xref: /dragonfly/sys/netinet/ip_encap.c (revision 1de703da)
1 /*	$FreeBSD: src/sys/netinet/ip_encap.c,v 1.1.2.5 2003/01/23 21:06:45 sam Exp $	*/
2 /*	$DragonFly: src/sys/netinet/ip_encap.c,v 1.2 2003/06/17 04:28:51 dillon Exp $	*/
3 /*	$KAME: ip_encap.c,v 1.41 2001/03/15 08:35:08 itojun Exp $	*/
4 
5 /*
6  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the name of the project nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 /*
34  * My grandfather said that there's a devil inside tunnelling technology...
35  *
36  * We have surprisingly many protocols that want packets with IP protocol
37  * #4 or #41.  Here's a list of protocols that want protocol #41:
38  *	RFC1933 configured tunnel
39  *	RFC1933 automatic tunnel
40  *	RFC2401 IPsec tunnel
41  *	RFC2473 IPv6 generic packet tunnelling
42  *	RFC2529 6over4 tunnel
43  *	mobile-ip6 (uses RFC2473)
44  *	RFC3056 6to4 tunnel
45  *	isatap tunnel
46  * Here's a list of protocol that want protocol #4:
47  *	RFC1853 IPv4-in-IPv4 tunnelling
48  *	RFC2003 IPv4 encapsulation within IPv4
49  *	RFC2344 reverse tunnelling for mobile-ip4
50  *	RFC2401 IPsec tunnel
51  * Well, what can I say.  They impose different en/decapsulation mechanism
52  * from each other, so they need separate protocol handler.  The only one
53  * we can easily determine by protocol # is IPsec, which always has
54  * AH/ESP/IPComp header right after outer IP header.
55  *
56  * So, clearly good old protosw does not work for protocol #4 and #41.
57  * The code will let you match protocol via src/dst address pair.
58  */
59 /* XXX is M_NETADDR correct? */
60 
61 #include "opt_inet.h"
62 #include "opt_inet6.h"
63 
64 #include <sys/param.h>
65 #include <sys/systm.h>
66 #include <sys/socket.h>
67 #include <sys/sockio.h>
68 #include <sys/mbuf.h>
69 #include <sys/errno.h>
70 #include <sys/protosw.h>
71 #include <sys/queue.h>
72 
73 #include <net/if.h>
74 #include <net/route.h>
75 
76 #include <netinet/in.h>
77 #include <netinet/in_systm.h>
78 #include <netinet/ip.h>
79 #include <netinet/ip_var.h>
80 #include <netinet/ip_encap.h>
81 #include <netinet/ipprotosw.h>
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 <net/net_osdep.h>
92 
93 #include <sys/kernel.h>
94 #include <sys/malloc.h>
95 MALLOC_DEFINE(M_NETADDR, "Export Host", "Export host address structure");
96 
97 static void encap_add __P((struct encaptab *));
98 static int mask_match __P((const struct encaptab *, const struct sockaddr *,
99 		const struct sockaddr *));
100 static void encap_fillarg __P((struct mbuf *, const struct encaptab *));
101 
102 #ifndef LIST_HEAD_INITIALIZER
103 /* rely upon BSS initialization */
104 LIST_HEAD(, encaptab) encaptab;
105 #else
106 LIST_HEAD(, encaptab) encaptab = LIST_HEAD_INITIALIZER(&encaptab);
107 #endif
108 
109 void     (*ipip_input)(struct mbuf *, int, int); /* hook for mrouting */
110 
111 void
112 encap_init()
113 {
114 	static int initialized = 0;
115 
116 	if (initialized)
117 		return;
118 	initialized++;
119 #if 0
120 	/*
121 	 * we cannot use LIST_INIT() here, since drivers may want to call
122 	 * encap_attach(), on driver attach.  encap_init() will be called
123 	 * on AF_INET{,6} initialization, which happens after driver
124 	 * initialization - using LIST_INIT() here can nuke encap_attach()
125 	 * from drivers.
126 	 */
127 	LIST_INIT(&encaptab);
128 #endif
129 }
130 
131 #ifdef INET
132 void
133 encap4_input(struct mbuf *m, int off, int proto)
134 {
135 	struct ip *ip;
136 	struct sockaddr_in s, d;
137 	const struct ipprotosw *psw;
138 	struct encaptab *ep, *match;
139 	int prio, matchprio;
140 
141 	ip = mtod(m, struct ip *);
142 
143 	bzero(&s, sizeof(s));
144 	s.sin_family = AF_INET;
145 	s.sin_len = sizeof(struct sockaddr_in);
146 	s.sin_addr = ip->ip_src;
147 	bzero(&d, sizeof(d));
148 	d.sin_family = AF_INET;
149 	d.sin_len = sizeof(struct sockaddr_in);
150 	d.sin_addr = ip->ip_dst;
151 
152 	match = NULL;
153 	matchprio = 0;
154 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
155 		if (ep->af != AF_INET)
156 			continue;
157 		if (ep->proto >= 0 && ep->proto != proto)
158 			continue;
159 		if (ep->func)
160 			prio = (*ep->func)(m, off, proto, ep->arg);
161 		else {
162 			/*
163 			 * it's inbound traffic, we need to match in reverse
164 			 * order
165 			 */
166 			prio = mask_match(ep, (struct sockaddr *)&d,
167 			    (struct sockaddr *)&s);
168 		}
169 
170 		/*
171 		 * We prioritize the matches by using bit length of the
172 		 * matches.  mask_match() and user-supplied matching function
173 		 * should return the bit length of the matches (for example,
174 		 * if both src/dst are matched for IPv4, 64 should be returned).
175 		 * 0 or negative return value means "it did not match".
176 		 *
177 		 * The question is, since we have two "mask" portion, we
178 		 * cannot really define total order between entries.
179 		 * For example, which of these should be preferred?
180 		 * mask_match() returns 48 (32 + 16) for both of them.
181 		 *	src=3ffe::/16, dst=3ffe:501::/32
182 		 *	src=3ffe:501::/32, dst=3ffe::/16
183 		 *
184 		 * We need to loop through all the possible candidates
185 		 * to get the best match - the search takes O(n) for
186 		 * n attachments (i.e. interfaces).
187 		 */
188 		if (prio <= 0)
189 			continue;
190 		if (prio > matchprio) {
191 			matchprio = prio;
192 			match = ep;
193 		}
194 	}
195 
196 	if (match) {
197 		/* found a match, "match" has the best one */
198 		psw = (const struct ipprotosw *)match->psw;
199 		if (psw && psw->pr_input) {
200 			encap_fillarg(m, match);
201 			(*psw->pr_input)(m, off, proto);
202 		} else
203 			m_freem(m);
204 		return;
205 	}
206 
207 	/* for backward compatibility */
208 	if (proto == IPPROTO_IPV4 && ipip_input) {
209 		ipip_input(m, off, proto);
210 		return;
211 	}
212 
213 	/* last resort: inject to raw socket */
214 	rip_input(m, off, proto);
215 }
216 #endif
217 
218 #ifdef INET6
219 int
220 encap6_input(mp, offp, proto)
221 	struct mbuf **mp;
222 	int *offp;
223 	int proto;
224 {
225 	struct mbuf *m = *mp;
226 	struct ip6_hdr *ip6;
227 	struct sockaddr_in6 s, d;
228 	const struct ip6protosw *psw;
229 	struct encaptab *ep, *match;
230 	int prio, matchprio;
231 
232 	ip6 = mtod(m, struct ip6_hdr *);
233 
234 	bzero(&s, sizeof(s));
235 	s.sin6_family = AF_INET6;
236 	s.sin6_len = sizeof(struct sockaddr_in6);
237 	s.sin6_addr = ip6->ip6_src;
238 	bzero(&d, sizeof(d));
239 	d.sin6_family = AF_INET6;
240 	d.sin6_len = sizeof(struct sockaddr_in6);
241 	d.sin6_addr = ip6->ip6_dst;
242 
243 	match = NULL;
244 	matchprio = 0;
245 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
246 		if (ep->af != AF_INET6)
247 			continue;
248 		if (ep->proto >= 0 && ep->proto != proto)
249 			continue;
250 		if (ep->func)
251 			prio = (*ep->func)(m, *offp, proto, ep->arg);
252 		else {
253 			/*
254 			 * it's inbound traffic, we need to match in reverse
255 			 * order
256 			 */
257 			prio = mask_match(ep, (struct sockaddr *)&d,
258 			    (struct sockaddr *)&s);
259 		}
260 
261 		/* see encap4_input() for issues here */
262 		if (prio <= 0)
263 			continue;
264 		if (prio > matchprio) {
265 			matchprio = prio;
266 			match = ep;
267 		}
268 	}
269 
270 	if (match) {
271 		/* found a match */
272 		psw = (const struct ip6protosw *)match->psw;
273 		if (psw && psw->pr_input) {
274 			encap_fillarg(m, match);
275 			return (*psw->pr_input)(mp, offp, proto);
276 		} else {
277 			m_freem(m);
278 			return IPPROTO_DONE;
279 		}
280 	}
281 
282 	/* last resort: inject to raw socket */
283 	return rip6_input(mp, offp, proto);
284 }
285 #endif
286 
287 static void
288 encap_add(ep)
289 	struct encaptab *ep;
290 {
291 
292 	LIST_INSERT_HEAD(&encaptab, ep, chain);
293 }
294 
295 /*
296  * sp (src ptr) is always my side, and dp (dst ptr) is always remote side.
297  * length of mask (sm and dm) is assumed to be same as sp/dp.
298  * Return value will be necessary as input (cookie) for encap_detach().
299  */
300 const struct encaptab *
301 encap_attach(af, proto, sp, sm, dp, dm, psw, arg)
302 	int af;
303 	int proto;
304 	const struct sockaddr *sp, *sm;
305 	const struct sockaddr *dp, *dm;
306 	const struct protosw *psw;
307 	void *arg;
308 {
309 	struct encaptab *ep;
310 	int error;
311 	int s;
312 
313 	s = splnet();
314 	/* sanity check on args */
315 	if (sp->sa_len > sizeof(ep->src) || dp->sa_len > sizeof(ep->dst)) {
316 		error = EINVAL;
317 		goto fail;
318 	}
319 	if (sp->sa_len != dp->sa_len) {
320 		error = EINVAL;
321 		goto fail;
322 	}
323 	if (af != sp->sa_family || af != dp->sa_family) {
324 		error = EINVAL;
325 		goto fail;
326 	}
327 
328 	/* check if anyone have already attached with exactly same config */
329 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
330 		if (ep->af != af)
331 			continue;
332 		if (ep->proto != proto)
333 			continue;
334 		if (ep->src.ss_len != sp->sa_len ||
335 		    bcmp(&ep->src, sp, sp->sa_len) != 0 ||
336 		    bcmp(&ep->srcmask, sm, sp->sa_len) != 0)
337 			continue;
338 		if (ep->dst.ss_len != dp->sa_len ||
339 		    bcmp(&ep->dst, dp, dp->sa_len) != 0 ||
340 		    bcmp(&ep->dstmask, dm, dp->sa_len) != 0)
341 			continue;
342 
343 		error = EEXIST;
344 		goto fail;
345 	}
346 
347 	ep = malloc(sizeof(*ep), M_NETADDR, M_NOWAIT);	/*XXX*/
348 	if (ep == NULL) {
349 		error = ENOBUFS;
350 		goto fail;
351 	}
352 	bzero(ep, sizeof(*ep));
353 
354 	ep->af = af;
355 	ep->proto = proto;
356 	bcopy(sp, &ep->src, sp->sa_len);
357 	bcopy(sm, &ep->srcmask, sp->sa_len);
358 	bcopy(dp, &ep->dst, dp->sa_len);
359 	bcopy(dm, &ep->dstmask, dp->sa_len);
360 	ep->psw = psw;
361 	ep->arg = arg;
362 
363 	encap_add(ep);
364 
365 	error = 0;
366 	splx(s);
367 	return ep;
368 
369 fail:
370 	splx(s);
371 	return NULL;
372 }
373 
374 const struct encaptab *
375 encap_attach_func(af, proto, func, psw, arg)
376 	int af;
377 	int proto;
378 	int (*func) __P((const struct mbuf *, int, int, void *));
379 	const struct protosw *psw;
380 	void *arg;
381 {
382 	struct encaptab *ep;
383 	int error;
384 	int s;
385 
386 	s = splnet();
387 	/* sanity check on args */
388 	if (!func) {
389 		error = EINVAL;
390 		goto fail;
391 	}
392 
393 	ep = malloc(sizeof(*ep), M_NETADDR, M_NOWAIT);	/*XXX*/
394 	if (ep == NULL) {
395 		error = ENOBUFS;
396 		goto fail;
397 	}
398 	bzero(ep, sizeof(*ep));
399 
400 	ep->af = af;
401 	ep->proto = proto;
402 	ep->func = func;
403 	ep->psw = psw;
404 	ep->arg = arg;
405 
406 	encap_add(ep);
407 
408 	error = 0;
409 	splx(s);
410 	return ep;
411 
412 fail:
413 	splx(s);
414 	return NULL;
415 }
416 
417 int
418 encap_detach(cookie)
419 	const struct encaptab *cookie;
420 {
421 	const struct encaptab *ep = cookie;
422 	struct encaptab *p;
423 
424 	for (p = LIST_FIRST(&encaptab); p; p = LIST_NEXT(p, chain)) {
425 		if (p == ep) {
426 			LIST_REMOVE(p, chain);
427 			free(p, M_NETADDR);	/*XXX*/
428 			return 0;
429 		}
430 	}
431 
432 	return EINVAL;
433 }
434 
435 static int
436 mask_match(ep, sp, dp)
437 	const struct encaptab *ep;
438 	const struct sockaddr *sp;
439 	const struct sockaddr *dp;
440 {
441 	struct sockaddr_storage s;
442 	struct sockaddr_storage d;
443 	int i;
444 	const u_int8_t *p, *q;
445 	u_int8_t *r;
446 	int matchlen;
447 
448 	if (sp->sa_len > sizeof(s) || dp->sa_len > sizeof(d))
449 		return 0;
450 	if (sp->sa_family != ep->af || dp->sa_family != ep->af)
451 		return 0;
452 	if (sp->sa_len != ep->src.ss_len || dp->sa_len != ep->dst.ss_len)
453 		return 0;
454 
455 	matchlen = 0;
456 
457 	p = (const u_int8_t *)sp;
458 	q = (const u_int8_t *)&ep->srcmask;
459 	r = (u_int8_t *)&s;
460 	for (i = 0 ; i < sp->sa_len; i++) {
461 		r[i] = p[i] & q[i];
462 		/* XXX estimate */
463 		matchlen += (q[i] ? 8 : 0);
464 	}
465 
466 	p = (const u_int8_t *)dp;
467 	q = (const u_int8_t *)&ep->dstmask;
468 	r = (u_int8_t *)&d;
469 	for (i = 0 ; i < dp->sa_len; i++) {
470 		r[i] = p[i] & q[i];
471 		/* XXX rough estimate */
472 		matchlen += (q[i] ? 8 : 0);
473 	}
474 
475 	/* need to overwrite len/family portion as we don't compare them */
476 	s.ss_len = sp->sa_len;
477 	s.ss_family = sp->sa_family;
478 	d.ss_len = dp->sa_len;
479 	d.ss_family = dp->sa_family;
480 
481 	if (bcmp(&s, &ep->src, ep->src.ss_len) == 0 &&
482 	    bcmp(&d, &ep->dst, ep->dst.ss_len) == 0) {
483 		return matchlen;
484 	} else
485 		return 0;
486 }
487 
488 static void
489 encap_fillarg(m, ep)
490 	struct mbuf *m;
491 	const struct encaptab *ep;
492 {
493 	struct m_tag *tag;
494 
495 	tag = m_tag_get(PACKET_TAG_ENCAP, sizeof (void*), M_NOWAIT);
496 	if (tag) {
497 		*(void**)(tag+1) = ep->arg;
498 		m_tag_prepend(m, tag);
499 	}
500 }
501 
502 void *
503 encap_getarg(m)
504 	struct mbuf *m;
505 {
506 	void *p = NULL;
507 	struct m_tag *tag;
508 
509 	tag = m_tag_find(m, PACKET_TAG_ENCAP, NULL);
510 	if (tag) {
511 		p = *(void**)(tag+1);
512 		m_tag_delete(m, tag);
513 	}
514 	return p;
515 }
516