xref: /freebsd/sys/net/if_ipsec.c (revision 00524fd475995d30780ce80ec75e085223206cac)
1 /*-
2  * Copyright (c) 2016-2018 Yandex LLC
3  * Copyright (c) 2016-2018 Andrey V. Elsukov <ae@FreeBSD.org>
4  * All rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  *
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 AUTHOR ``AS IS'' AND ANY EXPRESS OR
17  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
18  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
19  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
20  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
21  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
22  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
23  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
24  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
25  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
26  */
27 
28 #include <sys/cdefs.h>
29 #include "opt_inet.h"
30 #include "opt_inet6.h"
31 #include "opt_ipsec.h"
32 
33 #include <sys/param.h>
34 #include <sys/systm.h>
35 #include <sys/kernel.h>
36 #include <sys/fnv_hash.h>
37 #include <sys/jail.h>
38 #include <sys/lock.h>
39 #include <sys/malloc.h>
40 #include <sys/mbuf.h>
41 #include <sys/module.h>
42 #include <sys/socket.h>
43 #include <sys/sockio.h>
44 #include <sys/sx.h>
45 #include <sys/errno.h>
46 #include <sys/sysctl.h>
47 #include <sys/priv.h>
48 #include <sys/proc.h>
49 #include <sys/conf.h>
50 
51 #include <net/if.h>
52 #include <net/if_var.h>
53 #include <net/if_private.h>
54 #include <net/if_clone.h>
55 #include <net/if_types.h>
56 #include <net/bpf.h>
57 #include <net/route.h>
58 #include <net/vnet.h>
59 
60 #include <netinet/in.h>
61 #include <netinet/in_var.h>
62 #include <netinet/ip.h>
63 #include <netinet/ip_encap.h>
64 
65 #include <netinet/ip6.h>
66 #include <netinet6/in6_var.h>
67 #include <netinet6/scope6_var.h>
68 
69 #include <netipsec/ipsec.h>
70 #ifdef INET6
71 #include <netipsec/ipsec6.h>
72 #endif
73 
74 #include <net/if_ipsec.h>
75 #include <netipsec/key.h>
76 
77 #include <security/mac/mac_framework.h>
78 
79 static MALLOC_DEFINE(M_IPSEC, "ipsec", "IPsec Virtual Tunnel Interface");
80 static const char ipsecname[] = "ipsec";
81 
82 #if defined(INET) && defined(INET6)
83 #define	IPSEC_SPCOUNT		4
84 #else
85 #define	IPSEC_SPCOUNT		2
86 #endif
87 
88 struct ipsec_softc {
89 	struct ifnet		*ifp;
90 	struct secpolicy	*sp[IPSEC_SPCOUNT];
91 	uint32_t		reqid;
92 	u_int			family;
93 	u_int			fibnum;
94 
95 	CK_LIST_ENTRY(ipsec_softc) idhash;
96 	CK_LIST_ENTRY(ipsec_softc) srchash;
97 };
98 
99 #define	IPSEC_RLOCK_TRACKER	struct epoch_tracker ipsec_et
100 #define	IPSEC_RLOCK()	epoch_enter_preempt(net_epoch_preempt, &ipsec_et)
101 #define	IPSEC_RUNLOCK()	epoch_exit_preempt(net_epoch_preempt, &ipsec_et)
102 #define	IPSEC_WAIT()	epoch_wait_preempt(net_epoch_preempt)
103 
104 #ifndef IPSEC_HASH_SIZE
105 #define	IPSEC_HASH_SIZE	(1 << 5)
106 #endif
107 
108 CK_LIST_HEAD(ipsec_iflist, ipsec_softc);
109 VNET_DEFINE_STATIC(struct ipsec_iflist *, ipsec_idhtbl) = NULL;
110 #define	V_ipsec_idhtbl		VNET(ipsec_idhtbl)
111 
112 #ifdef INET
113 VNET_DEFINE_STATIC(struct ipsec_iflist *, ipsec4_srchtbl) = NULL;
114 #define	V_ipsec4_srchtbl	VNET(ipsec4_srchtbl)
115 static const struct srcaddrtab *ipsec4_srctab = NULL;
116 #endif
117 
118 #ifdef INET6
119 VNET_DEFINE_STATIC(struct ipsec_iflist *, ipsec6_srchtbl) = NULL;
120 #define	V_ipsec6_srchtbl	VNET(ipsec6_srchtbl)
121 static const struct srcaddrtab *ipsec6_srctab = NULL;
122 #endif
123 
124 static struct ipsec_iflist *
125 ipsec_idhash(uint32_t id)
126 {
127 
128 	return (&V_ipsec_idhtbl[fnv_32_buf(&id, sizeof(id),
129 	    FNV1_32_INIT) & (IPSEC_HASH_SIZE - 1)]);
130 }
131 
132 static struct ipsec_iflist *
133 ipsec_srchash(const struct sockaddr *sa)
134 {
135 	uint32_t hval;
136 
137 	switch (sa->sa_family) {
138 #ifdef INET
139 	case AF_INET:
140 		hval = fnv_32_buf(
141 		    &((const struct sockaddr_in *)sa)->sin_addr.s_addr,
142 		    sizeof(in_addr_t), FNV1_32_INIT);
143 		return (&V_ipsec4_srchtbl[hval & (IPSEC_HASH_SIZE - 1)]);
144 #endif
145 #ifdef INET6
146 	case AF_INET6:
147 		hval = fnv_32_buf(
148 		    &((const struct sockaddr_in6 *)sa)->sin6_addr,
149 		    sizeof(struct in6_addr), FNV1_32_INIT);
150 		return (&V_ipsec6_srchtbl[hval & (IPSEC_HASH_SIZE - 1)]);
151 #endif
152 	}
153 	return (NULL);
154 }
155 
156 /*
157  * ipsec_ioctl_sx protects from concurrent ioctls.
158  */
159 static struct sx ipsec_ioctl_sx;
160 SX_SYSINIT(ipsec_ioctl_sx, &ipsec_ioctl_sx, "ipsec_ioctl");
161 
162 static int	ipsec_init_reqid(struct ipsec_softc *);
163 static int	ipsec_set_tunnel(struct ipsec_softc *, struct sockaddr *,
164     struct sockaddr *, uint32_t);
165 static void	ipsec_delete_tunnel(struct ipsec_softc *);
166 
167 static int	ipsec_set_addresses(struct ifnet *, struct sockaddr *,
168     struct sockaddr *);
169 static int	ipsec_set_reqid(struct ipsec_softc *, uint32_t);
170 static void	ipsec_set_running(struct ipsec_softc *);
171 
172 #ifdef VIMAGE
173 static void	ipsec_reassign(struct ifnet *, struct vnet *, char *);
174 #endif
175 static void	ipsec_srcaddr(void *, const struct sockaddr *, int);
176 static int	ipsec_ioctl(struct ifnet *, u_long, caddr_t);
177 static int	ipsec_transmit(struct ifnet *, struct mbuf *);
178 static int	ipsec_output(struct ifnet *, struct mbuf *,
179     const struct sockaddr *, struct route *);
180 static void	ipsec_qflush(struct ifnet *);
181 static int	ipsec_clone_create(struct if_clone *, int, caddr_t);
182 static void	ipsec_clone_destroy(struct ifnet *);
183 
184 VNET_DEFINE_STATIC(struct if_clone *, ipsec_cloner);
185 #define	V_ipsec_cloner		VNET(ipsec_cloner)
186 
187 static int
188 ipsec_clone_create(struct if_clone *ifc, int unit, caddr_t params)
189 {
190 	struct ipsec_softc *sc;
191 	struct ifnet *ifp;
192 
193 	sc = malloc(sizeof(*sc), M_IPSEC, M_WAITOK | M_ZERO);
194 	sc->fibnum = curthread->td_proc->p_fibnum;
195 	sc->ifp = ifp = if_alloc(IFT_TUNNEL);
196 	ifp->if_softc = sc;
197 	if_initname(ifp, ipsecname, unit);
198 
199 	ifp->if_addrlen = 0;
200 	ifp->if_mtu = IPSEC_MTU;
201 	ifp->if_flags  = IFF_POINTOPOINT | IFF_MULTICAST;
202 	ifp->if_ioctl  = ipsec_ioctl;
203 	ifp->if_transmit  = ipsec_transmit;
204 	ifp->if_qflush  = ipsec_qflush;
205 	ifp->if_output = ipsec_output;
206 #ifdef VIMAGE
207 	ifp->if_reassign = ipsec_reassign;
208 #endif
209 	if_attach(ifp);
210 	bpfattach(ifp, DLT_NULL, sizeof(uint32_t));
211 
212 	return (0);
213 }
214 
215 #ifdef VIMAGE
216 static void
217 ipsec_reassign(struct ifnet *ifp, struct vnet *new_vnet __unused,
218     char *unused __unused)
219 {
220 	struct ipsec_softc *sc;
221 
222 	sx_xlock(&ipsec_ioctl_sx);
223 	sc = ifp->if_softc;
224 	if (sc != NULL)
225 		ipsec_delete_tunnel(sc);
226 	sx_xunlock(&ipsec_ioctl_sx);
227 }
228 #endif /* VIMAGE */
229 
230 static void
231 ipsec_clone_destroy(struct ifnet *ifp)
232 {
233 	struct ipsec_softc *sc;
234 
235 	sx_xlock(&ipsec_ioctl_sx);
236 	sc = ifp->if_softc;
237 	ipsec_delete_tunnel(sc);
238 	/*
239 	 * Delete softc from idhash on interface destroy, since
240 	 * ipsec_delete_tunnel() keeps reqid unchanged.
241 	 */
242 	if (sc->reqid != 0)
243 		CK_LIST_REMOVE(sc, idhash);
244 	bpfdetach(ifp);
245 	if_detach(ifp);
246 	ifp->if_softc = NULL;
247 	sx_xunlock(&ipsec_ioctl_sx);
248 
249 	IPSEC_WAIT();
250 	if_free(ifp);
251 	free(sc, M_IPSEC);
252 }
253 
254 static struct ipsec_iflist *
255 ipsec_hashinit(void)
256 {
257 	struct ipsec_iflist *hash;
258 	int i;
259 
260 	hash = malloc(sizeof(struct ipsec_iflist) * IPSEC_HASH_SIZE,
261 	    M_IPSEC, M_WAITOK);
262 	for (i = 0; i < IPSEC_HASH_SIZE; i++)
263 		CK_LIST_INIT(&hash[i]);
264 
265 	return (hash);
266 }
267 
268 static void
269 vnet_ipsec_init(const void *unused __unused)
270 {
271 
272 	V_ipsec_idhtbl = ipsec_hashinit();
273 #ifdef INET
274 	V_ipsec4_srchtbl = ipsec_hashinit();
275 	if (IS_DEFAULT_VNET(curvnet))
276 		ipsec4_srctab = ip_encap_register_srcaddr(ipsec_srcaddr,
277 		    NULL, M_WAITOK);
278 #endif
279 #ifdef INET6
280 	V_ipsec6_srchtbl = ipsec_hashinit();
281 	if (IS_DEFAULT_VNET(curvnet))
282 		ipsec6_srctab = ip6_encap_register_srcaddr(ipsec_srcaddr,
283 		    NULL, M_WAITOK);
284 #endif
285 	V_ipsec_cloner = if_clone_simple(ipsecname, ipsec_clone_create,
286 	    ipsec_clone_destroy, 0);
287 }
288 VNET_SYSINIT(vnet_ipsec_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
289     vnet_ipsec_init, NULL);
290 
291 static void
292 vnet_ipsec_uninit(const void *unused __unused)
293 {
294 
295 	if_clone_detach(V_ipsec_cloner);
296 	free(V_ipsec_idhtbl, M_IPSEC);
297 	/*
298 	 * Use V_ipsec_idhtbl pointer as indicator that VNET is going to be
299 	 * destroyed, it is used by ipsec_srcaddr() callback.
300 	 */
301 	V_ipsec_idhtbl = NULL;
302 	IPSEC_WAIT();
303 
304 #ifdef INET
305 	if (IS_DEFAULT_VNET(curvnet))
306 		ip_encap_unregister_srcaddr(ipsec4_srctab);
307 	free(V_ipsec4_srchtbl, M_IPSEC);
308 #endif
309 #ifdef INET6
310 	if (IS_DEFAULT_VNET(curvnet))
311 		ip6_encap_unregister_srcaddr(ipsec6_srctab);
312 	free(V_ipsec6_srchtbl, M_IPSEC);
313 #endif
314 }
315 VNET_SYSUNINIT(vnet_ipsec_uninit, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
316     vnet_ipsec_uninit, NULL);
317 
318 static struct secpolicy *
319 ipsec_getpolicy(struct ipsec_softc *sc, int dir, sa_family_t af)
320 {
321 
322 	switch (af) {
323 #ifdef INET
324 	case AF_INET:
325 		return (sc->sp[(dir == IPSEC_DIR_INBOUND ? 0: 1)]);
326 #endif
327 #ifdef INET6
328 	case AF_INET6:
329 		return (sc->sp[(dir == IPSEC_DIR_INBOUND ? 0: 1)
330 #ifdef INET
331 			+ 2
332 #endif
333 		]);
334 #endif
335 	}
336 	return (NULL);
337 }
338 
339 static struct secasindex *
340 ipsec_getsaidx(struct ipsec_softc *sc, int dir, sa_family_t af)
341 {
342 	struct secpolicy *sp;
343 
344 	sp = ipsec_getpolicy(sc, dir, af);
345 	if (sp == NULL)
346 		return (NULL);
347 	return (&sp->req[0]->saidx);
348 }
349 
350 static int
351 ipsec_transmit(struct ifnet *ifp, struct mbuf *m)
352 {
353 	IPSEC_RLOCK_TRACKER;
354 	struct ipsec_softc *sc;
355 	struct secpolicy *sp;
356 	struct ip *ip;
357 	uint32_t af;
358 	int error;
359 
360 	IPSEC_RLOCK();
361 #ifdef MAC
362 	error = mac_ifnet_check_transmit(ifp, m);
363 	if (error) {
364 		m_freem(m);
365 		goto err;
366 	}
367 #endif
368 	error = ENETDOWN;
369 	sc = ifp->if_softc;
370 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
371 	    (ifp->if_flags & IFF_MONITOR) != 0 ||
372 	    (ifp->if_flags & IFF_UP) == 0 || sc->family == 0) {
373 		m_freem(m);
374 		goto err;
375 	}
376 
377 	/* Determine address family to correctly handle packet in BPF */
378 	ip = mtod(m, struct ip *);
379 	switch (ip->ip_v) {
380 #ifdef INET
381 	case IPVERSION:
382 		af = AF_INET;
383 		break;
384 #endif
385 #ifdef INET6
386 	case (IPV6_VERSION >> 4):
387 		af = AF_INET6;
388 		break;
389 #endif
390 	default:
391 		error = EAFNOSUPPORT;
392 		m_freem(m);
393 		goto err;
394 	}
395 
396 	/*
397 	 * Loop prevention.
398 	 * XXX: for now just check presence of IPSEC_OUT_DONE mbuf tag.
399 	 *      We can read full chain and compare destination address,
400 	 *      proto and mode from xform_history with values from softc.
401 	 */
402 	if (m_tag_find(m, PACKET_TAG_IPSEC_OUT_DONE, NULL) != NULL) {
403 		m_freem(m);
404 		goto err;
405 	}
406 
407 	sp = ipsec_getpolicy(sc, IPSEC_DIR_OUTBOUND, af);
408 	key_addref(sp);
409 	M_SETFIB(m, sc->fibnum);
410 
411 	BPF_MTAP2(ifp, &af, sizeof(af), m);
412 	if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
413 	if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len);
414 
415 	switch (af) {
416 #ifdef INET
417 	case AF_INET:
418 		error = ipsec4_process_packet(ifp, m, sp, NULL, ifp->if_mtu);
419 		break;
420 #endif
421 #ifdef INET6
422 	case AF_INET6:
423 		error = ipsec6_process_packet(ifp, m, sp, NULL, ifp->if_mtu);
424 		break;
425 #endif
426 	default:
427 		panic("%s: unknown address family\n", __func__);
428 	}
429 err:
430 	if (error != 0)
431 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
432 	IPSEC_RUNLOCK();
433 	return (error);
434 }
435 
436 static void
437 ipsec_qflush(struct ifnet *ifp __unused)
438 {
439 
440 }
441 
442 static int
443 ipsec_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
444 	struct route *ro)
445 {
446 
447 	return (ifp->if_transmit(ifp, m));
448 }
449 
450 int
451 ipsec_if_input(struct mbuf *m, struct secasvar *sav, uint32_t af)
452 {
453 	IPSEC_RLOCK_TRACKER;
454 	struct secasindex *saidx;
455 	struct ipsec_softc *sc;
456 	struct ifnet *ifp;
457 
458 	if (sav->state != SADB_SASTATE_MATURE &&
459 	    sav->state != SADB_SASTATE_DYING) {
460 		m_freem(m);
461 		return (ENETDOWN);
462 	}
463 
464 	if (sav->sah->saidx.mode != IPSEC_MODE_TUNNEL ||
465 	    sav->sah->saidx.proto != IPPROTO_ESP)
466 		return (0);
467 
468 	IPSEC_RLOCK();
469 	CK_LIST_FOREACH(sc, ipsec_idhash(sav->sah->saidx.reqid), idhash) {
470 		if (sc->family == 0)
471 			continue;
472 		saidx = ipsec_getsaidx(sc, IPSEC_DIR_INBOUND,
473 		    sav->sah->saidx.src.sa.sa_family);
474 		/* SA's reqid should match reqid in SP */
475 		if (saidx == NULL ||
476 		    sav->sah->saidx.reqid != saidx->reqid)
477 			continue;
478 		/* SAH's addresses should match tunnel endpoints. */
479 		if (key_sockaddrcmp(&sav->sah->saidx.dst.sa,
480 		    &saidx->dst.sa, 0) != 0)
481 			continue;
482 		if (key_sockaddrcmp(&sav->sah->saidx.src.sa,
483 		    &saidx->src.sa, 0) == 0)
484 			break;
485 	}
486 	if (sc == NULL) {
487 		IPSEC_RUNLOCK();
488 		/* Tunnel was not found. Nothing to do. */
489 		return (0);
490 	}
491 	ifp = sc->ifp;
492 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
493 	    (ifp->if_flags & IFF_UP) == 0) {
494 		IPSEC_RUNLOCK();
495 		m_freem(m);
496 		return (ENETDOWN);
497 	}
498 	/*
499 	 * We found matching and working tunnel.
500 	 * Set its ifnet as receiving interface.
501 	 */
502 	m->m_pkthdr.rcvif = ifp;
503 
504 	m_clrprotoflags(m);
505 	M_SETFIB(m, ifp->if_fib);
506 	BPF_MTAP2(ifp, &af, sizeof(af), m);
507 	if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
508 	if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
509 	if ((ifp->if_flags & IFF_MONITOR) != 0) {
510 		IPSEC_RUNLOCK();
511 		m_freem(m);
512 		return (ENETDOWN);
513 	}
514 	IPSEC_RUNLOCK();
515 	return (0);
516 }
517 
518 static int
519 ipsec_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
520 {
521 	struct ifreq *ifr = (struct ifreq*)data;
522 	struct sockaddr *dst, *src;
523 	struct ipsec_softc *sc;
524 	struct secasindex *saidx;
525 #ifdef INET
526 	struct sockaddr_in *sin = NULL;
527 #endif
528 #ifdef INET6
529 	struct sockaddr_in6 *sin6 = NULL;
530 #endif
531 	uint32_t reqid;
532 	int error;
533 
534 	switch (cmd) {
535 	case SIOCSIFADDR:
536 		ifp->if_flags |= IFF_UP;
537 	case SIOCADDMULTI:
538 	case SIOCDELMULTI:
539 	case SIOCGIFMTU:
540 	case SIOCSIFFLAGS:
541 		return (0);
542 	case SIOCSIFMTU:
543 		if (ifr->ifr_mtu < IPSEC_MTU_MIN ||
544 		    ifr->ifr_mtu > IPSEC_MTU_MAX)
545 			return (EINVAL);
546 		else
547 			ifp->if_mtu = ifr->ifr_mtu;
548 		return (0);
549 	}
550 	sx_xlock(&ipsec_ioctl_sx);
551 	sc = ifp->if_softc;
552 	/* Check that softc is still here */
553 	if (sc == NULL) {
554 		error = ENXIO;
555 		goto bad;
556 	}
557 	error = 0;
558 	switch (cmd) {
559 	case SIOCSIFPHYADDR:
560 #ifdef INET6
561 	case SIOCSIFPHYADDR_IN6:
562 #endif
563 		error = EINVAL;
564 		switch (cmd) {
565 #ifdef INET
566 		case SIOCSIFPHYADDR:
567 			src = (struct sockaddr *)
568 				&(((struct in_aliasreq *)data)->ifra_addr);
569 			dst = (struct sockaddr *)
570 				&(((struct in_aliasreq *)data)->ifra_dstaddr);
571 			break;
572 #endif
573 #ifdef INET6
574 		case SIOCSIFPHYADDR_IN6:
575 			src = (struct sockaddr *)
576 				&(((struct in6_aliasreq *)data)->ifra_addr);
577 			dst = (struct sockaddr *)
578 				&(((struct in6_aliasreq *)data)->ifra_dstaddr);
579 			break;
580 #endif
581 		default:
582 			goto bad;
583 		}
584 		/* sa_family must be equal */
585 		if (src->sa_family != dst->sa_family ||
586 		    src->sa_len != dst->sa_len)
587 			goto bad;
588 
589 		/* validate sa_len */
590 		switch (src->sa_family) {
591 #ifdef INET
592 		case AF_INET:
593 			if (src->sa_len != sizeof(struct sockaddr_in))
594 				goto bad;
595 			break;
596 #endif
597 #ifdef INET6
598 		case AF_INET6:
599 			if (src->sa_len != sizeof(struct sockaddr_in6))
600 				goto bad;
601 			break;
602 #endif
603 		default:
604 			error = EAFNOSUPPORT;
605 			goto bad;
606 		}
607 		/* check sa_family looks sane for the cmd */
608 		error = EAFNOSUPPORT;
609 		switch (cmd) {
610 #ifdef INET
611 		case SIOCSIFPHYADDR:
612 			if (src->sa_family == AF_INET)
613 				break;
614 			goto bad;
615 #endif
616 #ifdef INET6
617 		case SIOCSIFPHYADDR_IN6:
618 			if (src->sa_family == AF_INET6)
619 				break;
620 			goto bad;
621 #endif
622 		}
623 		error = EADDRNOTAVAIL;
624 		switch (src->sa_family) {
625 #ifdef INET
626 		case AF_INET:
627 			if (satosin(src)->sin_addr.s_addr == INADDR_ANY ||
628 			    satosin(dst)->sin_addr.s_addr == INADDR_ANY)
629 				goto bad;
630 			break;
631 #endif
632 #ifdef INET6
633 		case AF_INET6:
634 			if (IN6_IS_ADDR_UNSPECIFIED(
635 			    &satosin6(src)->sin6_addr) ||
636 			    IN6_IS_ADDR_UNSPECIFIED(
637 			    &satosin6(dst)->sin6_addr))
638 				goto bad;
639 			/*
640 			 * Check validity of the scope zone ID of the
641 			 * addresses, and convert it into the kernel
642 			 * internal form if necessary.
643 			 */
644 			error = sa6_embedscope(satosin6(src), 0);
645 			if (error != 0)
646 				goto bad;
647 			error = sa6_embedscope(satosin6(dst), 0);
648 			if (error != 0)
649 				goto bad;
650 #endif
651 		};
652 		error = ipsec_set_addresses(ifp, src, dst);
653 		break;
654 	case SIOCDIFPHYADDR:
655 		ipsec_delete_tunnel(sc);
656 		break;
657 	case SIOCGIFPSRCADDR:
658 	case SIOCGIFPDSTADDR:
659 #ifdef INET6
660 	case SIOCGIFPSRCADDR_IN6:
661 	case SIOCGIFPDSTADDR_IN6:
662 #endif
663 		if (sc->family == 0) {
664 			error = EADDRNOTAVAIL;
665 			break;
666 		}
667 		saidx = ipsec_getsaidx(sc, IPSEC_DIR_OUTBOUND, sc->family);
668 		if (saidx == NULL) {
669 			error = ENXIO;
670 			break;
671 		}
672 		switch (cmd) {
673 #ifdef INET
674 		case SIOCGIFPSRCADDR:
675 		case SIOCGIFPDSTADDR:
676 			if (saidx->src.sa.sa_family != AF_INET) {
677 				error = EADDRNOTAVAIL;
678 				break;
679 			}
680 			sin = (struct sockaddr_in *)&ifr->ifr_addr;
681 			memset(sin, 0, sizeof(*sin));
682 			sin->sin_family = AF_INET;
683 			sin->sin_len = sizeof(*sin);
684 			break;
685 #endif
686 #ifdef INET6
687 		case SIOCGIFPSRCADDR_IN6:
688 		case SIOCGIFPDSTADDR_IN6:
689 			if (saidx->src.sa.sa_family != AF_INET6) {
690 				error = EADDRNOTAVAIL;
691 				break;
692 			}
693 			sin6 = (struct sockaddr_in6 *)
694 				&(((struct in6_ifreq *)data)->ifr_addr);
695 			memset(sin6, 0, sizeof(*sin6));
696 			sin6->sin6_family = AF_INET6;
697 			sin6->sin6_len = sizeof(*sin6);
698 			break;
699 #endif
700 		default:
701 			error = EAFNOSUPPORT;
702 		}
703 		if (error == 0) {
704 			switch (cmd) {
705 #ifdef INET
706 			case SIOCGIFPSRCADDR:
707 				sin->sin_addr = saidx->src.sin.sin_addr;
708 				break;
709 			case SIOCGIFPDSTADDR:
710 				sin->sin_addr = saidx->dst.sin.sin_addr;
711 				break;
712 #endif
713 #ifdef INET6
714 			case SIOCGIFPSRCADDR_IN6:
715 				sin6->sin6_addr = saidx->src.sin6.sin6_addr;
716 				break;
717 			case SIOCGIFPDSTADDR_IN6:
718 				sin6->sin6_addr = saidx->dst.sin6.sin6_addr;
719 				break;
720 #endif
721 			}
722 		}
723 		if (error != 0)
724 			break;
725 		switch (cmd) {
726 #ifdef INET
727 		case SIOCGIFPSRCADDR:
728 		case SIOCGIFPDSTADDR:
729 			error = prison_if(curthread->td_ucred,
730 			    (struct sockaddr *)sin);
731 			if (error != 0)
732 				memset(sin, 0, sizeof(*sin));
733 			break;
734 #endif
735 #ifdef INET6
736 		case SIOCGIFPSRCADDR_IN6:
737 		case SIOCGIFPDSTADDR_IN6:
738 			error = prison_if(curthread->td_ucred,
739 			    (struct sockaddr *)sin6);
740 			if (error == 0)
741 				error = sa6_recoverscope(sin6);
742 			if (error != 0)
743 				memset(sin6, 0, sizeof(*sin6));
744 #endif
745 		}
746 		break;
747 	case SIOCGTUNFIB:
748 		ifr->ifr_fib = sc->fibnum;
749 		break;
750 	case SIOCSTUNFIB:
751 		if ((error = priv_check(curthread, PRIV_NET_SETIFFIB)) != 0)
752 			break;
753 		if (ifr->ifr_fib >= rt_numfibs)
754 			error = EINVAL;
755 		else
756 			sc->fibnum = ifr->ifr_fib;
757 		break;
758 	case IPSECGREQID:
759 		reqid = sc->reqid;
760 		error = copyout(&reqid, ifr_data_get_ptr(ifr), sizeof(reqid));
761 		break;
762 	case IPSECSREQID:
763 		if ((error = priv_check(curthread, PRIV_NET_SETIFCAP)) != 0)
764 			break;
765 		error = copyin(ifr_data_get_ptr(ifr), &reqid, sizeof(reqid));
766 		if (error != 0)
767 			break;
768 		error = ipsec_set_reqid(sc, reqid);
769 		break;
770 	default:
771 		error = EINVAL;
772 		break;
773 	}
774 bad:
775 	sx_xunlock(&ipsec_ioctl_sx);
776 	return (error);
777 }
778 
779 /*
780  * Check that ingress address belongs to local host.
781  */
782 static void
783 ipsec_set_running(struct ipsec_softc *sc)
784 {
785 	struct secasindex *saidx;
786 	int localip;
787 
788 	saidx = ipsec_getsaidx(sc, IPSEC_DIR_OUTBOUND, sc->family);
789 	if (saidx == NULL)
790 		return;
791 	localip = 0;
792 	switch (sc->family) {
793 #ifdef INET
794 	case AF_INET:
795 		localip = in_localip(saidx->src.sin.sin_addr);
796 		break;
797 #endif
798 #ifdef INET6
799 	case AF_INET6:
800 		localip = in6_localip(&saidx->src.sin6.sin6_addr);
801 		break;
802 #endif
803 	}
804 	if (localip != 0)
805 		sc->ifp->if_drv_flags |= IFF_DRV_RUNNING;
806 	else
807 		sc->ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
808 }
809 
810 /*
811  * ifaddr_event handler.
812  * Clear IFF_DRV_RUNNING flag when ingress address disappears to prevent
813  * source address spoofing.
814  */
815 static void
816 ipsec_srcaddr(void *arg __unused, const struct sockaddr *sa,
817     int event __unused)
818 {
819 	struct ipsec_softc *sc;
820 	struct secasindex *saidx;
821 	struct ipsec_iflist *iflist;
822 
823 	/* Check that VNET is ready */
824 	if (V_ipsec_idhtbl == NULL)
825 		return;
826 
827 	NET_EPOCH_ASSERT();
828 	iflist = ipsec_srchash(sa);
829 	if (iflist == NULL)
830 		return;
831 	CK_LIST_FOREACH(sc, iflist, srchash) {
832 		if (sc->family == 0)
833 			continue;
834 		saidx = ipsec_getsaidx(sc, IPSEC_DIR_OUTBOUND, sa->sa_family);
835 		if (saidx == NULL ||
836 		    key_sockaddrcmp(&saidx->src.sa, sa, 0) != 0)
837 			continue;
838 		ipsec_set_running(sc);
839 	}
840 }
841 
842 /*
843  * Allocate new private security policies for tunneling interface.
844  * Each tunneling interface has following security policies for
845  * both AF:
846  *   0.0.0.0/0[any] 0.0.0.0/0[any] -P in \
847  *	ipsec esp/tunnel/RemoteIP-LocalIP/unique:reqid
848  *   0.0.0.0/0[any] 0.0.0.0/0[any] -P out \
849  *	ipsec esp/tunnel/LocalIP-RemoteIP/unique:reqid
850  */
851 static int
852 ipsec_newpolicies(struct ipsec_softc *sc, struct secpolicy *sp[IPSEC_SPCOUNT],
853     const struct sockaddr *src, const struct sockaddr *dst, uint32_t reqid)
854 {
855 	struct ipsecrequest *isr;
856 	int i;
857 
858 	memset(sp, 0, sizeof(struct secpolicy *) * IPSEC_SPCOUNT);
859 	for (i = 0; i < IPSEC_SPCOUNT; i++) {
860 		if ((sp[i] = key_newsp()) == NULL)
861 			goto fail;
862 		if ((isr = ipsec_newisr()) == NULL)
863 			goto fail;
864 
865 		sp[i]->policy = IPSEC_POLICY_IPSEC;
866 		sp[i]->state = IPSEC_SPSTATE_DEAD;
867 		sp[i]->req[sp[i]->tcount++] = isr;
868 		sp[i]->created = time_second;
869 		/* Use priority field to store if_index */
870 		sp[i]->priority = sc->ifp->if_index;
871 		isr->level = IPSEC_LEVEL_UNIQUE;
872 		isr->saidx.proto = IPPROTO_ESP;
873 		isr->saidx.mode = IPSEC_MODE_TUNNEL;
874 		isr->saidx.reqid = reqid;
875 		if (i % 2 == 0) {
876 			sp[i]->spidx.dir = IPSEC_DIR_INBOUND;
877 			bcopy(src, &isr->saidx.dst, src->sa_len);
878 			bcopy(dst, &isr->saidx.src, dst->sa_len);
879 		} else {
880 			sp[i]->spidx.dir = IPSEC_DIR_OUTBOUND;
881 			bcopy(src, &isr->saidx.src, src->sa_len);
882 			bcopy(dst, &isr->saidx.dst, dst->sa_len);
883 		}
884 		sp[i]->spidx.ul_proto = IPSEC_ULPROTO_ANY;
885 #ifdef INET
886 		if (i < 2) {
887 			sp[i]->spidx.src.sa.sa_family =
888 			    sp[i]->spidx.dst.sa.sa_family = AF_INET;
889 			sp[i]->spidx.src.sa.sa_len =
890 			    sp[i]->spidx.dst.sa.sa_len =
891 			    sizeof(struct sockaddr_in);
892 			continue;
893 		}
894 #endif
895 #ifdef INET6
896 		sp[i]->spidx.src.sa.sa_family =
897 		    sp[i]->spidx.dst.sa.sa_family = AF_INET6;
898 		sp[i]->spidx.src.sa.sa_len =
899 		    sp[i]->spidx.dst.sa.sa_len = sizeof(struct sockaddr_in6);
900 #endif
901 	}
902 	return (0);
903 fail:
904 	for (i = 0; i < IPSEC_SPCOUNT; i++) {
905 		if (sp[i] != NULL)
906 			key_freesp(&sp[i]);
907 	}
908 	return (ENOMEM);
909 }
910 
911 static int
912 ipsec_check_reqid(uint32_t reqid)
913 {
914 	struct ipsec_softc *sc;
915 
916 	sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
917 	CK_LIST_FOREACH(sc, ipsec_idhash(reqid), idhash) {
918 		if (sc->reqid == reqid)
919 			return (EEXIST);
920 	}
921 	return (0);
922 }
923 
924 /*
925  * We use key_newreqid() to automatically obtain unique reqid.
926  * Then we check that given id is unique, i.e. it is not used by
927  * another if_ipsec(4) interface. This macro limits the number of
928  * tries to get unique id.
929  */
930 #define	IPSEC_REQID_TRYCNT	64
931 static int
932 ipsec_init_reqid(struct ipsec_softc *sc)
933 {
934 	uint32_t reqid;
935 	int trycount;
936 
937 	sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
938 	if (sc->reqid != 0) /* already initialized */
939 		return (0);
940 
941 	trycount = IPSEC_REQID_TRYCNT;
942 	while (--trycount > 0) {
943 		reqid = key_newreqid();
944 		if (ipsec_check_reqid(reqid) == 0)
945 			break;
946 	}
947 	if (trycount == 0)
948 		return (EEXIST);
949 	sc->reqid = reqid;
950 	CK_LIST_INSERT_HEAD(ipsec_idhash(reqid), sc, idhash);
951 	return (0);
952 }
953 
954 /*
955  * Set or update reqid for given tunneling interface.
956  * When specified reqid is zero, generate new one.
957  * We are protected by ioctl_sx lock from concurrent id generation.
958  * Also softc would not disappear while we hold ioctl_sx lock.
959  */
960 static int
961 ipsec_set_reqid(struct ipsec_softc *sc, uint32_t reqid)
962 {
963 	struct secasindex *saidx;
964 
965 	sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
966 
967 	if (sc->reqid == reqid && reqid != 0)
968 		return (0);
969 
970 	if (reqid != 0) {
971 		/* Check that specified reqid doesn't exist */
972 		if (ipsec_check_reqid(reqid) != 0)
973 			return (EEXIST);
974 		if (sc->reqid != 0) {
975 			CK_LIST_REMOVE(sc, idhash);
976 			IPSEC_WAIT();
977 		}
978 		sc->reqid = reqid;
979 		CK_LIST_INSERT_HEAD(ipsec_idhash(reqid), sc, idhash);
980 	} else {
981 		/* Generate new reqid */
982 		if (ipsec_init_reqid(sc) != 0)
983 			return (EEXIST);
984 	}
985 
986 	/* Tunnel isn't fully configured, just return. */
987 	if (sc->family == 0)
988 		return (0);
989 
990 	saidx = ipsec_getsaidx(sc, IPSEC_DIR_OUTBOUND, sc->family);
991 	KASSERT(saidx != NULL,
992 	    ("saidx is NULL, but family is %d", sc->family));
993 	return (ipsec_set_tunnel(sc, &saidx->src.sa, &saidx->dst.sa,
994 	    sc->reqid));
995 }
996 
997 /*
998  * Set tunnel endpoints addresses.
999  */
1000 static int
1001 ipsec_set_addresses(struct ifnet *ifp, struct sockaddr *src,
1002     struct sockaddr *dst)
1003 {
1004 	struct ipsec_softc *sc;
1005 	struct secasindex *saidx;
1006 
1007 	sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
1008 
1009 	sc = ifp->if_softc;
1010 	if (sc->family != 0) {
1011 		saidx = ipsec_getsaidx(sc, IPSEC_DIR_OUTBOUND,
1012 		    src->sa_family);
1013 		if (saidx != NULL && saidx->reqid == sc->reqid &&
1014 		    key_sockaddrcmp(&saidx->src.sa, src, 0) == 0 &&
1015 		    key_sockaddrcmp(&saidx->dst.sa, dst, 0) == 0)
1016 			return (0); /* Nothing has been changed. */
1017 	}
1018 	/* If reqid is not set, generate new one. */
1019 	if (ipsec_init_reqid(sc) != 0)
1020 		return (EEXIST);
1021 	return (ipsec_set_tunnel(sc, src, dst, sc->reqid));
1022 }
1023 
1024 static int
1025 ipsec_set_tunnel(struct ipsec_softc *sc, struct sockaddr *src,
1026     struct sockaddr *dst, uint32_t reqid)
1027 {
1028 	struct epoch_tracker et;
1029 	struct ipsec_iflist *iflist;
1030 	struct secpolicy *sp[IPSEC_SPCOUNT];
1031 	int i;
1032 
1033 	sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
1034 
1035 	/* Allocate SP with new addresses. */
1036 	iflist = ipsec_srchash(src);
1037 	if (iflist == NULL) {
1038 		sc->ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1039 		return (EAFNOSUPPORT);
1040 	}
1041 	if (ipsec_newpolicies(sc, sp, src, dst, reqid) == 0) {
1042 		/* Add new policies to SPDB */
1043 		if (key_register_ifnet(sp, IPSEC_SPCOUNT) != 0) {
1044 			for (i = 0; i < IPSEC_SPCOUNT; i++)
1045 				key_freesp(&sp[i]);
1046 			return (EAGAIN);
1047 		}
1048 		if (sc->family != 0)
1049 			ipsec_delete_tunnel(sc);
1050 		for (i = 0; i < IPSEC_SPCOUNT; i++)
1051 			sc->sp[i] = sp[i];
1052 		sc->family = src->sa_family;
1053 		CK_LIST_INSERT_HEAD(iflist, sc, srchash);
1054 	} else {
1055 		sc->ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1056 		return (ENOMEM);
1057 	}
1058 	NET_EPOCH_ENTER(et);
1059 	ipsec_set_running(sc);
1060 	NET_EPOCH_EXIT(et);
1061 	return (0);
1062 }
1063 
1064 static void
1065 ipsec_delete_tunnel(struct ipsec_softc *sc)
1066 {
1067 	int i;
1068 
1069 	sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
1070 
1071 	sc->ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1072 	if (sc->family != 0) {
1073 		CK_LIST_REMOVE(sc, srchash);
1074 		sc->family = 0;
1075 		/*
1076 		 * Make sure that ipsec_if_input() will not do access
1077 		 * to softc's policies.
1078 		 */
1079 		IPSEC_WAIT();
1080 
1081 		key_unregister_ifnet(sc->sp, IPSEC_SPCOUNT);
1082 		for (i = 0; i < IPSEC_SPCOUNT; i++)
1083 			key_freesp(&sc->sp[i]);
1084 	}
1085 }
1086