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 #include <netinet/ip_carp.h>
  67 #ifdef INET6
  68 #include <netinet6/scope6_var.h>
  69 #endif
  70
  71 #if defined(INET) || defined(INET6)
  72 #ifdef SCTP
  73 extern void sctp_addr_change(struct ifaddr *ifa, int cmd);
  74 #endif /* SCTP */
  75 #endif
  76
  77 #ifdef COMPAT_FREEBSD32
  78 #include <sys/mount.h>
  79 #include <compat/freebsd32/freebsd32.h>
  80
  81 struct if_data32 {
  82         uint8_t ifi_type;
  83         uint8_t ifi_physical;
  84         uint8_t ifi_addrlen;
  85         uint8_t ifi_hdrlen;
  86         uint8_t ifi_link_state;
  87         uint8_t ifi_vhid;
  88         uint8_t ifi_spare_char2;
  89         uint8_t ifi_datalen;
  90         uint32_t ifi_mtu;
  91         uint32_t ifi_metric;
  92         uint32_t ifi_baudrate;
  93         uint32_t ifi_ipackets;
  94         uint32_t ifi_ierrors;
  95         uint32_t ifi_opackets;
  96         uint32_t ifi_oerrors;
  97         uint32_t ifi_collisions;
  98         uint32_t ifi_ibytes;
  99         uint32_t ifi_obytes;
 100         uint32_t ifi_imcasts;
 101         uint32_t ifi_omcasts;
 102         uint32_t ifi_iqdrops;
 103         uint32_t ifi_noproto;
 104         uint32_t ifi_hwassist;
 105         int32_t ifi_epoch;
 106         struct  timeval32 ifi_lastchange;
 107 };
 108
 109 struct if_msghdr32 {
 110         uint16_t ifm_msglen;
 111         uint8_t ifm_version;
 112         uint8_t ifm_type;
 113         int32_t ifm_addrs;
 114         int32_t ifm_flags;
 115         uint16_t ifm_index;
 116         struct  if_data32 ifm_data;
 117 };
 118
 119 struct if_msghdrl32 {
 120         uint16_t ifm_msglen;
 121         uint8_t ifm_version;
 122         uint8_t ifm_type;
 123         int32_t ifm_addrs;
 124         int32_t ifm_flags;
 125         uint16_t ifm_index;
 126         uint16_t _ifm_spare1;
 127         uint16_t ifm_len;
 128         uint16_t ifm_data_off;
 129         struct  if_data32 ifm_data;
 130 };
 131
 132 struct ifa_msghdrl32 {
 133         uint16_t ifam_msglen;
 134         uint8_t ifam_version;
 135         uint8_t ifam_type;
 136         int32_t ifam_addrs;
 137         int32_t ifam_flags;
 138         uint16_t ifam_index;
 139         uint16_t _ifam_spare1;
 140         uint16_t ifam_len;
 141         uint16_t ifam_data_off;
 142         int32_t ifam_metric;
 143         struct  if_data32 ifam_data;
 144 };
 145 #endif /* COMPAT_FREEBSD32 */
 146
 147 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
 148
 149 /* NB: these are not modified */
 150 static struct   sockaddr route_src = { 2, PF_ROUTE, };
 151 static struct   sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
 152
 153 /* These are external hooks for CARP. */
 154 int     (*carp_get_vhid_p)(struct ifaddr *);
 155
 156 /*
 157  * Used by rtsock/raw_input callback code to decide whether to filter the update
 158  * notification to a socket bound to a particular FIB.
 159  */
 160 #define RTS_FILTER_FIB  M_PROTO8
 161 #define RTS_ALLFIBS     -1
 162
 163 static struct {
 164         int     ip_count;       /* attached w/ AF_INET */
 165         int     ip6_count;      /* attached w/ AF_INET6 */
 166         int     ipx_count;      /* attached w/ AF_IPX */
 167         int     any_count;      /* total attached */
 168 } route_cb;
 169
 170 struct mtx rtsock_mtx;
 171 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
 172
 173 #define RTSOCK_LOCK()   mtx_lock(&rtsock_mtx)
 174 #define RTSOCK_UNLOCK() mtx_unlock(&rtsock_mtx)
 175 #define RTSOCK_LOCK_ASSERT()    mtx_assert(&rtsock_mtx, MA_OWNED)
 176
 177 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
 178
 179 struct walkarg {
 180         int     w_tmemsize;
 181         int     w_op, w_arg;
 182         caddr_t w_tmem;
 183         struct sysctl_req *w_req;
 184 };
 185
 186 static void     rts_input(struct mbuf *m);
 187 static struct mbuf *rt_msg1(int type, struct rt_addrinfo *rtinfo);
 188 static int      rt_msg2(int type, struct rt_addrinfo *rtinfo,
 189                         caddr_t cp, struct walkarg *w);
 190 static int      rt_xaddrs(caddr_t cp, caddr_t cplim,
 191                         struct rt_addrinfo *rtinfo);
 192 static int      sysctl_dumpentry(struct radix_node *rn, void *vw);
 193 static int      sysctl_iflist(int af, struct walkarg *w);
 194 static int      sysctl_ifmalist(int af, struct walkarg *w);
 195 static int      route_output(struct mbuf *m, struct socket *so);
 196 static void     rt_setmetrics(u_long which, const struct rt_metrics *in,
 197                         struct rt_metrics_lite *out);
 198 static void     rt_getmetrics(const struct rt_metrics_lite *in,
 199                         struct rt_metrics *out);
 200 static void     rt_dispatch(struct mbuf *, sa_family_t);
 201
 202 static struct netisr_handler rtsock_nh = {
 203         .nh_name = "rtsock",
 204         .nh_handler = rts_input,
 205         .nh_proto = NETISR_ROUTE,
 206         .nh_policy = NETISR_POLICY_SOURCE,
 207 };
 208
 209 static int
 210 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
 211 {
 212         int error, qlimit;
 213
 214         netisr_getqlimit(&rtsock_nh, &qlimit);
 215         error = sysctl_handle_int(oidp, &qlimit, 0, req);
 216         if (error || !req->newptr)
 217                 return (error);
 218         if (qlimit < 1)
 219                 return (EINVAL);
 220         return (netisr_setqlimit(&rtsock_nh, qlimit));
 221 }
 222 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
 223     0, 0, sysctl_route_netisr_maxqlen, "I",
 224     "maximum routing socket dispatch queue length");
 225
 226 static void
 227 rts_init(void)
 228 {
 229         int tmp;
 230
 231         if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
 232                 rtsock_nh.nh_qlimit = tmp;
 233         netisr_register(&rtsock_nh);
 234 }
 235 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0);
 236
 237 static int
 238 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
 239     struct rawcb *rp)
 240 {
 241         int fibnum;
 242
 243         KASSERT(m != NULL, ("%s: m is NULL", __func__));
 244         KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
 245         KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
 246
 247         /* No filtering requested. */
 248         if ((m->m_flags & RTS_FILTER_FIB) == 0)
 249                 return (0);
 250
 251         /* Check if it is a rts and the fib matches the one of the socket. */
 252         fibnum = M_GETFIB(m);
 253         if (proto->sp_family != PF_ROUTE ||
 254             rp->rcb_socket == NULL ||
 255             rp->rcb_socket->so_fibnum == fibnum)
 256                 return (0);
 257
 258         /* Filtering requested and no match, the socket shall be skipped. */
 259         return (1);
 260 }
 261
 262 static void
 263 rts_input(struct mbuf *m)
 264 {
 265         struct sockproto route_proto;
 266         unsigned short *family;
 267         struct m_tag *tag;
 268
 269         route_proto.sp_family = PF_ROUTE;
 270         tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
 271         if (tag != NULL) {
 272                 family = (unsigned short *)(tag + 1);
 273                 route_proto.sp_protocol = *family;
 274                 m_tag_delete(m, tag);
 275         } else
 276                 route_proto.sp_protocol = 0;
 277
 278         raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
 279 }
 280
 281 /*
 282  * It really doesn't make any sense at all for this code to share much
 283  * with raw_usrreq.c, since its functionality is so restricted.  XXX
 284  */
 285 static void
 286 rts_abort(struct socket *so)
 287 {
 288
 289         raw_usrreqs.pru_abort(so);
 290 }
 291
 292 static void
 293 rts_close(struct socket *so)
 294 {
 295
 296         raw_usrreqs.pru_close(so);
 297 }
 298
 299 /* pru_accept is EOPNOTSUPP */
 300
 301 static int
 302 rts_attach(struct socket *so, int proto, struct thread *td)
 303 {
 304         struct rawcb *rp;
 305         int error;
 306
 307         KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
 308
 309         /* XXX */
 310         rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
 311         if (rp == NULL)
 312                 return ENOBUFS;
 313
 314         so->so_pcb = (caddr_t)rp;
 315         so->so_fibnum = td->td_proc->p_fibnum;
 316         error = raw_attach(so, proto);
 317         rp = sotorawcb(so);
 318         if (error) {
 319                 so->so_pcb = NULL;
 320                 free(rp, M_PCB);
 321                 return error;
 322         }
 323         RTSOCK_LOCK();
 324         switch(rp->rcb_proto.sp_protocol) {
 325         case AF_INET:
 326                 route_cb.ip_count++;
 327                 break;
 328         case AF_INET6:
 329                 route_cb.ip6_count++;
 330                 break;
 331         case AF_IPX:
 332                 route_cb.ipx_count++;
 333                 break;
 334         }
 335         route_cb.any_count++;
 336         RTSOCK_UNLOCK();
 337         soisconnected(so);
 338         so->so_options |= SO_USELOOPBACK;
 339         return 0;
 340 }
 341
 342 static int
 343 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
 344 {
 345
 346         return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
 347 }
 348
 349 static int
 350 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
 351 {
 352
 353         return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
 354 }
 355
 356 /* pru_connect2 is EOPNOTSUPP */
 357 /* pru_control is EOPNOTSUPP */
 358
 359 static void
 360 rts_detach(struct socket *so)
 361 {
 362         struct rawcb *rp = sotorawcb(so);
 363
 364         KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
 365
 366         RTSOCK_LOCK();
 367         switch(rp->rcb_proto.sp_protocol) {
 368         case AF_INET:
 369                 route_cb.ip_count--;
 370                 break;
 371         case AF_INET6:
 372                 route_cb.ip6_count--;
 373                 break;
 374         case AF_IPX:
 375                 route_cb.ipx_count--;
 376                 break;
 377         }
 378         route_cb.any_count--;
 379         RTSOCK_UNLOCK();
 380         raw_usrreqs.pru_detach(so);
 381 }
 382
 383 static int
 384 rts_disconnect(struct socket *so)
 385 {
 386
 387         return (raw_usrreqs.pru_disconnect(so));
 388 }
 389
 390 /* pru_listen is EOPNOTSUPP */
 391
 392 static int
 393 rts_peeraddr(struct socket *so, struct sockaddr **nam)
 394 {
 395
 396         return (raw_usrreqs.pru_peeraddr(so, nam));
 397 }
 398
 399 /* pru_rcvd is EOPNOTSUPP */
 400 /* pru_rcvoob is EOPNOTSUPP */
 401
 402 static int
 403 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
 404          struct mbuf *control, struct thread *td)
 405 {
 406
 407         return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
 408 }
 409
 410 /* pru_sense is null */
 411
 412 static int
 413 rts_shutdown(struct socket *so)
 414 {
 415
 416         return (raw_usrreqs.pru_shutdown(so));
 417 }
 418
 419 static int
 420 rts_sockaddr(struct socket *so, struct sockaddr **nam)
 421 {
 422
 423         return (raw_usrreqs.pru_sockaddr(so, nam));
 424 }
 425
 426 static struct pr_usrreqs route_usrreqs = {
 427         .pru_abort =            rts_abort,
 428         .pru_attach =           rts_attach,
 429         .pru_bind =             rts_bind,
 430         .pru_connect =          rts_connect,
 431         .pru_detach =           rts_detach,
 432         .pru_disconnect =       rts_disconnect,
 433         .pru_peeraddr =         rts_peeraddr,
 434         .pru_send =             rts_send,
 435         .pru_shutdown =         rts_shutdown,
 436         .pru_sockaddr =         rts_sockaddr,
 437         .pru_close =            rts_close,
 438 };
 439
 440 #ifndef _SOCKADDR_UNION_DEFINED
 441 #define _SOCKADDR_UNION_DEFINED
 442 /*
 443  * The union of all possible address formats we handle.
 444  */
 445 union sockaddr_union {
 446         struct sockaddr         sa;
 447         struct sockaddr_in      sin;
 448         struct sockaddr_in6     sin6;
 449 };
 450 #endif /* _SOCKADDR_UNION_DEFINED */
 451
 452 static int
 453 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
 454     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
 455 {
 456
 457         /* First, see if the returned address is part of the jail. */
 458         if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
 459                 info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
 460                 return (0);
 461         }
 462
 463         switch (info->rti_info[RTAX_DST]->sa_family) {
 464 #ifdef INET
 465         case AF_INET:
 466         {
 467                 struct in_addr ia;
 468                 struct ifaddr *ifa;
 469                 int found;
 470
 471                 found = 0;
 472                 /*
 473                  * Try to find an address on the given outgoing interface
 474                  * that belongs to the jail.
 475                  */
 476                 IF_ADDR_RLOCK(ifp);
 477                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 478                         struct sockaddr *sa;
 479                         sa = ifa->ifa_addr;
 480                         if (sa->sa_family != AF_INET)
 481                                 continue;
 482                         ia = ((struct sockaddr_in *)sa)->sin_addr;
 483                         if (prison_check_ip4(cred, &ia) == 0) {
 484                                 found = 1;
 485                                 break;
 486                         }
 487                 }
 488                 IF_ADDR_RUNLOCK(ifp);
 489                 if (!found) {
 490                         /*
 491                          * As a last resort return the 'default' jail address.
 492                          */
 493                         ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
 494                             sin_addr;
 495                         if (prison_get_ip4(cred, &ia) != 0)
 496                                 return (ESRCH);
 497                 }
 498                 bzero(&saun->sin, sizeof(struct sockaddr_in));
 499                 saun->sin.sin_len = sizeof(struct sockaddr_in);
 500                 saun->sin.sin_family = AF_INET;
 501                 saun->sin.sin_addr.s_addr = ia.s_addr;
 502                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
 503                 break;
 504         }
 505 #endif
 506 #ifdef INET6
 507         case AF_INET6:
 508         {
 509                 struct in6_addr ia6;
 510                 struct ifaddr *ifa;
 511                 int found;
 512
 513                 found = 0;
 514                 /*
 515                  * Try to find an address on the given outgoing interface
 516                  * that belongs to the jail.
 517                  */
 518                 IF_ADDR_RLOCK(ifp);
 519                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 520                         struct sockaddr *sa;
 521                         sa = ifa->ifa_addr;
 522                         if (sa->sa_family != AF_INET6)
 523                                 continue;
 524                         bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
 525                             &ia6, sizeof(struct in6_addr));
 526                         if (prison_check_ip6(cred, &ia6) == 0) {
 527                                 found = 1;
 528                                 break;
 529                         }
 530                 }
 531                 IF_ADDR_RUNLOCK(ifp);
 532                 if (!found) {
 533                         /*
 534                          * As a last resort return the 'default' jail address.
 535                          */
 536                         ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
 537                             sin6_addr;
 538                         if (prison_get_ip6(cred, &ia6) != 0)
 539                                 return (ESRCH);
 540                 }
 541                 bzero(&saun->sin6, sizeof(struct sockaddr_in6));
 542                 saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
 543                 saun->sin6.sin6_family = AF_INET6;
 544                 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
 545                 if (sa6_recoverscope(&saun->sin6) != 0)
 546                         return (ESRCH);
 547                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
 548                 break;
 549         }
 550 #endif
 551         default:
 552                 return (ESRCH);
 553         }
 554         return (0);
 555 }
 556
 557 /*ARGSUSED*/
 558 static int
 559 route_output(struct mbuf *m, struct socket *so)
 560 {
 561 #define sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
 562         struct rt_msghdr *rtm = NULL;
 563         struct rtentry *rt = NULL;
 564         struct radix_node_head *rnh;
 565         struct rt_addrinfo info;
 566         int len, error = 0;
 567         struct ifnet *ifp = NULL;
 568         union sockaddr_union saun;
 569         sa_family_t saf = AF_UNSPEC;
 570
 571 #define senderr(e) { error = e; goto flush;}
 572         if (m == NULL || ((m->m_len < sizeof(long)) &&
 573                        (m = m_pullup(m, sizeof(long))) == NULL))
 574                 return (ENOBUFS);
 575         if ((m->m_flags & M_PKTHDR) == 0)
 576                 panic("route_output");
 577         len = m->m_pkthdr.len;
 578         if (len < sizeof(*rtm) ||
 579             len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
 580                 info.rti_info[RTAX_DST] = NULL;
 581                 senderr(EINVAL);
 582         }
 583         R_Malloc(rtm, struct rt_msghdr *, len);
 584         if (rtm == NULL) {
 585                 info.rti_info[RTAX_DST] = NULL;
 586                 senderr(ENOBUFS);
 587         }
 588         m_copydata(m, 0, len, (caddr_t)rtm);
 589         if (rtm->rtm_version != RTM_VERSION) {
 590                 info.rti_info[RTAX_DST] = NULL;
 591                 senderr(EPROTONOSUPPORT);
 592         }
 593         rtm->rtm_pid = curproc->p_pid;
 594         bzero(&info, sizeof(info));
 595         info.rti_addrs = rtm->rtm_addrs;
 596         if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
 597                 info.rti_info[RTAX_DST] = NULL;
 598                 senderr(EINVAL);
 599         }
 600         info.rti_flags = rtm->rtm_flags;
 601         if (info.rti_info[RTAX_DST] == NULL ||
 602             info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
 603             (info.rti_info[RTAX_GATEWAY] != NULL &&
 604              info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
 605                 senderr(EINVAL);
 606         saf = info.rti_info[RTAX_DST]->sa_family;
 607         /*
 608          * Verify that the caller has the appropriate privilege; RTM_GET
 609          * is the only operation the non-superuser is allowed.
 610          */
 611         if (rtm->rtm_type != RTM_GET) {
 612                 error = priv_check(curthread, PRIV_NET_ROUTE);
 613                 if (error)
 614                         senderr(error);
 615         }
 616
 617         /*
 618          * The given gateway address may be an interface address.
 619          * For example, issuing a "route change" command on a route
 620          * entry that was created from a tunnel, and the gateway
 621          * address given is the local end point. In this case the
 622          * RTF_GATEWAY flag must be cleared or the destination will
 623          * not be reachable even though there is no error message.
 624          */
 625         if (info.rti_info[RTAX_GATEWAY] != NULL &&
 626             info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
 627                 struct route gw_ro;
 628
 629                 bzero(&gw_ro, sizeof(gw_ro));
 630                 gw_ro.ro_dst = *info.rti_info[RTAX_GATEWAY];
 631                 rtalloc_ign_fib(&gw_ro, 0, so->so_fibnum);
 632                 /*
 633                  * A host route through the loopback interface is
 634                  * installed for each interface adddress. In pre 8.0
 635                  * releases the interface address of a PPP link type
 636                  * is not reachable locally. This behavior is fixed as
 637                  * part of the new L2/L3 redesign and rewrite work. The
 638                  * signature of this interface address route is the
 639                  * AF_LINK sa_family type of the rt_gateway, and the
 640                  * rt_ifp has the IFF_LOOPBACK flag set.
 641                  */
 642                 if (gw_ro.ro_rt != NULL &&
 643                     gw_ro.ro_rt->rt_gateway->sa_family == AF_LINK &&
 644                     gw_ro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK)
 645                         info.rti_flags &= ~RTF_GATEWAY;
 646                 if (gw_ro.ro_rt != NULL)
 647                         RTFREE(gw_ro.ro_rt);
 648         }
 649
 650         switch (rtm->rtm_type) {
 651                 struct rtentry *saved_nrt;
 652
 653         case RTM_ADD:
 654                 if (info.rti_info[RTAX_GATEWAY] == NULL)
 655                         senderr(EINVAL);
 656                 saved_nrt = NULL;
 657
 658                 /* support for new ARP code */
 659                 if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
 660                     (rtm->rtm_flags & RTF_LLDATA) != 0) {
 661                         error = lla_rt_output(rtm, &info);
 662                         break;
 663                 }
 664                 error = rtrequest1_fib(RTM_ADD, &info, &saved_nrt,
 665                     so->so_fibnum);
 666                 if (error == 0 && saved_nrt) {
 667                         RT_LOCK(saved_nrt);
 668                         rt_setmetrics(rtm->rtm_inits,
 669                                 &rtm->rtm_rmx, &saved_nrt->rt_rmx);
 670                         rtm->rtm_index = saved_nrt->rt_ifp->if_index;
 671                         RT_REMREF(saved_nrt);
 672                         RT_UNLOCK(saved_nrt);
 673                 }
 674                 break;
 675
 676         case RTM_DELETE:
 677                 saved_nrt = NULL;
 678                 /* support for new ARP code */
 679                 if (info.rti_info[RTAX_GATEWAY] &&
 680                     (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
 681                     (rtm->rtm_flags & RTF_LLDATA) != 0) {
 682                         error = lla_rt_output(rtm, &info);
 683                         break;
 684                 }
 685                 error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt,
 686                     so->so_fibnum);
 687                 if (error == 0) {
 688                         RT_LOCK(saved_nrt);
 689                         rt = saved_nrt;
 690                         goto report;
 691                 }
 692                 break;
 693
 694         case RTM_GET:
 695         case RTM_CHANGE:
 696         case RTM_LOCK:
 697                 rnh = rt_tables_get_rnh(so->so_fibnum,
 698                     info.rti_info[RTAX_DST]->sa_family);
 699                 if (rnh == NULL)
 700                         senderr(EAFNOSUPPORT);
 701                 RADIX_NODE_HEAD_RLOCK(rnh);
 702                 rt = (struct rtentry *) rnh->rnh_lookup(info.rti_info[RTAX_DST],
 703                         info.rti_info[RTAX_NETMASK], rnh);
 704                 if (rt == NULL) {       /* XXX looks bogus */
 705                         RADIX_NODE_HEAD_RUNLOCK(rnh);
 706                         senderr(ESRCH);
 707                 }
 708 #ifdef RADIX_MPATH
 709                 /*
 710                  * for RTM_CHANGE/LOCK, if we got multipath routes,
 711                  * we require users to specify a matching RTAX_GATEWAY.
 712                  *
 713                  * for RTM_GET, gate is optional even with multipath.
 714                  * if gate == NULL the first match is returned.
 715                  * (no need to call rt_mpath_matchgate if gate == NULL)
 716                  */
 717                 if (rn_mpath_capable(rnh) &&
 718                     (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
 719                         rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
 720                         if (!rt) {
 721                                 RADIX_NODE_HEAD_RUNLOCK(rnh);
 722                                 senderr(ESRCH);
 723                         }
 724                 }
 725 #endif
 726                 /*
 727                  * If performing proxied L2 entry insertion, and
 728                  * the actual PPP host entry is found, perform
 729                  * another search to retrieve the prefix route of
 730                  * the local end point of the PPP link.
 731                  */
 732                 if (rtm->rtm_flags & RTF_ANNOUNCE) {
 733                         struct sockaddr laddr;
 734
 735                         if (rt->rt_ifp != NULL &&
 736                             rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
 737                                 struct ifaddr *ifa;
 738
 739                                 ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1);
 740                                 if (ifa != NULL)
 741                                         rt_maskedcopy(ifa->ifa_addr,
 742                                                       &laddr,
 743                                                       ifa->ifa_netmask);
 744                         } else
 745                                 rt_maskedcopy(rt->rt_ifa->ifa_addr,
 746                                               &laddr,
 747                                               rt->rt_ifa->ifa_netmask);
 748                         /*
 749                          * refactor rt and no lock operation necessary
 750                          */
 751                         rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr, rnh);
 752                         if (rt == NULL) {
 753                                 RADIX_NODE_HEAD_RUNLOCK(rnh);
 754                                 senderr(ESRCH);
 755                         }
 756                 }
 757                 RT_LOCK(rt);
 758                 RT_ADDREF(rt);
 759                 RADIX_NODE_HEAD_RUNLOCK(rnh);
 760
 761                 /*
 762                  * Fix for PR: 82974
 763                  *
 764                  * RTM_CHANGE/LOCK need a perfect match, rn_lookup()
 765                  * returns a perfect match in case a netmask is
 766                  * specified.  For host routes only a longest prefix
 767                  * match is returned so it is necessary to compare the
 768                  * existence of the netmask.  If both have a netmask
 769                  * rnh_lookup() did a perfect match and if none of them
 770                  * have a netmask both are host routes which is also a
 771                  * perfect match.
 772                  */
 773
 774                 if (rtm->rtm_type != RTM_GET &&
 775                     (!rt_mask(rt) != !info.rti_info[RTAX_NETMASK])) {
 776                         RT_UNLOCK(rt);
 777                         senderr(ESRCH);
 778                 }
 779
 780                 switch(rtm->rtm_type) {
 781
 782                 case RTM_GET:
 783                 report:
 784                         RT_LOCK_ASSERT(rt);
 785                         if ((rt->rt_flags & RTF_HOST) == 0
 786                             ? jailed_without_vnet(curthread->td_ucred)
 787                             : prison_if(curthread->td_ucred,
 788                             rt_key(rt)) != 0) {
 789                                 RT_UNLOCK(rt);
 790                                 senderr(ESRCH);
 791                         }
 792                         info.rti_info[RTAX_DST] = rt_key(rt);
 793                         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
 794                         info.rti_info[RTAX_NETMASK] = rt_mask(rt);
 795                         info.rti_info[RTAX_GENMASK] = 0;
 796                         if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
 797                                 ifp = rt->rt_ifp;
 798                                 if (ifp) {
 799                                         info.rti_info[RTAX_IFP] =
 800                                             ifp->if_addr->ifa_addr;
 801                                         error = rtm_get_jailed(&info, ifp, rt,
 802                                             &saun, curthread->td_ucred);
 803                                         if (error != 0) {
 804                                                 RT_UNLOCK(rt);
 805                                                 senderr(error);
 806                                         }
 807                                         if (ifp->if_flags & IFF_POINTOPOINT)
 808                                                 info.rti_info[RTAX_BRD] =
 809                                                     rt->rt_ifa->ifa_dstaddr;
 810                                         rtm->rtm_index = ifp->if_index;
 811                                 } else {
 812                                         info.rti_info[RTAX_IFP] = NULL;
 813                                         info.rti_info[RTAX_IFA] = NULL;
 814                                 }
 815                         } else if ((ifp = rt->rt_ifp) != NULL) {
 816                                 rtm->rtm_index = ifp->if_index;
 817                         }
 818                         len = rt_msg2(rtm->rtm_type, &info, NULL, NULL);
 819                         if (len > rtm->rtm_msglen) {
 820                                 struct rt_msghdr *new_rtm;
 821                                 R_Malloc(new_rtm, struct rt_msghdr *, len);
 822                                 if (new_rtm == NULL) {
 823                                         RT_UNLOCK(rt);
 824                                         senderr(ENOBUFS);
 825                                 }
 826                                 bcopy(rtm, new_rtm, rtm->rtm_msglen);
 827                                 Free(rtm); rtm = new_rtm;
 828                         }
 829                         (void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm, NULL);
 830                         rtm->rtm_flags = rt->rt_flags;
 831                         rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
 832                         rtm->rtm_addrs = info.rti_addrs;
 833                         break;
 834
 835                 case RTM_CHANGE:
 836                         /*
 837                          * New gateway could require new ifaddr, ifp;
 838                          * flags may also be different; ifp may be specified
 839                          * by ll sockaddr when protocol address is ambiguous
 840                          */
 841                         if (((rt->rt_flags & RTF_GATEWAY) &&
 842                              info.rti_info[RTAX_GATEWAY] != NULL) ||
 843                             info.rti_info[RTAX_IFP] != NULL ||
 844                             (info.rti_info[RTAX_IFA] != NULL &&
 845                              !sa_equal(info.rti_info[RTAX_IFA],
 846                                        rt->rt_ifa->ifa_addr))) {
 847                                 RT_UNLOCK(rt);
 848                                 RADIX_NODE_HEAD_LOCK(rnh);
 849                                 error = rt_getifa_fib(&info, rt->rt_fibnum);
 850                                 /*
 851                                  * XXXRW: Really we should release this
 852                                  * reference later, but this maintains
 853                                  * historical behavior.
 854                                  */
 855                                 if (info.rti_ifa != NULL)
 856                                         ifa_free(info.rti_ifa);
 857                                 RADIX_NODE_HEAD_UNLOCK(rnh);
 858                                 if (error != 0)
 859                                         senderr(error);
 860                                 RT_LOCK(rt);
 861                         }
 862                         if (info.rti_ifa != NULL &&
 863                             info.rti_ifa != rt->rt_ifa &&
 864                             rt->rt_ifa != NULL &&
 865                             rt->rt_ifa->ifa_rtrequest != NULL) {
 866                                 rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
 867                                     &info);
 868                                 ifa_free(rt->rt_ifa);
 869                         }
 870                         if (info.rti_info[RTAX_GATEWAY] != NULL) {
 871                                 RT_UNLOCK(rt);
 872                                 RADIX_NODE_HEAD_LOCK(rnh);
 873                                 RT_LOCK(rt);
 874
 875                                 error = rt_setgate(rt, rt_key(rt),
 876                                     info.rti_info[RTAX_GATEWAY]);
 877                                 RADIX_NODE_HEAD_UNLOCK(rnh);
 878                                 if (error != 0) {
 879                                         RT_UNLOCK(rt);
 880                                         senderr(error);
 881                                 }
 882                                 rt->rt_flags |= (RTF_GATEWAY & info.rti_flags);
 883                         }
 884                         if (info.rti_ifa != NULL &&
 885                             info.rti_ifa != rt->rt_ifa) {
 886                                 ifa_ref(info.rti_ifa);
 887                                 rt->rt_ifa = info.rti_ifa;
 888                                 rt->rt_ifp = info.rti_ifp;
 889                         }
 890                         /* Allow some flags to be toggled on change. */
 891                         rt->rt_flags = (rt->rt_flags & ~RTF_FMASK) |
 892                                     (rtm->rtm_flags & RTF_FMASK);
 893                         rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
 894                                         &rt->rt_rmx);
 895                         rtm->rtm_index = rt->rt_ifp->if_index;
 896                         if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
 897                                rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
 898                         /* FALLTHROUGH */
 899                 case RTM_LOCK:
 900                         /* We don't support locks anymore */
 901                         break;
 902                 }
 903                 RT_UNLOCK(rt);
 904                 break;
 905
 906         default:
 907                 senderr(EOPNOTSUPP);
 908         }
 909
 910 flush:
 911         if (rtm) {
 912                 if (error)
 913                         rtm->rtm_errno = error;
 914                 else
 915                         rtm->rtm_flags |= RTF_DONE;
 916         }
 917         if (rt)         /* XXX can this be true? */
 918                 RTFREE(rt);
 919     {
 920         struct rawcb *rp = NULL;
 921         /*
 922          * Check to see if we don't want our own messages.
 923          */
 924         if ((so->so_options & SO_USELOOPBACK) == 0) {
 925                 if (route_cb.any_count <= 1) {
 926                         if (rtm)
 927                                 Free(rtm);
 928                         m_freem(m);
 929                         return (error);
 930                 }
 931                 /* There is another listener, so construct message */
 932                 rp = sotorawcb(so);
 933         }
 934         if (rtm) {
 935                 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
 936                 if (m->m_pkthdr.len < rtm->rtm_msglen) {
 937                         m_freem(m);
 938                         m = NULL;
 939                 } else if (m->m_pkthdr.len > rtm->rtm_msglen)
 940                         m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
 941         }
 942         if (m) {
 943                 M_SETFIB(m, so->so_fibnum);
 944                 m->m_flags |= RTS_FILTER_FIB;
 945                 if (rp) {
 946                         /*
 947                          * XXX insure we don't get a copy by
 948                          * invalidating our protocol
 949                          */
 950                         unsigned short family = rp->rcb_proto.sp_family;
 951                         rp->rcb_proto.sp_family = 0;
 952                         rt_dispatch(m, saf);
 953                         rp->rcb_proto.sp_family = family;
 954                 } else
 955                         rt_dispatch(m, saf);
 956         }
 957         /* info.rti_info[RTAX_DST] (used above) can point inside of rtm */
 958         if (rtm)
 959                 Free(rtm);
 960     }
 961         return (error);
 962 #undef  sa_equal
 963 }
 964
 965 static void
 966 rt_setmetrics(u_long which, const struct rt_metrics *in,
 967         struct rt_metrics_lite *out)
 968 {
 969 #define metric(f, e) if (which & (f)) out->e = in->e;
 970         /*
 971          * Only these are stored in the routing entry since introduction
 972          * of tcp hostcache. The rest is ignored.
 973          */
 974         metric(RTV_MTU, rmx_mtu);
 975         metric(RTV_WEIGHT, rmx_weight);
 976         /* Userland -> kernel timebase conversion. */
 977         if (which & RTV_EXPIRE)
 978                 out->rmx_expire = in->rmx_expire ?
 979                     in->rmx_expire - time_second + time_uptime : 0;
 980 #undef metric
 981 }
 982
 983 static void
 984 rt_getmetrics(const struct rt_metrics_lite *in, struct rt_metrics *out)
 985 {
 986 #define metric(e) out->e = in->e;
 987         bzero(out, sizeof(*out));
 988         metric(rmx_mtu);
 989         metric(rmx_weight);
 990         /* Kernel -> userland timebase conversion. */
 991         out->rmx_expire = in->rmx_expire ?
 992             in->rmx_expire - time_uptime + time_second : 0;
 993 #undef metric
 994 }
 995
 996 /*
 997  * Extract the addresses of the passed sockaddrs.
 998  * Do a little sanity checking so as to avoid bad memory references.
 999  * This data is derived straight from userland.
1000  */
1001 static int
1002 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1003 {
1004         struct sockaddr *sa;
1005         int i;
1006
1007         for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1008                 if ((rtinfo->rti_addrs & (1 << i)) == 0)
1009                         continue;
1010                 sa = (struct sockaddr *)cp;
1011                 /*
1012                  * It won't fit.
1013                  */
1014                 if (cp + sa->sa_len > cplim)
1015                         return (EINVAL);
1016                 /*
1017                  * there are no more.. quit now
1018                  * If there are more bits, they are in error.
1019                  * I've seen this. route(1) can evidently generate these.
1020                  * This causes kernel to core dump.
1021                  * for compatibility, If we see this, point to a safe address.
1022                  */
1023                 if (sa->sa_len == 0) {
1024                         rtinfo->rti_info[i] = &sa_zero;
1025                         return (0); /* should be EINVAL but for compat */
1026                 }
1027                 /* accept it */
1028                 rtinfo->rti_info[i] = sa;
1029                 cp += SA_SIZE(sa);
1030         }
1031         return (0);
1032 }
1033
1034 /*
1035  * Used by the routing socket.
1036  */
1037 static struct mbuf *
1038 rt_msg1(int type, struct rt_addrinfo *rtinfo)
1039 {
1040         struct rt_msghdr *rtm;
1041         struct mbuf *m;
1042         int i;
1043         struct sockaddr *sa;
1044         int len, dlen;
1045
1046         switch (type) {
1047
1048         case RTM_DELADDR:
1049         case RTM_NEWADDR:
1050                 len = sizeof(struct ifa_msghdr);
1051                 break;
1052
1053         case RTM_DELMADDR:
1054         case RTM_NEWMADDR:
1055                 len = sizeof(struct ifma_msghdr);
1056                 break;
1057
1058         case RTM_IFINFO:
1059                 len = sizeof(struct if_msghdr);
1060                 break;
1061
1062         case RTM_IFANNOUNCE:
1063         case RTM_IEEE80211:
1064                 len = sizeof(struct if_announcemsghdr);
1065                 break;
1066
1067         default:
1068                 len = sizeof(struct rt_msghdr);
1069         }
1070         if (len > MCLBYTES)
1071                 panic("rt_msg1");
1072         m = m_gethdr(M_DONTWAIT, MT_DATA);
1073         if (m && len > MHLEN) {
1074                 MCLGET(m, M_DONTWAIT);
1075                 if ((m->m_flags & M_EXT) == 0) {
1076                         m_free(m);
1077                         m = NULL;
1078                 }
1079         }
1080         if (m == NULL)
1081                 return (m);
1082         m->m_pkthdr.len = m->m_len = len;
1083         m->m_pkthdr.rcvif = NULL;
1084         rtm = mtod(m, struct rt_msghdr *);
1085         bzero((caddr_t)rtm, len);
1086         for (i = 0; i < RTAX_MAX; i++) {
1087                 if ((sa = rtinfo->rti_info[i]) == NULL)
1088                         continue;
1089                 rtinfo->rti_addrs |= (1 << i);
1090                 dlen = SA_SIZE(sa);
1091                 m_copyback(m, len, dlen, (caddr_t)sa);
1092                 len += dlen;
1093         }
1094         if (m->m_pkthdr.len != len) {
1095                 m_freem(m);
1096                 return (NULL);
1097         }
1098         rtm->rtm_msglen = len;
1099         rtm->rtm_version = RTM_VERSION;
1100         rtm->rtm_type = type;
1101         return (m);
1102 }
1103
1104 /*
1105  * Used by the sysctl code and routing socket.
1106  */
1107 static int
1108 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
1109 {
1110         int i;
1111         int len, dlen, second_time = 0;
1112         caddr_t cp0;
1113
1114         rtinfo->rti_addrs = 0;
1115 again:
1116         switch (type) {
1117
1118         case RTM_DELADDR:
1119         case RTM_NEWADDR:
1120                 if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1121 #ifdef COMPAT_FREEBSD32
1122                         if (w->w_req->flags & SCTL_MASK32)
1123                                 len = sizeof(struct ifa_msghdrl32);
1124                         else
1125 #endif
1126                                 len = sizeof(struct ifa_msghdrl);
1127                 } else
1128                         len = sizeof(struct ifa_msghdr);
1129                 break;
1130
1131         case RTM_IFINFO:
1132 #ifdef COMPAT_FREEBSD32
1133                 if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1134                         if (w->w_op == NET_RT_IFLISTL)
1135                                 len = sizeof(struct if_msghdrl32);
1136                         else
1137                                 len = sizeof(struct if_msghdr32);
1138                         break;
1139                 }
1140 #endif
1141                 if (w != NULL && w->w_op == NET_RT_IFLISTL)
1142                         len = sizeof(struct if_msghdrl);
1143                 else
1144                         len = sizeof(struct if_msghdr);
1145                 break;
1146
1147         case RTM_NEWMADDR:
1148                 len = sizeof(struct ifma_msghdr);
1149                 break;
1150
1151         default:
1152                 len = sizeof(struct rt_msghdr);
1153         }
1154         cp0 = cp;
1155         if (cp0)
1156                 cp += len;
1157         for (i = 0; i < RTAX_MAX; i++) {
1158                 struct sockaddr *sa;
1159
1160                 if ((sa = rtinfo->rti_info[i]) == NULL)
1161                         continue;
1162                 rtinfo->rti_addrs |= (1 << i);
1163                 dlen = SA_SIZE(sa);
1164                 if (cp) {
1165                         bcopy((caddr_t)sa, cp, (unsigned)dlen);
1166                         cp += dlen;
1167                 }
1168                 len += dlen;
1169         }
1170         len = ALIGN(len);
1171         if (cp == NULL && w != NULL && !second_time) {
1172                 struct walkarg *rw = w;
1173
1174                 if (rw->w_req) {
1175                         if (rw->w_tmemsize < len) {
1176                                 if (rw->w_tmem)
1177                                         free(rw->w_tmem, M_RTABLE);
1178                                 rw->w_tmem = (caddr_t)
1179                                         malloc(len, M_RTABLE, M_NOWAIT);
1180                                 if (rw->w_tmem)
1181                                         rw->w_tmemsize = len;
1182                         }
1183                         if (rw->w_tmem) {
1184                                 cp = rw->w_tmem;
1185                                 second_time = 1;
1186                                 goto again;
1187                         }
1188                 }
1189         }
1190         if (cp) {
1191                 struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
1192
1193                 rtm->rtm_version = RTM_VERSION;
1194                 rtm->rtm_type = type;
1195                 rtm->rtm_msglen = len;
1196         }
1197         return (len);
1198 }
1199
1200 /*
1201  * This routine is called to generate a message from the routing
1202  * socket indicating that a redirect has occured, a routing lookup
1203  * has failed, or that a protocol has detected timeouts to a particular
1204  * destination.
1205  */
1206 void
1207 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1208     int fibnum)
1209 {
1210         struct rt_msghdr *rtm;
1211         struct mbuf *m;
1212         struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1213
1214         if (route_cb.any_count == 0)
1215                 return;
1216         m = rt_msg1(type, rtinfo);
1217         if (m == NULL)
1218                 return;
1219
1220         if (fibnum != RTS_ALLFIBS) {
1221                 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1222                     "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1223                 M_SETFIB(m, fibnum);
1224                 m->m_flags |= RTS_FILTER_FIB;
1225         }
1226
1227         rtm = mtod(m, struct rt_msghdr *);
1228         rtm->rtm_flags = RTF_DONE | flags;
1229         rtm->rtm_errno = error;
1230         rtm->rtm_addrs = rtinfo->rti_addrs;
1231         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1232 }
1233
1234 void
1235 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1236 {
1237
1238         rt_missmsg_fib(type, rtinfo, flags, error, RTS_ALLFIBS);
1239 }
1240
1241 /*
1242  * This routine is called to generate a message from the routing
1243  * socket indicating that the status of a network interface has changed.
1244  */
1245 void
1246 rt_ifmsg(struct ifnet *ifp)
1247 {
1248         struct if_msghdr *ifm;
1249         struct mbuf *m;
1250         struct rt_addrinfo info;
1251
1252         if (route_cb.any_count == 0)
1253                 return;
1254         bzero((caddr_t)&info, sizeof(info));
1255         m = rt_msg1(RTM_IFINFO, &info);
1256         if (m == NULL)
1257                 return;
1258         ifm = mtod(m, struct if_msghdr *);
1259         ifm->ifm_index = ifp->if_index;
1260         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1261         ifm->ifm_data = ifp->if_data;
1262         ifm->ifm_addrs = 0;
1263         rt_dispatch(m, AF_UNSPEC);
1264 }
1265
1266 /*
1267  * This is called to generate messages from the routing socket
1268  * indicating a network interface has had addresses associated with it.
1269  * if we ever reverse the logic and replace messages TO the routing
1270  * socket indicate a request to configure interfaces, then it will
1271  * be unnecessary as the routing socket will automatically generate
1272  * copies of it.
1273  */
1274 void
1275 rt_newaddrmsg_fib(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt,
1276     int fibnum)
1277 {
1278         struct rt_addrinfo info;
1279         struct sockaddr *sa = NULL;
1280         int pass;
1281         struct mbuf *m = NULL;
1282         struct ifnet *ifp = ifa->ifa_ifp;
1283
1284         KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
1285                 ("unexpected cmd %u", cmd));
1286 #if defined(INET) || defined(INET6)
1287 #ifdef SCTP
1288         /*
1289          * notify the SCTP stack
1290          * this will only get called when an address is added/deleted
1291          * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1292          */
1293         sctp_addr_change(ifa, cmd);
1294 #endif /* SCTP */
1295 #endif
1296         if (route_cb.any_count == 0)
1297                 return;
1298         for (pass = 1; pass < 3; pass++) {
1299                 bzero((caddr_t)&info, sizeof(info));
1300                 if ((cmd == RTM_ADD && pass == 1) ||
1301                     (cmd == RTM_DELETE && pass == 2)) {
1302                         struct ifa_msghdr *ifam;
1303                         int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1304
1305                         info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1306                         info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1307                         info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1308                         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1309                         if ((m = rt_msg1(ncmd, &info)) == NULL)
1310                                 continue;
1311                         ifam = mtod(m, struct ifa_msghdr *);
1312                         ifam->ifam_index = ifp->if_index;
1313                         ifam->ifam_metric = ifa->ifa_metric;
1314                         ifam->ifam_flags = ifa->ifa_flags;
1315                         ifam->ifam_addrs = info.rti_addrs;
1316                 }
1317                 if ((cmd == RTM_ADD && pass == 2) ||
1318                     (cmd == RTM_DELETE && pass == 1)) {
1319                         struct rt_msghdr *rtm;
1320
1321                         if (rt == NULL)
1322                                 continue;
1323                         info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1324                         info.rti_info[RTAX_DST] = sa = rt_key(rt);
1325                         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1326                         if ((m = rt_msg1(cmd, &info)) == NULL)
1327                                 continue;
1328                         rtm = mtod(m, struct rt_msghdr *);
1329                         rtm->rtm_index = ifp->if_index;
1330                         rtm->rtm_flags |= rt->rt_flags;
1331                         rtm->rtm_errno = error;
1332                         rtm->rtm_addrs = info.rti_addrs;
1333                 }
1334                 if (fibnum != RTS_ALLFIBS) {
1335                         KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: "
1336                             "fibnum out of range 0 <= %d < %d", __func__,
1337                              fibnum, rt_numfibs));
1338                         M_SETFIB(m, fibnum);
1339                         m->m_flags |= RTS_FILTER_FIB;
1340                 }
1341                 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1342         }
1343 }
1344
1345 void
1346 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1347 {
1348
1349         rt_newaddrmsg_fib(cmd, ifa, error, rt, RTS_ALLFIBS);
1350 }
1351
1352 /*
1353  * This is the analogue to the rt_newaddrmsg which performs the same
1354  * function but for multicast group memberhips.  This is easier since
1355  * there is no route state to worry about.
1356  */
1357 void
1358 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1359 {
1360         struct rt_addrinfo info;
1361         struct mbuf *m = NULL;
1362         struct ifnet *ifp = ifma->ifma_ifp;
1363         struct ifma_msghdr *ifmam;
1364
1365         if (route_cb.any_count == 0)
1366                 return;
1367
1368         bzero((caddr_t)&info, sizeof(info));
1369         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1370         info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1371         /*
1372          * If a link-layer address is present, present it as a ``gateway''
1373          * (similarly to how ARP entries, e.g., are presented).
1374          */
1375         info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1376         m = rt_msg1(cmd, &info);
1377         if (m == NULL)
1378                 return;
1379         ifmam = mtod(m, struct ifma_msghdr *);
1380         KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1381             __func__));
1382         ifmam->ifmam_index = ifp->if_index;
1383         ifmam->ifmam_addrs = info.rti_addrs;
1384         rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1385 }
1386
1387 static struct mbuf *
1388 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1389         struct rt_addrinfo *info)
1390 {
1391         struct if_announcemsghdr *ifan;
1392         struct mbuf *m;
1393
1394         if (route_cb.any_count == 0)
1395                 return NULL;
1396         bzero((caddr_t)info, sizeof(*info));
1397         m = rt_msg1(type, info);
1398         if (m != NULL) {
1399                 ifan = mtod(m, struct if_announcemsghdr *);
1400                 ifan->ifan_index = ifp->if_index;
1401                 strlcpy(ifan->ifan_name, ifp->if_xname,
1402                         sizeof(ifan->ifan_name));
1403                 ifan->ifan_what = what;
1404         }
1405         return m;
1406 }
1407
1408 /*
1409  * This is called to generate routing socket messages indicating
1410  * IEEE80211 wireless events.
1411  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1412  */
1413 void
1414 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1415 {
1416         struct mbuf *m;
1417         struct rt_addrinfo info;
1418
1419         m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1420         if (m != NULL) {
1421                 /*
1422                  * Append the ieee80211 data.  Try to stick it in the
1423                  * mbuf containing the ifannounce msg; otherwise allocate
1424                  * a new mbuf and append.
1425                  *
1426                  * NB: we assume m is a single mbuf.
1427                  */
1428                 if (data_len > M_TRAILINGSPACE(m)) {
1429                         struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1430                         if (n == NULL) {
1431                                 m_freem(m);
1432                                 return;
1433                         }
1434                         bcopy(data, mtod(n, void *), data_len);
1435                         n->m_len = data_len;
1436                         m->m_next = n;
1437                 } else if (data_len > 0) {
1438                         bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1439                         m->m_len += data_len;
1440                 }
1441                 if (m->m_flags & M_PKTHDR)
1442                         m->m_pkthdr.len += data_len;
1443                 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1444                 rt_dispatch(m, AF_UNSPEC);
1445         }
1446 }
1447
1448 /*
1449  * This is called to generate routing socket messages indicating
1450  * network interface arrival and departure.
1451  */
1452 void
1453 rt_ifannouncemsg(struct ifnet *ifp, int what)
1454 {
1455         struct mbuf *m;
1456         struct rt_addrinfo info;
1457
1458         m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1459         if (m != NULL)
1460                 rt_dispatch(m, AF_UNSPEC);
1461 }
1462
1463 static void
1464 rt_dispatch(struct mbuf *m, sa_family_t saf)
1465 {
1466         struct m_tag *tag;
1467
1468         /*
1469          * Preserve the family from the sockaddr, if any, in an m_tag for
1470          * use when injecting the mbuf into the routing socket buffer from
1471          * the netisr.
1472          */
1473         if (saf != AF_UNSPEC) {
1474                 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1475                     M_NOWAIT);
1476                 if (tag == NULL) {
1477                         m_freem(m);
1478                         return;
1479                 }
1480                 *(unsigned short *)(tag + 1) = saf;
1481                 m_tag_prepend(m, tag);
1482         }
1483 #ifdef VIMAGE
1484         if (V_loif)
1485                 m->m_pkthdr.rcvif = V_loif;
1486         else {
1487                 m_freem(m);
1488                 return;
1489         }
1490 #endif
1491         netisr_queue(NETISR_ROUTE, m);  /* mbuf is free'd on failure. */
1492 }
1493
1494 /*
1495  * This is used in dumping the kernel table via sysctl().
1496  */
1497 static int
1498 sysctl_dumpentry(struct radix_node *rn, void *vw)
1499 {
1500         struct walkarg *w = vw;
1501         struct rtentry *rt = (struct rtentry *)rn;
1502         int error = 0, size;
1503         struct rt_addrinfo info;
1504
1505         if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1506                 return 0;
1507         if ((rt->rt_flags & RTF_HOST) == 0
1508             ? jailed_without_vnet(w->w_req->td->td_ucred)
1509             : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1510                 return (0);
1511         bzero((caddr_t)&info, sizeof(info));
1512         info.rti_info[RTAX_DST] = rt_key(rt);
1513         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1514         info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1515         info.rti_info[RTAX_GENMASK] = 0;
1516         if (rt->rt_ifp) {
1517                 info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1518                 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1519                 if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1520                         info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1521         }
1522         size = rt_msg2(RTM_GET, &info, NULL, w);
1523         if (w->w_req && w->w_tmem) {
1524                 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1525
1526                 rtm->rtm_flags = rt->rt_flags;
1527                 /*
1528                  * let's be honest about this being a retarded hack
1529                  */
1530                 rtm->rtm_fmask = rt->rt_rmx.rmx_pksent;
1531                 rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1532                 rtm->rtm_index = rt->rt_ifp->if_index;
1533                 rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1534                 rtm->rtm_addrs = info.rti_addrs;
1535                 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1536                 return (error);
1537         }
1538         return (error);
1539 }
1540
1541 #ifdef COMPAT_FREEBSD32
1542 static void
1543 copy_ifdata32(struct if_data *src, struct if_data32 *dst)
1544 {
1545
1546         bzero(dst, sizeof(*dst));
1547         CP(*src, *dst, ifi_type);
1548         CP(*src, *dst, ifi_physical);
1549         CP(*src, *dst, ifi_addrlen);
1550         CP(*src, *dst, ifi_hdrlen);
1551         CP(*src, *dst, ifi_link_state);
1552         CP(*src, *dst, ifi_vhid);
1553         dst->ifi_datalen = sizeof(struct if_data32);
1554         CP(*src, *dst, ifi_mtu);
1555         CP(*src, *dst, ifi_metric);
1556         CP(*src, *dst, ifi_baudrate);
1557         CP(*src, *dst, ifi_ipackets);
1558         CP(*src, *dst, ifi_ierrors);
1559         CP(*src, *dst, ifi_opackets);
1560         CP(*src, *dst, ifi_oerrors);
1561         CP(*src, *dst, ifi_collisions);
1562         CP(*src, *dst, ifi_ibytes);
1563         CP(*src, *dst, ifi_obytes);
1564         CP(*src, *dst, ifi_imcasts);
1565         CP(*src, *dst, ifi_omcasts);
1566         CP(*src, *dst, ifi_iqdrops);
1567         CP(*src, *dst, ifi_noproto);
1568         CP(*src, *dst, ifi_hwassist);
1569         CP(*src, *dst, ifi_epoch);
1570         TV_CP(*src, *dst, ifi_lastchange);
1571 }
1572 #endif
1573
1574 static int
1575 sysctl_iflist_ifml(struct ifnet *ifp, struct rt_addrinfo *info,
1576     struct walkarg *w, int len)
1577 {
1578         struct if_msghdrl *ifm;
1579
1580 #ifdef COMPAT_FREEBSD32
1581         if (w->w_req->flags & SCTL_MASK32) {
1582                 struct if_msghdrl32 *ifm32;
1583
1584                 ifm32 = (struct if_msghdrl32 *)w->w_tmem;
1585                 ifm32->ifm_addrs = info->rti_addrs;
1586                 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1587                 ifm32->ifm_index = ifp->if_index;
1588                 ifm32->_ifm_spare1 = 0;
1589                 ifm32->ifm_len = sizeof(*ifm32);
1590                 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1591
1592                 copy_ifdata32(&ifp->if_data, &ifm32->ifm_data);
1593                 /* Fixup if_data carp(4) vhid. */
1594                 if (carp_get_vhid_p != NULL)
1595                         ifm32->ifm_data.ifi_vhid =
1596                             (*carp_get_vhid_p)(ifp->if_addr);
1597
1598                 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm32, len));
1599         }
1600 #endif
1601         ifm = (struct if_msghdrl *)w->w_tmem;
1602         ifm->ifm_addrs = info->rti_addrs;
1603         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1604         ifm->ifm_index = ifp->if_index;
1605         ifm->_ifm_spare1 = 0;
1606         ifm->ifm_len = sizeof(*ifm);
1607         ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1608
1609         ifm->ifm_data = ifp->if_data;
1610         /* Fixup if_data carp(4) vhid. */
1611         if (carp_get_vhid_p != NULL)
1612                 ifm->ifm_data.ifi_vhid = (*carp_get_vhid_p)(ifp->if_addr);
1613
1614         return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1615 }
1616
1617 static int
1618 sysctl_iflist_ifm(struct ifnet *ifp, struct rt_addrinfo *info,
1619     struct walkarg *w, int len)
1620 {
1621         struct if_msghdr *ifm;
1622
1623 #ifdef COMPAT_FREEBSD32
1624         if (w->w_req->flags & SCTL_MASK32) {
1625                 struct if_msghdr32 *ifm32;
1626
1627                 ifm32 = (struct if_msghdr32 *)w->w_tmem;
1628                 ifm32->ifm_addrs = info->rti_addrs;
1629                 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1630                 ifm32->ifm_index = ifp->if_index;
1631
1632                 copy_ifdata32(&ifp->if_data, &ifm32->ifm_data);
1633                 /* Fixup if_data carp(4) vhid. */
1634                 if (carp_get_vhid_p != NULL)
1635                         ifm32->ifm_data.ifi_vhid =
1636                             (*carp_get_vhid_p)(ifp->if_addr);
1637
1638                 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm32, len));
1639         }
1640 #endif
1641         ifm = (struct if_msghdr *)w->w_tmem;
1642         ifm->ifm_addrs = info->rti_addrs;
1643         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1644         ifm->ifm_index = ifp->if_index;
1645
1646         ifm->ifm_data = ifp->if_data;
1647         /* Fixup if_data carp(4) vhid. */
1648         if (carp_get_vhid_p != NULL)
1649                 ifm->ifm_data.ifi_vhid = (*carp_get_vhid_p)(ifp->if_addr);
1650
1651         return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1652 }
1653
1654 static int
1655 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1656     struct walkarg *w, int len)
1657 {
1658         struct ifa_msghdrl *ifam;
1659
1660 #ifdef COMPAT_FREEBSD32
1661         if (w->w_req->flags & SCTL_MASK32) {
1662                 struct ifa_msghdrl32 *ifam32;
1663
1664                 ifam32 = (struct ifa_msghdrl32 *)w->w_tmem;
1665                 ifam32->ifam_addrs = info->rti_addrs;
1666                 ifam32->ifam_flags = ifa->ifa_flags;
1667                 ifam32->ifam_index = ifa->ifa_ifp->if_index;
1668                 ifam32->_ifam_spare1 = 0;
1669                 ifam32->ifam_len = sizeof(*ifam32);
1670                 ifam32->ifam_data_off =
1671                     offsetof(struct ifa_msghdrl32, ifam_data);
1672                 ifam32->ifam_metric = ifa->ifa_metric;
1673
1674                 copy_ifdata32(&ifa->ifa_ifp->if_data, &ifam32->ifam_data);
1675                 /* Fixup if_data carp(4) vhid. */
1676                 if (carp_get_vhid_p != NULL)
1677                         ifam32->ifam_data.ifi_vhid = (*carp_get_vhid_p)(ifa);
1678
1679                 return (SYSCTL_OUT(w->w_req, (caddr_t)ifam32, len));
1680         }
1681 #endif
1682
1683         ifam = (struct ifa_msghdrl *)w->w_tmem;
1684         ifam->ifam_addrs = info->rti_addrs;
1685         ifam->ifam_flags = ifa->ifa_flags;
1686         ifam->ifam_index = ifa->ifa_ifp->if_index;
1687         ifam->_ifam_spare1 = 0;
1688         ifam->ifam_len = sizeof(*ifam);
1689         ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1690         ifam->ifam_metric = ifa->ifa_metric;
1691
1692         ifam->ifam_data = ifa->if_data;
1693         /* Fixup if_data carp(4) vhid. */
1694         if (carp_get_vhid_p != NULL)
1695                 ifam->ifam_data.ifi_vhid = (*carp_get_vhid_p)(ifa);
1696
1697         return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1698 }
1699
1700 static int
1701 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1702     struct walkarg *w, int len)
1703 {
1704         struct ifa_msghdr *ifam;
1705
1706         ifam = (struct ifa_msghdr *)w->w_tmem;
1707         ifam->ifam_addrs = info->rti_addrs;
1708         ifam->ifam_flags = ifa->ifa_flags;
1709         ifam->ifam_index = ifa->ifa_ifp->if_index;
1710         ifam->ifam_metric = ifa->ifa_metric;
1711
1712         return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1713 }
1714
1715 static int
1716 sysctl_iflist(int af, struct walkarg *w)
1717 {
1718         struct ifnet *ifp;
1719         struct ifaddr *ifa;
1720         struct rt_addrinfo info;
1721         int len, error = 0;
1722
1723         bzero((caddr_t)&info, sizeof(info));
1724         IFNET_RLOCK();
1725         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1726                 if (w->w_arg && w->w_arg != ifp->if_index)
1727                         continue;
1728                 IF_ADDR_RLOCK(ifp);
1729                 ifa = ifp->if_addr;
1730                 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1731                 len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1732                 info.rti_info[RTAX_IFP] = NULL;
1733                 if (w->w_req && w->w_tmem) {
1734                         if (w->w_op == NET_RT_IFLISTL)
1735                                 error = sysctl_iflist_ifml(ifp, &info, w, len);
1736                         else
1737                                 error = sysctl_iflist_ifm(ifp, &info, w, len);
1738                         if (error)
1739                                 goto done;
1740                 }
1741                 while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1742                         if (af && af != ifa->ifa_addr->sa_family)
1743                                 continue;
1744                         if (prison_if(w->w_req->td->td_ucred,
1745                             ifa->ifa_addr) != 0)
1746                                 continue;
1747                         info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1748                         info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1749                         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1750                         len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1751                         if (w->w_req && w->w_tmem) {
1752                                 if (w->w_op == NET_RT_IFLISTL)
1753                                         error = sysctl_iflist_ifaml(ifa, &info,
1754                                             w, len);
1755                                 else
1756                                         error = sysctl_iflist_ifam(ifa, &info,
1757                                             w, len);
1758                                 if (error)
1759                                         goto done;
1760                         }
1761                 }
1762                 IF_ADDR_RUNLOCK(ifp);
1763                 info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1764                         info.rti_info[RTAX_BRD] = NULL;
1765         }
1766 done:
1767         if (ifp != NULL)
1768                 IF_ADDR_RUNLOCK(ifp);
1769         IFNET_RUNLOCK();
1770         return (error);
1771 }
1772
1773 static int
1774 sysctl_ifmalist(int af, struct walkarg *w)
1775 {
1776         struct ifnet *ifp;
1777         struct ifmultiaddr *ifma;
1778         struct  rt_addrinfo info;
1779         int     len, error = 0;
1780         struct ifaddr *ifa;
1781
1782         bzero((caddr_t)&info, sizeof(info));
1783         IFNET_RLOCK();
1784         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1785                 if (w->w_arg && w->w_arg != ifp->if_index)
1786                         continue;
1787                 ifa = ifp->if_addr;
1788                 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1789                 IF_ADDR_RLOCK(ifp);
1790                 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1791                         if (af && af != ifma->ifma_addr->sa_family)
1792                                 continue;
1793                         if (prison_if(w->w_req->td->td_ucred,
1794                             ifma->ifma_addr) != 0)
1795                                 continue;
1796                         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1797                         info.rti_info[RTAX_GATEWAY] =
1798                             (ifma->ifma_addr->sa_family != AF_LINK) ?
1799                             ifma->ifma_lladdr : NULL;
1800                         len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1801                         if (w->w_req && w->w_tmem) {
1802                                 struct ifma_msghdr *ifmam;
1803
1804                                 ifmam = (struct ifma_msghdr *)w->w_tmem;
1805                                 ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1806                                 ifmam->ifmam_flags = 0;
1807                                 ifmam->ifmam_addrs = info.rti_addrs;
1808                                 error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1809                                 if (error) {
1810                                         IF_ADDR_RUNLOCK(ifp);
1811                                         goto done;
1812                                 }
1813                         }
1814                 }
1815                 IF_ADDR_RUNLOCK(ifp);
1816         }
1817 done:
1818         IFNET_RUNLOCK();
1819         return (error);
1820 }
1821
1822 static int
1823 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1824 {
1825         int     *name = (int *)arg1;
1826         u_int   namelen = arg2;
1827         struct radix_node_head *rnh = NULL; /* silence compiler. */
1828         int     i, lim, error = EINVAL;
1829         u_char  af;
1830         struct  walkarg w;
1831
1832         name ++;
1833         namelen--;
1834         if (req->newptr)
1835                 return (EPERM);
1836         if (namelen != 3)
1837                 return ((namelen < 3) ? EISDIR : ENOTDIR);
1838         af = name[0];
1839         if (af > AF_MAX)
1840                 return (EINVAL);
1841         bzero(&w, sizeof(w));
1842         w.w_op = name[1];
1843         w.w_arg = name[2];
1844         w.w_req = req;
1845
1846         error = sysctl_wire_old_buffer(req, 0);
1847         if (error)
1848                 return (error);
1849         switch (w.w_op) {
1850
1851         case NET_RT_DUMP:
1852         case NET_RT_FLAGS:
1853                 if (af == 0) {                  /* dump all tables */
1854                         i = 1;
1855                         lim = AF_MAX;
1856                 } else                          /* dump only one table */
1857                         i = lim = af;
1858
1859                 /*
1860                  * take care of llinfo entries, the caller must
1861                  * specify an AF
1862                  */
1863                 if (w.w_op == NET_RT_FLAGS &&
1864                     (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1865                         if (af != 0)
1866                                 error = lltable_sysctl_dumparp(af, w.w_req);
1867                         else
1868                                 error = EINVAL;
1869                         break;
1870                 }
1871                 /*
1872                  * take care of routing entries
1873                  */
1874                 for (error = 0; error == 0 && i <= lim; i++) {
1875                         rnh = rt_tables_get_rnh(req->td->td_proc->p_fibnum, i);
1876                         if (rnh != NULL) {
1877                                 RADIX_NODE_HEAD_RLOCK(rnh);
1878                                 error = rnh->rnh_walktree(rnh,
1879                                     sysctl_dumpentry, &w);
1880                                 RADIX_NODE_HEAD_RUNLOCK(rnh);
1881                         } else if (af != 0)
1882                                 error = EAFNOSUPPORT;
1883                 }
1884                 break;
1885
1886         case NET_RT_IFLIST:
1887         case NET_RT_IFLISTL:
1888                 error = sysctl_iflist(af, &w);
1889                 break;
1890
1891         case NET_RT_IFMALIST:
1892                 error = sysctl_ifmalist(af, &w);
1893                 break;
1894         }
1895         if (w.w_tmem)
1896                 free(w.w_tmem, M_RTABLE);
1897         return (error);
1898 }
1899
1900 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1901
1902 /*
1903  * Definitions of protocols supported in the ROUTE domain.
1904  */
1905
1906 static struct domain routedomain;               /* or at least forward */
1907
1908 static struct protosw routesw[] = {
1909 {
1910         .pr_type =              SOCK_RAW,
1911         .pr_domain =            &routedomain,
1912         .pr_flags =             PR_ATOMIC|PR_ADDR,
1913         .pr_output =            route_output,
1914         .pr_ctlinput =          raw_ctlinput,
1915         .pr_init =              raw_init,
1916         .pr_usrreqs =           &route_usrreqs
1917 }
1918 };
1919
1920 static struct domain routedomain = {
1921         .dom_family =           PF_ROUTE,
1922         .dom_name =              "route",
1923         .dom_protosw =          routesw,
1924         .dom_protoswNPROTOSW =  &routesw[sizeof(routesw)/sizeof(routesw[0])]
1925 };
1926
1927 VNET_DOMAIN_SET(route);