xref: /linux/net/ipv4/ipip.c (revision 8211a26324667980a463c069469a818e71207e02)
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, 0, 0, 0,
364 			     RT_SCOPE_UNIVERSE);
365 	if (IS_ERR(rt))
366 		return PTR_ERR(rt);
367 
368 	path->type = DEV_PATH_TUN;
369 	path->tun.src_v4.s_addr = tiph->saddr;
370 	path->tun.dst_v4.s_addr = tiph->daddr;
371 	path->tun.l3_proto = IPPROTO_IPIP;
372 	path->dev = ctx->dev;
373 
374 	ctx->dev = rt->dst.dev;
375 	ip_rt_put(rt);
376 
377 	return 0;
378 }
379 
380 static const struct net_device_ops ipip_netdev_ops = {
381 	.ndo_init       = ipip_tunnel_init,
382 	.ndo_uninit     = ip_tunnel_uninit,
383 	.ndo_start_xmit	= ipip_tunnel_xmit,
384 	.ndo_siocdevprivate = ip_tunnel_siocdevprivate,
385 	.ndo_change_mtu = ip_tunnel_change_mtu,
386 	.ndo_get_stats64 = dev_get_tstats64,
387 	.ndo_get_iflink = ip_tunnel_get_iflink,
388 	.ndo_tunnel_ctl	= ipip_tunnel_ctl,
389 	.ndo_fill_forward_path = ipip_fill_forward_path,
390 };
391 
392 #define IPIP_FEATURES (NETIF_F_SG |		\
393 		       NETIF_F_FRAGLIST |	\
394 		       NETIF_F_HIGHDMA |	\
395 		       NETIF_F_GSO_SOFTWARE |	\
396 		       NETIF_F_HW_CSUM)
397 
398 static void ipip_tunnel_setup(struct net_device *dev)
399 {
400 	dev->netdev_ops		= &ipip_netdev_ops;
401 	dev->header_ops		= &ip_tunnel_header_ops;
402 
403 	dev->type		= ARPHRD_TUNNEL;
404 	dev->flags		= IFF_NOARP;
405 	dev->addr_len		= 4;
406 	dev->lltx		= true;
407 	netif_keep_dst(dev);
408 
409 	dev->features		|= IPIP_FEATURES;
410 	dev->hw_features	|= IPIP_FEATURES;
411 	ip_tunnel_setup(dev, ipip_net_id);
412 }
413 
414 static int ipip_tunnel_init(struct net_device *dev)
415 {
416 	struct ip_tunnel *tunnel = netdev_priv(dev);
417 
418 	__dev_addr_set(dev, &tunnel->parms.iph.saddr, 4);
419 	memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
420 
421 	tunnel->tun_hlen = 0;
422 	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
423 	return ip_tunnel_init(dev);
424 }
425 
426 static int ipip_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
427 				struct netlink_ext_ack *extack)
428 {
429 	u8 proto;
430 
431 	if (!data || !data[IFLA_IPTUN_PROTO])
432 		return 0;
433 
434 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
435 	if (proto != IPPROTO_IPIP && proto != IPPROTO_MPLS && proto != 0)
436 		return -EINVAL;
437 
438 	return 0;
439 }
440 
441 static void ipip_netlink_parms(struct nlattr *data[],
442 			       struct ip_tunnel_parm_kern *parms,
443 			       bool *collect_md, __u32 *fwmark)
444 {
445 	memset(parms, 0, sizeof(*parms));
446 
447 	parms->iph.version = 4;
448 	parms->iph.protocol = IPPROTO_IPIP;
449 	parms->iph.ihl = 5;
450 	*collect_md = false;
451 
452 	if (!data)
453 		return;
454 
455 	ip_tunnel_netlink_parms(data, parms);
456 
457 	if (data[IFLA_IPTUN_COLLECT_METADATA])
458 		*collect_md = true;
459 
460 	if (data[IFLA_IPTUN_FWMARK])
461 		*fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
462 }
463 
464 static int ipip_newlink(struct net_device *dev,
465 			struct rtnl_newlink_params *params,
466 			struct netlink_ext_ack *extack)
467 {
468 	struct ip_tunnel *t = netdev_priv(dev);
469 	struct nlattr **data = params->data;
470 	struct nlattr **tb = params->tb;
471 	struct ip_tunnel_encap ipencap;
472 	struct ip_tunnel_parm_kern p;
473 	__u32 fwmark = 0;
474 
475 	if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
476 		int err = ip_tunnel_encap_setup(t, &ipencap);
477 
478 		if (err < 0)
479 			return err;
480 	}
481 
482 	ipip_netlink_parms(data, &p, &t->collect_md, &fwmark);
483 	return ip_tunnel_newlink(params->link_net ? : dev_net(dev), dev, tb, &p,
484 				 fwmark);
485 }
486 
487 static int ipip_changelink(struct net_device *dev, struct nlattr *tb[],
488 			   struct nlattr *data[],
489 			   struct netlink_ext_ack *extack)
490 {
491 	struct ip_tunnel *t = netdev_priv(dev);
492 	struct ip_tunnel_encap ipencap;
493 	struct ip_tunnel_parm_kern p;
494 	bool collect_md;
495 	__u32 fwmark = t->fwmark;
496 
497 	if (!rtnl_dev_link_net_capable(dev, t->net))
498 		return -EPERM;
499 
500 	if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
501 		int err = ip_tunnel_encap_setup(t, &ipencap);
502 
503 		if (err < 0)
504 			return err;
505 	}
506 
507 	ipip_netlink_parms(data, &p, &collect_md, &fwmark);
508 	if (collect_md)
509 		return -EINVAL;
510 
511 	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
512 	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
513 		return -EINVAL;
514 
515 	return ip_tunnel_changelink(dev, tb, &p, fwmark);
516 }
517 
518 static size_t ipip_get_size(const struct net_device *dev)
519 {
520 	return
521 		/* IFLA_IPTUN_LINK */
522 		nla_total_size(4) +
523 		/* IFLA_IPTUN_LOCAL */
524 		nla_total_size(4) +
525 		/* IFLA_IPTUN_REMOTE */
526 		nla_total_size(4) +
527 		/* IFLA_IPTUN_TTL */
528 		nla_total_size(1) +
529 		/* IFLA_IPTUN_TOS */
530 		nla_total_size(1) +
531 		/* IFLA_IPTUN_PROTO */
532 		nla_total_size(1) +
533 		/* IFLA_IPTUN_PMTUDISC */
534 		nla_total_size(1) +
535 		/* IFLA_IPTUN_ENCAP_TYPE */
536 		nla_total_size(2) +
537 		/* IFLA_IPTUN_ENCAP_FLAGS */
538 		nla_total_size(2) +
539 		/* IFLA_IPTUN_ENCAP_SPORT */
540 		nla_total_size(2) +
541 		/* IFLA_IPTUN_ENCAP_DPORT */
542 		nla_total_size(2) +
543 		/* IFLA_IPTUN_COLLECT_METADATA */
544 		nla_total_size(0) +
545 		/* IFLA_IPTUN_FWMARK */
546 		nla_total_size(4) +
547 		0;
548 }
549 
550 static int ipip_fill_info(struct sk_buff *skb, const struct net_device *dev)
551 {
552 	struct ip_tunnel *tunnel = netdev_priv(dev);
553 	struct ip_tunnel_parm_kern *parm = &tunnel->parms;
554 
555 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
556 	    nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
557 	    nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
558 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
559 	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
560 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
561 	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
562 		       !!(parm->iph.frag_off & htons(IP_DF))) ||
563 	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
564 		goto nla_put_failure;
565 
566 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
567 			tunnel->encap.type) ||
568 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
569 			 tunnel->encap.sport) ||
570 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
571 			 tunnel->encap.dport) ||
572 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
573 			tunnel->encap.flags))
574 		goto nla_put_failure;
575 
576 	if (tunnel->collect_md)
577 		if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
578 			goto nla_put_failure;
579 	return 0;
580 
581 nla_put_failure:
582 	return -EMSGSIZE;
583 }
584 
585 static const struct nla_policy ipip_policy[IFLA_IPTUN_MAX + 1] = {
586 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
587 	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
588 	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
589 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
590 	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
591 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
592 	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
593 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
594 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
595 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
596 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
597 	[IFLA_IPTUN_COLLECT_METADATA]	= { .type = NLA_FLAG },
598 	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
599 };
600 
601 static struct rtnl_link_ops ipip_link_ops __read_mostly = {
602 	.kind		= "ipip",
603 	.maxtype	= IFLA_IPTUN_MAX,
604 	.policy		= ipip_policy,
605 	.priv_size	= sizeof(struct ip_tunnel),
606 	.setup		= ipip_tunnel_setup,
607 	.validate	= ipip_tunnel_validate,
608 	.newlink	= ipip_newlink,
609 	.changelink	= ipip_changelink,
610 	.dellink	= ip_tunnel_dellink,
611 	.get_size	= ipip_get_size,
612 	.fill_info	= ipip_fill_info,
613 	.get_link_net	= ip_tunnel_get_link_net,
614 };
615 
616 static struct xfrm_tunnel ipip_handler __read_mostly = {
617 	.handler	=	ipip_rcv,
618 	.err_handler	=	ipip_err,
619 	.priority	=	1,
620 };
621 
622 #if IS_ENABLED(CONFIG_MPLS)
623 static struct xfrm_tunnel mplsip_handler __read_mostly = {
624 	.handler	=	mplsip_rcv,
625 	.err_handler	=	ipip_err,
626 	.priority	=	1,
627 };
628 #endif
629 
630 static int __net_init ipip_init_net(struct net *net)
631 {
632 	return ip_tunnel_init_net(net, ipip_net_id, &ipip_link_ops, "tunl0");
633 }
634 
635 static void __net_exit ipip_exit_rtnl(struct net *net,
636 				      struct list_head *dev_to_kill)
637 {
638 	ip_tunnel_delete_net(net, ipip_net_id, &ipip_link_ops, dev_to_kill);
639 }
640 
641 static struct pernet_operations ipip_net_ops = {
642 	.init = ipip_init_net,
643 	.exit_rtnl = ipip_exit_rtnl,
644 	.id   = &ipip_net_id,
645 	.size = sizeof(struct ip_tunnel_net),
646 };
647 
648 static int __init ipip_init(void)
649 {
650 	int err;
651 
652 	pr_info("ipip: IPv4 and MPLS over IPv4 tunneling driver\n");
653 
654 	err = register_pernet_device(&ipip_net_ops);
655 	if (err < 0)
656 		return err;
657 	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
658 	if (err < 0) {
659 		pr_info("%s: can't register tunnel\n", __func__);
660 		goto xfrm_tunnel_ipip_failed;
661 	}
662 #if IS_ENABLED(CONFIG_MPLS)
663 	err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
664 	if (err < 0) {
665 		pr_info("%s: can't register tunnel\n", __func__);
666 		goto xfrm_tunnel_mplsip_failed;
667 	}
668 #endif
669 	err = rtnl_link_register(&ipip_link_ops);
670 	if (err < 0)
671 		goto rtnl_link_failed;
672 
673 out:
674 	return err;
675 
676 rtnl_link_failed:
677 #if IS_ENABLED(CONFIG_MPLS)
678 	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
679 xfrm_tunnel_mplsip_failed:
680 
681 #endif
682 	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
683 xfrm_tunnel_ipip_failed:
684 	unregister_pernet_device(&ipip_net_ops);
685 	goto out;
686 }
687 
688 static void __exit ipip_fini(void)
689 {
690 	rtnl_link_unregister(&ipip_link_ops);
691 	if (xfrm4_tunnel_deregister(&ipip_handler, AF_INET))
692 		pr_info("%s: can't deregister tunnel\n", __func__);
693 #if IS_ENABLED(CONFIG_MPLS)
694 	if (xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS))
695 		pr_info("%s: can't deregister tunnel\n", __func__);
696 #endif
697 	unregister_pernet_device(&ipip_net_ops);
698 }
699 
700 module_init(ipip_init);
701 module_exit(ipip_fini);
702 MODULE_DESCRIPTION("IP/IP protocol decoder library");
703 MODULE_LICENSE("GPL");
704 MODULE_ALIAS_RTNL_LINK("ipip");
705 MODULE_ALIAS_NETDEV("tunl0");
706