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/config.h>
22 #include <linux/module.h>
23 #include <linux/errno.h>
24 #include <linux/types.h>
25 #include <linux/socket.h>
26 #include <linux/sockios.h>
27 #include <linux/sched.h>
28 #include <linux/net.h>
29 #include <linux/in6.h>
30 #include <linux/netdevice.h>
31 #include <linux/if_arp.h>
32 #include <linux/icmp.h>
33 #include <asm/uaccess.h>
34 #include <linux/init.h>
35 #include <linux/netfilter_ipv4.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) ((addr^(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(u32 remote
, u32 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
*t
)
104 u32 remote
= t
->parms
.iph
.daddr
;
105 u32 local
= t
->parms
.iph
.saddr
;
117 return &tunnels
[prio
][h
];
120 static void ipip6_tunnel_unlink(struct ip_tunnel
*t
)
122 struct ip_tunnel
**tp
;
124 for (tp
= ipip6_bucket(t
); *tp
; tp
= &(*tp
)->next
) {
126 write_lock_bh(&ipip6_lock
);
128 write_unlock_bh(&ipip6_lock
);
134 static void ipip6_tunnel_link(struct ip_tunnel
*t
)
136 struct ip_tunnel
**tp
= ipip6_bucket(t
);
139 write_lock_bh(&ipip6_lock
);
141 write_unlock_bh(&ipip6_lock
);
144 static struct ip_tunnel
* ipip6_tunnel_locate(struct ip_tunnel_parm
*parms
, int create
)
146 u32 remote
= parms
->iph
.daddr
;
147 u32 local
= parms
->iph
.saddr
;
148 struct ip_tunnel
*t
, **tp
, *nt
;
149 struct net_device
*dev
;
162 for (tp
= &tunnels
[prio
][h
]; (t
= *tp
) != NULL
; tp
= &t
->next
) {
163 if (local
== t
->parms
.iph
.saddr
&& remote
== t
->parms
.iph
.daddr
)
170 strlcpy(name
, parms
->name
, IFNAMSIZ
);
173 for (i
=1; i
<100; i
++) {
174 sprintf(name
, "sit%d", i
);
175 if (__dev_get_by_name(name
) == NULL
)
182 dev
= alloc_netdev(sizeof(*t
), name
, ipip6_tunnel_setup
);
187 dev
->init
= ipip6_tunnel_init
;
190 if (register_netdevice(dev
) < 0) {
197 ipip6_tunnel_link(nt
);
198 /* Do not decrement MOD_USE_COUNT here. */
205 static void ipip6_tunnel_uninit(struct net_device
*dev
)
207 if (dev
== ipip6_fb_tunnel_dev
) {
208 write_lock_bh(&ipip6_lock
);
209 tunnels_wc
[0] = NULL
;
210 write_unlock_bh(&ipip6_lock
);
213 ipip6_tunnel_unlink((struct ip_tunnel
*)dev
->priv
);
219 static void ipip6_err(struct sk_buff
*skb
, u32 info
)
221 #ifndef I_WISH_WORLD_WERE_PERFECT
223 /* It is not :-( All the routers (except for Linux) return only
224 8 bytes of packet payload. It means, that precise relaying of
225 ICMP in the real Internet is absolutely infeasible.
227 struct iphdr
*iph
= (struct iphdr
*)skb
->data
;
228 int type
= skb
->h
.icmph
->type
;
229 int code
= skb
->h
.icmph
->code
;
234 case ICMP_PARAMETERPROB
:
237 case ICMP_DEST_UNREACH
:
240 case ICMP_PORT_UNREACH
:
241 /* Impossible event. */
243 case ICMP_FRAG_NEEDED
:
244 /* Soft state for pmtu is maintained by IP core. */
247 /* All others are translated to HOST_UNREACH.
248 rfc2003 contains "deep thoughts" about NET_UNREACH,
249 I believe they are just ether pollution. --ANK
254 case ICMP_TIME_EXCEEDED
:
255 if (code
!= ICMP_EXC_TTL
)
260 read_lock(&ipip6_lock
);
261 t
= ipip6_tunnel_lookup(iph
->daddr
, iph
->saddr
);
262 if (t
== NULL
|| t
->parms
.iph
.daddr
== 0)
264 if (t
->parms
.iph
.ttl
== 0 && type
== ICMP_TIME_EXCEEDED
)
267 if (jiffies
- t
->err_time
< IPTUNNEL_ERR_TIMEO
)
271 t
->err_time
= jiffies
;
273 read_unlock(&ipip6_lock
);
276 struct iphdr
*iph
= (struct iphdr
*)dp
;
277 int hlen
= iph
->ihl
<<2;
278 struct ipv6hdr
*iph6
;
279 int type
= skb
->h
.icmph
->type
;
280 int code
= skb
->h
.icmph
->code
;
284 struct sk_buff
*skb2
;
285 struct rt6_info
*rt6i
;
287 if (len
< hlen
+ sizeof(struct ipv6hdr
))
289 iph6
= (struct ipv6hdr
*)(dp
+ hlen
);
294 case ICMP_PARAMETERPROB
:
295 if (skb
->h
.icmph
->un
.gateway
< hlen
)
298 /* So... This guy found something strange INSIDE encapsulated
299 packet. Well, he is fool, but what can we do ?
301 rel_type
= ICMPV6_PARAMPROB
;
302 rel_info
= skb
->h
.icmph
->un
.gateway
- hlen
;
305 case ICMP_DEST_UNREACH
:
308 case ICMP_PORT_UNREACH
:
309 /* Impossible event. */
311 case ICMP_FRAG_NEEDED
:
312 /* Too complicated case ... */
315 /* All others are translated to HOST_UNREACH.
316 rfc2003 contains "deep thoughts" about NET_UNREACH,
317 I believe, it is just ether pollution. --ANK
319 rel_type
= ICMPV6_DEST_UNREACH
;
320 rel_code
= ICMPV6_ADDR_UNREACH
;
324 case ICMP_TIME_EXCEEDED
:
325 if (code
!= ICMP_EXC_TTL
)
327 rel_type
= ICMPV6_TIME_EXCEED
;
328 rel_code
= ICMPV6_EXC_HOPLIMIT
;
332 /* Prepare fake skb to feed it to icmpv6_send */
333 skb2
= skb_clone(skb
, GFP_ATOMIC
);
336 dst_release(skb2
->dst
);
338 skb_pull(skb2
, skb
->data
- (u8
*)iph6
);
339 skb2
->nh
.raw
= skb2
->data
;
341 /* Try to guess incoming interface */
342 rt6i
= rt6_lookup(&iph6
->saddr
, NULL
, NULL
, 0);
343 if (rt6i
&& rt6i
->rt6i_dev
) {
344 skb2
->dev
= rt6i
->rt6i_dev
;
346 rt6i
= rt6_lookup(&iph6
->daddr
, &iph6
->saddr
, NULL
, 0);
348 if (rt6i
&& rt6i
->rt6i_dev
&& rt6i
->rt6i_dev
->type
== ARPHRD_SIT
) {
349 struct ip_tunnel
* t
= (struct ip_tunnel
*)rt6i
->rt6i_dev
->priv
;
350 if (rel_type
== ICMPV6_TIME_EXCEED
&& t
->parms
.iph
.ttl
) {
351 rel_type
= ICMPV6_DEST_UNREACH
;
352 rel_code
= ICMPV6_ADDR_UNREACH
;
354 icmpv6_send(skb2
, rel_type
, rel_code
, rel_info
, skb2
->dev
);
362 static inline void ipip6_ecn_decapsulate(struct iphdr
*iph
, struct sk_buff
*skb
)
364 if (INET_ECN_is_ce(iph
->tos
))
365 IP6_ECN_set_ce(skb
->nh
.ipv6h
);
368 static int ipip6_rcv(struct sk_buff
*skb
)
371 struct ip_tunnel
*tunnel
;
373 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
378 read_lock(&ipip6_lock
);
379 if ((tunnel
= ipip6_tunnel_lookup(iph
->saddr
, iph
->daddr
)) != NULL
) {
381 skb
->mac
.raw
= skb
->nh
.raw
;
382 skb
->nh
.raw
= skb
->data
;
383 memset(&(IPCB(skb
)->opt
), 0, sizeof(struct ip_options
));
384 skb
->protocol
= htons(ETH_P_IPV6
);
385 skb
->pkt_type
= PACKET_HOST
;
386 tunnel
->stat
.rx_packets
++;
387 tunnel
->stat
.rx_bytes
+= skb
->len
;
388 skb
->dev
= tunnel
->dev
;
389 dst_release(skb
->dst
);
392 ipip6_ecn_decapsulate(iph
, skb
);
394 read_unlock(&ipip6_lock
);
398 icmp_send(skb
, ICMP_DEST_UNREACH
, ICMP_PROT_UNREACH
, 0);
400 read_unlock(&ipip6_lock
);
405 /* Returns the embedded IPv4 address if the IPv6 address
406 comes from 6to4 (RFC 3056) addr space */
408 static inline u32
try_6to4(struct in6_addr
*v6dst
)
412 if (v6dst
->s6_addr16
[0] == htons(0x2002)) {
413 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
414 memcpy(&dst
, &v6dst
->s6_addr16
[1], 4);
420 * This function assumes it is being called from dev_queue_xmit()
421 * and that skb is filled properly by that function.
424 static int ipip6_tunnel_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
426 struct ip_tunnel
*tunnel
= (struct ip_tunnel
*)dev
->priv
;
427 struct net_device_stats
*stats
= &tunnel
->stat
;
428 struct iphdr
*tiph
= &tunnel
->parms
.iph
;
429 struct ipv6hdr
*iph6
= skb
->nh
.ipv6h
;
430 u8 tos
= tunnel
->parms
.iph
.tos
;
431 struct rtable
*rt
; /* Route to the other host */
432 struct net_device
*tdev
; /* Device to other host */
433 struct iphdr
*iph
; /* Our new IP header */
434 int max_headroom
; /* The extra header space needed */
435 u32 dst
= tiph
->daddr
;
437 struct in6_addr
*addr6
;
440 if (tunnel
->recursion
++) {
441 tunnel
->stat
.collisions
++;
445 if (skb
->protocol
!= htons(ETH_P_IPV6
))
449 dst
= try_6to4(&iph6
->daddr
);
452 struct neighbour
*neigh
= NULL
;
455 neigh
= skb
->dst
->neighbour
;
459 printk(KERN_DEBUG
"sit: nexthop == NULL\n");
463 addr6
= (struct in6_addr
*)&neigh
->primary_key
;
464 addr_type
= ipv6_addr_type(addr6
);
466 if (addr_type
== IPV6_ADDR_ANY
) {
467 addr6
= &skb
->nh
.ipv6h
->daddr
;
468 addr_type
= ipv6_addr_type(addr6
);
471 if ((addr_type
& IPV6_ADDR_COMPATv4
) == 0)
474 dst
= addr6
->s6_addr32
[3];
478 struct flowi fl
= { .nl_u
= { .ip4_u
=
480 .saddr
= tiph
->saddr
,
481 .tos
= RT_TOS(tos
) } },
482 .oif
= tunnel
->parms
.link
,
483 .proto
= IPPROTO_IPV6
};
484 if (ip_route_output_key(&rt
, &fl
)) {
485 tunnel
->stat
.tx_carrier_errors
++;
489 if (rt
->rt_type
!= RTN_UNICAST
) {
491 tunnel
->stat
.tx_carrier_errors
++;
494 tdev
= rt
->u
.dst
.dev
;
498 tunnel
->stat
.collisions
++;
503 mtu
= dst_mtu(&rt
->u
.dst
) - sizeof(struct iphdr
);
505 mtu
= skb
->dst
? dst_mtu(skb
->dst
) : dev
->mtu
;
508 tunnel
->stat
.collisions
++;
512 if (mtu
< IPV6_MIN_MTU
)
514 if (tunnel
->parms
.iph
.daddr
&& skb
->dst
)
515 skb
->dst
->ops
->update_pmtu(skb
->dst
, mtu
);
517 if (skb
->len
> mtu
) {
518 icmpv6_send(skb
, ICMPV6_PKT_TOOBIG
, 0, mtu
, dev
);
523 if (tunnel
->err_count
> 0) {
524 if (jiffies
- tunnel
->err_time
< IPTUNNEL_ERR_TIMEO
) {
526 dst_link_failure(skb
);
528 tunnel
->err_count
= 0;
532 * Okay, now see if we can stuff it in the buffer as-is.
534 max_headroom
= LL_RESERVED_SPACE(tdev
)+sizeof(struct iphdr
);
536 if (skb_headroom(skb
) < max_headroom
|| skb_cloned(skb
) || skb_shared(skb
)) {
537 struct sk_buff
*new_skb
= skb_realloc_headroom(skb
, max_headroom
);
546 skb_set_owner_w(new_skb
, skb
->sk
);
549 iph6
= skb
->nh
.ipv6h
;
552 skb
->h
.raw
= skb
->nh
.raw
;
553 skb
->nh
.raw
= skb_push(skb
, sizeof(struct iphdr
));
554 memset(&(IPCB(skb
)->opt
), 0, sizeof(IPCB(skb
)->opt
));
555 dst_release(skb
->dst
);
556 skb
->dst
= &rt
->u
.dst
;
559 * Push down and install the IPIP header.
564 iph
->ihl
= sizeof(struct iphdr
)>>2;
565 if (mtu
> IPV6_MIN_MTU
)
566 iph
->frag_off
= htons(IP_DF
);
570 iph
->protocol
= IPPROTO_IPV6
;
571 iph
->tos
= INET_ECN_encapsulate(tos
, ipv6_get_dsfield(iph6
));
572 iph
->daddr
= rt
->rt_dst
;
573 iph
->saddr
= rt
->rt_src
;
575 if ((iph
->ttl
= tiph
->ttl
) == 0)
576 iph
->ttl
= iph6
->hop_limit
;
585 dst_link_failure(skb
);
594 ipip6_tunnel_ioctl (struct net_device
*dev
, struct ifreq
*ifr
, int cmd
)
597 struct ip_tunnel_parm p
;
603 if (dev
== ipip6_fb_tunnel_dev
) {
604 if (copy_from_user(&p
, ifr
->ifr_ifru
.ifru_data
, sizeof(p
))) {
608 t
= ipip6_tunnel_locate(&p
, 0);
611 t
= (struct ip_tunnel
*)dev
->priv
;
612 memcpy(&p
, &t
->parms
, sizeof(p
));
613 if (copy_to_user(ifr
->ifr_ifru
.ifru_data
, &p
, sizeof(p
)))
620 if (!capable(CAP_NET_ADMIN
))
624 if (copy_from_user(&p
, ifr
->ifr_ifru
.ifru_data
, sizeof(p
)))
628 if (p
.iph
.version
!= 4 || p
.iph
.protocol
!= IPPROTO_IPV6
||
629 p
.iph
.ihl
!= 5 || (p
.iph
.frag_off
&htons(~IP_DF
)))
632 p
.iph
.frag_off
|= htons(IP_DF
);
634 t
= ipip6_tunnel_locate(&p
, cmd
== SIOCADDTUNNEL
);
636 if (dev
!= ipip6_fb_tunnel_dev
&& cmd
== SIOCCHGTUNNEL
) {
643 if (((dev
->flags
&IFF_POINTOPOINT
) && !p
.iph
.daddr
) ||
644 (!(dev
->flags
&IFF_POINTOPOINT
) && p
.iph
.daddr
)) {
648 t
= (struct ip_tunnel
*)dev
->priv
;
649 ipip6_tunnel_unlink(t
);
650 t
->parms
.iph
.saddr
= p
.iph
.saddr
;
651 t
->parms
.iph
.daddr
= p
.iph
.daddr
;
652 memcpy(dev
->dev_addr
, &p
.iph
.saddr
, 4);
653 memcpy(dev
->broadcast
, &p
.iph
.daddr
, 4);
654 ipip6_tunnel_link(t
);
655 netdev_state_change(dev
);
661 if (cmd
== SIOCCHGTUNNEL
) {
662 t
->parms
.iph
.ttl
= p
.iph
.ttl
;
663 t
->parms
.iph
.tos
= p
.iph
.tos
;
665 if (copy_to_user(ifr
->ifr_ifru
.ifru_data
, &t
->parms
, sizeof(p
)))
668 err
= (cmd
== SIOCADDTUNNEL
? -ENOBUFS
: -ENOENT
);
673 if (!capable(CAP_NET_ADMIN
))
676 if (dev
== ipip6_fb_tunnel_dev
) {
678 if (copy_from_user(&p
, ifr
->ifr_ifru
.ifru_data
, sizeof(p
)))
681 if ((t
= ipip6_tunnel_locate(&p
, 0)) == NULL
)
684 if (t
== ipip6_fb_tunnel_dev
->priv
)
688 err
= unregister_netdevice(dev
);
699 static struct net_device_stats
*ipip6_tunnel_get_stats(struct net_device
*dev
)
701 return &(((struct ip_tunnel
*)dev
->priv
)->stat
);
704 static int ipip6_tunnel_change_mtu(struct net_device
*dev
, int new_mtu
)
706 if (new_mtu
< IPV6_MIN_MTU
|| new_mtu
> 0xFFF8 - sizeof(struct iphdr
))
712 static void ipip6_tunnel_setup(struct net_device
*dev
)
714 SET_MODULE_OWNER(dev
);
715 dev
->uninit
= ipip6_tunnel_uninit
;
716 dev
->destructor
= free_netdev
;
717 dev
->hard_start_xmit
= ipip6_tunnel_xmit
;
718 dev
->get_stats
= ipip6_tunnel_get_stats
;
719 dev
->do_ioctl
= ipip6_tunnel_ioctl
;
720 dev
->change_mtu
= ipip6_tunnel_change_mtu
;
722 dev
->type
= ARPHRD_SIT
;
723 dev
->hard_header_len
= LL_MAX_HEADER
+ sizeof(struct iphdr
);
724 dev
->mtu
= 1500 - sizeof(struct iphdr
);
725 dev
->flags
= IFF_NOARP
;
730 static int ipip6_tunnel_init(struct net_device
*dev
)
732 struct net_device
*tdev
= NULL
;
733 struct ip_tunnel
*tunnel
;
736 tunnel
= (struct ip_tunnel
*)dev
->priv
;
737 iph
= &tunnel
->parms
.iph
;
740 strcpy(tunnel
->parms
.name
, dev
->name
);
742 memcpy(dev
->dev_addr
, &tunnel
->parms
.iph
.saddr
, 4);
743 memcpy(dev
->broadcast
, &tunnel
->parms
.iph
.daddr
, 4);
746 struct flowi fl
= { .nl_u
= { .ip4_u
=
747 { .daddr
= iph
->daddr
,
749 .tos
= RT_TOS(iph
->tos
) } },
750 .oif
= tunnel
->parms
.link
,
751 .proto
= IPPROTO_IPV6
};
753 if (!ip_route_output_key(&rt
, &fl
)) {
754 tdev
= rt
->u
.dst
.dev
;
757 dev
->flags
|= IFF_POINTOPOINT
;
760 if (!tdev
&& tunnel
->parms
.link
)
761 tdev
= __dev_get_by_index(tunnel
->parms
.link
);
764 dev
->hard_header_len
= tdev
->hard_header_len
+ sizeof(struct iphdr
);
765 dev
->mtu
= tdev
->mtu
- sizeof(struct iphdr
);
766 if (dev
->mtu
< IPV6_MIN_MTU
)
767 dev
->mtu
= IPV6_MIN_MTU
;
769 dev
->iflink
= tunnel
->parms
.link
;
774 int __init
ipip6_fb_tunnel_init(struct net_device
*dev
)
776 struct ip_tunnel
*tunnel
= dev
->priv
;
777 struct iphdr
*iph
= &tunnel
->parms
.iph
;
780 strcpy(tunnel
->parms
.name
, dev
->name
);
783 iph
->protocol
= IPPROTO_IPV6
;
788 tunnels_wc
[0] = tunnel
;
792 static struct net_protocol sit_protocol
= {
793 .handler
= ipip6_rcv
,
794 .err_handler
= ipip6_err
,
797 void __exit
sit_cleanup(void)
799 inet_del_protocol(&sit_protocol
, IPPROTO_IPV6
);
800 unregister_netdev(ipip6_fb_tunnel_dev
);
803 int __init
sit_init(void)
807 printk(KERN_INFO
"IPv6 over IPv4 tunneling driver\n");
809 if (inet_add_protocol(&sit_protocol
, IPPROTO_IPV6
) < 0) {
810 printk(KERN_INFO
"sit init: Can't add protocol\n");
814 ipip6_fb_tunnel_dev
= alloc_netdev(sizeof(struct ip_tunnel
), "sit0",
816 if (!ipip6_fb_tunnel_dev
) {
821 ipip6_fb_tunnel_dev
->init
= ipip6_fb_tunnel_init
;
823 if ((err
= register_netdev(ipip6_fb_tunnel_dev
)))
829 free_netdev(ipip6_fb_tunnel_dev
);
831 inet_del_protocol(&sit_protocol
, IPPROTO_IPV6
);