xref: /openbmc/linux/net/ipv4/ipip.c (revision 609e478b)
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/mroute.h>
107 #include <linux/init.h>
108 #include <linux/netfilter_ipv4.h>
109 #include <linux/if_ether.h>
110 
111 #include <net/sock.h>
112 #include <net/ip.h>
113 #include <net/icmp.h>
114 #include <net/ip_tunnels.h>
115 #include <net/inet_ecn.h>
116 #include <net/xfrm.h>
117 #include <net/net_namespace.h>
118 #include <net/netns/generic.h>
119 
120 static bool log_ecn_error = true;
121 module_param(log_ecn_error, bool, 0644);
122 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
123 
124 static int ipip_net_id __read_mostly;
125 
126 static int ipip_tunnel_init(struct net_device *dev);
127 static struct rtnl_link_ops ipip_link_ops __read_mostly;
128 
129 static int ipip_err(struct sk_buff *skb, u32 info)
130 {
131 
132 /* All the routers (except for Linux) return only
133    8 bytes of packet payload. It means, that precise relaying of
134    ICMP in the real Internet is absolutely infeasible.
135  */
136 	struct net *net = dev_net(skb->dev);
137 	struct ip_tunnel_net *itn = net_generic(net, ipip_net_id);
138 	const struct iphdr *iph = (const struct iphdr *)skb->data;
139 	struct ip_tunnel *t;
140 	int err;
141 	const int type = icmp_hdr(skb)->type;
142 	const int code = icmp_hdr(skb)->code;
143 
144 	err = -ENOENT;
145 	t = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
146 			     iph->daddr, iph->saddr, 0);
147 	if (t == NULL)
148 		goto out;
149 
150 	if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
151 		ipv4_update_pmtu(skb, dev_net(skb->dev), info,
152 				 t->parms.link, 0, IPPROTO_IPIP, 0);
153 		err = 0;
154 		goto out;
155 	}
156 
157 	if (type == ICMP_REDIRECT) {
158 		ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 0,
159 			      IPPROTO_IPIP, 0);
160 		err = 0;
161 		goto out;
162 	}
163 
164 	if (t->parms.iph.daddr == 0)
165 		goto out;
166 
167 	err = 0;
168 	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
169 		goto out;
170 
171 	if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
172 		t->err_count++;
173 	else
174 		t->err_count = 1;
175 	t->err_time = jiffies;
176 
177 out:
178 	return err;
179 }
180 
181 static const struct tnl_ptk_info tpi = {
182 	/* no tunnel info required for ipip. */
183 	.proto = htons(ETH_P_IP),
184 };
185 
186 static int ipip_rcv(struct sk_buff *skb)
187 {
188 	struct net *net = dev_net(skb->dev);
189 	struct ip_tunnel_net *itn = net_generic(net, ipip_net_id);
190 	struct ip_tunnel *tunnel;
191 	const struct iphdr *iph;
192 
193 	iph = ip_hdr(skb);
194 	tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
195 			iph->saddr, iph->daddr, 0);
196 	if (tunnel) {
197 		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
198 			goto drop;
199 		if (iptunnel_pull_header(skb, 0, tpi.proto))
200 			goto drop;
201 		return ip_tunnel_rcv(tunnel, skb, &tpi, log_ecn_error);
202 	}
203 
204 	return -1;
205 
206 drop:
207 	kfree_skb(skb);
208 	return 0;
209 }
210 
211 /*
212  *	This function assumes it is being called from dev_queue_xmit()
213  *	and that skb is filled properly by that function.
214  */
215 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
216 {
217 	struct ip_tunnel *tunnel = netdev_priv(dev);
218 	const struct iphdr  *tiph = &tunnel->parms.iph;
219 
220 	if (unlikely(skb->protocol != htons(ETH_P_IP)))
221 		goto tx_error;
222 
223 	skb = iptunnel_handle_offloads(skb, false, SKB_GSO_IPIP);
224 	if (IS_ERR(skb))
225 		goto out;
226 
227 	skb_set_inner_ipproto(skb, IPPROTO_IPIP);
228 
229 	ip_tunnel_xmit(skb, dev, tiph, tiph->protocol);
230 	return NETDEV_TX_OK;
231 
232 tx_error:
233 	kfree_skb(skb);
234 out:
235 	dev->stats.tx_errors++;
236 	return NETDEV_TX_OK;
237 }
238 
239 static int
240 ipip_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
241 {
242 	int err = 0;
243 	struct ip_tunnel_parm p;
244 
245 	if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
246 		return -EFAULT;
247 
248 	if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
249 		if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPIP ||
250 		    p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
251 			return -EINVAL;
252 	}
253 
254 	p.i_key = p.o_key = p.i_flags = p.o_flags = 0;
255 	if (p.iph.ttl)
256 		p.iph.frag_off |= htons(IP_DF);
257 
258 	err = ip_tunnel_ioctl(dev, &p, cmd);
259 	if (err)
260 		return err;
261 
262 	if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
263 		return -EFAULT;
264 
265 	return 0;
266 }
267 
268 static const struct net_device_ops ipip_netdev_ops = {
269 	.ndo_init       = ipip_tunnel_init,
270 	.ndo_uninit     = ip_tunnel_uninit,
271 	.ndo_start_xmit	= ipip_tunnel_xmit,
272 	.ndo_do_ioctl	= ipip_tunnel_ioctl,
273 	.ndo_change_mtu = ip_tunnel_change_mtu,
274 	.ndo_get_stats64 = ip_tunnel_get_stats64,
275 };
276 
277 #define IPIP_FEATURES (NETIF_F_SG |		\
278 		       NETIF_F_FRAGLIST |	\
279 		       NETIF_F_HIGHDMA |	\
280 		       NETIF_F_GSO_SOFTWARE |	\
281 		       NETIF_F_HW_CSUM)
282 
283 static void ipip_tunnel_setup(struct net_device *dev)
284 {
285 	dev->netdev_ops		= &ipip_netdev_ops;
286 
287 	dev->type		= ARPHRD_TUNNEL;
288 	dev->flags		= IFF_NOARP;
289 	dev->iflink		= 0;
290 	dev->addr_len		= 4;
291 	dev->features		|= NETIF_F_LLTX;
292 	netif_keep_dst(dev);
293 
294 	dev->features		|= IPIP_FEATURES;
295 	dev->hw_features	|= IPIP_FEATURES;
296 	ip_tunnel_setup(dev, ipip_net_id);
297 }
298 
299 static int ipip_tunnel_init(struct net_device *dev)
300 {
301 	struct ip_tunnel *tunnel = netdev_priv(dev);
302 
303 	memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
304 	memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
305 
306 	tunnel->tun_hlen = 0;
307 	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
308 	tunnel->parms.iph.protocol = IPPROTO_IPIP;
309 	return ip_tunnel_init(dev);
310 }
311 
312 static void ipip_netlink_parms(struct nlattr *data[],
313 			       struct ip_tunnel_parm *parms)
314 {
315 	memset(parms, 0, sizeof(*parms));
316 
317 	parms->iph.version = 4;
318 	parms->iph.protocol = IPPROTO_IPIP;
319 	parms->iph.ihl = 5;
320 
321 	if (!data)
322 		return;
323 
324 	if (data[IFLA_IPTUN_LINK])
325 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
326 
327 	if (data[IFLA_IPTUN_LOCAL])
328 		parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
329 
330 	if (data[IFLA_IPTUN_REMOTE])
331 		parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
332 
333 	if (data[IFLA_IPTUN_TTL]) {
334 		parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
335 		if (parms->iph.ttl)
336 			parms->iph.frag_off = htons(IP_DF);
337 	}
338 
339 	if (data[IFLA_IPTUN_TOS])
340 		parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
341 
342 	if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
343 		parms->iph.frag_off = htons(IP_DF);
344 }
345 
346 /* This function returns true when ENCAP attributes are present in the nl msg */
347 static bool ipip_netlink_encap_parms(struct nlattr *data[],
348 				     struct ip_tunnel_encap *ipencap)
349 {
350 	bool ret = false;
351 
352 	memset(ipencap, 0, sizeof(*ipencap));
353 
354 	if (!data)
355 		return ret;
356 
357 	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
358 		ret = true;
359 		ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
360 	}
361 
362 	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
363 		ret = true;
364 		ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
365 	}
366 
367 	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
368 		ret = true;
369 		ipencap->sport = nla_get_u16(data[IFLA_IPTUN_ENCAP_SPORT]);
370 	}
371 
372 	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
373 		ret = true;
374 		ipencap->dport = nla_get_u16(data[IFLA_IPTUN_ENCAP_DPORT]);
375 	}
376 
377 	return ret;
378 }
379 
380 static int ipip_newlink(struct net *src_net, struct net_device *dev,
381 			struct nlattr *tb[], struct nlattr *data[])
382 {
383 	struct ip_tunnel_parm p;
384 	struct ip_tunnel_encap ipencap;
385 
386 	if (ipip_netlink_encap_parms(data, &ipencap)) {
387 		struct ip_tunnel *t = netdev_priv(dev);
388 		int err = ip_tunnel_encap_setup(t, &ipencap);
389 
390 		if (err < 0)
391 			return err;
392 	}
393 
394 	ipip_netlink_parms(data, &p);
395 	return ip_tunnel_newlink(dev, tb, &p);
396 }
397 
398 static int ipip_changelink(struct net_device *dev, struct nlattr *tb[],
399 			   struct nlattr *data[])
400 {
401 	struct ip_tunnel_parm p;
402 	struct ip_tunnel_encap ipencap;
403 
404 	if (ipip_netlink_encap_parms(data, &ipencap)) {
405 		struct ip_tunnel *t = netdev_priv(dev);
406 		int err = ip_tunnel_encap_setup(t, &ipencap);
407 
408 		if (err < 0)
409 			return err;
410 	}
411 
412 	ipip_netlink_parms(data, &p);
413 
414 	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
415 	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
416 		return -EINVAL;
417 
418 	return ip_tunnel_changelink(dev, tb, &p);
419 }
420 
421 static size_t ipip_get_size(const struct net_device *dev)
422 {
423 	return
424 		/* IFLA_IPTUN_LINK */
425 		nla_total_size(4) +
426 		/* IFLA_IPTUN_LOCAL */
427 		nla_total_size(4) +
428 		/* IFLA_IPTUN_REMOTE */
429 		nla_total_size(4) +
430 		/* IFLA_IPTUN_TTL */
431 		nla_total_size(1) +
432 		/* IFLA_IPTUN_TOS */
433 		nla_total_size(1) +
434 		/* IFLA_IPTUN_PMTUDISC */
435 		nla_total_size(1) +
436 		/* IFLA_IPTUN_ENCAP_TYPE */
437 		nla_total_size(2) +
438 		/* IFLA_IPTUN_ENCAP_FLAGS */
439 		nla_total_size(2) +
440 		/* IFLA_IPTUN_ENCAP_SPORT */
441 		nla_total_size(2) +
442 		/* IFLA_IPTUN_ENCAP_DPORT */
443 		nla_total_size(2) +
444 		0;
445 }
446 
447 static int ipip_fill_info(struct sk_buff *skb, const struct net_device *dev)
448 {
449 	struct ip_tunnel *tunnel = netdev_priv(dev);
450 	struct ip_tunnel_parm *parm = &tunnel->parms;
451 
452 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
453 	    nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
454 	    nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
455 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
456 	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
457 	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
458 		       !!(parm->iph.frag_off & htons(IP_DF))))
459 		goto nla_put_failure;
460 
461 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
462 			tunnel->encap.type) ||
463 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_SPORT,
464 			tunnel->encap.sport) ||
465 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_DPORT,
466 			tunnel->encap.dport) ||
467 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
468 			tunnel->encap.dport))
469 		goto nla_put_failure;
470 
471 	return 0;
472 
473 nla_put_failure:
474 	return -EMSGSIZE;
475 }
476 
477 static const struct nla_policy ipip_policy[IFLA_IPTUN_MAX + 1] = {
478 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
479 	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
480 	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
481 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
482 	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
483 	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
484 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
485 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
486 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
487 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
488 };
489 
490 static struct rtnl_link_ops ipip_link_ops __read_mostly = {
491 	.kind		= "ipip",
492 	.maxtype	= IFLA_IPTUN_MAX,
493 	.policy		= ipip_policy,
494 	.priv_size	= sizeof(struct ip_tunnel),
495 	.setup		= ipip_tunnel_setup,
496 	.newlink	= ipip_newlink,
497 	.changelink	= ipip_changelink,
498 	.dellink	= ip_tunnel_dellink,
499 	.get_size	= ipip_get_size,
500 	.fill_info	= ipip_fill_info,
501 };
502 
503 static struct xfrm_tunnel ipip_handler __read_mostly = {
504 	.handler	=	ipip_rcv,
505 	.err_handler	=	ipip_err,
506 	.priority	=	1,
507 };
508 
509 static int __net_init ipip_init_net(struct net *net)
510 {
511 	return ip_tunnel_init_net(net, ipip_net_id, &ipip_link_ops, "tunl0");
512 }
513 
514 static void __net_exit ipip_exit_net(struct net *net)
515 {
516 	struct ip_tunnel_net *itn = net_generic(net, ipip_net_id);
517 	ip_tunnel_delete_net(itn, &ipip_link_ops);
518 }
519 
520 static struct pernet_operations ipip_net_ops = {
521 	.init = ipip_init_net,
522 	.exit = ipip_exit_net,
523 	.id   = &ipip_net_id,
524 	.size = sizeof(struct ip_tunnel_net),
525 };
526 
527 static int __init ipip_init(void)
528 {
529 	int err;
530 
531 	pr_info("ipip: IPv4 over IPv4 tunneling driver\n");
532 
533 	err = register_pernet_device(&ipip_net_ops);
534 	if (err < 0)
535 		return err;
536 	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
537 	if (err < 0) {
538 		pr_info("%s: can't register tunnel\n", __func__);
539 		goto xfrm_tunnel_failed;
540 	}
541 	err = rtnl_link_register(&ipip_link_ops);
542 	if (err < 0)
543 		goto rtnl_link_failed;
544 
545 out:
546 	return err;
547 
548 rtnl_link_failed:
549 	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
550 xfrm_tunnel_failed:
551 	unregister_pernet_device(&ipip_net_ops);
552 	goto out;
553 }
554 
555 static void __exit ipip_fini(void)
556 {
557 	rtnl_link_unregister(&ipip_link_ops);
558 	if (xfrm4_tunnel_deregister(&ipip_handler, AF_INET))
559 		pr_info("%s: can't deregister tunnel\n", __func__);
560 
561 	unregister_pernet_device(&ipip_net_ops);
562 }
563 
564 module_init(ipip_init);
565 module_exit(ipip_fini);
566 MODULE_LICENSE("GPL");
567 MODULE_ALIAS_RTNL_LINK("ipip");
568 MODULE_ALIAS_NETDEV("tunl0");
569