1/*	$KAME: ip_encap.c,v 1.41 2001/03/15 08:35:08 itojun Exp $	*/
2
3/*-
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 *    notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in the
14 *    documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the project nor the names of its contributors
16 *    may be used to endorse or promote products derived from this software
17 *    without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 */
31/*
32 * My grandfather said that there's a devil inside tunnelling technology...
33 *
34 * We have surprisingly many protocols that want packets with IP protocol
35 * #4 or #41.  Here's a list of protocols that want protocol #41:
36 *	RFC1933 configured tunnel
37 *	RFC1933 automatic tunnel
38 *	RFC2401 IPsec tunnel
39 *	RFC2473 IPv6 generic packet tunnelling
40 *	RFC2529 6over4 tunnel
41 *	mobile-ip6 (uses RFC2473)
42 *	RFC3056 6to4 tunnel
43 *	isatap 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 <sys/cdefs.h>
60__FBSDID("$FreeBSD$");
61
62#include "opt_mrouting.h"
63#include "opt_inet.h"
64#include "opt_inet6.h"
65
66#include <sys/param.h>
67#include <sys/systm.h>
68#include <sys/socket.h>
69#include <sys/sockio.h>
70#include <sys/mbuf.h>
71#include <sys/errno.h>
72#include <sys/protosw.h>
73#include <sys/queue.h>
74
75#include <net/if.h>
76#include <net/route.h>
77
78#include <netinet/in.h>
79#include <netinet/in_systm.h>
80#include <netinet/ip.h>
81#include <netinet/ip_var.h>
82#include <netinet/ip_encap.h>
83
84#ifdef INET6
85#include <netinet/ip6.h>
86#include <netinet6/ip6_var.h>
87#include <netinet6/ip6protosw.h>
88#endif
89
90#include <machine/stdarg.h>
91
92#include <sys/kernel.h>
93#include <sys/malloc.h>
94static MALLOC_DEFINE(M_NETADDR, "encap_export_host", "Export host address structure");
95
96static void encap_add(struct encaptab *);
97static int mask_match(const struct encaptab *, const struct sockaddr *,
98		const struct sockaddr *);
99static void encap_fillarg(struct mbuf *, const struct encaptab *);
100
101/*
102 * All global variables in ip_encap.c are locked using encapmtx.
103 */
104static struct mtx encapmtx;
105MTX_SYSINIT(encapmtx, &encapmtx, "encapmtx", MTX_DEF);
106LIST_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 */
112void
113encap_init(void)
114{
115}
116
117#ifdef INET
118void
119encap4_input(struct mbuf *m, int off)
120{
121	struct ip *ip;
122	int proto;
123	struct sockaddr_in s, d;
124	const struct protosw *psw;
125	struct encaptab *ep, *match;
126	int prio, matchprio;
127
128	ip = mtod(m, struct ip *);
129	proto = ip->ip_p;
130
131	bzero(&s, sizeof(s));
132	s.sin_family = AF_INET;
133	s.sin_len = sizeof(struct sockaddr_in);
134	s.sin_addr = ip->ip_src;
135	bzero(&d, sizeof(d));
136	d.sin_family = AF_INET;
137	d.sin_len = sizeof(struct sockaddr_in);
138	d.sin_addr = ip->ip_dst;
139
140	match = NULL;
141	matchprio = 0;
142	mtx_lock(&encapmtx);
143	LIST_FOREACH(ep, &encaptab, chain) {
144		if (ep->af != AF_INET)
145			continue;
146		if (ep->proto >= 0 && ep->proto != proto)
147			continue;
148		if (ep->func)
149			prio = (*ep->func)(m, off, proto, ep->arg);
150		else {
151			/*
152			 * it's inbound traffic, we need to match in reverse
153			 * order
154			 */
155			prio = mask_match(ep, (struct sockaddr *)&d,
156			    (struct sockaddr *)&s);
157		}
158
159		/*
160		 * We prioritize the matches by using bit length of the
161		 * matches.  mask_match() and user-supplied matching function
162		 * should return the bit length of the matches (for example,
163		 * if both src/dst are matched for IPv4, 64 should be returned).
164		 * 0 or negative return value means "it did not match".
165		 *
166		 * The question is, since we have two "mask" portion, we
167		 * cannot really define total order between entries.
168		 * For example, which of these should be preferred?
169		 * mask_match() returns 48 (32 + 16) for both of them.
170		 *	src=3ffe::/16, dst=3ffe:501::/32
171		 *	src=3ffe:501::/32, dst=3ffe::/16
172		 *
173		 * We need to loop through all the possible candidates
174		 * to get the best match - the search takes O(n) for
175		 * n attachments (i.e. interfaces).
176		 */
177		if (prio <= 0)
178			continue;
179		if (prio > matchprio) {
180			matchprio = prio;
181			match = ep;
182		}
183	}
184	mtx_unlock(&encapmtx);
185
186	if (match) {
187		/* found a match, "match" has the best one */
188		psw = match->psw;
189		if (psw && psw->pr_input) {
190			encap_fillarg(m, match);
191			(*psw->pr_input)(m, off);
192		} else
193			m_freem(m);
194		return;
195	}
196
197	/* last resort: inject to raw socket */
198	rip_input(m, off);
199}
200#endif
201
202#ifdef INET6
203int
204encap6_input(struct mbuf **mp, int *offp, int proto)
205{
206	struct mbuf *m = *mp;
207	struct ip6_hdr *ip6;
208	struct sockaddr_in6 s, d;
209	const struct ip6protosw *psw;
210	struct encaptab *ep, *match;
211	int prio, matchprio;
212
213	ip6 = mtod(m, struct ip6_hdr *);
214
215	bzero(&s, sizeof(s));
216	s.sin6_family = AF_INET6;
217	s.sin6_len = sizeof(struct sockaddr_in6);
218	s.sin6_addr = ip6->ip6_src;
219	bzero(&d, sizeof(d));
220	d.sin6_family = AF_INET6;
221	d.sin6_len = sizeof(struct sockaddr_in6);
222	d.sin6_addr = ip6->ip6_dst;
223
224	match = NULL;
225	matchprio = 0;
226	mtx_lock(&encapmtx);
227	LIST_FOREACH(ep, &encaptab, chain) {
228		if (ep->af != AF_INET6)
229			continue;
230		if (ep->proto >= 0 && ep->proto != proto)
231			continue;
232		if (ep->func)
233			prio = (*ep->func)(m, *offp, proto, ep->arg);
234		else {
235			/*
236			 * it's inbound traffic, we need to match in reverse
237			 * order
238			 */
239			prio = mask_match(ep, (struct sockaddr *)&d,
240			    (struct sockaddr *)&s);
241		}
242
243		/* see encap4_input() for issues here */
244		if (prio <= 0)
245			continue;
246		if (prio > matchprio) {
247			matchprio = prio;
248			match = ep;
249		}
250	}
251	mtx_unlock(&encapmtx);
252
253	if (match) {
254		/* found a match */
255		psw = (const struct ip6protosw *)match->psw;
256		if (psw && psw->pr_input) {
257			encap_fillarg(m, match);
258			return (*psw->pr_input)(mp, offp, proto);
259		} else {
260			m_freem(m);
261			return IPPROTO_DONE;
262		}
263	}
264
265	/* last resort: inject to raw socket */
266	return rip6_input(mp, offp, proto);
267}
268#endif
269
270/*lint -sem(encap_add, custodial(1)) */
271static void
272encap_add(struct encaptab *ep)
273{
274
275	mtx_assert(&encapmtx, MA_OWNED);
276	LIST_INSERT_HEAD(&encaptab, ep, chain);
277}
278
279/*
280 * sp (src ptr) is always my side, and dp (dst ptr) is always remote side.
281 * length of mask (sm and dm) is assumed to be same as sp/dp.
282 * Return value will be necessary as input (cookie) for encap_detach().
283 */
284const struct encaptab *
285encap_attach(int af, int proto, const struct sockaddr *sp,
286    const struct sockaddr *sm, const struct sockaddr *dp,
287    const struct sockaddr *dm, const struct protosw *psw, void *arg)
288{
289	struct encaptab *ep;
290
291	/* sanity check on args */
292	if (sp->sa_len > sizeof(ep->src) || dp->sa_len > sizeof(ep->dst))
293		return (NULL);
294	if (sp->sa_len != dp->sa_len)
295		return (NULL);
296	if (af != sp->sa_family || af != dp->sa_family)
297		return (NULL);
298
299	/* check if anyone have already attached with exactly same config */
300	mtx_lock(&encapmtx);
301	LIST_FOREACH(ep, &encaptab, chain) {
302		if (ep->af != af)
303			continue;
304		if (ep->proto != proto)
305			continue;
306		if (ep->src.ss_len != sp->sa_len ||
307		    bcmp(&ep->src, sp, sp->sa_len) != 0 ||
308		    bcmp(&ep->srcmask, sm, sp->sa_len) != 0)
309			continue;
310		if (ep->dst.ss_len != dp->sa_len ||
311		    bcmp(&ep->dst, dp, dp->sa_len) != 0 ||
312		    bcmp(&ep->dstmask, dm, dp->sa_len) != 0)
313			continue;
314
315		mtx_unlock(&encapmtx);
316		return (NULL);
317	}
318
319	ep = malloc(sizeof(*ep), M_NETADDR, M_NOWAIT);	/*XXX*/
320	if (ep == NULL) {
321		mtx_unlock(&encapmtx);
322		return (NULL);
323	}
324	bzero(ep, sizeof(*ep));
325
326	ep->af = af;
327	ep->proto = proto;
328	bcopy(sp, &ep->src, sp->sa_len);
329	bcopy(sm, &ep->srcmask, sp->sa_len);
330	bcopy(dp, &ep->dst, dp->sa_len);
331	bcopy(dm, &ep->dstmask, dp->sa_len);
332	ep->psw = psw;
333	ep->arg = arg;
334
335	encap_add(ep);
336	mtx_unlock(&encapmtx);
337	return (ep);
338}
339
340const struct encaptab *
341encap_attach_func(int af, int proto,
342    int (*func)(const struct mbuf *, int, int, void *),
343    const struct protosw *psw, void *arg)
344{
345	struct encaptab *ep;
346
347	/* sanity check on args */
348	if (!func)
349		return (NULL);
350
351	ep = malloc(sizeof(*ep), M_NETADDR, M_NOWAIT);	/*XXX*/
352	if (ep == NULL)
353		return (NULL);
354	bzero(ep, sizeof(*ep));
355
356	ep->af = af;
357	ep->proto = proto;
358	ep->func = func;
359	ep->psw = psw;
360	ep->arg = arg;
361
362	mtx_lock(&encapmtx);
363	encap_add(ep);
364	mtx_unlock(&encapmtx);
365	return (ep);
366}
367
368int
369encap_detach(const struct encaptab *cookie)
370{
371	const struct encaptab *ep = cookie;
372	struct encaptab *p;
373
374	mtx_lock(&encapmtx);
375	LIST_FOREACH(p, &encaptab, chain) {
376		if (p == ep) {
377			LIST_REMOVE(p, chain);
378			mtx_unlock(&encapmtx);
379			free(p, M_NETADDR);	/*XXX*/
380			return 0;
381		}
382	}
383	mtx_unlock(&encapmtx);
384
385	return EINVAL;
386}
387
388static int
389mask_match(const struct encaptab *ep, const struct sockaddr *sp,
390    const struct sockaddr *dp)
391{
392	struct sockaddr_storage s;
393	struct sockaddr_storage d;
394	int i;
395	const u_int8_t *p, *q;
396	u_int8_t *r;
397	int matchlen;
398
399	if (sp->sa_len > sizeof(s) || dp->sa_len > sizeof(d))
400		return 0;
401	if (sp->sa_family != ep->af || dp->sa_family != ep->af)
402		return 0;
403	if (sp->sa_len != ep->src.ss_len || dp->sa_len != ep->dst.ss_len)
404		return 0;
405
406	matchlen = 0;
407
408	p = (const u_int8_t *)sp;
409	q = (const u_int8_t *)&ep->srcmask;
410	r = (u_int8_t *)&s;
411	for (i = 0 ; i < sp->sa_len; i++) {
412		r[i] = p[i] & q[i];
413		/* XXX estimate */
414		matchlen += (q[i] ? 8 : 0);
415	}
416
417	p = (const u_int8_t *)dp;
418	q = (const u_int8_t *)&ep->dstmask;
419	r = (u_int8_t *)&d;
420	for (i = 0 ; i < dp->sa_len; i++) {
421		r[i] = p[i] & q[i];
422		/* XXX rough estimate */
423		matchlen += (q[i] ? 8 : 0);
424	}
425
426	/* need to overwrite len/family portion as we don't compare them */
427	s.ss_len = sp->sa_len;
428	s.ss_family = sp->sa_family;
429	d.ss_len = dp->sa_len;
430	d.ss_family = dp->sa_family;
431
432	if (bcmp(&s, &ep->src, ep->src.ss_len) == 0 &&
433	    bcmp(&d, &ep->dst, ep->dst.ss_len) == 0) {
434		return matchlen;
435	} else
436		return 0;
437}
438
439static void
440encap_fillarg(struct mbuf *m, const struct encaptab *ep)
441{
442	struct m_tag *tag;
443
444	tag = m_tag_get(PACKET_TAG_ENCAP, sizeof (void*), M_NOWAIT);
445	if (tag) {
446		*(void**)(tag+1) = ep->arg;
447		m_tag_prepend(m, tag);
448	}
449}
450
451void *
452encap_getarg(struct mbuf *m)
453{
454	void *p = NULL;
455	struct m_tag *tag;
456
457	tag = m_tag_find(m, PACKET_TAG_ENCAP, NULL);
458	if (tag) {
459		p = *(void**)(tag+1);
460		m_tag_delete(m, tag);
461	}
462	return p;
463}
464