
127 struct ip_tunnel
**tp
;
129 for (tp
= ipip6_bucket(t
); *tp
; tp
= &(*tp
)->next
) {
131 write_lock_bh(&ipip6_lock
);
133 write_unlock_bh(&ipip6_lock
);
139 static void ipip6_tunnel_link(struct ip_tunnel
*t
)
141 struct ip_tunnel
**tp
= ipip6_bucket(t
);
144 write_lock_bh(&ipip6_lock
);
146 write_unlock_bh(&ipip6_lock
);
149 static struct ip_tunnel
* ipip6_tunnel_locate(struct ip_tunnel_parm
*parms
, int create
)
151 __be32 remote
= parms
->iph
.daddr
;
152 __be32 local
= parms
->iph
.saddr
;
153 struct ip_tunnel
*t
, **tp
, *nt
;
154 struct net_device
*dev
;
157 for (tp
= __ipip6_bucket(parms
); (t
= *tp
) != NULL
; tp
= &t
->next
) {
158 if (local
== t
->parms
.iph
.saddr
&& remote
== t
->parms
.iph
.daddr
)
165 strlcpy(name
, parms
->name
, IFNAMSIZ
);
168 for (i
=1; i
<100; i
++) {
169 sprintf(name
, "sit%d", i
);
170 if (__dev_get_by_name(&init_net
, name
) == NULL
)
177 dev
= alloc_netdev(sizeof(*t
), name
, ipip6_tunnel_setup
);
181 nt
= netdev_priv(dev
);
182 dev
->init
= ipip6_tunnel_init
;
185 if (register_netdevice(dev
) < 0) {
192 ipip6_tunnel_link(nt
);
199 static void ipip6_tunnel_uninit(struct net_device
*dev
)
201 if (dev
== ipip6_fb_tunnel_dev
) {
202 write_lock_bh(&ipip6_lock
);
203 tunnels_wc
[0] = NULL
;
204 write_unlock_bh(&ipip6_lock
);
207 ipip6_tunnel_unlink(netdev_priv(dev
));
213 static int ipip6_err(struct sk_buff
*skb
, u32 info
)
215 #ifndef I_WISH_WORLD_WERE_PERFECT
217 /* It is not :-( All the routers (except for Linux) return only
218 8 bytes of packet payload. It means, that precise relaying of
219 ICMP in the real Internet is absolutely infeasible.
221 struct iphdr
*iph
= (struct iphdr
*)skb
->data
;
222 const int type
= icmp_hdr(skb
)->type
;
223 const int code
= icmp_hdr(skb
)->code
;
229 case ICMP_PARAMETERPROB
:
232 case ICMP_DEST_UNREACH
:
235 case ICMP_PORT_UNREACH
:
236 /* Impossible event. */
238 case ICMP_FRAG_NEEDED
:
239 /* Soft state for pmtu is maintained by IP core. */
242 /* All others are translated to HOST_UNREACH.
243 rfc2003 contains "deep thoughts" about NET_UNREACH,
244 I believe they are just ether pollution. --ANK
249 case ICMP_TIME_EXCEEDED
:
250 if (code
!= ICMP_EXC_TTL
)
257 read_lock(&ipip6_lock
);
258 t
= ipip6_tunnel_lookup(iph
->daddr
, iph
->saddr
);
259 if (t
== NULL
|| t
->parms
.iph
.daddr
== 0)
263 if (t
->parms
.iph
.ttl
== 0 && type
== ICMP_TIME_EXCEEDED
)
266 if (jiffies
- t
->err_time
< IPTUNNEL_ERR_TIMEO
)
270 t
->err_time
= jiffies
;
272 read_unlock(&ipip6_lock
);
275 struct iphdr
*iph
= (struct iphdr
*)dp
;
276 int hlen
= iph
->ihl
<<2;
277 struct ipv6hdr
*iph6
;
278 const int type
= icmp_hdr(skb
)->type
;
279 const int code
= icmp_hdr(skb
)->code
;
283 struct sk_buff
*skb2
;
284 struct rt6_info
*rt6i
;
286 if (len
< hlen
+ sizeof(struct ipv6hdr
))
288 iph6
= (struct ipv6hdr
*)(dp
+ hlen
);
293 case ICMP_PARAMETERPROB
:
294 if (icmp_hdr(skb
)->un
.gateway
< hlen
)
297 /* So... This guy found something strange INSIDE encapsulated
298 packet. Well, he is fool, but what can we do ?
300 rel_type
= ICMPV6_PARAMPROB
;
301 rel_info
= icmp_hdr(skb
)->un
.gateway
- hlen
;
304 case ICMP_DEST_UNREACH
:
307 case ICMP_PORT_UNREACH
:
308 /* Impossible event. */
310 case ICMP_FRAG_NEEDED
:
311 /* Too complicated case ... */
314 /* All others are translated to HOST_UNREACH.
315 rfc2003 contains "deep thoughts" about NET_UNREACH,
316 I believe, it is just ether pollution. --ANK
318 rel_type
= ICMPV6_DEST_UNREACH
;
319 rel_code
= ICMPV6_ADDR_UNREACH
;
323 case ICMP_TIME_EXCEEDED
:
324 if (code
!= ICMP_EXC_TTL
)
326 rel_type
= ICMPV6_TIME_EXCEED
;
327 rel_code
= ICMPV6_EXC_HOPLIMIT
;
331 /* Prepare fake skb to feed it to icmpv6_send */
332 skb2
= skb_clone(skb
, GFP_ATOMIC
);
335 dst_release(skb2
->dst
);
337 skb_pull(skb2
, skb
->data
- (u8
*)iph6
);
338 skb_reset_network_header(skb2
);
340 /* Try to guess incoming interface */
341 rt6i
= rt6_lookup(&iph6
->saddr
, NULL
, NULL
, 0);
342 if (rt6i
&& rt6i
->rt6i_dev
) {
343 skb2
->dev
= rt6i
->rt6i_dev
;
345 rt6i
= rt6_lookup(&iph6
->daddr
, &iph6
->saddr
, NULL
, 0);
347 if (rt6i
&& rt6i
->rt6i_dev
&& rt6i
->rt6i_dev
->type
== ARPHRD_SIT
) {
348 struct ip_tunnel
*t
= netdev_priv(rt6i
->rt6i_dev
);
349 if (rel_type
== ICMPV6_TIME_EXCEED
&& t
->parms
.iph
.ttl
) {
350 rel_type
= ICMPV6_DEST_UNREACH
;
351 rel_code
= ICMPV6_ADDR_UNREACH
;
353 icmpv6_send(skb2
, rel_type
, rel_code
, rel_info
, skb2
->dev
);
361 static inline void ipip6_ecn_decapsulate(struct iphdr
*iph
, struct sk_buff
*skb
)
363 if (INET_ECN_is_ce(iph
->tos
))
364 IP6_ECN_set_ce(ipv6_hdr(skb
));
367 static int ipip6_rcv(struct sk_buff
*skb
)
370 struct ip_tunnel
*tunnel
;
372 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
377 read_lock(&ipip6_lock
);
378 if ((tunnel
= ipip6_tunnel_lookup(iph
->saddr
, iph
->daddr
)) != NULL
) {
380 skb
->mac_header
= skb
->network_header
;
381 skb_reset_network_header(skb
);
382 IPCB(skb
)->flags
= 0;
383 skb
->protocol
= htons(ETH_P_IPV6
);
384 skb
->pkt_type
= PACKET_HOST
;
385 tunnel
->stat
.rx_packets
++;
386 tunnel
->stat
.rx_bytes
+= skb
->len
;
387 skb
->dev
= tunnel
->dev
;
388 dst_release(skb
->dst
);
391 ipip6_ecn_decapsulate(iph
, skb
);
393 read_unlock(&ipip6_lock
);
397 icmp_send(skb
, ICMP_DEST_UNREACH
, ICMP_PORT_UNREACH
, 0);
399 read_unlock(&ipip6_lock
);
404 /* Returns the embedded IPv4 address if the IPv6 address
405 comes from 6to4 (RFC 3056) addr space */
407 static inline __be32
try_6to4(struct in6_addr
*v6dst
)
411 if (v6dst
->s6_addr16
[0] == htons(0x2002)) {
412 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
413 memcpy(&dst
, &v6dst
->s6_addr16
[1], 4);
419 * This function assumes it is being called from dev_queue_xmit()
420 * and that skb is filled properly by that function.
423 static int ipip6_tunnel_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
425 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
426 struct net_device_stats
*stats
= &tunnel
->stat
;
427 struct iphdr
*tiph
= &tunnel
->parms
.iph
;
428 struct ipv6hdr
*iph6
= ipv6_hdr(skb
);
429 u8 tos
= tunnel
->parms
.iph
.tos
;
430 struct rtable
*rt
; /* Route to the other host */
431 struct net_device
*tdev
; /* Device to other host */
432 struct iphdr
*iph
; /* Our new IP header */
433 int max_headroom
; /* The extra header space needed */
434 __be32 dst
= tiph
->daddr
;
436 struct in6_addr
*addr6
;
439 if (tunnel
->recursion
++) {
440 tunnel
->stat
.collisions
++;
444 if (skb
->protocol
!= htons(ETH_P_IPV6
))
448 dst
= try_6to4(&iph6
->daddr
);
451 struct neighbour
*neigh
= NULL
;
454 neigh
= skb
->dst
->neighbour
;
458 printk(KERN_DEBUG
"sit: nexthop == NULL\n");
462 addr6
= (struct in6_addr
*)&neigh
->primary_key
;
463 addr_type
= ipv6_addr_type(addr6
);
465 if (addr_type
== IPV6_ADDR_ANY
) {
466 addr6
= &ipv6_hdr(skb
)->daddr
;
467 addr_type
= ipv6_addr_type(addr6
);
470 if ((addr_type
& IPV6_ADDR_COMPATv4
) == 0)
473 dst
= addr6
->s6_addr32
[3];
477 struct flowi fl
= { .nl_u
= { .ip4_u
=
479 .saddr
= tiph
->saddr
,
480 .tos
= RT_TOS(tos
) } },
481 .oif
= tunnel
->parms
.link
,
482 .proto
= IPPROTO_IPV6
};
483 if (ip_route_output_key(&rt
, &fl
)) {
484 tunnel
->stat
.tx_carrier_errors
++;
488 if (rt
->rt_type
!= RTN_UNICAST
) {
490 tunnel
->stat
.tx_carrier_errors
++;
493 tdev
= rt
->u
.dst
.dev
;
497 tunnel
->stat
.collisions
++;
502 mtu
= dst_mtu(&rt
->u
.dst
) - sizeof(struct iphdr
);
504 mtu
= skb
->dst
? dst_mtu(skb
->dst
) : dev
->mtu
;
507 tunnel
->stat
.collisions
++;
511 if (mtu
< IPV6_MIN_MTU
)
513 if (tunnel
->parms
.iph
.daddr
&& skb
->dst
)
514 skb
->dst
->ops
->update_pmtu(skb
->dst
, mtu
);
516 if (skb
->len
> mtu
) {
517 icmpv6_send(skb
, ICMPV6_PKT_TOOBIG
, 0, mtu
, dev
);
522 if (tunnel
->err_count
> 0) {
523 if (jiffies
- tunnel
->err_time
< IPTUNNEL_ERR_TIMEO
) {
525 dst_link_failure(skb
);
527 tunnel
->err_count
= 0;
531 * Okay, now see if we can stuff it in the buffer as-is.
533 max_headroom
= LL_RESERVED_SPACE(tdev
)+sizeof(struct iphdr
);
535 if (skb_headroom(skb
) < max_headroom
|| skb_shared(skb
) ||
536 (skb_cloned(skb
) && !skb_clone_writable(skb
, 0))) {
537 struct sk_buff
*new_skb
= skb_realloc_headroom(skb
, max_headroom
);
546 skb_set_owner_w(new_skb
, skb
->sk
);
549 iph6
= ipv6_hdr(skb
);
552 skb
->transport_header
= skb
->network_header
;
553 skb_push(skb
, sizeof(struct iphdr
));
554 skb_reset_network_header(skb
);
555 memset(&(IPCB(skb
)->opt
), 0, sizeof(IPCB(skb
)->opt
));
556 IPCB(skb
)->flags
= 0;
557 dst_release(skb
->dst
);
558 skb
->dst
= &rt
->u
.dst
;
561 * Push down and install the IPIP header.
566 iph
->ihl
= sizeof(struct iphdr
)>>2;
567 if (mtu
> IPV6_MIN_MTU
)
568 iph
->frag_off
= htons(IP_DF
);
572 iph
->protocol
= IPPROTO_IPV6
;
573 iph
->tos
= INET_ECN_encapsulate(tos
, ipv6_get_dsfield(iph6
));
574 iph
->daddr
= rt
->rt_dst
;
575 iph
->saddr
= rt
->rt_src
;
577 if ((iph
->ttl
= tiph
->ttl
) == 0)
578 iph
->ttl
= iph6
->hop_limit
;
587 dst_link_failure(skb
);
596 ipip6_tunnel_ioctl (struct net_device
*dev
, struct ifreq
*ifr
, int cmd
)
599 struct ip_tunnel_parm p
;
605 if (dev
== ipip6_fb_tunnel_dev
) {
606 if (copy_from_user(&p
, ifr
->ifr_ifru
.ifru_data
, sizeof(p
))) {
610 t
= ipip6_tunnel_locate(&p
, 0);
613 t
= netdev_priv(dev
);
614 memcpy(&p
, &t
->parms
, sizeof(p
));
615 if (copy_to_user(ifr
->ifr_ifru
.ifru_data
, &p
, sizeof(p
)))
622 if (!capable(CAP_NET_ADMIN
))
626 if (copy_from_user(&p
, ifr
->ifr_ifru
.ifru_data
, sizeof(p
)))
630 if (p
.iph
.version
!= 4 || p
.iph
.protocol
!= IPPROTO_IPV6
||
631 p
.iph
.ihl
!= 5 || (p
.iph
.frag_off
&htons(~IP_DF
)))
634 p
.iph
.frag_off
|= htons(IP_DF
);
636 t
= ipip6_tunnel_locate(&p
, cmd
== SIOCADDTUNNEL
);
638 if (dev
!= ipip6_fb_tunnel_dev
&& cmd
== SIOCCHGTUNNEL
) {
645 if (((dev
->flags
&IFF_POINTOPOINT
) && !p
.iph
.daddr
) ||
646 (!(dev
->flags
&IFF_POINTOPOINT
) && p
.iph
.daddr
)) {
650 t
= netdev_priv(dev
);
651 ipip6_tunnel_unlink(t
);
652 t
->parms
.iph
.saddr
= p
.iph
.saddr
;
653 t
->parms
.iph
.daddr
= p
.iph
.daddr
;
654 memcpy(dev
->dev_addr
, &p
.iph
.saddr
, 4);
655 memcpy(dev
->broadcast
, &p
.iph
.daddr
, 4);
656 ipip6_tunnel_link(t
);
657 netdev_state_change(dev
);
663 if (cmd
== SIOCCHGTUNNEL
) {
664 t
->parms
.iph
.ttl
= p
.iph
.ttl
;
665 t
->parms
.iph
.tos
= p
.iph
.tos
;
667 if (copy_to_user(ifr
->ifr_ifru
.ifru_data
, &t
->parms
, sizeof(p
)))
670 err
= (cmd
== SIOCADDTUNNEL
? -ENOBUFS
: -ENOENT
);
675 if (!capable(CAP_NET_ADMIN
))
678 if (dev
== ipip6_fb_tunnel_dev
) {
680 if (copy_from_user(&p
, ifr
->ifr_ifru
.ifru_data
, sizeof(p
)))
683 if ((t
= ipip6_tunnel_locate(&p
, 0)) == NULL
)
686 if (t
== netdev_priv(ipip6_fb_tunnel_dev
))
690 unregister_netdevice(dev
);
702 static struct net_device_stats
*ipip6_tunnel_get_stats(struct net_device
*dev
)
704 return &(((struct ip_tunnel
*)netdev_priv(dev
))->stat
);
707 static int ipip6_tunnel_change_mtu(struct net_device
*dev
, int new_mtu
)
709 if (new_mtu
< IPV6_MIN_MTU
|| new_mtu
> 0xFFF8 - sizeof(struct iphdr
))
715 static void ipip6_tunnel_setup(struct net_device
*dev
)
717 dev
->uninit
= ipip6_tunnel_uninit
;
718 dev
->destructor
= free_netdev
;
719 dev
->hard_start_xmit
= ipip6_tunnel_xmit
;
720 dev
->get_stats
= ipip6_tunnel_get_stats
;
721 dev
->do_ioctl
= ipip6_tunnel_ioctl
;
722 dev
->change_mtu
= ipip6_tunnel_change_mtu
;
724 dev
->type
= ARPHRD_SIT
;
725 dev
->hard_header_len
= LL_MAX_HEADER
+ sizeof(struct iphdr
);
726 dev
->mtu
= ETH_DATA_LEN
- sizeof(struct iphdr
);
727 dev
->flags
= IFF_NOARP
;
732 static int ipip6_tunnel_init(struct net_device
*dev
)
734 struct net_device
*tdev
= NULL
;
735 struct ip_tunnel
*tunnel
;
738 tunnel
= netdev_priv(dev
);
739 iph
= &tunnel
->parms
.iph
;
742 strcpy(tunnel
->parms
.name
, dev
->name
);
744 memcpy(dev
->dev_addr
, &tunnel
->parms
.iph
.saddr
, 4);
745 memcpy(dev
->broadcast
, &tunnel
->parms
.iph
.daddr
, 4);
748 struct flowi fl
= { .nl_u
= { .ip4_u
=
749 { .daddr
= iph
->daddr
,
751 .tos
= RT_TOS(iph
->tos
) } },
752 .oif
= tunnel
->parms
.link
,
753 .proto
= IPPROTO_IPV6
};
755 if (!ip_route_output_key(&rt
, &fl
)) {
756 tdev
= rt
->u
.dst
.dev
;
759 dev
->flags
|= IFF_POINTOPOINT
;
762 if (!tdev
&& tunnel
->parms
.link
)
763 tdev
= __dev_get_by_index(&init_net
, tunnel
->parms
.link
);
766 dev
->hard_header_len
= tdev
->hard_header_len
+ sizeof(struct iphdr
);
767 dev
->mtu
= tdev
->mtu
- sizeof(struct iphdr
);
768 if (dev
->mtu
< IPV6_MIN_MTU
)
769 dev
->mtu
= IPV6_MIN_MTU
;
771 dev
->iflink
= tunnel
->parms
.link
;
776 static int __init
ipip6_fb_tunnel_init(struct net_device
*dev
)
778 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
779 struct iphdr
*iph
= &tunnel
->parms
.iph
;
782 strcpy(tunnel
->parms
.name
, dev
->name
);
785 iph
->protocol
= IPPROTO_IPV6
;
790 tunnels_wc
[0] = tunnel
;
794 static struct xfrm_tunnel sit_handler
= {
795 .handler
= ipip6_rcv
,
796 .err_handler
= ipip6_err
,
800 static void __exit
sit_destroy_tunnels(void)
804 for (prio
= 1; prio
< 4; prio
++) {
806 for (h
= 0; h
< HASH_SIZE
; h
++) {
808 while ((t
= tunnels
[prio
][h
]) != NULL
)
809 unregister_netdevice(t
->dev
);
814 static void __exit
sit_cleanup(void)
816 xfrm4_tunnel_deregister(&sit_handler
, AF_INET6
);
819 sit_destroy_tunnels();
820 unregister_netdevice(ipip6_fb_tunnel_dev
);
824 static int __init
sit_init(void)
828 printk(KERN_INFO
"IPv6 over IPv4 tunneling driver\n");
830 if (xfrm4_tunnel_register(&sit_handler
, AF_INET6
) < 0) {
831 printk(KERN_INFO
"sit init: Can't add protocol\n");
835 ipip6_fb_tunnel_dev
= alloc_netdev(sizeof(struct ip_tunnel
), "sit0",
837 if (!ipip6_fb_tunnel_dev
) {
842 ipip6_fb_tunnel_dev
->init
= ipip6_fb_tunnel_init
;
844 if ((err
= register_netdev(ipip6_fb_tunnel_dev
)))
850 free_netdev(ipip6_fb_tunnel_dev
);
852 xfrm4_tunnel_deregister(&sit_handler
, AF_INET6
);
856 module_init(sit_init
);
857 module_exit(sit_cleanup
);
858 MODULE_LICENSE("GPL");
859 MODULE_ALIAS("sit0");