xref: /netbsd-src/sys/netipsec/ipsec_netbsd.c (revision 73704c4ce4ee2a60eb617e693ce7e9f03902613e)
1 /*	$NetBSD: ipsec_netbsd.c,v 1.3 2003/10/06 22:05:15 tls Exp $	*/
2 /*	$KAME: esp_input.c,v 1.60 2001/09/04 08:43:19 itojun Exp $	*/
3 /*	$KAME: ah_input.c,v 1.64 2001/09/04 08:43:19 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 #include <sys/cdefs.h>
35 __KERNEL_RCSID(0, "$NetBSD: ipsec_netbsd.c,v 1.3 2003/10/06 22:05:15 tls Exp $");
36 
37 #include "opt_inet.h"
38 #include "opt_ipsec.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/malloc.h>
43 #include <sys/mbuf.h>
44 #include <sys/domain.h>
45 #include <sys/protosw.h>
46 #include <sys/socket.h>
47 #include <sys/errno.h>
48 #include <sys/time.h>
49 #include <sys/kernel.h>
50 #include <sys/sysctl.h>
51 
52 #include <net/if.h>
53 #include <net/route.h>
54 #include <net/netisr.h>
55 #include <machine/cpu.h>
56 
57 #include <netinet/in.h>
58 #include <netinet/in_systm.h>
59 #include <netinet/in_var.h>
60 #include <netinet/ip.h>
61 #include <netinet/ip_var.h>
62 #include <netinet/ip_ecn.h>
63 #include <netinet/ip_icmp.h>
64 
65 #ifdef IPSEC
66 #include <netkey/key.h>
67 #include <netkey/keydb.h>
68 #include <netkey/key_debug.h>
69 #endif
70 
71 #ifdef FAST_IPSEC
72 #include <netipsec/ipsec.h>
73 #include <netipsec/key.h>
74 #include <netipsec/keydb.h>
75 #include <netipsec/key_debug.h>
76 #include <netipsec/ah_var.h>
77 #endif
78 
79 
80 #include <machine/stdarg.h>
81 
82 
83 
84 #include <netipsec/key.h>
85 
86 /* assumes that ip header and ah header are contiguous on mbuf */
87 void *
88 ah4_ctlinput(cmd, sa, v)
89 	int cmd;
90 	struct sockaddr *sa;
91 	void *v;
92 {
93 	struct ip *ip = v;
94 	struct ah *ah;
95 	struct icmp *icp;
96 	struct secasvar *sav;
97 
98 	if (sa->sa_family != AF_INET ||
99 	    sa->sa_len != sizeof(struct sockaddr_in))
100 		return NULL;
101 	if ((unsigned)cmd >= PRC_NCMDS)
102 		return NULL;
103 #ifndef notyet
104 	(void) ip; (void) ah; (void) icp; (void) sav;
105 #else
106 	if (cmd == PRC_MSGSIZE && ip_mtudisc && ip && ip->ip_v == 4) {
107 		/*
108 		 * Check to see if we have a valid SA corresponding to
109 		 * the address in the ICMP message payload.
110 		 */
111 		ah = (struct ah *)((caddr_t)ip + (ip->ip_hl << 2));
112 		if ((sav = key_allocsa(AF_INET,
113 				       (caddr_t) &ip->ip_src,
114 				       (caddr_t) &ip->ip_dst,
115 				       IPPROTO_AH, ah->ah_spi)) == NULL)
116 			return NULL;
117 		if (sav->state != SADB_SASTATE_MATURE &&
118 		    sav->state != SADB_SASTATE_DYING) {
119 			key_freesav(sav);
120 			return NULL;
121 		}
122 
123 		/* XXX Further validation? */
124 
125 		key_freesav(sav);
126 
127 		/*
128 		 * Now that we've validated that we are actually communicating
129 		 * with the host indicated in the ICMP message, locate the
130 		 * ICMP header, recalculate the new MTU, and create the
131 		 * corresponding routing entry.
132 		 */
133 		icp = (struct icmp *)((caddr_t)ip -
134 		    offsetof(struct icmp, icmp_ip));
135 		icmp_mtudisc(icp, ip->ip_dst);
136 
137 		return NULL;
138 	}
139 #endif
140 
141 	return NULL;
142 }
143 
144 /* assumes that ip header and esp header are contiguous on mbuf */
145 void *
146 esp4_ctlinput(cmd, sa, v)
147 	int cmd;
148 	struct sockaddr *sa;
149 	void *v;
150 {
151 	struct ip *ip = v;
152 	struct esp *esp;
153 	struct icmp *icp;
154 	struct secasvar *sav;
155 
156 	if (sa->sa_family != AF_INET ||
157 	    sa->sa_len != sizeof(struct sockaddr_in))
158 		return NULL;
159 	if ((unsigned)cmd >= PRC_NCMDS)
160 		return NULL;
161 #ifndef notyet
162 	(void) ip; (void) esp; (void) icp; (void) sav;
163 #else
164 	if (cmd == PRC_MSGSIZE && ip_mtudisc && ip && ip->ip_v == 4) {
165 		/*
166 		 * Check to see if we have a valid SA corresponding to
167 		 * the address in the ICMP message payload.
168 		 */
169 		esp = (struct esp *)((caddr_t)ip + (ip->ip_hl << 2));
170 		if ((sav = key_allocsa(AF_INET,
171 				       (caddr_t) &ip->ip_src,
172 				       (caddr_t) &ip->ip_dst,
173 				       IPPROTO_ESP, esp->esp_spi)) == NULL)
174 			return NULL;
175 		if (sav->state != SADB_SASTATE_MATURE &&
176 		    sav->state != SADB_SASTATE_DYING) {
177 			key_freesav(sav);
178 			return NULL;
179 		}
180 
181 		/* XXX Further validation? */
182 
183 		key_freesav(sav);
184 
185 		/*
186 		 * Now that we've validated that we are actually communicating
187 		 * with the host indicated in the ICMP message, locate the
188 		 * ICMP header, recalculate the new MTU, and create the
189 		 * corresponding routing entry.
190 		 */
191 		icp = (struct icmp *)((caddr_t)ip -
192 		    offsetof(struct icmp, icmp_ip));
193 		icmp_mtudisc(icp, ip->ip_dst);
194 
195 		return NULL;
196 	}
197 #endif
198 
199 	return NULL;
200 }
201 
202 #ifdef INET6
203 void
204 esp6_ctlinput(cmd, sa, d)
205 	int cmd;
206 	struct sockaddr *sa;
207 	void *d;
208 {
209 	const struct newesp *espp;
210 	struct newesp esp;
211 	struct ip6ctlparam *ip6cp = NULL, ip6cp1;
212 	struct secasvar *sav;
213 	struct ip6_hdr *ip6;
214 	struct mbuf *m;
215 	int off;
216 	struct sockaddr_in6 *sa6_src, *sa6_dst;
217 
218 	if (sa->sa_family != AF_INET6 ||
219 	    sa->sa_len != sizeof(struct sockaddr_in6))
220 		return;
221 	if ((unsigned)cmd >= PRC_NCMDS)
222 		return;
223 
224 	/* if the parameter is from icmp6, decode it. */
225 	if (d != NULL) {
226 		ip6cp = (struct ip6ctlparam *)d;
227 		m = ip6cp->ip6c_m;
228 		ip6 = ip6cp->ip6c_ip6;
229 		off = ip6cp->ip6c_off;
230 	} else {
231 		m = NULL;
232 		ip6 = NULL;
233 	}
234 
235 	if (ip6) {
236 		/*
237 		 * Notify the error to all possible sockets via pfctlinput2.
238 		 * Since the upper layer information (such as protocol type,
239 		 * source and destination ports) is embedded in the encrypted
240 		 * data and might have been cut, we can't directly call
241 		 * an upper layer ctlinput function. However, the pcbnotify
242 		 * function will consider source and destination addresses
243 		 * as well as the flow info value, and may be able to find
244 		 * some PCB that should be notified.
245 		 * Although pfctlinput2 will call esp6_ctlinput(), there is
246 		 * no possibility of an infinite loop of function calls,
247 		 * because we don't pass the inner IPv6 header.
248 		 */
249 		bzero(&ip6cp1, sizeof(ip6cp1));
250 		ip6cp1.ip6c_src = ip6cp->ip6c_src;
251 		pfctlinput2(cmd, sa, (void *)&ip6cp1);
252 
253 		/*
254 		 * Then go to special cases that need ESP header information.
255 		 * XXX: We assume that when ip6 is non NULL,
256 		 * M and OFF are valid.
257 		 */
258 
259 		/* check if we can safely examine src and dst ports */
260 		if (m->m_pkthdr.len < off + sizeof(esp))
261 			return;
262 
263 		if (m->m_len < off + sizeof(esp)) {
264 			/*
265 			 * this should be rare case,
266 			 * so we compromise on this copy...
267 			 */
268 			m_copydata(m, off, sizeof(esp), (caddr_t)&esp);
269 			espp = &esp;
270 		} else
271 			espp = (struct newesp*)(mtod(m, caddr_t) + off);
272 
273 		if (cmd == PRC_MSGSIZE) {
274 			int valid = 0;
275 
276 			/*
277 			 * Check to see if we have a valid SA corresponding to
278 			 * the address in the ICMP message payload.
279 			 */
280 			sa6_src = ip6cp->ip6c_src;
281 			sa6_dst = (struct sockaddr_in6 *)sa;
282 			sav = key_allocsa(AF_INET6,
283 					  (caddr_t)&sa6_src->sin6_addr,
284 					  (caddr_t)&sa6_dst->sin6_addr,
285 					  IPPROTO_ESP, espp->esp_spi);
286 			if (sav) {
287 				if (sav->state == SADB_SASTATE_MATURE ||
288 				    sav->state == SADB_SASTATE_DYING)
289 					valid++;
290 				key_freesav(sav);
291 			}
292 
293 			/* XXX Further validation? */
294 
295 			/*
296 			 * Depending on the value of "valid" and routing table
297 			 * size (mtudisc_{hi,lo}wat), we will:
298 			 * - recalcurate the new MTU and create the
299 			 *   corresponding routing entry, or
300 			 * - ignore the MTU change notification.
301 			 */
302 			icmp6_mtudisc_update((struct ip6ctlparam *)d, valid);
303 		}
304 	} else {
305 		/* we normally notify any pcb here */
306 	}
307 }
308 #endif /* INET6 */
309 
310 
311 /*FIXME: placebo for invalpcbcacheall. Fast-IPsec has no pcb cache? */
312 
313 void	ipsec_invalpcbcacheall(void);
314 void
315 ipsec_invalpcbcacheall(void)
316 {
317 }
318 
319 /* XXX will need a different oid at parent */
320 int
321 ipsec_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
322 	int *name;
323 	u_int namelen;
324 	void *oldp;
325 	size_t *oldlenp;
326 	void *newp;
327 	size_t newlen;
328 {
329 	/* All sysctl names at this level are terminal. */
330 	if (namelen != 1)
331 		return ENOTDIR;
332 
333 	/* common sanity checks */
334 	switch (name[0]) {
335 	case IPSECCTL_DEF_ESP_TRANSLEV:
336 	case IPSECCTL_DEF_ESP_NETLEV:
337 	case IPSECCTL_DEF_AH_TRANSLEV:
338 	case IPSECCTL_DEF_AH_NETLEV:
339 		if (newp != NULL && newlen == sizeof(int)) {
340 			switch (*(int *)newp) {
341 			case IPSEC_LEVEL_USE:
342 			case IPSEC_LEVEL_REQUIRE:
343 				break;
344 			default:
345 				return EINVAL;
346 			}
347 		}
348 		break;
349 	case IPSECCTL_DEF_POLICY:
350 		if (newp != NULL && newlen == sizeof(int)) {
351 			switch (*(int *)newp) {
352 			case IPSEC_POLICY_DISCARD:
353 			case IPSEC_POLICY_NONE:
354 				break;
355 			default:
356 				return EINVAL;
357 			}
358 			ipsec_invalpcbcacheall();
359 		}
360 		break;
361 	}
362 
363 	switch (name[0]) {
364 	case IPSECCTL_STATS:
365 		return sysctl_struct(oldp, oldlenp, newp, newlen,
366 		    &ipsecstat, sizeof(ipsecstat));
367 	case IPSECCTL_DEF_POLICY:
368 		return sysctl_int(oldp, oldlenp, newp, newlen,
369 		    &ip4_def_policy.policy);
370 	case IPSECCTL_DEF_ESP_TRANSLEV:
371 		if (newp != NULL)
372 			ipsec_invalpcbcacheall();
373 		return sysctl_int(oldp, oldlenp, newp, newlen,
374 		    &ip4_esp_trans_deflev);
375 	case IPSECCTL_DEF_ESP_NETLEV:
376 		if (newp != NULL)
377 			ipsec_invalpcbcacheall();
378 		return sysctl_int(oldp, oldlenp, newp, newlen,
379 		    &ip4_esp_net_deflev);
380 	case IPSECCTL_DEF_AH_TRANSLEV:
381 		if (newp != NULL)
382 			ipsec_invalpcbcacheall();
383 		return sysctl_int(oldp, oldlenp, newp, newlen,
384 		    &ip4_ah_trans_deflev);
385 	case IPSECCTL_DEF_AH_NETLEV:
386 		if (newp != NULL)
387 			ipsec_invalpcbcacheall();
388 		return sysctl_int(oldp, oldlenp, newp, newlen,
389 		    &ip4_ah_net_deflev);
390 	case IPSECCTL_AH_CLEARTOS:
391 		return sysctl_int(oldp, oldlenp, newp, newlen,
392 		  &/*ip4_*/ah_cleartos);
393 	case IPSECCTL_AH_OFFSETMASK:
394 		return sysctl_int(oldp, oldlenp, newp, newlen,
395 		    &ip4_ah_offsetmask);
396 	case IPSECCTL_DFBIT:
397 		return sysctl_int(oldp, oldlenp, newp, newlen,
398 		    &ip4_ipsec_dfbit);
399 	case IPSECCTL_ECN:
400 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip4_ipsec_ecn);
401 	case IPSECCTL_DEBUG:
402 		return sysctl_int(oldp, oldlenp, newp, newlen, &ipsec_debug);
403 	default:
404 		return EOPNOTSUPP;
405 	}
406 	/* NOTREACHED */
407 }
408