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         if (rp == NULL)
 305                 return ENOBUFS;
 306
 307         so->so_pcb = (caddr_t)rp;
 308         so->so_fibnum = td->td_proc->p_fibnum;
 309         error = raw_attach(so, proto);
 310         rp = sotorawcb(so);
 311         if (error) {
 312                 so->so_pcb = NULL;
 313                 free(rp, M_PCB);
 314                 return error;
 315         }
 316         RTSOCK_LOCK();
 317         switch(rp->rcb_proto.sp_protocol) {
 318         case AF_INET:
 319                 route_cb.ip_count++;
 320                 break;
 321         case AF_INET6:
 322                 route_cb.ip6_count++;
 323                 break;
 324         case AF_IPX:
 325                 route_cb.ipx_count++;
 326                 break;
 327         }
 328         route_cb.any_count++;
 329         RTSOCK_UNLOCK();
 330         soisconnected(so);
 331         so->so_options |= SO_USELOOPBACK;
 332         return 0;
 333 }
 334
 335 static int
 336 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
 337 {
 338
 339         return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
 340 }
 341
 342 static int
 343 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
 344 {
 345
 346         return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
 347 }
 348
 349 /* pru_connect2 is EOPNOTSUPP */
 350 /* pru_control is EOPNOTSUPP */
 351
 352 static void
 353 rts_detach(struct socket *so)
 354 {
 355         struct rawcb *rp = sotorawcb(so);
 356
 357         KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
 358
 359         RTSOCK_LOCK();
 360         switch(rp->rcb_proto.sp_protocol) {
 361         case AF_INET:
 362                 route_cb.ip_count--;
 363                 break;
 364         case AF_INET6:
 365                 route_cb.ip6_count--;
 366                 break;
 367         case AF_IPX:
 368                 route_cb.ipx_count--;
 369                 break;
 370         }
 371         route_cb.any_count--;
 372         RTSOCK_UNLOCK();
 373         raw_usrreqs.pru_detach(so);
 374 }
 375
 376 static int
 377 rts_disconnect(struct socket *so)
 378 {
 379
 380         return (raw_usrreqs.pru_disconnect(so));
 381 }
 382
 383 /* pru_listen is EOPNOTSUPP */
 384
 385 static int
 386 rts_peeraddr(struct socket *so, struct sockaddr **nam)
 387 {
 388
 389         return (raw_usrreqs.pru_peeraddr(so, nam));
 390 }
 391
 392 /* pru_rcvd is EOPNOTSUPP */
 393 /* pru_rcvoob is EOPNOTSUPP */
 394
 395 static int
 396 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
 397          struct mbuf *control, struct thread *td)
 398 {
 399
 400         return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
 401 }
 402
 403 /* pru_sense is null */
 404
 405 static int
 406 rts_shutdown(struct socket *so)
 407 {
 408
 409         return (raw_usrreqs.pru_shutdown(so));
 410 }
 411
 412 static int
 413 rts_sockaddr(struct socket *so, struct sockaddr **nam)
 414 {
 415
 416         return (raw_usrreqs.pru_sockaddr(so, nam));
 417 }
 418
 419 static struct pr_usrreqs route_usrreqs = {
 420         .pru_abort =            rts_abort,
 421         .pru_attach =           rts_attach,
 422         .pru_bind =             rts_bind,
 423         .pru_connect =          rts_connect,
 424         .pru_detach =           rts_detach,
 425         .pru_disconnect =       rts_disconnect,
 426         .pru_peeraddr =         rts_peeraddr,
 427         .pru_send =             rts_send,
 428         .pru_shutdown =         rts_shutdown,
 429         .pru_sockaddr =         rts_sockaddr,
 430         .pru_close =            rts_close,
 431 };
 432
 433 #ifndef _SOCKADDR_UNION_DEFINED
 434 #define _SOCKADDR_UNION_DEFINED
 435 /*
 436  * The union of all possible address formats we handle.
 437  */
 438 union sockaddr_union {
 439         struct sockaddr         sa;
 440         struct sockaddr_in      sin;
 441         struct sockaddr_in6     sin6;
 442 };
 443 #endif /* _SOCKADDR_UNION_DEFINED */
 444
 445 static int
 446 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
 447     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
 448 {
 449
 450         /* First, see if the returned address is part of the jail. */
 451         if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
 452                 info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
 453                 return (0);
 454         }
 455
 456         switch (info->rti_info[RTAX_DST]->sa_family) {
 457 #ifdef INET
 458         case AF_INET:
 459         {
 460                 struct in_addr ia;
 461                 struct ifaddr *ifa;
 462                 int found;
 463
 464                 found = 0;
 465                 /*
 466                  * Try to find an address on the given outgoing interface
 467                  * that belongs to the jail.
 468                  */
 469                 IF_ADDR_RLOCK(ifp);
 470                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 471                         struct sockaddr *sa;
 472                         sa = ifa->ifa_addr;
 473                         if (sa->sa_family != AF_INET)
 474                                 continue;
 475                         ia = ((struct sockaddr_in *)sa)->sin_addr;
 476                         if (prison_check_ip4(cred, &ia) == 0) {
 477                                 found = 1;
 478                                 break;
 479                         }
 480                 }
 481                 IF_ADDR_RUNLOCK(ifp);
 482                 if (!found) {
 483                         /*
 484                          * As a last resort return the 'default' jail address.
 485                          */
 486                         ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
 487                             sin_addr;
 488                         if (prison_get_ip4(cred, &ia) != 0)
 489                                 return (ESRCH);
 490                 }
 491                 bzero(&saun->sin, sizeof(struct sockaddr_in));
 492                 saun->sin.sin_len = sizeof(struct sockaddr_in);
 493                 saun->sin.sin_family = AF_INET;
 494                 saun->sin.sin_addr.s_addr = ia.s_addr;
 495                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
 496                 break;
 497         }
 498 #endif
 499 #ifdef INET6
 500         case AF_INET6:
 501         {
 502                 struct in6_addr ia6;
 503                 struct ifaddr *ifa;
 504                 int found;
 505
 506                 found = 0;
 507                 /*
 508                  * Try to find an address on the given outgoing interface
 509                  * that belongs to the jail.
 510                  */
 511                 IF_ADDR_RLOCK(ifp);
 512                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 513                         struct sockaddr *sa;
 514                         sa = ifa->ifa_addr;
 515                         if (sa->sa_family != AF_INET6)
 516                                 continue;
 517                         bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
 518                             &ia6, sizeof(struct in6_addr));
 519                         if (prison_check_ip6(cred, &ia6) == 0) {
 520                                 found = 1;
 521                                 break;
 522                         }
 523                 }
 524                 IF_ADDR_RUNLOCK(ifp);
 525                 if (!found) {
 526                         /*
 527                          * As a last resort return the 'default' jail address.
 528                          */
 529                         ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
 530                             sin6_addr;
 531                         if (prison_get_ip6(cred, &ia6) != 0)
 532                                 return (ESRCH);
 533                 }
 534                 bzero(&saun->sin6, sizeof(struct sockaddr_in6));
 535                 saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
 536                 saun->sin6.sin6_family = AF_INET6;
 537                 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
 538                 if (sa6_recoverscope(&saun->sin6) != 0)
 539                         return (ESRCH);
 540                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
 541                 break;
 542         }
 543 #endif
 544         default:
 545                 return (ESRCH);
 546         }
 547         return (0);
 548 }
 549
 550 /*ARGSUSED*/
 551 static int
 552 route_output(struct mbuf *m, struct socket *so)
 553 {
 554 #define sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
 555         struct rt_msghdr *rtm = NULL;
 556         struct rtentry *rt = NULL;
 557         struct radix_node_head *rnh;
 558         struct rt_addrinfo info;
 559 #ifdef INET6
 560         struct sockaddr_storage ss;
 561         struct sockaddr_in6 *sin6;
 562         int i, rti_need_deembed = 0;
 563 #endif
 564         int len, error = 0;
 565         struct ifnet *ifp = NULL;
 566         union sockaddr_union saun;
 567         sa_family_t saf = AF_UNSPEC;
 568
 569 #define senderr(e) { error = e; goto flush;}
 570         if (m == NULL || ((m->m_len < sizeof(long)) &&
 571                        (m = m_pullup(m, sizeof(long))) == NULL))
 572                 return (ENOBUFS);
 573         if ((m->m_flags & M_PKTHDR) == 0)
 574                 panic("route_output");
 575         len = m->m_pkthdr.len;
 576         if (len < sizeof(*rtm) ||
 577             len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
 578                 info.rti_info[RTAX_DST] = NULL;
 579                 senderr(EINVAL);
 580         }
 581         R_Malloc(rtm, struct rt_msghdr *, len);
 582         if (rtm == NULL) {
 583                 info.rti_info[RTAX_DST] = NULL;
 584                 senderr(ENOBUFS);
 585         }
 586         m_copydata(m, 0, len, (caddr_t)rtm);
 587         if (rtm->rtm_version != RTM_VERSION) {
 588                 info.rti_info[RTAX_DST] = NULL;
 589                 senderr(EPROTONOSUPPORT);
 590         }
 591         rtm->rtm_pid = curproc->p_pid;
 592         bzero(&info, sizeof(info));
 593         info.rti_addrs = rtm->rtm_addrs;
 594         /*
 595          * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
 596          * link-local address because rtrequest requires addresses with
 597          * embedded scope id.
 598          */
 599         if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
 600                 info.rti_info[RTAX_DST] = NULL;
 601                 senderr(EINVAL);
 602         }
 603         info.rti_flags = rtm->rtm_flags;
 604         if (info.rti_info[RTAX_DST] == NULL ||
 605             info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
 606             (info.rti_info[RTAX_GATEWAY] != NULL &&
 607              info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
 608                 senderr(EINVAL);
 609         saf = info.rti_info[RTAX_DST]->sa_family;
 610         /*
 611          * Verify that the caller has the appropriate privilege; RTM_GET
 612          * is the only operation the non-superuser is allowed.
 613          */
 614         if (rtm->rtm_type != RTM_GET) {
 615                 error = priv_check(curthread, PRIV_NET_ROUTE);
 616                 if (error)
 617                         senderr(error);
 618         }
 619
 620         /*
 621          * The given gateway address may be an interface address.
 622          * For example, issuing a "route change" command on a route
 623          * entry that was created from a tunnel, and the gateway
 624          * address given is the local end point. In this case the
 625          * RTF_GATEWAY flag must be cleared or the destination will
 626          * not be reachable even though there is no error message.
 627          */
 628         if (info.rti_info[RTAX_GATEWAY] != NULL &&
 629             info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
 630                 struct route gw_ro;
 631
 632                 bzero(&gw_ro, sizeof(gw_ro));
 633                 gw_ro.ro_dst = *info.rti_info[RTAX_GATEWAY];
 634                 rtalloc_ign_fib(&gw_ro, 0, so->so_fibnum);
 635                 /*
 636                  * A host route through the loopback interface is
 637                  * installed for each interface adddress. In pre 8.0
 638                  * releases the interface address of a PPP link type
 639                  * is not reachable locally. This behavior is fixed as
 640                  * part of the new L2/L3 redesign and rewrite work. The
 641                  * signature of this interface address route is the
 642                  * AF_LINK sa_family type of the rt_gateway, and the
 643                  * rt_ifp has the IFF_LOOPBACK flag set.
 644                  */
 645                 if (gw_ro.ro_rt != NULL &&
 646                     gw_ro.ro_rt->rt_gateway->sa_family == AF_LINK &&
 647                     gw_ro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) {
 648                         info.rti_flags &= ~RTF_GATEWAY;
 649                         info.rti_flags |= RTF_GWFLAG_COMPAT;
 650                 }
 651                 if (gw_ro.ro_rt != NULL)
 652                         RTFREE(gw_ro.ro_rt);
 653         }
 654
 655         switch (rtm->rtm_type) {
 656                 struct rtentry *saved_nrt;
 657
 658         case RTM_ADD:
 659                 if (info.rti_info[RTAX_GATEWAY] == NULL)
 660                         senderr(EINVAL);
 661                 saved_nrt = NULL;
 662
 663                 /* support for new ARP code */
 664                 if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
 665                     (rtm->rtm_flags & RTF_LLDATA) != 0) {
 666                         error = lla_rt_output(rtm, &info);
 667 #ifdef INET6
 668                         if (error == 0)
 669                                 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 670 #endif
 671                         break;
 672                 }
 673                 error = rtrequest1_fib(RTM_ADD, &info, &saved_nrt,
 674                     so->so_fibnum);
 675                 if (error == 0 && saved_nrt) {
 676 #ifdef INET6
 677                         rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 678 #endif
 679                         RT_LOCK(saved_nrt);
 680                         rt_setmetrics(rtm, saved_nrt);
 681                         rtm->rtm_index = saved_nrt->rt_ifp->if_index;
 682                         RT_REMREF(saved_nrt);
 683                         RT_UNLOCK(saved_nrt);
 684                 }
 685                 break;
 686
 687         case RTM_DELETE:
 688                 saved_nrt = NULL;
 689                 /* support for new ARP code */
 690                 if (info.rti_info[RTAX_GATEWAY] &&
 691                     (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
 692                     (rtm->rtm_flags & RTF_LLDATA) != 0) {
 693                         error = lla_rt_output(rtm, &info);
 694 #ifdef INET6
 695                         if (error == 0)
 696                                 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 697 #endif
 698                         break;
 699                 }
 700                 error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt,
 701                     so->so_fibnum);
 702                 if (error == 0) {
 703                         RT_LOCK(saved_nrt);
 704                         rt = saved_nrt;
 705                         goto report;
 706                 }
 707 #ifdef INET6
 708                 /* rt_msg2() will not be used when RTM_DELETE fails. */
 709                 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 710 #endif
 711                 break;
 712
 713         case RTM_GET:
 714         case RTM_CHANGE:
 715         case RTM_LOCK:
 716                 rnh = rt_tables_get_rnh(so->so_fibnum,
 717                     info.rti_info[RTAX_DST]->sa_family);
 718                 if (rnh == NULL)
 719                         senderr(EAFNOSUPPORT);
 720
 721                 RADIX_NODE_HEAD_RLOCK(rnh);
 722
 723                 if (info.rti_info[RTAX_NETMASK] == NULL &&
 724                     rtm->rtm_type == RTM_GET) {
 725                         /*
 726                          * Provide logest prefix match for
 727                          * address lookup (no mask).
 728                          * 'route -n get addr'
 729                          */
 730                         rt = (struct rtentry *) rnh->rnh_matchaddr(
 731                             info.rti_info[RTAX_DST], rnh);
 732                 } else
 733                         rt = (struct rtentry *) rnh->rnh_lookup(
 734                             info.rti_info[RTAX_DST],
 735                             info.rti_info[RTAX_NETMASK], rnh);
 736
 737                 if (rt == NULL) {
 738                         RADIX_NODE_HEAD_RUNLOCK(rnh);
 739                         senderr(ESRCH);
 740                 }
 741 #ifdef RADIX_MPATH
 742                 /*
 743                  * for RTM_CHANGE/LOCK, if we got multipath routes,
 744                  * we require users to specify a matching RTAX_GATEWAY.
 745                  *
 746                  * for RTM_GET, gate is optional even with multipath.
 747                  * if gate == NULL the first match is returned.
 748                  * (no need to call rt_mpath_matchgate if gate == NULL)
 749                  */
 750                 if (rn_mpath_capable(rnh) &&
 751                     (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
 752                         rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
 753                         if (!rt) {
 754                                 RADIX_NODE_HEAD_RUNLOCK(rnh);
 755                                 senderr(ESRCH);
 756                         }
 757                 }
 758 #endif
 759                 /*
 760                  * If performing proxied L2 entry insertion, and
 761                  * the actual PPP host entry is found, perform
 762                  * another search to retrieve the prefix route of
 763                  * the local end point of the PPP link.
 764                  */
 765                 if (rtm->rtm_flags & RTF_ANNOUNCE) {
 766                         struct sockaddr laddr;
 767
 768                         if (rt->rt_ifp != NULL &&
 769                             rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
 770                                 struct ifaddr *ifa;
 771
 772                                 ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1);
 773                                 if (ifa != NULL)
 774                                         rt_maskedcopy(ifa->ifa_addr,
 775                                                       &laddr,
 776                                                       ifa->ifa_netmask);
 777                         } else
 778                                 rt_maskedcopy(rt->rt_ifa->ifa_addr,
 779                                               &laddr,
 780                                               rt->rt_ifa->ifa_netmask);
 781                         /*
 782                          * refactor rt and no lock operation necessary
 783                          */
 784                         rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr, rnh);
 785                         if (rt == NULL) {
 786                                 RADIX_NODE_HEAD_RUNLOCK(rnh);
 787                                 senderr(ESRCH);
 788                         }
 789                 }
 790                 RT_LOCK(rt);
 791                 RT_ADDREF(rt);
 792                 RADIX_NODE_HEAD_RUNLOCK(rnh);
 793
 794                 switch(rtm->rtm_type) {
 795
 796                 case RTM_GET:
 797                 report:
 798                         RT_LOCK_ASSERT(rt);
 799                         if ((rt->rt_flags & RTF_HOST) == 0
 800                             ? jailed_without_vnet(curthread->td_ucred)
 801                             : prison_if(curthread->td_ucred,
 802                             rt_key(rt)) != 0) {
 803                                 RT_UNLOCK(rt);
 804                                 senderr(ESRCH);
 805                         }
 806                         info.rti_info[RTAX_DST] = rt_key(rt);
 807                         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
 808                         info.rti_info[RTAX_NETMASK] = rt_mask(rt);
 809                         info.rti_info[RTAX_GENMASK] = 0;
 810                         if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
 811                                 ifp = rt->rt_ifp;
 812                                 if (ifp) {
 813                                         info.rti_info[RTAX_IFP] =
 814                                             ifp->if_addr->ifa_addr;
 815                                         error = rtm_get_jailed(&info, ifp, rt,
 816                                             &saun, curthread->td_ucred);
 817                                         if (error != 0) {
 818                                                 RT_UNLOCK(rt);
 819                                                 senderr(error);
 820                                         }
 821                                         if (ifp->if_flags & IFF_POINTOPOINT)
 822                                                 info.rti_info[RTAX_BRD] =
 823                                                     rt->rt_ifa->ifa_dstaddr;
 824                                         rtm->rtm_index = ifp->if_index;
 825                                 } else {
 826                                         info.rti_info[RTAX_IFP] = NULL;
 827                                         info.rti_info[RTAX_IFA] = NULL;
 828                                 }
 829                         } else if ((ifp = rt->rt_ifp) != NULL) {
 830                                 rtm->rtm_index = ifp->if_index;
 831                         }
 832                         len = rt_msg2(rtm->rtm_type, &info, NULL, NULL);
 833                         if (len > rtm->rtm_msglen) {
 834                                 struct rt_msghdr *new_rtm;
 835                                 R_Malloc(new_rtm, struct rt_msghdr *, len);
 836                                 if (new_rtm == NULL) {
 837                                         RT_UNLOCK(rt);
 838                                         senderr(ENOBUFS);
 839                                 }
 840                                 bcopy(rtm, new_rtm, rtm->rtm_msglen);
 841                                 Free(rtm); rtm = new_rtm;
 842                         }
 843                         (void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm, NULL);
 844                         if (rt->rt_flags & RTF_GWFLAG_COMPAT)
 845                                 rtm->rtm_flags = RTF_GATEWAY |
 846                                         (rt->rt_flags & ~RTF_GWFLAG_COMPAT);
 847                         else
 848                                 rtm->rtm_flags = rt->rt_flags;
 849                         rt_getmetrics(rt, &rtm->rtm_rmx);
 850                         rtm->rtm_addrs = info.rti_addrs;
 851                         break;
 852
 853                 case RTM_CHANGE:
 854                         /*
 855                          * New gateway could require new ifaddr, ifp;
 856                          * flags may also be different; ifp may be specified
 857                          * by ll sockaddr when protocol address is ambiguous
 858                          */
 859                         if (((rt->rt_flags & RTF_GATEWAY) &&
 860                              info.rti_info[RTAX_GATEWAY] != NULL) ||
 861                             info.rti_info[RTAX_IFP] != NULL ||
 862                             (info.rti_info[RTAX_IFA] != NULL &&
 863                              !sa_equal(info.rti_info[RTAX_IFA],
 864                                        rt->rt_ifa->ifa_addr))) {
 865                                 RT_UNLOCK(rt);
 866                                 RADIX_NODE_HEAD_LOCK(rnh);
 867                                 error = rt_getifa_fib(&info, rt->rt_fibnum);
 868                                 /*
 869                                  * XXXRW: Really we should release this
 870                                  * reference later, but this maintains
 871                                  * historical behavior.
 872                                  */
 873                                 if (info.rti_ifa != NULL)
 874                                         ifa_free(info.rti_ifa);
 875                                 RADIX_NODE_HEAD_UNLOCK(rnh);
 876                                 if (error != 0)
 877                                         senderr(error);
 878                                 RT_LOCK(rt);
 879                         }
 880                         if (info.rti_ifa != NULL &&
 881                             info.rti_ifa != rt->rt_ifa &&
 882                             rt->rt_ifa != NULL &&
 883                             rt->rt_ifa->ifa_rtrequest != NULL) {
 884                                 rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
 885                                     &info);
 886                                 ifa_free(rt->rt_ifa);
 887                         }
 888                         if (info.rti_info[RTAX_GATEWAY] != NULL) {
 889                                 RT_UNLOCK(rt);
 890                                 RADIX_NODE_HEAD_LOCK(rnh);
 891                                 RT_LOCK(rt);
 892
 893                                 error = rt_setgate(rt, rt_key(rt),
 894                                     info.rti_info[RTAX_GATEWAY]);
 895                                 RADIX_NODE_HEAD_UNLOCK(rnh);
 896                                 if (error != 0) {
 897                                         RT_UNLOCK(rt);
 898                                         senderr(error);
 899                                 }
 900                                 rt->rt_flags &= ~RTF_GATEWAY;
 901                                 rt->rt_flags |= (RTF_GATEWAY & info.rti_flags);
 902                         }
 903                         if (info.rti_ifa != NULL &&
 904                             info.rti_ifa != rt->rt_ifa) {
 905                                 ifa_ref(info.rti_ifa);
 906                                 rt->rt_ifa = info.rti_ifa;
 907                                 rt->rt_ifp = info.rti_ifp;
 908                         }
 909                         /* Allow some flags to be toggled on change. */
 910                         rt->rt_flags = (rt->rt_flags & ~RTF_FMASK) |
 911                                     (rtm->rtm_flags & RTF_FMASK);
 912                         rt_setmetrics(rtm, rt);
 913                         rtm->rtm_index = rt->rt_ifp->if_index;
 914                         if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
 915                                rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
 916                         /* FALLTHROUGH */
 917                 case RTM_LOCK:
 918                         /* We don't support locks anymore */
 919                         break;
 920                 }
 921                 RT_UNLOCK(rt);
 922                 break;
 923
 924         default:
 925                 senderr(EOPNOTSUPP);
 926         }
 927
 928 flush:
 929         if (rtm) {
 930                 if (error)
 931                         rtm->rtm_errno = error;
 932                 else
 933                         rtm->rtm_flags |= RTF_DONE;
 934         }
 935         if (rt)         /* XXX can this be true? */
 936                 RTFREE(rt);
 937     {
 938         struct rawcb *rp = NULL;
 939         /*
 940          * Check to see if we don't want our own messages.
 941          */
 942         if ((so->so_options & SO_USELOOPBACK) == 0) {
 943                 if (route_cb.any_count <= 1) {
 944                         if (rtm)
 945                                 Free(rtm);
 946                         m_freem(m);
 947                         return (error);
 948                 }
 949                 /* There is another listener, so construct message */
 950                 rp = sotorawcb(so);
 951         }
 952         if (rtm) {
 953 #ifdef INET6
 954                 if (rti_need_deembed) {
 955                         /* sin6_scope_id is recovered before sending rtm. */
 956                         sin6 = (struct sockaddr_in6 *)&ss;
 957                         for (i = 0; i < RTAX_MAX; i++) {
 958                                 if (info.rti_info[i] == NULL)
 959                                         continue;
 960                                 if (info.rti_info[i]->sa_family != AF_INET6)
 961                                         continue;
 962                                 bcopy(info.rti_info[i], sin6, sizeof(*sin6));
 963                                 if (sa6_recoverscope(sin6) == 0)
 964                                         bcopy(sin6, info.rti_info[i],
 965                                                     sizeof(*sin6));
 966                         }
 967                 }
 968 #endif
 969                 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
 970                 if (m->m_pkthdr.len < rtm->rtm_msglen) {
 971                         m_freem(m);
 972                         m = NULL;
 973                 } else if (m->m_pkthdr.len > rtm->rtm_msglen)
 974                         m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
 975         }
 976         if (m) {
 977                 M_SETFIB(m, so->so_fibnum);
 978                 m->m_flags |= RTS_FILTER_FIB;
 979                 if (rp) {
 980                         /*
 981                          * XXX insure we don't get a copy by
 982                          * invalidating our protocol
 983                          */
 984                         unsigned short family = rp->rcb_proto.sp_family;
 985                         rp->rcb_proto.sp_family = 0;
 986                         rt_dispatch(m, saf);
 987                         rp->rcb_proto.sp_family = family;
 988                 } else
 989                         rt_dispatch(m, saf);
 990         }
 991         /* info.rti_info[RTAX_DST] (used above) can point inside of rtm */
 992         if (rtm)
 993                 Free(rtm);
 994     }
 995         return (error);
 996 #undef  sa_equal
 997 }
 998
 999 static void
1000 rt_setmetrics(const struct rt_msghdr *rtm, struct rtentry *rt)
1001 {
1002
1003         if (rtm->rtm_inits & RTV_MTU)
1004                 rt->rt_mtu = rtm->rtm_rmx.rmx_mtu;
1005         if (rtm->rtm_inits & RTV_WEIGHT)
1006                 rt->rt_weight = rtm->rtm_rmx.rmx_weight;
1007         /* Kernel -> userland timebase conversion. */
1008         if (rtm->rtm_inits & RTV_EXPIRE)
1009                 rt->rt_expire = rtm->rtm_rmx.rmx_expire ?
1010                     rtm->rtm_rmx.rmx_expire - time_second + time_uptime : 0;
1011 }
1012
1013 static void
1014 rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out)
1015 {
1016
1017         bzero(out, sizeof(*out));
1018         out->rmx_mtu = rt->rt_mtu;
1019         out->rmx_weight = rt->rt_weight;
1020         out->rmx_pksent = counter_u64_fetch(rt->rt_pksent);
1021         /* Kernel -> userland timebase conversion. */
1022         out->rmx_expire = rt->rt_expire ?
1023             rt->rt_expire - time_uptime + time_second : 0;
1024 }
1025
1026 /*
1027  * Extract the addresses of the passed sockaddrs.
1028  * Do a little sanity checking so as to avoid bad memory references.
1029  * This data is derived straight from userland.
1030  */
1031 static int
1032 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1033 {
1034         struct sockaddr *sa;
1035         int i;
1036
1037         for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1038                 if ((rtinfo->rti_addrs & (1 << i)) == 0)
1039                         continue;
1040                 sa = (struct sockaddr *)cp;
1041                 /*
1042                  * It won't fit.
1043                  */
1044                 if (cp + sa->sa_len > cplim)
1045                         return (EINVAL);
1046                 /*
1047                  * there are no more.. quit now
1048                  * If there are more bits, they are in error.
1049                  * I've seen this. route(1) can evidently generate these.
1050                  * This causes kernel to core dump.
1051                  * for compatibility, If we see this, point to a safe address.
1052                  */
1053                 if (sa->sa_len == 0) {
1054                         rtinfo->rti_info[i] = &sa_zero;
1055                         return (0); /* should be EINVAL but for compat */
1056                 }
1057                 /* accept it */
1058 #ifdef INET6
1059                 if (sa->sa_family == AF_INET6)
1060                         sa6_embedscope((struct sockaddr_in6 *)sa,
1061                             V_ip6_use_defzone);
1062 #endif
1063                 rtinfo->rti_info[i] = sa;
1064                 cp += SA_SIZE(sa);
1065         }
1066         return (0);
1067 }
1068
1069 /*
1070  * Used by the routing socket.
1071  */
1072 static struct mbuf *
1073 rt_msg1(int type, struct rt_addrinfo *rtinfo)
1074 {
1075         struct rt_msghdr *rtm;
1076         struct mbuf *m;
1077         int i;
1078         struct sockaddr *sa;
1079 #ifdef INET6
1080         struct sockaddr_storage ss;
1081         struct sockaddr_in6 *sin6;
1082 #endif
1083         int len, dlen;
1084
1085         switch (type) {
1086
1087         case RTM_DELADDR:
1088         case RTM_NEWADDR:
1089                 len = sizeof(struct ifa_msghdr);
1090                 break;
1091
1092         case RTM_DELMADDR:
1093         case RTM_NEWMADDR:
1094                 len = sizeof(struct ifma_msghdr);
1095                 break;
1096
1097         case RTM_IFINFO:
1098                 len = sizeof(struct if_msghdr);
1099                 break;
1100
1101         case RTM_IFANNOUNCE:
1102         case RTM_IEEE80211:
1103                 len = sizeof(struct if_announcemsghdr);
1104                 break;
1105
1106         default:
1107                 len = sizeof(struct rt_msghdr);
1108         }
1109
1110         /* XXXGL: can we use MJUMPAGESIZE cluster here? */
1111         KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1112         if (len > MHLEN)
1113                 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1114         else
1115                 m = m_gethdr(M_NOWAIT, MT_DATA);
1116         if (m == NULL)
1117                 return (m);
1118
1119         m->m_pkthdr.len = m->m_len = len;
1120         rtm = mtod(m, struct rt_msghdr *);
1121         bzero((caddr_t)rtm, len);
1122         for (i = 0; i < RTAX_MAX; i++) {
1123                 if ((sa = rtinfo->rti_info[i]) == NULL)
1124                         continue;
1125                 rtinfo->rti_addrs |= (1 << i);
1126                 dlen = SA_SIZE(sa);
1127 #ifdef INET6
1128                 if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1129                         sin6 = (struct sockaddr_in6 *)&ss;
1130                         bcopy(sa, sin6, sizeof(*sin6));
1131                         if (sa6_recoverscope(sin6) == 0)
1132                                 sa = (struct sockaddr *)sin6;
1133                 }
1134 #endif
1135                 m_copyback(m, len, dlen, (caddr_t)sa);
1136                 len += dlen;
1137         }
1138         if (m->m_pkthdr.len != len) {
1139                 m_freem(m);
1140                 return (NULL);
1141         }
1142         rtm->rtm_msglen = len;
1143         rtm->rtm_version = RTM_VERSION;
1144         rtm->rtm_type = type;
1145         return (m);
1146 }
1147
1148 /*
1149  * Used by the sysctl code and routing socket.
1150  */
1151 static int
1152 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
1153 {
1154         int i;
1155         int len, dlen, second_time = 0;
1156         caddr_t cp0;
1157 #ifdef INET6
1158         struct sockaddr_storage ss;
1159         struct sockaddr_in6 *sin6;
1160 #endif
1161
1162         rtinfo->rti_addrs = 0;
1163 again:
1164         switch (type) {
1165
1166         case RTM_DELADDR:
1167         case RTM_NEWADDR:
1168                 if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1169 #ifdef COMPAT_FREEBSD32
1170                         if (w->w_req->flags & SCTL_MASK32)
1171                                 len = sizeof(struct ifa_msghdrl32);
1172                         else
1173 #endif
1174                                 len = sizeof(struct ifa_msghdrl);
1175                 } else
1176                         len = sizeof(struct ifa_msghdr);
1177                 break;
1178
1179         case RTM_IFINFO:
1180 #ifdef COMPAT_FREEBSD32
1181                 if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1182                         if (w->w_op == NET_RT_IFLISTL)
1183                                 len = sizeof(struct if_msghdrl32);
1184                         else
1185                                 len = sizeof(struct if_msghdr32);
1186                         break;
1187                 }
1188 #endif
1189                 if (w != NULL && w->w_op == NET_RT_IFLISTL)
1190                         len = sizeof(struct if_msghdrl);
1191                 else
1192                         len = sizeof(struct if_msghdr);
1193                 break;
1194
1195         case RTM_NEWMADDR:
1196                 len = sizeof(struct ifma_msghdr);
1197                 break;
1198
1199         default:
1200                 len = sizeof(struct rt_msghdr);
1201         }
1202         cp0 = cp;
1203         if (cp0)
1204                 cp += len;
1205         for (i = 0; i < RTAX_MAX; i++) {
1206                 struct sockaddr *sa;
1207
1208                 if ((sa = rtinfo->rti_info[i]) == NULL)
1209                         continue;
1210                 rtinfo->rti_addrs |= (1 << i);
1211                 dlen = SA_SIZE(sa);
1212                 if (cp) {
1213 #ifdef INET6
1214                         if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1215                                 sin6 = (struct sockaddr_in6 *)&ss;
1216                                 bcopy(sa, sin6, sizeof(*sin6));
1217                                 if (sa6_recoverscope(sin6) == 0)
1218                                         sa = (struct sockaddr *)sin6;
1219                         }
1220 #endif
1221                         bcopy((caddr_t)sa, cp, (unsigned)dlen);
1222                         cp += dlen;
1223                 }
1224                 len += dlen;
1225         }
1226         len = ALIGN(len);
1227         if (cp == NULL && w != NULL && !second_time) {
1228                 struct walkarg *rw = w;
1229
1230                 if (rw->w_req) {
1231                         if (rw->w_tmemsize < len) {
1232                                 if (rw->w_tmem)
1233                                         free(rw->w_tmem, M_RTABLE);
1234                                 rw->w_tmem = (caddr_t)
1235                                         malloc(len, M_RTABLE, M_NOWAIT);
1236                                 if (rw->w_tmem)
1237                                         rw->w_tmemsize = len;
1238                         }
1239                         if (rw->w_tmem) {
1240                                 cp = rw->w_tmem;
1241                                 second_time = 1;
1242                                 goto again;
1243                         }
1244                 }
1245         }
1246         if (cp) {
1247                 struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
1248
1249                 rtm->rtm_version = RTM_VERSION;
1250                 rtm->rtm_type = type;
1251                 rtm->rtm_msglen = len;
1252         }
1253         return (len);
1254 }
1255
1256 /*
1257  * This routine is called to generate a message from the routing
1258  * socket indicating that a redirect has occured, a routing lookup
1259  * has failed, or that a protocol has detected timeouts to a particular
1260  * destination.
1261  */
1262 void
1263 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1264     int fibnum)
1265 {
1266         struct rt_msghdr *rtm;
1267         struct mbuf *m;
1268         struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1269
1270         if (route_cb.any_count == 0)
1271                 return;
1272         m = rt_msg1(type, rtinfo);
1273         if (m == NULL)
1274                 return;
1275
1276         if (fibnum != RT_ALL_FIBS) {
1277                 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1278                     "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1279                 M_SETFIB(m, fibnum);
1280                 m->m_flags |= RTS_FILTER_FIB;
1281         }
1282
1283         rtm = mtod(m, struct rt_msghdr *);
1284         rtm->rtm_flags = RTF_DONE | flags;
1285         rtm->rtm_errno = error;
1286         rtm->rtm_addrs = rtinfo->rti_addrs;
1287         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1288 }
1289
1290 void
1291 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1292 {
1293
1294         rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1295 }
1296
1297 /*
1298  * This routine is called to generate a message from the routing
1299  * socket indicating that the status of a network interface has changed.
1300  */
1301 void
1302 rt_ifmsg(struct ifnet *ifp)
1303 {
1304         struct if_msghdr *ifm;
1305         struct mbuf *m;
1306         struct rt_addrinfo info;
1307
1308         if (route_cb.any_count == 0)
1309                 return;
1310         bzero((caddr_t)&info, sizeof(info));
1311         m = rt_msg1(RTM_IFINFO, &info);
1312         if (m == NULL)
1313                 return;
1314         ifm = mtod(m, struct if_msghdr *);
1315         ifm->ifm_index = ifp->if_index;
1316         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1317         ifm->ifm_data = ifp->if_data;
1318         ifm->ifm_addrs = 0;
1319         rt_dispatch(m, AF_UNSPEC);
1320 }
1321
1322 /*
1323  * Announce interface address arrival/withdraw.
1324  * Please do not call directly, use rt_addrmsg().
1325  * Assume input data to be valid.
1326  * Returns 0 on success.
1327  */
1328 int
1329 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1330 {
1331         struct rt_addrinfo info;
1332         struct sockaddr *sa;
1333         int ncmd;
1334         struct mbuf *m;
1335         struct ifa_msghdr *ifam;
1336         struct ifnet *ifp = ifa->ifa_ifp;
1337
1338         if (route_cb.any_count == 0)
1339                 return (0);
1340
1341         ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1342
1343         bzero((caddr_t)&info, sizeof(info));
1344         info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1345         info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1346         info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1347         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1348         if ((m = rt_msg1(ncmd, &info)) == NULL)
1349                 return (ENOBUFS);
1350         ifam = mtod(m, struct ifa_msghdr *);
1351         ifam->ifam_index = ifp->if_index;
1352         ifam->ifam_metric = ifa->ifa_metric;
1353         ifam->ifam_flags = ifa->ifa_flags;
1354         ifam->ifam_addrs = info.rti_addrs;
1355
1356         if (fibnum != RT_ALL_FIBS) {
1357                 M_SETFIB(m, fibnum);
1358                 m->m_flags |= RTS_FILTER_FIB;
1359         }
1360
1361         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1362
1363         return (0);
1364 }
1365
1366 /*
1367  * Announce route addition/removal.
1368  * Please do not call directly, use rt_routemsg().
1369  * Note that @rt data MAY be inconsistent/invalid:
1370  * if some userland app sends us "invalid" route message (invalid mask,
1371  * no dst, wrong address families, etc...) we need to pass it back
1372  * to app (and any other rtsock consumers) with rtm_errno field set to
1373  * non-zero value.
1374  *
1375  * Returns 0 on success.
1376  */
1377 int
1378 rtsock_routemsg(int cmd, struct ifnet *ifp, int error, struct rtentry *rt,
1379     int fibnum)
1380 {
1381         struct rt_addrinfo info;
1382         struct sockaddr *sa;
1383         struct mbuf *m;
1384         struct rt_msghdr *rtm;
1385
1386         if (route_cb.any_count == 0)
1387                 return (0);
1388
1389         bzero((caddr_t)&info, sizeof(info));
1390         info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1391         info.rti_info[RTAX_DST] = sa = rt_key(rt);
1392         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1393         if ((m = rt_msg1(cmd, &info)) == NULL)
1394                 return (ENOBUFS);
1395         rtm = mtod(m, struct rt_msghdr *);
1396         rtm->rtm_index = ifp->if_index;
1397         rtm->rtm_flags |= rt->rt_flags;
1398         rtm->rtm_errno = error;
1399         rtm->rtm_addrs = info.rti_addrs;
1400
1401         if (fibnum != RT_ALL_FIBS) {
1402                 M_SETFIB(m, fibnum);
1403                 m->m_flags |= RTS_FILTER_FIB;
1404         }
1405
1406         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1407
1408         return (0);
1409 }
1410
1411 /*
1412  * This is the analogue to the rt_newaddrmsg which performs the same
1413  * function but for multicast group memberhips.  This is easier since
1414  * there is no route state to worry about.
1415  */
1416 void
1417 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1418 {
1419         struct rt_addrinfo info;
1420         struct mbuf *m = NULL;
1421         struct ifnet *ifp = ifma->ifma_ifp;
1422         struct ifma_msghdr *ifmam;
1423
1424         if (route_cb.any_count == 0)
1425                 return;
1426
1427         bzero((caddr_t)&info, sizeof(info));
1428         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1429         info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1430         /*
1431          * If a link-layer address is present, present it as a ``gateway''
1432          * (similarly to how ARP entries, e.g., are presented).
1433          */
1434         info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1435         m = rt_msg1(cmd, &info);
1436         if (m == NULL)
1437                 return;
1438         ifmam = mtod(m, struct ifma_msghdr *);
1439         KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1440             __func__));
1441         ifmam->ifmam_index = ifp->if_index;
1442         ifmam->ifmam_addrs = info.rti_addrs;
1443         rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1444 }
1445
1446 static struct mbuf *
1447 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1448         struct rt_addrinfo *info)
1449 {
1450         struct if_announcemsghdr *ifan;
1451         struct mbuf *m;
1452
1453         if (route_cb.any_count == 0)
1454                 return NULL;
1455         bzero((caddr_t)info, sizeof(*info));
1456         m = rt_msg1(type, info);
1457         if (m != NULL) {
1458                 ifan = mtod(m, struct if_announcemsghdr *);
1459                 ifan->ifan_index = ifp->if_index;
1460                 strlcpy(ifan->ifan_name, ifp->if_xname,
1461                         sizeof(ifan->ifan_name));
1462                 ifan->ifan_what = what;
1463         }
1464         return m;
1465 }
1466
1467 /*
1468  * This is called to generate routing socket messages indicating
1469  * IEEE80211 wireless events.
1470  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1471  */
1472 void
1473 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1474 {
1475         struct mbuf *m;
1476         struct rt_addrinfo info;
1477
1478         m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1479         if (m != NULL) {
1480                 /*
1481                  * Append the ieee80211 data.  Try to stick it in the
1482                  * mbuf containing the ifannounce msg; otherwise allocate
1483                  * a new mbuf and append.
1484                  *
1485                  * NB: we assume m is a single mbuf.
1486                  */
1487                 if (data_len > M_TRAILINGSPACE(m)) {
1488                         struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1489                         if (n == NULL) {
1490                                 m_freem(m);
1491                                 return;
1492                         }
1493                         bcopy(data, mtod(n, void *), data_len);
1494                         n->m_len = data_len;
1495                         m->m_next = n;
1496                 } else if (data_len > 0) {
1497                         bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1498                         m->m_len += data_len;
1499                 }
1500                 if (m->m_flags & M_PKTHDR)
1501                         m->m_pkthdr.len += data_len;
1502                 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1503                 rt_dispatch(m, AF_UNSPEC);
1504         }
1505 }
1506
1507 /*
1508  * This is called to generate routing socket messages indicating
1509  * network interface arrival and departure.
1510  */
1511 void
1512 rt_ifannouncemsg(struct ifnet *ifp, int what)
1513 {
1514         struct mbuf *m;
1515         struct rt_addrinfo info;
1516
1517         m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1518         if (m != NULL)
1519                 rt_dispatch(m, AF_UNSPEC);
1520 }
1521
1522 static void
1523 rt_dispatch(struct mbuf *m, sa_family_t saf)
1524 {
1525         struct m_tag *tag;
1526
1527         /*
1528          * Preserve the family from the sockaddr, if any, in an m_tag for
1529          * use when injecting the mbuf into the routing socket buffer from
1530          * the netisr.
1531          */
1532         if (saf != AF_UNSPEC) {
1533                 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1534                     M_NOWAIT);
1535                 if (tag == NULL) {
1536                         m_freem(m);
1537                         return;
1538                 }
1539                 *(unsigned short *)(tag + 1) = saf;
1540                 m_tag_prepend(m, tag);
1541         }
1542 #ifdef VIMAGE
1543         if (V_loif)
1544                 m->m_pkthdr.rcvif = V_loif;
1545         else {
1546                 m_freem(m);
1547                 return;
1548         }
1549 #endif
1550         netisr_queue(NETISR_ROUTE, m);  /* mbuf is free'd on failure. */
1551 }
1552
1553 /*
1554  * This is used in dumping the kernel table via sysctl().
1555  */
1556 static int
1557 sysctl_dumpentry(struct radix_node *rn, void *vw)
1558 {
1559         struct walkarg *w = vw;
1560         struct rtentry *rt = (struct rtentry *)rn;
1561         int error = 0, size;
1562         struct rt_addrinfo info;
1563
1564         if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1565                 return 0;
1566         if ((rt->rt_flags & RTF_HOST) == 0
1567             ? jailed_without_vnet(w->w_req->td->td_ucred)
1568             : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1569                 return (0);
1570         bzero((caddr_t)&info, sizeof(info));
1571         info.rti_info[RTAX_DST] = rt_key(rt);
1572         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1573         info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1574         info.rti_info[RTAX_GENMASK] = 0;
1575         if (rt->rt_ifp) {
1576                 info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1577                 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1578                 if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1579                         info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1580         }
1581         size = rt_msg2(RTM_GET, &info, NULL, w);
1582         if (w->w_req && w->w_tmem) {
1583                 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1584
1585                 if (rt->rt_flags & RTF_GWFLAG_COMPAT)
1586                         rtm->rtm_flags = RTF_GATEWAY |
1587                                 (rt->rt_flags & ~RTF_GWFLAG_COMPAT);
1588                 else
1589                         rtm->rtm_flags = rt->rt_flags;
1590                 rt_getmetrics(rt, &rtm->rtm_rmx);
1591                 rtm->rtm_index = rt->rt_ifp->if_index;
1592                 rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1593                 rtm->rtm_addrs = info.rti_addrs;
1594                 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1595                 return (error);
1596         }
1597         return (error);
1598 }
1599
1600 #ifdef COMPAT_FREEBSD32
1601 static void
1602 copy_ifdata32(struct if_data *src, struct if_data32 *dst)
1603 {
1604
1605         bzero(dst, sizeof(*dst));
1606         CP(*src, *dst, ifi_type);
1607         CP(*src, *dst, ifi_physical);
1608         CP(*src, *dst, ifi_addrlen);
1609         CP(*src, *dst, ifi_hdrlen);
1610         CP(*src, *dst, ifi_link_state);
1611         CP(*src, *dst, ifi_vhid);
1612         CP(*src, *dst, ifi_baudrate_pf);
1613         dst->ifi_datalen = sizeof(struct if_data32);
1614         CP(*src, *dst, ifi_mtu);
1615         CP(*src, *dst, ifi_metric);
1616         CP(*src, *dst, ifi_baudrate);
1617         CP(*src, *dst, ifi_ipackets);
1618         CP(*src, *dst, ifi_ierrors);
1619         CP(*src, *dst, ifi_opackets);
1620         CP(*src, *dst, ifi_oerrors);
1621         CP(*src, *dst, ifi_collisions);
1622         CP(*src, *dst, ifi_ibytes);
1623         CP(*src, *dst, ifi_obytes);
1624         CP(*src, *dst, ifi_imcasts);
1625         CP(*src, *dst, ifi_omcasts);
1626         CP(*src, *dst, ifi_iqdrops);
1627         CP(*src, *dst, ifi_noproto);
1628         CP(*src, *dst, ifi_hwassist);
1629         CP(*src, *dst, ifi_epoch);
1630         TV_CP(*src, *dst, ifi_lastchange);
1631 }
1632 #endif
1633
1634 static int
1635 sysctl_iflist_ifml(struct ifnet *ifp, struct rt_addrinfo *info,
1636     struct walkarg *w, int len)
1637 {
1638         struct if_msghdrl *ifm;
1639
1640 #ifdef COMPAT_FREEBSD32
1641         if (w->w_req->flags & SCTL_MASK32) {
1642                 struct if_msghdrl32 *ifm32;
1643
1644                 ifm32 = (struct if_msghdrl32 *)w->w_tmem;
1645                 ifm32->ifm_addrs = info->rti_addrs;
1646                 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1647                 ifm32->ifm_index = ifp->if_index;
1648                 ifm32->_ifm_spare1 = 0;
1649                 ifm32->ifm_len = sizeof(*ifm32);
1650                 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1651
1652                 copy_ifdata32(&ifp->if_data, &ifm32->ifm_data);
1653                 /* Fixup if_data carp(4) vhid. */
1654                 if (carp_get_vhid_p != NULL)
1655                         ifm32->ifm_data.ifi_vhid =
1656                             (*carp_get_vhid_p)(ifp->if_addr);
1657                 ifm32->ifm_data.ifi_oqdrops = ifp->if_snd.ifq_drops;
1658
1659                 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm32, len));
1660         }
1661 #endif
1662         ifm = (struct if_msghdrl *)w->w_tmem;
1663         ifm->ifm_addrs = info->rti_addrs;
1664         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1665         ifm->ifm_index = ifp->if_index;
1666         ifm->_ifm_spare1 = 0;
1667         ifm->ifm_len = sizeof(*ifm);
1668         ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1669
1670         ifm->ifm_data = ifp->if_data;
1671         /* Fixup if_data carp(4) vhid. */
1672         if (carp_get_vhid_p != NULL)
1673                 ifm->ifm_data.ifi_vhid = (*carp_get_vhid_p)(ifp->if_addr);
1674
1675         ifm->ifm_data.ifi_datalen += sizeof(u_long);
1676         ifm->ifi_oqdrops = ifp->if_snd.ifq_drops;
1677
1678         return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1679 }
1680
1681 static int
1682 sysctl_iflist_ifm(struct ifnet *ifp, struct rt_addrinfo *info,
1683     struct walkarg *w, int len)
1684 {
1685         struct if_msghdr *ifm;
1686
1687 #ifdef COMPAT_FREEBSD32
1688         if (w->w_req->flags & SCTL_MASK32) {
1689                 struct if_msghdr32 *ifm32;
1690
1691                 ifm32 = (struct if_msghdr32 *)w->w_tmem;
1692                 ifm32->ifm_addrs = info->rti_addrs;
1693                 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1694                 ifm32->ifm_index = ifp->if_index;
1695
1696                 copy_ifdata32(&ifp->if_data, &ifm32->ifm_data);
1697                 /* Fixup if_data carp(4) vhid. */
1698                 if (carp_get_vhid_p != NULL)
1699                         ifm32->ifm_data.ifi_vhid =
1700                             (*carp_get_vhid_p)(ifp->if_addr);
1701
1702                 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm32, len));
1703         }
1704 #endif
1705         ifm = (struct if_msghdr *)w->w_tmem;
1706         ifm->ifm_addrs = info->rti_addrs;
1707         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1708         ifm->ifm_index = ifp->if_index;
1709
1710         ifm->ifm_data = ifp->if_data;
1711         /* Fixup if_data carp(4) vhid. */
1712         if (carp_get_vhid_p != NULL)
1713                 ifm->ifm_data.ifi_vhid = (*carp_get_vhid_p)(ifp->if_addr);
1714
1715         return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1716 }
1717
1718 static int
1719 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1720     struct walkarg *w, int len)
1721 {
1722         struct ifa_msghdrl *ifam;
1723
1724 #ifdef COMPAT_FREEBSD32
1725         if (w->w_req->flags & SCTL_MASK32) {
1726                 struct ifa_msghdrl32 *ifam32;
1727
1728                 ifam32 = (struct ifa_msghdrl32 *)w->w_tmem;
1729                 ifam32->ifam_addrs = info->rti_addrs;
1730                 ifam32->ifam_flags = ifa->ifa_flags;
1731                 ifam32->ifam_index = ifa->ifa_ifp->if_index;
1732                 ifam32->_ifam_spare1 = 0;
1733                 ifam32->ifam_len = sizeof(*ifam32);
1734                 ifam32->ifam_data_off =
1735                     offsetof(struct ifa_msghdrl32, ifam_data);
1736                 ifam32->ifam_metric = ifa->ifa_metric;
1737
1738                 copy_ifdata32(&ifa->ifa_ifp->if_data, &ifam32->ifam_data);
1739                 /* Fixup if_data carp(4) vhid. */
1740                 if (carp_get_vhid_p != NULL)
1741                         ifam32->ifam_data.ifi_vhid = (*carp_get_vhid_p)(ifa);
1742
1743                 return (SYSCTL_OUT(w->w_req, (caddr_t)ifam32, len));
1744         }
1745 #endif
1746
1747         ifam = (struct ifa_msghdrl *)w->w_tmem;
1748         ifam->ifam_addrs = info->rti_addrs;
1749         ifam->ifam_flags = ifa->ifa_flags;
1750         ifam->ifam_index = ifa->ifa_ifp->if_index;
1751         ifam->_ifam_spare1 = 0;
1752         ifam->ifam_len = sizeof(*ifam);
1753         ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1754         ifam->ifam_metric = ifa->ifa_metric;
1755
1756         ifam->ifam_data = ifa->if_data;
1757         /* Fixup if_data carp(4) vhid. */
1758         if (carp_get_vhid_p != NULL)
1759                 ifam->ifam_data.ifi_vhid = (*carp_get_vhid_p)(ifa);
1760
1761         return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1762 }
1763
1764 static int
1765 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1766     struct walkarg *w, int len)
1767 {
1768         struct ifa_msghdr *ifam;
1769
1770         ifam = (struct ifa_msghdr *)w->w_tmem;
1771         ifam->ifam_addrs = info->rti_addrs;
1772         ifam->ifam_flags = ifa->ifa_flags;
1773         ifam->ifam_index = ifa->ifa_ifp->if_index;
1774         ifam->ifam_metric = ifa->ifa_metric;
1775
1776         return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1777 }
1778
1779 static int
1780 sysctl_iflist(int af, struct walkarg *w)
1781 {
1782         struct ifnet *ifp;
1783         struct ifaddr *ifa;
1784         struct rt_addrinfo info;
1785         int len, error = 0;
1786
1787         bzero((caddr_t)&info, sizeof(info));
1788         IFNET_RLOCK_NOSLEEP();
1789         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1790                 if (w->w_arg && w->w_arg != ifp->if_index)
1791                         continue;
1792                 IF_ADDR_RLOCK(ifp);
1793                 ifa = ifp->if_addr;
1794                 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1795                 len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1796                 info.rti_info[RTAX_IFP] = NULL;
1797                 if (w->w_req && w->w_tmem) {
1798                         if (w->w_op == NET_RT_IFLISTL)
1799                                 error = sysctl_iflist_ifml(ifp, &info, w, len);
1800                         else
1801                                 error = sysctl_iflist_ifm(ifp, &info, w, len);
1802                         if (error)
1803                                 goto done;
1804                 }
1805                 while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1806                         if (af && af != ifa->ifa_addr->sa_family)
1807                                 continue;
1808                         if (prison_if(w->w_req->td->td_ucred,
1809                             ifa->ifa_addr) != 0)
1810                                 continue;
1811                         info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1812                         info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1813                         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1814                         len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1815                         if (w->w_req && w->w_tmem) {
1816                                 if (w->w_op == NET_RT_IFLISTL)
1817                                         error = sysctl_iflist_ifaml(ifa, &info,
1818                                             w, len);
1819                                 else
1820                                         error = sysctl_iflist_ifam(ifa, &info,
1821                                             w, len);
1822                                 if (error)
1823                                         goto done;
1824                         }
1825                 }
1826                 IF_ADDR_RUNLOCK(ifp);
1827                 info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1828                         info.rti_info[RTAX_BRD] = NULL;
1829         }
1830 done:
1831         if (ifp != NULL)
1832                 IF_ADDR_RUNLOCK(ifp);
1833         IFNET_RUNLOCK_NOSLEEP();
1834         return (error);
1835 }
1836
1837 static int
1838 sysctl_ifmalist(int af, struct walkarg *w)
1839 {
1840         struct ifnet *ifp;
1841         struct ifmultiaddr *ifma;
1842         struct  rt_addrinfo info;
1843         int     len, error = 0;
1844         struct ifaddr *ifa;
1845
1846         bzero((caddr_t)&info, sizeof(info));
1847         IFNET_RLOCK_NOSLEEP();
1848         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1849                 if (w->w_arg && w->w_arg != ifp->if_index)
1850                         continue;
1851                 ifa = ifp->if_addr;
1852                 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1853                 IF_ADDR_RLOCK(ifp);
1854                 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1855                         if (af && af != ifma->ifma_addr->sa_family)
1856                                 continue;
1857                         if (prison_if(w->w_req->td->td_ucred,
1858                             ifma->ifma_addr) != 0)
1859                                 continue;
1860                         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1861                         info.rti_info[RTAX_GATEWAY] =
1862                             (ifma->ifma_addr->sa_family != AF_LINK) ?
1863                             ifma->ifma_lladdr : NULL;
1864                         len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1865                         if (w->w_req && w->w_tmem) {
1866                                 struct ifma_msghdr *ifmam;
1867
1868                                 ifmam = (struct ifma_msghdr *)w->w_tmem;
1869                                 ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1870                                 ifmam->ifmam_flags = 0;
1871                                 ifmam->ifmam_addrs = info.rti_addrs;
1872                                 error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1873                                 if (error) {
1874                                         IF_ADDR_RUNLOCK(ifp);
1875                                         goto done;
1876                                 }
1877                         }
1878                 }
1879                 IF_ADDR_RUNLOCK(ifp);
1880         }
1881 done:
1882         IFNET_RUNLOCK_NOSLEEP();
1883         return (error);
1884 }
1885
1886 static int
1887 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1888 {
1889         int     *name = (int *)arg1;
1890         u_int   namelen = arg2;
1891         struct radix_node_head *rnh = NULL; /* silence compiler. */
1892         int     i, lim, error = EINVAL;
1893         int     fib = 0;
1894         u_char  af;
1895         struct  walkarg w;
1896
1897         name ++;
1898         namelen--;
1899         if (req->newptr)
1900                 return (EPERM);
1901         if (name[1] == NET_RT_DUMP) {
1902                 if (namelen == 3)
1903                         fib = req->td->td_proc->p_fibnum;
1904                 else if (namelen == 4)
1905                         fib = (name[3] == RT_ALL_FIBS) ?
1906                             req->td->td_proc->p_fibnum : name[3];
1907                 else
1908                         return ((namelen < 3) ? EISDIR : ENOTDIR);
1909                 if (fib < 0 || fib >= rt_numfibs)
1910                         return (EINVAL);
1911         } else if (namelen != 3)
1912                 return ((namelen < 3) ? EISDIR : ENOTDIR);
1913         af = name[0];
1914         if (af > AF_MAX)
1915                 return (EINVAL);
1916         bzero(&w, sizeof(w));
1917         w.w_op = name[1];
1918         w.w_arg = name[2];
1919         w.w_req = req;
1920
1921         error = sysctl_wire_old_buffer(req, 0);
1922         if (error)
1923                 return (error);
1924         switch (w.w_op) {
1925
1926         case NET_RT_DUMP:
1927         case NET_RT_FLAGS:
1928                 if (af == 0) {                  /* dump all tables */
1929                         i = 1;
1930                         lim = AF_MAX;
1931                 } else                          /* dump only one table */
1932                         i = lim = af;
1933
1934                 /*
1935                  * take care of llinfo entries, the caller must
1936                  * specify an AF
1937                  */
1938                 if (w.w_op == NET_RT_FLAGS &&
1939                     (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1940                         if (af != 0)
1941                                 error = lltable_sysctl_dumparp(af, w.w_req);
1942                         else
1943                                 error = EINVAL;
1944                         break;
1945                 }
1946                 /*
1947                  * take care of routing entries
1948                  */
1949                 for (error = 0; error == 0 && i <= lim; i++) {
1950                         rnh = rt_tables_get_rnh(fib, i);
1951                         if (rnh != NULL) {
1952                                 RADIX_NODE_HEAD_RLOCK(rnh);
1953                                 error = rnh->rnh_walktree(rnh,
1954                                     sysctl_dumpentry, &w);
1955                                 RADIX_NODE_HEAD_RUNLOCK(rnh);
1956                         } else if (af != 0)
1957                                 error = EAFNOSUPPORT;
1958                 }
1959                 break;
1960
1961         case NET_RT_IFLIST:
1962         case NET_RT_IFLISTL:
1963                 error = sysctl_iflist(af, &w);
1964                 break;
1965
1966         case NET_RT_IFMALIST:
1967                 error = sysctl_ifmalist(af, &w);
1968                 break;
1969         }
1970         if (w.w_tmem)
1971                 free(w.w_tmem, M_RTABLE);
1972         return (error);
1973 }
1974
1975 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1976
1977 /*
1978  * Definitions of protocols supported in the ROUTE domain.
1979  */
1980
1981 static struct domain routedomain;               /* or at least forward */
1982
1983 static struct protosw routesw[] = {
1984 {
1985         .pr_type =              SOCK_RAW,
1986         .pr_domain =            &routedomain,
1987         .pr_flags =             PR_ATOMIC|PR_ADDR,
1988         .pr_output =            route_output,
1989         .pr_ctlinput =          raw_ctlinput,
1990         .pr_init =              raw_init,
1991         .pr_usrreqs =           &route_usrreqs
1992 }
1993 };
1994
1995 static struct domain routedomain = {
1996         .dom_family =           PF_ROUTE,
1997         .dom_name =              "route",
1998         .dom_protosw =          routesw,
1999         .dom_protoswNPROTOSW =  &routesw[sizeof(routesw)/sizeof(routesw[0])]
2000 };
2001
2002 VNET_DOMAIN_SET(route);