Linux Audio

Check our new training course

Loading...
v4.17
 
  1/*
  2 *	Linux NET3:	IP/IP protocol decoder.
  3 *
  4 *	Authors:
  5 *		Sam Lantinga (slouken@cs.ucdavis.edu)  02/01/95
  6 *
  7 *	Fixes:
  8 *		Alan Cox	:	Merged and made usable non modular (its so tiny its silly as
  9 *					a module taking up 2 pages).
 10 *		Alan Cox	: 	Fixed bug with 1.3.18 and IPIP not working (now needs to set skb->h.iph)
 11 *					to keep ip_forward happy.
 12 *		Alan Cox	:	More fixes for 1.3.21, and firewall fix. Maybe this will work soon 8).
 13 *		Kai Schulte	:	Fixed #defines for IP_FIREWALL->FIREWALL
 14 *              David Woodhouse :       Perform some basic ICMP handling.
 15 *                                      IPIP Routing without decapsulation.
 16 *              Carlos Picoto   :       GRE over IP support
 17 *		Alexey Kuznetsov:	Reworked. Really, now it is truncated version of ipv4/ip_gre.c.
 18 *					I do not want to merge them together.
 19 *
 20 *	This program is free software; you can redistribute it and/or
 21 *	modify it under the terms of the GNU General Public License
 22 *	as published by the Free Software Foundation; either version
 23 *	2 of the License, or (at your option) any later version.
 24 *
 25 */
 26
 27/* tunnel.c: an IP tunnel driver
 28
 29	The purpose of this driver is to provide an IP tunnel through
 30	which you can tunnel network traffic transparently across subnets.
 31
 32	This was written by looking at Nick Holloway's dummy driver
 33	Thanks for the great code!
 34
 35		-Sam Lantinga	(slouken@cs.ucdavis.edu)  02/01/95
 36
 37	Minor tweaks:
 38		Cleaned up the code a little and added some pre-1.3.0 tweaks.
 39		dev->hard_header/hard_header_len changed to use no headers.
 40		Comments/bracketing tweaked.
 41		Made the tunnels use dev->name not tunnel: when error reporting.
 42		Added tx_dropped stat
 43
 44		-Alan Cox	(alan@lxorguk.ukuu.org.uk) 21 March 95
 45
 46	Reworked:
 47		Changed to tunnel to destination gateway in addition to the
 48			tunnel's pointopoint address
 49		Almost completely rewritten
 50		Note:  There is currently no firewall or ICMP handling done.
 51
 52		-Sam Lantinga	(slouken@cs.ucdavis.edu) 02/13/96
 53
 54*/
 55
 56/* Things I wish I had known when writing the tunnel driver:
 57
 58	When the tunnel_xmit() function is called, the skb contains the
 59	packet to be sent (plus a great deal of extra info), and dev
 60	contains the tunnel device that _we_ are.
 61
 62	When we are passed a packet, we are expected to fill in the
 63	source address with our source IP address.
 64
 65	What is the proper way to allocate, copy and free a buffer?
 66	After you allocate it, it is a "0 length" chunk of memory
 67	starting at zero.  If you want to add headers to the buffer
 68	later, you'll have to call "skb_reserve(skb, amount)" with
 69	the amount of memory you want reserved.  Then, you call
 70	"skb_put(skb, amount)" with the amount of space you want in
 71	the buffer.  skb_put() returns a pointer to the top (#0) of
 72	that buffer.  skb->len is set to the amount of space you have
 73	"allocated" with skb_put().  You can then write up to skb->len
 74	bytes to that buffer.  If you need more, you can call skb_put()
 75	again with the additional amount of space you need.  You can
 76	find out how much more space you can allocate by calling
 77	"skb_tailroom(skb)".
 78	Now, to add header space, call "skb_push(skb, header_len)".
 79	This creates space at the beginning of the buffer and returns
 80	a pointer to this new space.  If later you need to strip a
 81	header from a buffer, call "skb_pull(skb, header_len)".
 82	skb_headroom() will return how much space is left at the top
 83	of the buffer (before the main data).  Remember, this headroom
 84	space must be reserved before the skb_put() function is called.
 85	*/
 86
 87/*
 88   This version of net/ipv4/ipip.c is cloned of net/ipv4/ip_gre.c
 89
 90   For comments look at net/ipv4/ip_gre.c --ANK
 91 */
 92
 93
 94#include <linux/capability.h>
 95#include <linux/module.h>
 96#include <linux/types.h>
 97#include <linux/kernel.h>
 98#include <linux/slab.h>
 99#include <linux/uaccess.h>
100#include <linux/skbuff.h>
101#include <linux/netdevice.h>
102#include <linux/in.h>
103#include <linux/tcp.h>
104#include <linux/udp.h>
105#include <linux/if_arp.h>
106#include <linux/init.h>
107#include <linux/netfilter_ipv4.h>
108#include <linux/if_ether.h>
109
110#include <net/sock.h>
111#include <net/ip.h>
112#include <net/icmp.h>
113#include <net/ip_tunnels.h>
114#include <net/inet_ecn.h>
115#include <net/xfrm.h>
116#include <net/net_namespace.h>
117#include <net/netns/generic.h>
118#include <net/dst_metadata.h>
119
120static bool log_ecn_error = true;
121module_param(log_ecn_error, bool, 0644);
122MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
123
124static unsigned int ipip_net_id __read_mostly;
125
126static int ipip_tunnel_init(struct net_device *dev);
127static struct rtnl_link_ops ipip_link_ops __read_mostly;
128
129static int ipip_err(struct sk_buff *skb, u32 info)
130{
131	/* All the routers (except for Linux) return only
132	 * 8 bytes of packet payload. It means, that precise relaying of
133	 * ICMP in the real Internet is absolutely infeasible.
134	 */
135	struct net *net = dev_net(skb->dev);
136	struct ip_tunnel_net *itn = net_generic(net, ipip_net_id);
137	const struct iphdr *iph = (const struct iphdr *)skb->data;
138	const int type = icmp_hdr(skb)->type;
139	const int code = icmp_hdr(skb)->code;
140	struct ip_tunnel *t;
141	int err = 0;
142
 
 
 
 
 
 
 
143	switch (type) {
144	case ICMP_DEST_UNREACH:
145		switch (code) {
146		case ICMP_SR_FAILED:
147			/* Impossible event. */
148			goto out;
149		default:
150			/* All others are translated to HOST_UNREACH.
151			 * rfc2003 contains "deep thoughts" about NET_UNREACH,
152			 * I believe they are just ether pollution. --ANK
153			 */
154			break;
155		}
156		break;
157
158	case ICMP_TIME_EXCEEDED:
159		if (code != ICMP_EXC_TTL)
160			goto out;
161		break;
162
163	case ICMP_REDIRECT:
164		break;
165
166	default:
167		goto out;
168	}
169
170	t = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
171			     iph->daddr, iph->saddr, 0);
172	if (!t) {
173		err = -ENOENT;
174		goto out;
175	}
176
177	if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
178		ipv4_update_pmtu(skb, net, info, t->parms.link, 0,
179				 iph->protocol, 0);
180		goto out;
181	}
182
183	if (type == ICMP_REDIRECT) {
184		ipv4_redirect(skb, net, t->parms.link, 0, iph->protocol, 0);
185		goto out;
186	}
187
188	if (t->parms.iph.daddr == 0) {
189		err = -ENOENT;
190		goto out;
191	}
192
193	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
194		goto out;
195
196	if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
197		t->err_count++;
198	else
199		t->err_count = 1;
200	t->err_time = jiffies;
201
202out:
203	return err;
204}
205
206static const struct tnl_ptk_info ipip_tpi = {
207	/* no tunnel info required for ipip. */
208	.proto = htons(ETH_P_IP),
209};
210
211#if IS_ENABLED(CONFIG_MPLS)
212static const struct tnl_ptk_info mplsip_tpi = {
213	/* no tunnel info required for mplsip. */
214	.proto = htons(ETH_P_MPLS_UC),
215};
216#endif
217
218static int ipip_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
219{
220	struct net *net = dev_net(skb->dev);
221	struct ip_tunnel_net *itn = net_generic(net, ipip_net_id);
222	struct metadata_dst *tun_dst = NULL;
223	struct ip_tunnel *tunnel;
224	const struct iphdr *iph;
225
226	iph = ip_hdr(skb);
227	tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
228			iph->saddr, 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			tun_dst = ip_tun_rx_dst(skb, 0, 0, 0);
248			if (!tun_dst)
249				return 0;
250		}
 
 
251		return ip_tunnel_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
252	}
253
254	return -1;
255
256drop:
257	kfree_skb(skb);
258	return 0;
259}
260
261static int ipip_rcv(struct sk_buff *skb)
262{
263	return ipip_tunnel_rcv(skb, IPPROTO_IPIP);
264}
265
266#if IS_ENABLED(CONFIG_MPLS)
267static int mplsip_rcv(struct sk_buff *skb)
268{
269	return ipip_tunnel_rcv(skb, IPPROTO_MPLS);
270}
271#endif
272
273/*
274 *	This function assumes it is being called from dev_queue_xmit()
275 *	and that skb is filled properly by that function.
276 */
277static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb,
278				    struct net_device *dev)
279{
280	struct ip_tunnel *tunnel = netdev_priv(dev);
281	const struct iphdr  *tiph = &tunnel->parms.iph;
282	u8 ipproto;
283
 
 
 
284	switch (skb->protocol) {
285	case htons(ETH_P_IP):
286		ipproto = IPPROTO_IPIP;
287		break;
288#if IS_ENABLED(CONFIG_MPLS)
289	case htons(ETH_P_MPLS_UC):
290		ipproto = IPPROTO_MPLS;
291		break;
292#endif
293	default:
294		goto tx_error;
295	}
296
297	if (tiph->protocol != ipproto && tiph->protocol != 0)
298		goto tx_error;
299
300	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
301		goto tx_error;
302
303	skb_set_inner_ipproto(skb, ipproto);
304
305	if (tunnel->collect_md)
306		ip_md_tunnel_xmit(skb, dev, ipproto);
307	else
308		ip_tunnel_xmit(skb, dev, tiph, ipproto);
309	return NETDEV_TX_OK;
310
311tx_error:
312	kfree_skb(skb);
313
314	dev->stats.tx_errors++;
315	return NETDEV_TX_OK;
316}
317
318static bool ipip_tunnel_ioctl_verify_protocol(u8 ipproto)
319{
320	switch (ipproto) {
321	case 0:
322	case IPPROTO_IPIP:
323#if IS_ENABLED(CONFIG_MPLS)
324	case IPPROTO_MPLS:
325#endif
326		return true;
327	}
328
329	return false;
330}
331
332static int
333ipip_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
334{
335	int err = 0;
336	struct ip_tunnel_parm p;
337
338	if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
339		return -EFAULT;
340
341	if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
342		if (p.iph.version != 4 ||
343		    !ipip_tunnel_ioctl_verify_protocol(p.iph.protocol) ||
344		    p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
345			return -EINVAL;
346	}
347
348	p.i_key = p.o_key = 0;
349	p.i_flags = p.o_flags = 0;
350	err = ip_tunnel_ioctl(dev, &p, cmd);
351	if (err)
352		return err;
353
354	if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
355		return -EFAULT;
356
357	return 0;
358}
359
360static const struct net_device_ops ipip_netdev_ops = {
361	.ndo_init       = ipip_tunnel_init,
362	.ndo_uninit     = ip_tunnel_uninit,
363	.ndo_start_xmit	= ipip_tunnel_xmit,
364	.ndo_do_ioctl	= ipip_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,
 
368};
369
370#define IPIP_FEATURES (NETIF_F_SG |		\
371		       NETIF_F_FRAGLIST |	\
372		       NETIF_F_HIGHDMA |	\
373		       NETIF_F_GSO_SOFTWARE |	\
374		       NETIF_F_HW_CSUM)
375
376static void ipip_tunnel_setup(struct net_device *dev)
377{
378	dev->netdev_ops		= &ipip_netdev_ops;
 
379
380	dev->type		= ARPHRD_TUNNEL;
381	dev->flags		= IFF_NOARP;
382	dev->addr_len		= 4;
383	dev->features		|= NETIF_F_LLTX;
384	netif_keep_dst(dev);
385
386	dev->features		|= IPIP_FEATURES;
387	dev->hw_features	|= IPIP_FEATURES;
388	ip_tunnel_setup(dev, ipip_net_id);
389}
390
391static int ipip_tunnel_init(struct net_device *dev)
392{
393	struct ip_tunnel *tunnel = netdev_priv(dev);
394
395	memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
396	memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
397
398	tunnel->tun_hlen = 0;
399	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
400	return ip_tunnel_init(dev);
401}
402
403static int ipip_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
404				struct netlink_ext_ack *extack)
405{
406	u8 proto;
407
408	if (!data || !data[IFLA_IPTUN_PROTO])
409		return 0;
410
411	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
412	if (proto != IPPROTO_IPIP && proto != IPPROTO_MPLS && proto != 0)
413		return -EINVAL;
414
415	return 0;
416}
417
418static void ipip_netlink_parms(struct nlattr *data[],
419			       struct ip_tunnel_parm *parms, bool *collect_md,
420			       __u32 *fwmark)
421{
422	memset(parms, 0, sizeof(*parms));
423
424	parms->iph.version = 4;
425	parms->iph.protocol = IPPROTO_IPIP;
426	parms->iph.ihl = 5;
427	*collect_md = false;
428
429	if (!data)
430		return;
431
432	if (data[IFLA_IPTUN_LINK])
433		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
434
435	if (data[IFLA_IPTUN_LOCAL])
436		parms->iph.saddr = nla_get_in_addr(data[IFLA_IPTUN_LOCAL]);
437
438	if (data[IFLA_IPTUN_REMOTE])
439		parms->iph.daddr = nla_get_in_addr(data[IFLA_IPTUN_REMOTE]);
440
441	if (data[IFLA_IPTUN_TTL]) {
442		parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
443		if (parms->iph.ttl)
444			parms->iph.frag_off = htons(IP_DF);
445	}
446
447	if (data[IFLA_IPTUN_TOS])
448		parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
449
450	if (data[IFLA_IPTUN_PROTO])
451		parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
452
453	if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
454		parms->iph.frag_off = htons(IP_DF);
455
456	if (data[IFLA_IPTUN_COLLECT_METADATA])
457		*collect_md = true;
458
459	if (data[IFLA_IPTUN_FWMARK])
460		*fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
461}
462
463/* This function returns true when ENCAP attributes are present in the nl msg */
464static bool ipip_netlink_encap_parms(struct nlattr *data[],
465				     struct ip_tunnel_encap *ipencap)
466{
467	bool ret = false;
468
469	memset(ipencap, 0, sizeof(*ipencap));
470
471	if (!data)
472		return ret;
473
474	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
475		ret = true;
476		ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
477	}
478
479	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
480		ret = true;
481		ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
482	}
483
484	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
485		ret = true;
486		ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
487	}
488
489	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
490		ret = true;
491		ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
492	}
493
494	return ret;
495}
496
497static int ipip_newlink(struct net *src_net, struct net_device *dev,
498			struct nlattr *tb[], struct nlattr *data[],
499			struct netlink_ext_ack *extack)
500{
501	struct ip_tunnel *t = netdev_priv(dev);
502	struct ip_tunnel_parm p;
503	struct ip_tunnel_encap ipencap;
504	__u32 fwmark = 0;
505
506	if (ipip_netlink_encap_parms(data, &ipencap)) {
507		int err = ip_tunnel_encap_setup(t, &ipencap);
508
509		if (err < 0)
510			return err;
511	}
512
513	ipip_netlink_parms(data, &p, &t->collect_md, &fwmark);
514	return ip_tunnel_newlink(dev, tb, &p, fwmark);
515}
516
517static int ipip_changelink(struct net_device *dev, struct nlattr *tb[],
518			   struct nlattr *data[],
519			   struct netlink_ext_ack *extack)
520{
521	struct ip_tunnel *t = netdev_priv(dev);
522	struct ip_tunnel_parm p;
523	struct ip_tunnel_encap ipencap;
524	bool collect_md;
525	__u32 fwmark = t->fwmark;
526
527	if (ipip_netlink_encap_parms(data, &ipencap)) {
528		int err = ip_tunnel_encap_setup(t, &ipencap);
529
530		if (err < 0)
531			return err;
532	}
533
534	ipip_netlink_parms(data, &p, &collect_md, &fwmark);
535	if (collect_md)
536		return -EINVAL;
537
538	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
539	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
540		return -EINVAL;
541
542	return ip_tunnel_changelink(dev, tb, &p, fwmark);
543}
544
545static size_t ipip_get_size(const struct net_device *dev)
546{
547	return
548		/* IFLA_IPTUN_LINK */
549		nla_total_size(4) +
550		/* IFLA_IPTUN_LOCAL */
551		nla_total_size(4) +
552		/* IFLA_IPTUN_REMOTE */
553		nla_total_size(4) +
554		/* IFLA_IPTUN_TTL */
555		nla_total_size(1) +
556		/* IFLA_IPTUN_TOS */
557		nla_total_size(1) +
558		/* IFLA_IPTUN_PROTO */
559		nla_total_size(1) +
560		/* IFLA_IPTUN_PMTUDISC */
561		nla_total_size(1) +
562		/* IFLA_IPTUN_ENCAP_TYPE */
563		nla_total_size(2) +
564		/* IFLA_IPTUN_ENCAP_FLAGS */
565		nla_total_size(2) +
566		/* IFLA_IPTUN_ENCAP_SPORT */
567		nla_total_size(2) +
568		/* IFLA_IPTUN_ENCAP_DPORT */
569		nla_total_size(2) +
570		/* IFLA_IPTUN_COLLECT_METADATA */
571		nla_total_size(0) +
572		/* IFLA_IPTUN_FWMARK */
573		nla_total_size(4) +
574		0;
575}
576
577static int ipip_fill_info(struct sk_buff *skb, const struct net_device *dev)
578{
579	struct ip_tunnel *tunnel = netdev_priv(dev);
580	struct ip_tunnel_parm *parm = &tunnel->parms;
581
582	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
583	    nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
584	    nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
585	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
586	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
587	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
588	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
589		       !!(parm->iph.frag_off & htons(IP_DF))) ||
590	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
591		goto nla_put_failure;
592
593	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
594			tunnel->encap.type) ||
595	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
596			 tunnel->encap.sport) ||
597	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
598			 tunnel->encap.dport) ||
599	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
600			tunnel->encap.flags))
601		goto nla_put_failure;
602
603	if (tunnel->collect_md)
604		if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
605			goto nla_put_failure;
606	return 0;
607
608nla_put_failure:
609	return -EMSGSIZE;
610}
611
612static const struct nla_policy ipip_policy[IFLA_IPTUN_MAX + 1] = {
613	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
614	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
615	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
616	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
617	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
618	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
619	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
620	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
621	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
622	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
623	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
624	[IFLA_IPTUN_COLLECT_METADATA]	= { .type = NLA_FLAG },
625	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
626};
627
628static struct rtnl_link_ops ipip_link_ops __read_mostly = {
629	.kind		= "ipip",
630	.maxtype	= IFLA_IPTUN_MAX,
631	.policy		= ipip_policy,
632	.priv_size	= sizeof(struct ip_tunnel),
633	.setup		= ipip_tunnel_setup,
634	.validate	= ipip_tunnel_validate,
635	.newlink	= ipip_newlink,
636	.changelink	= ipip_changelink,
637	.dellink	= ip_tunnel_dellink,
638	.get_size	= ipip_get_size,
639	.fill_info	= ipip_fill_info,
640	.get_link_net	= ip_tunnel_get_link_net,
641};
642
643static struct xfrm_tunnel ipip_handler __read_mostly = {
644	.handler	=	ipip_rcv,
645	.err_handler	=	ipip_err,
646	.priority	=	1,
647};
648
649#if IS_ENABLED(CONFIG_MPLS)
650static struct xfrm_tunnel mplsip_handler __read_mostly = {
651	.handler	=	mplsip_rcv,
652	.err_handler	=	ipip_err,
653	.priority	=	1,
654};
655#endif
656
657static int __net_init ipip_init_net(struct net *net)
658{
659	return ip_tunnel_init_net(net, ipip_net_id, &ipip_link_ops, "tunl0");
660}
661
662static void __net_exit ipip_exit_batch_net(struct list_head *list_net)
663{
664	ip_tunnel_delete_nets(list_net, ipip_net_id, &ipip_link_ops);
665}
666
667static struct pernet_operations ipip_net_ops = {
668	.init = ipip_init_net,
669	.exit_batch = ipip_exit_batch_net,
670	.id   = &ipip_net_id,
671	.size = sizeof(struct ip_tunnel_net),
672};
673
674static int __init ipip_init(void)
675{
676	int err;
677
678	pr_info("ipip: IPv4 and MPLS over IPv4 tunneling driver\n");
679
680	err = register_pernet_device(&ipip_net_ops);
681	if (err < 0)
682		return err;
683	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
684	if (err < 0) {
685		pr_info("%s: can't register tunnel\n", __func__);
686		goto xfrm_tunnel_ipip_failed;
687	}
688#if IS_ENABLED(CONFIG_MPLS)
689	err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
690	if (err < 0) {
691		pr_info("%s: can't register tunnel\n", __func__);
692		goto xfrm_tunnel_mplsip_failed;
693	}
694#endif
695	err = rtnl_link_register(&ipip_link_ops);
696	if (err < 0)
697		goto rtnl_link_failed;
698
699out:
700	return err;
701
702rtnl_link_failed:
703#if IS_ENABLED(CONFIG_MPLS)
704	xfrm4_tunnel_deregister(&mplsip_handler, AF_INET);
705xfrm_tunnel_mplsip_failed:
706
707#endif
708	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
709xfrm_tunnel_ipip_failed:
710	unregister_pernet_device(&ipip_net_ops);
711	goto out;
712}
713
714static void __exit ipip_fini(void)
715{
716	rtnl_link_unregister(&ipip_link_ops);
717	if (xfrm4_tunnel_deregister(&ipip_handler, AF_INET))
718		pr_info("%s: can't deregister tunnel\n", __func__);
719#if IS_ENABLED(CONFIG_MPLS)
720	if (xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS))
721		pr_info("%s: can't deregister tunnel\n", __func__);
722#endif
723	unregister_pernet_device(&ipip_net_ops);
724}
725
726module_init(ipip_init);
727module_exit(ipip_fini);
728MODULE_LICENSE("GPL");
729MODULE_ALIAS_RTNL_LINK("ipip");
730MODULE_ALIAS_NETDEV("tunl0");
v5.14.15
  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
115static bool log_ecn_error = true;
116module_param(log_ecn_error, bool, 0644);
117MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
118
119static unsigned int ipip_net_id __read_mostly;
120
121static int ipip_tunnel_init(struct net_device *dev);
122static struct rtnl_link_ops ipip_link_ops __read_mostly;
123
124static 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	const int type = icmp_hdr(skb)->type;
134	const int code = icmp_hdr(skb)->code;
135	struct ip_tunnel *t;
136	int err = 0;
137
138	t = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
139			     iph->daddr, iph->saddr, 0);
140	if (!t) {
141		err = -ENOENT;
142		goto out;
143	}
144
145	switch (type) {
146	case ICMP_DEST_UNREACH:
147		switch (code) {
148		case ICMP_SR_FAILED:
149			/* Impossible event. */
150			goto out;
151		default:
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
155			 */
156			break;
157		}
158		break;
159
160	case ICMP_TIME_EXCEEDED:
161		if (code != ICMP_EXC_TTL)
162			goto out;
163		break;
164
165	case ICMP_REDIRECT:
166		break;
167
168	default:
169		goto out;
170	}
171
 
 
 
 
 
 
 
172	if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
173		ipv4_update_pmtu(skb, net, info, t->parms.link, iph->protocol);
 
174		goto out;
175	}
176
177	if (type == ICMP_REDIRECT) {
178		ipv4_redirect(skb, net, t->parms.link, iph->protocol);
179		goto out;
180	}
181
182	if (t->parms.iph.daddr == 0) {
183		err = -ENOENT;
184		goto out;
185	}
186
187	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
188		goto out;
189
190	if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
191		t->err_count++;
192	else
193		t->err_count = 1;
194	t->err_time = jiffies;
195
196out:
197	return err;
198}
199
200static const struct tnl_ptk_info ipip_tpi = {
201	/* no tunnel info required for ipip. */
202	.proto = htons(ETH_P_IP),
203};
204
205#if IS_ENABLED(CONFIG_MPLS)
206static const struct tnl_ptk_info mplsip_tpi = {
207	/* no tunnel info required for mplsip. */
208	.proto = htons(ETH_P_MPLS_UC),
209};
210#endif
211
212static int ipip_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
213{
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;
219
220	iph = ip_hdr(skb);
221	tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
222			iph->saddr, iph->daddr, 0);
223	if (tunnel) {
224		const struct tnl_ptk_info *tpi;
225
226		if (tunnel->parms.iph.protocol != ipproto &&
227		    tunnel->parms.iph.protocol != 0)
228			goto drop;
229
230		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
231			goto drop;
232#if IS_ENABLED(CONFIG_MPLS)
233		if (ipproto == IPPROTO_MPLS)
234			tpi = &mplsip_tpi;
235		else
236#endif
237			tpi = &ipip_tpi;
238		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
239			goto drop;
240		if (tunnel->collect_md) {
241			tun_dst = ip_tun_rx_dst(skb, 0, 0, 0);
242			if (!tun_dst)
243				return 0;
244		}
245		skb_reset_mac_header(skb);
246
247		return ip_tunnel_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
248	}
249
250	return -1;
251
252drop:
253	kfree_skb(skb);
254	return 0;
255}
256
257static int ipip_rcv(struct sk_buff *skb)
258{
259	return ipip_tunnel_rcv(skb, IPPROTO_IPIP);
260}
261
262#if IS_ENABLED(CONFIG_MPLS)
263static int mplsip_rcv(struct sk_buff *skb)
264{
265	return ipip_tunnel_rcv(skb, IPPROTO_MPLS);
266}
267#endif
268
269/*
270 *	This function assumes it is being called from dev_queue_xmit()
271 *	and that skb is filled properly by that function.
272 */
273static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb,
274				    struct net_device *dev)
275{
276	struct ip_tunnel *tunnel = netdev_priv(dev);
277	const struct iphdr  *tiph = &tunnel->parms.iph;
278	u8 ipproto;
279
280	if (!pskb_inet_may_pull(skb))
281		goto tx_error;
282
283	switch (skb->protocol) {
284	case htons(ETH_P_IP):
285		ipproto = IPPROTO_IPIP;
286		break;
287#if IS_ENABLED(CONFIG_MPLS)
288	case htons(ETH_P_MPLS_UC):
289		ipproto = IPPROTO_MPLS;
290		break;
291#endif
292	default:
293		goto tx_error;
294	}
295
296	if (tiph->protocol != ipproto && tiph->protocol != 0)
297		goto tx_error;
298
299	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
300		goto tx_error;
301
302	skb_set_inner_ipproto(skb, ipproto);
303
304	if (tunnel->collect_md)
305		ip_md_tunnel_xmit(skb, dev, ipproto, 0);
306	else
307		ip_tunnel_xmit(skb, dev, tiph, ipproto);
308	return NETDEV_TX_OK;
309
310tx_error:
311	kfree_skb(skb);
312
313	dev->stats.tx_errors++;
314	return NETDEV_TX_OK;
315}
316
317static bool ipip_tunnel_ioctl_verify_protocol(u8 ipproto)
318{
319	switch (ipproto) {
320	case 0:
321	case IPPROTO_IPIP:
322#if IS_ENABLED(CONFIG_MPLS)
323	case IPPROTO_MPLS:
324#endif
325		return true;
326	}
327
328	return false;
329}
330
331static int
332ipip_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd)
333{
 
 
 
 
 
 
334	if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
335		if (p->iph.version != 4 ||
336		    !ipip_tunnel_ioctl_verify_protocol(p->iph.protocol) ||
337		    p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)))
338			return -EINVAL;
339	}
340
341	p->i_key = p->o_key = 0;
342	p->i_flags = p->o_flags = 0;
343	return ip_tunnel_ctl(dev, p, cmd);
 
 
 
 
 
 
 
344}
345
346static const struct net_device_ops ipip_netdev_ops = {
347	.ndo_init       = ipip_tunnel_init,
348	.ndo_uninit     = ip_tunnel_uninit,
349	.ndo_start_xmit	= ipip_tunnel_xmit,
350	.ndo_do_ioctl	= ip_tunnel_ioctl,
351	.ndo_change_mtu = ip_tunnel_change_mtu,
352	.ndo_get_stats64 = dev_get_tstats64,
353	.ndo_get_iflink = ip_tunnel_get_iflink,
354	.ndo_tunnel_ctl	= ipip_tunnel_ctl,
355};
356
357#define IPIP_FEATURES (NETIF_F_SG |		\
358		       NETIF_F_FRAGLIST |	\
359		       NETIF_F_HIGHDMA |	\
360		       NETIF_F_GSO_SOFTWARE |	\
361		       NETIF_F_HW_CSUM)
362
363static void ipip_tunnel_setup(struct net_device *dev)
364{
365	dev->netdev_ops		= &ipip_netdev_ops;
366	dev->header_ops		= &ip_tunnel_header_ops;
367
368	dev->type		= ARPHRD_TUNNEL;
369	dev->flags		= IFF_NOARP;
370	dev->addr_len		= 4;
371	dev->features		|= NETIF_F_LLTX;
372	netif_keep_dst(dev);
373
374	dev->features		|= IPIP_FEATURES;
375	dev->hw_features	|= IPIP_FEATURES;
376	ip_tunnel_setup(dev, ipip_net_id);
377}
378
379static int ipip_tunnel_init(struct net_device *dev)
380{
381	struct ip_tunnel *tunnel = netdev_priv(dev);
382
383	memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
384	memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
385
386	tunnel->tun_hlen = 0;
387	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
388	return ip_tunnel_init(dev);
389}
390
391static int ipip_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
392				struct netlink_ext_ack *extack)
393{
394	u8 proto;
395
396	if (!data || !data[IFLA_IPTUN_PROTO])
397		return 0;
398
399	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
400	if (proto != IPPROTO_IPIP && proto != IPPROTO_MPLS && proto != 0)
401		return -EINVAL;
402
403	return 0;
404}
405
406static void ipip_netlink_parms(struct nlattr *data[],
407			       struct ip_tunnel_parm *parms, bool *collect_md,
408			       __u32 *fwmark)
409{
410	memset(parms, 0, sizeof(*parms));
411
412	parms->iph.version = 4;
413	parms->iph.protocol = IPPROTO_IPIP;
414	parms->iph.ihl = 5;
415	*collect_md = false;
416
417	if (!data)
418		return;
419
420	if (data[IFLA_IPTUN_LINK])
421		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
422
423	if (data[IFLA_IPTUN_LOCAL])
424		parms->iph.saddr = nla_get_in_addr(data[IFLA_IPTUN_LOCAL]);
425
426	if (data[IFLA_IPTUN_REMOTE])
427		parms->iph.daddr = nla_get_in_addr(data[IFLA_IPTUN_REMOTE]);
428
429	if (data[IFLA_IPTUN_TTL]) {
430		parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
431		if (parms->iph.ttl)
432			parms->iph.frag_off = htons(IP_DF);
433	}
434
435	if (data[IFLA_IPTUN_TOS])
436		parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
437
438	if (data[IFLA_IPTUN_PROTO])
439		parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
440
441	if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
442		parms->iph.frag_off = htons(IP_DF);
443
444	if (data[IFLA_IPTUN_COLLECT_METADATA])
445		*collect_md = true;
446
447	if (data[IFLA_IPTUN_FWMARK])
448		*fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
449}
450
451/* This function returns true when ENCAP attributes are present in the nl msg */
452static bool ipip_netlink_encap_parms(struct nlattr *data[],
453				     struct ip_tunnel_encap *ipencap)
454{
455	bool ret = false;
456
457	memset(ipencap, 0, sizeof(*ipencap));
458
459	if (!data)
460		return ret;
461
462	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
463		ret = true;
464		ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
465	}
466
467	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
468		ret = true;
469		ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
470	}
471
472	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
473		ret = true;
474		ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
475	}
476
477	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
478		ret = true;
479		ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
480	}
481
482	return ret;
483}
484
485static int ipip_newlink(struct net *src_net, struct net_device *dev,
486			struct nlattr *tb[], struct nlattr *data[],
487			struct netlink_ext_ack *extack)
488{
489	struct ip_tunnel *t = netdev_priv(dev);
490	struct ip_tunnel_parm p;
491	struct ip_tunnel_encap ipencap;
492	__u32 fwmark = 0;
493
494	if (ipip_netlink_encap_parms(data, &ipencap)) {
495		int err = ip_tunnel_encap_setup(t, &ipencap);
496
497		if (err < 0)
498			return err;
499	}
500
501	ipip_netlink_parms(data, &p, &t->collect_md, &fwmark);
502	return ip_tunnel_newlink(dev, tb, &p, fwmark);
503}
504
505static int ipip_changelink(struct net_device *dev, struct nlattr *tb[],
506			   struct nlattr *data[],
507			   struct netlink_ext_ack *extack)
508{
509	struct ip_tunnel *t = netdev_priv(dev);
510	struct ip_tunnel_parm p;
511	struct ip_tunnel_encap ipencap;
512	bool collect_md;
513	__u32 fwmark = t->fwmark;
514
515	if (ipip_netlink_encap_parms(data, &ipencap)) {
516		int err = ip_tunnel_encap_setup(t, &ipencap);
517
518		if (err < 0)
519			return err;
520	}
521
522	ipip_netlink_parms(data, &p, &collect_md, &fwmark);
523	if (collect_md)
524		return -EINVAL;
525
526	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
527	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
528		return -EINVAL;
529
530	return ip_tunnel_changelink(dev, tb, &p, fwmark);
531}
532
533static size_t ipip_get_size(const struct net_device *dev)
534{
535	return
536		/* IFLA_IPTUN_LINK */
537		nla_total_size(4) +
538		/* IFLA_IPTUN_LOCAL */
539		nla_total_size(4) +
540		/* IFLA_IPTUN_REMOTE */
541		nla_total_size(4) +
542		/* IFLA_IPTUN_TTL */
543		nla_total_size(1) +
544		/* IFLA_IPTUN_TOS */
545		nla_total_size(1) +
546		/* IFLA_IPTUN_PROTO */
547		nla_total_size(1) +
548		/* IFLA_IPTUN_PMTUDISC */
549		nla_total_size(1) +
550		/* IFLA_IPTUN_ENCAP_TYPE */
551		nla_total_size(2) +
552		/* IFLA_IPTUN_ENCAP_FLAGS */
553		nla_total_size(2) +
554		/* IFLA_IPTUN_ENCAP_SPORT */
555		nla_total_size(2) +
556		/* IFLA_IPTUN_ENCAP_DPORT */
557		nla_total_size(2) +
558		/* IFLA_IPTUN_COLLECT_METADATA */
559		nla_total_size(0) +
560		/* IFLA_IPTUN_FWMARK */
561		nla_total_size(4) +
562		0;
563}
564
565static int ipip_fill_info(struct sk_buff *skb, const struct net_device *dev)
566{
567	struct ip_tunnel *tunnel = netdev_priv(dev);
568	struct ip_tunnel_parm *parm = &tunnel->parms;
569
570	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
571	    nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
572	    nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
573	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
574	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
575	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
576	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
577		       !!(parm->iph.frag_off & htons(IP_DF))) ||
578	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
579		goto nla_put_failure;
580
581	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
582			tunnel->encap.type) ||
583	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
584			 tunnel->encap.sport) ||
585	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
586			 tunnel->encap.dport) ||
587	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
588			tunnel->encap.flags))
589		goto nla_put_failure;
590
591	if (tunnel->collect_md)
592		if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
593			goto nla_put_failure;
594	return 0;
595
596nla_put_failure:
597	return -EMSGSIZE;
598}
599
600static const struct nla_policy ipip_policy[IFLA_IPTUN_MAX + 1] = {
601	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
602	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
603	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
604	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
605	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
606	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
607	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
608	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
609	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
610	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
611	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
612	[IFLA_IPTUN_COLLECT_METADATA]	= { .type = NLA_FLAG },
613	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
614};
615
616static struct rtnl_link_ops ipip_link_ops __read_mostly = {
617	.kind		= "ipip",
618	.maxtype	= IFLA_IPTUN_MAX,
619	.policy		= ipip_policy,
620	.priv_size	= sizeof(struct ip_tunnel),
621	.setup		= ipip_tunnel_setup,
622	.validate	= ipip_tunnel_validate,
623	.newlink	= ipip_newlink,
624	.changelink	= ipip_changelink,
625	.dellink	= ip_tunnel_dellink,
626	.get_size	= ipip_get_size,
627	.fill_info	= ipip_fill_info,
628	.get_link_net	= ip_tunnel_get_link_net,
629};
630
631static struct xfrm_tunnel ipip_handler __read_mostly = {
632	.handler	=	ipip_rcv,
633	.err_handler	=	ipip_err,
634	.priority	=	1,
635};
636
637#if IS_ENABLED(CONFIG_MPLS)
638static struct xfrm_tunnel mplsip_handler __read_mostly = {
639	.handler	=	mplsip_rcv,
640	.err_handler	=	ipip_err,
641	.priority	=	1,
642};
643#endif
644
645static int __net_init ipip_init_net(struct net *net)
646{
647	return ip_tunnel_init_net(net, ipip_net_id, &ipip_link_ops, "tunl0");
648}
649
650static void __net_exit ipip_exit_batch_net(struct list_head *list_net)
651{
652	ip_tunnel_delete_nets(list_net, ipip_net_id, &ipip_link_ops);
653}
654
655static struct pernet_operations ipip_net_ops = {
656	.init = ipip_init_net,
657	.exit_batch = ipip_exit_batch_net,
658	.id   = &ipip_net_id,
659	.size = sizeof(struct ip_tunnel_net),
660};
661
662static int __init ipip_init(void)
663{
664	int err;
665
666	pr_info("ipip: IPv4 and MPLS over IPv4 tunneling driver\n");
667
668	err = register_pernet_device(&ipip_net_ops);
669	if (err < 0)
670		return err;
671	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
672	if (err < 0) {
673		pr_info("%s: can't register tunnel\n", __func__);
674		goto xfrm_tunnel_ipip_failed;
675	}
676#if IS_ENABLED(CONFIG_MPLS)
677	err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
678	if (err < 0) {
679		pr_info("%s: can't register tunnel\n", __func__);
680		goto xfrm_tunnel_mplsip_failed;
681	}
682#endif
683	err = rtnl_link_register(&ipip_link_ops);
684	if (err < 0)
685		goto rtnl_link_failed;
686
687out:
688	return err;
689
690rtnl_link_failed:
691#if IS_ENABLED(CONFIG_MPLS)
692	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
693xfrm_tunnel_mplsip_failed:
694
695#endif
696	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
697xfrm_tunnel_ipip_failed:
698	unregister_pernet_device(&ipip_net_ops);
699	goto out;
700}
701
702static void __exit ipip_fini(void)
703{
704	rtnl_link_unregister(&ipip_link_ops);
705	if (xfrm4_tunnel_deregister(&ipip_handler, AF_INET))
706		pr_info("%s: can't deregister tunnel\n", __func__);
707#if IS_ENABLED(CONFIG_MPLS)
708	if (xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS))
709		pr_info("%s: can't deregister tunnel\n", __func__);
710#endif
711	unregister_pernet_device(&ipip_net_ops);
712}
713
714module_init(ipip_init);
715module_exit(ipip_fini);
716MODULE_LICENSE("GPL");
717MODULE_ALIAS_RTNL_LINK("ipip");
718MODULE_ALIAS_NETDEV("tunl0");