2 * Linux NET3: IP/IP protocol decoder modified to support
3 * virtual tunnel interface
6 * Saurabh Mohan (saurabh.mohan@vyatta.com) 05/07/2012
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
16 This version of net/ipv4/ip_vti.c is cloned of net/ipv4/ipip.c
18 For comments look at net/ipv4/ip_gre.c --ANK
22 #include <linux/capability.h>
23 #include <linux/module.h>
24 #include <linux/types.h>
25 #include <linux/kernel.h>
26 #include <linux/uaccess.h>
27 #include <linux/skbuff.h>
28 #include <linux/netdevice.h>
30 #include <linux/tcp.h>
31 #include <linux/udp.h>
32 #include <linux/if_arp.h>
33 #include <linux/init.h>
34 #include <linux/netfilter_ipv4.h>
35 #include <linux/if_ether.h>
36 #include <linux/icmpv6.h>
41 #include <net/ip_tunnels.h>
42 #include <net/inet_ecn.h>
44 #include <net/net_namespace.h>
45 #include <net/netns/generic.h>
47 static struct rtnl_link_ops vti_link_ops __read_mostly
;
49 static int vti_net_id __read_mostly
;
50 static int vti_tunnel_init(struct net_device
*dev
);
52 static int vti_input(struct sk_buff
*skb
, int nexthdr
, __be32 spi
,
55 struct ip_tunnel
*tunnel
;
56 const struct iphdr
*iph
= ip_hdr(skb
);
57 struct net
*net
= dev_net(skb
->dev
);
58 struct ip_tunnel_net
*itn
= net_generic(net
, vti_net_id
);
60 tunnel
= ip_tunnel_lookup(itn
, skb
->dev
->ifindex
, TUNNEL_NO_KEY
,
61 iph
->saddr
, iph
->daddr
, 0);
63 if (!xfrm4_policy_check(NULL
, XFRM_POLICY_IN
, skb
))
66 XFRM_TUNNEL_SKB_CB(skb
)->tunnel
.ip4
= tunnel
;
68 return xfrm_input(skb
, nexthdr
, spi
, encap_type
);
77 static int vti_rcv(struct sk_buff
*skb
)
79 XFRM_SPI_SKB_CB(skb
)->family
= AF_INET
;
80 XFRM_SPI_SKB_CB(skb
)->daddroff
= offsetof(struct iphdr
, daddr
);
82 return vti_input(skb
, ip_hdr(skb
)->protocol
, 0, 0);
85 static int vti_rcv_cb(struct sk_buff
*skb
, int err
)
87 unsigned short family
;
88 struct net_device
*dev
;
89 struct pcpu_sw_netstats
*tstats
;
91 struct ip_tunnel
*tunnel
= XFRM_TUNNEL_SKB_CB(skb
)->tunnel
.ip4
;
92 u32 orig_mark
= skb
->mark
;
101 dev
->stats
.rx_errors
++;
102 dev
->stats
.rx_dropped
++;
107 x
= xfrm_input_state(skb
);
108 family
= x
->inner_mode
->afinfo
->family
;
110 skb
->mark
= be32_to_cpu(tunnel
->parms
.i_key
);
111 ret
= xfrm_policy_check(NULL
, XFRM_POLICY_IN
, skb
, family
);
112 skb
->mark
= orig_mark
;
117 skb_scrub_packet(skb
, !net_eq(tunnel
->net
, dev_net(skb
->dev
)));
120 tstats
= this_cpu_ptr(dev
->tstats
);
122 u64_stats_update_begin(&tstats
->syncp
);
123 tstats
->rx_packets
++;
124 tstats
->rx_bytes
+= skb
->len
;
125 u64_stats_update_end(&tstats
->syncp
);
130 static bool vti_state_check(const struct xfrm_state
*x
, __be32 dst
, __be32 src
)
132 xfrm_address_t
*daddr
= (xfrm_address_t
*)&dst
;
133 xfrm_address_t
*saddr
= (xfrm_address_t
*)&src
;
135 /* if there is no transform then this tunnel is not functional.
136 * Or if the xfrm is not mode tunnel.
138 if (!x
|| x
->props
.mode
!= XFRM_MODE_TUNNEL
||
139 x
->props
.family
!= AF_INET
)
143 return xfrm_addr_equal(saddr
, &x
->props
.saddr
, AF_INET
);
145 if (!xfrm_state_addr_check(x
, daddr
, saddr
, AF_INET
))
151 static netdev_tx_t
vti_xmit(struct sk_buff
*skb
, struct net_device
*dev
,
154 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
155 struct ip_tunnel_parm
*parms
= &tunnel
->parms
;
156 struct dst_entry
*dst
= skb_dst(skb
);
157 struct net_device
*tdev
; /* Device to other host */
162 dev
->stats
.tx_carrier_errors
++;
167 dst
= xfrm_lookup(tunnel
->net
, dst
, fl
, NULL
, 0);
169 dev
->stats
.tx_carrier_errors
++;
173 if (!vti_state_check(dst
->xfrm
, parms
->iph
.daddr
, parms
->iph
.saddr
)) {
174 dev
->stats
.tx_carrier_errors
++;
183 dev
->stats
.collisions
++;
187 if (tunnel
->err_count
> 0) {
188 if (time_before(jiffies
,
189 tunnel
->err_time
+ IPTUNNEL_ERR_TIMEO
)) {
191 dst_link_failure(skb
);
193 tunnel
->err_count
= 0;
197 if (skb
->len
> mtu
) {
198 skb_dst(skb
)->ops
->update_pmtu(skb_dst(skb
), NULL
, skb
, mtu
);
199 if (skb
->protocol
== htons(ETH_P_IP
)) {
200 icmp_send(skb
, ICMP_DEST_UNREACH
, ICMP_FRAG_NEEDED
,
203 if (mtu
< IPV6_MIN_MTU
)
206 icmpv6_send(skb
, ICMPV6_PKT_TOOBIG
, 0, mtu
);
213 skb_scrub_packet(skb
, !net_eq(tunnel
->net
, dev_net(dev
)));
214 skb_dst_set(skb
, dst
);
215 skb
->dev
= skb_dst(skb
)->dev
;
217 err
= dst_output(tunnel
->net
, skb
->sk
, skb
);
218 if (net_xmit_eval(err
) == 0)
220 iptunnel_xmit_stats(dev
, err
);
224 dst_link_failure(skb
);
226 dev
->stats
.tx_errors
++;
231 /* This function assumes it is being called from dev_queue_xmit()
232 * and that skb is filled properly by that function.
234 static netdev_tx_t
vti_tunnel_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
236 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
239 memset(&fl
, 0, sizeof(fl
));
241 switch (skb
->protocol
) {
242 case htons(ETH_P_IP
):
243 xfrm_decode_session(skb
, &fl
, AF_INET
);
244 memset(IPCB(skb
), 0, sizeof(*IPCB(skb
)));
246 case htons(ETH_P_IPV6
):
247 xfrm_decode_session(skb
, &fl
, AF_INET6
);
248 memset(IP6CB(skb
), 0, sizeof(*IP6CB(skb
)));
251 dev
->stats
.tx_errors
++;
256 /* override mark with tunnel output key */
257 fl
.flowi_mark
= be32_to_cpu(tunnel
->parms
.o_key
);
259 return vti_xmit(skb
, dev
, &fl
);
262 static int vti4_err(struct sk_buff
*skb
, u32 info
)
266 struct xfrm_state
*x
;
267 struct ip_tunnel
*tunnel
;
268 struct ip_esp_hdr
*esph
;
269 struct ip_auth_hdr
*ah
;
270 struct ip_comp_hdr
*ipch
;
271 struct net
*net
= dev_net(skb
->dev
);
272 const struct iphdr
*iph
= (const struct iphdr
*)skb
->data
;
273 int protocol
= iph
->protocol
;
274 struct ip_tunnel_net
*itn
= net_generic(net
, vti_net_id
);
276 tunnel
= ip_tunnel_lookup(itn
, skb
->dev
->ifindex
, TUNNEL_NO_KEY
,
277 iph
->daddr
, iph
->saddr
, 0);
281 mark
= be32_to_cpu(tunnel
->parms
.o_key
);
285 esph
= (struct ip_esp_hdr
*)(skb
->data
+(iph
->ihl
<<2));
289 ah
= (struct ip_auth_hdr
*)(skb
->data
+(iph
->ihl
<<2));
293 ipch
= (struct ip_comp_hdr
*)(skb
->data
+(iph
->ihl
<<2));
294 spi
= htonl(ntohs(ipch
->cpi
));
300 switch (icmp_hdr(skb
)->type
) {
301 case ICMP_DEST_UNREACH
:
302 if (icmp_hdr(skb
)->code
!= ICMP_FRAG_NEEDED
)
310 x
= xfrm_state_lookup(net
, mark
, (const xfrm_address_t
*)&iph
->daddr
,
311 spi
, protocol
, AF_INET
);
315 if (icmp_hdr(skb
)->type
== ICMP_DEST_UNREACH
)
316 ipv4_update_pmtu(skb
, net
, info
, 0, 0, protocol
, 0);
318 ipv4_redirect(skb
, net
, 0, 0, protocol
, 0);
325 vti_tunnel_ioctl(struct net_device
*dev
, struct ifreq
*ifr
, int cmd
)
328 struct ip_tunnel_parm p
;
330 if (copy_from_user(&p
, ifr
->ifr_ifru
.ifru_data
, sizeof(p
)))
333 if (cmd
== SIOCADDTUNNEL
|| cmd
== SIOCCHGTUNNEL
) {
334 if (p
.iph
.version
!= 4 || p
.iph
.protocol
!= IPPROTO_IPIP
||
339 if (!(p
.i_flags
& GRE_KEY
))
341 if (!(p
.o_flags
& GRE_KEY
))
344 p
.i_flags
= VTI_ISVTI
;
346 err
= ip_tunnel_ioctl(dev
, &p
, cmd
);
350 if (cmd
!= SIOCDELTUNNEL
) {
351 p
.i_flags
|= GRE_KEY
;
352 p
.o_flags
|= GRE_KEY
;
355 if (copy_to_user(ifr
->ifr_ifru
.ifru_data
, &p
, sizeof(p
)))
360 static const struct net_device_ops vti_netdev_ops
= {
361 .ndo_init
= vti_tunnel_init
,
362 .ndo_uninit
= ip_tunnel_uninit
,
363 .ndo_start_xmit
= vti_tunnel_xmit
,
364 .ndo_do_ioctl
= vti_tunnel_ioctl
,
365 .ndo_change_mtu
= ip_tunnel_change_mtu
,
366 .ndo_get_stats64
= ip_tunnel_get_stats64
,
367 .ndo_get_iflink
= ip_tunnel_get_iflink
,
370 static void vti_tunnel_setup(struct net_device
*dev
)
372 dev
->netdev_ops
= &vti_netdev_ops
;
373 dev
->type
= ARPHRD_TUNNEL
;
374 ip_tunnel_setup(dev
, vti_net_id
);
377 static int vti_tunnel_init(struct net_device
*dev
)
379 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
380 struct iphdr
*iph
= &tunnel
->parms
.iph
;
382 memcpy(dev
->dev_addr
, &iph
->saddr
, 4);
383 memcpy(dev
->broadcast
, &iph
->daddr
, 4);
385 dev
->hard_header_len
= LL_MAX_HEADER
+ sizeof(struct iphdr
);
386 dev
->mtu
= ETH_DATA_LEN
;
387 dev
->flags
= IFF_NOARP
;
389 dev
->features
|= NETIF_F_LLTX
;
392 return ip_tunnel_init(dev
);
395 static void __net_init
vti_fb_tunnel_init(struct net_device
*dev
)
397 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
398 struct iphdr
*iph
= &tunnel
->parms
.iph
;
401 iph
->protocol
= IPPROTO_IPIP
;
405 static struct xfrm4_protocol vti_esp4_protocol __read_mostly
= {
407 .input_handler
= vti_input
,
408 .cb_handler
= vti_rcv_cb
,
409 .err_handler
= vti4_err
,
413 static struct xfrm4_protocol vti_ah4_protocol __read_mostly
= {
415 .input_handler
= vti_input
,
416 .cb_handler
= vti_rcv_cb
,
417 .err_handler
= vti4_err
,
421 static struct xfrm4_protocol vti_ipcomp4_protocol __read_mostly
= {
423 .input_handler
= vti_input
,
424 .cb_handler
= vti_rcv_cb
,
425 .err_handler
= vti4_err
,
429 static int __net_init
vti_init_net(struct net
*net
)
432 struct ip_tunnel_net
*itn
;
434 err
= ip_tunnel_init_net(net
, vti_net_id
, &vti_link_ops
, "ip_vti0");
437 itn
= net_generic(net
, vti_net_id
);
438 vti_fb_tunnel_init(itn
->fb_tunnel_dev
);
442 static void __net_exit
vti_exit_net(struct net
*net
)
444 struct ip_tunnel_net
*itn
= net_generic(net
, vti_net_id
);
445 ip_tunnel_delete_net(itn
, &vti_link_ops
);
448 static struct pernet_operations vti_net_ops
= {
449 .init
= vti_init_net
,
450 .exit
= vti_exit_net
,
452 .size
= sizeof(struct ip_tunnel_net
),
455 static int vti_tunnel_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
460 static void vti_netlink_parms(struct nlattr
*data
[],
461 struct ip_tunnel_parm
*parms
)
463 memset(parms
, 0, sizeof(*parms
));
465 parms
->iph
.protocol
= IPPROTO_IPIP
;
470 parms
->i_flags
= VTI_ISVTI
;
472 if (data
[IFLA_VTI_LINK
])
473 parms
->link
= nla_get_u32(data
[IFLA_VTI_LINK
]);
475 if (data
[IFLA_VTI_IKEY
])
476 parms
->i_key
= nla_get_be32(data
[IFLA_VTI_IKEY
]);
478 if (data
[IFLA_VTI_OKEY
])
479 parms
->o_key
= nla_get_be32(data
[IFLA_VTI_OKEY
]);
481 if (data
[IFLA_VTI_LOCAL
])
482 parms
->iph
.saddr
= nla_get_in_addr(data
[IFLA_VTI_LOCAL
]);
484 if (data
[IFLA_VTI_REMOTE
])
485 parms
->iph
.daddr
= nla_get_in_addr(data
[IFLA_VTI_REMOTE
]);
489 static int vti_newlink(struct net
*src_net
, struct net_device
*dev
,
490 struct nlattr
*tb
[], struct nlattr
*data
[])
492 struct ip_tunnel_parm parms
;
494 vti_netlink_parms(data
, &parms
);
495 return ip_tunnel_newlink(dev
, tb
, &parms
);
498 static int vti_changelink(struct net_device
*dev
, struct nlattr
*tb
[],
499 struct nlattr
*data
[])
501 struct ip_tunnel_parm p
;
503 vti_netlink_parms(data
, &p
);
504 return ip_tunnel_changelink(dev
, tb
, &p
);
507 static size_t vti_get_size(const struct net_device
*dev
)
518 /* IFLA_VTI_REMOTE */
523 static int vti_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
525 struct ip_tunnel
*t
= netdev_priv(dev
);
526 struct ip_tunnel_parm
*p
= &t
->parms
;
528 nla_put_u32(skb
, IFLA_VTI_LINK
, p
->link
);
529 nla_put_be32(skb
, IFLA_VTI_IKEY
, p
->i_key
);
530 nla_put_be32(skb
, IFLA_VTI_OKEY
, p
->o_key
);
531 nla_put_in_addr(skb
, IFLA_VTI_LOCAL
, p
->iph
.saddr
);
532 nla_put_in_addr(skb
, IFLA_VTI_REMOTE
, p
->iph
.daddr
);
537 static const struct nla_policy vti_policy
[IFLA_VTI_MAX
+ 1] = {
538 [IFLA_VTI_LINK
] = { .type
= NLA_U32
},
539 [IFLA_VTI_IKEY
] = { .type
= NLA_U32
},
540 [IFLA_VTI_OKEY
] = { .type
= NLA_U32
},
541 [IFLA_VTI_LOCAL
] = { .len
= FIELD_SIZEOF(struct iphdr
, saddr
) },
542 [IFLA_VTI_REMOTE
] = { .len
= FIELD_SIZEOF(struct iphdr
, daddr
) },
545 static struct rtnl_link_ops vti_link_ops __read_mostly
= {
547 .maxtype
= IFLA_VTI_MAX
,
548 .policy
= vti_policy
,
549 .priv_size
= sizeof(struct ip_tunnel
),
550 .setup
= vti_tunnel_setup
,
551 .validate
= vti_tunnel_validate
,
552 .newlink
= vti_newlink
,
553 .changelink
= vti_changelink
,
554 .dellink
= ip_tunnel_dellink
,
555 .get_size
= vti_get_size
,
556 .fill_info
= vti_fill_info
,
557 .get_link_net
= ip_tunnel_get_link_net
,
560 static bool is_vti_tunnel(const struct net_device
*dev
)
562 return dev
->netdev_ops
== &vti_netdev_ops
;
565 static int vti_device_event(struct notifier_block
*unused
,
566 unsigned long event
, void *ptr
)
568 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
569 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
571 if (!is_vti_tunnel(dev
))
576 if (!net_eq(tunnel
->net
, dev_net(dev
)))
577 xfrm_garbage_collect(tunnel
->net
);
583 static struct notifier_block vti_notifier_block __read_mostly
= {
584 .notifier_call
= vti_device_event
,
587 static int __init
vti_init(void)
592 pr_info("IPv4 over IPsec tunneling driver\n");
594 register_netdevice_notifier(&vti_notifier_block
);
596 msg
= "tunnel device";
597 err
= register_pernet_device(&vti_net_ops
);
599 goto pernet_dev_failed
;
601 msg
= "tunnel protocols";
602 err
= xfrm4_protocol_register(&vti_esp4_protocol
, IPPROTO_ESP
);
604 goto xfrm_proto_esp_failed
;
605 err
= xfrm4_protocol_register(&vti_ah4_protocol
, IPPROTO_AH
);
607 goto xfrm_proto_ah_failed
;
608 err
= xfrm4_protocol_register(&vti_ipcomp4_protocol
, IPPROTO_COMP
);
610 goto xfrm_proto_comp_failed
;
612 msg
= "netlink interface";
613 err
= rtnl_link_register(&vti_link_ops
);
615 goto rtnl_link_failed
;
620 xfrm4_protocol_deregister(&vti_ipcomp4_protocol
, IPPROTO_COMP
);
621 xfrm_proto_comp_failed
:
622 xfrm4_protocol_deregister(&vti_ah4_protocol
, IPPROTO_AH
);
623 xfrm_proto_ah_failed
:
624 xfrm4_protocol_deregister(&vti_esp4_protocol
, IPPROTO_ESP
);
625 xfrm_proto_esp_failed
:
626 unregister_pernet_device(&vti_net_ops
);
628 unregister_netdevice_notifier(&vti_notifier_block
);
629 pr_err("vti init: failed to register %s\n", msg
);
633 static void __exit
vti_fini(void)
635 rtnl_link_unregister(&vti_link_ops
);
636 xfrm4_protocol_deregister(&vti_ipcomp4_protocol
, IPPROTO_COMP
);
637 xfrm4_protocol_deregister(&vti_ah4_protocol
, IPPROTO_AH
);
638 xfrm4_protocol_deregister(&vti_esp4_protocol
, IPPROTO_ESP
);
639 unregister_pernet_device(&vti_net_ops
);
640 unregister_netdevice_notifier(&vti_notifier_block
);
643 module_init(vti_init
);
644 module_exit(vti_fini
);
645 MODULE_LICENSE("GPL");
646 MODULE_ALIAS_RTNL_LINK("vti");
647 MODULE_ALIAS_NETDEV("ip_vti0");