xref: /dragonfly/sys/netinet/ip_encap.c (revision e98bdfd3) 
1 /*	$FreeBSD: src/sys/netinet/ip_encap.c,v 1.1.2.5 2003/01/23 21:06:45 sam Exp $	*/
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_inet.h"
61 #include "opt_inet6.h"
62 
63 #include <sys/param.h>
64 #include <sys/systm.h>
65 #include <sys/socket.h>
66 #include <sys/sockio.h>
67 #include <sys/mbuf.h>
68 #include <sys/errno.h>
69 #include <sys/protosw.h>
70 #include <sys/queue.h>
71 
72 #include <net/if.h>
73 #include <net/route.h>
74 
75 #include <netinet/in.h>
76 #include <netinet/in_systm.h>
77 #include <netinet/ip.h>
78 #include <netinet/ip_var.h>
79 #include <netinet/ip_encap.h>
80 
81 #ifdef INET6
82 #include <netinet/ip6.h>
83 #include <netinet6/ip6_var.h>
84 #include <netinet6/ip6protosw.h>
85 #endif
86 
87 #include <machine/stdarg.h>
88 
89 #include <net/net_osdep.h>
90 
91 #include <sys/kernel.h>
92 #include <sys/malloc.h>
93 #include <sys/thread2.h>
94 MALLOC_DEFINE(M_NETADDR, "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 #ifndef LIST_HEAD_INITIALIZER
102 /* rely upon BSS initialization */
103 LIST_HEAD(, encaptab) encaptab;
104 #else
105 LIST_HEAD(, encaptab) encaptab = LIST_HEAD_INITIALIZER(&encaptab);
106 #endif
107 
108 int     (*ipip_input)(struct mbuf **, int *, int); /* hook for mrouting */
109 
110 void
111 encap_init(void)
112 {
113 	static int initialized = 0;
114 
115 	if (initialized)
116 		return;
117 	initialized++;
118 #if 0
119 	/*
120 	 * we cannot use LIST_INIT() here, since drivers may want to call
121 	 * encap_attach(), on driver attach.  encap_init() will be called
122 	 * on AF_INET{,6} initialization, which happens after driver
123 	 * initialization - using LIST_INIT() here can nuke encap_attach()
124 	 * from drivers.
125 	 */
126 	LIST_INIT(&encaptab);
127 #endif
128 }
129 
130 #ifdef INET
131 int
132 encap4_input(struct mbuf **mp, int *offp, int proto)
133 {
134 	struct mbuf *m = *mp;
135 	int off = *offp;
136 	struct ip *ip;
137 	struct sockaddr_in s, d;
138 	const struct protosw *psw;
139 	struct encaptab *ep, *match;
140 	int prio, matchprio;
141 
142 	ip = mtod(m, struct ip *);
143 	*mp = NULL;
144 
145 	bzero(&s, sizeof s);
146 	s.sin_family = AF_INET;
147 	s.sin_len = sizeof(struct sockaddr_in);
148 	s.sin_addr = ip->ip_src;
149 	bzero(&d, sizeof d);
150 	d.sin_family = AF_INET;
151 	d.sin_len = sizeof(struct sockaddr_in);
152 	d.sin_addr = ip->ip_dst;
153 
154 	match = NULL;
155 	matchprio = 0;
156 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
157 		if (ep->af != AF_INET)
158 			continue;
159 		if (ep->proto >= 0 && ep->proto != proto)
160 			continue;
161 		if (ep->func)
162 			prio = (*ep->func)(m, off, proto, ep->arg);
163 		else {
164 			/*
165 			 * it's inbound traffic, we need to match in reverse
166 			 * order
167 			 */
168 			prio = mask_match(ep,
169 					  (struct sockaddr *)&d,
170 					  (struct sockaddr *)&s);
171 		}
172 
173 		/*
174 		 * We prioritize the matches by using bit length of the
175 		 * matches.  mask_match() and user-supplied matching function
176 		 * should return the bit length of the matches (for example,
177 		 * if both src/dst are matched for IPv4, 64 should be returned).
178 		 * 0 or negative return value means "it did not match".
179 		 *
180 		 * The question is, since we have two "mask" portion, we
181 		 * cannot really define total order between entries.
182 		 * For example, which of these should be preferred?
183 		 * mask_match() returns 48 (32 + 16) for both of them.
184 		 *	src=3ffe::/16, dst=3ffe:501::/32
185 		 *	src=3ffe:501::/32, dst=3ffe::/16
186 		 *
187 		 * We need to loop through all the possible candidates
188 		 * to get the best match - the search takes O(n) for
189 		 * n attachments (i.e. interfaces).
190 		 */
191 		if (prio <= 0)
192 			continue;
193 		if (prio > matchprio) {
194 			matchprio = prio;
195 			match = ep;
196 		}
197 	}
198 
199 	if (match) {
200 		/* found a match, "match" has the best one */
201 		psw = match->psw;
202 		if (psw && psw->pr_input) {
203 			encap_fillarg(m, match);
204 			*mp = m;
205 			(*psw->pr_input)(mp, offp, proto);
206 		} else {
207 			m_freem(m);
208 		}
209 		return(IPPROTO_DONE);
210 	}
211 
212 	/* for backward compatibility */
213 	if (proto == IPPROTO_IPV4 && ipip_input) {
214 		*mp = m;
215 		ipip_input(mp, offp, proto);
216 		return(IPPROTO_DONE);
217 	}
218 
219 	/* last resort: inject to raw socket */
220 	*mp = m;
221 	rip_input(mp, offp, proto);
222 	return(IPPROTO_DONE);
223 }
224 #endif
225 
226 #ifdef INET6
227 int
228 encap6_input(struct mbuf **mp, int *offp, int proto)
229 {
230 	struct mbuf *m = *mp;
231 	struct ip6_hdr *ip6;
232 	struct sockaddr_in6 s, d;
233 	const struct protosw *psw;
234 	struct encaptab *ep, *match;
235 	int prio, matchprio;
236 
237 	ip6 = mtod(m, struct ip6_hdr *);
238 
239 	bzero(&s, sizeof s);
240 	s.sin6_family = AF_INET6;
241 	s.sin6_len = sizeof(struct sockaddr_in6);
242 	s.sin6_addr = ip6->ip6_src;
243 	bzero(&d, sizeof d);
244 	d.sin6_family = AF_INET6;
245 	d.sin6_len = sizeof(struct sockaddr_in6);
246 	d.sin6_addr = ip6->ip6_dst;
247 
248 	match = NULL;
249 	matchprio = 0;
250 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
251 		if (ep->af != AF_INET6)
252 			continue;
253 		if (ep->proto >= 0 && ep->proto != proto)
254 			continue;
255 		if (ep->func)
256 			prio = (*ep->func)(m, *offp, proto, ep->arg);
257 		else {
258 			/*
259 			 * it's inbound traffic, we need to match in reverse
260 			 * order
261 			 */
262 			prio = mask_match(ep, (struct sockaddr *)&d,
263 			    (struct sockaddr *)&s);
264 		}
265 
266 		/* see encap4_input() for issues here */
267 		if (prio <= 0)
268 			continue;
269 		if (prio > matchprio) {
270 			matchprio = prio;
271 			match = ep;
272 		}
273 	}
274 
275 	if (match) {
276 		/* found a match */
277 		psw = match->psw;
278 		if (psw && psw->pr_input) {
279 			encap_fillarg(m, match);
280 			return (*psw->pr_input)(mp, offp, proto);
281 		} else {
282 			m_freem(m);
283 			return IPPROTO_DONE;
284 		}
285 	}
286 
287 	/* last resort: inject to raw socket */
288 	return rip6_input(mp, offp, proto);
289 }
290 #endif
291 
292 static void
293 encap_add(struct encaptab *ep)
294 {
295 
296 	LIST_INSERT_HEAD(&encaptab, ep, chain);
297 }
298 
299 /*
300  * sp (src ptr) is always my side, and dp (dst ptr) is always remote side.
301  * length of mask (sm and dm) is assumed to be same as sp/dp.
302  * Return value will be necessary as input (cookie) for encap_detach().
303  */
304 const struct encaptab *
305 encap_attach(int af, int proto, const struct sockaddr *sp,
306 	     const struct sockaddr *sm, const struct sockaddr *dp,
307 	     const struct sockaddr *dm, const struct protosw *psw, void *arg)
308 {
309 	struct encaptab *ep;
310 
311 	crit_enter();
312 	/* sanity check on args */
313 	if (sp->sa_len > sizeof ep->src || dp->sa_len > sizeof ep->dst)
314 		goto fail;
315 	if (sp->sa_len != dp->sa_len)
316 		goto fail;
317 	if (af != sp->sa_family || af != dp->sa_family)
318 		goto fail;
319 
320 	/* check if anyone have already attached with exactly same config */
321 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
322 		if (ep->af != af)
323 			continue;
324 		if (ep->proto != proto)
325 			continue;
326 		if (ep->src.ss_len != sp->sa_len ||
327 		    bcmp(&ep->src, sp, sp->sa_len) != 0 ||
328 		    bcmp(&ep->srcmask, sm, sp->sa_len) != 0)
329 			continue;
330 		if (ep->dst.ss_len != dp->sa_len ||
331 		    bcmp(&ep->dst, dp, dp->sa_len) != 0 ||
332 		    bcmp(&ep->dstmask, dm, dp->sa_len) != 0)
333 			continue;
334 
335 		goto fail;
336 	}
337 
338 	ep = kmalloc(sizeof *ep, M_NETADDR, M_INTWAIT | M_ZERO | M_NULLOK);
339 	if (ep == NULL)
340 		goto fail;
341 
342 	ep->af = af;
343 	ep->proto = proto;
344 	bcopy(sp, &ep->src, sp->sa_len);
345 	bcopy(sm, &ep->srcmask, sp->sa_len);
346 	bcopy(dp, &ep->dst, dp->sa_len);
347 	bcopy(dm, &ep->dstmask, dp->sa_len);
348 	ep->psw = psw;
349 	ep->arg = arg;
350 
351 	encap_add(ep);
352 
353 	crit_exit();
354 	return ep;
355 
356 fail:
357 	crit_exit();
358 	return NULL;
359 }
360 
361 const struct encaptab *
362 encap_attach_func(int af, int proto,
363 		  int (*func)(const struct mbuf *, int, int, void *),
364 		  const struct protosw *psw, void *arg)
365 {
366 	struct encaptab *ep;
367 
368 	crit_enter();
369 	/* sanity check on args */
370 	if (!func)
371 		goto fail;
372 
373 	ep = kmalloc(sizeof *ep, M_NETADDR, M_INTWAIT | M_ZERO | M_NULLOK);
374 	if (ep == NULL)
375 		goto fail;
376 
377 	ep->af = af;
378 	ep->proto = proto;
379 	ep->func = func;
380 	ep->psw = psw;
381 	ep->arg = arg;
382 
383 	encap_add(ep);
384 
385 	crit_exit();
386 	return ep;
387 
388 fail:
389 	crit_exit();
390 	return NULL;
391 }
392 
393 int
394 encap_detach(const struct encaptab *cookie)
395 {
396 	const struct encaptab *ep = cookie;
397 	struct encaptab *p;
398 
399 	for (p = LIST_FIRST(&encaptab); p; p = LIST_NEXT(p, chain)) {
400 		if (p == ep) {
401 			LIST_REMOVE(p, chain);
402 			kfree(p, M_NETADDR);	/*XXX*/
403 			return 0;
404 		}
405 	}
406 
407 	return EINVAL;
408 }
409 
410 static int
411 mask_match(const struct encaptab *ep, const struct sockaddr *sp,
412 	   const struct sockaddr *dp)
413 {
414 	struct sockaddr_storage s;
415 	struct sockaddr_storage d;
416 	int i;
417 	const u_int8_t *p, *q;
418 	u_int8_t *r;
419 	int matchlen;
420 
421 	if (sp->sa_len > sizeof s || dp->sa_len > sizeof d)
422 		return 0;
423 	if (sp->sa_family != ep->af || dp->sa_family != ep->af)
424 		return 0;
425 	if (sp->sa_len != ep->src.ss_len || dp->sa_len != ep->dst.ss_len)
426 		return 0;
427 
428 	matchlen = 0;
429 
430 	p = (const u_int8_t *)sp;
431 	q = (const u_int8_t *)&ep->srcmask;
432 	r = (u_int8_t *)&s;
433 	for (i = 0 ; i < sp->sa_len; i++) {
434 		r[i] = p[i] & q[i];
435 		/* XXX estimate */
436 		matchlen += (q[i] ? 8 : 0);
437 	}
438 
439 	p = (const u_int8_t *)dp;
440 	q = (const u_int8_t *)&ep->dstmask;
441 	r = (u_int8_t *)&d;
442 	for (i = 0 ; i < dp->sa_len; i++) {
443 		r[i] = p[i] & q[i];
444 		/* XXX rough estimate */
445 		matchlen += (q[i] ? 8 : 0);
446 	}
447 
448 	/* need to overwrite len/family portion as we don't compare them */
449 	s.ss_len = sp->sa_len;
450 	s.ss_family = sp->sa_family;
451 	d.ss_len = dp->sa_len;
452 	d.ss_family = dp->sa_family;
453 
454 	if (bcmp(&s, &ep->src, ep->src.ss_len) == 0 &&
455 	    bcmp(&d, &ep->dst, ep->dst.ss_len) == 0) {
456 		return matchlen;
457 	} else
458 		return 0;
459 }
460 
461 static void
462 encap_fillarg(struct mbuf *m, const struct encaptab *ep)
463 {
464 	struct m_tag *tag;
465 
466 	tag = m_tag_get(PACKET_TAG_ENCAP, sizeof(void *), M_NOWAIT);
467 	if (tag != NULL) {
468 		*(void **)m_tag_data(tag) = ep->arg;
469 		m_tag_prepend(m, tag);
470 	}
471 }
472 
473 void *
474 encap_getarg(struct mbuf *m)
475 {
476 	void *p = NULL;
477 	struct m_tag *tag;
478 
479 	tag = m_tag_find(m, PACKET_TAG_ENCAP, NULL);
480 	if (tag != NULL) {
481 		p = *(void **)m_tag_data(tag);
482 		m_tag_delete(m, tag);
483 	}
484 	return p;
485 }
486