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
 118 struct if_msghdrl32 {
 119         uint16_t ifm_msglen;
 120         uint8_t ifm_version;
 121         uint8_t ifm_type;
 122         int32_t ifm_addrs;
 123         int32_t ifm_flags;
 124         uint16_t ifm_index;
 125         uint16_t _ifm_spare1;
 126         uint16_t ifm_len;
 127         uint16_t ifm_data_off;
 128         struct  if_data32 ifm_data;
 129 };
 130
 131 struct ifa_msghdrl32 {
 132         uint16_t ifam_msglen;
 133         uint8_t ifam_version;
 134         uint8_t ifam_type;
 135         int32_t ifam_addrs;
 136         int32_t ifam_flags;
 137         uint16_t ifam_index;
 138         uint16_t _ifam_spare1;
 139         uint16_t ifam_len;
 140         uint16_t ifam_data_off;
 141         int32_t ifam_metric;
 142         struct  if_data32 ifam_data;
 143 };
 144 #endif /* COMPAT_FREEBSD32 */
 145
 146 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
 147
 148 /* NB: these are not modified */
 149 static struct   sockaddr route_src = { 2, PF_ROUTE, };
 150 static struct   sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
 151
 152 /*
 153  * Used by rtsock/raw_input callback code to decide whether to filter the update
 154  * notification to a socket bound to a particular FIB.
 155  */
 156 #define RTS_FILTER_FIB  M_PROTO8
 157 #define RTS_ALLFIBS     -1
 158
 159 static struct {
 160         int     ip_count;       /* attached w/ AF_INET */
 161         int     ip6_count;      /* attached w/ AF_INET6 */
 162         int     ipx_count;      /* attached w/ AF_IPX */
 163         int     any_count;      /* total attached */
 164 } route_cb;
 165
 166 struct mtx rtsock_mtx;
 167 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
 168
 169 #define RTSOCK_LOCK()   mtx_lock(&rtsock_mtx)
 170 #define RTSOCK_UNLOCK() mtx_unlock(&rtsock_mtx)
 171 #define RTSOCK_LOCK_ASSERT()    mtx_assert(&rtsock_mtx, MA_OWNED)
 172
 173 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
 174
 175 struct walkarg {
 176         int     w_tmemsize;
 177         int     w_op, w_arg;
 178         caddr_t w_tmem;
 179         struct sysctl_req *w_req;
 180 };
 181
 182 static void     rts_input(struct mbuf *m);
 183 static struct mbuf *rt_msg1(int type, struct rt_addrinfo *rtinfo);
 184 static int      rt_msg2(int type, struct rt_addrinfo *rtinfo,
 185                         caddr_t cp, struct walkarg *w);
 186 static int      rt_xaddrs(caddr_t cp, caddr_t cplim,
 187                         struct rt_addrinfo *rtinfo);
 188 static int      sysctl_dumpentry(struct radix_node *rn, void *vw);
 189 static int      sysctl_iflist(int af, struct walkarg *w);
 190 static int      sysctl_ifmalist(int af, struct walkarg *w);
 191 static int      route_output(struct mbuf *m, struct socket *so);
 192 static void     rt_setmetrics(u_long which, const struct rt_metrics *in,
 193                         struct rt_metrics_lite *out);
 194 static void     rt_getmetrics(const struct rt_metrics_lite *in,
 195                         struct rt_metrics *out);
 196 static void     rt_dispatch(struct mbuf *, sa_family_t);
 197
 198 static struct netisr_handler rtsock_nh = {
 199         .nh_name = "rtsock",
 200         .nh_handler = rts_input,
 201         .nh_proto = NETISR_ROUTE,
 202         .nh_policy = NETISR_POLICY_SOURCE,
 203 };
 204
 205 static int
 206 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
 207 {
 208         int error, qlimit;
 209
 210         netisr_getqlimit(&rtsock_nh, &qlimit);
 211         error = sysctl_handle_int(oidp, &qlimit, 0, req);
 212         if (error || !req->newptr)
 213                 return (error);
 214         if (qlimit < 1)
 215                 return (EINVAL);
 216         return (netisr_setqlimit(&rtsock_nh, qlimit));
 217 }
 218 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
 219     0, 0, sysctl_route_netisr_maxqlen, "I",
 220     "maximum routing socket dispatch queue length");
 221
 222 static void
 223 rts_init(void)
 224 {
 225         int tmp;
 226
 227         if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
 228                 rtsock_nh.nh_qlimit = tmp;
 229         netisr_register(&rtsock_nh);
 230 }
 231 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0);
 232
 233 static int
 234 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
 235     struct rawcb *rp)
 236 {
 237         int fibnum;
 238
 239         KASSERT(m != NULL, ("%s: m is NULL", __func__));
 240         KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
 241         KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
 242
 243         /* No filtering requested. */
 244         if ((m->m_flags & RTS_FILTER_FIB) == 0)
 245                 return (0);
 246
 247         /* Check if it is a rts and the fib matches the one of the socket. */
 248         fibnum = M_GETFIB(m);
 249         if (proto->sp_family != PF_ROUTE ||
 250             rp->rcb_socket == NULL ||
 251             rp->rcb_socket->so_fibnum == fibnum)
 252                 return (0);
 253
 254         /* Filtering requested and no match, the socket shall be skipped. */
 255         return (1);
 256 }
 257
 258 static void
 259 rts_input(struct mbuf *m)
 260 {
 261         struct sockproto route_proto;
 262         unsigned short *family;
 263         struct m_tag *tag;
 264
 265         route_proto.sp_family = PF_ROUTE;
 266         tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
 267         if (tag != NULL) {
 268                 family = (unsigned short *)(tag + 1);
 269                 route_proto.sp_protocol = *family;
 270                 m_tag_delete(m, tag);
 271         } else
 272                 route_proto.sp_protocol = 0;
 273
 274         raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
 275 }
 276
 277 /*
 278  * It really doesn't make any sense at all for this code to share much
 279  * with raw_usrreq.c, since its functionality is so restricted.  XXX
 280  */
 281 static void
 282 rts_abort(struct socket *so)
 283 {
 284
 285         raw_usrreqs.pru_abort(so);
 286 }
 287
 288 static void
 289 rts_close(struct socket *so)
 290 {
 291
 292         raw_usrreqs.pru_close(so);
 293 }
 294
 295 /* pru_accept is EOPNOTSUPP */
 296
 297 static int
 298 rts_attach(struct socket *so, int proto, struct thread *td)
 299 {
 300         struct rawcb *rp;
 301         int s, error;
 302
 303         KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
 304
 305         /* XXX */
 306         rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
 307         if (rp == NULL)
 308                 return ENOBUFS;
 309
 310         /*
 311          * The splnet() is necessary to block protocols from sending
 312          * error notifications (like RTM_REDIRECT or RTM_LOSING) while
 313          * this PCB is extant but incompletely initialized.
 314          * Probably we should try to do more of this work beforehand and
 315          * eliminate the spl.
 316          */
 317         s = splnet();
 318         so->so_pcb = (caddr_t)rp;
 319         so->so_fibnum = td->td_proc->p_fibnum;
 320         error = raw_attach(so, proto);
 321         rp = sotorawcb(so);
 322         if (error) {
 323                 splx(s);
 324                 so->so_pcb = NULL;
 325                 free(rp, M_PCB);
 326                 return error;
 327         }
 328         RTSOCK_LOCK();
 329         switch(rp->rcb_proto.sp_protocol) {
 330         case AF_INET:
 331                 route_cb.ip_count++;
 332                 break;
 333         case AF_INET6:
 334                 route_cb.ip6_count++;
 335                 break;
 336         case AF_IPX:
 337                 route_cb.ipx_count++;
 338                 break;
 339         }
 340         route_cb.any_count++;
 341         RTSOCK_UNLOCK();
 342         soisconnected(so);
 343         so->so_options |= SO_USELOOPBACK;
 344         splx(s);
 345         return 0;
 346 }
 347
 348 static int
 349 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
 350 {
 351
 352         return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
 353 }
 354
 355 static int
 356 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
 357 {
 358
 359         return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
 360 }
 361
 362 /* pru_connect2 is EOPNOTSUPP */
 363 /* pru_control is EOPNOTSUPP */
 364
 365 static void
 366 rts_detach(struct socket *so)
 367 {
 368         struct rawcb *rp = sotorawcb(so);
 369
 370         KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
 371
 372         RTSOCK_LOCK();
 373         switch(rp->rcb_proto.sp_protocol) {
 374         case AF_INET:
 375                 route_cb.ip_count--;
 376                 break;
 377         case AF_INET6:
 378                 route_cb.ip6_count--;
 379                 break;
 380         case AF_IPX:
 381                 route_cb.ipx_count--;
 382                 break;
 383         }
 384         route_cb.any_count--;
 385         RTSOCK_UNLOCK();
 386         raw_usrreqs.pru_detach(so);
 387 }
 388
 389 static int
 390 rts_disconnect(struct socket *so)
 391 {
 392
 393         return (raw_usrreqs.pru_disconnect(so));
 394 }
 395
 396 /* pru_listen is EOPNOTSUPP */
 397
 398 static int
 399 rts_peeraddr(struct socket *so, struct sockaddr **nam)
 400 {
 401
 402         return (raw_usrreqs.pru_peeraddr(so, nam));
 403 }
 404
 405 /* pru_rcvd is EOPNOTSUPP */
 406 /* pru_rcvoob is EOPNOTSUPP */
 407
 408 static int
 409 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
 410          struct mbuf *control, struct thread *td)
 411 {
 412
 413         return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
 414 }
 415
 416 /* pru_sense is null */
 417
 418 static int
 419 rts_shutdown(struct socket *so)
 420 {
 421
 422         return (raw_usrreqs.pru_shutdown(so));
 423 }
 424
 425 static int
 426 rts_sockaddr(struct socket *so, struct sockaddr **nam)
 427 {
 428
 429         return (raw_usrreqs.pru_sockaddr(so, nam));
 430 }
 431
 432 static struct pr_usrreqs route_usrreqs = {
 433         .pru_abort =            rts_abort,
 434         .pru_attach =           rts_attach,
 435         .pru_bind =             rts_bind,
 436         .pru_connect =          rts_connect,
 437         .pru_detach =           rts_detach,
 438         .pru_disconnect =       rts_disconnect,
 439         .pru_peeraddr =         rts_peeraddr,
 440         .pru_send =             rts_send,
 441         .pru_shutdown =         rts_shutdown,
 442         .pru_sockaddr =         rts_sockaddr,
 443         .pru_close =            rts_close,
 444 };
 445
 446 #ifndef _SOCKADDR_UNION_DEFINED
 447 #define _SOCKADDR_UNION_DEFINED
 448 /*
 449  * The union of all possible address formats we handle.
 450  */
 451 union sockaddr_union {
 452         struct sockaddr         sa;
 453         struct sockaddr_in      sin;
 454         struct sockaddr_in6     sin6;
 455 };
 456 #endif /* _SOCKADDR_UNION_DEFINED */
 457
 458 static int
 459 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
 460     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
 461 {
 462
 463         /* First, see if the returned address is part of the jail. */
 464         if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
 465                 info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
 466                 return (0);
 467         }
 468
 469         switch (info->rti_info[RTAX_DST]->sa_family) {
 470 #ifdef INET
 471         case AF_INET:
 472         {
 473                 struct in_addr ia;
 474                 struct ifaddr *ifa;
 475                 int found;
 476
 477                 found = 0;
 478                 /*
 479                  * Try to find an address on the given outgoing interface
 480                  * that belongs to the jail.
 481                  */
 482                 IF_ADDR_RLOCK(ifp);
 483                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 484                         struct sockaddr *sa;
 485                         sa = ifa->ifa_addr;
 486                         if (sa->sa_family != AF_INET)
 487                                 continue;
 488                         ia = ((struct sockaddr_in *)sa)->sin_addr;
 489                         if (prison_check_ip4(cred, &ia) == 0) {
 490                                 found = 1;
 491                                 break;
 492                         }
 493                 }
 494                 IF_ADDR_RUNLOCK(ifp);
 495                 if (!found) {
 496                         /*
 497                          * As a last resort return the 'default' jail address.
 498                          */
 499                         ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
 500                             sin_addr;
 501                         if (prison_get_ip4(cred, &ia) != 0)
 502                                 return (ESRCH);
 503                 }
 504                 bzero(&saun->sin, sizeof(struct sockaddr_in));
 505                 saun->sin.sin_len = sizeof(struct sockaddr_in);
 506                 saun->sin.sin_family = AF_INET;
 507                 saun->sin.sin_addr.s_addr = ia.s_addr;
 508                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
 509                 break;
 510         }
 511 #endif
 512 #ifdef INET6
 513         case AF_INET6:
 514         {
 515                 struct in6_addr ia6;
 516                 struct ifaddr *ifa;
 517                 int found;
 518
 519                 found = 0;
 520                 /*
 521                  * Try to find an address on the given outgoing interface
 522                  * that belongs to the jail.
 523                  */
 524                 IF_ADDR_RLOCK(ifp);
 525                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 526                         struct sockaddr *sa;
 527                         sa = ifa->ifa_addr;
 528                         if (sa->sa_family != AF_INET6)
 529                                 continue;
 530                         bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
 531                             &ia6, sizeof(struct in6_addr));
 532                         if (prison_check_ip6(cred, &ia6) == 0) {
 533                                 found = 1;
 534                                 break;
 535                         }
 536                 }
 537                 IF_ADDR_RUNLOCK(ifp);
 538                 if (!found) {
 539                         /*
 540                          * As a last resort return the 'default' jail address.
 541                          */
 542                         ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
 543                             sin6_addr;
 544                         if (prison_get_ip6(cred, &ia6) != 0)
 545                                 return (ESRCH);
 546                 }
 547                 bzero(&saun->sin6, sizeof(struct sockaddr_in6));
 548                 saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
 549                 saun->sin6.sin6_family = AF_INET6;
 550                 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
 551                 if (sa6_recoverscope(&saun->sin6) != 0)
 552                         return (ESRCH);
 553                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
 554                 break;
 555         }
 556 #endif
 557         default:
 558                 return (ESRCH);
 559         }
 560         return (0);
 561 }
 562
 563 /*ARGSUSED*/
 564 static int
 565 route_output(struct mbuf *m, struct socket *so)
 566 {
 567 #define sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
 568         struct rt_msghdr *rtm = NULL;
 569         struct rtentry *rt = NULL;
 570         struct radix_node_head *rnh;
 571         struct rt_addrinfo info;
 572         int len, error = 0;
 573         struct ifnet *ifp = NULL;
 574         union sockaddr_union saun;
 575         sa_family_t saf = AF_UNSPEC;
 576
 577 #define senderr(e) { error = e; goto flush;}
 578         if (m == NULL || ((m->m_len < sizeof(long)) &&
 579                        (m = m_pullup(m, sizeof(long))) == NULL))
 580                 return (ENOBUFS);
 581         if ((m->m_flags & M_PKTHDR) == 0)
 582                 panic("route_output");
 583         len = m->m_pkthdr.len;
 584         if (len < sizeof(*rtm) ||
 585             len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
 586                 info.rti_info[RTAX_DST] = NULL;
 587                 senderr(EINVAL);
 588         }
 589         R_Malloc(rtm, struct rt_msghdr *, len);
 590         if (rtm == NULL) {
 591                 info.rti_info[RTAX_DST] = NULL;
 592                 senderr(ENOBUFS);
 593         }
 594         m_copydata(m, 0, len, (caddr_t)rtm);
 595         if (rtm->rtm_version != RTM_VERSION) {
 596                 info.rti_info[RTAX_DST] = NULL;
 597                 senderr(EPROTONOSUPPORT);
 598         }
 599         rtm->rtm_pid = curproc->p_pid;
 600         bzero(&info, sizeof(info));
 601         info.rti_addrs = rtm->rtm_addrs;
 602         if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
 603                 info.rti_info[RTAX_DST] = NULL;
 604                 senderr(EINVAL);
 605         }
 606         info.rti_flags = rtm->rtm_flags;
 607         if (info.rti_info[RTAX_DST] == NULL ||
 608             info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
 609             (info.rti_info[RTAX_GATEWAY] != NULL &&
 610              info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
 611                 senderr(EINVAL);
 612         saf = info.rti_info[RTAX_DST]->sa_family;
 613         /*
 614          * Verify that the caller has the appropriate privilege; RTM_GET
 615          * is the only operation the non-superuser is allowed.
 616          */
 617         if (rtm->rtm_type != RTM_GET) {
 618                 error = priv_check(curthread, PRIV_NET_ROUTE);
 619                 if (error)
 620                         senderr(error);
 621         }
 622
 623         /*
 624          * The given gateway address may be an interface address.
 625          * For example, issuing a "route change" command on a route
 626          * entry that was created from a tunnel, and the gateway
 627          * address given is the local end point. In this case the
 628          * RTF_GATEWAY flag must be cleared or the destination will
 629          * not be reachable even though there is no error message.
 630          */
 631         if (info.rti_info[RTAX_GATEWAY] != NULL &&
 632             info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
 633                 struct route gw_ro;
 634
 635                 bzero(&gw_ro, sizeof(gw_ro));
 636                 gw_ro.ro_dst = *info.rti_info[RTAX_GATEWAY];
 637                 rtalloc_ign_fib(&gw_ro, 0, so->so_fibnum);
 638                 /*
 639                  * A host route through the loopback interface is
 640                  * installed for each interface adddress. In pre 8.0
 641                  * releases the interface address of a PPP link type
 642                  * is not reachable locally. This behavior is fixed as
 643                  * part of the new L2/L3 redesign and rewrite work. The
 644                  * signature of this interface address route is the
 645                  * AF_LINK sa_family type of the rt_gateway, and the
 646                  * rt_ifp has the IFF_LOOPBACK flag set.
 647                  */
 648                 if (gw_ro.ro_rt != NULL &&
 649                     gw_ro.ro_rt->rt_gateway->sa_family == AF_LINK &&
 650                     gw_ro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK)
 651                         info.rti_flags &= ~RTF_GATEWAY;
 652                 if (gw_ro.ro_rt != NULL)
 653                         RTFREE(gw_ro.ro_rt);
 654         }
 655
 656         switch (rtm->rtm_type) {
 657                 struct rtentry *saved_nrt;
 658
 659         case RTM_ADD:
 660                 if (info.rti_info[RTAX_GATEWAY] == NULL)
 661                         senderr(EINVAL);
 662                 saved_nrt = NULL;
 663
 664                 /* support for new ARP code */
 665                 if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
 666                     (rtm->rtm_flags & RTF_LLDATA) != 0) {
 667                         error = lla_rt_output(rtm, &info);
 668                         break;
 669                 }
 670                 error = rtrequest1_fib(RTM_ADD, &info, &saved_nrt,
 671                     so->so_fibnum);
 672                 if (error == 0 && saved_nrt) {
 673                         RT_LOCK(saved_nrt);
 674                         rt_setmetrics(rtm->rtm_inits,
 675                                 &rtm->rtm_rmx, &saved_nrt->rt_rmx);
 676                         rtm->rtm_index = saved_nrt->rt_ifp->if_index;
 677                         RT_REMREF(saved_nrt);
 678                         RT_UNLOCK(saved_nrt);
 679                 }
 680                 break;
 681
 682         case RTM_DELETE:
 683                 saved_nrt = NULL;
 684                 /* support for new ARP code */
 685                 if (info.rti_info[RTAX_GATEWAY] &&
 686                     (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
 687                     (rtm->rtm_flags & RTF_LLDATA) != 0) {
 688                         error = lla_rt_output(rtm, &info);
 689                         break;
 690                 }
 691                 error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt,
 692                     so->so_fibnum);
 693                 if (error == 0) {
 694                         RT_LOCK(saved_nrt);
 695                         rt = saved_nrt;
 696                         goto report;
 697                 }
 698                 break;
 699
 700         case RTM_GET:
 701         case RTM_CHANGE:
 702         case RTM_LOCK:
 703                 rnh = rt_tables_get_rnh(so->so_fibnum,
 704                     info.rti_info[RTAX_DST]->sa_family);
 705                 if (rnh == NULL)
 706                         senderr(EAFNOSUPPORT);
 707                 RADIX_NODE_HEAD_RLOCK(rnh);
 708                 rt = (struct rtentry *) rnh->rnh_lookup(info.rti_info[RTAX_DST],
 709                         info.rti_info[RTAX_NETMASK], rnh);
 710                 if (rt == NULL) {       /* XXX looks bogus */
 711                         RADIX_NODE_HEAD_RUNLOCK(rnh);
 712                         senderr(ESRCH);
 713                 }
 714 #ifdef RADIX_MPATH
 715                 /*
 716                  * for RTM_CHANGE/LOCK, if we got multipath routes,
 717                  * we require users to specify a matching RTAX_GATEWAY.
 718                  *
 719                  * for RTM_GET, gate is optional even with multipath.
 720                  * if gate == NULL the first match is returned.
 721                  * (no need to call rt_mpath_matchgate if gate == NULL)
 722                  */
 723                 if (rn_mpath_capable(rnh) &&
 724                     (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
 725                         rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
 726                         if (!rt) {
 727                                 RADIX_NODE_HEAD_RUNLOCK(rnh);
 728                                 senderr(ESRCH);
 729                         }
 730                 }
 731 #endif
 732                 /*
 733                  * If performing proxied L2 entry insertion, and
 734                  * the actual PPP host entry is found, perform
 735                  * another search to retrieve the prefix route of
 736                  * the local end point of the PPP link.
 737                  */
 738                 if (rtm->rtm_flags & RTF_ANNOUNCE) {
 739                         struct sockaddr laddr;
 740
 741                         if (rt->rt_ifp != NULL &&
 742                             rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
 743                                 struct ifaddr *ifa;
 744
 745                                 ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1);
 746                                 if (ifa != NULL)
 747                                         rt_maskedcopy(ifa->ifa_addr,
 748                                                       &laddr,
 749                                                       ifa->ifa_netmask);
 750                         } else
 751                                 rt_maskedcopy(rt->rt_ifa->ifa_addr,
 752                                               &laddr,
 753                                               rt->rt_ifa->ifa_netmask);
 754                         /*
 755                          * refactor rt and no lock operation necessary
 756                          */
 757                         rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr, rnh);
 758                         if (rt == NULL) {
 759                                 RADIX_NODE_HEAD_RUNLOCK(rnh);
 760                                 senderr(ESRCH);
 761                         }
 762                 }
 763                 RT_LOCK(rt);
 764                 RT_ADDREF(rt);
 765                 RADIX_NODE_HEAD_RUNLOCK(rnh);
 766
 767                 /*
 768                  * Fix for PR: 82974
 769                  *
 770                  * RTM_CHANGE/LOCK need a perfect match, rn_lookup()
 771                  * returns a perfect match in case a netmask is
 772                  * specified.  For host routes only a longest prefix
 773                  * match is returned so it is necessary to compare the
 774                  * existence of the netmask.  If both have a netmask
 775                  * rnh_lookup() did a perfect match and if none of them
 776                  * have a netmask both are host routes which is also a
 777                  * perfect match.
 778                  */
 779
 780                 if (rtm->rtm_type != RTM_GET &&
 781                     (!rt_mask(rt) != !info.rti_info[RTAX_NETMASK])) {
 782                         RT_UNLOCK(rt);
 783                         senderr(ESRCH);
 784                 }
 785
 786                 switch(rtm->rtm_type) {
 787
 788                 case RTM_GET:
 789                 report:
 790                         RT_LOCK_ASSERT(rt);
 791                         if ((rt->rt_flags & RTF_HOST) == 0
 792                             ? jailed_without_vnet(curthread->td_ucred)
 793                             : prison_if(curthread->td_ucred,
 794                             rt_key(rt)) != 0) {
 795                                 RT_UNLOCK(rt);
 796                                 senderr(ESRCH);
 797                         }
 798                         info.rti_info[RTAX_DST] = rt_key(rt);
 799                         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
 800                         info.rti_info[RTAX_NETMASK] = rt_mask(rt);
 801                         info.rti_info[RTAX_GENMASK] = 0;
 802                         if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
 803                                 ifp = rt->rt_ifp;
 804                                 if (ifp) {
 805                                         info.rti_info[RTAX_IFP] =
 806                                             ifp->if_addr->ifa_addr;
 807                                         error = rtm_get_jailed(&info, ifp, rt,
 808                                             &saun, curthread->td_ucred);
 809                                         if (error != 0) {
 810                                                 RT_UNLOCK(rt);
 811                                                 senderr(error);
 812                                         }
 813                                         if (ifp->if_flags & IFF_POINTOPOINT)
 814                                                 info.rti_info[RTAX_BRD] =
 815                                                     rt->rt_ifa->ifa_dstaddr;
 816                                         rtm->rtm_index = ifp->if_index;
 817                                 } else {
 818                                         info.rti_info[RTAX_IFP] = NULL;
 819                                         info.rti_info[RTAX_IFA] = NULL;
 820                                 }
 821                         } else if ((ifp = rt->rt_ifp) != NULL) {
 822                                 rtm->rtm_index = ifp->if_index;
 823                         }
 824                         len = rt_msg2(rtm->rtm_type, &info, NULL, NULL);
 825                         if (len > rtm->rtm_msglen) {
 826                                 struct rt_msghdr *new_rtm;
 827                                 R_Malloc(new_rtm, struct rt_msghdr *, len);
 828                                 if (new_rtm == NULL) {
 829                                         RT_UNLOCK(rt);
 830                                         senderr(ENOBUFS);
 831                                 }
 832                                 bcopy(rtm, new_rtm, rtm->rtm_msglen);
 833                                 Free(rtm); rtm = new_rtm;
 834                         }
 835                         (void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm, NULL);
 836                         rtm->rtm_flags = rt->rt_flags;
 837                         rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
 838                         rtm->rtm_addrs = info.rti_addrs;
 839                         break;
 840
 841                 case RTM_CHANGE:
 842                         /*
 843                          * New gateway could require new ifaddr, ifp;
 844                          * flags may also be different; ifp may be specified
 845                          * by ll sockaddr when protocol address is ambiguous
 846                          */
 847                         if (((rt->rt_flags & RTF_GATEWAY) &&
 848                              info.rti_info[RTAX_GATEWAY] != NULL) ||
 849                             info.rti_info[RTAX_IFP] != NULL ||
 850                             (info.rti_info[RTAX_IFA] != NULL &&
 851                              !sa_equal(info.rti_info[RTAX_IFA],
 852                                        rt->rt_ifa->ifa_addr))) {
 853                                 RT_UNLOCK(rt);
 854                                 RADIX_NODE_HEAD_LOCK(rnh);
 855                                 error = rt_getifa_fib(&info, rt->rt_fibnum);
 856                                 /*
 857                                  * XXXRW: Really we should release this
 858                                  * reference later, but this maintains
 859                                  * historical behavior.
 860                                  */
 861                                 if (info.rti_ifa != NULL)
 862                                         ifa_free(info.rti_ifa);
 863                                 RADIX_NODE_HEAD_UNLOCK(rnh);
 864                                 if (error != 0)
 865                                         senderr(error);
 866                                 RT_LOCK(rt);
 867                         }
 868                         if (info.rti_ifa != NULL &&
 869                             info.rti_ifa != rt->rt_ifa &&
 870                             rt->rt_ifa != NULL &&
 871                             rt->rt_ifa->ifa_rtrequest != NULL) {
 872                                 rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
 873                                     &info);
 874                                 ifa_free(rt->rt_ifa);
 875                         }
 876                         if (info.rti_info[RTAX_GATEWAY] != NULL) {
 877                                 RT_UNLOCK(rt);
 878                                 RADIX_NODE_HEAD_LOCK(rnh);
 879                                 RT_LOCK(rt);
 880
 881                                 error = rt_setgate(rt, rt_key(rt),
 882                                     info.rti_info[RTAX_GATEWAY]);
 883                                 RADIX_NODE_HEAD_UNLOCK(rnh);
 884                                 if (error != 0) {
 885                                         RT_UNLOCK(rt);
 886                                         senderr(error);
 887                                 }
 888                                 rt->rt_flags |= (RTF_GATEWAY & info.rti_flags);
 889                         }
 890                         if (info.rti_ifa != NULL &&
 891                             info.rti_ifa != rt->rt_ifa) {
 892                                 ifa_ref(info.rti_ifa);
 893                                 rt->rt_ifa = info.rti_ifa;
 894                                 rt->rt_ifp = info.rti_ifp;
 895                         }
 896                         /* Allow some flags to be toggled on change. */
 897                         rt->rt_flags = (rt->rt_flags & ~RTF_FMASK) |
 898                                     (rtm->rtm_flags & RTF_FMASK);
 899                         rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
 900                                         &rt->rt_rmx);
 901                         rtm->rtm_index = rt->rt_ifp->if_index;
 902                         if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
 903                                rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
 904                         /* FALLTHROUGH */
 905                 case RTM_LOCK:
 906                         /* We don't support locks anymore */
 907                         break;
 908                 }
 909                 RT_UNLOCK(rt);
 910                 break;
 911
 912         default:
 913                 senderr(EOPNOTSUPP);
 914         }
 915
 916 flush:
 917         if (rtm) {
 918                 if (error)
 919                         rtm->rtm_errno = error;
 920                 else
 921                         rtm->rtm_flags |= RTF_DONE;
 922         }
 923         if (rt)         /* XXX can this be true? */
 924                 RTFREE(rt);
 925     {
 926         struct rawcb *rp = NULL;
 927         /*
 928          * Check to see if we don't want our own messages.
 929          */
 930         if ((so->so_options & SO_USELOOPBACK) == 0) {
 931                 if (route_cb.any_count <= 1) {
 932                         if (rtm)
 933                                 Free(rtm);
 934                         m_freem(m);
 935                         return (error);
 936                 }
 937                 /* There is another listener, so construct message */
 938                 rp = sotorawcb(so);
 939         }
 940         if (rtm) {
 941                 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
 942                 if (m->m_pkthdr.len < rtm->rtm_msglen) {
 943                         m_freem(m);
 944                         m = NULL;
 945                 } else if (m->m_pkthdr.len > rtm->rtm_msglen)
 946                         m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
 947         }
 948         if (m) {
 949                 M_SETFIB(m, so->so_fibnum);
 950                 m->m_flags |= RTS_FILTER_FIB;
 951                 if (rp) {
 952                         /*
 953                          * XXX insure we don't get a copy by
 954                          * invalidating our protocol
 955                          */
 956                         unsigned short family = rp->rcb_proto.sp_family;
 957                         rp->rcb_proto.sp_family = 0;
 958                         rt_dispatch(m, saf);
 959                         rp->rcb_proto.sp_family = family;
 960                 } else
 961                         rt_dispatch(m, saf);
 962         }
 963         /* info.rti_info[RTAX_DST] (used above) can point inside of rtm */
 964         if (rtm)
 965                 Free(rtm);
 966     }
 967         return (error);
 968 #undef  sa_equal
 969 }
 970
 971 static void
 972 rt_setmetrics(u_long which, const struct rt_metrics *in,
 973         struct rt_metrics_lite *out)
 974 {
 975 #define metric(f, e) if (which & (f)) out->e = in->e;
 976         /*
 977          * Only these are stored in the routing entry since introduction
 978          * of tcp hostcache. The rest is ignored.
 979          */
 980         metric(RTV_MTU, rmx_mtu);
 981         metric(RTV_WEIGHT, rmx_weight);
 982         /* Userland -> kernel timebase conversion. */
 983         if (which & RTV_EXPIRE)
 984                 out->rmx_expire = in->rmx_expire ?
 985                     in->rmx_expire - time_second + time_uptime : 0;
 986 #undef metric
 987 }
 988
 989 static void
 990 rt_getmetrics(const struct rt_metrics_lite *in, struct rt_metrics *out)
 991 {
 992 #define metric(e) out->e = in->e;
 993         bzero(out, sizeof(*out));
 994         metric(rmx_mtu);
 995         metric(rmx_weight);
 996         /* Kernel -> userland timebase conversion. */
 997         out->rmx_expire = in->rmx_expire ?
 998             in->rmx_expire - time_uptime + time_second : 0;
 999 #undef metric
1000 }
1001
1002 /*
1003  * Extract the addresses of the passed sockaddrs.
1004  * Do a little sanity checking so as to avoid bad memory references.
1005  * This data is derived straight from userland.
1006  */
1007 static int
1008 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1009 {
1010         struct sockaddr *sa;
1011         int i;
1012
1013         for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1014                 if ((rtinfo->rti_addrs & (1 << i)) == 0)
1015                         continue;
1016                 sa = (struct sockaddr *)cp;
1017                 /*
1018                  * It won't fit.
1019                  */
1020                 if (cp + sa->sa_len > cplim)
1021                         return (EINVAL);
1022                 /*
1023                  * there are no more.. quit now
1024                  * If there are more bits, they are in error.
1025                  * I've seen this. route(1) can evidently generate these.
1026                  * This causes kernel to core dump.
1027                  * for compatibility, If we see this, point to a safe address.
1028                  */
1029                 if (sa->sa_len == 0) {
1030                         rtinfo->rti_info[i] = &sa_zero;
1031                         return (0); /* should be EINVAL but for compat */
1032                 }
1033                 /* accept it */
1034                 rtinfo->rti_info[i] = sa;
1035                 cp += SA_SIZE(sa);
1036         }
1037         return (0);
1038 }
1039
1040 /*
1041  * Used by the routing socket.
1042  */
1043 static struct mbuf *
1044 rt_msg1(int type, struct rt_addrinfo *rtinfo)
1045 {
1046         struct rt_msghdr *rtm;
1047         struct mbuf *m;
1048         int i;
1049         struct sockaddr *sa;
1050         int len, dlen;
1051
1052         switch (type) {
1053
1054         case RTM_DELADDR:
1055         case RTM_NEWADDR:
1056                 len = sizeof(struct ifa_msghdr);
1057                 break;
1058
1059         case RTM_DELMADDR:
1060         case RTM_NEWMADDR:
1061                 len = sizeof(struct ifma_msghdr);
1062                 break;
1063
1064         case RTM_IFINFO:
1065                 len = sizeof(struct if_msghdr);
1066                 break;
1067
1068         case RTM_IFANNOUNCE:
1069         case RTM_IEEE80211:
1070                 len = sizeof(struct if_announcemsghdr);
1071                 break;
1072
1073         default:
1074                 len = sizeof(struct rt_msghdr);
1075         }
1076         if (len > MCLBYTES)
1077                 panic("rt_msg1");
1078         m = m_gethdr(M_DONTWAIT, MT_DATA);
1079         if (m && len > MHLEN) {
1080                 MCLGET(m, M_DONTWAIT);
1081                 if ((m->m_flags & M_EXT) == 0) {
1082                         m_free(m);
1083                         m = NULL;
1084                 }
1085         }
1086         if (m == NULL)
1087                 return (m);
1088         m->m_pkthdr.len = m->m_len = len;
1089         m->m_pkthdr.rcvif = NULL;
1090         rtm = mtod(m, struct rt_msghdr *);
1091         bzero((caddr_t)rtm, len);
1092         for (i = 0; i < RTAX_MAX; i++) {
1093                 if ((sa = rtinfo->rti_info[i]) == NULL)
1094                         continue;
1095                 rtinfo->rti_addrs |= (1 << i);
1096                 dlen = SA_SIZE(sa);
1097                 m_copyback(m, len, dlen, (caddr_t)sa);
1098                 len += dlen;
1099         }
1100         if (m->m_pkthdr.len != len) {
1101                 m_freem(m);
1102                 return (NULL);
1103         }
1104         rtm->rtm_msglen = len;
1105         rtm->rtm_version = RTM_VERSION;
1106         rtm->rtm_type = type;
1107         return (m);
1108 }
1109
1110 /*
1111  * Used by the sysctl code and routing socket.
1112  */
1113 static int
1114 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
1115 {
1116         int i;
1117         int len, dlen, second_time = 0;
1118         caddr_t cp0;
1119
1120         rtinfo->rti_addrs = 0;
1121 again:
1122         switch (type) {
1123
1124         case RTM_DELADDR:
1125         case RTM_NEWADDR:
1126                 if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1127 #ifdef COMPAT_FREEBSD32
1128                         if (w->w_req->flags & SCTL_MASK32)
1129                                 len = sizeof(struct ifa_msghdrl32);
1130                         else
1131 #endif
1132                                 len = sizeof(struct ifa_msghdrl);
1133                 } else
1134                         len = sizeof(struct ifa_msghdr);
1135                 break;
1136
1137         case RTM_IFINFO:
1138 #ifdef COMPAT_FREEBSD32
1139                 if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1140                         if (w->w_op == NET_RT_IFLISTL)
1141                                 len = sizeof(struct if_msghdrl32);
1142                         else
1143                                 len = sizeof(struct if_msghdr32);
1144                         break;
1145                 }
1146 #endif
1147                 if (w != NULL && w->w_op == NET_RT_IFLISTL)
1148                         len = sizeof(struct if_msghdrl);
1149                 else
1150                         len = sizeof(struct if_msghdr);
1151                 break;
1152
1153         case RTM_NEWMADDR:
1154                 len = sizeof(struct ifma_msghdr);
1155                 break;
1156
1157         default:
1158                 len = sizeof(struct rt_msghdr);
1159         }
1160         cp0 = cp;
1161         if (cp0)
1162                 cp += len;
1163         for (i = 0; i < RTAX_MAX; i++) {
1164                 struct sockaddr *sa;
1165
1166                 if ((sa = rtinfo->rti_info[i]) == NULL)
1167                         continue;
1168                 rtinfo->rti_addrs |= (1 << i);
1169                 dlen = SA_SIZE(sa);
1170                 if (cp) {
1171                         bcopy((caddr_t)sa, cp, (unsigned)dlen);
1172                         cp += dlen;
1173                 }
1174                 len += dlen;
1175         }
1176         len = ALIGN(len);
1177         if (cp == NULL && w != NULL && !second_time) {
1178                 struct walkarg *rw = w;
1179
1180                 if (rw->w_req) {
1181                         if (rw->w_tmemsize < len) {
1182                                 if (rw->w_tmem)
1183                                         free(rw->w_tmem, M_RTABLE);
1184                                 rw->w_tmem = (caddr_t)
1185                                         malloc(len, M_RTABLE, M_NOWAIT);
1186                                 if (rw->w_tmem)
1187                                         rw->w_tmemsize = len;
1188                         }
1189                         if (rw->w_tmem) {
1190                                 cp = rw->w_tmem;
1191                                 second_time = 1;
1192                                 goto again;
1193                         }
1194                 }
1195         }
1196         if (cp) {
1197                 struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
1198
1199                 rtm->rtm_version = RTM_VERSION;
1200                 rtm->rtm_type = type;
1201                 rtm->rtm_msglen = len;
1202         }
1203         return (len);
1204 }
1205
1206 /*
1207  * This routine is called to generate a message from the routing
1208  * socket indicating that a redirect has occured, a routing lookup
1209  * has failed, or that a protocol has detected timeouts to a particular
1210  * destination.
1211  */
1212 void
1213 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1214     int fibnum)
1215 {
1216         struct rt_msghdr *rtm;
1217         struct mbuf *m;
1218         struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1219
1220         if (route_cb.any_count == 0)
1221                 return;
1222         m = rt_msg1(type, rtinfo);
1223         if (m == NULL)
1224                 return;
1225
1226         if (fibnum != RTS_ALLFIBS) {
1227                 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1228                     "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1229                 M_SETFIB(m, fibnum);
1230                 m->m_flags |= RTS_FILTER_FIB;
1231         }
1232
1233         rtm = mtod(m, struct rt_msghdr *);
1234         rtm->rtm_flags = RTF_DONE | flags;
1235         rtm->rtm_errno = error;
1236         rtm->rtm_addrs = rtinfo->rti_addrs;
1237         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1238 }
1239
1240 void
1241 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1242 {
1243
1244         rt_missmsg_fib(type, rtinfo, flags, error, RTS_ALLFIBS);
1245 }
1246
1247 /*
1248  * This routine is called to generate a message from the routing
1249  * socket indicating that the status of a network interface has changed.
1250  */
1251 void
1252 rt_ifmsg(struct ifnet *ifp)
1253 {
1254         struct if_msghdr *ifm;
1255         struct mbuf *m;
1256         struct rt_addrinfo info;
1257
1258         if (route_cb.any_count == 0)
1259                 return;
1260         bzero((caddr_t)&info, sizeof(info));
1261         m = rt_msg1(RTM_IFINFO, &info);
1262         if (m == NULL)
1263                 return;
1264         ifm = mtod(m, struct if_msghdr *);
1265         ifm->ifm_index = ifp->if_index;
1266         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1267         ifm->ifm_data = ifp->if_data;
1268         ifm->ifm_addrs = 0;
1269         rt_dispatch(m, AF_UNSPEC);
1270 }
1271
1272 /*
1273  * This is called to generate messages from the routing socket
1274  * indicating a network interface has had addresses associated with it.
1275  * if we ever reverse the logic and replace messages TO the routing
1276  * socket indicate a request to configure interfaces, then it will
1277  * be unnecessary as the routing socket will automatically generate
1278  * copies of it.
1279  */
1280 void
1281 rt_newaddrmsg_fib(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt,
1282     int fibnum)
1283 {
1284         struct rt_addrinfo info;
1285         struct sockaddr *sa = NULL;
1286         int pass;
1287         struct mbuf *m = NULL;
1288         struct ifnet *ifp = ifa->ifa_ifp;
1289
1290         KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
1291                 ("unexpected cmd %u", cmd));
1292 #if defined(INET) || defined(INET6)
1293 #ifdef SCTP
1294         /*
1295          * notify the SCTP stack
1296          * this will only get called when an address is added/deleted
1297          * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1298          */
1299         sctp_addr_change(ifa, cmd);
1300 #endif /* SCTP */
1301 #endif
1302         if (route_cb.any_count == 0)
1303                 return;
1304         for (pass = 1; pass < 3; pass++) {
1305                 bzero((caddr_t)&info, sizeof(info));
1306                 if ((cmd == RTM_ADD && pass == 1) ||
1307                     (cmd == RTM_DELETE && pass == 2)) {
1308                         struct ifa_msghdr *ifam;
1309                         int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1310
1311                         info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1312                         info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1313                         info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1314                         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1315                         if ((m = rt_msg1(ncmd, &info)) == NULL)
1316                                 continue;
1317                         ifam = mtod(m, struct ifa_msghdr *);
1318                         ifam->ifam_index = ifp->if_index;
1319                         ifam->ifam_metric = ifa->ifa_metric;
1320                         ifam->ifam_flags = ifa->ifa_flags;
1321                         ifam->ifam_addrs = info.rti_addrs;
1322                 }
1323                 if ((cmd == RTM_ADD && pass == 2) ||
1324                     (cmd == RTM_DELETE && pass == 1)) {
1325                         struct rt_msghdr *rtm;
1326
1327                         if (rt == NULL)
1328                                 continue;
1329                         info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1330                         info.rti_info[RTAX_DST] = sa = rt_key(rt);
1331                         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1332                         if ((m = rt_msg1(cmd, &info)) == NULL)
1333                                 continue;
1334                         rtm = mtod(m, struct rt_msghdr *);
1335                         rtm->rtm_index = ifp->if_index;
1336                         rtm->rtm_flags |= rt->rt_flags;
1337                         rtm->rtm_errno = error;
1338                         rtm->rtm_addrs = info.rti_addrs;
1339                 }
1340                 if (fibnum != RTS_ALLFIBS) {
1341                         KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: "
1342                             "fibnum out of range 0 <= %d < %d", __func__,
1343                              fibnum, rt_numfibs));
1344                         M_SETFIB(m, fibnum);
1345                         m->m_flags |= RTS_FILTER_FIB;
1346                 }
1347                 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1348         }
1349 }
1350
1351 void
1352 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1353 {
1354
1355         rt_newaddrmsg_fib(cmd, ifa, error, rt, RTS_ALLFIBS);
1356 }
1357
1358 /*
1359  * This is the analogue to the rt_newaddrmsg which performs the same
1360  * function but for multicast group memberhips.  This is easier since
1361  * there is no route state to worry about.
1362  */
1363 void
1364 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1365 {
1366         struct rt_addrinfo info;
1367         struct mbuf *m = NULL;
1368         struct ifnet *ifp = ifma->ifma_ifp;
1369         struct ifma_msghdr *ifmam;
1370
1371         if (route_cb.any_count == 0)
1372                 return;
1373
1374         bzero((caddr_t)&info, sizeof(info));
1375         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1376         info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1377         /*
1378          * If a link-layer address is present, present it as a ``gateway''
1379          * (similarly to how ARP entries, e.g., are presented).
1380          */
1381         info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1382         m = rt_msg1(cmd, &info);
1383         if (m == NULL)
1384                 return;
1385         ifmam = mtod(m, struct ifma_msghdr *);
1386         KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1387             __func__));
1388         ifmam->ifmam_index = ifp->if_index;
1389         ifmam->ifmam_addrs = info.rti_addrs;
1390         rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1391 }
1392
1393 static struct mbuf *
1394 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1395         struct rt_addrinfo *info)
1396 {
1397         struct if_announcemsghdr *ifan;
1398         struct mbuf *m;
1399
1400         if (route_cb.any_count == 0)
1401                 return NULL;
1402         bzero((caddr_t)info, sizeof(*info));
1403         m = rt_msg1(type, info);
1404         if (m != NULL) {
1405                 ifan = mtod(m, struct if_announcemsghdr *);
1406                 ifan->ifan_index = ifp->if_index;
1407                 strlcpy(ifan->ifan_name, ifp->if_xname,
1408                         sizeof(ifan->ifan_name));
1409                 ifan->ifan_what = what;
1410         }
1411         return m;
1412 }
1413
1414 /*
1415  * This is called to generate routing socket messages indicating
1416  * IEEE80211 wireless events.
1417  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1418  */
1419 void
1420 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1421 {
1422         struct mbuf *m;
1423         struct rt_addrinfo info;
1424
1425         m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1426         if (m != NULL) {
1427                 /*
1428                  * Append the ieee80211 data.  Try to stick it in the
1429                  * mbuf containing the ifannounce msg; otherwise allocate
1430                  * a new mbuf and append.
1431                  *
1432                  * NB: we assume m is a single mbuf.
1433                  */
1434                 if (data_len > M_TRAILINGSPACE(m)) {
1435                         struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1436                         if (n == NULL) {
1437                                 m_freem(m);
1438                                 return;
1439                         }
1440                         bcopy(data, mtod(n, void *), data_len);
1441                         n->m_len = data_len;
1442                         m->m_next = n;
1443                 } else if (data_len > 0) {
1444                         bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1445                         m->m_len += data_len;
1446                 }
1447                 if (m->m_flags & M_PKTHDR)
1448                         m->m_pkthdr.len += data_len;
1449                 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1450                 rt_dispatch(m, AF_UNSPEC);
1451         }
1452 }
1453
1454 /*
1455  * This is called to generate routing socket messages indicating
1456  * network interface arrival and departure.
1457  */
1458 void
1459 rt_ifannouncemsg(struct ifnet *ifp, int what)
1460 {
1461         struct mbuf *m;
1462         struct rt_addrinfo info;
1463
1464         m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1465         if (m != NULL)
1466                 rt_dispatch(m, AF_UNSPEC);
1467 }
1468
1469 static void
1470 rt_dispatch(struct mbuf *m, sa_family_t saf)
1471 {
1472         struct m_tag *tag;
1473
1474         /*
1475          * Preserve the family from the sockaddr, if any, in an m_tag for
1476          * use when injecting the mbuf into the routing socket buffer from
1477          * the netisr.
1478          */
1479         if (saf != AF_UNSPEC) {
1480                 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1481                     M_NOWAIT);
1482                 if (tag == NULL) {
1483                         m_freem(m);
1484                         return;
1485                 }
1486                 *(unsigned short *)(tag + 1) = saf;
1487                 m_tag_prepend(m, tag);
1488         }
1489 #ifdef VIMAGE
1490         if (V_loif)
1491                 m->m_pkthdr.rcvif = V_loif;
1492         else {
1493                 m_freem(m);
1494                 return;
1495         }
1496 #endif
1497         netisr_queue(NETISR_ROUTE, m);  /* mbuf is free'd on failure. */
1498 }
1499
1500 /*
1501  * This is used in dumping the kernel table via sysctl().
1502  */
1503 static int
1504 sysctl_dumpentry(struct radix_node *rn, void *vw)
1505 {
1506         struct walkarg *w = vw;
1507         struct rtentry *rt = (struct rtentry *)rn;
1508         int error = 0, size;
1509         struct rt_addrinfo info;
1510
1511         if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1512                 return 0;
1513         if ((rt->rt_flags & RTF_HOST) == 0
1514             ? jailed_without_vnet(w->w_req->td->td_ucred)
1515             : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1516                 return (0);
1517         bzero((caddr_t)&info, sizeof(info));
1518         info.rti_info[RTAX_DST] = rt_key(rt);
1519         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1520         info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1521         info.rti_info[RTAX_GENMASK] = 0;
1522         if (rt->rt_ifp) {
1523                 info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1524                 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1525                 if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1526                         info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1527         }
1528         size = rt_msg2(RTM_GET, &info, NULL, w);
1529         if (w->w_req && w->w_tmem) {
1530                 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1531
1532                 rtm->rtm_flags = rt->rt_flags;
1533                 /*
1534                  * let's be honest about this being a retarded hack
1535                  */
1536                 rtm->rtm_fmask = rt->rt_rmx.rmx_pksent;
1537                 rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1538                 rtm->rtm_index = rt->rt_ifp->if_index;
1539                 rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1540                 rtm->rtm_addrs = info.rti_addrs;
1541                 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1542                 return (error);
1543         }
1544         return (error);
1545 }
1546
1547 #ifdef COMPAT_FREEBSD32
1548 static void
1549 copy_ifdata32(struct if_data *src, struct if_data32 *dst)
1550 {
1551
1552         bzero(dst, sizeof(*dst));
1553         CP(*src, *dst, ifi_type);
1554         CP(*src, *dst, ifi_physical);
1555         CP(*src, *dst, ifi_addrlen);
1556         CP(*src, *dst, ifi_hdrlen);
1557         CP(*src, *dst, ifi_link_state);
1558         dst->ifi_datalen = sizeof(struct if_data32);
1559         CP(*src, *dst, ifi_mtu);
1560         CP(*src, *dst, ifi_metric);
1561         CP(*src, *dst, ifi_baudrate);
1562         CP(*src, *dst, ifi_ipackets);
1563         CP(*src, *dst, ifi_ierrors);
1564         CP(*src, *dst, ifi_opackets);
1565         CP(*src, *dst, ifi_oerrors);
1566         CP(*src, *dst, ifi_collisions);
1567         CP(*src, *dst, ifi_ibytes);
1568         CP(*src, *dst, ifi_obytes);
1569         CP(*src, *dst, ifi_imcasts);
1570         CP(*src, *dst, ifi_omcasts);
1571         CP(*src, *dst, ifi_iqdrops);
1572         CP(*src, *dst, ifi_noproto);
1573         CP(*src, *dst, ifi_hwassist);
1574         CP(*src, *dst, ifi_epoch);
1575         TV_CP(*src, *dst, ifi_lastchange);
1576 }
1577 #endif
1578
1579 static int
1580 sysctl_iflist_ifml(struct ifnet *ifp, struct rt_addrinfo *info,
1581     struct walkarg *w, int len)
1582 {
1583         struct if_msghdrl *ifm;
1584
1585 #ifdef COMPAT_FREEBSD32
1586         if (w->w_req->flags & SCTL_MASK32) {
1587                 struct if_msghdrl32 *ifm32;
1588
1589                 ifm32 = (struct if_msghdrl32 *)w->w_tmem;
1590                 ifm32->ifm_addrs = info->rti_addrs;
1591                 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1592                 ifm32->ifm_index = ifp->if_index;
1593                 ifm32->_ifm_spare1 = 0;
1594                 ifm32->ifm_len = sizeof(*ifm32);
1595                 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1596
1597                 copy_ifdata32(&ifp->if_data, &ifm32->ifm_data);
1598
1599                 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm32, len));
1600         }
1601 #endif
1602         ifm = (struct if_msghdrl *)w->w_tmem;
1603         ifm->ifm_addrs = info->rti_addrs;
1604         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1605         ifm->ifm_index = ifp->if_index;
1606         ifm->_ifm_spare1 = 0;
1607         ifm->ifm_len = sizeof(*ifm);
1608         ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1609
1610         ifm->ifm_data = ifp->if_data;
1611
1612         return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1613 }
1614
1615 static int
1616 sysctl_iflist_ifm(struct ifnet *ifp, struct rt_addrinfo *info,
1617     struct walkarg *w, int len)
1618 {
1619         struct if_msghdr *ifm;
1620
1621 #ifdef COMPAT_FREEBSD32
1622         if (w->w_req->flags & SCTL_MASK32) {
1623                 struct if_msghdr32 *ifm32;
1624
1625                 ifm32 = (struct if_msghdr32 *)w->w_tmem;
1626                 ifm32->ifm_addrs = info->rti_addrs;
1627                 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1628                 ifm32->ifm_index = ifp->if_index;
1629
1630                 copy_ifdata32(&ifp->if_data, &ifm32->ifm_data);
1631
1632                 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm32, len));
1633         }
1634 #endif
1635         ifm = (struct if_msghdr *)w->w_tmem;
1636         ifm->ifm_addrs = info->rti_addrs;
1637         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1638         ifm->ifm_index = ifp->if_index;
1639
1640         ifm->ifm_data = ifp->if_data;
1641
1642         return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1643 }
1644
1645 static int
1646 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1647     struct walkarg *w, int len)
1648 {
1649         struct ifa_msghdrl *ifam;
1650
1651 #ifdef COMPAT_FREEBSD32
1652         if (w->w_req->flags & SCTL_MASK32) {
1653                 struct ifa_msghdrl32 *ifam32;
1654
1655                 ifam32 = (struct ifa_msghdrl32 *)w->w_tmem;
1656                 ifam32->ifam_addrs = info->rti_addrs;
1657                 ifam32->ifam_flags = ifa->ifa_flags;
1658                 ifam32->ifam_index = ifa->ifa_ifp->if_index;
1659                 ifam32->_ifam_spare1 = 0;
1660                 ifam32->ifam_len = sizeof(*ifam32);
1661                 ifam32->ifam_data_off =
1662                     offsetof(struct ifa_msghdrl32, ifam_data);
1663                 ifam32->ifam_metric = ifa->ifa_metric;
1664
1665                 copy_ifdata32(&ifa->ifa_ifp->if_data, &ifam32->ifam_data);
1666
1667                 return (SYSCTL_OUT(w->w_req, (caddr_t)ifam32, len));
1668         }
1669 #endif
1670
1671         ifam = (struct ifa_msghdrl *)w->w_tmem;
1672         ifam->ifam_addrs = info->rti_addrs;
1673         ifam->ifam_flags = ifa->ifa_flags;
1674         ifam->ifam_index = ifa->ifa_ifp->if_index;
1675         ifam->_ifam_spare1 = 0;
1676         ifam->ifam_len = sizeof(*ifam);
1677         ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1678         ifam->ifam_metric = ifa->ifa_metric;
1679
1680         ifam->ifam_data = ifa->if_data;
1681
1682         return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1683 }
1684
1685 static int
1686 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1687     struct walkarg *w, int len)
1688 {
1689         struct ifa_msghdr *ifam;
1690
1691         ifam = (struct ifa_msghdr *)w->w_tmem;
1692         ifam->ifam_addrs = info->rti_addrs;
1693         ifam->ifam_flags = ifa->ifa_flags;
1694         ifam->ifam_index = ifa->ifa_ifp->if_index;
1695         ifam->ifam_metric = ifa->ifa_metric;
1696
1697         return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1698 }
1699
1700 static int
1701 sysctl_iflist(int af, struct walkarg *w)
1702 {
1703         struct ifnet *ifp;
1704         struct ifaddr *ifa;
1705         struct rt_addrinfo info;
1706         int len, error = 0;
1707
1708         bzero((caddr_t)&info, sizeof(info));
1709         IFNET_RLOCK();
1710         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1711                 if (w->w_arg && w->w_arg != ifp->if_index)
1712                         continue;
1713                 IF_ADDR_RLOCK(ifp);
1714                 ifa = ifp->if_addr;
1715                 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1716                 len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1717                 info.rti_info[RTAX_IFP] = NULL;
1718                 if (w->w_req && w->w_tmem) {
1719                         if (w->w_op == NET_RT_IFLISTL)
1720                                 error = sysctl_iflist_ifml(ifp, &info, w, len);
1721                         else
1722                                 error = sysctl_iflist_ifm(ifp, &info, w, len);
1723                         if (error)
1724                                 goto done;
1725                 }
1726                 while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1727                         if (af && af != ifa->ifa_addr->sa_family)
1728                                 continue;
1729                         if (prison_if(w->w_req->td->td_ucred,
1730                             ifa->ifa_addr) != 0)
1731                                 continue;
1732                         info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1733                         info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1734                         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1735                         len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1736                         if (w->w_req && w->w_tmem) {
1737                                 if (w->w_op == NET_RT_IFLISTL)
1738                                         error = sysctl_iflist_ifaml(ifa, &info,
1739                                             w, len);
1740                                 else
1741                                         error = sysctl_iflist_ifam(ifa, &info,
1742                                             w, len);
1743                                 if (error)
1744                                         goto done;
1745                         }
1746                 }
1747                 IF_ADDR_RUNLOCK(ifp);
1748                 info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1749                         info.rti_info[RTAX_BRD] = NULL;
1750         }
1751 done:
1752         if (ifp != NULL)
1753                 IF_ADDR_RUNLOCK(ifp);
1754         IFNET_RUNLOCK();
1755         return (error);
1756 }
1757
1758 static int
1759 sysctl_ifmalist(int af, struct walkarg *w)
1760 {
1761         struct ifnet *ifp;
1762         struct ifmultiaddr *ifma;
1763         struct  rt_addrinfo info;
1764         int     len, error = 0;
1765         struct ifaddr *ifa;
1766
1767         bzero((caddr_t)&info, sizeof(info));
1768         IFNET_RLOCK();
1769         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1770                 if (w->w_arg && w->w_arg != ifp->if_index)
1771                         continue;
1772                 ifa = ifp->if_addr;
1773                 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1774                 IF_ADDR_RLOCK(ifp);
1775                 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1776                         if (af && af != ifma->ifma_addr->sa_family)
1777                                 continue;
1778                         if (prison_if(w->w_req->td->td_ucred,
1779                             ifma->ifma_addr) != 0)
1780                                 continue;
1781                         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1782                         info.rti_info[RTAX_GATEWAY] =
1783                             (ifma->ifma_addr->sa_family != AF_LINK) ?
1784                             ifma->ifma_lladdr : NULL;
1785                         len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1786                         if (w->w_req && w->w_tmem) {
1787                                 struct ifma_msghdr *ifmam;
1788
1789                                 ifmam = (struct ifma_msghdr *)w->w_tmem;
1790                                 ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1791                                 ifmam->ifmam_flags = 0;
1792                                 ifmam->ifmam_addrs = info.rti_addrs;
1793                                 error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1794                                 if (error) {
1795                                         IF_ADDR_RUNLOCK(ifp);
1796                                         goto done;
1797                                 }
1798                         }
1799                 }
1800                 IF_ADDR_RUNLOCK(ifp);
1801         }
1802 done:
1803         IFNET_RUNLOCK();
1804         return (error);
1805 }
1806
1807 static int
1808 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1809 {
1810         int     *name = (int *)arg1;
1811         u_int   namelen = arg2;
1812         struct radix_node_head *rnh = NULL; /* silence compiler. */
1813         int     i, lim, error = EINVAL;
1814         int     fib = 0;
1815         u_char  af;
1816         struct  walkarg w;
1817
1818         name ++;
1819         namelen--;
1820         if (req->newptr)
1821                 return (EPERM);
1822         if (name[1] == NET_RT_DUMP) {
1823                 if (namelen == 3)
1824                         fib = req->td->td_proc->p_fibnum;
1825                 else if (namelen == 4)
1826                         fib = (name[3] == -1) ?
1827                             req->td->td_proc->p_fibnum : name[3];
1828                 else
1829                         return ((namelen < 3) ? EISDIR : ENOTDIR);
1830                 if (fib < 0 || fib >= rt_numfibs)
1831                         return (EINVAL);
1832         } else if (namelen != 3)
1833                 return ((namelen < 3) ? EISDIR : ENOTDIR);
1834         af = name[0];
1835         if (af > AF_MAX)
1836                 return (EINVAL);
1837         bzero(&w, sizeof(w));
1838         w.w_op = name[1];
1839         w.w_arg = name[2];
1840         w.w_req = req;
1841
1842         error = sysctl_wire_old_buffer(req, 0);
1843         if (error)
1844                 return (error);
1845         switch (w.w_op) {
1846
1847         case NET_RT_DUMP:
1848         case NET_RT_FLAGS:
1849                 if (af == 0) {                  /* dump all tables */
1850                         i = 1;
1851                         lim = AF_MAX;
1852                 } else                          /* dump only one table */
1853                         i = lim = af;
1854
1855                 /*
1856                  * take care of llinfo entries, the caller must
1857                  * specify an AF
1858                  */
1859                 if (w.w_op == NET_RT_FLAGS &&
1860                     (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1861                         if (af != 0)
1862                                 error = lltable_sysctl_dumparp(af, w.w_req);
1863                         else
1864                                 error = EINVAL;
1865                         break;
1866                 }
1867                 /*
1868                  * take care of routing entries
1869                  */
1870                 for (error = 0; error == 0 && i <= lim; i++) {
1871                         rnh = rt_tables_get_rnh(fib, i);
1872                         if (rnh != NULL) {
1873                                 RADIX_NODE_HEAD_RLOCK(rnh);
1874                                 error = rnh->rnh_walktree(rnh,
1875                                     sysctl_dumpentry, &w);
1876                                 RADIX_NODE_HEAD_RUNLOCK(rnh);
1877                         } else if (af != 0)
1878                                 error = EAFNOSUPPORT;
1879                 }
1880                 break;
1881
1882         case NET_RT_IFLIST:
1883         case NET_RT_IFLISTL:
1884                 error = sysctl_iflist(af, &w);
1885                 break;
1886
1887         case NET_RT_IFMALIST:
1888                 error = sysctl_ifmalist(af, &w);
1889                 break;
1890         }
1891         if (w.w_tmem)
1892                 free(w.w_tmem, M_RTABLE);
1893         return (error);
1894 }
1895
1896 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1897
1898 /*
1899  * Definitions of protocols supported in the ROUTE domain.
1900  */
1901
1902 static struct domain routedomain;               /* or at least forward */
1903
1904 static struct protosw routesw[] = {
1905 {
1906         .pr_type =              SOCK_RAW,
1907         .pr_domain =            &routedomain,
1908         .pr_flags =             PR_ATOMIC|PR_ADDR,
1909         .pr_output =            route_output,
1910         .pr_ctlinput =          raw_ctlinput,
1911         .pr_init =              raw_init,
1912         .pr_usrreqs =           &route_usrreqs
1913 }
1914 };
1915
1916 static struct domain routedomain = {
1917         .dom_family =           PF_ROUTE,
1918         .dom_name =              "route",
1919         .dom_protosw =          routesw,
1920         .dom_protoswNPROTOSW =  &routesw[sizeof(routesw)/sizeof(routesw[0])]
1921 };
1922
1923 VNET_DOMAIN_SET(route);