sys/net/if_ipsec.c

   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, iph;
 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 = &iph;
 379         m_copydata(m, 0, sizeof(*ip), (char *)ip);
 380         switch (ip->ip_v) {
 381 #ifdef INET
 382         case IPVERSION:
 383                 af = AF_INET;
 384                 break;
 385 #endif
 386 #ifdef INET6
 387         case (IPV6_VERSION >> 4):
 388                 af = AF_INET6;
 389                 break;
 390 #endif
 391         default:
 392                 error = EAFNOSUPPORT;
 393                 m_freem(m);
 394                 goto err;
 395         }
 396
 397         /*
 398          * Loop prevention.
 399          * XXX: for now just check presence of IPSEC_OUT_DONE mbuf tag.
 400          *      We can read full chain and compare destination address,
 401          *      proto and mode from xform_history with values from softc.
 402          */
 403         if (m_tag_find(m, PACKET_TAG_IPSEC_OUT_DONE, NULL) != NULL) {
 404                 m_freem(m);
 405                 goto err;
 406         }
 407
 408         sp = ipsec_getpolicy(sc, IPSEC_DIR_OUTBOUND, af);
 409         key_addref(sp);
 410         M_SETFIB(m, sc->fibnum);
 411
 412         BPF_MTAP2(ifp, &af, sizeof(af), m);
 413         if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
 414         if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len);
 415
 416         switch (af) {
 417 #ifdef INET
 418         case AF_INET:
 419                 error = ipsec4_process_packet(ifp, m, ip, sp, NULL,
 420                     ifp->if_mtu);
 421                 break;
 422 #endif
 423 #ifdef INET6
 424         case AF_INET6:
 425                 error = ipsec6_process_packet(ifp, m, sp, NULL, ifp->if_mtu);
 426                 break;
 427 #endif
 428         default:
 429                 panic("%s: unknown address family\n", __func__);
 430         }
 431 err:
 432         if (error != 0)
 433                 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
 434         IPSEC_RUNLOCK();
 435         return (error);
 436 }
 437
 438 static void
 439 ipsec_qflush(struct ifnet *ifp __unused)
 440 {
 441
 442 }
 443
 444 static int
 445 ipsec_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
 446         struct route *ro)
 447 {
 448
 449         return (ifp->if_transmit(ifp, m));
 450 }
 451
 452 int
 453 ipsec_if_input(struct mbuf *m, struct secasvar *sav, uint32_t af)
 454 {
 455         IPSEC_RLOCK_TRACKER;
 456         struct secasindex *saidx;
 457         struct ipsec_softc *sc;
 458         struct ifnet *ifp;
 459
 460         if (sav->state != SADB_SASTATE_MATURE &&
 461             sav->state != SADB_SASTATE_DYING) {
 462                 m_freem(m);
 463                 return (ENETDOWN);
 464         }
 465
 466         if (sav->sah->saidx.mode != IPSEC_MODE_TUNNEL ||
 467             sav->sah->saidx.proto != IPPROTO_ESP)
 468                 return (0);
 469
 470         IPSEC_RLOCK();
 471         CK_LIST_FOREACH(sc, ipsec_idhash(sav->sah->saidx.reqid), idhash) {
 472                 if (sc->family == 0)
 473                         continue;
 474                 saidx = ipsec_getsaidx(sc, IPSEC_DIR_INBOUND,
 475                     sav->sah->saidx.src.sa.sa_family);
 476                 /* SA's reqid should match reqid in SP */
 477                 if (saidx == NULL ||
 478                     sav->sah->saidx.reqid != saidx->reqid)
 479                         continue;
 480                 /* SAH's addresses should match tunnel endpoints. */
 481                 if (key_sockaddrcmp(&sav->sah->saidx.dst.sa,
 482                     &saidx->dst.sa, 0) != 0)
 483                         continue;
 484                 if (key_sockaddrcmp(&sav->sah->saidx.src.sa,
 485                     &saidx->src.sa, 0) == 0)
 486                         break;
 487         }
 488         if (sc == NULL) {
 489                 IPSEC_RUNLOCK();
 490                 /* Tunnel was not found. Nothing to do. */
 491                 return (0);
 492         }
 493         ifp = sc->ifp;
 494         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
 495             (ifp->if_flags & IFF_UP) == 0) {
 496                 IPSEC_RUNLOCK();
 497                 m_freem(m);
 498                 return (ENETDOWN);
 499         }
 500         /*
 501          * We found matching and working tunnel.
 502          * Set its ifnet as receiving interface.
 503          */
 504         m->m_pkthdr.rcvif = ifp;
 505
 506         m_clrprotoflags(m);
 507         M_SETFIB(m, ifp->if_fib);
 508         BPF_MTAP2(ifp, &af, sizeof(af), m);
 509         if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
 510         if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
 511         if ((ifp->if_flags & IFF_MONITOR) != 0) {
 512                 IPSEC_RUNLOCK();
 513                 m_freem(m);
 514                 return (ENETDOWN);
 515         }
 516         IPSEC_RUNLOCK();
 517         return (0);
 518 }
 519
 520 static int
 521 ipsec_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 522 {
 523         struct ifreq *ifr = (struct ifreq*)data;
 524         struct sockaddr *dst, *src;
 525         struct ipsec_softc *sc;
 526         struct secasindex *saidx;
 527 #ifdef INET
 528         struct sockaddr_in *sin = NULL;
 529 #endif
 530 #ifdef INET6
 531         struct sockaddr_in6 *sin6 = NULL;
 532 #endif
 533         uint32_t reqid;
 534         int error;
 535
 536         switch (cmd) {
 537         case SIOCSIFADDR:
 538                 ifp->if_flags |= IFF_UP;
 539         case SIOCADDMULTI:
 540         case SIOCDELMULTI:
 541         case SIOCGIFMTU:
 542         case SIOCSIFFLAGS:
 543                 return (0);
 544         case SIOCSIFMTU:
 545                 if (ifr->ifr_mtu < IPSEC_MTU_MIN ||
 546                     ifr->ifr_mtu > IPSEC_MTU_MAX)
 547                         return (EINVAL);
 548                 else
 549                         ifp->if_mtu = ifr->ifr_mtu;
 550                 return (0);
 551         }
 552         sx_xlock(&ipsec_ioctl_sx);
 553         sc = ifp->if_softc;
 554         /* Check that softc is still here */
 555         if (sc == NULL) {
 556                 error = ENXIO;
 557                 goto bad;
 558         }
 559         error = 0;
 560         switch (cmd) {
 561         case SIOCSIFPHYADDR:
 562 #ifdef INET6
 563         case SIOCSIFPHYADDR_IN6:
 564 #endif
 565                 error = EINVAL;
 566                 switch (cmd) {
 567 #ifdef INET
 568                 case SIOCSIFPHYADDR:
 569                         src = (struct sockaddr *)
 570                                 &(((struct in_aliasreq *)data)->ifra_addr);
 571                         dst = (struct sockaddr *)
 572                                 &(((struct in_aliasreq *)data)->ifra_dstaddr);
 573                         break;
 574 #endif
 575 #ifdef INET6
 576                 case SIOCSIFPHYADDR_IN6:
 577                         src = (struct sockaddr *)
 578                                 &(((struct in6_aliasreq *)data)->ifra_addr);
 579                         dst = (struct sockaddr *)
 580                                 &(((struct in6_aliasreq *)data)->ifra_dstaddr);
 581                         break;
 582 #endif
 583                 default:
 584                         goto bad;
 585                 }
 586                 /* sa_family must be equal */
 587                 if (src->sa_family != dst->sa_family ||
 588                     src->sa_len != dst->sa_len)
 589                         goto bad;
 590
 591                 /* validate sa_len */
 592                 switch (src->sa_family) {
 593 #ifdef INET
 594                 case AF_INET:
 595                         if (src->sa_len != sizeof(struct sockaddr_in))
 596                                 goto bad;
 597                         break;
 598 #endif
 599 #ifdef INET6
 600                 case AF_INET6:
 601                         if (src->sa_len != sizeof(struct sockaddr_in6))
 602                                 goto bad;
 603                         break;
 604 #endif
 605                 default:
 606                         error = EAFNOSUPPORT;
 607                         goto bad;
 608                 }
 609                 /* check sa_family looks sane for the cmd */
 610                 error = EAFNOSUPPORT;
 611                 switch (cmd) {
 612 #ifdef INET
 613                 case SIOCSIFPHYADDR:
 614                         if (src->sa_family == AF_INET)
 615                                 break;
 616                         goto bad;
 617 #endif
 618 #ifdef INET6
 619                 case SIOCSIFPHYADDR_IN6:
 620                         if (src->sa_family == AF_INET6)
 621                                 break;
 622                         goto bad;
 623 #endif
 624                 }
 625                 error = EADDRNOTAVAIL;
 626                 switch (src->sa_family) {
 627 #ifdef INET
 628                 case AF_INET:
 629                         if (satosin(src)->sin_addr.s_addr == INADDR_ANY ||
 630                             satosin(dst)->sin_addr.s_addr == INADDR_ANY)
 631                                 goto bad;
 632                         break;
 633 #endif
 634 #ifdef INET6
 635                 case AF_INET6:
 636                         if (IN6_IS_ADDR_UNSPECIFIED(
 637                             &satosin6(src)->sin6_addr) ||
 638                             IN6_IS_ADDR_UNSPECIFIED(
 639                             &satosin6(dst)->sin6_addr))
 640                                 goto bad;
 641                         /*
 642                          * Check validity of the scope zone ID of the
 643                          * addresses, and convert it into the kernel
 644                          * internal form if necessary.
 645                          */
 646                         error = sa6_embedscope(satosin6(src), 0);
 647                         if (error != 0)
 648                                 goto bad;
 649                         error = sa6_embedscope(satosin6(dst), 0);
 650                         if (error != 0)
 651                                 goto bad;
 652 #endif
 653                 };
 654                 error = ipsec_set_addresses(ifp, src, dst);
 655                 break;
 656         case SIOCDIFPHYADDR:
 657                 ipsec_delete_tunnel(sc);
 658                 break;
 659         case SIOCGIFPSRCADDR:
 660         case SIOCGIFPDSTADDR:
 661 #ifdef INET6
 662         case SIOCGIFPSRCADDR_IN6:
 663         case SIOCGIFPDSTADDR_IN6:
 664 #endif
 665                 if (sc->family == 0) {
 666                         error = EADDRNOTAVAIL;
 667                         break;
 668                 }
 669                 saidx = ipsec_getsaidx(sc, IPSEC_DIR_OUTBOUND, sc->family);
 670                 if (saidx == NULL) {
 671                         error = ENXIO;
 672                         break;
 673                 }
 674                 switch (cmd) {
 675 #ifdef INET
 676                 case SIOCGIFPSRCADDR:
 677                 case SIOCGIFPDSTADDR:
 678                         if (saidx->src.sa.sa_family != AF_INET) {
 679                                 error = EADDRNOTAVAIL;
 680                                 break;
 681                         }
 682                         sin = (struct sockaddr_in *)&ifr->ifr_addr;
 683                         memset(sin, 0, sizeof(*sin));
 684                         sin->sin_family = AF_INET;
 685                         sin->sin_len = sizeof(*sin);
 686                         break;
 687 #endif
 688 #ifdef INET6
 689                 case SIOCGIFPSRCADDR_IN6:
 690                 case SIOCGIFPDSTADDR_IN6:
 691                         if (saidx->src.sa.sa_family != AF_INET6) {
 692                                 error = EADDRNOTAVAIL;
 693                                 break;
 694                         }
 695                         sin6 = (struct sockaddr_in6 *)
 696                                 &(((struct in6_ifreq *)data)->ifr_addr);
 697                         memset(sin6, 0, sizeof(*sin6));
 698                         sin6->sin6_family = AF_INET6;
 699                         sin6->sin6_len = sizeof(*sin6);
 700                         break;
 701 #endif
 702                 default:
 703                         error = EAFNOSUPPORT;
 704                 }
 705                 if (error == 0) {
 706                         switch (cmd) {
 707 #ifdef INET
 708                         case SIOCGIFPSRCADDR:
 709                                 sin->sin_addr = saidx->src.sin.sin_addr;
 710                                 break;
 711                         case SIOCGIFPDSTADDR:
 712                                 sin->sin_addr = saidx->dst.sin.sin_addr;
 713                                 break;
 714 #endif
 715 #ifdef INET6
 716                         case SIOCGIFPSRCADDR_IN6:
 717                                 sin6->sin6_addr = saidx->src.sin6.sin6_addr;
 718                                 break;
 719                         case SIOCGIFPDSTADDR_IN6:
 720                                 sin6->sin6_addr = saidx->dst.sin6.sin6_addr;
 721                                 break;
 722 #endif
 723                         }
 724                 }
 725                 if (error != 0)
 726                         break;
 727                 switch (cmd) {
 728 #ifdef INET
 729                 case SIOCGIFPSRCADDR:
 730                 case SIOCGIFPDSTADDR:
 731                         error = prison_if(curthread->td_ucred,
 732                             (struct sockaddr *)sin);
 733                         if (error != 0)
 734                                 memset(sin, 0, sizeof(*sin));
 735                         break;
 736 #endif
 737 #ifdef INET6
 738                 case SIOCGIFPSRCADDR_IN6:
 739                 case SIOCGIFPDSTADDR_IN6:
 740                         error = prison_if(curthread->td_ucred,
 741                             (struct sockaddr *)sin6);
 742                         if (error == 0)
 743                                 error = sa6_recoverscope(sin6);
 744                         if (error != 0)
 745                                 memset(sin6, 0, sizeof(*sin6));
 746 #endif
 747                 }
 748                 break;
 749         case SIOCGTUNFIB:
 750                 ifr->ifr_fib = sc->fibnum;
 751                 break;
 752         case SIOCSTUNFIB:
 753                 if ((error = priv_check(curthread, PRIV_NET_SETIFFIB)) != 0)
 754                         break;
 755                 if (ifr->ifr_fib >= rt_numfibs)
 756                         error = EINVAL;
 757                 else
 758                         sc->fibnum = ifr->ifr_fib;
 759                 break;
 760         case IPSECGREQID:
 761                 reqid = sc->reqid;
 762                 error = copyout(&reqid, ifr_data_get_ptr(ifr), sizeof(reqid));
 763                 break;
 764         case IPSECSREQID:
 765                 if ((error = priv_check(curthread, PRIV_NET_SETIFCAP)) != 0)
 766                         break;
 767                 error = copyin(ifr_data_get_ptr(ifr), &reqid, sizeof(reqid));
 768                 if (error != 0)
 769                         break;
 770                 error = ipsec_set_reqid(sc, reqid);
 771                 break;
 772         default:
 773                 error = EINVAL;
 774                 break;
 775         }
 776 bad:
 777         sx_xunlock(&ipsec_ioctl_sx);
 778         return (error);
 779 }
 780
 781 /*
 782  * Check that ingress address belongs to local host.
 783  */
 784 static void
 785 ipsec_set_running(struct ipsec_softc *sc)
 786 {
 787         struct secasindex *saidx;
 788         int localip;
 789
 790         saidx = ipsec_getsaidx(sc, IPSEC_DIR_OUTBOUND, sc->family);
 791         if (saidx == NULL)
 792                 return;
 793         localip = 0;
 794         switch (sc->family) {
 795 #ifdef INET
 796         case AF_INET:
 797                 localip = in_localip(saidx->src.sin.sin_addr);
 798                 break;
 799 #endif
 800 #ifdef INET6
 801         case AF_INET6:
 802                 localip = in6_localip(&saidx->src.sin6.sin6_addr);
 803                 break;
 804 #endif
 805         }
 806         if (localip != 0)
 807                 sc->ifp->if_drv_flags |= IFF_DRV_RUNNING;
 808         else
 809                 sc->ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
 810 }
 811
 812 /*
 813  * ifaddr_event handler.
 814  * Clear IFF_DRV_RUNNING flag when ingress address disappears to prevent
 815  * source address spoofing.
 816  */
 817 static void
 818 ipsec_srcaddr(void *arg __unused, const struct sockaddr *sa,
 819     int event __unused)
 820 {
 821         struct ipsec_softc *sc;
 822         struct secasindex *saidx;
 823         struct ipsec_iflist *iflist;
 824
 825         /* Check that VNET is ready */
 826         if (V_ipsec_idhtbl == NULL)
 827                 return;
 828
 829         NET_EPOCH_ASSERT();
 830         iflist = ipsec_srchash(sa);
 831         if (iflist == NULL)
 832                 return;
 833         CK_LIST_FOREACH(sc, iflist, srchash) {
 834                 if (sc->family == 0)
 835                         continue;
 836                 saidx = ipsec_getsaidx(sc, IPSEC_DIR_OUTBOUND, sa->sa_family);
 837                 if (saidx == NULL ||
 838                     key_sockaddrcmp(&saidx->src.sa, sa, 0) != 0)
 839                         continue;
 840                 ipsec_set_running(sc);
 841         }
 842 }
 843
 844 /*
 845  * Allocate new private security policies for tunneling interface.
 846  * Each tunneling interface has following security policies for
 847  * both AF:
 848  *   0.0.0.0/0[any] 0.0.0.0/0[any] -P in \
 849  *      ipsec esp/tunnel/RemoteIP-LocalIP/unique:reqid
 850  *   0.0.0.0/0[any] 0.0.0.0/0[any] -P out \
 851  *      ipsec esp/tunnel/LocalIP-RemoteIP/unique:reqid
 852  */
 853 static int
 854 ipsec_newpolicies(struct ipsec_softc *sc, struct secpolicy *sp[IPSEC_SPCOUNT],
 855     const struct sockaddr *src, const struct sockaddr *dst, uint32_t reqid)
 856 {
 857         struct ipsecrequest *isr;
 858         int i;
 859
 860         memset(sp, 0, sizeof(struct secpolicy *) * IPSEC_SPCOUNT);
 861         for (i = 0; i < IPSEC_SPCOUNT; i++) {
 862                 if ((sp[i] = key_newsp()) == NULL)
 863                         goto fail;
 864                 if ((isr = ipsec_newisr()) == NULL)
 865                         goto fail;
 866
 867                 sp[i]->policy = IPSEC_POLICY_IPSEC;
 868                 sp[i]->state = IPSEC_SPSTATE_DEAD;
 869                 sp[i]->req[sp[i]->tcount++] = isr;
 870                 sp[i]->created = time_second;
 871                 /* Use priority field to store if_index */
 872                 sp[i]->priority = sc->ifp->if_index;
 873                 isr->level = IPSEC_LEVEL_UNIQUE;
 874                 isr->saidx.proto = IPPROTO_ESP;
 875                 isr->saidx.mode = IPSEC_MODE_TUNNEL;
 876                 isr->saidx.reqid = reqid;
 877                 if (i % 2 == 0) {
 878                         sp[i]->spidx.dir = IPSEC_DIR_INBOUND;
 879                         bcopy(src, &isr->saidx.dst, src->sa_len);
 880                         bcopy(dst, &isr->saidx.src, dst->sa_len);
 881                 } else {
 882                         sp[i]->spidx.dir = IPSEC_DIR_OUTBOUND;
 883                         bcopy(src, &isr->saidx.src, src->sa_len);
 884                         bcopy(dst, &isr->saidx.dst, dst->sa_len);
 885                 }
 886                 sp[i]->spidx.ul_proto = IPSEC_ULPROTO_ANY;
 887 #ifdef INET
 888                 if (i < 2) {
 889                         sp[i]->spidx.src.sa.sa_family =
 890                             sp[i]->spidx.dst.sa.sa_family = AF_INET;
 891                         sp[i]->spidx.src.sa.sa_len =
 892                             sp[i]->spidx.dst.sa.sa_len =
 893                             sizeof(struct sockaddr_in);
 894                         continue;
 895                 }
 896 #endif
 897 #ifdef INET6
 898                 sp[i]->spidx.src.sa.sa_family =
 899                     sp[i]->spidx.dst.sa.sa_family = AF_INET6;
 900                 sp[i]->spidx.src.sa.sa_len =
 901                     sp[i]->spidx.dst.sa.sa_len = sizeof(struct sockaddr_in6);
 902 #endif
 903         }
 904         return (0);
 905 fail:
 906         for (i = 0; i < IPSEC_SPCOUNT; i++) {
 907                 if (sp[i] != NULL)
 908                         key_freesp(&sp[i]);
 909         }
 910         return (ENOMEM);
 911 }
 912
 913 static int
 914 ipsec_check_reqid(uint32_t reqid)
 915 {
 916         struct ipsec_softc *sc;
 917
 918         sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
 919         CK_LIST_FOREACH(sc, ipsec_idhash(reqid), idhash) {
 920                 if (sc->reqid == reqid)
 921                         return (EEXIST);
 922         }
 923         return (0);
 924 }
 925
 926 /*
 927  * We use key_newreqid() to automatically obtain unique reqid.
 928  * Then we check that given id is unique, i.e. it is not used by
 929  * another if_ipsec(4) interface. This macro limits the number of
 930  * tries to get unique id.
 931  */
 932 #define IPSEC_REQID_TRYCNT      64
 933 static int
 934 ipsec_init_reqid(struct ipsec_softc *sc)
 935 {
 936         uint32_t reqid;
 937         int trycount;
 938
 939         sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
 940         if (sc->reqid != 0) /* already initialized */
 941                 return (0);
 942
 943         trycount = IPSEC_REQID_TRYCNT;
 944         while (--trycount > 0) {
 945                 reqid = key_newreqid();
 946                 if (ipsec_check_reqid(reqid) == 0)
 947                         break;
 948         }
 949         if (trycount == 0)
 950                 return (EEXIST);
 951         sc->reqid = reqid;
 952         CK_LIST_INSERT_HEAD(ipsec_idhash(reqid), sc, idhash);
 953         return (0);
 954 }
 955
 956 /*
 957  * Set or update reqid for given tunneling interface.
 958  * When specified reqid is zero, generate new one.
 959  * We are protected by ioctl_sx lock from concurrent id generation.
 960  * Also softc would not disappear while we hold ioctl_sx lock.
 961  */
 962 static int
 963 ipsec_set_reqid(struct ipsec_softc *sc, uint32_t reqid)
 964 {
 965         struct secasindex *saidx;
 966
 967         sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
 968
 969         if (sc->reqid == reqid && reqid != 0)
 970                 return (0);
 971
 972         if (reqid != 0) {
 973                 /* Check that specified reqid doesn't exist */
 974                 if (ipsec_check_reqid(reqid) != 0)
 975                         return (EEXIST);
 976                 if (sc->reqid != 0) {
 977                         CK_LIST_REMOVE(sc, idhash);
 978                         IPSEC_WAIT();
 979                 }
 980                 sc->reqid = reqid;
 981                 CK_LIST_INSERT_HEAD(ipsec_idhash(reqid), sc, idhash);
 982         } else {
 983                 /* Generate new reqid */
 984                 if (ipsec_init_reqid(sc) != 0)
 985                         return (EEXIST);
 986         }
 987
 988         /* Tunnel isn't fully configured, just return. */
 989         if (sc->family == 0)
 990                 return (0);
 991
 992         saidx = ipsec_getsaidx(sc, IPSEC_DIR_OUTBOUND, sc->family);
 993         KASSERT(saidx != NULL,
 994             ("saidx is NULL, but family is %d", sc->family));
 995         return (ipsec_set_tunnel(sc, &saidx->src.sa, &saidx->dst.sa,
 996             sc->reqid));
 997 }
 998
 999 /*
1000  * Set tunnel endpoints addresses.
1001  */
1002 static int
1003 ipsec_set_addresses(struct ifnet *ifp, struct sockaddr *src,
1004     struct sockaddr *dst)
1005 {
1006         struct ipsec_softc *sc;
1007         struct secasindex *saidx;
1008
1009         sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
1010
1011         sc = ifp->if_softc;
1012         if (sc->family != 0) {
1013                 saidx = ipsec_getsaidx(sc, IPSEC_DIR_OUTBOUND,
1014                     src->sa_family);
1015                 if (saidx != NULL && saidx->reqid == sc->reqid &&
1016                     key_sockaddrcmp(&saidx->src.sa, src, 0) == 0 &&
1017                     key_sockaddrcmp(&saidx->dst.sa, dst, 0) == 0)
1018                         return (0); /* Nothing has been changed. */
1019         }
1020         /* If reqid is not set, generate new one. */
1021         if (ipsec_init_reqid(sc) != 0)
1022                 return (EEXIST);
1023         return (ipsec_set_tunnel(sc, src, dst, sc->reqid));
1024 }
1025
1026 static int
1027 ipsec_set_tunnel(struct ipsec_softc *sc, struct sockaddr *src,
1028     struct sockaddr *dst, uint32_t reqid)
1029 {
1030         struct epoch_tracker et;
1031         struct ipsec_iflist *iflist;
1032         struct secpolicy *sp[IPSEC_SPCOUNT];
1033         int i;
1034
1035         sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
1036
1037         /* Allocate SP with new addresses. */
1038         iflist = ipsec_srchash(src);
1039         if (iflist == NULL) {
1040                 sc->ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1041                 return (EAFNOSUPPORT);
1042         }
1043         if (ipsec_newpolicies(sc, sp, src, dst, reqid) == 0) {
1044                 /* Add new policies to SPDB */
1045                 if (key_register_ifnet(sp, IPSEC_SPCOUNT) != 0) {
1046                         for (i = 0; i < IPSEC_SPCOUNT; i++)
1047                                 key_freesp(&sp[i]);
1048                         return (EAGAIN);
1049                 }
1050                 if (sc->family != 0)
1051                         ipsec_delete_tunnel(sc);
1052                 for (i = 0; i < IPSEC_SPCOUNT; i++)
1053                         sc->sp[i] = sp[i];
1054                 sc->family = src->sa_family;
1055                 CK_LIST_INSERT_HEAD(iflist, sc, srchash);
1056         } else {
1057                 sc->ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1058                 return (ENOMEM);
1059         }
1060         NET_EPOCH_ENTER(et);
1061         ipsec_set_running(sc);
1062         NET_EPOCH_EXIT(et);
1063         return (0);
1064 }
1065
1066 static void
1067 ipsec_delete_tunnel(struct ipsec_softc *sc)
1068 {
1069         int i;
1070
1071         sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
1072
1073         sc->ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1074         if (sc->family != 0) {
1075                 CK_LIST_REMOVE(sc, srchash);
1076                 sc->family = 0;
1077                 /*
1078                  * Make sure that ipsec_if_input() will not do access
1079                  * to softc's policies.
1080                  */
1081                 IPSEC_WAIT();
1082
1083                 key_unregister_ifnet(sc->sp, IPSEC_SPCOUNT);
1084                 for (i = 0; i < IPSEC_SPCOUNT; i++)
1085                         key_freesp(&sc->sp[i]);
1086         }
1087 }