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
 444         /* First, see if the returned address is part of the jail. */
 445         if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
 446                 info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
 447                 return (0);
 448         }
 449
 450         switch (info->rti_info[RTAX_DST]->sa_family) {
 451 #ifdef INET
 452         case AF_INET:
 453         {
 454                 struct in_addr ia;
 455                 struct ifaddr *ifa;
 456                 int found;
 457
 458                 found = 0;
 459                 /*
 460                  * Try to find an address on the given outgoing interface
 461                  * that belongs to the jail.
 462                  */
 463                 IF_ADDR_RLOCK(ifp);
 464                 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 465                         struct sockaddr *sa;
 466                         sa = ifa->ifa_addr;
 467                         if (sa->sa_family != AF_INET)
 468                                 continue;
 469                         ia = ((struct sockaddr_in *)sa)->sin_addr;
 470                         if (prison_check_ip4(cred, &ia) == 0) {
 471                                 found = 1;
 472                                 break;
 473                         }
 474                 }
 475                 IF_ADDR_RUNLOCK(ifp);
 476                 if (!found) {
 477                         /*
 478                          * As a last resort return the 'default' jail address.
 479                          */
 480                         ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
 481                             sin_addr;
 482                         if (prison_get_ip4(cred, &ia) != 0)
 483                                 return (ESRCH);
 484                 }
 485                 bzero(&saun->sin, sizeof(struct sockaddr_in));
 486                 saun->sin.sin_len = sizeof(struct sockaddr_in);
 487                 saun->sin.sin_family = AF_INET;
 488                 saun->sin.sin_addr.s_addr = ia.s_addr;
 489                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
 490                 break;
 491         }
 492 #endif
 493 #ifdef INET6
 494         case AF_INET6:
 495         {
 496                 struct in6_addr ia6;
 497                 struct ifaddr *ifa;
 498                 int found;
 499
 500                 found = 0;
 501                 /*
 502                  * Try to find an address on the given outgoing interface
 503                  * that belongs to the jail.
 504                  */
 505                 IF_ADDR_RLOCK(ifp);
 506                 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 507                         struct sockaddr *sa;
 508                         sa = ifa->ifa_addr;
 509                         if (sa->sa_family != AF_INET6)
 510                                 continue;
 511                         bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
 512                             &ia6, sizeof(struct in6_addr));
 513                         if (prison_check_ip6(cred, &ia6) == 0) {
 514                                 found = 1;
 515                                 break;
 516                         }
 517                 }
 518                 IF_ADDR_RUNLOCK(ifp);
 519                 if (!found) {
 520                         /*
 521                          * As a last resort return the 'default' jail address.
 522                          */
 523                         ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
 524                             sin6_addr;
 525                         if (prison_get_ip6(cred, &ia6) != 0)
 526                                 return (ESRCH);
 527                 }
 528                 bzero(&saun->sin6, sizeof(struct sockaddr_in6));
 529                 saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
 530                 saun->sin6.sin6_family = AF_INET6;
 531                 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
 532                 if (sa6_recoverscope(&saun->sin6) != 0)
 533                         return (ESRCH);
 534                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
 535                 break;
 536         }
 537 #endif
 538         default:
 539                 return (ESRCH);
 540         }
 541         return (0);
 542 }
 543
 544 /*ARGSUSED*/
 545 static int
 546 route_output(struct mbuf *m, struct socket *so, ...)
 547 {
 548         RIB_RLOCK_TRACKER;
 549         struct rt_msghdr *rtm = NULL;
 550         struct rtentry *rt = NULL;
 551         struct rib_head *rnh;
 552         struct rt_addrinfo info;
 553         struct sockaddr_storage ss;
 554 #ifdef INET6
 555         struct sockaddr_in6 *sin6;
 556         int i, rti_need_deembed = 0;
 557 #endif
 558         int alloc_len = 0, len, error = 0, fibnum;
 559         struct ifnet *ifp = NULL;
 560         union sockaddr_union saun;
 561         sa_family_t saf = AF_UNSPEC;
 562         struct rawcb *rp = NULL;
 563         struct walkarg w;
 564
 565         fibnum = so->so_fibnum;
 566
 567 #define senderr(e) { error = e; goto flush;}
 568         if (m == NULL || ((m->m_len < sizeof(long)) &&
 569                        (m = m_pullup(m, sizeof(long))) == NULL))
 570                 return (ENOBUFS);
 571         if ((m->m_flags & M_PKTHDR) == 0)
 572                 panic("route_output");
 573         len = m->m_pkthdr.len;
 574         if (len < sizeof(*rtm) ||
 575             len != mtod(m, struct rt_msghdr *)->rtm_msglen)
 576                 senderr(EINVAL);
 577
 578         /*
 579          * Most of current messages are in range 200-240 bytes,
 580          * minimize possible re-allocation on reply using larger size
 581          * buffer aligned on 1k boundaty.
 582          */
 583         alloc_len = roundup2(len, 1024);
 584         if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL)
 585                 senderr(ENOBUFS);
 586
 587         m_copydata(m, 0, len, (caddr_t)rtm);
 588         bzero(&info, sizeof(info));
 589         bzero(&w, sizeof(w));
 590
 591         if (rtm->rtm_version != RTM_VERSION) {
 592                 /* Do not touch message since format is unknown */
 593                 free(rtm, M_TEMP);
 594                 rtm = NULL;
 595                 senderr(EPROTONOSUPPORT);
 596         }
 597
 598         /*
 599          * Starting from here, it is possible
 600          * to alter original message and insert
 601          * caller PID and error value.
 602          */
 603
 604         rtm->rtm_pid = curproc->p_pid;
 605         info.rti_addrs = rtm->rtm_addrs;
 606
 607         info.rti_mflags = rtm->rtm_inits;
 608         info.rti_rmx = &rtm->rtm_rmx;
 609
 610         /*
 611          * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
 612          * link-local address because rtrequest requires addresses with
 613          * embedded scope id.
 614          */
 615         if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info))
 616                 senderr(EINVAL);
 617
 618         if (rtm->rtm_flags & RTF_RNH_LOCKED)
 619                 senderr(EINVAL);
 620         info.rti_flags = rtm->rtm_flags;
 621         if (info.rti_info[RTAX_DST] == NULL ||
 622             info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
 623             (info.rti_info[RTAX_GATEWAY] != NULL &&
 624              info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
 625                 senderr(EINVAL);
 626         saf = info.rti_info[RTAX_DST]->sa_family;
 627         /*
 628          * Verify that the caller has the appropriate privilege; RTM_GET
 629          * is the only operation the non-superuser is allowed.
 630          */
 631         if (rtm->rtm_type != RTM_GET) {
 632                 error = priv_check(curthread, PRIV_NET_ROUTE);
 633                 if (error)
 634                         senderr(error);
 635         }
 636
 637         /*
 638          * The given gateway address may be an interface address.
 639          * For example, issuing a "route change" command on a route
 640          * entry that was created from a tunnel, and the gateway
 641          * address given is the local end point. In this case the
 642          * RTF_GATEWAY flag must be cleared or the destination will
 643          * not be reachable even though there is no error message.
 644          */
 645         if (info.rti_info[RTAX_GATEWAY] != NULL &&
 646             info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
 647                 struct rt_addrinfo ginfo;
 648                 struct sockaddr *gdst;
 649
 650                 bzero(&ginfo, sizeof(ginfo));
 651                 bzero(&ss, sizeof(ss));
 652                 ss.ss_len = sizeof(ss);
 653
 654                 ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss;
 655                 gdst = info.rti_info[RTAX_GATEWAY];
 656
 657                 /*
 658                  * A host route through the loopback interface is
 659                  * installed for each interface adddress. In pre 8.0
 660                  * releases the interface address of a PPP link type
 661                  * is not reachable locally. This behavior is fixed as
 662                  * part of the new L2/L3 redesign and rewrite work. The
 663                  * signature of this interface address route is the
 664                  * AF_LINK sa_family type of the rt_gateway, and the
 665                  * rt_ifp has the IFF_LOOPBACK flag set.
 666                  */
 667                 if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) {
 668                         if (ss.ss_family == AF_LINK &&
 669                             ginfo.rti_ifp->if_flags & IFF_LOOPBACK) {
 670                                 info.rti_flags &= ~RTF_GATEWAY;
 671                                 info.rti_flags |= RTF_GWFLAG_COMPAT;
 672                         }
 673                         rib_free_info(&ginfo);
 674                 }
 675         }
 676
 677         switch (rtm->rtm_type) {
 678                 struct rtentry *saved_nrt;
 679
 680         case RTM_ADD:
 681         case RTM_CHANGE:
 682                 if (rtm->rtm_type == RTM_ADD) {
 683                         if (info.rti_info[RTAX_GATEWAY] == NULL)
 684                                 senderr(EINVAL);
 685                 }
 686                 saved_nrt = NULL;
 687
 688                 /* support for new ARP code */
 689                 if (info.rti_info[RTAX_GATEWAY] != NULL &&
 690                     info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
 691                     (rtm->rtm_flags & RTF_LLDATA) != 0) {
 692                         error = lla_rt_output(rtm, &info);
 693 #ifdef INET6
 694                         if (error == 0)
 695                                 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 696 #endif
 697                         break;
 698                 }
 699                 error = rtrequest1_fib(rtm->rtm_type, &info, &saved_nrt,
 700                     fibnum);
 701                 if (error == 0 && saved_nrt != NULL) {
 702 #ifdef INET6
 703                         rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 704 #endif
 705                         RT_LOCK(saved_nrt);
 706                         rtm->rtm_index = saved_nrt->rt_ifp->if_index;
 707                         RT_REMREF(saved_nrt);
 708                         RT_UNLOCK(saved_nrt);
 709                 }
 710                 break;
 711
 712         case RTM_DELETE:
 713                 saved_nrt = NULL;
 714                 /* support for new ARP code */
 715                 if (info.rti_info[RTAX_GATEWAY] &&
 716                     (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
 717                     (rtm->rtm_flags & RTF_LLDATA) != 0) {
 718                         error = lla_rt_output(rtm, &info);
 719 #ifdef INET6
 720                         if (error == 0)
 721                                 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 722 #endif
 723                         break;
 724                 }
 725                 error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt, fibnum);
 726                 if (error == 0) {
 727                         RT_LOCK(saved_nrt);
 728                         rt = saved_nrt;
 729                         goto report;
 730                 }
 731 #ifdef INET6
 732                 /* rt_msg2() will not be used when RTM_DELETE fails. */
 733                 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 734 #endif
 735                 break;
 736
 737         case RTM_GET:
 738                 rnh = rt_tables_get_rnh(fibnum, saf);
 739                 if (rnh == NULL)
 740                         senderr(EAFNOSUPPORT);
 741
 742                 RIB_RLOCK(rnh);
 743
 744                 if (info.rti_info[RTAX_NETMASK] == NULL &&
 745                     rtm->rtm_type == RTM_GET) {
 746                         /*
 747                          * Provide longest prefix match for
 748                          * address lookup (no mask).
 749                          * 'route -n get addr'
 750                          */
 751                         rt = (struct rtentry *) rnh->rnh_matchaddr(
 752                             info.rti_info[RTAX_DST], &rnh->head);
 753                 } else
 754                         rt = (struct rtentry *) rnh->rnh_lookup(
 755                             info.rti_info[RTAX_DST],
 756                             info.rti_info[RTAX_NETMASK], &rnh->head);
 757
 758                 if (rt == NULL) {
 759                         RIB_RUNLOCK(rnh);
 760                         senderr(ESRCH);
 761                 }
 762 #ifdef RADIX_MPATH
 763                 /*
 764                  * for RTM_CHANGE/LOCK, if we got multipath routes,
 765                  * we require users to specify a matching RTAX_GATEWAY.
 766                  *
 767                  * for RTM_GET, gate is optional even with multipath.
 768                  * if gate == NULL the first match is returned.
 769                  * (no need to call rt_mpath_matchgate if gate == NULL)
 770                  */
 771                 if (rt_mpath_capable(rnh) &&
 772                     (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
 773                         rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
 774                         if (!rt) {
 775                                 RIB_RUNLOCK(rnh);
 776                                 senderr(ESRCH);
 777                         }
 778                 }
 779 #endif
 780                 /*
 781                  * If performing proxied L2 entry insertion, and
 782                  * the actual PPP host entry is found, perform
 783                  * another search to retrieve the prefix route of
 784                  * the local end point of the PPP link.
 785                  */
 786                 if (rtm->rtm_flags & RTF_ANNOUNCE) {
 787                         struct sockaddr laddr;
 788
 789                         if (rt->rt_ifp != NULL &&
 790                             rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
 791                                 struct ifaddr *ifa;
 792
 793                                 NET_EPOCH_ENTER();
 794                                 ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1,
 795                                                 RT_ALL_FIBS);
 796                                 if (ifa != NULL)
 797                                         rt_maskedcopy(ifa->ifa_addr,
 798                                                       &laddr,
 799                                                       ifa->ifa_netmask);
 800                                 NET_EPOCH_EXIT();
 801                         } else
 802                                 rt_maskedcopy(rt->rt_ifa->ifa_addr,
 803                                               &laddr,
 804                                               rt->rt_ifa->ifa_netmask);
 805                         /*
 806                          * refactor rt and no lock operation necessary
 807                          */
 808                         rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr,
 809                             &rnh->head);
 810                         if (rt == NULL) {
 811                                 RIB_RUNLOCK(rnh);
 812                                 senderr(ESRCH);
 813                         }
 814                 }
 815                 RT_LOCK(rt);
 816                 RT_ADDREF(rt);
 817                 RIB_RUNLOCK(rnh);
 818
 819 report:
 820                 RT_LOCK_ASSERT(rt);
 821                 if ((rt->rt_flags & RTF_HOST) == 0
 822                     ? jailed_without_vnet(curthread->td_ucred)
 823                     : prison_if(curthread->td_ucred,
 824                     rt_key(rt)) != 0) {
 825                         RT_UNLOCK(rt);
 826                         senderr(ESRCH);
 827                 }
 828                 info.rti_info[RTAX_DST] = rt_key(rt);
 829                 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
 830                 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
 831                     rt_mask(rt), &ss);
 832                 info.rti_info[RTAX_GENMASK] = 0;
 833                 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
 834                         ifp = rt->rt_ifp;
 835                         if (ifp) {
 836                                 info.rti_info[RTAX_IFP] =
 837                                     ifp->if_addr->ifa_addr;
 838                                 error = rtm_get_jailed(&info, ifp, rt,
 839                                     &saun, curthread->td_ucred);
 840                                 if (error != 0) {
 841                                         RT_UNLOCK(rt);
 842                                         senderr(error);
 843                                 }
 844                                 if (ifp->if_flags & IFF_POINTOPOINT)
 845                                         info.rti_info[RTAX_BRD] =
 846                                             rt->rt_ifa->ifa_dstaddr;
 847                                 rtm->rtm_index = ifp->if_index;
 848                         } else {
 849                                 info.rti_info[RTAX_IFP] = NULL;
 850                                 info.rti_info[RTAX_IFA] = NULL;
 851                         }
 852                 } else if ((ifp = rt->rt_ifp) != NULL) {
 853                         rtm->rtm_index = ifp->if_index;
 854                 }
 855
 856                 /* Check if we need to realloc storage */
 857                 rtsock_msg_buffer(rtm->rtm_type, &info, NULL, &len);
 858                 if (len > alloc_len) {
 859                         struct rt_msghdr *new_rtm;
 860                         new_rtm = malloc(len, M_TEMP, M_NOWAIT);
 861                         if (new_rtm == NULL) {
 862                                 RT_UNLOCK(rt);
 863                                 senderr(ENOBUFS);
 864                         }
 865                         bcopy(rtm, new_rtm, rtm->rtm_msglen);
 866                         free(rtm, M_TEMP);
 867                         rtm = new_rtm;
 868                         alloc_len = len;
 869                 }
 870
 871                 w.w_tmem = (caddr_t)rtm;
 872                 w.w_tmemsize = alloc_len;
 873                 rtsock_msg_buffer(rtm->rtm_type, &info, &w, &len);
 874
 875                 if (rt->rt_flags & RTF_GWFLAG_COMPAT)
 876                         rtm->rtm_flags = RTF_GATEWAY |
 877                                 (rt->rt_flags & ~RTF_GWFLAG_COMPAT);
 878                 else
 879                         rtm->rtm_flags = rt->rt_flags;
 880                 rt_getmetrics(rt, &rtm->rtm_rmx);
 881                 rtm->rtm_addrs = info.rti_addrs;
 882
 883                 RT_UNLOCK(rt);
 884                 break;
 885
 886         default:
 887                 senderr(EOPNOTSUPP);
 888         }
 889
 890 flush:
 891         if (rt != NULL)
 892                 RTFREE(rt);
 893         /*
 894          * Check to see if we don't want our own messages.
 895          */
 896         if ((so->so_options & SO_USELOOPBACK) == 0) {
 897                 if (V_route_cb.any_count <= 1) {
 898                         if (rtm != NULL)
 899                                 free(rtm, M_TEMP);
 900                         m_freem(m);
 901                         return (error);
 902                 }
 903                 /* There is another listener, so construct message */
 904                 rp = sotorawcb(so);
 905         }
 906
 907         if (rtm != NULL) {
 908 #ifdef INET6
 909                 if (rti_need_deembed) {
 910                         /* sin6_scope_id is recovered before sending rtm. */
 911                         sin6 = (struct sockaddr_in6 *)&ss;
 912                         for (i = 0; i < RTAX_MAX; i++) {
 913                                 if (info.rti_info[i] == NULL)
 914                                         continue;
 915                                 if (info.rti_info[i]->sa_family != AF_INET6)
 916                                         continue;
 917                                 bcopy(info.rti_info[i], sin6, sizeof(*sin6));
 918                                 if (sa6_recoverscope(sin6) == 0)
 919                                         bcopy(sin6, info.rti_info[i],
 920                                                     sizeof(*sin6));
 921                         }
 922                 }
 923 #endif
 924                 if (error != 0)
 925                         rtm->rtm_errno = error;
 926                 else
 927                         rtm->rtm_flags |= RTF_DONE;
 928
 929                 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
 930                 if (m->m_pkthdr.len < rtm->rtm_msglen) {
 931                         m_freem(m);
 932                         m = NULL;
 933                 } else if (m->m_pkthdr.len > rtm->rtm_msglen)
 934                         m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
 935
 936                 free(rtm, M_TEMP);
 937         }
 938         if (m != NULL) {
 939                 M_SETFIB(m, fibnum);
 940                 m->m_flags |= RTS_FILTER_FIB;
 941                 if (rp) {
 942                         /*
 943                          * XXX insure we don't get a copy by
 944                          * invalidating our protocol
 945                          */
 946                         unsigned short family = rp->rcb_proto.sp_family;
 947                         rp->rcb_proto.sp_family = 0;
 948                         rt_dispatch(m, saf);
 949                         rp->rcb_proto.sp_family = family;
 950                 } else
 951                         rt_dispatch(m, saf);
 952         }
 953
 954         return (error);
 955 }
 956
 957 static void
 958 rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out)
 959 {
 960
 961         bzero(out, sizeof(*out));
 962         out->rmx_mtu = rt->rt_mtu;
 963         out->rmx_weight = rt->rt_weight;
 964         out->rmx_pksent = counter_u64_fetch(rt->rt_pksent);
 965         /* Kernel -> userland timebase conversion. */
 966         out->rmx_expire = rt->rt_expire ?
 967             rt->rt_expire - time_uptime + time_second : 0;
 968 }
 969
 970 /*
 971  * Extract the addresses of the passed sockaddrs.
 972  * Do a little sanity checking so as to avoid bad memory references.
 973  * This data is derived straight from userland.
 974  */
 975 static int
 976 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
 977 {
 978         struct sockaddr *sa;
 979         int i;
 980
 981         for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
 982                 if ((rtinfo->rti_addrs & (1 << i)) == 0)
 983                         continue;
 984                 sa = (struct sockaddr *)cp;
 985                 /*
 986                  * It won't fit.
 987                  */
 988                 if (cp + sa->sa_len > cplim)
 989                         return (EINVAL);
 990                 /*
 991                  * there are no more.. quit now
 992                  * If there are more bits, they are in error.
 993                  * I've seen this. route(1) can evidently generate these.
 994                  * This causes kernel to core dump.
 995                  * for compatibility, If we see this, point to a safe address.
 996                  */
 997                 if (sa->sa_len == 0) {
 998                         rtinfo->rti_info[i] = &sa_zero;
 999                         return (0); /* should be EINVAL but for compat */
1000                 }
1001                 /* accept it */
1002 #ifdef INET6
1003                 if (sa->sa_family == AF_INET6)
1004                         sa6_embedscope((struct sockaddr_in6 *)sa,
1005                             V_ip6_use_defzone);
1006 #endif
1007                 rtinfo->rti_info[i] = sa;
1008                 cp += SA_SIZE(sa);
1009         }
1010         return (0);
1011 }
1012
1013 /*
1014  * Fill in @dmask with valid netmask leaving original @smask
1015  * intact. Mostly used with radix netmasks.
1016  */
1017 static struct sockaddr *
1018 rtsock_fix_netmask(struct sockaddr *dst, struct sockaddr *smask,
1019     struct sockaddr_storage *dmask)
1020 {
1021         if (dst == NULL || smask == NULL)
1022                 return (NULL);
1023
1024         memset(dmask, 0, dst->sa_len);
1025         memcpy(dmask, smask, smask->sa_len);
1026         dmask->ss_len = dst->sa_len;
1027         dmask->ss_family = dst->sa_family;
1028
1029         return ((struct sockaddr *)dmask);
1030 }
1031
1032 /*
1033  * Writes information related to @rtinfo object to newly-allocated mbuf.
1034  * Assumes MCLBYTES is enough to construct any message.
1035  * Used for OS notifications of vaious events (if/ifa announces,etc)
1036  *
1037  * Returns allocated mbuf or NULL on failure.
1038  */
1039 static struct mbuf *
1040 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1041 {
1042         struct rt_msghdr *rtm;
1043         struct mbuf *m;
1044         int i;
1045         struct sockaddr *sa;
1046 #ifdef INET6
1047         struct sockaddr_storage ss;
1048         struct sockaddr_in6 *sin6;
1049 #endif
1050         int len, dlen;
1051
1052         switch (type) {
1053
1054         case RTM_DELADDR:
1055         case RTM_NEWADDR:
1056                 len = sizeof(struct ifa_msghdr);
1057                 break;
1058
1059         case RTM_DELMADDR:
1060         case RTM_NEWMADDR:
1061                 len = sizeof(struct ifma_msghdr);
1062                 break;
1063
1064         case RTM_IFINFO:
1065                 len = sizeof(struct if_msghdr);
1066                 break;
1067
1068         case RTM_IFANNOUNCE:
1069         case RTM_IEEE80211:
1070                 len = sizeof(struct if_announcemsghdr);
1071                 break;
1072
1073         default:
1074                 len = sizeof(struct rt_msghdr);
1075         }
1076
1077         /* XXXGL: can we use MJUMPAGESIZE cluster here? */
1078         KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1079         if (len > MHLEN)
1080                 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1081         else
1082                 m = m_gethdr(M_NOWAIT, MT_DATA);
1083         if (m == NULL)
1084                 return (m);
1085
1086         m->m_pkthdr.len = m->m_len = len;
1087         rtm = mtod(m, struct rt_msghdr *);
1088         bzero((caddr_t)rtm, len);
1089         for (i = 0; i < RTAX_MAX; i++) {
1090                 if ((sa = rtinfo->rti_info[i]) == NULL)
1091                         continue;
1092                 rtinfo->rti_addrs |= (1 << i);
1093                 dlen = SA_SIZE(sa);
1094 #ifdef INET6
1095                 if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1096                         sin6 = (struct sockaddr_in6 *)&ss;
1097                         bcopy(sa, sin6, sizeof(*sin6));
1098                         if (sa6_recoverscope(sin6) == 0)
1099                                 sa = (struct sockaddr *)sin6;
1100                 }
1101 #endif
1102                 m_copyback(m, len, dlen, (caddr_t)sa);
1103                 len += dlen;
1104         }
1105         if (m->m_pkthdr.len != len) {
1106                 m_freem(m);
1107                 return (NULL);
1108         }
1109         rtm->rtm_msglen = len;
1110         rtm->rtm_version = RTM_VERSION;
1111         rtm->rtm_type = type;
1112         return (m);
1113 }
1114
1115 /*
1116  * Writes information related to @rtinfo object to preallocated buffer.
1117  * Stores needed size in @plen. If @w is NULL, calculates size without
1118  * writing.
1119  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1120  *
1121  * Returns 0 on success.
1122  *
1123  */
1124 static int
1125 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1126 {
1127         int i;
1128         int len, buflen = 0, dlen;
1129         caddr_t cp = NULL;
1130         struct rt_msghdr *rtm = NULL;
1131 #ifdef INET6
1132         struct sockaddr_storage ss;
1133         struct sockaddr_in6 *sin6;
1134 #endif
1135 #ifdef COMPAT_FREEBSD32
1136         bool compat32 = false;
1137 #endif
1138
1139         switch (type) {
1140
1141         case RTM_DELADDR:
1142         case RTM_NEWADDR:
1143                 if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1144 #ifdef COMPAT_FREEBSD32
1145                         if (w->w_req->flags & SCTL_MASK32) {
1146                                 len = sizeof(struct ifa_msghdrl32);
1147                                 compat32 = true;
1148                         } else
1149 #endif
1150                                 len = sizeof(struct ifa_msghdrl);
1151                 } else
1152                         len = sizeof(struct ifa_msghdr);
1153                 break;
1154
1155         case RTM_IFINFO:
1156 #ifdef COMPAT_FREEBSD32
1157                 if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1158                         if (w->w_op == NET_RT_IFLISTL)
1159                                 len = sizeof(struct if_msghdrl32);
1160                         else
1161                                 len = sizeof(struct if_msghdr32);
1162                         compat32 = true;
1163                         break;
1164                 }
1165 #endif
1166                 if (w != NULL && w->w_op == NET_RT_IFLISTL)
1167                         len = sizeof(struct if_msghdrl);
1168                 else
1169                         len = sizeof(struct if_msghdr);
1170                 break;
1171
1172         case RTM_NEWMADDR:
1173                 len = sizeof(struct ifma_msghdr);
1174                 break;
1175
1176         default:
1177                 len = sizeof(struct rt_msghdr);
1178         }
1179
1180         if (w != NULL) {
1181                 rtm = (struct rt_msghdr *)w->w_tmem;
1182                 buflen = w->w_tmemsize - len;
1183                 cp = (caddr_t)w->w_tmem + len;
1184         }
1185
1186         rtinfo->rti_addrs = 0;
1187         for (i = 0; i < RTAX_MAX; i++) {
1188                 struct sockaddr *sa;
1189
1190                 if ((sa = rtinfo->rti_info[i]) == NULL)
1191                         continue;
1192                 rtinfo->rti_addrs |= (1 << i);
1193 #ifdef COMPAT_FREEBSD32
1194                 if (compat32)
1195                         dlen = SA_SIZE32(sa);
1196                 else
1197 #endif
1198                         dlen = SA_SIZE(sa);
1199                 if (cp != NULL && buflen >= dlen) {
1200 #ifdef INET6
1201                         if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1202                                 sin6 = (struct sockaddr_in6 *)&ss;
1203                                 bcopy(sa, sin6, sizeof(*sin6));
1204                                 if (sa6_recoverscope(sin6) == 0)
1205                                         sa = (struct sockaddr *)sin6;
1206                         }
1207 #endif
1208                         bcopy((caddr_t)sa, cp, (unsigned)dlen);
1209                         cp += dlen;
1210                         buflen -= dlen;
1211                 } else if (cp != NULL) {
1212                         /*
1213                          * Buffer too small. Count needed size
1214                          * and return with error.
1215                          */
1216                         cp = NULL;
1217                 }
1218
1219                 len += dlen;
1220         }
1221
1222         if (cp != NULL) {
1223                 dlen = ALIGN(len) - len;
1224                 if (buflen < dlen)
1225                         cp = NULL;
1226                 else {
1227                         bzero(cp, dlen);
1228                         cp += dlen;
1229                         buflen -= dlen;
1230                 }
1231         }
1232         len = ALIGN(len);
1233
1234         if (cp != NULL) {
1235                 /* fill header iff buffer is large enough */
1236                 rtm->rtm_version = RTM_VERSION;
1237                 rtm->rtm_type = type;
1238                 rtm->rtm_msglen = len;
1239         }
1240
1241         *plen = len;
1242
1243         if (w != NULL && cp == NULL)
1244                 return (ENOBUFS);
1245
1246         return (0);
1247 }
1248
1249 /*
1250  * This routine is called to generate a message from the routing
1251  * socket indicating that a redirect has occurred, a routing lookup
1252  * has failed, or that a protocol has detected timeouts to a particular
1253  * destination.
1254  */
1255 void
1256 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1257     int fibnum)
1258 {
1259         struct rt_msghdr *rtm;
1260         struct mbuf *m;
1261         struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1262
1263         if (V_route_cb.any_count == 0)
1264                 return;
1265         m = rtsock_msg_mbuf(type, rtinfo);
1266         if (m == NULL)
1267                 return;
1268
1269         if (fibnum != RT_ALL_FIBS) {
1270                 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1271                     "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1272                 M_SETFIB(m, fibnum);
1273                 m->m_flags |= RTS_FILTER_FIB;
1274         }
1275
1276         rtm = mtod(m, struct rt_msghdr *);
1277         rtm->rtm_flags = RTF_DONE | flags;
1278         rtm->rtm_errno = error;
1279         rtm->rtm_addrs = rtinfo->rti_addrs;
1280         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1281 }
1282
1283 void
1284 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1285 {
1286
1287         rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1288 }
1289
1290 /*
1291  * This routine is called to generate a message from the routing
1292  * socket indicating that the status of a network interface has changed.
1293  */
1294 void
1295 rt_ifmsg(struct ifnet *ifp)
1296 {
1297         struct if_msghdr *ifm;
1298         struct mbuf *m;
1299         struct rt_addrinfo info;
1300
1301         if (V_route_cb.any_count == 0)
1302                 return;
1303         bzero((caddr_t)&info, sizeof(info));
1304         m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1305         if (m == NULL)
1306                 return;
1307         ifm = mtod(m, struct if_msghdr *);
1308         ifm->ifm_index = ifp->if_index;
1309         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1310         if_data_copy(ifp, &ifm->ifm_data);
1311         ifm->ifm_addrs = 0;
1312         rt_dispatch(m, AF_UNSPEC);
1313 }
1314
1315 /*
1316  * Announce interface address arrival/withdraw.
1317  * Please do not call directly, use rt_addrmsg().
1318  * Assume input data to be valid.
1319  * Returns 0 on success.
1320  */
1321 int
1322 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1323 {
1324         struct rt_addrinfo info;
1325         struct sockaddr *sa;
1326         int ncmd;
1327         struct mbuf *m;
1328         struct ifa_msghdr *ifam;
1329         struct ifnet *ifp = ifa->ifa_ifp;
1330         struct sockaddr_storage ss;
1331
1332         if (V_route_cb.any_count == 0)
1333                 return (0);
1334
1335         ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1336
1337         bzero((caddr_t)&info, sizeof(info));
1338         info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1339         info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1340         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1341             info.rti_info[RTAX_IFP], ifa->ifa_netmask, &ss);
1342         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1343         if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1344                 return (ENOBUFS);
1345         ifam = mtod(m, struct ifa_msghdr *);
1346         ifam->ifam_index = ifp->if_index;
1347         ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1348         ifam->ifam_flags = ifa->ifa_flags;
1349         ifam->ifam_addrs = info.rti_addrs;
1350
1351         if (fibnum != RT_ALL_FIBS) {
1352                 M_SETFIB(m, fibnum);
1353                 m->m_flags |= RTS_FILTER_FIB;
1354         }
1355
1356         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1357
1358         return (0);
1359 }
1360
1361 /*
1362  * Announce route addition/removal.
1363  * Please do not call directly, use rt_routemsg().
1364  * Note that @rt data MAY be inconsistent/invalid:
1365  * if some userland app sends us "invalid" route message (invalid mask,
1366  * no dst, wrong address families, etc...) we need to pass it back
1367  * to app (and any other rtsock consumers) with rtm_errno field set to
1368  * non-zero value.
1369  *
1370  * Returns 0 on success.
1371  */
1372 int
1373 rtsock_routemsg(int cmd, struct ifnet *ifp, int error, struct rtentry *rt,
1374     int fibnum)
1375 {
1376         struct rt_addrinfo info;
1377         struct sockaddr *sa;
1378         struct mbuf *m;
1379         struct rt_msghdr *rtm;
1380         struct sockaddr_storage ss;
1381
1382         if (V_route_cb.any_count == 0)
1383                 return (0);
1384
1385         bzero((caddr_t)&info, sizeof(info));
1386         info.rti_info[RTAX_DST] = sa = rt_key(rt);
1387         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(sa, rt_mask(rt), &ss);
1388         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1389         if ((m = rtsock_msg_mbuf(cmd, &info)) == NULL)
1390                 return (ENOBUFS);
1391         rtm = mtod(m, struct rt_msghdr *);
1392         rtm->rtm_index = ifp->if_index;
1393         rtm->rtm_flags |= rt->rt_flags;
1394         rtm->rtm_errno = error;
1395         rtm->rtm_addrs = info.rti_addrs;
1396
1397         if (fibnum != RT_ALL_FIBS) {
1398                 M_SETFIB(m, fibnum);
1399                 m->m_flags |= RTS_FILTER_FIB;
1400         }
1401
1402         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1403
1404         return (0);
1405 }
1406
1407 /*
1408  * This is the analogue to the rt_newaddrmsg which performs the same
1409  * function but for multicast group memberhips.  This is easier since
1410  * there is no route state to worry about.
1411  */
1412 void
1413 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1414 {
1415         struct rt_addrinfo info;
1416         struct mbuf *m = NULL;
1417         struct ifnet *ifp = ifma->ifma_ifp;
1418         struct ifma_msghdr *ifmam;
1419
1420         if (V_route_cb.any_count == 0)
1421                 return;
1422
1423         bzero((caddr_t)&info, sizeof(info));
1424         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1425         if (ifp && ifp->if_addr)
1426                 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1427         else
1428                 info.rti_info[RTAX_IFP] = NULL;
1429         /*
1430          * If a link-layer address is present, present it as a ``gateway''
1431          * (similarly to how ARP entries, e.g., are presented).
1432          */
1433         info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1434         m = rtsock_msg_mbuf(cmd, &info);
1435         if (m == NULL)
1436                 return;
1437         ifmam = mtod(m, struct ifma_msghdr *);
1438         KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1439             __func__));
1440         ifmam->ifmam_index = ifp->if_index;
1441         ifmam->ifmam_addrs = info.rti_addrs;
1442         rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1443 }
1444
1445 static struct mbuf *
1446 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1447         struct rt_addrinfo *info)
1448 {
1449         struct if_announcemsghdr *ifan;
1450         struct mbuf *m;
1451
1452         if (V_route_cb.any_count == 0)
1453                 return NULL;
1454         bzero((caddr_t)info, sizeof(*info));
1455         m = rtsock_msg_mbuf(type, info);
1456         if (m != NULL) {
1457                 ifan = mtod(m, struct if_announcemsghdr *);
1458                 ifan->ifan_index = ifp->if_index;
1459                 strlcpy(ifan->ifan_name, ifp->if_xname,
1460                         sizeof(ifan->ifan_name));
1461                 ifan->ifan_what = what;
1462         }
1463         return m;
1464 }
1465
1466 /*
1467  * This is called to generate routing socket messages indicating
1468  * IEEE80211 wireless events.
1469  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1470  */
1471 void
1472 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1473 {
1474         struct mbuf *m;
1475         struct rt_addrinfo info;
1476
1477         m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1478         if (m != NULL) {
1479                 /*
1480                  * Append the ieee80211 data.  Try to stick it in the
1481                  * mbuf containing the ifannounce msg; otherwise allocate
1482                  * a new mbuf and append.
1483                  *
1484                  * NB: we assume m is a single mbuf.
1485                  */
1486                 if (data_len > M_TRAILINGSPACE(m)) {
1487                         struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1488                         if (n == NULL) {
1489                                 m_freem(m);
1490                                 return;
1491                         }
1492                         bcopy(data, mtod(n, void *), data_len);
1493                         n->m_len = data_len;
1494                         m->m_next = n;
1495                 } else if (data_len > 0) {
1496                         bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1497                         m->m_len += data_len;
1498                 }
1499                 if (m->m_flags & M_PKTHDR)
1500                         m->m_pkthdr.len += data_len;
1501                 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1502                 rt_dispatch(m, AF_UNSPEC);
1503         }
1504 }
1505
1506 /*
1507  * This is called to generate routing socket messages indicating
1508  * network interface arrival and departure.
1509  */
1510 void
1511 rt_ifannouncemsg(struct ifnet *ifp, int what)
1512 {
1513         struct mbuf *m;
1514         struct rt_addrinfo info;
1515
1516         m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1517         if (m != NULL)
1518                 rt_dispatch(m, AF_UNSPEC);
1519 }
1520
1521 static void
1522 rt_dispatch(struct mbuf *m, sa_family_t saf)
1523 {
1524         struct m_tag *tag;
1525
1526         /*
1527          * Preserve the family from the sockaddr, if any, in an m_tag for
1528          * use when injecting the mbuf into the routing socket buffer from
1529          * the netisr.
1530          */
1531         if (saf != AF_UNSPEC) {
1532                 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1533                     M_NOWAIT);
1534                 if (tag == NULL) {
1535                         m_freem(m);
1536                         return;
1537                 }
1538                 *(unsigned short *)(tag + 1) = saf;
1539                 m_tag_prepend(m, tag);
1540         }
1541 #ifdef VIMAGE
1542         if (V_loif)
1543                 m->m_pkthdr.rcvif = V_loif;
1544         else {
1545                 m_freem(m);
1546                 return;
1547         }
1548 #endif
1549         netisr_queue(NETISR_ROUTE, m);  /* mbuf is free'd on failure. */
1550 }
1551
1552 /*
1553  * This is used in dumping the kernel table via sysctl().
1554  */
1555 static int
1556 sysctl_dumpentry(struct radix_node *rn, void *vw)
1557 {
1558         struct walkarg *w = vw;
1559         struct rtentry *rt = (struct rtentry *)rn;
1560         int error = 0, size;
1561         struct rt_addrinfo info;
1562         struct sockaddr_storage ss;
1563
1564         IFNET_RLOCK_NOSLEEP_ASSERT();
1565
1566         if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1567                 return 0;
1568         if ((rt->rt_flags & RTF_HOST) == 0
1569             ? jailed_without_vnet(w->w_req->td->td_ucred)
1570             : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1571                 return (0);
1572         bzero((caddr_t)&info, sizeof(info));
1573         info.rti_info[RTAX_DST] = rt_key(rt);
1574         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1575         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
1576             rt_mask(rt), &ss);
1577         info.rti_info[RTAX_GENMASK] = 0;
1578         if (rt->rt_ifp && !(rt->rt_ifp->if_flags & IFF_DYING)) {
1579                 info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1580                 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1581                 if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1582                         info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1583         }
1584         if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
1585                 return (error);
1586         if (w->w_req && w->w_tmem) {
1587                 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1588
1589                 bzero(&rtm->rtm_index,
1590                     sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
1591                 if (rt->rt_flags & RTF_GWFLAG_COMPAT)
1592                         rtm->rtm_flags = RTF_GATEWAY |
1593                                 (rt->rt_flags & ~RTF_GWFLAG_COMPAT);
1594                 else
1595                         rtm->rtm_flags = rt->rt_flags;
1596                 rt_getmetrics(rt, &rtm->rtm_rmx);
1597                 rtm->rtm_index = rt->rt_ifp->if_index;
1598                 rtm->rtm_addrs = info.rti_addrs;
1599                 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1600                 return (error);
1601         }
1602         return (error);
1603 }
1604
1605 static int
1606 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
1607     struct rt_addrinfo *info, struct walkarg *w, int len)
1608 {
1609         struct if_msghdrl *ifm;
1610         struct if_data *ifd;
1611
1612         ifm = (struct if_msghdrl *)w->w_tmem;
1613
1614 #ifdef COMPAT_FREEBSD32
1615         if (w->w_req->flags & SCTL_MASK32) {
1616                 struct if_msghdrl32 *ifm32;
1617
1618                 ifm32 = (struct if_msghdrl32 *)ifm;
1619                 ifm32->ifm_addrs = info->rti_addrs;
1620                 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1621                 ifm32->ifm_index = ifp->if_index;
1622                 ifm32->_ifm_spare1 = 0;
1623                 ifm32->ifm_len = sizeof(*ifm32);
1624                 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1625                 ifm32->_ifm_spare2 = 0;
1626                 ifd = &ifm32->ifm_data;
1627         } else
1628 #endif
1629         {
1630                 ifm->ifm_addrs = info->rti_addrs;
1631                 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1632                 ifm->ifm_index = ifp->if_index;
1633                 ifm->_ifm_spare1 = 0;
1634                 ifm->ifm_len = sizeof(*ifm);
1635                 ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1636                 ifm->_ifm_spare2 = 0;
1637                 ifd = &ifm->ifm_data;
1638         }
1639
1640         memcpy(ifd, src_ifd, sizeof(*ifd));
1641
1642         return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1643 }
1644
1645 static int
1646 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
1647     struct rt_addrinfo *info, struct walkarg *w, int len)
1648 {
1649         struct if_msghdr *ifm;
1650         struct if_data *ifd;
1651
1652         ifm = (struct if_msghdr *)w->w_tmem;
1653
1654 #ifdef COMPAT_FREEBSD32
1655         if (w->w_req->flags & SCTL_MASK32) {
1656                 struct if_msghdr32 *ifm32;
1657
1658                 ifm32 = (struct if_msghdr32 *)ifm;
1659                 ifm32->ifm_addrs = info->rti_addrs;
1660                 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1661                 ifm32->ifm_index = ifp->if_index;
1662                 ifm32->_ifm_spare1 = 0;
1663                 ifd = &ifm32->ifm_data;
1664         } else
1665 #endif
1666         {
1667                 ifm->ifm_addrs = info->rti_addrs;
1668                 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1669                 ifm->ifm_index = ifp->if_index;
1670                 ifm->_ifm_spare1 = 0;
1671                 ifd = &ifm->ifm_data;
1672         }
1673
1674         memcpy(ifd, src_ifd, sizeof(*ifd));
1675
1676         return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1677 }
1678
1679 static int
1680 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1681     struct walkarg *w, int len)
1682 {
1683         struct ifa_msghdrl *ifam;
1684         struct if_data *ifd;
1685
1686         ifam = (struct ifa_msghdrl *)w->w_tmem;
1687
1688 #ifdef COMPAT_FREEBSD32
1689         if (w->w_req->flags & SCTL_MASK32) {
1690                 struct ifa_msghdrl32 *ifam32;
1691
1692                 ifam32 = (struct ifa_msghdrl32 *)ifam;
1693                 ifam32->ifam_addrs = info->rti_addrs;
1694                 ifam32->ifam_flags = ifa->ifa_flags;
1695                 ifam32->ifam_index = ifa->ifa_ifp->if_index;
1696                 ifam32->_ifam_spare1 = 0;
1697                 ifam32->ifam_len = sizeof(*ifam32);
1698                 ifam32->ifam_data_off =
1699                     offsetof(struct ifa_msghdrl32, ifam_data);
1700                 ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
1701                 ifd = &ifam32->ifam_data;
1702         } else
1703 #endif
1704         {
1705                 ifam->ifam_addrs = info->rti_addrs;
1706                 ifam->ifam_flags = ifa->ifa_flags;
1707                 ifam->ifam_index = ifa->ifa_ifp->if_index;
1708                 ifam->_ifam_spare1 = 0;
1709                 ifam->ifam_len = sizeof(*ifam);
1710                 ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1711                 ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1712                 ifd = &ifam->ifam_data;
1713         }
1714
1715         bzero(ifd, sizeof(*ifd));
1716         ifd->ifi_datalen = sizeof(struct if_data);
1717         ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
1718         ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
1719         ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
1720         ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
1721
1722         /* Fixup if_data carp(4) vhid. */
1723         if (carp_get_vhid_p != NULL)
1724                 ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
1725
1726         return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1727 }
1728
1729 static int
1730 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1731     struct walkarg *w, int len)
1732 {
1733         struct ifa_msghdr *ifam;
1734
1735         ifam = (struct ifa_msghdr *)w->w_tmem;
1736         ifam->ifam_addrs = info->rti_addrs;
1737         ifam->ifam_flags = ifa->ifa_flags;
1738         ifam->ifam_index = ifa->ifa_ifp->if_index;
1739         ifam->_ifam_spare1 = 0;
1740         ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1741
1742         return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1743 }
1744
1745 static int
1746 sysctl_iflist(int af, struct walkarg *w)
1747 {
1748         struct ifnet *ifp;
1749         struct ifaddr *ifa;
1750         struct if_data ifd;
1751         struct rt_addrinfo info;
1752         int len, error = 0;
1753         struct sockaddr_storage ss;
1754         struct epoch_tracker et;
1755
1756         bzero((caddr_t)&info, sizeof(info));
1757         bzero(&ifd, sizeof(ifd));
1758         NET_EPOCH_ENTER_ET(et);
1759         CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1760                 if (w->w_arg && w->w_arg != ifp->if_index)
1761                         continue;
1762                 if_data_copy(ifp, &ifd);
1763                 ifa = ifp->if_addr;
1764                 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1765                 error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
1766                 if (error != 0)
1767                         goto done;
1768                 info.rti_info[RTAX_IFP] = NULL;
1769                 if (w->w_req && w->w_tmem) {
1770                         if (w->w_op == NET_RT_IFLISTL)
1771                                 error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
1772                                     len);
1773                         else
1774                                 error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
1775                                     len);
1776                         if (error)
1777                                 goto done;
1778                 }
1779                 while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
1780                         if (af && af != ifa->ifa_addr->sa_family)
1781                                 continue;
1782                         if (prison_if(w->w_req->td->td_ucred,
1783                             ifa->ifa_addr) != 0)
1784                                 continue;
1785                         info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1786                         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1787                             ifa->ifa_addr, ifa->ifa_netmask, &ss);
1788                         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1789                         error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
1790                         if (error != 0)
1791                                 goto done;
1792                         if (w->w_req && w->w_tmem) {
1793                                 if (w->w_op == NET_RT_IFLISTL)
1794                                         error = sysctl_iflist_ifaml(ifa, &info,
1795                                             w, len);
1796                                 else
1797                                         error = sysctl_iflist_ifam(ifa, &info,
1798                                             w, len);
1799                                 if (error)
1800                                         goto done;
1801                         }
1802                 }
1803                 info.rti_info[RTAX_IFA] = NULL;
1804                 info.rti_info[RTAX_NETMASK] = NULL;
1805                 info.rti_info[RTAX_BRD] = NULL;
1806         }
1807 done:
1808         NET_EPOCH_EXIT_ET(et);
1809         return (error);
1810 }
1811
1812 static int
1813 sysctl_ifmalist(int af, struct walkarg *w)
1814 {
1815         struct rt_addrinfo info;
1816         struct ifaddr *ifa;
1817         struct ifmultiaddr *ifma;
1818         struct ifnet *ifp;
1819         int error, len;
1820
1821         error = 0;
1822         bzero((caddr_t)&info, sizeof(info));
1823
1824         IFNET_RLOCK_NOSLEEP();
1825         CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1826                 if (w->w_arg && w->w_arg != ifp->if_index)
1827                         continue;
1828                 ifa = ifp->if_addr;
1829                 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1830                 IF_ADDR_RLOCK(ifp);
1831                 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1832                         if (af && af != ifma->ifma_addr->sa_family)
1833                                 continue;
1834                         if (prison_if(w->w_req->td->td_ucred,
1835                             ifma->ifma_addr) != 0)
1836                                 continue;
1837                         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1838                         info.rti_info[RTAX_GATEWAY] =
1839                             (ifma->ifma_addr->sa_family != AF_LINK) ?
1840                             ifma->ifma_lladdr : NULL;
1841                         error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
1842                         if (error != 0)
1843                                 break;
1844                         if (w->w_req && w->w_tmem) {
1845                                 struct ifma_msghdr *ifmam;
1846
1847                                 ifmam = (struct ifma_msghdr *)w->w_tmem;
1848                                 ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1849                                 ifmam->ifmam_flags = 0;
1850                                 ifmam->ifmam_addrs = info.rti_addrs;
1851                                 ifmam->_ifmam_spare1 = 0;
1852                                 error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1853                                 if (error != 0)
1854                                         break;
1855                         }
1856                 }
1857                 IF_ADDR_RUNLOCK(ifp);
1858                 if (error != 0)
1859                         break;
1860         }
1861         IFNET_RUNLOCK_NOSLEEP();
1862         return (error);
1863 }
1864
1865 static int
1866 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1867 {
1868         RIB_RLOCK_TRACKER;
1869         int     *name = (int *)arg1;
1870         u_int   namelen = arg2;
1871         struct rib_head *rnh = NULL; /* silence compiler. */
1872         int     i, lim, error = EINVAL;
1873         int     fib = 0;
1874         u_char  af;
1875         struct  walkarg w;
1876
1877         name ++;
1878         namelen--;
1879         if (req->newptr)
1880                 return (EPERM);
1881         if (name[1] == NET_RT_DUMP) {
1882                 if (namelen == 3)
1883                         fib = req->td->td_proc->p_fibnum;
1884                 else if (namelen == 4)
1885                         fib = (name[3] == RT_ALL_FIBS) ?
1886                             req->td->td_proc->p_fibnum : name[3];
1887                 else
1888                         return ((namelen < 3) ? EISDIR : ENOTDIR);
1889                 if (fib < 0 || fib >= rt_numfibs)
1890                         return (EINVAL);
1891         } else if (namelen != 3)
1892                 return ((namelen < 3) ? EISDIR : ENOTDIR);
1893         af = name[0];
1894         if (af > AF_MAX)
1895                 return (EINVAL);
1896         bzero(&w, sizeof(w));
1897         w.w_op = name[1];
1898         w.w_arg = name[2];
1899         w.w_req = req;
1900
1901         error = sysctl_wire_old_buffer(req, 0);
1902         if (error)
1903                 return (error);
1904
1905         /*
1906          * Allocate reply buffer in advance.
1907          * All rtsock messages has maximum length of u_short.
1908          */
1909         w.w_tmemsize = 65536;
1910         w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
1911
1912         switch (w.w_op) {
1913
1914         case NET_RT_DUMP:
1915         case NET_RT_FLAGS:
1916                 if (af == 0) {                  /* dump all tables */
1917                         i = 1;
1918                         lim = AF_MAX;
1919                 } else                          /* dump only one table */
1920                         i = lim = af;
1921
1922                 /*
1923                  * take care of llinfo entries, the caller must
1924                  * specify an AF
1925                  */
1926                 if (w.w_op == NET_RT_FLAGS &&
1927                     (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1928                         if (af != 0)
1929                                 error = lltable_sysctl_dumparp(af, w.w_req);
1930                         else
1931                                 error = EINVAL;
1932                         break;
1933                 }
1934                 /*
1935                  * take care of routing entries
1936                  */
1937                 for (error = 0; error == 0 && i <= lim; i++) {
1938                         rnh = rt_tables_get_rnh(fib, i);
1939                         if (rnh != NULL) {
1940                                 RIB_RLOCK(rnh);
1941                                 IFNET_RLOCK_NOSLEEP();
1942                                 error = rnh->rnh_walktree(&rnh->head,
1943                                     sysctl_dumpentry, &w);
1944                                 IFNET_RUNLOCK_NOSLEEP();
1945                                 RIB_RUNLOCK(rnh);
1946                         } else if (af != 0)
1947                                 error = EAFNOSUPPORT;
1948                 }
1949                 break;
1950
1951         case NET_RT_IFLIST:
1952         case NET_RT_IFLISTL:
1953                 error = sysctl_iflist(af, &w);
1954                 break;
1955
1956         case NET_RT_IFMALIST:
1957                 error = sysctl_ifmalist(af, &w);
1958                 break;
1959         }
1960
1961         free(w.w_tmem, M_TEMP);
1962         return (error);
1963 }
1964
1965 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1966
1967 /*
1968  * Definitions of protocols supported in the ROUTE domain.
1969  */
1970
1971 static struct domain routedomain;               /* or at least forward */
1972
1973 static struct protosw routesw[] = {
1974 {
1975         .pr_type =              SOCK_RAW,
1976         .pr_domain =            &routedomain,
1977         .pr_flags =             PR_ATOMIC|PR_ADDR,
1978         .pr_output =            route_output,
1979         .pr_ctlinput =          raw_ctlinput,
1980         .pr_init =              raw_init,
1981         .pr_usrreqs =           &route_usrreqs
1982 }
1983 };
1984
1985 static struct domain routedomain = {
1986         .dom_family =           PF_ROUTE,
1987         .dom_name =              "route",
1988         .dom_protosw =          routesw,
1989         .dom_protoswNPROTOSW =  &routesw[nitems(routesw)]
1990 };
1991
1992 VNET_DOMAIN_SET(route);