1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * Linux NET3: IP/IP protocol decoder.
6 * Sam Lantinga (slouken@cs.ucdavis.edu) 02/01/95
9 * Alan Cox : Merged and made usable non modular (its so tiny its silly as
10 * a module taking up 2 pages).
11 * Alan Cox : Fixed bug with 1.3.18 and IPIP not working (now needs to set skb->h.iph)
12 * to keep ip_forward happy.
13 * Alan Cox : More fixes for 1.3.21, and firewall fix. Maybe this will work soon 8).
14 * Kai Schulte : Fixed #defines for IP_FIREWALL->FIREWALL
15 * David Woodhouse : Perform some basic ICMP handling.
16 * IPIP Routing without decapsulation.
17 * Carlos Picoto : GRE over IP support
18 * Alexey Kuznetsov: Reworked. Really, now it is truncated version of ipv4/ip_gre.c.
19 * I do not want to merge them together.
22 /* tunnel.c: an IP tunnel driver
24 The purpose of this driver is to provide an IP tunnel through
25 which you can tunnel network traffic transparently across subnets.
27 This was written by looking at Nick Holloway's dummy driver
28 Thanks for the great code!
30 -Sam Lantinga (slouken@cs.ucdavis.edu) 02/01/95
33 Cleaned up the code a little and added some pre-1.3.0 tweaks.
34 dev->hard_header/hard_header_len changed to use no headers.
35 Comments/bracketing tweaked.
36 Made the tunnels use dev->name not tunnel: when error reporting.
39 -Alan Cox (alan@lxorguk.ukuu.org.uk) 21 March 95
42 Changed to tunnel to destination gateway in addition to the
43 tunnel's pointopoint address
44 Almost completely rewritten
45 Note: There is currently no firewall or ICMP handling done.
47 -Sam Lantinga (slouken@cs.ucdavis.edu) 02/13/96
51 /* Things I wish I had known when writing the tunnel driver:
53 When the tunnel_xmit() function is called, the skb contains the
54 packet to be sent (plus a great deal of extra info), and dev
55 contains the tunnel device that _we_ are.
57 When we are passed a packet, we are expected to fill in the
58 source address with our source IP address.
60 What is the proper way to allocate, copy and free a buffer?
61 After you allocate it, it is a "0 length" chunk of memory
62 starting at zero. If you want to add headers to the buffer
63 later, you'll have to call "skb_reserve(skb, amount)" with
64 the amount of memory you want reserved. Then, you call
65 "skb_put(skb, amount)" with the amount of space you want in
66 the buffer. skb_put() returns a pointer to the top (#0) of
67 that buffer. skb->len is set to the amount of space you have
68 "allocated" with skb_put(). You can then write up to skb->len
69 bytes to that buffer. If you need more, you can call skb_put()
70 again with the additional amount of space you need. You can
71 find out how much more space you can allocate by calling
73 Now, to add header space, call "skb_push(skb, header_len)".
74 This creates space at the beginning of the buffer and returns
75 a pointer to this new space. If later you need to strip a
76 header from a buffer, call "skb_pull(skb, header_len)".
77 skb_headroom() will return how much space is left at the top
78 of the buffer (before the main data). Remember, this headroom
79 space must be reserved before the skb_put() function is called.
83 This version of net/ipv4/ipip.c is cloned of net/ipv4/ip_gre.c
85 For comments look at net/ipv4/ip_gre.c --ANK
89 #include <linux/capability.h>
90 #include <linux/module.h>
91 #include <linux/types.h>
92 #include <linux/kernel.h>
93 #include <linux/slab.h>
94 #include <linux/uaccess.h>
95 #include <linux/skbuff.h>
96 #include <linux/netdevice.h>
98 #include <linux/tcp.h>
99 #include <linux/udp.h>
100 #include <linux/if_arp.h>
101 #include <linux/init.h>
102 #include <linux/netfilter_ipv4.h>
103 #include <linux/if_ether.h>
105 #include <net/sock.h>
107 #include <net/icmp.h>
108 #include <net/ip_tunnels.h>
109 #include <net/inet_ecn.h>
110 #include <net/xfrm.h>
111 #include <net/net_namespace.h>
112 #include <net/netns/generic.h>
113 #include <net/dst_metadata.h>
115 static bool log_ecn_error
= true;
116 module_param(log_ecn_error
, bool, 0644);
117 MODULE_PARM_DESC(log_ecn_error
, "Log packets received with corrupted ECN");
119 static unsigned int ipip_net_id __read_mostly
;
121 static int ipip_tunnel_init(struct net_device
*dev
);
122 static struct rtnl_link_ops ipip_link_ops __read_mostly
;
124 static int ipip_err(struct sk_buff
*skb
, u32 info
)
126 /* All the routers (except for Linux) return only
127 * 8 bytes of packet payload. It means, that precise relaying of
128 * ICMP in the real Internet is absolutely infeasible.
130 struct net
*net
= dev_net(skb
->dev
);
131 struct ip_tunnel_net
*itn
= net_generic(net
, ipip_net_id
);
132 const struct iphdr
*iph
= (const struct iphdr
*)skb
->data
;
133 const int type
= icmp_hdr(skb
)->type
;
134 const int code
= icmp_hdr(skb
)->code
;
138 t
= ip_tunnel_lookup(itn
, skb
->dev
->ifindex
, TUNNEL_NO_KEY
,
139 iph
->daddr
, iph
->saddr
, 0);
146 case ICMP_DEST_UNREACH
:
149 /* Impossible event. */
152 /* All others are translated to HOST_UNREACH.
153 * rfc2003 contains "deep thoughts" about NET_UNREACH,
154 * I believe they are just ether pollution. --ANK
160 case ICMP_TIME_EXCEEDED
:
161 if (code
!= ICMP_EXC_TTL
)
172 if (type
== ICMP_DEST_UNREACH
&& code
== ICMP_FRAG_NEEDED
) {
173 ipv4_update_pmtu(skb
, net
, info
, t
->parms
.link
, iph
->protocol
);
177 if (type
== ICMP_REDIRECT
) {
178 ipv4_redirect(skb
, net
, t
->parms
.link
, iph
->protocol
);
182 if (t
->parms
.iph
.daddr
== 0) {
187 if (t
->parms
.iph
.ttl
== 0 && type
== ICMP_TIME_EXCEEDED
)
190 if (time_before(jiffies
, t
->err_time
+ IPTUNNEL_ERR_TIMEO
))
194 t
->err_time
= jiffies
;
200 static const struct tnl_ptk_info ipip_tpi
= {
201 /* no tunnel info required for ipip. */
202 .proto
= htons(ETH_P_IP
),
205 #if IS_ENABLED(CONFIG_MPLS)
206 static const struct tnl_ptk_info mplsip_tpi
= {
207 /* no tunnel info required for mplsip. */
208 .proto
= htons(ETH_P_MPLS_UC
),
212 static int ipip_tunnel_rcv(struct sk_buff
*skb
, u8 ipproto
)
214 struct net
*net
= dev_net(skb
->dev
);
215 struct ip_tunnel_net
*itn
= net_generic(net
, ipip_net_id
);
216 struct metadata_dst
*tun_dst
= NULL
;
217 struct ip_tunnel
*tunnel
;
218 const struct iphdr
*iph
;
221 tunnel
= ip_tunnel_lookup(itn
, skb
->dev
->ifindex
, TUNNEL_NO_KEY
,
222 iph
->saddr
, iph
->daddr
, 0);
224 const struct tnl_ptk_info
*tpi
;
226 if (tunnel
->parms
.iph
.protocol
!= ipproto
&&
227 tunnel
->parms
.iph
.protocol
!= 0)
230 if (!xfrm4_policy_check(NULL
, XFRM_POLICY_IN
, skb
))
232 #if IS_ENABLED(CONFIG_MPLS)
233 if (ipproto
== IPPROTO_MPLS
)
238 if (iptunnel_pull_header(skb
, 0, tpi
->proto
, false))
240 if (tunnel
->collect_md
) {
241 tun_dst
= ip_tun_rx_dst(skb
, 0, 0, 0);
245 return ip_tunnel_rcv(tunnel
, skb
, tpi
, tun_dst
, log_ecn_error
);
255 static int ipip_rcv(struct sk_buff
*skb
)
257 return ipip_tunnel_rcv(skb
, IPPROTO_IPIP
);
260 #if IS_ENABLED(CONFIG_MPLS)
261 static int mplsip_rcv(struct sk_buff
*skb
)
263 return ipip_tunnel_rcv(skb
, IPPROTO_MPLS
);
268 * This function assumes it is being called from dev_queue_xmit()
269 * and that skb is filled properly by that function.
271 static netdev_tx_t
ipip_tunnel_xmit(struct sk_buff
*skb
,
272 struct net_device
*dev
)
274 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
275 const struct iphdr
*tiph
= &tunnel
->parms
.iph
;
278 switch (skb
->protocol
) {
279 case htons(ETH_P_IP
):
280 ipproto
= IPPROTO_IPIP
;
282 #if IS_ENABLED(CONFIG_MPLS)
283 case htons(ETH_P_MPLS_UC
):
284 ipproto
= IPPROTO_MPLS
;
291 if (tiph
->protocol
!= ipproto
&& tiph
->protocol
!= 0)
294 if (iptunnel_handle_offloads(skb
, SKB_GSO_IPXIP4
))
297 skb_set_inner_ipproto(skb
, ipproto
);
299 if (tunnel
->collect_md
)
300 ip_md_tunnel_xmit(skb
, dev
, ipproto
, 0);
302 ip_tunnel_xmit(skb
, dev
, tiph
, ipproto
);
308 dev
->stats
.tx_errors
++;
312 static bool ipip_tunnel_ioctl_verify_protocol(u8 ipproto
)
317 #if IS_ENABLED(CONFIG_MPLS)
327 ipip_tunnel_ioctl(struct net_device
*dev
, struct ifreq
*ifr
, int cmd
)
330 struct ip_tunnel_parm p
;
332 if (copy_from_user(&p
, ifr
->ifr_ifru
.ifru_data
, sizeof(p
)))
335 if (cmd
== SIOCADDTUNNEL
|| cmd
== SIOCCHGTUNNEL
) {
336 if (p
.iph
.version
!= 4 ||
337 !ipip_tunnel_ioctl_verify_protocol(p
.iph
.protocol
) ||
338 p
.iph
.ihl
!= 5 || (p
.iph
.frag_off
&htons(~IP_DF
)))
342 p
.i_key
= p
.o_key
= 0;
343 p
.i_flags
= p
.o_flags
= 0;
344 err
= ip_tunnel_ioctl(dev
, &p
, cmd
);
348 if (copy_to_user(ifr
->ifr_ifru
.ifru_data
, &p
, sizeof(p
)))
354 static const struct net_device_ops ipip_netdev_ops
= {
355 .ndo_init
= ipip_tunnel_init
,
356 .ndo_uninit
= ip_tunnel_uninit
,
357 .ndo_start_xmit
= ipip_tunnel_xmit
,
358 .ndo_do_ioctl
= ipip_tunnel_ioctl
,
359 .ndo_change_mtu
= ip_tunnel_change_mtu
,
360 .ndo_get_stats64
= ip_tunnel_get_stats64
,
361 .ndo_get_iflink
= ip_tunnel_get_iflink
,
364 #define IPIP_FEATURES (NETIF_F_SG | \
367 NETIF_F_GSO_SOFTWARE | \
370 static void ipip_tunnel_setup(struct net_device
*dev
)
372 dev
->netdev_ops
= &ipip_netdev_ops
;
374 dev
->type
= ARPHRD_TUNNEL
;
375 dev
->flags
= IFF_NOARP
;
377 dev
->features
|= NETIF_F_LLTX
;
380 dev
->features
|= IPIP_FEATURES
;
381 dev
->hw_features
|= IPIP_FEATURES
;
382 ip_tunnel_setup(dev
, ipip_net_id
);
385 static int ipip_tunnel_init(struct net_device
*dev
)
387 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
389 memcpy(dev
->dev_addr
, &tunnel
->parms
.iph
.saddr
, 4);
390 memcpy(dev
->broadcast
, &tunnel
->parms
.iph
.daddr
, 4);
392 tunnel
->tun_hlen
= 0;
393 tunnel
->hlen
= tunnel
->tun_hlen
+ tunnel
->encap_hlen
;
394 return ip_tunnel_init(dev
);
397 static int ipip_tunnel_validate(struct nlattr
*tb
[], struct nlattr
*data
[],
398 struct netlink_ext_ack
*extack
)
402 if (!data
|| !data
[IFLA_IPTUN_PROTO
])
405 proto
= nla_get_u8(data
[IFLA_IPTUN_PROTO
]);
406 if (proto
!= IPPROTO_IPIP
&& proto
!= IPPROTO_MPLS
&& proto
!= 0)
412 static void ipip_netlink_parms(struct nlattr
*data
[],
413 struct ip_tunnel_parm
*parms
, bool *collect_md
,
416 memset(parms
, 0, sizeof(*parms
));
418 parms
->iph
.version
= 4;
419 parms
->iph
.protocol
= IPPROTO_IPIP
;
426 if (data
[IFLA_IPTUN_LINK
])
427 parms
->link
= nla_get_u32(data
[IFLA_IPTUN_LINK
]);
429 if (data
[IFLA_IPTUN_LOCAL
])
430 parms
->iph
.saddr
= nla_get_in_addr(data
[IFLA_IPTUN_LOCAL
]);
432 if (data
[IFLA_IPTUN_REMOTE
])
433 parms
->iph
.daddr
= nla_get_in_addr(data
[IFLA_IPTUN_REMOTE
]);
435 if (data
[IFLA_IPTUN_TTL
]) {
436 parms
->iph
.ttl
= nla_get_u8(data
[IFLA_IPTUN_TTL
]);
438 parms
->iph
.frag_off
= htons(IP_DF
);
441 if (data
[IFLA_IPTUN_TOS
])
442 parms
->iph
.tos
= nla_get_u8(data
[IFLA_IPTUN_TOS
]);
444 if (data
[IFLA_IPTUN_PROTO
])
445 parms
->iph
.protocol
= nla_get_u8(data
[IFLA_IPTUN_PROTO
]);
447 if (!data
[IFLA_IPTUN_PMTUDISC
] || nla_get_u8(data
[IFLA_IPTUN_PMTUDISC
]))
448 parms
->iph
.frag_off
= htons(IP_DF
);
450 if (data
[IFLA_IPTUN_COLLECT_METADATA
])
453 if (data
[IFLA_IPTUN_FWMARK
])
454 *fwmark
= nla_get_u32(data
[IFLA_IPTUN_FWMARK
]);
457 /* This function returns true when ENCAP attributes are present in the nl msg */
458 static bool ipip_netlink_encap_parms(struct nlattr
*data
[],
459 struct ip_tunnel_encap
*ipencap
)
463 memset(ipencap
, 0, sizeof(*ipencap
));
468 if (data
[IFLA_IPTUN_ENCAP_TYPE
]) {
470 ipencap
->type
= nla_get_u16(data
[IFLA_IPTUN_ENCAP_TYPE
]);
473 if (data
[IFLA_IPTUN_ENCAP_FLAGS
]) {
475 ipencap
->flags
= nla_get_u16(data
[IFLA_IPTUN_ENCAP_FLAGS
]);
478 if (data
[IFLA_IPTUN_ENCAP_SPORT
]) {
480 ipencap
->sport
= nla_get_be16(data
[IFLA_IPTUN_ENCAP_SPORT
]);
483 if (data
[IFLA_IPTUN_ENCAP_DPORT
]) {
485 ipencap
->dport
= nla_get_be16(data
[IFLA_IPTUN_ENCAP_DPORT
]);
491 static int ipip_newlink(struct net
*src_net
, struct net_device
*dev
,
492 struct nlattr
*tb
[], struct nlattr
*data
[],
493 struct netlink_ext_ack
*extack
)
495 struct ip_tunnel
*t
= netdev_priv(dev
);
496 struct ip_tunnel_parm p
;
497 struct ip_tunnel_encap ipencap
;
500 if (ipip_netlink_encap_parms(data
, &ipencap
)) {
501 int err
= ip_tunnel_encap_setup(t
, &ipencap
);
507 ipip_netlink_parms(data
, &p
, &t
->collect_md
, &fwmark
);
508 return ip_tunnel_newlink(dev
, tb
, &p
, fwmark
);
511 static int ipip_changelink(struct net_device
*dev
, struct nlattr
*tb
[],
512 struct nlattr
*data
[],
513 struct netlink_ext_ack
*extack
)
515 struct ip_tunnel
*t
= netdev_priv(dev
);
516 struct ip_tunnel_parm p
;
517 struct ip_tunnel_encap ipencap
;
519 __u32 fwmark
= t
->fwmark
;
521 if (ipip_netlink_encap_parms(data
, &ipencap
)) {
522 int err
= ip_tunnel_encap_setup(t
, &ipencap
);
528 ipip_netlink_parms(data
, &p
, &collect_md
, &fwmark
);
532 if (((dev
->flags
& IFF_POINTOPOINT
) && !p
.iph
.daddr
) ||
533 (!(dev
->flags
& IFF_POINTOPOINT
) && p
.iph
.daddr
))
536 return ip_tunnel_changelink(dev
, tb
, &p
, fwmark
);
539 static size_t ipip_get_size(const struct net_device
*dev
)
542 /* IFLA_IPTUN_LINK */
544 /* IFLA_IPTUN_LOCAL */
546 /* IFLA_IPTUN_REMOTE */
552 /* IFLA_IPTUN_PROTO */
554 /* IFLA_IPTUN_PMTUDISC */
556 /* IFLA_IPTUN_ENCAP_TYPE */
558 /* IFLA_IPTUN_ENCAP_FLAGS */
560 /* IFLA_IPTUN_ENCAP_SPORT */
562 /* IFLA_IPTUN_ENCAP_DPORT */
564 /* IFLA_IPTUN_COLLECT_METADATA */
566 /* IFLA_IPTUN_FWMARK */
571 static int ipip_fill_info(struct sk_buff
*skb
, const struct net_device
*dev
)
573 struct ip_tunnel
*tunnel
= netdev_priv(dev
);
574 struct ip_tunnel_parm
*parm
= &tunnel
->parms
;
576 if (nla_put_u32(skb
, IFLA_IPTUN_LINK
, parm
->link
) ||
577 nla_put_in_addr(skb
, IFLA_IPTUN_LOCAL
, parm
->iph
.saddr
) ||
578 nla_put_in_addr(skb
, IFLA_IPTUN_REMOTE
, parm
->iph
.daddr
) ||
579 nla_put_u8(skb
, IFLA_IPTUN_TTL
, parm
->iph
.ttl
) ||
580 nla_put_u8(skb
, IFLA_IPTUN_TOS
, parm
->iph
.tos
) ||
581 nla_put_u8(skb
, IFLA_IPTUN_PROTO
, parm
->iph
.protocol
) ||
582 nla_put_u8(skb
, IFLA_IPTUN_PMTUDISC
,
583 !!(parm
->iph
.frag_off
& htons(IP_DF
))) ||
584 nla_put_u32(skb
, IFLA_IPTUN_FWMARK
, tunnel
->fwmark
))
585 goto nla_put_failure
;
587 if (nla_put_u16(skb
, IFLA_IPTUN_ENCAP_TYPE
,
588 tunnel
->encap
.type
) ||
589 nla_put_be16(skb
, IFLA_IPTUN_ENCAP_SPORT
,
590 tunnel
->encap
.sport
) ||
591 nla_put_be16(skb
, IFLA_IPTUN_ENCAP_DPORT
,
592 tunnel
->encap
.dport
) ||
593 nla_put_u16(skb
, IFLA_IPTUN_ENCAP_FLAGS
,
594 tunnel
->encap
.flags
))
595 goto nla_put_failure
;
597 if (tunnel
->collect_md
)
598 if (nla_put_flag(skb
, IFLA_IPTUN_COLLECT_METADATA
))
599 goto nla_put_failure
;
606 static const struct nla_policy ipip_policy
[IFLA_IPTUN_MAX
+ 1] = {
607 [IFLA_IPTUN_LINK
] = { .type
= NLA_U32
},
608 [IFLA_IPTUN_LOCAL
] = { .type
= NLA_U32
},
609 [IFLA_IPTUN_REMOTE
] = { .type
= NLA_U32
},
610 [IFLA_IPTUN_TTL
] = { .type
= NLA_U8
},
611 [IFLA_IPTUN_TOS
] = { .type
= NLA_U8
},
612 [IFLA_IPTUN_PROTO
] = { .type
= NLA_U8
},
613 [IFLA_IPTUN_PMTUDISC
] = { .type
= NLA_U8
},
614 [IFLA_IPTUN_ENCAP_TYPE
] = { .type
= NLA_U16
},
615 [IFLA_IPTUN_ENCAP_FLAGS
] = { .type
= NLA_U16
},
616 [IFLA_IPTUN_ENCAP_SPORT
] = { .type
= NLA_U16
},
617 [IFLA_IPTUN_ENCAP_DPORT
] = { .type
= NLA_U16
},
618 [IFLA_IPTUN_COLLECT_METADATA
] = { .type
= NLA_FLAG
},
619 [IFLA_IPTUN_FWMARK
] = { .type
= NLA_U32
},
622 static struct rtnl_link_ops ipip_link_ops __read_mostly
= {
624 .maxtype
= IFLA_IPTUN_MAX
,
625 .policy
= ipip_policy
,
626 .priv_size
= sizeof(struct ip_tunnel
),
627 .setup
= ipip_tunnel_setup
,
628 .validate
= ipip_tunnel_validate
,
629 .newlink
= ipip_newlink
,
630 .changelink
= ipip_changelink
,
631 .dellink
= ip_tunnel_dellink
,
632 .get_size
= ipip_get_size
,
633 .fill_info
= ipip_fill_info
,
634 .get_link_net
= ip_tunnel_get_link_net
,
637 static struct xfrm_tunnel ipip_handler __read_mostly
= {
639 .err_handler
= ipip_err
,
643 #if IS_ENABLED(CONFIG_MPLS)
644 static struct xfrm_tunnel mplsip_handler __read_mostly
= {
645 .handler
= mplsip_rcv
,
646 .err_handler
= ipip_err
,
651 static int __net_init
ipip_init_net(struct net
*net
)
653 return ip_tunnel_init_net(net
, ipip_net_id
, &ipip_link_ops
, "tunl0");
656 static void __net_exit
ipip_exit_batch_net(struct list_head
*list_net
)
658 ip_tunnel_delete_nets(list_net
, ipip_net_id
, &ipip_link_ops
);
661 static struct pernet_operations ipip_net_ops
= {
662 .init
= ipip_init_net
,
663 .exit_batch
= ipip_exit_batch_net
,
665 .size
= sizeof(struct ip_tunnel_net
),
668 static int __init
ipip_init(void)
672 pr_info("ipip: IPv4 and MPLS over IPv4 tunneling driver\n");
674 err
= register_pernet_device(&ipip_net_ops
);
677 err
= xfrm4_tunnel_register(&ipip_handler
, AF_INET
);
679 pr_info("%s: can't register tunnel\n", __func__
);
680 goto xfrm_tunnel_ipip_failed
;
682 #if IS_ENABLED(CONFIG_MPLS)
683 err
= xfrm4_tunnel_register(&mplsip_handler
, AF_MPLS
);
685 pr_info("%s: can't register tunnel\n", __func__
);
686 goto xfrm_tunnel_mplsip_failed
;
689 err
= rtnl_link_register(&ipip_link_ops
);
691 goto rtnl_link_failed
;
697 #if IS_ENABLED(CONFIG_MPLS)
698 xfrm4_tunnel_deregister(&mplsip_handler
, AF_INET
);
699 xfrm_tunnel_mplsip_failed
:
702 xfrm4_tunnel_deregister(&ipip_handler
, AF_INET
);
703 xfrm_tunnel_ipip_failed
:
704 unregister_pernet_device(&ipip_net_ops
);
708 static void __exit
ipip_fini(void)
710 rtnl_link_unregister(&ipip_link_ops
);
711 if (xfrm4_tunnel_deregister(&ipip_handler
, AF_INET
))
712 pr_info("%s: can't deregister tunnel\n", __func__
);
713 #if IS_ENABLED(CONFIG_MPLS)
714 if (xfrm4_tunnel_deregister(&mplsip_handler
, AF_MPLS
))
715 pr_info("%s: can't deregister tunnel\n", __func__
);
717 unregister_pernet_device(&ipip_net_ops
);
720 module_init(ipip_init
);
721 module_exit(ipip_fini
);
722 MODULE_LICENSE("GPL");
723 MODULE_ALIAS_RTNL_LINK("ipip");
724 MODULE_ALIAS_NETDEV("tunl0");