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 #define _IN_NET_RTSOCK_C
  55 #include <net/if.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/vnet.h>
  63
  64 #include <netinet/in.h>
  65 #include <netinet/if_ether.h>
  66 #include <netinet/ip_carp.h>
  67 #ifdef INET6
  68 #include <netinet6/ip6_var.h>
  69 #include <netinet6/scope6_var.h>
  70 #endif
  71
  72 #ifdef COMPAT_FREEBSD32
  73 #include <sys/mount.h>
  74 #include <compat/freebsd32/freebsd32.h>
  75
  76 struct if_data32 {
  77         uint8_t ifi_type;
  78         uint8_t ifi_physical;
  79         uint8_t ifi_addrlen;
  80         uint8_t ifi_hdrlen;
  81         uint8_t ifi_link_state;
  82         uint8_t ifi_vhid;
  83         uint8_t ifi_baudrate_pf;
  84         uint8_t ifi_datalen;
  85         uint32_t ifi_mtu;
  86         uint32_t ifi_metric;
  87         uint32_t ifi_baudrate;
  88         uint32_t ifi_ipackets;
  89         uint32_t ifi_ierrors;
  90         uint32_t ifi_opackets;
  91         uint32_t ifi_oerrors;
  92         uint32_t ifi_collisions;
  93         uint32_t ifi_ibytes;
  94         uint32_t ifi_obytes;
  95         uint32_t ifi_imcasts;
  96         uint32_t ifi_omcasts;
  97         uint32_t ifi_iqdrops;
  98         uint32_t ifi_noproto;
  99         uint32_t ifi_hwassist;
 100         int32_t ifi_epoch;
 101         struct  timeval32 ifi_lastchange;
 102         uint32_t ifi_oqdrops;
 103 };
 104
 105 struct if_msghdr32 {
 106         uint16_t ifm_msglen;
 107         uint8_t ifm_version;
 108         uint8_t ifm_type;
 109         int32_t ifm_addrs;
 110         int32_t ifm_flags;
 111         uint16_t ifm_index;
 112         struct  if_data32 ifm_data;
 113 };
 114
 115 struct if_msghdrl32 {
 116         uint16_t ifm_msglen;
 117         uint8_t ifm_version;
 118         uint8_t ifm_type;
 119         int32_t ifm_addrs;
 120         int32_t ifm_flags;
 121         uint16_t ifm_index;
 122         uint16_t _ifm_spare1;
 123         uint16_t ifm_len;
 124         uint16_t ifm_data_off;
 125         struct  if_data32 ifm_data;
 126 };
 127
 128 struct ifa_msghdrl32 {
 129         uint16_t ifam_msglen;
 130         uint8_t ifam_version;
 131         uint8_t ifam_type;
 132         int32_t ifam_addrs;
 133         int32_t ifam_flags;
 134         uint16_t ifam_index;
 135         uint16_t _ifam_spare1;
 136         uint16_t ifam_len;
 137         uint16_t ifam_data_off;
 138         int32_t ifam_metric;
 139         struct  if_data32 ifam_data;
 140 };
 141 #endif /* COMPAT_FREEBSD32 */
 142
 143 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
 144
 145 /* NB: these are not modified */
 146 static struct   sockaddr route_src = { 2, PF_ROUTE, };
 147 static struct   sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
 148
 149 /* These are external hooks for CARP. */
 150 int     (*carp_get_vhid_p)(struct ifaddr *);
 151
 152 /*
 153  * Used by rtsock/raw_input callback code to decide whether to filter the update
 154  * notification to a socket bound to a particular FIB.
 155  */
 156 #define RTS_FILTER_FIB  M_PROTO8
 157
 158 static struct {
 159         int     ip_count;       /* attached w/ AF_INET */
 160         int     ip6_count;      /* attached w/ AF_INET6 */
 161         int     ipx_count;      /* attached w/ AF_IPX */
 162         int     any_count;      /* total attached */
 163 } route_cb;
 164
 165 struct mtx rtsock_mtx;
 166 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
 167
 168 #define RTSOCK_LOCK()   mtx_lock(&rtsock_mtx)
 169 #define RTSOCK_UNLOCK() mtx_unlock(&rtsock_mtx)
 170 #define RTSOCK_LOCK_ASSERT()    mtx_assert(&rtsock_mtx, MA_OWNED)
 171
 172 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
 173
 174 struct walkarg {
 175         int     w_tmemsize;
 176         int     w_op, w_arg;
 177         caddr_t w_tmem;
 178         struct sysctl_req *w_req;
 179 };
 180
 181 static void     rts_input(struct mbuf *m);
 182 static struct mbuf *rt_msg1(int type, struct rt_addrinfo *rtinfo);
 183 static int      rt_msg2(int type, struct rt_addrinfo *rtinfo,
 184                         caddr_t cp, struct walkarg *w);
 185 static int      rt_xaddrs(caddr_t cp, caddr_t cplim,
 186                         struct rt_addrinfo *rtinfo);
 187 static int      sysctl_dumpentry(struct radix_node *rn, void *vw);
 188 static int      sysctl_iflist(int af, struct walkarg *w);
 189 static int      sysctl_ifmalist(int af, struct walkarg *w);
 190 static int      route_output(struct mbuf *m, struct socket *so);
 191 static void     rt_setmetrics(const struct rt_msghdr *rtm, struct rtentry *rt);
 192 static void     rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out);
 193 static void     rt_dispatch(struct mbuf *, sa_family_t);
 194
 195 static struct netisr_handler rtsock_nh = {
 196         .nh_name = "rtsock",
 197         .nh_handler = rts_input,
 198         .nh_proto = NETISR_ROUTE,
 199         .nh_policy = NETISR_POLICY_SOURCE,
 200 };
 201
 202 static int
 203 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
 204 {
 205         int error, qlimit;
 206
 207         netisr_getqlimit(&rtsock_nh, &qlimit);
 208         error = sysctl_handle_int(oidp, &qlimit, 0, req);
 209         if (error || !req->newptr)
 210                 return (error);
 211         if (qlimit < 1)
 212                 return (EINVAL);
 213         return (netisr_setqlimit(&rtsock_nh, qlimit));
 214 }
 215 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
 216     0, 0, sysctl_route_netisr_maxqlen, "I",
 217     "maximum routing socket dispatch queue length");
 218
 219 static void
 220 rts_init(void)
 221 {
 222         int tmp;
 223
 224         if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
 225                 rtsock_nh.nh_qlimit = tmp;
 226         netisr_register(&rtsock_nh);
 227 }
 228 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0);
 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                 route_cb.ip_count++;
 318                 break;
 319         case AF_INET6:
 320                 route_cb.ip6_count++;
 321                 break;
 322         case AF_IPX:
 323                 route_cb.ipx_count++;
 324                 break;
 325         }
 326         route_cb.any_count++;
 327         RTSOCK_UNLOCK();
 328         soisconnected(so);
 329         so->so_options |= SO_USELOOPBACK;
 330         return 0;
 331 }
 332
 333 static int
 334 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
 335 {
 336
 337         return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
 338 }
 339
 340 static int
 341 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
 342 {
 343
 344         return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
 345 }
 346
 347 /* pru_connect2 is EOPNOTSUPP */
 348 /* pru_control is EOPNOTSUPP */
 349
 350 static void
 351 rts_detach(struct socket *so)
 352 {
 353         struct rawcb *rp = sotorawcb(so);
 354
 355         KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
 356
 357         RTSOCK_LOCK();
 358         switch(rp->rcb_proto.sp_protocol) {
 359         case AF_INET:
 360                 route_cb.ip_count--;
 361                 break;
 362         case AF_INET6:
 363                 route_cb.ip6_count--;
 364                 break;
 365         case AF_IPX:
 366                 route_cb.ipx_count--;
 367                 break;
 368         }
 369         route_cb.any_count--;
 370         RTSOCK_UNLOCK();
 371         raw_usrreqs.pru_detach(so);
 372 }
 373
 374 static int
 375 rts_disconnect(struct socket *so)
 376 {
 377
 378         return (raw_usrreqs.pru_disconnect(so));
 379 }
 380
 381 /* pru_listen is EOPNOTSUPP */
 382
 383 static int
 384 rts_peeraddr(struct socket *so, struct sockaddr **nam)
 385 {
 386
 387         return (raw_usrreqs.pru_peeraddr(so, nam));
 388 }
 389
 390 /* pru_rcvd is EOPNOTSUPP */
 391 /* pru_rcvoob is EOPNOTSUPP */
 392
 393 static int
 394 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
 395          struct mbuf *control, struct thread *td)
 396 {
 397
 398         return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
 399 }
 400
 401 /* pru_sense is null */
 402
 403 static int
 404 rts_shutdown(struct socket *so)
 405 {
 406
 407         return (raw_usrreqs.pru_shutdown(so));
 408 }
 409
 410 static int
 411 rts_sockaddr(struct socket *so, struct sockaddr **nam)
 412 {
 413
 414         return (raw_usrreqs.pru_sockaddr(so, nam));
 415 }
 416
 417 static struct pr_usrreqs route_usrreqs = {
 418         .pru_abort =            rts_abort,
 419         .pru_attach =           rts_attach,
 420         .pru_bind =             rts_bind,
 421         .pru_connect =          rts_connect,
 422         .pru_detach =           rts_detach,
 423         .pru_disconnect =       rts_disconnect,
 424         .pru_peeraddr =         rts_peeraddr,
 425         .pru_send =             rts_send,
 426         .pru_shutdown =         rts_shutdown,
 427         .pru_sockaddr =         rts_sockaddr,
 428         .pru_close =            rts_close,
 429 };
 430
 431 #ifndef _SOCKADDR_UNION_DEFINED
 432 #define _SOCKADDR_UNION_DEFINED
 433 /*
 434  * The union of all possible address formats we handle.
 435  */
 436 union sockaddr_union {
 437         struct sockaddr         sa;
 438         struct sockaddr_in      sin;
 439         struct sockaddr_in6     sin6;
 440 };
 441 #endif /* _SOCKADDR_UNION_DEFINED */
 442
 443 static int
 444 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
 445     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
 446 {
 447
 448         /* First, see if the returned address is part of the jail. */
 449         if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
 450                 info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
 451                 return (0);
 452         }
 453
 454         switch (info->rti_info[RTAX_DST]->sa_family) {
 455 #ifdef INET
 456         case AF_INET:
 457         {
 458                 struct in_addr ia;
 459                 struct ifaddr *ifa;
 460                 int found;
 461
 462                 found = 0;
 463                 /*
 464                  * Try to find an address on the given outgoing interface
 465                  * that belongs to the jail.
 466                  */
 467                 IF_ADDR_RLOCK(ifp);
 468                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 469                         struct sockaddr *sa;
 470                         sa = ifa->ifa_addr;
 471                         if (sa->sa_family != AF_INET)
 472                                 continue;
 473                         ia = ((struct sockaddr_in *)sa)->sin_addr;
 474                         if (prison_check_ip4(cred, &ia) == 0) {
 475                                 found = 1;
 476                                 break;
 477                         }
 478                 }
 479                 IF_ADDR_RUNLOCK(ifp);
 480                 if (!found) {
 481                         /*
 482                          * As a last resort return the 'default' jail address.
 483                          */
 484                         ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
 485                             sin_addr;
 486                         if (prison_get_ip4(cred, &ia) != 0)
 487                                 return (ESRCH);
 488                 }
 489                 bzero(&saun->sin, sizeof(struct sockaddr_in));
 490                 saun->sin.sin_len = sizeof(struct sockaddr_in);
 491                 saun->sin.sin_family = AF_INET;
 492                 saun->sin.sin_addr.s_addr = ia.s_addr;
 493                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
 494                 break;
 495         }
 496 #endif
 497 #ifdef INET6
 498         case AF_INET6:
 499         {
 500                 struct in6_addr ia6;
 501                 struct ifaddr *ifa;
 502                 int found;
 503
 504                 found = 0;
 505                 /*
 506                  * Try to find an address on the given outgoing interface
 507                  * that belongs to the jail.
 508                  */
 509                 IF_ADDR_RLOCK(ifp);
 510                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 511                         struct sockaddr *sa;
 512                         sa = ifa->ifa_addr;
 513                         if (sa->sa_family != AF_INET6)
 514                                 continue;
 515                         bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
 516                             &ia6, sizeof(struct in6_addr));
 517                         if (prison_check_ip6(cred, &ia6) == 0) {
 518                                 found = 1;
 519                                 break;
 520                         }
 521                 }
 522                 IF_ADDR_RUNLOCK(ifp);
 523                 if (!found) {
 524                         /*
 525                          * As a last resort return the 'default' jail address.
 526                          */
 527                         ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
 528                             sin6_addr;
 529                         if (prison_get_ip6(cred, &ia6) != 0)
 530                                 return (ESRCH);
 531                 }
 532                 bzero(&saun->sin6, sizeof(struct sockaddr_in6));
 533                 saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
 534                 saun->sin6.sin6_family = AF_INET6;
 535                 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
 536                 if (sa6_recoverscope(&saun->sin6) != 0)
 537                         return (ESRCH);
 538                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
 539                 break;
 540         }
 541 #endif
 542         default:
 543                 return (ESRCH);
 544         }
 545         return (0);
 546 }
 547
 548 /*ARGSUSED*/
 549 static int
 550 route_output(struct mbuf *m, struct socket *so)
 551 {
 552 #define sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
 553         struct rt_msghdr *rtm = NULL;
 554         struct rtentry *rt = NULL;
 555         struct radix_node_head *rnh;
 556         struct rt_addrinfo info;
 557 #ifdef INET6
 558         struct sockaddr_storage ss;
 559         struct sockaddr_in6 *sin6;
 560         int i, rti_need_deembed = 0;
 561 #endif
 562         int len, error = 0;
 563         struct ifnet *ifp = NULL;
 564         union sockaddr_union saun;
 565         sa_family_t saf = AF_UNSPEC;
 566
 567 #define senderr(e) { error = e; goto flush;}
 568         if (m == NULL || ((m->m_len < sizeof(long)) &&
 569                        (m = m_pullup(m, sizeof(long))) == NULL))
 570                 return (ENOBUFS);
 571         if ((m->m_flags & M_PKTHDR) == 0)
 572                 panic("route_output");
 573         len = m->m_pkthdr.len;
 574         if (len < sizeof(*rtm) ||
 575             len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
 576                 info.rti_info[RTAX_DST] = NULL;
 577                 senderr(EINVAL);
 578         }
 579         R_Malloc(rtm, struct rt_msghdr *, len);
 580         if (rtm == NULL) {
 581                 info.rti_info[RTAX_DST] = NULL;
 582                 senderr(ENOBUFS);
 583         }
 584         m_copydata(m, 0, len, (caddr_t)rtm);
 585         if (rtm->rtm_version != RTM_VERSION) {
 586                 info.rti_info[RTAX_DST] = NULL;
 587                 senderr(EPROTONOSUPPORT);
 588         }
 589         rtm->rtm_pid = curproc->p_pid;
 590         bzero(&info, sizeof(info));
 591         info.rti_addrs = rtm->rtm_addrs;
 592         /*
 593          * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
 594          * link-local address because rtrequest requires addresses with
 595          * embedded scope id.
 596          */
 597         if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
 598                 info.rti_info[RTAX_DST] = NULL;
 599                 senderr(EINVAL);
 600         }
 601         info.rti_flags = rtm->rtm_flags;
 602         if (info.rti_info[RTAX_DST] == NULL ||
 603             info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
 604             (info.rti_info[RTAX_GATEWAY] != NULL &&
 605              info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
 606                 senderr(EINVAL);
 607         saf = info.rti_info[RTAX_DST]->sa_family;
 608         /*
 609          * Verify that the caller has the appropriate privilege; RTM_GET
 610          * is the only operation the non-superuser is allowed.
 611          */
 612         if (rtm->rtm_type != RTM_GET) {
 613                 error = priv_check(curthread, PRIV_NET_ROUTE);
 614                 if (error)
 615                         senderr(error);
 616         }
 617
 618         /*
 619          * The given gateway address may be an interface address.
 620          * For example, issuing a "route change" command on a route
 621          * entry that was created from a tunnel, and the gateway
 622          * address given is the local end point. In this case the
 623          * RTF_GATEWAY flag must be cleared or the destination will
 624          * not be reachable even though there is no error message.
 625          */
 626         if (info.rti_info[RTAX_GATEWAY] != NULL &&
 627             info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
 628                 struct route gw_ro;
 629
 630                 bzero(&gw_ro, sizeof(gw_ro));
 631                 gw_ro.ro_dst = *info.rti_info[RTAX_GATEWAY];
 632                 rtalloc_ign_fib(&gw_ro, 0, so->so_fibnum);
 633                 /*
 634                  * A host route through the loopback interface is
 635                  * installed for each interface adddress. In pre 8.0
 636                  * releases the interface address of a PPP link type
 637                  * is not reachable locally. This behavior is fixed as
 638                  * part of the new L2/L3 redesign and rewrite work. The
 639                  * signature of this interface address route is the
 640                  * AF_LINK sa_family type of the rt_gateway, and the
 641                  * rt_ifp has the IFF_LOOPBACK flag set.
 642                  */
 643                 if (gw_ro.ro_rt != NULL &&
 644                     gw_ro.ro_rt->rt_gateway->sa_family == AF_LINK &&
 645                     gw_ro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) {
 646                         info.rti_flags &= ~RTF_GATEWAY;
 647                         info.rti_flags |= RTF_GWFLAG_COMPAT;
 648                 }
 649                 if (gw_ro.ro_rt != NULL)
 650                         RTFREE(gw_ro.ro_rt);
 651         }
 652
 653         switch (rtm->rtm_type) {
 654                 struct rtentry *saved_nrt;
 655
 656         case RTM_ADD:
 657                 if (info.rti_info[RTAX_GATEWAY] == NULL)
 658                         senderr(EINVAL);
 659                 saved_nrt = NULL;
 660
 661                 /* support for new ARP code */
 662                 if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
 663                     (rtm->rtm_flags & RTF_LLDATA) != 0) {
 664                         error = lla_rt_output(rtm, &info);
 665 #ifdef INET6
 666                         if (error == 0)
 667                                 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 668 #endif
 669                         break;
 670                 }
 671                 error = rtrequest1_fib(RTM_ADD, &info, &saved_nrt,
 672                     so->so_fibnum);
 673                 if (error == 0 && saved_nrt) {
 674 #ifdef INET6
 675                         rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 676 #endif
 677                         RT_LOCK(saved_nrt);
 678                         rt_setmetrics(rtm, saved_nrt);
 679                         rtm->rtm_index = saved_nrt->rt_ifp->if_index;
 680                         RT_REMREF(saved_nrt);
 681                         RT_UNLOCK(saved_nrt);
 682                 }
 683                 break;
 684
 685         case RTM_DELETE:
 686                 saved_nrt = NULL;
 687                 /* support for new ARP code */
 688                 if (info.rti_info[RTAX_GATEWAY] &&
 689                     (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
 690                     (rtm->rtm_flags & RTF_LLDATA) != 0) {
 691                         error = lla_rt_output(rtm, &info);
 692 #ifdef INET6
 693                         if (error == 0)
 694                                 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 695 #endif
 696                         break;
 697                 }
 698                 error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt,
 699                     so->so_fibnum);
 700                 if (error == 0) {
 701                         RT_LOCK(saved_nrt);
 702                         rt = saved_nrt;
 703                         goto report;
 704                 }
 705 #ifdef INET6
 706                 /* rt_msg2() will not be used when RTM_DELETE fails. */
 707                 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 708 #endif
 709                 break;
 710
 711         case RTM_GET:
 712         case RTM_CHANGE:
 713         case RTM_LOCK:
 714                 rnh = rt_tables_get_rnh(so->so_fibnum,
 715                     info.rti_info[RTAX_DST]->sa_family);
 716                 if (rnh == NULL)
 717                         senderr(EAFNOSUPPORT);
 718
 719                 RADIX_NODE_HEAD_RLOCK(rnh);
 720
 721                 if (info.rti_info[RTAX_NETMASK] == NULL &&
 722                     rtm->rtm_type == RTM_GET) {
 723                         /*
 724                          * Provide logest prefix match for
 725                          * address lookup (no mask).
 726                          * 'route -n get addr'
 727                          */
 728                         rt = (struct rtentry *) rnh->rnh_matchaddr(
 729                             info.rti_info[RTAX_DST], rnh);
 730                 } else
 731                         rt = (struct rtentry *) rnh->rnh_lookup(
 732                             info.rti_info[RTAX_DST],
 733                             info.rti_info[RTAX_NETMASK], rnh);
 734
 735                 if (rt == NULL) {
 736                         RADIX_NODE_HEAD_RUNLOCK(rnh);
 737                         senderr(ESRCH);
 738                 }
 739 #ifdef RADIX_MPATH
 740                 /*
 741                  * for RTM_CHANGE/LOCK, if we got multipath routes,
 742                  * we require users to specify a matching RTAX_GATEWAY.
 743                  *
 744                  * for RTM_GET, gate is optional even with multipath.
 745                  * if gate == NULL the first match is returned.
 746                  * (no need to call rt_mpath_matchgate if gate == NULL)
 747                  */
 748                 if (rn_mpath_capable(rnh) &&
 749                     (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
 750                         rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
 751                         if (!rt) {
 752                                 RADIX_NODE_HEAD_RUNLOCK(rnh);
 753                                 senderr(ESRCH);
 754                         }
 755                 }
 756 #endif
 757                 /*
 758                  * If performing proxied L2 entry insertion, and
 759                  * the actual PPP host entry is found, perform
 760                  * another search to retrieve the prefix route of
 761                  * the local end point of the PPP link.
 762                  */
 763                 if (rtm->rtm_flags & RTF_ANNOUNCE) {
 764                         struct sockaddr laddr;
 765
 766                         if (rt->rt_ifp != NULL &&
 767                             rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
 768                                 struct ifaddr *ifa;
 769
 770                                 ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1);
 771                                 if (ifa != NULL)
 772                                         rt_maskedcopy(ifa->ifa_addr,
 773                                                       &laddr,
 774                                                       ifa->ifa_netmask);
 775                         } else
 776                                 rt_maskedcopy(rt->rt_ifa->ifa_addr,
 777                                               &laddr,
 778                                               rt->rt_ifa->ifa_netmask);
 779                         /*
 780                          * refactor rt and no lock operation necessary
 781                          */
 782                         rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr, rnh);
 783                         if (rt == NULL) {
 784                                 RADIX_NODE_HEAD_RUNLOCK(rnh);
 785                                 senderr(ESRCH);
 786                         }
 787                 }
 788                 RT_LOCK(rt);
 789                 RT_ADDREF(rt);
 790                 RADIX_NODE_HEAD_RUNLOCK(rnh);
 791
 792                 switch(rtm->rtm_type) {
 793
 794                 case RTM_GET:
 795                 report:
 796                         RT_LOCK_ASSERT(rt);
 797                         if ((rt->rt_flags & RTF_HOST) == 0
 798                             ? jailed_without_vnet(curthread->td_ucred)
 799                             : prison_if(curthread->td_ucred,
 800                             rt_key(rt)) != 0) {
 801                                 RT_UNLOCK(rt);
 802                                 senderr(ESRCH);
 803                         }
 804                         info.rti_info[RTAX_DST] = rt_key(rt);
 805                         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
 806                         info.rti_info[RTAX_NETMASK] = rt_mask(rt);
 807                         info.rti_info[RTAX_GENMASK] = 0;
 808                         if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
 809                                 ifp = rt->rt_ifp;
 810                                 if (ifp) {
 811                                         info.rti_info[RTAX_IFP] =
 812                                             ifp->if_addr->ifa_addr;
 813                                         error = rtm_get_jailed(&info, ifp, rt,
 814                                             &saun, curthread->td_ucred);
 815                                         if (error != 0) {
 816                                                 RT_UNLOCK(rt);
 817                                                 senderr(error);
 818                                         }
 819                                         if (ifp->if_flags & IFF_POINTOPOINT)
 820                                                 info.rti_info[RTAX_BRD] =
 821                                                     rt->rt_ifa->ifa_dstaddr;
 822                                         rtm->rtm_index = ifp->if_index;
 823                                 } else {
 824                                         info.rti_info[RTAX_IFP] = NULL;
 825                                         info.rti_info[RTAX_IFA] = NULL;
 826                                 }
 827                         } else if ((ifp = rt->rt_ifp) != NULL) {
 828                                 rtm->rtm_index = ifp->if_index;
 829                         }
 830                         len = rt_msg2(rtm->rtm_type, &info, NULL, NULL);
 831                         if (len > rtm->rtm_msglen) {
 832                                 struct rt_msghdr *new_rtm;
 833                                 R_Malloc(new_rtm, struct rt_msghdr *, len);
 834                                 if (new_rtm == NULL) {
 835                                         RT_UNLOCK(rt);
 836                                         senderr(ENOBUFS);
 837                                 }
 838                                 bcopy(rtm, new_rtm, rtm->rtm_msglen);
 839                                 Free(rtm); rtm = new_rtm;
 840                         }
 841                         (void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm, NULL);
 842                         if (rt->rt_flags & RTF_GWFLAG_COMPAT)
 843                                 rtm->rtm_flags = RTF_GATEWAY |
 844                                         (rt->rt_flags & ~RTF_GWFLAG_COMPAT);
 845                         else
 846                                 rtm->rtm_flags = rt->rt_flags;
 847                         rt_getmetrics(rt, &rtm->rtm_rmx);
 848                         rtm->rtm_addrs = info.rti_addrs;
 849                         break;
 850
 851                 case RTM_CHANGE:
 852                         /*
 853                          * New gateway could require new ifaddr, ifp;
 854                          * flags may also be different; ifp may be specified
 855                          * by ll sockaddr when protocol address is ambiguous
 856                          */
 857                         if (((rt->rt_flags & RTF_GATEWAY) &&
 858                              info.rti_info[RTAX_GATEWAY] != NULL) ||
 859                             info.rti_info[RTAX_IFP] != NULL ||
 860                             (info.rti_info[RTAX_IFA] != NULL &&
 861                              !sa_equal(info.rti_info[RTAX_IFA],
 862                                        rt->rt_ifa->ifa_addr))) {
 863                                 RT_UNLOCK(rt);
 864                                 RADIX_NODE_HEAD_LOCK(rnh);
 865                                 error = rt_getifa_fib(&info, rt->rt_fibnum);
 866                                 /*
 867                                  * XXXRW: Really we should release this
 868                                  * reference later, but this maintains
 869                                  * historical behavior.
 870                                  */
 871                                 if (info.rti_ifa != NULL)
 872                                         ifa_free(info.rti_ifa);
 873                                 RADIX_NODE_HEAD_UNLOCK(rnh);
 874                                 if (error != 0)
 875                                         senderr(error);
 876                                 RT_LOCK(rt);
 877                         }
 878                         if (info.rti_ifa != NULL &&
 879                             info.rti_ifa != rt->rt_ifa &&
 880                             rt->rt_ifa != NULL &&
 881                             rt->rt_ifa->ifa_rtrequest != NULL) {
 882                                 rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
 883                                     &info);
 884                                 ifa_free(rt->rt_ifa);
 885                         }
 886                         if (info.rti_info[RTAX_GATEWAY] != NULL) {
 887                                 RT_UNLOCK(rt);
 888                                 RADIX_NODE_HEAD_LOCK(rnh);
 889                                 RT_LOCK(rt);
 890
 891                                 error = rt_setgate(rt, rt_key(rt),
 892                                     info.rti_info[RTAX_GATEWAY]);
 893                                 RADIX_NODE_HEAD_UNLOCK(rnh);
 894                                 if (error != 0) {
 895                                         RT_UNLOCK(rt);
 896                                         senderr(error);
 897                                 }
 898                                 rt->rt_flags &= ~RTF_GATEWAY;
 899                                 rt->rt_flags |= (RTF_GATEWAY & info.rti_flags);
 900                         }
 901                         if (info.rti_ifa != NULL &&
 902                             info.rti_ifa != rt->rt_ifa) {
 903                                 ifa_ref(info.rti_ifa);
 904                                 rt->rt_ifa = info.rti_ifa;
 905                                 rt->rt_ifp = info.rti_ifp;
 906                         }
 907                         /* Allow some flags to be toggled on change. */
 908                         rt->rt_flags = (rt->rt_flags & ~RTF_FMASK) |
 909                                     (rtm->rtm_flags & RTF_FMASK);
 910                         rt_setmetrics(rtm, rt);
 911                         rtm->rtm_index = rt->rt_ifp->if_index;
 912                         if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
 913                                rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
 914                         /* FALLTHROUGH */
 915                 case RTM_LOCK:
 916                         /* We don't support locks anymore */
 917                         break;
 918                 }
 919                 RT_UNLOCK(rt);
 920                 break;
 921
 922         default:
 923                 senderr(EOPNOTSUPP);
 924         }
 925
 926 flush:
 927         if (rtm) {
 928                 if (error)
 929                         rtm->rtm_errno = error;
 930                 else
 931                         rtm->rtm_flags |= RTF_DONE;
 932         }
 933         if (rt)         /* XXX can this be true? */
 934                 RTFREE(rt);
 935     {
 936         struct rawcb *rp = NULL;
 937         /*
 938          * Check to see if we don't want our own messages.
 939          */
 940         if ((so->so_options & SO_USELOOPBACK) == 0) {
 941                 if (route_cb.any_count <= 1) {
 942                         if (rtm)
 943                                 Free(rtm);
 944                         m_freem(m);
 945                         return (error);
 946                 }
 947                 /* There is another listener, so construct message */
 948                 rp = sotorawcb(so);
 949         }
 950         if (rtm) {
 951 #ifdef INET6
 952                 if (rti_need_deembed) {
 953                         /* sin6_scope_id is recovered before sending rtm. */
 954                         sin6 = (struct sockaddr_in6 *)&ss;
 955                         for (i = 0; i < RTAX_MAX; i++) {
 956                                 if (info.rti_info[i] == NULL)
 957                                         continue;
 958                                 if (info.rti_info[i]->sa_family != AF_INET6)
 959                                         continue;
 960                                 bcopy(info.rti_info[i], sin6, sizeof(*sin6));
 961                                 if (sa6_recoverscope(sin6) == 0)
 962                                         bcopy(sin6, info.rti_info[i],
 963                                                     sizeof(*sin6));
 964                         }
 965                 }
 966 #endif
 967                 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
 968                 if (m->m_pkthdr.len < rtm->rtm_msglen) {
 969                         m_freem(m);
 970                         m = NULL;
 971                 } else if (m->m_pkthdr.len > rtm->rtm_msglen)
 972                         m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
 973         }
 974         if (m) {
 975                 M_SETFIB(m, so->so_fibnum);
 976                 m->m_flags |= RTS_FILTER_FIB;
 977                 if (rp) {
 978                         /*
 979                          * XXX insure we don't get a copy by
 980                          * invalidating our protocol
 981                          */
 982                         unsigned short family = rp->rcb_proto.sp_family;
 983                         rp->rcb_proto.sp_family = 0;
 984                         rt_dispatch(m, saf);
 985                         rp->rcb_proto.sp_family = family;
 986                 } else
 987                         rt_dispatch(m, saf);
 988         }
 989         /* info.rti_info[RTAX_DST] (used above) can point inside of rtm */
 990         if (rtm)
 991                 Free(rtm);
 992     }
 993         return (error);
 994 #undef  sa_equal
 995 }
 996
 997 static void
 998 rt_setmetrics(const struct rt_msghdr *rtm, struct rtentry *rt)
 999 {
1000
1001         if (rtm->rtm_inits & RTV_MTU)
1002                 rt->rt_mtu = rtm->rtm_rmx.rmx_mtu;
1003         if (rtm->rtm_inits & RTV_WEIGHT)
1004                 rt->rt_weight = rtm->rtm_rmx.rmx_weight;
1005         /* Kernel -> userland timebase conversion. */
1006         if (rtm->rtm_inits & RTV_EXPIRE)
1007                 rt->rt_expire = rtm->rtm_rmx.rmx_expire ?
1008                     rtm->rtm_rmx.rmx_expire - time_second + time_uptime : 0;
1009 }
1010
1011 static void
1012 rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out)
1013 {
1014
1015         bzero(out, sizeof(*out));
1016         out->rmx_mtu = rt->rt_mtu;
1017         out->rmx_weight = rt->rt_weight;
1018         out->rmx_pksent = counter_u64_fetch(rt->rt_pksent);
1019         /* Kernel -> userland timebase conversion. */
1020         out->rmx_expire = rt->rt_expire ?
1021             rt->rt_expire - time_uptime + time_second : 0;
1022 }
1023
1024 /*
1025  * Extract the addresses of the passed sockaddrs.
1026  * Do a little sanity checking so as to avoid bad memory references.
1027  * This data is derived straight from userland.
1028  */
1029 static int
1030 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1031 {
1032         struct sockaddr *sa;
1033         int i;
1034
1035         for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1036                 if ((rtinfo->rti_addrs & (1 << i)) == 0)
1037                         continue;
1038                 sa = (struct sockaddr *)cp;
1039                 /*
1040                  * It won't fit.
1041                  */
1042                 if (cp + sa->sa_len > cplim)
1043                         return (EINVAL);
1044                 /*
1045                  * there are no more.. quit now
1046                  * If there are more bits, they are in error.
1047                  * I've seen this. route(1) can evidently generate these.
1048                  * This causes kernel to core dump.
1049                  * for compatibility, If we see this, point to a safe address.
1050                  */
1051                 if (sa->sa_len == 0) {
1052                         rtinfo->rti_info[i] = &sa_zero;
1053                         return (0); /* should be EINVAL but for compat */
1054                 }
1055                 /* accept it */
1056 #ifdef INET6
1057                 if (sa->sa_family == AF_INET6)
1058                         sa6_embedscope((struct sockaddr_in6 *)sa,
1059                             V_ip6_use_defzone);
1060 #endif
1061                 rtinfo->rti_info[i] = sa;
1062                 cp += SA_SIZE(sa);
1063         }
1064         return (0);
1065 }
1066
1067 /*
1068  * Used by the routing socket.
1069  */
1070 static struct mbuf *
1071 rt_msg1(int type, struct rt_addrinfo *rtinfo)
1072 {
1073         struct rt_msghdr *rtm;
1074         struct mbuf *m;
1075         int i;
1076         struct sockaddr *sa;
1077 #ifdef INET6
1078         struct sockaddr_storage ss;
1079         struct sockaddr_in6 *sin6;
1080 #endif
1081         int len, dlen;
1082
1083         switch (type) {
1084
1085         case RTM_DELADDR:
1086         case RTM_NEWADDR:
1087                 len = sizeof(struct ifa_msghdr);
1088                 break;
1089
1090         case RTM_DELMADDR:
1091         case RTM_NEWMADDR:
1092                 len = sizeof(struct ifma_msghdr);
1093                 break;
1094
1095         case RTM_IFINFO:
1096                 len = sizeof(struct if_msghdr);
1097                 break;
1098
1099         case RTM_IFANNOUNCE:
1100         case RTM_IEEE80211:
1101                 len = sizeof(struct if_announcemsghdr);
1102                 break;
1103
1104         default:
1105                 len = sizeof(struct rt_msghdr);
1106         }
1107
1108         /* XXXGL: can we use MJUMPAGESIZE cluster here? */
1109         KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1110         if (len > MHLEN)
1111                 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1112         else
1113                 m = m_gethdr(M_NOWAIT, MT_DATA);
1114         if (m == NULL)
1115                 return (m);
1116
1117         m->m_pkthdr.len = m->m_len = len;
1118         rtm = mtod(m, struct rt_msghdr *);
1119         bzero((caddr_t)rtm, len);
1120         for (i = 0; i < RTAX_MAX; i++) {
1121                 if ((sa = rtinfo->rti_info[i]) == NULL)
1122                         continue;
1123                 rtinfo->rti_addrs |= (1 << i);
1124                 dlen = SA_SIZE(sa);
1125 #ifdef INET6
1126                 if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1127                         sin6 = (struct sockaddr_in6 *)&ss;
1128                         bcopy(sa, sin6, sizeof(*sin6));
1129                         if (sa6_recoverscope(sin6) == 0)
1130                                 sa = (struct sockaddr *)sin6;
1131                 }
1132 #endif
1133                 m_copyback(m, len, dlen, (caddr_t)sa);
1134                 len += dlen;
1135         }
1136         if (m->m_pkthdr.len != len) {
1137                 m_freem(m);
1138                 return (NULL);
1139         }
1140         rtm->rtm_msglen = len;
1141         rtm->rtm_version = RTM_VERSION;
1142         rtm->rtm_type = type;
1143         return (m);
1144 }
1145
1146 /*
1147  * Used by the sysctl code and routing socket.
1148  */
1149 static int
1150 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
1151 {
1152         int i;
1153         int len, dlen, second_time = 0;
1154         caddr_t cp0;
1155 #ifdef INET6
1156         struct sockaddr_storage ss;
1157         struct sockaddr_in6 *sin6;
1158 #endif
1159
1160         rtinfo->rti_addrs = 0;
1161 again:
1162         switch (type) {
1163
1164         case RTM_DELADDR:
1165         case RTM_NEWADDR:
1166                 if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1167 #ifdef COMPAT_FREEBSD32
1168                         if (w->w_req->flags & SCTL_MASK32)
1169                                 len = sizeof(struct ifa_msghdrl32);
1170                         else
1171 #endif
1172                                 len = sizeof(struct ifa_msghdrl);
1173                 } else
1174                         len = sizeof(struct ifa_msghdr);
1175                 break;
1176
1177         case RTM_IFINFO:
1178 #ifdef COMPAT_FREEBSD32
1179                 if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1180                         if (w->w_op == NET_RT_IFLISTL)
1181                                 len = sizeof(struct if_msghdrl32);
1182                         else
1183                                 len = sizeof(struct if_msghdr32);
1184                         break;
1185                 }
1186 #endif
1187                 if (w != NULL && w->w_op == NET_RT_IFLISTL)
1188                         len = sizeof(struct if_msghdrl);
1189                 else
1190                         len = sizeof(struct if_msghdr);
1191                 break;
1192
1193         case RTM_NEWMADDR:
1194                 len = sizeof(struct ifma_msghdr);
1195                 break;
1196
1197         default:
1198                 len = sizeof(struct rt_msghdr);
1199         }
1200         cp0 = cp;
1201         if (cp0)
1202                 cp += len;
1203         for (i = 0; i < RTAX_MAX; i++) {
1204                 struct sockaddr *sa;
1205
1206                 if ((sa = rtinfo->rti_info[i]) == NULL)
1207                         continue;
1208                 rtinfo->rti_addrs |= (1 << i);
1209                 dlen = SA_SIZE(sa);
1210                 if (cp) {
1211 #ifdef INET6
1212                         if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1213                                 sin6 = (struct sockaddr_in6 *)&ss;
1214                                 bcopy(sa, sin6, sizeof(*sin6));
1215                                 if (sa6_recoverscope(sin6) == 0)
1216                                         sa = (struct sockaddr *)sin6;
1217                         }
1218 #endif
1219                         bcopy((caddr_t)sa, cp, (unsigned)dlen);
1220                         cp += dlen;
1221                 }
1222                 len += dlen;
1223         }
1224         len = ALIGN(len);
1225         if (cp == NULL && w != NULL && !second_time) {
1226                 struct walkarg *rw = w;
1227
1228                 if (rw->w_req) {
1229                         if (rw->w_tmemsize < len) {
1230                                 if (rw->w_tmem)
1231                                         free(rw->w_tmem, M_RTABLE);
1232                                 rw->w_tmem = (caddr_t)
1233                                         malloc(len, M_RTABLE, M_NOWAIT);
1234                                 if (rw->w_tmem)
1235                                         rw->w_tmemsize = len;
1236                         }
1237                         if (rw->w_tmem) {
1238                                 cp = rw->w_tmem;
1239                                 second_time = 1;
1240                                 goto again;
1241                         }
1242                 }
1243         }
1244         if (cp) {
1245                 struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
1246
1247                 rtm->rtm_version = RTM_VERSION;
1248                 rtm->rtm_type = type;
1249                 rtm->rtm_msglen = len;
1250         }
1251         return (len);
1252 }
1253
1254 /*
1255  * This routine is called to generate a message from the routing
1256  * socket indicating that a redirect has occured, a routing lookup
1257  * has failed, or that a protocol has detected timeouts to a particular
1258  * destination.
1259  */
1260 void
1261 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1262     int fibnum)
1263 {
1264         struct rt_msghdr *rtm;
1265         struct mbuf *m;
1266         struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1267
1268         if (route_cb.any_count == 0)
1269                 return;
1270         m = rt_msg1(type, rtinfo);
1271         if (m == NULL)
1272                 return;
1273
1274         if (fibnum != RT_ALL_FIBS) {
1275                 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1276                     "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1277                 M_SETFIB(m, fibnum);
1278                 m->m_flags |= RTS_FILTER_FIB;
1279         }
1280
1281         rtm = mtod(m, struct rt_msghdr *);
1282         rtm->rtm_flags = RTF_DONE | flags;
1283         rtm->rtm_errno = error;
1284         rtm->rtm_addrs = rtinfo->rti_addrs;
1285         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1286 }
1287
1288 void
1289 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1290 {
1291
1292         rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1293 }
1294
1295 /*
1296  * This routine is called to generate a message from the routing
1297  * socket indicating that the status of a network interface has changed.
1298  */
1299 void
1300 rt_ifmsg(struct ifnet *ifp)
1301 {
1302         struct if_msghdr *ifm;
1303         struct mbuf *m;
1304         struct rt_addrinfo info;
1305
1306         if (route_cb.any_count == 0)
1307                 return;
1308         bzero((caddr_t)&info, sizeof(info));
1309         m = rt_msg1(RTM_IFINFO, &info);
1310         if (m == NULL)
1311                 return;
1312         ifm = mtod(m, struct if_msghdr *);
1313         ifm->ifm_index = ifp->if_index;
1314         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1315         ifm->ifm_data = ifp->if_data;
1316         ifm->ifm_addrs = 0;
1317         rt_dispatch(m, AF_UNSPEC);
1318 }
1319
1320 /*
1321  * Announce interface address arrival/withdraw.
1322  * Please do not call directly, use rt_addrmsg().
1323  * Assume input data to be valid.
1324  * Returns 0 on success.
1325  */
1326 int
1327 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1328 {
1329         struct rt_addrinfo info;
1330         struct sockaddr *sa;
1331         int ncmd;
1332         struct mbuf *m;
1333         struct ifa_msghdr *ifam;
1334         struct ifnet *ifp = ifa->ifa_ifp;
1335
1336         if (route_cb.any_count == 0)
1337                 return (0);
1338
1339         ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1340
1341         bzero((caddr_t)&info, sizeof(info));
1342         info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1343         info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1344         info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1345         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1346         if ((m = rt_msg1(ncmd, &info)) == NULL)
1347                 return (ENOBUFS);
1348         ifam = mtod(m, struct ifa_msghdr *);
1349         ifam->ifam_index = ifp->if_index;
1350         ifam->ifam_metric = ifa->ifa_metric;
1351         ifam->ifam_flags = ifa->ifa_flags;
1352         ifam->ifam_addrs = info.rti_addrs;
1353
1354         if (fibnum != RT_ALL_FIBS) {
1355                 M_SETFIB(m, fibnum);
1356                 m->m_flags |= RTS_FILTER_FIB;
1357         }
1358
1359         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1360
1361         return (0);
1362 }
1363
1364 /*
1365  * Announce route addition/removal.
1366  * Please do not call directly, use rt_routemsg().
1367  * Note that @rt data MAY be inconsistent/invalid:
1368  * if some userland app sends us "invalid" route message (invalid mask,
1369  * no dst, wrong address families, etc...) we need to pass it back
1370  * to app (and any other rtsock consumers) with rtm_errno field set to
1371  * non-zero value.
1372  *
1373  * Returns 0 on success.
1374  */
1375 int
1376 rtsock_routemsg(int cmd, struct ifnet *ifp, int error, struct rtentry *rt,
1377     int fibnum)
1378 {
1379         struct rt_addrinfo info;
1380         struct sockaddr *sa;
1381         struct mbuf *m;
1382         struct rt_msghdr *rtm;
1383
1384         if (route_cb.any_count == 0)
1385                 return (0);
1386
1387         bzero((caddr_t)&info, sizeof(info));
1388         info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1389         info.rti_info[RTAX_DST] = sa = rt_key(rt);
1390         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1391         if ((m = rt_msg1(cmd, &info)) == NULL)
1392                 return (ENOBUFS);
1393         rtm = mtod(m, struct rt_msghdr *);
1394         rtm->rtm_index = ifp->if_index;
1395         rtm->rtm_flags |= rt->rt_flags;
1396         rtm->rtm_errno = error;
1397         rtm->rtm_addrs = info.rti_addrs;
1398
1399         if (fibnum != RT_ALL_FIBS) {
1400                 M_SETFIB(m, fibnum);
1401                 m->m_flags |= RTS_FILTER_FIB;
1402         }
1403
1404         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1405
1406         return (0);
1407 }
1408
1409 /*
1410  * This is the analogue to the rt_newaddrmsg which performs the same
1411  * function but for multicast group memberhips.  This is easier since
1412  * there is no route state to worry about.
1413  */
1414 void
1415 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1416 {
1417         struct rt_addrinfo info;
1418         struct mbuf *m = NULL;
1419         struct ifnet *ifp = ifma->ifma_ifp;
1420         struct ifma_msghdr *ifmam;
1421
1422         if (route_cb.any_count == 0)
1423                 return;
1424
1425         bzero((caddr_t)&info, sizeof(info));
1426         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1427         info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1428         /*
1429          * If a link-layer address is present, present it as a ``gateway''
1430          * (similarly to how ARP entries, e.g., are presented).
1431          */
1432         info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1433         m = rt_msg1(cmd, &info);
1434         if (m == NULL)
1435                 return;
1436         ifmam = mtod(m, struct ifma_msghdr *);
1437         KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1438             __func__));
1439         ifmam->ifmam_index = ifp->if_index;
1440         ifmam->ifmam_addrs = info.rti_addrs;
1441         rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1442 }
1443
1444 static struct mbuf *
1445 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1446         struct rt_addrinfo *info)
1447 {
1448         struct if_announcemsghdr *ifan;
1449         struct mbuf *m;
1450
1451         if (route_cb.any_count == 0)
1452                 return NULL;
1453         bzero((caddr_t)info, sizeof(*info));
1454         m = rt_msg1(type, info);
1455         if (m != NULL) {
1456                 ifan = mtod(m, struct if_announcemsghdr *);
1457                 ifan->ifan_index = ifp->if_index;
1458                 strlcpy(ifan->ifan_name, ifp->if_xname,
1459                         sizeof(ifan->ifan_name));
1460                 ifan->ifan_what = what;
1461         }
1462         return m;
1463 }
1464
1465 /*
1466  * This is called to generate routing socket messages indicating
1467  * IEEE80211 wireless events.
1468  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1469  */
1470 void
1471 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1472 {
1473         struct mbuf *m;
1474         struct rt_addrinfo info;
1475
1476         m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1477         if (m != NULL) {
1478                 /*
1479                  * Append the ieee80211 data.  Try to stick it in the
1480                  * mbuf containing the ifannounce msg; otherwise allocate
1481                  * a new mbuf and append.
1482                  *
1483                  * NB: we assume m is a single mbuf.
1484                  */
1485                 if (data_len > M_TRAILINGSPACE(m)) {
1486                         struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1487                         if (n == NULL) {
1488                                 m_freem(m);
1489                                 return;
1490                         }
1491                         bcopy(data, mtod(n, void *), data_len);
1492                         n->m_len = data_len;
1493                         m->m_next = n;
1494                 } else if (data_len > 0) {
1495                         bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1496                         m->m_len += data_len;
1497                 }
1498                 if (m->m_flags & M_PKTHDR)
1499                         m->m_pkthdr.len += data_len;
1500                 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1501                 rt_dispatch(m, AF_UNSPEC);
1502         }
1503 }
1504
1505 /*
1506  * This is called to generate routing socket messages indicating
1507  * network interface arrival and departure.
1508  */
1509 void
1510 rt_ifannouncemsg(struct ifnet *ifp, int what)
1511 {
1512         struct mbuf *m;
1513         struct rt_addrinfo info;
1514
1515         m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1516         if (m != NULL)
1517                 rt_dispatch(m, AF_UNSPEC);
1518 }
1519
1520 static void
1521 rt_dispatch(struct mbuf *m, sa_family_t saf)
1522 {
1523         struct m_tag *tag;
1524
1525         /*
1526          * Preserve the family from the sockaddr, if any, in an m_tag for
1527          * use when injecting the mbuf into the routing socket buffer from
1528          * the netisr.
1529          */
1530         if (saf != AF_UNSPEC) {
1531                 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1532                     M_NOWAIT);
1533                 if (tag == NULL) {
1534                         m_freem(m);
1535                         return;
1536                 }
1537                 *(unsigned short *)(tag + 1) = saf;
1538                 m_tag_prepend(m, tag);
1539         }
1540 #ifdef VIMAGE
1541         if (V_loif)
1542                 m->m_pkthdr.rcvif = V_loif;
1543         else {
1544                 m_freem(m);
1545                 return;
1546         }
1547 #endif
1548         netisr_queue(NETISR_ROUTE, m);  /* mbuf is free'd on failure. */
1549 }
1550
1551 /*
1552  * This is used in dumping the kernel table via sysctl().
1553  */
1554 static int
1555 sysctl_dumpentry(struct radix_node *rn, void *vw)
1556 {
1557         struct walkarg *w = vw;
1558         struct rtentry *rt = (struct rtentry *)rn;
1559         int error = 0, size;
1560         struct rt_addrinfo info;
1561
1562         if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1563                 return 0;
1564         if ((rt->rt_flags & RTF_HOST) == 0
1565             ? jailed_without_vnet(w->w_req->td->td_ucred)
1566             : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1567                 return (0);
1568         bzero((caddr_t)&info, sizeof(info));
1569         info.rti_info[RTAX_DST] = rt_key(rt);
1570         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1571         info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1572         info.rti_info[RTAX_GENMASK] = 0;
1573         if (rt->rt_ifp) {
1574                 info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1575                 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1576                 if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1577                         info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1578         }
1579         size = rt_msg2(RTM_GET, &info, NULL, w);
1580         if (w->w_req && w->w_tmem) {
1581                 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1582
1583                 if (rt->rt_flags & RTF_GWFLAG_COMPAT)
1584                         rtm->rtm_flags = RTF_GATEWAY |
1585                                 (rt->rt_flags & ~RTF_GWFLAG_COMPAT);
1586                 else
1587                         rtm->rtm_flags = rt->rt_flags;
1588                 rt_getmetrics(rt, &rtm->rtm_rmx);
1589                 rtm->rtm_index = rt->rt_ifp->if_index;
1590                 rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1591                 rtm->rtm_addrs = info.rti_addrs;
1592                 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1593                 return (error);
1594         }
1595         return (error);
1596 }
1597
1598 #ifdef COMPAT_FREEBSD32
1599 static void
1600 copy_ifdata32(struct if_data *src, struct if_data32 *dst)
1601 {
1602
1603         bzero(dst, sizeof(*dst));
1604         CP(*src, *dst, ifi_type);
1605         CP(*src, *dst, ifi_physical);
1606         CP(*src, *dst, ifi_addrlen);
1607         CP(*src, *dst, ifi_hdrlen);
1608         CP(*src, *dst, ifi_link_state);
1609         CP(*src, *dst, ifi_vhid);
1610         CP(*src, *dst, ifi_baudrate_pf);
1611         dst->ifi_datalen = sizeof(struct if_data32);
1612         CP(*src, *dst, ifi_mtu);
1613         CP(*src, *dst, ifi_metric);
1614         CP(*src, *dst, ifi_baudrate);
1615         CP(*src, *dst, ifi_ipackets);
1616         CP(*src, *dst, ifi_ierrors);
1617         CP(*src, *dst, ifi_opackets);
1618         CP(*src, *dst, ifi_oerrors);
1619         CP(*src, *dst, ifi_collisions);
1620         CP(*src, *dst, ifi_ibytes);
1621         CP(*src, *dst, ifi_obytes);
1622         CP(*src, *dst, ifi_imcasts);
1623         CP(*src, *dst, ifi_omcasts);
1624         CP(*src, *dst, ifi_iqdrops);
1625         CP(*src, *dst, ifi_noproto);
1626         CP(*src, *dst, ifi_hwassist);
1627         CP(*src, *dst, ifi_epoch);
1628         TV_CP(*src, *dst, ifi_lastchange);
1629 }
1630 #endif
1631
1632 static int
1633 sysctl_iflist_ifml(struct ifnet *ifp, struct rt_addrinfo *info,
1634     struct walkarg *w, int len)
1635 {
1636         struct if_msghdrl *ifm;
1637
1638 #ifdef COMPAT_FREEBSD32
1639         if (w->w_req->flags & SCTL_MASK32) {
1640                 struct if_msghdrl32 *ifm32;
1641
1642                 ifm32 = (struct if_msghdrl32 *)w->w_tmem;
1643                 ifm32->ifm_addrs = info->rti_addrs;
1644                 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1645                 ifm32->ifm_index = ifp->if_index;
1646                 ifm32->_ifm_spare1 = 0;
1647                 ifm32->ifm_len = sizeof(*ifm32);
1648                 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1649
1650                 copy_ifdata32(&ifp->if_data, &ifm32->ifm_data);
1651                 /* Fixup if_data carp(4) vhid. */
1652                 if (carp_get_vhid_p != NULL)
1653                         ifm32->ifm_data.ifi_vhid =
1654                             (*carp_get_vhid_p)(ifp->if_addr);
1655                 ifm32->ifm_data.ifi_oqdrops = ifp->if_snd.ifq_drops;
1656
1657                 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm32, len));
1658         }
1659 #endif
1660         ifm = (struct if_msghdrl *)w->w_tmem;
1661         ifm->ifm_addrs = info->rti_addrs;
1662         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1663         ifm->ifm_index = ifp->if_index;
1664         ifm->_ifm_spare1 = 0;
1665         ifm->ifm_len = sizeof(*ifm);
1666         ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1667
1668         ifm->ifm_data = ifp->if_data;
1669         /* Fixup if_data carp(4) vhid. */
1670         if (carp_get_vhid_p != NULL)
1671                 ifm->ifm_data.ifi_vhid = (*carp_get_vhid_p)(ifp->if_addr);
1672
1673         ifm->ifm_data.ifi_datalen += sizeof(u_long);
1674         ifm->ifi_oqdrops = ifp->if_snd.ifq_drops;
1675
1676         return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1677 }
1678
1679 static int
1680 sysctl_iflist_ifm(struct ifnet *ifp, struct rt_addrinfo *info,
1681     struct walkarg *w, int len)
1682 {
1683         struct if_msghdr *ifm;
1684
1685 #ifdef COMPAT_FREEBSD32
1686         if (w->w_req->flags & SCTL_MASK32) {
1687                 struct if_msghdr32 *ifm32;
1688
1689                 ifm32 = (struct if_msghdr32 *)w->w_tmem;
1690                 ifm32->ifm_addrs = info->rti_addrs;
1691                 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1692                 ifm32->ifm_index = ifp->if_index;
1693
1694                 copy_ifdata32(&ifp->if_data, &ifm32->ifm_data);
1695                 /* Fixup if_data carp(4) vhid. */
1696                 if (carp_get_vhid_p != NULL)
1697                         ifm32->ifm_data.ifi_vhid =
1698                             (*carp_get_vhid_p)(ifp->if_addr);
1699
1700                 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm32, len));
1701         }
1702 #endif
1703         ifm = (struct if_msghdr *)w->w_tmem;
1704         ifm->ifm_addrs = info->rti_addrs;
1705         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1706         ifm->ifm_index = ifp->if_index;
1707
1708         ifm->ifm_data = ifp->if_data;
1709         /* Fixup if_data carp(4) vhid. */
1710         if (carp_get_vhid_p != NULL)
1711                 ifm->ifm_data.ifi_vhid = (*carp_get_vhid_p)(ifp->if_addr);
1712
1713         return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1714 }
1715
1716 static int
1717 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1718     struct walkarg *w, int len)
1719 {
1720         struct ifa_msghdrl *ifam;
1721
1722 #ifdef COMPAT_FREEBSD32
1723         if (w->w_req->flags & SCTL_MASK32) {
1724                 struct ifa_msghdrl32 *ifam32;
1725
1726                 ifam32 = (struct ifa_msghdrl32 *)w->w_tmem;
1727                 ifam32->ifam_addrs = info->rti_addrs;
1728                 ifam32->ifam_flags = ifa->ifa_flags;
1729                 ifam32->ifam_index = ifa->ifa_ifp->if_index;
1730                 ifam32->_ifam_spare1 = 0;
1731                 ifam32->ifam_len = sizeof(*ifam32);
1732                 ifam32->ifam_data_off =
1733                     offsetof(struct ifa_msghdrl32, ifam_data);
1734                 ifam32->ifam_metric = ifa->ifa_metric;
1735
1736                 copy_ifdata32(&ifa->ifa_ifp->if_data, &ifam32->ifam_data);
1737                 /* Fixup if_data carp(4) vhid. */
1738                 if (carp_get_vhid_p != NULL)
1739                         ifam32->ifam_data.ifi_vhid = (*carp_get_vhid_p)(ifa);
1740
1741                 return (SYSCTL_OUT(w->w_req, (caddr_t)ifam32, len));
1742         }
1743 #endif
1744
1745         ifam = (struct ifa_msghdrl *)w->w_tmem;
1746         ifam->ifam_addrs = info->rti_addrs;
1747         ifam->ifam_flags = ifa->ifa_flags;
1748         ifam->ifam_index = ifa->ifa_ifp->if_index;
1749         ifam->_ifam_spare1 = 0;
1750         ifam->ifam_len = sizeof(*ifam);
1751         ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1752         ifam->ifam_metric = ifa->ifa_metric;
1753
1754         ifam->ifam_data = ifa->if_data;
1755         /* Fixup if_data carp(4) vhid. */
1756         if (carp_get_vhid_p != NULL)
1757                 ifam->ifam_data.ifi_vhid = (*carp_get_vhid_p)(ifa);
1758
1759         return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1760 }
1761
1762 static int
1763 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1764     struct walkarg *w, int len)
1765 {
1766         struct ifa_msghdr *ifam;
1767
1768         ifam = (struct ifa_msghdr *)w->w_tmem;
1769         ifam->ifam_addrs = info->rti_addrs;
1770         ifam->ifam_flags = ifa->ifa_flags;
1771         ifam->ifam_index = ifa->ifa_ifp->if_index;
1772         ifam->ifam_metric = ifa->ifa_metric;
1773
1774         return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1775 }
1776
1777 static int
1778 sysctl_iflist(int af, struct walkarg *w)
1779 {
1780         struct ifnet *ifp;
1781         struct ifaddr *ifa;
1782         struct rt_addrinfo info;
1783         int len, error = 0;
1784
1785         bzero((caddr_t)&info, sizeof(info));
1786         IFNET_RLOCK_NOSLEEP();
1787         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1788                 if (w->w_arg && w->w_arg != ifp->if_index)
1789                         continue;
1790                 IF_ADDR_RLOCK(ifp);
1791                 ifa = ifp->if_addr;
1792                 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1793                 len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1794                 info.rti_info[RTAX_IFP] = NULL;
1795                 if (w->w_req && w->w_tmem) {
1796                         if (w->w_op == NET_RT_IFLISTL)
1797                                 error = sysctl_iflist_ifml(ifp, &info, w, len);
1798                         else
1799                                 error = sysctl_iflist_ifm(ifp, &info, w, len);
1800                         if (error)
1801                                 goto done;
1802                 }
1803                 while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1804                         if (af && af != ifa->ifa_addr->sa_family)
1805                                 continue;
1806                         if (prison_if(w->w_req->td->td_ucred,
1807                             ifa->ifa_addr) != 0)
1808                                 continue;
1809                         info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1810                         info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1811                         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1812                         len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1813                         if (w->w_req && w->w_tmem) {
1814                                 if (w->w_op == NET_RT_IFLISTL)
1815                                         error = sysctl_iflist_ifaml(ifa, &info,
1816                                             w, len);
1817                                 else
1818                                         error = sysctl_iflist_ifam(ifa, &info,
1819                                             w, len);
1820                                 if (error)
1821                                         goto done;
1822                         }
1823                 }
1824                 IF_ADDR_RUNLOCK(ifp);
1825                 info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1826                         info.rti_info[RTAX_BRD] = NULL;
1827         }
1828 done:
1829         if (ifp != NULL)
1830                 IF_ADDR_RUNLOCK(ifp);
1831         IFNET_RUNLOCK_NOSLEEP();
1832         return (error);
1833 }
1834
1835 static int
1836 sysctl_ifmalist(int af, struct walkarg *w)
1837 {
1838         struct ifnet *ifp;
1839         struct ifmultiaddr *ifma;
1840         struct  rt_addrinfo info;
1841         int     len, error = 0;
1842         struct ifaddr *ifa;
1843
1844         bzero((caddr_t)&info, sizeof(info));
1845         IFNET_RLOCK_NOSLEEP();
1846         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1847                 if (w->w_arg && w->w_arg != ifp->if_index)
1848                         continue;
1849                 ifa = ifp->if_addr;
1850                 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1851                 IF_ADDR_RLOCK(ifp);
1852                 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1853                         if (af && af != ifma->ifma_addr->sa_family)
1854                                 continue;
1855                         if (prison_if(w->w_req->td->td_ucred,
1856                             ifma->ifma_addr) != 0)
1857                                 continue;
1858                         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1859                         info.rti_info[RTAX_GATEWAY] =
1860                             (ifma->ifma_addr->sa_family != AF_LINK) ?
1861                             ifma->ifma_lladdr : NULL;
1862                         len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1863                         if (w->w_req && w->w_tmem) {
1864                                 struct ifma_msghdr *ifmam;
1865
1866                                 ifmam = (struct ifma_msghdr *)w->w_tmem;
1867                                 ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1868                                 ifmam->ifmam_flags = 0;
1869                                 ifmam->ifmam_addrs = info.rti_addrs;
1870                                 error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1871                                 if (error) {
1872                                         IF_ADDR_RUNLOCK(ifp);
1873                                         goto done;
1874                                 }
1875                         }
1876                 }
1877                 IF_ADDR_RUNLOCK(ifp);
1878         }
1879 done:
1880         IFNET_RUNLOCK_NOSLEEP();
1881         return (error);
1882 }
1883
1884 static int
1885 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1886 {
1887         int     *name = (int *)arg1;
1888         u_int   namelen = arg2;
1889         struct radix_node_head *rnh = NULL; /* silence compiler. */
1890         int     i, lim, error = EINVAL;
1891         int     fib = 0;
1892         u_char  af;
1893         struct  walkarg w;
1894
1895         name ++;
1896         namelen--;
1897         if (req->newptr)
1898                 return (EPERM);
1899         if (name[1] == NET_RT_DUMP) {
1900                 if (namelen == 3)
1901                         fib = req->td->td_proc->p_fibnum;
1902                 else if (namelen == 4)
1903                         fib = (name[3] == RT_ALL_FIBS) ?
1904                             req->td->td_proc->p_fibnum : name[3];
1905                 else
1906                         return ((namelen < 3) ? EISDIR : ENOTDIR);
1907                 if (fib < 0 || fib >= rt_numfibs)
1908                         return (EINVAL);
1909         } else if (namelen != 3)
1910                 return ((namelen < 3) ? EISDIR : ENOTDIR);
1911         af = name[0];
1912         if (af > AF_MAX)
1913                 return (EINVAL);
1914         bzero(&w, sizeof(w));
1915         w.w_op = name[1];
1916         w.w_arg = name[2];
1917         w.w_req = req;
1918
1919         error = sysctl_wire_old_buffer(req, 0);
1920         if (error)
1921                 return (error);
1922         switch (w.w_op) {
1923
1924         case NET_RT_DUMP:
1925         case NET_RT_FLAGS:
1926                 if (af == 0) {                  /* dump all tables */
1927                         i = 1;
1928                         lim = AF_MAX;
1929                 } else                          /* dump only one table */
1930                         i = lim = af;
1931
1932                 /*
1933                  * take care of llinfo entries, the caller must
1934                  * specify an AF
1935                  */
1936                 if (w.w_op == NET_RT_FLAGS &&
1937                     (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1938                         if (af != 0)
1939                                 error = lltable_sysctl_dumparp(af, w.w_req);
1940                         else
1941                                 error = EINVAL;
1942                         break;
1943                 }
1944                 /*
1945                  * take care of routing entries
1946                  */
1947                 for (error = 0; error == 0 && i <= lim; i++) {
1948                         rnh = rt_tables_get_rnh(fib, i);
1949                         if (rnh != NULL) {
1950                                 RADIX_NODE_HEAD_RLOCK(rnh);
1951                                 error = rnh->rnh_walktree(rnh,
1952                                     sysctl_dumpentry, &w);
1953                                 RADIX_NODE_HEAD_RUNLOCK(rnh);
1954                         } else if (af != 0)
1955                                 error = EAFNOSUPPORT;
1956                 }
1957                 break;
1958
1959         case NET_RT_IFLIST:
1960         case NET_RT_IFLISTL:
1961                 error = sysctl_iflist(af, &w);
1962                 break;
1963
1964         case NET_RT_IFMALIST:
1965                 error = sysctl_ifmalist(af, &w);
1966                 break;
1967         }
1968         if (w.w_tmem)
1969                 free(w.w_tmem, M_RTABLE);
1970         return (error);
1971 }
1972
1973 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1974
1975 /*
1976  * Definitions of protocols supported in the ROUTE domain.
1977  */
1978
1979 static struct domain routedomain;               /* or at least forward */
1980
1981 static struct protosw routesw[] = {
1982 {
1983         .pr_type =              SOCK_RAW,
1984         .pr_domain =            &routedomain,
1985         .pr_flags =             PR_ATOMIC|PR_ADDR,
1986         .pr_output =            route_output,
1987         .pr_ctlinput =          raw_ctlinput,
1988         .pr_init =              raw_init,
1989         .pr_usrreqs =           &route_usrreqs
1990 }
1991 };
1992
1993 static struct domain routedomain = {
1994         .dom_family =           PF_ROUTE,
1995         .dom_name =              "route",
1996         .dom_protosw =          routesw,
1997         .dom_protoswNPROTOSW =  &routesw[sizeof(routesw)/sizeof(routesw[0])]
1998 };
1999
2000 VNET_DOMAIN_SET(route);