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_mpath.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 #ifdef DDB
  58 #include <ddb/ddb.h>
  59 #include <ddb/db_lex.h>
  60 #endif
  61
  62 #include <net/if.h>
  63 #include <net/if_var.h>
  64 #include <net/if_dl.h>
  65 #include <net/if_llatbl.h>
  66 #include <net/if_types.h>
  67 #include <net/netisr.h>
  68 #include <net/raw_cb.h>
  69 #include <net/route.h>
  70 #include <net/route_var.h>
  71 #include <net/vnet.h>
  72
  73 #include <netinet/in.h>
  74 #include <netinet/if_ether.h>
  75 #include <netinet/ip_carp.h>
  76 #ifdef INET6
  77 #include <netinet6/ip6_var.h>
  78 #include <netinet6/scope6_var.h>
  79 #endif
  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 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 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_iflist(int af, struct walkarg *w);
 178 static int      sysctl_ifmalist(int af, struct walkarg *w);
 179 static int      route_output(struct mbuf *m, struct socket *so, ...);
 180 static void     rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out);
 181 static void     rt_dispatch(struct mbuf *, sa_family_t);
 182 static struct sockaddr  *rtsock_fix_netmask(struct sockaddr *dst,
 183                         struct sockaddr *smask, struct sockaddr_storage *dmask);
 184
 185 static struct netisr_handler rtsock_nh = {
 186         .nh_name = "rtsock",
 187         .nh_handler = rts_input,
 188         .nh_proto = NETISR_ROUTE,
 189         .nh_policy = NETISR_POLICY_SOURCE,
 190 };
 191
 192 static int
 193 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
 194 {
 195         int error, qlimit;
 196
 197         netisr_getqlimit(&rtsock_nh, &qlimit);
 198         error = sysctl_handle_int(oidp, &qlimit, 0, req);
 199         if (error || !req->newptr)
 200                 return (error);
 201         if (qlimit < 1)
 202                 return (EINVAL);
 203         return (netisr_setqlimit(&rtsock_nh, qlimit));
 204 }
 205 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
 206     0, 0, sysctl_route_netisr_maxqlen, "I",
 207     "maximum routing socket dispatch queue length");
 208
 209 static void
 210 vnet_rts_init(void)
 211 {
 212         int tmp;
 213
 214         if (IS_DEFAULT_VNET(curvnet)) {
 215                 if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
 216                         rtsock_nh.nh_qlimit = tmp;
 217                 netisr_register(&rtsock_nh);
 218         }
 219 #ifdef VIMAGE
 220          else
 221                 netisr_register_vnet(&rtsock_nh);
 222 #endif
 223 }
 224 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
 225     vnet_rts_init, 0);
 226
 227 #ifdef VIMAGE
 228 static void
 229 vnet_rts_uninit(void)
 230 {
 231
 232         netisr_unregister_vnet(&rtsock_nh);
 233 }
 234 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
 235     vnet_rts_uninit, 0);
 236 #endif
 237
 238 static int
 239 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
 240     struct rawcb *rp)
 241 {
 242         int fibnum;
 243
 244         KASSERT(m != NULL, ("%s: m is NULL", __func__));
 245         KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
 246         KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
 247
 248         /* No filtering requested. */
 249         if ((m->m_flags & RTS_FILTER_FIB) == 0)
 250                 return (0);
 251
 252         /* Check if it is a rts and the fib matches the one of the socket. */
 253         fibnum = M_GETFIB(m);
 254         if (proto->sp_family != PF_ROUTE ||
 255             rp->rcb_socket == NULL ||
 256             rp->rcb_socket->so_fibnum == fibnum)
 257                 return (0);
 258
 259         /* Filtering requested and no match, the socket shall be skipped. */
 260         return (1);
 261 }
 262
 263 static void
 264 rts_input(struct mbuf *m)
 265 {
 266         struct sockproto route_proto;
 267         unsigned short *family;
 268         struct m_tag *tag;
 269
 270         route_proto.sp_family = PF_ROUTE;
 271         tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
 272         if (tag != NULL) {
 273                 family = (unsigned short *)(tag + 1);
 274                 route_proto.sp_protocol = *family;
 275                 m_tag_delete(m, tag);
 276         } else
 277                 route_proto.sp_protocol = 0;
 278
 279         raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
 280 }
 281
 282 /*
 283  * It really doesn't make any sense at all for this code to share much
 284  * with raw_usrreq.c, since its functionality is so restricted.  XXX
 285  */
 286 static void
 287 rts_abort(struct socket *so)
 288 {
 289
 290         raw_usrreqs.pru_abort(so);
 291 }
 292
 293 static void
 294 rts_close(struct socket *so)
 295 {
 296
 297         raw_usrreqs.pru_close(so);
 298 }
 299
 300 /* pru_accept is EOPNOTSUPP */
 301
 302 static int
 303 rts_attach(struct socket *so, int proto, struct thread *td)
 304 {
 305         struct rawcb *rp;
 306         int error;
 307
 308         KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
 309
 310         /* XXX */
 311         rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
 312
 313         so->so_pcb = (caddr_t)rp;
 314         so->so_fibnum = td->td_proc->p_fibnum;
 315         error = raw_attach(so, proto);
 316         rp = sotorawcb(so);
 317         if (error) {
 318                 so->so_pcb = NULL;
 319                 free(rp, M_PCB);
 320                 return error;
 321         }
 322         RTSOCK_LOCK();
 323         switch(rp->rcb_proto.sp_protocol) {
 324         case AF_INET:
 325                 V_route_cb.ip_count++;
 326                 break;
 327         case AF_INET6:
 328                 V_route_cb.ip6_count++;
 329                 break;
 330         }
 331         V_route_cb.any_count++;
 332         RTSOCK_UNLOCK();
 333         soisconnected(so);
 334         so->so_options |= SO_USELOOPBACK;
 335         return 0;
 336 }
 337
 338 static int
 339 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
 340 {
 341
 342         return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
 343 }
 344
 345 static int
 346 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
 347 {
 348
 349         return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
 350 }
 351
 352 /* pru_connect2 is EOPNOTSUPP */
 353 /* pru_control is EOPNOTSUPP */
 354
 355 static void
 356 rts_detach(struct socket *so)
 357 {
 358         struct rawcb *rp = sotorawcb(so);
 359
 360         KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
 361
 362         RTSOCK_LOCK();
 363         switch(rp->rcb_proto.sp_protocol) {
 364         case AF_INET:
 365                 V_route_cb.ip_count--;
 366                 break;
 367         case AF_INET6:
 368                 V_route_cb.ip6_count--;
 369                 break;
 370         }
 371         V_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 #if defined(INET) || defined(INET6)
 450         struct epoch_tracker et;
 451 #endif
 452
 453         /* First, see if the returned address is part of the jail. */
 454         if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
 455                 info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
 456                 return (0);
 457         }
 458
 459         switch (info->rti_info[RTAX_DST]->sa_family) {
 460 #ifdef INET
 461         case AF_INET:
 462         {
 463                 struct in_addr ia;
 464                 struct ifaddr *ifa;
 465                 int found;
 466
 467                 found = 0;
 468                 /*
 469                  * Try to find an address on the given outgoing interface
 470                  * that belongs to the jail.
 471                  */
 472                 NET_EPOCH_ENTER(et);
 473                 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 474                         struct sockaddr *sa;
 475                         sa = ifa->ifa_addr;
 476                         if (sa->sa_family != AF_INET)
 477                                 continue;
 478                         ia = ((struct sockaddr_in *)sa)->sin_addr;
 479                         if (prison_check_ip4(cred, &ia) == 0) {
 480                                 found = 1;
 481                                 break;
 482                         }
 483                 }
 484                 NET_EPOCH_EXIT(et);
 485                 if (!found) {
 486                         /*
 487                          * As a last resort return the 'default' jail address.
 488                          */
 489                         ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
 490                             sin_addr;
 491                         if (prison_get_ip4(cred, &ia) != 0)
 492                                 return (ESRCH);
 493                 }
 494                 bzero(&saun->sin, sizeof(struct sockaddr_in));
 495                 saun->sin.sin_len = sizeof(struct sockaddr_in);
 496                 saun->sin.sin_family = AF_INET;
 497                 saun->sin.sin_addr.s_addr = ia.s_addr;
 498                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
 499                 break;
 500         }
 501 #endif
 502 #ifdef INET6
 503         case AF_INET6:
 504         {
 505                 struct in6_addr ia6;
 506                 struct ifaddr *ifa;
 507                 int found;
 508
 509                 found = 0;
 510                 /*
 511                  * Try to find an address on the given outgoing interface
 512                  * that belongs to the jail.
 513                  */
 514                 NET_EPOCH_ENTER(et);
 515                 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 516                         struct sockaddr *sa;
 517                         sa = ifa->ifa_addr;
 518                         if (sa->sa_family != AF_INET6)
 519                                 continue;
 520                         bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
 521                             &ia6, sizeof(struct in6_addr));
 522                         if (prison_check_ip6(cred, &ia6) == 0) {
 523                                 found = 1;
 524                                 break;
 525                         }
 526                 }
 527                 NET_EPOCH_EXIT(et);
 528                 if (!found) {
 529                         /*
 530                          * As a last resort return the 'default' jail address.
 531                          */
 532                         ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
 533                             sin6_addr;
 534                         if (prison_get_ip6(cred, &ia6) != 0)
 535                                 return (ESRCH);
 536                 }
 537                 bzero(&saun->sin6, sizeof(struct sockaddr_in6));
 538                 saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
 539                 saun->sin6.sin6_family = AF_INET6;
 540                 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
 541                 if (sa6_recoverscope(&saun->sin6) != 0)
 542                         return (ESRCH);
 543                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
 544                 break;
 545         }
 546 #endif
 547         default:
 548                 return (ESRCH);
 549         }
 550         return (0);
 551 }
 552
 553 /*ARGSUSED*/
 554 static int
 555 route_output(struct mbuf *m, struct socket *so, ...)
 556 {
 557         RIB_RLOCK_TRACKER;
 558         struct rt_msghdr *rtm = NULL;
 559         struct rtentry *rt = NULL;
 560         struct rib_head *rnh;
 561         struct rt_addrinfo info;
 562         struct sockaddr_storage ss;
 563         struct epoch_tracker et;
 564 #ifdef INET6
 565         struct sockaddr_in6 *sin6;
 566         int i, rti_need_deembed = 0;
 567 #endif
 568         int alloc_len = 0, len, error = 0, fibnum;
 569         struct ifnet *ifp = NULL;
 570         union sockaddr_union saun;
 571         sa_family_t saf = AF_UNSPEC;
 572         struct rawcb *rp = NULL;
 573         struct walkarg w;
 574
 575         fibnum = so->so_fibnum;
 576
 577 #define senderr(e) { error = e; goto flush;}
 578         if (m == NULL || ((m->m_len < sizeof(long)) &&
 579                        (m = m_pullup(m, sizeof(long))) == NULL))
 580                 return (ENOBUFS);
 581         if ((m->m_flags & M_PKTHDR) == 0)
 582                 panic("route_output");
 583         NET_EPOCH_ENTER(et);
 584         len = m->m_pkthdr.len;
 585         if (len < sizeof(*rtm) ||
 586             len != mtod(m, struct rt_msghdr *)->rtm_msglen)
 587                 senderr(EINVAL);
 588
 589         /*
 590          * Most of current messages are in range 200-240 bytes,
 591          * minimize possible re-allocation on reply using larger size
 592          * buffer aligned on 1k boundaty.
 593          */
 594         alloc_len = roundup2(len, 1024);
 595         if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL)
 596                 senderr(ENOBUFS);
 597
 598         m_copydata(m, 0, len, (caddr_t)rtm);
 599         bzero(&info, sizeof(info));
 600         bzero(&w, sizeof(w));
 601
 602         if (rtm->rtm_version != RTM_VERSION) {
 603                 /* Do not touch message since format is unknown */
 604                 free(rtm, M_TEMP);
 605                 rtm = NULL;
 606                 senderr(EPROTONOSUPPORT);
 607         }
 608
 609         /*
 610          * Starting from here, it is possible
 611          * to alter original message and insert
 612          * caller PID and error value.
 613          */
 614
 615         rtm->rtm_pid = curproc->p_pid;
 616         info.rti_addrs = rtm->rtm_addrs;
 617
 618         info.rti_mflags = rtm->rtm_inits;
 619         info.rti_rmx = &rtm->rtm_rmx;
 620
 621         /*
 622          * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
 623          * link-local address because rtrequest requires addresses with
 624          * embedded scope id.
 625          */
 626         if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info))
 627                 senderr(EINVAL);
 628
 629         if (rtm->rtm_flags & RTF_RNH_LOCKED)
 630                 senderr(EINVAL);
 631         info.rti_flags = rtm->rtm_flags;
 632         if (info.rti_info[RTAX_DST] == NULL ||
 633             info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
 634             (info.rti_info[RTAX_GATEWAY] != NULL &&
 635              info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
 636                 senderr(EINVAL);
 637         saf = info.rti_info[RTAX_DST]->sa_family;
 638         /*
 639          * Verify that the caller has the appropriate privilege; RTM_GET
 640          * is the only operation the non-superuser is allowed.
 641          */
 642         if (rtm->rtm_type != RTM_GET) {
 643                 error = priv_check(curthread, PRIV_NET_ROUTE);
 644                 if (error)
 645                         senderr(error);
 646         }
 647
 648         /*
 649          * The given gateway address may be an interface address.
 650          * For example, issuing a "route change" command on a route
 651          * entry that was created from a tunnel, and the gateway
 652          * address given is the local end point. In this case the
 653          * RTF_GATEWAY flag must be cleared or the destination will
 654          * not be reachable even though there is no error message.
 655          */
 656         if (info.rti_info[RTAX_GATEWAY] != NULL &&
 657             info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
 658                 struct rt_addrinfo ginfo;
 659                 struct sockaddr *gdst;
 660
 661                 bzero(&ginfo, sizeof(ginfo));
 662                 bzero(&ss, sizeof(ss));
 663                 ss.ss_len = sizeof(ss);
 664
 665                 ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss;
 666                 gdst = info.rti_info[RTAX_GATEWAY];
 667
 668                 /*
 669                  * A host route through the loopback interface is
 670                  * installed for each interface adddress. In pre 8.0
 671                  * releases the interface address of a PPP link type
 672                  * is not reachable locally. This behavior is fixed as
 673                  * part of the new L2/L3 redesign and rewrite work. The
 674                  * signature of this interface address route is the
 675                  * AF_LINK sa_family type of the rt_gateway, and the
 676                  * rt_ifp has the IFF_LOOPBACK flag set.
 677                  */
 678                 if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) {
 679                         if (ss.ss_family == AF_LINK &&
 680                             ginfo.rti_ifp->if_flags & IFF_LOOPBACK) {
 681                                 info.rti_flags &= ~RTF_GATEWAY;
 682                                 info.rti_flags |= RTF_GWFLAG_COMPAT;
 683                         }
 684                         rib_free_info(&ginfo);
 685                 }
 686         }
 687
 688         switch (rtm->rtm_type) {
 689                 struct rtentry *saved_nrt;
 690
 691         case RTM_ADD:
 692         case RTM_CHANGE:
 693                 if (rtm->rtm_type == RTM_ADD) {
 694                         if (info.rti_info[RTAX_GATEWAY] == NULL)
 695                                 senderr(EINVAL);
 696                 }
 697                 saved_nrt = NULL;
 698
 699                 /* support for new ARP code */
 700                 if (info.rti_info[RTAX_GATEWAY] != NULL &&
 701                     info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
 702                     (rtm->rtm_flags & RTF_LLDATA) != 0) {
 703                         error = lla_rt_output(rtm, &info);
 704 #ifdef INET6
 705                         if (error == 0)
 706                                 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 707 #endif
 708                         break;
 709                 }
 710                 error = rtrequest1_fib(rtm->rtm_type, &info, &saved_nrt,
 711                     fibnum);
 712                 if (error == 0 && saved_nrt != NULL) {
 713 #ifdef INET6
 714                         rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 715 #endif
 716                         RT_LOCK(saved_nrt);
 717                         rtm->rtm_index = saved_nrt->rt_ifp->if_index;
 718                         RT_REMREF(saved_nrt);
 719                         RT_UNLOCK(saved_nrt);
 720                 }
 721                 break;
 722
 723         case RTM_DELETE:
 724                 saved_nrt = NULL;
 725                 /* support for new ARP code */
 726                 if (info.rti_info[RTAX_GATEWAY] &&
 727                     (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
 728                     (rtm->rtm_flags & RTF_LLDATA) != 0) {
 729                         error = lla_rt_output(rtm, &info);
 730 #ifdef INET6
 731                         if (error == 0)
 732                                 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 733 #endif
 734                         break;
 735                 }
 736                 error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt, fibnum);
 737                 if (error == 0) {
 738                         RT_LOCK(saved_nrt);
 739                         rt = saved_nrt;
 740                         goto report;
 741                 }
 742 #ifdef INET6
 743                 /* rt_msg2() will not be used when RTM_DELETE fails. */
 744                 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
 745 #endif
 746                 break;
 747
 748         case RTM_GET:
 749                 rnh = rt_tables_get_rnh(fibnum, saf);
 750                 if (rnh == NULL)
 751                         senderr(EAFNOSUPPORT);
 752
 753                 RIB_RLOCK(rnh);
 754
 755                 if (info.rti_info[RTAX_NETMASK] == NULL &&
 756                     rtm->rtm_type == RTM_GET) {
 757                         /*
 758                          * Provide longest prefix match for
 759                          * address lookup (no mask).
 760                          * 'route -n get addr'
 761                          */
 762                         rt = (struct rtentry *) rnh->rnh_matchaddr(
 763                             info.rti_info[RTAX_DST], &rnh->head);
 764                 } else
 765                         rt = (struct rtentry *) rnh->rnh_lookup(
 766                             info.rti_info[RTAX_DST],
 767                             info.rti_info[RTAX_NETMASK], &rnh->head);
 768
 769                 if (rt == NULL) {
 770                         RIB_RUNLOCK(rnh);
 771                         senderr(ESRCH);
 772                 }
 773 #ifdef RADIX_MPATH
 774                 /*
 775                  * for RTM_CHANGE/LOCK, if we got multipath routes,
 776                  * we require users to specify a matching RTAX_GATEWAY.
 777                  *
 778                  * for RTM_GET, gate is optional even with multipath.
 779                  * if gate == NULL the first match is returned.
 780                  * (no need to call rt_mpath_matchgate if gate == NULL)
 781                  */
 782                 if (rt_mpath_capable(rnh) &&
 783                     (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
 784                         rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
 785                         if (!rt) {
 786                                 RIB_RUNLOCK(rnh);
 787                                 senderr(ESRCH);
 788                         }
 789                 }
 790 #endif
 791                 /*
 792                  * If performing proxied L2 entry insertion, and
 793                  * the actual PPP host entry is found, perform
 794                  * another search to retrieve the prefix route of
 795                  * the local end point of the PPP link.
 796                  */
 797                 if (rtm->rtm_flags & RTF_ANNOUNCE) {
 798                         struct sockaddr laddr;
 799
 800                         if (rt->rt_ifp != NULL &&
 801                             rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
 802                                 struct epoch_tracker et;
 803                                 struct ifaddr *ifa;
 804
 805                                 NET_EPOCH_ENTER(et);
 806                                 ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1,
 807                                                 RT_ALL_FIBS);
 808                                 NET_EPOCH_EXIT(et);
 809                                 if (ifa != NULL)
 810                                         rt_maskedcopy(ifa->ifa_addr,
 811                                                       &laddr,
 812                                                       ifa->ifa_netmask);
 813                         } else
 814                                 rt_maskedcopy(rt->rt_ifa->ifa_addr,
 815                                               &laddr,
 816                                               rt->rt_ifa->ifa_netmask);
 817                         /*
 818                          * refactor rt and no lock operation necessary
 819                          */
 820                         rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr,
 821                             &rnh->head);
 822                         if (rt == NULL) {
 823                                 RIB_RUNLOCK(rnh);
 824                                 senderr(ESRCH);
 825                         }
 826                 }
 827                 RT_LOCK(rt);
 828                 RT_ADDREF(rt);
 829                 RIB_RUNLOCK(rnh);
 830
 831 report:
 832                 RT_LOCK_ASSERT(rt);
 833                 if ((rt->rt_flags & RTF_HOST) == 0
 834                     ? jailed_without_vnet(curthread->td_ucred)
 835                     : prison_if(curthread->td_ucred,
 836                     rt_key(rt)) != 0) {
 837                         RT_UNLOCK(rt);
 838                         senderr(ESRCH);
 839                 }
 840                 info.rti_info[RTAX_DST] = rt_key(rt);
 841                 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
 842                 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
 843                     rt_mask(rt), &ss);
 844                 info.rti_info[RTAX_GENMASK] = 0;
 845                 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
 846                         ifp = rt->rt_ifp;
 847                         if (ifp) {
 848                                 info.rti_info[RTAX_IFP] =
 849                                     ifp->if_addr->ifa_addr;
 850                                 error = rtm_get_jailed(&info, ifp, rt,
 851                                     &saun, curthread->td_ucred);
 852                                 if (error != 0) {
 853                                         RT_UNLOCK(rt);
 854                                         senderr(error);
 855                                 }
 856                                 if (ifp->if_flags & IFF_POINTOPOINT)
 857                                         info.rti_info[RTAX_BRD] =
 858                                             rt->rt_ifa->ifa_dstaddr;
 859                                 rtm->rtm_index = ifp->if_index;
 860                         } else {
 861                                 info.rti_info[RTAX_IFP] = NULL;
 862                                 info.rti_info[RTAX_IFA] = NULL;
 863                         }
 864                 } else if ((ifp = rt->rt_ifp) != NULL) {
 865                         rtm->rtm_index = ifp->if_index;
 866                 }
 867
 868                 /* Check if we need to realloc storage */
 869                 rtsock_msg_buffer(rtm->rtm_type, &info, NULL, &len);
 870                 if (len > alloc_len) {
 871                         struct rt_msghdr *new_rtm;
 872                         new_rtm = malloc(len, M_TEMP, M_NOWAIT);
 873                         if (new_rtm == NULL) {
 874                                 RT_UNLOCK(rt);
 875                                 senderr(ENOBUFS);
 876                         }
 877                         bcopy(rtm, new_rtm, rtm->rtm_msglen);
 878                         free(rtm, M_TEMP);
 879                         rtm = new_rtm;
 880                         alloc_len = len;
 881                 }
 882
 883                 w.w_tmem = (caddr_t)rtm;
 884                 w.w_tmemsize = alloc_len;
 885                 rtsock_msg_buffer(rtm->rtm_type, &info, &w, &len);
 886
 887                 if (rt->rt_flags & RTF_GWFLAG_COMPAT)
 888                         rtm->rtm_flags = RTF_GATEWAY |
 889                                 (rt->rt_flags & ~RTF_GWFLAG_COMPAT);
 890                 else
 891                         rtm->rtm_flags = rt->rt_flags;
 892                 rt_getmetrics(rt, &rtm->rtm_rmx);
 893                 rtm->rtm_addrs = info.rti_addrs;
 894
 895                 RT_UNLOCK(rt);
 896                 break;
 897
 898         default:
 899                 senderr(EOPNOTSUPP);
 900         }
 901
 902 flush:
 903         NET_EPOCH_EXIT(et);
 904         if (rt != NULL)
 905                 RTFREE(rt);
 906         /*
 907          * Check to see if we don't want our own messages.
 908          */
 909         if ((so->so_options & SO_USELOOPBACK) == 0) {
 910                 if (V_route_cb.any_count <= 1) {
 911                         if (rtm != NULL)
 912                                 free(rtm, M_TEMP);
 913                         m_freem(m);
 914                         return (error);
 915                 }
 916                 /* There is another listener, so construct message */
 917                 rp = sotorawcb(so);
 918         }
 919
 920         if (rtm != NULL) {
 921 #ifdef INET6
 922                 if (rti_need_deembed) {
 923                         /* sin6_scope_id is recovered before sending rtm. */
 924                         sin6 = (struct sockaddr_in6 *)&ss;
 925                         for (i = 0; i < RTAX_MAX; i++) {
 926                                 if (info.rti_info[i] == NULL)
 927                                         continue;
 928                                 if (info.rti_info[i]->sa_family != AF_INET6)
 929                                         continue;
 930                                 bcopy(info.rti_info[i], sin6, sizeof(*sin6));
 931                                 if (sa6_recoverscope(sin6) == 0)
 932                                         bcopy(sin6, info.rti_info[i],
 933                                                     sizeof(*sin6));
 934                         }
 935                 }
 936 #endif
 937                 if (error != 0)
 938                         rtm->rtm_errno = error;
 939                 else
 940                         rtm->rtm_flags |= RTF_DONE;
 941
 942                 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
 943                 if (m->m_pkthdr.len < rtm->rtm_msglen) {
 944                         m_freem(m);
 945                         m = NULL;
 946                 } else if (m->m_pkthdr.len > rtm->rtm_msglen)
 947                         m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
 948
 949                 free(rtm, M_TEMP);
 950         }
 951         if (m != NULL) {
 952                 M_SETFIB(m, fibnum);
 953                 m->m_flags |= RTS_FILTER_FIB;
 954                 if (rp) {
 955                         /*
 956                          * XXX insure we don't get a copy by
 957                          * invalidating our protocol
 958                          */
 959                         unsigned short family = rp->rcb_proto.sp_family;
 960                         rp->rcb_proto.sp_family = 0;
 961                         rt_dispatch(m, saf);
 962                         rp->rcb_proto.sp_family = family;
 963                 } else
 964                         rt_dispatch(m, saf);
 965         }
 966
 967         return (error);
 968 }
 969
 970 static void
 971 rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out)
 972 {
 973
 974         bzero(out, sizeof(*out));
 975         out->rmx_mtu = rt->rt_mtu;
 976         out->rmx_weight = rt->rt_weight;
 977         out->rmx_pksent = counter_u64_fetch(rt->rt_pksent);
 978         /* Kernel -> userland timebase conversion. */
 979         out->rmx_expire = rt->rt_expire ?
 980             rt->rt_expire - time_uptime + time_second : 0;
 981 }
 982
 983 /*
 984  * Extract the addresses of the passed sockaddrs.
 985  * Do a little sanity checking so as to avoid bad memory references.
 986  * This data is derived straight from userland.
 987  */
 988 static int
 989 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
 990 {
 991         struct sockaddr *sa;
 992         int i;
 993
 994         for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
 995                 if ((rtinfo->rti_addrs & (1 << i)) == 0)
 996                         continue;
 997                 sa = (struct sockaddr *)cp;
 998                 /*
 999                  * It won't fit.
1000                  */
1001                 if (cp + sa->sa_len > cplim)
1002                         return (EINVAL);
1003                 /*
1004                  * there are no more.. quit now
1005                  * If there are more bits, they are in error.
1006                  * I've seen this. route(1) can evidently generate these.
1007                  * This causes kernel to core dump.
1008                  * for compatibility, If we see this, point to a safe address.
1009                  */
1010                 if (sa->sa_len == 0) {
1011                         rtinfo->rti_info[i] = &sa_zero;
1012                         return (0); /* should be EINVAL but for compat */
1013                 }
1014                 /* accept it */
1015 #ifdef INET6
1016                 if (sa->sa_family == AF_INET6)
1017                         sa6_embedscope((struct sockaddr_in6 *)sa,
1018                             V_ip6_use_defzone);
1019 #endif
1020                 rtinfo->rti_info[i] = sa;
1021                 cp += SA_SIZE(sa);
1022         }
1023         return (0);
1024 }
1025
1026 /*
1027  * Fill in @dmask with valid netmask leaving original @smask
1028  * intact. Mostly used with radix netmasks.
1029  */
1030 static struct sockaddr *
1031 rtsock_fix_netmask(struct sockaddr *dst, struct sockaddr *smask,
1032     struct sockaddr_storage *dmask)
1033 {
1034         if (dst == NULL || smask == NULL)
1035                 return (NULL);
1036
1037         memset(dmask, 0, dst->sa_len);
1038         memcpy(dmask, smask, smask->sa_len);
1039         dmask->ss_len = dst->sa_len;
1040         dmask->ss_family = dst->sa_family;
1041
1042         return ((struct sockaddr *)dmask);
1043 }
1044
1045 /*
1046  * Writes information related to @rtinfo object to newly-allocated mbuf.
1047  * Assumes MCLBYTES is enough to construct any message.
1048  * Used for OS notifications of vaious events (if/ifa announces,etc)
1049  *
1050  * Returns allocated mbuf or NULL on failure.
1051  */
1052 static struct mbuf *
1053 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1054 {
1055         struct rt_msghdr *rtm;
1056         struct mbuf *m;
1057         int i;
1058         struct sockaddr *sa;
1059 #ifdef INET6
1060         struct sockaddr_storage ss;
1061         struct sockaddr_in6 *sin6;
1062 #endif
1063         int len, dlen;
1064
1065         switch (type) {
1066
1067         case RTM_DELADDR:
1068         case RTM_NEWADDR:
1069                 len = sizeof(struct ifa_msghdr);
1070                 break;
1071
1072         case RTM_DELMADDR:
1073         case RTM_NEWMADDR:
1074                 len = sizeof(struct ifma_msghdr);
1075                 break;
1076
1077         case RTM_IFINFO:
1078                 len = sizeof(struct if_msghdr);
1079                 break;
1080
1081         case RTM_IFANNOUNCE:
1082         case RTM_IEEE80211:
1083                 len = sizeof(struct if_announcemsghdr);
1084                 break;
1085
1086         default:
1087                 len = sizeof(struct rt_msghdr);
1088         }
1089
1090         /* XXXGL: can we use MJUMPAGESIZE cluster here? */
1091         KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1092         if (len > MHLEN)
1093                 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1094         else
1095                 m = m_gethdr(M_NOWAIT, MT_DATA);
1096         if (m == NULL)
1097                 return (m);
1098
1099         m->m_pkthdr.len = m->m_len = len;
1100         rtm = mtod(m, struct rt_msghdr *);
1101         bzero((caddr_t)rtm, len);
1102         for (i = 0; i < RTAX_MAX; i++) {
1103                 if ((sa = rtinfo->rti_info[i]) == NULL)
1104                         continue;
1105                 rtinfo->rti_addrs |= (1 << i);
1106                 dlen = SA_SIZE(sa);
1107 #ifdef INET6
1108                 if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1109                         sin6 = (struct sockaddr_in6 *)&ss;
1110                         bcopy(sa, sin6, sizeof(*sin6));
1111                         if (sa6_recoverscope(sin6) == 0)
1112                                 sa = (struct sockaddr *)sin6;
1113                 }
1114 #endif
1115                 m_copyback(m, len, dlen, (caddr_t)sa);
1116                 len += dlen;
1117         }
1118         if (m->m_pkthdr.len != len) {
1119                 m_freem(m);
1120                 return (NULL);
1121         }
1122         rtm->rtm_msglen = len;
1123         rtm->rtm_version = RTM_VERSION;
1124         rtm->rtm_type = type;
1125         return (m);
1126 }
1127
1128 /*
1129  * Writes information related to @rtinfo object to preallocated buffer.
1130  * Stores needed size in @plen. If @w is NULL, calculates size without
1131  * writing.
1132  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1133  *
1134  * Returns 0 on success.
1135  *
1136  */
1137 static int
1138 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1139 {
1140         int i;
1141         int len, buflen = 0, dlen;
1142         caddr_t cp = NULL;
1143         struct rt_msghdr *rtm = NULL;
1144 #ifdef INET6
1145         struct sockaddr_storage ss;
1146         struct sockaddr_in6 *sin6;
1147 #endif
1148 #ifdef COMPAT_FREEBSD32
1149         bool compat32 = false;
1150 #endif
1151
1152         switch (type) {
1153
1154         case RTM_DELADDR:
1155         case RTM_NEWADDR:
1156                 if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1157 #ifdef COMPAT_FREEBSD32
1158                         if (w->w_req->flags & SCTL_MASK32) {
1159                                 len = sizeof(struct ifa_msghdrl32);
1160                                 compat32 = true;
1161                         } else
1162 #endif
1163                                 len = sizeof(struct ifa_msghdrl);
1164                 } else
1165                         len = sizeof(struct ifa_msghdr);
1166                 break;
1167
1168         case RTM_IFINFO:
1169 #ifdef COMPAT_FREEBSD32
1170                 if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1171                         if (w->w_op == NET_RT_IFLISTL)
1172                                 len = sizeof(struct if_msghdrl32);
1173                         else
1174                                 len = sizeof(struct if_msghdr32);
1175                         compat32 = true;
1176                         break;
1177                 }
1178 #endif
1179                 if (w != NULL && w->w_op == NET_RT_IFLISTL)
1180                         len = sizeof(struct if_msghdrl);
1181                 else
1182                         len = sizeof(struct if_msghdr);
1183                 break;
1184
1185         case RTM_NEWMADDR:
1186                 len = sizeof(struct ifma_msghdr);
1187                 break;
1188
1189         default:
1190                 len = sizeof(struct rt_msghdr);
1191         }
1192
1193         if (w != NULL) {
1194                 rtm = (struct rt_msghdr *)w->w_tmem;
1195                 buflen = w->w_tmemsize - len;
1196                 cp = (caddr_t)w->w_tmem + len;
1197         }
1198
1199         rtinfo->rti_addrs = 0;
1200         for (i = 0; i < RTAX_MAX; i++) {
1201                 struct sockaddr *sa;
1202
1203                 if ((sa = rtinfo->rti_info[i]) == NULL)
1204                         continue;
1205                 rtinfo->rti_addrs |= (1 << i);
1206 #ifdef COMPAT_FREEBSD32
1207                 if (compat32)
1208                         dlen = SA_SIZE32(sa);
1209                 else
1210 #endif
1211                         dlen = SA_SIZE(sa);
1212                 if (cp != NULL && buflen >= dlen) {
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                         buflen -= dlen;
1224                 } else if (cp != NULL) {
1225                         /*
1226                          * Buffer too small. Count needed size
1227                          * and return with error.
1228                          */
1229                         cp = NULL;
1230                 }
1231
1232                 len += dlen;
1233         }
1234
1235         if (cp != NULL) {
1236                 dlen = ALIGN(len) - len;
1237                 if (buflen < dlen)
1238                         cp = NULL;
1239                 else {
1240                         bzero(cp, dlen);
1241                         cp += dlen;
1242                         buflen -= dlen;
1243                 }
1244         }
1245         len = ALIGN(len);
1246
1247         if (cp != NULL) {
1248                 /* fill header iff buffer is large enough */
1249                 rtm->rtm_version = RTM_VERSION;
1250                 rtm->rtm_type = type;
1251                 rtm->rtm_msglen = len;
1252         }
1253
1254         *plen = len;
1255
1256         if (w != NULL && cp == NULL)
1257                 return (ENOBUFS);
1258
1259         return (0);
1260 }
1261
1262 /*
1263  * This routine is called to generate a message from the routing
1264  * socket indicating that a redirect has occurred, a routing lookup
1265  * has failed, or that a protocol has detected timeouts to a particular
1266  * destination.
1267  */
1268 void
1269 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1270     int fibnum)
1271 {
1272         struct rt_msghdr *rtm;
1273         struct mbuf *m;
1274         struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1275
1276         if (V_route_cb.any_count == 0)
1277                 return;
1278         m = rtsock_msg_mbuf(type, rtinfo);
1279         if (m == NULL)
1280                 return;
1281
1282         if (fibnum != RT_ALL_FIBS) {
1283                 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1284                     "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1285                 M_SETFIB(m, fibnum);
1286                 m->m_flags |= RTS_FILTER_FIB;
1287         }
1288
1289         rtm = mtod(m, struct rt_msghdr *);
1290         rtm->rtm_flags = RTF_DONE | flags;
1291         rtm->rtm_errno = error;
1292         rtm->rtm_addrs = rtinfo->rti_addrs;
1293         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1294 }
1295
1296 void
1297 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1298 {
1299
1300         rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1301 }
1302
1303 /*
1304  * This routine is called to generate a message from the routing
1305  * socket indicating that the status of a network interface has changed.
1306  */
1307 void
1308 rt_ifmsg(struct ifnet *ifp)
1309 {
1310         struct if_msghdr *ifm;
1311         struct mbuf *m;
1312         struct rt_addrinfo info;
1313
1314         if (V_route_cb.any_count == 0)
1315                 return;
1316         bzero((caddr_t)&info, sizeof(info));
1317         m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1318         if (m == NULL)
1319                 return;
1320         ifm = mtod(m, struct if_msghdr *);
1321         ifm->ifm_index = ifp->if_index;
1322         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1323         if_data_copy(ifp, &ifm->ifm_data);
1324         ifm->ifm_addrs = 0;
1325         rt_dispatch(m, AF_UNSPEC);
1326 }
1327
1328 /*
1329  * Announce interface address arrival/withdraw.
1330  * Please do not call directly, use rt_addrmsg().
1331  * Assume input data to be valid.
1332  * Returns 0 on success.
1333  */
1334 int
1335 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1336 {
1337         struct rt_addrinfo info;
1338         struct sockaddr *sa;
1339         int ncmd;
1340         struct mbuf *m;
1341         struct ifa_msghdr *ifam;
1342         struct ifnet *ifp = ifa->ifa_ifp;
1343         struct sockaddr_storage ss;
1344
1345         if (V_route_cb.any_count == 0)
1346                 return (0);
1347
1348         ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1349
1350         bzero((caddr_t)&info, sizeof(info));
1351         info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1352         info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1353         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1354             info.rti_info[RTAX_IFP], ifa->ifa_netmask, &ss);
1355         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1356         if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1357                 return (ENOBUFS);
1358         ifam = mtod(m, struct ifa_msghdr *);
1359         ifam->ifam_index = ifp->if_index;
1360         ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1361         ifam->ifam_flags = ifa->ifa_flags;
1362         ifam->ifam_addrs = info.rti_addrs;
1363
1364         if (fibnum != RT_ALL_FIBS) {
1365                 M_SETFIB(m, fibnum);
1366                 m->m_flags |= RTS_FILTER_FIB;
1367         }
1368
1369         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1370
1371         return (0);
1372 }
1373
1374 /*
1375  * Announce route addition/removal.
1376  * Please do not call directly, use rt_routemsg().
1377  * Note that @rt data MAY be inconsistent/invalid:
1378  * if some userland app sends us "invalid" route message (invalid mask,
1379  * no dst, wrong address families, etc...) we need to pass it back
1380  * to app (and any other rtsock consumers) with rtm_errno field set to
1381  * non-zero value.
1382  *
1383  * Returns 0 on success.
1384  */
1385 int
1386 rtsock_routemsg(int cmd, struct ifnet *ifp, int error, struct rtentry *rt,
1387     int fibnum)
1388 {
1389         struct rt_addrinfo info;
1390         struct sockaddr *sa;
1391         struct mbuf *m;
1392         struct rt_msghdr *rtm;
1393         struct sockaddr_storage ss;
1394
1395         if (V_route_cb.any_count == 0)
1396                 return (0);
1397
1398         bzero((caddr_t)&info, sizeof(info));
1399         info.rti_info[RTAX_DST] = sa = rt_key(rt);
1400         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(sa, rt_mask(rt), &ss);
1401         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1402         if ((m = rtsock_msg_mbuf(cmd, &info)) == NULL)
1403                 return (ENOBUFS);
1404         rtm = mtod(m, struct rt_msghdr *);
1405         rtm->rtm_index = ifp->if_index;
1406         rtm->rtm_flags |= rt->rt_flags;
1407         rtm->rtm_errno = error;
1408         rtm->rtm_addrs = info.rti_addrs;
1409
1410         if (fibnum != RT_ALL_FIBS) {
1411                 M_SETFIB(m, fibnum);
1412                 m->m_flags |= RTS_FILTER_FIB;
1413         }
1414
1415         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1416
1417         return (0);
1418 }
1419
1420 /*
1421  * This is the analogue to the rt_newaddrmsg which performs the same
1422  * function but for multicast group memberhips.  This is easier since
1423  * there is no route state to worry about.
1424  */
1425 void
1426 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1427 {
1428         struct rt_addrinfo info;
1429         struct mbuf *m = NULL;
1430         struct ifnet *ifp = ifma->ifma_ifp;
1431         struct ifma_msghdr *ifmam;
1432
1433         if (V_route_cb.any_count == 0)
1434                 return;
1435
1436         bzero((caddr_t)&info, sizeof(info));
1437         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1438         if (ifp && ifp->if_addr)
1439                 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1440         else
1441                 info.rti_info[RTAX_IFP] = NULL;
1442         /*
1443          * If a link-layer address is present, present it as a ``gateway''
1444          * (similarly to how ARP entries, e.g., are presented).
1445          */
1446         info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1447         m = rtsock_msg_mbuf(cmd, &info);
1448         if (m == NULL)
1449                 return;
1450         ifmam = mtod(m, struct ifma_msghdr *);
1451         KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1452             __func__));
1453         ifmam->ifmam_index = ifp->if_index;
1454         ifmam->ifmam_addrs = info.rti_addrs;
1455         rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1456 }
1457
1458 static struct mbuf *
1459 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1460         struct rt_addrinfo *info)
1461 {
1462         struct if_announcemsghdr *ifan;
1463         struct mbuf *m;
1464
1465         if (V_route_cb.any_count == 0)
1466                 return NULL;
1467         bzero((caddr_t)info, sizeof(*info));
1468         m = rtsock_msg_mbuf(type, info);
1469         if (m != NULL) {
1470                 ifan = mtod(m, struct if_announcemsghdr *);
1471                 ifan->ifan_index = ifp->if_index;
1472                 strlcpy(ifan->ifan_name, ifp->if_xname,
1473                         sizeof(ifan->ifan_name));
1474                 ifan->ifan_what = what;
1475         }
1476         return m;
1477 }
1478
1479 /*
1480  * This is called to generate routing socket messages indicating
1481  * IEEE80211 wireless events.
1482  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1483  */
1484 void
1485 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1486 {
1487         struct mbuf *m;
1488         struct rt_addrinfo info;
1489
1490         m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1491         if (m != NULL) {
1492                 /*
1493                  * Append the ieee80211 data.  Try to stick it in the
1494                  * mbuf containing the ifannounce msg; otherwise allocate
1495                  * a new mbuf and append.
1496                  *
1497                  * NB: we assume m is a single mbuf.
1498                  */
1499                 if (data_len > M_TRAILINGSPACE(m)) {
1500                         struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1501                         if (n == NULL) {
1502                                 m_freem(m);
1503                                 return;
1504                         }
1505                         bcopy(data, mtod(n, void *), data_len);
1506                         n->m_len = data_len;
1507                         m->m_next = n;
1508                 } else if (data_len > 0) {
1509                         bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1510                         m->m_len += data_len;
1511                 }
1512                 if (m->m_flags & M_PKTHDR)
1513                         m->m_pkthdr.len += data_len;
1514                 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1515                 rt_dispatch(m, AF_UNSPEC);
1516         }
1517 }
1518
1519 /*
1520  * This is called to generate routing socket messages indicating
1521  * network interface arrival and departure.
1522  */
1523 void
1524 rt_ifannouncemsg(struct ifnet *ifp, int what)
1525 {
1526         struct mbuf *m;
1527         struct rt_addrinfo info;
1528
1529         m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1530         if (m != NULL)
1531                 rt_dispatch(m, AF_UNSPEC);
1532 }
1533
1534 static void
1535 rt_dispatch(struct mbuf *m, sa_family_t saf)
1536 {
1537         struct m_tag *tag;
1538
1539         /*
1540          * Preserve the family from the sockaddr, if any, in an m_tag for
1541          * use when injecting the mbuf into the routing socket buffer from
1542          * the netisr.
1543          */
1544         if (saf != AF_UNSPEC) {
1545                 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1546                     M_NOWAIT);
1547                 if (tag == NULL) {
1548                         m_freem(m);
1549                         return;
1550                 }
1551                 *(unsigned short *)(tag + 1) = saf;
1552                 m_tag_prepend(m, tag);
1553         }
1554 #ifdef VIMAGE
1555         if (V_loif)
1556                 m->m_pkthdr.rcvif = V_loif;
1557         else {
1558                 m_freem(m);
1559                 return;
1560         }
1561 #endif
1562         netisr_queue(NETISR_ROUTE, m);  /* mbuf is free'd on failure. */
1563 }
1564
1565 /*
1566  * This is used in dumping the kernel table via sysctl().
1567  */
1568 static int
1569 sysctl_dumpentry(struct radix_node *rn, void *vw)
1570 {
1571         struct walkarg *w = vw;
1572         struct rtentry *rt = (struct rtentry *)rn;
1573         int error = 0, size;
1574         struct rt_addrinfo info;
1575         struct sockaddr_storage ss;
1576
1577         NET_EPOCH_ASSERT();
1578
1579         if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1580                 return 0;
1581         if ((rt->rt_flags & RTF_HOST) == 0
1582             ? jailed_without_vnet(w->w_req->td->td_ucred)
1583             : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1584                 return (0);
1585         bzero((caddr_t)&info, sizeof(info));
1586         info.rti_info[RTAX_DST] = rt_key(rt);
1587         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1588         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
1589             rt_mask(rt), &ss);
1590         info.rti_info[RTAX_GENMASK] = 0;
1591         if (rt->rt_ifp && !(rt->rt_ifp->if_flags & IFF_DYING)) {
1592                 info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1593                 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1594                 if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1595                         info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1596         }
1597         if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
1598                 return (error);
1599         if (w->w_req && w->w_tmem) {
1600                 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1601
1602                 bzero(&rtm->rtm_index,
1603                     sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
1604                 if (rt->rt_flags & RTF_GWFLAG_COMPAT)
1605                         rtm->rtm_flags = RTF_GATEWAY |
1606                                 (rt->rt_flags & ~RTF_GWFLAG_COMPAT);
1607                 else
1608                         rtm->rtm_flags = rt->rt_flags;
1609                 rt_getmetrics(rt, &rtm->rtm_rmx);
1610                 rtm->rtm_index = rt->rt_ifp->if_index;
1611                 rtm->rtm_addrs = info.rti_addrs;
1612                 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1613                 return (error);
1614         }
1615         return (error);
1616 }
1617
1618 static int
1619 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
1620     struct rt_addrinfo *info, struct walkarg *w, int len)
1621 {
1622         struct if_msghdrl *ifm;
1623         struct if_data *ifd;
1624
1625         ifm = (struct if_msghdrl *)w->w_tmem;
1626
1627 #ifdef COMPAT_FREEBSD32
1628         if (w->w_req->flags & SCTL_MASK32) {
1629                 struct if_msghdrl32 *ifm32;
1630
1631                 ifm32 = (struct if_msghdrl32 *)ifm;
1632                 ifm32->ifm_addrs = info->rti_addrs;
1633                 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1634                 ifm32->ifm_index = ifp->if_index;
1635                 ifm32->_ifm_spare1 = 0;
1636                 ifm32->ifm_len = sizeof(*ifm32);
1637                 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1638                 ifm32->_ifm_spare2 = 0;
1639                 ifd = &ifm32->ifm_data;
1640         } else
1641 #endif
1642         {
1643                 ifm->ifm_addrs = info->rti_addrs;
1644                 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1645                 ifm->ifm_index = ifp->if_index;
1646                 ifm->_ifm_spare1 = 0;
1647                 ifm->ifm_len = sizeof(*ifm);
1648                 ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1649                 ifm->_ifm_spare2 = 0;
1650                 ifd = &ifm->ifm_data;
1651         }
1652
1653         memcpy(ifd, src_ifd, sizeof(*ifd));
1654
1655         return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1656 }
1657
1658 static int
1659 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
1660     struct rt_addrinfo *info, struct walkarg *w, int len)
1661 {
1662         struct if_msghdr *ifm;
1663         struct if_data *ifd;
1664
1665         ifm = (struct if_msghdr *)w->w_tmem;
1666
1667 #ifdef COMPAT_FREEBSD32
1668         if (w->w_req->flags & SCTL_MASK32) {
1669                 struct if_msghdr32 *ifm32;
1670
1671                 ifm32 = (struct if_msghdr32 *)ifm;
1672                 ifm32->ifm_addrs = info->rti_addrs;
1673                 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1674                 ifm32->ifm_index = ifp->if_index;
1675                 ifm32->_ifm_spare1 = 0;
1676                 ifd = &ifm32->ifm_data;
1677         } else
1678 #endif
1679         {
1680                 ifm->ifm_addrs = info->rti_addrs;
1681                 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1682                 ifm->ifm_index = ifp->if_index;
1683                 ifm->_ifm_spare1 = 0;
1684                 ifd = &ifm->ifm_data;
1685         }
1686
1687         memcpy(ifd, src_ifd, sizeof(*ifd));
1688
1689         return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1690 }
1691
1692 static int
1693 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1694     struct walkarg *w, int len)
1695 {
1696         struct ifa_msghdrl *ifam;
1697         struct if_data *ifd;
1698
1699         ifam = (struct ifa_msghdrl *)w->w_tmem;
1700
1701 #ifdef COMPAT_FREEBSD32
1702         if (w->w_req->flags & SCTL_MASK32) {
1703                 struct ifa_msghdrl32 *ifam32;
1704
1705                 ifam32 = (struct ifa_msghdrl32 *)ifam;
1706                 ifam32->ifam_addrs = info->rti_addrs;
1707                 ifam32->ifam_flags = ifa->ifa_flags;
1708                 ifam32->ifam_index = ifa->ifa_ifp->if_index;
1709                 ifam32->_ifam_spare1 = 0;
1710                 ifam32->ifam_len = sizeof(*ifam32);
1711                 ifam32->ifam_data_off =
1712                     offsetof(struct ifa_msghdrl32, ifam_data);
1713                 ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
1714                 ifd = &ifam32->ifam_data;
1715         } else
1716 #endif
1717         {
1718                 ifam->ifam_addrs = info->rti_addrs;
1719                 ifam->ifam_flags = ifa->ifa_flags;
1720                 ifam->ifam_index = ifa->ifa_ifp->if_index;
1721                 ifam->_ifam_spare1 = 0;
1722                 ifam->ifam_len = sizeof(*ifam);
1723                 ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1724                 ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1725                 ifd = &ifam->ifam_data;
1726         }
1727
1728         bzero(ifd, sizeof(*ifd));
1729         ifd->ifi_datalen = sizeof(struct if_data);
1730         ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
1731         ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
1732         ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
1733         ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
1734
1735         /* Fixup if_data carp(4) vhid. */
1736         if (carp_get_vhid_p != NULL)
1737                 ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
1738
1739         return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1740 }
1741
1742 static int
1743 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1744     struct walkarg *w, int len)
1745 {
1746         struct ifa_msghdr *ifam;
1747
1748         ifam = (struct ifa_msghdr *)w->w_tmem;
1749         ifam->ifam_addrs = info->rti_addrs;
1750         ifam->ifam_flags = ifa->ifa_flags;
1751         ifam->ifam_index = ifa->ifa_ifp->if_index;
1752         ifam->_ifam_spare1 = 0;
1753         ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1754
1755         return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1756 }
1757
1758 static int
1759 sysctl_iflist(int af, struct walkarg *w)
1760 {
1761         struct ifnet *ifp;
1762         struct ifaddr *ifa;
1763         struct if_data ifd;
1764         struct rt_addrinfo info;
1765         int len, error = 0;
1766         struct sockaddr_storage ss;
1767
1768         bzero((caddr_t)&info, sizeof(info));
1769         bzero(&ifd, sizeof(ifd));
1770         CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1771                 if (w->w_arg && w->w_arg != ifp->if_index)
1772                         continue;
1773                 if_data_copy(ifp, &ifd);
1774                 ifa = ifp->if_addr;
1775                 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1776                 error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
1777                 if (error != 0)
1778                         goto done;
1779                 info.rti_info[RTAX_IFP] = NULL;
1780                 if (w->w_req && w->w_tmem) {
1781                         if (w->w_op == NET_RT_IFLISTL)
1782                                 error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
1783                                     len);
1784                         else
1785                                 error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
1786                                     len);
1787                         if (error)
1788                                 goto done;
1789                 }
1790                 while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
1791                         if (af && af != ifa->ifa_addr->sa_family)
1792                                 continue;
1793                         if (prison_if(w->w_req->td->td_ucred,
1794                             ifa->ifa_addr) != 0)
1795                                 continue;
1796                         info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1797                         info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1798                             ifa->ifa_addr, ifa->ifa_netmask, &ss);
1799                         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1800                         error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
1801                         if (error != 0)
1802                                 goto done;
1803                         if (w->w_req && w->w_tmem) {
1804                                 if (w->w_op == NET_RT_IFLISTL)
1805                                         error = sysctl_iflist_ifaml(ifa, &info,
1806                                             w, len);
1807                                 else
1808                                         error = sysctl_iflist_ifam(ifa, &info,
1809                                             w, len);
1810                                 if (error)
1811                                         goto done;
1812                         }
1813                 }
1814                 info.rti_info[RTAX_IFA] = NULL;
1815                 info.rti_info[RTAX_NETMASK] = NULL;
1816                 info.rti_info[RTAX_BRD] = NULL;
1817         }
1818 done:
1819         return (error);
1820 }
1821
1822 static int
1823 sysctl_ifmalist(int af, struct walkarg *w)
1824 {
1825         struct rt_addrinfo info;
1826         struct ifaddr *ifa;
1827         struct ifmultiaddr *ifma;
1828         struct ifnet *ifp;
1829         int error, len;
1830
1831         NET_EPOCH_ASSERT();
1832
1833         error = 0;
1834         bzero((caddr_t)&info, sizeof(info));
1835
1836         CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1837                 if (w->w_arg && w->w_arg != ifp->if_index)
1838                         continue;
1839                 ifa = ifp->if_addr;
1840                 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1841                 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1842                         if (af && af != ifma->ifma_addr->sa_family)
1843                                 continue;
1844                         if (prison_if(w->w_req->td->td_ucred,
1845                             ifma->ifma_addr) != 0)
1846                                 continue;
1847                         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1848                         info.rti_info[RTAX_GATEWAY] =
1849                             (ifma->ifma_addr->sa_family != AF_LINK) ?
1850                             ifma->ifma_lladdr : NULL;
1851                         error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
1852                         if (error != 0)
1853                                 break;
1854                         if (w->w_req && w->w_tmem) {
1855                                 struct ifma_msghdr *ifmam;
1856
1857                                 ifmam = (struct ifma_msghdr *)w->w_tmem;
1858                                 ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1859                                 ifmam->ifmam_flags = 0;
1860                                 ifmam->ifmam_addrs = info.rti_addrs;
1861                                 ifmam->_ifmam_spare1 = 0;
1862                                 error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1863                                 if (error != 0)
1864                                         break;
1865                         }
1866                 }
1867                 if (error != 0)
1868                         break;
1869         }
1870         return (error);
1871 }
1872
1873 static int
1874 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1875 {
1876         RIB_RLOCK_TRACKER;
1877         struct epoch_tracker et;
1878         int     *name = (int *)arg1;
1879         u_int   namelen = arg2;
1880         struct rib_head *rnh = NULL; /* silence compiler. */
1881         int     i, lim, error = EINVAL;
1882         int     fib = 0;
1883         u_char  af;
1884         struct  walkarg w;
1885
1886         name ++;
1887         namelen--;
1888         if (req->newptr)
1889                 return (EPERM);
1890         if (name[1] == NET_RT_DUMP) {
1891                 if (namelen == 3)
1892                         fib = req->td->td_proc->p_fibnum;
1893                 else if (namelen == 4)
1894                         fib = (name[3] == RT_ALL_FIBS) ?
1895                             req->td->td_proc->p_fibnum : name[3];
1896                 else
1897                         return ((namelen < 3) ? EISDIR : ENOTDIR);
1898                 if (fib < 0 || fib >= rt_numfibs)
1899                         return (EINVAL);
1900         } else if (namelen != 3)
1901                 return ((namelen < 3) ? EISDIR : ENOTDIR);
1902         af = name[0];
1903         if (af > AF_MAX)
1904                 return (EINVAL);
1905         bzero(&w, sizeof(w));
1906         w.w_op = name[1];
1907         w.w_arg = name[2];
1908         w.w_req = req;
1909
1910         error = sysctl_wire_old_buffer(req, 0);
1911         if (error)
1912                 return (error);
1913
1914         /*
1915          * Allocate reply buffer in advance.
1916          * All rtsock messages has maximum length of u_short.
1917          */
1918         w.w_tmemsize = 65536;
1919         w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
1920
1921         NET_EPOCH_ENTER(et);
1922         switch (w.w_op) {
1923         case NET_RT_DUMP:
1924         case NET_RT_FLAGS:
1925                 if (af == 0) {                  /* dump all tables */
1926                         i = 1;
1927                         lim = AF_MAX;
1928                 } else                          /* dump only one table */
1929                         i = lim = af;
1930
1931                 /*
1932                  * take care of llinfo entries, the caller must
1933                  * specify an AF
1934                  */
1935                 if (w.w_op == NET_RT_FLAGS &&
1936                     (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1937                         if (af != 0)
1938                                 error = lltable_sysctl_dumparp(af, w.w_req);
1939                         else
1940                                 error = EINVAL;
1941                         break;
1942                 }
1943                 /*
1944                  * take care of routing entries
1945                  */
1946                 for (error = 0; error == 0 && i <= lim; i++) {
1947                         rnh = rt_tables_get_rnh(fib, i);
1948                         if (rnh != NULL) {
1949                                 RIB_RLOCK(rnh);
1950                                 error = rnh->rnh_walktree(&rnh->head,
1951                                     sysctl_dumpentry, &w);
1952                                 RIB_RUNLOCK(rnh);
1953                         } else if (af != 0)
1954                                 error = EAFNOSUPPORT;
1955                 }
1956                 break;
1957
1958         case NET_RT_IFLIST:
1959         case NET_RT_IFLISTL:
1960                 error = sysctl_iflist(af, &w);
1961                 break;
1962
1963         case NET_RT_IFMALIST:
1964                 error = sysctl_ifmalist(af, &w);
1965                 break;
1966         }
1967         NET_EPOCH_EXIT(et);
1968
1969         free(w.w_tmem, M_TEMP);
1970         return (error);
1971 }
1972
1973 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1974
1975 /*
1976  * Definitions of protocols supported in the ROUTE domain.
1977  */
1978
1979 static struct domain routedomain;               /* or at least forward */
1980
1981 static struct protosw routesw[] = {
1982 {
1983         .pr_type =              SOCK_RAW,
1984         .pr_domain =            &routedomain,
1985         .pr_flags =             PR_ATOMIC|PR_ADDR,
1986         .pr_output =            route_output,
1987         .pr_ctlinput =          raw_ctlinput,
1988         .pr_init =              raw_init,
1989         .pr_usrreqs =           &route_usrreqs
1990 }
1991 };
1992
1993 static struct domain routedomain = {
1994         .dom_family =           PF_ROUTE,
1995         .dom_name =              "route",
1996         .dom_protosw =          routesw,
1997         .dom_protoswNPROTOSW =  &routesw[nitems(routesw)]
1998 };
1999
2000 VNET_DOMAIN_SET(route);
2001
2002 #ifdef DDB
2003 /*
2004  * Unfortunately, RTF_ values are expressed as raw masks rather than powers of
2005  * 2, so we cannot use them as nice C99 initializer indices below.
2006  */
2007 static const char * const rtf_flag_strings[] = {
2008         "UP",
2009         "GATEWAY",
2010         "HOST",
2011         "REJECT",
2012         "DYNAMIC",
2013         "MODIFIED",
2014         "DONE",
2015         "UNUSED_0x80",
2016         "UNUSED_0x100",
2017         "XRESOLVE",
2018         "LLDATA",
2019         "STATIC",
2020         "BLACKHOLE",
2021         "UNUSED_0x2000",
2022         "PROTO2",
2023         "PROTO1",
2024         "UNUSED_0x10000",
2025         "UNUSED_0x20000",
2026         "PROTO3",
2027         "FIXEDMTU",
2028         "PINNED",
2029         "LOCAL",
2030         "BROADCAST",
2031         "MULTICAST",
2032         /* Big gap. */
2033         [28] = "STICKY",
2034         [30] = "RNH_LOCKED",
2035         [31] = "GWFLAG_COMPAT",
2036 };
2037
2038 static const char * __pure
2039 rt_flag_name(unsigned idx)
2040 {
2041         if (idx >= nitems(rtf_flag_strings))
2042                 return ("INVALID_FLAG");
2043         if (rtf_flag_strings[idx] == NULL)
2044                 return ("UNKNOWN");
2045         return (rtf_flag_strings[idx]);
2046 }
2047
2048 static void
2049 rt_dumpaddr_ddb(const char *name, const struct sockaddr *sa)
2050 {
2051         char buf[INET6_ADDRSTRLEN], *res;
2052
2053         res = NULL;
2054         if (sa == NULL)
2055                 res = "NULL";
2056         else if (sa->sa_family == AF_INET) {
2057                 res = inet_ntop(AF_INET,
2058                     &((const struct sockaddr_in *)sa)->sin_addr,
2059                     buf, sizeof(buf));
2060         } else if (sa->sa_family == AF_INET6) {
2061                 res = inet_ntop(AF_INET6,
2062                     &((const struct sockaddr_in6 *)sa)->sin6_addr,
2063                     buf, sizeof(buf));
2064         } else if (sa->sa_family == AF_LINK) {
2065                 res = "on link";
2066         }
2067
2068         if (res != NULL) {
2069                 db_printf("%s <%s> ", name, res);
2070                 return;
2071         }
2072
2073         db_printf("%s <af:%d> ", name, sa->sa_family);
2074 }
2075
2076 static int
2077 rt_dumpentry_ddb(struct radix_node *rn, void *arg __unused)
2078 {
2079         struct sockaddr_storage ss;
2080         struct rtentry *rt;
2081         int flags, idx;
2082
2083         /* If RNTORT is important, put it in a header. */
2084         rt = (void *)rn;
2085
2086         rt_dumpaddr_ddb("dst", rt_key(rt));
2087         rt_dumpaddr_ddb("gateway", rt->rt_gateway);
2088         rt_dumpaddr_ddb("netmask", rtsock_fix_netmask(rt_key(rt), rt_mask(rt),
2089             &ss));
2090         if (rt->rt_ifp != NULL && (rt->rt_ifp->if_flags & IFF_DYING) == 0) {
2091                 rt_dumpaddr_ddb("ifp", rt->rt_ifp->if_addr->ifa_addr);
2092                 rt_dumpaddr_ddb("ifa", rt->rt_ifa->ifa_addr);
2093         }
2094
2095         db_printf("flags ");
2096         flags = rt->rt_flags;
2097         if (flags == 0)
2098                 db_printf("none");
2099
2100         while ((idx = ffs(flags)) > 0) {
2101                 idx--;
2102
2103                 if (flags != rt->rt_flags)
2104                         db_printf(",");
2105                 db_printf("%s", rt_flag_name(idx));
2106
2107                 flags &= ~(1ul << idx);
2108         }
2109
2110         db_printf("\n");
2111         return (0);
2112 }
2113
2114 DB_SHOW_COMMAND(routetable, db_show_routetable_cmd)
2115 {
2116         struct rib_head *rnh;
2117         int error, i, lim;
2118
2119         if (have_addr)
2120                 i = lim = addr;
2121         else {
2122                 i = 1;
2123                 lim = AF_MAX;
2124         }
2125
2126         for (; i <= lim; i++) {
2127                 rnh = rt_tables_get_rnh(0, i);
2128                 if (rnh == NULL) {
2129                         if (have_addr) {
2130                                 db_printf("%s: AF %d not supported?\n",
2131                                     __func__, i);
2132                                 break;
2133                         }
2134                         continue;
2135                 }
2136
2137                 if (!have_addr && i > 1)
2138                         db_printf("\n");
2139
2140                 db_printf("Route table for AF %d%s%s%s:\n", i,
2141                     (i == AF_INET || i == AF_INET6) ? " (" : "",
2142                     (i == AF_INET) ? "INET" : (i == AF_INET6) ? "INET6" : "",
2143                     (i == AF_INET || i == AF_INET6) ? ")" : "");
2144
2145                 error = rnh->rnh_walktree(&rnh->head, rt_dumpentry_ddb, NULL);
2146                 if (error != 0)
2147                         db_printf("%s: walktree(%d): %d\n", __func__, i,
2148                             error);
2149         }
2150 }
2151
2152 _DB_FUNC(_show, route, db_show_route_cmd, db_show_table, CS_OWN, NULL)
2153 {
2154         char buf[INET6_ADDRSTRLEN], *bp;
2155         const void *dst_addrp;
2156         struct sockaddr *dstp;
2157         struct rtentry *rt;
2158         union {
2159                 struct sockaddr_in dest_sin;
2160                 struct sockaddr_in6 dest_sin6;
2161         } u;
2162         uint16_t hextets[8];
2163         unsigned i, tets;
2164         int t, af, exp, tokflags;
2165
2166         /*
2167          * Undecoded address family.  No double-colon expansion seen yet.
2168          */
2169         af = -1;
2170         exp = -1;
2171         /* Assume INET6 to start; we can work back if guess was wrong. */
2172         tokflags = DRT_WSPACE | DRT_HEX | DRT_HEXADECIMAL;
2173
2174         /*
2175          * db_command has lexed 'show route' for us.
2176          */
2177         t = db_read_token_flags(tokflags);
2178         if (t == tWSPACE)
2179                 t = db_read_token_flags(tokflags);
2180
2181         /*
2182          * tEOL: Just 'show route' isn't a valid mode.
2183          * tMINUS: It's either '-h' or some invalid option.  Regardless, usage.
2184          */
2185         if (t == tEOL || t == tMINUS)
2186                 goto usage;
2187
2188         db_unread_token(t);
2189
2190         tets = nitems(hextets);
2191
2192         /*
2193          * Each loop iteration, we expect to read one octet (v4) or hextet
2194          * (v6), followed by an appropriate field separator ('.' or ':' or
2195          * '::').
2196          *
2197          * At the start of each loop, we're looking for a number (octet or
2198          * hextet).
2199          *
2200          * INET6 addresses have a special case where they may begin with '::'.
2201          */
2202         for (i = 0; i < tets; i++) {
2203                 t = db_read_token_flags(tokflags);
2204
2205                 if (t == tCOLONCOLON) {
2206                         /* INET6 with leading '::' or invalid. */
2207                         if (i != 0) {
2208                                 db_printf("Parse error: unexpected extra "
2209                                     "colons.\n");
2210                                 goto exit;
2211                         }
2212
2213                         af = AF_INET6;
2214                         exp = i;
2215                         hextets[i] = 0;
2216                         continue;
2217                 } else if (t == tNUMBER) {
2218                         /*
2219                          * Lexer separates out '-' as tMINUS, but make the
2220                          * assumption explicit here.
2221                          */
2222                         MPASS(db_tok_number >= 0);
2223
2224                         if (af == AF_INET && db_tok_number > UINT8_MAX) {
2225                                 db_printf("Not a valid v4 octet: %ld\n",
2226                                     (long)db_tok_number);
2227                                 goto exit;
2228                         }
2229                         hextets[i] = db_tok_number;
2230                 } else if (t == tEOL) {
2231                         /*
2232                          * We can only detect the end of an IPv6 address in
2233                          * compact representation with EOL.
2234                          */
2235                         if (af != AF_INET6 || exp < 0) {
2236                                 db_printf("Parse failed.  Got unexpected EOF "
2237                                     "when the address is not a compact-"
2238                                     "representation IPv6 address.\n");
2239                                 goto exit;
2240                         }
2241                         break;
2242                 } else {
2243                         db_printf("Parse failed.  Unexpected token %d.\n", t);
2244                         goto exit;
2245                 }
2246
2247                 /* Next, look for a separator, if appropriate. */
2248                 if (i == tets - 1)
2249                         continue;
2250
2251                 t = db_read_token_flags(tokflags);
2252                 if (af < 0) {
2253                         if (t == tCOLON) {
2254                                 af = AF_INET6;
2255                                 continue;
2256                         }
2257                         if (t == tCOLONCOLON) {
2258                                 af = AF_INET6;
2259                                 i++;
2260                                 hextets[i] = 0;
2261                                 exp = i;
2262                                 continue;
2263                         }
2264                         if (t == tDOT) {
2265                                 unsigned hn, dn;
2266
2267                                 af = AF_INET;
2268                                 /* Need to fixup the first parsed number. */
2269                                 if (hextets[0] > 0x255 ||
2270                                     (hextets[0] & 0xf0) > 0x90 ||
2271                                     (hextets[0] & 0xf) > 9) {
2272                                         db_printf("Not a valid v4 octet: %x\n",
2273                                             hextets[0]);
2274                                         goto exit;
2275                                 }
2276
2277                                 hn = hextets[0];
2278                                 dn = (hn >> 8) * 100 +
2279                                     ((hn >> 4) & 0xf) * 10 +
2280                                     (hn & 0xf);
2281
2282                                 hextets[0] = dn;
2283
2284                                 /* Switch to decimal for remaining octets. */
2285                                 tokflags &= ~DRT_RADIX_MASK;
2286                                 tokflags |= DRT_DECIMAL;
2287
2288                                 tets = 4;
2289                                 continue;
2290                         }
2291
2292                         db_printf("Parse error.  Unexpected token %d.\n", t);
2293                         goto exit;
2294                 } else if (af == AF_INET) {
2295                         if (t == tDOT)
2296                                 continue;
2297                         db_printf("Expected '.' (%d) between octets but got "
2298                             "(%d).\n", tDOT, t);
2299                         goto exit;
2300
2301                 } else if (af == AF_INET6) {
2302                         if (t == tCOLON)
2303                                 continue;
2304                         if (t == tCOLONCOLON) {
2305                                 if (exp < 0) {
2306                                         i++;
2307                                         hextets[i] = 0;
2308                                         exp = i;
2309                                         continue;
2310                                 }
2311                                 db_printf("Got bogus second '::' in v6 "
2312                                     "address.\n");
2313                                 goto exit;
2314                         }
2315                         if (t == tEOL) {
2316                                 /*
2317                                  * Handle in the earlier part of the loop
2318                                  * because we need to handle trailing :: too.
2319                                  */
2320                                 db_unread_token(t);
2321                                 continue;
2322                         }
2323
2324                         db_printf("Expected ':' (%d) or '::' (%d) between "
2325                             "hextets but got (%d).\n", tCOLON, tCOLONCOLON, t);
2326                         goto exit;
2327                 }
2328         }
2329
2330         /* Check for trailing garbage. */
2331         if (i == tets) {
2332                 t = db_read_token_flags(tokflags);
2333                 if (t != tEOL) {
2334                         db_printf("Got unexpected garbage after address "
2335                             "(%d).\n", t);
2336                         goto exit;
2337                 }
2338         }
2339
2340         /*
2341          * Need to expand compact INET6 addresses.
2342          *
2343          * Technically '::' for a single ':0:' is MUST NOT but just in case,
2344          * don't bother expanding that form (exp >= 0 && i == tets case).
2345          */
2346         if (af == AF_INET6 && exp >= 0 && i < tets) {
2347                 if (exp + 1 < i) {
2348                         memmove(&hextets[exp + 1 + (nitems(hextets) - i)],
2349                             &hextets[exp + 1],
2350                             (i - (exp + 1)) * sizeof(hextets[0]));
2351                 }
2352                 memset(&hextets[exp + 1], 0, (nitems(hextets) - i) *
2353                     sizeof(hextets[0]));
2354         }
2355
2356         memset(&u, 0, sizeof(u));
2357         if (af == AF_INET) {
2358                 u.dest_sin.sin_family = AF_INET;
2359                 u.dest_sin.sin_len = sizeof(u.dest_sin);
2360                 u.dest_sin.sin_addr.s_addr = htonl(
2361                     ((uint32_t)hextets[0] << 24) |
2362                     ((uint32_t)hextets[1] << 16) |
2363                     ((uint32_t)hextets[2] << 8) |
2364                     (uint32_t)hextets[3]);
2365                 dstp = (void *)&u.dest_sin;
2366                 dst_addrp = &u.dest_sin.sin_addr;
2367         } else if (af == AF_INET6) {
2368                 u.dest_sin6.sin6_family = AF_INET6;
2369                 u.dest_sin6.sin6_len = sizeof(u.dest_sin6);
2370                 for (i = 0; i < nitems(hextets); i++)
2371                         u.dest_sin6.sin6_addr.s6_addr16[i] = htons(hextets[i]);
2372                 dstp = (void *)&u.dest_sin6;
2373                 dst_addrp = &u.dest_sin6.sin6_addr;
2374         } else {
2375                 MPASS(false);
2376                 /* UNREACHABLE */
2377                 /* Appease Clang false positive: */
2378                 dstp = NULL;
2379         }
2380
2381         bp = inet_ntop(af, dst_addrp, buf, sizeof(buf));
2382         if (bp != NULL)
2383                 db_printf("Looking up route to destination '%s'\n", bp);
2384
2385         CURVNET_SET(vnet0);
2386         rt = rtalloc1(dstp, 0, RTF_RNH_LOCKED);
2387         CURVNET_RESTORE();
2388
2389         if (rt == NULL) {
2390                 db_printf("Could not get route for that server.\n");
2391                 return;
2392         }
2393
2394         rt_dumpentry_ddb((void *)rt, NULL);
2395         RTFREE_LOCKED(rt);
2396
2397         return;
2398 usage:
2399         db_printf("Usage: 'show route <address>'\n"
2400             "  Currently accepts only dotted-decimal INET or colon-separated\n"
2401             "  hextet INET6 addresses.\n");
2402 exit:
2403         db_skip_to_eol();
2404 }
2405 #endif