2 * IPv6 tunneling device
3 * Linux INET6 implementation
6 * Ville Nuorvala <vnuorval@tcs.hut.fi>
7 * Yasuyuki Kozakai <kozakai@linux-ipv6.org>
10 * linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
14 * This program is free software; you can redistribute it and/or
15 * modify it under the terms of the GNU General Public License
16 * as published by the Free Software Foundation; either version
17 * 2 of the License, or (at your option) any later version.
21 #include <linux/module.h>
22 #include <linux/capability.h>
23 #include <linux/errno.h>
24 #include <linux/types.h>
25 #include <linux/sockios.h>
26 #include <linux/icmp.h>
30 #include <linux/if_tunnel.h>
31 #include <linux/net.h>
32 #include <linux/in6.h>
33 #include <linux/netdevice.h>
34 #include <linux/if_arp.h>
35 #include <linux/icmpv6.h>
36 #include <linux/init.h>
37 #include <linux/route.h>
38 #include <linux/rtnetlink.h>
39 #include <linux/netfilter_ipv6.h>
40 #include <linux/slab.h>
42 #include <asm/uaccess.h>
43 #include <asm/atomic.h>
48 #include <net/ip6_route.h>
49 #include <net/addrconf.h>
50 #include <net/ip6_tunnel.h>
52 #include <net/dsfield.h>
53 #include <net/inet_ecn.h>
54 #include <net/net_namespace.h>
55 #include <net/netns/generic.h>
57 MODULE_AUTHOR("Ville Nuorvala");
58 MODULE_DESCRIPTION("IPv6 tunneling device");
59 MODULE_LICENSE("GPL");
62 #define IP6_TNL_TRACE(x...) printk(KERN_DEBUG "%s:" x "\n", __func__)
64 #define IP6_TNL_TRACE(x...) do {;} while(0)
67 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
68 #define IPV6_TCLASS_SHIFT 20
72 #define HASH(addr) ((__force u32)((addr)->s6_addr32[0] ^ (addr)->s6_addr32[1] ^ \
73 (addr)->s6_addr32[2] ^ (addr)->s6_addr32[3]) & \
76 static int ip6_tnl_dev_init(struct net_device
*dev
);
77 static void ip6_tnl_dev_setup(struct net_device
*dev
);
79 static int ip6_tnl_net_id __read_mostly
;
81 /* the IPv6 tunnel fallback device */
82 struct net_device
*fb_tnl_dev
;
83 /* lists for storing tunnels in use */
84 struct ip6_tnl __rcu
*tnls_r_l
[HASH_SIZE
];
85 struct ip6_tnl __rcu
*tnls_wc
[1];
86 struct ip6_tnl __rcu
**tnls
[2];
89 /* often modified stats are per cpu, other are shared (netdev->stats) */
91 unsigned long rx_packets
;
92 unsigned long rx_bytes
;
93 unsigned long tx_packets
;
94 unsigned long tx_bytes
;
97 static struct net_device_stats
*ip6_get_stats(struct net_device
*dev
)
99 struct pcpu_tstats sum
= { 0 };
102 for_each_possible_cpu(i
) {
103 const struct pcpu_tstats
*tstats
= per_cpu_ptr(dev
->tstats
, i
);
105 sum
.rx_packets
+= tstats
->rx_packets
;
106 sum
.rx_bytes
+= tstats
->rx_bytes
;
107 sum
.tx_packets
+= tstats
->tx_packets
;
108 sum
.tx_bytes
+= tstats
->tx_bytes
;
110 dev
->stats
.rx_packets
= sum
.rx_packets
;
111 dev
->stats
.rx_bytes
= sum
.rx_bytes
;
112 dev
->stats
.tx_packets
= sum
.tx_packets
;
113 dev
->stats
.tx_bytes
= sum
.tx_bytes
;
118 * Locking : hash tables are protected by RCU and RTNL
121 static inline struct dst_entry
*ip6_tnl_dst_check(struct ip6_tnl
*t
)
123 struct dst_entry
*dst
= t
->dst_cache
;
125 if (dst
&& dst
->obsolete
&&
126 dst
->ops
->check(dst
, t
->dst_cookie
) == NULL
) {
135 static inline void ip6_tnl_dst_reset(struct ip6_tnl
*t
)
137 dst_release(t
->dst_cache
);
141 static inline void ip6_tnl_dst_store(struct ip6_tnl
*t
, struct dst_entry
*dst
)
143 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
144 t
->dst_cookie
= rt
->rt6i_node
? rt
->rt6i_node
->fn_sernum
: 0;
145 dst_release(t
->dst_cache
);
150 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
151 * @remote: the address of the tunnel exit-point
152 * @local: the address of the tunnel entry-point
155 * tunnel matching given end-points if found,
156 * else fallback tunnel if its device is up,
160 #define for_each_ip6_tunnel_rcu(start) \
161 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
163 static struct ip6_tnl
*
164 ip6_tnl_lookup(struct net
*net
, struct in6_addr
*remote
, struct in6_addr
*local
)
166 unsigned int h0
= HASH(remote
);
167 unsigned int h1
= HASH(local
);
169 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
171 for_each_ip6_tunnel_rcu(ip6n
->tnls_r_l
[h0
^ h1
]) {
172 if (ipv6_addr_equal(local
, &t
->parms
.laddr
) &&
173 ipv6_addr_equal(remote
, &t
->parms
.raddr
) &&
174 (t
->dev
->flags
& IFF_UP
))
177 t
= rcu_dereference(ip6n
->tnls_wc
[0]);
178 if (t
&& (t
->dev
->flags
& IFF_UP
))
185 * ip6_tnl_bucket - get head of list matching given tunnel parameters
186 * @p: parameters containing tunnel end-points
189 * ip6_tnl_bucket() returns the head of the list matching the
190 * &struct in6_addr entries laddr and raddr in @p.
192 * Return: head of IPv6 tunnel list
195 static struct ip6_tnl __rcu
**
196 ip6_tnl_bucket(struct ip6_tnl_net
*ip6n
, struct ip6_tnl_parm
*p
)
198 struct in6_addr
*remote
= &p
->raddr
;
199 struct in6_addr
*local
= &p
->laddr
;
203 if (!ipv6_addr_any(remote
) || !ipv6_addr_any(local
)) {
205 h
= HASH(remote
) ^ HASH(local
);
207 return &ip6n
->tnls
[prio
][h
];
211 * ip6_tnl_link - add tunnel to hash table
212 * @t: tunnel to be added
216 ip6_tnl_link(struct ip6_tnl_net
*ip6n
, struct ip6_tnl
*t
)
218 struct ip6_tnl __rcu
**tp
= ip6_tnl_bucket(ip6n
, &t
->parms
);
220 rcu_assign_pointer(t
->next
, rtnl_dereference(*tp
));
221 rcu_assign_pointer(*tp
, t
);
225 * ip6_tnl_unlink - remove tunnel from hash table
226 * @t: tunnel to be removed
230 ip6_tnl_unlink(struct ip6_tnl_net
*ip6n
, struct ip6_tnl
*t
)
232 struct ip6_tnl __rcu
**tp
;
233 struct ip6_tnl
*iter
;
235 for (tp
= ip6_tnl_bucket(ip6n
, &t
->parms
);
236 (iter
= rtnl_dereference(*tp
)) != NULL
;
239 rcu_assign_pointer(*tp
, t
->next
);
245 static void ip6_dev_free(struct net_device
*dev
)
247 free_percpu(dev
->tstats
);
252 * ip6_tnl_create() - create a new tunnel
253 * @p: tunnel parameters
254 * @pt: pointer to new tunnel
257 * Create tunnel matching given parameters.
260 * created tunnel or NULL
263 static struct ip6_tnl
*ip6_tnl_create(struct net
*net
, struct ip6_tnl_parm
*p
)
265 struct net_device
*dev
;
269 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
272 strlcpy(name
, p
->name
, IFNAMSIZ
);
274 sprintf(name
, "ip6tnl%%d");
276 dev
= alloc_netdev(sizeof (*t
), name
, ip6_tnl_dev_setup
);
280 dev_net_set(dev
, net
);
282 if (strchr(name
, '%')) {
283 if (dev_alloc_name(dev
, name
) < 0)
287 t
= netdev_priv(dev
);
289 err
= ip6_tnl_dev_init(dev
);
293 if ((err
= register_netdevice(dev
)) < 0)
297 ip6_tnl_link(ip6n
, t
);
307 * ip6_tnl_locate - find or create tunnel matching given parameters
308 * @p: tunnel parameters
309 * @create: != 0 if allowed to create new tunnel if no match found
312 * ip6_tnl_locate() first tries to locate an existing tunnel
313 * based on @parms. If this is unsuccessful, but @create is set a new
314 * tunnel device is created and registered for use.
317 * matching tunnel or NULL
320 static struct ip6_tnl
*ip6_tnl_locate(struct net
*net
,
321 struct ip6_tnl_parm
*p
, int create
)
323 struct in6_addr
*remote
= &p
->raddr
;
324 struct in6_addr
*local
= &p
->laddr
;
325 struct ip6_tnl __rcu
**tp
;
327 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
329 for (tp
= ip6_tnl_bucket(ip6n
, p
);
330 (t
= rtnl_dereference(*tp
)) != NULL
;
332 if (ipv6_addr_equal(local
, &t
->parms
.laddr
) &&
333 ipv6_addr_equal(remote
, &t
->parms
.raddr
))
338 return ip6_tnl_create(net
, p
);
342 * ip6_tnl_dev_uninit - tunnel device uninitializer
343 * @dev: the device to be destroyed
346 * ip6_tnl_dev_uninit() removes tunnel from its list
350 ip6_tnl_dev_uninit(struct net_device
*dev
)
352 struct ip6_tnl
*t
= netdev_priv(dev
);
353 struct net
*net
= dev_net(dev
);
354 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
356 if (dev
== ip6n
->fb_tnl_dev
)
357 rcu_assign_pointer(ip6n
->tnls_wc
[0], NULL
);
359 ip6_tnl_unlink(ip6n
, t
);
360 ip6_tnl_dst_reset(t
);
365 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
366 * @skb: received socket buffer
369 * 0 if none was found,
370 * else index to encapsulation limit
374 parse_tlv_tnl_enc_lim(struct sk_buff
*skb
, __u8
* raw
)
376 struct ipv6hdr
*ipv6h
= (struct ipv6hdr
*) raw
;
377 __u8 nexthdr
= ipv6h
->nexthdr
;
378 __u16 off
= sizeof (*ipv6h
);
380 while (ipv6_ext_hdr(nexthdr
) && nexthdr
!= NEXTHDR_NONE
) {
382 struct ipv6_opt_hdr
*hdr
;
383 if (raw
+ off
+ sizeof (*hdr
) > skb
->data
&&
384 !pskb_may_pull(skb
, raw
- skb
->data
+ off
+ sizeof (*hdr
)))
387 hdr
= (struct ipv6_opt_hdr
*) (raw
+ off
);
388 if (nexthdr
== NEXTHDR_FRAGMENT
) {
389 struct frag_hdr
*frag_hdr
= (struct frag_hdr
*) hdr
;
390 if (frag_hdr
->frag_off
)
393 } else if (nexthdr
== NEXTHDR_AUTH
) {
394 optlen
= (hdr
->hdrlen
+ 2) << 2;
396 optlen
= ipv6_optlen(hdr
);
398 if (nexthdr
== NEXTHDR_DEST
) {
401 struct ipv6_tlv_tnl_enc_lim
*tel
;
403 /* No more room for encapsulation limit */
404 if (i
+ sizeof (*tel
) > off
+ optlen
)
407 tel
= (struct ipv6_tlv_tnl_enc_lim
*) &raw
[i
];
408 /* return index of option if found and valid */
409 if (tel
->type
== IPV6_TLV_TNL_ENCAP_LIMIT
&&
412 /* else jump to next option */
414 i
+= tel
->length
+ 2;
419 nexthdr
= hdr
->nexthdr
;
426 * ip6_tnl_err - tunnel error handler
429 * ip6_tnl_err() should handle errors in the tunnel according
430 * to the specifications in RFC 2473.
434 ip6_tnl_err(struct sk_buff
*skb
, __u8 ipproto
, struct inet6_skb_parm
*opt
,
435 u8
*type
, u8
*code
, int *msg
, __u32
*info
, int offset
)
437 struct ipv6hdr
*ipv6h
= (struct ipv6hdr
*) skb
->data
;
440 u8 rel_type
= ICMPV6_DEST_UNREACH
;
441 u8 rel_code
= ICMPV6_ADDR_UNREACH
;
446 /* If the packet doesn't contain the original IPv6 header we are
447 in trouble since we might need the source address for further
448 processing of the error. */
451 if ((t
= ip6_tnl_lookup(dev_net(skb
->dev
), &ipv6h
->daddr
,
452 &ipv6h
->saddr
)) == NULL
)
455 if (t
->parms
.proto
!= ipproto
&& t
->parms
.proto
!= 0)
462 struct ipv6_tlv_tnl_enc_lim
*tel
;
464 case ICMPV6_DEST_UNREACH
:
467 "%s: Path to destination invalid "
468 "or inactive!\n", t
->parms
.name
);
471 case ICMPV6_TIME_EXCEED
:
472 if ((*code
) == ICMPV6_EXC_HOPLIMIT
) {
475 "%s: Too small hop limit or "
476 "routing loop in tunnel!\n",
481 case ICMPV6_PARAMPROB
:
483 if ((*code
) == ICMPV6_HDR_FIELD
)
484 teli
= parse_tlv_tnl_enc_lim(skb
, skb
->data
);
486 if (teli
&& teli
== *info
- 2) {
487 tel
= (struct ipv6_tlv_tnl_enc_lim
*) &skb
->data
[teli
];
488 if (tel
->encap_limit
== 0) {
491 "%s: Too small encapsulation "
492 "limit or routing loop in "
493 "tunnel!\n", t
->parms
.name
);
496 } else if (net_ratelimit()) {
498 "%s: Recipient unable to parse tunneled "
499 "packet!\n ", t
->parms
.name
);
502 case ICMPV6_PKT_TOOBIG
:
503 mtu
= *info
- offset
;
504 if (mtu
< IPV6_MIN_MTU
)
508 if ((len
= sizeof (*ipv6h
) + ntohs(ipv6h
->payload_len
)) > mtu
) {
509 rel_type
= ICMPV6_PKT_TOOBIG
;
528 ip4ip6_err(struct sk_buff
*skb
, struct inet6_skb_parm
*opt
,
529 u8 type
, u8 code
, int offset
, __be32 info
)
534 __u32 rel_info
= ntohl(info
);
536 struct sk_buff
*skb2
;
541 err
= ip6_tnl_err(skb
, IPPROTO_IPIP
, opt
, &rel_type
, &rel_code
,
542 &rel_msg
, &rel_info
, offset
);
550 case ICMPV6_DEST_UNREACH
:
551 if (rel_code
!= ICMPV6_ADDR_UNREACH
)
553 rel_type
= ICMP_DEST_UNREACH
;
554 rel_code
= ICMP_HOST_UNREACH
;
556 case ICMPV6_PKT_TOOBIG
:
559 rel_type
= ICMP_DEST_UNREACH
;
560 rel_code
= ICMP_FRAG_NEEDED
;
566 if (!pskb_may_pull(skb
, offset
+ sizeof(struct iphdr
)))
569 skb2
= skb_clone(skb
, GFP_ATOMIC
);
575 skb_pull(skb2
, offset
);
576 skb_reset_network_header(skb2
);
579 /* Try to guess incoming interface */
580 memset(&fl
, 0, sizeof(fl
));
581 fl
.fl4_dst
= eiph
->saddr
;
582 fl
.fl4_tos
= RT_TOS(eiph
->tos
);
583 fl
.proto
= IPPROTO_IPIP
;
584 if (ip_route_output_key(dev_net(skb
->dev
), &rt
, &fl
))
587 skb2
->dev
= rt
->dst
.dev
;
589 /* route "incoming" packet */
590 if (rt
->rt_flags
& RTCF_LOCAL
) {
593 fl
.fl4_dst
= eiph
->daddr
;
594 fl
.fl4_src
= eiph
->saddr
;
595 fl
.fl4_tos
= eiph
->tos
;
596 if (ip_route_output_key(dev_net(skb
->dev
), &rt
, &fl
) ||
597 rt
->dst
.dev
->type
!= ARPHRD_TUNNEL
) {
601 skb_dst_set(skb2
, (struct dst_entry
*)rt
);
604 if (ip_route_input(skb2
, eiph
->daddr
, eiph
->saddr
, eiph
->tos
,
606 skb_dst(skb2
)->dev
->type
!= ARPHRD_TUNNEL
)
610 /* change mtu on this route */
611 if (rel_type
== ICMP_DEST_UNREACH
&& rel_code
== ICMP_FRAG_NEEDED
) {
612 if (rel_info
> dst_mtu(skb_dst(skb2
)))
615 skb_dst(skb2
)->ops
->update_pmtu(skb_dst(skb2
), rel_info
);
618 icmp_send(skb2
, rel_type
, rel_code
, htonl(rel_info
));
626 ip6ip6_err(struct sk_buff
*skb
, struct inet6_skb_parm
*opt
,
627 u8 type
, u8 code
, int offset
, __be32 info
)
632 __u32 rel_info
= ntohl(info
);
635 err
= ip6_tnl_err(skb
, IPPROTO_IPV6
, opt
, &rel_type
, &rel_code
,
636 &rel_msg
, &rel_info
, offset
);
640 if (rel_msg
&& pskb_may_pull(skb
, offset
+ sizeof(struct ipv6hdr
))) {
642 struct sk_buff
*skb2
= skb_clone(skb
, GFP_ATOMIC
);
648 skb_pull(skb2
, offset
);
649 skb_reset_network_header(skb2
);
651 /* Try to guess incoming interface */
652 rt
= rt6_lookup(dev_net(skb
->dev
), &ipv6_hdr(skb2
)->saddr
,
655 if (rt
&& rt
->rt6i_dev
)
656 skb2
->dev
= rt
->rt6i_dev
;
658 icmpv6_send(skb2
, rel_type
, rel_code
, rel_info
);
661 dst_release(&rt
->dst
);
669 static void ip4ip6_dscp_ecn_decapsulate(struct ip6_tnl
*t
,
670 struct ipv6hdr
*ipv6h
,
673 __u8 dsfield
= ipv6_get_dsfield(ipv6h
) & ~INET_ECN_MASK
;
675 if (t
->parms
.flags
& IP6_TNL_F_RCV_DSCP_COPY
)
676 ipv4_change_dsfield(ip_hdr(skb
), INET_ECN_MASK
, dsfield
);
678 if (INET_ECN_is_ce(dsfield
))
679 IP_ECN_set_ce(ip_hdr(skb
));
682 static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl
*t
,
683 struct ipv6hdr
*ipv6h
,
686 if (t
->parms
.flags
& IP6_TNL_F_RCV_DSCP_COPY
)
687 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h
), ipv6_hdr(skb
));
689 if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h
)))
690 IP6_ECN_set_ce(ipv6_hdr(skb
));
693 /* called with rcu_read_lock() */
694 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl
*t
)
696 struct ip6_tnl_parm
*p
= &t
->parms
;
698 struct net
*net
= dev_net(t
->dev
);
700 if (p
->flags
& IP6_TNL_F_CAP_RCV
) {
701 struct net_device
*ldev
= NULL
;
704 ldev
= dev_get_by_index_rcu(net
, p
->link
);
706 if ((ipv6_addr_is_multicast(&p
->laddr
) ||
707 likely(ipv6_chk_addr(net
, &p
->laddr
, ldev
, 0))) &&
708 likely(!ipv6_chk_addr(net
, &p
->raddr
, NULL
, 0)))
716 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
717 * @skb: received socket buffer
718 * @protocol: ethernet protocol ID
719 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
724 static int ip6_tnl_rcv(struct sk_buff
*skb
, __u16 protocol
,
726 void (*dscp_ecn_decapsulate
)(struct ip6_tnl
*t
,
727 struct ipv6hdr
*ipv6h
,
728 struct sk_buff
*skb
))
731 struct ipv6hdr
*ipv6h
= ipv6_hdr(skb
);
735 if ((t
= ip6_tnl_lookup(dev_net(skb
->dev
), &ipv6h
->saddr
,
736 &ipv6h
->daddr
)) != NULL
) {
737 struct pcpu_tstats
*tstats
;
739 if (t
->parms
.proto
!= ipproto
&& t
->parms
.proto
!= 0) {
744 if (!xfrm6_policy_check(NULL
, XFRM_POLICY_IN
, skb
)) {
749 if (!ip6_tnl_rcv_ctl(t
)) {
750 t
->dev
->stats
.rx_dropped
++;
755 skb
->mac_header
= skb
->network_header
;
756 skb_reset_network_header(skb
);
757 skb
->protocol
= htons(protocol
);
758 skb
->pkt_type
= PACKET_HOST
;
759 memset(skb
->cb
, 0, sizeof(struct inet6_skb_parm
));
761 tstats
= this_cpu_ptr(t
->dev
->tstats
);
762 tstats
->rx_packets
++;
763 tstats
->rx_bytes
+= skb
->len
;
765 __skb_tunnel_rx(skb
, t
->dev
);
767 dscp_ecn_decapsulate(t
, ipv6h
, skb
);
782 static int ip4ip6_rcv(struct sk_buff
*skb
)
784 return ip6_tnl_rcv(skb
, ETH_P_IP
, IPPROTO_IPIP
,
785 ip4ip6_dscp_ecn_decapsulate
);
788 static int ip6ip6_rcv(struct sk_buff
*skb
)
790 return ip6_tnl_rcv(skb
, ETH_P_IPV6
, IPPROTO_IPV6
,
791 ip6ip6_dscp_ecn_decapsulate
);
794 struct ipv6_tel_txoption
{
795 struct ipv6_txoptions ops
;
799 static void init_tel_txopt(struct ipv6_tel_txoption
*opt
, __u8 encap_limit
)
801 memset(opt
, 0, sizeof(struct ipv6_tel_txoption
));
803 opt
->dst_opt
[2] = IPV6_TLV_TNL_ENCAP_LIMIT
;
805 opt
->dst_opt
[4] = encap_limit
;
806 opt
->dst_opt
[5] = IPV6_TLV_PADN
;
809 opt
->ops
.dst0opt
= (struct ipv6_opt_hdr
*) opt
->dst_opt
;
810 opt
->ops
.opt_nflen
= 8;
814 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
815 * @t: the outgoing tunnel device
816 * @hdr: IPv6 header from the incoming packet
819 * Avoid trivial tunneling loop by checking that tunnel exit-point
820 * doesn't match source of incoming packet.
828 ip6_tnl_addr_conflict(struct ip6_tnl
*t
, struct ipv6hdr
*hdr
)
830 return ipv6_addr_equal(&t
->parms
.raddr
, &hdr
->saddr
);
833 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl
*t
)
835 struct ip6_tnl_parm
*p
= &t
->parms
;
837 struct net
*net
= dev_net(t
->dev
);
839 if (p
->flags
& IP6_TNL_F_CAP_XMIT
) {
840 struct net_device
*ldev
= NULL
;
844 ldev
= dev_get_by_index_rcu(net
, p
->link
);
846 if (unlikely(!ipv6_chk_addr(net
, &p
->laddr
, ldev
, 0)))
848 "%s xmit: Local address not yet configured!\n",
850 else if (!ipv6_addr_is_multicast(&p
->raddr
) &&
851 unlikely(ipv6_chk_addr(net
, &p
->raddr
, NULL
, 0)))
853 "%s xmit: Routing loop! "
854 "Remote address found on this node!\n",
863 * ip6_tnl_xmit2 - encapsulate packet and send
864 * @skb: the outgoing socket buffer
865 * @dev: the outgoing tunnel device
866 * @dsfield: dscp code for outer header
867 * @fl: flow of tunneled packet
868 * @encap_limit: encapsulation limit
869 * @pmtu: Path MTU is stored if packet is too big
872 * Build new header and do some sanity checks on the packet before sending
878 * %-EMSGSIZE message too big. return mtu in this case.
881 static int ip6_tnl_xmit2(struct sk_buff
*skb
,
882 struct net_device
*dev
,
888 struct net
*net
= dev_net(dev
);
889 struct ip6_tnl
*t
= netdev_priv(dev
);
890 struct net_device_stats
*stats
= &t
->dev
->stats
;
891 struct ipv6hdr
*ipv6h
= ipv6_hdr(skb
);
892 struct ipv6_tel_txoption opt
;
893 struct dst_entry
*dst
;
894 struct net_device
*tdev
;
896 unsigned int max_headroom
= sizeof(struct ipv6hdr
);
901 if ((dst
= ip6_tnl_dst_check(t
)) != NULL
)
904 dst
= ip6_route_output(net
, NULL
, fl
);
906 if (dst
->error
|| xfrm_lookup(net
, &dst
, fl
, NULL
, 0) < 0)
907 goto tx_err_link_failure
;
916 "%s: Local routing loop detected!\n",
918 goto tx_err_dst_release
;
920 mtu
= dst_mtu(dst
) - sizeof (*ipv6h
);
921 if (encap_limit
>= 0) {
925 if (mtu
< IPV6_MIN_MTU
)
928 skb_dst(skb
)->ops
->update_pmtu(skb_dst(skb
), mtu
);
929 if (skb
->len
> mtu
) {
932 goto tx_err_dst_release
;
936 * Okay, now see if we can stuff it in the buffer as-is.
938 max_headroom
+= LL_RESERVED_SPACE(tdev
);
940 if (skb_headroom(skb
) < max_headroom
|| skb_shared(skb
) ||
941 (skb_cloned(skb
) && !skb_clone_writable(skb
, 0))) {
942 struct sk_buff
*new_skb
;
944 if (!(new_skb
= skb_realloc_headroom(skb
, max_headroom
)))
945 goto tx_err_dst_release
;
948 skb_set_owner_w(new_skb
, skb
->sk
);
953 skb_dst_set(skb
, dst_clone(dst
));
955 skb
->transport_header
= skb
->network_header
;
958 if (encap_limit
>= 0) {
959 init_tel_txopt(&opt
, encap_limit
);
960 ipv6_push_nfrag_opts(skb
, &opt
.ops
, &proto
, NULL
);
962 skb_push(skb
, sizeof(struct ipv6hdr
));
963 skb_reset_network_header(skb
);
964 ipv6h
= ipv6_hdr(skb
);
965 *(__be32
*)ipv6h
= fl
->fl6_flowlabel
| htonl(0x60000000);
966 dsfield
= INET_ECN_encapsulate(0, dsfield
);
967 ipv6_change_dsfield(ipv6h
, ~INET_ECN_MASK
, dsfield
);
968 ipv6h
->hop_limit
= t
->parms
.hop_limit
;
969 ipv6h
->nexthdr
= proto
;
970 ipv6_addr_copy(&ipv6h
->saddr
, &fl
->fl6_src
);
971 ipv6_addr_copy(&ipv6h
->daddr
, &fl
->fl6_dst
);
974 err
= ip6_local_out(skb
);
976 if (net_xmit_eval(err
) == 0) {
977 struct pcpu_tstats
*tstats
= this_cpu_ptr(t
->dev
->tstats
);
979 tstats
->tx_bytes
+= pkt_len
;
980 tstats
->tx_packets
++;
983 stats
->tx_aborted_errors
++;
985 ip6_tnl_dst_store(t
, dst
);
988 stats
->tx_carrier_errors
++;
989 dst_link_failure(skb
);
996 ip4ip6_tnl_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
998 struct ip6_tnl
*t
= netdev_priv(dev
);
999 struct iphdr
*iph
= ip_hdr(skb
);
1000 int encap_limit
= -1;
1006 if ((t
->parms
.proto
!= IPPROTO_IPIP
&& t
->parms
.proto
!= 0) ||
1007 !ip6_tnl_xmit_ctl(t
))
1010 if (!(t
->parms
.flags
& IP6_TNL_F_IGN_ENCAP_LIMIT
))
1011 encap_limit
= t
->parms
.encap_limit
;
1013 memcpy(&fl
, &t
->fl
, sizeof (fl
));
1014 fl
.proto
= IPPROTO_IPIP
;
1016 dsfield
= ipv4_get_dsfield(iph
);
1018 if ((t
->parms
.flags
& IP6_TNL_F_USE_ORIG_TCLASS
))
1019 fl
.fl6_flowlabel
|= htonl((__u32
)iph
->tos
<< IPV6_TCLASS_SHIFT
)
1022 err
= ip6_tnl_xmit2(skb
, dev
, dsfield
, &fl
, encap_limit
, &mtu
);
1024 /* XXX: send ICMP error even if DF is not set. */
1025 if (err
== -EMSGSIZE
)
1026 icmp_send(skb
, ICMP_DEST_UNREACH
, ICMP_FRAG_NEEDED
,
1035 ip6ip6_tnl_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
1037 struct ip6_tnl
*t
= netdev_priv(dev
);
1038 struct ipv6hdr
*ipv6h
= ipv6_hdr(skb
);
1039 int encap_limit
= -1;
1046 if ((t
->parms
.proto
!= IPPROTO_IPV6
&& t
->parms
.proto
!= 0) ||
1047 !ip6_tnl_xmit_ctl(t
) || ip6_tnl_addr_conflict(t
, ipv6h
))
1050 offset
= parse_tlv_tnl_enc_lim(skb
, skb_network_header(skb
));
1052 struct ipv6_tlv_tnl_enc_lim
*tel
;
1053 tel
= (struct ipv6_tlv_tnl_enc_lim
*)&skb_network_header(skb
)[offset
];
1054 if (tel
->encap_limit
== 0) {
1055 icmpv6_send(skb
, ICMPV6_PARAMPROB
,
1056 ICMPV6_HDR_FIELD
, offset
+ 2);
1059 encap_limit
= tel
->encap_limit
- 1;
1060 } else if (!(t
->parms
.flags
& IP6_TNL_F_IGN_ENCAP_LIMIT
))
1061 encap_limit
= t
->parms
.encap_limit
;
1063 memcpy(&fl
, &t
->fl
, sizeof (fl
));
1064 fl
.proto
= IPPROTO_IPV6
;
1066 dsfield
= ipv6_get_dsfield(ipv6h
);
1067 if ((t
->parms
.flags
& IP6_TNL_F_USE_ORIG_TCLASS
))
1068 fl
.fl6_flowlabel
|= (*(__be32
*) ipv6h
& IPV6_TCLASS_MASK
);
1069 if ((t
->parms
.flags
& IP6_TNL_F_USE_ORIG_FLOWLABEL
))
1070 fl
.fl6_flowlabel
|= (*(__be32
*) ipv6h
& IPV6_FLOWLABEL_MASK
);
1072 err
= ip6_tnl_xmit2(skb
, dev
, dsfield
, &fl
, encap_limit
, &mtu
);
1074 if (err
== -EMSGSIZE
)
1075 icmpv6_send(skb
, ICMPV6_PKT_TOOBIG
, 0, mtu
);
1083 ip6_tnl_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
1085 struct ip6_tnl
*t
= netdev_priv(dev
);
1086 struct net_device_stats
*stats
= &t
->dev
->stats
;
1089 switch (skb
->protocol
) {
1090 case htons(ETH_P_IP
):
1091 ret
= ip4ip6_tnl_xmit(skb
, dev
);
1093 case htons(ETH_P_IPV6
):
1094 ret
= ip6ip6_tnl_xmit(skb
, dev
);
1103 return NETDEV_TX_OK
;
1107 stats
->tx_dropped
++;
1109 return NETDEV_TX_OK
;
1112 static void ip6_tnl_set_cap(struct ip6_tnl
*t
)
1114 struct ip6_tnl_parm
*p
= &t
->parms
;
1115 int ltype
= ipv6_addr_type(&p
->laddr
);
1116 int rtype
= ipv6_addr_type(&p
->raddr
);
1118 p
->flags
&= ~(IP6_TNL_F_CAP_XMIT
|IP6_TNL_F_CAP_RCV
);
1120 if (ltype
& (IPV6_ADDR_UNICAST
|IPV6_ADDR_MULTICAST
) &&
1121 rtype
& (IPV6_ADDR_UNICAST
|IPV6_ADDR_MULTICAST
) &&
1122 !((ltype
|rtype
) & IPV6_ADDR_LOOPBACK
) &&
1123 (!((ltype
|rtype
) & IPV6_ADDR_LINKLOCAL
) || p
->link
)) {
1124 if (ltype
&IPV6_ADDR_UNICAST
)
1125 p
->flags
|= IP6_TNL_F_CAP_XMIT
;
1126 if (rtype
&IPV6_ADDR_UNICAST
)
1127 p
->flags
|= IP6_TNL_F_CAP_RCV
;
1131 static void ip6_tnl_link_config(struct ip6_tnl
*t
)
1133 struct net_device
*dev
= t
->dev
;
1134 struct ip6_tnl_parm
*p
= &t
->parms
;
1135 struct flowi
*fl
= &t
->fl
;
1137 memcpy(dev
->dev_addr
, &p
->laddr
, sizeof(struct in6_addr
));
1138 memcpy(dev
->broadcast
, &p
->raddr
, sizeof(struct in6_addr
));
1140 /* Set up flowi template */
1141 ipv6_addr_copy(&fl
->fl6_src
, &p
->laddr
);
1142 ipv6_addr_copy(&fl
->fl6_dst
, &p
->raddr
);
1144 fl
->fl6_flowlabel
= 0;
1146 if (!(p
->flags
&IP6_TNL_F_USE_ORIG_TCLASS
))
1147 fl
->fl6_flowlabel
|= IPV6_TCLASS_MASK
& p
->flowinfo
;
1148 if (!(p
->flags
&IP6_TNL_F_USE_ORIG_FLOWLABEL
))
1149 fl
->fl6_flowlabel
|= IPV6_FLOWLABEL_MASK
& p
->flowinfo
;
1153 if (p
->flags
&IP6_TNL_F_CAP_XMIT
&& p
->flags
&IP6_TNL_F_CAP_RCV
)
1154 dev
->flags
|= IFF_POINTOPOINT
;
1156 dev
->flags
&= ~IFF_POINTOPOINT
;
1158 dev
->iflink
= p
->link
;
1160 if (p
->flags
& IP6_TNL_F_CAP_XMIT
) {
1161 int strict
= (ipv6_addr_type(&p
->raddr
) &
1162 (IPV6_ADDR_MULTICAST
|IPV6_ADDR_LINKLOCAL
));
1164 struct rt6_info
*rt
= rt6_lookup(dev_net(dev
),
1165 &p
->raddr
, &p
->laddr
,
1172 dev
->hard_header_len
= rt
->rt6i_dev
->hard_header_len
+
1173 sizeof (struct ipv6hdr
);
1175 dev
->mtu
= rt
->rt6i_dev
->mtu
- sizeof (struct ipv6hdr
);
1176 if (!(t
->parms
.flags
& IP6_TNL_F_IGN_ENCAP_LIMIT
))
1179 if (dev
->mtu
< IPV6_MIN_MTU
)
1180 dev
->mtu
= IPV6_MIN_MTU
;
1182 dst_release(&rt
->dst
);
1187 * ip6_tnl_change - update the tunnel parameters
1188 * @t: tunnel to be changed
1189 * @p: tunnel configuration parameters
1192 * ip6_tnl_change() updates the tunnel parameters
1196 ip6_tnl_change(struct ip6_tnl
*t
, struct ip6_tnl_parm
*p
)
1198 ipv6_addr_copy(&t
->parms
.laddr
, &p
->laddr
);
1199 ipv6_addr_copy(&t
->parms
.raddr
, &p
->raddr
);
1200 t
->parms
.flags
= p
->flags
;
1201 t
->parms
.hop_limit
= p
->hop_limit
;
1202 t
->parms
.encap_limit
= p
->encap_limit
;
1203 t
->parms
.flowinfo
= p
->flowinfo
;
1204 t
->parms
.link
= p
->link
;
1205 t
->parms
.proto
= p
->proto
;
1206 ip6_tnl_dst_reset(t
);
1207 ip6_tnl_link_config(t
);
1212 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1213 * @dev: virtual device associated with tunnel
1214 * @ifr: parameters passed from userspace
1215 * @cmd: command to be performed
1218 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1221 * The possible commands are the following:
1222 * %SIOCGETTUNNEL: get tunnel parameters for device
1223 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1224 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1225 * %SIOCDELTUNNEL: delete tunnel
1227 * The fallback device "ip6tnl0", created during module
1228 * initialization, can be used for creating other tunnel devices.
1232 * %-EFAULT if unable to copy data to or from userspace,
1233 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1234 * %-EINVAL if passed tunnel parameters are invalid,
1235 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1236 * %-ENODEV if attempting to change or delete a nonexisting device
1240 ip6_tnl_ioctl(struct net_device
*dev
, struct ifreq
*ifr
, int cmd
)
1243 struct ip6_tnl_parm p
;
1244 struct ip6_tnl
*t
= NULL
;
1245 struct net
*net
= dev_net(dev
);
1246 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
1250 if (dev
== ip6n
->fb_tnl_dev
) {
1251 if (copy_from_user(&p
, ifr
->ifr_ifru
.ifru_data
, sizeof (p
))) {
1255 t
= ip6_tnl_locate(net
, &p
, 0);
1258 t
= netdev_priv(dev
);
1259 memcpy(&p
, &t
->parms
, sizeof (p
));
1260 if (copy_to_user(ifr
->ifr_ifru
.ifru_data
, &p
, sizeof (p
))) {
1267 if (!capable(CAP_NET_ADMIN
))
1270 if (copy_from_user(&p
, ifr
->ifr_ifru
.ifru_data
, sizeof (p
)))
1273 if (p
.proto
!= IPPROTO_IPV6
&& p
.proto
!= IPPROTO_IPIP
&&
1276 t
= ip6_tnl_locate(net
, &p
, cmd
== SIOCADDTUNNEL
);
1277 if (dev
!= ip6n
->fb_tnl_dev
&& cmd
== SIOCCHGTUNNEL
) {
1279 if (t
->dev
!= dev
) {
1284 t
= netdev_priv(dev
);
1286 ip6_tnl_unlink(ip6n
, t
);
1288 err
= ip6_tnl_change(t
, &p
);
1289 ip6_tnl_link(ip6n
, t
);
1290 netdev_state_change(dev
);
1294 if (copy_to_user(ifr
->ifr_ifru
.ifru_data
, &t
->parms
, sizeof (p
)))
1298 err
= (cmd
== SIOCADDTUNNEL
? -ENOBUFS
: -ENOENT
);
1302 if (!capable(CAP_NET_ADMIN
))
1305 if (dev
== ip6n
->fb_tnl_dev
) {
1307 if (copy_from_user(&p
, ifr
->ifr_ifru
.ifru_data
, sizeof (p
)))
1310 if ((t
= ip6_tnl_locate(net
, &p
, 0)) == NULL
)
1313 if (t
->dev
== ip6n
->fb_tnl_dev
)
1318 unregister_netdevice(dev
);
1327 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1328 * @dev: virtual device associated with tunnel
1329 * @new_mtu: the new mtu
1333 * %-EINVAL if mtu too small
1337 ip6_tnl_change_mtu(struct net_device
*dev
, int new_mtu
)
1339 if (new_mtu
< IPV6_MIN_MTU
) {
1347 static const struct net_device_ops ip6_tnl_netdev_ops
= {
1348 .ndo_uninit
= ip6_tnl_dev_uninit
,
1349 .ndo_start_xmit
= ip6_tnl_xmit
,
1350 .ndo_do_ioctl
= ip6_tnl_ioctl
,
1351 .ndo_change_mtu
= ip6_tnl_change_mtu
,
1352 .ndo_get_stats
= ip6_get_stats
,
1357 * ip6_tnl_dev_setup - setup virtual tunnel device
1358 * @dev: virtual device associated with tunnel
1361 * Initialize function pointers and device parameters
1364 static void ip6_tnl_dev_setup(struct net_device
*dev
)
1368 dev
->netdev_ops
= &ip6_tnl_netdev_ops
;
1369 dev
->destructor
= ip6_dev_free
;
1371 dev
->type
= ARPHRD_TUNNEL6
;
1372 dev
->hard_header_len
= LL_MAX_HEADER
+ sizeof (struct ipv6hdr
);
1373 dev
->mtu
= ETH_DATA_LEN
- sizeof (struct ipv6hdr
);
1374 t
= netdev_priv(dev
);
1375 if (!(t
->parms
.flags
& IP6_TNL_F_IGN_ENCAP_LIMIT
))
1377 dev
->flags
|= IFF_NOARP
;
1378 dev
->addr_len
= sizeof(struct in6_addr
);
1379 dev
->features
|= NETIF_F_NETNS_LOCAL
;
1380 dev
->priv_flags
&= ~IFF_XMIT_DST_RELEASE
;
1385 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1386 * @dev: virtual device associated with tunnel
1390 ip6_tnl_dev_init_gen(struct net_device
*dev
)
1392 struct ip6_tnl
*t
= netdev_priv(dev
);
1395 strcpy(t
->parms
.name
, dev
->name
);
1396 dev
->tstats
= alloc_percpu(struct pcpu_tstats
);
1403 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1404 * @dev: virtual device associated with tunnel
1407 static int ip6_tnl_dev_init(struct net_device
*dev
)
1409 struct ip6_tnl
*t
= netdev_priv(dev
);
1410 int err
= ip6_tnl_dev_init_gen(dev
);
1414 ip6_tnl_link_config(t
);
1419 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1420 * @dev: fallback device
1425 static int __net_init
ip6_fb_tnl_dev_init(struct net_device
*dev
)
1427 struct ip6_tnl
*t
= netdev_priv(dev
);
1428 struct net
*net
= dev_net(dev
);
1429 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
1430 int err
= ip6_tnl_dev_init_gen(dev
);
1435 t
->parms
.proto
= IPPROTO_IPV6
;
1437 rcu_assign_pointer(ip6n
->tnls_wc
[0], t
);
1441 static struct xfrm6_tunnel ip4ip6_handler __read_mostly
= {
1442 .handler
= ip4ip6_rcv
,
1443 .err_handler
= ip4ip6_err
,
1447 static struct xfrm6_tunnel ip6ip6_handler __read_mostly
= {
1448 .handler
= ip6ip6_rcv
,
1449 .err_handler
= ip6ip6_err
,
1453 static void __net_exit
ip6_tnl_destroy_tunnels(struct ip6_tnl_net
*ip6n
)
1459 for (h
= 0; h
< HASH_SIZE
; h
++) {
1460 t
= rtnl_dereference(ip6n
->tnls_r_l
[h
]);
1462 unregister_netdevice_queue(t
->dev
, &list
);
1463 t
= rtnl_dereference(t
->next
);
1467 t
= rtnl_dereference(ip6n
->tnls_wc
[0]);
1468 unregister_netdevice_queue(t
->dev
, &list
);
1469 unregister_netdevice_many(&list
);
1472 static int __net_init
ip6_tnl_init_net(struct net
*net
)
1474 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
1477 ip6n
->tnls
[0] = ip6n
->tnls_wc
;
1478 ip6n
->tnls
[1] = ip6n
->tnls_r_l
;
1481 ip6n
->fb_tnl_dev
= alloc_netdev(sizeof(struct ip6_tnl
), "ip6tnl0",
1484 if (!ip6n
->fb_tnl_dev
)
1486 dev_net_set(ip6n
->fb_tnl_dev
, net
);
1488 err
= ip6_fb_tnl_dev_init(ip6n
->fb_tnl_dev
);
1492 err
= register_netdev(ip6n
->fb_tnl_dev
);
1498 ip6_dev_free(ip6n
->fb_tnl_dev
);
1503 static void __net_exit
ip6_tnl_exit_net(struct net
*net
)
1505 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
1508 ip6_tnl_destroy_tunnels(ip6n
);
1512 static struct pernet_operations ip6_tnl_net_ops
= {
1513 .init
= ip6_tnl_init_net
,
1514 .exit
= ip6_tnl_exit_net
,
1515 .id
= &ip6_tnl_net_id
,
1516 .size
= sizeof(struct ip6_tnl_net
),
1520 * ip6_tunnel_init - register protocol and reserve needed resources
1522 * Return: 0 on success
1525 static int __init
ip6_tunnel_init(void)
1529 err
= register_pernet_device(&ip6_tnl_net_ops
);
1533 err
= xfrm6_tunnel_register(&ip4ip6_handler
, AF_INET
);
1535 printk(KERN_ERR
"ip6_tunnel init: can't register ip4ip6\n");
1539 err
= xfrm6_tunnel_register(&ip6ip6_handler
, AF_INET6
);
1541 printk(KERN_ERR
"ip6_tunnel init: can't register ip6ip6\n");
1548 xfrm6_tunnel_deregister(&ip4ip6_handler
, AF_INET
);
1550 unregister_pernet_device(&ip6_tnl_net_ops
);
1556 * ip6_tunnel_cleanup - free resources and unregister protocol
1559 static void __exit
ip6_tunnel_cleanup(void)
1561 if (xfrm6_tunnel_deregister(&ip4ip6_handler
, AF_INET
))
1562 printk(KERN_INFO
"ip6_tunnel close: can't deregister ip4ip6\n");
1564 if (xfrm6_tunnel_deregister(&ip6ip6_handler
, AF_INET6
))
1565 printk(KERN_INFO
"ip6_tunnel close: can't deregister ip6ip6\n");
1567 unregister_pernet_device(&ip6_tnl_net_ops
);
1570 module_init(ip6_tunnel_init
);
1571 module_exit(ip6_tunnel_cleanup
);