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 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
23 #include <linux/module.h>
24 #include <linux/capability.h>
25 #include <linux/errno.h>
26 #include <linux/types.h>
27 #include <linux/sockios.h>
28 #include <linux/icmp.h>
32 #include <linux/net.h>
33 #include <linux/in6.h>
34 #include <linux/netdevice.h>
35 #include <linux/if_arp.h>
36 #include <linux/icmpv6.h>
37 #include <linux/init.h>
38 #include <linux/route.h>
39 #include <linux/rtnetlink.h>
40 #include <linux/netfilter_ipv6.h>
41 #include <linux/slab.h>
42 #include <linux/hash.h>
43 #include <linux/etherdevice.h>
45 #include <asm/uaccess.h>
46 #include <linux/atomic.h>
50 #include <net/ip_tunnels.h>
52 #include <net/ip6_route.h>
53 #include <net/addrconf.h>
54 #include <net/ip6_tunnel.h>
56 #include <net/dsfield.h>
57 #include <net/inet_ecn.h>
58 #include <net/net_namespace.h>
59 #include <net/netns/generic.h>
61 MODULE_AUTHOR("Ville Nuorvala");
62 MODULE_DESCRIPTION("IPv6 tunneling device");
63 MODULE_LICENSE("GPL");
64 MODULE_ALIAS_RTNL_LINK("ip6tnl");
65 MODULE_ALIAS_NETDEV("ip6tnl0");
68 #define IP6_TNL_TRACE(x...) pr_debug("%s:" x "\n", __func__)
70 #define IP6_TNL_TRACE(x...) do {;} while(0)
73 #define HASH_SIZE_SHIFT 5
74 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
76 static bool log_ecn_error
= true;
77 module_param(log_ecn_error
, bool, 0644);
78 MODULE_PARM_DESC(log_ecn_error
, "Log packets received with corrupted ECN");
80 static u32
HASH(const struct in6_addr
*addr1
, const struct in6_addr
*addr2
)
82 u32 hash
= ipv6_addr_hash(addr1
) ^ ipv6_addr_hash(addr2
);
84 return hash_32(hash
, HASH_SIZE_SHIFT
);
87 static int ip6_tnl_dev_init(struct net_device
*dev
);
88 static void ip6_tnl_dev_setup(struct net_device
*dev
);
89 static struct rtnl_link_ops ip6_link_ops __read_mostly
;
91 static int ip6_tnl_net_id __read_mostly
;
93 /* the IPv6 tunnel fallback device */
94 struct net_device
*fb_tnl_dev
;
95 /* lists for storing tunnels in use */
96 struct ip6_tnl __rcu
*tnls_r_l
[HASH_SIZE
];
97 struct ip6_tnl __rcu
*tnls_wc
[1];
98 struct ip6_tnl __rcu
**tnls
[2];
101 static struct net_device_stats
*ip6_get_stats(struct net_device
*dev
)
103 struct pcpu_sw_netstats tmp
, sum
= { 0 };
106 for_each_possible_cpu(i
) {
108 const struct pcpu_sw_netstats
*tstats
=
109 per_cpu_ptr(dev
->tstats
, i
);
112 start
= u64_stats_fetch_begin_irq(&tstats
->syncp
);
113 tmp
.rx_packets
= tstats
->rx_packets
;
114 tmp
.rx_bytes
= tstats
->rx_bytes
;
115 tmp
.tx_packets
= tstats
->tx_packets
;
116 tmp
.tx_bytes
= tstats
->tx_bytes
;
117 } while (u64_stats_fetch_retry_irq(&tstats
->syncp
, start
));
119 sum
.rx_packets
+= tmp
.rx_packets
;
120 sum
.rx_bytes
+= tmp
.rx_bytes
;
121 sum
.tx_packets
+= tmp
.tx_packets
;
122 sum
.tx_bytes
+= tmp
.tx_bytes
;
124 dev
->stats
.rx_packets
= sum
.rx_packets
;
125 dev
->stats
.rx_bytes
= sum
.rx_bytes
;
126 dev
->stats
.tx_packets
= sum
.tx_packets
;
127 dev
->stats
.tx_bytes
= sum
.tx_bytes
;
132 * Locking : hash tables are protected by RCU and RTNL
135 struct dst_entry
*ip6_tnl_dst_check(struct ip6_tnl
*t
)
137 struct dst_entry
*dst
= t
->dst_cache
;
139 if (dst
&& dst
->obsolete
&&
140 dst
->ops
->check(dst
, t
->dst_cookie
) == NULL
) {
148 EXPORT_SYMBOL_GPL(ip6_tnl_dst_check
);
150 void ip6_tnl_dst_reset(struct ip6_tnl
*t
)
152 dst_release(t
->dst_cache
);
155 EXPORT_SYMBOL_GPL(ip6_tnl_dst_reset
);
157 void ip6_tnl_dst_store(struct ip6_tnl
*t
, struct dst_entry
*dst
)
159 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
160 t
->dst_cookie
= rt
->rt6i_node
? rt
->rt6i_node
->fn_sernum
: 0;
161 dst_release(t
->dst_cache
);
164 EXPORT_SYMBOL_GPL(ip6_tnl_dst_store
);
167 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
168 * @remote: the address of the tunnel exit-point
169 * @local: the address of the tunnel entry-point
172 * tunnel matching given end-points if found,
173 * else fallback tunnel if its device is up,
177 #define for_each_ip6_tunnel_rcu(start) \
178 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
180 static struct ip6_tnl
*
181 ip6_tnl_lookup(struct net
*net
, const struct in6_addr
*remote
, const struct in6_addr
*local
)
183 unsigned int hash
= HASH(remote
, local
);
185 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
187 for_each_ip6_tunnel_rcu(ip6n
->tnls_r_l
[hash
]) {
188 if (ipv6_addr_equal(local
, &t
->parms
.laddr
) &&
189 ipv6_addr_equal(remote
, &t
->parms
.raddr
) &&
190 (t
->dev
->flags
& IFF_UP
))
193 t
= rcu_dereference(ip6n
->tnls_wc
[0]);
194 if (t
&& (t
->dev
->flags
& IFF_UP
))
201 * ip6_tnl_bucket - get head of list matching given tunnel parameters
202 * @p: parameters containing tunnel end-points
205 * ip6_tnl_bucket() returns the head of the list matching the
206 * &struct in6_addr entries laddr and raddr in @p.
208 * Return: head of IPv6 tunnel list
211 static struct ip6_tnl __rcu
**
212 ip6_tnl_bucket(struct ip6_tnl_net
*ip6n
, const struct __ip6_tnl_parm
*p
)
214 const struct in6_addr
*remote
= &p
->raddr
;
215 const struct in6_addr
*local
= &p
->laddr
;
219 if (!ipv6_addr_any(remote
) || !ipv6_addr_any(local
)) {
221 h
= HASH(remote
, local
);
223 return &ip6n
->tnls
[prio
][h
];
227 * ip6_tnl_link - add tunnel to hash table
228 * @t: tunnel to be added
232 ip6_tnl_link(struct ip6_tnl_net
*ip6n
, struct ip6_tnl
*t
)
234 struct ip6_tnl __rcu
**tp
= ip6_tnl_bucket(ip6n
, &t
->parms
);
236 rcu_assign_pointer(t
->next
, rtnl_dereference(*tp
));
237 rcu_assign_pointer(*tp
, t
);
241 * ip6_tnl_unlink - remove tunnel from hash table
242 * @t: tunnel to be removed
246 ip6_tnl_unlink(struct ip6_tnl_net
*ip6n
, struct ip6_tnl
*t
)
248 struct ip6_tnl __rcu
**tp
;
249 struct ip6_tnl
*iter
;
251 for (tp
= ip6_tnl_bucket(ip6n
, &t
->parms
);
252 (iter
= rtnl_dereference(*tp
)) != NULL
;
255 rcu_assign_pointer(*tp
, t
->next
);
261 static void ip6_dev_free(struct net_device
*dev
)
263 free_percpu(dev
->tstats
);
267 static int ip6_tnl_create2(struct net_device
*dev
)
269 struct ip6_tnl
*t
= netdev_priv(dev
);
270 struct net
*net
= dev_net(dev
);
271 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
274 t
= netdev_priv(dev
);
275 err
= ip6_tnl_dev_init(dev
);
279 err
= register_netdevice(dev
);
283 strcpy(t
->parms
.name
, dev
->name
);
284 dev
->rtnl_link_ops
= &ip6_link_ops
;
287 ip6_tnl_link(ip6n
, t
);
295 * ip6_tnl_create - create a new tunnel
296 * @p: tunnel parameters
297 * @pt: pointer to new tunnel
300 * Create tunnel matching given parameters.
303 * created tunnel or NULL
306 static struct ip6_tnl
*ip6_tnl_create(struct net
*net
, struct __ip6_tnl_parm
*p
)
308 struct net_device
*dev
;
314 strlcpy(name
, p
->name
, IFNAMSIZ
);
316 sprintf(name
, "ip6tnl%%d");
318 dev
= alloc_netdev(sizeof(*t
), name
, NET_NAME_UNKNOWN
,
323 dev_net_set(dev
, net
);
325 t
= netdev_priv(dev
);
327 t
->net
= dev_net(dev
);
328 err
= ip6_tnl_create2(dev
);
341 * ip6_tnl_locate - find or create tunnel matching given parameters
342 * @p: tunnel parameters
343 * @create: != 0 if allowed to create new tunnel if no match found
346 * ip6_tnl_locate() first tries to locate an existing tunnel
347 * based on @parms. If this is unsuccessful, but @create is set a new
348 * tunnel device is created and registered for use.
351 * matching tunnel or NULL
354 static struct ip6_tnl
*ip6_tnl_locate(struct net
*net
,
355 struct __ip6_tnl_parm
*p
, int create
)
357 const struct in6_addr
*remote
= &p
->raddr
;
358 const struct in6_addr
*local
= &p
->laddr
;
359 struct ip6_tnl __rcu
**tp
;
361 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
363 for (tp
= ip6_tnl_bucket(ip6n
, p
);
364 (t
= rtnl_dereference(*tp
)) != NULL
;
366 if (ipv6_addr_equal(local
, &t
->parms
.laddr
) &&
367 ipv6_addr_equal(remote
, &t
->parms
.raddr
)) {
376 return ip6_tnl_create(net
, p
);
380 * ip6_tnl_dev_uninit - tunnel device uninitializer
381 * @dev: the device to be destroyed
384 * ip6_tnl_dev_uninit() removes tunnel from its list
388 ip6_tnl_dev_uninit(struct net_device
*dev
)
390 struct ip6_tnl
*t
= netdev_priv(dev
);
391 struct net
*net
= t
->net
;
392 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
394 if (dev
== ip6n
->fb_tnl_dev
)
395 RCU_INIT_POINTER(ip6n
->tnls_wc
[0], NULL
);
397 ip6_tnl_unlink(ip6n
, t
);
398 ip6_tnl_dst_reset(t
);
403 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
404 * @skb: received socket buffer
407 * 0 if none was found,
408 * else index to encapsulation limit
411 __u16
ip6_tnl_parse_tlv_enc_lim(struct sk_buff
*skb
, __u8
*raw
)
413 const struct ipv6hdr
*ipv6h
= (const struct ipv6hdr
*) raw
;
414 __u8 nexthdr
= ipv6h
->nexthdr
;
415 __u16 off
= sizeof(*ipv6h
);
417 while (ipv6_ext_hdr(nexthdr
) && nexthdr
!= NEXTHDR_NONE
) {
419 struct ipv6_opt_hdr
*hdr
;
420 if (raw
+ off
+ sizeof(*hdr
) > skb
->data
&&
421 !pskb_may_pull(skb
, raw
- skb
->data
+ off
+ sizeof (*hdr
)))
424 hdr
= (struct ipv6_opt_hdr
*) (raw
+ off
);
425 if (nexthdr
== NEXTHDR_FRAGMENT
) {
426 struct frag_hdr
*frag_hdr
= (struct frag_hdr
*) hdr
;
427 if (frag_hdr
->frag_off
)
430 } else if (nexthdr
== NEXTHDR_AUTH
) {
431 optlen
= (hdr
->hdrlen
+ 2) << 2;
433 optlen
= ipv6_optlen(hdr
);
435 if (nexthdr
== NEXTHDR_DEST
) {
438 struct ipv6_tlv_tnl_enc_lim
*tel
;
440 /* No more room for encapsulation limit */
441 if (i
+ sizeof (*tel
) > off
+ optlen
)
444 tel
= (struct ipv6_tlv_tnl_enc_lim
*) &raw
[i
];
445 /* return index of option if found and valid */
446 if (tel
->type
== IPV6_TLV_TNL_ENCAP_LIMIT
&&
449 /* else jump to next option */
451 i
+= tel
->length
+ 2;
456 nexthdr
= hdr
->nexthdr
;
461 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim
);
464 * ip6_tnl_err - tunnel error handler
467 * ip6_tnl_err() should handle errors in the tunnel according
468 * to the specifications in RFC 2473.
472 ip6_tnl_err(struct sk_buff
*skb
, __u8 ipproto
, struct inet6_skb_parm
*opt
,
473 u8
*type
, u8
*code
, int *msg
, __u32
*info
, int offset
)
475 const struct ipv6hdr
*ipv6h
= (const struct ipv6hdr
*) skb
->data
;
478 u8 rel_type
= ICMPV6_DEST_UNREACH
;
479 u8 rel_code
= ICMPV6_ADDR_UNREACH
;
485 /* If the packet doesn't contain the original IPv6 header we are
486 in trouble since we might need the source address for further
487 processing of the error. */
490 if ((t
= ip6_tnl_lookup(dev_net(skb
->dev
), &ipv6h
->daddr
,
491 &ipv6h
->saddr
)) == NULL
)
494 tproto
= ACCESS_ONCE(t
->parms
.proto
);
495 if (tproto
!= ipproto
&& tproto
!= 0)
502 struct ipv6_tlv_tnl_enc_lim
*tel
;
504 case ICMPV6_DEST_UNREACH
:
505 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
509 case ICMPV6_TIME_EXCEED
:
510 if ((*code
) == ICMPV6_EXC_HOPLIMIT
) {
511 net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
516 case ICMPV6_PARAMPROB
:
518 if ((*code
) == ICMPV6_HDR_FIELD
)
519 teli
= ip6_tnl_parse_tlv_enc_lim(skb
, skb
->data
);
521 if (teli
&& teli
== *info
- 2) {
522 tel
= (struct ipv6_tlv_tnl_enc_lim
*) &skb
->data
[teli
];
523 if (tel
->encap_limit
== 0) {
524 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
529 net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
533 case ICMPV6_PKT_TOOBIG
:
534 mtu
= *info
- offset
;
535 if (mtu
< IPV6_MIN_MTU
)
539 if ((len
= sizeof(*ipv6h
) + ntohs(ipv6h
->payload_len
)) > mtu
) {
540 rel_type
= ICMPV6_PKT_TOOBIG
;
559 ip4ip6_err(struct sk_buff
*skb
, struct inet6_skb_parm
*opt
,
560 u8 type
, u8 code
, int offset
, __be32 info
)
565 __u32 rel_info
= ntohl(info
);
567 struct sk_buff
*skb2
;
568 const struct iphdr
*eiph
;
572 err
= ip6_tnl_err(skb
, IPPROTO_IPIP
, opt
, &rel_type
, &rel_code
,
573 &rel_msg
, &rel_info
, offset
);
581 case ICMPV6_DEST_UNREACH
:
582 if (rel_code
!= ICMPV6_ADDR_UNREACH
)
584 rel_type
= ICMP_DEST_UNREACH
;
585 rel_code
= ICMP_HOST_UNREACH
;
587 case ICMPV6_PKT_TOOBIG
:
590 rel_type
= ICMP_DEST_UNREACH
;
591 rel_code
= ICMP_FRAG_NEEDED
;
594 rel_type
= ICMP_REDIRECT
;
595 rel_code
= ICMP_REDIR_HOST
;
600 if (!pskb_may_pull(skb
, offset
+ sizeof(struct iphdr
)))
603 skb2
= skb_clone(skb
, GFP_ATOMIC
);
609 skb_pull(skb2
, offset
);
610 skb_reset_network_header(skb2
);
613 /* Try to guess incoming interface */
614 rt
= ip_route_output_ports(dev_net(skb
->dev
), &fl4
, NULL
,
617 IPPROTO_IPIP
, RT_TOS(eiph
->tos
), 0);
621 skb2
->dev
= rt
->dst
.dev
;
623 /* route "incoming" packet */
624 if (rt
->rt_flags
& RTCF_LOCAL
) {
627 rt
= ip_route_output_ports(dev_net(skb
->dev
), &fl4
, NULL
,
628 eiph
->daddr
, eiph
->saddr
,
631 RT_TOS(eiph
->tos
), 0);
633 rt
->dst
.dev
->type
!= ARPHRD_TUNNEL
) {
638 skb_dst_set(skb2
, &rt
->dst
);
641 if (ip_route_input(skb2
, eiph
->daddr
, eiph
->saddr
, eiph
->tos
,
643 skb_dst(skb2
)->dev
->type
!= ARPHRD_TUNNEL
)
647 /* change mtu on this route */
648 if (rel_type
== ICMP_DEST_UNREACH
&& rel_code
== ICMP_FRAG_NEEDED
) {
649 if (rel_info
> dst_mtu(skb_dst(skb2
)))
652 skb_dst(skb2
)->ops
->update_pmtu(skb_dst(skb2
), NULL
, skb2
, rel_info
);
654 if (rel_type
== ICMP_REDIRECT
)
655 skb_dst(skb2
)->ops
->redirect(skb_dst(skb2
), NULL
, skb2
);
657 icmp_send(skb2
, rel_type
, rel_code
, htonl(rel_info
));
665 ip6ip6_err(struct sk_buff
*skb
, struct inet6_skb_parm
*opt
,
666 u8 type
, u8 code
, int offset
, __be32 info
)
671 __u32 rel_info
= ntohl(info
);
674 err
= ip6_tnl_err(skb
, IPPROTO_IPV6
, opt
, &rel_type
, &rel_code
,
675 &rel_msg
, &rel_info
, offset
);
679 if (rel_msg
&& pskb_may_pull(skb
, offset
+ sizeof(struct ipv6hdr
))) {
681 struct sk_buff
*skb2
= skb_clone(skb
, GFP_ATOMIC
);
687 skb_pull(skb2
, offset
);
688 skb_reset_network_header(skb2
);
690 /* Try to guess incoming interface */
691 rt
= rt6_lookup(dev_net(skb
->dev
), &ipv6_hdr(skb2
)->saddr
,
694 if (rt
&& rt
->dst
.dev
)
695 skb2
->dev
= rt
->dst
.dev
;
697 icmpv6_send(skb2
, rel_type
, rel_code
, rel_info
);
707 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl
*t
,
708 const struct ipv6hdr
*ipv6h
,
711 __u8 dsfield
= ipv6_get_dsfield(ipv6h
) & ~INET_ECN_MASK
;
713 if (t
->parms
.flags
& IP6_TNL_F_RCV_DSCP_COPY
)
714 ipv4_change_dsfield(ip_hdr(skb
), INET_ECN_MASK
, dsfield
);
716 return IP6_ECN_decapsulate(ipv6h
, skb
);
719 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl
*t
,
720 const struct ipv6hdr
*ipv6h
,
723 if (t
->parms
.flags
& IP6_TNL_F_RCV_DSCP_COPY
)
724 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h
), ipv6_hdr(skb
));
726 return IP6_ECN_decapsulate(ipv6h
, skb
);
729 __u32
ip6_tnl_get_cap(struct ip6_tnl
*t
,
730 const struct in6_addr
*laddr
,
731 const struct in6_addr
*raddr
)
733 struct __ip6_tnl_parm
*p
= &t
->parms
;
734 int ltype
= ipv6_addr_type(laddr
);
735 int rtype
= ipv6_addr_type(raddr
);
738 if (ltype
== IPV6_ADDR_ANY
|| rtype
== IPV6_ADDR_ANY
) {
739 flags
= IP6_TNL_F_CAP_PER_PACKET
;
740 } else if (ltype
& (IPV6_ADDR_UNICAST
|IPV6_ADDR_MULTICAST
) &&
741 rtype
& (IPV6_ADDR_UNICAST
|IPV6_ADDR_MULTICAST
) &&
742 !((ltype
|rtype
) & IPV6_ADDR_LOOPBACK
) &&
743 (!((ltype
|rtype
) & IPV6_ADDR_LINKLOCAL
) || p
->link
)) {
744 if (ltype
&IPV6_ADDR_UNICAST
)
745 flags
|= IP6_TNL_F_CAP_XMIT
;
746 if (rtype
&IPV6_ADDR_UNICAST
)
747 flags
|= IP6_TNL_F_CAP_RCV
;
751 EXPORT_SYMBOL(ip6_tnl_get_cap
);
753 /* called with rcu_read_lock() */
754 int ip6_tnl_rcv_ctl(struct ip6_tnl
*t
,
755 const struct in6_addr
*laddr
,
756 const struct in6_addr
*raddr
)
758 struct __ip6_tnl_parm
*p
= &t
->parms
;
760 struct net
*net
= t
->net
;
762 if ((p
->flags
& IP6_TNL_F_CAP_RCV
) ||
763 ((p
->flags
& IP6_TNL_F_CAP_PER_PACKET
) &&
764 (ip6_tnl_get_cap(t
, laddr
, raddr
) & IP6_TNL_F_CAP_RCV
))) {
765 struct net_device
*ldev
= NULL
;
768 ldev
= dev_get_by_index_rcu(net
, p
->link
);
770 if ((ipv6_addr_is_multicast(laddr
) ||
771 likely(ipv6_chk_addr(net
, laddr
, ldev
, 0))) &&
772 likely(!ipv6_chk_addr(net
, raddr
, NULL
, 0)))
777 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl
);
780 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
781 * @skb: received socket buffer
782 * @protocol: ethernet protocol ID
783 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
788 static int ip6_tnl_rcv(struct sk_buff
*skb
, __u16 protocol
,
790 int (*dscp_ecn_decapsulate
)(const struct ip6_tnl
*t
,
791 const struct ipv6hdr
*ipv6h
,
792 struct sk_buff
*skb
))
795 const struct ipv6hdr
*ipv6h
= ipv6_hdr(skb
);
801 if ((t
= ip6_tnl_lookup(dev_net(skb
->dev
), &ipv6h
->saddr
,
802 &ipv6h
->daddr
)) != NULL
) {
803 struct pcpu_sw_netstats
*tstats
;
805 tproto
= ACCESS_ONCE(t
->parms
.proto
);
806 if (tproto
!= ipproto
&& tproto
!= 0) {
811 if (!xfrm6_policy_check(NULL
, XFRM_POLICY_IN
, skb
)) {
816 if (!ip6_tnl_rcv_ctl(t
, &ipv6h
->daddr
, &ipv6h
->saddr
)) {
817 t
->dev
->stats
.rx_dropped
++;
821 skb
->mac_header
= skb
->network_header
;
822 skb_reset_network_header(skb
);
823 skb
->protocol
= htons(protocol
);
824 memset(skb
->cb
, 0, sizeof(struct inet6_skb_parm
));
826 __skb_tunnel_rx(skb
, t
->dev
, t
->net
);
828 err
= dscp_ecn_decapsulate(t
, ipv6h
, skb
);
831 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
833 ipv6_get_dsfield(ipv6h
));
835 ++t
->dev
->stats
.rx_frame_errors
;
836 ++t
->dev
->stats
.rx_errors
;
842 tstats
= this_cpu_ptr(t
->dev
->tstats
);
843 u64_stats_update_begin(&tstats
->syncp
);
844 tstats
->rx_packets
++;
845 tstats
->rx_bytes
+= skb
->len
;
846 u64_stats_update_end(&tstats
->syncp
);
861 static int ip4ip6_rcv(struct sk_buff
*skb
)
863 return ip6_tnl_rcv(skb
, ETH_P_IP
, IPPROTO_IPIP
,
864 ip4ip6_dscp_ecn_decapsulate
);
867 static int ip6ip6_rcv(struct sk_buff
*skb
)
869 return ip6_tnl_rcv(skb
, ETH_P_IPV6
, IPPROTO_IPV6
,
870 ip6ip6_dscp_ecn_decapsulate
);
873 struct ipv6_tel_txoption
{
874 struct ipv6_txoptions ops
;
878 static void init_tel_txopt(struct ipv6_tel_txoption
*opt
, __u8 encap_limit
)
880 memset(opt
, 0, sizeof(struct ipv6_tel_txoption
));
882 opt
->dst_opt
[2] = IPV6_TLV_TNL_ENCAP_LIMIT
;
884 opt
->dst_opt
[4] = encap_limit
;
885 opt
->dst_opt
[5] = IPV6_TLV_PADN
;
888 opt
->ops
.dst0opt
= (struct ipv6_opt_hdr
*) opt
->dst_opt
;
889 opt
->ops
.opt_nflen
= 8;
893 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
894 * @t: the outgoing tunnel device
895 * @hdr: IPv6 header from the incoming packet
898 * Avoid trivial tunneling loop by checking that tunnel exit-point
899 * doesn't match source of incoming packet.
907 ip6_tnl_addr_conflict(const struct ip6_tnl
*t
, const struct ipv6hdr
*hdr
)
909 return ipv6_addr_equal(&t
->parms
.raddr
, &hdr
->saddr
);
912 int ip6_tnl_xmit_ctl(struct ip6_tnl
*t
)
914 struct __ip6_tnl_parm
*p
= &t
->parms
;
916 struct net
*net
= t
->net
;
918 if (p
->flags
& IP6_TNL_F_CAP_XMIT
) {
919 struct net_device
*ldev
= NULL
;
923 ldev
= dev_get_by_index_rcu(net
, p
->link
);
925 if (unlikely(!ipv6_chk_addr(net
, &p
->laddr
, ldev
, 0)))
926 pr_warn("%s xmit: Local address not yet configured!\n",
928 else if (!ipv6_addr_is_multicast(&p
->raddr
) &&
929 unlikely(ipv6_chk_addr(net
, &p
->raddr
, NULL
, 0)))
930 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
938 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl
);
941 * ip6_tnl_xmit2 - encapsulate packet and send
942 * @skb: the outgoing socket buffer
943 * @dev: the outgoing tunnel device
944 * @dsfield: dscp code for outer header
945 * @fl: flow of tunneled packet
946 * @encap_limit: encapsulation limit
947 * @pmtu: Path MTU is stored if packet is too big
950 * Build new header and do some sanity checks on the packet before sending
956 * %-EMSGSIZE message too big. return mtu in this case.
959 static int ip6_tnl_xmit2(struct sk_buff
*skb
,
960 struct net_device
*dev
,
966 struct ip6_tnl
*t
= netdev_priv(dev
);
967 struct net
*net
= t
->net
;
968 struct net_device_stats
*stats
= &t
->dev
->stats
;
969 struct ipv6hdr
*ipv6h
= ipv6_hdr(skb
);
970 struct ipv6_tel_txoption opt
;
971 struct dst_entry
*dst
= NULL
, *ndst
= NULL
;
972 struct net_device
*tdev
;
974 unsigned int max_headroom
= sizeof(struct ipv6hdr
);
978 if (!fl6
->flowi6_mark
)
979 dst
= ip6_tnl_dst_check(t
);
981 ndst
= ip6_route_output(net
, NULL
, fl6
);
984 goto tx_err_link_failure
;
985 ndst
= xfrm_lookup(net
, ndst
, flowi6_to_flowi(fl6
), NULL
, 0);
989 goto tx_err_link_failure
;
998 net_warn_ratelimited("%s: Local routing loop detected!\n",
1000 goto tx_err_dst_release
;
1002 mtu
= dst_mtu(dst
) - sizeof(*ipv6h
);
1003 if (encap_limit
>= 0) {
1007 if (mtu
< IPV6_MIN_MTU
)
1010 skb_dst(skb
)->ops
->update_pmtu(skb_dst(skb
), NULL
, skb
, mtu
);
1011 if (skb
->len
> mtu
) {
1014 goto tx_err_dst_release
;
1017 skb_scrub_packet(skb
, !net_eq(t
->net
, dev_net(dev
)));
1020 * Okay, now see if we can stuff it in the buffer as-is.
1022 max_headroom
+= LL_RESERVED_SPACE(tdev
);
1024 if (skb_headroom(skb
) < max_headroom
|| skb_shared(skb
) ||
1025 (skb_cloned(skb
) && !skb_clone_writable(skb
, 0))) {
1026 struct sk_buff
*new_skb
;
1028 if (!(new_skb
= skb_realloc_headroom(skb
, max_headroom
)))
1029 goto tx_err_dst_release
;
1032 skb_set_owner_w(new_skb
, skb
->sk
);
1036 if (fl6
->flowi6_mark
) {
1037 skb_dst_set(skb
, dst
);
1040 skb_dst_set_noref(skb
, dst
);
1042 skb
->transport_header
= skb
->network_header
;
1044 proto
= fl6
->flowi6_proto
;
1045 if (encap_limit
>= 0) {
1046 init_tel_txopt(&opt
, encap_limit
);
1047 ipv6_push_nfrag_opts(skb
, &opt
.ops
, &proto
, NULL
);
1050 if (likely(!skb
->encapsulation
)) {
1051 skb_reset_inner_headers(skb
);
1052 skb
->encapsulation
= 1;
1055 skb_push(skb
, sizeof(struct ipv6hdr
));
1056 skb_reset_network_header(skb
);
1057 ipv6h
= ipv6_hdr(skb
);
1058 ip6_flow_hdr(ipv6h
, INET_ECN_encapsulate(0, dsfield
),
1059 ip6_make_flowlabel(net
, skb
, fl6
->flowlabel
, false));
1060 ipv6h
->hop_limit
= t
->parms
.hop_limit
;
1061 ipv6h
->nexthdr
= proto
;
1062 ipv6h
->saddr
= fl6
->saddr
;
1063 ipv6h
->daddr
= fl6
->daddr
;
1064 ip6tunnel_xmit(skb
, dev
);
1066 ip6_tnl_dst_store(t
, ndst
);
1068 tx_err_link_failure
:
1069 stats
->tx_carrier_errors
++;
1070 dst_link_failure(skb
);
1077 ip4ip6_tnl_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
1079 struct ip6_tnl
*t
= netdev_priv(dev
);
1080 const struct iphdr
*iph
= ip_hdr(skb
);
1081 int encap_limit
= -1;
1088 tproto
= ACCESS_ONCE(t
->parms
.proto
);
1089 if ((tproto
!= IPPROTO_IPIP
&& tproto
!= 0) ||
1090 !ip6_tnl_xmit_ctl(t
))
1093 if (!(t
->parms
.flags
& IP6_TNL_F_IGN_ENCAP_LIMIT
))
1094 encap_limit
= t
->parms
.encap_limit
;
1096 memcpy(&fl6
, &t
->fl
.u
.ip6
, sizeof(fl6
));
1097 fl6
.flowi6_proto
= IPPROTO_IPIP
;
1099 dsfield
= ipv4_get_dsfield(iph
);
1101 if (t
->parms
.flags
& IP6_TNL_F_USE_ORIG_TCLASS
)
1102 fl6
.flowlabel
|= htonl((__u32
)iph
->tos
<< IPV6_TCLASS_SHIFT
)
1104 if (t
->parms
.flags
& IP6_TNL_F_USE_ORIG_FWMARK
)
1105 fl6
.flowi6_mark
= skb
->mark
;
1107 err
= ip6_tnl_xmit2(skb
, dev
, dsfield
, &fl6
, encap_limit
, &mtu
);
1109 /* XXX: send ICMP error even if DF is not set. */
1110 if (err
== -EMSGSIZE
)
1111 icmp_send(skb
, ICMP_DEST_UNREACH
, ICMP_FRAG_NEEDED
,
1120 ip6ip6_tnl_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
1122 struct ip6_tnl
*t
= netdev_priv(dev
);
1123 struct ipv6hdr
*ipv6h
= ipv6_hdr(skb
);
1124 int encap_limit
= -1;
1132 tproto
= ACCESS_ONCE(t
->parms
.proto
);
1133 if ((tproto
!= IPPROTO_IPV6
&& tproto
!= 0) ||
1134 !ip6_tnl_xmit_ctl(t
) || ip6_tnl_addr_conflict(t
, ipv6h
))
1137 offset
= ip6_tnl_parse_tlv_enc_lim(skb
, skb_network_header(skb
));
1139 struct ipv6_tlv_tnl_enc_lim
*tel
;
1140 tel
= (struct ipv6_tlv_tnl_enc_lim
*)&skb_network_header(skb
)[offset
];
1141 if (tel
->encap_limit
== 0) {
1142 icmpv6_send(skb
, ICMPV6_PARAMPROB
,
1143 ICMPV6_HDR_FIELD
, offset
+ 2);
1146 encap_limit
= tel
->encap_limit
- 1;
1147 } else if (!(t
->parms
.flags
& IP6_TNL_F_IGN_ENCAP_LIMIT
))
1148 encap_limit
= t
->parms
.encap_limit
;
1150 memcpy(&fl6
, &t
->fl
.u
.ip6
, sizeof(fl6
));
1151 fl6
.flowi6_proto
= IPPROTO_IPV6
;
1153 dsfield
= ipv6_get_dsfield(ipv6h
);
1154 if (t
->parms
.flags
& IP6_TNL_F_USE_ORIG_TCLASS
)
1155 fl6
.flowlabel
|= (*(__be32
*) ipv6h
& IPV6_TCLASS_MASK
);
1156 if (t
->parms
.flags
& IP6_TNL_F_USE_ORIG_FLOWLABEL
)
1157 fl6
.flowlabel
|= ip6_flowlabel(ipv6h
);
1158 if (t
->parms
.flags
& IP6_TNL_F_USE_ORIG_FWMARK
)
1159 fl6
.flowi6_mark
= skb
->mark
;
1161 err
= ip6_tnl_xmit2(skb
, dev
, dsfield
, &fl6
, encap_limit
, &mtu
);
1163 if (err
== -EMSGSIZE
)
1164 icmpv6_send(skb
, ICMPV6_PKT_TOOBIG
, 0, mtu
);
1172 ip6_tnl_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
1174 struct ip6_tnl
*t
= netdev_priv(dev
);
1175 struct net_device_stats
*stats
= &t
->dev
->stats
;
1178 switch (skb
->protocol
) {
1179 case htons(ETH_P_IP
):
1180 ret
= ip4ip6_tnl_xmit(skb
, dev
);
1182 case htons(ETH_P_IPV6
):
1183 ret
= ip6ip6_tnl_xmit(skb
, dev
);
1192 return NETDEV_TX_OK
;
1196 stats
->tx_dropped
++;
1198 return NETDEV_TX_OK
;
1201 static void ip6_tnl_link_config(struct ip6_tnl
*t
)
1203 struct net_device
*dev
= t
->dev
;
1204 struct __ip6_tnl_parm
*p
= &t
->parms
;
1205 struct flowi6
*fl6
= &t
->fl
.u
.ip6
;
1207 memcpy(dev
->dev_addr
, &p
->laddr
, sizeof(struct in6_addr
));
1208 memcpy(dev
->broadcast
, &p
->raddr
, sizeof(struct in6_addr
));
1210 /* Set up flowi template */
1211 fl6
->saddr
= p
->laddr
;
1212 fl6
->daddr
= p
->raddr
;
1213 fl6
->flowi6_oif
= p
->link
;
1216 if (!(p
->flags
&IP6_TNL_F_USE_ORIG_TCLASS
))
1217 fl6
->flowlabel
|= IPV6_TCLASS_MASK
& p
->flowinfo
;
1218 if (!(p
->flags
&IP6_TNL_F_USE_ORIG_FLOWLABEL
))
1219 fl6
->flowlabel
|= IPV6_FLOWLABEL_MASK
& p
->flowinfo
;
1221 p
->flags
&= ~(IP6_TNL_F_CAP_XMIT
|IP6_TNL_F_CAP_RCV
|IP6_TNL_F_CAP_PER_PACKET
);
1222 p
->flags
|= ip6_tnl_get_cap(t
, &p
->laddr
, &p
->raddr
);
1224 if (p
->flags
&IP6_TNL_F_CAP_XMIT
&& p
->flags
&IP6_TNL_F_CAP_RCV
)
1225 dev
->flags
|= IFF_POINTOPOINT
;
1227 dev
->flags
&= ~IFF_POINTOPOINT
;
1229 dev
->iflink
= p
->link
;
1231 if (p
->flags
& IP6_TNL_F_CAP_XMIT
) {
1232 int strict
= (ipv6_addr_type(&p
->raddr
) &
1233 (IPV6_ADDR_MULTICAST
|IPV6_ADDR_LINKLOCAL
));
1235 struct rt6_info
*rt
= rt6_lookup(t
->net
,
1236 &p
->raddr
, &p
->laddr
,
1243 dev
->hard_header_len
= rt
->dst
.dev
->hard_header_len
+
1244 sizeof(struct ipv6hdr
);
1246 dev
->mtu
= rt
->dst
.dev
->mtu
- sizeof(struct ipv6hdr
);
1247 if (!(t
->parms
.flags
& IP6_TNL_F_IGN_ENCAP_LIMIT
))
1250 if (dev
->mtu
< IPV6_MIN_MTU
)
1251 dev
->mtu
= IPV6_MIN_MTU
;
1258 * ip6_tnl_change - update the tunnel parameters
1259 * @t: tunnel to be changed
1260 * @p: tunnel configuration parameters
1263 * ip6_tnl_change() updates the tunnel parameters
1267 ip6_tnl_change(struct ip6_tnl
*t
, const struct __ip6_tnl_parm
*p
)
1269 t
->parms
.laddr
= p
->laddr
;
1270 t
->parms
.raddr
= p
->raddr
;
1271 t
->parms
.flags
= p
->flags
;
1272 t
->parms
.hop_limit
= p
->hop_limit
;
1273 t
->parms
.encap_limit
= p
->encap_limit
;
1274 t
->parms
.flowinfo
= p
->flowinfo
;
1275 t
->parms
.link
= p
->link
;
1276 t
->parms
.proto
= p
->proto
;
1277 ip6_tnl_dst_reset(t
);
1278 ip6_tnl_link_config(t
);
1282 static int ip6_tnl_update(struct ip6_tnl
*t
, struct __ip6_tnl_parm
*p
)
1284 struct net
*net
= t
->net
;
1285 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
1288 ip6_tnl_unlink(ip6n
, t
);
1290 err
= ip6_tnl_change(t
, p
);
1291 ip6_tnl_link(ip6n
, t
);
1292 netdev_state_change(t
->dev
);
1296 static int ip6_tnl0_update(struct ip6_tnl
*t
, struct __ip6_tnl_parm
*p
)
1298 /* for default tnl0 device allow to change only the proto */
1299 t
->parms
.proto
= p
->proto
;
1300 netdev_state_change(t
->dev
);
1305 ip6_tnl_parm_from_user(struct __ip6_tnl_parm
*p
, const struct ip6_tnl_parm
*u
)
1307 p
->laddr
= u
->laddr
;
1308 p
->raddr
= u
->raddr
;
1309 p
->flags
= u
->flags
;
1310 p
->hop_limit
= u
->hop_limit
;
1311 p
->encap_limit
= u
->encap_limit
;
1312 p
->flowinfo
= u
->flowinfo
;
1314 p
->proto
= u
->proto
;
1315 memcpy(p
->name
, u
->name
, sizeof(u
->name
));
1319 ip6_tnl_parm_to_user(struct ip6_tnl_parm
*u
, const struct __ip6_tnl_parm
*p
)
1321 u
->laddr
= p
->laddr
;
1322 u
->raddr
= p
->raddr
;
1323 u
->flags
= p
->flags
;
1324 u
->hop_limit
= p
->hop_limit
;
1325 u
->encap_limit
= p
->encap_limit
;
1326 u
->flowinfo
= p
->flowinfo
;
1328 u
->proto
= p
->proto
;
1329 memcpy(u
->name
, p
->name
, sizeof(u
->name
));
1333 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1334 * @dev: virtual device associated with tunnel
1335 * @ifr: parameters passed from userspace
1336 * @cmd: command to be performed
1339 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1342 * The possible commands are the following:
1343 * %SIOCGETTUNNEL: get tunnel parameters for device
1344 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1345 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1346 * %SIOCDELTUNNEL: delete tunnel
1348 * The fallback device "ip6tnl0", created during module
1349 * initialization, can be used for creating other tunnel devices.
1353 * %-EFAULT if unable to copy data to or from userspace,
1354 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1355 * %-EINVAL if passed tunnel parameters are invalid,
1356 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1357 * %-ENODEV if attempting to change or delete a nonexisting device
1361 ip6_tnl_ioctl(struct net_device
*dev
, struct ifreq
*ifr
, int cmd
)
1364 struct ip6_tnl_parm p
;
1365 struct __ip6_tnl_parm p1
;
1366 struct ip6_tnl
*t
= netdev_priv(dev
);
1367 struct net
*net
= t
->net
;
1368 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
1372 if (dev
== ip6n
->fb_tnl_dev
) {
1373 if (copy_from_user(&p
, ifr
->ifr_ifru
.ifru_data
, sizeof(p
))) {
1377 ip6_tnl_parm_from_user(&p1
, &p
);
1378 t
= ip6_tnl_locate(net
, &p1
, 0);
1380 t
= netdev_priv(dev
);
1382 memset(&p
, 0, sizeof(p
));
1384 ip6_tnl_parm_to_user(&p
, &t
->parms
);
1385 if (copy_to_user(ifr
->ifr_ifru
.ifru_data
, &p
, sizeof(p
))) {
1392 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
1395 if (copy_from_user(&p
, ifr
->ifr_ifru
.ifru_data
, sizeof(p
)))
1398 if (p
.proto
!= IPPROTO_IPV6
&& p
.proto
!= IPPROTO_IPIP
&&
1401 ip6_tnl_parm_from_user(&p1
, &p
);
1402 t
= ip6_tnl_locate(net
, &p1
, cmd
== SIOCADDTUNNEL
);
1403 if (cmd
== SIOCCHGTUNNEL
) {
1405 if (t
->dev
!= dev
) {
1410 t
= netdev_priv(dev
);
1411 if (dev
== ip6n
->fb_tnl_dev
)
1412 err
= ip6_tnl0_update(t
, &p1
);
1414 err
= ip6_tnl_update(t
, &p1
);
1418 ip6_tnl_parm_to_user(&p
, &t
->parms
);
1419 if (copy_to_user(ifr
->ifr_ifru
.ifru_data
, &p
, sizeof(p
)))
1423 err
= (cmd
== SIOCADDTUNNEL
? -ENOBUFS
: -ENOENT
);
1427 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
1430 if (dev
== ip6n
->fb_tnl_dev
) {
1432 if (copy_from_user(&p
, ifr
->ifr_ifru
.ifru_data
, sizeof(p
)))
1435 ip6_tnl_parm_from_user(&p1
, &p
);
1436 t
= ip6_tnl_locate(net
, &p1
, 0);
1440 if (t
->dev
== ip6n
->fb_tnl_dev
)
1445 unregister_netdevice(dev
);
1454 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1455 * @dev: virtual device associated with tunnel
1456 * @new_mtu: the new mtu
1460 * %-EINVAL if mtu too small
1464 ip6_tnl_change_mtu(struct net_device
*dev
, int new_mtu
)
1466 struct ip6_tnl
*tnl
= netdev_priv(dev
);
1468 if (tnl
->parms
.proto
== IPPROTO_IPIP
) {
1472 if (new_mtu
< IPV6_MIN_MTU
)
1475 if (new_mtu
> 0xFFF8 - dev
->hard_header_len
)
1482 static const struct net_device_ops ip6_tnl_netdev_ops
= {
1483 .ndo_uninit
= ip6_tnl_dev_uninit
,
1484 .ndo_start_xmit
= ip6_tnl_xmit
,
1485 .ndo_do_ioctl
= ip6_tnl_ioctl
,
1486 .ndo_change_mtu
= ip6_tnl_change_mtu
,
1487 .ndo_get_stats
= ip6_get_stats
,
1492 * ip6_tnl_dev_setup - setup virtual tunnel device
1493 * @dev: virtual device associated with tunnel
1496 * Initialize function pointers and device parameters
1499 static void ip6_tnl_dev_setup(struct net_device
*dev
)
1503 dev
->netdev_ops
= &ip6_tnl_netdev_ops
;
1504 dev
->destructor
= ip6_dev_free
;
1506 dev
->type
= ARPHRD_TUNNEL6
;
1507 dev
->hard_header_len
= LL_MAX_HEADER
+ sizeof(struct ipv6hdr
);
1508 dev
->mtu
= ETH_DATA_LEN
- sizeof(struct ipv6hdr
);
1509 t
= netdev_priv(dev
);
1510 if (!(t
->parms
.flags
& IP6_TNL_F_IGN_ENCAP_LIMIT
))
1512 dev
->flags
|= IFF_NOARP
;
1513 dev
->addr_len
= sizeof(struct in6_addr
);
1514 netif_keep_dst(dev
);
1515 /* This perm addr will be used as interface identifier by IPv6 */
1516 dev
->addr_assign_type
= NET_ADDR_RANDOM
;
1517 eth_random_addr(dev
->perm_addr
);
1522 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1523 * @dev: virtual device associated with tunnel
1527 ip6_tnl_dev_init_gen(struct net_device
*dev
)
1529 struct ip6_tnl
*t
= netdev_priv(dev
);
1532 t
->net
= dev_net(dev
);
1533 dev
->tstats
= netdev_alloc_pcpu_stats(struct pcpu_sw_netstats
);
1540 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1541 * @dev: virtual device associated with tunnel
1544 static int ip6_tnl_dev_init(struct net_device
*dev
)
1546 struct ip6_tnl
*t
= netdev_priv(dev
);
1547 int err
= ip6_tnl_dev_init_gen(dev
);
1551 ip6_tnl_link_config(t
);
1556 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1557 * @dev: fallback device
1562 static int __net_init
ip6_fb_tnl_dev_init(struct net_device
*dev
)
1564 struct ip6_tnl
*t
= netdev_priv(dev
);
1565 struct net
*net
= dev_net(dev
);
1566 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
1567 int err
= ip6_tnl_dev_init_gen(dev
);
1572 t
->parms
.proto
= IPPROTO_IPV6
;
1575 ip6_tnl_link_config(t
);
1577 rcu_assign_pointer(ip6n
->tnls_wc
[0], t
);
1581 static int ip6_tnl_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
1585 if (!data
|| !data
[IFLA_IPTUN_PROTO
])
1588 proto
= nla_get_u8(data
[IFLA_IPTUN_PROTO
]);
1589 if (proto
!= IPPROTO_IPV6
&&
1590 proto
!= IPPROTO_IPIP
&&
1597 static void ip6_tnl_netlink_parms(struct nlattr
*data
[],
1598 struct __ip6_tnl_parm
*parms
)
1600 memset(parms
, 0, sizeof(*parms
));
1605 if (data
[IFLA_IPTUN_LINK
])
1606 parms
->link
= nla_get_u32(data
[IFLA_IPTUN_LINK
]);
1608 if (data
[IFLA_IPTUN_LOCAL
])
1609 nla_memcpy(&parms
->laddr
, data
[IFLA_IPTUN_LOCAL
],
1610 sizeof(struct in6_addr
));
1612 if (data
[IFLA_IPTUN_REMOTE
])
1613 nla_memcpy(&parms
->raddr
, data
[IFLA_IPTUN_REMOTE
],
1614 sizeof(struct in6_addr
));
1616 if (data
[IFLA_IPTUN_TTL
])
1617 parms
->hop_limit
= nla_get_u8(data
[IFLA_IPTUN_TTL
]);
1619 if (data
[IFLA_IPTUN_ENCAP_LIMIT
])
1620 parms
->encap_limit
= nla_get_u8(data
[IFLA_IPTUN_ENCAP_LIMIT
]);
1622 if (data
[IFLA_IPTUN_FLOWINFO
])
1623 parms
->flowinfo
= nla_get_be32(data
[IFLA_IPTUN_FLOWINFO
]);
1625 if (data
[IFLA_IPTUN_FLAGS
])
1626 parms
->flags
= nla_get_u32(data
[IFLA_IPTUN_FLAGS
]);
1628 if (data
[IFLA_IPTUN_PROTO
])
1629 parms
->proto
= nla_get_u8(data
[IFLA_IPTUN_PROTO
]);
1632 static int ip6_tnl_newlink(struct net
*src_net
, struct net_device
*dev
,
1633 struct nlattr
*tb
[], struct nlattr
*data
[])
1635 struct net
*net
= dev_net(dev
);
1638 nt
= netdev_priv(dev
);
1639 ip6_tnl_netlink_parms(data
, &nt
->parms
);
1641 if (ip6_tnl_locate(net
, &nt
->parms
, 0))
1644 return ip6_tnl_create2(dev
);
1647 static int ip6_tnl_changelink(struct net_device
*dev
, struct nlattr
*tb
[],
1648 struct nlattr
*data
[])
1650 struct ip6_tnl
*t
= netdev_priv(dev
);
1651 struct __ip6_tnl_parm p
;
1652 struct net
*net
= t
->net
;
1653 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
1655 if (dev
== ip6n
->fb_tnl_dev
)
1658 ip6_tnl_netlink_parms(data
, &p
);
1660 t
= ip6_tnl_locate(net
, &p
, 0);
1666 t
= netdev_priv(dev
);
1668 return ip6_tnl_update(t
, &p
);
1671 static void ip6_tnl_dellink(struct net_device
*dev
, struct list_head
*head
)
1673 struct net
*net
= dev_net(dev
);
1674 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
1676 if (dev
!= ip6n
->fb_tnl_dev
)
1677 unregister_netdevice_queue(dev
, head
);
1680 static size_t ip6_tnl_get_size(const struct net_device
*dev
)
1683 /* IFLA_IPTUN_LINK */
1685 /* IFLA_IPTUN_LOCAL */
1686 nla_total_size(sizeof(struct in6_addr
)) +
1687 /* IFLA_IPTUN_REMOTE */
1688 nla_total_size(sizeof(struct in6_addr
)) +
1689 /* IFLA_IPTUN_TTL */
1691 /* IFLA_IPTUN_ENCAP_LIMIT */
1693 /* IFLA_IPTUN_FLOWINFO */
1695 /* IFLA_IPTUN_FLAGS */
1697 /* IFLA_IPTUN_PROTO */
1702 static int ip6_tnl_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
1704 struct ip6_tnl
*tunnel
= netdev_priv(dev
);
1705 struct __ip6_tnl_parm
*parm
= &tunnel
->parms
;
1707 if (nla_put_u32(skb
, IFLA_IPTUN_LINK
, parm
->link
) ||
1708 nla_put(skb
, IFLA_IPTUN_LOCAL
, sizeof(struct in6_addr
),
1710 nla_put(skb
, IFLA_IPTUN_REMOTE
, sizeof(struct in6_addr
),
1712 nla_put_u8(skb
, IFLA_IPTUN_TTL
, parm
->hop_limit
) ||
1713 nla_put_u8(skb
, IFLA_IPTUN_ENCAP_LIMIT
, parm
->encap_limit
) ||
1714 nla_put_be32(skb
, IFLA_IPTUN_FLOWINFO
, parm
->flowinfo
) ||
1715 nla_put_u32(skb
, IFLA_IPTUN_FLAGS
, parm
->flags
) ||
1716 nla_put_u8(skb
, IFLA_IPTUN_PROTO
, parm
->proto
))
1717 goto nla_put_failure
;
1724 static const struct nla_policy ip6_tnl_policy
[IFLA_IPTUN_MAX
+ 1] = {
1725 [IFLA_IPTUN_LINK
] = { .type
= NLA_U32
},
1726 [IFLA_IPTUN_LOCAL
] = { .len
= sizeof(struct in6_addr
) },
1727 [IFLA_IPTUN_REMOTE
] = { .len
= sizeof(struct in6_addr
) },
1728 [IFLA_IPTUN_TTL
] = { .type
= NLA_U8
},
1729 [IFLA_IPTUN_ENCAP_LIMIT
] = { .type
= NLA_U8
},
1730 [IFLA_IPTUN_FLOWINFO
] = { .type
= NLA_U32
},
1731 [IFLA_IPTUN_FLAGS
] = { .type
= NLA_U32
},
1732 [IFLA_IPTUN_PROTO
] = { .type
= NLA_U8
},
1735 static struct rtnl_link_ops ip6_link_ops __read_mostly
= {
1737 .maxtype
= IFLA_IPTUN_MAX
,
1738 .policy
= ip6_tnl_policy
,
1739 .priv_size
= sizeof(struct ip6_tnl
),
1740 .setup
= ip6_tnl_dev_setup
,
1741 .validate
= ip6_tnl_validate
,
1742 .newlink
= ip6_tnl_newlink
,
1743 .changelink
= ip6_tnl_changelink
,
1744 .dellink
= ip6_tnl_dellink
,
1745 .get_size
= ip6_tnl_get_size
,
1746 .fill_info
= ip6_tnl_fill_info
,
1749 static struct xfrm6_tunnel ip4ip6_handler __read_mostly
= {
1750 .handler
= ip4ip6_rcv
,
1751 .err_handler
= ip4ip6_err
,
1755 static struct xfrm6_tunnel ip6ip6_handler __read_mostly
= {
1756 .handler
= ip6ip6_rcv
,
1757 .err_handler
= ip6ip6_err
,
1761 static void __net_exit
ip6_tnl_destroy_tunnels(struct net
*net
)
1763 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
1764 struct net_device
*dev
, *aux
;
1769 for_each_netdev_safe(net
, dev
, aux
)
1770 if (dev
->rtnl_link_ops
== &ip6_link_ops
)
1771 unregister_netdevice_queue(dev
, &list
);
1773 for (h
= 0; h
< HASH_SIZE
; h
++) {
1774 t
= rtnl_dereference(ip6n
->tnls_r_l
[h
]);
1776 /* If dev is in the same netns, it has already
1777 * been added to the list by the previous loop.
1779 if (!net_eq(dev_net(t
->dev
), net
))
1780 unregister_netdevice_queue(t
->dev
, &list
);
1781 t
= rtnl_dereference(t
->next
);
1785 unregister_netdevice_many(&list
);
1788 static int __net_init
ip6_tnl_init_net(struct net
*net
)
1790 struct ip6_tnl_net
*ip6n
= net_generic(net
, ip6_tnl_net_id
);
1791 struct ip6_tnl
*t
= NULL
;
1794 ip6n
->tnls
[0] = ip6n
->tnls_wc
;
1795 ip6n
->tnls
[1] = ip6n
->tnls_r_l
;
1798 ip6n
->fb_tnl_dev
= alloc_netdev(sizeof(struct ip6_tnl
), "ip6tnl0",
1799 NET_NAME_UNKNOWN
, ip6_tnl_dev_setup
);
1801 if (!ip6n
->fb_tnl_dev
)
1803 dev_net_set(ip6n
->fb_tnl_dev
, net
);
1804 ip6n
->fb_tnl_dev
->rtnl_link_ops
= &ip6_link_ops
;
1805 /* FB netdevice is special: we have one, and only one per netns.
1806 * Allowing to move it to another netns is clearly unsafe.
1808 ip6n
->fb_tnl_dev
->features
|= NETIF_F_NETNS_LOCAL
;
1810 err
= ip6_fb_tnl_dev_init(ip6n
->fb_tnl_dev
);
1814 err
= register_netdev(ip6n
->fb_tnl_dev
);
1818 t
= netdev_priv(ip6n
->fb_tnl_dev
);
1820 strcpy(t
->parms
.name
, ip6n
->fb_tnl_dev
->name
);
1824 ip6_dev_free(ip6n
->fb_tnl_dev
);
1829 static void __net_exit
ip6_tnl_exit_net(struct net
*net
)
1832 ip6_tnl_destroy_tunnels(net
);
1836 static struct pernet_operations ip6_tnl_net_ops
= {
1837 .init
= ip6_tnl_init_net
,
1838 .exit
= ip6_tnl_exit_net
,
1839 .id
= &ip6_tnl_net_id
,
1840 .size
= sizeof(struct ip6_tnl_net
),
1844 * ip6_tunnel_init - register protocol and reserve needed resources
1846 * Return: 0 on success
1849 static int __init
ip6_tunnel_init(void)
1853 err
= register_pernet_device(&ip6_tnl_net_ops
);
1857 err
= xfrm6_tunnel_register(&ip4ip6_handler
, AF_INET
);
1859 pr_err("%s: can't register ip4ip6\n", __func__
);
1863 err
= xfrm6_tunnel_register(&ip6ip6_handler
, AF_INET6
);
1865 pr_err("%s: can't register ip6ip6\n", __func__
);
1868 err
= rtnl_link_register(&ip6_link_ops
);
1870 goto rtnl_link_failed
;
1875 xfrm6_tunnel_deregister(&ip6ip6_handler
, AF_INET6
);
1877 xfrm6_tunnel_deregister(&ip4ip6_handler
, AF_INET
);
1879 unregister_pernet_device(&ip6_tnl_net_ops
);
1885 * ip6_tunnel_cleanup - free resources and unregister protocol
1888 static void __exit
ip6_tunnel_cleanup(void)
1890 rtnl_link_unregister(&ip6_link_ops
);
1891 if (xfrm6_tunnel_deregister(&ip4ip6_handler
, AF_INET
))
1892 pr_info("%s: can't deregister ip4ip6\n", __func__
);
1894 if (xfrm6_tunnel_deregister(&ip6ip6_handler
, AF_INET6
))
1895 pr_info("%s: can't deregister ip6ip6\n", __func__
);
1897 unregister_pernet_device(&ip6_tnl_net_ops
);
1900 module_init(ip6_tunnel_init
);
1901 module_exit(ip6_tunnel_cleanup
);