xref: /netbsd-src/sys/netinet/ip_encap.c (revision 481fca6e59249d8ffcf24fef7cfbe7b131bfb080)
1 /*	$NetBSD: ip_encap.c,v 1.3 2000/07/05 21:32:51 thorpej Exp $	*/
2 /*	$KAME: ip_encap.c,v 1.30 2000/04/19 04:29:37 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 /*
33  * My grandfather said that there's a devil inside tunnelling technology...
34  *
35  * We have surprisingly many protocols that want packets with IP protocol
36  * #4 or #41.  Here's a list of protocols that want protocol #41:
37  *	RFC1933 configured tunnel
38  *	RFC1933 automatic tunnel
39  *	RFC2401 IPsec tunnel
40  *	RFC2473 IPv6 generic packet tunnelling
41  *	RFC2529 6over4 tunnel
42  *	mobile-ip6 (uses RFC2473)
43  *	6to4 tunnel
44  * Here's a list of protocol that want protocol #4:
45  *	RFC1853 IPv4-in-IPv4 tunnelling
46  *	RFC2003 IPv4 encapsulation within IPv4
47  *	RFC2344 reverse tunnelling for mobile-ip4
48  *	RFC2401 IPsec tunnel
49  * Well, what can I say.  They impose different en/decapsulation mechanism
50  * from each other, so they need separate protocol handler.  The only one
51  * we can easily determine by protocol # is IPsec, which always has
52  * AH/ESP/IPComp header right after outer IP header.
53  *
54  * So, clearly good old protosw does not work for protocol #4 and #41.
55  * The code will let you match protocol via src/dst address pair.
56  */
57 /* XXX is M_NETADDR correct? */
58 
59 #ifdef __FreeBSD__
60 # include "opt_mrouting.h"
61 # if __FreeBSD__ == 3
62 #  include "opt_inet.h"
63 # endif
64 # if __FreeBSD__ >= 4
65 #  include "opt_inet.h"
66 #  include "opt_inet6.h"
67 # endif
68 #else
69 # ifdef __NetBSD__
70 #  include "opt_mrouting.h"
71 #  include "opt_inet.h"
72 # endif
73 #endif
74 
75 #include <sys/param.h>
76 #include <sys/systm.h>
77 #include <sys/socket.h>
78 #include <sys/sockio.h>
79 #include <sys/mbuf.h>
80 #include <sys/errno.h>
81 #include <sys/protosw.h>
82 
83 #include <net/if.h>
84 #include <net/route.h>
85 
86 #include <netinet/in.h>
87 #include <netinet/in_systm.h>
88 #include <netinet/ip.h>
89 #include <netinet/ip_var.h>
90 #include <netinet/ip_encap.h>
91 #ifdef MROUTING
92 #include <netinet/ip_mroute.h>
93 #endif /* MROUTING */
94 #ifdef __OpenBSD__
95 #include <netinet/ip_ipsp.h>
96 #endif
97 
98 #ifdef INET6
99 #include <netinet/ip6.h>
100 #include <netinet6/ip6_var.h>
101 #include <netinet6/ip6protosw.h>
102 #endif
103 
104 #include <machine/stdarg.h>
105 
106 #ifdef __NetBSD__
107 # include "ipip.h"
108 # if NIPIP > 0
109 #  include <netinet/ip_ipip.h>
110 # else
111 #  ifdef MROUTING
112 #   include <netinet/ip_mroute.h>
113 #  endif
114 # endif
115 #endif
116 
117 #include <net/net_osdep.h>
118 
119 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
120 #include <sys/kernel.h>
121 #include <sys/malloc.h>
122 MALLOC_DEFINE(M_NETADDR, "Export Host", "Export host address structure");
123 #endif
124 
125 static void encap_add __P((struct encaptab *));
126 static int mask_match __P((const struct encaptab *, const struct sockaddr *,
127 		const struct sockaddr *));
128 static void encap_fillarg __P((struct mbuf *, const struct encaptab *));
129 
130 /* rely upon BSS initialization */
131 LIST_HEAD(, encaptab) encaptab = LIST_HEAD_INITIALIZER(&encaptab);
132 
133 void
134 encap_init()
135 {
136 #if 0
137 	/*
138 	 * we cannot use LIST_INIT() here, since drivers may want to call
139 	 * encap_attach(), on driver attach.  encap_init() wlil be called
140 	 * on AF_INET{,6} initialization, which happens after driver
141 	 * initialization - using LIST_INIT() here can nuke encap_attach()
142 	 * from drivers.
143 	 */
144 	LIST_INIT(&encaptab);
145 #endif
146 }
147 
148 #if !(defined(__FreeBSD__) && __FreeBSD__ >= 4)
149 void
150 #if __STDC__
151 encap4_input(struct mbuf *m, ...)
152 #else
153 encap4_input(m, va_alist)
154 	struct mbuf *m;
155 	va_dcl
156 #endif
157 {
158 	int off, proto;
159 	struct ip *ip;
160 	struct sockaddr_in s, d;
161 	struct encaptab *ep, *match;
162 	va_list ap;
163 	int prio, matchprio;
164 
165 	va_start(ap, m);
166 	off = va_arg(ap, int);
167 #ifndef __OpenBSD__
168 	proto = va_arg(ap, int);
169 #endif
170 	va_end(ap);
171 
172 	ip = mtod(m, struct ip *);
173 #ifdef __OpenBSD__
174 	proto = ip->ip_p;
175 #endif
176 
177 	bzero(&s, sizeof(s));
178 	s.sin_family = AF_INET;
179 	s.sin_len = sizeof(struct sockaddr_in);
180 	s.sin_addr = ip->ip_src;
181 	bzero(&d, sizeof(d));
182 	d.sin_family = AF_INET;
183 	d.sin_len = sizeof(struct sockaddr_in);
184 	d.sin_addr = ip->ip_dst;
185 
186 	match = NULL;
187 	matchprio = 0;
188 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
189 		if (ep->af != AF_INET)
190 			continue;
191 		if (ep->proto >= 0 && ep->proto != proto)
192 			continue;
193 		if (ep->func)
194 			prio = (*ep->func)(m, off, proto, ep->arg);
195 		else {
196 			/*
197 			 * it's inbound traffic, we need to match in reverse
198 			 * order
199 			 */
200 			prio = mask_match(ep, (struct sockaddr *)&d,
201 			    (struct sockaddr *)&s);
202 		}
203 
204 		/*
205 		 * We prioritize the matches by using bit length of the
206 		 * matches.  mask_match() and user-supplied matching function
207 		 * should return the bit length of the matches (for example,
208 		 * if both src/dst are matched for IPv4, 64 should be returned).
209 		 * 0 or negative return value means "it did not match".
210 		 *
211 		 * The question is, since we have two "mask" portion, we
212 		 * cannot really define total order between entries.
213 		 * For example, which of these should be preferred?
214 		 * mask_match() returns 48 (32 + 16) for both of them.
215 		 *	src=3ffe::/16, dst=3ffe:501::/32
216 		 *	src=3ffe:501::/32, dst=3ffe::/16
217 		 *
218 		 * We need to loop through all the possible candidates
219 		 * to get the best match - the search takes O(n) for
220 		 * n attachments (i.e. interfaces).
221 		 */
222 		if (prio <= 0)
223 			continue;
224 		if (prio > matchprio) {
225 			matchprio = prio;
226 			match = ep;
227 		}
228 	}
229 
230 	if (match) {
231 		/* found a match, "match" has the best one */
232 		if (match->psw && match->psw->pr_input) {
233 			encap_fillarg(m, match);
234 			(*match->psw->pr_input)(m, off, proto);
235 		} else
236 			m_freem(m);
237 		return;
238 	}
239 
240 	/* backward compatibility clauses */
241 #ifdef MROUTING
242 	if (proto == IPPROTO_IPV4 && mrt_ipip_input(m, off)) {
243 		/*
244 		 * Multicast routing code claimed this one.  No
245 		 * more processing at this level.
246 		 */
247 		return;
248 	}
249 #endif
250 
251 	/* last resort: inject to raw socket */
252 	rip_input(m, off, proto);
253 }
254 #endif
255 
256 #ifdef INET6
257 int
258 encap6_input(mp, offp, proto)
259 	struct mbuf **mp;
260 	int *offp;
261 	int proto;
262 {
263 	struct mbuf *m = *mp;
264 	struct ip6_hdr *ip6;
265 	struct sockaddr_in6 s, d;
266 	struct ip6protosw *psw;
267 	struct encaptab *ep, *match;
268 	int prio, matchprio;
269 
270 	ip6 = mtod(m, struct ip6_hdr *);
271 
272 	bzero(&s, sizeof(s));
273 	s.sin6_family = AF_INET6;
274 	s.sin6_len = sizeof(struct sockaddr_in6);
275 	s.sin6_addr = ip6->ip6_src;
276 	bzero(&d, sizeof(d));
277 	d.sin6_family = AF_INET6;
278 	d.sin6_len = sizeof(struct sockaddr_in6);
279 	d.sin6_addr = ip6->ip6_dst;
280 
281 	match = NULL;
282 	matchprio = 0;
283 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
284 		if (ep->af != AF_INET6)
285 			continue;
286 		if (ep->proto >= 0 && ep->proto != proto)
287 			continue;
288 		if (ep->func)
289 			prio = (*ep->func)(m, *offp, proto, ep->arg);
290 		else {
291 			/*
292 			 * it's inbound traffic, we need to match in reverse
293 			 * order
294 			 */
295 			prio = mask_match(ep, (struct sockaddr *)&d,
296 			    (struct sockaddr *)&s);
297 		}
298 
299 		/* see encap4_input() for issues here */
300 		if (prio <= 0)
301 			continue;
302 		if (prio > matchprio) {
303 			matchprio = prio;
304 			match = ep;
305 		}
306 	}
307 
308 	if (match) {
309 		/* found a match */
310 		psw = (struct ip6protosw *)match->psw;
311 		if (psw && psw->pr_input) {
312 			encap_fillarg(m, match);
313 			return (*psw->pr_input)(mp, offp, proto);
314 		} else {
315 			m_freem(m);
316 			return IPPROTO_DONE;
317 		}
318 	}
319 
320 	/* last resort: inject to raw socket */
321 	return rip6_input(mp, offp, proto);
322 }
323 #endif
324 
325 static void
326 encap_add(ep)
327 	struct encaptab *ep;
328 {
329 
330 	LIST_INSERT_HEAD(&encaptab, ep, chain);
331 }
332 
333 /*
334  * sp (src ptr) is always my side, and dp (dst ptr) is always remote side.
335  * length of mask (sm and dm) is assumed to be same as sp/dp.
336  * Return value will be necessary as input (cookie) for encap_detach().
337  */
338 const struct encaptab *
339 encap_attach(af, proto, sp, sm, dp, dm, psw, arg)
340 	int af;
341 	int proto;
342 	const struct sockaddr *sp, *sm;
343 	const struct sockaddr *dp, *dm;
344 	const struct protosw *psw;
345 	void *arg;
346 {
347 	struct encaptab *ep;
348 	int error;
349 	int s;
350 
351 #if defined(__NetBSD__) || defined(__OpenBSD__)
352 	s = splsoftnet();
353 #else
354 	s = splnet();
355 #endif
356 	/* sanity check on args */
357 	if (sp->sa_len > sizeof(ep->src) || dp->sa_len > sizeof(ep->dst)) {
358 		error = EINVAL;
359 		goto fail;
360 	}
361 	if (sp->sa_len != dp->sa_len) {
362 		error = EINVAL;
363 		goto fail;
364 	}
365 	if (af != sp->sa_family || af != dp->sa_family) {
366 		error = EINVAL;
367 		goto fail;
368 	}
369 
370 	/* check if anyone have already attached with exactly same config */
371 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
372 		if (ep->af != af)
373 			continue;
374 		if (ep->proto != proto)
375 			continue;
376 		if (ep->src.ss_len != sp->sa_len ||
377 		    bcmp(&ep->src, sp, sp->sa_len) != 0 ||
378 		    bcmp(&ep->srcmask, sm, sp->sa_len) != 0)
379 			continue;
380 		if (ep->dst.ss_len != dp->sa_len ||
381 		    bcmp(&ep->dst, dp, dp->sa_len) != 0 ||
382 		    bcmp(&ep->dstmask, dm, dp->sa_len) != 0)
383 			continue;
384 
385 		error = EEXIST;
386 		goto fail;
387 	}
388 
389 	/*
390 	 * XXX NEED TO CHECK viftable IN THE ip_mroute CODE!!!
391 	 * XXX Actually, that code needs to be replaced with
392 	 * XXX new code that uses `gif' tunnels.
393 	 */
394 
395 	ep = malloc(sizeof(*ep), M_NETADDR, M_NOWAIT);	/*XXX*/
396 	if (ep == NULL) {
397 		error = ENOBUFS;
398 		goto fail;
399 	}
400 	bzero(ep, sizeof(*ep));
401 
402 	ep->af = af;
403 	ep->proto = proto;
404 	bcopy(sp, &ep->src, sp->sa_len);
405 	bcopy(sm, &ep->srcmask, sp->sa_len);
406 	bcopy(dp, &ep->dst, dp->sa_len);
407 	bcopy(dm, &ep->dstmask, dp->sa_len);
408 	ep->psw = psw;
409 	ep->arg = arg;
410 
411 	encap_add(ep);
412 
413 	error = 0;
414 	splx(s);
415 	return ep;
416 
417 fail:
418 	splx(s);
419 	return NULL;
420 }
421 
422 const struct encaptab *
423 encap_attach_func(af, proto, func, psw, arg)
424 	int af;
425 	int proto;
426 	int (*func) __P((const struct mbuf *, int, int, void *));
427 	const struct protosw *psw;
428 	void *arg;
429 {
430 	struct encaptab *ep;
431 	int error;
432 	int s;
433 
434 #if defined(__NetBSD__) || defined(__OpenBSD__)
435 	s = splsoftnet();
436 #else
437 	s = splnet();
438 #endif
439 	/* sanity check on args */
440 	if (!func) {
441 		error = EINVAL;
442 		goto fail;
443 	}
444 
445 	ep = malloc(sizeof(*ep), M_NETADDR, M_NOWAIT);	/*XXX*/
446 	if (ep == NULL) {
447 		error = ENOBUFS;
448 		goto fail;
449 	}
450 	bzero(ep, sizeof(*ep));
451 
452 	ep->af = af;
453 	ep->proto = proto;
454 	ep->func = func;
455 	ep->psw = psw;
456 	ep->arg = arg;
457 
458 	encap_add(ep);
459 
460 	error = 0;
461 	splx(s);
462 	return ep;
463 
464 fail:
465 	splx(s);
466 	return NULL;
467 }
468 
469 int
470 encap_detach(cookie)
471 	const struct encaptab *cookie;
472 {
473 	const struct encaptab *ep = cookie;
474 	struct encaptab *p;
475 
476 	for (p = LIST_FIRST(&encaptab); p; p = LIST_NEXT(p, chain)) {
477 		if (p == ep) {
478 			LIST_REMOVE(p, chain);
479 			free(p, M_NETADDR);	/*XXX*/
480 			return 0;
481 		}
482 	}
483 
484 	return EINVAL;
485 }
486 
487 static int
488 mask_match(ep, sp, dp)
489 	const struct encaptab *ep;
490 	const struct sockaddr *sp;
491 	const struct sockaddr *dp;
492 {
493 	struct sockaddr_storage s;
494 	struct sockaddr_storage d;
495 	int i;
496 	u_int8_t *p, *q, *r;
497 	int matchlen;
498 
499 	if (sp->sa_len > sizeof(s) || dp->sa_len > sizeof(d))
500 		return 0;
501 	if (sp->sa_family != ep->af || dp->sa_family != ep->af)
502 		return 0;
503 	if (sp->sa_len != ep->src.ss_len || dp->sa_len != ep->dst.ss_len)
504 		return 0;
505 
506 	matchlen = 0;
507 
508 	p = (u_int8_t *)sp;
509 	q = (u_int8_t *)&ep->srcmask;
510 	r = (u_int8_t *)&s;
511 	for (i = 0 ; i < sp->sa_len; i++) {
512 		r[i] = p[i] & q[i];
513 		/* XXX estimate */
514 		matchlen += (q[i] ? 8 : 0);
515 	}
516 
517 	p = (u_int8_t *)dp;
518 	q = (u_int8_t *)&ep->dstmask;
519 	r = (u_int8_t *)&d;
520 	for (i = 0 ; i < dp->sa_len; i++) {
521 		r[i] = p[i] & q[i];
522 		/* XXX rough estimate */
523 		matchlen += (q[i] ? 8 : 0);
524 	}
525 
526 	/* need to overwrite len/family portion as we don't compare them */
527 	s.ss_len = sp->sa_len;
528 	s.ss_family = sp->sa_family;
529 	d.ss_len = dp->sa_len;
530 	d.ss_family = dp->sa_family;
531 
532 	if (bcmp(&s, &ep->src, ep->src.ss_len) == 0 &&
533 	    bcmp(&d, &ep->dst, ep->dst.ss_len) == 0) {
534 		return matchlen;
535 	} else
536 		return 0;
537 }
538 
539 static void
540 encap_fillarg(m, ep)
541 	struct mbuf *m;
542 	const struct encaptab *ep;
543 {
544 #if 0
545 	m->m_pkthdr.aux = ep->arg;
546 #else
547 	struct mbuf *n;
548 
549 	n = m_aux_add(m, AF_INET, IPPROTO_IPV4);
550 	if (n) {
551 		*mtod(n, void **) = ep->arg;
552 		n->m_len = sizeof(void *);
553 	}
554 #endif
555 }
556 
557 void *
558 encap_getarg(m)
559 	struct mbuf *m;
560 {
561 	void *p;
562 #if 0
563 	p = m->m_pkthdr.aux;
564 	m->m_pkthdr.aux = NULL;
565 	return p;
566 #else
567 	struct mbuf *n;
568 
569 	p = NULL;
570 	n = m_aux_find(m, AF_INET, IPPROTO_IPV4);
571 	if (n) {
572 		if (n->m_len == sizeof(void *))
573 			p = *mtod(n, void **);
574 		m_aux_delete(m, n);
575 	}
576 	return p;
577 #endif
578 }
579