xref: /linux/net/ipv4/ipip.c (revision 333f7de560e1196034b67db16916b10a0c529e1d)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	Linux NET3:	IP/IP protocol decoder.
4  *
5  *	Authors:
6  *		Sam Lantinga (slouken@cs.ucdavis.edu)  02/01/95
7  *
8  *	Fixes:
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.
20  */
21 
22 /* tunnel.c: an IP tunnel driver
23 
24 	The purpose of this driver is to provide an IP tunnel through
25 	which you can tunnel network traffic transparently across subnets.
26 
27 	This was written by looking at Nick Holloway's dummy driver
28 	Thanks for the great code!
29 
30 		-Sam Lantinga	(slouken@cs.ucdavis.edu)  02/01/95
31 
32 	Minor tweaks:
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.
37 		Added tx_dropped stat
38 
39 		-Alan Cox	(alan@lxorguk.ukuu.org.uk) 21 March 95
40 
41 	Reworked:
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.
46 
47 		-Sam Lantinga	(slouken@cs.ucdavis.edu) 02/13/96
48 
49 */
50 
51 /* Things I wish I had known when writing the tunnel driver:
52 
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.
56 
57 	When we are passed a packet, we are expected to fill in the
58 	source address with our source IP address.
59 
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
72 	"skb_tailroom(skb)".
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.
80 	*/
81 
82 /*
83    This version of net/ipv4/ipip.c is cloned of net/ipv4/ip_gre.c
84 
85    For comments look at net/ipv4/ip_gre.c --ANK
86  */
87 
88 
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>
97 #include <linux/in.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>
104 
105 #include <net/sock.h>
106 #include <net/ip.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>
114 
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");
118 
119 static unsigned int ipip_net_id __read_mostly;
120 
121 static int ipip_tunnel_init(struct net_device *dev);
122 static struct rtnl_link_ops ipip_link_ops __read_mostly;
123 
124 static int ipip_err(struct sk_buff *skb, u32 info)
125 {
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.
129 	 */
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 	IP_TUNNEL_DECLARE_FLAGS(flags) = { };
134 	const int type = icmp_hdr(skb)->type;
135 	const int code = icmp_hdr(skb)->code;
136 	struct ip_tunnel *t;
137 	int err = 0;
138 
139 	__set_bit(IP_TUNNEL_NO_KEY_BIT, flags);
140 
141 	t = ip_tunnel_lookup(itn, skb->dev->ifindex, flags, iph->daddr,
142 			     iph->saddr, 0);
143 	if (!t) {
144 		err = -ENOENT;
145 		goto out;
146 	}
147 
148 	switch (type) {
149 	case ICMP_DEST_UNREACH:
150 		switch (code) {
151 		case ICMP_SR_FAILED:
152 			/* Impossible event. */
153 			goto out;
154 		default:
155 			/* All others are translated to HOST_UNREACH.
156 			 * rfc2003 contains "deep thoughts" about NET_UNREACH,
157 			 * I believe they are just ether pollution. --ANK
158 			 */
159 			break;
160 		}
161 		break;
162 
163 	case ICMP_TIME_EXCEEDED:
164 		if (code != ICMP_EXC_TTL)
165 			goto out;
166 		break;
167 
168 	case ICMP_REDIRECT:
169 		break;
170 
171 	default:
172 		goto out;
173 	}
174 
175 	if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
176 		ipv4_update_pmtu(skb, net, info, t->parms.link, iph->protocol);
177 		goto out;
178 	}
179 
180 	if (type == ICMP_REDIRECT) {
181 		ipv4_redirect(skb, net, t->parms.link, iph->protocol);
182 		goto out;
183 	}
184 
185 	if (t->parms.iph.daddr == 0) {
186 		err = -ENOENT;
187 		goto out;
188 	}
189 
190 	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
191 		goto out;
192 
193 	if (time_before(jiffies, READ_ONCE(t->err_time) + IPTUNNEL_ERR_TIMEO))
194 		WRITE_ONCE(t->err_count, READ_ONCE(t->err_count) + 1);
195 	else
196 		WRITE_ONCE(t->err_count, 1);
197 	WRITE_ONCE(t->err_time, jiffies);
198 
199 out:
200 	return err;
201 }
202 
203 static const struct tnl_ptk_info ipip_tpi = {
204 	/* no tunnel info required for ipip. */
205 	.proto = htons(ETH_P_IP),
206 };
207 
208 #if IS_ENABLED(CONFIG_MPLS)
209 static const struct tnl_ptk_info mplsip_tpi = {
210 	/* no tunnel info required for mplsip. */
211 	.proto = htons(ETH_P_MPLS_UC),
212 };
213 #endif
214 
215 static int ipip_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
216 {
217 	struct net *net = dev_net(skb->dev);
218 	struct ip_tunnel_net *itn = net_generic(net, ipip_net_id);
219 	IP_TUNNEL_DECLARE_FLAGS(flags) = { };
220 	struct metadata_dst *tun_dst = NULL;
221 	struct ip_tunnel *tunnel;
222 	const struct iphdr *iph;
223 
224 	__set_bit(IP_TUNNEL_NO_KEY_BIT, flags);
225 
226 	iph = ip_hdr(skb);
227 	tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, flags, iph->saddr,
228 				  iph->daddr, 0);
229 	if (tunnel) {
230 		const struct tnl_ptk_info *tpi;
231 
232 		if (tunnel->parms.iph.protocol != ipproto &&
233 		    tunnel->parms.iph.protocol != 0)
234 			goto drop;
235 
236 		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
237 			goto drop;
238 #if IS_ENABLED(CONFIG_MPLS)
239 		if (ipproto == IPPROTO_MPLS)
240 			tpi = &mplsip_tpi;
241 		else
242 #endif
243 			tpi = &ipip_tpi;
244 		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
245 			goto drop;
246 		if (tunnel->collect_md) {
247 			ip_tunnel_flags_zero(flags);
248 
249 			tun_dst = ip_tun_rx_dst(skb, flags, 0, 0);
250 			if (!tun_dst)
251 				return 0;
252 			ip_tunnel_md_udp_encap(skb, &tun_dst->u.tun_info);
253 		}
254 		skb_reset_mac_header(skb);
255 
256 		return ip_tunnel_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
257 	}
258 
259 	return -1;
260 
261 drop:
262 	kfree_skb(skb);
263 	return 0;
264 }
265 
266 static int ipip_rcv(struct sk_buff *skb)
267 {
268 	return ipip_tunnel_rcv(skb, IPPROTO_IPIP);
269 }
270 
271 #if IS_ENABLED(CONFIG_MPLS)
272 static int mplsip_rcv(struct sk_buff *skb)
273 {
274 	return ipip_tunnel_rcv(skb, IPPROTO_MPLS);
275 }
276 #endif
277 
278 /*
279  *	This function assumes it is being called from dev_queue_xmit()
280  *	and that skb is filled properly by that function.
281  */
282 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb,
283 				    struct net_device *dev)
284 {
285 	struct ip_tunnel *tunnel = netdev_priv(dev);
286 	const struct iphdr  *tiph = &tunnel->parms.iph;
287 	u8 ipproto;
288 
289 	if (!pskb_inet_may_pull(skb))
290 		goto tx_error;
291 
292 	switch (skb->protocol) {
293 	case htons(ETH_P_IP):
294 		ipproto = IPPROTO_IPIP;
295 		break;
296 #if IS_ENABLED(CONFIG_MPLS)
297 	case htons(ETH_P_MPLS_UC):
298 		ipproto = IPPROTO_MPLS;
299 		break;
300 #endif
301 	default:
302 		goto tx_error;
303 	}
304 
305 	if (tiph->protocol != ipproto && tiph->protocol != 0)
306 		goto tx_error;
307 
308 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
309 		goto tx_error;
310 
311 	skb_set_inner_ipproto(skb, ipproto);
312 
313 	if (tunnel->collect_md)
314 		ip_md_tunnel_xmit(skb, dev, ipproto, 0);
315 	else
316 		ip_tunnel_xmit(skb, dev, tiph, ipproto);
317 	return NETDEV_TX_OK;
318 
319 tx_error:
320 	kfree_skb(skb);
321 
322 	DEV_STATS_INC(dev, tx_errors);
323 	return NETDEV_TX_OK;
324 }
325 
326 static bool ipip_tunnel_ioctl_verify_protocol(u8 ipproto)
327 {
328 	switch (ipproto) {
329 	case 0:
330 	case IPPROTO_IPIP:
331 #if IS_ENABLED(CONFIG_MPLS)
332 	case IPPROTO_MPLS:
333 #endif
334 		return true;
335 	}
336 
337 	return false;
338 }
339 
340 static int
341 ipip_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm_kern *p, int cmd)
342 {
343 	if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
344 		if (p->iph.version != 4 ||
345 		    !ipip_tunnel_ioctl_verify_protocol(p->iph.protocol) ||
346 		    p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)))
347 			return -EINVAL;
348 	}
349 
350 	p->i_key = p->o_key = 0;
351 	ip_tunnel_flags_zero(p->i_flags);
352 	ip_tunnel_flags_zero(p->o_flags);
353 	return ip_tunnel_ctl(dev, p, cmd);
354 }
355 
356 static int ipip_fill_forward_path(struct net_device_path_ctx *ctx,
357 				  struct net_device_path *path)
358 {
359 	struct ip_tunnel *tunnel = netdev_priv(ctx->dev);
360 	const struct iphdr *tiph = &tunnel->parms.iph;
361 	struct rtable *rt;
362 
363 	rt = ip_route_output(dev_net(ctx->dev), tiph->daddr, tiph->saddr,
364 			     inet_dsfield_to_dscp(tiph->tos),
365 			     tunnel->parms.link, RT_SCOPE_UNIVERSE);
366 	if (IS_ERR(rt))
367 		return PTR_ERR(rt);
368 
369 	path->type = DEV_PATH_TUN;
370 	path->tun.src_v4.s_addr = tiph->saddr;
371 	path->tun.dst_v4.s_addr = tiph->daddr;
372 	path->tun.l3_proto = IPPROTO_IPIP;
373 	path->dev = ctx->dev;
374 
375 	ctx->dev = rt->dst.dev;
376 	ip_rt_put(rt);
377 
378 	return 0;
379 }
380 
381 static const struct net_device_ops ipip_netdev_ops = {
382 	.ndo_init       = ipip_tunnel_init,
383 	.ndo_uninit     = ip_tunnel_uninit,
384 	.ndo_start_xmit	= ipip_tunnel_xmit,
385 	.ndo_siocdevprivate = ip_tunnel_siocdevprivate,
386 	.ndo_change_mtu = ip_tunnel_change_mtu,
387 	.ndo_get_stats64 = dev_get_tstats64,
388 	.ndo_get_iflink = ip_tunnel_get_iflink,
389 	.ndo_tunnel_ctl	= ipip_tunnel_ctl,
390 	.ndo_fill_forward_path = ipip_fill_forward_path,
391 };
392 
393 #define IPIP_FEATURES (NETIF_F_SG |		\
394 		       NETIF_F_FRAGLIST |	\
395 		       NETIF_F_HIGHDMA |	\
396 		       NETIF_F_GSO_SOFTWARE |	\
397 		       NETIF_F_HW_CSUM)
398 
399 static void ipip_tunnel_setup(struct net_device *dev)
400 {
401 	dev->netdev_ops		= &ipip_netdev_ops;
402 	dev->header_ops		= &ip_tunnel_header_ops;
403 
404 	dev->type		= ARPHRD_TUNNEL;
405 	dev->flags		= IFF_NOARP;
406 	dev->addr_len		= 4;
407 	dev->lltx		= true;
408 	netif_keep_dst(dev);
409 
410 	dev->features		|= IPIP_FEATURES;
411 	dev->hw_features	|= IPIP_FEATURES;
412 	ip_tunnel_setup(dev, ipip_net_id);
413 }
414 
415 static int ipip_tunnel_init(struct net_device *dev)
416 {
417 	struct ip_tunnel *tunnel = netdev_priv(dev);
418 
419 	__dev_addr_set(dev, &tunnel->parms.iph.saddr, 4);
420 	memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
421 
422 	tunnel->tun_hlen = 0;
423 	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
424 	return ip_tunnel_init(dev);
425 }
426 
427 static int ipip_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
428 				struct netlink_ext_ack *extack)
429 {
430 	u8 proto;
431 
432 	if (!data || !data[IFLA_IPTUN_PROTO])
433 		return 0;
434 
435 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
436 	if (proto != IPPROTO_IPIP && proto != IPPROTO_MPLS && proto != 0)
437 		return -EINVAL;
438 
439 	return 0;
440 }
441 
442 static void ipip_netlink_parms(struct nlattr *data[],
443 			       struct ip_tunnel_parm_kern *parms,
444 			       bool *collect_md, __u32 *fwmark)
445 {
446 	memset(parms, 0, sizeof(*parms));
447 
448 	parms->iph.version = 4;
449 	parms->iph.protocol = IPPROTO_IPIP;
450 	parms->iph.ihl = 5;
451 	*collect_md = false;
452 
453 	if (!data)
454 		return;
455 
456 	ip_tunnel_netlink_parms(data, parms);
457 
458 	if (data[IFLA_IPTUN_COLLECT_METADATA])
459 		*collect_md = true;
460 
461 	if (data[IFLA_IPTUN_FWMARK])
462 		*fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
463 }
464 
465 static int ipip_newlink(struct net_device *dev,
466 			struct rtnl_newlink_params *params,
467 			struct netlink_ext_ack *extack)
468 {
469 	struct ip_tunnel *t = netdev_priv(dev);
470 	struct nlattr **data = params->data;
471 	struct nlattr **tb = params->tb;
472 	struct ip_tunnel_encap ipencap;
473 	struct ip_tunnel_parm_kern p;
474 	__u32 fwmark = 0;
475 
476 	if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
477 		int err = ip_tunnel_encap_setup(t, &ipencap);
478 
479 		if (err < 0)
480 			return err;
481 	}
482 
483 	ipip_netlink_parms(data, &p, &t->collect_md, &fwmark);
484 	return ip_tunnel_newlink(params->link_net ? : dev_net(dev), dev, tb, &p,
485 				 fwmark);
486 }
487 
488 static int ipip_changelink(struct net_device *dev, struct nlattr *tb[],
489 			   struct nlattr *data[],
490 			   struct netlink_ext_ack *extack)
491 {
492 	struct ip_tunnel *t = netdev_priv(dev);
493 	struct ip_tunnel_encap ipencap;
494 	struct ip_tunnel_parm_kern p;
495 	bool collect_md;
496 	__u32 fwmark = t->fwmark;
497 
498 	if (!rtnl_dev_link_net_capable(dev, t->net))
499 		return -EPERM;
500 
501 	if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
502 		int err = ip_tunnel_encap_setup(t, &ipencap);
503 
504 		if (err < 0)
505 			return err;
506 	}
507 
508 	ipip_netlink_parms(data, &p, &collect_md, &fwmark);
509 	if (collect_md)
510 		return -EINVAL;
511 
512 	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
513 	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
514 		return -EINVAL;
515 
516 	return ip_tunnel_changelink(dev, tb, &p, fwmark);
517 }
518 
519 static size_t ipip_get_size(const struct net_device *dev)
520 {
521 	return
522 		/* IFLA_IPTUN_LINK */
523 		nla_total_size(4) +
524 		/* IFLA_IPTUN_LOCAL */
525 		nla_total_size(4) +
526 		/* IFLA_IPTUN_REMOTE */
527 		nla_total_size(4) +
528 		/* IFLA_IPTUN_TTL */
529 		nla_total_size(1) +
530 		/* IFLA_IPTUN_TOS */
531 		nla_total_size(1) +
532 		/* IFLA_IPTUN_PROTO */
533 		nla_total_size(1) +
534 		/* IFLA_IPTUN_PMTUDISC */
535 		nla_total_size(1) +
536 		/* IFLA_IPTUN_ENCAP_TYPE */
537 		nla_total_size(2) +
538 		/* IFLA_IPTUN_ENCAP_FLAGS */
539 		nla_total_size(2) +
540 		/* IFLA_IPTUN_ENCAP_SPORT */
541 		nla_total_size(2) +
542 		/* IFLA_IPTUN_ENCAP_DPORT */
543 		nla_total_size(2) +
544 		/* IFLA_IPTUN_COLLECT_METADATA */
545 		nla_total_size(0) +
546 		/* IFLA_IPTUN_FWMARK */
547 		nla_total_size(4) +
548 		0;
549 }
550 
551 static int ipip_fill_info(struct sk_buff *skb, const struct net_device *dev)
552 {
553 	struct ip_tunnel *tunnel = netdev_priv(dev);
554 	struct ip_tunnel_parm_kern *parm = &tunnel->parms;
555 
556 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
557 	    nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
558 	    nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
559 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
560 	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
561 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
562 	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
563 		       !!(parm->iph.frag_off & htons(IP_DF))) ||
564 	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
565 		goto nla_put_failure;
566 
567 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
568 			tunnel->encap.type) ||
569 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
570 			 tunnel->encap.sport) ||
571 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
572 			 tunnel->encap.dport) ||
573 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
574 			tunnel->encap.flags))
575 		goto nla_put_failure;
576 
577 	if (tunnel->collect_md)
578 		if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
579 			goto nla_put_failure;
580 	return 0;
581 
582 nla_put_failure:
583 	return -EMSGSIZE;
584 }
585 
586 static const struct nla_policy ipip_policy[IFLA_IPTUN_MAX + 1] = {
587 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
588 	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
589 	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
590 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
591 	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
592 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
593 	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
594 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
595 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
596 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
597 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
598 	[IFLA_IPTUN_COLLECT_METADATA]	= { .type = NLA_FLAG },
599 	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
600 };
601 
602 static struct rtnl_link_ops ipip_link_ops __read_mostly = {
603 	.kind		= "ipip",
604 	.maxtype	= IFLA_IPTUN_MAX,
605 	.policy		= ipip_policy,
606 	.priv_size	= sizeof(struct ip_tunnel),
607 	.setup		= ipip_tunnel_setup,
608 	.validate	= ipip_tunnel_validate,
609 	.newlink	= ipip_newlink,
610 	.changelink	= ipip_changelink,
611 	.dellink	= ip_tunnel_dellink,
612 	.get_size	= ipip_get_size,
613 	.fill_info	= ipip_fill_info,
614 	.get_link_net	= ip_tunnel_get_link_net,
615 };
616 
617 static struct xfrm_tunnel ipip_handler __read_mostly = {
618 	.handler	=	ipip_rcv,
619 	.err_handler	=	ipip_err,
620 	.priority	=	1,
621 };
622 
623 #if IS_ENABLED(CONFIG_MPLS)
624 static struct xfrm_tunnel mplsip_handler __read_mostly = {
625 	.handler	=	mplsip_rcv,
626 	.err_handler	=	ipip_err,
627 	.priority	=	1,
628 };
629 #endif
630 
631 static int __net_init ipip_init_net(struct net *net)
632 {
633 	return ip_tunnel_init_net(net, ipip_net_id, &ipip_link_ops, "tunl0");
634 }
635 
636 static void __net_exit ipip_exit_rtnl(struct net *net,
637 				      struct list_head *dev_to_kill)
638 {
639 	ip_tunnel_delete_net(net, ipip_net_id, &ipip_link_ops, dev_to_kill);
640 }
641 
642 static struct pernet_operations ipip_net_ops = {
643 	.init = ipip_init_net,
644 	.exit_rtnl = ipip_exit_rtnl,
645 	.id   = &ipip_net_id,
646 	.size = sizeof(struct ip_tunnel_net),
647 };
648 
649 static int __init ipip_init(void)
650 {
651 	int err;
652 
653 	pr_info("ipip: IPv4 and MPLS over IPv4 tunneling driver\n");
654 
655 	err = register_pernet_device(&ipip_net_ops);
656 	if (err < 0)
657 		return err;
658 	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
659 	if (err < 0) {
660 		pr_info("%s: can't register tunnel\n", __func__);
661 		goto xfrm_tunnel_ipip_failed;
662 	}
663 #if IS_ENABLED(CONFIG_MPLS)
664 	err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
665 	if (err < 0) {
666 		pr_info("%s: can't register tunnel\n", __func__);
667 		goto xfrm_tunnel_mplsip_failed;
668 	}
669 #endif
670 	err = rtnl_link_register(&ipip_link_ops);
671 	if (err < 0)
672 		goto rtnl_link_failed;
673 
674 out:
675 	return err;
676 
677 rtnl_link_failed:
678 #if IS_ENABLED(CONFIG_MPLS)
679 	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
680 xfrm_tunnel_mplsip_failed:
681 
682 #endif
683 	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
684 xfrm_tunnel_ipip_failed:
685 	unregister_pernet_device(&ipip_net_ops);
686 	goto out;
687 }
688 
689 static void __exit ipip_fini(void)
690 {
691 	rtnl_link_unregister(&ipip_link_ops);
692 	if (xfrm4_tunnel_deregister(&ipip_handler, AF_INET))
693 		pr_info("%s: can't deregister tunnel\n", __func__);
694 #if IS_ENABLED(CONFIG_MPLS)
695 	if (xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS))
696 		pr_info("%s: can't deregister tunnel\n", __func__);
697 #endif
698 	unregister_pernet_device(&ipip_net_ops);
699 }
700 
701 module_init(ipip_init);
702 module_exit(ipip_fini);
703 MODULE_DESCRIPTION("IP/IP protocol decoder library");
704 MODULE_LICENSE("GPL");
705 MODULE_ALIAS_RTNL_LINK("ipip");
706 MODULE_ALIAS_NETDEV("tunl0");
707