sys/net/rtsock.c

   1 /*-
   2  * Copyright (c) 1988, 1991, 1993
   3  *      The Regents of the University of California.  All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  * 4. Neither the name of the University nor the names of its contributors
  14  *    may be used to endorse or promote products derived from this software
  15  *    without specific prior written permission.
  16  *
  17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  27  * SUCH DAMAGE.
  28  *
  29  *      @(#)rtsock.c    8.7 (Berkeley) 10/12/95
  30  * $FreeBSD$
  31  */
  32 #include "opt_compat.h"
  33 #include "opt_mpath.h"
  34 #include "opt_inet.h"
  35 #include "opt_inet6.h"
  36
  37 #include <sys/param.h>
  38 #include <sys/jail.h>
  39 #include <sys/kernel.h>
  40 #include <sys/domain.h>
  41 #include <sys/lock.h>
  42 #include <sys/malloc.h>
  43 #include <sys/mbuf.h>
  44 #include <sys/priv.h>
  45 #include <sys/proc.h>
  46 #include <sys/protosw.h>
  47 #include <sys/rwlock.h>
  48 #include <sys/signalvar.h>
  49 #include <sys/socket.h>
  50 #include <sys/socketvar.h>
  51 #include <sys/sysctl.h>
  52 #include <sys/systm.h>
  53
  54 #include <net/if.h>
  55 #include <net/if_var.h>
  56 #include <net/if_dl.h>
  57 #include <net/if_llatbl.h>
  58 #include <net/if_types.h>
  59 #include <net/netisr.h>
  60 #include <net/raw_cb.h>
  61 #include <net/route.h>
  62 #include <net/route_var.h>
  63 #include <net/vnet.h>
  64
  65 #include <netinet/in.h>
  66 #include <netinet/if_ether.h>
  67 #include <netinet/ip_carp.h>
  68 #ifdef INET6
  69 #include <netinet6/ip6_var.h>
  70 #include <netinet6/scope6_var.h>
  71 #endif
  72
  73 #ifdef COMPAT_FREEBSD32
  74 #include <sys/mount.h>
  75 #include <compat/freebsd32/freebsd32.h>
  76
  77 struct if_msghdr32 {
  78         uint16_t ifm_msglen;
  79         uint8_t ifm_version;
  80         uint8_t ifm_type;
  81         int32_t ifm_addrs;
  82         int32_t ifm_flags;
  83         uint16_t ifm_index;
  84         uint16_t _ifm_spare1;
  85         struct  if_data ifm_data;
  86 };
  87
  88 struct if_msghdrl32 {
  89         uint16_t ifm_msglen;
  90         uint8_t ifm_version;
  91         uint8_t ifm_type;
  92         int32_t ifm_addrs;
  93         int32_t ifm_flags;
  94         uint16_t ifm_index;
  95         uint16_t _ifm_spare1;
  96         uint16_t ifm_len;
  97         uint16_t ifm_data_off;
  98         uint32_t _ifm_spare2;
  99         struct  if_data ifm_data;
 100 };
 101
 102 struct ifa_msghdrl32 {
 103         uint16_t ifam_msglen;
 104         uint8_t ifam_version;
 105         uint8_t ifam_type;
 106         int32_t ifam_addrs;
 107         int32_t ifam_flags;
 108         uint16_t ifam_index;
 109         uint16_t _ifam_spare1;
 110         uint16_t ifam_len;
 111         uint16_t ifam_data_off;
 112         int32_t ifam_metric;
 113         struct  if_data ifam_data;
 114 };
 115
 116 #define SA_SIZE32(sa)                                           \
 117     (  (((struct sockaddr *)(sa))->sa_len == 0) ?               \
 118         sizeof(int)             :                               \
 119         1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) )
 120
 121 #endif /* COMPAT_FREEBSD32 */
 122
 123 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
 124
 125 /* NB: these are not modified */
 126 static struct   sockaddr route_src = { 2, PF_ROUTE, };
 127 static struct   sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
 128
 129 /* These are external hooks for CARP. */
 130 int     (*carp_get_vhid_p)(struct ifaddr *);
 131
 132 /*
 133  * Used by rtsock/raw_input callback code to decide whether to filter the update
 134  * notification to a socket bound to a particular FIB.
 135  */
 136 #define RTS_FILTER_FIB  M_PROTO8
 137
 138 typedef struct {
 139         int     ip_count;       /* attached w/ AF_INET */
 140         int     ip6_count;      /* attached w/ AF_INET6 */
 141         int     any_count;      /* total attached */
 142 } route_cb_t;
 143 static VNET_DEFINE(route_cb_t, route_cb);
 144 #define V_route_cb VNET(route_cb)
 145
 146 struct mtx rtsock_mtx;
 147 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
 148
 149 #define RTSOCK_LOCK()   mtx_lock(&rtsock_mtx)
 150 #define RTSOCK_UNLOCK() mtx_unlock(&rtsock_mtx)
 151 #define RTSOCK_LOCK_ASSERT()    mtx_assert(&rtsock_mtx, MA_OWNED)
 152
 153 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
 154
 155 struct walkarg {
 156         int     w_tmemsize;
 157         int     w_op, w_arg;
 158         caddr_t w_tmem;
 159         struct sysctl_req *w_req;
 160 };
 161
 162 static void     rts_input(struct mbuf *m);
 163 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
 164 static int      rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
 165                         struct walkarg *w, int *plen);
 166 static int      rt_xaddrs(caddr_t cp, caddr_t cplim,
 167                         struct rt_addrinfo *rtinfo);
 168 static int      sysctl_dumpentry(struct radix_node *rn, void *vw);
 169 static int      sysctl_iflist(int af, struct walkarg *w);
 170 static int      sysctl_ifmalist(int af, struct walkarg *w);
 171 static int      route_output(struct mbuf *m, struct socket *so, ...);
 172 static void     rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out);
 173 static void     rt_dispatch(struct mbuf *, sa_family_t);
 174 static struct sockaddr  *rtsock_fix_netmask(struct sockaddr *dst,
 175                         struct sockaddr *smask, struct sockaddr_storage *dmask);
 176
 177 static struct netisr_handler rtsock_nh = {
 178         .nh_name = "rtsock",
 179         .nh_handler = rts_input,
 180         .nh_proto = NETISR_ROUTE,
 181         .nh_policy = NETISR_POLICY_SOURCE,
 182 };
 183
 184 static int
 185 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
 186 {
 187         int error, qlimit;
 188
 189         netisr_getqlimit(&rtsock_nh, &qlimit);
 190         error = sysctl_handle_int(oidp, &qlimit, 0, req);
 191         if (error || !req->newptr)
 192                 return (error);
 193         if (qlimit < 1)
 194                 return (EINVAL);
 195         return (netisr_setqlimit(&rtsock_nh, qlimit));
 196 }
 197 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
 198     0, 0, sysctl_route_netisr_maxqlen, "I",
 199     "maximum routing socket dispatch queue length");
 200
 201 static void
 202 vnet_rts_init(void)
 203 {
 204         int tmp;
 205
 206         if (IS_DEFAULT_VNET(curvnet)) {
 207                 if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
 208                         rtsock_nh.nh_qlimit = tmp;
 209                 netisr_register(&rtsock_nh);
 210         }
 211 #ifdef VIMAGE
 212          else
 213                 netisr_register_vnet(&rtsock_nh);
 214 #endif
 215 }
 216 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
 217     vnet_rts_init, 0);
 218
 219 #ifdef VIMAGE
 220 static void
 221 vnet_rts_uninit(void)
 222 {
 223
 224         netisr_unregister_vnet(&rtsock_nh);
 225 }
 226 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
 227     vnet_rts_uninit, 0);
 228 #endif
 229
 230 static int
 231 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
 232     struct rawcb *rp)
 233 {
 234         int fibnum;
 235
 236         KASSERT(m != NULL, ("%s: m is NULL", __func__));
 237         KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
 238         KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
 239
 240         /* No filtering requested. */
 241         if ((m->m_flags & RTS_FILTER_FIB) == 0)
 242                 return (0);
 243
 244         /* Check if it is a rts and the fib matches the one of the socket. */
 245         fibnum = M_GETFIB(m);
 246         if (proto->sp_family != PF_ROUTE ||
 247             rp->rcb_socket == NULL ||
 248             rp->rcb_socket->so_fibnum == fibnum)
 249                 return (0);
 250
 251         /* Filtering requested and no match, the socket shall be skipped. */
 252         return (1);
 253 }
 254
 255 static void
 256 rts_input(struct mbuf *m)
 257 {
 258         struct sockproto route_proto;
 259         unsigned short *family;
 260         struct m_tag *tag;
 261
 262         route_proto.sp_family = PF_ROUTE;
 263         tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
 264         if (tag != NULL) {
 265                 family = (unsigned short *)(tag + 1);
 266                 route_proto.sp_protocol = *family;
 267                 m_tag_delete(m, tag);
 268         } else
 269                 route_proto.sp_protocol = 0;
 270
 271         raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
 272 }
 273
 274 /*
 275  * It really doesn't make any sense at all for this code to share much
 276  * with raw_usrreq.c, since its functionality is so restricted.  XXX
 277  */
 278 static void
 279 rts_abort(struct socket *so)
 280 {
 281
 282         raw_usrreqs.pru_abort(so);
 283 }
 284
 285 static void
 286 rts_close(struct socket *so)
 287 {
 288
 289         raw_usrreqs.pru_close(so);
 290 }
 291
 292 /* pru_accept is EOPNOTSUPP */
 293
 294 static int
 295 rts_attach(struct socket *so, int proto, struct thread *td)
 296 {
 297         struct rawcb *rp;
 298         int error;
 299
 300         KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
 301
 302         /* XXX */
 303         rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
 304
 305         so->so_pcb = (caddr_t)rp;
 306         so->so_fibnum = td->td_proc->p_fibnum;
 307         error = raw_attach(so, proto);
 308         rp = sotorawcb(so);
 309         if (error) {
 310                 so->so_pcb = NULL;
 311                 free(rp, M_PCB);
 312                 return error;
 313         }
 314         RTSOCK_LOCK();
 315         switch(rp->rcb_proto.sp_protocol) {
 316         case AF_INET:
 317                 V_route_cb.ip_count++;
 318                 break;
 319         case AF_INET6:
 320                 V_route_cb.ip6_count++;
 321                 break;
 322         }
 323         V_route_cb.any_count++;
 324         RTSOCK_UNLOCK();
 325         soisconnected(so);
 326         so->so_options |= SO_USELOOPBACK;
 327         return 0;
 328 }
 329
 330 static int
 331 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
 332 {
 333
 334         return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
 335 }
 336
 337 static int
 338 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
 339 {
 340
 341         return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
 342 }
 343
 344 /* pru_connect2 is EOPNOTSUPP */
 345 /* pru_control is EOPNOTSUPP */
 346
 347 static void
 348 rts_detach(struct socket *so)
 349 {
 350         struct rawcb *rp = sotorawcb(so);
 351
 352         KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
 353
 354         RTSOCK_LOCK();
 355         switch(rp->rcb_proto.sp_protocol) {
 356         case AF_INET:
 357                 V_route_cb.ip_count--;
 358                 break;
 359         case AF_INET6:
 360                 V_route_cb.ip6_count--;
 361                 break;
 362         }
 363         V_route_cb.any_count--;
 364         RTSOCK_UNLOCK();
 365         raw_usrreqs.pru_detach(so);
 366 }
 367
 368 static int
 369 rts_disconnect(struct socket *so)
 370 {
 371
 372         return (raw_usrreqs.pru_disconnect(so));
 373 }
 374
 375 /* pru_listen is EOPNOTSUPP */
 376
 377 static int
 378 rts_peeraddr(struct socket *so, struct sockaddr **nam)
 379 {
 380
 381         return (raw_usrreqs.pru_peeraddr(so, nam));
 382 }
 383
 384 /* pru_rcvd is EOPNOTSUPP */
 385 /* pru_rcvoob is EOPNOTSUPP */
 386
 387 static int
 388 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
 389          struct mbuf *control, struct thread *td)
 390 {
 391
 392         return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
 393 }
 394
 395 /* pru_sense is null */
 396
 397 static int
 398 rts_shutdown(struct socket *so)
 399 {
 400
 401         return (raw_usrreqs.pru_shutdown(so));
 402 }
 403
 404 static int
 405 rts_sockaddr(struct socket *so, struct sockaddr **nam)
 406 {
 407
 408         return (raw_usrreqs.pru_sockaddr(so, nam));
 409 }
 410
 411 static struct pr_usrreqs route_usrreqs = {
 412         .pru_abort =            rts_abort,
 413         .pru_attach =           rts_attach,
 414         .pru_bind =             rts_bind,
 415         .pru_connect =          rts_connect,
 416         .pru_detach =           rts_detach,
 417         .pru_disconnect =       rts_disconnect,
 418         .pru_peeraddr =         rts_peeraddr,
 419         .pru_send =             rts_send,
 420         .pru_shutdown =         rts_shutdown,
 421         .pru_sockaddr =         rts_sockaddr,
 422         .pru_close =            rts_close,
 423 };
 424
 425 #ifndef _SOCKADDR_UNION_DEFINED
 426 #define _SOCKADDR_UNION_DEFINED
 427 /*
 428  * The union of all possible address formats we handle.
 429  */
 430 union sockaddr_union {
 431         struct sockaddr         sa;
 432         struct sockaddr_in      sin;
 433         struct sockaddr_in6     sin6;
 434 };
 435 #endif /* _SOCKADDR_UNION_DEFINED */
 436
 437 static int
 438 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
 439     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
 440 {
 441
 442         /* First, see if the returned address is part of the jail. */
 443         if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
 444                 info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
 445                 return (0);
 446         }
 447
 448         switch (info->rti_info[RTAX_DST]->sa_family) {
 449 #ifdef INET
 450         case AF_INET:
 451         {
 452                 struct in_addr ia;
 453                 struct ifaddr *ifa;
 454                 int found;
 455
 456                 found = 0;
 457                 /*
 458                  * Try to find an address on the given outgoing interface
 459                  * that belongs to the jail.
 460                  */
 461                 IF_ADDR_RLOCK(ifp);
 462                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 463                         struct sockaddr *sa;
 464                         sa = ifa->ifa_addr;
 465                         if (sa->sa_family != AF_INET)
 466                                 continue;
 467                         ia = ((struct sockaddr_in *)sa)->sin_addr;
 468                         if (prison_check_ip4(cred, &ia) == 0) {
 469                                 found = 1;
 470                                 break;
 471                         }
 472                 }
 473                 IF_ADDR_RUNLOCK(ifp);
 474                 if (!found) {
 475                         /*
 476                          * As a last resort return the 'default' jail address.
 477                          */
 478                         ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
 479                             sin_addr;
 480                         if (prison_get_ip4(cred, &ia) != 0)
 481                                 return (ESRCH);
 482                 }
 483                 bzero(&saun->sin, sizeof(struct sockaddr_in));
 484                 saun->sin.sin_len = sizeof(struct sockaddr_in);
 485                 saun->sin.sin_family = AF_INET;
 486                 saun->sin.sin_addr.s_addr = ia.s_addr;
 487                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
 488                 break;
 489         }
 490 #endif
 491 #ifdef INET6
 492         case AF_INET6:
 493         {
 494                 struct in6_addr ia6;
 495                 struct ifaddr *ifa;
 496                 int found;
 497
 498                 found = 0;
 499                 /*
 500                  * Try to find an address on the given outgoing interface
 501                  * that belongs to the jail.
 502                  */
 503                 IF_ADDR_RLOCK(ifp);
 504                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 505                         struct sockaddr *sa;
 506                         sa = ifa->ifa_addr;
 507                         if (sa->sa_family != AF_INET6)
 508                                 continue;
 509                         bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
 510                             &ia6, sizeof(struct in6_addr));
 511                         if (prison_check_ip6(cred, &ia6) == 0) {
 512                                 found = 1;
 513                                 break;
 514                         }
 515                 }
 516                 IF_ADDR_RUNLOCK(ifp);
 517                 if (!found) {
 518                         /*
 519                          * As a last resort return the 'default' jail address.
 520                          */
 521                         ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
 522                             sin6_addr;
 523                         if (prison_get_ip6(cred, &ia6) != 0)
 524                                 return (ESRCH);
 525                 }
 526                 bzero(&saun->sin6, sizeof(struct sockaddr_in6));
 527                 saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
 528                 saun->sin6.sin6_family = AF_INET6;
 529                 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
 530                 if (sa6_recoverscope(&saun->sin6) != 0)
 531                         return (ESRCH);
 532                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
 533                 break;
 534         }
 535 #endif
 536         default:
 537                 return (ESRCH);
 538         }
 539         return (0);
 540 }
 541
 542 /*ARGSUSED*/
 543 static int
 544 route_output(struct mbuf *m, struct socket *so, ...)
 545 {
 546         struct rt_msghdr *rtm = NULL;
 547         struct rtentry *rt = NULL;
 548         struct rib_head *rnh;
 549         struct rt_addrinfo info;
 550         struct sockaddr_storage ss;
 551 #ifdef INET6
 552         struct sockaddr_in6 *sin6;
 553         int i, rti_need_deembed = 0;
 554 #endif
 555         int alloc_len = 0, len, error = 0, fibnum;
 556         struct ifnet *ifp = NULL;
 557         union sockaddr_union saun;
 558         sa_family_t saf = AF_UNSPEC;
 559         struct rawcb *rp = NULL;
 560         struct walkarg w;
 561
 562         fibnum = so->so_fibnum;
 563
 564 #define senderr(e) { error = e; goto flush;}
 565         if (m == NULL || ((m->m_len < sizeof(long)) &&
 566                        (m = m_pullup(m, sizeof(long))) == NULL))
 567                 return (ENOBUFS);
 568         if ((m->m_flags & M_PKTHDR) == 0)
 569                 panic("route_output");
 570         len = m->m_pkthdr.len;
 571         if (len < sizeof(*rtm) ||
 572             len != mtod(m, struct rt_msghdr *)->rtm_msglen)
 573                 senderr(EINVAL);
 574
 575         /*
 576          * Most of current messages are in range 200-240 bytes,
 577          * minimize possible re-allocation on reply using larger size
 578          * buffer aligned on 1k boundaty.
 579          */
 580         alloc_len = roundup2(len, 1024);
 581         if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL)
 582                 senderr(ENOBUFS);
 583
 584         m_copydata(m, 0, len, (caddr_t)rtm);
 585         bzero(&info, sizeof(info));
 586         bzero(&w, sizeof(w));
 587
 588         if (rtm->rtm_version != RTM_VERSION) {
 589                 /* Do not touch message since format is unknown */
 590                 free(rtm, M_TEMP);
 591                 rtm = NULL;
 592                 senderr(EPROTONOSUPPORT);
 593         }
 594
 595         /*
 596          * Starting from here, it is possible
 597          * to alter original message and insert
 598          * caller PID and error value.
 599          */
 600
 601         rtm->rtm_pid = curproc->p_pid;
 602         info.rti_addrs = rtm->rtm_addrs;
 603
 604         info.rti_mflags = rtm->rtm_inits;
 605         info.rti_rmx = &rtm->rtm_rmx;
 606
 607         /*
 608          * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
 609          * link-local address because rtrequest requires addresses with
 610          * embedded scope id.
 611          */
 612         if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info))
 613                 senderr(EINVAL);
 614
 615         info.rti_flags = rtm->rtm_flags;
 616         if (info.rti_info[RTAX_DST] == NULL ||
 617             info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
 618             (info.rti_info[RTAX_GATEWAY] != NULL &&
 619              info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
 620                 senderr(EINVAL);
 621         saf = info.rti_info[RTAX_DST]->sa_family;
 622         /*
 623          * Verify that the caller has the appropriate privilege; RTM_GET
 624          * is the only operation the non-superuser is allowed.
 625          */
 626         if (rtm->rtm_type != RTM_GET) {
 627                 error = priv_check(curthread, PRIV_NET_ROUTE);
 628                 if (error)
 629                         senderr(error);
 630         }
 631
 632         /*
 633          * The given gateway address may be an interface address.
 634          * For example, issuing a "route change" command on a route
 635          * entry that was created from a tunnel, and the gateway
 636          * address given is the local end point. In this case the
 637          * RTF_GATEWAY flag must be cleared or the destination will
 638          * not be reachable even though there is no error message.
 639          */
 640         if (info.rti_info[RTAX_GATEWAY] != NULL &&
 641             info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
 642                 struct rt_addrinfo ginfo;
 643                 struct sockaddr *gdst;
 644
 645                 bzero(&ginfo, sizeof(ginfo));
 646                 bzero(&ss, sizeof(ss));
 647                 ss.ss_len = sizeof(ss);
 648
 649                 ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss;
 650                 gdst = info.rti_info[RTAX_GATEWAY];
 651
 652                 /*
 653                  * A host route through the loopback interface is
 654                  * installed for each interface adddress. In pre 8.0
 655                  * releases the interface address of a PPP link type
 656                  * is not reachable locally. This behavior is fixed as
 657                  * part of the new L2/L3 redesign and rewrite work. The
 658                  * signature of this interface address route is the
 659                  * AF_LINK sa_family type of the rt_gateway, and the
 660                  * rt_ifp has the IFF_LOOPBACK flag set.
 661                  */
 662                 if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) {
 663                         if (ss.ss_family == AF_LINK &&
 664                             ginfo.rti_ifp->if_flags & IFF_LOOPBACK) {
 665                                 info.rti_flags &= ~RTF_GATEWAY;
 666                                 info.rti_flags |= RTF_GWFLAG_COMPAT;
 667                         }
 668                         rib_free_info(&ginfo);
 669                 }
 670         }
 671
 672         switch (rtm->rtm_type) {
 673                 struct rtentry *saved_nrt;
 674
 675         case RTM_ADD:
 676         case RTM_CHANGE:
 677                 if (info.rti_info[RTAX_GATEWAY] == NULL)
 678                         senderr(EINVAL);
 679                 saved_nrt = NULL;
 680
 681                 /* support for new ARP code */
 682                 if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
 683                     (rtm->rtm_flags & RTF_LLDATA) != 0) {
 684                         error = lla_rt_output(rtm, &info);
 685 #ifdef INET6
 686                         if (error == 0)
 687                                 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 688 #endif
 689                         break;
 690                 }
 691                 error = rtrequest1_fib(rtm->rtm_type, &info, &saved_nrt,
 692                     fibnum);
 693                 if (error == 0 && saved_nrt != NULL) {
 694 #ifdef INET6
 695                         rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 696 #endif
 697                         RT_LOCK(saved_nrt);
 698                         rtm->rtm_index = saved_nrt->rt_ifp->if_index;
 699                         RT_REMREF(saved_nrt);
 700                         RT_UNLOCK(saved_nrt);
 701                 }
 702                 break;
 703
 704         case RTM_DELETE:
 705                 saved_nrt = NULL;
 706                 /* support for new ARP code */
 707                 if (info.rti_info[RTAX_GATEWAY] &&
 708                     (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
 709                     (rtm->rtm_flags & RTF_LLDATA) != 0) {
 710                         error = lla_rt_output(rtm, &info);
 711 #ifdef INET6
 712                         if (error == 0)
 713                                 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 714 #endif
 715                         break;
 716                 }
 717                 error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt, fibnum);
 718                 if (error == 0) {
 719                         RT_LOCK(saved_nrt);
 720                         rt = saved_nrt;
 721                         goto report;
 722                 }
 723 #ifdef INET6
 724                 /* rt_msg2() will not be used when RTM_DELETE fails. */
 725                 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 726 #endif
 727                 break;
 728
 729         case RTM_GET:
 730                 rnh = rt_tables_get_rnh(fibnum, saf);
 731                 if (rnh == NULL)
 732                         senderr(EAFNOSUPPORT);
 733
 734                 RIB_RLOCK(rnh);
 735
 736                 if (info.rti_info[RTAX_NETMASK] == NULL &&
 737                     rtm->rtm_type == RTM_GET) {
 738                         /*
 739                          * Provide longest prefix match for
 740                          * address lookup (no mask).
 741                          * 'route -n get addr'
 742                          */
 743                         rt = (struct rtentry *) rnh->rnh_matchaddr(
 744                             info.rti_info[RTAX_DST], &rnh->head);
 745                 } else
 746                         rt = (struct rtentry *) rnh->rnh_lookup(
 747                             info.rti_info[RTAX_DST],
 748                             info.rti_info[RTAX_NETMASK], &rnh->head);
 749
 750                 if (rt == NULL) {
 751                         RIB_RUNLOCK(rnh);
 752                         senderr(ESRCH);
 753                 }
 754 #ifdef RADIX_MPATH
 755                 /*
 756                  * for RTM_CHANGE/LOCK, if we got multipath routes,
 757                  * we require users to specify a matching RTAX_GATEWAY.
 758                  *
 759                  * for RTM_GET, gate is optional even with multipath.
 760                  * if gate == NULL the first match is returned.
 761                  * (no need to call rt_mpath_matchgate if gate == NULL)
 762                  */
 763                 if (rt_mpath_capable(rnh) &&
 764                     (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
 765                         rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
 766                         if (!rt) {
 767                                 RIB_RUNLOCK(rnh);
 768                                 senderr(ESRCH);
 769                         }
 770                 }
 771 #endif
 772                 /*
 773                  * If performing proxied L2 entry insertion, and
 774                  * the actual PPP host entry is found, perform
 775                  * another search to retrieve the prefix route of
 776                  * the local end point of the PPP link.
 777                  */
 778                 if (rtm->rtm_flags & RTF_ANNOUNCE) {
 779                         struct sockaddr laddr;
 780
 781                         if (rt->rt_ifp != NULL &&
 782                             rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
 783                                 struct ifaddr *ifa;
 784
 785                                 ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1,
 786                                                 RT_ALL_FIBS);
 787                                 if (ifa != NULL)
 788                                         rt_maskedcopy(ifa->ifa_addr,
 789                                                       &laddr,
 790                                                       ifa->ifa_netmask);
 791                         } else
 792                                 rt_maskedcopy(rt->rt_ifa->ifa_addr,
 793                                               &laddr,
 794                                               rt->rt_ifa->ifa_netmask);
 795                         /*
 796                          * refactor rt and no lock operation necessary
 797                          */
 798                         rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr,
 799                             &rnh->head);
 800                         if (rt == NULL) {
 801                                 RIB_RUNLOCK(rnh);
 802                                 senderr(ESRCH);
 803                         }
 804                 }
 805                 RT_LOCK(rt);
 806                 RT_ADDREF(rt);
 807                 RIB_RUNLOCK(rnh);
 808
 809 report:
 810                 RT_LOCK_ASSERT(rt);
 811                 if ((rt->rt_flags & RTF_HOST) == 0
 812                     ? jailed_without_vnet(curthread->td_ucred)
 813                     : prison_if(curthread->td_ucred,
 814                     rt_key(rt)) != 0) {
 815                         RT_UNLOCK(rt);
 816                         senderr(ESRCH);
 817                 }
 818                 info.rti_info[RTAX_DST] = rt_key(rt);
 819                 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
 820                 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
 821                     rt_mask(rt), &ss);
 822                 info.rti_info[RTAX_GENMASK] = 0;
 823                 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
 824                         ifp = rt->rt_ifp;
 825                         if (ifp) {
 826                                 info.rti_info[RTAX_IFP] =
 827                                     ifp->if_addr->ifa_addr;
 828                                 error = rtm_get_jailed(&info, ifp, rt,
 829                                     &saun, curthread->td_ucred);
 830                                 if (error != 0) {
 831                                         RT_UNLOCK(rt);
 832                                         senderr(error);
 833                                 }
 834                                 if (ifp->if_flags & IFF_POINTOPOINT)
 835                                         info.rti_info[RTAX_BRD] =
 836                                             rt->rt_ifa->ifa_dstaddr;
 837                                 rtm->rtm_index = ifp->if_index;
 838                         } else {
 839                                 info.rti_info[RTAX_IFP] = NULL;
 840                                 info.rti_info[RTAX_IFA] = NULL;
 841                         }
 842                 } else if ((ifp = rt->rt_ifp) != NULL) {
 843                         rtm->rtm_index = ifp->if_index;
 844                 }
 845
 846                 /* Check if we need to realloc storage */
 847                 rtsock_msg_buffer(rtm->rtm_type, &info, NULL, &len);
 848                 if (len > alloc_len) {
 849                         struct rt_msghdr *new_rtm;
 850                         new_rtm = malloc(len, M_TEMP, M_NOWAIT);
 851                         if (new_rtm == NULL) {
 852                                 RT_UNLOCK(rt);
 853                                 senderr(ENOBUFS);
 854                         }
 855                         bcopy(rtm, new_rtm, rtm->rtm_msglen);
 856                         free(rtm, M_TEMP);
 857                         rtm = new_rtm;
 858                         alloc_len = len;
 859                 }
 860
 861                 w.w_tmem = (caddr_t)rtm;
 862                 w.w_tmemsize = alloc_len;
 863                 rtsock_msg_buffer(rtm->rtm_type, &info, &w, &len);
 864
 865                 if (rt->rt_flags & RTF_GWFLAG_COMPAT)
 866                         rtm->rtm_flags = RTF_GATEWAY |
 867                                 (rt->rt_flags & ~RTF_GWFLAG_COMPAT);
 868                 else
 869                         rtm->rtm_flags = rt->rt_flags;
 870                 rt_getmetrics(rt, &rtm->rtm_rmx);
 871                 rtm->rtm_addrs = info.rti_addrs;
 872
 873                 RT_UNLOCK(rt);
 874                 break;
 875
 876         default:
 877                 senderr(EOPNOTSUPP);
 878         }
 879
 880 flush:
 881         if (rt != NULL)
 882                 RTFREE(rt);
 883         /*
 884          * Check to see if we don't want our own messages.
 885          */
 886         if ((so->so_options & SO_USELOOPBACK) == 0) {
 887                 if (V_route_cb.any_count <= 1) {
 888                         if (rtm != NULL)
 889                                 free(rtm, M_TEMP);
 890                         m_freem(m);
 891                         return (error);
 892                 }
 893                 /* There is another listener, so construct message */
 894                 rp = sotorawcb(so);
 895         }
 896
 897         if (rtm != NULL) {
 898 #ifdef INET6
 899                 if (rti_need_deembed) {
 900                         /* sin6_scope_id is recovered before sending rtm. */
 901                         sin6 = (struct sockaddr_in6 *)&ss;
 902                         for (i = 0; i < RTAX_MAX; i++) {
 903                                 if (info.rti_info[i] == NULL)
 904                                         continue;
 905                                 if (info.rti_info[i]->sa_family != AF_INET6)
 906                                         continue;
 907                                 bcopy(info.rti_info[i], sin6, sizeof(*sin6));
 908                                 if (sa6_recoverscope(sin6) == 0)
 909                                         bcopy(sin6, info.rti_info[i],
 910                                                     sizeof(*sin6));
 911                         }
 912                 }
 913 #endif
 914                 if (error != 0)
 915                         rtm->rtm_errno = error;
 916                 else
 917                         rtm->rtm_flags |= RTF_DONE;
 918
 919                 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
 920                 if (m->m_pkthdr.len < rtm->rtm_msglen) {
 921                         m_freem(m);
 922                         m = NULL;
 923                 } else if (m->m_pkthdr.len > rtm->rtm_msglen)
 924                         m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
 925
 926                 free(rtm, M_TEMP);
 927         }
 928         if (m != NULL) {
 929                 M_SETFIB(m, fibnum);
 930                 m->m_flags |= RTS_FILTER_FIB;
 931                 if (rp) {
 932                         /*
 933                          * XXX insure we don't get a copy by
 934                          * invalidating our protocol
 935                          */
 936                         unsigned short family = rp->rcb_proto.sp_family;
 937                         rp->rcb_proto.sp_family = 0;
 938                         rt_dispatch(m, saf);
 939                         rp->rcb_proto.sp_family = family;
 940                 } else
 941                         rt_dispatch(m, saf);
 942         }
 943
 944         return (error);
 945 }
 946
 947 static void
 948 rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out)
 949 {
 950
 951         bzero(out, sizeof(*out));
 952         out->rmx_mtu = rt->rt_mtu;
 953         out->rmx_weight = rt->rt_weight;
 954         out->rmx_pksent = counter_u64_fetch(rt->rt_pksent);
 955         /* Kernel -> userland timebase conversion. */
 956         out->rmx_expire = rt->rt_expire ?
 957             rt->rt_expire - time_uptime + time_second : 0;
 958 }
 959
 960 /*
 961  * Extract the addresses of the passed sockaddrs.
 962  * Do a little sanity checking so as to avoid bad memory references.
 963  * This data is derived straight from userland.
 964  */
 965 static int
 966 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
 967 {
 968         struct sockaddr *sa;
 969         int i;
 970
 971         for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
 972                 if ((rtinfo->rti_addrs & (1 << i)) == 0)
 973                         continue;
 974                 sa = (struct sockaddr *)cp;
 975                 /*
 976                  * It won't fit.
 977                  */
 978                 if (cp + sa->sa_len > cplim)
 979                         return (EINVAL);
 980                 /*
 981                  * there are no more.. quit now
 982                  * If there are more bits, they are in error.
 983                  * I've seen this. route(1) can evidently generate these.
 984                  * This causes kernel to core dump.
 985                  * for compatibility, If we see this, point to a safe address.
 986                  */
 987                 if (sa->sa_len == 0) {
 988                         rtinfo->rti_info[i] = &sa_zero;
 989                         return (0); /* should be EINVAL but for compat */
 990                 }
 991                 /* accept it */
 992 #ifdef INET6
 993                 if (sa->sa_family == AF_INET6)
 994                         sa6_embedscope((struct sockaddr_in6 *)sa,
 995                             V_ip6_use_defzone);
 996 #endif
 997                 rtinfo->rti_info[i] = sa;
 998                 cp += SA_SIZE(sa);
 999         }
1000         return (0);
1001 }
1002
1003 /*
1004  * Fill in @dmask with valid netmask leaving original @smask
1005  * intact. Mostly used with radix netmasks.
1006  */
1007 static struct sockaddr *
1008 rtsock_fix_netmask(struct sockaddr *dst, struct sockaddr *smask,
1009     struct sockaddr_storage *dmask)
1010 {
1011         if (dst == NULL || smask == NULL)
1012                 return (NULL);
1013
1014         memset(dmask, 0, dst->sa_len);
1015         memcpy(dmask, smask, smask->sa_len);
1016         dmask->ss_len = dst->sa_len;
1017         dmask->ss_family = dst->sa_family;
1018
1019         return ((struct sockaddr *)dmask);
1020 }
1021
1022 /*
1023  * Writes information related to @rtinfo object to newly-allocated mbuf.
1024  * Assumes MCLBYTES is enough to construct any message.
1025  * Used for OS notifications of vaious events (if/ifa announces,etc)
1026  *
1027  * Returns allocated mbuf or NULL on failure.
1028  */
1029 static struct mbuf *
1030 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1031 {
1032         struct rt_msghdr *rtm;
1033         struct mbuf *m;
1034         int i;
1035         struct sockaddr *sa;
1036 #ifdef INET6
1037         struct sockaddr_storage ss;
1038         struct sockaddr_in6 *sin6;
1039 #endif
1040         int len, dlen;
1041
1042         switch (type) {
1043
1044         case RTM_DELADDR:
1045         case RTM_NEWADDR:
1046                 len = sizeof(struct ifa_msghdr);
1047                 break;
1048
1049         case RTM_DELMADDR:
1050         case RTM_NEWMADDR:
1051                 len = sizeof(struct ifma_msghdr);
1052                 break;
1053
1054         case RTM_IFINFO:
1055                 len = sizeof(struct if_msghdr);
1056                 break;
1057
1058         case RTM_IFANNOUNCE:
1059         case RTM_IEEE80211:
1060                 len = sizeof(struct if_announcemsghdr);
1061                 break;
1062
1063         default:
1064                 len = sizeof(struct rt_msghdr);
1065         }
1066
1067         /* XXXGL: can we use MJUMPAGESIZE cluster here? */
1068         KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1069         if (len > MHLEN)
1070                 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1071         else
1072                 m = m_gethdr(M_NOWAIT, MT_DATA);
1073         if (m == NULL)
1074                 return (m);
1075
1076         m->m_pkthdr.len = m->m_len = len;
1077         rtm = mtod(m, struct rt_msghdr *);
1078         bzero((caddr_t)rtm, len);
1079         for (i = 0; i < RTAX_MAX; i++) {
1080                 if ((sa = rtinfo->rti_info[i]) == NULL)
1081                         continue;
1082                 rtinfo->rti_addrs |= (1 << i);
1083                 dlen = SA_SIZE(sa);
1084 #ifdef INET6
1085                 if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1086                         sin6 = (struct sockaddr_in6 *)&ss;
1087                         bcopy(sa, sin6, sizeof(*sin6));
1088                         if (sa6_recoverscope(sin6) == 0)
1089                                 sa = (struct sockaddr *)sin6;
1090                 }
1091 #endif
1092                 m_copyback(m, len, dlen, (caddr_t)sa);
1093                 len += dlen;
1094         }
1095         if (m->m_pkthdr.len != len) {
1096                 m_freem(m);
1097                 return (NULL);
1098         }
1099         rtm->rtm_msglen = len;
1100         rtm->rtm_version = RTM_VERSION;
1101         rtm->rtm_type = type;
1102         return (m);
1103 }
1104
1105 /*
1106  * Writes information related to @rtinfo object to preallocated buffer.
1107  * Stores needed size in @plen. If @w is NULL, calculates size without
1108  * writing.
1109  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1110  *
1111  * Returns 0 on success.
1112  *
1113  */
1114 static int
1115 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1116 {
1117         int i;
1118         int len, buflen = 0, dlen;
1119         caddr_t cp = NULL;
1120         struct rt_msghdr *rtm = NULL;
1121 #ifdef INET6
1122         struct sockaddr_storage ss;
1123         struct sockaddr_in6 *sin6;
1124 #endif
1125 #ifdef COMPAT_FREEBSD32
1126         bool compat32 = false;
1127 #endif
1128
1129         switch (type) {
1130
1131         case RTM_DELADDR:
1132         case RTM_NEWADDR:
1133                 if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1134 #ifdef COMPAT_FREEBSD32
1135                         if (w->w_req->flags & SCTL_MASK32) {
1136                                 len = sizeof(struct ifa_msghdrl32);
1137                                 compat32 = true;
1138                         } else
1139 #endif
1140                                 len = sizeof(struct ifa_msghdrl);
1141                 } else
1142                         len = sizeof(struct ifa_msghdr);
1143                 break;
1144
1145         case RTM_IFINFO:
1146 #ifdef COMPAT_FREEBSD32
1147                 if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1148                         if (w->w_op == NET_RT_IFLISTL)
1149                                 len = sizeof(struct if_msghdrl32);
1150                         else
1151                                 len = sizeof(struct if_msghdr32);
1152                         compat32 = true;
1153                         break;
1154                 }
1155 #endif
1156                 if (w != NULL && w->w_op == NET_RT_IFLISTL)
1157                         len = sizeof(struct if_msghdrl);
1158                 else
1159                         len = sizeof(struct if_msghdr);
1160                 break;
1161
1162         case RTM_NEWMADDR:
1163                 len = sizeof(struct ifma_msghdr);
1164                 break;
1165
1166         default:
1167                 len = sizeof(struct rt_msghdr);
1168         }
1169
1170         if (w != NULL) {
1171                 rtm = (struct rt_msghdr *)w->w_tmem;
1172                 buflen = w->w_tmemsize - len;
1173                 cp = (caddr_t)w->w_tmem + len;
1174         }
1175
1176         rtinfo->rti_addrs = 0;
1177         for (i = 0; i < RTAX_MAX; i++) {
1178                 struct sockaddr *sa;
1179
1180                 if ((sa = rtinfo->rti_info[i]) == NULL)
1181                         continue;
1182                 rtinfo->rti_addrs |= (1 << i);
1183 #ifdef COMPAT_FREEBSD32
1184                 if (compat32)
1185                         dlen = SA_SIZE32(sa);
1186                 else
1187 #endif
1188                         dlen = SA_SIZE(sa);
1189                 if (cp != NULL && buflen >= dlen) {
1190 #ifdef INET6
1191                         if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1192                                 sin6 = (struct sockaddr_in6 *)&ss;
1193                                 bcopy(sa, sin6, sizeof(*sin6));
1194                                 if (sa6_recoverscope(sin6) == 0)
1195                                         sa = (struct sockaddr *)sin6;
1196                         }
1197 #endif
1198                         bcopy((caddr_t)sa, cp, (unsigned)dlen);
1199                         cp += dlen;
1200                         buflen -= dlen;
1201                 } else if (cp != NULL) {
1202                         /*
1203                          * Buffer too small. Count needed size
1204                          * and return with error.
1205                          */
1206                         cp = NULL;
1207                 }
1208
1209                 len += dlen;
1210         }
1211
1212         if (cp != NULL) {
1213                 dlen = ALIGN(len) - len;
1214                 if (buflen < dlen)
1215                         cp = NULL;
1216                 else {
1217                         bzero(cp, dlen);
1218                         cp += dlen;
1219                         buflen -= dlen;
1220                 }
1221         }
1222         len = ALIGN(len);
1223
1224         if (cp != NULL) {
1225                 /* fill header iff buffer is large enough */
1226                 rtm->rtm_version = RTM_VERSION;
1227                 rtm->rtm_type = type;
1228                 rtm->rtm_msglen = len;
1229         }
1230
1231         *plen = len;
1232
1233         if (w != NULL && cp == NULL)
1234                 return (ENOBUFS);
1235
1236         return (0);
1237 }
1238
1239 /*
1240  * This routine is called to generate a message from the routing
1241  * socket indicating that a redirect has occurred, a routing lookup
1242  * has failed, or that a protocol has detected timeouts to a particular
1243  * destination.
1244  */
1245 void
1246 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1247     int fibnum)
1248 {
1249         struct rt_msghdr *rtm;
1250         struct mbuf *m;
1251         struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1252
1253         if (V_route_cb.any_count == 0)
1254                 return;
1255         m = rtsock_msg_mbuf(type, rtinfo);
1256         if (m == NULL)
1257                 return;
1258
1259         if (fibnum != RT_ALL_FIBS) {
1260                 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1261                     "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1262                 M_SETFIB(m, fibnum);
1263                 m->m_flags |= RTS_FILTER_FIB;
1264         }
1265
1266         rtm = mtod(m, struct rt_msghdr *);
1267         rtm->rtm_flags = RTF_DONE | flags;
1268         rtm->rtm_errno = error;
1269         rtm->rtm_addrs = rtinfo->rti_addrs;
1270         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1271 }
1272
1273 void
1274 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1275 {
1276
1277         rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1278 }
1279
1280 /*
1281  * This routine is called to generate a message from the routing
1282  * socket indicating that the status of a network interface has changed.
1283  */
1284 void
1285 rt_ifmsg(struct ifnet *ifp)
1286 {
1287         struct if_msghdr *ifm;
1288         struct mbuf *m;
1289         struct rt_addrinfo info;
1290
1291         if (V_route_cb.any_count == 0)
1292                 return;
1293         bzero((caddr_t)&info, sizeof(info));
1294         m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1295         if (m == NULL)
1296                 return;
1297         ifm = mtod(m, struct if_msghdr *);
1298         ifm->ifm_index = ifp->if_index;
1299         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1300         if_data_copy(ifp, &ifm->ifm_data);
1301         ifm->ifm_addrs = 0;
1302         rt_dispatch(m, AF_UNSPEC);
1303 }
1304
1305 /*
1306  * Announce interface address arrival/withdraw.
1307  * Please do not call directly, use rt_addrmsg().
1308  * Assume input data to be valid.
1309  * Returns 0 on success.
1310  */
1311 int
1312 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1313 {
1314         struct rt_addrinfo info;
1315         struct sockaddr *sa;
1316         int ncmd;
1317         struct mbuf *m;
1318         struct ifa_msghdr *ifam;
1319         struct ifnet *ifp = ifa->ifa_ifp;
1320         struct sockaddr_storage ss;
1321
1322         if (V_route_cb.any_count == 0)
1323                 return (0);
1324
1325         ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1326
1327         bzero((caddr_t)&info, sizeof(info));
1328         info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1329         info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1330         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1331             info.rti_info[RTAX_IFP], ifa->ifa_netmask, &ss);
1332         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1333         if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1334                 return (ENOBUFS);
1335         ifam = mtod(m, struct ifa_msghdr *);
1336         ifam->ifam_index = ifp->if_index;
1337         ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1338         ifam->ifam_flags = ifa->ifa_flags;
1339         ifam->ifam_addrs = info.rti_addrs;
1340
1341         if (fibnum != RT_ALL_FIBS) {
1342                 M_SETFIB(m, fibnum);
1343                 m->m_flags |= RTS_FILTER_FIB;
1344         }
1345
1346         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1347
1348         return (0);
1349 }
1350
1351 /*
1352  * Announce route addition/removal.
1353  * Please do not call directly, use rt_routemsg().
1354  * Note that @rt data MAY be inconsistent/invalid:
1355  * if some userland app sends us "invalid" route message (invalid mask,
1356  * no dst, wrong address families, etc...) we need to pass it back
1357  * to app (and any other rtsock consumers) with rtm_errno field set to
1358  * non-zero value.
1359  *
1360  * Returns 0 on success.
1361  */
1362 int
1363 rtsock_routemsg(int cmd, struct ifnet *ifp, int error, struct rtentry *rt,
1364     int fibnum)
1365 {
1366         struct rt_addrinfo info;
1367         struct sockaddr *sa;
1368         struct mbuf *m;
1369         struct rt_msghdr *rtm;
1370         struct sockaddr_storage ss;
1371
1372         if (V_route_cb.any_count == 0)
1373                 return (0);
1374
1375         bzero((caddr_t)&info, sizeof(info));
1376         info.rti_info[RTAX_DST] = sa = rt_key(rt);
1377         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(sa, rt_mask(rt), &ss);
1378         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1379         if ((m = rtsock_msg_mbuf(cmd, &info)) == NULL)
1380                 return (ENOBUFS);
1381         rtm = mtod(m, struct rt_msghdr *);
1382         rtm->rtm_index = ifp->if_index;
1383         rtm->rtm_flags |= rt->rt_flags;
1384         rtm->rtm_errno = error;
1385         rtm->rtm_addrs = info.rti_addrs;
1386
1387         if (fibnum != RT_ALL_FIBS) {
1388                 M_SETFIB(m, fibnum);
1389                 m->m_flags |= RTS_FILTER_FIB;
1390         }
1391
1392         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1393
1394         return (0);
1395 }
1396
1397 /*
1398  * This is the analogue to the rt_newaddrmsg which performs the same
1399  * function but for multicast group memberhips.  This is easier since
1400  * there is no route state to worry about.
1401  */
1402 void
1403 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1404 {
1405         struct rt_addrinfo info;
1406         struct mbuf *m = NULL;
1407         struct ifnet *ifp = ifma->ifma_ifp;
1408         struct ifma_msghdr *ifmam;
1409
1410         if (V_route_cb.any_count == 0)
1411                 return;
1412
1413         bzero((caddr_t)&info, sizeof(info));
1414         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1415         info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1416         /*
1417          * If a link-layer address is present, present it as a ``gateway''
1418          * (similarly to how ARP entries, e.g., are presented).
1419          */
1420         info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1421         m = rtsock_msg_mbuf(cmd, &info);
1422         if (m == NULL)
1423                 return;
1424         ifmam = mtod(m, struct ifma_msghdr *);
1425         KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1426             __func__));
1427         ifmam->ifmam_index = ifp->if_index;
1428         ifmam->ifmam_addrs = info.rti_addrs;
1429         rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1430 }
1431
1432 static struct mbuf *
1433 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1434         struct rt_addrinfo *info)
1435 {
1436         struct if_announcemsghdr *ifan;
1437         struct mbuf *m;
1438
1439         if (V_route_cb.any_count == 0)
1440                 return NULL;
1441         bzero((caddr_t)info, sizeof(*info));
1442         m = rtsock_msg_mbuf(type, info);
1443         if (m != NULL) {
1444                 ifan = mtod(m, struct if_announcemsghdr *);
1445                 ifan->ifan_index = ifp->if_index;
1446                 strlcpy(ifan->ifan_name, ifp->if_xname,
1447                         sizeof(ifan->ifan_name));
1448                 ifan->ifan_what = what;
1449         }
1450         return m;
1451 }
1452
1453 /*
1454  * This is called to generate routing socket messages indicating
1455  * IEEE80211 wireless events.
1456  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1457  */
1458 void
1459 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1460 {
1461         struct mbuf *m;
1462         struct rt_addrinfo info;
1463
1464         m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1465         if (m != NULL) {
1466                 /*
1467                  * Append the ieee80211 data.  Try to stick it in the
1468                  * mbuf containing the ifannounce msg; otherwise allocate
1469                  * a new mbuf and append.
1470                  *
1471                  * NB: we assume m is a single mbuf.
1472                  */
1473                 if (data_len > M_TRAILINGSPACE(m)) {
1474                         struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1475                         if (n == NULL) {
1476                                 m_freem(m);
1477                                 return;
1478                         }
1479                         bcopy(data, mtod(n, void *), data_len);
1480                         n->m_len = data_len;
1481                         m->m_next = n;
1482                 } else if (data_len > 0) {
1483                         bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1484                         m->m_len += data_len;
1485                 }
1486                 if (m->m_flags & M_PKTHDR)
1487                         m->m_pkthdr.len += data_len;
1488                 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1489                 rt_dispatch(m, AF_UNSPEC);
1490         }
1491 }
1492
1493 /*
1494  * This is called to generate routing socket messages indicating
1495  * network interface arrival and departure.
1496  */
1497 void
1498 rt_ifannouncemsg(struct ifnet *ifp, int what)
1499 {
1500         struct mbuf *m;
1501         struct rt_addrinfo info;
1502
1503         m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1504         if (m != NULL)
1505                 rt_dispatch(m, AF_UNSPEC);
1506 }
1507
1508 static void
1509 rt_dispatch(struct mbuf *m, sa_family_t saf)
1510 {
1511         struct m_tag *tag;
1512
1513         /*
1514          * Preserve the family from the sockaddr, if any, in an m_tag for
1515          * use when injecting the mbuf into the routing socket buffer from
1516          * the netisr.
1517          */
1518         if (saf != AF_UNSPEC) {
1519                 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1520                     M_NOWAIT);
1521                 if (tag == NULL) {
1522                         m_freem(m);
1523                         return;
1524                 }
1525                 *(unsigned short *)(tag + 1) = saf;
1526                 m_tag_prepend(m, tag);
1527         }
1528 #ifdef VIMAGE
1529         if (V_loif)
1530                 m->m_pkthdr.rcvif = V_loif;
1531         else {
1532                 m_freem(m);
1533                 return;
1534         }
1535 #endif
1536         netisr_queue(NETISR_ROUTE, m);  /* mbuf is free'd on failure. */
1537 }
1538
1539 /*
1540  * This is used in dumping the kernel table via sysctl().
1541  */
1542 static int
1543 sysctl_dumpentry(struct radix_node *rn, void *vw)
1544 {
1545         struct walkarg *w = vw;
1546         struct rtentry *rt = (struct rtentry *)rn;
1547         int error = 0, size;
1548         struct rt_addrinfo info;
1549         struct sockaddr_storage ss;
1550
1551         IFNET_RLOCK_NOSLEEP_ASSERT();
1552
1553         if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1554                 return 0;
1555         if ((rt->rt_flags & RTF_HOST) == 0
1556             ? jailed_without_vnet(w->w_req->td->td_ucred)
1557             : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1558                 return (0);
1559         bzero((caddr_t)&info, sizeof(info));
1560         info.rti_info[RTAX_DST] = rt_key(rt);
1561         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1562         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
1563             rt_mask(rt), &ss);
1564         info.rti_info[RTAX_GENMASK] = 0;
1565         if (rt->rt_ifp && !(rt->rt_ifp->if_flags & IFF_DYING)) {
1566                 info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1567                 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1568                 if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1569                         info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1570         }
1571         if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
1572                 return (error);
1573         if (w->w_req && w->w_tmem) {
1574                 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1575
1576                 bzero(&rtm->rtm_index,
1577                     sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
1578                 if (rt->rt_flags & RTF_GWFLAG_COMPAT)
1579                         rtm->rtm_flags = RTF_GATEWAY |
1580                                 (rt->rt_flags & ~RTF_GWFLAG_COMPAT);
1581                 else
1582                         rtm->rtm_flags = rt->rt_flags;
1583                 rt_getmetrics(rt, &rtm->rtm_rmx);
1584                 rtm->rtm_index = rt->rt_ifp->if_index;
1585                 rtm->rtm_addrs = info.rti_addrs;
1586                 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1587                 return (error);
1588         }
1589         return (error);
1590 }
1591
1592 static int
1593 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
1594     struct rt_addrinfo *info, struct walkarg *w, int len)
1595 {
1596         struct if_msghdrl *ifm;
1597         struct if_data *ifd;
1598
1599         ifm = (struct if_msghdrl *)w->w_tmem;
1600
1601 #ifdef COMPAT_FREEBSD32
1602         if (w->w_req->flags & SCTL_MASK32) {
1603                 struct if_msghdrl32 *ifm32;
1604
1605                 ifm32 = (struct if_msghdrl32 *)ifm;
1606                 ifm32->ifm_addrs = info->rti_addrs;
1607                 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1608                 ifm32->ifm_index = ifp->if_index;
1609                 ifm32->_ifm_spare1 = 0;
1610                 ifm32->ifm_len = sizeof(*ifm32);
1611                 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1612                 ifm32->_ifm_spare2 = 0;
1613                 ifd = &ifm32->ifm_data;
1614         } else
1615 #endif
1616         {
1617                 ifm->ifm_addrs = info->rti_addrs;
1618                 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1619                 ifm->ifm_index = ifp->if_index;
1620                 ifm->_ifm_spare1 = 0;
1621                 ifm->ifm_len = sizeof(*ifm);
1622                 ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1623                 ifm->_ifm_spare2 = 0;
1624                 ifd = &ifm->ifm_data;
1625         }
1626
1627         memcpy(ifd, src_ifd, sizeof(*ifd));
1628
1629         return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1630 }
1631
1632 static int
1633 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
1634     struct rt_addrinfo *info, struct walkarg *w, int len)
1635 {
1636         struct if_msghdr *ifm;
1637         struct if_data *ifd;
1638
1639         ifm = (struct if_msghdr *)w->w_tmem;
1640
1641 #ifdef COMPAT_FREEBSD32
1642         if (w->w_req->flags & SCTL_MASK32) {
1643                 struct if_msghdr32 *ifm32;
1644
1645                 ifm32 = (struct if_msghdr32 *)ifm;
1646                 ifm32->ifm_addrs = info->rti_addrs;
1647                 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1648                 ifm32->ifm_index = ifp->if_index;
1649                 ifm32->_ifm_spare1 = 0;
1650                 ifd = &ifm32->ifm_data;
1651         } else
1652 #endif
1653         {
1654                 ifm->ifm_addrs = info->rti_addrs;
1655                 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1656                 ifm->ifm_index = ifp->if_index;
1657                 ifm->_ifm_spare1 = 0;
1658                 ifd = &ifm->ifm_data;
1659         }
1660
1661         memcpy(ifd, src_ifd, sizeof(*ifd));
1662
1663         return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1664 }
1665
1666 static int
1667 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1668     struct walkarg *w, int len)
1669 {
1670         struct ifa_msghdrl *ifam;
1671         struct if_data *ifd;
1672
1673         ifam = (struct ifa_msghdrl *)w->w_tmem;
1674
1675 #ifdef COMPAT_FREEBSD32
1676         if (w->w_req->flags & SCTL_MASK32) {
1677                 struct ifa_msghdrl32 *ifam32;
1678
1679                 ifam32 = (struct ifa_msghdrl32 *)ifam;
1680                 ifam32->ifam_addrs = info->rti_addrs;
1681                 ifam32->ifam_flags = ifa->ifa_flags;
1682                 ifam32->ifam_index = ifa->ifa_ifp->if_index;
1683                 ifam32->_ifam_spare1 = 0;
1684                 ifam32->ifam_len = sizeof(*ifam32);
1685                 ifam32->ifam_data_off =
1686                     offsetof(struct ifa_msghdrl32, ifam_data);
1687                 ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
1688                 ifd = &ifam32->ifam_data;
1689         } else
1690 #endif
1691         {
1692                 ifam->ifam_addrs = info->rti_addrs;
1693                 ifam->ifam_flags = ifa->ifa_flags;
1694                 ifam->ifam_index = ifa->ifa_ifp->if_index;
1695                 ifam->_ifam_spare1 = 0;
1696                 ifam->ifam_len = sizeof(*ifam);
1697                 ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1698                 ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1699                 ifd = &ifam->ifam_data;
1700         }
1701
1702         bzero(ifd, sizeof(*ifd));
1703         ifd->ifi_datalen = sizeof(struct if_data);
1704         ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
1705         ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
1706         ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
1707         ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
1708
1709         /* Fixup if_data carp(4) vhid. */
1710         if (carp_get_vhid_p != NULL)
1711                 ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
1712
1713         return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1714 }
1715
1716 static int
1717 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1718     struct walkarg *w, int len)
1719 {
1720         struct ifa_msghdr *ifam;
1721
1722         ifam = (struct ifa_msghdr *)w->w_tmem;
1723         ifam->ifam_addrs = info->rti_addrs;
1724         ifam->ifam_flags = ifa->ifa_flags;
1725         ifam->ifam_index = ifa->ifa_ifp->if_index;
1726         ifam->_ifam_spare1 = 0;
1727         ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1728
1729         return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1730 }
1731
1732 static int
1733 sysctl_iflist(int af, struct walkarg *w)
1734 {
1735         struct ifnet *ifp;
1736         struct ifaddr *ifa;
1737         struct if_data ifd;
1738         struct rt_addrinfo info;
1739         int len, error = 0;
1740         struct sockaddr_storage ss;
1741
1742         bzero((caddr_t)&info, sizeof(info));
1743         bzero(&ifd, sizeof(ifd));
1744         IFNET_RLOCK_NOSLEEP();
1745         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1746                 if (w->w_arg && w->w_arg != ifp->if_index)
1747                         continue;
1748                 if_data_copy(ifp, &ifd);
1749                 IF_ADDR_RLOCK(ifp);
1750                 ifa = ifp->if_addr;
1751                 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1752                 error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
1753                 if (error != 0)
1754                         goto done;
1755                 info.rti_info[RTAX_IFP] = NULL;
1756                 if (w->w_req && w->w_tmem) {
1757                         if (w->w_op == NET_RT_IFLISTL)
1758                                 error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
1759                                     len);
1760                         else
1761                                 error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
1762                                     len);
1763                         if (error)
1764                                 goto done;
1765                 }
1766                 while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1767                         if (af && af != ifa->ifa_addr->sa_family)
1768                                 continue;
1769                         if (prison_if(w->w_req->td->td_ucred,
1770                             ifa->ifa_addr) != 0)
1771                                 continue;
1772                         info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1773                         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1774                             ifa->ifa_addr, ifa->ifa_netmask, &ss);
1775                         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1776                         error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
1777                         if (error != 0)
1778                                 goto done;
1779                         if (w->w_req && w->w_tmem) {
1780                                 if (w->w_op == NET_RT_IFLISTL)
1781                                         error = sysctl_iflist_ifaml(ifa, &info,
1782                                             w, len);
1783                                 else
1784                                         error = sysctl_iflist_ifam(ifa, &info,
1785                                             w, len);
1786                                 if (error)
1787                                         goto done;
1788                         }
1789                 }
1790                 IF_ADDR_RUNLOCK(ifp);
1791                 info.rti_info[RTAX_IFA] = NULL;
1792                 info.rti_info[RTAX_NETMASK] = NULL;
1793                 info.rti_info[RTAX_BRD] = NULL;
1794         }
1795 done:
1796         if (ifp != NULL)
1797                 IF_ADDR_RUNLOCK(ifp);
1798         IFNET_RUNLOCK_NOSLEEP();
1799         return (error);
1800 }
1801
1802 static int
1803 sysctl_ifmalist(int af, struct walkarg *w)
1804 {
1805         struct rt_addrinfo info;
1806         struct ifaddr *ifa;
1807         struct ifmultiaddr *ifma;
1808         struct ifnet *ifp;
1809         int error, len;
1810
1811         error = 0;
1812         bzero((caddr_t)&info, sizeof(info));
1813
1814         IFNET_RLOCK_NOSLEEP();
1815         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1816                 if (w->w_arg && w->w_arg != ifp->if_index)
1817                         continue;
1818                 ifa = ifp->if_addr;
1819                 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1820                 IF_ADDR_RLOCK(ifp);
1821                 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1822                         if (af && af != ifma->ifma_addr->sa_family)
1823                                 continue;
1824                         if (prison_if(w->w_req->td->td_ucred,
1825                             ifma->ifma_addr) != 0)
1826                                 continue;
1827                         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1828                         info.rti_info[RTAX_GATEWAY] =
1829                             (ifma->ifma_addr->sa_family != AF_LINK) ?
1830                             ifma->ifma_lladdr : NULL;
1831                         error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
1832                         if (error != 0)
1833                                 break;
1834                         if (w->w_req && w->w_tmem) {
1835                                 struct ifma_msghdr *ifmam;
1836
1837                                 ifmam = (struct ifma_msghdr *)w->w_tmem;
1838                                 ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1839                                 ifmam->ifmam_flags = 0;
1840                                 ifmam->ifmam_addrs = info.rti_addrs;
1841                                 ifmam->_ifmam_spare1 = 0;
1842                                 error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1843                                 if (error != 0)
1844                                         break;
1845                         }
1846                 }
1847                 IF_ADDR_RUNLOCK(ifp);
1848                 if (error != 0)
1849                         break;
1850         }
1851         IFNET_RUNLOCK_NOSLEEP();
1852         return (error);
1853 }
1854
1855 static int
1856 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1857 {
1858         int     *name = (int *)arg1;
1859         u_int   namelen = arg2;
1860         struct rib_head *rnh = NULL; /* silence compiler. */
1861         int     i, lim, error = EINVAL;
1862         int     fib = 0;
1863         u_char  af;
1864         struct  walkarg w;
1865
1866         name ++;
1867         namelen--;
1868         if (req->newptr)
1869                 return (EPERM);
1870         if (name[1] == NET_RT_DUMP) {
1871                 if (namelen == 3)
1872                         fib = req->td->td_proc->p_fibnum;
1873                 else if (namelen == 4)
1874                         fib = (name[3] == RT_ALL_FIBS) ?
1875                             req->td->td_proc->p_fibnum : name[3];
1876                 else
1877                         return ((namelen < 3) ? EISDIR : ENOTDIR);
1878                 if (fib < 0 || fib >= rt_numfibs)
1879                         return (EINVAL);
1880         } else if (namelen != 3)
1881                 return ((namelen < 3) ? EISDIR : ENOTDIR);
1882         af = name[0];
1883         if (af > AF_MAX)
1884                 return (EINVAL);
1885         bzero(&w, sizeof(w));
1886         w.w_op = name[1];
1887         w.w_arg = name[2];
1888         w.w_req = req;
1889
1890         error = sysctl_wire_old_buffer(req, 0);
1891         if (error)
1892                 return (error);
1893
1894         /*
1895          * Allocate reply buffer in advance.
1896          * All rtsock messages has maximum length of u_short.
1897          */
1898         w.w_tmemsize = 65536;
1899         w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
1900
1901         switch (w.w_op) {
1902
1903         case NET_RT_DUMP:
1904         case NET_RT_FLAGS:
1905                 if (af == 0) {                  /* dump all tables */
1906                         i = 1;
1907                         lim = AF_MAX;
1908                 } else                          /* dump only one table */
1909                         i = lim = af;
1910
1911                 /*
1912                  * take care of llinfo entries, the caller must
1913                  * specify an AF
1914                  */
1915                 if (w.w_op == NET_RT_FLAGS &&
1916                     (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1917                         if (af != 0)
1918                                 error = lltable_sysctl_dumparp(af, w.w_req);
1919                         else
1920                                 error = EINVAL;
1921                         break;
1922                 }
1923                 /*
1924                  * take care of routing entries
1925                  */
1926                 for (error = 0; error == 0 && i <= lim; i++) {
1927                         rnh = rt_tables_get_rnh(fib, i);
1928                         if (rnh != NULL) {
1929                                 RIB_RLOCK(rnh);
1930                                 IFNET_RLOCK_NOSLEEP();
1931                                 error = rnh->rnh_walktree(&rnh->head,
1932                                     sysctl_dumpentry, &w);
1933                                 IFNET_RUNLOCK_NOSLEEP();
1934                                 RIB_RUNLOCK(rnh);
1935                         } else if (af != 0)
1936                                 error = EAFNOSUPPORT;
1937                 }
1938                 break;
1939
1940         case NET_RT_IFLIST:
1941         case NET_RT_IFLISTL:
1942                 error = sysctl_iflist(af, &w);
1943                 break;
1944
1945         case NET_RT_IFMALIST:
1946                 error = sysctl_ifmalist(af, &w);
1947                 break;
1948         }
1949
1950         free(w.w_tmem, M_TEMP);
1951         return (error);
1952 }
1953
1954 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1955
1956 /*
1957  * Definitions of protocols supported in the ROUTE domain.
1958  */
1959
1960 static struct domain routedomain;               /* or at least forward */
1961
1962 static struct protosw routesw[] = {
1963 {
1964         .pr_type =              SOCK_RAW,
1965         .pr_domain =            &routedomain,
1966         .pr_flags =             PR_ATOMIC|PR_ADDR,
1967         .pr_output =            route_output,
1968         .pr_ctlinput =          raw_ctlinput,
1969         .pr_init =              raw_init,
1970         .pr_usrreqs =           &route_usrreqs
1971 }
1972 };
1973
1974 static struct domain routedomain = {
1975         .dom_family =           PF_ROUTE,
1976         .dom_name =              "route",
1977         .dom_protosw =          routesw,
1978         .dom_protoswNPROTOSW =  &routesw[nitems(routesw)]
1979 };
1980
1981 VNET_DOMAIN_SET(route);