xref: /openbmc/linux/net/ipv4/ipip.c (revision 78700c0a)
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 <asm/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 
119 static bool log_ecn_error = true;
120 module_param(log_ecn_error, bool, 0644);
121 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
122 
123 static int ipip_net_id __read_mostly;
124 
125 static int ipip_tunnel_init(struct net_device *dev);
126 static struct rtnl_link_ops ipip_link_ops __read_mostly;
127 
128 static int ipip_err(struct sk_buff *skb, u32 info)
129 {
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 	struct ip_tunnel *t;
139 	int err;
140 	const int type = icmp_hdr(skb)->type;
141 	const int code = icmp_hdr(skb)->code;
142 
143 	err = -ENOENT;
144 	t = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
145 			     iph->daddr, iph->saddr, 0);
146 	if (!t)
147 		goto out;
148 
149 	if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
150 		ipv4_update_pmtu(skb, dev_net(skb->dev), info,
151 				 t->parms.link, 0, IPPROTO_IPIP, 0);
152 		err = 0;
153 		goto out;
154 	}
155 
156 	if (type == ICMP_REDIRECT) {
157 		ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 0,
158 			      IPPROTO_IPIP, 0);
159 		err = 0;
160 		goto out;
161 	}
162 
163 	if (t->parms.iph.daddr == 0)
164 		goto out;
165 
166 	err = 0;
167 	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
168 		goto out;
169 
170 	if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
171 		t->err_count++;
172 	else
173 		t->err_count = 1;
174 	t->err_time = jiffies;
175 
176 out:
177 	return err;
178 }
179 
180 static const struct tnl_ptk_info tpi = {
181 	/* no tunnel info required for ipip. */
182 	.proto = htons(ETH_P_IP),
183 };
184 
185 static int ipip_rcv(struct sk_buff *skb)
186 {
187 	struct net *net = dev_net(skb->dev);
188 	struct ip_tunnel_net *itn = net_generic(net, ipip_net_id);
189 	struct ip_tunnel *tunnel;
190 	const struct iphdr *iph;
191 
192 	iph = ip_hdr(skb);
193 	tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
194 			iph->saddr, iph->daddr, 0);
195 	if (tunnel) {
196 		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
197 			goto drop;
198 		if (iptunnel_pull_header(skb, 0, tpi.proto, false))
199 			goto drop;
200 		return ip_tunnel_rcv(tunnel, skb, &tpi, NULL, log_ecn_error);
201 	}
202 
203 	return -1;
204 
205 drop:
206 	kfree_skb(skb);
207 	return 0;
208 }
209 
210 /*
211  *	This function assumes it is being called from dev_queue_xmit()
212  *	and that skb is filled properly by that function.
213  */
214 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
215 {
216 	struct ip_tunnel *tunnel = netdev_priv(dev);
217 	const struct iphdr  *tiph = &tunnel->parms.iph;
218 
219 	if (unlikely(skb->protocol != htons(ETH_P_IP)))
220 		goto tx_error;
221 
222 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
223 		goto tx_error;
224 
225 	skb_set_inner_ipproto(skb, IPPROTO_IPIP);
226 
227 	ip_tunnel_xmit(skb, dev, tiph, tiph->protocol);
228 	return NETDEV_TX_OK;
229 
230 tx_error:
231 	kfree_skb(skb);
232 
233 	dev->stats.tx_errors++;
234 	return NETDEV_TX_OK;
235 }
236 
237 static int
238 ipip_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
239 {
240 	int err = 0;
241 	struct ip_tunnel_parm p;
242 
243 	if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
244 		return -EFAULT;
245 
246 	if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
247 		if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPIP ||
248 		    p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
249 			return -EINVAL;
250 	}
251 
252 	p.i_key = p.o_key = 0;
253 	p.i_flags = p.o_flags = 0;
254 	err = ip_tunnel_ioctl(dev, &p, cmd);
255 	if (err)
256 		return err;
257 
258 	if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
259 		return -EFAULT;
260 
261 	return 0;
262 }
263 
264 static const struct net_device_ops ipip_netdev_ops = {
265 	.ndo_init       = ipip_tunnel_init,
266 	.ndo_uninit     = ip_tunnel_uninit,
267 	.ndo_start_xmit	= ipip_tunnel_xmit,
268 	.ndo_do_ioctl	= ipip_tunnel_ioctl,
269 	.ndo_change_mtu = ip_tunnel_change_mtu,
270 	.ndo_get_stats64 = ip_tunnel_get_stats64,
271 	.ndo_get_iflink = ip_tunnel_get_iflink,
272 };
273 
274 #define IPIP_FEATURES (NETIF_F_SG |		\
275 		       NETIF_F_FRAGLIST |	\
276 		       NETIF_F_HIGHDMA |	\
277 		       NETIF_F_GSO_SOFTWARE |	\
278 		       NETIF_F_HW_CSUM)
279 
280 static void ipip_tunnel_setup(struct net_device *dev)
281 {
282 	dev->netdev_ops		= &ipip_netdev_ops;
283 
284 	dev->type		= ARPHRD_TUNNEL;
285 	dev->flags		= IFF_NOARP;
286 	dev->addr_len		= 4;
287 	dev->features		|= NETIF_F_LLTX;
288 	netif_keep_dst(dev);
289 
290 	dev->features		|= IPIP_FEATURES;
291 	dev->hw_features	|= IPIP_FEATURES;
292 	ip_tunnel_setup(dev, ipip_net_id);
293 }
294 
295 static int ipip_tunnel_init(struct net_device *dev)
296 {
297 	struct ip_tunnel *tunnel = netdev_priv(dev);
298 
299 	memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
300 	memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
301 
302 	tunnel->tun_hlen = 0;
303 	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
304 	tunnel->parms.iph.protocol = IPPROTO_IPIP;
305 	return ip_tunnel_init(dev);
306 }
307 
308 static void ipip_netlink_parms(struct nlattr *data[],
309 			       struct ip_tunnel_parm *parms)
310 {
311 	memset(parms, 0, sizeof(*parms));
312 
313 	parms->iph.version = 4;
314 	parms->iph.protocol = IPPROTO_IPIP;
315 	parms->iph.ihl = 5;
316 
317 	if (!data)
318 		return;
319 
320 	if (data[IFLA_IPTUN_LINK])
321 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
322 
323 	if (data[IFLA_IPTUN_LOCAL])
324 		parms->iph.saddr = nla_get_in_addr(data[IFLA_IPTUN_LOCAL]);
325 
326 	if (data[IFLA_IPTUN_REMOTE])
327 		parms->iph.daddr = nla_get_in_addr(data[IFLA_IPTUN_REMOTE]);
328 
329 	if (data[IFLA_IPTUN_TTL]) {
330 		parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
331 		if (parms->iph.ttl)
332 			parms->iph.frag_off = htons(IP_DF);
333 	}
334 
335 	if (data[IFLA_IPTUN_TOS])
336 		parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
337 
338 	if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
339 		parms->iph.frag_off = htons(IP_DF);
340 }
341 
342 /* This function returns true when ENCAP attributes are present in the nl msg */
343 static bool ipip_netlink_encap_parms(struct nlattr *data[],
344 				     struct ip_tunnel_encap *ipencap)
345 {
346 	bool ret = false;
347 
348 	memset(ipencap, 0, sizeof(*ipencap));
349 
350 	if (!data)
351 		return ret;
352 
353 	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
354 		ret = true;
355 		ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
356 	}
357 
358 	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
359 		ret = true;
360 		ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
361 	}
362 
363 	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
364 		ret = true;
365 		ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
366 	}
367 
368 	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
369 		ret = true;
370 		ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
371 	}
372 
373 	return ret;
374 }
375 
376 static int ipip_newlink(struct net *src_net, struct net_device *dev,
377 			struct nlattr *tb[], struct nlattr *data[])
378 {
379 	struct ip_tunnel_parm p;
380 	struct ip_tunnel_encap ipencap;
381 
382 	if (ipip_netlink_encap_parms(data, &ipencap)) {
383 		struct ip_tunnel *t = netdev_priv(dev);
384 		int err = ip_tunnel_encap_setup(t, &ipencap);
385 
386 		if (err < 0)
387 			return err;
388 	}
389 
390 	ipip_netlink_parms(data, &p);
391 	return ip_tunnel_newlink(dev, tb, &p);
392 }
393 
394 static int ipip_changelink(struct net_device *dev, struct nlattr *tb[],
395 			   struct nlattr *data[])
396 {
397 	struct ip_tunnel_parm p;
398 	struct ip_tunnel_encap ipencap;
399 
400 	if (ipip_netlink_encap_parms(data, &ipencap)) {
401 		struct ip_tunnel *t = netdev_priv(dev);
402 		int err = ip_tunnel_encap_setup(t, &ipencap);
403 
404 		if (err < 0)
405 			return err;
406 	}
407 
408 	ipip_netlink_parms(data, &p);
409 
410 	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
411 	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
412 		return -EINVAL;
413 
414 	return ip_tunnel_changelink(dev, tb, &p);
415 }
416 
417 static size_t ipip_get_size(const struct net_device *dev)
418 {
419 	return
420 		/* IFLA_IPTUN_LINK */
421 		nla_total_size(4) +
422 		/* IFLA_IPTUN_LOCAL */
423 		nla_total_size(4) +
424 		/* IFLA_IPTUN_REMOTE */
425 		nla_total_size(4) +
426 		/* IFLA_IPTUN_TTL */
427 		nla_total_size(1) +
428 		/* IFLA_IPTUN_TOS */
429 		nla_total_size(1) +
430 		/* IFLA_IPTUN_PMTUDISC */
431 		nla_total_size(1) +
432 		/* IFLA_IPTUN_ENCAP_TYPE */
433 		nla_total_size(2) +
434 		/* IFLA_IPTUN_ENCAP_FLAGS */
435 		nla_total_size(2) +
436 		/* IFLA_IPTUN_ENCAP_SPORT */
437 		nla_total_size(2) +
438 		/* IFLA_IPTUN_ENCAP_DPORT */
439 		nla_total_size(2) +
440 		0;
441 }
442 
443 static int ipip_fill_info(struct sk_buff *skb, const struct net_device *dev)
444 {
445 	struct ip_tunnel *tunnel = netdev_priv(dev);
446 	struct ip_tunnel_parm *parm = &tunnel->parms;
447 
448 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
449 	    nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
450 	    nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
451 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
452 	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
453 	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
454 		       !!(parm->iph.frag_off & htons(IP_DF))))
455 		goto nla_put_failure;
456 
457 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
458 			tunnel->encap.type) ||
459 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
460 			 tunnel->encap.sport) ||
461 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
462 			 tunnel->encap.dport) ||
463 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
464 			tunnel->encap.flags))
465 		goto nla_put_failure;
466 
467 	return 0;
468 
469 nla_put_failure:
470 	return -EMSGSIZE;
471 }
472 
473 static const struct nla_policy ipip_policy[IFLA_IPTUN_MAX + 1] = {
474 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
475 	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
476 	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
477 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
478 	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
479 	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
480 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
481 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
482 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
483 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
484 };
485 
486 static struct rtnl_link_ops ipip_link_ops __read_mostly = {
487 	.kind		= "ipip",
488 	.maxtype	= IFLA_IPTUN_MAX,
489 	.policy		= ipip_policy,
490 	.priv_size	= sizeof(struct ip_tunnel),
491 	.setup		= ipip_tunnel_setup,
492 	.newlink	= ipip_newlink,
493 	.changelink	= ipip_changelink,
494 	.dellink	= ip_tunnel_dellink,
495 	.get_size	= ipip_get_size,
496 	.fill_info	= ipip_fill_info,
497 	.get_link_net	= ip_tunnel_get_link_net,
498 };
499 
500 static struct xfrm_tunnel ipip_handler __read_mostly = {
501 	.handler	=	ipip_rcv,
502 	.err_handler	=	ipip_err,
503 	.priority	=	1,
504 };
505 
506 static int __net_init ipip_init_net(struct net *net)
507 {
508 	return ip_tunnel_init_net(net, ipip_net_id, &ipip_link_ops, "tunl0");
509 }
510 
511 static void __net_exit ipip_exit_net(struct net *net)
512 {
513 	struct ip_tunnel_net *itn = net_generic(net, ipip_net_id);
514 	ip_tunnel_delete_net(itn, &ipip_link_ops);
515 }
516 
517 static struct pernet_operations ipip_net_ops = {
518 	.init = ipip_init_net,
519 	.exit = ipip_exit_net,
520 	.id   = &ipip_net_id,
521 	.size = sizeof(struct ip_tunnel_net),
522 };
523 
524 static int __init ipip_init(void)
525 {
526 	int err;
527 
528 	pr_info("ipip: IPv4 over IPv4 tunneling driver\n");
529 
530 	err = register_pernet_device(&ipip_net_ops);
531 	if (err < 0)
532 		return err;
533 	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
534 	if (err < 0) {
535 		pr_info("%s: can't register tunnel\n", __func__);
536 		goto xfrm_tunnel_failed;
537 	}
538 	err = rtnl_link_register(&ipip_link_ops);
539 	if (err < 0)
540 		goto rtnl_link_failed;
541 
542 out:
543 	return err;
544 
545 rtnl_link_failed:
546 	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
547 xfrm_tunnel_failed:
548 	unregister_pernet_device(&ipip_net_ops);
549 	goto out;
550 }
551 
552 static void __exit ipip_fini(void)
553 {
554 	rtnl_link_unregister(&ipip_link_ops);
555 	if (xfrm4_tunnel_deregister(&ipip_handler, AF_INET))
556 		pr_info("%s: can't deregister tunnel\n", __func__);
557 
558 	unregister_pernet_device(&ipip_net_ops);
559 }
560 
561 module_init(ipip_init);
562 module_exit(ipip_fini);
563 MODULE_LICENSE("GPL");
564 MODULE_ALIAS_RTNL_LINK("ipip");
565 MODULE_ALIAS_NETDEV("tunl0");
566