xref: /netbsd-src/sys/netinet/ip_encap.c (revision 3b01aba77a7a698587faaae455bbfe740923c1f5)
1 /*	$NetBSD: ip_encap.c,v 1.5 2001/05/08 10:07:15 itojun Exp $	*/
2 /*	$KAME: ip_encap.c,v 1.39 2000/10/01 12:37:18 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 #include "opt_mrouting.h"
60 #include "opt_inet.h"
61 
62 #include <sys/param.h>
63 #include <sys/systm.h>
64 #include <sys/socket.h>
65 #include <sys/sockio.h>
66 #include <sys/mbuf.h>
67 #include <sys/errno.h>
68 #include <sys/protosw.h>
69 #include <sys/queue.h>
70 
71 #include <net/if.h>
72 #include <net/route.h>
73 
74 #include <netinet/in.h>
75 #include <netinet/in_systm.h>
76 #include <netinet/ip.h>
77 #include <netinet/ip_var.h>
78 #include <netinet/ip_encap.h>
79 #ifdef MROUTING
80 #include <netinet/ip_mroute.h>
81 #endif /* MROUTING */
82 
83 #ifdef INET6
84 #include <netinet/ip6.h>
85 #include <netinet6/ip6_var.h>
86 #include <netinet6/ip6protosw.h>
87 #endif
88 
89 #include <machine/stdarg.h>
90 
91 #include "ipip.h"
92 #if NIPIP > 0
93 # include <netinet/ip_ipip.h>
94 #else
95 # ifdef MROUTING
96 #  include <netinet/ip_mroute.h>
97 # endif
98 #endif
99 
100 #include <net/net_osdep.h>
101 
102 static void encap_add __P((struct encaptab *));
103 static int mask_match __P((const struct encaptab *, const struct sockaddr *,
104 		const struct sockaddr *));
105 static void encap_fillarg __P((struct mbuf *, const struct encaptab *));
106 
107 /* rely upon BSS initialization */
108 LIST_HEAD(, encaptab) encaptab = LIST_HEAD_INITIALIZER(&encaptab);
109 
110 void
111 encap_init()
112 {
113 #if 0
114 	/*
115 	 * we cannot use LIST_INIT() here, since drivers may want to call
116 	 * encap_attach(), on driver attach.  encap_init() will be called
117 	 * on AF_INET{,6} initialization, which happens after driver
118 	 * initialization - using LIST_INIT() here can nuke encap_attach()
119 	 * from drivers.
120 	 */
121 	LIST_INIT(&encaptab);
122 #endif
123 }
124 
125 #ifdef INET
126 void
127 #if __STDC__
128 encap4_input(struct mbuf *m, ...)
129 #else
130 encap4_input(m, va_alist)
131 	struct mbuf *m;
132 	va_dcl
133 #endif
134 {
135 	int off, proto;
136 	struct ip *ip;
137 	struct sockaddr_in s, d;
138 	const struct protosw *psw;
139 	struct encaptab *ep, *match;
140 	va_list ap;
141 	int prio, matchprio;
142 
143 	va_start(ap, m);
144 	off = va_arg(ap, int);
145 	proto = va_arg(ap, int);
146 	va_end(ap);
147 
148 	ip = mtod(m, struct ip *);
149 
150 	bzero(&s, sizeof(s));
151 	s.sin_family = AF_INET;
152 	s.sin_len = sizeof(struct sockaddr_in);
153 	s.sin_addr = ip->ip_src;
154 	bzero(&d, sizeof(d));
155 	d.sin_family = AF_INET;
156 	d.sin_len = sizeof(struct sockaddr_in);
157 	d.sin_addr = ip->ip_dst;
158 
159 	match = NULL;
160 	matchprio = 0;
161 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
162 		if (ep->af != AF_INET)
163 			continue;
164 		if (ep->proto >= 0 && ep->proto != proto)
165 			continue;
166 		if (ep->func)
167 			prio = (*ep->func)(m, off, proto, ep->arg);
168 		else {
169 			/*
170 			 * it's inbound traffic, we need to match in reverse
171 			 * order
172 			 */
173 			prio = mask_match(ep, (struct sockaddr *)&d,
174 			    (struct sockaddr *)&s);
175 		}
176 
177 		/*
178 		 * We prioritize the matches by using bit length of the
179 		 * matches.  mask_match() and user-supplied matching function
180 		 * should return the bit length of the matches (for example,
181 		 * if both src/dst are matched for IPv4, 64 should be returned).
182 		 * 0 or negative return value means "it did not match".
183 		 *
184 		 * The question is, since we have two "mask" portion, we
185 		 * cannot really define total order between entries.
186 		 * For example, which of these should be preferred?
187 		 * mask_match() returns 48 (32 + 16) for both of them.
188 		 *	src=3ffe::/16, dst=3ffe:501::/32
189 		 *	src=3ffe:501::/32, dst=3ffe::/16
190 		 *
191 		 * We need to loop through all the possible candidates
192 		 * to get the best match - the search takes O(n) for
193 		 * n attachments (i.e. interfaces).
194 		 */
195 		if (prio <= 0)
196 			continue;
197 		if (prio > matchprio) {
198 			matchprio = prio;
199 			match = ep;
200 		}
201 	}
202 
203 	if (match) {
204 		/* found a match, "match" has the best one */
205 		psw = match->psw;
206 		if (psw && psw->pr_input) {
207 			encap_fillarg(m, match);
208 			(*psw->pr_input)(m, off, proto);
209 		} else
210 			m_freem(m);
211 		return;
212 	}
213 
214 	/* last resort: inject to raw socket */
215 	rip_input(m, off, proto);
216 }
217 #endif
218 
219 #ifdef INET6
220 int
221 encap6_input(mp, offp, proto)
222 	struct mbuf **mp;
223 	int *offp;
224 	int proto;
225 {
226 	struct mbuf *m = *mp;
227 	struct ip6_hdr *ip6;
228 	struct sockaddr_in6 s, d;
229 	const struct ip6protosw *psw;
230 	struct encaptab *ep, *match;
231 	int prio, matchprio;
232 
233 	ip6 = mtod(m, struct ip6_hdr *);
234 
235 	bzero(&s, sizeof(s));
236 	s.sin6_family = AF_INET6;
237 	s.sin6_len = sizeof(struct sockaddr_in6);
238 	s.sin6_addr = ip6->ip6_src;
239 	bzero(&d, sizeof(d));
240 	d.sin6_family = AF_INET6;
241 	d.sin6_len = sizeof(struct sockaddr_in6);
242 	d.sin6_addr = ip6->ip6_dst;
243 
244 	match = NULL;
245 	matchprio = 0;
246 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
247 		if (ep->af != AF_INET6)
248 			continue;
249 		if (ep->proto >= 0 && ep->proto != proto)
250 			continue;
251 		if (ep->func)
252 			prio = (*ep->func)(m, *offp, proto, ep->arg);
253 		else {
254 			/*
255 			 * it's inbound traffic, we need to match in reverse
256 			 * order
257 			 */
258 			prio = mask_match(ep, (struct sockaddr *)&d,
259 			    (struct sockaddr *)&s);
260 		}
261 
262 		/* see encap4_input() for issues here */
263 		if (prio <= 0)
264 			continue;
265 		if (prio > matchprio) {
266 			matchprio = prio;
267 			match = ep;
268 		}
269 	}
270 
271 	if (match) {
272 		/* found a match */
273 		psw = (const struct ip6protosw *)match->psw;
274 		if (psw && psw->pr_input) {
275 			encap_fillarg(m, match);
276 			return (*psw->pr_input)(mp, offp, proto);
277 		} else {
278 			m_freem(m);
279 			return IPPROTO_DONE;
280 		}
281 	}
282 
283 	/* last resort: inject to raw socket */
284 	return rip6_input(mp, offp, proto);
285 }
286 #endif
287 
288 static void
289 encap_add(ep)
290 	struct encaptab *ep;
291 {
292 
293 	LIST_INSERT_HEAD(&encaptab, ep, chain);
294 }
295 
296 /*
297  * sp (src ptr) is always my side, and dp (dst ptr) is always remote side.
298  * length of mask (sm and dm) is assumed to be same as sp/dp.
299  * Return value will be necessary as input (cookie) for encap_detach().
300  */
301 const struct encaptab *
302 encap_attach(af, proto, sp, sm, dp, dm, psw, arg)
303 	int af;
304 	int proto;
305 	const struct sockaddr *sp, *sm;
306 	const struct sockaddr *dp, *dm;
307 	const struct protosw *psw;
308 	void *arg;
309 {
310 	struct encaptab *ep;
311 	int error;
312 	int s;
313 
314 	s = splsoftnet();
315 	/* sanity check on args */
316 	if (sp->sa_len > sizeof(ep->src) || dp->sa_len > sizeof(ep->dst)) {
317 		error = EINVAL;
318 		goto fail;
319 	}
320 	if (sp->sa_len != dp->sa_len) {
321 		error = EINVAL;
322 		goto fail;
323 	}
324 	if (af != sp->sa_family || af != dp->sa_family) {
325 		error = EINVAL;
326 		goto fail;
327 	}
328 
329 	/* check if anyone have already attached with exactly same config */
330 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
331 		if (ep->af != af)
332 			continue;
333 		if (ep->proto != proto)
334 			continue;
335 		if (ep->src.ss_len != sp->sa_len ||
336 		    bcmp(&ep->src, sp, sp->sa_len) != 0 ||
337 		    bcmp(&ep->srcmask, sm, sp->sa_len) != 0)
338 			continue;
339 		if (ep->dst.ss_len != dp->sa_len ||
340 		    bcmp(&ep->dst, dp, dp->sa_len) != 0 ||
341 		    bcmp(&ep->dstmask, dm, dp->sa_len) != 0)
342 			continue;
343 
344 		error = EEXIST;
345 		goto fail;
346 	}
347 
348 	/*
349 	 * XXX NEED TO CHECK viftable IN THE ip_mroute CODE!!!
350 	 * XXX Actually, that code needs to be replaced with
351 	 * XXX new code that uses `gif' tunnels.
352 	 */
353 
354 	ep = malloc(sizeof(*ep), M_NETADDR, M_NOWAIT);	/*XXX*/
355 	if (ep == NULL) {
356 		error = ENOBUFS;
357 		goto fail;
358 	}
359 	bzero(ep, sizeof(*ep));
360 
361 	ep->af = af;
362 	ep->proto = proto;
363 	bcopy(sp, &ep->src, sp->sa_len);
364 	bcopy(sm, &ep->srcmask, sp->sa_len);
365 	bcopy(dp, &ep->dst, dp->sa_len);
366 	bcopy(dm, &ep->dstmask, dp->sa_len);
367 	ep->psw = psw;
368 	ep->arg = arg;
369 
370 	encap_add(ep);
371 
372 	error = 0;
373 	splx(s);
374 	return ep;
375 
376 fail:
377 	splx(s);
378 	return NULL;
379 }
380 
381 const struct encaptab *
382 encap_attach_func(af, proto, func, psw, arg)
383 	int af;
384 	int proto;
385 	int (*func) __P((const struct mbuf *, int, int, void *));
386 	const struct protosw *psw;
387 	void *arg;
388 {
389 	struct encaptab *ep;
390 	int error;
391 	int s;
392 
393 	s = splsoftnet();
394 	/* sanity check on args */
395 	if (!func) {
396 		error = EINVAL;
397 		goto fail;
398 	}
399 
400 	ep = malloc(sizeof(*ep), M_NETADDR, M_NOWAIT);	/*XXX*/
401 	if (ep == NULL) {
402 		error = ENOBUFS;
403 		goto fail;
404 	}
405 	bzero(ep, sizeof(*ep));
406 
407 	ep->af = af;
408 	ep->proto = proto;
409 	ep->func = func;
410 	ep->psw = psw;
411 	ep->arg = arg;
412 
413 	encap_add(ep);
414 
415 	error = 0;
416 	splx(s);
417 	return ep;
418 
419 fail:
420 	splx(s);
421 	return NULL;
422 }
423 
424 int
425 encap_detach(cookie)
426 	const struct encaptab *cookie;
427 {
428 	const struct encaptab *ep = cookie;
429 	struct encaptab *p;
430 
431 	for (p = LIST_FIRST(&encaptab); p; p = LIST_NEXT(p, chain)) {
432 		if (p == ep) {
433 			LIST_REMOVE(p, chain);
434 			free(p, M_NETADDR);	/*XXX*/
435 			return 0;
436 		}
437 	}
438 
439 	return EINVAL;
440 }
441 
442 static int
443 mask_match(ep, sp, dp)
444 	const struct encaptab *ep;
445 	const struct sockaddr *sp;
446 	const struct sockaddr *dp;
447 {
448 	struct sockaddr_storage s;
449 	struct sockaddr_storage d;
450 	int i;
451 	const u_int8_t *p, *q;
452 	u_int8_t *r;
453 	int matchlen;
454 
455 	if (sp->sa_len > sizeof(s) || dp->sa_len > sizeof(d))
456 		return 0;
457 	if (sp->sa_family != ep->af || dp->sa_family != ep->af)
458 		return 0;
459 	if (sp->sa_len != ep->src.ss_len || dp->sa_len != ep->dst.ss_len)
460 		return 0;
461 
462 	matchlen = 0;
463 
464 	p = (const u_int8_t *)sp;
465 	q = (const u_int8_t *)&ep->srcmask;
466 	r = (u_int8_t *)&s;
467 	for (i = 0 ; i < sp->sa_len; i++) {
468 		r[i] = p[i] & q[i];
469 		/* XXX estimate */
470 		matchlen += (q[i] ? 8 : 0);
471 	}
472 
473 	p = (const u_int8_t *)dp;
474 	q = (const u_int8_t *)&ep->dstmask;
475 	r = (u_int8_t *)&d;
476 	for (i = 0 ; i < dp->sa_len; i++) {
477 		r[i] = p[i] & q[i];
478 		/* XXX rough estimate */
479 		matchlen += (q[i] ? 8 : 0);
480 	}
481 
482 	/* need to overwrite len/family portion as we don't compare them */
483 	s.ss_len = sp->sa_len;
484 	s.ss_family = sp->sa_family;
485 	d.ss_len = dp->sa_len;
486 	d.ss_family = dp->sa_family;
487 
488 	if (bcmp(&s, &ep->src, ep->src.ss_len) == 0 &&
489 	    bcmp(&d, &ep->dst, ep->dst.ss_len) == 0) {
490 		return matchlen;
491 	} else
492 		return 0;
493 }
494 
495 static void
496 encap_fillarg(m, ep)
497 	struct mbuf *m;
498 	const struct encaptab *ep;
499 {
500 #if 0
501 	m->m_pkthdr.aux = ep->arg;
502 #else
503 	struct mbuf *n;
504 
505 	n = m_aux_add(m, AF_INET, IPPROTO_IPV4);
506 	if (n) {
507 		*mtod(n, void **) = ep->arg;
508 		n->m_len = sizeof(void *);
509 	}
510 #endif
511 }
512 
513 void *
514 encap_getarg(m)
515 	struct mbuf *m;
516 {
517 	void *p;
518 #if 0
519 	p = m->m_pkthdr.aux;
520 	m->m_pkthdr.aux = NULL;
521 	return p;
522 #else
523 	struct mbuf *n;
524 
525 	p = NULL;
526 	n = m_aux_find(m, AF_INET, IPPROTO_IPV4);
527 	if (n) {
528 		if (n->m_len == sizeof(void *))
529 			p = *mtod(n, void **);
530 		m_aux_delete(m, n);
531 	}
532 	return p;
533 #endif
534 }
535