ipsecif.c revision 1.15 1 /* $NetBSD: ipsecif.c,v 1.15 2019/04/12 07:12:12 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 1.15 2019/04/12 07:12:12 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 int ipsecif_set_natt_ports(struct ipsec_variant *, struct mbuf *);
75 static void ipsecif4_input(struct mbuf *, int, int, void *);
76 static int ipsecif4_output(struct ipsec_variant *, int, struct mbuf *);
77 static int ipsecif4_filter4(const struct ip *, struct ipsec_variant *,
78 struct ifnet *);
79
80 #ifdef INET6
81 static int ipsecif6_input(struct mbuf **, int *, int, void *);
82 static int ipsecif6_output(struct ipsec_variant *, int, struct mbuf *);
83 static int ipsecif6_filter6(const struct ip6_hdr *, struct ipsec_variant *,
84 struct ifnet *);
85 #endif
86
87 static int ip_ipsec_ttl = IPSEC_TTL;
88 static int ip_ipsec_copy_tos = 0;
89 #ifdef INET6
90 static int ip6_ipsec_hlim = IPSEC_HLIM;
91 static int ip6_ipsec_pmtu = 0; /* XXX: per interface configuration?? */
92 static int ip6_ipsec_copy_tos = 0;
93 #endif
94
95 static const struct encapsw ipsecif4_encapsw = {
96 .encapsw4 = {
97 .pr_input = ipsecif4_input,
98 .pr_ctlinput = NULL,
99 }
100 };
101
102 #ifdef INET6
103 static const struct encapsw ipsecif6_encapsw;
104 #endif
105
106 static int
107 ipsecif_set_natt_ports(struct ipsec_variant *var, struct mbuf *m)
108 {
109
110 KASSERT(if_ipsec_heldref_variant(var));
111
112 if (var->iv_sport || var->iv_dport) {
113 struct m_tag *mtag;
114
115 mtag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
116 sizeof(uint16_t) + sizeof(uint16_t), M_DONTWAIT);
117 if (mtag) {
118 uint16_t *natt_port;
119
120 natt_port = (uint16_t *)(mtag + 1);
121 natt_port[0] = var->iv_dport;
122 natt_port[1] = var->iv_sport;
123 m_tag_prepend(m, mtag);
124 } else {
125 return ENOBUFS;
126 }
127 }
128
129 return 0;
130 }
131
132 static struct mbuf *
133 ipsecif4_prepend_hdr(struct ipsec_variant *var, struct mbuf *m,
134 uint8_t proto, uint8_t tos)
135 {
136 struct ip *ip;
137 struct sockaddr_in *src, *dst;
138
139 src = satosin(var->iv_psrc);
140 dst = satosin(var->iv_pdst);
141
142 if (in_nullhost(src->sin_addr) || in_nullhost(src->sin_addr) ||
143 src->sin_addr.s_addr == INADDR_BROADCAST ||
144 dst->sin_addr.s_addr == INADDR_BROADCAST) {
145 m_freem(m);
146 return NULL;
147 }
148 m->m_flags &= ~M_BCAST;
149
150 if (IN_MULTICAST(src->sin_addr.s_addr) ||
151 IN_MULTICAST(dst->sin_addr.s_addr)) {
152 m_freem(m);
153 return NULL;
154 }
155
156 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
157 if (m && M_UNWRITABLE(m, sizeof(struct ip)))
158 m = m_pullup(m, sizeof(struct ip));
159 if (m == NULL)
160 return NULL;
161
162 ip = mtod(m, struct ip *);
163 ip->ip_v = IPVERSION;
164 ip->ip_off = htons(0);
165 if (m->m_pkthdr.len < IP_MINFRAGSIZE)
166 ip->ip_id = 0;
167 else
168 ip->ip_id = ip_newid(NULL);
169 ip->ip_hl = sizeof(*ip) >> 2;
170 if (ip_ipsec_copy_tos)
171 ip->ip_tos = tos;
172 else
173 ip->ip_tos = 0;
174 ip->ip_sum = 0;
175 ip->ip_src = src->sin_addr;
176 ip->ip_dst = dst->sin_addr;
177 ip->ip_p = proto;
178 ip->ip_ttl = ip_ipsec_ttl;
179 ip->ip_len = htons(m->m_pkthdr.len);
180 #ifndef IPSEC_TX_TOS_CLEAR
181 struct ifnet *ifp = &var->iv_softc->ipsec_if;
182 if (ifp->if_flags & IFF_ECN)
183 ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
184 else
185 ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
186 #endif
187
188 return m;
189 }
190
191 static int
192 ipsecif4_needfrag(struct mbuf *m, struct ipsecrequest *isr)
193 {
194 struct ip ip0;
195 struct ip *ip;
196 int mtu;
197 struct secasvar *sav;
198
199 sav = key_lookup_sa_bysaidx(&isr->saidx);
200 if (sav == NULL)
201 return 0;
202
203 if (!(sav->natt_type & UDP_ENCAP_ESPINUDP)) {
204 mtu = 0;
205 goto out;
206 }
207
208 if (m->m_len < sizeof(struct ip)) {
209 m_copydata(m, 0, sizeof(ip0), &ip0);
210 ip = &ip0;
211 } else {
212 ip = mtod(m, struct ip *);
213 }
214 mtu = sav->esp_frag;
215 if (ntohs(ip->ip_len) <= mtu)
216 mtu = 0;
217
218 out:
219 KEY_SA_UNREF(&sav);
220 return mtu;
221 }
222
223 static struct mbuf *
224 ipsecif4_flowinfo(struct mbuf *m, int family, int *proto0, u_int8_t *tos0)
225 {
226 const struct ip *ip;
227 int proto;
228 int tos;
229
230 KASSERT(proto0 != NULL);
231 KASSERT(tos0 != NULL);
232
233 switch (family) {
234 case AF_INET:
235 proto = IPPROTO_IPV4;
236 if (m->m_len < sizeof(*ip)) {
237 m = m_pullup(m, sizeof(*ip));
238 if (m == NULL) {
239 *tos0 = 0;
240 *proto0 = 0;
241 return NULL;
242 }
243 }
244 ip = mtod(m, const struct ip *);
245 tos = ip->ip_tos;
246 /* TODO: support ALTQ for innner packet */
247 break;
248 #ifdef INET6
249 case AF_INET6: {
250 const struct ip6_hdr *ip6;
251 proto = IPPROTO_IPV6;
252 if (m->m_len < sizeof(*ip6)) {
253 m = m_pullup(m, sizeof(*ip6));
254 if (m == NULL) {
255 *tos0 = 0;
256 *proto0 = 0;
257 return NULL;
258 }
259 }
260 ip6 = mtod(m, const struct ip6_hdr *);
261 tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
262 /* TODO: support ALTQ for innner packet */
263 break;
264 }
265 #endif /* INET6 */
266 default:
267 *tos0 = 0;
268 *proto0 = 0;
269 return NULL;
270 }
271
272 *proto0 = proto;
273 *tos0 = tos;
274 return m;
275 }
276
277 static int
278 ipsecif4_fragout(struct ipsec_variant *var, int family, struct mbuf *m, int mtu)
279 {
280 struct ifnet *ifp = &var->iv_softc->ipsec_if;
281 struct mbuf *next;
282 struct m_tag *mtag;
283 int error;
284
285 KASSERT(if_ipsec_heldref_variant(var));
286
287 mtag = m_tag_find(m, PACKET_TAG_IPSEC_NAT_T_PORTS);
288 if (mtag)
289 m_tag_delete(m, mtag);
290
291 /* consider new IP header prepended in ipsecif4_output() */
292 if (mtu <= sizeof(struct ip)) {
293 m_freem(m);
294 return ENETUNREACH;
295 }
296 m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
297 error = ip_fragment(m, ifp, mtu - sizeof(struct ip));
298 if (error)
299 return error;
300
301 for (error = 0; m; m = next) {
302 next = m->m_nextpkt;
303 m->m_nextpkt = NULL;
304 if (error) {
305 m_freem(m);
306 continue;
307 }
308
309 error = ipsecif4_output(var, family, m);
310 }
311 if (error == 0)
312 IP_STATINC(IP_STAT_FRAGMENTED);
313
314 return error;
315 }
316
317 int
318 ipsecif4_encap_func(struct mbuf *m, struct ip *ip, struct ipsec_variant *var)
319 {
320 struct m_tag *mtag;
321 struct sockaddr_in *src, *dst;
322 u_int16_t src_port = 0;
323 u_int16_t dst_port = 0;
324
325 KASSERT(var != NULL);
326
327 src = satosin(var->iv_psrc);
328 dst = satosin(var->iv_pdst);
329 mtag = m_tag_find(m, PACKET_TAG_IPSEC_NAT_T_PORTS);
330 if (mtag) {
331 u_int16_t *ports;
332
333 ports = (u_int16_t *)(mtag + 1);
334 src_port = ports[0];
335 dst_port = ports[1];
336 }
337
338 /* address match */
339 if (src->sin_addr.s_addr != ip->ip_dst.s_addr ||
340 dst->sin_addr.s_addr != ip->ip_src.s_addr)
341 return 0;
342
343 /* UDP encap? */
344 if (mtag == NULL && var->iv_sport == 0 && var->iv_dport == 0)
345 goto match;
346
347 /* port match */
348 if (src_port != var->iv_dport ||
349 dst_port != var->iv_sport) {
350 #ifdef DEBUG
351 printf("%s: port mismatch: pkt(%u, %u), if(%u, %u)\n",
352 __func__, ntohs(src_port), ntohs(dst_port),
353 ntohs(var->iv_sport), ntohs(var->iv_dport));
354 #endif
355 return 0;
356 }
357
358 match:
359 /*
360 * hide NAT-T information from encapsulated traffics.
361 * they don't know about IPsec.
362 */
363 if (mtag)
364 m_tag_delete(m, mtag);
365 return sizeof(src->sin_addr) + sizeof(dst->sin_addr);
366 }
367
368 static int
369 ipsecif4_output(struct ipsec_variant *var, int family, struct mbuf *m)
370 {
371 struct secpolicy *sp = NULL;
372 u_int8_t tos;
373 int proto;
374 int error;
375 int mtu;
376 u_long sa_mtu = 0;
377
378 KASSERT(if_ipsec_heldref_variant(var));
379 KASSERT(if_ipsec_variant_is_configured(var));
380 KASSERT(var->iv_psrc->sa_family == AF_INET);
381 KASSERT(var->iv_pdst->sa_family == AF_INET);
382
383 sp = IV_SP_OUT(var);
384 KASSERT(sp != NULL);
385 /*
386 * The SPs in ipsec_variant are prevented from freed by
387 * ipsec_variant->iv_psref. So, KEY_SP_REF() is unnecessary here.
388 */
389
390 KASSERT(sp->policy != IPSEC_POLICY_NONE);
391 KASSERT(sp->policy != IPSEC_POLICY_ENTRUST);
392 KASSERT(sp->policy != IPSEC_POLICY_BYPASS);
393 if (sp->policy != IPSEC_POLICY_IPSEC) {
394 m_freem(m);
395 error = ENETUNREACH;
396 goto done;
397 }
398
399 /* get flowinfo */
400 m = ipsecif4_flowinfo(m, family, &proto, &tos);
401 if (m == NULL) {
402 error = ENETUNREACH;
403 goto done;
404 }
405
406 /* prepend new IP header */
407 m = ipsecif4_prepend_hdr(var, m, proto, tos);
408 if (m == NULL) {
409 error = ENETUNREACH;
410 goto done;
411 }
412
413 /*
414 * Normal netipsec's NAT-T fragmentation is done in ip_output().
415 * See "natt_frag" processing.
416 * However, ipsec(4) interface's one is not done in the same way,
417 * so we must do NAT-T fragmentation by own code.
418 */
419 /* NAT-T ESP fragmentation */
420 mtu = ipsecif4_needfrag(m, sp->req);
421 if (mtu > 0)
422 return ipsecif4_fragout(var, family, m, mtu);
423
424 /* set NAT-T ports */
425 error = ipsecif_set_natt_ports(var, m);
426 if (error) {
427 m_freem(m);
428 goto done;
429 }
430
431 /* IPsec output */
432 IP_STATINC(IP_STAT_LOCALOUT);
433 error = ipsec4_process_packet(m, sp->req, &sa_mtu);
434 if (error == ENOENT)
435 error = 0;
436 /*
437 * frangmentation is already done in ipsecif4_fragout(),
438 * so ipsec4_process_packet() must not do fragmentation here.
439 */
440 KASSERT(sa_mtu == 0);
441
442 done:
443 return error;
444 }
445
446 #ifdef INET6
447 int
448 ipsecif6_encap_func(struct mbuf *m, struct ip6_hdr *ip6, struct ipsec_variant *var)
449 {
450 struct m_tag *mtag;
451 struct sockaddr_in6 *src, *dst;
452 u_int16_t src_port = 0;
453 u_int16_t dst_port = 0;
454
455 KASSERT(var != NULL);
456
457 src = satosin6(var->iv_psrc);
458 dst = satosin6(var->iv_pdst);
459 mtag = m_tag_find(m, PACKET_TAG_IPSEC_NAT_T_PORTS);
460 if (mtag) {
461 u_int16_t *ports;
462
463 ports = (u_int16_t *)(mtag + 1);
464 src_port = ports[0];
465 dst_port = ports[1];
466 }
467
468 /* address match */
469 if (!IN6_ARE_ADDR_EQUAL(&src->sin6_addr, &ip6->ip6_dst) ||
470 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_src))
471 return 0;
472
473 /* UDP encap? */
474 if (mtag == NULL && var->iv_sport == 0 && var->iv_dport == 0)
475 goto match;
476
477 /* port match */
478 if (src_port != var->iv_dport ||
479 dst_port != var->iv_sport) {
480 #ifdef DEBUG
481 printf("%s: port mismatch: pkt(%u, %u), if(%u, %u)\n",
482 __func__, ntohs(src_port), ntohs(dst_port),
483 ntohs(var->iv_sport), ntohs(var->iv_dport));
484 #endif
485 return 0;
486 }
487
488 match:
489 /*
490 * hide NAT-T information from encapsulated traffics.
491 * they don't know about IPsec.
492 */
493 if (mtag)
494 m_tag_delete(m, mtag);
495 return sizeof(src->sin6_addr) + sizeof(dst->sin6_addr);
496 }
497
498 static int
499 ipsecif6_output(struct ipsec_variant *var, int family, struct mbuf *m)
500 {
501 struct ifnet *ifp = &var->iv_softc->ipsec_if;
502 struct ipsec_softc *sc = ifp->if_softc;
503 struct ipsec_ro *iro;
504 struct rtentry *rt;
505 struct sockaddr_in6 *sin6_src;
506 struct sockaddr_in6 *sin6_dst;
507 struct ip6_hdr *ip6;
508 int proto, error;
509 u_int8_t itos, otos;
510 union {
511 struct sockaddr dst;
512 struct sockaddr_in6 dst6;
513 } u;
514
515 KASSERT(if_ipsec_heldref_variant(var));
516 KASSERT(if_ipsec_variant_is_configured(var));
517
518 sin6_src = satosin6(var->iv_psrc);
519 sin6_dst = satosin6(var->iv_pdst);
520
521 KASSERT(sin6_src->sin6_family == AF_INET6);
522 KASSERT(sin6_dst->sin6_family == AF_INET6);
523
524 switch (family) {
525 #ifdef INET
526 case AF_INET:
527 {
528 struct ip *ip;
529
530 proto = IPPROTO_IPV4;
531 if (m->m_len < sizeof(*ip)) {
532 m = m_pullup(m, sizeof(*ip));
533 if (m == NULL)
534 return ENOBUFS;
535 }
536 ip = mtod(m, struct ip *);
537 itos = ip->ip_tos;
538 /* TODO: support ALTQ for innner packet */
539 break;
540 }
541 #endif /* INET */
542 case AF_INET6:
543 {
544 struct ip6_hdr *xip6;
545 proto = IPPROTO_IPV6;
546 if (m->m_len < sizeof(*xip6)) {
547 m = m_pullup(m, sizeof(*xip6));
548 if (m == NULL)
549 return ENOBUFS;
550 }
551 xip6 = mtod(m, struct ip6_hdr *);
552 itos = (ntohl(xip6->ip6_flow) >> 20) & 0xff;
553 /* TODO: support ALTQ for innner packet */
554 break;
555 }
556 default:
557 m_freem(m);
558 return EAFNOSUPPORT;
559 }
560
561 /* prepend new IP header */
562 M_PREPEND(m, sizeof(struct ip6_hdr), M_DONTWAIT);
563 if (m && M_UNWRITABLE(m, sizeof(struct ip6_hdr)))
564 m = m_pullup(m, sizeof(struct ip6_hdr));
565 if (m == NULL)
566 return ENOBUFS;
567
568 ip6 = mtod(m, struct ip6_hdr *);
569 ip6->ip6_flow = 0;
570 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
571 ip6->ip6_vfc |= IPV6_VERSION;
572 #if 0 /* ip6->ip6_plen will be filled by ip6_output */
573 ip6->ip6_plen = htons((u_short)m->m_pkthdr.len - sizeof(*ip6));
574 #endif
575 ip6->ip6_nxt = proto;
576 ip6->ip6_hlim = ip6_ipsec_hlim;
577 ip6->ip6_src = sin6_src->sin6_addr;
578 /* bidirectional configured tunnel mode */
579 if (!IN6_IS_ADDR_UNSPECIFIED(&sin6_dst->sin6_addr)) {
580 ip6->ip6_dst = sin6_dst->sin6_addr;
581 } else {
582 m_freem(m);
583 return ENETUNREACH;
584 }
585 #ifndef IPSEC_TX_TOS_CLEAR
586 if (ifp->if_flags & IFF_ECN)
587 ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
588 else
589 ip_ecn_ingress(ECN_NOCARE, &otos, &itos);
590
591 if (!ip6_ipsec_copy_tos)
592 otos = 0;
593 #else
594 if (ip6_ipsec_copy_tos)
595 otos = itos;
596 else
597 otos = 0;
598 #endif
599 ip6->ip6_flow &= ~ntohl(0xff00000);
600 ip6->ip6_flow |= htonl((u_int32_t)otos << 20);
601
602 sockaddr_in6_init(&u.dst6, &sin6_dst->sin6_addr, 0, 0, 0);
603
604 iro = percpu_getref(sc->ipsec_ro_percpu);
605 mutex_enter(iro->ir_lock);
606 if ((rt = rtcache_lookup(&iro->ir_ro, &u.dst)) == NULL) {
607 mutex_exit(iro->ir_lock);
608 percpu_putref(sc->ipsec_ro_percpu);
609 m_freem(m);
610 return ENETUNREACH;
611 }
612
613 if (rt->rt_ifp == ifp) {
614 rtcache_unref(rt, &iro->ir_ro);
615 rtcache_free(&iro->ir_ro);
616 mutex_exit(iro->ir_lock);
617 percpu_putref(sc->ipsec_ro_percpu);
618 m_freem(m);
619 return ENETUNREACH;
620 }
621 rtcache_unref(rt, &iro->ir_ro);
622
623 /* set NAT-T ports */
624 error = ipsecif_set_natt_ports(var, m);
625 if (error) {
626 m_freem(m);
627 goto out;
628 }
629
630 /*
631 * force fragmentation to minimum MTU, to avoid path MTU discovery.
632 * it is too painful to ask for resend of inner packet, to achieve
633 * path MTU discovery for encapsulated packets.
634 */
635 error = ip6_output(m, 0, &iro->ir_ro,
636 ip6_ipsec_pmtu ? 0 : IPV6_MINMTU, 0, NULL, NULL);
637
638 out:
639 if (error)
640 rtcache_free(&iro->ir_ro);
641 mutex_exit(iro->ir_lock);
642 percpu_putref(sc->ipsec_ro_percpu);
643
644 return error;
645 }
646 #endif /* INET6 */
647
648 static void
649 ipsecif4_input(struct mbuf *m, int off, int proto, void *eparg)
650 {
651 struct ifnet *ipsecp;
652 struct ipsec_softc *sc = eparg;
653 struct ipsec_variant *var;
654 const struct ip *ip;
655 int af;
656 #ifndef IPSEC_TX_TOS_CLEAR
657 u_int8_t otos;
658 #endif
659 struct psref psref_rcvif;
660 struct psref psref_var;
661 struct ifnet *rcvif;
662
663 KASSERT(sc != NULL);
664
665 ipsecp = &sc->ipsec_if;
666 if ((ipsecp->if_flags & IFF_UP) == 0) {
667 m_freem(m);
668 ip_statinc(IP_STAT_NOIPSEC);
669 return;
670 }
671
672 var = if_ipsec_getref_variant(sc, &psref_var);
673 if (if_ipsec_variant_is_unconfigured(var)) {
674 if_ipsec_putref_variant(var, &psref_var);
675 m_freem(m);
676 ip_statinc(IP_STAT_NOIPSEC);
677 return;
678 }
679
680 ip = mtod(m, const struct ip *);
681
682 rcvif = m_get_rcvif_psref(m, &psref_rcvif);
683 if (rcvif == NULL || !ipsecif4_filter4(ip, var, rcvif)) {
684 m_put_rcvif_psref(rcvif, &psref_rcvif);
685 if_ipsec_putref_variant(var, &psref_var);
686 m_freem(m);
687 ip_statinc(IP_STAT_NOIPSEC);
688 return;
689 }
690 m_put_rcvif_psref(rcvif, &psref_rcvif);
691 if_ipsec_putref_variant(var, &psref_var);
692 #ifndef IPSEC_TX_TOS_CLEAR
693 otos = ip->ip_tos;
694 #endif
695 m_adj(m, off);
696
697 switch (proto) {
698 case IPPROTO_IPV4:
699 {
700 struct ip *xip;
701 af = AF_INET;
702 if (M_UNWRITABLE(m, sizeof(*xip))) {
703 m = m_pullup(m, sizeof(*xip));
704 if (m == NULL)
705 return;
706 }
707 xip = mtod(m, struct ip *);
708 #ifndef IPSEC_TX_TOS_CLEAR
709 if (ipsecp->if_flags & IFF_ECN)
710 ip_ecn_egress(ECN_ALLOWED, &otos, &xip->ip_tos);
711 else
712 ip_ecn_egress(ECN_NOCARE, &otos, &xip->ip_tos);
713 #endif
714 break;
715 }
716 #ifdef INET6
717 case IPPROTO_IPV6:
718 {
719 struct ip6_hdr *ip6;
720 u_int8_t itos;
721 af = AF_INET6;
722 if (M_UNWRITABLE(m, sizeof(*ip6))) {
723 m = m_pullup(m, sizeof(*ip6));
724 if (m == NULL)
725 return;
726 }
727 ip6 = mtod(m, struct ip6_hdr *);
728 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
729 #ifndef IPSEC_TX_TOS_CLEAR
730 if (ipsecp->if_flags & IFF_ECN)
731 ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
732 else
733 ip_ecn_egress(ECN_NOCARE, &otos, &itos);
734 #endif
735 ip6->ip6_flow &= ~htonl(0xff << 20);
736 ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
737 break;
738 }
739 #endif /* INET6 */
740 default:
741 ip_statinc(IP_STAT_NOIPSEC);
742 m_freem(m);
743 return;
744 }
745 if_ipsec_input(m, af, ipsecp);
746
747 return;
748 }
749
750 /*
751 * validate and filter the packet
752 */
753 static int
754 ipsecif4_filter4(const struct ip *ip, struct ipsec_variant *var,
755 struct ifnet *ifp)
756 {
757 struct sockaddr_in *src, *dst;
758
759 src = satosin(var->iv_psrc);
760 dst = satosin(var->iv_pdst);
761
762 return in_tunnel_validate(ip, src->sin_addr, dst->sin_addr);
763 }
764
765 #ifdef INET6
766 static int
767 ipsecif6_input(struct mbuf **mp, int *offp, int proto, void *eparg)
768 {
769 struct mbuf *m = *mp;
770 struct ifnet *ipsecp;
771 struct ipsec_softc *sc = eparg;
772 struct ipsec_variant *var;
773 struct ip6_hdr *ip6;
774 int af = 0;
775 #ifndef IPSEC_TX_TOS_CLEAR
776 u_int32_t otos;
777 #endif
778 struct psref psref_rcvif;
779 struct psref psref_var;
780 struct ifnet *rcvif;
781
782 KASSERT(eparg != NULL);
783
784 ipsecp = &sc->ipsec_if;
785 if ((ipsecp->if_flags & IFF_UP) == 0) {
786 m_freem(m);
787 IP6_STATINC(IP6_STAT_NOIPSEC);
788 return IPPROTO_DONE;
789 }
790
791 var = if_ipsec_getref_variant(sc, &psref_var);
792 if (if_ipsec_variant_is_unconfigured(var)) {
793 if_ipsec_putref_variant(var, &psref_var);
794 m_freem(m);
795 IP6_STATINC(IP6_STAT_NOIPSEC);
796 return IPPROTO_DONE;
797 }
798
799 ip6 = mtod(m, struct ip6_hdr *);
800
801 rcvif = m_get_rcvif_psref(m, &psref_rcvif);
802 if (rcvif == NULL || !ipsecif6_filter6(ip6, var, rcvif)) {
803 m_put_rcvif_psref(rcvif, &psref_rcvif);
804 if_ipsec_putref_variant(var, &psref_var);
805 m_freem(m);
806 IP6_STATINC(IP6_STAT_NOIPSEC);
807 return IPPROTO_DONE;
808 }
809 m_put_rcvif_psref(rcvif, &psref_rcvif);
810 if_ipsec_putref_variant(var, &psref_var);
811
812 #ifndef IPSEC_TX_TOS_CLEAR
813 otos = ip6->ip6_flow;
814 #endif
815 m_adj(m, *offp);
816
817 switch (proto) {
818 #ifdef INET
819 case IPPROTO_IPV4:
820 {
821 af = AF_INET;
822 #ifndef IPSEC_TX_TOS_CLEAR
823 struct ip *ip;
824 u_int8_t otos8;
825 otos8 = (ntohl(otos) >> 20) & 0xff;
826
827 if (M_UNWRITABLE(m, sizeof(*ip))) {
828 m = m_pullup(m, sizeof(*ip));
829 if (m == NULL)
830 return IPPROTO_DONE;
831 }
832 ip = mtod(m, struct ip *);
833 if (ipsecp->if_flags & IFF_ECN)
834 ip_ecn_egress(ECN_ALLOWED, &otos8, &ip->ip_tos);
835 else
836 ip_ecn_egress(ECN_NOCARE, &otos8, &ip->ip_tos);
837 #endif
838 break;
839 }
840 #endif /* INET */
841 case IPPROTO_IPV6:
842 {
843 af = AF_INET6;
844 #ifndef IPSEC_TX_TOS_CLEAR
845 struct ip6_hdr *xip6;
846
847 if (M_UNWRITABLE(m, sizeof(*xip6))) {
848 m = m_pullup(m, sizeof(*xip6));
849 if (m == NULL)
850 return IPPROTO_DONE;
851 }
852 xip6 = mtod(m, struct ip6_hdr *);
853 if (ipsecp->if_flags & IFF_ECN)
854 ip6_ecn_egress(ECN_ALLOWED, &otos, &xip6->ip6_flow);
855 else
856 ip6_ecn_egress(ECN_NOCARE, &otos, &xip6->ip6_flow);
857 break;
858 #endif
859 }
860 default:
861 IP6_STATINC(IP6_STAT_NOIPSEC);
862 m_freem(m);
863 return IPPROTO_DONE;
864 }
865
866 if_ipsec_input(m, af, ipsecp);
867 return IPPROTO_DONE;
868 }
869
870 /*
871 * validate and filter the packet.
872 */
873 static int
874 ipsecif6_filter6(const struct ip6_hdr *ip6, struct ipsec_variant *var,
875 struct ifnet *ifp)
876 {
877 struct sockaddr_in6 *src, *dst;
878
879 src = satosin6(var->iv_psrc);
880 dst = satosin6(var->iv_pdst);
881
882 return in6_tunnel_validate(ip6, &src->sin6_addr, &dst->sin6_addr);
883 }
884 #endif /* INET6 */
885
886 int
887 ipsecif4_attach(struct ipsec_variant *var)
888 {
889 struct ipsec_softc *sc = var->iv_softc;
890
891 KASSERT(if_ipsec_variant_is_configured(var));
892
893 if (var->iv_encap_cookie4 != NULL)
894 return EALREADY;
895 var->iv_encap_cookie4 = encap_attach_func(AF_INET, -1, if_ipsec_encap_func,
896 &ipsecif4_encapsw, sc);
897 if (var->iv_encap_cookie4 == NULL)
898 return EEXIST;
899
900 var->iv_output = ipsecif4_output;
901 return 0;
902 }
903
904 int
905 ipsecif4_detach(struct ipsec_variant *var)
906 {
907 int error;
908
909 if (var->iv_encap_cookie4 == NULL)
910 return 0;
911
912 var->iv_output = NULL;
913 error = encap_detach(var->iv_encap_cookie4);
914 if (error == 0)
915 var->iv_encap_cookie4 = NULL;
916
917 return error;
918 }
919
920 #ifdef INET6
921 int
922 ipsecif6_attach(struct ipsec_variant *var)
923 {
924 struct ipsec_softc *sc = var->iv_softc;
925
926 KASSERT(if_ipsec_variant_is_configured(var));
927 KASSERT(var->iv_encap_cookie6 == NULL);
928
929 var->iv_encap_cookie6 = encap_attach_func(AF_INET6, -1, if_ipsec_encap_func,
930 &ipsecif6_encapsw, sc);
931 if (var->iv_encap_cookie6 == NULL)
932 return EEXIST;
933
934 var->iv_output = ipsecif6_output;
935 return 0;
936 }
937
938 static void
939 ipsecif6_rtcache_free_pc(void *p, void *arg __unused, struct cpu_info *ci __unused)
940 {
941 struct ipsec_ro *iro = p;
942
943 mutex_enter(iro->ir_lock);
944 rtcache_free(&iro->ir_ro);
945 mutex_exit(iro->ir_lock);
946 }
947
948 int
949 ipsecif6_detach(struct ipsec_variant *var)
950 {
951 struct ipsec_softc *sc = var->iv_softc;
952 int error;
953
954 KASSERT(var->iv_encap_cookie6 != NULL);
955
956 percpu_foreach(sc->ipsec_ro_percpu, ipsecif6_rtcache_free_pc, NULL);
957
958 var->iv_output = NULL;
959 error = encap_detach(var->iv_encap_cookie6);
960 if (error == 0)
961 var->iv_encap_cookie6 = NULL;
962 return error;
963 }
964
965 void *
966 ipsecif6_ctlinput(int cmd, const struct sockaddr *sa, void *d, void *eparg)
967 {
968 struct ipsec_softc *sc = eparg;
969 struct ip6ctlparam *ip6cp = NULL;
970 struct ip6_hdr *ip6;
971 const struct sockaddr_in6 *dst6;
972 struct ipsec_ro *iro;
973
974 if (sa->sa_family != AF_INET6 ||
975 sa->sa_len != sizeof(struct sockaddr_in6))
976 return NULL;
977
978 if ((unsigned)cmd >= PRC_NCMDS)
979 return NULL;
980 if (cmd == PRC_HOSTDEAD)
981 d = NULL;
982 else if (inet6ctlerrmap[cmd] == 0)
983 return NULL;
984
985 /* if the parameter is from icmp6, decode it. */
986 if (d != NULL) {
987 ip6cp = (struct ip6ctlparam *)d;
988 ip6 = ip6cp->ip6c_ip6;
989 } else {
990 ip6 = NULL;
991 }
992
993 if (!ip6)
994 return NULL;
995
996 iro = percpu_getref(sc->ipsec_ro_percpu);
997 mutex_enter(iro->ir_lock);
998 dst6 = satocsin6(rtcache_getdst(&iro->ir_ro));
999 /* XXX scope */
1000 if (dst6 == NULL)
1001 ;
1002 else if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst6->sin6_addr))
1003 /* flush route cache */
1004 rtcache_free(&iro->ir_ro);
1005
1006 mutex_exit(iro->ir_lock);
1007 percpu_putref(sc->ipsec_ro_percpu);
1008
1009 return NULL;
1010 }
1011
1012 ENCAP_PR_WRAP_CTLINPUT(ipsecif6_ctlinput)
1013 #define ipsecif6_ctlinput ipsecif6_ctlinput_wrapper
1014
1015 static const struct encapsw ipsecif6_encapsw = {
1016 .encapsw6 = {
1017 .pr_input = ipsecif6_input,
1018 .pr_ctlinput = ipsecif6_ctlinput,
1019 }
1020 };
1021 #endif /* INET6 */
1022