xref: /dflybsd-src/sys/netinet/ip_encap.c (revision 48d201a5a8c1dab4aa7166b0812594c101fc43c3)
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 1.8 2004/06/03 18:30:03 joerg 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 (struct encaptab *);
98 static int mask_match (const struct encaptab *, const struct sockaddr *,
99 		const struct sockaddr *);
100 static void encap_fillarg (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, ...)
134 {
135 	int off, proto;
136 	struct ip *ip;
137 	struct sockaddr_in s, d;
138 	const struct ipprotosw *psw;
139 	struct encaptab *ep, *match;
140 	int prio, matchprio;
141 	__va_list ap;
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 = (const struct ipprotosw *)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 	/* for backward compatibility */
215 	if (proto == IPPROTO_IPV4 && ipip_input) {
216 		ipip_input(m, off, proto);
217 		return;
218 	}
219 
220 	/* last resort: inject to raw socket */
221 	rip_input(m, off, proto);
222 }
223 #endif
224 
225 #ifdef INET6
226 int
227 encap6_input(mp, offp, proto)
228 	struct mbuf **mp;
229 	int *offp;
230 	int proto;
231 {
232 	struct mbuf *m = *mp;
233 	struct ip6_hdr *ip6;
234 	struct sockaddr_in6 s, d;
235 	const struct ip6protosw *psw;
236 	struct encaptab *ep, *match;
237 	int prio, matchprio;
238 
239 	ip6 = mtod(m, struct ip6_hdr *);
240 
241 	bzero(&s, sizeof(s));
242 	s.sin6_family = AF_INET6;
243 	s.sin6_len = sizeof(struct sockaddr_in6);
244 	s.sin6_addr = ip6->ip6_src;
245 	bzero(&d, sizeof(d));
246 	d.sin6_family = AF_INET6;
247 	d.sin6_len = sizeof(struct sockaddr_in6);
248 	d.sin6_addr = ip6->ip6_dst;
249 
250 	match = NULL;
251 	matchprio = 0;
252 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
253 		if (ep->af != AF_INET6)
254 			continue;
255 		if (ep->proto >= 0 && ep->proto != proto)
256 			continue;
257 		if (ep->func)
258 			prio = (*ep->func)(m, *offp, proto, ep->arg);
259 		else {
260 			/*
261 			 * it's inbound traffic, we need to match in reverse
262 			 * order
263 			 */
264 			prio = mask_match(ep, (struct sockaddr *)&d,
265 			    (struct sockaddr *)&s);
266 		}
267 
268 		/* see encap4_input() for issues here */
269 		if (prio <= 0)
270 			continue;
271 		if (prio > matchprio) {
272 			matchprio = prio;
273 			match = ep;
274 		}
275 	}
276 
277 	if (match) {
278 		/* found a match */
279 		psw = (const struct ip6protosw *)match->psw;
280 		if (psw && psw->pr_input) {
281 			encap_fillarg(m, match);
282 			return (*psw->pr_input)(mp, offp, proto);
283 		} else {
284 			m_freem(m);
285 			return IPPROTO_DONE;
286 		}
287 	}
288 
289 	/* last resort: inject to raw socket */
290 	return rip6_input(mp, offp, proto);
291 }
292 #endif
293 
294 static void
295 encap_add(ep)
296 	struct encaptab *ep;
297 {
298 
299 	LIST_INSERT_HEAD(&encaptab, ep, chain);
300 }
301 
302 /*
303  * sp (src ptr) is always my side, and dp (dst ptr) is always remote side.
304  * length of mask (sm and dm) is assumed to be same as sp/dp.
305  * Return value will be necessary as input (cookie) for encap_detach().
306  */
307 const struct encaptab *
308 encap_attach(af, proto, sp, sm, dp, dm, psw, arg)
309 	int af;
310 	int proto;
311 	const struct sockaddr *sp, *sm;
312 	const struct sockaddr *dp, *dm;
313 	const struct protosw *psw;
314 	void *arg;
315 {
316 	struct encaptab *ep;
317 	int error;
318 	int s;
319 
320 	s = splnet();
321 	/* sanity check on args */
322 	if (sp->sa_len > sizeof(ep->src) || dp->sa_len > sizeof(ep->dst)) {
323 		error = EINVAL;
324 		goto fail;
325 	}
326 	if (sp->sa_len != dp->sa_len) {
327 		error = EINVAL;
328 		goto fail;
329 	}
330 	if (af != sp->sa_family || af != dp->sa_family) {
331 		error = EINVAL;
332 		goto fail;
333 	}
334 
335 	/* check if anyone have already attached with exactly same config */
336 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
337 		if (ep->af != af)
338 			continue;
339 		if (ep->proto != proto)
340 			continue;
341 		if (ep->src.ss_len != sp->sa_len ||
342 		    bcmp(&ep->src, sp, sp->sa_len) != 0 ||
343 		    bcmp(&ep->srcmask, sm, sp->sa_len) != 0)
344 			continue;
345 		if (ep->dst.ss_len != dp->sa_len ||
346 		    bcmp(&ep->dst, dp, dp->sa_len) != 0 ||
347 		    bcmp(&ep->dstmask, dm, dp->sa_len) != 0)
348 			continue;
349 
350 		error = EEXIST;
351 		goto fail;
352 	}
353 
354 	ep = malloc(sizeof(*ep), M_NETADDR, M_INTWAIT | M_ZERO | M_NULLOK);
355 	if (ep == NULL) {
356 		error = ENOBUFS;
357 		goto fail;
358 	}
359 
360 	ep->af = af;
361 	ep->proto = proto;
362 	bcopy(sp, &ep->src, sp->sa_len);
363 	bcopy(sm, &ep->srcmask, sp->sa_len);
364 	bcopy(dp, &ep->dst, dp->sa_len);
365 	bcopy(dm, &ep->dstmask, dp->sa_len);
366 	ep->psw = psw;
367 	ep->arg = arg;
368 
369 	encap_add(ep);
370 
371 	error = 0;
372 	splx(s);
373 	return ep;
374 
375 fail:
376 	splx(s);
377 	return NULL;
378 }
379 
380 const struct encaptab *
381 encap_attach_func(af, proto, func, psw, arg)
382 	int af;
383 	int proto;
384 	int (*func) (const struct mbuf *, int, int, void *);
385 	const struct protosw *psw;
386 	void *arg;
387 {
388 	struct encaptab *ep;
389 	int error;
390 	int s;
391 
392 	s = splnet();
393 	/* sanity check on args */
394 	if (!func) {
395 		error = EINVAL;
396 		goto fail;
397 	}
398 
399 	ep = malloc(sizeof(*ep), M_NETADDR, M_INTWAIT | M_ZERO | M_NULLOK);
400 	if (ep == NULL) {
401 		error = ENOBUFS;
402 		goto fail;
403 	}
404 
405 	ep->af = af;
406 	ep->proto = proto;
407 	ep->func = func;
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 int
423 encap_detach(cookie)
424 	const struct encaptab *cookie;
425 {
426 	const struct encaptab *ep = cookie;
427 	struct encaptab *p;
428 
429 	for (p = LIST_FIRST(&encaptab); p; p = LIST_NEXT(p, chain)) {
430 		if (p == ep) {
431 			LIST_REMOVE(p, chain);
432 			free(p, M_NETADDR);	/*XXX*/
433 			return 0;
434 		}
435 	}
436 
437 	return EINVAL;
438 }
439 
440 static int
441 mask_match(ep, sp, dp)
442 	const struct encaptab *ep;
443 	const struct sockaddr *sp;
444 	const struct sockaddr *dp;
445 {
446 	struct sockaddr_storage s;
447 	struct sockaddr_storage d;
448 	int i;
449 	const u_int8_t *p, *q;
450 	u_int8_t *r;
451 	int matchlen;
452 
453 	if (sp->sa_len > sizeof(s) || dp->sa_len > sizeof(d))
454 		return 0;
455 	if (sp->sa_family != ep->af || dp->sa_family != ep->af)
456 		return 0;
457 	if (sp->sa_len != ep->src.ss_len || dp->sa_len != ep->dst.ss_len)
458 		return 0;
459 
460 	matchlen = 0;
461 
462 	p = (const u_int8_t *)sp;
463 	q = (const u_int8_t *)&ep->srcmask;
464 	r = (u_int8_t *)&s;
465 	for (i = 0 ; i < sp->sa_len; i++) {
466 		r[i] = p[i] & q[i];
467 		/* XXX estimate */
468 		matchlen += (q[i] ? 8 : 0);
469 	}
470 
471 	p = (const u_int8_t *)dp;
472 	q = (const u_int8_t *)&ep->dstmask;
473 	r = (u_int8_t *)&d;
474 	for (i = 0 ; i < dp->sa_len; i++) {
475 		r[i] = p[i] & q[i];
476 		/* XXX rough estimate */
477 		matchlen += (q[i] ? 8 : 0);
478 	}
479 
480 	/* need to overwrite len/family portion as we don't compare them */
481 	s.ss_len = sp->sa_len;
482 	s.ss_family = sp->sa_family;
483 	d.ss_len = dp->sa_len;
484 	d.ss_family = dp->sa_family;
485 
486 	if (bcmp(&s, &ep->src, ep->src.ss_len) == 0 &&
487 	    bcmp(&d, &ep->dst, ep->dst.ss_len) == 0) {
488 		return matchlen;
489 	} else
490 		return 0;
491 }
492 
493 static void
494 encap_fillarg(m, ep)
495 	struct mbuf *m;
496 	const struct encaptab *ep;
497 {
498 	struct m_tag *tag;
499 
500 	tag = m_tag_get(PACKET_TAG_ENCAP, sizeof (void*), MB_DONTWAIT);
501 	if (tag) {
502 		*(void**)(tag+1) = ep->arg;
503 		m_tag_prepend(m, tag);
504 	}
505 }
506 
507 void *
508 encap_getarg(m)
509 	struct mbuf *m;
510 {
511 	void *p = NULL;
512 	struct m_tag *tag;
513 
514 	tag = m_tag_find(m, PACKET_TAG_ENCAP, NULL);
515 	if (tag) {
516 		p = *(void**)(tag+1);
517 		m_tag_delete(m, tag);
518 	}
519 	return p;
520 }
521