Linux Audio

Check our new training course

Buildroot integration, development and maintenance

Need a Buildroot system for your embedded project?
Loading...
v6.8
  1// SPDX-License-Identifier: GPL-2.0-or-later
  2/*
  3 *	Linux NET3: IP/IP protocol decoder modified to support
  4 *		    virtual tunnel interface
  5 *
  6 *	Authors:
  7 *		Saurabh Mohan (saurabh.mohan@vyatta.com) 05/07/2012
  8 */
  9
 10/*
 11   This version of net/ipv4/ip_vti.c is cloned of net/ipv4/ipip.c
 12
 13   For comments look at net/ipv4/ip_gre.c --ANK
 14 */
 15
 16
 17#include <linux/capability.h>
 18#include <linux/module.h>
 19#include <linux/types.h>
 20#include <linux/kernel.h>
 21#include <linux/uaccess.h>
 22#include <linux/skbuff.h>
 23#include <linux/netdevice.h>
 24#include <linux/in.h>
 25#include <linux/tcp.h>
 26#include <linux/udp.h>
 27#include <linux/if_arp.h>
 28#include <linux/init.h>
 29#include <linux/netfilter_ipv4.h>
 30#include <linux/if_ether.h>
 31#include <linux/icmpv6.h>
 32
 33#include <net/sock.h>
 34#include <net/ip.h>
 35#include <net/icmp.h>
 36#include <net/ip_tunnels.h>
 37#include <net/inet_ecn.h>
 38#include <net/xfrm.h>
 39#include <net/net_namespace.h>
 40#include <net/netns/generic.h>
 41
 42static struct rtnl_link_ops vti_link_ops __read_mostly;
 43
 44static unsigned int vti_net_id __read_mostly;
 45static int vti_tunnel_init(struct net_device *dev);
 46
 47static int vti_input(struct sk_buff *skb, int nexthdr, __be32 spi,
 48		     int encap_type, bool update_skb_dev)
 49{
 50	struct ip_tunnel *tunnel;
 51	const struct iphdr *iph = ip_hdr(skb);
 52	struct net *net = dev_net(skb->dev);
 53	struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
 
 54
 55	tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
 
 
 56				  iph->saddr, iph->daddr, 0);
 57	if (tunnel) {
 58		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 59			goto drop;
 60
 61		XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4 = tunnel;
 62
 63		if (update_skb_dev)
 64			skb->dev = tunnel->dev;
 65
 66		return xfrm_input(skb, nexthdr, spi, encap_type);
 67	}
 68
 69	return -EINVAL;
 70drop:
 71	kfree_skb(skb);
 72	return 0;
 73}
 74
 75static int vti_input_proto(struct sk_buff *skb, int nexthdr, __be32 spi,
 76			   int encap_type)
 77{
 78	return vti_input(skb, nexthdr, spi, encap_type, false);
 79}
 80
 81static int vti_rcv(struct sk_buff *skb, __be32 spi, bool update_skb_dev)
 82{
 83	XFRM_SPI_SKB_CB(skb)->family = AF_INET;
 84	XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct iphdr, daddr);
 85
 86	return vti_input(skb, ip_hdr(skb)->protocol, spi, 0, update_skb_dev);
 87}
 88
 89static int vti_rcv_proto(struct sk_buff *skb)
 90{
 91	return vti_rcv(skb, 0, false);
 92}
 93
 94static int vti_rcv_cb(struct sk_buff *skb, int err)
 95{
 96	unsigned short family;
 97	struct net_device *dev;
 98	struct xfrm_state *x;
 99	const struct xfrm_mode *inner_mode;
100	struct ip_tunnel *tunnel = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4;
101	u32 orig_mark = skb->mark;
102	int ret;
103
104	if (!tunnel)
105		return 1;
106
107	dev = tunnel->dev;
108
109	if (err) {
110		DEV_STATS_INC(dev, rx_errors);
111		DEV_STATS_INC(dev, rx_dropped);
112
113		return 0;
114	}
115
116	x = xfrm_input_state(skb);
117
118	inner_mode = &x->inner_mode;
119
120	if (x->sel.family == AF_UNSPEC) {
121		inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
122		if (inner_mode == NULL) {
123			XFRM_INC_STATS(dev_net(skb->dev),
124				       LINUX_MIB_XFRMINSTATEMODEERROR);
125			return -EINVAL;
126		}
127	}
128
129	family = inner_mode->family;
130
131	skb->mark = be32_to_cpu(tunnel->parms.i_key);
132	ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
133	skb->mark = orig_mark;
134
135	if (!ret)
136		return -EPERM;
137
138	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(skb->dev)));
139	skb->dev = dev;
140	dev_sw_netstats_rx_add(dev, skb->len);
141
142	return 0;
143}
144
145static bool vti_state_check(const struct xfrm_state *x, __be32 dst, __be32 src)
146{
147	xfrm_address_t *daddr = (xfrm_address_t *)&dst;
148	xfrm_address_t *saddr = (xfrm_address_t *)&src;
149
150	/* if there is no transform then this tunnel is not functional.
151	 * Or if the xfrm is not mode tunnel.
152	 */
153	if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
154	    x->props.family != AF_INET)
155		return false;
156
157	if (!dst)
158		return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET);
159
160	if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET))
161		return false;
162
163	return true;
164}
165
166static netdev_tx_t vti_xmit(struct sk_buff *skb, struct net_device *dev,
167			    struct flowi *fl)
168{
169	struct ip_tunnel *tunnel = netdev_priv(dev);
170	struct ip_tunnel_parm *parms = &tunnel->parms;
171	struct dst_entry *dst = skb_dst(skb);
172	struct net_device *tdev;	/* Device to other host */
173	int pkt_len = skb->len;
174	int err;
175	int mtu;
176
177	if (!dst) {
178		switch (skb->protocol) {
179		case htons(ETH_P_IP): {
180			struct rtable *rt;
181
182			fl->u.ip4.flowi4_oif = dev->ifindex;
183			fl->u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC;
184			rt = __ip_route_output_key(dev_net(dev), &fl->u.ip4);
185			if (IS_ERR(rt)) {
186				DEV_STATS_INC(dev, tx_carrier_errors);
187				goto tx_error_icmp;
188			}
189			dst = &rt->dst;
190			skb_dst_set(skb, dst);
191			break;
192		}
193#if IS_ENABLED(CONFIG_IPV6)
194		case htons(ETH_P_IPV6):
195			fl->u.ip6.flowi6_oif = dev->ifindex;
196			fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
197			dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6);
198			if (dst->error) {
199				dst_release(dst);
200				dst = NULL;
201				DEV_STATS_INC(dev, tx_carrier_errors);
202				goto tx_error_icmp;
203			}
204			skb_dst_set(skb, dst);
205			break;
206#endif
207		default:
208			DEV_STATS_INC(dev, tx_carrier_errors);
209			goto tx_error_icmp;
210		}
211	}
212
213	dst_hold(dst);
214	dst = xfrm_lookup_route(tunnel->net, dst, fl, NULL, 0);
215	if (IS_ERR(dst)) {
216		DEV_STATS_INC(dev, tx_carrier_errors);
217		goto tx_error_icmp;
218	}
219
220	if (dst->flags & DST_XFRM_QUEUE)
221		goto xmit;
222
223	if (!vti_state_check(dst->xfrm, parms->iph.daddr, parms->iph.saddr)) {
224		DEV_STATS_INC(dev, tx_carrier_errors);
225		dst_release(dst);
226		goto tx_error_icmp;
227	}
228
229	tdev = dst->dev;
230
231	if (tdev == dev) {
232		dst_release(dst);
233		DEV_STATS_INC(dev, collisions);
234		goto tx_error;
235	}
236
237	mtu = dst_mtu(dst);
238	if (skb->len > mtu) {
239		skb_dst_update_pmtu_no_confirm(skb, mtu);
240		if (skb->protocol == htons(ETH_P_IP)) {
241			if (!(ip_hdr(skb)->frag_off & htons(IP_DF)))
242				goto xmit;
243			icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
244				      htonl(mtu));
245		} else {
246			if (mtu < IPV6_MIN_MTU)
247				mtu = IPV6_MIN_MTU;
248
249			icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
250		}
251
252		dst_release(dst);
253		goto tx_error;
254	}
255
256xmit:
257	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(dev)));
258	skb_dst_set(skb, dst);
259	skb->dev = skb_dst(skb)->dev;
260
261	err = dst_output(tunnel->net, skb->sk, skb);
262	if (net_xmit_eval(err) == 0)
263		err = pkt_len;
264	iptunnel_xmit_stats(dev, err);
265	return NETDEV_TX_OK;
266
267tx_error_icmp:
268	dst_link_failure(skb);
269tx_error:
270	DEV_STATS_INC(dev, tx_errors);
271	kfree_skb(skb);
272	return NETDEV_TX_OK;
273}
274
275/* This function assumes it is being called from dev_queue_xmit()
276 * and that skb is filled properly by that function.
277 */
278static netdev_tx_t vti_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
279{
280	struct ip_tunnel *tunnel = netdev_priv(dev);
281	struct flowi fl;
282
283	if (!pskb_inet_may_pull(skb))
284		goto tx_err;
285
286	memset(&fl, 0, sizeof(fl));
287
288	switch (skb->protocol) {
289	case htons(ETH_P_IP):
290		memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
291		xfrm_decode_session(dev_net(dev), skb, &fl, AF_INET);
292		break;
293	case htons(ETH_P_IPV6):
294		memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
295		xfrm_decode_session(dev_net(dev), skb, &fl, AF_INET6);
296		break;
297	default:
298		goto tx_err;
299	}
300
301	/* override mark with tunnel output key */
302	fl.flowi_mark = be32_to_cpu(tunnel->parms.o_key);
303
304	return vti_xmit(skb, dev, &fl);
305
306tx_err:
307	DEV_STATS_INC(dev, tx_errors);
308	kfree_skb(skb);
309	return NETDEV_TX_OK;
310}
311
312static int vti4_err(struct sk_buff *skb, u32 info)
313{
314	__be32 spi;
315	__u32 mark;
316	struct xfrm_state *x;
317	struct ip_tunnel *tunnel;
318	struct ip_esp_hdr *esph;
319	struct ip_auth_hdr *ah ;
320	struct ip_comp_hdr *ipch;
321	struct net *net = dev_net(skb->dev);
322	const struct iphdr *iph = (const struct iphdr *)skb->data;
323	int protocol = iph->protocol;
324	struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
 
 
 
325
326	tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
327				  iph->daddr, iph->saddr, 0);
328	if (!tunnel)
329		return -1;
330
331	mark = be32_to_cpu(tunnel->parms.o_key);
332
333	switch (protocol) {
334	case IPPROTO_ESP:
335		esph = (struct ip_esp_hdr *)(skb->data+(iph->ihl<<2));
336		spi = esph->spi;
337		break;
338	case IPPROTO_AH:
339		ah = (struct ip_auth_hdr *)(skb->data+(iph->ihl<<2));
340		spi = ah->spi;
341		break;
342	case IPPROTO_COMP:
343		ipch = (struct ip_comp_hdr *)(skb->data+(iph->ihl<<2));
344		spi = htonl(ntohs(ipch->cpi));
345		break;
346	default:
347		return 0;
348	}
349
350	switch (icmp_hdr(skb)->type) {
351	case ICMP_DEST_UNREACH:
352		if (icmp_hdr(skb)->code != ICMP_FRAG_NEEDED)
353			return 0;
354		break;
355	case ICMP_REDIRECT:
356		break;
357	default:
358		return 0;
359	}
360
361	x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
362			      spi, protocol, AF_INET);
363	if (!x)
364		return 0;
365
366	if (icmp_hdr(skb)->type == ICMP_DEST_UNREACH)
367		ipv4_update_pmtu(skb, net, info, 0, protocol);
368	else
369		ipv4_redirect(skb, net, 0, protocol);
370	xfrm_state_put(x);
371
372	return 0;
373}
374
375static int
376vti_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd)
377{
 
378	int err = 0;
379
380	if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
381		if (p->iph.version != 4 || p->iph.protocol != IPPROTO_IPIP ||
382		    p->iph.ihl != 5)
383			return -EINVAL;
384	}
385
386	if (!(p->i_flags & GRE_KEY))
 
 
 
 
387		p->i_key = 0;
388	if (!(p->o_flags & GRE_KEY))
389		p->o_key = 0;
390
391	p->i_flags = VTI_ISVTI;
 
392
393	err = ip_tunnel_ctl(dev, p, cmd);
394	if (err)
395		return err;
396
397	if (cmd != SIOCDELTUNNEL) {
398		p->i_flags |= GRE_KEY;
399		p->o_flags |= GRE_KEY;
 
400	}
401	return 0;
402}
403
404static const struct net_device_ops vti_netdev_ops = {
405	.ndo_init	= vti_tunnel_init,
406	.ndo_uninit	= ip_tunnel_uninit,
407	.ndo_start_xmit	= vti_tunnel_xmit,
408	.ndo_siocdevprivate = ip_tunnel_siocdevprivate,
409	.ndo_change_mtu	= ip_tunnel_change_mtu,
410	.ndo_get_stats64 = dev_get_tstats64,
411	.ndo_get_iflink = ip_tunnel_get_iflink,
412	.ndo_tunnel_ctl	= vti_tunnel_ctl,
413};
414
415static void vti_tunnel_setup(struct net_device *dev)
416{
417	dev->netdev_ops		= &vti_netdev_ops;
418	dev->header_ops		= &ip_tunnel_header_ops;
419	dev->type		= ARPHRD_TUNNEL;
420	ip_tunnel_setup(dev, vti_net_id);
421}
422
423static int vti_tunnel_init(struct net_device *dev)
424{
425	struct ip_tunnel *tunnel = netdev_priv(dev);
426	struct iphdr *iph = &tunnel->parms.iph;
427
428	__dev_addr_set(dev, &iph->saddr, 4);
429	memcpy(dev->broadcast, &iph->daddr, 4);
430
431	dev->flags		= IFF_NOARP;
432	dev->addr_len		= 4;
433	dev->features		|= NETIF_F_LLTX;
434	netif_keep_dst(dev);
435
436	return ip_tunnel_init(dev);
437}
438
439static void __net_init vti_fb_tunnel_init(struct net_device *dev)
440{
441	struct ip_tunnel *tunnel = netdev_priv(dev);
442	struct iphdr *iph = &tunnel->parms.iph;
443
444	iph->version		= 4;
445	iph->protocol		= IPPROTO_IPIP;
446	iph->ihl		= 5;
447}
448
449static struct xfrm4_protocol vti_esp4_protocol __read_mostly = {
450	.handler	=	vti_rcv_proto,
451	.input_handler	=	vti_input_proto,
452	.cb_handler	=	vti_rcv_cb,
453	.err_handler	=	vti4_err,
454	.priority	=	100,
455};
456
457static struct xfrm4_protocol vti_ah4_protocol __read_mostly = {
458	.handler	=	vti_rcv_proto,
459	.input_handler	=	vti_input_proto,
460	.cb_handler	=	vti_rcv_cb,
461	.err_handler	=	vti4_err,
462	.priority	=	100,
463};
464
465static struct xfrm4_protocol vti_ipcomp4_protocol __read_mostly = {
466	.handler	=	vti_rcv_proto,
467	.input_handler	=	vti_input_proto,
468	.cb_handler	=	vti_rcv_cb,
469	.err_handler	=	vti4_err,
470	.priority	=	100,
471};
472
473#if IS_ENABLED(CONFIG_INET_XFRM_TUNNEL)
474static int vti_rcv_tunnel(struct sk_buff *skb)
475{
476	XFRM_SPI_SKB_CB(skb)->family = AF_INET;
477	XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct iphdr, daddr);
478
479	return vti_input(skb, IPPROTO_IPIP, ip_hdr(skb)->saddr, 0, false);
480}
481
482static struct xfrm_tunnel vti_ipip_handler __read_mostly = {
483	.handler	=	vti_rcv_tunnel,
484	.cb_handler	=	vti_rcv_cb,
485	.err_handler	=	vti4_err,
486	.priority	=	0,
487};
488
489#if IS_ENABLED(CONFIG_IPV6)
490static struct xfrm_tunnel vti_ipip6_handler __read_mostly = {
491	.handler	=	vti_rcv_tunnel,
492	.cb_handler	=	vti_rcv_cb,
493	.err_handler	=	vti4_err,
494	.priority	=	0,
495};
496#endif
497#endif
498
499static int __net_init vti_init_net(struct net *net)
500{
501	int err;
502	struct ip_tunnel_net *itn;
503
504	err = ip_tunnel_init_net(net, vti_net_id, &vti_link_ops, "ip_vti0");
505	if (err)
506		return err;
507	itn = net_generic(net, vti_net_id);
508	if (itn->fb_tunnel_dev)
509		vti_fb_tunnel_init(itn->fb_tunnel_dev);
510	return 0;
511}
512
513static void __net_exit vti_exit_batch_net(struct list_head *list_net)
 
514{
515	ip_tunnel_delete_nets(list_net, vti_net_id, &vti_link_ops);
 
516}
517
518static struct pernet_operations vti_net_ops = {
519	.init = vti_init_net,
520	.exit_batch = vti_exit_batch_net,
521	.id   = &vti_net_id,
522	.size = sizeof(struct ip_tunnel_net),
523};
524
525static int vti_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
526			       struct netlink_ext_ack *extack)
527{
528	return 0;
529}
530
531static void vti_netlink_parms(struct nlattr *data[],
532			      struct ip_tunnel_parm *parms,
533			      __u32 *fwmark)
534{
535	memset(parms, 0, sizeof(*parms));
536
537	parms->iph.protocol = IPPROTO_IPIP;
538
539	if (!data)
540		return;
541
542	parms->i_flags = VTI_ISVTI;
543
544	if (data[IFLA_VTI_LINK])
545		parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
546
547	if (data[IFLA_VTI_IKEY])
548		parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
549
550	if (data[IFLA_VTI_OKEY])
551		parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
552
553	if (data[IFLA_VTI_LOCAL])
554		parms->iph.saddr = nla_get_in_addr(data[IFLA_VTI_LOCAL]);
555
556	if (data[IFLA_VTI_REMOTE])
557		parms->iph.daddr = nla_get_in_addr(data[IFLA_VTI_REMOTE]);
558
559	if (data[IFLA_VTI_FWMARK])
560		*fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]);
561}
562
563static int vti_newlink(struct net *src_net, struct net_device *dev,
564		       struct nlattr *tb[], struct nlattr *data[],
565		       struct netlink_ext_ack *extack)
566{
567	struct ip_tunnel_parm parms;
568	__u32 fwmark = 0;
569
570	vti_netlink_parms(data, &parms, &fwmark);
571	return ip_tunnel_newlink(dev, tb, &parms, fwmark);
572}
573
574static int vti_changelink(struct net_device *dev, struct nlattr *tb[],
575			  struct nlattr *data[],
576			  struct netlink_ext_ack *extack)
577{
578	struct ip_tunnel *t = netdev_priv(dev);
 
579	__u32 fwmark = t->fwmark;
580	struct ip_tunnel_parm p;
581
582	vti_netlink_parms(data, &p, &fwmark);
583	return ip_tunnel_changelink(dev, tb, &p, fwmark);
584}
585
586static size_t vti_get_size(const struct net_device *dev)
587{
588	return
589		/* IFLA_VTI_LINK */
590		nla_total_size(4) +
591		/* IFLA_VTI_IKEY */
592		nla_total_size(4) +
593		/* IFLA_VTI_OKEY */
594		nla_total_size(4) +
595		/* IFLA_VTI_LOCAL */
596		nla_total_size(4) +
597		/* IFLA_VTI_REMOTE */
598		nla_total_size(4) +
599		/* IFLA_VTI_FWMARK */
600		nla_total_size(4) +
601		0;
602}
603
604static int vti_fill_info(struct sk_buff *skb, const struct net_device *dev)
605{
606	struct ip_tunnel *t = netdev_priv(dev);
607	struct ip_tunnel_parm *p = &t->parms;
608
609	if (nla_put_u32(skb, IFLA_VTI_LINK, p->link) ||
610	    nla_put_be32(skb, IFLA_VTI_IKEY, p->i_key) ||
611	    nla_put_be32(skb, IFLA_VTI_OKEY, p->o_key) ||
612	    nla_put_in_addr(skb, IFLA_VTI_LOCAL, p->iph.saddr) ||
613	    nla_put_in_addr(skb, IFLA_VTI_REMOTE, p->iph.daddr) ||
614	    nla_put_u32(skb, IFLA_VTI_FWMARK, t->fwmark))
615		return -EMSGSIZE;
616
617	return 0;
618}
619
620static const struct nla_policy vti_policy[IFLA_VTI_MAX + 1] = {
621	[IFLA_VTI_LINK]		= { .type = NLA_U32 },
622	[IFLA_VTI_IKEY]		= { .type = NLA_U32 },
623	[IFLA_VTI_OKEY]		= { .type = NLA_U32 },
624	[IFLA_VTI_LOCAL]	= { .len = sizeof_field(struct iphdr, saddr) },
625	[IFLA_VTI_REMOTE]	= { .len = sizeof_field(struct iphdr, daddr) },
626	[IFLA_VTI_FWMARK]	= { .type = NLA_U32 },
627};
628
629static struct rtnl_link_ops vti_link_ops __read_mostly = {
630	.kind		= "vti",
631	.maxtype	= IFLA_VTI_MAX,
632	.policy		= vti_policy,
633	.priv_size	= sizeof(struct ip_tunnel),
634	.setup		= vti_tunnel_setup,
635	.validate	= vti_tunnel_validate,
636	.newlink	= vti_newlink,
637	.changelink	= vti_changelink,
638	.dellink        = ip_tunnel_dellink,
639	.get_size	= vti_get_size,
640	.fill_info	= vti_fill_info,
641	.get_link_net	= ip_tunnel_get_link_net,
642};
643
644static int __init vti_init(void)
645{
646	const char *msg;
647	int err;
648
649	pr_info("IPv4 over IPsec tunneling driver\n");
650
651	msg = "tunnel device";
652	err = register_pernet_device(&vti_net_ops);
653	if (err < 0)
654		goto pernet_dev_failed;
655
656	msg = "tunnel protocols";
657	err = xfrm4_protocol_register(&vti_esp4_protocol, IPPROTO_ESP);
658	if (err < 0)
659		goto xfrm_proto_esp_failed;
660	err = xfrm4_protocol_register(&vti_ah4_protocol, IPPROTO_AH);
661	if (err < 0)
662		goto xfrm_proto_ah_failed;
663	err = xfrm4_protocol_register(&vti_ipcomp4_protocol, IPPROTO_COMP);
664	if (err < 0)
665		goto xfrm_proto_comp_failed;
666
667#if IS_ENABLED(CONFIG_INET_XFRM_TUNNEL)
668	msg = "ipip tunnel";
669	err = xfrm4_tunnel_register(&vti_ipip_handler, AF_INET);
670	if (err < 0)
671		goto xfrm_tunnel_ipip_failed;
672#if IS_ENABLED(CONFIG_IPV6)
673	err = xfrm4_tunnel_register(&vti_ipip6_handler, AF_INET6);
674	if (err < 0)
675		goto xfrm_tunnel_ipip6_failed;
676#endif
677#endif
678
679	msg = "netlink interface";
680	err = rtnl_link_register(&vti_link_ops);
681	if (err < 0)
682		goto rtnl_link_failed;
683
684	return err;
685
686rtnl_link_failed:
687#if IS_ENABLED(CONFIG_INET_XFRM_TUNNEL)
688#if IS_ENABLED(CONFIG_IPV6)
689	xfrm4_tunnel_deregister(&vti_ipip6_handler, AF_INET6);
690xfrm_tunnel_ipip6_failed:
691#endif
692	xfrm4_tunnel_deregister(&vti_ipip_handler, AF_INET);
693xfrm_tunnel_ipip_failed:
694#endif
695	xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP);
696xfrm_proto_comp_failed:
697	xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
698xfrm_proto_ah_failed:
699	xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
700xfrm_proto_esp_failed:
701	unregister_pernet_device(&vti_net_ops);
702pernet_dev_failed:
703	pr_err("vti init: failed to register %s\n", msg);
704	return err;
705}
706
707static void __exit vti_fini(void)
708{
709	rtnl_link_unregister(&vti_link_ops);
710#if IS_ENABLED(CONFIG_INET_XFRM_TUNNEL)
711#if IS_ENABLED(CONFIG_IPV6)
712	xfrm4_tunnel_deregister(&vti_ipip6_handler, AF_INET6);
713#endif
714	xfrm4_tunnel_deregister(&vti_ipip_handler, AF_INET);
715#endif
716	xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP);
717	xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
718	xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
719	unregister_pernet_device(&vti_net_ops);
720}
721
722module_init(vti_init);
723module_exit(vti_fini);
724MODULE_DESCRIPTION("Virtual (secure) IP tunneling library");
725MODULE_LICENSE("GPL");
726MODULE_ALIAS_RTNL_LINK("vti");
727MODULE_ALIAS_NETDEV("ip_vti0");
v6.13.7
  1// SPDX-License-Identifier: GPL-2.0-or-later
  2/*
  3 *	Linux NET3: IP/IP protocol decoder modified to support
  4 *		    virtual tunnel interface
  5 *
  6 *	Authors:
  7 *		Saurabh Mohan (saurabh.mohan@vyatta.com) 05/07/2012
  8 */
  9
 10/*
 11   This version of net/ipv4/ip_vti.c is cloned of net/ipv4/ipip.c
 12
 13   For comments look at net/ipv4/ip_gre.c --ANK
 14 */
 15
 16
 17#include <linux/capability.h>
 18#include <linux/module.h>
 19#include <linux/types.h>
 20#include <linux/kernel.h>
 21#include <linux/uaccess.h>
 22#include <linux/skbuff.h>
 23#include <linux/netdevice.h>
 24#include <linux/in.h>
 25#include <linux/tcp.h>
 26#include <linux/udp.h>
 27#include <linux/if_arp.h>
 28#include <linux/init.h>
 29#include <linux/netfilter_ipv4.h>
 30#include <linux/if_ether.h>
 31#include <linux/icmpv6.h>
 32
 33#include <net/sock.h>
 34#include <net/ip.h>
 35#include <net/icmp.h>
 36#include <net/ip_tunnels.h>
 37#include <net/inet_ecn.h>
 38#include <net/xfrm.h>
 39#include <net/net_namespace.h>
 40#include <net/netns/generic.h>
 41
 42static struct rtnl_link_ops vti_link_ops __read_mostly;
 43
 44static unsigned int vti_net_id __read_mostly;
 45static int vti_tunnel_init(struct net_device *dev);
 46
 47static int vti_input(struct sk_buff *skb, int nexthdr, __be32 spi,
 48		     int encap_type, bool update_skb_dev)
 49{
 50	struct ip_tunnel *tunnel;
 51	const struct iphdr *iph = ip_hdr(skb);
 52	struct net *net = dev_net(skb->dev);
 53	struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
 54	IP_TUNNEL_DECLARE_FLAGS(flags) = { };
 55
 56	__set_bit(IP_TUNNEL_NO_KEY_BIT, flags);
 57
 58	tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, flags,
 59				  iph->saddr, iph->daddr, 0);
 60	if (tunnel) {
 61		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 62			goto drop;
 63
 64		XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4 = tunnel;
 65
 66		if (update_skb_dev)
 67			skb->dev = tunnel->dev;
 68
 69		return xfrm_input(skb, nexthdr, spi, encap_type);
 70	}
 71
 72	return -EINVAL;
 73drop:
 74	kfree_skb(skb);
 75	return 0;
 76}
 77
 78static int vti_input_proto(struct sk_buff *skb, int nexthdr, __be32 spi,
 79			   int encap_type)
 80{
 81	return vti_input(skb, nexthdr, spi, encap_type, false);
 82}
 83
 84static int vti_rcv(struct sk_buff *skb, __be32 spi, bool update_skb_dev)
 85{
 86	XFRM_SPI_SKB_CB(skb)->family = AF_INET;
 87	XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct iphdr, daddr);
 88
 89	return vti_input(skb, ip_hdr(skb)->protocol, spi, 0, update_skb_dev);
 90}
 91
 92static int vti_rcv_proto(struct sk_buff *skb)
 93{
 94	return vti_rcv(skb, 0, false);
 95}
 96
 97static int vti_rcv_cb(struct sk_buff *skb, int err)
 98{
 99	unsigned short family;
100	struct net_device *dev;
101	struct xfrm_state *x;
102	const struct xfrm_mode *inner_mode;
103	struct ip_tunnel *tunnel = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4;
104	u32 orig_mark = skb->mark;
105	int ret;
106
107	if (!tunnel)
108		return 1;
109
110	dev = tunnel->dev;
111
112	if (err) {
113		DEV_STATS_INC(dev, rx_errors);
114		DEV_STATS_INC(dev, rx_dropped);
115
116		return 0;
117	}
118
119	x = xfrm_input_state(skb);
120
121	inner_mode = &x->inner_mode;
122
123	if (x->sel.family == AF_UNSPEC) {
124		inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
125		if (inner_mode == NULL) {
126			XFRM_INC_STATS(dev_net(skb->dev),
127				       LINUX_MIB_XFRMINSTATEMODEERROR);
128			return -EINVAL;
129		}
130	}
131
132	family = inner_mode->family;
133
134	skb->mark = be32_to_cpu(tunnel->parms.i_key);
135	ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
136	skb->mark = orig_mark;
137
138	if (!ret)
139		return -EPERM;
140
141	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(skb->dev)));
142	skb->dev = dev;
143	dev_sw_netstats_rx_add(dev, skb->len);
144
145	return 0;
146}
147
148static bool vti_state_check(const struct xfrm_state *x, __be32 dst, __be32 src)
149{
150	xfrm_address_t *daddr = (xfrm_address_t *)&dst;
151	xfrm_address_t *saddr = (xfrm_address_t *)&src;
152
153	/* if there is no transform then this tunnel is not functional.
154	 * Or if the xfrm is not mode tunnel.
155	 */
156	if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
157	    x->props.family != AF_INET)
158		return false;
159
160	if (!dst)
161		return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET);
162
163	if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET))
164		return false;
165
166	return true;
167}
168
169static netdev_tx_t vti_xmit(struct sk_buff *skb, struct net_device *dev,
170			    struct flowi *fl)
171{
172	struct ip_tunnel *tunnel = netdev_priv(dev);
173	struct ip_tunnel_parm_kern *parms = &tunnel->parms;
174	struct dst_entry *dst = skb_dst(skb);
175	struct net_device *tdev;	/* Device to other host */
176	int pkt_len = skb->len;
177	int err;
178	int mtu;
179
180	if (!dst) {
181		switch (skb->protocol) {
182		case htons(ETH_P_IP): {
183			struct rtable *rt;
184
185			fl->u.ip4.flowi4_oif = dev->ifindex;
186			fl->u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC;
187			rt = __ip_route_output_key(dev_net(dev), &fl->u.ip4);
188			if (IS_ERR(rt)) {
189				DEV_STATS_INC(dev, tx_carrier_errors);
190				goto tx_error_icmp;
191			}
192			dst = &rt->dst;
193			skb_dst_set(skb, dst);
194			break;
195		}
196#if IS_ENABLED(CONFIG_IPV6)
197		case htons(ETH_P_IPV6):
198			fl->u.ip6.flowi6_oif = dev->ifindex;
199			fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
200			dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6);
201			if (dst->error) {
202				dst_release(dst);
203				dst = NULL;
204				DEV_STATS_INC(dev, tx_carrier_errors);
205				goto tx_error_icmp;
206			}
207			skb_dst_set(skb, dst);
208			break;
209#endif
210		default:
211			DEV_STATS_INC(dev, tx_carrier_errors);
212			goto tx_error_icmp;
213		}
214	}
215
216	dst_hold(dst);
217	dst = xfrm_lookup_route(tunnel->net, dst, fl, NULL, 0);
218	if (IS_ERR(dst)) {
219		DEV_STATS_INC(dev, tx_carrier_errors);
220		goto tx_error_icmp;
221	}
222
223	if (dst->flags & DST_XFRM_QUEUE)
224		goto xmit;
225
226	if (!vti_state_check(dst->xfrm, parms->iph.daddr, parms->iph.saddr)) {
227		DEV_STATS_INC(dev, tx_carrier_errors);
228		dst_release(dst);
229		goto tx_error_icmp;
230	}
231
232	tdev = dst->dev;
233
234	if (tdev == dev) {
235		dst_release(dst);
236		DEV_STATS_INC(dev, collisions);
237		goto tx_error;
238	}
239
240	mtu = dst_mtu(dst);
241	if (skb->len > mtu) {
242		skb_dst_update_pmtu_no_confirm(skb, mtu);
243		if (skb->protocol == htons(ETH_P_IP)) {
244			if (!(ip_hdr(skb)->frag_off & htons(IP_DF)))
245				goto xmit;
246			icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
247				      htonl(mtu));
248		} else {
249			if (mtu < IPV6_MIN_MTU)
250				mtu = IPV6_MIN_MTU;
251
252			icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
253		}
254
255		dst_release(dst);
256		goto tx_error;
257	}
258
259xmit:
260	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(dev)));
261	skb_dst_set(skb, dst);
262	skb->dev = skb_dst(skb)->dev;
263
264	err = dst_output(tunnel->net, skb->sk, skb);
265	if (net_xmit_eval(err) == 0)
266		err = pkt_len;
267	iptunnel_xmit_stats(dev, err);
268	return NETDEV_TX_OK;
269
270tx_error_icmp:
271	dst_link_failure(skb);
272tx_error:
273	DEV_STATS_INC(dev, tx_errors);
274	kfree_skb(skb);
275	return NETDEV_TX_OK;
276}
277
278/* This function assumes it is being called from dev_queue_xmit()
279 * and that skb is filled properly by that function.
280 */
281static netdev_tx_t vti_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
282{
283	struct ip_tunnel *tunnel = netdev_priv(dev);
284	struct flowi fl;
285
286	if (!pskb_inet_may_pull(skb))
287		goto tx_err;
288
289	memset(&fl, 0, sizeof(fl));
290
291	switch (skb->protocol) {
292	case htons(ETH_P_IP):
293		memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
294		xfrm_decode_session(dev_net(dev), skb, &fl, AF_INET);
295		break;
296	case htons(ETH_P_IPV6):
297		memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
298		xfrm_decode_session(dev_net(dev), skb, &fl, AF_INET6);
299		break;
300	default:
301		goto tx_err;
302	}
303
304	/* override mark with tunnel output key */
305	fl.flowi_mark = be32_to_cpu(tunnel->parms.o_key);
306
307	return vti_xmit(skb, dev, &fl);
308
309tx_err:
310	DEV_STATS_INC(dev, tx_errors);
311	kfree_skb(skb);
312	return NETDEV_TX_OK;
313}
314
315static int vti4_err(struct sk_buff *skb, u32 info)
316{
317	__be32 spi;
318	__u32 mark;
319	struct xfrm_state *x;
320	struct ip_tunnel *tunnel;
321	struct ip_esp_hdr *esph;
322	struct ip_auth_hdr *ah ;
323	struct ip_comp_hdr *ipch;
324	struct net *net = dev_net(skb->dev);
325	const struct iphdr *iph = (const struct iphdr *)skb->data;
326	int protocol = iph->protocol;
327	struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
328	IP_TUNNEL_DECLARE_FLAGS(flags) = { };
329
330	__set_bit(IP_TUNNEL_NO_KEY_BIT, flags);
331
332	tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, flags,
333				  iph->daddr, iph->saddr, 0);
334	if (!tunnel)
335		return -1;
336
337	mark = be32_to_cpu(tunnel->parms.o_key);
338
339	switch (protocol) {
340	case IPPROTO_ESP:
341		esph = (struct ip_esp_hdr *)(skb->data+(iph->ihl<<2));
342		spi = esph->spi;
343		break;
344	case IPPROTO_AH:
345		ah = (struct ip_auth_hdr *)(skb->data+(iph->ihl<<2));
346		spi = ah->spi;
347		break;
348	case IPPROTO_COMP:
349		ipch = (struct ip_comp_hdr *)(skb->data+(iph->ihl<<2));
350		spi = htonl(ntohs(ipch->cpi));
351		break;
352	default:
353		return 0;
354	}
355
356	switch (icmp_hdr(skb)->type) {
357	case ICMP_DEST_UNREACH:
358		if (icmp_hdr(skb)->code != ICMP_FRAG_NEEDED)
359			return 0;
360		break;
361	case ICMP_REDIRECT:
362		break;
363	default:
364		return 0;
365	}
366
367	x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
368			      spi, protocol, AF_INET);
369	if (!x)
370		return 0;
371
372	if (icmp_hdr(skb)->type == ICMP_DEST_UNREACH)
373		ipv4_update_pmtu(skb, net, info, 0, protocol);
374	else
375		ipv4_redirect(skb, net, 0, protocol);
376	xfrm_state_put(x);
377
378	return 0;
379}
380
381static int
382vti_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm_kern *p, int cmd)
383{
384	IP_TUNNEL_DECLARE_FLAGS(flags) = { };
385	int err = 0;
386
387	if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
388		if (p->iph.version != 4 || p->iph.protocol != IPPROTO_IPIP ||
389		    p->iph.ihl != 5)
390			return -EINVAL;
391	}
392
393	if (!ip_tunnel_flags_is_be16_compat(p->i_flags) ||
394	    !ip_tunnel_flags_is_be16_compat(p->o_flags))
395		return -EOVERFLOW;
396
397	if (!(ip_tunnel_flags_to_be16(p->i_flags) & GRE_KEY))
398		p->i_key = 0;
399	if (!(ip_tunnel_flags_to_be16(p->o_flags) & GRE_KEY))
400		p->o_key = 0;
401
402	__set_bit(IP_TUNNEL_VTI_BIT, flags);
403	ip_tunnel_flags_copy(p->i_flags, flags);
404
405	err = ip_tunnel_ctl(dev, p, cmd);
406	if (err)
407		return err;
408
409	if (cmd != SIOCDELTUNNEL) {
410		ip_tunnel_flags_from_be16(flags, GRE_KEY);
411		ip_tunnel_flags_or(p->i_flags, p->i_flags, flags);
412		ip_tunnel_flags_or(p->o_flags, p->o_flags, flags);
413	}
414	return 0;
415}
416
417static const struct net_device_ops vti_netdev_ops = {
418	.ndo_init	= vti_tunnel_init,
419	.ndo_uninit	= ip_tunnel_uninit,
420	.ndo_start_xmit	= vti_tunnel_xmit,
421	.ndo_siocdevprivate = ip_tunnel_siocdevprivate,
422	.ndo_change_mtu	= ip_tunnel_change_mtu,
423	.ndo_get_stats64 = dev_get_tstats64,
424	.ndo_get_iflink = ip_tunnel_get_iflink,
425	.ndo_tunnel_ctl	= vti_tunnel_ctl,
426};
427
428static void vti_tunnel_setup(struct net_device *dev)
429{
430	dev->netdev_ops		= &vti_netdev_ops;
431	dev->header_ops		= &ip_tunnel_header_ops;
432	dev->type		= ARPHRD_TUNNEL;
433	ip_tunnel_setup(dev, vti_net_id);
434}
435
436static int vti_tunnel_init(struct net_device *dev)
437{
438	struct ip_tunnel *tunnel = netdev_priv(dev);
439	struct iphdr *iph = &tunnel->parms.iph;
440
441	__dev_addr_set(dev, &iph->saddr, 4);
442	memcpy(dev->broadcast, &iph->daddr, 4);
443
444	dev->flags		= IFF_NOARP;
445	dev->addr_len		= 4;
446	dev->lltx		= true;
447	netif_keep_dst(dev);
448
449	return ip_tunnel_init(dev);
450}
451
452static void __net_init vti_fb_tunnel_init(struct net_device *dev)
453{
454	struct ip_tunnel *tunnel = netdev_priv(dev);
455	struct iphdr *iph = &tunnel->parms.iph;
456
457	iph->version		= 4;
458	iph->protocol		= IPPROTO_IPIP;
459	iph->ihl		= 5;
460}
461
462static struct xfrm4_protocol vti_esp4_protocol __read_mostly = {
463	.handler	=	vti_rcv_proto,
464	.input_handler	=	vti_input_proto,
465	.cb_handler	=	vti_rcv_cb,
466	.err_handler	=	vti4_err,
467	.priority	=	100,
468};
469
470static struct xfrm4_protocol vti_ah4_protocol __read_mostly = {
471	.handler	=	vti_rcv_proto,
472	.input_handler	=	vti_input_proto,
473	.cb_handler	=	vti_rcv_cb,
474	.err_handler	=	vti4_err,
475	.priority	=	100,
476};
477
478static struct xfrm4_protocol vti_ipcomp4_protocol __read_mostly = {
479	.handler	=	vti_rcv_proto,
480	.input_handler	=	vti_input_proto,
481	.cb_handler	=	vti_rcv_cb,
482	.err_handler	=	vti4_err,
483	.priority	=	100,
484};
485
486#if IS_ENABLED(CONFIG_INET_XFRM_TUNNEL)
487static int vti_rcv_tunnel(struct sk_buff *skb)
488{
489	XFRM_SPI_SKB_CB(skb)->family = AF_INET;
490	XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct iphdr, daddr);
491
492	return vti_input(skb, IPPROTO_IPIP, ip_hdr(skb)->saddr, 0, false);
493}
494
495static struct xfrm_tunnel vti_ipip_handler __read_mostly = {
496	.handler	=	vti_rcv_tunnel,
497	.cb_handler	=	vti_rcv_cb,
498	.err_handler	=	vti4_err,
499	.priority	=	0,
500};
501
502#if IS_ENABLED(CONFIG_IPV6)
503static struct xfrm_tunnel vti_ipip6_handler __read_mostly = {
504	.handler	=	vti_rcv_tunnel,
505	.cb_handler	=	vti_rcv_cb,
506	.err_handler	=	vti4_err,
507	.priority	=	0,
508};
509#endif
510#endif
511
512static int __net_init vti_init_net(struct net *net)
513{
514	int err;
515	struct ip_tunnel_net *itn;
516
517	err = ip_tunnel_init_net(net, vti_net_id, &vti_link_ops, "ip_vti0");
518	if (err)
519		return err;
520	itn = net_generic(net, vti_net_id);
521	if (itn->fb_tunnel_dev)
522		vti_fb_tunnel_init(itn->fb_tunnel_dev);
523	return 0;
524}
525
526static void __net_exit vti_exit_batch_rtnl(struct list_head *list_net,
527					   struct list_head *dev_to_kill)
528{
529	ip_tunnel_delete_nets(list_net, vti_net_id, &vti_link_ops,
530			      dev_to_kill);
531}
532
533static struct pernet_operations vti_net_ops = {
534	.init = vti_init_net,
535	.exit_batch_rtnl = vti_exit_batch_rtnl,
536	.id   = &vti_net_id,
537	.size = sizeof(struct ip_tunnel_net),
538};
539
540static int vti_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
541			       struct netlink_ext_ack *extack)
542{
543	return 0;
544}
545
546static void vti_netlink_parms(struct nlattr *data[],
547			      struct ip_tunnel_parm_kern *parms,
548			      __u32 *fwmark)
549{
550	memset(parms, 0, sizeof(*parms));
551
552	parms->iph.protocol = IPPROTO_IPIP;
553
554	if (!data)
555		return;
556
557	__set_bit(IP_TUNNEL_VTI_BIT, parms->i_flags);
558
559	if (data[IFLA_VTI_LINK])
560		parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
561
562	if (data[IFLA_VTI_IKEY])
563		parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
564
565	if (data[IFLA_VTI_OKEY])
566		parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
567
568	if (data[IFLA_VTI_LOCAL])
569		parms->iph.saddr = nla_get_in_addr(data[IFLA_VTI_LOCAL]);
570
571	if (data[IFLA_VTI_REMOTE])
572		parms->iph.daddr = nla_get_in_addr(data[IFLA_VTI_REMOTE]);
573
574	if (data[IFLA_VTI_FWMARK])
575		*fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]);
576}
577
578static int vti_newlink(struct net *src_net, struct net_device *dev,
579		       struct nlattr *tb[], struct nlattr *data[],
580		       struct netlink_ext_ack *extack)
581{
582	struct ip_tunnel_parm_kern parms;
583	__u32 fwmark = 0;
584
585	vti_netlink_parms(data, &parms, &fwmark);
586	return ip_tunnel_newlink(dev, tb, &parms, fwmark);
587}
588
589static int vti_changelink(struct net_device *dev, struct nlattr *tb[],
590			  struct nlattr *data[],
591			  struct netlink_ext_ack *extack)
592{
593	struct ip_tunnel *t = netdev_priv(dev);
594	struct ip_tunnel_parm_kern p;
595	__u32 fwmark = t->fwmark;
 
596
597	vti_netlink_parms(data, &p, &fwmark);
598	return ip_tunnel_changelink(dev, tb, &p, fwmark);
599}
600
601static size_t vti_get_size(const struct net_device *dev)
602{
603	return
604		/* IFLA_VTI_LINK */
605		nla_total_size(4) +
606		/* IFLA_VTI_IKEY */
607		nla_total_size(4) +
608		/* IFLA_VTI_OKEY */
609		nla_total_size(4) +
610		/* IFLA_VTI_LOCAL */
611		nla_total_size(4) +
612		/* IFLA_VTI_REMOTE */
613		nla_total_size(4) +
614		/* IFLA_VTI_FWMARK */
615		nla_total_size(4) +
616		0;
617}
618
619static int vti_fill_info(struct sk_buff *skb, const struct net_device *dev)
620{
621	struct ip_tunnel *t = netdev_priv(dev);
622	struct ip_tunnel_parm_kern *p = &t->parms;
623
624	if (nla_put_u32(skb, IFLA_VTI_LINK, p->link) ||
625	    nla_put_be32(skb, IFLA_VTI_IKEY, p->i_key) ||
626	    nla_put_be32(skb, IFLA_VTI_OKEY, p->o_key) ||
627	    nla_put_in_addr(skb, IFLA_VTI_LOCAL, p->iph.saddr) ||
628	    nla_put_in_addr(skb, IFLA_VTI_REMOTE, p->iph.daddr) ||
629	    nla_put_u32(skb, IFLA_VTI_FWMARK, t->fwmark))
630		return -EMSGSIZE;
631
632	return 0;
633}
634
635static const struct nla_policy vti_policy[IFLA_VTI_MAX + 1] = {
636	[IFLA_VTI_LINK]		= { .type = NLA_U32 },
637	[IFLA_VTI_IKEY]		= { .type = NLA_U32 },
638	[IFLA_VTI_OKEY]		= { .type = NLA_U32 },
639	[IFLA_VTI_LOCAL]	= { .len = sizeof_field(struct iphdr, saddr) },
640	[IFLA_VTI_REMOTE]	= { .len = sizeof_field(struct iphdr, daddr) },
641	[IFLA_VTI_FWMARK]	= { .type = NLA_U32 },
642};
643
644static struct rtnl_link_ops vti_link_ops __read_mostly = {
645	.kind		= "vti",
646	.maxtype	= IFLA_VTI_MAX,
647	.policy		= vti_policy,
648	.priv_size	= sizeof(struct ip_tunnel),
649	.setup		= vti_tunnel_setup,
650	.validate	= vti_tunnel_validate,
651	.newlink	= vti_newlink,
652	.changelink	= vti_changelink,
653	.dellink        = ip_tunnel_dellink,
654	.get_size	= vti_get_size,
655	.fill_info	= vti_fill_info,
656	.get_link_net	= ip_tunnel_get_link_net,
657};
658
659static int __init vti_init(void)
660{
661	const char *msg;
662	int err;
663
664	pr_info("IPv4 over IPsec tunneling driver\n");
665
666	msg = "tunnel device";
667	err = register_pernet_device(&vti_net_ops);
668	if (err < 0)
669		goto pernet_dev_failed;
670
671	msg = "tunnel protocols";
672	err = xfrm4_protocol_register(&vti_esp4_protocol, IPPROTO_ESP);
673	if (err < 0)
674		goto xfrm_proto_esp_failed;
675	err = xfrm4_protocol_register(&vti_ah4_protocol, IPPROTO_AH);
676	if (err < 0)
677		goto xfrm_proto_ah_failed;
678	err = xfrm4_protocol_register(&vti_ipcomp4_protocol, IPPROTO_COMP);
679	if (err < 0)
680		goto xfrm_proto_comp_failed;
681
682#if IS_ENABLED(CONFIG_INET_XFRM_TUNNEL)
683	msg = "ipip tunnel";
684	err = xfrm4_tunnel_register(&vti_ipip_handler, AF_INET);
685	if (err < 0)
686		goto xfrm_tunnel_ipip_failed;
687#if IS_ENABLED(CONFIG_IPV6)
688	err = xfrm4_tunnel_register(&vti_ipip6_handler, AF_INET6);
689	if (err < 0)
690		goto xfrm_tunnel_ipip6_failed;
691#endif
692#endif
693
694	msg = "netlink interface";
695	err = rtnl_link_register(&vti_link_ops);
696	if (err < 0)
697		goto rtnl_link_failed;
698
699	return err;
700
701rtnl_link_failed:
702#if IS_ENABLED(CONFIG_INET_XFRM_TUNNEL)
703#if IS_ENABLED(CONFIG_IPV6)
704	xfrm4_tunnel_deregister(&vti_ipip6_handler, AF_INET6);
705xfrm_tunnel_ipip6_failed:
706#endif
707	xfrm4_tunnel_deregister(&vti_ipip_handler, AF_INET);
708xfrm_tunnel_ipip_failed:
709#endif
710	xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP);
711xfrm_proto_comp_failed:
712	xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
713xfrm_proto_ah_failed:
714	xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
715xfrm_proto_esp_failed:
716	unregister_pernet_device(&vti_net_ops);
717pernet_dev_failed:
718	pr_err("vti init: failed to register %s\n", msg);
719	return err;
720}
721
722static void __exit vti_fini(void)
723{
724	rtnl_link_unregister(&vti_link_ops);
725#if IS_ENABLED(CONFIG_INET_XFRM_TUNNEL)
726#if IS_ENABLED(CONFIG_IPV6)
727	xfrm4_tunnel_deregister(&vti_ipip6_handler, AF_INET6);
728#endif
729	xfrm4_tunnel_deregister(&vti_ipip_handler, AF_INET);
730#endif
731	xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP);
732	xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
733	xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
734	unregister_pernet_device(&vti_net_ops);
735}
736
737module_init(vti_init);
738module_exit(vti_fini);
739MODULE_DESCRIPTION("Virtual (secure) IP tunneling library");
740MODULE_LICENSE("GPL");
741MODULE_ALIAS_RTNL_LINK("vti");
742MODULE_ALIAS_NETDEV("ip_vti0");