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_ddb.h"
  35 #include "opt_mpath.h"
  36 #include "opt_inet.h"
  37 #include "opt_inet6.h"
  38
  39 #include <sys/param.h>
  40 #include <sys/jail.h>
  41 #include <sys/kernel.h>
  42 #include <sys/domain.h>
  43 #include <sys/lock.h>
  44 #include <sys/malloc.h>
  45 #include <sys/mbuf.h>
  46 #include <sys/priv.h>
  47 #include <sys/proc.h>
  48 #include <sys/protosw.h>
  49 #include <sys/rmlock.h>
  50 #include <sys/rwlock.h>
  51 #include <sys/signalvar.h>
  52 #include <sys/socket.h>
  53 #include <sys/socketvar.h>
  54 #include <sys/sysctl.h>
  55 #include <sys/systm.h>
  56
  57 #include <net/if.h>
  58 #include <net/if_var.h>
  59 #include <net/if_dl.h>
  60 #include <net/if_llatbl.h>
  61 #include <net/if_types.h>
  62 #include <net/netisr.h>
  63 #include <net/raw_cb.h>
  64 #include <net/route.h>
  65 #include <net/route/route_var.h>
  66 #ifdef RADIX_MPATH
  67 #include <net/radix_mpath.h>
  68 #endif
  69 #include <net/vnet.h>
  70
  71 #include <netinet/in.h>
  72 #include <netinet/if_ether.h>
  73 #include <netinet/ip_carp.h>
  74 #ifdef INET6
  75 #include <netinet6/ip6_var.h>
  76 #include <netinet6/scope6_var.h>
  77 #endif
  78 #include <net/route/nhop.h>
  79 #include <net/route/shared.h>
  80
  81 #ifdef COMPAT_FREEBSD32
  82 #include <sys/mount.h>
  83 #include <compat/freebsd32/freebsd32.h>
  84
  85 struct if_msghdr32 {
  86         uint16_t ifm_msglen;
  87         uint8_t ifm_version;
  88         uint8_t ifm_type;
  89         int32_t ifm_addrs;
  90         int32_t ifm_flags;
  91         uint16_t ifm_index;
  92         uint16_t _ifm_spare1;
  93         struct  if_data ifm_data;
  94 };
  95
  96 struct if_msghdrl32 {
  97         uint16_t ifm_msglen;
  98         uint8_t ifm_version;
  99         uint8_t ifm_type;
 100         int32_t ifm_addrs;
 101         int32_t ifm_flags;
 102         uint16_t ifm_index;
 103         uint16_t _ifm_spare1;
 104         uint16_t ifm_len;
 105         uint16_t ifm_data_off;
 106         uint32_t _ifm_spare2;
 107         struct  if_data ifm_data;
 108 };
 109
 110 struct ifa_msghdrl32 {
 111         uint16_t ifam_msglen;
 112         uint8_t ifam_version;
 113         uint8_t ifam_type;
 114         int32_t ifam_addrs;
 115         int32_t ifam_flags;
 116         uint16_t ifam_index;
 117         uint16_t _ifam_spare1;
 118         uint16_t ifam_len;
 119         uint16_t ifam_data_off;
 120         int32_t ifam_metric;
 121         struct  if_data ifam_data;
 122 };
 123
 124 #define SA_SIZE32(sa)                                           \
 125     (  (((struct sockaddr *)(sa))->sa_len == 0) ?               \
 126         sizeof(int)             :                               \
 127         1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) )
 128
 129 #endif /* COMPAT_FREEBSD32 */
 130
 131 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
 132
 133 /* NB: these are not modified */
 134 static struct   sockaddr route_src = { 2, PF_ROUTE, };
 135 static struct   sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
 136
 137 /* These are external hooks for CARP. */
 138 int     (*carp_get_vhid_p)(struct ifaddr *);
 139
 140 /*
 141  * Used by rtsock/raw_input callback code to decide whether to filter the update
 142  * notification to a socket bound to a particular FIB.
 143  */
 144 #define RTS_FILTER_FIB  M_PROTO8
 145
 146 typedef struct {
 147         int     ip_count;       /* attached w/ AF_INET */
 148         int     ip6_count;      /* attached w/ AF_INET6 */
 149         int     any_count;      /* total attached */
 150 } route_cb_t;
 151 VNET_DEFINE_STATIC(route_cb_t, route_cb);
 152 #define V_route_cb VNET(route_cb)
 153
 154 struct mtx rtsock_mtx;
 155 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
 156
 157 #define RTSOCK_LOCK()   mtx_lock(&rtsock_mtx)
 158 #define RTSOCK_UNLOCK() mtx_unlock(&rtsock_mtx)
 159 #define RTSOCK_LOCK_ASSERT()    mtx_assert(&rtsock_mtx, MA_OWNED)
 160
 161 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD | CTLFLAG_MPSAFE, 0,
 162     "");
 163
 164 struct walkarg {
 165         int     w_tmemsize;
 166         int     w_op, w_arg;
 167         caddr_t w_tmem;
 168         struct sysctl_req *w_req;
 169 };
 170
 171 static void     rts_input(struct mbuf *m);
 172 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
 173 static int      rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
 174                         struct walkarg *w, int *plen);
 175 static int      rt_xaddrs(caddr_t cp, caddr_t cplim,
 176                         struct rt_addrinfo *rtinfo);
 177 static int      sysctl_dumpentry(struct radix_node *rn, void *vw);
 178 static int      sysctl_iflist(int af, struct walkarg *w);
 179 static int      sysctl_ifmalist(int af, struct walkarg *w);
 180 static int      route_output(struct mbuf *m, struct socket *so, ...);
 181 static void     rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out);
 182 static void     rt_dispatch(struct mbuf *, sa_family_t);
 183 static int      handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
 184                         struct rt_msghdr *rtm, struct rtentry **ret_nrt);
 185 static int      update_rtm_from_rte(struct rt_addrinfo *info,
 186                         struct rt_msghdr **prtm, int alloc_len,
 187                         struct rtentry *rt);
 188 static void     send_rtm_reply(struct socket *so, struct rt_msghdr *rtm,
 189                         struct mbuf *m, sa_family_t saf, u_int fibnum,
 190                         int rtm_errno);
 191 static int      can_export_rte(struct ucred *td_ucred, const struct rtentry *rt);
 192
 193 static struct netisr_handler rtsock_nh = {
 194         .nh_name = "rtsock",
 195         .nh_handler = rts_input,
 196         .nh_proto = NETISR_ROUTE,
 197         .nh_policy = NETISR_POLICY_SOURCE,
 198 };
 199
 200 static int
 201 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
 202 {
 203         int error, qlimit;
 204
 205         netisr_getqlimit(&rtsock_nh, &qlimit);
 206         error = sysctl_handle_int(oidp, &qlimit, 0, req);
 207         if (error || !req->newptr)
 208                 return (error);
 209         if (qlimit < 1)
 210                 return (EINVAL);
 211         return (netisr_setqlimit(&rtsock_nh, qlimit));
 212 }
 213 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen,
 214     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
 215     0, 0, sysctl_route_netisr_maxqlen, "I",
 216     "maximum routing socket dispatch queue length");
 217
 218 static void
 219 vnet_rts_init(void)
 220 {
 221         int tmp;
 222
 223         if (IS_DEFAULT_VNET(curvnet)) {
 224                 if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
 225                         rtsock_nh.nh_qlimit = tmp;
 226                 netisr_register(&rtsock_nh);
 227         }
 228 #ifdef VIMAGE
 229          else
 230                 netisr_register_vnet(&rtsock_nh);
 231 #endif
 232 }
 233 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
 234     vnet_rts_init, 0);
 235
 236 #ifdef VIMAGE
 237 static void
 238 vnet_rts_uninit(void)
 239 {
 240
 241         netisr_unregister_vnet(&rtsock_nh);
 242 }
 243 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
 244     vnet_rts_uninit, 0);
 245 #endif
 246
 247 static int
 248 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
 249     struct rawcb *rp)
 250 {
 251         int fibnum;
 252
 253         KASSERT(m != NULL, ("%s: m is NULL", __func__));
 254         KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
 255         KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
 256
 257         /* No filtering requested. */
 258         if ((m->m_flags & RTS_FILTER_FIB) == 0)
 259                 return (0);
 260
 261         /* Check if it is a rts and the fib matches the one of the socket. */
 262         fibnum = M_GETFIB(m);
 263         if (proto->sp_family != PF_ROUTE ||
 264             rp->rcb_socket == NULL ||
 265             rp->rcb_socket->so_fibnum == fibnum)
 266                 return (0);
 267
 268         /* Filtering requested and no match, the socket shall be skipped. */
 269         return (1);
 270 }
 271
 272 static void
 273 rts_input(struct mbuf *m)
 274 {
 275         struct sockproto route_proto;
 276         unsigned short *family;
 277         struct m_tag *tag;
 278
 279         route_proto.sp_family = PF_ROUTE;
 280         tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
 281         if (tag != NULL) {
 282                 family = (unsigned short *)(tag + 1);
 283                 route_proto.sp_protocol = *family;
 284                 m_tag_delete(m, tag);
 285         } else
 286                 route_proto.sp_protocol = 0;
 287
 288         raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
 289 }
 290
 291 /*
 292  * It really doesn't make any sense at all for this code to share much
 293  * with raw_usrreq.c, since its functionality is so restricted.  XXX
 294  */
 295 static void
 296 rts_abort(struct socket *so)
 297 {
 298
 299         raw_usrreqs.pru_abort(so);
 300 }
 301
 302 static void
 303 rts_close(struct socket *so)
 304 {
 305
 306         raw_usrreqs.pru_close(so);
 307 }
 308
 309 /* pru_accept is EOPNOTSUPP */
 310
 311 static int
 312 rts_attach(struct socket *so, int proto, struct thread *td)
 313 {
 314         struct rawcb *rp;
 315         int error;
 316
 317         KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
 318
 319         /* XXX */
 320         rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
 321
 322         so->so_pcb = (caddr_t)rp;
 323         so->so_fibnum = td->td_proc->p_fibnum;
 324         error = raw_attach(so, proto);
 325         rp = sotorawcb(so);
 326         if (error) {
 327                 so->so_pcb = NULL;
 328                 free(rp, M_PCB);
 329                 return error;
 330         }
 331         RTSOCK_LOCK();
 332         switch(rp->rcb_proto.sp_protocol) {
 333         case AF_INET:
 334                 V_route_cb.ip_count++;
 335                 break;
 336         case AF_INET6:
 337                 V_route_cb.ip6_count++;
 338                 break;
 339         }
 340         V_route_cb.any_count++;
 341         RTSOCK_UNLOCK();
 342         soisconnected(so);
 343         so->so_options |= SO_USELOOPBACK;
 344         return 0;
 345 }
 346
 347 static int
 348 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
 349 {
 350
 351         return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
 352 }
 353
 354 static int
 355 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
 356 {
 357
 358         return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
 359 }
 360
 361 /* pru_connect2 is EOPNOTSUPP */
 362 /* pru_control is EOPNOTSUPP */
 363
 364 static void
 365 rts_detach(struct socket *so)
 366 {
 367         struct rawcb *rp = sotorawcb(so);
 368
 369         KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
 370
 371         RTSOCK_LOCK();
 372         switch(rp->rcb_proto.sp_protocol) {
 373         case AF_INET:
 374                 V_route_cb.ip_count--;
 375                 break;
 376         case AF_INET6:
 377                 V_route_cb.ip6_count--;
 378                 break;
 379         }
 380         V_route_cb.any_count--;
 381         RTSOCK_UNLOCK();
 382         raw_usrreqs.pru_detach(so);
 383 }
 384
 385 static int
 386 rts_disconnect(struct socket *so)
 387 {
 388
 389         return (raw_usrreqs.pru_disconnect(so));
 390 }
 391
 392 /* pru_listen is EOPNOTSUPP */
 393
 394 static int
 395 rts_peeraddr(struct socket *so, struct sockaddr **nam)
 396 {
 397
 398         return (raw_usrreqs.pru_peeraddr(so, nam));
 399 }
 400
 401 /* pru_rcvd is EOPNOTSUPP */
 402 /* pru_rcvoob is EOPNOTSUPP */
 403
 404 static int
 405 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
 406          struct mbuf *control, struct thread *td)
 407 {
 408
 409         return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
 410 }
 411
 412 /* pru_sense is null */
 413
 414 static int
 415 rts_shutdown(struct socket *so)
 416 {
 417
 418         return (raw_usrreqs.pru_shutdown(so));
 419 }
 420
 421 static int
 422 rts_sockaddr(struct socket *so, struct sockaddr **nam)
 423 {
 424
 425         return (raw_usrreqs.pru_sockaddr(so, nam));
 426 }
 427
 428 static struct pr_usrreqs route_usrreqs = {
 429         .pru_abort =            rts_abort,
 430         .pru_attach =           rts_attach,
 431         .pru_bind =             rts_bind,
 432         .pru_connect =          rts_connect,
 433         .pru_detach =           rts_detach,
 434         .pru_disconnect =       rts_disconnect,
 435         .pru_peeraddr =         rts_peeraddr,
 436         .pru_send =             rts_send,
 437         .pru_shutdown =         rts_shutdown,
 438         .pru_sockaddr =         rts_sockaddr,
 439         .pru_close =            rts_close,
 440 };
 441
 442 #ifndef _SOCKADDR_UNION_DEFINED
 443 #define _SOCKADDR_UNION_DEFINED
 444 /*
 445  * The union of all possible address formats we handle.
 446  */
 447 union sockaddr_union {
 448         struct sockaddr         sa;
 449         struct sockaddr_in      sin;
 450         struct sockaddr_in6     sin6;
 451 };
 452 #endif /* _SOCKADDR_UNION_DEFINED */
 453
 454 static int
 455 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
 456     struct nhop_object *nh, union sockaddr_union *saun, struct ucred *cred)
 457 {
 458 #if defined(INET) || defined(INET6)
 459         struct epoch_tracker et;
 460 #endif
 461
 462         /* First, see if the returned address is part of the jail. */
 463         if (prison_if(cred, nh->nh_ifa->ifa_addr) == 0) {
 464                 info->rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
 465                 return (0);
 466         }
 467
 468         switch (info->rti_info[RTAX_DST]->sa_family) {
 469 #ifdef INET
 470         case AF_INET:
 471         {
 472                 struct in_addr ia;
 473                 struct ifaddr *ifa;
 474                 int found;
 475
 476                 found = 0;
 477                 /*
 478                  * Try to find an address on the given outgoing interface
 479                  * that belongs to the jail.
 480                  */
 481                 NET_EPOCH_ENTER(et);
 482                 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 483                         struct sockaddr *sa;
 484                         sa = ifa->ifa_addr;
 485                         if (sa->sa_family != AF_INET)
 486                                 continue;
 487                         ia = ((struct sockaddr_in *)sa)->sin_addr;
 488                         if (prison_check_ip4(cred, &ia) == 0) {
 489                                 found = 1;
 490                                 break;
 491                         }
 492                 }
 493                 NET_EPOCH_EXIT(et);
 494                 if (!found) {
 495                         /*
 496                          * As a last resort return the 'default' jail address.
 497                          */
 498                         ia = ((struct sockaddr_in *)nh->nh_ifa->ifa_addr)->
 499                             sin_addr;
 500                         if (prison_get_ip4(cred, &ia) != 0)
 501                                 return (ESRCH);
 502                 }
 503                 bzero(&saun->sin, sizeof(struct sockaddr_in));
 504                 saun->sin.sin_len = sizeof(struct sockaddr_in);
 505                 saun->sin.sin_family = AF_INET;
 506                 saun->sin.sin_addr.s_addr = ia.s_addr;
 507                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
 508                 break;
 509         }
 510 #endif
 511 #ifdef INET6
 512         case AF_INET6:
 513         {
 514                 struct in6_addr ia6;
 515                 struct ifaddr *ifa;
 516                 int found;
 517
 518                 found = 0;
 519                 /*
 520                  * Try to find an address on the given outgoing interface
 521                  * that belongs to the jail.
 522                  */
 523                 NET_EPOCH_ENTER(et);
 524                 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 525                         struct sockaddr *sa;
 526                         sa = ifa->ifa_addr;
 527                         if (sa->sa_family != AF_INET6)
 528                                 continue;
 529                         bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
 530                             &ia6, sizeof(struct in6_addr));
 531                         if (prison_check_ip6(cred, &ia6) == 0) {
 532                                 found = 1;
 533                                 break;
 534                         }
 535                 }
 536                 NET_EPOCH_EXIT(et);
 537                 if (!found) {
 538                         /*
 539                          * As a last resort return the 'default' jail address.
 540                          */
 541                         ia6 = ((struct sockaddr_in6 *)nh->nh_ifa->ifa_addr)->
 542                             sin6_addr;
 543                         if (prison_get_ip6(cred, &ia6) != 0)
 544                                 return (ESRCH);
 545                 }
 546                 bzero(&saun->sin6, sizeof(struct sockaddr_in6));
 547                 saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
 548                 saun->sin6.sin6_family = AF_INET6;
 549                 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
 550                 if (sa6_recoverscope(&saun->sin6) != 0)
 551                         return (ESRCH);
 552                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
 553                 break;
 554         }
 555 #endif
 556         default:
 557                 return (ESRCH);
 558         }
 559         return (0);
 560 }
 561
 562 /*
 563  * Fills in @info based on userland-provided @rtm message.
 564  *
 565  * Returns 0 on success.
 566  */
 567 static int
 568 fill_addrinfo(struct rt_msghdr *rtm, int len, u_int fibnum, struct rt_addrinfo *info)
 569 {
 570         int error;
 571         sa_family_t saf;
 572
 573         rtm->rtm_pid = curproc->p_pid;
 574         info->rti_addrs = rtm->rtm_addrs;
 575
 576         info->rti_mflags = rtm->rtm_inits;
 577         info->rti_rmx = &rtm->rtm_rmx;
 578
 579         /*
 580          * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
 581          * link-local address because rtrequest requires addresses with
 582          * embedded scope id.
 583          */
 584         if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info))
 585                 return (EINVAL);
 586
 587         if (rtm->rtm_flags & RTF_RNH_LOCKED)
 588                 return (EINVAL);
 589         info->rti_flags = rtm->rtm_flags;
 590         if (info->rti_info[RTAX_DST] == NULL ||
 591             info->rti_info[RTAX_DST]->sa_family >= AF_MAX ||
 592             (info->rti_info[RTAX_GATEWAY] != NULL &&
 593              info->rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
 594                 return (EINVAL);
 595         saf = info->rti_info[RTAX_DST]->sa_family;
 596         /*
 597          * Verify that the caller has the appropriate privilege; RTM_GET
 598          * is the only operation the non-superuser is allowed.
 599          */
 600         if (rtm->rtm_type != RTM_GET) {
 601                 error = priv_check(curthread, PRIV_NET_ROUTE);
 602                 if (error != 0)
 603                         return (error);
 604         }
 605
 606         /*
 607          * The given gateway address may be an interface address.
 608          * For example, issuing a "route change" command on a route
 609          * entry that was created from a tunnel, and the gateway
 610          * address given is the local end point. In this case the
 611          * RTF_GATEWAY flag must be cleared or the destination will
 612          * not be reachable even though there is no error message.
 613          */
 614         if (info->rti_info[RTAX_GATEWAY] != NULL &&
 615             info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
 616                 struct rt_addrinfo ginfo;
 617                 struct sockaddr *gdst;
 618                 struct sockaddr_storage ss;
 619
 620                 bzero(&ginfo, sizeof(ginfo));
 621                 bzero(&ss, sizeof(ss));
 622                 ss.ss_len = sizeof(ss);
 623
 624                 ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss;
 625                 gdst = info->rti_info[RTAX_GATEWAY];
 626
 627                 /*
 628                  * A host route through the loopback interface is
 629                  * installed for each interface adddress. In pre 8.0
 630                  * releases the interface address of a PPP link type
 631                  * is not reachable locally. This behavior is fixed as
 632                  * part of the new L2/L3 redesign and rewrite work. The
 633                  * signature of this interface address route is the
 634                  * AF_LINK sa_family type of the gateway, and the
 635                  * rt_ifp has the IFF_LOOPBACK flag set.
 636                  */
 637                 if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) {
 638                         if (ss.ss_family == AF_LINK &&
 639                             ginfo.rti_ifp->if_flags & IFF_LOOPBACK) {
 640                                 info->rti_flags &= ~RTF_GATEWAY;
 641                                 info->rti_flags |= RTF_GWFLAG_COMPAT;
 642                         }
 643                         rib_free_info(&ginfo);
 644                 }
 645         }
 646
 647         return (0);
 648 }
 649
 650 /*
 651  * Handles RTM_GET message from routing socket, returning matching rt.
 652  *
 653  * Returns:
 654  * 0 on success, with locked and referenced matching rt in @rt_nrt
 655  * errno of failure
 656  */
 657 static int
 658 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
 659     struct rt_msghdr *rtm, struct rtentry **ret_nrt)
 660 {
 661         RIB_RLOCK_TRACKER;
 662         struct rtentry *rt;
 663         struct rib_head *rnh;
 664         sa_family_t saf;
 665
 666         saf = info->rti_info[RTAX_DST]->sa_family;
 667
 668         rnh = rt_tables_get_rnh(fibnum, saf);
 669         if (rnh == NULL)
 670                 return (EAFNOSUPPORT);
 671
 672         RIB_RLOCK(rnh);
 673
 674         if (info->rti_info[RTAX_NETMASK] == NULL) {
 675                 /*
 676                  * Provide longest prefix match for
 677                  * address lookup (no mask).
 678                  * 'route -n get addr'
 679                  */
 680                 rt = (struct rtentry *) rnh->rnh_matchaddr(
 681                     info->rti_info[RTAX_DST], &rnh->head);
 682         } else
 683                 rt = (struct rtentry *) rnh->rnh_lookup(
 684                     info->rti_info[RTAX_DST],
 685                     info->rti_info[RTAX_NETMASK], &rnh->head);
 686
 687         if (rt == NULL) {
 688                 RIB_RUNLOCK(rnh);
 689                 return (ESRCH);
 690         }
 691 #ifdef RADIX_MPATH
 692         /*
 693          * for RTM_GET, gate is optional even with multipath.
 694          * if gate == NULL the first match is returned.
 695          * (no need to call rt_mpath_matchgate if gate == NULL)
 696          */
 697         if (rt_mpath_capable(rnh) && info->rti_info[RTAX_GATEWAY]) {
 698                 rt = rt_mpath_matchgate(rt, info->rti_info[RTAX_GATEWAY]);
 699                 if (!rt) {
 700                         RIB_RUNLOCK(rnh);
 701                         return (ESRCH);
 702                 }
 703         }
 704 #endif
 705         /*
 706          * If performing proxied L2 entry insertion, and
 707          * the actual PPP host entry is found, perform
 708          * another search to retrieve the prefix route of
 709          * the local end point of the PPP link.
 710          * TODO: move this logic to userland.
 711          */
 712         if (rtm->rtm_flags & RTF_ANNOUNCE) {
 713                 struct sockaddr laddr;
 714                 struct nhop_object *nh;
 715
 716                 nh = rt->rt_nhop;
 717                 if (nh->nh_ifp != NULL &&
 718                     nh->nh_ifp->if_type == IFT_PROPVIRTUAL) {
 719                         struct epoch_tracker et;
 720                         struct ifaddr *ifa;
 721
 722                         NET_EPOCH_ENTER(et);
 723                         ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1,
 724                                         RT_ALL_FIBS);
 725                         NET_EPOCH_EXIT(et);
 726                         if (ifa != NULL)
 727                                 rt_maskedcopy(ifa->ifa_addr,
 728                                               &laddr,
 729                                               ifa->ifa_netmask);
 730                 } else
 731                         rt_maskedcopy(nh->nh_ifa->ifa_addr,
 732                                       &laddr,
 733                                       nh->nh_ifa->ifa_netmask);
 734                 /*
 735                  * refactor rt and no lock operation necessary
 736                  */
 737                 rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr,
 738                     &rnh->head);
 739                 if (rt == NULL) {
 740                         RIB_RUNLOCK(rnh);
 741                         return (ESRCH);
 742                 }
 743         }
 744         RT_LOCK(rt);
 745         RT_ADDREF(rt);
 746         RIB_RUNLOCK(rnh);
 747
 748         *ret_nrt = rt;
 749
 750         return (0);
 751 }
 752
 753 /*
 754  * Update sockaddrs, flags, etc in @prtm based on @rt data.
 755  * Assumes @rt is locked.
 756  * rtm can be reallocated.
 757  *
 758  * Returns 0 on success, along with pointer to (potentially reallocated)
 759  *  rtm.
 760  *
 761  */
 762 static int
 763 update_rtm_from_rte(struct rt_addrinfo *info, struct rt_msghdr **prtm,
 764     int alloc_len, struct rtentry *rt)
 765 {
 766         struct sockaddr_storage netmask_ss;
 767         struct walkarg w;
 768         union sockaddr_union saun;
 769         struct rt_msghdr *rtm, *orig_rtm = NULL;
 770         struct nhop_object *nh;
 771         struct ifnet *ifp;
 772         int error, len;
 773
 774         RT_LOCK_ASSERT(rt);
 775
 776         rtm = *prtm;
 777
 778         nh = rt->rt_nhop;
 779         info->rti_info[RTAX_DST] = rt_key(rt);
 780         info->rti_info[RTAX_GATEWAY] = &nh->gw_sa;
 781         info->rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
 782             rt_mask(rt), &netmask_ss);
 783         info->rti_info[RTAX_GENMASK] = 0;
 784         ifp = nh->nh_ifp;
 785         if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
 786                 if (ifp) {
 787                         info->rti_info[RTAX_IFP] =
 788                             ifp->if_addr->ifa_addr;
 789                         error = rtm_get_jailed(info, ifp, nh,
 790                             &saun, curthread->td_ucred);
 791                         if (error != 0)
 792                                 return (error);
 793                         if (ifp->if_flags & IFF_POINTOPOINT)
 794                                 info->rti_info[RTAX_BRD] =
 795                                     nh->nh_ifa->ifa_dstaddr;
 796                         rtm->rtm_index = ifp->if_index;
 797                 } else {
 798                         info->rti_info[RTAX_IFP] = NULL;
 799                         info->rti_info[RTAX_IFA] = NULL;
 800                 }
 801         } else if (ifp != NULL)
 802                 rtm->rtm_index = ifp->if_index;
 803
 804         /* Check if we need to realloc storage */
 805         rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len);
 806         if (len > alloc_len) {
 807                 struct rt_msghdr *tmp_rtm;
 808
 809                 tmp_rtm = malloc(len, M_TEMP, M_NOWAIT);
 810                 if (tmp_rtm == NULL)
 811                         return (ENOBUFS);
 812                 bcopy(rtm, tmp_rtm, rtm->rtm_msglen);
 813                 orig_rtm = rtm;
 814                 rtm = tmp_rtm;
 815                 alloc_len = len;
 816
 817                 /*
 818                  * Delay freeing original rtm as info contains
 819                  * data referencing it.
 820                  */
 821         }
 822
 823         w.w_tmem = (caddr_t)rtm;
 824         w.w_tmemsize = alloc_len;
 825         rtsock_msg_buffer(rtm->rtm_type, info, &w, &len);
 826
 827         if (rt->rt_flags & RTF_GWFLAG_COMPAT)
 828                 rtm->rtm_flags = RTF_GATEWAY |
 829                         (rt->rt_flags & ~RTF_GWFLAG_COMPAT);
 830         else
 831                 rtm->rtm_flags = rt->rt_flags;
 832         rt_getmetrics(rt, &rtm->rtm_rmx);
 833         rtm->rtm_addrs = info->rti_addrs;
 834
 835         if (orig_rtm != NULL)
 836                 free(orig_rtm, M_TEMP);
 837         *prtm = rtm;
 838
 839         return (0);
 840 }
 841
 842 /*ARGSUSED*/
 843 static int
 844 route_output(struct mbuf *m, struct socket *so, ...)
 845 {
 846         struct rt_msghdr *rtm = NULL;
 847         struct rtentry *rt = NULL;
 848         struct rt_addrinfo info;
 849         struct epoch_tracker et;
 850 #ifdef INET6
 851         struct sockaddr_storage ss;
 852         struct sockaddr_in6 *sin6;
 853         int i, rti_need_deembed = 0;
 854 #endif
 855         int alloc_len = 0, len, error = 0, fibnum;
 856         sa_family_t saf = AF_UNSPEC;
 857         struct walkarg w;
 858
 859         fibnum = so->so_fibnum;
 860
 861 #define senderr(e) { error = e; goto flush;}
 862         if (m == NULL || ((m->m_len < sizeof(long)) &&
 863                        (m = m_pullup(m, sizeof(long))) == NULL))
 864                 return (ENOBUFS);
 865         if ((m->m_flags & M_PKTHDR) == 0)
 866                 panic("route_output");
 867         NET_EPOCH_ENTER(et);
 868         len = m->m_pkthdr.len;
 869         if (len < sizeof(*rtm) ||
 870             len != mtod(m, struct rt_msghdr *)->rtm_msglen)
 871                 senderr(EINVAL);
 872
 873         /*
 874          * Most of current messages are in range 200-240 bytes,
 875          * minimize possible re-allocation on reply using larger size
 876          * buffer aligned on 1k boundaty.
 877          */
 878         alloc_len = roundup2(len, 1024);
 879         if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL)
 880                 senderr(ENOBUFS);
 881
 882         m_copydata(m, 0, len, (caddr_t)rtm);
 883         bzero(&info, sizeof(info));
 884         bzero(&w, sizeof(w));
 885
 886         if (rtm->rtm_version != RTM_VERSION) {
 887                 /* Do not touch message since format is unknown */
 888                 free(rtm, M_TEMP);
 889                 rtm = NULL;
 890                 senderr(EPROTONOSUPPORT);
 891         }
 892
 893         /*
 894          * Starting from here, it is possible
 895          * to alter original message and insert
 896          * caller PID and error value.
 897          */
 898
 899         if ((error = fill_addrinfo(rtm, len, fibnum, &info)) != 0) {
 900                 senderr(error);
 901         }
 902
 903         saf = info.rti_info[RTAX_DST]->sa_family;
 904
 905         /* support for new ARP code */
 906         if (rtm->rtm_flags & RTF_LLDATA) {
 907                 error = lla_rt_output(rtm, &info);
 908 #ifdef INET6
 909                 if (error == 0)
 910                         rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 911 #endif
 912                 goto flush;
 913         }
 914
 915         switch (rtm->rtm_type) {
 916                 struct rtentry *saved_nrt;
 917
 918         case RTM_ADD:
 919         case RTM_CHANGE:
 920                 if (rtm->rtm_type == RTM_ADD) {
 921                         if (info.rti_info[RTAX_GATEWAY] == NULL)
 922                                 senderr(EINVAL);
 923                 }
 924                 saved_nrt = NULL;
 925                 error = rtrequest1_fib(rtm->rtm_type, &info, &saved_nrt,
 926                     fibnum);
 927                 if (error == 0 && saved_nrt != NULL) {
 928 #ifdef INET6
 929                         rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 930 #endif
 931                         RT_LOCK(saved_nrt);
 932                         rtm->rtm_index = saved_nrt->rt_nhop->nh_ifp->if_index;
 933                         RT_REMREF(saved_nrt);
 934                         RT_UNLOCK(saved_nrt);
 935                 }
 936                 break;
 937
 938         case RTM_DELETE:
 939                 saved_nrt = NULL;
 940                 error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt, fibnum);
 941                 if (error == 0) {
 942                         RT_LOCK(saved_nrt);
 943                         rt = saved_nrt;
 944                         goto report;
 945                 }
 946 #ifdef INET6
 947                 /* rt_msg2() will not be used when RTM_DELETE fails. */
 948                 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 949 #endif
 950                 break;
 951
 952         case RTM_GET:
 953                 error = handle_rtm_get(&info, fibnum, rtm, &rt);
 954                 if (error != 0)
 955                         senderr(error);
 956
 957 report:
 958                 RT_LOCK_ASSERT(rt);
 959                 if (!can_export_rte(curthread->td_ucred, rt)) {
 960                         RT_UNLOCK(rt);
 961                         senderr(ESRCH);
 962                 }
 963                 error = update_rtm_from_rte(&info, &rtm, alloc_len, rt);
 964                 /*
 965                  * Note that some sockaddr pointers may have changed to
 966                  * point to memory outsize @rtm. Some may be pointing
 967                  * to the on-stack variables.
 968                  * Given that, any pointer in @info CANNOT BE USED.
 969                  */
 970
 971                 /*
 972                  * scopeid deembedding has been performed while
 973                  * writing updated rtm in rtsock_msg_buffer().
 974                  * With that in mind, skip deembedding procedure below.
 975                  */
 976 #ifdef INET6
 977                 rti_need_deembed = 0;
 978 #endif
 979                 RT_UNLOCK(rt);
 980                 if (error != 0)
 981                         senderr(error);
 982                 break;
 983
 984         default:
 985                 senderr(EOPNOTSUPP);
 986         }
 987
 988 flush:
 989         NET_EPOCH_EXIT(et);
 990         if (rt != NULL)
 991                 RTFREE(rt);
 992
 993 #ifdef INET6
 994         if (rtm != NULL) {
 995                 if (rti_need_deembed) {
 996                         /* sin6_scope_id is recovered before sending rtm. */
 997                         sin6 = (struct sockaddr_in6 *)&ss;
 998                         for (i = 0; i < RTAX_MAX; i++) {
 999                                 if (info.rti_info[i] == NULL)
1000                                         continue;
1001                                 if (info.rti_info[i]->sa_family != AF_INET6)
1002                                         continue;
1003                                 bcopy(info.rti_info[i], sin6, sizeof(*sin6));
1004                                 if (sa6_recoverscope(sin6) == 0)
1005                                         bcopy(sin6, info.rti_info[i],
1006                                                     sizeof(*sin6));
1007                         }
1008                 }
1009         }
1010 #endif
1011         send_rtm_reply(so, rtm, m, saf, fibnum, error);
1012
1013         return (error);
1014 }
1015
1016 /*
1017  * Sends the prepared reply message in @rtm to all rtsock clients.
1018  * Frees @m and @rtm.
1019  *
1020  */
1021 static void
1022 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m,
1023     sa_family_t saf, u_int fibnum, int rtm_errno)
1024 {
1025         struct rawcb *rp = NULL;
1026
1027         /*
1028          * Check to see if we don't want our own messages.
1029          */
1030         if ((so->so_options & SO_USELOOPBACK) == 0) {
1031                 if (V_route_cb.any_count <= 1) {
1032                         if (rtm != NULL)
1033                                 free(rtm, M_TEMP);
1034                         m_freem(m);
1035                         return;
1036                 }
1037                 /* There is another listener, so construct message */
1038                 rp = sotorawcb(so);
1039         }
1040
1041         if (rtm != NULL) {
1042                 if (rtm_errno!= 0)
1043                         rtm->rtm_errno = rtm_errno;
1044                 else
1045                         rtm->rtm_flags |= RTF_DONE;
1046
1047                 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
1048                 if (m->m_pkthdr.len < rtm->rtm_msglen) {
1049                         m_freem(m);
1050                         m = NULL;
1051                 } else if (m->m_pkthdr.len > rtm->rtm_msglen)
1052                         m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
1053
1054                 free(rtm, M_TEMP);
1055         }
1056         if (m != NULL) {
1057                 M_SETFIB(m, fibnum);
1058                 m->m_flags |= RTS_FILTER_FIB;
1059                 if (rp) {
1060                         /*
1061                          * XXX insure we don't get a copy by
1062                          * invalidating our protocol
1063                          */
1064                         unsigned short family = rp->rcb_proto.sp_family;
1065                         rp->rcb_proto.sp_family = 0;
1066                         rt_dispatch(m, saf);
1067                         rp->rcb_proto.sp_family = family;
1068                 } else
1069                         rt_dispatch(m, saf);
1070         }
1071 }
1072
1073
1074 static void
1075 rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out)
1076 {
1077
1078         bzero(out, sizeof(*out));
1079         out->rmx_mtu = rt->rt_nhop->nh_mtu;
1080         out->rmx_weight = rt->rt_weight;
1081         out->rmx_nhidx = nhop_get_idx(rt->rt_nhop);
1082         /* Kernel -> userland timebase conversion. */
1083         out->rmx_expire = rt->rt_expire ?
1084             rt->rt_expire - time_uptime + time_second : 0;
1085 }
1086
1087 /*
1088  * Extract the addresses of the passed sockaddrs.
1089  * Do a little sanity checking so as to avoid bad memory references.
1090  * This data is derived straight from userland.
1091  */
1092 static int
1093 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1094 {
1095         struct sockaddr *sa;
1096         int i;
1097
1098         for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1099                 if ((rtinfo->rti_addrs & (1 << i)) == 0)
1100                         continue;
1101                 sa = (struct sockaddr *)cp;
1102                 /*
1103                  * It won't fit.
1104                  */
1105                 if (cp + sa->sa_len > cplim)
1106                         return (EINVAL);
1107                 /*
1108                  * there are no more.. quit now
1109                  * If there are more bits, they are in error.
1110                  * I've seen this. route(1) can evidently generate these.
1111                  * This causes kernel to core dump.
1112                  * for compatibility, If we see this, point to a safe address.
1113                  */
1114                 if (sa->sa_len == 0) {
1115                         rtinfo->rti_info[i] = &sa_zero;
1116                         return (0); /* should be EINVAL but for compat */
1117                 }
1118                 /* accept it */
1119 #ifdef INET6
1120                 if (sa->sa_family == AF_INET6)
1121                         sa6_embedscope((struct sockaddr_in6 *)sa,
1122                             V_ip6_use_defzone);
1123 #endif
1124                 rtinfo->rti_info[i] = sa;
1125                 cp += SA_SIZE(sa);
1126         }
1127         return (0);
1128 }
1129
1130 /*
1131  * Fill in @dmask with valid netmask leaving original @smask
1132  * intact. Mostly used with radix netmasks.
1133  */
1134 struct sockaddr *
1135 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask,
1136     struct sockaddr_storage *dmask)
1137 {
1138         if (dst == NULL || smask == NULL)
1139                 return (NULL);
1140
1141         memset(dmask, 0, dst->sa_len);
1142         memcpy(dmask, smask, smask->sa_len);
1143         dmask->ss_len = dst->sa_len;
1144         dmask->ss_family = dst->sa_family;
1145
1146         return ((struct sockaddr *)dmask);
1147 }
1148
1149 /*
1150  * Writes information related to @rtinfo object to newly-allocated mbuf.
1151  * Assumes MCLBYTES is enough to construct any message.
1152  * Used for OS notifications of vaious events (if/ifa announces,etc)
1153  *
1154  * Returns allocated mbuf or NULL on failure.
1155  */
1156 static struct mbuf *
1157 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1158 {
1159         struct rt_msghdr *rtm;
1160         struct mbuf *m;
1161         int i;
1162         struct sockaddr *sa;
1163 #ifdef INET6
1164         struct sockaddr_storage ss;
1165         struct sockaddr_in6 *sin6;
1166 #endif
1167         int len, dlen;
1168
1169         switch (type) {
1170
1171         case RTM_DELADDR:
1172         case RTM_NEWADDR:
1173                 len = sizeof(struct ifa_msghdr);
1174                 break;
1175
1176         case RTM_DELMADDR:
1177         case RTM_NEWMADDR:
1178                 len = sizeof(struct ifma_msghdr);
1179                 break;
1180
1181         case RTM_IFINFO:
1182                 len = sizeof(struct if_msghdr);
1183                 break;
1184
1185         case RTM_IFANNOUNCE:
1186         case RTM_IEEE80211:
1187                 len = sizeof(struct if_announcemsghdr);
1188                 break;
1189
1190         default:
1191                 len = sizeof(struct rt_msghdr);
1192         }
1193
1194         /* XXXGL: can we use MJUMPAGESIZE cluster here? */
1195         KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1196         if (len > MHLEN)
1197                 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1198         else
1199                 m = m_gethdr(M_NOWAIT, MT_DATA);
1200         if (m == NULL)
1201                 return (m);
1202
1203         m->m_pkthdr.len = m->m_len = len;
1204         rtm = mtod(m, struct rt_msghdr *);
1205         bzero((caddr_t)rtm, len);
1206         for (i = 0; i < RTAX_MAX; i++) {
1207                 if ((sa = rtinfo->rti_info[i]) == NULL)
1208                         continue;
1209                 rtinfo->rti_addrs |= (1 << i);
1210                 dlen = SA_SIZE(sa);
1211 #ifdef INET6
1212                 if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1213                         sin6 = (struct sockaddr_in6 *)&ss;
1214                         bcopy(sa, sin6, sizeof(*sin6));
1215                         if (sa6_recoverscope(sin6) == 0)
1216                                 sa = (struct sockaddr *)sin6;
1217                 }
1218 #endif
1219                 m_copyback(m, len, dlen, (caddr_t)sa);
1220                 len += dlen;
1221         }
1222         if (m->m_pkthdr.len != len) {
1223                 m_freem(m);
1224                 return (NULL);
1225         }
1226         rtm->rtm_msglen = len;
1227         rtm->rtm_version = RTM_VERSION;
1228         rtm->rtm_type = type;
1229         return (m);
1230 }
1231
1232 /*
1233  * Writes information related to @rtinfo object to preallocated buffer.
1234  * Stores needed size in @plen. If @w is NULL, calculates size without
1235  * writing.
1236  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1237  *
1238  * Returns 0 on success.
1239  *
1240  */
1241 static int
1242 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1243 {
1244         int i;
1245         int len, buflen = 0, dlen;
1246         caddr_t cp = NULL;
1247         struct rt_msghdr *rtm = NULL;
1248 #ifdef INET6
1249         struct sockaddr_storage ss;
1250         struct sockaddr_in6 *sin6;
1251 #endif
1252 #ifdef COMPAT_FREEBSD32
1253         bool compat32 = false;
1254 #endif
1255
1256         switch (type) {
1257
1258         case RTM_DELADDR:
1259         case RTM_NEWADDR:
1260                 if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1261 #ifdef COMPAT_FREEBSD32
1262                         if (w->w_req->flags & SCTL_MASK32) {
1263                                 len = sizeof(struct ifa_msghdrl32);
1264                                 compat32 = true;
1265                         } else
1266 #endif
1267                                 len = sizeof(struct ifa_msghdrl);
1268                 } else
1269                         len = sizeof(struct ifa_msghdr);
1270                 break;
1271
1272         case RTM_IFINFO:
1273 #ifdef COMPAT_FREEBSD32
1274                 if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1275                         if (w->w_op == NET_RT_IFLISTL)
1276                                 len = sizeof(struct if_msghdrl32);
1277                         else
1278                                 len = sizeof(struct if_msghdr32);
1279                         compat32 = true;
1280                         break;
1281                 }
1282 #endif
1283                 if (w != NULL && w->w_op == NET_RT_IFLISTL)
1284                         len = sizeof(struct if_msghdrl);
1285                 else
1286                         len = sizeof(struct if_msghdr);
1287                 break;
1288
1289         case RTM_NEWMADDR:
1290                 len = sizeof(struct ifma_msghdr);
1291                 break;
1292
1293         default:
1294                 len = sizeof(struct rt_msghdr);
1295         }
1296
1297         if (w != NULL) {
1298                 rtm = (struct rt_msghdr *)w->w_tmem;
1299                 buflen = w->w_tmemsize - len;
1300                 cp = (caddr_t)w->w_tmem + len;
1301         }
1302
1303         rtinfo->rti_addrs = 0;
1304         for (i = 0; i < RTAX_MAX; i++) {
1305                 struct sockaddr *sa;
1306
1307                 if ((sa = rtinfo->rti_info[i]) == NULL)
1308                         continue;
1309                 rtinfo->rti_addrs |= (1 << i);
1310 #ifdef COMPAT_FREEBSD32
1311                 if (compat32)
1312                         dlen = SA_SIZE32(sa);
1313                 else
1314 #endif
1315                         dlen = SA_SIZE(sa);
1316                 if (cp != NULL && buflen >= dlen) {
1317 #ifdef INET6
1318                         if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1319                                 sin6 = (struct sockaddr_in6 *)&ss;
1320                                 bcopy(sa, sin6, sizeof(*sin6));
1321                                 if (sa6_recoverscope(sin6) == 0)
1322                                         sa = (struct sockaddr *)sin6;
1323                         }
1324 #endif
1325                         bcopy((caddr_t)sa, cp, (unsigned)dlen);
1326                         cp += dlen;
1327                         buflen -= dlen;
1328                 } else if (cp != NULL) {
1329                         /*
1330                          * Buffer too small. Count needed size
1331                          * and return with error.
1332                          */
1333                         cp = NULL;
1334                 }
1335
1336                 len += dlen;
1337         }
1338
1339         if (cp != NULL) {
1340                 dlen = ALIGN(len) - len;
1341                 if (buflen < dlen)
1342                         cp = NULL;
1343                 else {
1344                         bzero(cp, dlen);
1345                         cp += dlen;
1346                         buflen -= dlen;
1347                 }
1348         }
1349         len = ALIGN(len);
1350
1351         if (cp != NULL) {
1352                 /* fill header iff buffer is large enough */
1353                 rtm->rtm_version = RTM_VERSION;
1354                 rtm->rtm_type = type;
1355                 rtm->rtm_msglen = len;
1356         }
1357
1358         *plen = len;
1359
1360         if (w != NULL && cp == NULL)
1361                 return (ENOBUFS);
1362
1363         return (0);
1364 }
1365
1366 /*
1367  * This routine is called to generate a message from the routing
1368  * socket indicating that a redirect has occurred, a routing lookup
1369  * has failed, or that a protocol has detected timeouts to a particular
1370  * destination.
1371  */
1372 void
1373 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1374     int fibnum)
1375 {
1376         struct rt_msghdr *rtm;
1377         struct mbuf *m;
1378         struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1379
1380         if (V_route_cb.any_count == 0)
1381                 return;
1382         m = rtsock_msg_mbuf(type, rtinfo);
1383         if (m == NULL)
1384                 return;
1385
1386         if (fibnum != RT_ALL_FIBS) {
1387                 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1388                     "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1389                 M_SETFIB(m, fibnum);
1390                 m->m_flags |= RTS_FILTER_FIB;
1391         }
1392
1393         rtm = mtod(m, struct rt_msghdr *);
1394         rtm->rtm_flags = RTF_DONE | flags;
1395         rtm->rtm_errno = error;
1396         rtm->rtm_addrs = rtinfo->rti_addrs;
1397         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1398 }
1399
1400 void
1401 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1402 {
1403
1404         rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1405 }
1406
1407 /*
1408  * This routine is called to generate a message from the routing
1409  * socket indicating that the status of a network interface has changed.
1410  */
1411 void
1412 rt_ifmsg(struct ifnet *ifp)
1413 {
1414         struct if_msghdr *ifm;
1415         struct mbuf *m;
1416         struct rt_addrinfo info;
1417
1418         if (V_route_cb.any_count == 0)
1419                 return;
1420         bzero((caddr_t)&info, sizeof(info));
1421         m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1422         if (m == NULL)
1423                 return;
1424         ifm = mtod(m, struct if_msghdr *);
1425         ifm->ifm_index = ifp->if_index;
1426         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1427         if_data_copy(ifp, &ifm->ifm_data);
1428         ifm->ifm_addrs = 0;
1429         rt_dispatch(m, AF_UNSPEC);
1430 }
1431
1432 /*
1433  * Announce interface address arrival/withdraw.
1434  * Please do not call directly, use rt_addrmsg().
1435  * Assume input data to be valid.
1436  * Returns 0 on success.
1437  */
1438 int
1439 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1440 {
1441         struct rt_addrinfo info;
1442         struct sockaddr *sa;
1443         int ncmd;
1444         struct mbuf *m;
1445         struct ifa_msghdr *ifam;
1446         struct ifnet *ifp = ifa->ifa_ifp;
1447         struct sockaddr_storage ss;
1448
1449         if (V_route_cb.any_count == 0)
1450                 return (0);
1451
1452         ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1453
1454         bzero((caddr_t)&info, sizeof(info));
1455         info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1456         info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1457         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1458             info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss);
1459         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1460         if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1461                 return (ENOBUFS);
1462         ifam = mtod(m, struct ifa_msghdr *);
1463         ifam->ifam_index = ifp->if_index;
1464         ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1465         ifam->ifam_flags = ifa->ifa_flags;
1466         ifam->ifam_addrs = info.rti_addrs;
1467
1468         if (fibnum != RT_ALL_FIBS) {
1469                 M_SETFIB(m, fibnum);
1470                 m->m_flags |= RTS_FILTER_FIB;
1471         }
1472
1473         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1474
1475         return (0);
1476 }
1477
1478 /*
1479  * Announce route addition/removal to rtsock based on @rt data.
1480  * Callers are advives to use rt_routemsg() instead of using this
1481  *  function directly.
1482  * Assume @rt data is consistent.
1483  *
1484  * Returns 0 on success.
1485  */
1486 int
1487 rtsock_routemsg(int cmd, struct rtentry *rt, struct ifnet *ifp, int rti_addrs,
1488     int fibnum)
1489 {
1490         struct sockaddr_storage ss;
1491         struct rt_addrinfo info;
1492
1493         if (V_route_cb.any_count == 0)
1494                 return (0);
1495
1496         bzero((caddr_t)&info, sizeof(info));
1497         info.rti_info[RTAX_DST] = rt_key(rt);
1498         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt), rt_mask(rt), &ss);
1499         info.rti_info[RTAX_GATEWAY] = &rt->rt_nhop->gw_sa;
1500         info.rti_flags = rt->rt_flags;
1501         info.rti_ifp = ifp;
1502
1503         return (rtsock_routemsg_info(cmd, &info, fibnum));
1504 }
1505
1506 int
1507 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
1508 {
1509         struct rt_msghdr *rtm;
1510         struct sockaddr *sa;
1511         struct mbuf *m;
1512
1513         if (V_route_cb.any_count == 0)
1514                 return (0);
1515
1516         if (info->rti_flags & RTF_HOST)
1517                 info->rti_info[RTAX_NETMASK] = NULL;
1518
1519         m = rtsock_msg_mbuf(cmd, info);
1520         if (m == NULL)
1521                 return (ENOBUFS);
1522
1523         if (fibnum != RT_ALL_FIBS) {
1524                 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1525                     "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1526                 M_SETFIB(m, fibnum);
1527                 m->m_flags |= RTS_FILTER_FIB;
1528         }
1529
1530         rtm = mtod(m, struct rt_msghdr *);
1531         rtm->rtm_addrs = info->rti_addrs;
1532         if (info->rti_ifp != NULL)
1533                 rtm->rtm_index = info->rti_ifp->if_index;
1534         /* Add RTF_DONE to indicate command 'completion' required by API */
1535         info->rti_flags |= RTF_DONE;
1536         /* Reported routes has to be up */
1537         if (cmd == RTM_ADD || cmd == RTM_CHANGE)
1538                 info->rti_flags |= RTF_UP;
1539         rtm->rtm_flags = info->rti_flags;
1540
1541         sa = info->rti_info[RTAX_DST];
1542         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1543
1544         return (0);
1545 }
1546
1547 /*
1548  * This is the analogue to the rt_newaddrmsg which performs the same
1549  * function but for multicast group memberhips.  This is easier since
1550  * there is no route state to worry about.
1551  */
1552 void
1553 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1554 {
1555         struct rt_addrinfo info;
1556         struct mbuf *m = NULL;
1557         struct ifnet *ifp = ifma->ifma_ifp;
1558         struct ifma_msghdr *ifmam;
1559
1560         if (V_route_cb.any_count == 0)
1561                 return;
1562
1563         bzero((caddr_t)&info, sizeof(info));
1564         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1565         if (ifp && ifp->if_addr)
1566                 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1567         else
1568                 info.rti_info[RTAX_IFP] = NULL;
1569         /*
1570          * If a link-layer address is present, present it as a ``gateway''
1571          * (similarly to how ARP entries, e.g., are presented).
1572          */
1573         info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1574         m = rtsock_msg_mbuf(cmd, &info);
1575         if (m == NULL)
1576                 return;
1577         ifmam = mtod(m, struct ifma_msghdr *);
1578         KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1579             __func__));
1580         ifmam->ifmam_index = ifp->if_index;
1581         ifmam->ifmam_addrs = info.rti_addrs;
1582         rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1583 }
1584
1585 static struct mbuf *
1586 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1587         struct rt_addrinfo *info)
1588 {
1589         struct if_announcemsghdr *ifan;
1590         struct mbuf *m;
1591
1592         if (V_route_cb.any_count == 0)
1593                 return NULL;
1594         bzero((caddr_t)info, sizeof(*info));
1595         m = rtsock_msg_mbuf(type, info);
1596         if (m != NULL) {
1597                 ifan = mtod(m, struct if_announcemsghdr *);
1598                 ifan->ifan_index = ifp->if_index;
1599                 strlcpy(ifan->ifan_name, ifp->if_xname,
1600                         sizeof(ifan->ifan_name));
1601                 ifan->ifan_what = what;
1602         }
1603         return m;
1604 }
1605
1606 /*
1607  * This is called to generate routing socket messages indicating
1608  * IEEE80211 wireless events.
1609  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1610  */
1611 void
1612 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1613 {
1614         struct mbuf *m;
1615         struct rt_addrinfo info;
1616
1617         m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1618         if (m != NULL) {
1619                 /*
1620                  * Append the ieee80211 data.  Try to stick it in the
1621                  * mbuf containing the ifannounce msg; otherwise allocate
1622                  * a new mbuf and append.
1623                  *
1624                  * NB: we assume m is a single mbuf.
1625                  */
1626                 if (data_len > M_TRAILINGSPACE(m)) {
1627                         struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1628                         if (n == NULL) {
1629                                 m_freem(m);
1630                                 return;
1631                         }
1632                         bcopy(data, mtod(n, void *), data_len);
1633                         n->m_len = data_len;
1634                         m->m_next = n;
1635                 } else if (data_len > 0) {
1636                         bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1637                         m->m_len += data_len;
1638                 }
1639                 if (m->m_flags & M_PKTHDR)
1640                         m->m_pkthdr.len += data_len;
1641                 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1642                 rt_dispatch(m, AF_UNSPEC);
1643         }
1644 }
1645
1646 /*
1647  * This is called to generate routing socket messages indicating
1648  * network interface arrival and departure.
1649  */
1650 void
1651 rt_ifannouncemsg(struct ifnet *ifp, int what)
1652 {
1653         struct mbuf *m;
1654         struct rt_addrinfo info;
1655
1656         m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1657         if (m != NULL)
1658                 rt_dispatch(m, AF_UNSPEC);
1659 }
1660
1661 static void
1662 rt_dispatch(struct mbuf *m, sa_family_t saf)
1663 {
1664         struct m_tag *tag;
1665
1666         /*
1667          * Preserve the family from the sockaddr, if any, in an m_tag for
1668          * use when injecting the mbuf into the routing socket buffer from
1669          * the netisr.
1670          */
1671         if (saf != AF_UNSPEC) {
1672                 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1673                     M_NOWAIT);
1674                 if (tag == NULL) {
1675                         m_freem(m);
1676                         return;
1677                 }
1678                 *(unsigned short *)(tag + 1) = saf;
1679                 m_tag_prepend(m, tag);
1680         }
1681 #ifdef VIMAGE
1682         if (V_loif)
1683                 m->m_pkthdr.rcvif = V_loif;
1684         else {
1685                 m_freem(m);
1686                 return;
1687         }
1688 #endif
1689         netisr_queue(NETISR_ROUTE, m);  /* mbuf is free'd on failure. */
1690 }
1691
1692 /*
1693  * Checks if rte can be exported v.r.t jails/vnets.
1694  *
1695  * Returns 1 if it can, 0 otherwise.
1696  */
1697 static int
1698 can_export_rte(struct ucred *td_ucred, const struct rtentry *rt)
1699 {
1700
1701         if ((rt->rt_flags & RTF_HOST) == 0
1702             ? jailed_without_vnet(td_ucred)
1703             : prison_if(td_ucred, rt_key_const(rt)) != 0)
1704                 return (0);
1705         return (1);
1706 }
1707
1708 /*
1709  * This is used in dumping the kernel table via sysctl().
1710  */
1711 static int
1712 sysctl_dumpentry(struct radix_node *rn, void *vw)
1713 {
1714         struct walkarg *w = vw;
1715         struct rtentry *rt = (struct rtentry *)rn;
1716         struct nhop_object *nh;
1717         int error = 0, size;
1718         struct rt_addrinfo info;
1719         struct sockaddr_storage ss;
1720
1721         NET_EPOCH_ASSERT();
1722
1723         if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1724                 return 0;
1725         if (!can_export_rte(w->w_req->td->td_ucred, rt))
1726                 return (0);
1727         bzero((caddr_t)&info, sizeof(info));
1728         info.rti_info[RTAX_DST] = rt_key(rt);
1729         info.rti_info[RTAX_GATEWAY] = &rt->rt_nhop->gw_sa;
1730         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
1731             rt_mask(rt), &ss);
1732         info.rti_info[RTAX_GENMASK] = 0;
1733         nh = rt->rt_nhop;
1734         if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) {
1735                 info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr;
1736                 info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
1737                 if (nh->nh_ifp->if_flags & IFF_POINTOPOINT)
1738                         info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr;
1739         }
1740         if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
1741                 return (error);
1742         if (w->w_req && w->w_tmem) {
1743                 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1744
1745                 bzero(&rtm->rtm_index,
1746                     sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
1747                 if (rt->rt_flags & RTF_GWFLAG_COMPAT)
1748                         rtm->rtm_flags = RTF_GATEWAY |
1749                                 (rt->rt_flags & ~RTF_GWFLAG_COMPAT);
1750                 else
1751                         rtm->rtm_flags = rt->rt_flags;
1752                 rtm->rtm_flags |= nhop_get_rtflags(nh);
1753                 rt_getmetrics(rt, &rtm->rtm_rmx);
1754                 rtm->rtm_index = nh->nh_ifp->if_index;
1755                 rtm->rtm_addrs = info.rti_addrs;
1756                 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1757                 return (error);
1758         }
1759         return (error);
1760 }
1761
1762 static int
1763 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
1764     struct rt_addrinfo *info, struct walkarg *w, int len)
1765 {
1766         struct if_msghdrl *ifm;
1767         struct if_data *ifd;
1768
1769         ifm = (struct if_msghdrl *)w->w_tmem;
1770
1771 #ifdef COMPAT_FREEBSD32
1772         if (w->w_req->flags & SCTL_MASK32) {
1773                 struct if_msghdrl32 *ifm32;
1774
1775                 ifm32 = (struct if_msghdrl32 *)ifm;
1776                 ifm32->ifm_addrs = info->rti_addrs;
1777                 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1778                 ifm32->ifm_index = ifp->if_index;
1779                 ifm32->_ifm_spare1 = 0;
1780                 ifm32->ifm_len = sizeof(*ifm32);
1781                 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1782                 ifm32->_ifm_spare2 = 0;
1783                 ifd = &ifm32->ifm_data;
1784         } else
1785 #endif
1786         {
1787                 ifm->ifm_addrs = info->rti_addrs;
1788                 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1789                 ifm->ifm_index = ifp->if_index;
1790                 ifm->_ifm_spare1 = 0;
1791                 ifm->ifm_len = sizeof(*ifm);
1792                 ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1793                 ifm->_ifm_spare2 = 0;
1794                 ifd = &ifm->ifm_data;
1795         }
1796
1797         memcpy(ifd, src_ifd, sizeof(*ifd));
1798
1799         return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1800 }
1801
1802 static int
1803 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
1804     struct rt_addrinfo *info, struct walkarg *w, int len)
1805 {
1806         struct if_msghdr *ifm;
1807         struct if_data *ifd;
1808
1809         ifm = (struct if_msghdr *)w->w_tmem;
1810
1811 #ifdef COMPAT_FREEBSD32
1812         if (w->w_req->flags & SCTL_MASK32) {
1813                 struct if_msghdr32 *ifm32;
1814
1815                 ifm32 = (struct if_msghdr32 *)ifm;
1816                 ifm32->ifm_addrs = info->rti_addrs;
1817                 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1818                 ifm32->ifm_index = ifp->if_index;
1819                 ifm32->_ifm_spare1 = 0;
1820                 ifd = &ifm32->ifm_data;
1821         } else
1822 #endif
1823         {
1824                 ifm->ifm_addrs = info->rti_addrs;
1825                 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1826                 ifm->ifm_index = ifp->if_index;
1827                 ifm->_ifm_spare1 = 0;
1828                 ifd = &ifm->ifm_data;
1829         }
1830
1831         memcpy(ifd, src_ifd, sizeof(*ifd));
1832
1833         return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1834 }
1835
1836 static int
1837 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1838     struct walkarg *w, int len)
1839 {
1840         struct ifa_msghdrl *ifam;
1841         struct if_data *ifd;
1842
1843         ifam = (struct ifa_msghdrl *)w->w_tmem;
1844
1845 #ifdef COMPAT_FREEBSD32
1846         if (w->w_req->flags & SCTL_MASK32) {
1847                 struct ifa_msghdrl32 *ifam32;
1848
1849                 ifam32 = (struct ifa_msghdrl32 *)ifam;
1850                 ifam32->ifam_addrs = info->rti_addrs;
1851                 ifam32->ifam_flags = ifa->ifa_flags;
1852                 ifam32->ifam_index = ifa->ifa_ifp->if_index;
1853                 ifam32->_ifam_spare1 = 0;
1854                 ifam32->ifam_len = sizeof(*ifam32);
1855                 ifam32->ifam_data_off =
1856                     offsetof(struct ifa_msghdrl32, ifam_data);
1857                 ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
1858                 ifd = &ifam32->ifam_data;
1859         } else
1860 #endif
1861         {
1862                 ifam->ifam_addrs = info->rti_addrs;
1863                 ifam->ifam_flags = ifa->ifa_flags;
1864                 ifam->ifam_index = ifa->ifa_ifp->if_index;
1865                 ifam->_ifam_spare1 = 0;
1866                 ifam->ifam_len = sizeof(*ifam);
1867                 ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1868                 ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1869                 ifd = &ifam->ifam_data;
1870         }
1871
1872         bzero(ifd, sizeof(*ifd));
1873         ifd->ifi_datalen = sizeof(struct if_data);
1874         ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
1875         ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
1876         ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
1877         ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
1878
1879         /* Fixup if_data carp(4) vhid. */
1880         if (carp_get_vhid_p != NULL)
1881                 ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
1882
1883         return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1884 }
1885
1886 static int
1887 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1888     struct walkarg *w, int len)
1889 {
1890         struct ifa_msghdr *ifam;
1891
1892         ifam = (struct ifa_msghdr *)w->w_tmem;
1893         ifam->ifam_addrs = info->rti_addrs;
1894         ifam->ifam_flags = ifa->ifa_flags;
1895         ifam->ifam_index = ifa->ifa_ifp->if_index;
1896         ifam->_ifam_spare1 = 0;
1897         ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1898
1899         return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1900 }
1901
1902 static int
1903 sysctl_iflist(int af, struct walkarg *w)
1904 {
1905         struct ifnet *ifp;
1906         struct ifaddr *ifa;
1907         struct if_data ifd;
1908         struct rt_addrinfo info;
1909         int len, error = 0;
1910         struct sockaddr_storage ss;
1911
1912         bzero((caddr_t)&info, sizeof(info));
1913         bzero(&ifd, sizeof(ifd));
1914         CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1915                 if (w->w_arg && w->w_arg != ifp->if_index)
1916                         continue;
1917                 if_data_copy(ifp, &ifd);
1918                 ifa = ifp->if_addr;
1919                 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1920                 error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
1921                 if (error != 0)
1922                         goto done;
1923                 info.rti_info[RTAX_IFP] = NULL;
1924                 if (w->w_req && w->w_tmem) {
1925                         if (w->w_op == NET_RT_IFLISTL)
1926                                 error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
1927                                     len);
1928                         else
1929                                 error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
1930                                     len);
1931                         if (error)
1932                                 goto done;
1933                 }
1934                 while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
1935                         if (af && af != ifa->ifa_addr->sa_family)
1936                                 continue;
1937                         if (prison_if(w->w_req->td->td_ucred,
1938                             ifa->ifa_addr) != 0)
1939                                 continue;
1940                         info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1941                         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1942                             ifa->ifa_addr, ifa->ifa_netmask, &ss);
1943                         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1944                         error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
1945                         if (error != 0)
1946                                 goto done;
1947                         if (w->w_req && w->w_tmem) {
1948                                 if (w->w_op == NET_RT_IFLISTL)
1949                                         error = sysctl_iflist_ifaml(ifa, &info,
1950                                             w, len);
1951                                 else
1952                                         error = sysctl_iflist_ifam(ifa, &info,
1953                                             w, len);
1954                                 if (error)
1955                                         goto done;
1956                         }
1957                 }
1958                 info.rti_info[RTAX_IFA] = NULL;
1959                 info.rti_info[RTAX_NETMASK] = NULL;
1960                 info.rti_info[RTAX_BRD] = NULL;
1961         }
1962 done:
1963         return (error);
1964 }
1965
1966 static int
1967 sysctl_ifmalist(int af, struct walkarg *w)
1968 {
1969         struct rt_addrinfo info;
1970         struct ifaddr *ifa;
1971         struct ifmultiaddr *ifma;
1972         struct ifnet *ifp;
1973         int error, len;
1974
1975         NET_EPOCH_ASSERT();
1976
1977         error = 0;
1978         bzero((caddr_t)&info, sizeof(info));
1979
1980         CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1981                 if (w->w_arg && w->w_arg != ifp->if_index)
1982                         continue;
1983                 ifa = ifp->if_addr;
1984                 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1985                 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1986                         if (af && af != ifma->ifma_addr->sa_family)
1987                                 continue;
1988                         if (prison_if(w->w_req->td->td_ucred,
1989                             ifma->ifma_addr) != 0)
1990                                 continue;
1991                         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1992                         info.rti_info[RTAX_GATEWAY] =
1993                             (ifma->ifma_addr->sa_family != AF_LINK) ?
1994                             ifma->ifma_lladdr : NULL;
1995                         error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
1996                         if (error != 0)
1997                                 break;
1998                         if (w->w_req && w->w_tmem) {
1999                                 struct ifma_msghdr *ifmam;
2000
2001                                 ifmam = (struct ifma_msghdr *)w->w_tmem;
2002                                 ifmam->ifmam_index = ifma->ifma_ifp->if_index;
2003                                 ifmam->ifmam_flags = 0;
2004                                 ifmam->ifmam_addrs = info.rti_addrs;
2005                                 ifmam->_ifmam_spare1 = 0;
2006                                 error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
2007                                 if (error != 0)
2008                                         break;
2009                         }
2010                 }
2011                 if (error != 0)
2012                         break;
2013         }
2014         return (error);
2015 }
2016
2017 static int
2018 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
2019 {
2020         RIB_RLOCK_TRACKER;
2021         struct epoch_tracker et;
2022         int     *name = (int *)arg1;
2023         u_int   namelen = arg2;
2024         struct rib_head *rnh = NULL; /* silence compiler. */
2025         int     i, lim, error = EINVAL;
2026         int     fib = 0;
2027         u_char  af;
2028         struct  walkarg w;
2029
2030         name ++;
2031         namelen--;
2032         if (req->newptr)
2033                 return (EPERM);
2034         if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP) {
2035                 if (namelen == 3)
2036                         fib = req->td->td_proc->p_fibnum;
2037                 else if (namelen == 4)
2038                         fib = (name[3] == RT_ALL_FIBS) ?
2039                             req->td->td_proc->p_fibnum : name[3];
2040                 else
2041                         return ((namelen < 3) ? EISDIR : ENOTDIR);
2042                 if (fib < 0 || fib >= rt_numfibs)
2043                         return (EINVAL);
2044         } else if (namelen != 3)
2045                 return ((namelen < 3) ? EISDIR : ENOTDIR);
2046         af = name[0];
2047         if (af > AF_MAX)
2048                 return (EINVAL);
2049         bzero(&w, sizeof(w));
2050         w.w_op = name[1];
2051         w.w_arg = name[2];
2052         w.w_req = req;
2053
2054         error = sysctl_wire_old_buffer(req, 0);
2055         if (error)
2056                 return (error);
2057
2058         /*
2059          * Allocate reply buffer in advance.
2060          * All rtsock messages has maximum length of u_short.
2061          */
2062         w.w_tmemsize = 65536;
2063         w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
2064
2065         NET_EPOCH_ENTER(et);
2066         switch (w.w_op) {
2067         case NET_RT_DUMP:
2068         case NET_RT_FLAGS:
2069                 if (af == 0) {                  /* dump all tables */
2070                         i = 1;
2071                         lim = AF_MAX;
2072                 } else                          /* dump only one table */
2073                         i = lim = af;
2074
2075                 /*
2076                  * take care of llinfo entries, the caller must
2077                  * specify an AF
2078                  */
2079                 if (w.w_op == NET_RT_FLAGS &&
2080                     (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
2081                         if (af != 0)
2082                                 error = lltable_sysctl_dumparp(af, w.w_req);
2083                         else
2084                                 error = EINVAL;
2085                         break;
2086                 }
2087                 /*
2088                  * take care of routing entries
2089                  */
2090                 for (error = 0; error == 0 && i <= lim; i++) {
2091                         rnh = rt_tables_get_rnh(fib, i);
2092                         if (rnh != NULL) {
2093                                 RIB_RLOCK(rnh);
2094                                 error = rnh->rnh_walktree(&rnh->head,
2095                                     sysctl_dumpentry, &w);
2096                                 RIB_RUNLOCK(rnh);
2097                         } else if (af != 0)
2098                                 error = EAFNOSUPPORT;
2099                 }
2100                 break;
2101         case NET_RT_NHOP:
2102                 /* Allow dumping one specific af/fib at a time */
2103                 if (namelen < 4) {
2104                         error = EINVAL;
2105                         break;
2106                 }
2107                 fib = name[3];
2108                 if (fib < 0 || fib > rt_numfibs) {
2109                         error = EINVAL;
2110                         break;
2111                 }
2112                 rnh = rt_tables_get_rnh(fib, af);
2113                 if (rnh == NULL) {
2114                         error = EAFNOSUPPORT;
2115                         break;
2116                 }
2117                 if (w.w_op == NET_RT_NHOP)
2118                         error = nhops_dump_sysctl(rnh, w.w_req);
2119                 break;
2120         case NET_RT_IFLIST:
2121         case NET_RT_IFLISTL:
2122                 error = sysctl_iflist(af, &w);
2123                 break;
2124
2125         case NET_RT_IFMALIST:
2126                 error = sysctl_ifmalist(af, &w);
2127                 break;
2128         }
2129         NET_EPOCH_EXIT(et);
2130
2131         free(w.w_tmem, M_TEMP);
2132         return (error);
2133 }
2134
2135 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE,
2136     sysctl_rtsock, "Return route tables and interface/address lists");
2137
2138 /*
2139  * Definitions of protocols supported in the ROUTE domain.
2140  */
2141
2142 static struct domain routedomain;               /* or at least forward */
2143
2144 static struct protosw routesw[] = {
2145 {
2146         .pr_type =              SOCK_RAW,
2147         .pr_domain =            &routedomain,
2148         .pr_flags =             PR_ATOMIC|PR_ADDR,
2149         .pr_output =            route_output,
2150         .pr_ctlinput =          raw_ctlinput,
2151         .pr_init =              raw_init,
2152         .pr_usrreqs =           &route_usrreqs
2153 }
2154 };
2155
2156 static struct domain routedomain = {
2157         .dom_family =           PF_ROUTE,
2158         .dom_name =              "route",
2159         .dom_protosw =          routesw,
2160         .dom_protoswNPROTOSW =  &routesw[nitems(routesw)]
2161 };
2162
2163 VNET_DOMAIN_SET(route);
2164