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");
60 MODULE_ALIAS_NETDEV("ip6tnl0");
62 #define IPV6_TLV_TEL_DST_SIZE 8
65 #define IP6_TNL_TRACE(x...) printk(KERN_DEBUG "%s:" x "\n", __func__)
67 #define IP6_TNL_TRACE(x...) do {;} while(0)
70 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
71 #define IPV6_TCLASS_SHIFT 20
75 #define HASH(addr) ((__force u32)((addr)->s6_addr32[0] ^ (addr)->s6_addr32[1] ^ \
76 (addr)->s6_addr32[2] ^ (addr)->s6_addr32[3]) & \
79 static int ip6_tnl_dev_init(struct net_device
*dev
);
80 static void ip6_tnl_dev_setup(struct net_device
*dev
);
82 static int ip6_tnl_net_id __read_mostly
;
84 /* the IPv6 tunnel fallback device */
85 struct net_device
*fb_tnl_dev
;
86 /* lists for storing tunnels in use */
87 struct ip6_tnl __rcu
*tnls_r_l
[HASH_SIZE
];
88 struct ip6_tnl __rcu
*tnls_wc
[1];
89 struct ip6_tnl __rcu
**tnls
[2];
92 /* often modified stats are per cpu, other are shared (netdev->stats) */
94 unsigned long rx_packets
;
95 unsigned long rx_bytes
;
96 unsigned long tx_packets
;
97 unsigned long tx_bytes
;
100 static struct net_device_stats
*ip6_get_stats(struct net_device
*dev
)
102 struct pcpu_tstats sum
= { 0 };
105 for_each_possible_cpu(i
) {
106 const struct pcpu_tstats
*tstats
= per_cpu_ptr(dev
->tstats
, i
);
108 sum
.rx_packets
+= tstats
->rx_packets
;
109 sum
.rx_bytes
+= tstats
->rx_bytes
;
110 sum
.tx_packets
+= tstats
->tx_packets
;
111 sum
.tx_bytes
+= tstats
->tx_bytes
;
113 dev
->stats
.rx_packets
= sum
.rx_packets
;
114 dev
->stats
.rx_bytes
= sum
.rx_bytes
;
115 dev
->stats
.tx_packets
= sum
.tx_packets
;
116 dev
->stats
.tx_bytes
= sum
.tx_bytes
;
121 * Locking : hash tables are protected by RCU and RTNL
124 static inline struct dst_entry
*ip6_tnl_dst_check(struct ip6_tnl
*t
)
126 struct dst_entry
*dst
= t
->dst_cache
;
128 if (dst
&& dst
->obsolete
&&
129 dst
->ops
->check(dst
, t
->dst_cookie
) == NULL
) {
138 static inline void ip6_tnl_dst_reset(struct ip6_tnl
*t
)
140 dst_release(t
->dst_cache
);
144 static inline void ip6_tnl_dst_store(struct ip6_tnl
*t
, struct dst_entry
*dst
)
146 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
147 t
->dst_cookie
= rt
->rt6i_node
? rt
->rt6i_node
->fn_sernum
: 0;
148 dst_release(t
->dst_cache
);
153 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
154 * @remote: the address of the tunnel exit-point
155 * @local: the address of the tunnel entry-point
158 * tunnel matching given end-points if found,
159 * else fallback tunnel if its device is up,
163 #define for_each_ip6_tunnel_rcu(start) \
164 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
166 static struct ip6_tnl
*
167 ip6_tnl_lookup(struct net
*net
, struct in6_addr
*remote
, struct in6_addr
*local
)
169 unsigned int h0
= HASH(remote
);
170 unsigned int h1
= HASH(local
);
172 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
174 for_each_ip6_tunnel_rcu(ip6n
->tnls_r_l
[h0
^ h1
]) {
175 if (ipv6_addr_equal(local
, &t
->parms
.laddr
) &&
176 ipv6_addr_equal(remote
, &t
->parms
.raddr
) &&
177 (t
->dev
->flags
& IFF_UP
))
180 t
= rcu_dereference(ip6n
->tnls_wc
[0]);
181 if (t
&& (t
->dev
->flags
& IFF_UP
))
188 * ip6_tnl_bucket - get head of list matching given tunnel parameters
189 * @p: parameters containing tunnel end-points
192 * ip6_tnl_bucket() returns the head of the list matching the
193 * &struct in6_addr entries laddr and raddr in @p.
195 * Return: head of IPv6 tunnel list
198 static struct ip6_tnl __rcu
**
199 ip6_tnl_bucket(struct ip6_tnl_net
*ip6n
, struct ip6_tnl_parm
*p
)
201 struct in6_addr
*remote
= &p
->raddr
;
202 struct in6_addr
*local
= &p
->laddr
;
206 if (!ipv6_addr_any(remote
) || !ipv6_addr_any(local
)) {
208 h
= HASH(remote
) ^ HASH(local
);
210 return &ip6n
->tnls
[prio
][h
];
214 * ip6_tnl_link - add tunnel to hash table
215 * @t: tunnel to be added
219 ip6_tnl_link(struct ip6_tnl_net
*ip6n
, struct ip6_tnl
*t
)
221 struct ip6_tnl __rcu
**tp
= ip6_tnl_bucket(ip6n
, &t
->parms
);
223 rcu_assign_pointer(t
->next
, rtnl_dereference(*tp
));
224 rcu_assign_pointer(*tp
, t
);
228 * ip6_tnl_unlink - remove tunnel from hash table
229 * @t: tunnel to be removed
233 ip6_tnl_unlink(struct ip6_tnl_net
*ip6n
, struct ip6_tnl
*t
)
235 struct ip6_tnl __rcu
**tp
;
236 struct ip6_tnl
*iter
;
238 for (tp
= ip6_tnl_bucket(ip6n
, &t
->parms
);
239 (iter
= rtnl_dereference(*tp
)) != NULL
;
242 rcu_assign_pointer(*tp
, t
->next
);
248 static void ip6_dev_free(struct net_device
*dev
)
250 free_percpu(dev
->tstats
);
255 * ip6_tnl_create() - create a new tunnel
256 * @p: tunnel parameters
257 * @pt: pointer to new tunnel
260 * Create tunnel matching given parameters.
263 * created tunnel or NULL
266 static struct ip6_tnl
*ip6_tnl_create(struct net
*net
, struct ip6_tnl_parm
*p
)
268 struct net_device
*dev
;
272 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
275 strlcpy(name
, p
->name
, IFNAMSIZ
);
277 sprintf(name
, "ip6tnl%%d");
279 dev
= alloc_netdev(sizeof (*t
), name
, ip6_tnl_dev_setup
);
283 dev_net_set(dev
, net
);
285 if (strchr(name
, '%')) {
286 if (dev_alloc_name(dev
, name
) < 0)
290 t
= netdev_priv(dev
);
292 err
= ip6_tnl_dev_init(dev
);
296 if ((err
= register_netdevice(dev
)) < 0)
300 ip6_tnl_link(ip6n
, t
);
310 * ip6_tnl_locate - find or create tunnel matching given parameters
311 * @p: tunnel parameters
312 * @create: != 0 if allowed to create new tunnel if no match found
315 * ip6_tnl_locate() first tries to locate an existing tunnel
316 * based on @parms. If this is unsuccessful, but @create is set a new
317 * tunnel device is created and registered for use.
320 * matching tunnel or NULL
323 static struct ip6_tnl
*ip6_tnl_locate(struct net
*net
,
324 struct ip6_tnl_parm
*p
, int create
)
326 struct in6_addr
*remote
= &p
->raddr
;
327 struct in6_addr
*local
= &p
->laddr
;
328 struct ip6_tnl __rcu
**tp
;
330 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
332 for (tp
= ip6_tnl_bucket(ip6n
, p
);
333 (t
= rtnl_dereference(*tp
)) != NULL
;
335 if (ipv6_addr_equal(local
, &t
->parms
.laddr
) &&
336 ipv6_addr_equal(remote
, &t
->parms
.raddr
))
341 return ip6_tnl_create(net
, p
);
345 * ip6_tnl_dev_uninit - tunnel device uninitializer
346 * @dev: the device to be destroyed
349 * ip6_tnl_dev_uninit() removes tunnel from its list
353 ip6_tnl_dev_uninit(struct net_device
*dev
)
355 struct ip6_tnl
*t
= netdev_priv(dev
);
356 struct net
*net
= dev_net(dev
);
357 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
359 if (dev
== ip6n
->fb_tnl_dev
)
360 rcu_assign_pointer(ip6n
->tnls_wc
[0], NULL
);
362 ip6_tnl_unlink(ip6n
, t
);
363 ip6_tnl_dst_reset(t
);
368 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
369 * @skb: received socket buffer
372 * 0 if none was found,
373 * else index to encapsulation limit
377 parse_tlv_tnl_enc_lim(struct sk_buff
*skb
, __u8
* raw
)
379 struct ipv6hdr
*ipv6h
= (struct ipv6hdr
*) raw
;
380 __u8 nexthdr
= ipv6h
->nexthdr
;
381 __u16 off
= sizeof (*ipv6h
);
383 while (ipv6_ext_hdr(nexthdr
) && nexthdr
!= NEXTHDR_NONE
) {
385 struct ipv6_opt_hdr
*hdr
;
386 if (raw
+ off
+ sizeof (*hdr
) > skb
->data
&&
387 !pskb_may_pull(skb
, raw
- skb
->data
+ off
+ sizeof (*hdr
)))
390 hdr
= (struct ipv6_opt_hdr
*) (raw
+ off
);
391 if (nexthdr
== NEXTHDR_FRAGMENT
) {
392 struct frag_hdr
*frag_hdr
= (struct frag_hdr
*) hdr
;
393 if (frag_hdr
->frag_off
)
396 } else if (nexthdr
== NEXTHDR_AUTH
) {
397 optlen
= (hdr
->hdrlen
+ 2) << 2;
399 optlen
= ipv6_optlen(hdr
);
401 if (nexthdr
== NEXTHDR_DEST
) {
404 struct ipv6_tlv_tnl_enc_lim
*tel
;
406 /* No more room for encapsulation limit */
407 if (i
+ sizeof (*tel
) > off
+ optlen
)
410 tel
= (struct ipv6_tlv_tnl_enc_lim
*) &raw
[i
];
411 /* return index of option if found and valid */
412 if (tel
->type
== IPV6_TLV_TNL_ENCAP_LIMIT
&&
415 /* else jump to next option */
417 i
+= tel
->length
+ 2;
422 nexthdr
= hdr
->nexthdr
;
429 * ip6_tnl_err - tunnel error handler
432 * ip6_tnl_err() should handle errors in the tunnel according
433 * to the specifications in RFC 2473.
437 ip6_tnl_err(struct sk_buff
*skb
, __u8 ipproto
, struct inet6_skb_parm
*opt
,
438 u8
*type
, u8
*code
, int *msg
, __u32
*info
, int offset
)
440 struct ipv6hdr
*ipv6h
= (struct ipv6hdr
*) skb
->data
;
443 u8 rel_type
= ICMPV6_DEST_UNREACH
;
444 u8 rel_code
= ICMPV6_ADDR_UNREACH
;
449 /* If the packet doesn't contain the original IPv6 header we are
450 in trouble since we might need the source address for further
451 processing of the error. */
454 if ((t
= ip6_tnl_lookup(dev_net(skb
->dev
), &ipv6h
->daddr
,
455 &ipv6h
->saddr
)) == NULL
)
458 if (t
->parms
.proto
!= ipproto
&& t
->parms
.proto
!= 0)
465 struct ipv6_tlv_tnl_enc_lim
*tel
;
467 case ICMPV6_DEST_UNREACH
:
470 "%s: Path to destination invalid "
471 "or inactive!\n", t
->parms
.name
);
474 case ICMPV6_TIME_EXCEED
:
475 if ((*code
) == ICMPV6_EXC_HOPLIMIT
) {
478 "%s: Too small hop limit or "
479 "routing loop in tunnel!\n",
484 case ICMPV6_PARAMPROB
:
486 if ((*code
) == ICMPV6_HDR_FIELD
)
487 teli
= parse_tlv_tnl_enc_lim(skb
, skb
->data
);
489 if (teli
&& teli
== *info
- 2) {
490 tel
= (struct ipv6_tlv_tnl_enc_lim
*) &skb
->data
[teli
];
491 if (tel
->encap_limit
== 0) {
494 "%s: Too small encapsulation "
495 "limit or routing loop in "
496 "tunnel!\n", t
->parms
.name
);
499 } else if (net_ratelimit()) {
501 "%s: Recipient unable to parse tunneled "
502 "packet!\n ", t
->parms
.name
);
505 case ICMPV6_PKT_TOOBIG
:
506 mtu
= *info
- offset
;
507 if (mtu
< IPV6_MIN_MTU
)
511 if ((len
= sizeof (*ipv6h
) + ntohs(ipv6h
->payload_len
)) > mtu
) {
512 rel_type
= ICMPV6_PKT_TOOBIG
;
531 ip4ip6_err(struct sk_buff
*skb
, struct inet6_skb_parm
*opt
,
532 u8 type
, u8 code
, int offset
, __be32 info
)
537 __u32 rel_info
= ntohl(info
);
539 struct sk_buff
*skb2
;
544 err
= ip6_tnl_err(skb
, IPPROTO_IPIP
, opt
, &rel_type
, &rel_code
,
545 &rel_msg
, &rel_info
, offset
);
553 case ICMPV6_DEST_UNREACH
:
554 if (rel_code
!= ICMPV6_ADDR_UNREACH
)
556 rel_type
= ICMP_DEST_UNREACH
;
557 rel_code
= ICMP_HOST_UNREACH
;
559 case ICMPV6_PKT_TOOBIG
:
562 rel_type
= ICMP_DEST_UNREACH
;
563 rel_code
= ICMP_FRAG_NEEDED
;
569 if (!pskb_may_pull(skb
, offset
+ sizeof(struct iphdr
)))
572 skb2
= skb_clone(skb
, GFP_ATOMIC
);
578 skb_pull(skb2
, offset
);
579 skb_reset_network_header(skb2
);
582 /* Try to guess incoming interface */
583 memset(&fl
, 0, sizeof(fl
));
584 fl
.fl4_dst
= eiph
->saddr
;
585 fl
.fl4_tos
= RT_TOS(eiph
->tos
);
586 fl
.proto
= IPPROTO_IPIP
;
587 if (ip_route_output_key(dev_net(skb
->dev
), &rt
, &fl
))
590 skb2
->dev
= rt
->dst
.dev
;
592 /* route "incoming" packet */
593 if (rt
->rt_flags
& RTCF_LOCAL
) {
596 fl
.fl4_dst
= eiph
->daddr
;
597 fl
.fl4_src
= eiph
->saddr
;
598 fl
.fl4_tos
= eiph
->tos
;
599 if (ip_route_output_key(dev_net(skb
->dev
), &rt
, &fl
) ||
600 rt
->dst
.dev
->type
!= ARPHRD_TUNNEL
) {
604 skb_dst_set(skb2
, (struct dst_entry
*)rt
);
607 if (ip_route_input(skb2
, eiph
->daddr
, eiph
->saddr
, eiph
->tos
,
609 skb_dst(skb2
)->dev
->type
!= ARPHRD_TUNNEL
)
613 /* change mtu on this route */
614 if (rel_type
== ICMP_DEST_UNREACH
&& rel_code
== ICMP_FRAG_NEEDED
) {
615 if (rel_info
> dst_mtu(skb_dst(skb2
)))
618 skb_dst(skb2
)->ops
->update_pmtu(skb_dst(skb2
), rel_info
);
621 icmp_send(skb2
, rel_type
, rel_code
, htonl(rel_info
));
629 ip6ip6_err(struct sk_buff
*skb
, struct inet6_skb_parm
*opt
,
630 u8 type
, u8 code
, int offset
, __be32 info
)
635 __u32 rel_info
= ntohl(info
);
638 err
= ip6_tnl_err(skb
, IPPROTO_IPV6
, opt
, &rel_type
, &rel_code
,
639 &rel_msg
, &rel_info
, offset
);
643 if (rel_msg
&& pskb_may_pull(skb
, offset
+ sizeof(struct ipv6hdr
))) {
645 struct sk_buff
*skb2
= skb_clone(skb
, GFP_ATOMIC
);
651 skb_pull(skb2
, offset
);
652 skb_reset_network_header(skb2
);
654 /* Try to guess incoming interface */
655 rt
= rt6_lookup(dev_net(skb
->dev
), &ipv6_hdr(skb2
)->saddr
,
658 if (rt
&& rt
->rt6i_dev
)
659 skb2
->dev
= rt
->rt6i_dev
;
661 icmpv6_send(skb2
, rel_type
, rel_code
, rel_info
);
664 dst_release(&rt
->dst
);
672 static void ip4ip6_dscp_ecn_decapsulate(struct ip6_tnl
*t
,
673 struct ipv6hdr
*ipv6h
,
676 __u8 dsfield
= ipv6_get_dsfield(ipv6h
) & ~INET_ECN_MASK
;
678 if (t
->parms
.flags
& IP6_TNL_F_RCV_DSCP_COPY
)
679 ipv4_change_dsfield(ip_hdr(skb
), INET_ECN_MASK
, dsfield
);
681 if (INET_ECN_is_ce(dsfield
))
682 IP_ECN_set_ce(ip_hdr(skb
));
685 static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl
*t
,
686 struct ipv6hdr
*ipv6h
,
689 if (t
->parms
.flags
& IP6_TNL_F_RCV_DSCP_COPY
)
690 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h
), ipv6_hdr(skb
));
692 if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h
)))
693 IP6_ECN_set_ce(ipv6_hdr(skb
));
696 /* called with rcu_read_lock() */
697 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl
*t
)
699 struct ip6_tnl_parm
*p
= &t
->parms
;
701 struct net
*net
= dev_net(t
->dev
);
703 if (p
->flags
& IP6_TNL_F_CAP_RCV
) {
704 struct net_device
*ldev
= NULL
;
707 ldev
= dev_get_by_index_rcu(net
, p
->link
);
709 if ((ipv6_addr_is_multicast(&p
->laddr
) ||
710 likely(ipv6_chk_addr(net
, &p
->laddr
, ldev
, 0))) &&
711 likely(!ipv6_chk_addr(net
, &p
->raddr
, NULL
, 0)))
719 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
720 * @skb: received socket buffer
721 * @protocol: ethernet protocol ID
722 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
727 static int ip6_tnl_rcv(struct sk_buff
*skb
, __u16 protocol
,
729 void (*dscp_ecn_decapsulate
)(struct ip6_tnl
*t
,
730 struct ipv6hdr
*ipv6h
,
731 struct sk_buff
*skb
))
734 struct ipv6hdr
*ipv6h
= ipv6_hdr(skb
);
738 if ((t
= ip6_tnl_lookup(dev_net(skb
->dev
), &ipv6h
->saddr
,
739 &ipv6h
->daddr
)) != NULL
) {
740 struct pcpu_tstats
*tstats
;
742 if (t
->parms
.proto
!= ipproto
&& t
->parms
.proto
!= 0) {
747 if (!xfrm6_policy_check(NULL
, XFRM_POLICY_IN
, skb
)) {
752 if (!ip6_tnl_rcv_ctl(t
)) {
753 t
->dev
->stats
.rx_dropped
++;
758 skb
->mac_header
= skb
->network_header
;
759 skb_reset_network_header(skb
);
760 skb
->protocol
= htons(protocol
);
761 skb
->pkt_type
= PACKET_HOST
;
762 memset(skb
->cb
, 0, sizeof(struct inet6_skb_parm
));
764 tstats
= this_cpu_ptr(t
->dev
->tstats
);
765 tstats
->rx_packets
++;
766 tstats
->rx_bytes
+= skb
->len
;
768 __skb_tunnel_rx(skb
, t
->dev
);
770 dscp_ecn_decapsulate(t
, ipv6h
, skb
);
785 static int ip4ip6_rcv(struct sk_buff
*skb
)
787 return ip6_tnl_rcv(skb
, ETH_P_IP
, IPPROTO_IPIP
,
788 ip4ip6_dscp_ecn_decapsulate
);
791 static int ip6ip6_rcv(struct sk_buff
*skb
)
793 return ip6_tnl_rcv(skb
, ETH_P_IPV6
, IPPROTO_IPV6
,
794 ip6ip6_dscp_ecn_decapsulate
);
797 struct ipv6_tel_txoption
{
798 struct ipv6_txoptions ops
;
802 static void init_tel_txopt(struct ipv6_tel_txoption
*opt
, __u8 encap_limit
)
804 memset(opt
, 0, sizeof(struct ipv6_tel_txoption
));
806 opt
->dst_opt
[2] = IPV6_TLV_TNL_ENCAP_LIMIT
;
808 opt
->dst_opt
[4] = encap_limit
;
809 opt
->dst_opt
[5] = IPV6_TLV_PADN
;
812 opt
->ops
.dst0opt
= (struct ipv6_opt_hdr
*) opt
->dst_opt
;
813 opt
->ops
.opt_nflen
= 8;
817 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
818 * @t: the outgoing tunnel device
819 * @hdr: IPv6 header from the incoming packet
822 * Avoid trivial tunneling loop by checking that tunnel exit-point
823 * doesn't match source of incoming packet.
831 ip6_tnl_addr_conflict(struct ip6_tnl
*t
, struct ipv6hdr
*hdr
)
833 return ipv6_addr_equal(&t
->parms
.raddr
, &hdr
->saddr
);
836 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl
*t
)
838 struct ip6_tnl_parm
*p
= &t
->parms
;
840 struct net
*net
= dev_net(t
->dev
);
842 if (p
->flags
& IP6_TNL_F_CAP_XMIT
) {
843 struct net_device
*ldev
= NULL
;
847 ldev
= dev_get_by_index_rcu(net
, p
->link
);
849 if (unlikely(!ipv6_chk_addr(net
, &p
->laddr
, ldev
, 0)))
851 "%s xmit: Local address not yet configured!\n",
853 else if (!ipv6_addr_is_multicast(&p
->raddr
) &&
854 unlikely(ipv6_chk_addr(net
, &p
->raddr
, NULL
, 0)))
856 "%s xmit: Routing loop! "
857 "Remote address found on this node!\n",
866 * ip6_tnl_xmit2 - encapsulate packet and send
867 * @skb: the outgoing socket buffer
868 * @dev: the outgoing tunnel device
869 * @dsfield: dscp code for outer header
870 * @fl: flow of tunneled packet
871 * @encap_limit: encapsulation limit
872 * @pmtu: Path MTU is stored if packet is too big
875 * Build new header and do some sanity checks on the packet before sending
881 * %-EMSGSIZE message too big. return mtu in this case.
884 static int ip6_tnl_xmit2(struct sk_buff
*skb
,
885 struct net_device
*dev
,
891 struct net
*net
= dev_net(dev
);
892 struct ip6_tnl
*t
= netdev_priv(dev
);
893 struct net_device_stats
*stats
= &t
->dev
->stats
;
894 struct ipv6hdr
*ipv6h
= ipv6_hdr(skb
);
895 struct ipv6_tel_txoption opt
;
896 struct dst_entry
*dst
;
897 struct net_device
*tdev
;
899 unsigned int max_headroom
= sizeof(struct ipv6hdr
);
904 if ((dst
= ip6_tnl_dst_check(t
)) != NULL
)
907 dst
= ip6_route_output(net
, NULL
, fl
);
909 if (dst
->error
|| xfrm_lookup(net
, &dst
, fl
, NULL
, 0) < 0)
910 goto tx_err_link_failure
;
919 "%s: Local routing loop detected!\n",
921 goto tx_err_dst_release
;
923 mtu
= dst_mtu(dst
) - sizeof (*ipv6h
);
924 if (encap_limit
>= 0) {
928 if (mtu
< IPV6_MIN_MTU
)
931 skb_dst(skb
)->ops
->update_pmtu(skb_dst(skb
), mtu
);
932 if (skb
->len
> mtu
) {
935 goto tx_err_dst_release
;
939 * Okay, now see if we can stuff it in the buffer as-is.
941 max_headroom
+= LL_RESERVED_SPACE(tdev
);
943 if (skb_headroom(skb
) < max_headroom
|| skb_shared(skb
) ||
944 (skb_cloned(skb
) && !skb_clone_writable(skb
, 0))) {
945 struct sk_buff
*new_skb
;
947 if (!(new_skb
= skb_realloc_headroom(skb
, max_headroom
)))
948 goto tx_err_dst_release
;
951 skb_set_owner_w(new_skb
, skb
->sk
);
956 skb_dst_set(skb
, dst_clone(dst
));
958 skb
->transport_header
= skb
->network_header
;
961 if (encap_limit
>= 0) {
962 init_tel_txopt(&opt
, encap_limit
);
963 ipv6_push_nfrag_opts(skb
, &opt
.ops
, &proto
, NULL
);
965 skb_push(skb
, sizeof(struct ipv6hdr
));
966 skb_reset_network_header(skb
);
967 ipv6h
= ipv6_hdr(skb
);
968 *(__be32
*)ipv6h
= fl
->fl6_flowlabel
| htonl(0x60000000);
969 dsfield
= INET_ECN_encapsulate(0, dsfield
);
970 ipv6_change_dsfield(ipv6h
, ~INET_ECN_MASK
, dsfield
);
971 ipv6h
->hop_limit
= t
->parms
.hop_limit
;
972 ipv6h
->nexthdr
= proto
;
973 ipv6_addr_copy(&ipv6h
->saddr
, &fl
->fl6_src
);
974 ipv6_addr_copy(&ipv6h
->daddr
, &fl
->fl6_dst
);
977 err
= ip6_local_out(skb
);
979 if (net_xmit_eval(err
) == 0) {
980 struct pcpu_tstats
*tstats
= this_cpu_ptr(t
->dev
->tstats
);
982 tstats
->tx_bytes
+= pkt_len
;
983 tstats
->tx_packets
++;
986 stats
->tx_aborted_errors
++;
988 ip6_tnl_dst_store(t
, dst
);
991 stats
->tx_carrier_errors
++;
992 dst_link_failure(skb
);
999 ip4ip6_tnl_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
1001 struct ip6_tnl
*t
= netdev_priv(dev
);
1002 struct iphdr
*iph
= ip_hdr(skb
);
1003 int encap_limit
= -1;
1009 if ((t
->parms
.proto
!= IPPROTO_IPIP
&& t
->parms
.proto
!= 0) ||
1010 !ip6_tnl_xmit_ctl(t
))
1013 if (!(t
->parms
.flags
& IP6_TNL_F_IGN_ENCAP_LIMIT
))
1014 encap_limit
= t
->parms
.encap_limit
;
1016 memcpy(&fl
, &t
->fl
, sizeof (fl
));
1017 fl
.proto
= IPPROTO_IPIP
;
1019 dsfield
= ipv4_get_dsfield(iph
);
1021 if ((t
->parms
.flags
& IP6_TNL_F_USE_ORIG_TCLASS
))
1022 fl
.fl6_flowlabel
|= htonl((__u32
)iph
->tos
<< IPV6_TCLASS_SHIFT
)
1025 err
= ip6_tnl_xmit2(skb
, dev
, dsfield
, &fl
, encap_limit
, &mtu
);
1027 /* XXX: send ICMP error even if DF is not set. */
1028 if (err
== -EMSGSIZE
)
1029 icmp_send(skb
, ICMP_DEST_UNREACH
, ICMP_FRAG_NEEDED
,
1038 ip6ip6_tnl_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
1040 struct ip6_tnl
*t
= netdev_priv(dev
);
1041 struct ipv6hdr
*ipv6h
= ipv6_hdr(skb
);
1042 int encap_limit
= -1;
1049 if ((t
->parms
.proto
!= IPPROTO_IPV6
&& t
->parms
.proto
!= 0) ||
1050 !ip6_tnl_xmit_ctl(t
) || ip6_tnl_addr_conflict(t
, ipv6h
))
1053 offset
= parse_tlv_tnl_enc_lim(skb
, skb_network_header(skb
));
1055 struct ipv6_tlv_tnl_enc_lim
*tel
;
1056 tel
= (struct ipv6_tlv_tnl_enc_lim
*)&skb_network_header(skb
)[offset
];
1057 if (tel
->encap_limit
== 0) {
1058 icmpv6_send(skb
, ICMPV6_PARAMPROB
,
1059 ICMPV6_HDR_FIELD
, offset
+ 2);
1062 encap_limit
= tel
->encap_limit
- 1;
1063 } else if (!(t
->parms
.flags
& IP6_TNL_F_IGN_ENCAP_LIMIT
))
1064 encap_limit
= t
->parms
.encap_limit
;
1066 memcpy(&fl
, &t
->fl
, sizeof (fl
));
1067 fl
.proto
= IPPROTO_IPV6
;
1069 dsfield
= ipv6_get_dsfield(ipv6h
);
1070 if ((t
->parms
.flags
& IP6_TNL_F_USE_ORIG_TCLASS
))
1071 fl
.fl6_flowlabel
|= (*(__be32
*) ipv6h
& IPV6_TCLASS_MASK
);
1072 if ((t
->parms
.flags
& IP6_TNL_F_USE_ORIG_FLOWLABEL
))
1073 fl
.fl6_flowlabel
|= (*(__be32
*) ipv6h
& IPV6_FLOWLABEL_MASK
);
1075 err
= ip6_tnl_xmit2(skb
, dev
, dsfield
, &fl
, encap_limit
, &mtu
);
1077 if (err
== -EMSGSIZE
)
1078 icmpv6_send(skb
, ICMPV6_PKT_TOOBIG
, 0, mtu
);
1086 ip6_tnl_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
1088 struct ip6_tnl
*t
= netdev_priv(dev
);
1089 struct net_device_stats
*stats
= &t
->dev
->stats
;
1092 switch (skb
->protocol
) {
1093 case htons(ETH_P_IP
):
1094 ret
= ip4ip6_tnl_xmit(skb
, dev
);
1096 case htons(ETH_P_IPV6
):
1097 ret
= ip6ip6_tnl_xmit(skb
, dev
);
1106 return NETDEV_TX_OK
;
1110 stats
->tx_dropped
++;
1112 return NETDEV_TX_OK
;
1115 static void ip6_tnl_set_cap(struct ip6_tnl
*t
)
1117 struct ip6_tnl_parm
*p
= &t
->parms
;
1118 int ltype
= ipv6_addr_type(&p
->laddr
);
1119 int rtype
= ipv6_addr_type(&p
->raddr
);
1121 p
->flags
&= ~(IP6_TNL_F_CAP_XMIT
|IP6_TNL_F_CAP_RCV
);
1123 if (ltype
& (IPV6_ADDR_UNICAST
|IPV6_ADDR_MULTICAST
) &&
1124 rtype
& (IPV6_ADDR_UNICAST
|IPV6_ADDR_MULTICAST
) &&
1125 !((ltype
|rtype
) & IPV6_ADDR_LOOPBACK
) &&
1126 (!((ltype
|rtype
) & IPV6_ADDR_LINKLOCAL
) || p
->link
)) {
1127 if (ltype
&IPV6_ADDR_UNICAST
)
1128 p
->flags
|= IP6_TNL_F_CAP_XMIT
;
1129 if (rtype
&IPV6_ADDR_UNICAST
)
1130 p
->flags
|= IP6_TNL_F_CAP_RCV
;
1134 static void ip6_tnl_link_config(struct ip6_tnl
*t
)
1136 struct net_device
*dev
= t
->dev
;
1137 struct ip6_tnl_parm
*p
= &t
->parms
;
1138 struct flowi
*fl
= &t
->fl
;
1140 memcpy(dev
->dev_addr
, &p
->laddr
, sizeof(struct in6_addr
));
1141 memcpy(dev
->broadcast
, &p
->raddr
, sizeof(struct in6_addr
));
1143 /* Set up flowi template */
1144 ipv6_addr_copy(&fl
->fl6_src
, &p
->laddr
);
1145 ipv6_addr_copy(&fl
->fl6_dst
, &p
->raddr
);
1147 fl
->fl6_flowlabel
= 0;
1149 if (!(p
->flags
&IP6_TNL_F_USE_ORIG_TCLASS
))
1150 fl
->fl6_flowlabel
|= IPV6_TCLASS_MASK
& p
->flowinfo
;
1151 if (!(p
->flags
&IP6_TNL_F_USE_ORIG_FLOWLABEL
))
1152 fl
->fl6_flowlabel
|= IPV6_FLOWLABEL_MASK
& p
->flowinfo
;
1156 if (p
->flags
&IP6_TNL_F_CAP_XMIT
&& p
->flags
&IP6_TNL_F_CAP_RCV
)
1157 dev
->flags
|= IFF_POINTOPOINT
;
1159 dev
->flags
&= ~IFF_POINTOPOINT
;
1161 dev
->iflink
= p
->link
;
1163 if (p
->flags
& IP6_TNL_F_CAP_XMIT
) {
1164 int strict
= (ipv6_addr_type(&p
->raddr
) &
1165 (IPV6_ADDR_MULTICAST
|IPV6_ADDR_LINKLOCAL
));
1167 struct rt6_info
*rt
= rt6_lookup(dev_net(dev
),
1168 &p
->raddr
, &p
->laddr
,
1175 dev
->hard_header_len
= rt
->rt6i_dev
->hard_header_len
+
1176 sizeof (struct ipv6hdr
);
1178 dev
->mtu
= rt
->rt6i_dev
->mtu
- sizeof (struct ipv6hdr
);
1179 if (!(t
->parms
.flags
& IP6_TNL_F_IGN_ENCAP_LIMIT
))
1182 if (dev
->mtu
< IPV6_MIN_MTU
)
1183 dev
->mtu
= IPV6_MIN_MTU
;
1185 dst_release(&rt
->dst
);
1190 * ip6_tnl_change - update the tunnel parameters
1191 * @t: tunnel to be changed
1192 * @p: tunnel configuration parameters
1195 * ip6_tnl_change() updates the tunnel parameters
1199 ip6_tnl_change(struct ip6_tnl
*t
, struct ip6_tnl_parm
*p
)
1201 ipv6_addr_copy(&t
->parms
.laddr
, &p
->laddr
);
1202 ipv6_addr_copy(&t
->parms
.raddr
, &p
->raddr
);
1203 t
->parms
.flags
= p
->flags
;
1204 t
->parms
.hop_limit
= p
->hop_limit
;
1205 t
->parms
.encap_limit
= p
->encap_limit
;
1206 t
->parms
.flowinfo
= p
->flowinfo
;
1207 t
->parms
.link
= p
->link
;
1208 t
->parms
.proto
= p
->proto
;
1209 ip6_tnl_dst_reset(t
);
1210 ip6_tnl_link_config(t
);
1215 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1216 * @dev: virtual device associated with tunnel
1217 * @ifr: parameters passed from userspace
1218 * @cmd: command to be performed
1221 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1224 * The possible commands are the following:
1225 * %SIOCGETTUNNEL: get tunnel parameters for device
1226 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1227 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1228 * %SIOCDELTUNNEL: delete tunnel
1230 * The fallback device "ip6tnl0", created during module
1231 * initialization, can be used for creating other tunnel devices.
1235 * %-EFAULT if unable to copy data to or from userspace,
1236 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1237 * %-EINVAL if passed tunnel parameters are invalid,
1238 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1239 * %-ENODEV if attempting to change or delete a nonexisting device
1243 ip6_tnl_ioctl(struct net_device
*dev
, struct ifreq
*ifr
, int cmd
)
1246 struct ip6_tnl_parm p
;
1247 struct ip6_tnl
*t
= NULL
;
1248 struct net
*net
= dev_net(dev
);
1249 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
1253 if (dev
== ip6n
->fb_tnl_dev
) {
1254 if (copy_from_user(&p
, ifr
->ifr_ifru
.ifru_data
, sizeof (p
))) {
1258 t
= ip6_tnl_locate(net
, &p
, 0);
1261 t
= netdev_priv(dev
);
1262 memcpy(&p
, &t
->parms
, sizeof (p
));
1263 if (copy_to_user(ifr
->ifr_ifru
.ifru_data
, &p
, sizeof (p
))) {
1270 if (!capable(CAP_NET_ADMIN
))
1273 if (copy_from_user(&p
, ifr
->ifr_ifru
.ifru_data
, sizeof (p
)))
1276 if (p
.proto
!= IPPROTO_IPV6
&& p
.proto
!= IPPROTO_IPIP
&&
1279 t
= ip6_tnl_locate(net
, &p
, cmd
== SIOCADDTUNNEL
);
1280 if (dev
!= ip6n
->fb_tnl_dev
&& cmd
== SIOCCHGTUNNEL
) {
1282 if (t
->dev
!= dev
) {
1287 t
= netdev_priv(dev
);
1289 ip6_tnl_unlink(ip6n
, t
);
1291 err
= ip6_tnl_change(t
, &p
);
1292 ip6_tnl_link(ip6n
, t
);
1293 netdev_state_change(dev
);
1297 if (copy_to_user(ifr
->ifr_ifru
.ifru_data
, &t
->parms
, sizeof (p
)))
1301 err
= (cmd
== SIOCADDTUNNEL
? -ENOBUFS
: -ENOENT
);
1305 if (!capable(CAP_NET_ADMIN
))
1308 if (dev
== ip6n
->fb_tnl_dev
) {
1310 if (copy_from_user(&p
, ifr
->ifr_ifru
.ifru_data
, sizeof (p
)))
1313 if ((t
= ip6_tnl_locate(net
, &p
, 0)) == NULL
)
1316 if (t
->dev
== ip6n
->fb_tnl_dev
)
1321 unregister_netdevice(dev
);
1330 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1331 * @dev: virtual device associated with tunnel
1332 * @new_mtu: the new mtu
1336 * %-EINVAL if mtu too small
1340 ip6_tnl_change_mtu(struct net_device
*dev
, int new_mtu
)
1342 if (new_mtu
< IPV6_MIN_MTU
) {
1350 static const struct net_device_ops ip6_tnl_netdev_ops
= {
1351 .ndo_uninit
= ip6_tnl_dev_uninit
,
1352 .ndo_start_xmit
= ip6_tnl_xmit
,
1353 .ndo_do_ioctl
= ip6_tnl_ioctl
,
1354 .ndo_change_mtu
= ip6_tnl_change_mtu
,
1355 .ndo_get_stats
= ip6_get_stats
,
1360 * ip6_tnl_dev_setup - setup virtual tunnel device
1361 * @dev: virtual device associated with tunnel
1364 * Initialize function pointers and device parameters
1367 static void ip6_tnl_dev_setup(struct net_device
*dev
)
1371 dev
->netdev_ops
= &ip6_tnl_netdev_ops
;
1372 dev
->destructor
= ip6_dev_free
;
1374 dev
->type
= ARPHRD_TUNNEL6
;
1375 dev
->hard_header_len
= LL_MAX_HEADER
+ sizeof (struct ipv6hdr
);
1376 dev
->mtu
= ETH_DATA_LEN
- sizeof (struct ipv6hdr
);
1377 t
= netdev_priv(dev
);
1378 if (!(t
->parms
.flags
& IP6_TNL_F_IGN_ENCAP_LIMIT
))
1380 dev
->flags
|= IFF_NOARP
;
1381 dev
->addr_len
= sizeof(struct in6_addr
);
1382 dev
->features
|= NETIF_F_NETNS_LOCAL
;
1383 dev
->priv_flags
&= ~IFF_XMIT_DST_RELEASE
;
1388 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1389 * @dev: virtual device associated with tunnel
1393 ip6_tnl_dev_init_gen(struct net_device
*dev
)
1395 struct ip6_tnl
*t
= netdev_priv(dev
);
1398 strcpy(t
->parms
.name
, dev
->name
);
1399 dev
->tstats
= alloc_percpu(struct pcpu_tstats
);
1406 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1407 * @dev: virtual device associated with tunnel
1410 static int ip6_tnl_dev_init(struct net_device
*dev
)
1412 struct ip6_tnl
*t
= netdev_priv(dev
);
1413 int err
= ip6_tnl_dev_init_gen(dev
);
1417 ip6_tnl_link_config(t
);
1422 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1423 * @dev: fallback device
1428 static int __net_init
ip6_fb_tnl_dev_init(struct net_device
*dev
)
1430 struct ip6_tnl
*t
= netdev_priv(dev
);
1431 struct net
*net
= dev_net(dev
);
1432 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
1433 int err
= ip6_tnl_dev_init_gen(dev
);
1438 t
->parms
.proto
= IPPROTO_IPV6
;
1440 rcu_assign_pointer(ip6n
->tnls_wc
[0], t
);
1444 static struct xfrm6_tunnel ip4ip6_handler __read_mostly
= {
1445 .handler
= ip4ip6_rcv
,
1446 .err_handler
= ip4ip6_err
,
1450 static struct xfrm6_tunnel ip6ip6_handler __read_mostly
= {
1451 .handler
= ip6ip6_rcv
,
1452 .err_handler
= ip6ip6_err
,
1456 static void __net_exit
ip6_tnl_destroy_tunnels(struct ip6_tnl_net
*ip6n
)
1462 for (h
= 0; h
< HASH_SIZE
; h
++) {
1463 t
= rtnl_dereference(ip6n
->tnls_r_l
[h
]);
1465 unregister_netdevice_queue(t
->dev
, &list
);
1466 t
= rtnl_dereference(t
->next
);
1470 t
= rtnl_dereference(ip6n
->tnls_wc
[0]);
1471 unregister_netdevice_queue(t
->dev
, &list
);
1472 unregister_netdevice_many(&list
);
1475 static int __net_init
ip6_tnl_init_net(struct net
*net
)
1477 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
1480 ip6n
->tnls
[0] = ip6n
->tnls_wc
;
1481 ip6n
->tnls
[1] = ip6n
->tnls_r_l
;
1484 ip6n
->fb_tnl_dev
= alloc_netdev(sizeof(struct ip6_tnl
), "ip6tnl0",
1487 if (!ip6n
->fb_tnl_dev
)
1489 dev_net_set(ip6n
->fb_tnl_dev
, net
);
1491 err
= ip6_fb_tnl_dev_init(ip6n
->fb_tnl_dev
);
1495 err
= register_netdev(ip6n
->fb_tnl_dev
);
1501 ip6_dev_free(ip6n
->fb_tnl_dev
);
1506 static void __net_exit
ip6_tnl_exit_net(struct net
*net
)
1508 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
1511 ip6_tnl_destroy_tunnels(ip6n
);
1515 static struct pernet_operations ip6_tnl_net_ops
= {
1516 .init
= ip6_tnl_init_net
,
1517 .exit
= ip6_tnl_exit_net
,
1518 .id
= &ip6_tnl_net_id
,
1519 .size
= sizeof(struct ip6_tnl_net
),
1523 * ip6_tunnel_init - register protocol and reserve needed resources
1525 * Return: 0 on success
1528 static int __init
ip6_tunnel_init(void)
1532 err
= register_pernet_device(&ip6_tnl_net_ops
);
1536 err
= xfrm6_tunnel_register(&ip4ip6_handler
, AF_INET
);
1538 printk(KERN_ERR
"ip6_tunnel init: can't register ip4ip6\n");
1542 err
= xfrm6_tunnel_register(&ip6ip6_handler
, AF_INET6
);
1544 printk(KERN_ERR
"ip6_tunnel init: can't register ip6ip6\n");
1551 xfrm6_tunnel_deregister(&ip4ip6_handler
, AF_INET
);
1553 unregister_pernet_device(&ip6_tnl_net_ops
);
1559 * ip6_tunnel_cleanup - free resources and unregister protocol
1562 static void __exit
ip6_tunnel_cleanup(void)
1564 if (xfrm6_tunnel_deregister(&ip4ip6_handler
, AF_INET
))
1565 printk(KERN_INFO
"ip6_tunnel close: can't deregister ip4ip6\n");
1567 if (xfrm6_tunnel_deregister(&ip6ip6_handler
, AF_INET6
))
1568 printk(KERN_INFO
"ip6_tunnel close: can't deregister ip6ip6\n");
1570 unregister_pernet_device(&ip6_tnl_net_ops
);
1573 module_init(ip6_tunnel_init
);
1574 module_exit(ip6_tunnel_cleanup
);