sys/net/rtsock.c

   1 /*-
   2  * Copyright (c) 1988, 1991, 1993
   3  *      The Regents of the University of California.  All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  * 4. Neither the name of the University nor the names of its contributors
  14  *    may be used to endorse or promote products derived from this software
  15  *    without specific prior written permission.
  16  *
  17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  27  * SUCH DAMAGE.
  28  *
  29  *      @(#)rtsock.c    8.7 (Berkeley) 10/12/95
  30  * $FreeBSD$
  31  */
  32 #include "opt_compat.h"
  33 #include "opt_sctp.h"
  34 #include "opt_mpath.h"
  35 #include "opt_inet.h"
  36 #include "opt_inet6.h"
  37
  38 #include <sys/param.h>
  39 #include <sys/jail.h>
  40 #include <sys/kernel.h>
  41 #include <sys/domain.h>
  42 #include <sys/lock.h>
  43 #include <sys/malloc.h>
  44 #include <sys/mbuf.h>
  45 #include <sys/priv.h>
  46 #include <sys/proc.h>
  47 #include <sys/protosw.h>
  48 #include <sys/rwlock.h>
  49 #include <sys/signalvar.h>
  50 #include <sys/socket.h>
  51 #include <sys/socketvar.h>
  52 #include <sys/sysctl.h>
  53 #include <sys/systm.h>
  54
  55 #include <net/if.h>
  56 #include <net/if_dl.h>
  57 #include <net/if_llatbl.h>
  58 #include <net/if_types.h>
  59 #include <net/netisr.h>
  60 #include <net/raw_cb.h>
  61 #include <net/route.h>
  62 #include <net/vnet.h>
  63
  64 #include <netinet/in.h>
  65 #include <netinet/if_ether.h>
  66 #ifdef INET6
  67 #include <netinet6/scope6_var.h>
  68 #endif
  69
  70 #if defined(INET) || defined(INET6)
  71 #ifdef SCTP
  72 extern void sctp_addr_change(struct ifaddr *ifa, int cmd);
  73 #endif /* SCTP */
  74 #endif
  75
  76 #ifdef COMPAT_FREEBSD32
  77 #include <sys/mount.h>
  78 #include <compat/freebsd32/freebsd32.h>
  79
  80 struct if_data32 {
  81         uint8_t ifi_type;
  82         uint8_t ifi_physical;
  83         uint8_t ifi_addrlen;
  84         uint8_t ifi_hdrlen;
  85         uint8_t ifi_link_state;
  86         uint8_t ifi_spare_char1;
  87         uint8_t ifi_spare_char2;
  88         uint8_t ifi_datalen;
  89         uint32_t ifi_mtu;
  90         uint32_t ifi_metric;
  91         uint32_t ifi_baudrate;
  92         uint32_t ifi_ipackets;
  93         uint32_t ifi_ierrors;
  94         uint32_t ifi_opackets;
  95         uint32_t ifi_oerrors;
  96         uint32_t ifi_collisions;
  97         uint32_t ifi_ibytes;
  98         uint32_t ifi_obytes;
  99         uint32_t ifi_imcasts;
 100         uint32_t ifi_omcasts;
 101         uint32_t ifi_iqdrops;
 102         uint32_t ifi_noproto;
 103         uint32_t ifi_hwassist;
 104         int32_t ifi_epoch;
 105         struct  timeval32 ifi_lastchange;
 106 };
 107
 108 struct if_msghdr32 {
 109         uint16_t ifm_msglen;
 110         uint8_t ifm_version;
 111         uint8_t ifm_type;
 112         int32_t ifm_addrs;
 113         int32_t ifm_flags;
 114         uint16_t ifm_index;
 115         struct  if_data32 ifm_data;
 116 };
 117 #endif
 118
 119 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
 120
 121 /* NB: these are not modified */
 122 static struct   sockaddr route_src = { 2, PF_ROUTE, };
 123 static struct   sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
 124
 125 static struct {
 126         int     ip_count;       /* attached w/ AF_INET */
 127         int     ip6_count;      /* attached w/ AF_INET6 */
 128         int     ipx_count;      /* attached w/ AF_IPX */
 129         int     any_count;      /* total attached */
 130 } route_cb;
 131
 132 struct mtx rtsock_mtx;
 133 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
 134
 135 #define RTSOCK_LOCK()   mtx_lock(&rtsock_mtx)
 136 #define RTSOCK_UNLOCK() mtx_unlock(&rtsock_mtx)
 137 #define RTSOCK_LOCK_ASSERT()    mtx_assert(&rtsock_mtx, MA_OWNED)
 138
 139 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
 140
 141 struct walkarg {
 142         int     w_tmemsize;
 143         int     w_op, w_arg;
 144         caddr_t w_tmem;
 145         struct sysctl_req *w_req;
 146 };
 147
 148 static void     rts_input(struct mbuf *m);
 149 static struct mbuf *rt_msg1(int type, struct rt_addrinfo *rtinfo);
 150 static int      rt_msg2(int type, struct rt_addrinfo *rtinfo,
 151                         caddr_t cp, struct walkarg *w);
 152 static int      rt_xaddrs(caddr_t cp, caddr_t cplim,
 153                         struct rt_addrinfo *rtinfo);
 154 static int      sysctl_dumpentry(struct radix_node *rn, void *vw);
 155 static int      sysctl_iflist(int af, struct walkarg *w);
 156 static int      sysctl_ifmalist(int af, struct walkarg *w);
 157 static int      route_output(struct mbuf *m, struct socket *so);
 158 static void     rt_setmetrics(u_long which, const struct rt_metrics *in,
 159                         struct rt_metrics_lite *out);
 160 static void     rt_getmetrics(const struct rt_metrics_lite *in,
 161                         struct rt_metrics *out);
 162 static void     rt_dispatch(struct mbuf *, const struct sockaddr *);
 163
 164 static struct netisr_handler rtsock_nh = {
 165         .nh_name = "rtsock",
 166         .nh_handler = rts_input,
 167         .nh_proto = NETISR_ROUTE,
 168         .nh_policy = NETISR_POLICY_SOURCE,
 169 };
 170
 171 static int
 172 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
 173 {
 174         int error, qlimit;
 175
 176         netisr_getqlimit(&rtsock_nh, &qlimit);
 177         error = sysctl_handle_int(oidp, &qlimit, 0, req);
 178         if (error || !req->newptr)
 179                 return (error);
 180         if (qlimit < 1)
 181                 return (EINVAL);
 182         return (netisr_setqlimit(&rtsock_nh, qlimit));
 183 }
 184 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
 185     0, 0, sysctl_route_netisr_maxqlen, "I",
 186     "maximum routing socket dispatch queue length");
 187
 188 static void
 189 rts_init(void)
 190 {
 191         int tmp;
 192
 193         if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
 194                 rtsock_nh.nh_qlimit = tmp;
 195         netisr_register(&rtsock_nh);
 196 }
 197 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0);
 198
 199 static void
 200 rts_input(struct mbuf *m)
 201 {
 202         struct sockproto route_proto;
 203         unsigned short *family;
 204         struct m_tag *tag;
 205
 206         route_proto.sp_family = PF_ROUTE;
 207         tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
 208         if (tag != NULL) {
 209                 family = (unsigned short *)(tag + 1);
 210                 route_proto.sp_protocol = *family;
 211                 m_tag_delete(m, tag);
 212         } else
 213                 route_proto.sp_protocol = 0;
 214
 215         raw_input(m, &route_proto, &route_src);
 216 }
 217
 218 /*
 219  * It really doesn't make any sense at all for this code to share much
 220  * with raw_usrreq.c, since its functionality is so restricted.  XXX
 221  */
 222 static void
 223 rts_abort(struct socket *so)
 224 {
 225
 226         raw_usrreqs.pru_abort(so);
 227 }
 228
 229 static void
 230 rts_close(struct socket *so)
 231 {
 232
 233         raw_usrreqs.pru_close(so);
 234 }
 235
 236 /* pru_accept is EOPNOTSUPP */
 237
 238 static int
 239 rts_attach(struct socket *so, int proto, struct thread *td)
 240 {
 241         struct rawcb *rp;
 242         int s, error;
 243
 244         KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
 245
 246         /* XXX */
 247         rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
 248         if (rp == NULL)
 249                 return ENOBUFS;
 250
 251         /*
 252          * The splnet() is necessary to block protocols from sending
 253          * error notifications (like RTM_REDIRECT or RTM_LOSING) while
 254          * this PCB is extant but incompletely initialized.
 255          * Probably we should try to do more of this work beforehand and
 256          * eliminate the spl.
 257          */
 258         s = splnet();
 259         so->so_pcb = (caddr_t)rp;
 260         so->so_fibnum = td->td_proc->p_fibnum;
 261         error = raw_attach(so, proto);
 262         rp = sotorawcb(so);
 263         if (error) {
 264                 splx(s);
 265                 so->so_pcb = NULL;
 266                 free(rp, M_PCB);
 267                 return error;
 268         }
 269         RTSOCK_LOCK();
 270         switch(rp->rcb_proto.sp_protocol) {
 271         case AF_INET:
 272                 route_cb.ip_count++;
 273                 break;
 274         case AF_INET6:
 275                 route_cb.ip6_count++;
 276                 break;
 277         case AF_IPX:
 278                 route_cb.ipx_count++;
 279                 break;
 280         }
 281         route_cb.any_count++;
 282         RTSOCK_UNLOCK();
 283         soisconnected(so);
 284         so->so_options |= SO_USELOOPBACK;
 285         splx(s);
 286         return 0;
 287 }
 288
 289 static int
 290 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
 291 {
 292
 293         return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
 294 }
 295
 296 static int
 297 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
 298 {
 299
 300         return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
 301 }
 302
 303 /* pru_connect2 is EOPNOTSUPP */
 304 /* pru_control is EOPNOTSUPP */
 305
 306 static void
 307 rts_detach(struct socket *so)
 308 {
 309         struct rawcb *rp = sotorawcb(so);
 310
 311         KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
 312
 313         RTSOCK_LOCK();
 314         switch(rp->rcb_proto.sp_protocol) {
 315         case AF_INET:
 316                 route_cb.ip_count--;
 317                 break;
 318         case AF_INET6:
 319                 route_cb.ip6_count--;
 320                 break;
 321         case AF_IPX:
 322                 route_cb.ipx_count--;
 323                 break;
 324         }
 325         route_cb.any_count--;
 326         RTSOCK_UNLOCK();
 327         raw_usrreqs.pru_detach(so);
 328 }
 329
 330 static int
 331 rts_disconnect(struct socket *so)
 332 {
 333
 334         return (raw_usrreqs.pru_disconnect(so));
 335 }
 336
 337 /* pru_listen is EOPNOTSUPP */
 338
 339 static int
 340 rts_peeraddr(struct socket *so, struct sockaddr **nam)
 341 {
 342
 343         return (raw_usrreqs.pru_peeraddr(so, nam));
 344 }
 345
 346 /* pru_rcvd is EOPNOTSUPP */
 347 /* pru_rcvoob is EOPNOTSUPP */
 348
 349 static int
 350 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
 351          struct mbuf *control, struct thread *td)
 352 {
 353
 354         return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
 355 }
 356
 357 /* pru_sense is null */
 358
 359 static int
 360 rts_shutdown(struct socket *so)
 361 {
 362
 363         return (raw_usrreqs.pru_shutdown(so));
 364 }
 365
 366 static int
 367 rts_sockaddr(struct socket *so, struct sockaddr **nam)
 368 {
 369
 370         return (raw_usrreqs.pru_sockaddr(so, nam));
 371 }
 372
 373 static struct pr_usrreqs route_usrreqs = {
 374         .pru_abort =            rts_abort,
 375         .pru_attach =           rts_attach,
 376         .pru_bind =             rts_bind,
 377         .pru_connect =          rts_connect,
 378         .pru_detach =           rts_detach,
 379         .pru_disconnect =       rts_disconnect,
 380         .pru_peeraddr =         rts_peeraddr,
 381         .pru_send =             rts_send,
 382         .pru_shutdown =         rts_shutdown,
 383         .pru_sockaddr =         rts_sockaddr,
 384         .pru_close =            rts_close,
 385 };
 386
 387 #ifndef _SOCKADDR_UNION_DEFINED
 388 #define _SOCKADDR_UNION_DEFINED
 389 /*
 390  * The union of all possible address formats we handle.
 391  */
 392 union sockaddr_union {
 393         struct sockaddr         sa;
 394         struct sockaddr_in      sin;
 395         struct sockaddr_in6     sin6;
 396 };
 397 #endif /* _SOCKADDR_UNION_DEFINED */
 398
 399 static int
 400 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
 401     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
 402 {
 403
 404         /* First, see if the returned address is part of the jail. */
 405         if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
 406                 info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
 407                 return (0);
 408         }
 409
 410         switch (info->rti_info[RTAX_DST]->sa_family) {
 411 #ifdef INET
 412         case AF_INET:
 413         {
 414                 struct in_addr ia;
 415                 struct ifaddr *ifa;
 416                 int found;
 417
 418                 found = 0;
 419                 /*
 420                  * Try to find an address on the given outgoing interface
 421                  * that belongs to the jail.
 422                  */
 423                 IF_ADDR_LOCK(ifp);
 424                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 425                         struct sockaddr *sa;
 426                         sa = ifa->ifa_addr;
 427                         if (sa->sa_family != AF_INET)
 428                                 continue;
 429                         ia = ((struct sockaddr_in *)sa)->sin_addr;
 430                         if (prison_check_ip4(cred, &ia) == 0) {
 431                                 found = 1;
 432                                 break;
 433                         }
 434                 }
 435                 IF_ADDR_UNLOCK(ifp);
 436                 if (!found) {
 437                         /*
 438                          * As a last resort return the 'default' jail address.
 439                          */
 440                         ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
 441                             sin_addr;
 442                         if (prison_get_ip4(cred, &ia) != 0)
 443                                 return (ESRCH);
 444                 }
 445                 bzero(&saun->sin, sizeof(struct sockaddr_in));
 446                 saun->sin.sin_len = sizeof(struct sockaddr_in);
 447                 saun->sin.sin_family = AF_INET;
 448                 saun->sin.sin_addr.s_addr = ia.s_addr;
 449                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
 450                 break;
 451         }
 452 #endif
 453 #ifdef INET6
 454         case AF_INET6:
 455         {
 456                 struct in6_addr ia6;
 457                 struct ifaddr *ifa;
 458                 int found;
 459
 460                 found = 0;
 461                 /*
 462                  * Try to find an address on the given outgoing interface
 463                  * that belongs to the jail.
 464                  */
 465                 IF_ADDR_LOCK(ifp);
 466                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 467                         struct sockaddr *sa;
 468                         sa = ifa->ifa_addr;
 469                         if (sa->sa_family != AF_INET6)
 470                                 continue;
 471                         bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
 472                             &ia6, sizeof(struct in6_addr));
 473                         if (prison_check_ip6(cred, &ia6) == 0) {
 474                                 found = 1;
 475                                 break;
 476                         }
 477                 }
 478                 IF_ADDR_UNLOCK(ifp);
 479                 if (!found) {
 480                         /*
 481                          * As a last resort return the 'default' jail address.
 482                          */
 483                         ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
 484                             sin6_addr;
 485                         if (prison_get_ip6(cred, &ia6) != 0)
 486                                 return (ESRCH);
 487                 }
 488                 bzero(&saun->sin6, sizeof(struct sockaddr_in6));
 489                 saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
 490                 saun->sin6.sin6_family = AF_INET6;
 491                 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
 492                 if (sa6_recoverscope(&saun->sin6) != 0)
 493                         return (ESRCH);
 494                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
 495                 break;
 496         }
 497 #endif
 498         default:
 499                 return (ESRCH);
 500         }
 501         return (0);
 502 }
 503
 504 /*ARGSUSED*/
 505 static int
 506 route_output(struct mbuf *m, struct socket *so)
 507 {
 508 #define sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
 509         struct rt_msghdr *rtm = NULL;
 510         struct rtentry *rt = NULL;
 511         struct radix_node_head *rnh;
 512         struct rt_addrinfo info;
 513         int len, error = 0;
 514         struct ifnet *ifp = NULL;
 515         union sockaddr_union saun;
 516
 517 #define senderr(e) { error = e; goto flush;}
 518         if (m == NULL || ((m->m_len < sizeof(long)) &&
 519                        (m = m_pullup(m, sizeof(long))) == NULL))
 520                 return (ENOBUFS);
 521         if ((m->m_flags & M_PKTHDR) == 0)
 522                 panic("route_output");
 523         len = m->m_pkthdr.len;
 524         if (len < sizeof(*rtm) ||
 525             len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
 526                 info.rti_info[RTAX_DST] = NULL;
 527                 senderr(EINVAL);
 528         }
 529         R_Malloc(rtm, struct rt_msghdr *, len);
 530         if (rtm == NULL) {
 531                 info.rti_info[RTAX_DST] = NULL;
 532                 senderr(ENOBUFS);
 533         }
 534         m_copydata(m, 0, len, (caddr_t)rtm);
 535         if (rtm->rtm_version != RTM_VERSION) {
 536                 info.rti_info[RTAX_DST] = NULL;
 537                 senderr(EPROTONOSUPPORT);
 538         }
 539         rtm->rtm_pid = curproc->p_pid;
 540         bzero(&info, sizeof(info));
 541         info.rti_addrs = rtm->rtm_addrs;
 542         if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
 543                 info.rti_info[RTAX_DST] = NULL;
 544                 senderr(EINVAL);
 545         }
 546         info.rti_flags = rtm->rtm_flags;
 547         if (info.rti_info[RTAX_DST] == NULL ||
 548             info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
 549             (info.rti_info[RTAX_GATEWAY] != NULL &&
 550              info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
 551                 senderr(EINVAL);
 552         /*
 553          * Verify that the caller has the appropriate privilege; RTM_GET
 554          * is the only operation the non-superuser is allowed.
 555          */
 556         if (rtm->rtm_type != RTM_GET) {
 557                 error = priv_check(curthread, PRIV_NET_ROUTE);
 558                 if (error)
 559                         senderr(error);
 560         }
 561
 562         /*
 563          * The given gateway address may be an interface address.
 564          * For example, issuing a "route change" command on a route
 565          * entry that was created from a tunnel, and the gateway
 566          * address given is the local end point. In this case the
 567          * RTF_GATEWAY flag must be cleared or the destination will
 568          * not be reachable even though there is no error message.
 569          */
 570         if (info.rti_info[RTAX_GATEWAY] != NULL &&
 571             info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
 572                 struct route gw_ro;
 573
 574                 bzero(&gw_ro, sizeof(gw_ro));
 575                 gw_ro.ro_dst = *info.rti_info[RTAX_GATEWAY];
 576                 rtalloc_ign_fib(&gw_ro, 0, so->so_fibnum);
 577                 /*
 578                  * A host route through the loopback interface is
 579                  * installed for each interface adddress. In pre 8.0
 580                  * releases the interface address of a PPP link type
 581                  * is not reachable locally. This behavior is fixed as
 582                  * part of the new L2/L3 redesign and rewrite work. The
 583                  * signature of this interface address route is the
 584                  * AF_LINK sa_family type of the rt_gateway, and the
 585                  * rt_ifp has the IFF_LOOPBACK flag set.
 586                  */
 587                 if (gw_ro.ro_rt != NULL &&
 588                     gw_ro.ro_rt->rt_gateway->sa_family == AF_LINK &&
 589                     gw_ro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK)
 590                         info.rti_flags &= ~RTF_GATEWAY;
 591                 if (gw_ro.ro_rt != NULL)
 592                         RTFREE(gw_ro.ro_rt);
 593         }
 594
 595         switch (rtm->rtm_type) {
 596                 struct rtentry *saved_nrt;
 597
 598         case RTM_ADD:
 599                 if (info.rti_info[RTAX_GATEWAY] == NULL)
 600                         senderr(EINVAL);
 601                 saved_nrt = NULL;
 602
 603                 /* support for new ARP code */
 604                 if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
 605                     (rtm->rtm_flags & RTF_LLDATA) != 0) {
 606                         error = lla_rt_output(rtm, &info);
 607                         break;
 608                 }
 609                 error = rtrequest1_fib(RTM_ADD, &info, &saved_nrt,
 610                     so->so_fibnum);
 611                 if (error == 0 && saved_nrt) {
 612                         RT_LOCK(saved_nrt);
 613                         rt_setmetrics(rtm->rtm_inits,
 614                                 &rtm->rtm_rmx, &saved_nrt->rt_rmx);
 615                         rtm->rtm_index = saved_nrt->rt_ifp->if_index;
 616                         RT_REMREF(saved_nrt);
 617                         RT_UNLOCK(saved_nrt);
 618                 }
 619                 break;
 620
 621         case RTM_DELETE:
 622                 saved_nrt = NULL;
 623                 /* support for new ARP code */
 624                 if (info.rti_info[RTAX_GATEWAY] &&
 625                     (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
 626                     (rtm->rtm_flags & RTF_LLDATA) != 0) {
 627                         error = lla_rt_output(rtm, &info);
 628                         break;
 629                 }
 630                 error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt,
 631                     so->so_fibnum);
 632                 if (error == 0) {
 633                         RT_LOCK(saved_nrt);
 634                         rt = saved_nrt;
 635                         goto report;
 636                 }
 637                 break;
 638
 639         case RTM_GET:
 640         case RTM_CHANGE:
 641         case RTM_LOCK:
 642                 rnh = rt_tables_get_rnh(so->so_fibnum,
 643                     info.rti_info[RTAX_DST]->sa_family);
 644                 if (rnh == NULL)
 645                         senderr(EAFNOSUPPORT);
 646                 RADIX_NODE_HEAD_RLOCK(rnh);
 647                 rt = (struct rtentry *) rnh->rnh_lookup(info.rti_info[RTAX_DST],
 648                         info.rti_info[RTAX_NETMASK], rnh);
 649                 if (rt == NULL) {       /* XXX looks bogus */
 650                         RADIX_NODE_HEAD_RUNLOCK(rnh);
 651                         senderr(ESRCH);
 652                 }
 653 #ifdef RADIX_MPATH
 654                 /*
 655                  * for RTM_CHANGE/LOCK, if we got multipath routes,
 656                  * we require users to specify a matching RTAX_GATEWAY.
 657                  *
 658                  * for RTM_GET, gate is optional even with multipath.
 659                  * if gate == NULL the first match is returned.
 660                  * (no need to call rt_mpath_matchgate if gate == NULL)
 661                  */
 662                 if (rn_mpath_capable(rnh) &&
 663                     (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
 664                         rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
 665                         if (!rt) {
 666                                 RADIX_NODE_HEAD_RUNLOCK(rnh);
 667                                 senderr(ESRCH);
 668                         }
 669                 }
 670 #endif
 671                 /*
 672                  * If performing proxied L2 entry insertion, and
 673                  * the actual PPP host entry is found, perform
 674                  * another search to retrieve the prefix route of
 675                  * the local end point of the PPP link.
 676                  */
 677                 if (rtm->rtm_flags & RTF_ANNOUNCE) {
 678                         struct sockaddr laddr;
 679
 680                         if (rt->rt_ifp != NULL &&
 681                             rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
 682                                 struct ifaddr *ifa;
 683
 684                                 ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1);
 685                                 if (ifa != NULL)
 686                                         rt_maskedcopy(ifa->ifa_addr,
 687                                                       &laddr,
 688                                                       ifa->ifa_netmask);
 689                         } else
 690                                 rt_maskedcopy(rt->rt_ifa->ifa_addr,
 691                                               &laddr,
 692                                               rt->rt_ifa->ifa_netmask);
 693                         /*
 694                          * refactor rt and no lock operation necessary
 695                          */
 696                         rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr, rnh);
 697                         if (rt == NULL) {
 698                                 RADIX_NODE_HEAD_RUNLOCK(rnh);
 699                                 senderr(ESRCH);
 700                         }
 701                 }
 702                 RT_LOCK(rt);
 703                 RT_ADDREF(rt);
 704                 RADIX_NODE_HEAD_RUNLOCK(rnh);
 705
 706                 /*
 707                  * Fix for PR: 82974
 708                  *
 709                  * RTM_CHANGE/LOCK need a perfect match, rn_lookup()
 710                  * returns a perfect match in case a netmask is
 711                  * specified.  For host routes only a longest prefix
 712                  * match is returned so it is necessary to compare the
 713                  * existence of the netmask.  If both have a netmask
 714                  * rnh_lookup() did a perfect match and if none of them
 715                  * have a netmask both are host routes which is also a
 716                  * perfect match.
 717                  */
 718
 719                 if (rtm->rtm_type != RTM_GET &&
 720                     (!rt_mask(rt) != !info.rti_info[RTAX_NETMASK])) {
 721                         RT_UNLOCK(rt);
 722                         senderr(ESRCH);
 723                 }
 724
 725                 switch(rtm->rtm_type) {
 726
 727                 case RTM_GET:
 728                 report:
 729                         RT_LOCK_ASSERT(rt);
 730                         if ((rt->rt_flags & RTF_HOST) == 0
 731                             ? jailed_without_vnet(curthread->td_ucred)
 732                             : prison_if(curthread->td_ucred,
 733                             rt_key(rt)) != 0) {
 734                                 RT_UNLOCK(rt);
 735                                 senderr(ESRCH);
 736                         }
 737                         info.rti_info[RTAX_DST] = rt_key(rt);
 738                         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
 739                         info.rti_info[RTAX_NETMASK] = rt_mask(rt);
 740                         info.rti_info[RTAX_GENMASK] = 0;
 741                         if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
 742                                 ifp = rt->rt_ifp;
 743                                 if (ifp) {
 744                                         info.rti_info[RTAX_IFP] =
 745                                             ifp->if_addr->ifa_addr;
 746                                         error = rtm_get_jailed(&info, ifp, rt,
 747                                             &saun, curthread->td_ucred);
 748                                         if (error != 0) {
 749                                                 RT_UNLOCK(rt);
 750                                                 senderr(error);
 751                                         }
 752                                         if (ifp->if_flags & IFF_POINTOPOINT)
 753                                                 info.rti_info[RTAX_BRD] =
 754                                                     rt->rt_ifa->ifa_dstaddr;
 755                                         rtm->rtm_index = ifp->if_index;
 756                                 } else {
 757                                         info.rti_info[RTAX_IFP] = NULL;
 758                                         info.rti_info[RTAX_IFA] = NULL;
 759                                 }
 760                         } else if ((ifp = rt->rt_ifp) != NULL) {
 761                                 rtm->rtm_index = ifp->if_index;
 762                         }
 763                         len = rt_msg2(rtm->rtm_type, &info, NULL, NULL);
 764                         if (len > rtm->rtm_msglen) {
 765                                 struct rt_msghdr *new_rtm;
 766                                 R_Malloc(new_rtm, struct rt_msghdr *, len);
 767                                 if (new_rtm == NULL) {
 768                                         RT_UNLOCK(rt);
 769                                         senderr(ENOBUFS);
 770                                 }
 771                                 bcopy(rtm, new_rtm, rtm->rtm_msglen);
 772                                 Free(rtm); rtm = new_rtm;
 773                         }
 774                         (void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm, NULL);
 775                         rtm->rtm_flags = rt->rt_flags;
 776                         rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
 777                         rtm->rtm_addrs = info.rti_addrs;
 778                         break;
 779
 780                 case RTM_CHANGE:
 781                         /*
 782                          * New gateway could require new ifaddr, ifp;
 783                          * flags may also be different; ifp may be specified
 784                          * by ll sockaddr when protocol address is ambiguous
 785                          */
 786                         if (((rt->rt_flags & RTF_GATEWAY) &&
 787                              info.rti_info[RTAX_GATEWAY] != NULL) ||
 788                             info.rti_info[RTAX_IFP] != NULL ||
 789                             (info.rti_info[RTAX_IFA] != NULL &&
 790                              !sa_equal(info.rti_info[RTAX_IFA],
 791                                        rt->rt_ifa->ifa_addr))) {
 792                                 RT_UNLOCK(rt);
 793                                 RADIX_NODE_HEAD_LOCK(rnh);
 794                                 error = rt_getifa_fib(&info, rt->rt_fibnum);
 795                                 /*
 796                                  * XXXRW: Really we should release this
 797                                  * reference later, but this maintains
 798                                  * historical behavior.
 799                                  */
 800                                 if (info.rti_ifa != NULL)
 801                                         ifa_free(info.rti_ifa);
 802                                 RADIX_NODE_HEAD_UNLOCK(rnh);
 803                                 if (error != 0)
 804                                         senderr(error);
 805                                 RT_LOCK(rt);
 806                         }
 807                         if (info.rti_ifa != NULL &&
 808                             info.rti_ifa != rt->rt_ifa &&
 809                             rt->rt_ifa != NULL &&
 810                             rt->rt_ifa->ifa_rtrequest != NULL) {
 811                                 rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
 812                                     &info);
 813                                 ifa_free(rt->rt_ifa);
 814                         }
 815                         if (info.rti_info[RTAX_GATEWAY] != NULL) {
 816                                 RT_UNLOCK(rt);
 817                                 RADIX_NODE_HEAD_LOCK(rnh);
 818                                 RT_LOCK(rt);
 819
 820                                 error = rt_setgate(rt, rt_key(rt),
 821                                     info.rti_info[RTAX_GATEWAY]);
 822                                 RADIX_NODE_HEAD_UNLOCK(rnh);
 823                                 if (error != 0) {
 824                                         RT_UNLOCK(rt);
 825                                         senderr(error);
 826                                 }
 827                                 rt->rt_flags |= (RTF_GATEWAY & info.rti_flags);
 828                         }
 829                         if (info.rti_ifa != NULL &&
 830                             info.rti_ifa != rt->rt_ifa) {
 831                                 ifa_ref(info.rti_ifa);
 832                                 rt->rt_ifa = info.rti_ifa;
 833                                 rt->rt_ifp = info.rti_ifp;
 834                         }
 835                         /* Allow some flags to be toggled on change. */
 836                         rt->rt_flags = (rt->rt_flags & ~RTF_FMASK) |
 837                                     (rtm->rtm_flags & RTF_FMASK);
 838                         rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
 839                                         &rt->rt_rmx);
 840                         rtm->rtm_index = rt->rt_ifp->if_index;
 841                         if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
 842                                rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
 843                         /* FALLTHROUGH */
 844                 case RTM_LOCK:
 845                         /* We don't support locks anymore */
 846                         break;
 847                 }
 848                 RT_UNLOCK(rt);
 849                 break;
 850
 851         default:
 852                 senderr(EOPNOTSUPP);
 853         }
 854
 855 flush:
 856         if (rtm) {
 857                 if (error)
 858                         rtm->rtm_errno = error;
 859                 else
 860                         rtm->rtm_flags |= RTF_DONE;
 861         }
 862         if (rt)         /* XXX can this be true? */
 863                 RTFREE(rt);
 864     {
 865         struct rawcb *rp = NULL;
 866         /*
 867          * Check to see if we don't want our own messages.
 868          */
 869         if ((so->so_options & SO_USELOOPBACK) == 0) {
 870                 if (route_cb.any_count <= 1) {
 871                         if (rtm)
 872                                 Free(rtm);
 873                         m_freem(m);
 874                         return (error);
 875                 }
 876                 /* There is another listener, so construct message */
 877                 rp = sotorawcb(so);
 878         }
 879         if (rtm) {
 880                 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
 881                 if (m->m_pkthdr.len < rtm->rtm_msglen) {
 882                         m_freem(m);
 883                         m = NULL;
 884                 } else if (m->m_pkthdr.len > rtm->rtm_msglen)
 885                         m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
 886         }
 887         if (m) {
 888                 if (rp) {
 889                         /*
 890                          * XXX insure we don't get a copy by
 891                          * invalidating our protocol
 892                          */
 893                         unsigned short family = rp->rcb_proto.sp_family;
 894                         rp->rcb_proto.sp_family = 0;
 895                         rt_dispatch(m, info.rti_info[RTAX_DST]);
 896                         rp->rcb_proto.sp_family = family;
 897                 } else
 898                         rt_dispatch(m, info.rti_info[RTAX_DST]);
 899         }
 900         /* info.rti_info[RTAX_DST] (used above) can point inside of rtm */
 901         if (rtm)
 902                 Free(rtm);
 903     }
 904         return (error);
 905 #undef  sa_equal
 906 }
 907
 908 static void
 909 rt_setmetrics(u_long which, const struct rt_metrics *in,
 910         struct rt_metrics_lite *out)
 911 {
 912 #define metric(f, e) if (which & (f)) out->e = in->e;
 913         /*
 914          * Only these are stored in the routing entry since introduction
 915          * of tcp hostcache. The rest is ignored.
 916          */
 917         metric(RTV_MTU, rmx_mtu);
 918         metric(RTV_WEIGHT, rmx_weight);
 919         /* Userland -> kernel timebase conversion. */
 920         if (which & RTV_EXPIRE)
 921                 out->rmx_expire = in->rmx_expire ?
 922                     in->rmx_expire - time_second + time_uptime : 0;
 923 #undef metric
 924 }
 925
 926 static void
 927 rt_getmetrics(const struct rt_metrics_lite *in, struct rt_metrics *out)
 928 {
 929 #define metric(e) out->e = in->e;
 930         bzero(out, sizeof(*out));
 931         metric(rmx_mtu);
 932         metric(rmx_weight);
 933         /* Kernel -> userland timebase conversion. */
 934         out->rmx_expire = in->rmx_expire ?
 935             in->rmx_expire - time_uptime + time_second : 0;
 936 #undef metric
 937 }
 938
 939 /*
 940  * Extract the addresses of the passed sockaddrs.
 941  * Do a little sanity checking so as to avoid bad memory references.
 942  * This data is derived straight from userland.
 943  */
 944 static int
 945 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
 946 {
 947         struct sockaddr *sa;
 948         int i;
 949
 950         for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
 951                 if ((rtinfo->rti_addrs & (1 << i)) == 0)
 952                         continue;
 953                 sa = (struct sockaddr *)cp;
 954                 /*
 955                  * It won't fit.
 956                  */
 957                 if (cp + sa->sa_len > cplim)
 958                         return (EINVAL);
 959                 /*
 960                  * there are no more.. quit now
 961                  * If there are more bits, they are in error.
 962                  * I've seen this. route(1) can evidently generate these.
 963                  * This causes kernel to core dump.
 964                  * for compatibility, If we see this, point to a safe address.
 965                  */
 966                 if (sa->sa_len == 0) {
 967                         rtinfo->rti_info[i] = &sa_zero;
 968                         return (0); /* should be EINVAL but for compat */
 969                 }
 970                 /* accept it */
 971                 rtinfo->rti_info[i] = sa;
 972                 cp += SA_SIZE(sa);
 973         }
 974         return (0);
 975 }
 976
 977 static struct mbuf *
 978 rt_msg1(int type, struct rt_addrinfo *rtinfo)
 979 {
 980         struct rt_msghdr *rtm;
 981         struct mbuf *m;
 982         int i;
 983         struct sockaddr *sa;
 984         int len, dlen;
 985
 986         switch (type) {
 987
 988         case RTM_DELADDR:
 989         case RTM_NEWADDR:
 990                 len = sizeof(struct ifa_msghdr);
 991                 break;
 992
 993         case RTM_DELMADDR:
 994         case RTM_NEWMADDR:
 995                 len = sizeof(struct ifma_msghdr);
 996                 break;
 997
 998         case RTM_IFINFO:
 999                 len = sizeof(struct if_msghdr);
1000                 break;
1001
1002         case RTM_IFANNOUNCE:
1003         case RTM_IEEE80211:
1004                 len = sizeof(struct if_announcemsghdr);
1005                 break;
1006
1007         default:
1008                 len = sizeof(struct rt_msghdr);
1009         }
1010         if (len > MCLBYTES)
1011                 panic("rt_msg1");
1012         m = m_gethdr(M_DONTWAIT, MT_DATA);
1013         if (m && len > MHLEN) {
1014                 MCLGET(m, M_DONTWAIT);
1015                 if ((m->m_flags & M_EXT) == 0) {
1016                         m_free(m);
1017                         m = NULL;
1018                 }
1019         }
1020         if (m == NULL)
1021                 return (m);
1022         m->m_pkthdr.len = m->m_len = len;
1023         m->m_pkthdr.rcvif = NULL;
1024         rtm = mtod(m, struct rt_msghdr *);
1025         bzero((caddr_t)rtm, len);
1026         for (i = 0; i < RTAX_MAX; i++) {
1027                 if ((sa = rtinfo->rti_info[i]) == NULL)
1028                         continue;
1029                 rtinfo->rti_addrs |= (1 << i);
1030                 dlen = SA_SIZE(sa);
1031                 m_copyback(m, len, dlen, (caddr_t)sa);
1032                 len += dlen;
1033         }
1034         if (m->m_pkthdr.len != len) {
1035                 m_freem(m);
1036                 return (NULL);
1037         }
1038         rtm->rtm_msglen = len;
1039         rtm->rtm_version = RTM_VERSION;
1040         rtm->rtm_type = type;
1041         return (m);
1042 }
1043
1044 static int
1045 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
1046 {
1047         int i;
1048         int len, dlen, second_time = 0;
1049         caddr_t cp0;
1050
1051         rtinfo->rti_addrs = 0;
1052 again:
1053         switch (type) {
1054
1055         case RTM_DELADDR:
1056         case RTM_NEWADDR:
1057                 len = sizeof(struct ifa_msghdr);
1058                 break;
1059
1060         case RTM_IFINFO:
1061 #ifdef COMPAT_FREEBSD32
1062                 if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1063                         len = sizeof(struct if_msghdr32);
1064                         break;
1065                 }
1066 #endif
1067                 len = sizeof(struct if_msghdr);
1068                 break;
1069
1070         case RTM_NEWMADDR:
1071                 len = sizeof(struct ifma_msghdr);
1072                 break;
1073
1074         default:
1075                 len = sizeof(struct rt_msghdr);
1076         }
1077         cp0 = cp;
1078         if (cp0)
1079                 cp += len;
1080         for (i = 0; i < RTAX_MAX; i++) {
1081                 struct sockaddr *sa;
1082
1083                 if ((sa = rtinfo->rti_info[i]) == NULL)
1084                         continue;
1085                 rtinfo->rti_addrs |= (1 << i);
1086                 dlen = SA_SIZE(sa);
1087                 if (cp) {
1088                         bcopy((caddr_t)sa, cp, (unsigned)dlen);
1089                         cp += dlen;
1090                 }
1091                 len += dlen;
1092         }
1093         len = ALIGN(len);
1094         if (cp == NULL && w != NULL && !second_time) {
1095                 struct walkarg *rw = w;
1096
1097                 if (rw->w_req) {
1098                         if (rw->w_tmemsize < len) {
1099                                 if (rw->w_tmem)
1100                                         free(rw->w_tmem, M_RTABLE);
1101                                 rw->w_tmem = (caddr_t)
1102                                         malloc(len, M_RTABLE, M_NOWAIT);
1103                                 if (rw->w_tmem)
1104                                         rw->w_tmemsize = len;
1105                         }
1106                         if (rw->w_tmem) {
1107                                 cp = rw->w_tmem;
1108                                 second_time = 1;
1109                                 goto again;
1110                         }
1111                 }
1112         }
1113         if (cp) {
1114                 struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
1115
1116                 rtm->rtm_version = RTM_VERSION;
1117                 rtm->rtm_type = type;
1118                 rtm->rtm_msglen = len;
1119         }
1120         return (len);
1121 }
1122
1123 /*
1124  * This routine is called to generate a message from the routing
1125  * socket indicating that a redirect has occured, a routing lookup
1126  * has failed, or that a protocol has detected timeouts to a particular
1127  * destination.
1128  */
1129 void
1130 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1131 {
1132         struct rt_msghdr *rtm;
1133         struct mbuf *m;
1134         struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1135
1136         if (route_cb.any_count == 0)
1137                 return;
1138         m = rt_msg1(type, rtinfo);
1139         if (m == NULL)
1140                 return;
1141         rtm = mtod(m, struct rt_msghdr *);
1142         rtm->rtm_flags = RTF_DONE | flags;
1143         rtm->rtm_errno = error;
1144         rtm->rtm_addrs = rtinfo->rti_addrs;
1145         rt_dispatch(m, sa);
1146 }
1147
1148 /*
1149  * This routine is called to generate a message from the routing
1150  * socket indicating that the status of a network interface has changed.
1151  */
1152 void
1153 rt_ifmsg(struct ifnet *ifp)
1154 {
1155         struct if_msghdr *ifm;
1156         struct mbuf *m;
1157         struct rt_addrinfo info;
1158
1159         if (route_cb.any_count == 0)
1160                 return;
1161         bzero((caddr_t)&info, sizeof(info));
1162         m = rt_msg1(RTM_IFINFO, &info);
1163         if (m == NULL)
1164                 return;
1165         ifm = mtod(m, struct if_msghdr *);
1166         ifm->ifm_index = ifp->if_index;
1167         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1168         ifm->ifm_data = ifp->if_data;
1169         ifm->ifm_addrs = 0;
1170         rt_dispatch(m, NULL);
1171 }
1172
1173 /*
1174  * This is called to generate messages from the routing socket
1175  * indicating a network interface has had addresses associated with it.
1176  * if we ever reverse the logic and replace messages TO the routing
1177  * socket indicate a request to configure interfaces, then it will
1178  * be unnecessary as the routing socket will automatically generate
1179  * copies of it.
1180  */
1181 void
1182 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1183 {
1184         struct rt_addrinfo info;
1185         struct sockaddr *sa = NULL;
1186         int pass;
1187         struct mbuf *m = NULL;
1188         struct ifnet *ifp = ifa->ifa_ifp;
1189
1190         KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
1191                 ("unexpected cmd %u", cmd));
1192 #if defined(INET) || defined(INET6)
1193 #ifdef SCTP
1194         /*
1195          * notify the SCTP stack
1196          * this will only get called when an address is added/deleted
1197          * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1198          */
1199         sctp_addr_change(ifa, cmd);
1200 #endif /* SCTP */
1201 #endif
1202         if (route_cb.any_count == 0)
1203                 return;
1204         for (pass = 1; pass < 3; pass++) {
1205                 bzero((caddr_t)&info, sizeof(info));
1206                 if ((cmd == RTM_ADD && pass == 1) ||
1207                     (cmd == RTM_DELETE && pass == 2)) {
1208                         struct ifa_msghdr *ifam;
1209                         int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1210
1211                         info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1212                         info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1213                         info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1214                         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1215                         if ((m = rt_msg1(ncmd, &info)) == NULL)
1216                                 continue;
1217                         ifam = mtod(m, struct ifa_msghdr *);
1218                         ifam->ifam_index = ifp->if_index;
1219                         ifam->ifam_metric = ifa->ifa_metric;
1220                         ifam->ifam_flags = ifa->ifa_flags;
1221                         ifam->ifam_addrs = info.rti_addrs;
1222                 }
1223                 if ((cmd == RTM_ADD && pass == 2) ||
1224                     (cmd == RTM_DELETE && pass == 1)) {
1225                         struct rt_msghdr *rtm;
1226
1227                         if (rt == NULL)
1228                                 continue;
1229                         info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1230                         info.rti_info[RTAX_DST] = sa = rt_key(rt);
1231                         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1232                         if ((m = rt_msg1(cmd, &info)) == NULL)
1233                                 continue;
1234                         rtm = mtod(m, struct rt_msghdr *);
1235                         rtm->rtm_index = ifp->if_index;
1236                         rtm->rtm_flags |= rt->rt_flags;
1237                         rtm->rtm_errno = error;
1238                         rtm->rtm_addrs = info.rti_addrs;
1239                 }
1240                 rt_dispatch(m, sa);
1241         }
1242 }
1243
1244 /*
1245  * This is the analogue to the rt_newaddrmsg which performs the same
1246  * function but for multicast group memberhips.  This is easier since
1247  * there is no route state to worry about.
1248  */
1249 void
1250 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1251 {
1252         struct rt_addrinfo info;
1253         struct mbuf *m = NULL;
1254         struct ifnet *ifp = ifma->ifma_ifp;
1255         struct ifma_msghdr *ifmam;
1256
1257         if (route_cb.any_count == 0)
1258                 return;
1259
1260         bzero((caddr_t)&info, sizeof(info));
1261         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1262         info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1263         /*
1264          * If a link-layer address is present, present it as a ``gateway''
1265          * (similarly to how ARP entries, e.g., are presented).
1266          */
1267         info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1268         m = rt_msg1(cmd, &info);
1269         if (m == NULL)
1270                 return;
1271         ifmam = mtod(m, struct ifma_msghdr *);
1272         KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1273             __func__));
1274         ifmam->ifmam_index = ifp->if_index;
1275         ifmam->ifmam_addrs = info.rti_addrs;
1276         rt_dispatch(m, ifma->ifma_addr);
1277 }
1278
1279 static struct mbuf *
1280 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1281         struct rt_addrinfo *info)
1282 {
1283         struct if_announcemsghdr *ifan;
1284         struct mbuf *m;
1285
1286         if (route_cb.any_count == 0)
1287                 return NULL;
1288         bzero((caddr_t)info, sizeof(*info));
1289         m = rt_msg1(type, info);
1290         if (m != NULL) {
1291                 ifan = mtod(m, struct if_announcemsghdr *);
1292                 ifan->ifan_index = ifp->if_index;
1293                 strlcpy(ifan->ifan_name, ifp->if_xname,
1294                         sizeof(ifan->ifan_name));
1295                 ifan->ifan_what = what;
1296         }
1297         return m;
1298 }
1299
1300 /*
1301  * This is called to generate routing socket messages indicating
1302  * IEEE80211 wireless events.
1303  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1304  */
1305 void
1306 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1307 {
1308         struct mbuf *m;
1309         struct rt_addrinfo info;
1310
1311         m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1312         if (m != NULL) {
1313                 /*
1314                  * Append the ieee80211 data.  Try to stick it in the
1315                  * mbuf containing the ifannounce msg; otherwise allocate
1316                  * a new mbuf and append.
1317                  *
1318                  * NB: we assume m is a single mbuf.
1319                  */
1320                 if (data_len > M_TRAILINGSPACE(m)) {
1321                         struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1322                         if (n == NULL) {
1323                                 m_freem(m);
1324                                 return;
1325                         }
1326                         bcopy(data, mtod(n, void *), data_len);
1327                         n->m_len = data_len;
1328                         m->m_next = n;
1329                 } else if (data_len > 0) {
1330                         bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1331                         m->m_len += data_len;
1332                 }
1333                 if (m->m_flags & M_PKTHDR)
1334                         m->m_pkthdr.len += data_len;
1335                 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1336                 rt_dispatch(m, NULL);
1337         }
1338 }
1339
1340 /*
1341  * This is called to generate routing socket messages indicating
1342  * network interface arrival and departure.
1343  */
1344 void
1345 rt_ifannouncemsg(struct ifnet *ifp, int what)
1346 {
1347         struct mbuf *m;
1348         struct rt_addrinfo info;
1349
1350         m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1351         if (m != NULL)
1352                 rt_dispatch(m, NULL);
1353 }
1354
1355 static void
1356 rt_dispatch(struct mbuf *m, const struct sockaddr *sa)
1357 {
1358         struct m_tag *tag;
1359
1360         /*
1361          * Preserve the family from the sockaddr, if any, in an m_tag for
1362          * use when injecting the mbuf into the routing socket buffer from
1363          * the netisr.
1364          */
1365         if (sa != NULL) {
1366                 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1367                     M_NOWAIT);
1368                 if (tag == NULL) {
1369                         m_freem(m);
1370                         return;
1371                 }
1372                 *(unsigned short *)(tag + 1) = sa->sa_family;
1373                 m_tag_prepend(m, tag);
1374         }
1375 #ifdef VIMAGE
1376         if (V_loif)
1377                 m->m_pkthdr.rcvif = V_loif;
1378         else {
1379                 m_freem(m);
1380                 return;
1381         }
1382 #endif
1383         netisr_queue(NETISR_ROUTE, m);  /* mbuf is free'd on failure. */
1384 }
1385
1386 /*
1387  * This is used in dumping the kernel table via sysctl().
1388  */
1389 static int
1390 sysctl_dumpentry(struct radix_node *rn, void *vw)
1391 {
1392         struct walkarg *w = vw;
1393         struct rtentry *rt = (struct rtentry *)rn;
1394         int error = 0, size;
1395         struct rt_addrinfo info;
1396
1397         if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1398                 return 0;
1399         if ((rt->rt_flags & RTF_HOST) == 0
1400             ? jailed_without_vnet(w->w_req->td->td_ucred)
1401             : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1402                 return (0);
1403         bzero((caddr_t)&info, sizeof(info));
1404         info.rti_info[RTAX_DST] = rt_key(rt);
1405         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1406         info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1407         info.rti_info[RTAX_GENMASK] = 0;
1408         if (rt->rt_ifp) {
1409                 info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1410                 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1411                 if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1412                         info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1413         }
1414         size = rt_msg2(RTM_GET, &info, NULL, w);
1415         if (w->w_req && w->w_tmem) {
1416                 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1417
1418                 rtm->rtm_flags = rt->rt_flags;
1419                 /*
1420                  * let's be honest about this being a retarded hack
1421                  */
1422                 rtm->rtm_fmask = rt->rt_rmx.rmx_pksent;
1423                 rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1424                 rtm->rtm_index = rt->rt_ifp->if_index;
1425                 rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1426                 rtm->rtm_addrs = info.rti_addrs;
1427                 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1428                 return (error);
1429         }
1430         return (error);
1431 }
1432
1433 #ifdef COMPAT_FREEBSD32
1434 static void
1435 copy_ifdata32(struct if_data *src, struct if_data32 *dst)
1436 {
1437
1438         bzero(dst, sizeof(*dst));
1439         CP(*src, *dst, ifi_type);
1440         CP(*src, *dst, ifi_physical);
1441         CP(*src, *dst, ifi_addrlen);
1442         CP(*src, *dst, ifi_hdrlen);
1443         CP(*src, *dst, ifi_link_state);
1444         dst->ifi_datalen = sizeof(struct if_data32);
1445         CP(*src, *dst, ifi_mtu);
1446         CP(*src, *dst, ifi_metric);
1447         CP(*src, *dst, ifi_baudrate);
1448         CP(*src, *dst, ifi_ipackets);
1449         CP(*src, *dst, ifi_ierrors);
1450         CP(*src, *dst, ifi_opackets);
1451         CP(*src, *dst, ifi_oerrors);
1452         CP(*src, *dst, ifi_collisions);
1453         CP(*src, *dst, ifi_ibytes);
1454         CP(*src, *dst, ifi_obytes);
1455         CP(*src, *dst, ifi_imcasts);
1456         CP(*src, *dst, ifi_omcasts);
1457         CP(*src, *dst, ifi_iqdrops);
1458         CP(*src, *dst, ifi_noproto);
1459         CP(*src, *dst, ifi_hwassist);
1460         CP(*src, *dst, ifi_epoch);
1461         TV_CP(*src, *dst, ifi_lastchange);
1462 }
1463 #endif
1464
1465 static int
1466 sysctl_iflist(int af, struct walkarg *w)
1467 {
1468         struct ifnet *ifp;
1469         struct ifaddr *ifa;
1470         struct rt_addrinfo info;
1471         int len, error = 0;
1472
1473         bzero((caddr_t)&info, sizeof(info));
1474         IFNET_RLOCK();
1475         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1476                 if (w->w_arg && w->w_arg != ifp->if_index)
1477                         continue;
1478                 IF_ADDR_LOCK(ifp);
1479                 ifa = ifp->if_addr;
1480                 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1481                 len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1482                 info.rti_info[RTAX_IFP] = NULL;
1483                 if (w->w_req && w->w_tmem) {
1484                         struct if_msghdr *ifm;
1485
1486 #ifdef COMPAT_FREEBSD32
1487                         if (w->w_req->flags & SCTL_MASK32) {
1488                                 struct if_msghdr32 *ifm32;
1489
1490                                 ifm32 = (struct if_msghdr32 *)w->w_tmem;
1491                                 ifm32->ifm_index = ifp->if_index;
1492                                 ifm32->ifm_flags = ifp->if_flags |
1493                                     ifp->if_drv_flags;
1494                                 copy_ifdata32(&ifp->if_data, &ifm32->ifm_data);
1495                                 ifm32->ifm_addrs = info.rti_addrs;
1496                                 error = SYSCTL_OUT(w->w_req, (caddr_t)ifm32,
1497                                     len);
1498                                 goto sysctl_out;
1499                         }
1500 #endif
1501                         ifm = (struct if_msghdr *)w->w_tmem;
1502                         ifm->ifm_index = ifp->if_index;
1503                         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1504                         ifm->ifm_data = ifp->if_data;
1505                         ifm->ifm_addrs = info.rti_addrs;
1506                         error = SYSCTL_OUT(w->w_req, (caddr_t)ifm, len);
1507 #ifdef COMPAT_FREEBSD32
1508                 sysctl_out:
1509 #endif
1510                         if (error)
1511                                 goto done;
1512                 }
1513                 while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1514                         if (af && af != ifa->ifa_addr->sa_family)
1515                                 continue;
1516                         if (prison_if(w->w_req->td->td_ucred,
1517                             ifa->ifa_addr) != 0)
1518                                 continue;
1519                         info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1520                         info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1521                         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1522                         len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1523                         if (w->w_req && w->w_tmem) {
1524                                 struct ifa_msghdr *ifam;
1525
1526                                 ifam = (struct ifa_msghdr *)w->w_tmem;
1527                                 ifam->ifam_index = ifa->ifa_ifp->if_index;
1528                                 ifam->ifam_flags = ifa->ifa_flags;
1529                                 ifam->ifam_metric = ifa->ifa_metric;
1530                                 ifam->ifam_addrs = info.rti_addrs;
1531                                 error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1532                                 if (error)
1533                                         goto done;
1534                         }
1535                 }
1536                 IF_ADDR_UNLOCK(ifp);
1537                 info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1538                         info.rti_info[RTAX_BRD] = NULL;
1539         }
1540 done:
1541         if (ifp != NULL)
1542                 IF_ADDR_UNLOCK(ifp);
1543         IFNET_RUNLOCK();
1544         return (error);
1545 }
1546
1547 static int
1548 sysctl_ifmalist(int af, struct walkarg *w)
1549 {
1550         struct ifnet *ifp;
1551         struct ifmultiaddr *ifma;
1552         struct  rt_addrinfo info;
1553         int     len, error = 0;
1554         struct ifaddr *ifa;
1555
1556         bzero((caddr_t)&info, sizeof(info));
1557         IFNET_RLOCK();
1558         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1559                 if (w->w_arg && w->w_arg != ifp->if_index)
1560                         continue;
1561                 ifa = ifp->if_addr;
1562                 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1563                 IF_ADDR_LOCK(ifp);
1564                 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1565                         if (af && af != ifma->ifma_addr->sa_family)
1566                                 continue;
1567                         if (prison_if(w->w_req->td->td_ucred,
1568                             ifma->ifma_addr) != 0)
1569                                 continue;
1570                         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1571                         info.rti_info[RTAX_GATEWAY] =
1572                             (ifma->ifma_addr->sa_family != AF_LINK) ?
1573                             ifma->ifma_lladdr : NULL;
1574                         len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1575                         if (w->w_req && w->w_tmem) {
1576                                 struct ifma_msghdr *ifmam;
1577
1578                                 ifmam = (struct ifma_msghdr *)w->w_tmem;
1579                                 ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1580                                 ifmam->ifmam_flags = 0;
1581                                 ifmam->ifmam_addrs = info.rti_addrs;
1582                                 error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1583                                 if (error) {
1584                                         IF_ADDR_UNLOCK(ifp);
1585                                         goto done;
1586                                 }
1587                         }
1588                 }
1589                 IF_ADDR_UNLOCK(ifp);
1590         }
1591 done:
1592         IFNET_RUNLOCK();
1593         return (error);
1594 }
1595
1596 static int
1597 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1598 {
1599         int     *name = (int *)arg1;
1600         u_int   namelen = arg2;
1601         struct radix_node_head *rnh = NULL; /* silence compiler. */
1602         int     i, lim, error = EINVAL;
1603         u_char  af;
1604         struct  walkarg w;
1605
1606         name ++;
1607         namelen--;
1608         if (req->newptr)
1609                 return (EPERM);
1610         if (namelen != 3)
1611                 return ((namelen < 3) ? EISDIR : ENOTDIR);
1612         af = name[0];
1613         if (af > AF_MAX)
1614                 return (EINVAL);
1615         bzero(&w, sizeof(w));
1616         w.w_op = name[1];
1617         w.w_arg = name[2];
1618         w.w_req = req;
1619
1620         error = sysctl_wire_old_buffer(req, 0);
1621         if (error)
1622                 return (error);
1623         switch (w.w_op) {
1624
1625         case NET_RT_DUMP:
1626         case NET_RT_FLAGS:
1627                 if (af == 0) {                  /* dump all tables */
1628                         i = 1;
1629                         lim = AF_MAX;
1630                 } else                          /* dump only one table */
1631                         i = lim = af;
1632
1633                 /*
1634                  * take care of llinfo entries, the caller must
1635                  * specify an AF
1636                  */
1637                 if (w.w_op == NET_RT_FLAGS &&
1638                     (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1639                         if (af != 0)
1640                                 error = lltable_sysctl_dumparp(af, w.w_req);
1641                         else
1642                                 error = EINVAL;
1643                         break;
1644                 }
1645                 /*
1646                  * take care of routing entries
1647                  */
1648                 for (error = 0; error == 0 && i <= lim; i++) {
1649                         rnh = rt_tables_get_rnh(req->td->td_proc->p_fibnum, i);
1650                         if (rnh != NULL) {
1651                                 RADIX_NODE_HEAD_LOCK(rnh);
1652                                 error = rnh->rnh_walktree(rnh,
1653                                     sysctl_dumpentry, &w);
1654                                 RADIX_NODE_HEAD_UNLOCK(rnh);
1655                         } else if (af != 0)
1656                                 error = EAFNOSUPPORT;
1657                 }
1658                 break;
1659
1660         case NET_RT_IFLIST:
1661                 error = sysctl_iflist(af, &w);
1662                 break;
1663
1664         case NET_RT_IFMALIST:
1665                 error = sysctl_ifmalist(af, &w);
1666                 break;
1667         }
1668         if (w.w_tmem)
1669                 free(w.w_tmem, M_RTABLE);
1670         return (error);
1671 }
1672
1673 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1674
1675 /*
1676  * Definitions of protocols supported in the ROUTE domain.
1677  */
1678
1679 static struct domain routedomain;               /* or at least forward */
1680
1681 static struct protosw routesw[] = {
1682 {
1683         .pr_type =              SOCK_RAW,
1684         .pr_domain =            &routedomain,
1685         .pr_flags =             PR_ATOMIC|PR_ADDR,
1686         .pr_output =            route_output,
1687         .pr_ctlinput =          raw_ctlinput,
1688         .pr_init =              raw_init,
1689         .pr_usrreqs =           &route_usrreqs
1690 }
1691 };
1692
1693 static struct domain routedomain = {
1694         .dom_family =           PF_ROUTE,
1695         .dom_name =              "route",
1696         .dom_protosw =          routesw,
1697         .dom_protoswNPROTOSW =  &routesw[sizeof(routesw)/sizeof(routesw[0])]
1698 };
1699
1700 VNET_DOMAIN_SET(route);