xref: /freebsd/sys/netinet/ip_encap.c (revision 4f29da19bd44f0e99f021510460a81bf754c21d2) 
1 /*	$FreeBSD$	*/
2 /*	$KAME: ip_encap.c,v 1.41 2001/03/15 08:35:08 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  *	RFC3056 6to4 tunnel
44  *	isatap tunnel
45  * Here's a list of protocol that want protocol #4:
46  *	RFC1853 IPv4-in-IPv4 tunnelling
47  *	RFC2003 IPv4 encapsulation within IPv4
48  *	RFC2344 reverse tunnelling for mobile-ip4
49  *	RFC2401 IPsec tunnel
50  * Well, what can I say.  They impose different en/decapsulation mechanism
51  * from each other, so they need separate protocol handler.  The only one
52  * we can easily determine by protocol # is IPsec, which always has
53  * AH/ESP/IPComp header right after outer IP header.
54  *
55  * So, clearly good old protosw does not work for protocol #4 and #41.
56  * The code will let you match protocol via src/dst address pair.
57  */
58 /* XXX is M_NETADDR correct? */
59 
60 #include "opt_mrouting.h"
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 
82 #ifdef INET6
83 #include <netinet/ip6.h>
84 #include <netinet6/ip6_var.h>
85 #include <netinet6/ip6protosw.h>
86 #endif
87 
88 #include <machine/stdarg.h>
89 
90 #include <net/net_osdep.h>
91 
92 #include <sys/kernel.h>
93 #include <sys/malloc.h>
94 static MALLOC_DEFINE(M_NETADDR, "encap_export_host", "Export host address structure");
95 
96 static void encap_add(struct encaptab *);
97 static int mask_match(const struct encaptab *, const struct sockaddr *,
98 		const struct sockaddr *);
99 static void encap_fillarg(struct mbuf *, const struct encaptab *);
100 
101 /*
102  * All global variables in ip_encap.c are locked using encapmtx.
103  */
104 static struct mtx encapmtx;
105 MTX_SYSINIT(encapmtx, &encapmtx, "encapmtx", MTX_DEF);
106 LIST_HEAD(, encaptab) encaptab = LIST_HEAD_INITIALIZER(&encaptab);
107 
108 /*
109  * We currently keey encap_init() for source code compatibility reasons --
110  * it's referenced by KAME pieces in netinet6.
111  */
112 void
113 encap_init()
114 {
115 }
116 
117 #ifdef INET
118 void
119 encap4_input(m, off)
120 	struct mbuf *m;
121 	int off;
122 {
123 	struct ip *ip;
124 	int proto;
125 	struct sockaddr_in s, d;
126 	const struct protosw *psw;
127 	struct encaptab *ep, *match;
128 	int prio, matchprio;
129 
130 	ip = mtod(m, struct ip *);
131 	proto = ip->ip_p;
132 
133 	bzero(&s, sizeof(s));
134 	s.sin_family = AF_INET;
135 	s.sin_len = sizeof(struct sockaddr_in);
136 	s.sin_addr = ip->ip_src;
137 	bzero(&d, sizeof(d));
138 	d.sin_family = AF_INET;
139 	d.sin_len = sizeof(struct sockaddr_in);
140 	d.sin_addr = ip->ip_dst;
141 
142 	match = NULL;
143 	matchprio = 0;
144 	mtx_lock(&encapmtx);
145 	LIST_FOREACH(ep, &encaptab, chain) {
146 		if (ep->af != AF_INET)
147 			continue;
148 		if (ep->proto >= 0 && ep->proto != proto)
149 			continue;
150 		if (ep->func)
151 			prio = (*ep->func)(m, off, proto, ep->arg);
152 		else {
153 			/*
154 			 * it's inbound traffic, we need to match in reverse
155 			 * order
156 			 */
157 			prio = mask_match(ep, (struct sockaddr *)&d,
158 			    (struct sockaddr *)&s);
159 		}
160 
161 		/*
162 		 * We prioritize the matches by using bit length of the
163 		 * matches.  mask_match() and user-supplied matching function
164 		 * should return the bit length of the matches (for example,
165 		 * if both src/dst are matched for IPv4, 64 should be returned).
166 		 * 0 or negative return value means "it did not match".
167 		 *
168 		 * The question is, since we have two "mask" portion, we
169 		 * cannot really define total order between entries.
170 		 * For example, which of these should be preferred?
171 		 * mask_match() returns 48 (32 + 16) for both of them.
172 		 *	src=3ffe::/16, dst=3ffe:501::/32
173 		 *	src=3ffe:501::/32, dst=3ffe::/16
174 		 *
175 		 * We need to loop through all the possible candidates
176 		 * to get the best match - the search takes O(n) for
177 		 * n attachments (i.e. interfaces).
178 		 */
179 		if (prio <= 0)
180 			continue;
181 		if (prio > matchprio) {
182 			matchprio = prio;
183 			match = ep;
184 		}
185 	}
186 	mtx_unlock(&encapmtx);
187 
188 	if (match) {
189 		/* found a match, "match" has the best one */
190 		psw = match->psw;
191 		if (psw && psw->pr_input) {
192 			encap_fillarg(m, match);
193 			(*psw->pr_input)(m, off);
194 		} else
195 			m_freem(m);
196 		return;
197 	}
198 
199 	/* last resort: inject to raw socket */
200 	rip_input(m, off);
201 }
202 #endif
203 
204 #ifdef INET6
205 int
206 encap6_input(mp, offp, proto)
207 	struct mbuf **mp;
208 	int *offp;
209 	int proto;
210 {
211 	struct mbuf *m = *mp;
212 	struct ip6_hdr *ip6;
213 	struct sockaddr_in6 s, d;
214 	const struct ip6protosw *psw;
215 	struct encaptab *ep, *match;
216 	int prio, matchprio;
217 
218 	ip6 = mtod(m, struct ip6_hdr *);
219 
220 	bzero(&s, sizeof(s));
221 	s.sin6_family = AF_INET6;
222 	s.sin6_len = sizeof(struct sockaddr_in6);
223 	s.sin6_addr = ip6->ip6_src;
224 	bzero(&d, sizeof(d));
225 	d.sin6_family = AF_INET6;
226 	d.sin6_len = sizeof(struct sockaddr_in6);
227 	d.sin6_addr = ip6->ip6_dst;
228 
229 	match = NULL;
230 	matchprio = 0;
231 	mtx_lock(&encapmtx);
232 	LIST_FOREACH(ep, &encaptab, chain) {
233 		if (ep->af != AF_INET6)
234 			continue;
235 		if (ep->proto >= 0 && ep->proto != proto)
236 			continue;
237 		if (ep->func)
238 			prio = (*ep->func)(m, *offp, proto, ep->arg);
239 		else {
240 			/*
241 			 * it's inbound traffic, we need to match in reverse
242 			 * order
243 			 */
244 			prio = mask_match(ep, (struct sockaddr *)&d,
245 			    (struct sockaddr *)&s);
246 		}
247 
248 		/* see encap4_input() for issues here */
249 		if (prio <= 0)
250 			continue;
251 		if (prio > matchprio) {
252 			matchprio = prio;
253 			match = ep;
254 		}
255 	}
256 	mtx_unlock(&encapmtx);
257 
258 	if (match) {
259 		/* found a match */
260 		psw = (const struct ip6protosw *)match->psw;
261 		if (psw && psw->pr_input) {
262 			encap_fillarg(m, match);
263 			return (*psw->pr_input)(mp, offp, proto);
264 		} else {
265 			m_freem(m);
266 			return IPPROTO_DONE;
267 		}
268 	}
269 
270 	/* last resort: inject to raw socket */
271 	return rip6_input(mp, offp, proto);
272 }
273 #endif
274 
275 /*lint -sem(encap_add, custodial(1)) */
276 static void
277 encap_add(ep)
278 	struct encaptab *ep;
279 {
280 
281 	mtx_assert(&encapmtx, MA_OWNED);
282 	LIST_INSERT_HEAD(&encaptab, ep, chain);
283 }
284 
285 /*
286  * sp (src ptr) is always my side, and dp (dst ptr) is always remote side.
287  * length of mask (sm and dm) is assumed to be same as sp/dp.
288  * Return value will be necessary as input (cookie) for encap_detach().
289  */
290 const struct encaptab *
291 encap_attach(af, proto, sp, sm, dp, dm, psw, arg)
292 	int af;
293 	int proto;
294 	const struct sockaddr *sp, *sm;
295 	const struct sockaddr *dp, *dm;
296 	const struct protosw *psw;
297 	void *arg;
298 {
299 	struct encaptab *ep;
300 
301 	/* sanity check on args */
302 	if (sp->sa_len > sizeof(ep->src) || dp->sa_len > sizeof(ep->dst))
303 		return (NULL);
304 	if (sp->sa_len != dp->sa_len)
305 		return (NULL);
306 	if (af != sp->sa_family || af != dp->sa_family)
307 		return (NULL);
308 
309 	/* check if anyone have already attached with exactly same config */
310 	mtx_lock(&encapmtx);
311 	LIST_FOREACH(ep, &encaptab, chain) {
312 		if (ep->af != af)
313 			continue;
314 		if (ep->proto != proto)
315 			continue;
316 		if (ep->src.ss_len != sp->sa_len ||
317 		    bcmp(&ep->src, sp, sp->sa_len) != 0 ||
318 		    bcmp(&ep->srcmask, sm, sp->sa_len) != 0)
319 			continue;
320 		if (ep->dst.ss_len != dp->sa_len ||
321 		    bcmp(&ep->dst, dp, dp->sa_len) != 0 ||
322 		    bcmp(&ep->dstmask, dm, dp->sa_len) != 0)
323 			continue;
324 
325 		mtx_unlock(&encapmtx);
326 		return (NULL);
327 	}
328 
329 	ep = malloc(sizeof(*ep), M_NETADDR, M_NOWAIT);	/*XXX*/
330 	if (ep == NULL) {
331 		mtx_unlock(&encapmtx);
332 		return (NULL);
333 	}
334 	bzero(ep, sizeof(*ep));
335 
336 	ep->af = af;
337 	ep->proto = proto;
338 	bcopy(sp, &ep->src, sp->sa_len);
339 	bcopy(sm, &ep->srcmask, sp->sa_len);
340 	bcopy(dp, &ep->dst, dp->sa_len);
341 	bcopy(dm, &ep->dstmask, dp->sa_len);
342 	ep->psw = psw;
343 	ep->arg = arg;
344 
345 	encap_add(ep);
346 	mtx_unlock(&encapmtx);
347 	return (ep);
348 }
349 
350 const struct encaptab *
351 encap_attach_func(af, proto, func, psw, arg)
352 	int af;
353 	int proto;
354 	int (*func)(const struct mbuf *, int, int, void *);
355 	const struct protosw *psw;
356 	void *arg;
357 {
358 	struct encaptab *ep;
359 
360 	/* sanity check on args */
361 	if (!func)
362 		return (NULL);
363 
364 	ep = malloc(sizeof(*ep), M_NETADDR, M_NOWAIT);	/*XXX*/
365 	if (ep == NULL)
366 		return (NULL);
367 	bzero(ep, sizeof(*ep));
368 
369 	ep->af = af;
370 	ep->proto = proto;
371 	ep->func = func;
372 	ep->psw = psw;
373 	ep->arg = arg;
374 
375 	mtx_lock(&encapmtx);
376 	encap_add(ep);
377 	mtx_unlock(&encapmtx);
378 	return (ep);
379 }
380 
381 int
382 encap_detach(cookie)
383 	const struct encaptab *cookie;
384 {
385 	const struct encaptab *ep = cookie;
386 	struct encaptab *p;
387 
388 	mtx_lock(&encapmtx);
389 	LIST_FOREACH(p, &encaptab, chain) {
390 		if (p == ep) {
391 			LIST_REMOVE(p, chain);
392 			mtx_unlock(&encapmtx);
393 			free(p, M_NETADDR);	/*XXX*/
394 			return 0;
395 		}
396 	}
397 	mtx_unlock(&encapmtx);
398 
399 	return EINVAL;
400 }
401 
402 static int
403 mask_match(ep, sp, dp)
404 	const struct encaptab *ep;
405 	const struct sockaddr *sp;
406 	const struct sockaddr *dp;
407 {
408 	struct sockaddr_storage s;
409 	struct sockaddr_storage d;
410 	int i;
411 	const u_int8_t *p, *q;
412 	u_int8_t *r;
413 	int matchlen;
414 
415 	if (sp->sa_len > sizeof(s) || dp->sa_len > sizeof(d))
416 		return 0;
417 	if (sp->sa_family != ep->af || dp->sa_family != ep->af)
418 		return 0;
419 	if (sp->sa_len != ep->src.ss_len || dp->sa_len != ep->dst.ss_len)
420 		return 0;
421 
422 	matchlen = 0;
423 
424 	p = (const u_int8_t *)sp;
425 	q = (const u_int8_t *)&ep->srcmask;
426 	r = (u_int8_t *)&s;
427 	for (i = 0 ; i < sp->sa_len; i++) {
428 		r[i] = p[i] & q[i];
429 		/* XXX estimate */
430 		matchlen += (q[i] ? 8 : 0);
431 	}
432 
433 	p = (const u_int8_t *)dp;
434 	q = (const u_int8_t *)&ep->dstmask;
435 	r = (u_int8_t *)&d;
436 	for (i = 0 ; i < dp->sa_len; i++) {
437 		r[i] = p[i] & q[i];
438 		/* XXX rough estimate */
439 		matchlen += (q[i] ? 8 : 0);
440 	}
441 
442 	/* need to overwrite len/family portion as we don't compare them */
443 	s.ss_len = sp->sa_len;
444 	s.ss_family = sp->sa_family;
445 	d.ss_len = dp->sa_len;
446 	d.ss_family = dp->sa_family;
447 
448 	if (bcmp(&s, &ep->src, ep->src.ss_len) == 0 &&
449 	    bcmp(&d, &ep->dst, ep->dst.ss_len) == 0) {
450 		return matchlen;
451 	} else
452 		return 0;
453 }
454 
455 static void
456 encap_fillarg(m, ep)
457 	struct mbuf *m;
458 	const struct encaptab *ep;
459 {
460 	struct m_tag *tag;
461 
462 	tag = m_tag_get(PACKET_TAG_ENCAP, sizeof (void*), M_NOWAIT);
463 	if (tag) {
464 		*(void**)(tag+1) = ep->arg;
465 		m_tag_prepend(m, tag);
466 	}
467 }
468 
469 void *
470 encap_getarg(m)
471 	struct mbuf *m;
472 {
473 	void *p = NULL;
474 	struct m_tag *tag;
475 
476 	tag = m_tag_find(m, PACKET_TAG_ENCAP, NULL);
477 	if (tag) {
478 		p = *(void**)(tag+1);
479 		m_tag_delete(m, tag);
480 	}
481 	return p;
482 }
483