ipsecif.c revision 1.4 1 /* $NetBSD: ipsecif.c,v 1.4 2018/03/09 11:05:21 knakahara Exp $ */
2
3 /*
4 * Copyright (c) 2017 Internet Initiative Japan Inc.
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 *
16 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
17 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
18 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
19 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
20 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
21 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
22 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
23 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
24 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
25 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
26 * POSSIBILITY OF SUCH DAMAGE.
27 */
28
29 #include <sys/cdefs.h>
30 __KERNEL_RCSID(0, "$NetBSD: ipsecif.c,v 1.4 2018/03/09 11:05:21 knakahara Exp $");
31
32 #ifdef _KERNEL_OPT
33 #include "opt_inet.h"
34 #include "opt_ipsec.h"
35 #endif
36
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/socket.h>
40 #include <sys/sockio.h>
41 #include <sys/mbuf.h>
42 #include <sys/errno.h>
43 #include <sys/ioctl.h>
44 #include <sys/syslog.h>
45 #include <sys/kernel.h>
46
47 #include <net/if.h>
48 #include <net/route.h>
49
50 #include <netinet/in.h>
51 #include <netinet/in_systm.h>
52 #include <netinet/ip.h>
53 #include <netinet/ip_var.h>
54 #include <netinet/in_var.h>
55 #include <netinet/ip_encap.h>
56 #include <netinet/ip_ecn.h>
57 #include <netinet/ip_private.h>
58 #include <netinet/udp.h>
59
60 #ifdef INET6
61 #include <netinet/ip6.h>
62 #include <netinet6/ip6_var.h>
63 #include <netinet6/ip6_private.h>
64 #include <netinet6/in6_var.h>
65 #include <netinet6/ip6protosw.h> /* for struct ip6ctlparam */
66 #include <netinet/ip_ecn.h>
67 #endif
68
69 #include <netipsec/key.h>
70 #include <netipsec/ipsecif.h>
71
72 #include <net/if_ipsec.h>
73
74 static void ipsecif4_input(struct mbuf *, int, int, void *);
75 static int ipsecif4_output(struct ipsec_variant *, int, struct mbuf *);
76 static int ipsecif4_filter4(const struct ip *, struct ipsec_variant *,
77 struct ifnet *);
78
79 #ifdef INET6
80 static int ipsecif6_input(struct mbuf **, int *, int, void *);
81 static int ipsecif6_output(struct ipsec_variant *, int, struct mbuf *);
82 static int ipsecif6_filter6(const struct ip6_hdr *, struct ipsec_variant *,
83 struct ifnet *);
84 #endif
85
86 static int ip_ipsec_ttl = IPSEC_TTL;
87 static int ip_ipsec_copy_tos = 0;
88 #ifdef INET6
89 static int ip6_ipsec_hlim = IPSEC_HLIM;
90 static int ip6_ipsec_pmtu = 0; /* XXX: per interface configuration?? */
91 static int ip6_ipsec_copy_tos = 0;
92 #endif
93
94 struct encapsw ipsecif4_encapsw = {
95 .encapsw4 = {
96 .pr_input = ipsecif4_input,
97 .pr_ctlinput = NULL,
98 }
99 };
100
101 #ifdef INET6
102 static const struct encapsw ipsecif6_encapsw;
103 #endif
104
105 static struct mbuf *
106 ipsecif4_prepend_hdr(struct ipsec_variant *var, struct mbuf *m,
107 uint8_t proto, uint8_t tos)
108 {
109 struct ip *ip;
110 struct sockaddr_in *src, *dst;
111
112 src = satosin(var->iv_psrc);
113 dst = satosin(var->iv_pdst);
114
115 if (in_nullhost(src->sin_addr) || in_nullhost(src->sin_addr) ||
116 src->sin_addr.s_addr == INADDR_BROADCAST ||
117 dst->sin_addr.s_addr == INADDR_BROADCAST) {
118 m_freem(m);
119 return NULL;
120 }
121 m->m_flags &= ~M_BCAST;
122
123 if (IN_MULTICAST(src->sin_addr.s_addr) ||
124 IN_MULTICAST(dst->sin_addr.s_addr)) {
125 m_freem(m);
126 return NULL;
127 }
128
129 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
130 if (m && M_UNWRITABLE(m, sizeof(struct ip)))
131 m = m_pullup(m, sizeof(struct ip));
132 if (m == NULL)
133 return NULL;
134
135 ip = mtod(m, struct ip *);
136 ip->ip_v = IPVERSION;
137 ip->ip_off = htons(0);
138 ip->ip_id = 0;
139 ip->ip_hl = sizeof(*ip) >> 2;
140 if (ip_ipsec_copy_tos)
141 ip->ip_tos = tos;
142 else
143 ip->ip_tos = 0;
144 ip->ip_sum = 0;
145 ip->ip_src = src->sin_addr;
146 ip->ip_dst = dst->sin_addr;
147 ip->ip_p = proto;
148 ip->ip_ttl = ip_ipsec_ttl;
149 ip->ip_len = htons(m->m_pkthdr.len);
150 #ifndef IPSEC_TX_TOS_CLEAR
151 struct ifnet *ifp = &var->iv_softc->ipsec_if;
152 if (ifp->if_flags & IFF_ECN)
153 ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
154 else
155 ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
156 #endif
157
158 return m;
159 }
160
161 static int
162 ipsecif4_needfrag(struct mbuf *m, struct ipsecrequest *isr)
163 {
164 struct ip ip0;
165 struct ip *ip;
166 int mtu;
167 struct secasvar *sav;
168
169 sav = key_lookup_sa_bysaidx(&isr->saidx);
170 if (sav == NULL)
171 return 0;
172
173 if (!(sav->natt_type & UDP_ENCAP_ESPINUDP) &&
174 !(sav->natt_type & UDP_ENCAP_ESPINUDP_NON_IKE)) {
175 mtu = 0;
176 goto out;
177 }
178
179 if (m->m_len < sizeof(struct ip)) {
180 m_copydata(m, 0, sizeof(ip0), &ip0);
181 ip = &ip0;
182 } else {
183 ip = mtod(m, struct ip *);
184 }
185 mtu = sav->esp_frag;
186 if (ntohs(ip->ip_len) <= mtu)
187 mtu = 0;
188
189 out:
190 KEY_SA_UNREF(&sav);
191 return mtu;
192 }
193
194 static struct mbuf *
195 ipsecif4_flowinfo(struct mbuf *m, int family, int *proto0, u_int8_t *tos0)
196 {
197 const struct ip *ip;
198 int proto;
199 int tos;
200
201 KASSERT(proto0 != NULL);
202 KASSERT(tos0 != NULL);
203
204 switch (family) {
205 case AF_INET:
206 proto = IPPROTO_IPV4;
207 if (m->m_len < sizeof(*ip)) {
208 m = m_pullup(m, sizeof(*ip));
209 if (m == NULL) {
210 *tos0 = 0;
211 *proto0 = 0;
212 return NULL;
213 }
214 }
215 ip = mtod(m, const struct ip *);
216 tos = ip->ip_tos;
217 /* TODO: support ALTQ for innner packet */
218 break;
219 #ifdef INET6
220 case AF_INET6: {
221 const struct ip6_hdr *ip6;
222 proto = IPPROTO_IPV6;
223 if (m->m_len < sizeof(*ip6)) {
224 m = m_pullup(m, sizeof(*ip6));
225 if (m == NULL) {
226 *tos0 = 0;
227 *proto0 = 0;
228 return NULL;
229 }
230 }
231 ip6 = mtod(m, const struct ip6_hdr *);
232 tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
233 /* TODO: support ALTQ for innner packet */
234 break;
235 }
236 #endif /* INET6 */
237 default:
238 *tos0 = 0;
239 *proto0 = 0;
240 return NULL;
241 }
242
243 *proto0 = proto;
244 *tos0 = tos;
245 return m;
246 }
247
248 static int
249 ipsecif4_fragout(struct ipsec_variant *var, int family, struct mbuf *m, int mtu)
250 {
251 struct ifnet *ifp = &var->iv_softc->ipsec_if;
252 struct mbuf *next;
253 struct m_tag *mtag;
254 int error;
255
256 KASSERT(if_ipsec_heldref_variant(var));
257
258 mtag = m_tag_find(m, PACKET_TAG_IPSEC_NAT_T_PORTS, NULL);
259 if (mtag)
260 m_tag_delete(m, mtag);
261
262 /* consider new IP header prepended in ipsecif4_output() */
263 if (mtu <= sizeof(struct ip)) {
264 m_freem(m);
265 return ENETUNREACH;
266 }
267 m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
268 error = ip_fragment(m, ifp, mtu - sizeof(struct ip));
269 if (error)
270 return error;
271
272 for (error = 0; m; m = next) {
273 next = m->m_nextpkt;
274 m->m_nextpkt = NULL;
275 if (error) {
276 m_freem(m);
277 continue;
278 }
279
280 error = ipsecif4_output(var, family, m);
281 }
282 if (error == 0)
283 IP_STATINC(IP_STAT_FRAGMENTED);
284
285 return error;
286 }
287
288 int
289 ipsecif4_encap_func(struct mbuf *m, struct ip *ip, struct ipsec_variant *var)
290 {
291 struct m_tag *mtag;
292 struct sockaddr_in *src, *dst;
293 u_int16_t src_port = 0;
294 u_int16_t dst_port = 0;
295
296 KASSERT(var != NULL);
297
298 src = satosin(var->iv_psrc);
299 dst = satosin(var->iv_pdst);
300 mtag = m_tag_find(m, PACKET_TAG_IPSEC_NAT_T_PORTS, NULL);
301 if (mtag) {
302 u_int16_t *ports;
303
304 ports = (u_int16_t *)(mtag + 1);
305 src_port = ports[0];
306 dst_port = ports[1];
307 }
308
309 /* address match */
310 if (src->sin_addr.s_addr != ip->ip_dst.s_addr ||
311 dst->sin_addr.s_addr != ip->ip_src.s_addr)
312 return 0;
313
314 /* UDP encap? */
315 if (mtag == NULL && var->iv_sport == 0 && var->iv_dport == 0)
316 goto match;
317
318 /* port match */
319 if (src_port != var->iv_dport ||
320 dst_port != var->iv_sport) {
321 #ifdef DEBUG
322 printf("%s: port mismatch: pkt(%u, %u), if(%u, %u)\n",
323 __func__, ntohs(src_port), ntohs(dst_port),
324 ntohs(var->iv_sport), ntohs(var->iv_dport));
325 #endif
326 return 0;
327 }
328
329 match:
330 /*
331 * hide NAT-T information from encapsulated traffics.
332 * they don't know about IPsec.
333 */
334 if (mtag)
335 m_tag_delete(m, mtag);
336 return sizeof(src->sin_addr) + sizeof(dst->sin_addr);
337 }
338
339 static int
340 ipsecif4_output(struct ipsec_variant *var, int family, struct mbuf *m)
341 {
342 struct secpolicy *sp = NULL;
343 u_int8_t tos;
344 int proto;
345 int error;
346 int mtu;
347 u_long sa_mtu = 0;
348
349 KASSERT(if_ipsec_heldref_variant(var));
350 KASSERT(if_ipsec_variant_is_configured(var));
351 KASSERT(var->iv_psrc->sa_family == AF_INET);
352 KASSERT(var->iv_pdst->sa_family == AF_INET);
353
354 sp = IV_SP_OUT(var);
355 KASSERT(sp != NULL);
356 /*
357 * The SPs in ipsec_variant are prevented from freed by
358 * ipsec_variant->iv_psref. So, KEY_SP_REF() is unnecessary here.
359 */
360
361 KASSERT(sp->policy != IPSEC_POLICY_NONE);
362 KASSERT(sp->policy != IPSEC_POLICY_ENTRUST);
363 KASSERT(sp->policy != IPSEC_POLICY_BYPASS);
364 if (sp->policy != IPSEC_POLICY_IPSEC) {
365 struct ifnet *ifp = &var->iv_softc->ipsec_if;
366 m_freem(m);
367 IF_DROP(&ifp->if_snd);
368 return 0;
369 }
370
371 /* get flowinfo */
372 m = ipsecif4_flowinfo(m, family, &proto, &tos);
373 if (m == NULL) {
374 error = ENETUNREACH;
375 goto done;
376 }
377
378 /* prepend new IP header */
379 m = ipsecif4_prepend_hdr(var, m, proto, tos);
380 if (m == NULL) {
381 error = ENETUNREACH;
382 goto done;
383 }
384
385 /*
386 * Normal netipsec's NAT-T fragmentation is done in ip_output().
387 * See "natt_frag" processing.
388 * However, ipsec(4) interface's one is not done in the same way,
389 * so we must do NAT-T fragmentation by own code.
390 */
391 /* NAT-T ESP fragmentation */
392 mtu = ipsecif4_needfrag(m, sp->req);
393 if (mtu > 0)
394 return ipsecif4_fragout(var, family, m, mtu);
395
396 /* IPsec output */
397 IP_STATINC(IP_STAT_LOCALOUT);
398 error = ipsec4_process_packet(m, sp->req, &sa_mtu);
399 if (error == ENOENT)
400 error = 0;
401 /*
402 * frangmentation is already done in ipsecif4_fragout(),
403 * so ipsec4_process_packet() must not do fragmentation here.
404 */
405 KASSERT(sa_mtu == 0);
406
407 done:
408 return error;
409 }
410
411 #ifdef INET6
412 static int
413 ipsecif6_output(struct ipsec_variant *var, int family, struct mbuf *m)
414 {
415 struct ifnet *ifp = &var->iv_softc->ipsec_if;
416 struct ipsec_softc *sc = ifp->if_softc;
417 struct ipsec_ro *iro;
418 struct rtentry *rt;
419 struct sockaddr_in6 *sin6_src;
420 struct sockaddr_in6 *sin6_dst;
421 struct ip6_hdr *ip6;
422 int proto, error;
423 u_int8_t itos, otos;
424 union {
425 struct sockaddr dst;
426 struct sockaddr_in6 dst6;
427 } u;
428
429 KASSERT(if_ipsec_heldref_variant(var));
430 KASSERT(if_ipsec_variant_is_configured(var));
431
432 sin6_src = satosin6(var->iv_psrc);
433 sin6_dst = satosin6(var->iv_pdst);
434
435 KASSERT(sin6_src->sin6_family == AF_INET6);
436 KASSERT(sin6_dst->sin6_family == AF_INET6);
437
438 switch (family) {
439 #ifdef INET
440 case AF_INET:
441 {
442 struct ip *ip;
443
444 proto = IPPROTO_IPV4;
445 if (m->m_len < sizeof(*ip)) {
446 m = m_pullup(m, sizeof(*ip));
447 if (m == NULL)
448 return ENOBUFS;
449 }
450 ip = mtod(m, struct ip *);
451 itos = ip->ip_tos;
452 /* TODO: support ALTQ for innner packet */
453 break;
454 }
455 #endif /* INET */
456 case AF_INET6:
457 {
458 struct ip6_hdr *xip6;
459 proto = IPPROTO_IPV6;
460 if (m->m_len < sizeof(*xip6)) {
461 m = m_pullup(m, sizeof(*xip6));
462 if (m == NULL)
463 return ENOBUFS;
464 }
465 xip6 = mtod(m, struct ip6_hdr *);
466 itos = (ntohl(xip6->ip6_flow) >> 20) & 0xff;
467 /* TODO: support ALTQ for innner packet */
468 break;
469 }
470 default:
471 m_freem(m);
472 return EAFNOSUPPORT;
473 }
474
475 /* prepend new IP header */
476 M_PREPEND(m, sizeof(struct ip6_hdr), M_DONTWAIT);
477 if (m && M_UNWRITABLE(m, sizeof(struct ip6_hdr)))
478 m = m_pullup(m, sizeof(struct ip6_hdr));
479 if (m == NULL)
480 return ENOBUFS;
481
482 ip6 = mtod(m, struct ip6_hdr *);
483 ip6->ip6_flow = 0;
484 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
485 ip6->ip6_vfc |= IPV6_VERSION;
486 ip6->ip6_plen = htons((u_short)m->m_pkthdr.len);
487 ip6->ip6_nxt = proto;
488 ip6->ip6_hlim = ip6_ipsec_hlim;
489 ip6->ip6_src = sin6_src->sin6_addr;
490 /* bidirectional configured tunnel mode */
491 if (!IN6_IS_ADDR_UNSPECIFIED(&sin6_dst->sin6_addr)) {
492 ip6->ip6_dst = sin6_dst->sin6_addr;
493 } else {
494 m_freem(m);
495 return ENETUNREACH;
496 }
497 #ifndef IPSEC_TX_TOS_CLEAR
498 if (ifp->if_flags & IFF_ECN)
499 ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
500 else
501 ip_ecn_ingress(ECN_NOCARE, &otos, &itos);
502
503 if (!ip6_ipsec_copy_tos)
504 otos = 0;
505 #else
506 if (ip6_ipsec_copy_tos)
507 otos = itos;
508 else
509 otos = 0;
510 #endif
511 ip6->ip6_flow &= ~ntohl(0xff00000);
512 ip6->ip6_flow |= htonl((u_int32_t)otos << 20);
513
514 sockaddr_in6_init(&u.dst6, &sin6_dst->sin6_addr, 0, 0, 0);
515
516 iro = percpu_getref(sc->ipsec_ro_percpu);
517 mutex_enter(&iro->ir_lock);
518 if ((rt = rtcache_lookup(&iro->ir_ro, &u.dst)) == NULL) {
519 mutex_exit(&iro->ir_lock);
520 percpu_putref(sc->ipsec_ro_percpu);
521 m_freem(m);
522 return ENETUNREACH;
523 }
524
525 if (rt->rt_ifp == ifp) {
526 rtcache_unref(rt, &iro->ir_ro);
527 rtcache_free(&iro->ir_ro);
528 mutex_exit(&iro->ir_lock);
529 percpu_putref(sc->ipsec_ro_percpu);
530 m_freem(m);
531 return ENETUNREACH;
532 }
533 rtcache_unref(rt, &iro->ir_ro);
534
535 /*
536 * force fragmentation to minimum MTU, to avoid path MTU discovery.
537 * it is too painful to ask for resend of inner packet, to achieve
538 * path MTU discovery for encapsulated packets.
539 */
540 error = ip6_output(m, 0, &iro->ir_ro,
541 ip6_ipsec_pmtu ? 0 : IPV6_MINMTU, 0, NULL, NULL);
542 if (error)
543 rtcache_free(&iro->ir_ro);
544
545 mutex_exit(&iro->ir_lock);
546 percpu_putref(sc->ipsec_ro_percpu);
547
548 return error;
549 }
550 #endif /* INET6 */
551
552 static void
553 ipsecif4_input(struct mbuf *m, int off, int proto, void *eparg)
554 {
555 struct ifnet *ipsecp;
556 struct ipsec_softc *sc = eparg;
557 struct ipsec_variant *var;
558 const struct ip *ip;
559 int af;
560 #ifndef IPSEC_TX_TOS_CLEAR
561 u_int8_t otos;
562 #endif
563 struct psref psref_rcvif;
564 struct psref psref_var;
565 struct ifnet *rcvif;
566
567 KASSERT(sc != NULL);
568
569 ipsecp = &sc->ipsec_if;
570 if ((ipsecp->if_flags & IFF_UP) == 0) {
571 m_freem(m);
572 ip_statinc(IP_STAT_NOIPSEC);
573 return;
574 }
575
576 var = if_ipsec_getref_variant(sc, &psref_var);
577 if (if_ipsec_variant_is_unconfigured(var)) {
578 if_ipsec_putref_variant(var, &psref_var);
579 m_freem(m);
580 ip_statinc(IP_STAT_NOIPSEC);
581 return;
582 }
583
584 ip = mtod(m, const struct ip *);
585
586 rcvif = m_get_rcvif_psref(m, &psref_rcvif);
587 if (rcvif == NULL || !ipsecif4_filter4(ip, var, rcvif)) {
588 m_put_rcvif_psref(rcvif, &psref_rcvif);
589 if_ipsec_putref_variant(var, &psref_var);
590 m_freem(m);
591 ip_statinc(IP_STAT_NOIPSEC);
592 return;
593 }
594 m_put_rcvif_psref(rcvif, &psref_rcvif);
595 if_ipsec_putref_variant(var, &psref_var);
596 #ifndef IPSEC_TX_TOS_CLEAR
597 otos = ip->ip_tos;
598 #endif
599 m_adj(m, off);
600
601 switch (proto) {
602 case IPPROTO_IPV4:
603 {
604 struct ip *xip;
605 af = AF_INET;
606 if (M_UNWRITABLE(m, sizeof(*xip))) {
607 m = m_pullup(m, sizeof(*xip));
608 if (m == NULL)
609 return;
610 }
611 xip = mtod(m, struct ip *);
612 #ifndef IPSEC_TX_TOS_CLEAR
613 if (ipsecp->if_flags & IFF_ECN)
614 ip_ecn_egress(ECN_ALLOWED, &otos, &xip->ip_tos);
615 else
616 ip_ecn_egress(ECN_NOCARE, &otos, &xip->ip_tos);
617 #endif
618 break;
619 }
620 #ifdef INET6
621 case IPPROTO_IPV6:
622 {
623 struct ip6_hdr *ip6;
624 u_int8_t itos;
625 af = AF_INET6;
626 if (M_UNWRITABLE(m, sizeof(*ip6))) {
627 m = m_pullup(m, sizeof(*ip6));
628 if (m == NULL)
629 return;
630 }
631 ip6 = mtod(m, struct ip6_hdr *);
632 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
633 #ifndef IPSEC_TX_TOS_CLEAR
634 if (ipsecp->if_flags & IFF_ECN)
635 ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
636 else
637 ip_ecn_egress(ECN_NOCARE, &otos, &itos);
638 #endif
639 ip6->ip6_flow &= ~htonl(0xff << 20);
640 ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
641 break;
642 }
643 #endif /* INET6 */
644 default:
645 ip_statinc(IP_STAT_NOIPSEC);
646 m_freem(m);
647 return;
648 }
649 if_ipsec_input(m, af, ipsecp);
650
651 return;
652 }
653
654 /*
655 * validate and filter the pakcet
656 */
657 static int
658 ipsecif4_filter4(const struct ip *ip, struct ipsec_variant *var,
659 struct ifnet *ifp)
660 {
661 struct sockaddr_in *src, *dst;
662
663 src = satosin(var->iv_psrc);
664 dst = satosin(var->iv_pdst);
665
666 return in_tunnel_validate(ip, src->sin_addr, dst->sin_addr);
667 }
668
669 #ifdef INET6
670 static int
671 ipsecif6_input(struct mbuf **mp, int *offp, int proto, void *eparg)
672 {
673 struct mbuf *m = *mp;
674 struct ifnet *ipsecp;
675 struct ipsec_softc *sc = eparg;
676 struct ipsec_variant *var;
677 struct ip6_hdr *ip6;
678 int af = 0;
679 #ifndef IPSEC_TX_TOS_CLEAR
680 u_int32_t otos;
681 #endif
682 struct psref psref_rcvif;
683 struct psref psref_var;
684 struct ifnet *rcvif;
685
686 KASSERT(eparg != NULL);
687
688 ipsecp = &sc->ipsec_if;
689 if ((ipsecp->if_flags & IFF_UP) == 0) {
690 m_freem(m);
691 IP6_STATINC(IP6_STAT_NOIPSEC);
692 return IPPROTO_DONE;
693 }
694
695 var = if_ipsec_getref_variant(sc, &psref_var);
696 if (if_ipsec_variant_is_unconfigured(var)) {
697 if_ipsec_putref_variant(var, &psref_var);
698 m_freem(m);
699 IP6_STATINC(IP6_STAT_NOIPSEC);
700 return IPPROTO_DONE;
701 }
702
703 ip6 = mtod(m, struct ip6_hdr *);
704
705 rcvif = m_get_rcvif_psref(m, &psref_rcvif);
706 if (rcvif == NULL || !ipsecif6_filter6(ip6, var, rcvif)) {
707 m_put_rcvif_psref(rcvif, &psref_rcvif);
708 if_ipsec_putref_variant(var, &psref_var);
709 m_freem(m);
710 IP6_STATINC(IP6_STAT_NOIPSEC);
711 return IPPROTO_DONE;
712 }
713 m_put_rcvif_psref(rcvif, &psref_rcvif);
714 if_ipsec_putref_variant(var, &psref_var);
715
716 #ifndef IPSEC_TX_TOS_CLEAR
717 otos = ip6->ip6_flow;
718 #endif
719 m_adj(m, *offp);
720
721 switch (proto) {
722 #ifdef INET
723 case IPPROTO_IPV4:
724 {
725 af = AF_INET;
726 #ifndef IPSEC_TX_TOS_CLEAR
727 struct ip *ip;
728 u_int8_t otos8;
729 otos8 = (ntohl(otos) >> 20) & 0xff;
730
731 if (M_UNWRITABLE(m, sizeof(*ip))) {
732 m = m_pullup(m, sizeof(*ip));
733 if (m == NULL)
734 return IPPROTO_DONE;
735 }
736 ip = mtod(m, struct ip *);
737 if (ipsecp->if_flags & IFF_ECN)
738 ip_ecn_egress(ECN_ALLOWED, &otos8, &ip->ip_tos);
739 else
740 ip_ecn_egress(ECN_NOCARE, &otos8, &ip->ip_tos);
741 #endif
742 break;
743 }
744 #endif /* INET */
745 case IPPROTO_IPV6:
746 {
747 af = AF_INET6;
748 #ifndef IPSEC_TX_TOS_CLEAR
749 struct ip6_hdr *xip6;
750
751 if (M_UNWRITABLE(m, sizeof(*xip6))) {
752 m = m_pullup(m, sizeof(*xip6));
753 if (m == NULL)
754 return IPPROTO_DONE;
755 }
756 xip6 = mtod(m, struct ip6_hdr *);
757 if (ipsecp->if_flags & IFF_ECN)
758 ip6_ecn_egress(ECN_ALLOWED, &otos, &xip6->ip6_flow);
759 else
760 ip6_ecn_egress(ECN_NOCARE, &otos, &xip6->ip6_flow);
761 break;
762 #endif
763 }
764 default:
765 IP6_STATINC(IP6_STAT_NOIPSEC);
766 m_freem(m);
767 return IPPROTO_DONE;
768 }
769
770 if_ipsec_input(m, af, ipsecp);
771 return IPPROTO_DONE;
772 }
773
774 /*
775 * validate and filter the packet.
776 */
777 static int
778 ipsecif6_filter6(const struct ip6_hdr *ip6, struct ipsec_variant *var,
779 struct ifnet *ifp)
780 {
781 struct sockaddr_in6 *src, *dst;
782
783 src = satosin6(var->iv_psrc);
784 dst = satosin6(var->iv_pdst);
785
786 return in6_tunnel_validate(ip6, &src->sin6_addr, &dst->sin6_addr);
787 }
788 #endif /* INET6 */
789
790 int
791 ipsecif4_attach(struct ipsec_variant *var)
792 {
793 struct ipsec_softc *sc = var->iv_softc;
794
795 KASSERT(if_ipsec_variant_is_configured(var));
796
797 if (var->iv_encap_cookie4 != NULL)
798 return EALREADY;
799 var->iv_encap_cookie4 = encap_attach_func(AF_INET, -1, if_ipsec_encap_func,
800 &ipsecif4_encapsw, sc);
801 if (var->iv_encap_cookie4 == NULL)
802 return EEXIST;
803
804 var->iv_output = ipsecif4_output;
805 return 0;
806 }
807
808 int
809 ipsecif4_detach(struct ipsec_variant *var)
810 {
811 int error;
812
813 if (var->iv_encap_cookie4 == NULL)
814 return 0;
815
816 var->iv_output = NULL;
817 error = encap_detach(var->iv_encap_cookie4);
818 if (error == 0)
819 var->iv_encap_cookie4 = NULL;
820
821 return error;
822 }
823
824 #ifdef INET6
825 int
826 ipsecif6_attach(struct ipsec_variant *var)
827 {
828 struct sockaddr_in6 mask6;
829 struct ipsec_softc *sc = var->iv_softc;
830
831 KASSERT(if_ipsec_variant_is_configured(var));
832 KASSERT(var->iv_encap_cookie6 == NULL);
833
834 memset(&mask6, 0, sizeof(mask6));
835 mask6.sin6_len = sizeof(struct sockaddr_in6);
836 mask6.sin6_addr.s6_addr32[0] = mask6.sin6_addr.s6_addr32[1] =
837 mask6.sin6_addr.s6_addr32[2] = mask6.sin6_addr.s6_addr32[3] = ~0;
838
839 var->iv_encap_cookie6 = encap_attach(AF_INET6, -1,
840 var->iv_psrc, (struct sockaddr *)&mask6,
841 var->iv_pdst, (struct sockaddr *)&mask6,
842 &ipsecif6_encapsw, sc);
843 if (var->iv_encap_cookie6 == NULL)
844 return EEXIST;
845
846 var->iv_output = ipsecif6_output;
847 return 0;
848 }
849
850 static void
851 ipsecif6_rtcache_free_pc(void *p, void *arg __unused, struct cpu_info *ci __unused)
852 {
853 struct ipsec_ro *iro = p;
854
855 mutex_enter(&iro->ir_lock);
856 rtcache_free(&iro->ir_ro);
857 mutex_exit(&iro->ir_lock);
858 }
859
860 int
861 ipsecif6_detach(struct ipsec_variant *var)
862 {
863 struct ipsec_softc *sc = var->iv_softc;
864 int error;
865
866 KASSERT(var->iv_encap_cookie6 != NULL);
867
868 percpu_foreach(sc->ipsec_ro_percpu, ipsecif6_rtcache_free_pc, NULL);
869
870 var->iv_output = NULL;
871 error = encap_detach(var->iv_encap_cookie6);
872 if (error == 0)
873 var->iv_encap_cookie6 = NULL;
874 return error;
875 }
876
877 void *
878 ipsecif6_ctlinput(int cmd, const struct sockaddr *sa, void *d, void *eparg)
879 {
880 struct ipsec_softc *sc = eparg;
881 struct ip6ctlparam *ip6cp = NULL;
882 struct ip6_hdr *ip6;
883 const struct sockaddr_in6 *dst6;
884 struct ipsec_ro *iro;
885
886 if (sa->sa_family != AF_INET6 ||
887 sa->sa_len != sizeof(struct sockaddr_in6))
888 return NULL;
889
890 if ((unsigned)cmd >= PRC_NCMDS)
891 return NULL;
892 if (cmd == PRC_HOSTDEAD)
893 d = NULL;
894 else if (inet6ctlerrmap[cmd] == 0)
895 return NULL;
896
897 /* if the parameter is from icmp6, decode it. */
898 if (d != NULL) {
899 ip6cp = (struct ip6ctlparam *)d;
900 ip6 = ip6cp->ip6c_ip6;
901 } else {
902 ip6 = NULL;
903 }
904
905 if (!ip6)
906 return NULL;
907
908 iro = percpu_getref(sc->ipsec_ro_percpu);
909 mutex_enter(&iro->ir_lock);
910 dst6 = satocsin6(rtcache_getdst(&iro->ir_ro));
911 /* XXX scope */
912 if (dst6 == NULL)
913 ;
914 else if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst6->sin6_addr))
915 /* flush route cache */
916 rtcache_free(&iro->ir_ro);
917
918 mutex_exit(&iro->ir_lock);
919 percpu_putref(sc->ipsec_ro_percpu);
920
921 return NULL;
922 }
923
924 ENCAP_PR_WRAP_CTLINPUT(ipsecif6_ctlinput)
925 #define ipsecif6_ctlinput ipsecif6_ctlinput_wrapper
926
927 static const struct encapsw ipsecif6_encapsw = {
928 .encapsw6 = {
929 .pr_input = ipsecif6_input,
930 .pr_ctlinput = ipsecif6_ctlinput,
931 }
932 };
933 #endif /* INET6 */
934