xref: /openbmc/linux/net/ipv4/ip_vti.c (revision b34e08d5)
1 /*
2  *	Linux NET3: IP/IP protocol decoder modified to support
3  *		    virtual tunnel interface
4  *
5  *	Authors:
6  *		Saurabh Mohan (saurabh.mohan@vyatta.com) 05/07/2012
7  *
8  *	This program is free software; you can redistribute it and/or
9  *	modify it under the terms of the GNU General Public License
10  *	as published by the Free Software Foundation; either version
11  *	2 of the License, or (at your option) any later version.
12  *
13  */
14 
15 /*
16    This version of net/ipv4/ip_vti.c is cloned of net/ipv4/ipip.c
17 
18    For comments look at net/ipv4/ip_gre.c --ANK
19  */
20 
21 
22 #include <linux/capability.h>
23 #include <linux/module.h>
24 #include <linux/types.h>
25 #include <linux/kernel.h>
26 #include <linux/uaccess.h>
27 #include <linux/skbuff.h>
28 #include <linux/netdevice.h>
29 #include <linux/in.h>
30 #include <linux/tcp.h>
31 #include <linux/udp.h>
32 #include <linux/if_arp.h>
33 #include <linux/mroute.h>
34 #include <linux/init.h>
35 #include <linux/netfilter_ipv4.h>
36 #include <linux/if_ether.h>
37 #include <linux/icmpv6.h>
38 
39 #include <net/sock.h>
40 #include <net/ip.h>
41 #include <net/icmp.h>
42 #include <net/ip_tunnels.h>
43 #include <net/inet_ecn.h>
44 #include <net/xfrm.h>
45 #include <net/net_namespace.h>
46 #include <net/netns/generic.h>
47 
48 static struct rtnl_link_ops vti_link_ops __read_mostly;
49 
50 static int vti_net_id __read_mostly;
51 static int vti_tunnel_init(struct net_device *dev);
52 
53 static int vti_input(struct sk_buff *skb, int nexthdr, __be32 spi,
54 		     int encap_type)
55 {
56 	struct ip_tunnel *tunnel;
57 	const struct iphdr *iph = ip_hdr(skb);
58 	struct net *net = dev_net(skb->dev);
59 	struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
60 
61 	tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
62 				  iph->saddr, iph->daddr, 0);
63 	if (tunnel != NULL) {
64 		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
65 			goto drop;
66 
67 		XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4 = tunnel;
68 		skb->mark = be32_to_cpu(tunnel->parms.i_key);
69 
70 		return xfrm_input(skb, nexthdr, spi, encap_type);
71 	}
72 
73 	return -EINVAL;
74 drop:
75 	kfree_skb(skb);
76 	return 0;
77 }
78 
79 static int vti_rcv(struct sk_buff *skb)
80 {
81 	XFRM_SPI_SKB_CB(skb)->family = AF_INET;
82 	XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct iphdr, daddr);
83 
84 	return vti_input(skb, ip_hdr(skb)->protocol, 0, 0);
85 }
86 
87 static int vti_rcv_cb(struct sk_buff *skb, int err)
88 {
89 	unsigned short family;
90 	struct net_device *dev;
91 	struct pcpu_sw_netstats *tstats;
92 	struct xfrm_state *x;
93 	struct ip_tunnel *tunnel = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4;
94 
95 	if (!tunnel)
96 		return 1;
97 
98 	dev = tunnel->dev;
99 
100 	if (err) {
101 		dev->stats.rx_errors++;
102 		dev->stats.rx_dropped++;
103 
104 		return 0;
105 	}
106 
107 	x = xfrm_input_state(skb);
108 	family = x->inner_mode->afinfo->family;
109 
110 	if (!xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family))
111 		return -EPERM;
112 
113 	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(skb->dev)));
114 	skb->dev = dev;
115 
116 	tstats = this_cpu_ptr(dev->tstats);
117 
118 	u64_stats_update_begin(&tstats->syncp);
119 	tstats->rx_packets++;
120 	tstats->rx_bytes += skb->len;
121 	u64_stats_update_end(&tstats->syncp);
122 
123 	return 0;
124 }
125 
126 static bool vti_state_check(const struct xfrm_state *x, __be32 dst, __be32 src)
127 {
128 	xfrm_address_t *daddr = (xfrm_address_t *)&dst;
129 	xfrm_address_t *saddr = (xfrm_address_t *)&src;
130 
131 	/* if there is no transform then this tunnel is not functional.
132 	 * Or if the xfrm is not mode tunnel.
133 	 */
134 	if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
135 	    x->props.family != AF_INET)
136 		return false;
137 
138 	if (!dst)
139 		return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET);
140 
141 	if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET))
142 		return false;
143 
144 	return true;
145 }
146 
147 static netdev_tx_t vti_xmit(struct sk_buff *skb, struct net_device *dev,
148 			    struct flowi *fl)
149 {
150 	struct ip_tunnel *tunnel = netdev_priv(dev);
151 	struct ip_tunnel_parm *parms = &tunnel->parms;
152 	struct dst_entry *dst = skb_dst(skb);
153 	struct net_device *tdev;	/* Device to other host */
154 	int err;
155 
156 	if (!dst) {
157 		dev->stats.tx_carrier_errors++;
158 		goto tx_error_icmp;
159 	}
160 
161 	dst_hold(dst);
162 	dst = xfrm_lookup(tunnel->net, dst, fl, NULL, 0);
163 	if (IS_ERR(dst)) {
164 		dev->stats.tx_carrier_errors++;
165 		goto tx_error_icmp;
166 	}
167 
168 	if (!vti_state_check(dst->xfrm, parms->iph.daddr, parms->iph.saddr)) {
169 		dev->stats.tx_carrier_errors++;
170 		dst_release(dst);
171 		goto tx_error_icmp;
172 	}
173 
174 	tdev = dst->dev;
175 
176 	if (tdev == dev) {
177 		dst_release(dst);
178 		dev->stats.collisions++;
179 		goto tx_error;
180 	}
181 
182 	if (tunnel->err_count > 0) {
183 		if (time_before(jiffies,
184 				tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
185 			tunnel->err_count--;
186 			dst_link_failure(skb);
187 		} else
188 			tunnel->err_count = 0;
189 	}
190 
191 	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(dev)));
192 	skb_dst_set(skb, dst);
193 	skb->dev = skb_dst(skb)->dev;
194 
195 	err = dst_output(skb);
196 	if (net_xmit_eval(err) == 0)
197 		err = skb->len;
198 	iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
199 	return NETDEV_TX_OK;
200 
201 tx_error_icmp:
202 	dst_link_failure(skb);
203 tx_error:
204 	dev->stats.tx_errors++;
205 	kfree_skb(skb);
206 	return NETDEV_TX_OK;
207 }
208 
209 /* This function assumes it is being called from dev_queue_xmit()
210  * and that skb is filled properly by that function.
211  */
212 static netdev_tx_t vti_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
213 {
214 	struct ip_tunnel *tunnel = netdev_priv(dev);
215 	struct flowi fl;
216 
217 	memset(&fl, 0, sizeof(fl));
218 
219 	skb->mark = be32_to_cpu(tunnel->parms.o_key);
220 
221 	switch (skb->protocol) {
222 	case htons(ETH_P_IP):
223 		xfrm_decode_session(skb, &fl, AF_INET);
224 		memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
225 		break;
226 	case htons(ETH_P_IPV6):
227 		xfrm_decode_session(skb, &fl, AF_INET6);
228 		memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
229 		break;
230 	default:
231 		dev->stats.tx_errors++;
232 		dev_kfree_skb(skb);
233 		return NETDEV_TX_OK;
234 	}
235 
236 	return vti_xmit(skb, dev, &fl);
237 }
238 
239 static int vti4_err(struct sk_buff *skb, u32 info)
240 {
241 	__be32 spi;
242 	struct xfrm_state *x;
243 	struct ip_tunnel *tunnel;
244 	struct ip_esp_hdr *esph;
245 	struct ip_auth_hdr *ah ;
246 	struct ip_comp_hdr *ipch;
247 	struct net *net = dev_net(skb->dev);
248 	const struct iphdr *iph = (const struct iphdr *)skb->data;
249 	int protocol = iph->protocol;
250 	struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
251 
252 	tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
253 				  iph->daddr, iph->saddr, 0);
254 	if (!tunnel)
255 		return -1;
256 
257 	switch (protocol) {
258 	case IPPROTO_ESP:
259 		esph = (struct ip_esp_hdr *)(skb->data+(iph->ihl<<2));
260 		spi = esph->spi;
261 		break;
262 	case IPPROTO_AH:
263 		ah = (struct ip_auth_hdr *)(skb->data+(iph->ihl<<2));
264 		spi = ah->spi;
265 		break;
266 	case IPPROTO_COMP:
267 		ipch = (struct ip_comp_hdr *)(skb->data+(iph->ihl<<2));
268 		spi = htonl(ntohs(ipch->cpi));
269 		break;
270 	default:
271 		return 0;
272 	}
273 
274 	switch (icmp_hdr(skb)->type) {
275 	case ICMP_DEST_UNREACH:
276 		if (icmp_hdr(skb)->code != ICMP_FRAG_NEEDED)
277 			return 0;
278 	case ICMP_REDIRECT:
279 		break;
280 	default:
281 		return 0;
282 	}
283 
284 	x = xfrm_state_lookup(net, skb->mark, (const xfrm_address_t *)&iph->daddr,
285 			      spi, protocol, AF_INET);
286 	if (!x)
287 		return 0;
288 
289 	if (icmp_hdr(skb)->type == ICMP_DEST_UNREACH)
290 		ipv4_update_pmtu(skb, net, info, 0, 0, protocol, 0);
291 	else
292 		ipv4_redirect(skb, net, 0, 0, protocol, 0);
293 	xfrm_state_put(x);
294 
295 	return 0;
296 }
297 
298 static int
299 vti_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
300 {
301 	int err = 0;
302 	struct ip_tunnel_parm p;
303 
304 	if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
305 		return -EFAULT;
306 
307 	if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
308 		if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPIP ||
309 		    p.iph.ihl != 5)
310 			return -EINVAL;
311 	}
312 
313 	p.i_flags |= VTI_ISVTI;
314 	err = ip_tunnel_ioctl(dev, &p, cmd);
315 	if (err)
316 		return err;
317 
318 	if (cmd != SIOCDELTUNNEL) {
319 		p.i_flags |= GRE_KEY;
320 		p.o_flags |= GRE_KEY;
321 	}
322 
323 	if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
324 		return -EFAULT;
325 	return 0;
326 }
327 
328 static const struct net_device_ops vti_netdev_ops = {
329 	.ndo_init	= vti_tunnel_init,
330 	.ndo_uninit	= ip_tunnel_uninit,
331 	.ndo_start_xmit	= vti_tunnel_xmit,
332 	.ndo_do_ioctl	= vti_tunnel_ioctl,
333 	.ndo_change_mtu	= ip_tunnel_change_mtu,
334 	.ndo_get_stats64 = ip_tunnel_get_stats64,
335 };
336 
337 static void vti_tunnel_setup(struct net_device *dev)
338 {
339 	dev->netdev_ops		= &vti_netdev_ops;
340 	dev->type		= ARPHRD_TUNNEL;
341 	ip_tunnel_setup(dev, vti_net_id);
342 }
343 
344 static int vti_tunnel_init(struct net_device *dev)
345 {
346 	struct ip_tunnel *tunnel = netdev_priv(dev);
347 	struct iphdr *iph = &tunnel->parms.iph;
348 
349 	memcpy(dev->dev_addr, &iph->saddr, 4);
350 	memcpy(dev->broadcast, &iph->daddr, 4);
351 
352 	dev->hard_header_len	= LL_MAX_HEADER + sizeof(struct iphdr);
353 	dev->mtu		= ETH_DATA_LEN;
354 	dev->flags		= IFF_NOARP;
355 	dev->iflink		= 0;
356 	dev->addr_len		= 4;
357 	dev->features		|= NETIF_F_LLTX;
358 	dev->priv_flags		&= ~IFF_XMIT_DST_RELEASE;
359 
360 	return ip_tunnel_init(dev);
361 }
362 
363 static void __net_init vti_fb_tunnel_init(struct net_device *dev)
364 {
365 	struct ip_tunnel *tunnel = netdev_priv(dev);
366 	struct iphdr *iph = &tunnel->parms.iph;
367 
368 	iph->version		= 4;
369 	iph->protocol		= IPPROTO_IPIP;
370 	iph->ihl		= 5;
371 }
372 
373 static struct xfrm4_protocol vti_esp4_protocol __read_mostly = {
374 	.handler	=	vti_rcv,
375 	.input_handler	=	vti_input,
376 	.cb_handler	=	vti_rcv_cb,
377 	.err_handler	=	vti4_err,
378 	.priority	=	100,
379 };
380 
381 static struct xfrm4_protocol vti_ah4_protocol __read_mostly = {
382 	.handler	=	vti_rcv,
383 	.input_handler	=	vti_input,
384 	.cb_handler	=	vti_rcv_cb,
385 	.err_handler	=	vti4_err,
386 	.priority	=	100,
387 };
388 
389 static struct xfrm4_protocol vti_ipcomp4_protocol __read_mostly = {
390 	.handler	=	vti_rcv,
391 	.input_handler	=	vti_input,
392 	.cb_handler	=	vti_rcv_cb,
393 	.err_handler	=	vti4_err,
394 	.priority	=	100,
395 };
396 
397 static int __net_init vti_init_net(struct net *net)
398 {
399 	int err;
400 	struct ip_tunnel_net *itn;
401 
402 	err = ip_tunnel_init_net(net, vti_net_id, &vti_link_ops, "ip_vti0");
403 	if (err)
404 		return err;
405 	itn = net_generic(net, vti_net_id);
406 	vti_fb_tunnel_init(itn->fb_tunnel_dev);
407 	return 0;
408 }
409 
410 static void __net_exit vti_exit_net(struct net *net)
411 {
412 	struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
413 	ip_tunnel_delete_net(itn, &vti_link_ops);
414 }
415 
416 static struct pernet_operations vti_net_ops = {
417 	.init = vti_init_net,
418 	.exit = vti_exit_net,
419 	.id   = &vti_net_id,
420 	.size = sizeof(struct ip_tunnel_net),
421 };
422 
423 static int vti_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
424 {
425 	return 0;
426 }
427 
428 static void vti_netlink_parms(struct nlattr *data[],
429 			      struct ip_tunnel_parm *parms)
430 {
431 	memset(parms, 0, sizeof(*parms));
432 
433 	parms->iph.protocol = IPPROTO_IPIP;
434 
435 	if (!data)
436 		return;
437 
438 	parms->i_flags = VTI_ISVTI;
439 
440 	if (data[IFLA_VTI_LINK])
441 		parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
442 
443 	if (data[IFLA_VTI_IKEY])
444 		parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
445 
446 	if (data[IFLA_VTI_OKEY])
447 		parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
448 
449 	if (data[IFLA_VTI_LOCAL])
450 		parms->iph.saddr = nla_get_be32(data[IFLA_VTI_LOCAL]);
451 
452 	if (data[IFLA_VTI_REMOTE])
453 		parms->iph.daddr = nla_get_be32(data[IFLA_VTI_REMOTE]);
454 
455 }
456 
457 static int vti_newlink(struct net *src_net, struct net_device *dev,
458 		       struct nlattr *tb[], struct nlattr *data[])
459 {
460 	struct ip_tunnel_parm parms;
461 
462 	vti_netlink_parms(data, &parms);
463 	return ip_tunnel_newlink(dev, tb, &parms);
464 }
465 
466 static int vti_changelink(struct net_device *dev, struct nlattr *tb[],
467 			  struct nlattr *data[])
468 {
469 	struct ip_tunnel_parm p;
470 
471 	vti_netlink_parms(data, &p);
472 	return ip_tunnel_changelink(dev, tb, &p);
473 }
474 
475 static size_t vti_get_size(const struct net_device *dev)
476 {
477 	return
478 		/* IFLA_VTI_LINK */
479 		nla_total_size(4) +
480 		/* IFLA_VTI_IKEY */
481 		nla_total_size(4) +
482 		/* IFLA_VTI_OKEY */
483 		nla_total_size(4) +
484 		/* IFLA_VTI_LOCAL */
485 		nla_total_size(4) +
486 		/* IFLA_VTI_REMOTE */
487 		nla_total_size(4) +
488 		0;
489 }
490 
491 static int vti_fill_info(struct sk_buff *skb, const struct net_device *dev)
492 {
493 	struct ip_tunnel *t = netdev_priv(dev);
494 	struct ip_tunnel_parm *p = &t->parms;
495 
496 	nla_put_u32(skb, IFLA_VTI_LINK, p->link);
497 	nla_put_be32(skb, IFLA_VTI_IKEY, p->i_key);
498 	nla_put_be32(skb, IFLA_VTI_OKEY, p->o_key);
499 	nla_put_be32(skb, IFLA_VTI_LOCAL, p->iph.saddr);
500 	nla_put_be32(skb, IFLA_VTI_REMOTE, p->iph.daddr);
501 
502 	return 0;
503 }
504 
505 static const struct nla_policy vti_policy[IFLA_VTI_MAX + 1] = {
506 	[IFLA_VTI_LINK]		= { .type = NLA_U32 },
507 	[IFLA_VTI_IKEY]		= { .type = NLA_U32 },
508 	[IFLA_VTI_OKEY]		= { .type = NLA_U32 },
509 	[IFLA_VTI_LOCAL]	= { .len = FIELD_SIZEOF(struct iphdr, saddr) },
510 	[IFLA_VTI_REMOTE]	= { .len = FIELD_SIZEOF(struct iphdr, daddr) },
511 };
512 
513 static struct rtnl_link_ops vti_link_ops __read_mostly = {
514 	.kind		= "vti",
515 	.maxtype	= IFLA_VTI_MAX,
516 	.policy		= vti_policy,
517 	.priv_size	= sizeof(struct ip_tunnel),
518 	.setup		= vti_tunnel_setup,
519 	.validate	= vti_tunnel_validate,
520 	.newlink	= vti_newlink,
521 	.changelink	= vti_changelink,
522 	.get_size	= vti_get_size,
523 	.fill_info	= vti_fill_info,
524 };
525 
526 static int __init vti_init(void)
527 {
528 	int err;
529 
530 	pr_info("IPv4 over IPSec tunneling driver\n");
531 
532 	err = register_pernet_device(&vti_net_ops);
533 	if (err < 0)
534 		return err;
535 	err = xfrm4_protocol_register(&vti_esp4_protocol, IPPROTO_ESP);
536 	if (err < 0) {
537 		unregister_pernet_device(&vti_net_ops);
538 		pr_info("vti init: can't register tunnel\n");
539 
540 		return err;
541 	}
542 
543 	err = xfrm4_protocol_register(&vti_ah4_protocol, IPPROTO_AH);
544 	if (err < 0) {
545 		xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
546 		unregister_pernet_device(&vti_net_ops);
547 		pr_info("vti init: can't register tunnel\n");
548 
549 		return err;
550 	}
551 
552 	err = xfrm4_protocol_register(&vti_ipcomp4_protocol, IPPROTO_COMP);
553 	if (err < 0) {
554 		xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
555 		xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
556 		unregister_pernet_device(&vti_net_ops);
557 		pr_info("vti init: can't register tunnel\n");
558 
559 		return err;
560 	}
561 
562 	err = rtnl_link_register(&vti_link_ops);
563 	if (err < 0)
564 		goto rtnl_link_failed;
565 
566 	return err;
567 
568 rtnl_link_failed:
569 	xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP);
570 	xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
571 	xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
572 	unregister_pernet_device(&vti_net_ops);
573 	return err;
574 }
575 
576 static void __exit vti_fini(void)
577 {
578 	rtnl_link_unregister(&vti_link_ops);
579 	if (xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP))
580 		pr_info("vti close: can't deregister tunnel\n");
581 	if (xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH))
582 		pr_info("vti close: can't deregister tunnel\n");
583 	if (xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP))
584 		pr_info("vti close: can't deregister tunnel\n");
585 
586 
587 	unregister_pernet_device(&vti_net_ops);
588 }
589 
590 module_init(vti_init);
591 module_exit(vti_fini);
592 MODULE_LICENSE("GPL");
593 MODULE_ALIAS_RTNL_LINK("vti");
594 MODULE_ALIAS_NETDEV("ip_vti0");
595