sys/net/rtsock.c

   1 /*-
   2  * SPDX-License-Identifier: BSD-3-Clause
   3  *
   4  * Copyright (c) 1988, 1991, 1993
   5  *      The Regents of the University of California.  All rights reserved.
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  10  * 1. Redistributions of source code must retain the above copyright
  11  *    notice, this list of conditions and the following disclaimer.
  12  * 2. Redistributions in binary form must reproduce the above copyright
  13  *    notice, this list of conditions and the following disclaimer in the
  14  *    documentation and/or other materials provided with the distribution.
  15  * 3. Neither the name of the University nor the names of its contributors
  16  *    may be used to endorse or promote products derived from this software
  17  *    without specific prior written permission.
  18  *
  19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  29  * SUCH DAMAGE.
  30  *
  31  *      @(#)rtsock.c    8.7 (Berkeley) 10/12/95
  32  * $FreeBSD$
  33  */
  34 #include "opt_ddb.h"
  35 #include "opt_route.h"
  36 #include "opt_inet.h"
  37 #include "opt_inet6.h"
  38
  39 #include <sys/param.h>
  40 #include <sys/jail.h>
  41 #include <sys/kernel.h>
  42 #include <sys/domain.h>
  43 #include <sys/lock.h>
  44 #include <sys/malloc.h>
  45 #include <sys/mbuf.h>
  46 #include <sys/priv.h>
  47 #include <sys/proc.h>
  48 #include <sys/protosw.h>
  49 #include <sys/rmlock.h>
  50 #include <sys/rwlock.h>
  51 #include <sys/signalvar.h>
  52 #include <sys/socket.h>
  53 #include <sys/socketvar.h>
  54 #include <sys/sysctl.h>
  55 #include <sys/systm.h>
  56
  57 #include <net/if.h>
  58 #include <net/if_var.h>
  59 #include <net/if_dl.h>
  60 #include <net/if_llatbl.h>
  61 #include <net/if_types.h>
  62 #include <net/netisr.h>
  63 #include <net/raw_cb.h>
  64 #include <net/route.h>
  65 #include <net/route/route_ctl.h>
  66 #include <net/route/route_var.h>
  67 #ifdef RADIX_MPATH
  68 #include <net/radix_mpath.h>
  69 #endif
  70 #include <net/vnet.h>
  71
  72 #include <netinet/in.h>
  73 #include <netinet/if_ether.h>
  74 #include <netinet/ip_carp.h>
  75 #ifdef INET6
  76 #include <netinet6/ip6_var.h>
  77 #include <netinet6/scope6_var.h>
  78 #endif
  79 #include <net/route/nhop.h>
  80
  81 #ifdef COMPAT_FREEBSD32
  82 #include <sys/mount.h>
  83 #include <compat/freebsd32/freebsd32.h>
  84
  85 struct if_msghdr32 {
  86         uint16_t ifm_msglen;
  87         uint8_t ifm_version;
  88         uint8_t ifm_type;
  89         int32_t ifm_addrs;
  90         int32_t ifm_flags;
  91         uint16_t ifm_index;
  92         uint16_t _ifm_spare1;
  93         struct  if_data ifm_data;
  94 };
  95
  96 struct if_msghdrl32 {
  97         uint16_t ifm_msglen;
  98         uint8_t ifm_version;
  99         uint8_t ifm_type;
 100         int32_t ifm_addrs;
 101         int32_t ifm_flags;
 102         uint16_t ifm_index;
 103         uint16_t _ifm_spare1;
 104         uint16_t ifm_len;
 105         uint16_t ifm_data_off;
 106         uint32_t _ifm_spare2;
 107         struct  if_data ifm_data;
 108 };
 109
 110 struct ifa_msghdrl32 {
 111         uint16_t ifam_msglen;
 112         uint8_t ifam_version;
 113         uint8_t ifam_type;
 114         int32_t ifam_addrs;
 115         int32_t ifam_flags;
 116         uint16_t ifam_index;
 117         uint16_t _ifam_spare1;
 118         uint16_t ifam_len;
 119         uint16_t ifam_data_off;
 120         int32_t ifam_metric;
 121         struct  if_data ifam_data;
 122 };
 123
 124 #define SA_SIZE32(sa)                                           \
 125     (  (((struct sockaddr *)(sa))->sa_len == 0) ?               \
 126         sizeof(int)             :                               \
 127         1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) )
 128
 129 #endif /* COMPAT_FREEBSD32 */
 130
 131 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
 132
 133 /* NB: these are not modified */
 134 static struct   sockaddr route_src = { 2, PF_ROUTE, };
 135 static struct   sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
 136
 137 /* These are external hooks for CARP. */
 138 int     (*carp_get_vhid_p)(struct ifaddr *);
 139
 140 /*
 141  * Used by rtsock/raw_input callback code to decide whether to filter the update
 142  * notification to a socket bound to a particular FIB.
 143  */
 144 #define RTS_FILTER_FIB  M_PROTO8
 145
 146 typedef struct {
 147         int     ip_count;       /* attached w/ AF_INET */
 148         int     ip6_count;      /* attached w/ AF_INET6 */
 149         int     any_count;      /* total attached */
 150 } route_cb_t;
 151 VNET_DEFINE_STATIC(route_cb_t, route_cb);
 152 #define V_route_cb VNET(route_cb)
 153
 154 struct mtx rtsock_mtx;
 155 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
 156
 157 #define RTSOCK_LOCK()   mtx_lock(&rtsock_mtx)
 158 #define RTSOCK_UNLOCK() mtx_unlock(&rtsock_mtx)
 159 #define RTSOCK_LOCK_ASSERT()    mtx_assert(&rtsock_mtx, MA_OWNED)
 160
 161 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, "");
 162
 163 struct walkarg {
 164         int     w_tmemsize;
 165         int     w_op, w_arg;
 166         caddr_t w_tmem;
 167         struct sysctl_req *w_req;
 168 };
 169
 170 static void     rts_input(struct mbuf *m);
 171 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
 172 static int      rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
 173                         struct walkarg *w, int *plen);
 174 static int      rt_xaddrs(caddr_t cp, caddr_t cplim,
 175                         struct rt_addrinfo *rtinfo);
 176 static int      sysctl_dumpentry(struct radix_node *rn, void *vw);
 177 static int      sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh,
 178                         uint32_t weight, struct walkarg *w);
 179 static int      sysctl_iflist(int af, struct walkarg *w);
 180 static int      sysctl_ifmalist(int af, struct walkarg *w);
 181 static int      route_output(struct mbuf *m, struct socket *so, ...);
 182 static void     rt_getmetrics(const struct rtentry *rt,
 183                         const struct nhop_object *nh, struct rt_metrics *out);
 184 static void     rt_dispatch(struct mbuf *, sa_family_t);
 185 static int      handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
 186                         struct rt_msghdr *rtm, struct rib_cmd_info *rc);
 187 static int      update_rtm_from_rc(struct rt_addrinfo *info,
 188                         struct rt_msghdr **prtm, int alloc_len,
 189                         struct rib_cmd_info *rc, struct nhop_object *nh);
 190 static void     send_rtm_reply(struct socket *so, struct rt_msghdr *rtm,
 191                         struct mbuf *m, sa_family_t saf, u_int fibnum,
 192                         int rtm_errno);
 193 static int      can_export_rte(struct ucred *td_ucred, const struct rtentry *rt);
 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,
 216     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
 217     0, 0, sysctl_route_netisr_maxqlen, "I",
 218     "maximum routing socket dispatch queue length");
 219
 220 static void
 221 vnet_rts_init(void)
 222 {
 223         int tmp;
 224
 225         if (IS_DEFAULT_VNET(curvnet)) {
 226                 if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
 227                         rtsock_nh.nh_qlimit = tmp;
 228                 netisr_register(&rtsock_nh);
 229         }
 230 #ifdef VIMAGE
 231          else
 232                 netisr_register_vnet(&rtsock_nh);
 233 #endif
 234 }
 235 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
 236     vnet_rts_init, 0);
 237
 238 #ifdef VIMAGE
 239 static void
 240 vnet_rts_uninit(void)
 241 {
 242
 243         netisr_unregister_vnet(&rtsock_nh);
 244 }
 245 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
 246     vnet_rts_uninit, 0);
 247 #endif
 248
 249 static int
 250 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
 251     struct rawcb *rp)
 252 {
 253         int fibnum;
 254
 255         KASSERT(m != NULL, ("%s: m is NULL", __func__));
 256         KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
 257         KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
 258
 259         /* No filtering requested. */
 260         if ((m->m_flags & RTS_FILTER_FIB) == 0)
 261                 return (0);
 262
 263         /* Check if it is a rts and the fib matches the one of the socket. */
 264         fibnum = M_GETFIB(m);
 265         if (proto->sp_family != PF_ROUTE ||
 266             rp->rcb_socket == NULL ||
 267             rp->rcb_socket->so_fibnum == fibnum)
 268                 return (0);
 269
 270         /* Filtering requested and no match, the socket shall be skipped. */
 271         return (1);
 272 }
 273
 274 static void
 275 rts_input(struct mbuf *m)
 276 {
 277         struct sockproto route_proto;
 278         unsigned short *family;
 279         struct m_tag *tag;
 280
 281         route_proto.sp_family = PF_ROUTE;
 282         tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
 283         if (tag != NULL) {
 284                 family = (unsigned short *)(tag + 1);
 285                 route_proto.sp_protocol = *family;
 286                 m_tag_delete(m, tag);
 287         } else
 288                 route_proto.sp_protocol = 0;
 289
 290         raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
 291 }
 292
 293 /*
 294  * It really doesn't make any sense at all for this code to share much
 295  * with raw_usrreq.c, since its functionality is so restricted.  XXX
 296  */
 297 static void
 298 rts_abort(struct socket *so)
 299 {
 300
 301         raw_usrreqs.pru_abort(so);
 302 }
 303
 304 static void
 305 rts_close(struct socket *so)
 306 {
 307
 308         raw_usrreqs.pru_close(so);
 309 }
 310
 311 /* pru_accept is EOPNOTSUPP */
 312
 313 static int
 314 rts_attach(struct socket *so, int proto, struct thread *td)
 315 {
 316         struct rawcb *rp;
 317         int error;
 318
 319         KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
 320
 321         /* XXX */
 322         rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
 323
 324         so->so_pcb = (caddr_t)rp;
 325         so->so_fibnum = td->td_proc->p_fibnum;
 326         error = raw_attach(so, proto);
 327         rp = sotorawcb(so);
 328         if (error) {
 329                 so->so_pcb = NULL;
 330                 free(rp, M_PCB);
 331                 return error;
 332         }
 333         RTSOCK_LOCK();
 334         switch(rp->rcb_proto.sp_protocol) {
 335         case AF_INET:
 336                 V_route_cb.ip_count++;
 337                 break;
 338         case AF_INET6:
 339                 V_route_cb.ip6_count++;
 340                 break;
 341         }
 342         V_route_cb.any_count++;
 343         RTSOCK_UNLOCK();
 344         soisconnected(so);
 345         so->so_options |= SO_USELOOPBACK;
 346         return 0;
 347 }
 348
 349 static int
 350 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
 351 {
 352
 353         return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
 354 }
 355
 356 static int
 357 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
 358 {
 359
 360         return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
 361 }
 362
 363 /* pru_connect2 is EOPNOTSUPP */
 364 /* pru_control is EOPNOTSUPP */
 365
 366 static void
 367 rts_detach(struct socket *so)
 368 {
 369         struct rawcb *rp = sotorawcb(so);
 370
 371         KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
 372
 373         RTSOCK_LOCK();
 374         switch(rp->rcb_proto.sp_protocol) {
 375         case AF_INET:
 376                 V_route_cb.ip_count--;
 377                 break;
 378         case AF_INET6:
 379                 V_route_cb.ip6_count--;
 380                 break;
 381         }
 382         V_route_cb.any_count--;
 383         RTSOCK_UNLOCK();
 384         raw_usrreqs.pru_detach(so);
 385 }
 386
 387 static int
 388 rts_disconnect(struct socket *so)
 389 {
 390
 391         return (raw_usrreqs.pru_disconnect(so));
 392 }
 393
 394 /* pru_listen is EOPNOTSUPP */
 395
 396 static int
 397 rts_peeraddr(struct socket *so, struct sockaddr **nam)
 398 {
 399
 400         return (raw_usrreqs.pru_peeraddr(so, nam));
 401 }
 402
 403 /* pru_rcvd is EOPNOTSUPP */
 404 /* pru_rcvoob is EOPNOTSUPP */
 405
 406 static int
 407 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
 408          struct mbuf *control, struct thread *td)
 409 {
 410
 411         return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
 412 }
 413
 414 /* pru_sense is null */
 415
 416 static int
 417 rts_shutdown(struct socket *so)
 418 {
 419
 420         return (raw_usrreqs.pru_shutdown(so));
 421 }
 422
 423 static int
 424 rts_sockaddr(struct socket *so, struct sockaddr **nam)
 425 {
 426
 427         return (raw_usrreqs.pru_sockaddr(so, nam));
 428 }
 429
 430 static struct pr_usrreqs route_usrreqs = {
 431         .pru_abort =            rts_abort,
 432         .pru_attach =           rts_attach,
 433         .pru_bind =             rts_bind,
 434         .pru_connect =          rts_connect,
 435         .pru_detach =           rts_detach,
 436         .pru_disconnect =       rts_disconnect,
 437         .pru_peeraddr =         rts_peeraddr,
 438         .pru_send =             rts_send,
 439         .pru_shutdown =         rts_shutdown,
 440         .pru_sockaddr =         rts_sockaddr,
 441         .pru_close =            rts_close,
 442 };
 443
 444 #ifndef _SOCKADDR_UNION_DEFINED
 445 #define _SOCKADDR_UNION_DEFINED
 446 /*
 447  * The union of all possible address formats we handle.
 448  */
 449 union sockaddr_union {
 450         struct sockaddr         sa;
 451         struct sockaddr_in      sin;
 452         struct sockaddr_in6     sin6;
 453 };
 454 #endif /* _SOCKADDR_UNION_DEFINED */
 455
 456 static int
 457 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
 458     struct nhop_object *nh, union sockaddr_union *saun, struct ucred *cred)
 459 {
 460 #if defined(INET) || defined(INET6)
 461         struct epoch_tracker et;
 462 #endif
 463
 464         /* First, see if the returned address is part of the jail. */
 465         if (prison_if(cred, nh->nh_ifa->ifa_addr) == 0) {
 466                 info->rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
 467                 return (0);
 468         }
 469
 470         switch (info->rti_info[RTAX_DST]->sa_family) {
 471 #ifdef INET
 472         case AF_INET:
 473         {
 474                 struct in_addr ia;
 475                 struct ifaddr *ifa;
 476                 int found;
 477
 478                 found = 0;
 479                 /*
 480                  * Try to find an address on the given outgoing interface
 481                  * that belongs to the jail.
 482                  */
 483                 NET_EPOCH_ENTER(et);
 484                 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 485                         struct sockaddr *sa;
 486                         sa = ifa->ifa_addr;
 487                         if (sa->sa_family != AF_INET)
 488                                 continue;
 489                         ia = ((struct sockaddr_in *)sa)->sin_addr;
 490                         if (prison_check_ip4(cred, &ia) == 0) {
 491                                 found = 1;
 492                                 break;
 493                         }
 494                 }
 495                 NET_EPOCH_EXIT(et);
 496                 if (!found) {
 497                         /*
 498                          * As a last resort return the 'default' jail address.
 499                          */
 500                         ia = ((struct sockaddr_in *)nh->nh_ifa->ifa_addr)->
 501                             sin_addr;
 502                         if (prison_get_ip4(cred, &ia) != 0)
 503                                 return (ESRCH);
 504                 }
 505                 bzero(&saun->sin, sizeof(struct sockaddr_in));
 506                 saun->sin.sin_len = sizeof(struct sockaddr_in);
 507                 saun->sin.sin_family = AF_INET;
 508                 saun->sin.sin_addr.s_addr = ia.s_addr;
 509                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
 510                 break;
 511         }
 512 #endif
 513 #ifdef INET6
 514         case AF_INET6:
 515         {
 516                 struct in6_addr ia6;
 517                 struct ifaddr *ifa;
 518                 int found;
 519
 520                 found = 0;
 521                 /*
 522                  * Try to find an address on the given outgoing interface
 523                  * that belongs to the jail.
 524                  */
 525                 NET_EPOCH_ENTER(et);
 526                 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 527                         struct sockaddr *sa;
 528                         sa = ifa->ifa_addr;
 529                         if (sa->sa_family != AF_INET6)
 530                                 continue;
 531                         bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
 532                             &ia6, sizeof(struct in6_addr));
 533                         if (prison_check_ip6(cred, &ia6) == 0) {
 534                                 found = 1;
 535                                 break;
 536                         }
 537                 }
 538                 NET_EPOCH_EXIT(et);
 539                 if (!found) {
 540                         /*
 541                          * As a last resort return the 'default' jail address.
 542                          */
 543                         ia6 = ((struct sockaddr_in6 *)nh->nh_ifa->ifa_addr)->
 544                             sin6_addr;
 545                         if (prison_get_ip6(cred, &ia6) != 0)
 546                                 return (ESRCH);
 547                 }
 548                 bzero(&saun->sin6, sizeof(struct sockaddr_in6));
 549                 saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
 550                 saun->sin6.sin6_family = AF_INET6;
 551                 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
 552                 if (sa6_recoverscope(&saun->sin6) != 0)
 553                         return (ESRCH);
 554                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
 555                 break;
 556         }
 557 #endif
 558         default:
 559                 return (ESRCH);
 560         }
 561         return (0);
 562 }
 563
 564 /*
 565  * Fills in @info based on userland-provided @rtm message.
 566  *
 567  * Returns 0 on success.
 568  */
 569 static int
 570 fill_addrinfo(struct rt_msghdr *rtm, int len, u_int fibnum, struct rt_addrinfo *info)
 571 {
 572         int error;
 573         sa_family_t saf;
 574
 575         rtm->rtm_pid = curproc->p_pid;
 576         info->rti_addrs = rtm->rtm_addrs;
 577
 578         info->rti_mflags = rtm->rtm_inits;
 579         info->rti_rmx = &rtm->rtm_rmx;
 580
 581         /*
 582          * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
 583          * link-local address because rtrequest requires addresses with
 584          * embedded scope id.
 585          */
 586         if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info))
 587                 return (EINVAL);
 588
 589         if (rtm->rtm_flags & RTF_RNH_LOCKED)
 590                 return (EINVAL);
 591         info->rti_flags = rtm->rtm_flags;
 592         if (info->rti_info[RTAX_DST] == NULL ||
 593             info->rti_info[RTAX_DST]->sa_family >= AF_MAX ||
 594             (info->rti_info[RTAX_GATEWAY] != NULL &&
 595              info->rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
 596                 return (EINVAL);
 597         saf = info->rti_info[RTAX_DST]->sa_family;
 598         /*
 599          * Verify that the caller has the appropriate privilege; RTM_GET
 600          * is the only operation the non-superuser is allowed.
 601          */
 602         if (rtm->rtm_type != RTM_GET) {
 603                 error = priv_check(curthread, PRIV_NET_ROUTE);
 604                 if (error != 0)
 605                         return (error);
 606         }
 607
 608         /*
 609          * The given gateway address may be an interface address.
 610          * For example, issuing a "route change" command on a route
 611          * entry that was created from a tunnel, and the gateway
 612          * address given is the local end point. In this case the
 613          * RTF_GATEWAY flag must be cleared or the destination will
 614          * not be reachable even though there is no error message.
 615          */
 616         if (info->rti_info[RTAX_GATEWAY] != NULL &&
 617             info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
 618                 struct rt_addrinfo ginfo;
 619                 struct sockaddr *gdst;
 620                 struct sockaddr_storage ss;
 621
 622                 bzero(&ginfo, sizeof(ginfo));
 623                 bzero(&ss, sizeof(ss));
 624                 ss.ss_len = sizeof(ss);
 625
 626                 ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss;
 627                 gdst = info->rti_info[RTAX_GATEWAY];
 628
 629                 /*
 630                  * A host route through the loopback interface is
 631                  * installed for each interface adddress. In pre 8.0
 632                  * releases the interface address of a PPP link type
 633                  * is not reachable locally. This behavior is fixed as
 634                  * part of the new L2/L3 redesign and rewrite work. The
 635                  * signature of this interface address route is the
 636                  * AF_LINK sa_family type of the gateway, and the
 637                  * rt_ifp has the IFF_LOOPBACK flag set.
 638                  */
 639                 if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) {
 640                         if (ss.ss_family == AF_LINK &&
 641                             ginfo.rti_ifp->if_flags & IFF_LOOPBACK) {
 642                                 info->rti_flags &= ~RTF_GATEWAY;
 643                                 info->rti_flags |= RTF_GWFLAG_COMPAT;
 644                         }
 645                         rib_free_info(&ginfo);
 646                 }
 647         }
 648
 649         return (0);
 650 }
 651
 652 static struct nhop_object *
 653 select_nhop(struct nhop_object *nh, const struct sockaddr *gw)
 654 {
 655         if (!NH_IS_NHGRP(nh))
 656                 return (nh);
 657 #ifdef ROUTE_MPATH
 658         struct weightened_nhop *wn;
 659         uint32_t num_nhops;
 660         wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
 661         if (gw == NULL)
 662                 return (wn[0].nh);
 663         for (int i = 0; i < num_nhops; i++) {
 664                 if (match_nhop_gw(wn[i].nh, gw))
 665                         return (wn[i].nh);
 666         }
 667 #endif
 668         return (NULL);
 669 }
 670
 671 /*
 672  * Handles RTM_GET message from routing socket, returning matching rt.
 673  *
 674  * Returns:
 675  * 0 on success, with locked and referenced matching rt in @rt_nrt
 676  * errno of failure
 677  */
 678 static int
 679 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
 680     struct rt_msghdr *rtm, struct rib_cmd_info *rc)
 681 {
 682         RIB_RLOCK_TRACKER;
 683         struct rib_head *rnh;
 684         struct nhop_object *nh;
 685         sa_family_t saf;
 686
 687         saf = info->rti_info[RTAX_DST]->sa_family;
 688
 689         rnh = rt_tables_get_rnh(fibnum, saf);
 690         if (rnh == NULL)
 691                 return (EAFNOSUPPORT);
 692
 693         RIB_RLOCK(rnh);
 694
 695         if (info->rti_info[RTAX_NETMASK] == NULL) {
 696                 /*
 697                  * Provide longest prefix match for
 698                  * address lookup (no mask).
 699                  * 'route -n get addr'
 700                  */
 701                 rc->rc_rt = (struct rtentry *) rnh->rnh_matchaddr(
 702                     info->rti_info[RTAX_DST], &rnh->head);
 703         } else
 704                 rc->rc_rt = (struct rtentry *) rnh->rnh_lookup(
 705                     info->rti_info[RTAX_DST],
 706                     info->rti_info[RTAX_NETMASK], &rnh->head);
 707
 708         if (rc->rc_rt == NULL) {
 709                 RIB_RUNLOCK(rnh);
 710                 return (ESRCH);
 711         }
 712
 713         nh = select_nhop(rc->rc_rt->rt_nhop, info->rti_info[RTAX_GATEWAY]);
 714         if (nh == NULL) {
 715                 RIB_RUNLOCK(rnh);
 716                 return (ESRCH);
 717         }
 718         /*
 719          * If performing proxied L2 entry insertion, and
 720          * the actual PPP host entry is found, perform
 721          * another search to retrieve the prefix route of
 722          * the local end point of the PPP link.
 723          * TODO: move this logic to userland.
 724          */
 725         if (rtm->rtm_flags & RTF_ANNOUNCE) {
 726                 struct sockaddr laddr;
 727                 struct nhop_object *nh;
 728
 729                 nh = rc->rc_rt->rt_nhop;
 730                 if (nh->nh_ifp != NULL &&
 731                     nh->nh_ifp->if_type == IFT_PROPVIRTUAL) {
 732                         struct ifaddr *ifa;
 733
 734                         ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1,
 735                                         RT_ALL_FIBS);
 736                         if (ifa != NULL)
 737                                 rt_maskedcopy(ifa->ifa_addr,
 738                                               &laddr,
 739                                               ifa->ifa_netmask);
 740                 } else
 741                         rt_maskedcopy(nh->nh_ifa->ifa_addr,
 742                                       &laddr,
 743                                       nh->nh_ifa->ifa_netmask);
 744                 /*
 745                  * refactor rt and no lock operation necessary
 746                  */
 747                 rc->rc_rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr,
 748                     &rnh->head);
 749                 if (rc->rc_rt == NULL) {
 750                         RIB_RUNLOCK(rnh);
 751                         return (ESRCH);
 752                 }
 753                 nh = select_nhop(rc->rc_rt->rt_nhop, info->rti_info[RTAX_GATEWAY]);
 754                 if (nh == NULL) {
 755                         RIB_RUNLOCK(rnh);
 756                         return (ESRCH);
 757                 }
 758         }
 759         rc->rc_nh_new = nh;
 760         rc->rc_nh_weight = rc->rc_rt->rt_weight;
 761         RIB_RUNLOCK(rnh);
 762
 763         return (0);
 764 }
 765
 766 /*
 767  * Update sockaddrs, flags, etc in @prtm based on @rc data.
 768  * rtm can be reallocated.
 769  *
 770  * Returns 0 on success, along with pointer to (potentially reallocated)
 771  *  rtm.
 772  *
 773  */
 774 static int
 775 update_rtm_from_rc(struct rt_addrinfo *info, struct rt_msghdr **prtm,
 776     int alloc_len, struct rib_cmd_info *rc, struct nhop_object *nh)
 777 {
 778         struct sockaddr_storage netmask_ss;
 779         struct walkarg w;
 780         union sockaddr_union saun;
 781         struct rt_msghdr *rtm, *orig_rtm = NULL;
 782         struct ifnet *ifp;
 783         int error, len;
 784
 785         rtm = *prtm;
 786
 787         info->rti_info[RTAX_DST] = rt_key(rc->rc_rt);
 788         info->rti_info[RTAX_GATEWAY] = &nh->gw_sa;
 789         info->rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rc->rc_rt),
 790             rt_mask(rc->rc_rt), &netmask_ss);
 791         info->rti_info[RTAX_GENMASK] = 0;
 792         ifp = nh->nh_ifp;
 793         if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
 794                 if (ifp) {
 795                         info->rti_info[RTAX_IFP] =
 796                             ifp->if_addr->ifa_addr;
 797                         error = rtm_get_jailed(info, ifp, nh,
 798                             &saun, curthread->td_ucred);
 799                         if (error != 0)
 800                                 return (error);
 801                         if (ifp->if_flags & IFF_POINTOPOINT)
 802                                 info->rti_info[RTAX_BRD] =
 803                                     nh->nh_ifa->ifa_dstaddr;
 804                         rtm->rtm_index = ifp->if_index;
 805                 } else {
 806                         info->rti_info[RTAX_IFP] = NULL;
 807                         info->rti_info[RTAX_IFA] = NULL;
 808                 }
 809         } else if (ifp != NULL)
 810                 rtm->rtm_index = ifp->if_index;
 811
 812         /* Check if we need to realloc storage */
 813         rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len);
 814         if (len > alloc_len) {
 815                 struct rt_msghdr *tmp_rtm;
 816
 817                 tmp_rtm = malloc(len, M_TEMP, M_NOWAIT);
 818                 if (tmp_rtm == NULL)
 819                         return (ENOBUFS);
 820                 bcopy(rtm, tmp_rtm, rtm->rtm_msglen);
 821                 orig_rtm = rtm;
 822                 rtm = tmp_rtm;
 823                 alloc_len = len;
 824
 825                 /*
 826                  * Delay freeing original rtm as info contains
 827                  * data referencing it.
 828                  */
 829         }
 830
 831         w.w_tmem = (caddr_t)rtm;
 832         w.w_tmemsize = alloc_len;
 833         rtsock_msg_buffer(rtm->rtm_type, info, &w, &len);
 834
 835         rtm->rtm_flags = rc->rc_rt->rte_flags | nhop_get_rtflags(nh);
 836         if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
 837                 rtm->rtm_flags = RTF_GATEWAY |
 838                         (rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
 839         rt_getmetrics(rc->rc_rt, nh, &rtm->rtm_rmx);
 840         rtm->rtm_rmx.rmx_weight = rc->rc_nh_weight;
 841         rtm->rtm_addrs = info->rti_addrs;
 842
 843         if (orig_rtm != NULL)
 844                 free(orig_rtm, M_TEMP);
 845         *prtm = rtm;
 846
 847         return (0);
 848 }
 849
 850 #ifdef ROUTE_MPATH
 851 static void
 852 save_del_notification(struct rib_cmd_info *rc, void *_cbdata)
 853 {
 854         struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
 855
 856         if (rc->rc_cmd == RTM_DELETE)
 857                 *rc_new = *rc;
 858 }
 859
 860 static void
 861 save_add_notification(struct rib_cmd_info *rc, void *_cbdata)
 862 {
 863         struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
 864
 865         if (rc->rc_cmd == RTM_ADD)
 866                 *rc_new = *rc;
 867 }
 868 #endif
 869
 870 /*ARGSUSED*/
 871 static int
 872 route_output(struct mbuf *m, struct socket *so, ...)
 873 {
 874         struct rt_msghdr *rtm = NULL;
 875         struct rtentry *rt = NULL;
 876         struct rt_addrinfo info;
 877         struct epoch_tracker et;
 878 #ifdef INET6
 879         struct sockaddr_storage ss;
 880         struct sockaddr_in6 *sin6;
 881         int i, rti_need_deembed = 0;
 882 #endif
 883         int alloc_len = 0, len, error = 0, fibnum;
 884         sa_family_t saf = AF_UNSPEC;
 885         struct walkarg w;
 886         struct rib_cmd_info rc;
 887         struct nhop_object *nh;
 888
 889         fibnum = so->so_fibnum;
 890 #define senderr(e) { error = e; goto flush;}
 891         if (m == NULL || ((m->m_len < sizeof(long)) &&
 892                        (m = m_pullup(m, sizeof(long))) == NULL))
 893                 return (ENOBUFS);
 894         if ((m->m_flags & M_PKTHDR) == 0)
 895                 panic("route_output");
 896         NET_EPOCH_ENTER(et);
 897         len = m->m_pkthdr.len;
 898         if (len < sizeof(*rtm) ||
 899             len != mtod(m, struct rt_msghdr *)->rtm_msglen)
 900                 senderr(EINVAL);
 901
 902         /*
 903          * Most of current messages are in range 200-240 bytes,
 904          * minimize possible re-allocation on reply using larger size
 905          * buffer aligned on 1k boundaty.
 906          */
 907         alloc_len = roundup2(len, 1024);
 908         if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL)
 909                 senderr(ENOBUFS);
 910
 911         m_copydata(m, 0, len, (caddr_t)rtm);
 912         bzero(&info, sizeof(info));
 913         bzero(&w, sizeof(w));
 914         nh = NULL;
 915
 916         if (rtm->rtm_version != RTM_VERSION) {
 917                 /* Do not touch message since format is unknown */
 918                 free(rtm, M_TEMP);
 919                 rtm = NULL;
 920                 senderr(EPROTONOSUPPORT);
 921         }
 922
 923         /*
 924          * Starting from here, it is possible
 925          * to alter original message and insert
 926          * caller PID and error value.
 927          */
 928
 929         if ((error = fill_addrinfo(rtm, len, fibnum, &info)) != 0) {
 930                 senderr(error);
 931         }
 932
 933         saf = info.rti_info[RTAX_DST]->sa_family;
 934
 935         /* support for new ARP code */
 936         if (rtm->rtm_flags & RTF_LLDATA) {
 937                 error = lla_rt_output(rtm, &info);
 938 #ifdef INET6
 939                 if (error == 0)
 940                         rti_need_deembed = 1;
 941 #endif
 942                 goto flush;
 943         }
 944
 945         switch (rtm->rtm_type) {
 946         case RTM_ADD:
 947         case RTM_CHANGE:
 948                 if (rtm->rtm_type == RTM_ADD) {
 949                         if (info.rti_info[RTAX_GATEWAY] == NULL)
 950                                 senderr(EINVAL);
 951                 }
 952                 error = rib_action(fibnum, rtm->rtm_type, &info, &rc);
 953                 if (error == 0) {
 954 #ifdef INET6
 955                         rti_need_deembed = 1;
 956 #endif
 957 #ifdef ROUTE_MPATH
 958                         if (NH_IS_NHGRP(rc.rc_nh_new) ||
 959                             (rc.rc_nh_old && NH_IS_NHGRP(rc.rc_nh_old))) {
 960                                 struct rib_cmd_info rc_simple = {};
 961                                 rib_decompose_notification(&rc,
 962                                     save_add_notification, (void *)&rc_simple);
 963                                 rc = rc_simple;
 964                         }
 965 #endif
 966                         nh = rc.rc_nh_new;
 967                         rtm->rtm_index = nh->nh_ifp->if_index;
 968                 }
 969                 break;
 970
 971         case RTM_DELETE:
 972                 error = rib_action(fibnum, RTM_DELETE, &info, &rc);
 973                 if (error == 0) {
 974 #ifdef ROUTE_MPATH
 975                         if (NH_IS_NHGRP(rc.rc_nh_old) ||
 976                             (rc.rc_nh_new && NH_IS_NHGRP(rc.rc_nh_new))) {
 977                                 struct rib_cmd_info rc_simple = {};
 978                                 rib_decompose_notification(&rc,
 979                                     save_del_notification, (void *)&rc_simple);
 980                                 rc = rc_simple;
 981                         }
 982 #endif
 983                         nh = rc.rc_nh_old;
 984                         goto report;
 985                 }
 986 #ifdef INET6
 987                 /* rt_msg2() will not be used when RTM_DELETE fails. */
 988                 rti_need_deembed = 1;
 989 #endif
 990                 break;
 991
 992         case RTM_GET:
 993                 error = handle_rtm_get(&info, fibnum, rtm, &rc);
 994                 if (error != 0)
 995                         senderr(error);
 996                 nh = rc.rc_nh_new;
 997
 998 report:
 999                 if (!can_export_rte(curthread->td_ucred, rc.rc_rt)) {
1000                         senderr(ESRCH);
1001                 }
1002
1003                 error = update_rtm_from_rc(&info, &rtm, alloc_len, &rc, nh);
1004                 /*
1005                  * Note that some sockaddr pointers may have changed to
1006                  * point to memory outsize @rtm. Some may be pointing
1007                  * to the on-stack variables.
1008                  * Given that, any pointer in @info CANNOT BE USED.
1009                  */
1010
1011                 /*
1012                  * scopeid deembedding has been performed while
1013                  * writing updated rtm in rtsock_msg_buffer().
1014                  * With that in mind, skip deembedding procedure below.
1015                  */
1016 #ifdef INET6
1017                 rti_need_deembed = 0;
1018 #endif
1019                 if (error != 0)
1020                         senderr(error);
1021                 break;
1022
1023         default:
1024                 senderr(EOPNOTSUPP);
1025         }
1026
1027 flush:
1028         NET_EPOCH_EXIT(et);
1029         rt = NULL;
1030
1031 #ifdef INET6
1032         if (rtm != NULL) {
1033                 if (rti_need_deembed) {
1034                         /* sin6_scope_id is recovered before sending rtm. */
1035                         sin6 = (struct sockaddr_in6 *)&ss;
1036                         for (i = 0; i < RTAX_MAX; i++) {
1037                                 if (info.rti_info[i] == NULL)
1038                                         continue;
1039                                 if (info.rti_info[i]->sa_family != AF_INET6)
1040                                         continue;
1041                                 bcopy(info.rti_info[i], sin6, sizeof(*sin6));
1042                                 if (sa6_recoverscope(sin6) == 0)
1043                                         bcopy(sin6, info.rti_info[i],
1044                                                     sizeof(*sin6));
1045                         }
1046                 }
1047         }
1048 #endif
1049         send_rtm_reply(so, rtm, m, saf, fibnum, error);
1050
1051         return (error);
1052 }
1053
1054 /*
1055  * Sends the prepared reply message in @rtm to all rtsock clients.
1056  * Frees @m and @rtm.
1057  *
1058  */
1059 static void
1060 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m,
1061     sa_family_t saf, u_int fibnum, int rtm_errno)
1062 {
1063         struct rawcb *rp = NULL;
1064
1065         /*
1066          * Check to see if we don't want our own messages.
1067          */
1068         if ((so->so_options & SO_USELOOPBACK) == 0) {
1069                 if (V_route_cb.any_count <= 1) {
1070                         if (rtm != NULL)
1071                                 free(rtm, M_TEMP);
1072                         m_freem(m);
1073                         return;
1074                 }
1075                 /* There is another listener, so construct message */
1076                 rp = sotorawcb(so);
1077         }
1078
1079         if (rtm != NULL) {
1080                 if (rtm_errno!= 0)
1081                         rtm->rtm_errno = rtm_errno;
1082                 else
1083                         rtm->rtm_flags |= RTF_DONE;
1084
1085                 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
1086                 if (m->m_pkthdr.len < rtm->rtm_msglen) {
1087                         m_freem(m);
1088                         m = NULL;
1089                 } else if (m->m_pkthdr.len > rtm->rtm_msglen)
1090                         m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
1091
1092                 free(rtm, M_TEMP);
1093         }
1094         if (m != NULL) {
1095                 M_SETFIB(m, fibnum);
1096                 m->m_flags |= RTS_FILTER_FIB;
1097                 if (rp) {
1098                         /*
1099                          * XXX insure we don't get a copy by
1100                          * invalidating our protocol
1101                          */
1102                         unsigned short family = rp->rcb_proto.sp_family;
1103                         rp->rcb_proto.sp_family = 0;
1104                         rt_dispatch(m, saf);
1105                         rp->rcb_proto.sp_family = family;
1106                 } else
1107                         rt_dispatch(m, saf);
1108         }
1109 }
1110
1111 static void
1112 rt_getmetrics(const struct rtentry *rt, const struct nhop_object *nh,
1113     struct rt_metrics *out)
1114 {
1115
1116         bzero(out, sizeof(*out));
1117         out->rmx_mtu = nh->nh_mtu;
1118         out->rmx_weight = rt->rt_weight;
1119         out->rmx_nhidx = nhop_get_idx(nh);
1120         /* Kernel -> userland timebase conversion. */
1121         out->rmx_expire = rt->rt_expire ?
1122             rt->rt_expire - time_uptime + time_second : 0;
1123 }
1124
1125 /*
1126  * Extract the addresses of the passed sockaddrs.
1127  * Do a little sanity checking so as to avoid bad memory references.
1128  * This data is derived straight from userland.
1129  */
1130 static int
1131 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1132 {
1133         struct sockaddr *sa;
1134         int i;
1135
1136         for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1137                 if ((rtinfo->rti_addrs & (1 << i)) == 0)
1138                         continue;
1139                 sa = (struct sockaddr *)cp;
1140                 /*
1141                  * It won't fit.
1142                  */
1143                 if (cp + sa->sa_len > cplim)
1144                         return (EINVAL);
1145                 /*
1146                  * there are no more.. quit now
1147                  * If there are more bits, they are in error.
1148                  * I've seen this. route(1) can evidently generate these.
1149                  * This causes kernel to core dump.
1150                  * for compatibility, If we see this, point to a safe address.
1151                  */
1152                 if (sa->sa_len == 0) {
1153                         rtinfo->rti_info[i] = &sa_zero;
1154                         return (0); /* should be EINVAL but for compat */
1155                 }
1156                 /* accept it */
1157 #ifdef INET6
1158                 if (sa->sa_family == AF_INET6)
1159                         sa6_embedscope((struct sockaddr_in6 *)sa,
1160                             V_ip6_use_defzone);
1161 #endif
1162                 rtinfo->rti_info[i] = sa;
1163                 cp += SA_SIZE(sa);
1164         }
1165         return (0);
1166 }
1167
1168 /*
1169  * Fill in @dmask with valid netmask leaving original @smask
1170  * intact. Mostly used with radix netmasks.
1171  */
1172 struct sockaddr *
1173 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask,
1174     struct sockaddr_storage *dmask)
1175 {
1176         if (dst == NULL || smask == NULL)
1177                 return (NULL);
1178
1179         memset(dmask, 0, dst->sa_len);
1180         memcpy(dmask, smask, smask->sa_len);
1181         dmask->ss_len = dst->sa_len;
1182         dmask->ss_family = dst->sa_family;
1183
1184         return ((struct sockaddr *)dmask);
1185 }
1186
1187 /*
1188  * Writes information related to @rtinfo object to newly-allocated mbuf.
1189  * Assumes MCLBYTES is enough to construct any message.
1190  * Used for OS notifications of vaious events (if/ifa announces,etc)
1191  *
1192  * Returns allocated mbuf or NULL on failure.
1193  */
1194 static struct mbuf *
1195 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1196 {
1197         struct rt_msghdr *rtm;
1198         struct mbuf *m;
1199         int i;
1200         struct sockaddr *sa;
1201 #ifdef INET6
1202         struct sockaddr_storage ss;
1203         struct sockaddr_in6 *sin6;
1204 #endif
1205         int len, dlen;
1206
1207         switch (type) {
1208         case RTM_DELADDR:
1209         case RTM_NEWADDR:
1210                 len = sizeof(struct ifa_msghdr);
1211                 break;
1212
1213         case RTM_DELMADDR:
1214         case RTM_NEWMADDR:
1215                 len = sizeof(struct ifma_msghdr);
1216                 break;
1217
1218         case RTM_IFINFO:
1219                 len = sizeof(struct if_msghdr);
1220                 break;
1221
1222         case RTM_IFANNOUNCE:
1223         case RTM_IEEE80211:
1224                 len = sizeof(struct if_announcemsghdr);
1225                 break;
1226
1227         default:
1228                 len = sizeof(struct rt_msghdr);
1229         }
1230
1231         /* XXXGL: can we use MJUMPAGESIZE cluster here? */
1232         KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1233         if (len > MHLEN)
1234                 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1235         else
1236                 m = m_gethdr(M_NOWAIT, MT_DATA);
1237         if (m == NULL)
1238                 return (m);
1239
1240         m->m_pkthdr.len = m->m_len = len;
1241         rtm = mtod(m, struct rt_msghdr *);
1242         bzero((caddr_t)rtm, len);
1243         for (i = 0; i < RTAX_MAX; i++) {
1244                 if ((sa = rtinfo->rti_info[i]) == NULL)
1245                         continue;
1246                 rtinfo->rti_addrs |= (1 << i);
1247                 dlen = SA_SIZE(sa);
1248 #ifdef INET6
1249                 if (sa->sa_family == AF_INET6) {
1250                         sin6 = (struct sockaddr_in6 *)&ss;
1251                         bcopy(sa, sin6, sizeof(*sin6));
1252                         if (sa6_recoverscope(sin6) == 0)
1253                                 sa = (struct sockaddr *)sin6;
1254                 }
1255 #endif
1256                 m_copyback(m, len, dlen, (caddr_t)sa);
1257                 len += dlen;
1258         }
1259         if (m->m_pkthdr.len != len) {
1260                 m_freem(m);
1261                 return (NULL);
1262         }
1263         rtm->rtm_msglen = len;
1264         rtm->rtm_version = RTM_VERSION;
1265         rtm->rtm_type = type;
1266         return (m);
1267 }
1268
1269 /*
1270  * Writes information related to @rtinfo object to preallocated buffer.
1271  * Stores needed size in @plen. If @w is NULL, calculates size without
1272  * writing.
1273  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1274  *
1275  * Returns 0 on success.
1276  *
1277  */
1278 static int
1279 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1280 {
1281         int i;
1282         int len, buflen = 0, dlen;
1283         caddr_t cp = NULL;
1284         struct rt_msghdr *rtm = NULL;
1285 #ifdef INET6
1286         struct sockaddr_storage ss;
1287         struct sockaddr_in6 *sin6;
1288 #endif
1289 #ifdef COMPAT_FREEBSD32
1290         bool compat32 = false;
1291 #endif
1292
1293         switch (type) {
1294         case RTM_DELADDR:
1295         case RTM_NEWADDR:
1296                 if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1297 #ifdef COMPAT_FREEBSD32
1298                         if (w->w_req->flags & SCTL_MASK32) {
1299                                 len = sizeof(struct ifa_msghdrl32);
1300                                 compat32 = true;
1301                         } else
1302 #endif
1303                                 len = sizeof(struct ifa_msghdrl);
1304                 } else
1305                         len = sizeof(struct ifa_msghdr);
1306                 break;
1307
1308         case RTM_IFINFO:
1309 #ifdef COMPAT_FREEBSD32
1310                 if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1311                         if (w->w_op == NET_RT_IFLISTL)
1312                                 len = sizeof(struct if_msghdrl32);
1313                         else
1314                                 len = sizeof(struct if_msghdr32);
1315                         compat32 = true;
1316                         break;
1317                 }
1318 #endif
1319                 if (w != NULL && w->w_op == NET_RT_IFLISTL)
1320                         len = sizeof(struct if_msghdrl);
1321                 else
1322                         len = sizeof(struct if_msghdr);
1323                 break;
1324
1325         case RTM_NEWMADDR:
1326                 len = sizeof(struct ifma_msghdr);
1327                 break;
1328
1329         default:
1330                 len = sizeof(struct rt_msghdr);
1331         }
1332
1333         if (w != NULL) {
1334                 rtm = (struct rt_msghdr *)w->w_tmem;
1335                 buflen = w->w_tmemsize - len;
1336                 cp = (caddr_t)w->w_tmem + len;
1337         }
1338
1339         rtinfo->rti_addrs = 0;
1340         for (i = 0; i < RTAX_MAX; i++) {
1341                 struct sockaddr *sa;
1342
1343                 if ((sa = rtinfo->rti_info[i]) == NULL)
1344                         continue;
1345                 rtinfo->rti_addrs |= (1 << i);
1346 #ifdef COMPAT_FREEBSD32
1347                 if (compat32)
1348                         dlen = SA_SIZE32(sa);
1349                 else
1350 #endif
1351                         dlen = SA_SIZE(sa);
1352                 if (cp != NULL && buflen >= dlen) {
1353 #ifdef INET6
1354                         if (sa->sa_family == AF_INET6) {
1355                                 sin6 = (struct sockaddr_in6 *)&ss;
1356                                 bcopy(sa, sin6, sizeof(*sin6));
1357                                 if (sa6_recoverscope(sin6) == 0)
1358                                         sa = (struct sockaddr *)sin6;
1359                         }
1360 #endif
1361                         bcopy((caddr_t)sa, cp, (unsigned)dlen);
1362                         cp += dlen;
1363                         buflen -= dlen;
1364                 } else if (cp != NULL) {
1365                         /*
1366                          * Buffer too small. Count needed size
1367                          * and return with error.
1368                          */
1369                         cp = NULL;
1370                 }
1371
1372                 len += dlen;
1373         }
1374
1375         if (cp != NULL) {
1376                 dlen = ALIGN(len) - len;
1377                 if (buflen < dlen)
1378                         cp = NULL;
1379                 else {
1380                         bzero(cp, dlen);
1381                         cp += dlen;
1382                         buflen -= dlen;
1383                 }
1384         }
1385         len = ALIGN(len);
1386
1387         if (cp != NULL) {
1388                 /* fill header iff buffer is large enough */
1389                 rtm->rtm_version = RTM_VERSION;
1390                 rtm->rtm_type = type;
1391                 rtm->rtm_msglen = len;
1392         }
1393
1394         *plen = len;
1395
1396         if (w != NULL && cp == NULL)
1397                 return (ENOBUFS);
1398
1399         return (0);
1400 }
1401
1402 /*
1403  * This routine is called to generate a message from the routing
1404  * socket indicating that a redirect has occurred, a routing lookup
1405  * has failed, or that a protocol has detected timeouts to a particular
1406  * destination.
1407  */
1408 void
1409 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1410     int fibnum)
1411 {
1412         struct rt_msghdr *rtm;
1413         struct mbuf *m;
1414         struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1415
1416         if (V_route_cb.any_count == 0)
1417                 return;
1418         m = rtsock_msg_mbuf(type, rtinfo);
1419         if (m == NULL)
1420                 return;
1421
1422         if (fibnum != RT_ALL_FIBS) {
1423                 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1424                     "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1425                 M_SETFIB(m, fibnum);
1426                 m->m_flags |= RTS_FILTER_FIB;
1427         }
1428
1429         rtm = mtod(m, struct rt_msghdr *);
1430         rtm->rtm_flags = RTF_DONE | flags;
1431         rtm->rtm_errno = error;
1432         rtm->rtm_addrs = rtinfo->rti_addrs;
1433         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1434 }
1435
1436 void
1437 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1438 {
1439
1440         rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1441 }
1442
1443 /*
1444  * This routine is called to generate a message from the routing
1445  * socket indicating that the status of a network interface has changed.
1446  */
1447 void
1448 rt_ifmsg(struct ifnet *ifp)
1449 {
1450         struct if_msghdr *ifm;
1451         struct mbuf *m;
1452         struct rt_addrinfo info;
1453
1454         if (V_route_cb.any_count == 0)
1455                 return;
1456         bzero((caddr_t)&info, sizeof(info));
1457         m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1458         if (m == NULL)
1459                 return;
1460         ifm = mtod(m, struct if_msghdr *);
1461         ifm->ifm_index = ifp->if_index;
1462         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1463         if_data_copy(ifp, &ifm->ifm_data);
1464         ifm->ifm_addrs = 0;
1465         rt_dispatch(m, AF_UNSPEC);
1466 }
1467
1468 /*
1469  * Announce interface address arrival/withdraw.
1470  * Please do not call directly, use rt_addrmsg().
1471  * Assume input data to be valid.
1472  * Returns 0 on success.
1473  */
1474 int
1475 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1476 {
1477         struct rt_addrinfo info;
1478         struct sockaddr *sa;
1479         int ncmd;
1480         struct mbuf *m;
1481         struct ifa_msghdr *ifam;
1482         struct ifnet *ifp = ifa->ifa_ifp;
1483         struct sockaddr_storage ss;
1484
1485         if (V_route_cb.any_count == 0)
1486                 return (0);
1487
1488         ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1489
1490         bzero((caddr_t)&info, sizeof(info));
1491         info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1492         info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1493         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1494             info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss);
1495         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1496         if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1497                 return (ENOBUFS);
1498         ifam = mtod(m, struct ifa_msghdr *);
1499         ifam->ifam_index = ifp->if_index;
1500         ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1501         ifam->ifam_flags = ifa->ifa_flags;
1502         ifam->ifam_addrs = info.rti_addrs;
1503
1504         if (fibnum != RT_ALL_FIBS) {
1505                 M_SETFIB(m, fibnum);
1506                 m->m_flags |= RTS_FILTER_FIB;
1507         }
1508
1509         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1510
1511         return (0);
1512 }
1513
1514 /*
1515  * Announce route addition/removal to rtsock based on @rt data.
1516  * Callers are advives to use rt_routemsg() instead of using this
1517  *  function directly.
1518  * Assume @rt data is consistent.
1519  *
1520  * Returns 0 on success.
1521  */
1522 int
1523 rtsock_routemsg(int cmd, struct rtentry *rt, struct ifnet *ifp, int rti_addrs,
1524     int fibnum)
1525 {
1526         struct sockaddr_storage ss;
1527         struct rt_addrinfo info;
1528         struct nhop_object *nh;
1529
1530         if (V_route_cb.any_count == 0)
1531                 return (0);
1532
1533         nh = rt->rt_nhop;
1534         bzero((caddr_t)&info, sizeof(info));
1535         info.rti_info[RTAX_DST] = rt_key(rt);
1536         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt), rt_mask(rt), &ss);
1537         info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
1538         info.rti_flags = rt->rte_flags | nhop_get_rtflags(nh);
1539         info.rti_ifp = ifp;
1540
1541         return (rtsock_routemsg_info(cmd, &info, fibnum));
1542 }
1543
1544 int
1545 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
1546 {
1547         struct rt_msghdr *rtm;
1548         struct sockaddr *sa;
1549         struct mbuf *m;
1550
1551         if (V_route_cb.any_count == 0)
1552                 return (0);
1553
1554         if (info->rti_flags & RTF_HOST)
1555                 info->rti_info[RTAX_NETMASK] = NULL;
1556
1557         m = rtsock_msg_mbuf(cmd, info);
1558         if (m == NULL)
1559                 return (ENOBUFS);
1560
1561         if (fibnum != RT_ALL_FIBS) {
1562                 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1563                     "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1564                 M_SETFIB(m, fibnum);
1565                 m->m_flags |= RTS_FILTER_FIB;
1566         }
1567
1568         rtm = mtod(m, struct rt_msghdr *);
1569         rtm->rtm_addrs = info->rti_addrs;
1570         if (info->rti_ifp != NULL)
1571                 rtm->rtm_index = info->rti_ifp->if_index;
1572         /* Add RTF_DONE to indicate command 'completion' required by API */
1573         info->rti_flags |= RTF_DONE;
1574         /* Reported routes has to be up */
1575         if (cmd == RTM_ADD || cmd == RTM_CHANGE)
1576                 info->rti_flags |= RTF_UP;
1577         rtm->rtm_flags = info->rti_flags;
1578
1579         sa = info->rti_info[RTAX_DST];
1580         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1581
1582         return (0);
1583 }
1584
1585 /*
1586  * This is the analogue to the rt_newaddrmsg which performs the same
1587  * function but for multicast group memberhips.  This is easier since
1588  * there is no route state to worry about.
1589  */
1590 void
1591 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1592 {
1593         struct rt_addrinfo info;
1594         struct mbuf *m = NULL;
1595         struct ifnet *ifp = ifma->ifma_ifp;
1596         struct ifma_msghdr *ifmam;
1597
1598         if (V_route_cb.any_count == 0)
1599                 return;
1600
1601         bzero((caddr_t)&info, sizeof(info));
1602         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1603         if (ifp && ifp->if_addr)
1604                 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1605         else
1606                 info.rti_info[RTAX_IFP] = NULL;
1607         /*
1608          * If a link-layer address is present, present it as a ``gateway''
1609          * (similarly to how ARP entries, e.g., are presented).
1610          */
1611         info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1612         m = rtsock_msg_mbuf(cmd, &info);
1613         if (m == NULL)
1614                 return;
1615         ifmam = mtod(m, struct ifma_msghdr *);
1616         KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1617             __func__));
1618         ifmam->ifmam_index = ifp->if_index;
1619         ifmam->ifmam_addrs = info.rti_addrs;
1620         rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1621 }
1622
1623 static struct mbuf *
1624 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1625         struct rt_addrinfo *info)
1626 {
1627         struct if_announcemsghdr *ifan;
1628         struct mbuf *m;
1629
1630         if (V_route_cb.any_count == 0)
1631                 return NULL;
1632         bzero((caddr_t)info, sizeof(*info));
1633         m = rtsock_msg_mbuf(type, info);
1634         if (m != NULL) {
1635                 ifan = mtod(m, struct if_announcemsghdr *);
1636                 ifan->ifan_index = ifp->if_index;
1637                 strlcpy(ifan->ifan_name, ifp->if_xname,
1638                         sizeof(ifan->ifan_name));
1639                 ifan->ifan_what = what;
1640         }
1641         return m;
1642 }
1643
1644 /*
1645  * This is called to generate routing socket messages indicating
1646  * IEEE80211 wireless events.
1647  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1648  */
1649 void
1650 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1651 {
1652         struct mbuf *m;
1653         struct rt_addrinfo info;
1654
1655         m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1656         if (m != NULL) {
1657                 /*
1658                  * Append the ieee80211 data.  Try to stick it in the
1659                  * mbuf containing the ifannounce msg; otherwise allocate
1660                  * a new mbuf and append.
1661                  *
1662                  * NB: we assume m is a single mbuf.
1663                  */
1664                 if (data_len > M_TRAILINGSPACE(m)) {
1665                         struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1666                         if (n == NULL) {
1667                                 m_freem(m);
1668                                 return;
1669                         }
1670                         bcopy(data, mtod(n, void *), data_len);
1671                         n->m_len = data_len;
1672                         m->m_next = n;
1673                 } else if (data_len > 0) {
1674                         bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1675                         m->m_len += data_len;
1676                 }
1677                 if (m->m_flags & M_PKTHDR)
1678                         m->m_pkthdr.len += data_len;
1679                 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1680                 rt_dispatch(m, AF_UNSPEC);
1681         }
1682 }
1683
1684 /*
1685  * This is called to generate routing socket messages indicating
1686  * network interface arrival and departure.
1687  */
1688 void
1689 rt_ifannouncemsg(struct ifnet *ifp, int what)
1690 {
1691         struct mbuf *m;
1692         struct rt_addrinfo info;
1693
1694         m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1695         if (m != NULL)
1696                 rt_dispatch(m, AF_UNSPEC);
1697 }
1698
1699 static void
1700 rt_dispatch(struct mbuf *m, sa_family_t saf)
1701 {
1702         struct m_tag *tag;
1703
1704         /*
1705          * Preserve the family from the sockaddr, if any, in an m_tag for
1706          * use when injecting the mbuf into the routing socket buffer from
1707          * the netisr.
1708          */
1709         if (saf != AF_UNSPEC) {
1710                 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1711                     M_NOWAIT);
1712                 if (tag == NULL) {
1713                         m_freem(m);
1714                         return;
1715                 }
1716                 *(unsigned short *)(tag + 1) = saf;
1717                 m_tag_prepend(m, tag);
1718         }
1719 #ifdef VIMAGE
1720         if (V_loif)
1721                 m->m_pkthdr.rcvif = V_loif;
1722         else {
1723                 m_freem(m);
1724                 return;
1725         }
1726 #endif
1727         netisr_queue(NETISR_ROUTE, m);  /* mbuf is free'd on failure. */
1728 }
1729
1730 /*
1731  * Checks if rte can be exported v.r.t jails/vnets.
1732  *
1733  * Returns 1 if it can, 0 otherwise.
1734  */
1735 static int
1736 can_export_rte(struct ucred *td_ucred, const struct rtentry *rt)
1737 {
1738
1739         if ((rt->rte_flags & RTF_HOST) == 0
1740             ? jailed_without_vnet(td_ucred)
1741             : prison_if(td_ucred, rt_key_const(rt)) != 0)
1742                 return (0);
1743         return (1);
1744 }
1745
1746 /*
1747  * This is used in dumping the kernel table via sysctl().
1748  */
1749 static int
1750 sysctl_dumpentry(struct radix_node *rn, void *vw)
1751 {
1752         struct walkarg *w = vw;
1753         struct rtentry *rt = (struct rtentry *)rn;
1754         struct nhop_object *nh;
1755         int error = 0;
1756
1757         NET_EPOCH_ASSERT();
1758
1759         if (w->w_op == NET_RT_FLAGS && !(rt->rte_flags & w->w_arg))
1760                 return 0;
1761         if (!can_export_rte(w->w_req->td->td_ucred, rt))
1762                 return (0);
1763         nh = rt->rt_nhop;
1764 #ifdef ROUTE_MPATH
1765         if (NH_IS_NHGRP(nh)) {
1766                 struct weightened_nhop *wn;
1767                 uint32_t num_nhops;
1768                 wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
1769                 for (int i = 0; i < num_nhops; i++) {
1770                         error = sysctl_dumpnhop(rt, wn[i].nh, wn[i].weight, w);
1771                         if (error != 0)
1772                                 return (error);
1773                 }
1774         } else
1775 #endif
1776                 error = sysctl_dumpnhop(rt, nh, rt->rt_weight, w);
1777
1778         return (0);
1779 }
1780
1781
1782 static int
1783 sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, uint32_t weight,
1784     struct walkarg *w)
1785 {
1786         struct rt_addrinfo info;
1787         int error = 0, size;
1788         struct sockaddr_storage ss;
1789
1790         bzero((caddr_t)&info, sizeof(info));
1791         info.rti_info[RTAX_DST] = rt_key(rt);
1792         info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
1793         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
1794             rt_mask(rt), &ss);
1795         info.rti_info[RTAX_GENMASK] = 0;
1796         if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) {
1797                 info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr;
1798                 info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
1799                 if (nh->nh_ifp->if_flags & IFF_POINTOPOINT)
1800                         info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr;
1801         }
1802         if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
1803                 return (error);
1804         if (w->w_req && w->w_tmem) {
1805                 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1806
1807                 bzero(&rtm->rtm_index,
1808                     sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
1809                 if (rt->rte_flags & RTF_GWFLAG_COMPAT)
1810                         rtm->rtm_flags = RTF_GATEWAY |
1811                                 (rt->rte_flags & ~RTF_GWFLAG_COMPAT);
1812                 else
1813                         rtm->rtm_flags = rt->rte_flags;
1814                 rtm->rtm_flags |= nhop_get_rtflags(nh);
1815                 rt_getmetrics(rt, nh, &rtm->rtm_rmx);
1816                 rtm->rtm_rmx.rmx_weight = weight;
1817                 rtm->rtm_index = nh->nh_ifp->if_index;
1818                 rtm->rtm_addrs = info.rti_addrs;
1819                 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1820                 return (error);
1821         }
1822         return (error);
1823 }
1824
1825 static int
1826 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
1827     struct rt_addrinfo *info, struct walkarg *w, int len)
1828 {
1829         struct if_msghdrl *ifm;
1830         struct if_data *ifd;
1831
1832         ifm = (struct if_msghdrl *)w->w_tmem;
1833
1834 #ifdef COMPAT_FREEBSD32
1835         if (w->w_req->flags & SCTL_MASK32) {
1836                 struct if_msghdrl32 *ifm32;
1837
1838                 ifm32 = (struct if_msghdrl32 *)ifm;
1839                 ifm32->ifm_addrs = info->rti_addrs;
1840                 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1841                 ifm32->ifm_index = ifp->if_index;
1842                 ifm32->_ifm_spare1 = 0;
1843                 ifm32->ifm_len = sizeof(*ifm32);
1844                 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1845                 ifm32->_ifm_spare2 = 0;
1846                 ifd = &ifm32->ifm_data;
1847         } else
1848 #endif
1849         {
1850                 ifm->ifm_addrs = info->rti_addrs;
1851                 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1852                 ifm->ifm_index = ifp->if_index;
1853                 ifm->_ifm_spare1 = 0;
1854                 ifm->ifm_len = sizeof(*ifm);
1855                 ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1856                 ifm->_ifm_spare2 = 0;
1857                 ifd = &ifm->ifm_data;
1858         }
1859
1860         memcpy(ifd, src_ifd, sizeof(*ifd));
1861
1862         return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1863 }
1864
1865 static int
1866 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
1867     struct rt_addrinfo *info, struct walkarg *w, int len)
1868 {
1869         struct if_msghdr *ifm;
1870         struct if_data *ifd;
1871
1872         ifm = (struct if_msghdr *)w->w_tmem;
1873
1874 #ifdef COMPAT_FREEBSD32
1875         if (w->w_req->flags & SCTL_MASK32) {
1876                 struct if_msghdr32 *ifm32;
1877
1878                 ifm32 = (struct if_msghdr32 *)ifm;
1879                 ifm32->ifm_addrs = info->rti_addrs;
1880                 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1881                 ifm32->ifm_index = ifp->if_index;
1882                 ifm32->_ifm_spare1 = 0;
1883                 ifd = &ifm32->ifm_data;
1884         } else
1885 #endif
1886         {
1887                 ifm->ifm_addrs = info->rti_addrs;
1888                 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1889                 ifm->ifm_index = ifp->if_index;
1890                 ifm->_ifm_spare1 = 0;
1891                 ifd = &ifm->ifm_data;
1892         }
1893
1894         memcpy(ifd, src_ifd, sizeof(*ifd));
1895
1896         return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1897 }
1898
1899 static int
1900 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1901     struct walkarg *w, int len)
1902 {
1903         struct ifa_msghdrl *ifam;
1904         struct if_data *ifd;
1905
1906         ifam = (struct ifa_msghdrl *)w->w_tmem;
1907
1908 #ifdef COMPAT_FREEBSD32
1909         if (w->w_req->flags & SCTL_MASK32) {
1910                 struct ifa_msghdrl32 *ifam32;
1911
1912                 ifam32 = (struct ifa_msghdrl32 *)ifam;
1913                 ifam32->ifam_addrs = info->rti_addrs;
1914                 ifam32->ifam_flags = ifa->ifa_flags;
1915                 ifam32->ifam_index = ifa->ifa_ifp->if_index;
1916                 ifam32->_ifam_spare1 = 0;
1917                 ifam32->ifam_len = sizeof(*ifam32);
1918                 ifam32->ifam_data_off =
1919                     offsetof(struct ifa_msghdrl32, ifam_data);
1920                 ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
1921                 ifd = &ifam32->ifam_data;
1922         } else
1923 #endif
1924         {
1925                 ifam->ifam_addrs = info->rti_addrs;
1926                 ifam->ifam_flags = ifa->ifa_flags;
1927                 ifam->ifam_index = ifa->ifa_ifp->if_index;
1928                 ifam->_ifam_spare1 = 0;
1929                 ifam->ifam_len = sizeof(*ifam);
1930                 ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1931                 ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1932                 ifd = &ifam->ifam_data;
1933         }
1934
1935         bzero(ifd, sizeof(*ifd));
1936         ifd->ifi_datalen = sizeof(struct if_data);
1937         ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
1938         ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
1939         ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
1940         ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
1941
1942         /* Fixup if_data carp(4) vhid. */
1943         if (carp_get_vhid_p != NULL)
1944                 ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
1945
1946         return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1947 }
1948
1949 static int
1950 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1951     struct walkarg *w, int len)
1952 {
1953         struct ifa_msghdr *ifam;
1954
1955         ifam = (struct ifa_msghdr *)w->w_tmem;
1956         ifam->ifam_addrs = info->rti_addrs;
1957         ifam->ifam_flags = ifa->ifa_flags;
1958         ifam->ifam_index = ifa->ifa_ifp->if_index;
1959         ifam->_ifam_spare1 = 0;
1960         ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1961
1962         return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1963 }
1964
1965 static int
1966 sysctl_iflist(int af, struct walkarg *w)
1967 {
1968         struct ifnet *ifp;
1969         struct ifaddr *ifa;
1970         struct if_data ifd;
1971         struct rt_addrinfo info;
1972         int len, error = 0;
1973         struct sockaddr_storage ss;
1974
1975         bzero((caddr_t)&info, sizeof(info));
1976         bzero(&ifd, sizeof(ifd));
1977         CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1978                 if (w->w_arg && w->w_arg != ifp->if_index)
1979                         continue;
1980                 if_data_copy(ifp, &ifd);
1981                 ifa = ifp->if_addr;
1982                 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1983                 error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
1984                 if (error != 0)
1985                         goto done;
1986                 info.rti_info[RTAX_IFP] = NULL;
1987                 if (w->w_req && w->w_tmem) {
1988                         if (w->w_op == NET_RT_IFLISTL)
1989                                 error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
1990                                     len);
1991                         else
1992                                 error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
1993                                     len);
1994                         if (error)
1995                                 goto done;
1996                 }
1997                 while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
1998                         if (af && af != ifa->ifa_addr->sa_family)
1999                                 continue;
2000                         if (prison_if(w->w_req->td->td_ucred,
2001                             ifa->ifa_addr) != 0)
2002                                 continue;
2003                         info.rti_info[RTAX_IFA] = ifa->ifa_addr;
2004                         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
2005                             ifa->ifa_addr, ifa->ifa_netmask, &ss);
2006                         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
2007                         error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
2008                         if (error != 0)
2009                                 goto done;
2010                         if (w->w_req && w->w_tmem) {
2011                                 if (w->w_op == NET_RT_IFLISTL)
2012                                         error = sysctl_iflist_ifaml(ifa, &info,
2013                                             w, len);
2014                                 else
2015                                         error = sysctl_iflist_ifam(ifa, &info,
2016                                             w, len);
2017                                 if (error)
2018                                         goto done;
2019                         }
2020                 }
2021                 info.rti_info[RTAX_IFA] = NULL;
2022                 info.rti_info[RTAX_NETMASK] = NULL;
2023                 info.rti_info[RTAX_BRD] = NULL;
2024         }
2025 done:
2026         return (error);
2027 }
2028
2029 static int
2030 sysctl_ifmalist(int af, struct walkarg *w)
2031 {
2032         struct rt_addrinfo info;
2033         struct ifaddr *ifa;
2034         struct ifmultiaddr *ifma;
2035         struct ifnet *ifp;
2036         int error, len;
2037
2038         NET_EPOCH_ASSERT();
2039
2040         error = 0;
2041         bzero((caddr_t)&info, sizeof(info));
2042
2043         CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2044                 if (w->w_arg && w->w_arg != ifp->if_index)
2045                         continue;
2046                 ifa = ifp->if_addr;
2047                 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
2048                 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2049                         if (af && af != ifma->ifma_addr->sa_family)
2050                                 continue;
2051                         if (prison_if(w->w_req->td->td_ucred,
2052                             ifma->ifma_addr) != 0)
2053                                 continue;
2054                         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2055                         info.rti_info[RTAX_GATEWAY] =
2056                             (ifma->ifma_addr->sa_family != AF_LINK) ?
2057                             ifma->ifma_lladdr : NULL;
2058                         error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
2059                         if (error != 0)
2060                                 break;
2061                         if (w->w_req && w->w_tmem) {
2062                                 struct ifma_msghdr *ifmam;
2063
2064                                 ifmam = (struct ifma_msghdr *)w->w_tmem;
2065                                 ifmam->ifmam_index = ifma->ifma_ifp->if_index;
2066                                 ifmam->ifmam_flags = 0;
2067                                 ifmam->ifmam_addrs = info.rti_addrs;
2068                                 ifmam->_ifmam_spare1 = 0;
2069                                 error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
2070                                 if (error != 0)
2071                                         break;
2072                         }
2073                 }
2074                 if (error != 0)
2075                         break;
2076         }
2077         return (error);
2078 }
2079
2080 static int
2081 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
2082 {
2083         RIB_RLOCK_TRACKER;
2084         struct epoch_tracker et;
2085         int     *name = (int *)arg1;
2086         u_int   namelen = arg2;
2087         struct rib_head *rnh = NULL; /* silence compiler. */
2088         int     i, lim, error = EINVAL;
2089         int     fib = 0;
2090         u_char  af;
2091         struct  walkarg w;
2092
2093         name ++;
2094         namelen--;
2095         if (req->newptr)
2096                 return (EPERM);
2097         if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP || name[1] == NET_RT_NHGRP) {
2098                 if (namelen == 3)
2099                         fib = req->td->td_proc->p_fibnum;
2100                 else if (namelen == 4)
2101                         fib = (name[3] == RT_ALL_FIBS) ?
2102                             req->td->td_proc->p_fibnum : name[3];
2103                 else
2104                         return ((namelen < 3) ? EISDIR : ENOTDIR);
2105                 if (fib < 0 || fib >= rt_numfibs)
2106                         return (EINVAL);
2107         } else if (namelen != 3)
2108                 return ((namelen < 3) ? EISDIR : ENOTDIR);
2109         af = name[0];
2110         if (af > AF_MAX)
2111                 return (EINVAL);
2112         bzero(&w, sizeof(w));
2113         w.w_op = name[1];
2114         w.w_arg = name[2];
2115         w.w_req = req;
2116
2117         error = sysctl_wire_old_buffer(req, 0);
2118         if (error)
2119                 return (error);
2120
2121         /*
2122          * Allocate reply buffer in advance.
2123          * All rtsock messages has maximum length of u_short.
2124          */
2125         w.w_tmemsize = 65536;
2126         w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
2127
2128         NET_EPOCH_ENTER(et);
2129         switch (w.w_op) {
2130         case NET_RT_DUMP:
2131         case NET_RT_FLAGS:
2132                 if (af == 0) {                  /* dump all tables */
2133                         i = 1;
2134                         lim = AF_MAX;
2135                 } else                          /* dump only one table */
2136                         i = lim = af;
2137
2138                 /*
2139                  * take care of llinfo entries, the caller must
2140                  * specify an AF
2141                  */
2142                 if (w.w_op == NET_RT_FLAGS &&
2143                     (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
2144                         if (af != 0)
2145                                 error = lltable_sysctl_dumparp(af, w.w_req);
2146                         else
2147                                 error = EINVAL;
2148                         break;
2149                 }
2150                 /*
2151                  * take care of routing entries
2152                  */
2153                 for (error = 0; error == 0 && i <= lim; i++) {
2154                         rnh = rt_tables_get_rnh(fib, i);
2155                         if (rnh != NULL) {
2156                                 RIB_RLOCK(rnh);
2157                                 error = rnh->rnh_walktree(&rnh->head,
2158                                     sysctl_dumpentry, &w);
2159                                 RIB_RUNLOCK(rnh);
2160                         } else if (af != 0)
2161                                 error = EAFNOSUPPORT;
2162                 }
2163                 break;
2164         case NET_RT_NHOP:
2165         case NET_RT_NHGRP:
2166                 /* Allow dumping one specific af/fib at a time */
2167                 if (namelen < 4) {
2168                         error = EINVAL;
2169                         break;
2170                 }
2171                 fib = name[3];
2172                 if (fib < 0 || fib > rt_numfibs) {
2173                         error = EINVAL;
2174                         break;
2175                 }
2176                 rnh = rt_tables_get_rnh(fib, af);
2177                 if (rnh == NULL) {
2178                         error = EAFNOSUPPORT;
2179                         break;
2180                 }
2181                 if (w.w_op == NET_RT_NHOP)
2182                         error = nhops_dump_sysctl(rnh, w.w_req);
2183                 else
2184 #ifdef ROUTE_MPATH
2185                         error = nhgrp_dump_sysctl(rnh, w.w_req);
2186 #else
2187                         error = ENOTSUP;
2188 #endif
2189                 break;
2190         case NET_RT_IFLIST:
2191         case NET_RT_IFLISTL:
2192                 error = sysctl_iflist(af, &w);
2193                 break;
2194
2195         case NET_RT_IFMALIST:
2196                 error = sysctl_ifmalist(af, &w);
2197                 break;
2198         }
2199         NET_EPOCH_EXIT(et);
2200
2201         free(w.w_tmem, M_TEMP);
2202         return (error);
2203 }
2204
2205 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE,
2206     sysctl_rtsock, "Return route tables and interface/address lists");
2207
2208 /*
2209  * Definitions of protocols supported in the ROUTE domain.
2210  */
2211
2212 static struct domain routedomain;               /* or at least forward */
2213
2214 static struct protosw routesw[] = {
2215 {
2216         .pr_type =              SOCK_RAW,
2217         .pr_domain =            &routedomain,
2218         .pr_flags =             PR_ATOMIC|PR_ADDR,
2219         .pr_output =            route_output,
2220         .pr_ctlinput =          raw_ctlinput,
2221         .pr_init =              raw_init,
2222         .pr_usrreqs =           &route_usrreqs
2223 }
2224 };
2225
2226 static struct domain routedomain = {
2227         .dom_family =           PF_ROUTE,
2228         .dom_name =             "route",
2229         .dom_protosw =          routesw,
2230         .dom_protoswNPROTOSW =  &routesw[nitems(routesw)]
2231 };
2232
2233 VNET_DOMAIN_SET(route);