sys/net/rtsock.c

   1 /*-
   2  * Copyright (c) 1988, 1991, 1993
   3  *      The Regents of the University of California.  All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  * 4. Neither the name of the University nor the names of its contributors
  14  *    may be used to endorse or promote products derived from this software
  15  *    without specific prior written permission.
  16  *
  17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  27  * SUCH DAMAGE.
  28  *
  29  *      @(#)rtsock.c    8.7 (Berkeley) 10/12/95
  30  * $FreeBSD$
  31  */
  32 #include "opt_compat.h"
  33 #include "opt_sctp.h"
  34 #include "opt_mpath.h"
  35 #include "opt_inet.h"
  36 #include "opt_inet6.h"
  37
  38 #include <sys/param.h>
  39 #include <sys/jail.h>
  40 #include <sys/kernel.h>
  41 #include <sys/domain.h>
  42 #include <sys/lock.h>
  43 #include <sys/malloc.h>
  44 #include <sys/mbuf.h>
  45 #include <sys/priv.h>
  46 #include <sys/proc.h>
  47 #include <sys/protosw.h>
  48 #include <sys/rwlock.h>
  49 #include <sys/signalvar.h>
  50 #include <sys/socket.h>
  51 #include <sys/socketvar.h>
  52 #include <sys/sysctl.h>
  53 #include <sys/systm.h>
  54
  55 #define _IN_NET_RTSOCK_C
  56 #include <net/if.h>
  57 #include <net/if_dl.h>
  58 #include <net/if_llatbl.h>
  59 #include <net/if_types.h>
  60 #include <net/netisr.h>
  61 #include <net/raw_cb.h>
  62 #include <net/route.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 #if defined(INET) || defined(INET6)
  74 #ifdef SCTP
  75 extern void sctp_addr_change(struct ifaddr *ifa, int cmd);
  76 #endif /* SCTP */
  77 #endif
  78
  79 #ifdef COMPAT_FREEBSD32
  80 #include <sys/mount.h>
  81 #include <compat/freebsd32/freebsd32.h>
  82
  83 struct if_data32 {
  84         uint8_t ifi_type;
  85         uint8_t ifi_physical;
  86         uint8_t ifi_addrlen;
  87         uint8_t ifi_hdrlen;
  88         uint8_t ifi_link_state;
  89         uint8_t ifi_vhid;
  90         uint8_t ifi_baudrate_pf;
  91         uint8_t ifi_datalen;
  92         uint32_t ifi_mtu;
  93         uint32_t ifi_metric;
  94         uint32_t ifi_baudrate;
  95         uint32_t ifi_ipackets;
  96         uint32_t ifi_ierrors;
  97         uint32_t ifi_opackets;
  98         uint32_t ifi_oerrors;
  99         uint32_t ifi_collisions;
 100         uint32_t ifi_ibytes;
 101         uint32_t ifi_obytes;
 102         uint32_t ifi_imcasts;
 103         uint32_t ifi_omcasts;
 104         uint32_t ifi_iqdrops;
 105         uint32_t ifi_noproto;
 106         uint32_t ifi_hwassist;
 107         int32_t ifi_epoch;
 108         struct  timeval32 ifi_lastchange;
 109         uint32_t ifi_oqdrops;
 110 };
 111
 112 struct if_msghdr32 {
 113         uint16_t ifm_msglen;
 114         uint8_t ifm_version;
 115         uint8_t ifm_type;
 116         int32_t ifm_addrs;
 117         int32_t ifm_flags;
 118         uint16_t ifm_index;
 119         struct  if_data32 ifm_data;
 120 };
 121
 122 struct if_msghdrl32 {
 123         uint16_t ifm_msglen;
 124         uint8_t ifm_version;
 125         uint8_t ifm_type;
 126         int32_t ifm_addrs;
 127         int32_t ifm_flags;
 128         uint16_t ifm_index;
 129         uint16_t _ifm_spare1;
 130         uint16_t ifm_len;
 131         uint16_t ifm_data_off;
 132         struct  if_data32 ifm_data;
 133 };
 134
 135 struct ifa_msghdrl32 {
 136         uint16_t ifam_msglen;
 137         uint8_t ifam_version;
 138         uint8_t ifam_type;
 139         int32_t ifam_addrs;
 140         int32_t ifam_flags;
 141         uint16_t ifam_index;
 142         uint16_t _ifam_spare1;
 143         uint16_t ifam_len;
 144         uint16_t ifam_data_off;
 145         int32_t ifam_metric;
 146         struct  if_data32 ifam_data;
 147 };
 148 #endif /* COMPAT_FREEBSD32 */
 149
 150 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
 151
 152 /* NB: these are not modified */
 153 static struct   sockaddr route_src = { 2, PF_ROUTE, };
 154 static struct   sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
 155
 156 /* These are external hooks for CARP. */
 157 int     (*carp_get_vhid_p)(struct ifaddr *);
 158
 159 /*
 160  * Used by rtsock/raw_input callback code to decide whether to filter the update
 161  * notification to a socket bound to a particular FIB.
 162  */
 163 #define RTS_FILTER_FIB  M_PROTO8
 164
 165 static struct {
 166         int     ip_count;       /* attached w/ AF_INET */
 167         int     ip6_count;      /* attached w/ AF_INET6 */
 168         int     ipx_count;      /* attached w/ AF_IPX */
 169         int     any_count;      /* total attached */
 170 } route_cb;
 171
 172 struct mtx rtsock_mtx;
 173 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
 174
 175 #define RTSOCK_LOCK()   mtx_lock(&rtsock_mtx)
 176 #define RTSOCK_UNLOCK() mtx_unlock(&rtsock_mtx)
 177 #define RTSOCK_LOCK_ASSERT()    mtx_assert(&rtsock_mtx, MA_OWNED)
 178
 179 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
 180
 181 struct walkarg {
 182         int     w_tmemsize;
 183         int     w_op, w_arg;
 184         caddr_t w_tmem;
 185         struct sysctl_req *w_req;
 186 };
 187
 188 static void     rts_input(struct mbuf *m);
 189 static struct mbuf *rt_msg1(int type, struct rt_addrinfo *rtinfo);
 190 static int      rt_msg2(int type, struct rt_addrinfo *rtinfo,
 191                         caddr_t cp, struct walkarg *w);
 192 static int      rt_xaddrs(caddr_t cp, caddr_t cplim,
 193                         struct rt_addrinfo *rtinfo);
 194 static int      sysctl_dumpentry(struct radix_node *rn, void *vw);
 195 static int      sysctl_iflist(int af, struct walkarg *w);
 196 static int      sysctl_ifmalist(int af, struct walkarg *w);
 197 static int      route_output(struct mbuf *m, struct socket *so);
 198 static void     rt_setmetrics(const struct rt_msghdr *rtm, struct rtentry *rt);
 199 static void     rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out);
 200 static void     rt_dispatch(struct mbuf *, sa_family_t);
 201
 202 static struct netisr_handler rtsock_nh = {
 203         .nh_name = "rtsock",
 204         .nh_handler = rts_input,
 205         .nh_proto = NETISR_ROUTE,
 206         .nh_policy = NETISR_POLICY_SOURCE,
 207 };
 208
 209 static int
 210 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
 211 {
 212         int error, qlimit;
 213
 214         netisr_getqlimit(&rtsock_nh, &qlimit);
 215         error = sysctl_handle_int(oidp, &qlimit, 0, req);
 216         if (error || !req->newptr)
 217                 return (error);
 218         if (qlimit < 1)
 219                 return (EINVAL);
 220         return (netisr_setqlimit(&rtsock_nh, qlimit));
 221 }
 222 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
 223     0, 0, sysctl_route_netisr_maxqlen, "I",
 224     "maximum routing socket dispatch queue length");
 225
 226 static void
 227 rts_init(void)
 228 {
 229         int tmp;
 230
 231         if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
 232                 rtsock_nh.nh_qlimit = tmp;
 233         netisr_register(&rtsock_nh);
 234 }
 235 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0);
 236
 237 static int
 238 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
 239     struct rawcb *rp)
 240 {
 241         int fibnum;
 242
 243         KASSERT(m != NULL, ("%s: m is NULL", __func__));
 244         KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
 245         KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
 246
 247         /* No filtering requested. */
 248         if ((m->m_flags & RTS_FILTER_FIB) == 0)
 249                 return (0);
 250
 251         /* Check if it is a rts and the fib matches the one of the socket. */
 252         fibnum = M_GETFIB(m);
 253         if (proto->sp_family != PF_ROUTE ||
 254             rp->rcb_socket == NULL ||
 255             rp->rcb_socket->so_fibnum == fibnum)
 256                 return (0);
 257
 258         /* Filtering requested and no match, the socket shall be skipped. */
 259         return (1);
 260 }
 261
 262 static void
 263 rts_input(struct mbuf *m)
 264 {
 265         struct sockproto route_proto;
 266         unsigned short *family;
 267         struct m_tag *tag;
 268
 269         route_proto.sp_family = PF_ROUTE;
 270         tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
 271         if (tag != NULL) {
 272                 family = (unsigned short *)(tag + 1);
 273                 route_proto.sp_protocol = *family;
 274                 m_tag_delete(m, tag);
 275         } else
 276                 route_proto.sp_protocol = 0;
 277
 278         raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
 279 }
 280
 281 /*
 282  * It really doesn't make any sense at all for this code to share much
 283  * with raw_usrreq.c, since its functionality is so restricted.  XXX
 284  */
 285 static void
 286 rts_abort(struct socket *so)
 287 {
 288
 289         raw_usrreqs.pru_abort(so);
 290 }
 291
 292 static void
 293 rts_close(struct socket *so)
 294 {
 295
 296         raw_usrreqs.pru_close(so);
 297 }
 298
 299 /* pru_accept is EOPNOTSUPP */
 300
 301 static int
 302 rts_attach(struct socket *so, int proto, struct thread *td)
 303 {
 304         struct rawcb *rp;
 305         int error;
 306
 307         KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
 308
 309         /* XXX */
 310         rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
 311         if (rp == NULL)
 312                 return ENOBUFS;
 313
 314         so->so_pcb = (caddr_t)rp;
 315         so->so_fibnum = td->td_proc->p_fibnum;
 316         error = raw_attach(so, proto);
 317         rp = sotorawcb(so);
 318         if (error) {
 319                 so->so_pcb = NULL;
 320                 free(rp, M_PCB);
 321                 return error;
 322         }
 323         RTSOCK_LOCK();
 324         switch(rp->rcb_proto.sp_protocol) {
 325         case AF_INET:
 326                 route_cb.ip_count++;
 327                 break;
 328         case AF_INET6:
 329                 route_cb.ip6_count++;
 330                 break;
 331         case AF_IPX:
 332                 route_cb.ipx_count++;
 333                 break;
 334         }
 335         route_cb.any_count++;
 336         RTSOCK_UNLOCK();
 337         soisconnected(so);
 338         so->so_options |= SO_USELOOPBACK;
 339         return 0;
 340 }
 341
 342 static int
 343 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
 344 {
 345
 346         return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
 347 }
 348
 349 static int
 350 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
 351 {
 352
 353         return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
 354 }
 355
 356 /* pru_connect2 is EOPNOTSUPP */
 357 /* pru_control is EOPNOTSUPP */
 358
 359 static void
 360 rts_detach(struct socket *so)
 361 {
 362         struct rawcb *rp = sotorawcb(so);
 363
 364         KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
 365
 366         RTSOCK_LOCK();
 367         switch(rp->rcb_proto.sp_protocol) {
 368         case AF_INET:
 369                 route_cb.ip_count--;
 370                 break;
 371         case AF_INET6:
 372                 route_cb.ip6_count--;
 373                 break;
 374         case AF_IPX:
 375                 route_cb.ipx_count--;
 376                 break;
 377         }
 378         route_cb.any_count--;
 379         RTSOCK_UNLOCK();
 380         raw_usrreqs.pru_detach(so);
 381 }
 382
 383 static int
 384 rts_disconnect(struct socket *so)
 385 {
 386
 387         return (raw_usrreqs.pru_disconnect(so));
 388 }
 389
 390 /* pru_listen is EOPNOTSUPP */
 391
 392 static int
 393 rts_peeraddr(struct socket *so, struct sockaddr **nam)
 394 {
 395
 396         return (raw_usrreqs.pru_peeraddr(so, nam));
 397 }
 398
 399 /* pru_rcvd is EOPNOTSUPP */
 400 /* pru_rcvoob is EOPNOTSUPP */
 401
 402 static int
 403 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
 404          struct mbuf *control, struct thread *td)
 405 {
 406
 407         return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
 408 }
 409
 410 /* pru_sense is null */
 411
 412 static int
 413 rts_shutdown(struct socket *so)
 414 {
 415
 416         return (raw_usrreqs.pru_shutdown(so));
 417 }
 418
 419 static int
 420 rts_sockaddr(struct socket *so, struct sockaddr **nam)
 421 {
 422
 423         return (raw_usrreqs.pru_sockaddr(so, nam));
 424 }
 425
 426 static struct pr_usrreqs route_usrreqs = {
 427         .pru_abort =            rts_abort,
 428         .pru_attach =           rts_attach,
 429         .pru_bind =             rts_bind,
 430         .pru_connect =          rts_connect,
 431         .pru_detach =           rts_detach,
 432         .pru_disconnect =       rts_disconnect,
 433         .pru_peeraddr =         rts_peeraddr,
 434         .pru_send =             rts_send,
 435         .pru_shutdown =         rts_shutdown,
 436         .pru_sockaddr =         rts_sockaddr,
 437         .pru_close =            rts_close,
 438 };
 439
 440 #ifndef _SOCKADDR_UNION_DEFINED
 441 #define _SOCKADDR_UNION_DEFINED
 442 /*
 443  * The union of all possible address formats we handle.
 444  */
 445 union sockaddr_union {
 446         struct sockaddr         sa;
 447         struct sockaddr_in      sin;
 448         struct sockaddr_in6     sin6;
 449 };
 450 #endif /* _SOCKADDR_UNION_DEFINED */
 451
 452 static int
 453 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
 454     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
 455 {
 456
 457         /* First, see if the returned address is part of the jail. */
 458         if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
 459                 info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
 460                 return (0);
 461         }
 462
 463         switch (info->rti_info[RTAX_DST]->sa_family) {
 464 #ifdef INET
 465         case AF_INET:
 466         {
 467                 struct in_addr ia;
 468                 struct ifaddr *ifa;
 469                 int found;
 470
 471                 found = 0;
 472                 /*
 473                  * Try to find an address on the given outgoing interface
 474                  * that belongs to the jail.
 475                  */
 476                 IF_ADDR_RLOCK(ifp);
 477                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 478                         struct sockaddr *sa;
 479                         sa = ifa->ifa_addr;
 480                         if (sa->sa_family != AF_INET)
 481                                 continue;
 482                         ia = ((struct sockaddr_in *)sa)->sin_addr;
 483                         if (prison_check_ip4(cred, &ia) == 0) {
 484                                 found = 1;
 485                                 break;
 486                         }
 487                 }
 488                 IF_ADDR_RUNLOCK(ifp);
 489                 if (!found) {
 490                         /*
 491                          * As a last resort return the 'default' jail address.
 492                          */
 493                         ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
 494                             sin_addr;
 495                         if (prison_get_ip4(cred, &ia) != 0)
 496                                 return (ESRCH);
 497                 }
 498                 bzero(&saun->sin, sizeof(struct sockaddr_in));
 499                 saun->sin.sin_len = sizeof(struct sockaddr_in);
 500                 saun->sin.sin_family = AF_INET;
 501                 saun->sin.sin_addr.s_addr = ia.s_addr;
 502                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
 503                 break;
 504         }
 505 #endif
 506 #ifdef INET6
 507         case AF_INET6:
 508         {
 509                 struct in6_addr ia6;
 510                 struct ifaddr *ifa;
 511                 int found;
 512
 513                 found = 0;
 514                 /*
 515                  * Try to find an address on the given outgoing interface
 516                  * that belongs to the jail.
 517                  */
 518                 IF_ADDR_RLOCK(ifp);
 519                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 520                         struct sockaddr *sa;
 521                         sa = ifa->ifa_addr;
 522                         if (sa->sa_family != AF_INET6)
 523                                 continue;
 524                         bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
 525                             &ia6, sizeof(struct in6_addr));
 526                         if (prison_check_ip6(cred, &ia6) == 0) {
 527                                 found = 1;
 528                                 break;
 529                         }
 530                 }
 531                 IF_ADDR_RUNLOCK(ifp);
 532                 if (!found) {
 533                         /*
 534                          * As a last resort return the 'default' jail address.
 535                          */
 536                         ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
 537                             sin6_addr;
 538                         if (prison_get_ip6(cred, &ia6) != 0)
 539                                 return (ESRCH);
 540                 }
 541                 bzero(&saun->sin6, sizeof(struct sockaddr_in6));
 542                 saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
 543                 saun->sin6.sin6_family = AF_INET6;
 544                 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
 545                 if (sa6_recoverscope(&saun->sin6) != 0)
 546                         return (ESRCH);
 547                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
 548                 break;
 549         }
 550 #endif
 551         default:
 552                 return (ESRCH);
 553         }
 554         return (0);
 555 }
 556
 557 /*ARGSUSED*/
 558 static int
 559 route_output(struct mbuf *m, struct socket *so)
 560 {
 561 #define sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
 562         struct rt_msghdr *rtm = NULL;
 563         struct rtentry *rt = NULL;
 564         struct radix_node_head *rnh;
 565         struct rt_addrinfo info;
 566 #ifdef INET6
 567         struct sockaddr_storage ss;
 568         struct sockaddr_in6 *sin6;
 569         int i, rti_need_deembed = 0;
 570 #endif
 571         int len, error = 0;
 572         struct ifnet *ifp = NULL;
 573         union sockaddr_union saun;
 574         sa_family_t saf = AF_UNSPEC;
 575
 576 #define senderr(e) { error = e; goto flush;}
 577         if (m == NULL || ((m->m_len < sizeof(long)) &&
 578                        (m = m_pullup(m, sizeof(long))) == NULL))
 579                 return (ENOBUFS);
 580         if ((m->m_flags & M_PKTHDR) == 0)
 581                 panic("route_output");
 582         len = m->m_pkthdr.len;
 583         if (len < sizeof(*rtm) ||
 584             len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
 585                 info.rti_info[RTAX_DST] = NULL;
 586                 senderr(EINVAL);
 587         }
 588         R_Malloc(rtm, struct rt_msghdr *, len);
 589         if (rtm == NULL) {
 590                 info.rti_info[RTAX_DST] = NULL;
 591                 senderr(ENOBUFS);
 592         }
 593         m_copydata(m, 0, len, (caddr_t)rtm);
 594         if (rtm->rtm_version != RTM_VERSION) {
 595                 info.rti_info[RTAX_DST] = NULL;
 596                 senderr(EPROTONOSUPPORT);
 597         }
 598         rtm->rtm_pid = curproc->p_pid;
 599         bzero(&info, sizeof(info));
 600         info.rti_addrs = rtm->rtm_addrs;
 601         /*
 602          * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
 603          * link-local address because rtrequest requires addresses with
 604          * embedded scope id.
 605          */
 606         if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
 607                 info.rti_info[RTAX_DST] = NULL;
 608                 senderr(EINVAL);
 609         }
 610         info.rti_flags = rtm->rtm_flags;
 611         if (info.rti_info[RTAX_DST] == NULL ||
 612             info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
 613             (info.rti_info[RTAX_GATEWAY] != NULL &&
 614              info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
 615                 senderr(EINVAL);
 616         saf = info.rti_info[RTAX_DST]->sa_family;
 617         /*
 618          * Verify that the caller has the appropriate privilege; RTM_GET
 619          * is the only operation the non-superuser is allowed.
 620          */
 621         if (rtm->rtm_type != RTM_GET) {
 622                 error = priv_check(curthread, PRIV_NET_ROUTE);
 623                 if (error)
 624                         senderr(error);
 625         }
 626
 627         /*
 628          * The given gateway address may be an interface address.
 629          * For example, issuing a "route change" command on a route
 630          * entry that was created from a tunnel, and the gateway
 631          * address given is the local end point. In this case the
 632          * RTF_GATEWAY flag must be cleared or the destination will
 633          * not be reachable even though there is no error message.
 634          */
 635         if (info.rti_info[RTAX_GATEWAY] != NULL &&
 636             info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
 637                 struct route gw_ro;
 638
 639                 bzero(&gw_ro, sizeof(gw_ro));
 640                 gw_ro.ro_dst = *info.rti_info[RTAX_GATEWAY];
 641                 rtalloc_ign_fib(&gw_ro, 0, so->so_fibnum);
 642                 /*
 643                  * A host route through the loopback interface is
 644                  * installed for each interface adddress. In pre 8.0
 645                  * releases the interface address of a PPP link type
 646                  * is not reachable locally. This behavior is fixed as
 647                  * part of the new L2/L3 redesign and rewrite work. The
 648                  * signature of this interface address route is the
 649                  * AF_LINK sa_family type of the rt_gateway, and the
 650                  * rt_ifp has the IFF_LOOPBACK flag set.
 651                  */
 652                 if (gw_ro.ro_rt != NULL &&
 653                     gw_ro.ro_rt->rt_gateway->sa_family == AF_LINK &&
 654                     gw_ro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) {
 655                         info.rti_flags &= ~RTF_GATEWAY;
 656                         info.rti_flags |= RTF_GWFLAG_COMPAT;
 657                 }
 658                 if (gw_ro.ro_rt != NULL)
 659                         RTFREE(gw_ro.ro_rt);
 660         }
 661
 662         switch (rtm->rtm_type) {
 663                 struct rtentry *saved_nrt;
 664
 665         case RTM_ADD:
 666                 if (info.rti_info[RTAX_GATEWAY] == NULL)
 667                         senderr(EINVAL);
 668                 saved_nrt = NULL;
 669
 670                 /* support for new ARP code */
 671                 if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
 672                     (rtm->rtm_flags & RTF_LLDATA) != 0) {
 673                         error = lla_rt_output(rtm, &info);
 674 #ifdef INET6
 675                         if (error == 0)
 676                                 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 677 #endif
 678                         break;
 679                 }
 680                 error = rtrequest1_fib(RTM_ADD, &info, &saved_nrt,
 681                     so->so_fibnum);
 682                 if (error == 0 && saved_nrt) {
 683 #ifdef INET6
 684                         rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 685 #endif
 686                         RT_LOCK(saved_nrt);
 687                         rt_setmetrics(rtm, saved_nrt);
 688                         rtm->rtm_index = saved_nrt->rt_ifp->if_index;
 689                         RT_REMREF(saved_nrt);
 690                         RT_UNLOCK(saved_nrt);
 691                 }
 692                 break;
 693
 694         case RTM_DELETE:
 695                 saved_nrt = NULL;
 696                 /* support for new ARP code */
 697                 if (info.rti_info[RTAX_GATEWAY] &&
 698                     (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
 699                     (rtm->rtm_flags & RTF_LLDATA) != 0) {
 700                         error = lla_rt_output(rtm, &info);
 701 #ifdef INET6
 702                         if (error == 0)
 703                                 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 704 #endif
 705                         break;
 706                 }
 707                 error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt,
 708                     so->so_fibnum);
 709                 if (error == 0) {
 710                         RT_LOCK(saved_nrt);
 711                         rt = saved_nrt;
 712                         goto report;
 713                 }
 714 #ifdef INET6
 715                 /* rt_msg2() will not be used when RTM_DELETE fails. */
 716                 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 717 #endif
 718                 break;
 719
 720         case RTM_GET:
 721         case RTM_CHANGE:
 722         case RTM_LOCK:
 723                 rnh = rt_tables_get_rnh(so->so_fibnum,
 724                     info.rti_info[RTAX_DST]->sa_family);
 725                 if (rnh == NULL)
 726                         senderr(EAFNOSUPPORT);
 727
 728                 RADIX_NODE_HEAD_RLOCK(rnh);
 729
 730                 if (info.rti_info[RTAX_NETMASK] == NULL &&
 731                     rtm->rtm_type == RTM_GET) {
 732                         /*
 733                          * Provide logest prefix match for
 734                          * address lookup (no mask).
 735                          * 'route -n get addr'
 736                          */
 737                         rt = (struct rtentry *) rnh->rnh_matchaddr(
 738                             info.rti_info[RTAX_DST], rnh);
 739                 } else
 740                         rt = (struct rtentry *) rnh->rnh_lookup(
 741                             info.rti_info[RTAX_DST],
 742                             info.rti_info[RTAX_NETMASK], rnh);
 743
 744                 if (rt == NULL) {
 745                         RADIX_NODE_HEAD_RUNLOCK(rnh);
 746                         senderr(ESRCH);
 747                 }
 748 #ifdef RADIX_MPATH
 749                 /*
 750                  * for RTM_CHANGE/LOCK, if we got multipath routes,
 751                  * we require users to specify a matching RTAX_GATEWAY.
 752                  *
 753                  * for RTM_GET, gate is optional even with multipath.
 754                  * if gate == NULL the first match is returned.
 755                  * (no need to call rt_mpath_matchgate if gate == NULL)
 756                  */
 757                 if (rn_mpath_capable(rnh) &&
 758                     (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
 759                         rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
 760                         if (!rt) {
 761                                 RADIX_NODE_HEAD_RUNLOCK(rnh);
 762                                 senderr(ESRCH);
 763                         }
 764                 }
 765 #endif
 766                 /*
 767                  * If performing proxied L2 entry insertion, and
 768                  * the actual PPP host entry is found, perform
 769                  * another search to retrieve the prefix route of
 770                  * the local end point of the PPP link.
 771                  */
 772                 if (rtm->rtm_flags & RTF_ANNOUNCE) {
 773                         struct sockaddr laddr;
 774
 775                         if (rt->rt_ifp != NULL &&
 776                             rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
 777                                 struct ifaddr *ifa;
 778
 779                                 ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1);
 780                                 if (ifa != NULL)
 781                                         rt_maskedcopy(ifa->ifa_addr,
 782                                                       &laddr,
 783                                                       ifa->ifa_netmask);
 784                         } else
 785                                 rt_maskedcopy(rt->rt_ifa->ifa_addr,
 786                                               &laddr,
 787                                               rt->rt_ifa->ifa_netmask);
 788                         /*
 789                          * refactor rt and no lock operation necessary
 790                          */
 791                         rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr, rnh);
 792                         if (rt == NULL) {
 793                                 RADIX_NODE_HEAD_RUNLOCK(rnh);
 794                                 senderr(ESRCH);
 795                         }
 796                 }
 797                 RT_LOCK(rt);
 798                 RT_ADDREF(rt);
 799                 RADIX_NODE_HEAD_RUNLOCK(rnh);
 800
 801                 switch(rtm->rtm_type) {
 802
 803                 case RTM_GET:
 804                 report:
 805                         RT_LOCK_ASSERT(rt);
 806                         if ((rt->rt_flags & RTF_HOST) == 0
 807                             ? jailed_without_vnet(curthread->td_ucred)
 808                             : prison_if(curthread->td_ucred,
 809                             rt_key(rt)) != 0) {
 810                                 RT_UNLOCK(rt);
 811                                 senderr(ESRCH);
 812                         }
 813                         info.rti_info[RTAX_DST] = rt_key(rt);
 814                         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
 815                         info.rti_info[RTAX_NETMASK] = rt_mask(rt);
 816                         info.rti_info[RTAX_GENMASK] = 0;
 817                         if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
 818                                 ifp = rt->rt_ifp;
 819                                 if (ifp) {
 820                                         info.rti_info[RTAX_IFP] =
 821                                             ifp->if_addr->ifa_addr;
 822                                         error = rtm_get_jailed(&info, ifp, rt,
 823                                             &saun, curthread->td_ucred);
 824                                         if (error != 0) {
 825                                                 RT_UNLOCK(rt);
 826                                                 senderr(error);
 827                                         }
 828                                         if (ifp->if_flags & IFF_POINTOPOINT)
 829                                                 info.rti_info[RTAX_BRD] =
 830                                                     rt->rt_ifa->ifa_dstaddr;
 831                                         rtm->rtm_index = ifp->if_index;
 832                                 } else {
 833                                         info.rti_info[RTAX_IFP] = NULL;
 834                                         info.rti_info[RTAX_IFA] = NULL;
 835                                 }
 836                         } else if ((ifp = rt->rt_ifp) != NULL) {
 837                                 rtm->rtm_index = ifp->if_index;
 838                         }
 839                         len = rt_msg2(rtm->rtm_type, &info, NULL, NULL);
 840                         if (len > rtm->rtm_msglen) {
 841                                 struct rt_msghdr *new_rtm;
 842                                 R_Malloc(new_rtm, struct rt_msghdr *, len);
 843                                 if (new_rtm == NULL) {
 844                                         RT_UNLOCK(rt);
 845                                         senderr(ENOBUFS);
 846                                 }
 847                                 bcopy(rtm, new_rtm, rtm->rtm_msglen);
 848                                 Free(rtm); rtm = new_rtm;
 849                         }
 850                         (void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm, NULL);
 851                         if (rt->rt_flags & RTF_GWFLAG_COMPAT)
 852                                 rtm->rtm_flags = RTF_GATEWAY |
 853                                         (rt->rt_flags & ~RTF_GWFLAG_COMPAT);
 854                         else
 855                                 rtm->rtm_flags = rt->rt_flags;
 856                         rt_getmetrics(rt, &rtm->rtm_rmx);
 857                         rtm->rtm_addrs = info.rti_addrs;
 858                         break;
 859
 860                 case RTM_CHANGE:
 861                         /*
 862                          * New gateway could require new ifaddr, ifp;
 863                          * flags may also be different; ifp may be specified
 864                          * by ll sockaddr when protocol address is ambiguous
 865                          */
 866                         if (((rt->rt_flags & RTF_GATEWAY) &&
 867                              info.rti_info[RTAX_GATEWAY] != NULL) ||
 868                             info.rti_info[RTAX_IFP] != NULL ||
 869                             (info.rti_info[RTAX_IFA] != NULL &&
 870                              !sa_equal(info.rti_info[RTAX_IFA],
 871                                        rt->rt_ifa->ifa_addr))) {
 872                                 RT_UNLOCK(rt);
 873                                 RADIX_NODE_HEAD_LOCK(rnh);
 874                                 error = rt_getifa_fib(&info, rt->rt_fibnum);
 875                                 /*
 876                                  * XXXRW: Really we should release this
 877                                  * reference later, but this maintains
 878                                  * historical behavior.
 879                                  */
 880                                 if (info.rti_ifa != NULL)
 881                                         ifa_free(info.rti_ifa);
 882                                 RADIX_NODE_HEAD_UNLOCK(rnh);
 883                                 if (error != 0)
 884                                         senderr(error);
 885                                 RT_LOCK(rt);
 886                         }
 887                         if (info.rti_ifa != NULL &&
 888                             info.rti_ifa != rt->rt_ifa &&
 889                             rt->rt_ifa != NULL &&
 890                             rt->rt_ifa->ifa_rtrequest != NULL) {
 891                                 rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
 892                                     &info);
 893                                 ifa_free(rt->rt_ifa);
 894                         }
 895                         if (info.rti_info[RTAX_GATEWAY] != NULL) {
 896                                 RT_UNLOCK(rt);
 897                                 RADIX_NODE_HEAD_LOCK(rnh);
 898                                 RT_LOCK(rt);
 899
 900                                 error = rt_setgate(rt, rt_key(rt),
 901                                     info.rti_info[RTAX_GATEWAY]);
 902                                 RADIX_NODE_HEAD_UNLOCK(rnh);
 903                                 if (error != 0) {
 904                                         RT_UNLOCK(rt);
 905                                         senderr(error);
 906                                 }
 907                                 rt->rt_flags &= ~RTF_GATEWAY;
 908                                 rt->rt_flags |= (RTF_GATEWAY & info.rti_flags);
 909                         }
 910                         if (info.rti_ifa != NULL &&
 911                             info.rti_ifa != rt->rt_ifa) {
 912                                 ifa_ref(info.rti_ifa);
 913                                 rt->rt_ifa = info.rti_ifa;
 914                                 rt->rt_ifp = info.rti_ifp;
 915                         }
 916                         /* Allow some flags to be toggled on change. */
 917                         rt->rt_flags = (rt->rt_flags & ~RTF_FMASK) |
 918                                     (rtm->rtm_flags & RTF_FMASK);
 919                         rt_setmetrics(rtm, rt);
 920                         rtm->rtm_index = rt->rt_ifp->if_index;
 921                         if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
 922                                rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
 923                         /* FALLTHROUGH */
 924                 case RTM_LOCK:
 925                         /* We don't support locks anymore */
 926                         break;
 927                 }
 928                 RT_UNLOCK(rt);
 929                 break;
 930
 931         default:
 932                 senderr(EOPNOTSUPP);
 933         }
 934
 935 flush:
 936         if (rtm) {
 937                 if (error)
 938                         rtm->rtm_errno = error;
 939                 else
 940                         rtm->rtm_flags |= RTF_DONE;
 941         }
 942         if (rt)         /* XXX can this be true? */
 943                 RTFREE(rt);
 944     {
 945         struct rawcb *rp = NULL;
 946         /*
 947          * Check to see if we don't want our own messages.
 948          */
 949         if ((so->so_options & SO_USELOOPBACK) == 0) {
 950                 if (route_cb.any_count <= 1) {
 951                         if (rtm)
 952                                 Free(rtm);
 953                         m_freem(m);
 954                         return (error);
 955                 }
 956                 /* There is another listener, so construct message */
 957                 rp = sotorawcb(so);
 958         }
 959         if (rtm) {
 960 #ifdef INET6
 961                 if (rti_need_deembed) {
 962                         /* sin6_scope_id is recovered before sending rtm. */
 963                         sin6 = (struct sockaddr_in6 *)&ss;
 964                         for (i = 0; i < RTAX_MAX; i++) {
 965                                 if (info.rti_info[i] == NULL)
 966                                         continue;
 967                                 if (info.rti_info[i]->sa_family != AF_INET6)
 968                                         continue;
 969                                 bcopy(info.rti_info[i], sin6, sizeof(*sin6));
 970                                 if (sa6_recoverscope(sin6) == 0)
 971                                         bcopy(sin6, info.rti_info[i],
 972                                                     sizeof(*sin6));
 973                         }
 974                 }
 975 #endif
 976                 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
 977                 if (m->m_pkthdr.len < rtm->rtm_msglen) {
 978                         m_freem(m);
 979                         m = NULL;
 980                 } else if (m->m_pkthdr.len > rtm->rtm_msglen)
 981                         m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
 982         }
 983         if (m) {
 984                 M_SETFIB(m, so->so_fibnum);
 985                 m->m_flags |= RTS_FILTER_FIB;
 986                 if (rp) {
 987                         /*
 988                          * XXX insure we don't get a copy by
 989                          * invalidating our protocol
 990                          */
 991                         unsigned short family = rp->rcb_proto.sp_family;
 992                         rp->rcb_proto.sp_family = 0;
 993                         rt_dispatch(m, saf);
 994                         rp->rcb_proto.sp_family = family;
 995                 } else
 996                         rt_dispatch(m, saf);
 997         }
 998         /* info.rti_info[RTAX_DST] (used above) can point inside of rtm */
 999         if (rtm)
1000                 Free(rtm);
1001     }
1002         return (error);
1003 #undef  sa_equal
1004 }
1005
1006 static void
1007 rt_setmetrics(const struct rt_msghdr *rtm, struct rtentry *rt)
1008 {
1009
1010         if (rtm->rtm_inits & RTV_MTU)
1011                 rt->rt_mtu = rtm->rtm_rmx.rmx_mtu;
1012         if (rtm->rtm_inits & RTV_WEIGHT)
1013                 rt->rt_weight = rtm->rtm_rmx.rmx_weight;
1014         /* Kernel -> userland timebase conversion. */
1015         if (rtm->rtm_inits & RTV_EXPIRE)
1016                 rt->rt_expire = rtm->rtm_rmx.rmx_expire ?
1017                     rtm->rtm_rmx.rmx_expire - time_second + time_uptime : 0;
1018 }
1019
1020 static void
1021 rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out)
1022 {
1023
1024         bzero(out, sizeof(*out));
1025         out->rmx_mtu = rt->rt_mtu;
1026         out->rmx_weight = rt->rt_weight;
1027         out->rmx_pksent = counter_u64_fetch(rt->rt_pksent);
1028         /* Kernel -> userland timebase conversion. */
1029         out->rmx_expire = rt->rt_expire ?
1030             rt->rt_expire - time_uptime + time_second : 0;
1031 }
1032
1033 /*
1034  * Extract the addresses of the passed sockaddrs.
1035  * Do a little sanity checking so as to avoid bad memory references.
1036  * This data is derived straight from userland.
1037  */
1038 static int
1039 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1040 {
1041         struct sockaddr *sa;
1042         int i;
1043
1044         for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1045                 if ((rtinfo->rti_addrs & (1 << i)) == 0)
1046                         continue;
1047                 sa = (struct sockaddr *)cp;
1048                 /*
1049                  * It won't fit.
1050                  */
1051                 if (cp + sa->sa_len > cplim)
1052                         return (EINVAL);
1053                 /*
1054                  * there are no more.. quit now
1055                  * If there are more bits, they are in error.
1056                  * I've seen this. route(1) can evidently generate these.
1057                  * This causes kernel to core dump.
1058                  * for compatibility, If we see this, point to a safe address.
1059                  */
1060                 if (sa->sa_len == 0) {
1061                         rtinfo->rti_info[i] = &sa_zero;
1062                         return (0); /* should be EINVAL but for compat */
1063                 }
1064                 /* accept it */
1065 #ifdef INET6
1066                 if (sa->sa_family == AF_INET6)
1067                         sa6_embedscope((struct sockaddr_in6 *)sa,
1068                             V_ip6_use_defzone);
1069 #endif
1070                 rtinfo->rti_info[i] = sa;
1071                 cp += SA_SIZE(sa);
1072         }
1073         return (0);
1074 }
1075
1076 /*
1077  * Used by the routing socket.
1078  */
1079 static struct mbuf *
1080 rt_msg1(int type, struct rt_addrinfo *rtinfo)
1081 {
1082         struct rt_msghdr *rtm;
1083         struct mbuf *m;
1084         int i;
1085         struct sockaddr *sa;
1086 #ifdef INET6
1087         struct sockaddr_storage ss;
1088         struct sockaddr_in6 *sin6;
1089 #endif
1090         int len, dlen;
1091
1092         switch (type) {
1093
1094         case RTM_DELADDR:
1095         case RTM_NEWADDR:
1096                 len = sizeof(struct ifa_msghdr);
1097                 break;
1098
1099         case RTM_DELMADDR:
1100         case RTM_NEWMADDR:
1101                 len = sizeof(struct ifma_msghdr);
1102                 break;
1103
1104         case RTM_IFINFO:
1105                 len = sizeof(struct if_msghdr);
1106                 break;
1107
1108         case RTM_IFANNOUNCE:
1109         case RTM_IEEE80211:
1110                 len = sizeof(struct if_announcemsghdr);
1111                 break;
1112
1113         default:
1114                 len = sizeof(struct rt_msghdr);
1115         }
1116
1117         /* XXXGL: can we use MJUMPAGESIZE cluster here? */
1118         KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1119         if (len > MHLEN)
1120                 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1121         else
1122                 m = m_gethdr(M_NOWAIT, MT_DATA);
1123         if (m == NULL)
1124                 return (m);
1125
1126         m->m_pkthdr.len = m->m_len = len;
1127         rtm = mtod(m, struct rt_msghdr *);
1128         bzero((caddr_t)rtm, len);
1129         for (i = 0; i < RTAX_MAX; i++) {
1130                 if ((sa = rtinfo->rti_info[i]) == NULL)
1131                         continue;
1132                 rtinfo->rti_addrs |= (1 << i);
1133                 dlen = SA_SIZE(sa);
1134 #ifdef INET6
1135                 if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1136                         sin6 = (struct sockaddr_in6 *)&ss;
1137                         bcopy(sa, sin6, sizeof(*sin6));
1138                         if (sa6_recoverscope(sin6) == 0)
1139                                 sa = (struct sockaddr *)sin6;
1140                 }
1141 #endif
1142                 m_copyback(m, len, dlen, (caddr_t)sa);
1143                 len += dlen;
1144         }
1145         if (m->m_pkthdr.len != len) {
1146                 m_freem(m);
1147                 return (NULL);
1148         }
1149         rtm->rtm_msglen = len;
1150         rtm->rtm_version = RTM_VERSION;
1151         rtm->rtm_type = type;
1152         return (m);
1153 }
1154
1155 /*
1156  * Used by the sysctl code and routing socket.
1157  */
1158 static int
1159 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
1160 {
1161         int i;
1162         int len, dlen, second_time = 0;
1163         caddr_t cp0;
1164 #ifdef INET6
1165         struct sockaddr_storage ss;
1166         struct sockaddr_in6 *sin6;
1167 #endif
1168
1169         rtinfo->rti_addrs = 0;
1170 again:
1171         switch (type) {
1172
1173         case RTM_DELADDR:
1174         case RTM_NEWADDR:
1175                 if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1176 #ifdef COMPAT_FREEBSD32
1177                         if (w->w_req->flags & SCTL_MASK32)
1178                                 len = sizeof(struct ifa_msghdrl32);
1179                         else
1180 #endif
1181                                 len = sizeof(struct ifa_msghdrl);
1182                 } else
1183                         len = sizeof(struct ifa_msghdr);
1184                 break;
1185
1186         case RTM_IFINFO:
1187 #ifdef COMPAT_FREEBSD32
1188                 if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1189                         if (w->w_op == NET_RT_IFLISTL)
1190                                 len = sizeof(struct if_msghdrl32);
1191                         else
1192                                 len = sizeof(struct if_msghdr32);
1193                         break;
1194                 }
1195 #endif
1196                 if (w != NULL && w->w_op == NET_RT_IFLISTL)
1197                         len = sizeof(struct if_msghdrl);
1198                 else
1199                         len = sizeof(struct if_msghdr);
1200                 break;
1201
1202         case RTM_NEWMADDR:
1203                 len = sizeof(struct ifma_msghdr);
1204                 break;
1205
1206         default:
1207                 len = sizeof(struct rt_msghdr);
1208         }
1209         cp0 = cp;
1210         if (cp0)
1211                 cp += len;
1212         for (i = 0; i < RTAX_MAX; i++) {
1213                 struct sockaddr *sa;
1214
1215                 if ((sa = rtinfo->rti_info[i]) == NULL)
1216                         continue;
1217                 rtinfo->rti_addrs |= (1 << i);
1218                 dlen = SA_SIZE(sa);
1219                 if (cp) {
1220 #ifdef INET6
1221                         if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1222                                 sin6 = (struct sockaddr_in6 *)&ss;
1223                                 bcopy(sa, sin6, sizeof(*sin6));
1224                                 if (sa6_recoverscope(sin6) == 0)
1225                                         sa = (struct sockaddr *)sin6;
1226                         }
1227 #endif
1228                         bcopy((caddr_t)sa, cp, (unsigned)dlen);
1229                         cp += dlen;
1230                 }
1231                 len += dlen;
1232         }
1233         len = ALIGN(len);
1234         if (cp == NULL && w != NULL && !second_time) {
1235                 struct walkarg *rw = w;
1236
1237                 if (rw->w_req) {
1238                         if (rw->w_tmemsize < len) {
1239                                 if (rw->w_tmem)
1240                                         free(rw->w_tmem, M_RTABLE);
1241                                 rw->w_tmem = (caddr_t)
1242                                         malloc(len, M_RTABLE, M_NOWAIT);
1243                                 if (rw->w_tmem)
1244                                         rw->w_tmemsize = len;
1245                         }
1246                         if (rw->w_tmem) {
1247                                 cp = rw->w_tmem;
1248                                 second_time = 1;
1249                                 goto again;
1250                         }
1251                 }
1252         }
1253         if (cp) {
1254                 struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
1255
1256                 rtm->rtm_version = RTM_VERSION;
1257                 rtm->rtm_type = type;
1258                 rtm->rtm_msglen = len;
1259         }
1260         return (len);
1261 }
1262
1263 /*
1264  * This routine is called to generate a message from the routing
1265  * socket indicating that a redirect has occured, a routing lookup
1266  * has failed, or that a protocol has detected timeouts to a particular
1267  * destination.
1268  */
1269 void
1270 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1271     int fibnum)
1272 {
1273         struct rt_msghdr *rtm;
1274         struct mbuf *m;
1275         struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1276
1277         if (route_cb.any_count == 0)
1278                 return;
1279         m = rt_msg1(type, rtinfo);
1280         if (m == NULL)
1281                 return;
1282
1283         if (fibnum != RT_ALL_FIBS) {
1284                 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1285                     "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1286                 M_SETFIB(m, fibnum);
1287                 m->m_flags |= RTS_FILTER_FIB;
1288         }
1289
1290         rtm = mtod(m, struct rt_msghdr *);
1291         rtm->rtm_flags = RTF_DONE | flags;
1292         rtm->rtm_errno = error;
1293         rtm->rtm_addrs = rtinfo->rti_addrs;
1294         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1295 }
1296
1297 void
1298 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1299 {
1300
1301         rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1302 }
1303
1304 /*
1305  * This routine is called to generate a message from the routing
1306  * socket indicating that the status of a network interface has changed.
1307  */
1308 void
1309 rt_ifmsg(struct ifnet *ifp)
1310 {
1311         struct if_msghdr *ifm;
1312         struct mbuf *m;
1313         struct rt_addrinfo info;
1314
1315         if (route_cb.any_count == 0)
1316                 return;
1317         bzero((caddr_t)&info, sizeof(info));
1318         m = rt_msg1(RTM_IFINFO, &info);
1319         if (m == NULL)
1320                 return;
1321         ifm = mtod(m, struct if_msghdr *);
1322         ifm->ifm_index = ifp->if_index;
1323         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1324         ifm->ifm_data = ifp->if_data;
1325         ifm->ifm_addrs = 0;
1326         rt_dispatch(m, AF_UNSPEC);
1327 }
1328
1329 /*
1330  * This is called to generate messages from the routing socket
1331  * indicating a network interface has had addresses associated with it.
1332  * if we ever reverse the logic and replace messages TO the routing
1333  * socket indicate a request to configure interfaces, then it will
1334  * be unnecessary as the routing socket will automatically generate
1335  * copies of it.
1336  */
1337 void
1338 rt_newaddrmsg_fib(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt,
1339     int fibnum)
1340 {
1341         struct rt_addrinfo info;
1342         struct sockaddr *sa = NULL;
1343         int pass;
1344         struct mbuf *m = NULL;
1345         struct ifnet *ifp = ifa->ifa_ifp;
1346
1347         KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
1348                 ("unexpected cmd %u", cmd));
1349 #if defined(INET) || defined(INET6)
1350 #ifdef SCTP
1351         /*
1352          * notify the SCTP stack
1353          * this will only get called when an address is added/deleted
1354          * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1355          */
1356         sctp_addr_change(ifa, cmd);
1357 #endif /* SCTP */
1358 #endif
1359         if (route_cb.any_count == 0)
1360                 return;
1361         for (pass = 1; pass < 3; pass++) {
1362                 bzero((caddr_t)&info, sizeof(info));
1363                 if ((cmd == RTM_ADD && pass == 1) ||
1364                     (cmd == RTM_DELETE && pass == 2)) {
1365                         struct ifa_msghdr *ifam;
1366                         int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1367
1368                         info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1369                         info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1370                         info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1371                         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1372                         if ((m = rt_msg1(ncmd, &info)) == NULL)
1373                                 continue;
1374                         ifam = mtod(m, struct ifa_msghdr *);
1375                         ifam->ifam_index = ifp->if_index;
1376                         ifam->ifam_metric = ifa->ifa_metric;
1377                         ifam->ifam_flags = ifa->ifa_flags;
1378                         ifam->ifam_addrs = info.rti_addrs;
1379                 }
1380                 if ((cmd == RTM_ADD && pass == 2) ||
1381                     (cmd == RTM_DELETE && pass == 1)) {
1382                         struct rt_msghdr *rtm;
1383
1384                         if (rt == NULL)
1385                                 continue;
1386                         info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1387                         info.rti_info[RTAX_DST] = sa = rt_key(rt);
1388                         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1389                         if ((m = rt_msg1(cmd, &info)) == NULL)
1390                                 continue;
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                         KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: "
1399                             "fibnum out of range 0 <= %d < %d", __func__,
1400                              fibnum, rt_numfibs));
1401                         M_SETFIB(m, fibnum);
1402                         m->m_flags |= RTS_FILTER_FIB;
1403                 }
1404                 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1405         }
1406 }
1407
1408 void
1409 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1410 {
1411
1412         rt_newaddrmsg_fib(cmd, ifa, error, rt, RT_ALL_FIBS);
1413 }
1414
1415 /*
1416  * This is the analogue to the rt_newaddrmsg which performs the same
1417  * function but for multicast group memberhips.  This is easier since
1418  * there is no route state to worry about.
1419  */
1420 void
1421 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1422 {
1423         struct rt_addrinfo info;
1424         struct mbuf *m = NULL;
1425         struct ifnet *ifp = ifma->ifma_ifp;
1426         struct ifma_msghdr *ifmam;
1427
1428         if (route_cb.any_count == 0)
1429                 return;
1430
1431         bzero((caddr_t)&info, sizeof(info));
1432         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1433         info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1434         /*
1435          * If a link-layer address is present, present it as a ``gateway''
1436          * (similarly to how ARP entries, e.g., are presented).
1437          */
1438         info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1439         m = rt_msg1(cmd, &info);
1440         if (m == NULL)
1441                 return;
1442         ifmam = mtod(m, struct ifma_msghdr *);
1443         KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1444             __func__));
1445         ifmam->ifmam_index = ifp->if_index;
1446         ifmam->ifmam_addrs = info.rti_addrs;
1447         rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1448 }
1449
1450 static struct mbuf *
1451 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1452         struct rt_addrinfo *info)
1453 {
1454         struct if_announcemsghdr *ifan;
1455         struct mbuf *m;
1456
1457         if (route_cb.any_count == 0)
1458                 return NULL;
1459         bzero((caddr_t)info, sizeof(*info));
1460         m = rt_msg1(type, info);
1461         if (m != NULL) {
1462                 ifan = mtod(m, struct if_announcemsghdr *);
1463                 ifan->ifan_index = ifp->if_index;
1464                 strlcpy(ifan->ifan_name, ifp->if_xname,
1465                         sizeof(ifan->ifan_name));
1466                 ifan->ifan_what = what;
1467         }
1468         return m;
1469 }
1470
1471 /*
1472  * This is called to generate routing socket messages indicating
1473  * IEEE80211 wireless events.
1474  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1475  */
1476 void
1477 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1478 {
1479         struct mbuf *m;
1480         struct rt_addrinfo info;
1481
1482         m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1483         if (m != NULL) {
1484                 /*
1485                  * Append the ieee80211 data.  Try to stick it in the
1486                  * mbuf containing the ifannounce msg; otherwise allocate
1487                  * a new mbuf and append.
1488                  *
1489                  * NB: we assume m is a single mbuf.
1490                  */
1491                 if (data_len > M_TRAILINGSPACE(m)) {
1492                         struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1493                         if (n == NULL) {
1494                                 m_freem(m);
1495                                 return;
1496                         }
1497                         bcopy(data, mtod(n, void *), data_len);
1498                         n->m_len = data_len;
1499                         m->m_next = n;
1500                 } else if (data_len > 0) {
1501                         bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1502                         m->m_len += data_len;
1503                 }
1504                 if (m->m_flags & M_PKTHDR)
1505                         m->m_pkthdr.len += data_len;
1506                 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1507                 rt_dispatch(m, AF_UNSPEC);
1508         }
1509 }
1510
1511 /*
1512  * This is called to generate routing socket messages indicating
1513  * network interface arrival and departure.
1514  */
1515 void
1516 rt_ifannouncemsg(struct ifnet *ifp, int what)
1517 {
1518         struct mbuf *m;
1519         struct rt_addrinfo info;
1520
1521         m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1522         if (m != NULL)
1523                 rt_dispatch(m, AF_UNSPEC);
1524 }
1525
1526 static void
1527 rt_dispatch(struct mbuf *m, sa_family_t saf)
1528 {
1529         struct m_tag *tag;
1530
1531         /*
1532          * Preserve the family from the sockaddr, if any, in an m_tag for
1533          * use when injecting the mbuf into the routing socket buffer from
1534          * the netisr.
1535          */
1536         if (saf != AF_UNSPEC) {
1537                 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1538                     M_NOWAIT);
1539                 if (tag == NULL) {
1540                         m_freem(m);
1541                         return;
1542                 }
1543                 *(unsigned short *)(tag + 1) = saf;
1544                 m_tag_prepend(m, tag);
1545         }
1546 #ifdef VIMAGE
1547         if (V_loif)
1548                 m->m_pkthdr.rcvif = V_loif;
1549         else {
1550                 m_freem(m);
1551                 return;
1552         }
1553 #endif
1554         netisr_queue(NETISR_ROUTE, m);  /* mbuf is free'd on failure. */
1555 }
1556
1557 /*
1558  * This is used in dumping the kernel table via sysctl().
1559  */
1560 static int
1561 sysctl_dumpentry(struct radix_node *rn, void *vw)
1562 {
1563         struct walkarg *w = vw;
1564         struct rtentry *rt = (struct rtentry *)rn;
1565         int error = 0, size;
1566         struct rt_addrinfo info;
1567
1568         if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1569                 return 0;
1570         if ((rt->rt_flags & RTF_HOST) == 0
1571             ? jailed_without_vnet(w->w_req->td->td_ucred)
1572             : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1573                 return (0);
1574         bzero((caddr_t)&info, sizeof(info));
1575         info.rti_info[RTAX_DST] = rt_key(rt);
1576         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1577         info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1578         info.rti_info[RTAX_GENMASK] = 0;
1579         if (rt->rt_ifp) {
1580                 info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1581                 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1582                 if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1583                         info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1584         }
1585         size = rt_msg2(RTM_GET, &info, NULL, w);
1586         if (w->w_req && w->w_tmem) {
1587                 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1588
1589                 if (rt->rt_flags & RTF_GWFLAG_COMPAT)
1590                         rtm->rtm_flags = RTF_GATEWAY |
1591                                 (rt->rt_flags & ~RTF_GWFLAG_COMPAT);
1592                 else
1593                         rtm->rtm_flags = rt->rt_flags;
1594                 rt_getmetrics(rt, &rtm->rtm_rmx);
1595                 rtm->rtm_index = rt->rt_ifp->if_index;
1596                 rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1597                 rtm->rtm_addrs = info.rti_addrs;
1598                 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1599                 return (error);
1600         }
1601         return (error);
1602 }
1603
1604 #ifdef COMPAT_FREEBSD32
1605 static void
1606 copy_ifdata32(struct if_data *src, struct if_data32 *dst)
1607 {
1608
1609         bzero(dst, sizeof(*dst));
1610         CP(*src, *dst, ifi_type);
1611         CP(*src, *dst, ifi_physical);
1612         CP(*src, *dst, ifi_addrlen);
1613         CP(*src, *dst, ifi_hdrlen);
1614         CP(*src, *dst, ifi_link_state);
1615         CP(*src, *dst, ifi_vhid);
1616         CP(*src, *dst, ifi_baudrate_pf);
1617         dst->ifi_datalen = sizeof(struct if_data32);
1618         CP(*src, *dst, ifi_mtu);
1619         CP(*src, *dst, ifi_metric);
1620         CP(*src, *dst, ifi_baudrate);
1621         CP(*src, *dst, ifi_ipackets);
1622         CP(*src, *dst, ifi_ierrors);
1623         CP(*src, *dst, ifi_opackets);
1624         CP(*src, *dst, ifi_oerrors);
1625         CP(*src, *dst, ifi_collisions);
1626         CP(*src, *dst, ifi_ibytes);
1627         CP(*src, *dst, ifi_obytes);
1628         CP(*src, *dst, ifi_imcasts);
1629         CP(*src, *dst, ifi_omcasts);
1630         CP(*src, *dst, ifi_iqdrops);
1631         CP(*src, *dst, ifi_noproto);
1632         CP(*src, *dst, ifi_hwassist);
1633         CP(*src, *dst, ifi_epoch);
1634         TV_CP(*src, *dst, ifi_lastchange);
1635 }
1636 #endif
1637
1638 static int
1639 sysctl_iflist_ifml(struct ifnet *ifp, struct rt_addrinfo *info,
1640     struct walkarg *w, int len)
1641 {
1642         struct if_msghdrl *ifm;
1643
1644 #ifdef COMPAT_FREEBSD32
1645         if (w->w_req->flags & SCTL_MASK32) {
1646                 struct if_msghdrl32 *ifm32;
1647
1648                 ifm32 = (struct if_msghdrl32 *)w->w_tmem;
1649                 ifm32->ifm_addrs = info->rti_addrs;
1650                 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1651                 ifm32->ifm_index = ifp->if_index;
1652                 ifm32->_ifm_spare1 = 0;
1653                 ifm32->ifm_len = sizeof(*ifm32);
1654                 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1655
1656                 copy_ifdata32(&ifp->if_data, &ifm32->ifm_data);
1657                 /* Fixup if_data carp(4) vhid. */
1658                 if (carp_get_vhid_p != NULL)
1659                         ifm32->ifm_data.ifi_vhid =
1660                             (*carp_get_vhid_p)(ifp->if_addr);
1661                 ifm32->ifm_data.ifi_oqdrops = ifp->if_snd.ifq_drops;
1662
1663                 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm32, len));
1664         }
1665 #endif
1666         ifm = (struct if_msghdrl *)w->w_tmem;
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         ifm->ifm_len = sizeof(*ifm);
1672         ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1673
1674         ifm->ifm_data = ifp->if_data;
1675         /* Fixup if_data carp(4) vhid. */
1676         if (carp_get_vhid_p != NULL)
1677                 ifm->ifm_data.ifi_vhid = (*carp_get_vhid_p)(ifp->if_addr);
1678
1679         ifm->ifm_data.ifi_datalen += sizeof(u_long);
1680         ifm->ifi_oqdrops = ifp->if_snd.ifq_drops;
1681
1682         return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1683 }
1684
1685 static int
1686 sysctl_iflist_ifm(struct ifnet *ifp, struct rt_addrinfo *info,
1687     struct walkarg *w, int len)
1688 {
1689         struct if_msghdr *ifm;
1690
1691 #ifdef COMPAT_FREEBSD32
1692         if (w->w_req->flags & SCTL_MASK32) {
1693                 struct if_msghdr32 *ifm32;
1694
1695                 ifm32 = (struct if_msghdr32 *)w->w_tmem;
1696                 ifm32->ifm_addrs = info->rti_addrs;
1697                 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1698                 ifm32->ifm_index = ifp->if_index;
1699
1700                 copy_ifdata32(&ifp->if_data, &ifm32->ifm_data);
1701                 /* Fixup if_data carp(4) vhid. */
1702                 if (carp_get_vhid_p != NULL)
1703                         ifm32->ifm_data.ifi_vhid =
1704                             (*carp_get_vhid_p)(ifp->if_addr);
1705
1706                 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm32, len));
1707         }
1708 #endif
1709         ifm = (struct if_msghdr *)w->w_tmem;
1710         ifm->ifm_addrs = info->rti_addrs;
1711         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1712         ifm->ifm_index = ifp->if_index;
1713
1714         ifm->ifm_data = ifp->if_data;
1715         /* Fixup if_data carp(4) vhid. */
1716         if (carp_get_vhid_p != NULL)
1717                 ifm->ifm_data.ifi_vhid = (*carp_get_vhid_p)(ifp->if_addr);
1718
1719         return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1720 }
1721
1722 static int
1723 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1724     struct walkarg *w, int len)
1725 {
1726         struct ifa_msghdrl *ifam;
1727
1728 #ifdef COMPAT_FREEBSD32
1729         if (w->w_req->flags & SCTL_MASK32) {
1730                 struct ifa_msghdrl32 *ifam32;
1731
1732                 ifam32 = (struct ifa_msghdrl32 *)w->w_tmem;
1733                 ifam32->ifam_addrs = info->rti_addrs;
1734                 ifam32->ifam_flags = ifa->ifa_flags;
1735                 ifam32->ifam_index = ifa->ifa_ifp->if_index;
1736                 ifam32->_ifam_spare1 = 0;
1737                 ifam32->ifam_len = sizeof(*ifam32);
1738                 ifam32->ifam_data_off =
1739                     offsetof(struct ifa_msghdrl32, ifam_data);
1740                 ifam32->ifam_metric = ifa->ifa_metric;
1741
1742                 copy_ifdata32(&ifa->ifa_ifp->if_data, &ifam32->ifam_data);
1743                 /* Fixup if_data carp(4) vhid. */
1744                 if (carp_get_vhid_p != NULL)
1745                         ifam32->ifam_data.ifi_vhid = (*carp_get_vhid_p)(ifa);
1746
1747                 return (SYSCTL_OUT(w->w_req, (caddr_t)ifam32, len));
1748         }
1749 #endif
1750
1751         ifam = (struct ifa_msghdrl *)w->w_tmem;
1752         ifam->ifam_addrs = info->rti_addrs;
1753         ifam->ifam_flags = ifa->ifa_flags;
1754         ifam->ifam_index = ifa->ifa_ifp->if_index;
1755         ifam->_ifam_spare1 = 0;
1756         ifam->ifam_len = sizeof(*ifam);
1757         ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1758         ifam->ifam_metric = ifa->ifa_metric;
1759
1760         ifam->ifam_data = ifa->if_data;
1761         /* Fixup if_data carp(4) vhid. */
1762         if (carp_get_vhid_p != NULL)
1763                 ifam->ifam_data.ifi_vhid = (*carp_get_vhid_p)(ifa);
1764
1765         return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1766 }
1767
1768 static int
1769 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1770     struct walkarg *w, int len)
1771 {
1772         struct ifa_msghdr *ifam;
1773
1774         ifam = (struct ifa_msghdr *)w->w_tmem;
1775         ifam->ifam_addrs = info->rti_addrs;
1776         ifam->ifam_flags = ifa->ifa_flags;
1777         ifam->ifam_index = ifa->ifa_ifp->if_index;
1778         ifam->ifam_metric = ifa->ifa_metric;
1779
1780         return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1781 }
1782
1783 static int
1784 sysctl_iflist(int af, struct walkarg *w)
1785 {
1786         struct ifnet *ifp;
1787         struct ifaddr *ifa;
1788         struct rt_addrinfo info;
1789         int len, error = 0;
1790
1791         bzero((caddr_t)&info, sizeof(info));
1792         IFNET_RLOCK_NOSLEEP();
1793         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1794                 if (w->w_arg && w->w_arg != ifp->if_index)
1795                         continue;
1796                 IF_ADDR_RLOCK(ifp);
1797                 ifa = ifp->if_addr;
1798                 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1799                 len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1800                 info.rti_info[RTAX_IFP] = NULL;
1801                 if (w->w_req && w->w_tmem) {
1802                         if (w->w_op == NET_RT_IFLISTL)
1803                                 error = sysctl_iflist_ifml(ifp, &info, w, len);
1804                         else
1805                                 error = sysctl_iflist_ifm(ifp, &info, w, len);
1806                         if (error)
1807                                 goto done;
1808                 }
1809                 while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1810                         if (af && af != ifa->ifa_addr->sa_family)
1811                                 continue;
1812                         if (prison_if(w->w_req->td->td_ucred,
1813                             ifa->ifa_addr) != 0)
1814                                 continue;
1815                         info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1816                         info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1817                         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1818                         len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1819                         if (w->w_req && w->w_tmem) {
1820                                 if (w->w_op == NET_RT_IFLISTL)
1821                                         error = sysctl_iflist_ifaml(ifa, &info,
1822                                             w, len);
1823                                 else
1824                                         error = sysctl_iflist_ifam(ifa, &info,
1825                                             w, len);
1826                                 if (error)
1827                                         goto done;
1828                         }
1829                 }
1830                 IF_ADDR_RUNLOCK(ifp);
1831                 info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1832                         info.rti_info[RTAX_BRD] = NULL;
1833         }
1834 done:
1835         if (ifp != NULL)
1836                 IF_ADDR_RUNLOCK(ifp);
1837         IFNET_RUNLOCK_NOSLEEP();
1838         return (error);
1839 }
1840
1841 static int
1842 sysctl_ifmalist(int af, struct walkarg *w)
1843 {
1844         struct ifnet *ifp;
1845         struct ifmultiaddr *ifma;
1846         struct  rt_addrinfo info;
1847         int     len, error = 0;
1848         struct ifaddr *ifa;
1849
1850         bzero((caddr_t)&info, sizeof(info));
1851         IFNET_RLOCK_NOSLEEP();
1852         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1853                 if (w->w_arg && w->w_arg != ifp->if_index)
1854                         continue;
1855                 ifa = ifp->if_addr;
1856                 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1857                 IF_ADDR_RLOCK(ifp);
1858                 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1859                         if (af && af != ifma->ifma_addr->sa_family)
1860                                 continue;
1861                         if (prison_if(w->w_req->td->td_ucred,
1862                             ifma->ifma_addr) != 0)
1863                                 continue;
1864                         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1865                         info.rti_info[RTAX_GATEWAY] =
1866                             (ifma->ifma_addr->sa_family != AF_LINK) ?
1867                             ifma->ifma_lladdr : NULL;
1868                         len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1869                         if (w->w_req && w->w_tmem) {
1870                                 struct ifma_msghdr *ifmam;
1871
1872                                 ifmam = (struct ifma_msghdr *)w->w_tmem;
1873                                 ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1874                                 ifmam->ifmam_flags = 0;
1875                                 ifmam->ifmam_addrs = info.rti_addrs;
1876                                 error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1877                                 if (error) {
1878                                         IF_ADDR_RUNLOCK(ifp);
1879                                         goto done;
1880                                 }
1881                         }
1882                 }
1883                 IF_ADDR_RUNLOCK(ifp);
1884         }
1885 done:
1886         IFNET_RUNLOCK_NOSLEEP();
1887         return (error);
1888 }
1889
1890 static int
1891 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1892 {
1893         int     *name = (int *)arg1;
1894         u_int   namelen = arg2;
1895         struct radix_node_head *rnh = NULL; /* silence compiler. */
1896         int     i, lim, error = EINVAL;
1897         int     fib = 0;
1898         u_char  af;
1899         struct  walkarg w;
1900
1901         name ++;
1902         namelen--;
1903         if (req->newptr)
1904                 return (EPERM);
1905         if (name[1] == NET_RT_DUMP) {
1906                 if (namelen == 3)
1907                         fib = req->td->td_proc->p_fibnum;
1908                 else if (namelen == 4)
1909                         fib = (name[3] == RT_ALL_FIBS) ?
1910                             req->td->td_proc->p_fibnum : name[3];
1911                 else
1912                         return ((namelen < 3) ? EISDIR : ENOTDIR);
1913                 if (fib < 0 || fib >= rt_numfibs)
1914                         return (EINVAL);
1915         } else if (namelen != 3)
1916                 return ((namelen < 3) ? EISDIR : ENOTDIR);
1917         af = name[0];
1918         if (af > AF_MAX)
1919                 return (EINVAL);
1920         bzero(&w, sizeof(w));
1921         w.w_op = name[1];
1922         w.w_arg = name[2];
1923         w.w_req = req;
1924
1925         error = sysctl_wire_old_buffer(req, 0);
1926         if (error)
1927                 return (error);
1928         switch (w.w_op) {
1929
1930         case NET_RT_DUMP:
1931         case NET_RT_FLAGS:
1932                 if (af == 0) {                  /* dump all tables */
1933                         i = 1;
1934                         lim = AF_MAX;
1935                 } else                          /* dump only one table */
1936                         i = lim = af;
1937
1938                 /*
1939                  * take care of llinfo entries, the caller must
1940                  * specify an AF
1941                  */
1942                 if (w.w_op == NET_RT_FLAGS &&
1943                     (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1944                         if (af != 0)
1945                                 error = lltable_sysctl_dumparp(af, w.w_req);
1946                         else
1947                                 error = EINVAL;
1948                         break;
1949                 }
1950                 /*
1951                  * take care of routing entries
1952                  */
1953                 for (error = 0; error == 0 && i <= lim; i++) {
1954                         rnh = rt_tables_get_rnh(fib, i);
1955                         if (rnh != NULL) {
1956                                 RADIX_NODE_HEAD_RLOCK(rnh);
1957                                 error = rnh->rnh_walktree(rnh,
1958                                     sysctl_dumpentry, &w);
1959                                 RADIX_NODE_HEAD_RUNLOCK(rnh);
1960                         } else if (af != 0)
1961                                 error = EAFNOSUPPORT;
1962                 }
1963                 break;
1964
1965         case NET_RT_IFLIST:
1966         case NET_RT_IFLISTL:
1967                 error = sysctl_iflist(af, &w);
1968                 break;
1969
1970         case NET_RT_IFMALIST:
1971                 error = sysctl_ifmalist(af, &w);
1972                 break;
1973         }
1974         if (w.w_tmem)
1975                 free(w.w_tmem, M_RTABLE);
1976         return (error);
1977 }
1978
1979 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1980
1981 /*
1982  * Definitions of protocols supported in the ROUTE domain.
1983  */
1984
1985 static struct domain routedomain;               /* or at least forward */
1986
1987 static struct protosw routesw[] = {
1988 {
1989         .pr_type =              SOCK_RAW,
1990         .pr_domain =            &routedomain,
1991         .pr_flags =             PR_ATOMIC|PR_ADDR,
1992         .pr_output =            route_output,
1993         .pr_ctlinput =          raw_ctlinput,
1994         .pr_init =              raw_init,
1995         .pr_usrreqs =           &route_usrreqs
1996 }
1997 };
1998
1999 static struct domain routedomain = {
2000         .dom_family =           PF_ROUTE,
2001         .dom_name =              "route",
2002         .dom_protosw =          routesw,
2003         .dom_protoswNPROTOSW =  &routesw[sizeof(routesw)/sizeof(routesw[0])]
2004 };
2005
2006 VNET_DOMAIN_SET(route);