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 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
15 * Roger Venning <r.venning@telstra.com>: 6to4 support
16 * Nate Thompson <nate@thebog.net>: 6to4 support
17 * Fred Templin <fred.l.templin@boeing.com>: isatap support
20 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
22 #include <linux/module.h>
23 #include <linux/capability.h>
24 #include <linux/errno.h>
25 #include <linux/types.h>
26 #include <linux/socket.h>
27 #include <linux/sockios.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 <linux/slab.h>
34 #include <asm/uaccess.h>
35 #include <linux/init.h>
36 #include <linux/netfilter_ipv4.h>
37 #include <linux/if_ether.h>
43 #include <net/protocol.h>
44 #include <net/transp_v6.h>
45 #include <net/ip6_fib.h>
46 #include <net/ip6_route.h>
47 #include <net/ndisc.h>
48 #include <net/addrconf.h>
52 #include <net/ip_tunnels.h>
53 #include <net/inet_ecn.h>
55 #include <net/dsfield.h>
56 #include <net/net_namespace.h>
57 #include <net/netns/generic.h>
60 This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
62 For comments look at net/ipv4/ip_gre.c --ANK
66 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
68 static bool log_ecn_error
= true;
69 module_param(log_ecn_error
, bool, 0644);
70 MODULE_PARM_DESC(log_ecn_error
, "Log packets received with corrupted ECN");
72 static int ipip6_tunnel_init(struct net_device
*dev
);
73 static void ipip6_tunnel_setup(struct net_device
*dev
);
74 static void ipip6_dev_free(struct net_device
*dev
);
75 static bool check_6rd(struct ip_tunnel
*tunnel
, const struct in6_addr
*v6dst
,
77 static struct rtnl_link_ops sit_link_ops __read_mostly
;
79 static int sit_net_id __read_mostly
;
81 struct ip_tunnel __rcu
*tunnels_r_l
[HASH_SIZE
];
82 struct ip_tunnel __rcu
*tunnels_r
[HASH_SIZE
];
83 struct ip_tunnel __rcu
*tunnels_l
[HASH_SIZE
];
84 struct ip_tunnel __rcu
*tunnels_wc
[1];
85 struct ip_tunnel __rcu
**tunnels
[4];
87 struct net_device
*fb_tunnel_dev
;
91 * Must be invoked with rcu_read_lock
93 static struct ip_tunnel
*ipip6_tunnel_lookup(struct net
*net
,
94 struct net_device
*dev
, __be32 remote
, __be32 local
)
96 unsigned int h0
= HASH(remote
);
97 unsigned int h1
= HASH(local
);
99 struct sit_net
*sitn
= net_generic(net
, sit_net_id
);
101 for_each_ip_tunnel_rcu(t
, sitn
->tunnels_r_l
[h0
^ h1
]) {
102 if (local
== t
->parms
.iph
.saddr
&&
103 remote
== t
->parms
.iph
.daddr
&&
104 (!dev
|| !t
->parms
.link
|| dev
->iflink
== t
->parms
.link
) &&
105 (t
->dev
->flags
& IFF_UP
))
108 for_each_ip_tunnel_rcu(t
, sitn
->tunnels_r
[h0
]) {
109 if (remote
== t
->parms
.iph
.daddr
&&
110 (!dev
|| !t
->parms
.link
|| dev
->iflink
== t
->parms
.link
) &&
111 (t
->dev
->flags
& IFF_UP
))
114 for_each_ip_tunnel_rcu(t
, sitn
->tunnels_l
[h1
]) {
115 if (local
== t
->parms
.iph
.saddr
&&
116 (!dev
|| !t
->parms
.link
|| dev
->iflink
== t
->parms
.link
) &&
117 (t
->dev
->flags
& IFF_UP
))
120 t
= rcu_dereference(sitn
->tunnels_wc
[0]);
121 if ((t
!= NULL
) && (t
->dev
->flags
& IFF_UP
))
126 static struct ip_tunnel __rcu
**__ipip6_bucket(struct sit_net
*sitn
,
127 struct ip_tunnel_parm
*parms
)
129 __be32 remote
= parms
->iph
.daddr
;
130 __be32 local
= parms
->iph
.saddr
;
142 return &sitn
->tunnels
[prio
][h
];
145 static inline struct ip_tunnel __rcu
**ipip6_bucket(struct sit_net
*sitn
,
148 return __ipip6_bucket(sitn
, &t
->parms
);
151 static void ipip6_tunnel_unlink(struct sit_net
*sitn
, struct ip_tunnel
*t
)
153 struct ip_tunnel __rcu
**tp
;
154 struct ip_tunnel
*iter
;
156 for (tp
= ipip6_bucket(sitn
, t
);
157 (iter
= rtnl_dereference(*tp
)) != NULL
;
160 rcu_assign_pointer(*tp
, t
->next
);
166 static void ipip6_tunnel_link(struct sit_net
*sitn
, struct ip_tunnel
*t
)
168 struct ip_tunnel __rcu
**tp
= ipip6_bucket(sitn
, t
);
170 rcu_assign_pointer(t
->next
, rtnl_dereference(*tp
));
171 rcu_assign_pointer(*tp
, t
);
174 static void ipip6_tunnel_clone_6rd(struct net_device
*dev
, struct sit_net
*sitn
)
176 #ifdef CONFIG_IPV6_SIT_6RD
177 struct ip_tunnel
*t
= netdev_priv(dev
);
179 if (t
->dev
== sitn
->fb_tunnel_dev
) {
180 ipv6_addr_set(&t
->ip6rd
.prefix
, htonl(0x20020000), 0, 0, 0);
181 t
->ip6rd
.relay_prefix
= 0;
182 t
->ip6rd
.prefixlen
= 16;
183 t
->ip6rd
.relay_prefixlen
= 0;
185 struct ip_tunnel
*t0
= netdev_priv(sitn
->fb_tunnel_dev
);
186 memcpy(&t
->ip6rd
, &t0
->ip6rd
, sizeof(t
->ip6rd
));
191 static int ipip6_tunnel_create(struct net_device
*dev
)
193 struct ip_tunnel
*t
= netdev_priv(dev
);
194 struct net
*net
= dev_net(dev
);
195 struct sit_net
*sitn
= net_generic(net
, sit_net_id
);
198 err
= ipip6_tunnel_init(dev
);
201 ipip6_tunnel_clone_6rd(dev
, sitn
);
203 if ((__force u16
)t
->parms
.i_flags
& SIT_ISATAP
)
204 dev
->priv_flags
|= IFF_ISATAP
;
206 err
= register_netdevice(dev
);
210 strcpy(t
->parms
.name
, dev
->name
);
211 dev
->rtnl_link_ops
= &sit_link_ops
;
215 ipip6_tunnel_link(sitn
, t
);
222 static struct ip_tunnel
*ipip6_tunnel_locate(struct net
*net
,
223 struct ip_tunnel_parm
*parms
, int create
)
225 __be32 remote
= parms
->iph
.daddr
;
226 __be32 local
= parms
->iph
.saddr
;
227 struct ip_tunnel
*t
, *nt
;
228 struct ip_tunnel __rcu
**tp
;
229 struct net_device
*dev
;
231 struct sit_net
*sitn
= net_generic(net
, sit_net_id
);
233 for (tp
= __ipip6_bucket(sitn
, parms
);
234 (t
= rtnl_dereference(*tp
)) != NULL
;
236 if (local
== t
->parms
.iph
.saddr
&&
237 remote
== t
->parms
.iph
.daddr
&&
238 parms
->link
== t
->parms
.link
) {
249 strlcpy(name
, parms
->name
, IFNAMSIZ
);
251 strcpy(name
, "sit%d");
253 dev
= alloc_netdev(sizeof(*t
), name
, ipip6_tunnel_setup
);
257 dev_net_set(dev
, net
);
259 nt
= netdev_priv(dev
);
262 if (ipip6_tunnel_create(dev
) < 0)
273 #define for_each_prl_rcu(start) \
274 for (prl = rcu_dereference(start); \
276 prl = rcu_dereference(prl->next))
278 static struct ip_tunnel_prl_entry
*
279 __ipip6_tunnel_locate_prl(struct ip_tunnel
*t
, __be32 addr
)
281 struct ip_tunnel_prl_entry
*prl
;
283 for_each_prl_rcu(t
->prl
)
284 if (prl
->addr
== addr
)
290 static int ipip6_tunnel_get_prl(struct ip_tunnel
*t
,
291 struct ip_tunnel_prl __user
*a
)
293 struct ip_tunnel_prl kprl
, *kp
;
294 struct ip_tunnel_prl_entry
*prl
;
295 unsigned int cmax
, c
= 0, ca
, len
;
298 if (copy_from_user(&kprl
, a
, sizeof(kprl
)))
300 cmax
= kprl
.datalen
/ sizeof(kprl
);
301 if (cmax
> 1 && kprl
.addr
!= htonl(INADDR_ANY
))
304 /* For simple GET or for root users,
305 * we try harder to allocate.
307 kp
= (cmax
<= 1 || capable(CAP_NET_ADMIN
)) ?
308 kcalloc(cmax
, sizeof(*kp
), GFP_KERNEL
) :
313 ca
= t
->prl_count
< cmax
? t
->prl_count
: cmax
;
316 /* We don't try hard to allocate much memory for
318 * For root users, retry allocating enough memory for
321 kp
= kcalloc(ca
, sizeof(*kp
), GFP_ATOMIC
);
329 for_each_prl_rcu(t
->prl
) {
332 if (kprl
.addr
!= htonl(INADDR_ANY
) && prl
->addr
!= kprl
.addr
)
334 kp
[c
].addr
= prl
->addr
;
335 kp
[c
].flags
= prl
->flags
;
337 if (kprl
.addr
!= htonl(INADDR_ANY
))
343 len
= sizeof(*kp
) * c
;
345 if ((len
&& copy_to_user(a
+ 1, kp
, len
)) || put_user(len
, &a
->datalen
))
354 ipip6_tunnel_add_prl(struct ip_tunnel
*t
, struct ip_tunnel_prl
*a
, int chg
)
356 struct ip_tunnel_prl_entry
*p
;
359 if (a
->addr
== htonl(INADDR_ANY
))
364 for (p
= rtnl_dereference(t
->prl
); p
; p
= rtnl_dereference(p
->next
)) {
365 if (p
->addr
== a
->addr
) {
380 p
= kzalloc(sizeof(struct ip_tunnel_prl_entry
), GFP_KERNEL
);
390 rcu_assign_pointer(t
->prl
, p
);
395 static void prl_list_destroy_rcu(struct rcu_head
*head
)
397 struct ip_tunnel_prl_entry
*p
, *n
;
399 p
= container_of(head
, struct ip_tunnel_prl_entry
, rcu_head
);
401 n
= rcu_dereference_protected(p
->next
, 1);
408 ipip6_tunnel_del_prl(struct ip_tunnel
*t
, struct ip_tunnel_prl
*a
)
410 struct ip_tunnel_prl_entry
*x
;
411 struct ip_tunnel_prl_entry __rcu
**p
;
416 if (a
&& a
->addr
!= htonl(INADDR_ANY
)) {
418 (x
= rtnl_dereference(*p
)) != NULL
;
420 if (x
->addr
== a
->addr
) {
422 kfree_rcu(x
, rcu_head
);
429 x
= rtnl_dereference(t
->prl
);
432 call_rcu(&x
->rcu_head
, prl_list_destroy_rcu
);
441 isatap_chksrc(struct sk_buff
*skb
, const struct iphdr
*iph
, struct ip_tunnel
*t
)
443 struct ip_tunnel_prl_entry
*p
;
447 p
= __ipip6_tunnel_locate_prl(t
, iph
->saddr
);
449 if (p
->flags
& PRL_DEFAULT
)
450 skb
->ndisc_nodetype
= NDISC_NODETYPE_DEFAULT
;
452 skb
->ndisc_nodetype
= NDISC_NODETYPE_NODEFAULT
;
454 const struct in6_addr
*addr6
= &ipv6_hdr(skb
)->saddr
;
456 if (ipv6_addr_is_isatap(addr6
) &&
457 (addr6
->s6_addr32
[3] == iph
->saddr
) &&
458 ipv6_chk_prefix(addr6
, t
->dev
))
459 skb
->ndisc_nodetype
= NDISC_NODETYPE_HOST
;
467 static void ipip6_tunnel_uninit(struct net_device
*dev
)
469 struct net
*net
= dev_net(dev
);
470 struct sit_net
*sitn
= net_generic(net
, sit_net_id
);
472 if (dev
== sitn
->fb_tunnel_dev
) {
473 RCU_INIT_POINTER(sitn
->tunnels_wc
[0], NULL
);
475 ipip6_tunnel_unlink(sitn
, netdev_priv(dev
));
476 ipip6_tunnel_del_prl(netdev_priv(dev
), NULL
);
482 static int ipip6_err(struct sk_buff
*skb
, u32 info
)
485 /* All the routers (except for Linux) return only
486 8 bytes of packet payload. It means, that precise relaying of
487 ICMP in the real Internet is absolutely infeasible.
489 const struct iphdr
*iph
= (const struct iphdr
*)skb
->data
;
490 const int type
= icmp_hdr(skb
)->type
;
491 const int code
= icmp_hdr(skb
)->code
;
497 case ICMP_PARAMETERPROB
:
500 case ICMP_DEST_UNREACH
:
503 case ICMP_PORT_UNREACH
:
504 /* Impossible event. */
507 /* All others are translated to HOST_UNREACH.
508 rfc2003 contains "deep thoughts" about NET_UNREACH,
509 I believe they are just ether pollution. --ANK
514 case ICMP_TIME_EXCEEDED
:
515 if (code
!= ICMP_EXC_TTL
)
524 t
= ipip6_tunnel_lookup(dev_net(skb
->dev
),
531 if (type
== ICMP_DEST_UNREACH
&& code
== ICMP_FRAG_NEEDED
) {
532 ipv4_update_pmtu(skb
, dev_net(skb
->dev
), info
,
533 t
->dev
->ifindex
, 0, IPPROTO_IPV6
, 0);
537 if (type
== ICMP_REDIRECT
) {
538 ipv4_redirect(skb
, dev_net(skb
->dev
), t
->dev
->ifindex
, 0,
544 if (t
->parms
.iph
.daddr
== 0)
548 if (t
->parms
.iph
.ttl
== 0 && type
== ICMP_TIME_EXCEEDED
)
551 if (time_before(jiffies
, t
->err_time
+ IPTUNNEL_ERR_TIMEO
))
555 t
->err_time
= jiffies
;
560 static inline bool is_spoofed_6rd(struct ip_tunnel
*tunnel
, const __be32 v4addr
,
561 const struct in6_addr
*v6addr
)
564 if (check_6rd(tunnel
, v6addr
, &v4embed
) && v4addr
!= v4embed
)
569 static int ipip6_rcv(struct sk_buff
*skb
)
571 const struct iphdr
*iph
= ip_hdr(skb
);
572 struct ip_tunnel
*tunnel
;
575 tunnel
= ipip6_tunnel_lookup(dev_net(skb
->dev
), skb
->dev
,
576 iph
->saddr
, iph
->daddr
);
577 if (tunnel
!= NULL
) {
578 struct pcpu_tstats
*tstats
;
581 skb
->mac_header
= skb
->network_header
;
582 skb_reset_network_header(skb
);
583 IPCB(skb
)->flags
= 0;
584 skb
->protocol
= htons(ETH_P_IPV6
);
585 skb
->pkt_type
= PACKET_HOST
;
587 if (tunnel
->dev
->priv_flags
& IFF_ISATAP
) {
588 if (!isatap_chksrc(skb
, iph
, tunnel
)) {
589 tunnel
->dev
->stats
.rx_errors
++;
593 if (is_spoofed_6rd(tunnel
, iph
->saddr
,
594 &ipv6_hdr(skb
)->saddr
) ||
595 is_spoofed_6rd(tunnel
, iph
->daddr
,
596 &ipv6_hdr(skb
)->daddr
)) {
597 tunnel
->dev
->stats
.rx_errors
++;
602 __skb_tunnel_rx(skb
, tunnel
->dev
);
604 err
= IP_ECN_decapsulate(iph
, skb
);
607 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
608 &iph
->saddr
, iph
->tos
);
610 ++tunnel
->dev
->stats
.rx_frame_errors
;
611 ++tunnel
->dev
->stats
.rx_errors
;
616 tstats
= this_cpu_ptr(tunnel
->dev
->tstats
);
617 tstats
->rx_packets
++;
618 tstats
->rx_bytes
+= skb
->len
;
625 /* no tunnel matched, let upstream know, ipsec may handle it */
633 * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
634 * stores the embedded IPv4 address in v4dst and returns true.
636 static bool check_6rd(struct ip_tunnel
*tunnel
, const struct in6_addr
*v6dst
,
639 #ifdef CONFIG_IPV6_SIT_6RD
640 if (ipv6_prefix_equal(v6dst
, &tunnel
->ip6rd
.prefix
,
641 tunnel
->ip6rd
.prefixlen
)) {
642 unsigned int pbw0
, pbi0
;
646 pbw0
= tunnel
->ip6rd
.prefixlen
>> 5;
647 pbi0
= tunnel
->ip6rd
.prefixlen
& 0x1f;
649 d
= (ntohl(v6dst
->s6_addr32
[pbw0
]) << pbi0
) >>
650 tunnel
->ip6rd
.relay_prefixlen
;
652 pbi1
= pbi0
- tunnel
->ip6rd
.relay_prefixlen
;
654 d
|= ntohl(v6dst
->s6_addr32
[pbw0
+ 1]) >>
657 *v4dst
= tunnel
->ip6rd
.relay_prefix
| htonl(d
);
661 if (v6dst
->s6_addr16
[0] == htons(0x2002)) {
662 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
663 memcpy(v4dst
, &v6dst
->s6_addr16
[1], 4);
670 static inline __be32
try_6rd(struct ip_tunnel
*tunnel
,
671 const struct in6_addr
*v6dst
)
674 check_6rd(tunnel
, v6dst
, &dst
);
679 * This function assumes it is being called from dev_queue_xmit()
680 * and that skb is filled properly by that function.
683 static netdev_tx_t
ipip6_tunnel_xmit(struct sk_buff
*skb
,
684 struct net_device
*dev
)
686 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
687 const struct iphdr
*tiph
= &tunnel
->parms
.iph
;
688 const struct ipv6hdr
*iph6
= ipv6_hdr(skb
);
689 u8 tos
= tunnel
->parms
.iph
.tos
;
690 __be16 df
= tiph
->frag_off
;
691 struct rtable
*rt
; /* Route to the other host */
692 struct net_device
*tdev
; /* Device to other host */
693 struct iphdr
*iph
; /* Our new IP header */
694 unsigned int max_headroom
; /* The extra header space needed */
695 __be32 dst
= tiph
->daddr
;
698 const struct in6_addr
*addr6
;
701 if (skb
->protocol
!= htons(ETH_P_IPV6
))
705 tos
= ipv6_get_dsfield(iph6
);
707 /* ISATAP (RFC4214) - must come before 6to4 */
708 if (dev
->priv_flags
& IFF_ISATAP
) {
709 struct neighbour
*neigh
= NULL
;
710 bool do_tx_error
= false;
713 neigh
= dst_neigh_lookup(skb_dst(skb
), &iph6
->daddr
);
716 net_dbg_ratelimited("sit: nexthop == NULL\n");
720 addr6
= (const struct in6_addr
*)&neigh
->primary_key
;
721 addr_type
= ipv6_addr_type(addr6
);
723 if ((addr_type
& IPV6_ADDR_UNICAST
) &&
724 ipv6_addr_is_isatap(addr6
))
725 dst
= addr6
->s6_addr32
[3];
729 neigh_release(neigh
);
735 dst
= try_6rd(tunnel
, &iph6
->daddr
);
738 struct neighbour
*neigh
= NULL
;
739 bool do_tx_error
= false;
742 neigh
= dst_neigh_lookup(skb_dst(skb
), &iph6
->daddr
);
745 net_dbg_ratelimited("sit: nexthop == NULL\n");
749 addr6
= (const struct in6_addr
*)&neigh
->primary_key
;
750 addr_type
= ipv6_addr_type(addr6
);
752 if (addr_type
== IPV6_ADDR_ANY
) {
753 addr6
= &ipv6_hdr(skb
)->daddr
;
754 addr_type
= ipv6_addr_type(addr6
);
757 if ((addr_type
& IPV6_ADDR_COMPATv4
) != 0)
758 dst
= addr6
->s6_addr32
[3];
762 neigh_release(neigh
);
767 rt
= ip_route_output_ports(dev_net(dev
), &fl4
, NULL
,
770 IPPROTO_IPV6
, RT_TOS(tos
),
773 dev
->stats
.tx_carrier_errors
++;
776 if (rt
->rt_type
!= RTN_UNICAST
) {
778 dev
->stats
.tx_carrier_errors
++;
785 dev
->stats
.collisions
++;
790 mtu
= dst_mtu(&rt
->dst
) - sizeof(struct iphdr
);
793 dev
->stats
.collisions
++;
798 if (mtu
< IPV6_MIN_MTU
) {
803 if (tunnel
->parms
.iph
.daddr
&& skb_dst(skb
))
804 skb_dst(skb
)->ops
->update_pmtu(skb_dst(skb
), NULL
, skb
, mtu
);
806 if (skb
->len
> mtu
) {
807 icmpv6_send(skb
, ICMPV6_PKT_TOOBIG
, 0, mtu
);
813 if (tunnel
->err_count
> 0) {
814 if (time_before(jiffies
,
815 tunnel
->err_time
+ IPTUNNEL_ERR_TIMEO
)) {
817 dst_link_failure(skb
);
819 tunnel
->err_count
= 0;
823 * Okay, now see if we can stuff it in the buffer as-is.
825 max_headroom
= LL_RESERVED_SPACE(tdev
)+sizeof(struct iphdr
);
827 if (skb_headroom(skb
) < max_headroom
|| skb_shared(skb
) ||
828 (skb_cloned(skb
) && !skb_clone_writable(skb
, 0))) {
829 struct sk_buff
*new_skb
= skb_realloc_headroom(skb
, max_headroom
);
832 dev
->stats
.tx_dropped
++;
837 skb_set_owner_w(new_skb
, skb
->sk
);
840 iph6
= ipv6_hdr(skb
);
843 skb
->transport_header
= skb
->network_header
;
844 skb_push(skb
, sizeof(struct iphdr
));
845 skb_reset_network_header(skb
);
846 memset(&(IPCB(skb
)->opt
), 0, sizeof(IPCB(skb
)->opt
));
847 IPCB(skb
)->flags
= 0;
849 skb_dst_set(skb
, &rt
->dst
);
852 * Push down and install the IPIP header.
857 iph
->ihl
= sizeof(struct iphdr
)>>2;
859 iph
->protocol
= IPPROTO_IPV6
;
860 iph
->tos
= INET_ECN_encapsulate(tos
, ipv6_get_dsfield(iph6
));
861 iph
->daddr
= fl4
.daddr
;
862 iph
->saddr
= fl4
.saddr
;
864 if ((iph
->ttl
= tiph
->ttl
) == 0)
865 iph
->ttl
= iph6
->hop_limit
;
867 skb
->ip_summed
= CHECKSUM_NONE
;
868 ip_select_ident(iph
, skb_dst(skb
), NULL
);
869 iptunnel_xmit(skb
, dev
);
873 dst_link_failure(skb
);
875 dev
->stats
.tx_errors
++;
880 static void ipip6_tunnel_bind_dev(struct net_device
*dev
)
882 struct net_device
*tdev
= NULL
;
883 struct ip_tunnel
*tunnel
;
884 const struct iphdr
*iph
;
887 tunnel
= netdev_priv(dev
);
888 iph
= &tunnel
->parms
.iph
;
891 struct rtable
*rt
= ip_route_output_ports(dev_net(dev
), &fl4
, NULL
,
892 iph
->daddr
, iph
->saddr
,
902 dev
->flags
|= IFF_POINTOPOINT
;
905 if (!tdev
&& tunnel
->parms
.link
)
906 tdev
= __dev_get_by_index(dev_net(dev
), tunnel
->parms
.link
);
909 dev
->hard_header_len
= tdev
->hard_header_len
+ sizeof(struct iphdr
);
910 dev
->mtu
= tdev
->mtu
- sizeof(struct iphdr
);
911 if (dev
->mtu
< IPV6_MIN_MTU
)
912 dev
->mtu
= IPV6_MIN_MTU
;
914 dev
->iflink
= tunnel
->parms
.link
;
917 static void ipip6_tunnel_update(struct ip_tunnel
*t
, struct ip_tunnel_parm
*p
)
919 struct net
*net
= dev_net(t
->dev
);
920 struct sit_net
*sitn
= net_generic(net
, sit_net_id
);
922 ipip6_tunnel_unlink(sitn
, t
);
924 t
->parms
.iph
.saddr
= p
->iph
.saddr
;
925 t
->parms
.iph
.daddr
= p
->iph
.daddr
;
926 memcpy(t
->dev
->dev_addr
, &p
->iph
.saddr
, 4);
927 memcpy(t
->dev
->broadcast
, &p
->iph
.daddr
, 4);
928 ipip6_tunnel_link(sitn
, t
);
929 t
->parms
.iph
.ttl
= p
->iph
.ttl
;
930 t
->parms
.iph
.tos
= p
->iph
.tos
;
931 if (t
->parms
.link
!= p
->link
) {
932 t
->parms
.link
= p
->link
;
933 ipip6_tunnel_bind_dev(t
->dev
);
935 netdev_state_change(t
->dev
);
938 #ifdef CONFIG_IPV6_SIT_6RD
939 static int ipip6_tunnel_update_6rd(struct ip_tunnel
*t
,
940 struct ip_tunnel_6rd
*ip6rd
)
942 struct in6_addr prefix
;
945 if (ip6rd
->relay_prefixlen
> 32 ||
946 ip6rd
->prefixlen
+ (32 - ip6rd
->relay_prefixlen
) > 64)
949 ipv6_addr_prefix(&prefix
, &ip6rd
->prefix
, ip6rd
->prefixlen
);
950 if (!ipv6_addr_equal(&prefix
, &ip6rd
->prefix
))
952 if (ip6rd
->relay_prefixlen
)
953 relay_prefix
= ip6rd
->relay_prefix
&
954 htonl(0xffffffffUL
<<
955 (32 - ip6rd
->relay_prefixlen
));
958 if (relay_prefix
!= ip6rd
->relay_prefix
)
961 t
->ip6rd
.prefix
= prefix
;
962 t
->ip6rd
.relay_prefix
= relay_prefix
;
963 t
->ip6rd
.prefixlen
= ip6rd
->prefixlen
;
964 t
->ip6rd
.relay_prefixlen
= ip6rd
->relay_prefixlen
;
965 netdev_state_change(t
->dev
);
971 ipip6_tunnel_ioctl (struct net_device
*dev
, struct ifreq
*ifr
, int cmd
)
974 struct ip_tunnel_parm p
;
975 struct ip_tunnel_prl prl
;
977 struct net
*net
= dev_net(dev
);
978 struct sit_net
*sitn
= net_generic(net
, sit_net_id
);
979 #ifdef CONFIG_IPV6_SIT_6RD
980 struct ip_tunnel_6rd ip6rd
;
985 #ifdef CONFIG_IPV6_SIT_6RD
989 if (dev
== sitn
->fb_tunnel_dev
) {
990 if (copy_from_user(&p
, ifr
->ifr_ifru
.ifru_data
, sizeof(p
))) {
994 t
= ipip6_tunnel_locate(net
, &p
, 0);
997 t
= netdev_priv(dev
);
1000 if (cmd
== SIOCGETTUNNEL
) {
1001 memcpy(&p
, &t
->parms
, sizeof(p
));
1002 if (copy_to_user(ifr
->ifr_ifru
.ifru_data
, &p
,
1005 #ifdef CONFIG_IPV6_SIT_6RD
1007 ip6rd
.prefix
= t
->ip6rd
.prefix
;
1008 ip6rd
.relay_prefix
= t
->ip6rd
.relay_prefix
;
1009 ip6rd
.prefixlen
= t
->ip6rd
.prefixlen
;
1010 ip6rd
.relay_prefixlen
= t
->ip6rd
.relay_prefixlen
;
1011 if (copy_to_user(ifr
->ifr_ifru
.ifru_data
, &ip6rd
,
1022 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
1026 if (copy_from_user(&p
, ifr
->ifr_ifru
.ifru_data
, sizeof(p
)))
1030 if (p
.iph
.version
!= 4 || p
.iph
.protocol
!= IPPROTO_IPV6
||
1031 p
.iph
.ihl
!= 5 || (p
.iph
.frag_off
&htons(~IP_DF
)))
1034 p
.iph
.frag_off
|= htons(IP_DF
);
1036 t
= ipip6_tunnel_locate(net
, &p
, cmd
== SIOCADDTUNNEL
);
1038 if (dev
!= sitn
->fb_tunnel_dev
&& cmd
== SIOCCHGTUNNEL
) {
1040 if (t
->dev
!= dev
) {
1045 if (((dev
->flags
&IFF_POINTOPOINT
) && !p
.iph
.daddr
) ||
1046 (!(dev
->flags
&IFF_POINTOPOINT
) && p
.iph
.daddr
)) {
1050 t
= netdev_priv(dev
);
1053 ipip6_tunnel_update(t
, &p
);
1058 if (copy_to_user(ifr
->ifr_ifru
.ifru_data
, &t
->parms
, sizeof(p
)))
1061 err
= (cmd
== SIOCADDTUNNEL
? -ENOBUFS
: -ENOENT
);
1066 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
1069 if (dev
== sitn
->fb_tunnel_dev
) {
1071 if (copy_from_user(&p
, ifr
->ifr_ifru
.ifru_data
, sizeof(p
)))
1074 if ((t
= ipip6_tunnel_locate(net
, &p
, 0)) == NULL
)
1077 if (t
== netdev_priv(sitn
->fb_tunnel_dev
))
1081 unregister_netdevice(dev
);
1087 if (dev
== sitn
->fb_tunnel_dev
)
1090 if (!(t
= netdev_priv(dev
)))
1092 err
= ipip6_tunnel_get_prl(t
, ifr
->ifr_ifru
.ifru_data
);
1099 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
1102 if (dev
== sitn
->fb_tunnel_dev
)
1105 if (copy_from_user(&prl
, ifr
->ifr_ifru
.ifru_data
, sizeof(prl
)))
1108 if (!(t
= netdev_priv(dev
)))
1113 err
= ipip6_tunnel_del_prl(t
, &prl
);
1117 err
= ipip6_tunnel_add_prl(t
, &prl
, cmd
== SIOCCHGPRL
);
1120 netdev_state_change(dev
);
1123 #ifdef CONFIG_IPV6_SIT_6RD
1128 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
1132 if (copy_from_user(&ip6rd
, ifr
->ifr_ifru
.ifru_data
,
1136 t
= netdev_priv(dev
);
1138 if (cmd
!= SIOCDEL6RD
) {
1139 err
= ipip6_tunnel_update_6rd(t
, &ip6rd
);
1143 ipip6_tunnel_clone_6rd(dev
, sitn
);
1157 static int ipip6_tunnel_change_mtu(struct net_device
*dev
, int new_mtu
)
1159 if (new_mtu
< IPV6_MIN_MTU
|| new_mtu
> 0xFFF8 - sizeof(struct iphdr
))
1165 static const struct net_device_ops ipip6_netdev_ops
= {
1166 .ndo_uninit
= ipip6_tunnel_uninit
,
1167 .ndo_start_xmit
= ipip6_tunnel_xmit
,
1168 .ndo_do_ioctl
= ipip6_tunnel_ioctl
,
1169 .ndo_change_mtu
= ipip6_tunnel_change_mtu
,
1170 .ndo_get_stats64
= ip_tunnel_get_stats64
,
1173 static void ipip6_dev_free(struct net_device
*dev
)
1175 free_percpu(dev
->tstats
);
1179 static void ipip6_tunnel_setup(struct net_device
*dev
)
1181 dev
->netdev_ops
= &ipip6_netdev_ops
;
1182 dev
->destructor
= ipip6_dev_free
;
1184 dev
->type
= ARPHRD_SIT
;
1185 dev
->hard_header_len
= LL_MAX_HEADER
+ sizeof(struct iphdr
);
1186 dev
->mtu
= ETH_DATA_LEN
- sizeof(struct iphdr
);
1187 dev
->flags
= IFF_NOARP
;
1188 dev
->priv_flags
&= ~IFF_XMIT_DST_RELEASE
;
1191 dev
->features
|= NETIF_F_NETNS_LOCAL
;
1192 dev
->features
|= NETIF_F_LLTX
;
1195 static int ipip6_tunnel_init(struct net_device
*dev
)
1197 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
1201 memcpy(dev
->dev_addr
, &tunnel
->parms
.iph
.saddr
, 4);
1202 memcpy(dev
->broadcast
, &tunnel
->parms
.iph
.daddr
, 4);
1204 ipip6_tunnel_bind_dev(dev
);
1205 dev
->tstats
= alloc_percpu(struct pcpu_tstats
);
1212 static int __net_init
ipip6_fb_tunnel_init(struct net_device
*dev
)
1214 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
1215 struct iphdr
*iph
= &tunnel
->parms
.iph
;
1216 struct net
*net
= dev_net(dev
);
1217 struct sit_net
*sitn
= net_generic(net
, sit_net_id
);
1220 strcpy(tunnel
->parms
.name
, dev
->name
);
1223 iph
->protocol
= IPPROTO_IPV6
;
1227 dev
->tstats
= alloc_percpu(struct pcpu_tstats
);
1231 rcu_assign_pointer(sitn
->tunnels_wc
[0], tunnel
);
1235 static void ipip6_netlink_parms(struct nlattr
*data
[],
1236 struct ip_tunnel_parm
*parms
)
1238 memset(parms
, 0, sizeof(*parms
));
1240 parms
->iph
.version
= 4;
1241 parms
->iph
.protocol
= IPPROTO_IPV6
;
1243 parms
->iph
.ttl
= 64;
1248 if (data
[IFLA_IPTUN_LINK
])
1249 parms
->link
= nla_get_u32(data
[IFLA_IPTUN_LINK
]);
1251 if (data
[IFLA_IPTUN_LOCAL
])
1252 parms
->iph
.saddr
= nla_get_be32(data
[IFLA_IPTUN_LOCAL
]);
1254 if (data
[IFLA_IPTUN_REMOTE
])
1255 parms
->iph
.daddr
= nla_get_be32(data
[IFLA_IPTUN_REMOTE
]);
1257 if (data
[IFLA_IPTUN_TTL
]) {
1258 parms
->iph
.ttl
= nla_get_u8(data
[IFLA_IPTUN_TTL
]);
1260 parms
->iph
.frag_off
= htons(IP_DF
);
1263 if (data
[IFLA_IPTUN_TOS
])
1264 parms
->iph
.tos
= nla_get_u8(data
[IFLA_IPTUN_TOS
]);
1266 if (!data
[IFLA_IPTUN_PMTUDISC
] || nla_get_u8(data
[IFLA_IPTUN_PMTUDISC
]))
1267 parms
->iph
.frag_off
= htons(IP_DF
);
1269 if (data
[IFLA_IPTUN_FLAGS
])
1270 parms
->i_flags
= nla_get_be16(data
[IFLA_IPTUN_FLAGS
]);
1273 #ifdef CONFIG_IPV6_SIT_6RD
1274 /* This function returns true when 6RD attributes are present in the nl msg */
1275 static bool ipip6_netlink_6rd_parms(struct nlattr
*data
[],
1276 struct ip_tunnel_6rd
*ip6rd
)
1279 memset(ip6rd
, 0, sizeof(*ip6rd
));
1284 if (data
[IFLA_IPTUN_6RD_PREFIX
]) {
1286 nla_memcpy(&ip6rd
->prefix
, data
[IFLA_IPTUN_6RD_PREFIX
],
1287 sizeof(struct in6_addr
));
1290 if (data
[IFLA_IPTUN_6RD_RELAY_PREFIX
]) {
1292 ip6rd
->relay_prefix
=
1293 nla_get_be32(data
[IFLA_IPTUN_6RD_RELAY_PREFIX
]);
1296 if (data
[IFLA_IPTUN_6RD_PREFIXLEN
]) {
1298 ip6rd
->prefixlen
= nla_get_u16(data
[IFLA_IPTUN_6RD_PREFIXLEN
]);
1301 if (data
[IFLA_IPTUN_6RD_RELAY_PREFIXLEN
]) {
1303 ip6rd
->relay_prefixlen
=
1304 nla_get_u16(data
[IFLA_IPTUN_6RD_RELAY_PREFIXLEN
]);
1311 static int ipip6_newlink(struct net
*src_net
, struct net_device
*dev
,
1312 struct nlattr
*tb
[], struct nlattr
*data
[])
1314 struct net
*net
= dev_net(dev
);
1315 struct ip_tunnel
*nt
;
1316 #ifdef CONFIG_IPV6_SIT_6RD
1317 struct ip_tunnel_6rd ip6rd
;
1321 nt
= netdev_priv(dev
);
1322 ipip6_netlink_parms(data
, &nt
->parms
);
1324 if (ipip6_tunnel_locate(net
, &nt
->parms
, 0))
1327 err
= ipip6_tunnel_create(dev
);
1331 #ifdef CONFIG_IPV6_SIT_6RD
1332 if (ipip6_netlink_6rd_parms(data
, &ip6rd
))
1333 err
= ipip6_tunnel_update_6rd(nt
, &ip6rd
);
1339 static int ipip6_changelink(struct net_device
*dev
, struct nlattr
*tb
[],
1340 struct nlattr
*data
[])
1342 struct ip_tunnel
*t
;
1343 struct ip_tunnel_parm p
;
1344 struct net
*net
= dev_net(dev
);
1345 struct sit_net
*sitn
= net_generic(net
, sit_net_id
);
1346 #ifdef CONFIG_IPV6_SIT_6RD
1347 struct ip_tunnel_6rd ip6rd
;
1350 if (dev
== sitn
->fb_tunnel_dev
)
1353 ipip6_netlink_parms(data
, &p
);
1355 if (((dev
->flags
& IFF_POINTOPOINT
) && !p
.iph
.daddr
) ||
1356 (!(dev
->flags
& IFF_POINTOPOINT
) && p
.iph
.daddr
))
1359 t
= ipip6_tunnel_locate(net
, &p
, 0);
1365 t
= netdev_priv(dev
);
1367 ipip6_tunnel_update(t
, &p
);
1369 #ifdef CONFIG_IPV6_SIT_6RD
1370 if (ipip6_netlink_6rd_parms(data
, &ip6rd
))
1371 return ipip6_tunnel_update_6rd(t
, &ip6rd
);
1377 static size_t ipip6_get_size(const struct net_device
*dev
)
1380 /* IFLA_IPTUN_LINK */
1382 /* IFLA_IPTUN_LOCAL */
1384 /* IFLA_IPTUN_REMOTE */
1386 /* IFLA_IPTUN_TTL */
1388 /* IFLA_IPTUN_TOS */
1390 /* IFLA_IPTUN_PMTUDISC */
1392 /* IFLA_IPTUN_FLAGS */
1394 #ifdef CONFIG_IPV6_SIT_6RD
1395 /* IFLA_IPTUN_6RD_PREFIX */
1396 nla_total_size(sizeof(struct in6_addr
)) +
1397 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1399 /* IFLA_IPTUN_6RD_PREFIXLEN */
1401 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1407 static int ipip6_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
1409 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
1410 struct ip_tunnel_parm
*parm
= &tunnel
->parms
;
1412 if (nla_put_u32(skb
, IFLA_IPTUN_LINK
, parm
->link
) ||
1413 nla_put_be32(skb
, IFLA_IPTUN_LOCAL
, parm
->iph
.saddr
) ||
1414 nla_put_be32(skb
, IFLA_IPTUN_REMOTE
, parm
->iph
.daddr
) ||
1415 nla_put_u8(skb
, IFLA_IPTUN_TTL
, parm
->iph
.ttl
) ||
1416 nla_put_u8(skb
, IFLA_IPTUN_TOS
, parm
->iph
.tos
) ||
1417 nla_put_u8(skb
, IFLA_IPTUN_PMTUDISC
,
1418 !!(parm
->iph
.frag_off
& htons(IP_DF
))) ||
1419 nla_put_be16(skb
, IFLA_IPTUN_FLAGS
, parm
->i_flags
))
1420 goto nla_put_failure
;
1422 #ifdef CONFIG_IPV6_SIT_6RD
1423 if (nla_put(skb
, IFLA_IPTUN_6RD_PREFIX
, sizeof(struct in6_addr
),
1424 &tunnel
->ip6rd
.prefix
) ||
1425 nla_put_be32(skb
, IFLA_IPTUN_6RD_RELAY_PREFIX
,
1426 tunnel
->ip6rd
.relay_prefix
) ||
1427 nla_put_u16(skb
, IFLA_IPTUN_6RD_PREFIXLEN
,
1428 tunnel
->ip6rd
.prefixlen
) ||
1429 nla_put_u16(skb
, IFLA_IPTUN_6RD_RELAY_PREFIXLEN
,
1430 tunnel
->ip6rd
.relay_prefixlen
))
1431 goto nla_put_failure
;
1440 static const struct nla_policy ipip6_policy
[IFLA_IPTUN_MAX
+ 1] = {
1441 [IFLA_IPTUN_LINK
] = { .type
= NLA_U32
},
1442 [IFLA_IPTUN_LOCAL
] = { .type
= NLA_U32
},
1443 [IFLA_IPTUN_REMOTE
] = { .type
= NLA_U32
},
1444 [IFLA_IPTUN_TTL
] = { .type
= NLA_U8
},
1445 [IFLA_IPTUN_TOS
] = { .type
= NLA_U8
},
1446 [IFLA_IPTUN_PMTUDISC
] = { .type
= NLA_U8
},
1447 [IFLA_IPTUN_FLAGS
] = { .type
= NLA_U16
},
1448 #ifdef CONFIG_IPV6_SIT_6RD
1449 [IFLA_IPTUN_6RD_PREFIX
] = { .len
= sizeof(struct in6_addr
) },
1450 [IFLA_IPTUN_6RD_RELAY_PREFIX
] = { .type
= NLA_U32
},
1451 [IFLA_IPTUN_6RD_PREFIXLEN
] = { .type
= NLA_U16
},
1452 [IFLA_IPTUN_6RD_RELAY_PREFIXLEN
] = { .type
= NLA_U16
},
1456 static struct rtnl_link_ops sit_link_ops __read_mostly
= {
1458 .maxtype
= IFLA_IPTUN_MAX
,
1459 .policy
= ipip6_policy
,
1460 .priv_size
= sizeof(struct ip_tunnel
),
1461 .setup
= ipip6_tunnel_setup
,
1462 .newlink
= ipip6_newlink
,
1463 .changelink
= ipip6_changelink
,
1464 .get_size
= ipip6_get_size
,
1465 .fill_info
= ipip6_fill_info
,
1468 static struct xfrm_tunnel sit_handler __read_mostly
= {
1469 .handler
= ipip6_rcv
,
1470 .err_handler
= ipip6_err
,
1474 static void __net_exit
sit_destroy_tunnels(struct sit_net
*sitn
, struct list_head
*head
)
1478 for (prio
= 1; prio
< 4; prio
++) {
1480 for (h
= 0; h
< HASH_SIZE
; h
++) {
1481 struct ip_tunnel
*t
;
1483 t
= rtnl_dereference(sitn
->tunnels
[prio
][h
]);
1485 unregister_netdevice_queue(t
->dev
, head
);
1486 t
= rtnl_dereference(t
->next
);
1492 static int __net_init
sit_init_net(struct net
*net
)
1494 struct sit_net
*sitn
= net_generic(net
, sit_net_id
);
1495 struct ip_tunnel
*t
;
1498 sitn
->tunnels
[0] = sitn
->tunnels_wc
;
1499 sitn
->tunnels
[1] = sitn
->tunnels_l
;
1500 sitn
->tunnels
[2] = sitn
->tunnels_r
;
1501 sitn
->tunnels
[3] = sitn
->tunnels_r_l
;
1503 sitn
->fb_tunnel_dev
= alloc_netdev(sizeof(struct ip_tunnel
), "sit0",
1504 ipip6_tunnel_setup
);
1505 if (!sitn
->fb_tunnel_dev
) {
1509 dev_net_set(sitn
->fb_tunnel_dev
, net
);
1511 err
= ipip6_fb_tunnel_init(sitn
->fb_tunnel_dev
);
1515 ipip6_tunnel_clone_6rd(sitn
->fb_tunnel_dev
, sitn
);
1517 if ((err
= register_netdev(sitn
->fb_tunnel_dev
)))
1520 t
= netdev_priv(sitn
->fb_tunnel_dev
);
1522 strcpy(t
->parms
.name
, sitn
->fb_tunnel_dev
->name
);
1526 dev_put(sitn
->fb_tunnel_dev
);
1528 ipip6_dev_free(sitn
->fb_tunnel_dev
);
1533 static void __net_exit
sit_exit_net(struct net
*net
)
1535 struct sit_net
*sitn
= net_generic(net
, sit_net_id
);
1539 sit_destroy_tunnels(sitn
, &list
);
1540 unregister_netdevice_queue(sitn
->fb_tunnel_dev
, &list
);
1541 unregister_netdevice_many(&list
);
1545 static struct pernet_operations sit_net_ops
= {
1546 .init
= sit_init_net
,
1547 .exit
= sit_exit_net
,
1549 .size
= sizeof(struct sit_net
),
1552 static void __exit
sit_cleanup(void)
1554 rtnl_link_unregister(&sit_link_ops
);
1555 xfrm4_tunnel_deregister(&sit_handler
, AF_INET6
);
1557 unregister_pernet_device(&sit_net_ops
);
1558 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1561 static int __init
sit_init(void)
1565 pr_info("IPv6 over IPv4 tunneling driver\n");
1567 err
= register_pernet_device(&sit_net_ops
);
1570 err
= xfrm4_tunnel_register(&sit_handler
, AF_INET6
);
1572 pr_info("%s: can't add protocol\n", __func__
);
1573 goto xfrm_tunnel_failed
;
1575 err
= rtnl_link_register(&sit_link_ops
);
1577 goto rtnl_link_failed
;
1583 xfrm4_tunnel_deregister(&sit_handler
, AF_INET6
);
1585 unregister_pernet_device(&sit_net_ops
);
1589 module_init(sit_init
);
1590 module_exit(sit_cleanup
);
1591 MODULE_LICENSE("GPL");
1592 MODULE_ALIAS_NETDEV("sit0");