sys/net/rtsock.c

   1 /*-
   2  * SPDX-License-Identifier: BSD-3-Clause
   3  *
   4  * Copyright (c) 1988, 1991, 1993
   5  *      The Regents of the University of California.  All rights reserved.
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  10  * 1. Redistributions of source code must retain the above copyright
  11  *    notice, this list of conditions and the following disclaimer.
  12  * 2. Redistributions in binary form must reproduce the above copyright
  13  *    notice, this list of conditions and the following disclaimer in the
  14  *    documentation and/or other materials provided with the distribution.
  15  * 3. Neither the name of the University nor the names of its contributors
  16  *    may be used to endorse or promote products derived from this software
  17  *    without specific prior written permission.
  18  *
  19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  29  * SUCH DAMAGE.
  30  *
  31  *      @(#)rtsock.c    8.7 (Berkeley) 10/12/95
  32  * $FreeBSD$
  33  */
  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/rmlock.h>
  49 #include <sys/rwlock.h>
  50 #include <sys/signalvar.h>
  51 #include <sys/socket.h>
  52 #include <sys/socketvar.h>
  53 #include <sys/sysctl.h>
  54 #include <sys/systm.h>
  55
  56 #include <net/if.h>
  57 #include <net/if_var.h>
  58 #include <net/if_dl.h>
  59 #include <net/if_llatbl.h>
  60 #include <net/if_types.h>
  61 #include <net/netisr.h>
  62 #include <net/raw_cb.h>
  63 #include <net/route.h>
  64 #include <net/route_var.h>
  65 #include <net/vnet.h>
  66
  67 #include <netinet/in.h>
  68 #include <netinet/if_ether.h>
  69 #include <netinet/ip_carp.h>
  70 #ifdef INET6
  71 #include <netinet6/ip6_var.h>
  72 #include <netinet6/scope6_var.h>
  73 #endif
  74
  75 #ifdef COMPAT_FREEBSD32
  76 #include <sys/mount.h>
  77 #include <compat/freebsd32/freebsd32.h>
  78
  79 struct if_msghdr32 {
  80         uint16_t ifm_msglen;
  81         uint8_t ifm_version;
  82         uint8_t ifm_type;
  83         int32_t ifm_addrs;
  84         int32_t ifm_flags;
  85         uint16_t ifm_index;
  86         uint16_t _ifm_spare1;
  87         struct  if_data ifm_data;
  88 };
  89
  90 struct if_msghdrl32 {
  91         uint16_t ifm_msglen;
  92         uint8_t ifm_version;
  93         uint8_t ifm_type;
  94         int32_t ifm_addrs;
  95         int32_t ifm_flags;
  96         uint16_t ifm_index;
  97         uint16_t _ifm_spare1;
  98         uint16_t ifm_len;
  99         uint16_t ifm_data_off;
 100         uint32_t _ifm_spare2;
 101         struct  if_data ifm_data;
 102 };
 103
 104 struct ifa_msghdrl32 {
 105         uint16_t ifam_msglen;
 106         uint8_t ifam_version;
 107         uint8_t ifam_type;
 108         int32_t ifam_addrs;
 109         int32_t ifam_flags;
 110         uint16_t ifam_index;
 111         uint16_t _ifam_spare1;
 112         uint16_t ifam_len;
 113         uint16_t ifam_data_off;
 114         int32_t ifam_metric;
 115         struct  if_data ifam_data;
 116 };
 117
 118 #define SA_SIZE32(sa)                                           \
 119     (  (((struct sockaddr *)(sa))->sa_len == 0) ?               \
 120         sizeof(int)             :                               \
 121         1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) )
 122
 123 #endif /* COMPAT_FREEBSD32 */
 124
 125 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
 126
 127 /* NB: these are not modified */
 128 static struct   sockaddr route_src = { 2, PF_ROUTE, };
 129 static struct   sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
 130
 131 /* These are external hooks for CARP. */
 132 int     (*carp_get_vhid_p)(struct ifaddr *);
 133
 134 /*
 135  * Used by rtsock/raw_input callback code to decide whether to filter the update
 136  * notification to a socket bound to a particular FIB.
 137  */
 138 #define RTS_FILTER_FIB  M_PROTO8
 139
 140 typedef struct {
 141         int     ip_count;       /* attached w/ AF_INET */
 142         int     ip6_count;      /* attached w/ AF_INET6 */
 143         int     any_count;      /* total attached */
 144 } route_cb_t;
 145 VNET_DEFINE_STATIC(route_cb_t, route_cb);
 146 #define V_route_cb VNET(route_cb)
 147
 148 struct mtx rtsock_mtx;
 149 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
 150
 151 #define RTSOCK_LOCK()   mtx_lock(&rtsock_mtx)
 152 #define RTSOCK_UNLOCK() mtx_unlock(&rtsock_mtx)
 153 #define RTSOCK_LOCK_ASSERT()    mtx_assert(&rtsock_mtx, MA_OWNED)
 154
 155 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
 156
 157 struct walkarg {
 158         int     w_tmemsize;
 159         int     w_op, w_arg;
 160         caddr_t w_tmem;
 161         struct sysctl_req *w_req;
 162 };
 163
 164 static void     rts_input(struct mbuf *m);
 165 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
 166 static int      rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
 167                         struct walkarg *w, int *plen);
 168 static int      rt_xaddrs(caddr_t cp, caddr_t cplim,
 169                         struct rt_addrinfo *rtinfo);
 170 static int      sysctl_dumpentry(struct radix_node *rn, void *vw);
 171 static int      sysctl_iflist(int af, struct walkarg *w);
 172 static int      sysctl_ifmalist(int af, struct walkarg *w);
 173 static int      route_output(struct mbuf *m, struct socket *so, ...);
 174 static void     rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out);
 175 static void     rt_dispatch(struct mbuf *, sa_family_t);
 176 static struct sockaddr  *rtsock_fix_netmask(struct sockaddr *dst,
 177                         struct sockaddr *smask, struct sockaddr_storage *dmask);
 178
 179 static struct netisr_handler rtsock_nh = {
 180         .nh_name = "rtsock",
 181         .nh_handler = rts_input,
 182         .nh_proto = NETISR_ROUTE,
 183         .nh_policy = NETISR_POLICY_SOURCE,
 184 };
 185
 186 static int
 187 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
 188 {
 189         int error, qlimit;
 190
 191         netisr_getqlimit(&rtsock_nh, &qlimit);
 192         error = sysctl_handle_int(oidp, &qlimit, 0, req);
 193         if (error || !req->newptr)
 194                 return (error);
 195         if (qlimit < 1)
 196                 return (EINVAL);
 197         return (netisr_setqlimit(&rtsock_nh, qlimit));
 198 }
 199 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
 200     0, 0, sysctl_route_netisr_maxqlen, "I",
 201     "maximum routing socket dispatch queue length");
 202
 203 static void
 204 vnet_rts_init(void)
 205 {
 206         int tmp;
 207
 208         if (IS_DEFAULT_VNET(curvnet)) {
 209                 if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
 210                         rtsock_nh.nh_qlimit = tmp;
 211                 netisr_register(&rtsock_nh);
 212         }
 213 #ifdef VIMAGE
 214          else
 215                 netisr_register_vnet(&rtsock_nh);
 216 #endif
 217 }
 218 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
 219     vnet_rts_init, 0);
 220
 221 #ifdef VIMAGE
 222 static void
 223 vnet_rts_uninit(void)
 224 {
 225
 226         netisr_unregister_vnet(&rtsock_nh);
 227 }
 228 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
 229     vnet_rts_uninit, 0);
 230 #endif
 231
 232 static int
 233 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
 234     struct rawcb *rp)
 235 {
 236         int fibnum;
 237
 238         KASSERT(m != NULL, ("%s: m is NULL", __func__));
 239         KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
 240         KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
 241
 242         /* No filtering requested. */
 243         if ((m->m_flags & RTS_FILTER_FIB) == 0)
 244                 return (0);
 245
 246         /* Check if it is a rts and the fib matches the one of the socket. */
 247         fibnum = M_GETFIB(m);
 248         if (proto->sp_family != PF_ROUTE ||
 249             rp->rcb_socket == NULL ||
 250             rp->rcb_socket->so_fibnum == fibnum)
 251                 return (0);
 252
 253         /* Filtering requested and no match, the socket shall be skipped. */
 254         return (1);
 255 }
 256
 257 static void
 258 rts_input(struct mbuf *m)
 259 {
 260         struct sockproto route_proto;
 261         unsigned short *family;
 262         struct m_tag *tag;
 263
 264         route_proto.sp_family = PF_ROUTE;
 265         tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
 266         if (tag != NULL) {
 267                 family = (unsigned short *)(tag + 1);
 268                 route_proto.sp_protocol = *family;
 269                 m_tag_delete(m, tag);
 270         } else
 271                 route_proto.sp_protocol = 0;
 272
 273         raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
 274 }
 275
 276 /*
 277  * It really doesn't make any sense at all for this code to share much
 278  * with raw_usrreq.c, since its functionality is so restricted.  XXX
 279  */
 280 static void
 281 rts_abort(struct socket *so)
 282 {
 283
 284         raw_usrreqs.pru_abort(so);
 285 }
 286
 287 static void
 288 rts_close(struct socket *so)
 289 {
 290
 291         raw_usrreqs.pru_close(so);
 292 }
 293
 294 /* pru_accept is EOPNOTSUPP */
 295
 296 static int
 297 rts_attach(struct socket *so, int proto, struct thread *td)
 298 {
 299         struct rawcb *rp;
 300         int error;
 301
 302         KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
 303
 304         /* XXX */
 305         rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
 306
 307         so->so_pcb = (caddr_t)rp;
 308         so->so_fibnum = td->td_proc->p_fibnum;
 309         error = raw_attach(so, proto);
 310         rp = sotorawcb(so);
 311         if (error) {
 312                 so->so_pcb = NULL;
 313                 free(rp, M_PCB);
 314                 return error;
 315         }
 316         RTSOCK_LOCK();
 317         switch(rp->rcb_proto.sp_protocol) {
 318         case AF_INET:
 319                 V_route_cb.ip_count++;
 320                 break;
 321         case AF_INET6:
 322                 V_route_cb.ip6_count++;
 323                 break;
 324         }
 325         V_route_cb.any_count++;
 326         RTSOCK_UNLOCK();
 327         soisconnected(so);
 328         so->so_options |= SO_USELOOPBACK;
 329         return 0;
 330 }
 331
 332 static int
 333 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
 334 {
 335
 336         return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
 337 }
 338
 339 static int
 340 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
 341 {
 342
 343         return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
 344 }
 345
 346 /* pru_connect2 is EOPNOTSUPP */
 347 /* pru_control is EOPNOTSUPP */
 348
 349 static void
 350 rts_detach(struct socket *so)
 351 {
 352         struct rawcb *rp = sotorawcb(so);
 353
 354         KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
 355
 356         RTSOCK_LOCK();
 357         switch(rp->rcb_proto.sp_protocol) {
 358         case AF_INET:
 359                 V_route_cb.ip_count--;
 360                 break;
 361         case AF_INET6:
 362                 V_route_cb.ip6_count--;
 363                 break;
 364         }
 365         V_route_cb.any_count--;
 366         RTSOCK_UNLOCK();
 367         raw_usrreqs.pru_detach(so);
 368 }
 369
 370 static int
 371 rts_disconnect(struct socket *so)
 372 {
 373
 374         return (raw_usrreqs.pru_disconnect(so));
 375 }
 376
 377 /* pru_listen is EOPNOTSUPP */
 378
 379 static int
 380 rts_peeraddr(struct socket *so, struct sockaddr **nam)
 381 {
 382
 383         return (raw_usrreqs.pru_peeraddr(so, nam));
 384 }
 385
 386 /* pru_rcvd is EOPNOTSUPP */
 387 /* pru_rcvoob is EOPNOTSUPP */
 388
 389 static int
 390 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
 391          struct mbuf *control, struct thread *td)
 392 {
 393
 394         return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
 395 }
 396
 397 /* pru_sense is null */
 398
 399 static int
 400 rts_shutdown(struct socket *so)
 401 {
 402
 403         return (raw_usrreqs.pru_shutdown(so));
 404 }
 405
 406 static int
 407 rts_sockaddr(struct socket *so, struct sockaddr **nam)
 408 {
 409
 410         return (raw_usrreqs.pru_sockaddr(so, nam));
 411 }
 412
 413 static struct pr_usrreqs route_usrreqs = {
 414         .pru_abort =            rts_abort,
 415         .pru_attach =           rts_attach,
 416         .pru_bind =             rts_bind,
 417         .pru_connect =          rts_connect,
 418         .pru_detach =           rts_detach,
 419         .pru_disconnect =       rts_disconnect,
 420         .pru_peeraddr =         rts_peeraddr,
 421         .pru_send =             rts_send,
 422         .pru_shutdown =         rts_shutdown,
 423         .pru_sockaddr =         rts_sockaddr,
 424         .pru_close =            rts_close,
 425 };
 426
 427 #ifndef _SOCKADDR_UNION_DEFINED
 428 #define _SOCKADDR_UNION_DEFINED
 429 /*
 430  * The union of all possible address formats we handle.
 431  */
 432 union sockaddr_union {
 433         struct sockaddr         sa;
 434         struct sockaddr_in      sin;
 435         struct sockaddr_in6     sin6;
 436 };
 437 #endif /* _SOCKADDR_UNION_DEFINED */
 438
 439 static int
 440 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
 441     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
 442 {
 443 #if defined(INET) || defined(INET6)
 444         struct epoch_tracker et;
 445 #endif
 446
 447         /* First, see if the returned address is part of the jail. */
 448         if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
 449                 info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
 450                 return (0);
 451         }
 452
 453         switch (info->rti_info[RTAX_DST]->sa_family) {
 454 #ifdef INET
 455         case AF_INET:
 456         {
 457                 struct in_addr ia;
 458                 struct ifaddr *ifa;
 459                 int found;
 460
 461                 found = 0;
 462                 /*
 463                  * Try to find an address on the given outgoing interface
 464                  * that belongs to the jail.
 465                  */
 466                 NET_EPOCH_ENTER(et);
 467                 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 468                         struct sockaddr *sa;
 469                         sa = ifa->ifa_addr;
 470                         if (sa->sa_family != AF_INET)
 471                                 continue;
 472                         ia = ((struct sockaddr_in *)sa)->sin_addr;
 473                         if (prison_check_ip4(cred, &ia) == 0) {
 474                                 found = 1;
 475                                 break;
 476                         }
 477                 }
 478                 NET_EPOCH_EXIT(et);
 479                 if (!found) {
 480                         /*
 481                          * As a last resort return the 'default' jail address.
 482                          */
 483                         ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
 484                             sin_addr;
 485                         if (prison_get_ip4(cred, &ia) != 0)
 486                                 return (ESRCH);
 487                 }
 488                 bzero(&saun->sin, sizeof(struct sockaddr_in));
 489                 saun->sin.sin_len = sizeof(struct sockaddr_in);
 490                 saun->sin.sin_family = AF_INET;
 491                 saun->sin.sin_addr.s_addr = ia.s_addr;
 492                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
 493                 break;
 494         }
 495 #endif
 496 #ifdef INET6
 497         case AF_INET6:
 498         {
 499                 struct in6_addr ia6;
 500                 struct ifaddr *ifa;
 501                 int found;
 502
 503                 found = 0;
 504                 /*
 505                  * Try to find an address on the given outgoing interface
 506                  * that belongs to the jail.
 507                  */
 508                 NET_EPOCH_ENTER(et);
 509                 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 510                         struct sockaddr *sa;
 511                         sa = ifa->ifa_addr;
 512                         if (sa->sa_family != AF_INET6)
 513                                 continue;
 514                         bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
 515                             &ia6, sizeof(struct in6_addr));
 516                         if (prison_check_ip6(cred, &ia6) == 0) {
 517                                 found = 1;
 518                                 break;
 519                         }
 520                 }
 521                 NET_EPOCH_EXIT(et);
 522                 if (!found) {
 523                         /*
 524                          * As a last resort return the 'default' jail address.
 525                          */
 526                         ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
 527                             sin6_addr;
 528                         if (prison_get_ip6(cred, &ia6) != 0)
 529                                 return (ESRCH);
 530                 }
 531                 bzero(&saun->sin6, sizeof(struct sockaddr_in6));
 532                 saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
 533                 saun->sin6.sin6_family = AF_INET6;
 534                 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
 535                 if (sa6_recoverscope(&saun->sin6) != 0)
 536                         return (ESRCH);
 537                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
 538                 break;
 539         }
 540 #endif
 541         default:
 542                 return (ESRCH);
 543         }
 544         return (0);
 545 }
 546
 547 /*ARGSUSED*/
 548 static int
 549 route_output(struct mbuf *m, struct socket *so, ...)
 550 {
 551         RIB_RLOCK_TRACKER;
 552         struct rt_msghdr *rtm = NULL;
 553         struct rtentry *rt = NULL;
 554         struct rib_head *rnh;
 555         struct rt_addrinfo info;
 556         struct sockaddr_storage ss;
 557 #ifdef INET6
 558         struct sockaddr_in6 *sin6;
 559         int i, rti_need_deembed = 0;
 560 #endif
 561         int alloc_len = 0, len, error = 0, fibnum;
 562         struct ifnet *ifp = NULL;
 563         union sockaddr_union saun;
 564         sa_family_t saf = AF_UNSPEC;
 565         struct rawcb *rp = NULL;
 566         struct walkarg w;
 567
 568         fibnum = so->so_fibnum;
 569
 570 #define senderr(e) { error = e; goto flush;}
 571         if (m == NULL || ((m->m_len < sizeof(long)) &&
 572                        (m = m_pullup(m, sizeof(long))) == NULL))
 573                 return (ENOBUFS);
 574         if ((m->m_flags & M_PKTHDR) == 0)
 575                 panic("route_output");
 576         len = m->m_pkthdr.len;
 577         if (len < sizeof(*rtm) ||
 578             len != mtod(m, struct rt_msghdr *)->rtm_msglen)
 579                 senderr(EINVAL);
 580
 581         /*
 582          * Most of current messages are in range 200-240 bytes,
 583          * minimize possible re-allocation on reply using larger size
 584          * buffer aligned on 1k boundaty.
 585          */
 586         alloc_len = roundup2(len, 1024);
 587         if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL)
 588                 senderr(ENOBUFS);
 589
 590         m_copydata(m, 0, len, (caddr_t)rtm);
 591         bzero(&info, sizeof(info));
 592         bzero(&w, sizeof(w));
 593
 594         if (rtm->rtm_version != RTM_VERSION) {
 595                 /* Do not touch message since format is unknown */
 596                 free(rtm, M_TEMP);
 597                 rtm = NULL;
 598                 senderr(EPROTONOSUPPORT);
 599         }
 600
 601         /*
 602          * Starting from here, it is possible
 603          * to alter original message and insert
 604          * caller PID and error value.
 605          */
 606
 607         rtm->rtm_pid = curproc->p_pid;
 608         info.rti_addrs = rtm->rtm_addrs;
 609
 610         info.rti_mflags = rtm->rtm_inits;
 611         info.rti_rmx = &rtm->rtm_rmx;
 612
 613         /*
 614          * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
 615          * link-local address because rtrequest requires addresses with
 616          * embedded scope id.
 617          */
 618         if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info))
 619                 senderr(EINVAL);
 620
 621         info.rti_flags = rtm->rtm_flags;
 622         if (info.rti_info[RTAX_DST] == NULL ||
 623             info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
 624             (info.rti_info[RTAX_GATEWAY] != NULL &&
 625              info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
 626                 senderr(EINVAL);
 627         saf = info.rti_info[RTAX_DST]->sa_family;
 628         /*
 629          * Verify that the caller has the appropriate privilege; RTM_GET
 630          * is the only operation the non-superuser is allowed.
 631          */
 632         if (rtm->rtm_type != RTM_GET) {
 633                 error = priv_check(curthread, PRIV_NET_ROUTE);
 634                 if (error)
 635                         senderr(error);
 636         }
 637
 638         /*
 639          * The given gateway address may be an interface address.
 640          * For example, issuing a "route change" command on a route
 641          * entry that was created from a tunnel, and the gateway
 642          * address given is the local end point. In this case the
 643          * RTF_GATEWAY flag must be cleared or the destination will
 644          * not be reachable even though there is no error message.
 645          */
 646         if (info.rti_info[RTAX_GATEWAY] != NULL &&
 647             info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
 648                 struct rt_addrinfo ginfo;
 649                 struct sockaddr *gdst;
 650
 651                 bzero(&ginfo, sizeof(ginfo));
 652                 bzero(&ss, sizeof(ss));
 653                 ss.ss_len = sizeof(ss);
 654
 655                 ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss;
 656                 gdst = info.rti_info[RTAX_GATEWAY];
 657
 658                 /*
 659                  * A host route through the loopback interface is
 660                  * installed for each interface adddress. In pre 8.0
 661                  * releases the interface address of a PPP link type
 662                  * is not reachable locally. This behavior is fixed as
 663                  * part of the new L2/L3 redesign and rewrite work. The
 664                  * signature of this interface address route is the
 665                  * AF_LINK sa_family type of the rt_gateway, and the
 666                  * rt_ifp has the IFF_LOOPBACK flag set.
 667                  */
 668                 if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) {
 669                         if (ss.ss_family == AF_LINK &&
 670                             ginfo.rti_ifp->if_flags & IFF_LOOPBACK) {
 671                                 info.rti_flags &= ~RTF_GATEWAY;
 672                                 info.rti_flags |= RTF_GWFLAG_COMPAT;
 673                         }
 674                         rib_free_info(&ginfo);
 675                 }
 676         }
 677
 678         switch (rtm->rtm_type) {
 679                 struct rtentry *saved_nrt;
 680
 681         case RTM_ADD:
 682         case RTM_CHANGE:
 683                 if (rtm->rtm_type == RTM_ADD) {
 684                         if (info.rti_info[RTAX_GATEWAY] == NULL)
 685                                 senderr(EINVAL);
 686                 }
 687                 saved_nrt = NULL;
 688
 689                 /* support for new ARP code */
 690                 if (info.rti_info[RTAX_GATEWAY] != NULL &&
 691                     info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
 692                     (rtm->rtm_flags & RTF_LLDATA) != 0) {
 693                         error = lla_rt_output(rtm, &info);
 694 #ifdef INET6
 695                         if (error == 0)
 696                                 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 697 #endif
 698                         break;
 699                 }
 700                 error = rtrequest1_fib(rtm->rtm_type, &info, &saved_nrt,
 701                     fibnum);
 702                 if (error == 0 && saved_nrt != NULL) {
 703 #ifdef INET6
 704                         rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 705 #endif
 706                         RT_LOCK(saved_nrt);
 707                         rtm->rtm_index = saved_nrt->rt_ifp->if_index;
 708                         RT_REMREF(saved_nrt);
 709                         RT_UNLOCK(saved_nrt);
 710                 }
 711                 break;
 712
 713         case RTM_DELETE:
 714                 saved_nrt = NULL;
 715                 /* support for new ARP code */
 716                 if (info.rti_info[RTAX_GATEWAY] &&
 717                     (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
 718                     (rtm->rtm_flags & RTF_LLDATA) != 0) {
 719                         error = lla_rt_output(rtm, &info);
 720 #ifdef INET6
 721                         if (error == 0)
 722                                 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 723 #endif
 724                         break;
 725                 }
 726                 error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt, fibnum);
 727                 if (error == 0) {
 728                         RT_LOCK(saved_nrt);
 729                         rt = saved_nrt;
 730                         goto report;
 731                 }
 732 #ifdef INET6
 733                 /* rt_msg2() will not be used when RTM_DELETE fails. */
 734                 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 735 #endif
 736                 break;
 737
 738         case RTM_GET:
 739                 rnh = rt_tables_get_rnh(fibnum, saf);
 740                 if (rnh == NULL)
 741                         senderr(EAFNOSUPPORT);
 742
 743                 RIB_RLOCK(rnh);
 744
 745                 if (info.rti_info[RTAX_NETMASK] == NULL &&
 746                     rtm->rtm_type == RTM_GET) {
 747                         /*
 748                          * Provide longest prefix match for
 749                          * address lookup (no mask).
 750                          * 'route -n get addr'
 751                          */
 752                         rt = (struct rtentry *) rnh->rnh_matchaddr(
 753                             info.rti_info[RTAX_DST], &rnh->head);
 754                 } else
 755                         rt = (struct rtentry *) rnh->rnh_lookup(
 756                             info.rti_info[RTAX_DST],
 757                             info.rti_info[RTAX_NETMASK], &rnh->head);
 758
 759                 if (rt == NULL) {
 760                         RIB_RUNLOCK(rnh);
 761                         senderr(ESRCH);
 762                 }
 763 #ifdef RADIX_MPATH
 764                 /*
 765                  * for RTM_CHANGE/LOCK, if we got multipath routes,
 766                  * we require users to specify a matching RTAX_GATEWAY.
 767                  *
 768                  * for RTM_GET, gate is optional even with multipath.
 769                  * if gate == NULL the first match is returned.
 770                  * (no need to call rt_mpath_matchgate if gate == NULL)
 771                  */
 772                 if (rt_mpath_capable(rnh) &&
 773                     (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
 774                         rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
 775                         if (!rt) {
 776                                 RIB_RUNLOCK(rnh);
 777                                 senderr(ESRCH);
 778                         }
 779                 }
 780 #endif
 781                 /*
 782                  * If performing proxied L2 entry insertion, and
 783                  * the actual PPP host entry is found, perform
 784                  * another search to retrieve the prefix route of
 785                  * the local end point of the PPP link.
 786                  */
 787                 if (rtm->rtm_flags & RTF_ANNOUNCE) {
 788                         struct sockaddr laddr;
 789
 790                         if (rt->rt_ifp != NULL &&
 791                             rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
 792                                 struct epoch_tracker et;
 793                                 struct ifaddr *ifa;
 794
 795                                 NET_EPOCH_ENTER(et);
 796                                 ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1,
 797                                                 RT_ALL_FIBS);
 798                                 if (ifa != NULL)
 799                                         rt_maskedcopy(ifa->ifa_addr,
 800                                                       &laddr,
 801                                                       ifa->ifa_netmask);
 802                                 NET_EPOCH_EXIT(et);
 803                         } else
 804                                 rt_maskedcopy(rt->rt_ifa->ifa_addr,
 805                                               &laddr,
 806                                               rt->rt_ifa->ifa_netmask);
 807                         /*
 808                          * refactor rt and no lock operation necessary
 809                          */
 810                         rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr,
 811                             &rnh->head);
 812                         if (rt == NULL) {
 813                                 RIB_RUNLOCK(rnh);
 814                                 senderr(ESRCH);
 815                         }
 816                 }
 817                 RT_LOCK(rt);
 818                 RT_ADDREF(rt);
 819                 RIB_RUNLOCK(rnh);
 820
 821 report:
 822                 RT_LOCK_ASSERT(rt);
 823                 if ((rt->rt_flags & RTF_HOST) == 0
 824                     ? jailed_without_vnet(curthread->td_ucred)
 825                     : prison_if(curthread->td_ucred,
 826                     rt_key(rt)) != 0) {
 827                         RT_UNLOCK(rt);
 828                         senderr(ESRCH);
 829                 }
 830                 info.rti_info[RTAX_DST] = rt_key(rt);
 831                 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
 832                 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
 833                     rt_mask(rt), &ss);
 834                 info.rti_info[RTAX_GENMASK] = 0;
 835                 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
 836                         ifp = rt->rt_ifp;
 837                         if (ifp) {
 838                                 info.rti_info[RTAX_IFP] =
 839                                     ifp->if_addr->ifa_addr;
 840                                 error = rtm_get_jailed(&info, ifp, rt,
 841                                     &saun, curthread->td_ucred);
 842                                 if (error != 0) {
 843                                         RT_UNLOCK(rt);
 844                                         senderr(error);
 845                                 }
 846                                 if (ifp->if_flags & IFF_POINTOPOINT)
 847                                         info.rti_info[RTAX_BRD] =
 848                                             rt->rt_ifa->ifa_dstaddr;
 849                                 rtm->rtm_index = ifp->if_index;
 850                         } else {
 851                                 info.rti_info[RTAX_IFP] = NULL;
 852                                 info.rti_info[RTAX_IFA] = NULL;
 853                         }
 854                 } else if ((ifp = rt->rt_ifp) != NULL) {
 855                         rtm->rtm_index = ifp->if_index;
 856                 }
 857
 858                 /* Check if we need to realloc storage */
 859                 rtsock_msg_buffer(rtm->rtm_type, &info, NULL, &len);
 860                 if (len > alloc_len) {
 861                         struct rt_msghdr *new_rtm;
 862                         new_rtm = malloc(len, M_TEMP, M_NOWAIT);
 863                         if (new_rtm == NULL) {
 864                                 RT_UNLOCK(rt);
 865                                 senderr(ENOBUFS);
 866                         }
 867                         bcopy(rtm, new_rtm, rtm->rtm_msglen);
 868                         free(rtm, M_TEMP);
 869                         rtm = new_rtm;
 870                         alloc_len = len;
 871                 }
 872
 873                 w.w_tmem = (caddr_t)rtm;
 874                 w.w_tmemsize = alloc_len;
 875                 rtsock_msg_buffer(rtm->rtm_type, &info, &w, &len);
 876
 877                 if (rt->rt_flags & RTF_GWFLAG_COMPAT)
 878                         rtm->rtm_flags = RTF_GATEWAY |
 879                                 (rt->rt_flags & ~RTF_GWFLAG_COMPAT);
 880                 else
 881                         rtm->rtm_flags = rt->rt_flags;
 882                 rt_getmetrics(rt, &rtm->rtm_rmx);
 883                 rtm->rtm_addrs = info.rti_addrs;
 884
 885                 RT_UNLOCK(rt);
 886                 break;
 887
 888         default:
 889                 senderr(EOPNOTSUPP);
 890         }
 891
 892 flush:
 893         if (rt != NULL)
 894                 RTFREE(rt);
 895         /*
 896          * Check to see if we don't want our own messages.
 897          */
 898         if ((so->so_options & SO_USELOOPBACK) == 0) {
 899                 if (V_route_cb.any_count <= 1) {
 900                         if (rtm != NULL)
 901                                 free(rtm, M_TEMP);
 902                         m_freem(m);
 903                         return (error);
 904                 }
 905                 /* There is another listener, so construct message */
 906                 rp = sotorawcb(so);
 907         }
 908
 909         if (rtm != NULL) {
 910 #ifdef INET6
 911                 if (rti_need_deembed) {
 912                         /* sin6_scope_id is recovered before sending rtm. */
 913                         sin6 = (struct sockaddr_in6 *)&ss;
 914                         for (i = 0; i < RTAX_MAX; i++) {
 915                                 if (info.rti_info[i] == NULL)
 916                                         continue;
 917                                 if (info.rti_info[i]->sa_family != AF_INET6)
 918                                         continue;
 919                                 bcopy(info.rti_info[i], sin6, sizeof(*sin6));
 920                                 if (sa6_recoverscope(sin6) == 0)
 921                                         bcopy(sin6, info.rti_info[i],
 922                                                     sizeof(*sin6));
 923                         }
 924                 }
 925 #endif
 926                 if (error != 0)
 927                         rtm->rtm_errno = error;
 928                 else
 929                         rtm->rtm_flags |= RTF_DONE;
 930
 931                 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
 932                 if (m->m_pkthdr.len < rtm->rtm_msglen) {
 933                         m_freem(m);
 934                         m = NULL;
 935                 } else if (m->m_pkthdr.len > rtm->rtm_msglen)
 936                         m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
 937
 938                 free(rtm, M_TEMP);
 939         }
 940         if (m != NULL) {
 941                 M_SETFIB(m, fibnum);
 942                 m->m_flags |= RTS_FILTER_FIB;
 943                 if (rp) {
 944                         /*
 945                          * XXX insure we don't get a copy by
 946                          * invalidating our protocol
 947                          */
 948                         unsigned short family = rp->rcb_proto.sp_family;
 949                         rp->rcb_proto.sp_family = 0;
 950                         rt_dispatch(m, saf);
 951                         rp->rcb_proto.sp_family = family;
 952                 } else
 953                         rt_dispatch(m, saf);
 954         }
 955
 956         return (error);
 957 }
 958
 959 static void
 960 rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out)
 961 {
 962
 963         bzero(out, sizeof(*out));
 964         out->rmx_mtu = rt->rt_mtu;
 965         out->rmx_weight = rt->rt_weight;
 966         out->rmx_pksent = counter_u64_fetch(rt->rt_pksent);
 967         /* Kernel -> userland timebase conversion. */
 968         out->rmx_expire = rt->rt_expire ?
 969             rt->rt_expire - time_uptime + time_second : 0;
 970 }
 971
 972 /*
 973  * Extract the addresses of the passed sockaddrs.
 974  * Do a little sanity checking so as to avoid bad memory references.
 975  * This data is derived straight from userland.
 976  */
 977 static int
 978 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
 979 {
 980         struct sockaddr *sa;
 981         int i;
 982
 983         for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
 984                 if ((rtinfo->rti_addrs & (1 << i)) == 0)
 985                         continue;
 986                 sa = (struct sockaddr *)cp;
 987                 /*
 988                  * It won't fit.
 989                  */
 990                 if (cp + sa->sa_len > cplim)
 991                         return (EINVAL);
 992                 /*
 993                  * there are no more.. quit now
 994                  * If there are more bits, they are in error.
 995                  * I've seen this. route(1) can evidently generate these.
 996                  * This causes kernel to core dump.
 997                  * for compatibility, If we see this, point to a safe address.
 998                  */
 999                 if (sa->sa_len == 0) {
1000                         rtinfo->rti_info[i] = &sa_zero;
1001                         return (0); /* should be EINVAL but for compat */
1002                 }
1003                 /* accept it */
1004 #ifdef INET6
1005                 if (sa->sa_family == AF_INET6)
1006                         sa6_embedscope((struct sockaddr_in6 *)sa,
1007                             V_ip6_use_defzone);
1008 #endif
1009                 rtinfo->rti_info[i] = sa;
1010                 cp += SA_SIZE(sa);
1011         }
1012         return (0);
1013 }
1014
1015 /*
1016  * Fill in @dmask with valid netmask leaving original @smask
1017  * intact. Mostly used with radix netmasks.
1018  */
1019 static struct sockaddr *
1020 rtsock_fix_netmask(struct sockaddr *dst, struct sockaddr *smask,
1021     struct sockaddr_storage *dmask)
1022 {
1023         if (dst == NULL || smask == NULL)
1024                 return (NULL);
1025
1026         memset(dmask, 0, dst->sa_len);
1027         memcpy(dmask, smask, smask->sa_len);
1028         dmask->ss_len = dst->sa_len;
1029         dmask->ss_family = dst->sa_family;
1030
1031         return ((struct sockaddr *)dmask);
1032 }
1033
1034 /*
1035  * Writes information related to @rtinfo object to newly-allocated mbuf.
1036  * Assumes MCLBYTES is enough to construct any message.
1037  * Used for OS notifications of vaious events (if/ifa announces,etc)
1038  *
1039  * Returns allocated mbuf or NULL on failure.
1040  */
1041 static struct mbuf *
1042 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1043 {
1044         struct rt_msghdr *rtm;
1045         struct mbuf *m;
1046         int i;
1047         struct sockaddr *sa;
1048 #ifdef INET6
1049         struct sockaddr_storage ss;
1050         struct sockaddr_in6 *sin6;
1051 #endif
1052         int len, dlen;
1053
1054         switch (type) {
1055
1056         case RTM_DELADDR:
1057         case RTM_NEWADDR:
1058                 len = sizeof(struct ifa_msghdr);
1059                 break;
1060
1061         case RTM_DELMADDR:
1062         case RTM_NEWMADDR:
1063                 len = sizeof(struct ifma_msghdr);
1064                 break;
1065
1066         case RTM_IFINFO:
1067                 len = sizeof(struct if_msghdr);
1068                 break;
1069
1070         case RTM_IFANNOUNCE:
1071         case RTM_IEEE80211:
1072                 len = sizeof(struct if_announcemsghdr);
1073                 break;
1074
1075         default:
1076                 len = sizeof(struct rt_msghdr);
1077         }
1078
1079         /* XXXGL: can we use MJUMPAGESIZE cluster here? */
1080         KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1081         if (len > MHLEN)
1082                 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1083         else
1084                 m = m_gethdr(M_NOWAIT, MT_DATA);
1085         if (m == NULL)
1086                 return (m);
1087
1088         m->m_pkthdr.len = m->m_len = len;
1089         rtm = mtod(m, struct rt_msghdr *);
1090         bzero((caddr_t)rtm, len);
1091         for (i = 0; i < RTAX_MAX; i++) {
1092                 if ((sa = rtinfo->rti_info[i]) == NULL)
1093                         continue;
1094                 rtinfo->rti_addrs |= (1 << i);
1095                 dlen = SA_SIZE(sa);
1096 #ifdef INET6
1097                 if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1098                         sin6 = (struct sockaddr_in6 *)&ss;
1099                         bcopy(sa, sin6, sizeof(*sin6));
1100                         if (sa6_recoverscope(sin6) == 0)
1101                                 sa = (struct sockaddr *)sin6;
1102                 }
1103 #endif
1104                 m_copyback(m, len, dlen, (caddr_t)sa);
1105                 len += dlen;
1106         }
1107         if (m->m_pkthdr.len != len) {
1108                 m_freem(m);
1109                 return (NULL);
1110         }
1111         rtm->rtm_msglen = len;
1112         rtm->rtm_version = RTM_VERSION;
1113         rtm->rtm_type = type;
1114         return (m);
1115 }
1116
1117 /*
1118  * Writes information related to @rtinfo object to preallocated buffer.
1119  * Stores needed size in @plen. If @w is NULL, calculates size without
1120  * writing.
1121  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1122  *
1123  * Returns 0 on success.
1124  *
1125  */
1126 static int
1127 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1128 {
1129         int i;
1130         int len, buflen = 0, dlen;
1131         caddr_t cp = NULL;
1132         struct rt_msghdr *rtm = NULL;
1133 #ifdef INET6
1134         struct sockaddr_storage ss;
1135         struct sockaddr_in6 *sin6;
1136 #endif
1137 #ifdef COMPAT_FREEBSD32
1138         bool compat32 = false;
1139 #endif
1140
1141         switch (type) {
1142
1143         case RTM_DELADDR:
1144         case RTM_NEWADDR:
1145                 if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1146 #ifdef COMPAT_FREEBSD32
1147                         if (w->w_req->flags & SCTL_MASK32) {
1148                                 len = sizeof(struct ifa_msghdrl32);
1149                                 compat32 = true;
1150                         } else
1151 #endif
1152                                 len = sizeof(struct ifa_msghdrl);
1153                 } else
1154                         len = sizeof(struct ifa_msghdr);
1155                 break;
1156
1157         case RTM_IFINFO:
1158 #ifdef COMPAT_FREEBSD32
1159                 if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1160                         if (w->w_op == NET_RT_IFLISTL)
1161                                 len = sizeof(struct if_msghdrl32);
1162                         else
1163                                 len = sizeof(struct if_msghdr32);
1164                         compat32 = true;
1165                         break;
1166                 }
1167 #endif
1168                 if (w != NULL && w->w_op == NET_RT_IFLISTL)
1169                         len = sizeof(struct if_msghdrl);
1170                 else
1171                         len = sizeof(struct if_msghdr);
1172                 break;
1173
1174         case RTM_NEWMADDR:
1175                 len = sizeof(struct ifma_msghdr);
1176                 break;
1177
1178         default:
1179                 len = sizeof(struct rt_msghdr);
1180         }
1181
1182         if (w != NULL) {
1183                 rtm = (struct rt_msghdr *)w->w_tmem;
1184                 buflen = w->w_tmemsize - len;
1185                 cp = (caddr_t)w->w_tmem + len;
1186         }
1187
1188         rtinfo->rti_addrs = 0;
1189         for (i = 0; i < RTAX_MAX; i++) {
1190                 struct sockaddr *sa;
1191
1192                 if ((sa = rtinfo->rti_info[i]) == NULL)
1193                         continue;
1194                 rtinfo->rti_addrs |= (1 << i);
1195 #ifdef COMPAT_FREEBSD32
1196                 if (compat32)
1197                         dlen = SA_SIZE32(sa);
1198                 else
1199 #endif
1200                         dlen = SA_SIZE(sa);
1201                 if (cp != NULL && buflen >= dlen) {
1202 #ifdef INET6
1203                         if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1204                                 sin6 = (struct sockaddr_in6 *)&ss;
1205                                 bcopy(sa, sin6, sizeof(*sin6));
1206                                 if (sa6_recoverscope(sin6) == 0)
1207                                         sa = (struct sockaddr *)sin6;
1208                         }
1209 #endif
1210                         bcopy((caddr_t)sa, cp, (unsigned)dlen);
1211                         cp += dlen;
1212                         buflen -= dlen;
1213                 } else if (cp != NULL) {
1214                         /*
1215                          * Buffer too small. Count needed size
1216                          * and return with error.
1217                          */
1218                         cp = NULL;
1219                 }
1220
1221                 len += dlen;
1222         }
1223
1224         if (cp != NULL) {
1225                 dlen = ALIGN(len) - len;
1226                 if (buflen < dlen)
1227                         cp = NULL;
1228                 else {
1229                         bzero(cp, dlen);
1230                         cp += dlen;
1231                         buflen -= dlen;
1232                 }
1233         }
1234         len = ALIGN(len);
1235
1236         if (cp != NULL) {
1237                 /* fill header iff buffer is large enough */
1238                 rtm->rtm_version = RTM_VERSION;
1239                 rtm->rtm_type = type;
1240                 rtm->rtm_msglen = len;
1241         }
1242
1243         *plen = len;
1244
1245         if (w != NULL && cp == NULL)
1246                 return (ENOBUFS);
1247
1248         return (0);
1249 }
1250
1251 /*
1252  * This routine is called to generate a message from the routing
1253  * socket indicating that a redirect has occurred, a routing lookup
1254  * has failed, or that a protocol has detected timeouts to a particular
1255  * destination.
1256  */
1257 void
1258 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1259     int fibnum)
1260 {
1261         struct rt_msghdr *rtm;
1262         struct mbuf *m;
1263         struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1264
1265         if (V_route_cb.any_count == 0)
1266                 return;
1267         m = rtsock_msg_mbuf(type, rtinfo);
1268         if (m == NULL)
1269                 return;
1270
1271         if (fibnum != RT_ALL_FIBS) {
1272                 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1273                     "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1274                 M_SETFIB(m, fibnum);
1275                 m->m_flags |= RTS_FILTER_FIB;
1276         }
1277
1278         rtm = mtod(m, struct rt_msghdr *);
1279         rtm->rtm_flags = RTF_DONE | flags;
1280         rtm->rtm_errno = error;
1281         rtm->rtm_addrs = rtinfo->rti_addrs;
1282         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1283 }
1284
1285 void
1286 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1287 {
1288
1289         rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1290 }
1291
1292 /*
1293  * This routine is called to generate a message from the routing
1294  * socket indicating that the status of a network interface has changed.
1295  */
1296 void
1297 rt_ifmsg(struct ifnet *ifp)
1298 {
1299         struct if_msghdr *ifm;
1300         struct mbuf *m;
1301         struct rt_addrinfo info;
1302
1303         if (V_route_cb.any_count == 0)
1304                 return;
1305         bzero((caddr_t)&info, sizeof(info));
1306         m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1307         if (m == NULL)
1308                 return;
1309         ifm = mtod(m, struct if_msghdr *);
1310         ifm->ifm_index = ifp->if_index;
1311         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1312         if_data_copy(ifp, &ifm->ifm_data);
1313         ifm->ifm_addrs = 0;
1314         rt_dispatch(m, AF_UNSPEC);
1315 }
1316
1317 /*
1318  * Announce interface address arrival/withdraw.
1319  * Please do not call directly, use rt_addrmsg().
1320  * Assume input data to be valid.
1321  * Returns 0 on success.
1322  */
1323 int
1324 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1325 {
1326         struct rt_addrinfo info;
1327         struct sockaddr *sa;
1328         int ncmd;
1329         struct mbuf *m;
1330         struct ifa_msghdr *ifam;
1331         struct ifnet *ifp = ifa->ifa_ifp;
1332         struct sockaddr_storage ss;
1333
1334         if (V_route_cb.any_count == 0)
1335                 return (0);
1336
1337         ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1338
1339         bzero((caddr_t)&info, sizeof(info));
1340         info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1341         info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1342         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1343             info.rti_info[RTAX_IFP], ifa->ifa_netmask, &ss);
1344         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1345         if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1346                 return (ENOBUFS);
1347         ifam = mtod(m, struct ifa_msghdr *);
1348         ifam->ifam_index = ifp->if_index;
1349         ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1350         ifam->ifam_flags = ifa->ifa_flags;
1351         ifam->ifam_addrs = info.rti_addrs;
1352
1353         if (fibnum != RT_ALL_FIBS) {
1354                 M_SETFIB(m, fibnum);
1355                 m->m_flags |= RTS_FILTER_FIB;
1356         }
1357
1358         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1359
1360         return (0);
1361 }
1362
1363 /*
1364  * Announce route addition/removal.
1365  * Please do not call directly, use rt_routemsg().
1366  * Note that @rt data MAY be inconsistent/invalid:
1367  * if some userland app sends us "invalid" route message (invalid mask,
1368  * no dst, wrong address families, etc...) we need to pass it back
1369  * to app (and any other rtsock consumers) with rtm_errno field set to
1370  * non-zero value.
1371  *
1372  * Returns 0 on success.
1373  */
1374 int
1375 rtsock_routemsg(int cmd, struct ifnet *ifp, int error, struct rtentry *rt,
1376     int fibnum)
1377 {
1378         struct rt_addrinfo info;
1379         struct sockaddr *sa;
1380         struct mbuf *m;
1381         struct rt_msghdr *rtm;
1382         struct sockaddr_storage ss;
1383
1384         if (V_route_cb.any_count == 0)
1385                 return (0);
1386
1387         bzero((caddr_t)&info, sizeof(info));
1388         info.rti_info[RTAX_DST] = sa = rt_key(rt);
1389         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(sa, rt_mask(rt), &ss);
1390         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1391         if ((m = rtsock_msg_mbuf(cmd, &info)) == NULL)
1392                 return (ENOBUFS);
1393         rtm = mtod(m, struct rt_msghdr *);
1394         rtm->rtm_index = ifp->if_index;
1395         rtm->rtm_flags |= rt->rt_flags;
1396         rtm->rtm_errno = error;
1397         rtm->rtm_addrs = info.rti_addrs;
1398
1399         if (fibnum != RT_ALL_FIBS) {
1400                 M_SETFIB(m, fibnum);
1401                 m->m_flags |= RTS_FILTER_FIB;
1402         }
1403
1404         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1405
1406         return (0);
1407 }
1408
1409 /*
1410  * This is the analogue to the rt_newaddrmsg which performs the same
1411  * function but for multicast group memberhips.  This is easier since
1412  * there is no route state to worry about.
1413  */
1414 void
1415 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1416 {
1417         struct rt_addrinfo info;
1418         struct mbuf *m = NULL;
1419         struct ifnet *ifp = ifma->ifma_ifp;
1420         struct ifma_msghdr *ifmam;
1421
1422         if (V_route_cb.any_count == 0)
1423                 return;
1424
1425         bzero((caddr_t)&info, sizeof(info));
1426         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1427         if (ifp && ifp->if_addr)
1428                 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1429         else
1430                 info.rti_info[RTAX_IFP] = NULL;
1431         /*
1432          * If a link-layer address is present, present it as a ``gateway''
1433          * (similarly to how ARP entries, e.g., are presented).
1434          */
1435         info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1436         m = rtsock_msg_mbuf(cmd, &info);
1437         if (m == NULL)
1438                 return;
1439         ifmam = mtod(m, struct ifma_msghdr *);
1440         KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1441             __func__));
1442         ifmam->ifmam_index = ifp->if_index;
1443         ifmam->ifmam_addrs = info.rti_addrs;
1444         rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1445 }
1446
1447 static struct mbuf *
1448 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1449         struct rt_addrinfo *info)
1450 {
1451         struct if_announcemsghdr *ifan;
1452         struct mbuf *m;
1453
1454         if (V_route_cb.any_count == 0)
1455                 return NULL;
1456         bzero((caddr_t)info, sizeof(*info));
1457         m = rtsock_msg_mbuf(type, info);
1458         if (m != NULL) {
1459                 ifan = mtod(m, struct if_announcemsghdr *);
1460                 ifan->ifan_index = ifp->if_index;
1461                 strlcpy(ifan->ifan_name, ifp->if_xname,
1462                         sizeof(ifan->ifan_name));
1463                 ifan->ifan_what = what;
1464         }
1465         return m;
1466 }
1467
1468 /*
1469  * This is called to generate routing socket messages indicating
1470  * IEEE80211 wireless events.
1471  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1472  */
1473 void
1474 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1475 {
1476         struct mbuf *m;
1477         struct rt_addrinfo info;
1478
1479         m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1480         if (m != NULL) {
1481                 /*
1482                  * Append the ieee80211 data.  Try to stick it in the
1483                  * mbuf containing the ifannounce msg; otherwise allocate
1484                  * a new mbuf and append.
1485                  *
1486                  * NB: we assume m is a single mbuf.
1487                  */
1488                 if (data_len > M_TRAILINGSPACE(m)) {
1489                         struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1490                         if (n == NULL) {
1491                                 m_freem(m);
1492                                 return;
1493                         }
1494                         bcopy(data, mtod(n, void *), data_len);
1495                         n->m_len = data_len;
1496                         m->m_next = n;
1497                 } else if (data_len > 0) {
1498                         bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1499                         m->m_len += data_len;
1500                 }
1501                 if (m->m_flags & M_PKTHDR)
1502                         m->m_pkthdr.len += data_len;
1503                 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1504                 rt_dispatch(m, AF_UNSPEC);
1505         }
1506 }
1507
1508 /*
1509  * This is called to generate routing socket messages indicating
1510  * network interface arrival and departure.
1511  */
1512 void
1513 rt_ifannouncemsg(struct ifnet *ifp, int what)
1514 {
1515         struct mbuf *m;
1516         struct rt_addrinfo info;
1517
1518         m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1519         if (m != NULL)
1520                 rt_dispatch(m, AF_UNSPEC);
1521 }
1522
1523 static void
1524 rt_dispatch(struct mbuf *m, sa_family_t saf)
1525 {
1526         struct m_tag *tag;
1527
1528         /*
1529          * Preserve the family from the sockaddr, if any, in an m_tag for
1530          * use when injecting the mbuf into the routing socket buffer from
1531          * the netisr.
1532          */
1533         if (saf != AF_UNSPEC) {
1534                 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1535                     M_NOWAIT);
1536                 if (tag == NULL) {
1537                         m_freem(m);
1538                         return;
1539                 }
1540                 *(unsigned short *)(tag + 1) = saf;
1541                 m_tag_prepend(m, tag);
1542         }
1543 #ifdef VIMAGE
1544         if (V_loif)
1545                 m->m_pkthdr.rcvif = V_loif;
1546         else {
1547                 m_freem(m);
1548                 return;
1549         }
1550 #endif
1551         netisr_queue(NETISR_ROUTE, m);  /* mbuf is free'd on failure. */
1552 }
1553
1554 /*
1555  * This is used in dumping the kernel table via sysctl().
1556  */
1557 static int
1558 sysctl_dumpentry(struct radix_node *rn, void *vw)
1559 {
1560         struct walkarg *w = vw;
1561         struct rtentry *rt = (struct rtentry *)rn;
1562         int error = 0, size;
1563         struct rt_addrinfo info;
1564         struct sockaddr_storage ss;
1565
1566         NET_EPOCH_ASSERT();
1567
1568         if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1569                 return 0;
1570         if ((rt->rt_flags & RTF_HOST) == 0
1571             ? jailed_without_vnet(w->w_req->td->td_ucred)
1572             : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1573                 return (0);
1574         bzero((caddr_t)&info, sizeof(info));
1575         info.rti_info[RTAX_DST] = rt_key(rt);
1576         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1577         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
1578             rt_mask(rt), &ss);
1579         info.rti_info[RTAX_GENMASK] = 0;
1580         if (rt->rt_ifp && !(rt->rt_ifp->if_flags & IFF_DYING)) {
1581                 info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1582                 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1583                 if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1584                         info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1585         }
1586         if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
1587                 return (error);
1588         if (w->w_req && w->w_tmem) {
1589                 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1590
1591                 bzero(&rtm->rtm_index,
1592                     sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
1593                 if (rt->rt_flags & RTF_GWFLAG_COMPAT)
1594                         rtm->rtm_flags = RTF_GATEWAY |
1595                                 (rt->rt_flags & ~RTF_GWFLAG_COMPAT);
1596                 else
1597                         rtm->rtm_flags = rt->rt_flags;
1598                 rt_getmetrics(rt, &rtm->rtm_rmx);
1599                 rtm->rtm_index = rt->rt_ifp->if_index;
1600                 rtm->rtm_addrs = info.rti_addrs;
1601                 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1602                 return (error);
1603         }
1604         return (error);
1605 }
1606
1607 static int
1608 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
1609     struct rt_addrinfo *info, struct walkarg *w, int len)
1610 {
1611         struct if_msghdrl *ifm;
1612         struct if_data *ifd;
1613
1614         ifm = (struct if_msghdrl *)w->w_tmem;
1615
1616 #ifdef COMPAT_FREEBSD32
1617         if (w->w_req->flags & SCTL_MASK32) {
1618                 struct if_msghdrl32 *ifm32;
1619
1620                 ifm32 = (struct if_msghdrl32 *)ifm;
1621                 ifm32->ifm_addrs = info->rti_addrs;
1622                 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1623                 ifm32->ifm_index = ifp->if_index;
1624                 ifm32->_ifm_spare1 = 0;
1625                 ifm32->ifm_len = sizeof(*ifm32);
1626                 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1627                 ifm32->_ifm_spare2 = 0;
1628                 ifd = &ifm32->ifm_data;
1629         } else
1630 #endif
1631         {
1632                 ifm->ifm_addrs = info->rti_addrs;
1633                 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1634                 ifm->ifm_index = ifp->if_index;
1635                 ifm->_ifm_spare1 = 0;
1636                 ifm->ifm_len = sizeof(*ifm);
1637                 ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1638                 ifm->_ifm_spare2 = 0;
1639                 ifd = &ifm->ifm_data;
1640         }
1641
1642         memcpy(ifd, src_ifd, sizeof(*ifd));
1643
1644         return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1645 }
1646
1647 static int
1648 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
1649     struct rt_addrinfo *info, struct walkarg *w, int len)
1650 {
1651         struct if_msghdr *ifm;
1652         struct if_data *ifd;
1653
1654         ifm = (struct if_msghdr *)w->w_tmem;
1655
1656 #ifdef COMPAT_FREEBSD32
1657         if (w->w_req->flags & SCTL_MASK32) {
1658                 struct if_msghdr32 *ifm32;
1659
1660                 ifm32 = (struct if_msghdr32 *)ifm;
1661                 ifm32->ifm_addrs = info->rti_addrs;
1662                 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1663                 ifm32->ifm_index = ifp->if_index;
1664                 ifm32->_ifm_spare1 = 0;
1665                 ifd = &ifm32->ifm_data;
1666         } else
1667 #endif
1668         {
1669                 ifm->ifm_addrs = info->rti_addrs;
1670                 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1671                 ifm->ifm_index = ifp->if_index;
1672                 ifm->_ifm_spare1 = 0;
1673                 ifd = &ifm->ifm_data;
1674         }
1675
1676         memcpy(ifd, src_ifd, sizeof(*ifd));
1677
1678         return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1679 }
1680
1681 static int
1682 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1683     struct walkarg *w, int len)
1684 {
1685         struct ifa_msghdrl *ifam;
1686         struct if_data *ifd;
1687
1688         ifam = (struct ifa_msghdrl *)w->w_tmem;
1689
1690 #ifdef COMPAT_FREEBSD32
1691         if (w->w_req->flags & SCTL_MASK32) {
1692                 struct ifa_msghdrl32 *ifam32;
1693
1694                 ifam32 = (struct ifa_msghdrl32 *)ifam;
1695                 ifam32->ifam_addrs = info->rti_addrs;
1696                 ifam32->ifam_flags = ifa->ifa_flags;
1697                 ifam32->ifam_index = ifa->ifa_ifp->if_index;
1698                 ifam32->_ifam_spare1 = 0;
1699                 ifam32->ifam_len = sizeof(*ifam32);
1700                 ifam32->ifam_data_off =
1701                     offsetof(struct ifa_msghdrl32, ifam_data);
1702                 ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
1703                 ifd = &ifam32->ifam_data;
1704         } else
1705 #endif
1706         {
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_spare1 = 0;
1711                 ifam->ifam_len = sizeof(*ifam);
1712                 ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1713                 ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1714                 ifd = &ifam->ifam_data;
1715         }
1716
1717         bzero(ifd, sizeof(*ifd));
1718         ifd->ifi_datalen = sizeof(struct if_data);
1719         ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
1720         ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
1721         ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
1722         ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
1723
1724         /* Fixup if_data carp(4) vhid. */
1725         if (carp_get_vhid_p != NULL)
1726                 ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
1727
1728         return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1729 }
1730
1731 static int
1732 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1733     struct walkarg *w, int len)
1734 {
1735         struct ifa_msghdr *ifam;
1736
1737         ifam = (struct ifa_msghdr *)w->w_tmem;
1738         ifam->ifam_addrs = info->rti_addrs;
1739         ifam->ifam_flags = ifa->ifa_flags;
1740         ifam->ifam_index = ifa->ifa_ifp->if_index;
1741         ifam->_ifam_spare1 = 0;
1742         ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1743
1744         return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1745 }
1746
1747 static int
1748 sysctl_iflist(int af, struct walkarg *w)
1749 {
1750         struct ifnet *ifp;
1751         struct ifaddr *ifa;
1752         struct if_data ifd;
1753         struct rt_addrinfo info;
1754         int len, error = 0;
1755         struct sockaddr_storage ss;
1756         struct epoch_tracker et;
1757
1758         bzero((caddr_t)&info, sizeof(info));
1759         bzero(&ifd, sizeof(ifd));
1760         NET_EPOCH_ENTER(et);
1761         CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1762                 if (w->w_arg && w->w_arg != ifp->if_index)
1763                         continue;
1764                 if_data_copy(ifp, &ifd);
1765                 ifa = ifp->if_addr;
1766                 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1767                 error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
1768                 if (error != 0)
1769                         goto done;
1770                 info.rti_info[RTAX_IFP] = NULL;
1771                 if (w->w_req && w->w_tmem) {
1772                         if (w->w_op == NET_RT_IFLISTL)
1773                                 error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
1774                                     len);
1775                         else
1776                                 error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
1777                                     len);
1778                         if (error)
1779                                 goto done;
1780                 }
1781                 while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
1782                         if (af && af != ifa->ifa_addr->sa_family)
1783                                 continue;
1784                         if (prison_if(w->w_req->td->td_ucred,
1785                             ifa->ifa_addr) != 0)
1786                                 continue;
1787                         info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1788                         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1789                             ifa->ifa_addr, ifa->ifa_netmask, &ss);
1790                         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1791                         error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
1792                         if (error != 0)
1793                                 goto done;
1794                         if (w->w_req && w->w_tmem) {
1795                                 if (w->w_op == NET_RT_IFLISTL)
1796                                         error = sysctl_iflist_ifaml(ifa, &info,
1797                                             w, len);
1798                                 else
1799                                         error = sysctl_iflist_ifam(ifa, &info,
1800                                             w, len);
1801                                 if (error)
1802                                         goto done;
1803                         }
1804                 }
1805                 info.rti_info[RTAX_IFA] = NULL;
1806                 info.rti_info[RTAX_NETMASK] = NULL;
1807                 info.rti_info[RTAX_BRD] = NULL;
1808         }
1809 done:
1810         NET_EPOCH_EXIT(et);
1811         return (error);
1812 }
1813
1814 static int
1815 sysctl_ifmalist(int af, struct walkarg *w)
1816 {
1817         struct rt_addrinfo info;
1818         struct epoch_tracker et;
1819         struct ifaddr *ifa;
1820         struct ifmultiaddr *ifma;
1821         struct ifnet *ifp;
1822         int error, len;
1823
1824         error = 0;
1825         bzero((caddr_t)&info, sizeof(info));
1826
1827         NET_EPOCH_ENTER(et);
1828         CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1829                 if (w->w_arg && w->w_arg != ifp->if_index)
1830                         continue;
1831                 ifa = ifp->if_addr;
1832                 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1833                 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1834                         if (af && af != ifma->ifma_addr->sa_family)
1835                                 continue;
1836                         if (prison_if(w->w_req->td->td_ucred,
1837                             ifma->ifma_addr) != 0)
1838                                 continue;
1839                         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1840                         info.rti_info[RTAX_GATEWAY] =
1841                             (ifma->ifma_addr->sa_family != AF_LINK) ?
1842                             ifma->ifma_lladdr : NULL;
1843                         error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
1844                         if (error != 0)
1845                                 break;
1846                         if (w->w_req && w->w_tmem) {
1847                                 struct ifma_msghdr *ifmam;
1848
1849                                 ifmam = (struct ifma_msghdr *)w->w_tmem;
1850                                 ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1851                                 ifmam->ifmam_flags = 0;
1852                                 ifmam->ifmam_addrs = info.rti_addrs;
1853                                 ifmam->_ifmam_spare1 = 0;
1854                                 error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1855                                 if (error != 0)
1856                                         break;
1857                         }
1858                 }
1859                 if (error != 0)
1860                         break;
1861         }
1862         NET_EPOCH_EXIT(et);
1863         return (error);
1864 }
1865
1866 static int
1867 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1868 {
1869         RIB_RLOCK_TRACKER;
1870         int     *name = (int *)arg1;
1871         u_int   namelen = arg2;
1872         struct rib_head *rnh = NULL; /* silence compiler. */
1873         int     i, lim, error = EINVAL;
1874         int     fib = 0;
1875         u_char  af;
1876         struct  walkarg w;
1877
1878         name ++;
1879         namelen--;
1880         if (req->newptr)
1881                 return (EPERM);
1882         if (name[1] == NET_RT_DUMP) {
1883                 if (namelen == 3)
1884                         fib = req->td->td_proc->p_fibnum;
1885                 else if (namelen == 4)
1886                         fib = (name[3] == RT_ALL_FIBS) ?
1887                             req->td->td_proc->p_fibnum : name[3];
1888                 else
1889                         return ((namelen < 3) ? EISDIR : ENOTDIR);
1890                 if (fib < 0 || fib >= rt_numfibs)
1891                         return (EINVAL);
1892         } else if (namelen != 3)
1893                 return ((namelen < 3) ? EISDIR : ENOTDIR);
1894         af = name[0];
1895         if (af > AF_MAX)
1896                 return (EINVAL);
1897         bzero(&w, sizeof(w));
1898         w.w_op = name[1];
1899         w.w_arg = name[2];
1900         w.w_req = req;
1901
1902         error = sysctl_wire_old_buffer(req, 0);
1903         if (error)
1904                 return (error);
1905
1906         /*
1907          * Allocate reply buffer in advance.
1908          * All rtsock messages has maximum length of u_short.
1909          */
1910         w.w_tmemsize = 65536;
1911         w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
1912
1913         switch (w.w_op) {
1914
1915         case NET_RT_DUMP:
1916         case NET_RT_FLAGS:
1917                 if (af == 0) {                  /* dump all tables */
1918                         i = 1;
1919                         lim = AF_MAX;
1920                 } else                          /* dump only one table */
1921                         i = lim = af;
1922
1923                 /*
1924                  * take care of llinfo entries, the caller must
1925                  * specify an AF
1926                  */
1927                 if (w.w_op == NET_RT_FLAGS &&
1928                     (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1929                         if (af != 0)
1930                                 error = lltable_sysctl_dumparp(af, w.w_req);
1931                         else
1932                                 error = EINVAL;
1933                         break;
1934                 }
1935                 /*
1936                  * take care of routing entries
1937                  */
1938                 for (error = 0; error == 0 && i <= lim; i++) {
1939                         rnh = rt_tables_get_rnh(fib, i);
1940                         if (rnh != NULL) {
1941                                 struct epoch_tracker et;
1942
1943                                 RIB_RLOCK(rnh);
1944                                 NET_EPOCH_ENTER(et);
1945                                 error = rnh->rnh_walktree(&rnh->head,
1946                                     sysctl_dumpentry, &w);
1947                                 NET_EPOCH_EXIT(et);
1948                                 RIB_RUNLOCK(rnh);
1949                         } else if (af != 0)
1950                                 error = EAFNOSUPPORT;
1951                 }
1952                 break;
1953
1954         case NET_RT_IFLIST:
1955         case NET_RT_IFLISTL:
1956                 error = sysctl_iflist(af, &w);
1957                 break;
1958
1959         case NET_RT_IFMALIST:
1960                 error = sysctl_ifmalist(af, &w);
1961                 break;
1962         }
1963
1964         free(w.w_tmem, M_TEMP);
1965         return (error);
1966 }
1967
1968 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1969
1970 /*
1971  * Definitions of protocols supported in the ROUTE domain.
1972  */
1973
1974 static struct domain routedomain;               /* or at least forward */
1975
1976 static struct protosw routesw[] = {
1977 {
1978         .pr_type =              SOCK_RAW,
1979         .pr_domain =            &routedomain,
1980         .pr_flags =             PR_ATOMIC|PR_ADDR,
1981         .pr_output =            route_output,
1982         .pr_ctlinput =          raw_ctlinput,
1983         .pr_init =              raw_init,
1984         .pr_usrreqs =           &route_usrreqs
1985 }
1986 };
1987
1988 static struct domain routedomain = {
1989         .dom_family =           PF_ROUTE,
1990         .dom_name =              "route",
1991         .dom_protosw =          routesw,
1992         .dom_protoswNPROTOSW =  &routesw[nitems(routesw)]
1993 };
1994
1995 VNET_DOMAIN_SET(route);