1 /*- 2 * Copyright (c) 1988, 1991, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)rtsock.c 8.7 (Berkeley) 10/12/95 30 * $FreeBSD$ 31 */ 32 #include "opt_compat.h" 33 #include "opt_mpath.h" 34 #include "opt_inet.h" 35 #include "opt_inet6.h" 36 37 #include <sys/param.h> 38 #include <sys/jail.h> 39 #include <sys/kernel.h> 40 #include <sys/domain.h> 41 #include <sys/lock.h> 42 #include <sys/malloc.h> 43 #include <sys/mbuf.h> 44 #include <sys/priv.h> 45 #include <sys/proc.h> 46 #include <sys/protosw.h> 47 #include <sys/rwlock.h> 48 #include <sys/signalvar.h> 49 #include <sys/socket.h> 50 #include <sys/socketvar.h> 51 #include <sys/sysctl.h> 52 #include <sys/systm.h> 53 54 #include <net/if.h> 55 #include <net/if_var.h> 56 #include <net/if_dl.h> 57 #include <net/if_llatbl.h> 58 #include <net/if_types.h> 59 #include <net/netisr.h> 60 #include <net/raw_cb.h> 61 #include <net/route.h> 62 #include <net/vnet.h> 63 64 #include <netinet/in.h> 65 #include <netinet/if_ether.h> 66 #include <netinet/ip_carp.h> 67 #ifdef INET6 68 #include <netinet6/ip6_var.h> 69 #include <netinet6/scope6_var.h> 70 #endif 71 72 #ifdef COMPAT_FREEBSD32 73 #include <sys/mount.h> 74 #include <compat/freebsd32/freebsd32.h> 75 76 struct if_msghdr32 { 77 uint16_t ifm_msglen; 78 uint8_t ifm_version; 79 uint8_t ifm_type; 80 int32_t ifm_addrs; 81 int32_t ifm_flags; 82 uint16_t ifm_index; 83 struct if_data ifm_data; 84 }; 85 86 struct if_msghdrl32 { 87 uint16_t ifm_msglen; 88 uint8_t ifm_version; 89 uint8_t ifm_type; 90 int32_t ifm_addrs; 91 int32_t ifm_flags; 92 uint16_t ifm_index; 93 uint16_t _ifm_spare1; 94 uint16_t ifm_len; 95 uint16_t ifm_data_off; 96 struct if_data ifm_data; 97 }; 98 99 struct ifa_msghdrl32 { 100 uint16_t ifam_msglen; 101 uint8_t ifam_version; 102 uint8_t ifam_type; 103 int32_t ifam_addrs; 104 int32_t ifam_flags; 105 uint16_t ifam_index; 106 uint16_t _ifam_spare1; 107 uint16_t ifam_len; 108 uint16_t ifam_data_off; 109 int32_t ifam_metric; 110 struct if_data ifam_data; 111 }; 112 #endif /* COMPAT_FREEBSD32 */ 113 114 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables"); 115 116 /* NB: these are not modified */ 117 static struct sockaddr route_src = { 2, PF_ROUTE, }; 118 static struct sockaddr sa_zero = { sizeof(sa_zero), AF_INET, }; 119 120 /* These are external hooks for CARP. */ 121 int (*carp_get_vhid_p)(struct ifaddr *); 122 123 /* 124 * Used by rtsock/raw_input callback code to decide whether to filter the update 125 * notification to a socket bound to a particular FIB. 126 */ 127 #define RTS_FILTER_FIB M_PROTO8 128 129 typedef struct { 130 int ip_count; /* attached w/ AF_INET */ 131 int ip6_count; /* attached w/ AF_INET6 */ 132 int any_count; /* total attached */ 133 } route_cb_t; 134 static VNET_DEFINE(route_cb_t, route_cb); 135 #define V_route_cb VNET(route_cb) 136 137 struct mtx rtsock_mtx; 138 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF); 139 140 #define RTSOCK_LOCK() mtx_lock(&rtsock_mtx) 141 #define RTSOCK_UNLOCK() mtx_unlock(&rtsock_mtx) 142 #define RTSOCK_LOCK_ASSERT() mtx_assert(&rtsock_mtx, MA_OWNED) 143 144 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, ""); 145 146 struct walkarg { 147 int w_tmemsize; 148 int w_op, w_arg; 149 caddr_t w_tmem; 150 struct sysctl_req *w_req; 151 }; 152 153 static void rts_input(struct mbuf *m); 154 static struct mbuf *rt_msg1(int type, struct rt_addrinfo *rtinfo); 155 static int rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, 156 struct walkarg *w, int *plen); 157 static int rt_xaddrs(caddr_t cp, caddr_t cplim, 158 struct rt_addrinfo *rtinfo); 159 static int sysctl_dumpentry(struct radix_node *rn, void *vw); 160 static int sysctl_iflist(int af, struct walkarg *w); 161 static int sysctl_ifmalist(int af, struct walkarg *w); 162 static int route_output(struct mbuf *m, struct socket *so); 163 static void rt_setmetrics(const struct rt_msghdr *rtm, struct rtentry *rt); 164 static void rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out); 165 static void rt_dispatch(struct mbuf *, sa_family_t); 166 167 static struct netisr_handler rtsock_nh = { 168 .nh_name = "rtsock", 169 .nh_handler = rts_input, 170 .nh_proto = NETISR_ROUTE, 171 .nh_policy = NETISR_POLICY_SOURCE, 172 }; 173 174 static int 175 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS) 176 { 177 int error, qlimit; 178 179 netisr_getqlimit(&rtsock_nh, &qlimit); 180 error = sysctl_handle_int(oidp, &qlimit, 0, req); 181 if (error || !req->newptr) 182 return (error); 183 if (qlimit < 1) 184 return (EINVAL); 185 return (netisr_setqlimit(&rtsock_nh, qlimit)); 186 } 187 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW, 188 0, 0, sysctl_route_netisr_maxqlen, "I", 189 "maximum routing socket dispatch queue length"); 190 191 static void 192 rts_init(void) 193 { 194 int tmp; 195 196 if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp)) 197 rtsock_nh.nh_qlimit = tmp; 198 netisr_register(&rtsock_nh); 199 } 200 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0); 201 202 static int 203 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src, 204 struct rawcb *rp) 205 { 206 int fibnum; 207 208 KASSERT(m != NULL, ("%s: m is NULL", __func__)); 209 KASSERT(proto != NULL, ("%s: proto is NULL", __func__)); 210 KASSERT(rp != NULL, ("%s: rp is NULL", __func__)); 211 212 /* No filtering requested. */ 213 if ((m->m_flags & RTS_FILTER_FIB) == 0) 214 return (0); 215 216 /* Check if it is a rts and the fib matches the one of the socket. */ 217 fibnum = M_GETFIB(m); 218 if (proto->sp_family != PF_ROUTE || 219 rp->rcb_socket == NULL || 220 rp->rcb_socket->so_fibnum == fibnum) 221 return (0); 222 223 /* Filtering requested and no match, the socket shall be skipped. */ 224 return (1); 225 } 226 227 static void 228 rts_input(struct mbuf *m) 229 { 230 struct sockproto route_proto; 231 unsigned short *family; 232 struct m_tag *tag; 233 234 route_proto.sp_family = PF_ROUTE; 235 tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL); 236 if (tag != NULL) { 237 family = (unsigned short *)(tag + 1); 238 route_proto.sp_protocol = *family; 239 m_tag_delete(m, tag); 240 } else 241 route_proto.sp_protocol = 0; 242 243 raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb); 244 } 245 246 /* 247 * It really doesn't make any sense at all for this code to share much 248 * with raw_usrreq.c, since its functionality is so restricted. XXX 249 */ 250 static void 251 rts_abort(struct socket *so) 252 { 253 254 raw_usrreqs.pru_abort(so); 255 } 256 257 static void 258 rts_close(struct socket *so) 259 { 260 261 raw_usrreqs.pru_close(so); 262 } 263 264 /* pru_accept is EOPNOTSUPP */ 265 266 static int 267 rts_attach(struct socket *so, int proto, struct thread *td) 268 { 269 struct rawcb *rp; 270 int error; 271 272 KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL")); 273 274 /* XXX */ 275 rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO); 276 if (rp == NULL) 277 return ENOBUFS; 278 279 so->so_pcb = (caddr_t)rp; 280 so->so_fibnum = td->td_proc->p_fibnum; 281 error = raw_attach(so, proto); 282 rp = sotorawcb(so); 283 if (error) { 284 so->so_pcb = NULL; 285 free(rp, M_PCB); 286 return error; 287 } 288 RTSOCK_LOCK(); 289 switch(rp->rcb_proto.sp_protocol) { 290 case AF_INET: 291 V_route_cb.ip_count++; 292 break; 293 case AF_INET6: 294 V_route_cb.ip6_count++; 295 break; 296 } 297 V_route_cb.any_count++; 298 RTSOCK_UNLOCK(); 299 soisconnected(so); 300 so->so_options |= SO_USELOOPBACK; 301 return 0; 302 } 303 304 static int 305 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 306 { 307 308 return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */ 309 } 310 311 static int 312 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 313 { 314 315 return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */ 316 } 317 318 /* pru_connect2 is EOPNOTSUPP */ 319 /* pru_control is EOPNOTSUPP */ 320 321 static void 322 rts_detach(struct socket *so) 323 { 324 struct rawcb *rp = sotorawcb(so); 325 326 KASSERT(rp != NULL, ("rts_detach: rp == NULL")); 327 328 RTSOCK_LOCK(); 329 switch(rp->rcb_proto.sp_protocol) { 330 case AF_INET: 331 V_route_cb.ip_count--; 332 break; 333 case AF_INET6: 334 V_route_cb.ip6_count--; 335 break; 336 } 337 V_route_cb.any_count--; 338 RTSOCK_UNLOCK(); 339 raw_usrreqs.pru_detach(so); 340 } 341 342 static int 343 rts_disconnect(struct socket *so) 344 { 345 346 return (raw_usrreqs.pru_disconnect(so)); 347 } 348 349 /* pru_listen is EOPNOTSUPP */ 350 351 static int 352 rts_peeraddr(struct socket *so, struct sockaddr **nam) 353 { 354 355 return (raw_usrreqs.pru_peeraddr(so, nam)); 356 } 357 358 /* pru_rcvd is EOPNOTSUPP */ 359 /* pru_rcvoob is EOPNOTSUPP */ 360 361 static int 362 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 363 struct mbuf *control, struct thread *td) 364 { 365 366 return (raw_usrreqs.pru_send(so, flags, m, nam, control, td)); 367 } 368 369 /* pru_sense is null */ 370 371 static int 372 rts_shutdown(struct socket *so) 373 { 374 375 return (raw_usrreqs.pru_shutdown(so)); 376 } 377 378 static int 379 rts_sockaddr(struct socket *so, struct sockaddr **nam) 380 { 381 382 return (raw_usrreqs.pru_sockaddr(so, nam)); 383 } 384 385 static struct pr_usrreqs route_usrreqs = { 386 .pru_abort = rts_abort, 387 .pru_attach = rts_attach, 388 .pru_bind = rts_bind, 389 .pru_connect = rts_connect, 390 .pru_detach = rts_detach, 391 .pru_disconnect = rts_disconnect, 392 .pru_peeraddr = rts_peeraddr, 393 .pru_send = rts_send, 394 .pru_shutdown = rts_shutdown, 395 .pru_sockaddr = rts_sockaddr, 396 .pru_close = rts_close, 397 }; 398 399 #ifndef _SOCKADDR_UNION_DEFINED 400 #define _SOCKADDR_UNION_DEFINED 401 /* 402 * The union of all possible address formats we handle. 403 */ 404 union sockaddr_union { 405 struct sockaddr sa; 406 struct sockaddr_in sin; 407 struct sockaddr_in6 sin6; 408 }; 409 #endif /* _SOCKADDR_UNION_DEFINED */ 410 411 static int 412 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp, 413 struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred) 414 { 415 416 /* First, see if the returned address is part of the jail. */ 417 if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) { 418 info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr; 419 return (0); 420 } 421 422 switch (info->rti_info[RTAX_DST]->sa_family) { 423 #ifdef INET 424 case AF_INET: 425 { 426 struct in_addr ia; 427 struct ifaddr *ifa; 428 int found; 429 430 found = 0; 431 /* 432 * Try to find an address on the given outgoing interface 433 * that belongs to the jail. 434 */ 435 IF_ADDR_RLOCK(ifp); 436 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 437 struct sockaddr *sa; 438 sa = ifa->ifa_addr; 439 if (sa->sa_family != AF_INET) 440 continue; 441 ia = ((struct sockaddr_in *)sa)->sin_addr; 442 if (prison_check_ip4(cred, &ia) == 0) { 443 found = 1; 444 break; 445 } 446 } 447 IF_ADDR_RUNLOCK(ifp); 448 if (!found) { 449 /* 450 * As a last resort return the 'default' jail address. 451 */ 452 ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)-> 453 sin_addr; 454 if (prison_get_ip4(cred, &ia) != 0) 455 return (ESRCH); 456 } 457 bzero(&saun->sin, sizeof(struct sockaddr_in)); 458 saun->sin.sin_len = sizeof(struct sockaddr_in); 459 saun->sin.sin_family = AF_INET; 460 saun->sin.sin_addr.s_addr = ia.s_addr; 461 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin; 462 break; 463 } 464 #endif 465 #ifdef INET6 466 case AF_INET6: 467 { 468 struct in6_addr ia6; 469 struct ifaddr *ifa; 470 int found; 471 472 found = 0; 473 /* 474 * Try to find an address on the given outgoing interface 475 * that belongs to the jail. 476 */ 477 IF_ADDR_RLOCK(ifp); 478 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 479 struct sockaddr *sa; 480 sa = ifa->ifa_addr; 481 if (sa->sa_family != AF_INET6) 482 continue; 483 bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr, 484 &ia6, sizeof(struct in6_addr)); 485 if (prison_check_ip6(cred, &ia6) == 0) { 486 found = 1; 487 break; 488 } 489 } 490 IF_ADDR_RUNLOCK(ifp); 491 if (!found) { 492 /* 493 * As a last resort return the 'default' jail address. 494 */ 495 ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)-> 496 sin6_addr; 497 if (prison_get_ip6(cred, &ia6) != 0) 498 return (ESRCH); 499 } 500 bzero(&saun->sin6, sizeof(struct sockaddr_in6)); 501 saun->sin6.sin6_len = sizeof(struct sockaddr_in6); 502 saun->sin6.sin6_family = AF_INET6; 503 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr)); 504 if (sa6_recoverscope(&saun->sin6) != 0) 505 return (ESRCH); 506 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6; 507 break; 508 } 509 #endif 510 default: 511 return (ESRCH); 512 } 513 return (0); 514 } 515 516 /*ARGSUSED*/ 517 static int 518 route_output(struct mbuf *m, struct socket *so) 519 { 520 struct rt_msghdr *rtm = NULL; 521 struct rtentry *rt = NULL; 522 struct radix_node_head *rnh; 523 struct rt_addrinfo info; 524 #ifdef INET6 525 struct sockaddr_storage ss; 526 struct sockaddr_in6 *sin6; 527 int i, rti_need_deembed = 0; 528 #endif 529 int alloc_len = 0, len, error = 0, fibnum; 530 struct ifnet *ifp = NULL; 531 union sockaddr_union saun; 532 sa_family_t saf = AF_UNSPEC; 533 struct rawcb *rp = NULL; 534 struct walkarg w; 535 char msgbuf[512]; 536 537 fibnum = so->so_fibnum; 538 539 #define senderr(e) { error = e; goto flush;} 540 if (m == NULL || ((m->m_len < sizeof(long)) && 541 (m = m_pullup(m, sizeof(long))) == NULL)) 542 return (ENOBUFS); 543 if ((m->m_flags & M_PKTHDR) == 0) 544 panic("route_output"); 545 len = m->m_pkthdr.len; 546 if (len < sizeof(*rtm) || 547 len != mtod(m, struct rt_msghdr *)->rtm_msglen) 548 senderr(EINVAL); 549 550 /* 551 * Most of current messages are in range 200-240 bytes, 552 * minimize possible failures by using on-stack buffer 553 * which should fit for most messages. 554 * However, use stable memory if we need to handle 555 * something large. 556 */ 557 if (len < sizeof(msgbuf)) { 558 alloc_len = sizeof(msgbuf); 559 rtm = (struct rt_msghdr *)msgbuf; 560 } else { 561 alloc_len = roundup2(len, 1024); 562 rtm = malloc(alloc_len, M_TEMP, M_NOWAIT); 563 if (rtm == NULL) 564 senderr(ENOBUFS); 565 } 566 567 m_copydata(m, 0, len, (caddr_t)rtm); 568 bzero(&info, sizeof(info)); 569 bzero(&w, sizeof(w)); 570 571 if (rtm->rtm_version != RTM_VERSION) { 572 /* Do not touch message since format is unknown */ 573 if ((char *)rtm != msgbuf) 574 free(rtm, M_TEMP); 575 rtm = NULL; 576 senderr(EPROTONOSUPPORT); 577 } 578 579 /* 580 * Starting from here, it is possible 581 * to alter original message and insert 582 * caller PID and error value. 583 */ 584 585 rtm->rtm_pid = curproc->p_pid; 586 info.rti_addrs = rtm->rtm_addrs; 587 /* 588 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6 589 * link-local address because rtrequest requires addresses with 590 * embedded scope id. 591 */ 592 if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) 593 senderr(EINVAL); 594 595 info.rti_flags = rtm->rtm_flags; 596 if (info.rti_info[RTAX_DST] == NULL || 597 info.rti_info[RTAX_DST]->sa_family >= AF_MAX || 598 (info.rti_info[RTAX_GATEWAY] != NULL && 599 info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) 600 senderr(EINVAL); 601 saf = info.rti_info[RTAX_DST]->sa_family; 602 /* 603 * Verify that the caller has the appropriate privilege; RTM_GET 604 * is the only operation the non-superuser is allowed. 605 */ 606 if (rtm->rtm_type != RTM_GET) { 607 error = priv_check(curthread, PRIV_NET_ROUTE); 608 if (error) 609 senderr(error); 610 } 611 612 /* 613 * The given gateway address may be an interface address. 614 * For example, issuing a "route change" command on a route 615 * entry that was created from a tunnel, and the gateway 616 * address given is the local end point. In this case the 617 * RTF_GATEWAY flag must be cleared or the destination will 618 * not be reachable even though there is no error message. 619 */ 620 if (info.rti_info[RTAX_GATEWAY] != NULL && 621 info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) { 622 struct route gw_ro; 623 624 bzero(&gw_ro, sizeof(gw_ro)); 625 gw_ro.ro_dst = *info.rti_info[RTAX_GATEWAY]; 626 rtalloc_ign_fib(&gw_ro, 0, fibnum); 627 /* 628 * A host route through the loopback interface is 629 * installed for each interface adddress. In pre 8.0 630 * releases the interface address of a PPP link type 631 * is not reachable locally. This behavior is fixed as 632 * part of the new L2/L3 redesign and rewrite work. The 633 * signature of this interface address route is the 634 * AF_LINK sa_family type of the rt_gateway, and the 635 * rt_ifp has the IFF_LOOPBACK flag set. 636 */ 637 if (gw_ro.ro_rt != NULL && 638 gw_ro.ro_rt->rt_gateway->sa_family == AF_LINK && 639 gw_ro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) { 640 info.rti_flags &= ~RTF_GATEWAY; 641 info.rti_flags |= RTF_GWFLAG_COMPAT; 642 } 643 if (gw_ro.ro_rt != NULL) 644 RTFREE(gw_ro.ro_rt); 645 } 646 647 switch (rtm->rtm_type) { 648 struct rtentry *saved_nrt; 649 650 case RTM_ADD: 651 case RTM_CHANGE: 652 if (info.rti_info[RTAX_GATEWAY] == NULL) 653 senderr(EINVAL); 654 saved_nrt = NULL; 655 656 /* support for new ARP code */ 657 if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK && 658 (rtm->rtm_flags & RTF_LLDATA) != 0) { 659 error = lla_rt_output(rtm, &info); 660 #ifdef INET6 661 if (error == 0) 662 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0; 663 #endif 664 break; 665 } 666 error = rtrequest1_fib(rtm->rtm_type, &info, &saved_nrt, 667 fibnum); 668 if (error == 0 && saved_nrt != NULL) { 669 #ifdef INET6 670 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0; 671 #endif 672 RT_LOCK(saved_nrt); 673 rt_setmetrics(rtm, saved_nrt); 674 rtm->rtm_index = saved_nrt->rt_ifp->if_index; 675 RT_REMREF(saved_nrt); 676 RT_UNLOCK(saved_nrt); 677 } 678 break; 679 680 case RTM_DELETE: 681 saved_nrt = NULL; 682 /* support for new ARP code */ 683 if (info.rti_info[RTAX_GATEWAY] && 684 (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) && 685 (rtm->rtm_flags & RTF_LLDATA) != 0) { 686 error = lla_rt_output(rtm, &info); 687 #ifdef INET6 688 if (error == 0) 689 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0; 690 #endif 691 break; 692 } 693 error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt, fibnum); 694 if (error == 0) { 695 RT_LOCK(saved_nrt); 696 rt = saved_nrt; 697 goto report; 698 } 699 #ifdef INET6 700 /* rt_msg2() will not be used when RTM_DELETE fails. */ 701 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0; 702 #endif 703 break; 704 705 case RTM_GET: 706 rnh = rt_tables_get_rnh(fibnum, saf); 707 if (rnh == NULL) 708 senderr(EAFNOSUPPORT); 709 710 RADIX_NODE_HEAD_RLOCK(rnh); 711 712 if (info.rti_info[RTAX_NETMASK] == NULL && 713 rtm->rtm_type == RTM_GET) { 714 /* 715 * Provide logest prefix match for 716 * address lookup (no mask). 717 * 'route -n get addr' 718 */ 719 rt = (struct rtentry *) rnh->rnh_matchaddr( 720 info.rti_info[RTAX_DST], rnh); 721 } else 722 rt = (struct rtentry *) rnh->rnh_lookup( 723 info.rti_info[RTAX_DST], 724 info.rti_info[RTAX_NETMASK], rnh); 725 726 if (rt == NULL) { 727 RADIX_NODE_HEAD_RUNLOCK(rnh); 728 senderr(ESRCH); 729 } 730 #ifdef RADIX_MPATH 731 /* 732 * for RTM_CHANGE/LOCK, if we got multipath routes, 733 * we require users to specify a matching RTAX_GATEWAY. 734 * 735 * for RTM_GET, gate is optional even with multipath. 736 * if gate == NULL the first match is returned. 737 * (no need to call rt_mpath_matchgate if gate == NULL) 738 */ 739 if (rn_mpath_capable(rnh) && 740 (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) { 741 rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]); 742 if (!rt) { 743 RADIX_NODE_HEAD_RUNLOCK(rnh); 744 senderr(ESRCH); 745 } 746 } 747 #endif 748 /* 749 * If performing proxied L2 entry insertion, and 750 * the actual PPP host entry is found, perform 751 * another search to retrieve the prefix route of 752 * the local end point of the PPP link. 753 */ 754 if (rtm->rtm_flags & RTF_ANNOUNCE) { 755 struct sockaddr laddr; 756 757 if (rt->rt_ifp != NULL && 758 rt->rt_ifp->if_type == IFT_PROPVIRTUAL) { 759 struct ifaddr *ifa; 760 761 ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1, 762 RT_DEFAULT_FIB); 763 if (ifa != NULL) 764 rt_maskedcopy(ifa->ifa_addr, 765 &laddr, 766 ifa->ifa_netmask); 767 } else 768 rt_maskedcopy(rt->rt_ifa->ifa_addr, 769 &laddr, 770 rt->rt_ifa->ifa_netmask); 771 /* 772 * refactor rt and no lock operation necessary 773 */ 774 rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr, rnh); 775 if (rt == NULL) { 776 RADIX_NODE_HEAD_RUNLOCK(rnh); 777 senderr(ESRCH); 778 } 779 } 780 RT_LOCK(rt); 781 RT_ADDREF(rt); 782 RADIX_NODE_HEAD_RUNLOCK(rnh); 783 784 report: 785 RT_LOCK_ASSERT(rt); 786 if ((rt->rt_flags & RTF_HOST) == 0 787 ? jailed_without_vnet(curthread->td_ucred) 788 : prison_if(curthread->td_ucred, 789 rt_key(rt)) != 0) { 790 RT_UNLOCK(rt); 791 senderr(ESRCH); 792 } 793 info.rti_info[RTAX_DST] = rt_key(rt); 794 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 795 info.rti_info[RTAX_NETMASK] = rt_mask(rt); 796 info.rti_info[RTAX_GENMASK] = 0; 797 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) { 798 ifp = rt->rt_ifp; 799 if (ifp) { 800 info.rti_info[RTAX_IFP] = 801 ifp->if_addr->ifa_addr; 802 error = rtm_get_jailed(&info, ifp, rt, 803 &saun, curthread->td_ucred); 804 if (error != 0) { 805 RT_UNLOCK(rt); 806 senderr(error); 807 } 808 if (ifp->if_flags & IFF_POINTOPOINT) 809 info.rti_info[RTAX_BRD] = 810 rt->rt_ifa->ifa_dstaddr; 811 rtm->rtm_index = ifp->if_index; 812 } else { 813 info.rti_info[RTAX_IFP] = NULL; 814 info.rti_info[RTAX_IFA] = NULL; 815 } 816 } else if ((ifp = rt->rt_ifp) != NULL) { 817 rtm->rtm_index = ifp->if_index; 818 } 819 820 /* Check if we need to realloc storage */ 821 rtsock_msg_buffer(rtm->rtm_type, &info, NULL, &len); 822 if (len > alloc_len) { 823 struct rt_msghdr *new_rtm; 824 new_rtm = malloc(len, M_TEMP, M_NOWAIT); 825 if (new_rtm == NULL) { 826 RT_UNLOCK(rt); 827 senderr(ENOBUFS); 828 } 829 bcopy(rtm, new_rtm, rtm->rtm_msglen); 830 free(rtm, M_TEMP); 831 rtm = new_rtm; 832 alloc_len = len; 833 } 834 835 w.w_tmem = (caddr_t)rtm; 836 w.w_tmemsize = alloc_len; 837 rtsock_msg_buffer(rtm->rtm_type, &info, &w, &len); 838 839 if (rt->rt_flags & RTF_GWFLAG_COMPAT) 840 rtm->rtm_flags = RTF_GATEWAY | 841 (rt->rt_flags & ~RTF_GWFLAG_COMPAT); 842 else 843 rtm->rtm_flags = rt->rt_flags; 844 rt_getmetrics(rt, &rtm->rtm_rmx); 845 rtm->rtm_addrs = info.rti_addrs; 846 847 RT_UNLOCK(rt); 848 break; 849 850 default: 851 senderr(EOPNOTSUPP); 852 } 853 854 flush: 855 if (rt != NULL) 856 RTFREE(rt); 857 /* 858 * Check to see if we don't want our own messages. 859 */ 860 if ((so->so_options & SO_USELOOPBACK) == 0) { 861 if (V_route_cb.any_count <= 1) { 862 if (rtm != NULL && (char *)rtm != msgbuf) 863 free(rtm, M_TEMP); 864 m_freem(m); 865 return (error); 866 } 867 /* There is another listener, so construct message */ 868 rp = sotorawcb(so); 869 } 870 871 if (rtm != NULL) { 872 #ifdef INET6 873 if (rti_need_deembed) { 874 /* sin6_scope_id is recovered before sending rtm. */ 875 sin6 = (struct sockaddr_in6 *)&ss; 876 for (i = 0; i < RTAX_MAX; i++) { 877 if (info.rti_info[i] == NULL) 878 continue; 879 if (info.rti_info[i]->sa_family != AF_INET6) 880 continue; 881 bcopy(info.rti_info[i], sin6, sizeof(*sin6)); 882 if (sa6_recoverscope(sin6) == 0) 883 bcopy(sin6, info.rti_info[i], 884 sizeof(*sin6)); 885 } 886 } 887 #endif 888 if (error != 0) 889 rtm->rtm_errno = error; 890 else 891 rtm->rtm_flags |= RTF_DONE; 892 893 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm); 894 if (m->m_pkthdr.len < rtm->rtm_msglen) { 895 m_freem(m); 896 m = NULL; 897 } else if (m->m_pkthdr.len > rtm->rtm_msglen) 898 m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len); 899 900 if ((char *)rtm != msgbuf) 901 free(rtm, M_TEMP); 902 } 903 if (m != NULL) { 904 M_SETFIB(m, fibnum); 905 m->m_flags |= RTS_FILTER_FIB; 906 if (rp) { 907 /* 908 * XXX insure we don't get a copy by 909 * invalidating our protocol 910 */ 911 unsigned short family = rp->rcb_proto.sp_family; 912 rp->rcb_proto.sp_family = 0; 913 rt_dispatch(m, saf); 914 rp->rcb_proto.sp_family = family; 915 } else 916 rt_dispatch(m, saf); 917 } 918 919 return (error); 920 } 921 922 static void 923 rt_setmetrics(const struct rt_msghdr *rtm, struct rtentry *rt) 924 { 925 926 if (rtm->rtm_inits & RTV_MTU) 927 rt->rt_mtu = rtm->rtm_rmx.rmx_mtu; 928 if (rtm->rtm_inits & RTV_WEIGHT) 929 rt->rt_weight = rtm->rtm_rmx.rmx_weight; 930 /* Kernel -> userland timebase conversion. */ 931 if (rtm->rtm_inits & RTV_EXPIRE) 932 rt->rt_expire = rtm->rtm_rmx.rmx_expire ? 933 rtm->rtm_rmx.rmx_expire - time_second + time_uptime : 0; 934 } 935 936 static void 937 rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out) 938 { 939 940 bzero(out, sizeof(*out)); 941 out->rmx_mtu = rt->rt_mtu; 942 out->rmx_weight = rt->rt_weight; 943 out->rmx_pksent = counter_u64_fetch(rt->rt_pksent); 944 /* Kernel -> userland timebase conversion. */ 945 out->rmx_expire = rt->rt_expire ? 946 rt->rt_expire - time_uptime + time_second : 0; 947 } 948 949 /* 950 * Extract the addresses of the passed sockaddrs. 951 * Do a little sanity checking so as to avoid bad memory references. 952 * This data is derived straight from userland. 953 */ 954 static int 955 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo) 956 { 957 struct sockaddr *sa; 958 int i; 959 960 for (i = 0; i < RTAX_MAX && cp < cplim; i++) { 961 if ((rtinfo->rti_addrs & (1 << i)) == 0) 962 continue; 963 sa = (struct sockaddr *)cp; 964 /* 965 * It won't fit. 966 */ 967 if (cp + sa->sa_len > cplim) 968 return (EINVAL); 969 /* 970 * there are no more.. quit now 971 * If there are more bits, they are in error. 972 * I've seen this. route(1) can evidently generate these. 973 * This causes kernel to core dump. 974 * for compatibility, If we see this, point to a safe address. 975 */ 976 if (sa->sa_len == 0) { 977 rtinfo->rti_info[i] = &sa_zero; 978 return (0); /* should be EINVAL but for compat */ 979 } 980 /* accept it */ 981 #ifdef INET6 982 if (sa->sa_family == AF_INET6) 983 sa6_embedscope((struct sockaddr_in6 *)sa, 984 V_ip6_use_defzone); 985 #endif 986 rtinfo->rti_info[i] = sa; 987 cp += SA_SIZE(sa); 988 } 989 return (0); 990 } 991 992 /* 993 * Used by the routing socket. 994 */ 995 static struct mbuf * 996 rt_msg1(int type, struct rt_addrinfo *rtinfo) 997 { 998 struct rt_msghdr *rtm; 999 struct mbuf *m; 1000 int i; 1001 struct sockaddr *sa; 1002 #ifdef INET6 1003 struct sockaddr_storage ss; 1004 struct sockaddr_in6 *sin6; 1005 #endif 1006 int len, dlen; 1007 1008 switch (type) { 1009 1010 case RTM_DELADDR: 1011 case RTM_NEWADDR: 1012 len = sizeof(struct ifa_msghdr); 1013 break; 1014 1015 case RTM_DELMADDR: 1016 case RTM_NEWMADDR: 1017 len = sizeof(struct ifma_msghdr); 1018 break; 1019 1020 case RTM_IFINFO: 1021 len = sizeof(struct if_msghdr); 1022 break; 1023 1024 case RTM_IFANNOUNCE: 1025 case RTM_IEEE80211: 1026 len = sizeof(struct if_announcemsghdr); 1027 break; 1028 1029 default: 1030 len = sizeof(struct rt_msghdr); 1031 } 1032 1033 /* XXXGL: can we use MJUMPAGESIZE cluster here? */ 1034 KASSERT(len <= MCLBYTES, ("%s: message too big", __func__)); 1035 if (len > MHLEN) 1036 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 1037 else 1038 m = m_gethdr(M_NOWAIT, MT_DATA); 1039 if (m == NULL) 1040 return (m); 1041 1042 m->m_pkthdr.len = m->m_len = len; 1043 rtm = mtod(m, struct rt_msghdr *); 1044 bzero((caddr_t)rtm, len); 1045 for (i = 0; i < RTAX_MAX; i++) { 1046 if ((sa = rtinfo->rti_info[i]) == NULL) 1047 continue; 1048 rtinfo->rti_addrs |= (1 << i); 1049 dlen = SA_SIZE(sa); 1050 #ifdef INET6 1051 if (V_deembed_scopeid && sa->sa_family == AF_INET6) { 1052 sin6 = (struct sockaddr_in6 *)&ss; 1053 bcopy(sa, sin6, sizeof(*sin6)); 1054 if (sa6_recoverscope(sin6) == 0) 1055 sa = (struct sockaddr *)sin6; 1056 } 1057 #endif 1058 m_copyback(m, len, dlen, (caddr_t)sa); 1059 len += dlen; 1060 } 1061 if (m->m_pkthdr.len != len) { 1062 m_freem(m); 1063 return (NULL); 1064 } 1065 rtm->rtm_msglen = len; 1066 rtm->rtm_version = RTM_VERSION; 1067 rtm->rtm_type = type; 1068 return (m); 1069 } 1070 1071 /* 1072 * Writes information related to @rtinfo object to preallocated buffer. 1073 * Stores needed size in @plen. If @w is NULL, calculates size without 1074 * writing. 1075 * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation. 1076 * 1077 * Returns 0 on success. 1078 * 1079 */ 1080 static int 1081 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen) 1082 { 1083 int i; 1084 int len, buflen = 0, dlen; 1085 caddr_t cp = NULL; 1086 struct rt_msghdr *rtm = NULL; 1087 #ifdef INET6 1088 struct sockaddr_storage ss; 1089 struct sockaddr_in6 *sin6; 1090 #endif 1091 1092 switch (type) { 1093 1094 case RTM_DELADDR: 1095 case RTM_NEWADDR: 1096 if (w != NULL && w->w_op == NET_RT_IFLISTL) { 1097 #ifdef COMPAT_FREEBSD32 1098 if (w->w_req->flags & SCTL_MASK32) 1099 len = sizeof(struct ifa_msghdrl32); 1100 else 1101 #endif 1102 len = sizeof(struct ifa_msghdrl); 1103 } else 1104 len = sizeof(struct ifa_msghdr); 1105 break; 1106 1107 case RTM_IFINFO: 1108 #ifdef COMPAT_FREEBSD32 1109 if (w != NULL && w->w_req->flags & SCTL_MASK32) { 1110 if (w->w_op == NET_RT_IFLISTL) 1111 len = sizeof(struct if_msghdrl32); 1112 else 1113 len = sizeof(struct if_msghdr32); 1114 break; 1115 } 1116 #endif 1117 if (w != NULL && w->w_op == NET_RT_IFLISTL) 1118 len = sizeof(struct if_msghdrl); 1119 else 1120 len = sizeof(struct if_msghdr); 1121 break; 1122 1123 case RTM_NEWMADDR: 1124 len = sizeof(struct ifma_msghdr); 1125 break; 1126 1127 default: 1128 len = sizeof(struct rt_msghdr); 1129 } 1130 1131 if (w != NULL) { 1132 rtm = (struct rt_msghdr *)w->w_tmem; 1133 buflen = w->w_tmemsize - len; 1134 cp = (caddr_t)w->w_tmem + len; 1135 } 1136 1137 rtinfo->rti_addrs = 0; 1138 for (i = 0; i < RTAX_MAX; i++) { 1139 struct sockaddr *sa; 1140 1141 if ((sa = rtinfo->rti_info[i]) == NULL) 1142 continue; 1143 rtinfo->rti_addrs |= (1 << i); 1144 dlen = SA_SIZE(sa); 1145 if (cp != NULL && buflen >= dlen) { 1146 #ifdef INET6 1147 if (V_deembed_scopeid && sa->sa_family == AF_INET6) { 1148 sin6 = (struct sockaddr_in6 *)&ss; 1149 bcopy(sa, sin6, sizeof(*sin6)); 1150 if (sa6_recoverscope(sin6) == 0) 1151 sa = (struct sockaddr *)sin6; 1152 } 1153 #endif 1154 bcopy((caddr_t)sa, cp, (unsigned)dlen); 1155 cp += dlen; 1156 buflen -= dlen; 1157 } else if (cp != NULL) { 1158 /* 1159 * Buffer too small. Count needed size 1160 * and return with error. 1161 */ 1162 cp = NULL; 1163 } 1164 1165 len += dlen; 1166 } 1167 1168 if (cp != NULL) { 1169 dlen = ALIGN(len) - len; 1170 if (buflen < dlen) 1171 cp = NULL; 1172 else 1173 buflen -= dlen; 1174 } 1175 len = ALIGN(len); 1176 1177 if (cp != NULL) { 1178 /* fill header iff buffer is large enough */ 1179 rtm->rtm_version = RTM_VERSION; 1180 rtm->rtm_type = type; 1181 rtm->rtm_msglen = len; 1182 } 1183 1184 *plen = len; 1185 1186 if (w != NULL && cp == NULL) 1187 return (ENOBUFS); 1188 1189 return (0); 1190 } 1191 1192 /* 1193 * This routine is called to generate a message from the routing 1194 * socket indicating that a redirect has occured, a routing lookup 1195 * has failed, or that a protocol has detected timeouts to a particular 1196 * destination. 1197 */ 1198 void 1199 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error, 1200 int fibnum) 1201 { 1202 struct rt_msghdr *rtm; 1203 struct mbuf *m; 1204 struct sockaddr *sa = rtinfo->rti_info[RTAX_DST]; 1205 1206 if (V_route_cb.any_count == 0) 1207 return; 1208 m = rt_msg1(type, rtinfo); 1209 if (m == NULL) 1210 return; 1211 1212 if (fibnum != RT_ALL_FIBS) { 1213 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out " 1214 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs)); 1215 M_SETFIB(m, fibnum); 1216 m->m_flags |= RTS_FILTER_FIB; 1217 } 1218 1219 rtm = mtod(m, struct rt_msghdr *); 1220 rtm->rtm_flags = RTF_DONE | flags; 1221 rtm->rtm_errno = error; 1222 rtm->rtm_addrs = rtinfo->rti_addrs; 1223 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1224 } 1225 1226 void 1227 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error) 1228 { 1229 1230 rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS); 1231 } 1232 1233 /* 1234 * This routine is called to generate a message from the routing 1235 * socket indicating that the status of a network interface has changed. 1236 */ 1237 void 1238 rt_ifmsg(struct ifnet *ifp) 1239 { 1240 struct if_msghdr *ifm; 1241 struct mbuf *m; 1242 struct rt_addrinfo info; 1243 1244 if (V_route_cb.any_count == 0) 1245 return; 1246 bzero((caddr_t)&info, sizeof(info)); 1247 m = rt_msg1(RTM_IFINFO, &info); 1248 if (m == NULL) 1249 return; 1250 ifm = mtod(m, struct if_msghdr *); 1251 ifm->ifm_index = ifp->if_index; 1252 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1253 ifm->ifm_data = ifp->if_data; 1254 ifm->ifm_addrs = 0; 1255 rt_dispatch(m, AF_UNSPEC); 1256 } 1257 1258 /* 1259 * Announce interface address arrival/withdraw. 1260 * Please do not call directly, use rt_addrmsg(). 1261 * Assume input data to be valid. 1262 * Returns 0 on success. 1263 */ 1264 int 1265 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum) 1266 { 1267 struct rt_addrinfo info; 1268 struct sockaddr *sa; 1269 int ncmd; 1270 struct mbuf *m; 1271 struct ifa_msghdr *ifam; 1272 struct ifnet *ifp = ifa->ifa_ifp; 1273 1274 if (V_route_cb.any_count == 0) 1275 return (0); 1276 1277 ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR; 1278 1279 bzero((caddr_t)&info, sizeof(info)); 1280 info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr; 1281 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr; 1282 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask; 1283 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr; 1284 if ((m = rt_msg1(ncmd, &info)) == NULL) 1285 return (ENOBUFS); 1286 ifam = mtod(m, struct ifa_msghdr *); 1287 ifam->ifam_index = ifp->if_index; 1288 ifam->ifam_metric = ifa->ifa_metric; 1289 ifam->ifam_flags = ifa->ifa_flags; 1290 ifam->ifam_addrs = info.rti_addrs; 1291 1292 if (fibnum != RT_ALL_FIBS) { 1293 M_SETFIB(m, fibnum); 1294 m->m_flags |= RTS_FILTER_FIB; 1295 } 1296 1297 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1298 1299 return (0); 1300 } 1301 1302 /* 1303 * Announce route addition/removal. 1304 * Please do not call directly, use rt_routemsg(). 1305 * Note that @rt data MAY be inconsistent/invalid: 1306 * if some userland app sends us "invalid" route message (invalid mask, 1307 * no dst, wrong address families, etc...) we need to pass it back 1308 * to app (and any other rtsock consumers) with rtm_errno field set to 1309 * non-zero value. 1310 * 1311 * Returns 0 on success. 1312 */ 1313 int 1314 rtsock_routemsg(int cmd, struct ifnet *ifp, int error, struct rtentry *rt, 1315 int fibnum) 1316 { 1317 struct rt_addrinfo info; 1318 struct sockaddr *sa; 1319 struct mbuf *m; 1320 struct rt_msghdr *rtm; 1321 1322 if (V_route_cb.any_count == 0) 1323 return (0); 1324 1325 bzero((caddr_t)&info, sizeof(info)); 1326 info.rti_info[RTAX_NETMASK] = rt_mask(rt); 1327 info.rti_info[RTAX_DST] = sa = rt_key(rt); 1328 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 1329 if ((m = rt_msg1(cmd, &info)) == NULL) 1330 return (ENOBUFS); 1331 rtm = mtod(m, struct rt_msghdr *); 1332 rtm->rtm_index = ifp->if_index; 1333 rtm->rtm_flags |= rt->rt_flags; 1334 rtm->rtm_errno = error; 1335 rtm->rtm_addrs = info.rti_addrs; 1336 1337 if (fibnum != RT_ALL_FIBS) { 1338 M_SETFIB(m, fibnum); 1339 m->m_flags |= RTS_FILTER_FIB; 1340 } 1341 1342 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1343 1344 return (0); 1345 } 1346 1347 /* 1348 * This is the analogue to the rt_newaddrmsg which performs the same 1349 * function but for multicast group memberhips. This is easier since 1350 * there is no route state to worry about. 1351 */ 1352 void 1353 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma) 1354 { 1355 struct rt_addrinfo info; 1356 struct mbuf *m = NULL; 1357 struct ifnet *ifp = ifma->ifma_ifp; 1358 struct ifma_msghdr *ifmam; 1359 1360 if (V_route_cb.any_count == 0) 1361 return; 1362 1363 bzero((caddr_t)&info, sizeof(info)); 1364 info.rti_info[RTAX_IFA] = ifma->ifma_addr; 1365 info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL; 1366 /* 1367 * If a link-layer address is present, present it as a ``gateway'' 1368 * (similarly to how ARP entries, e.g., are presented). 1369 */ 1370 info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr; 1371 m = rt_msg1(cmd, &info); 1372 if (m == NULL) 1373 return; 1374 ifmam = mtod(m, struct ifma_msghdr *); 1375 KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n", 1376 __func__)); 1377 ifmam->ifmam_index = ifp->if_index; 1378 ifmam->ifmam_addrs = info.rti_addrs; 1379 rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC); 1380 } 1381 1382 static struct mbuf * 1383 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what, 1384 struct rt_addrinfo *info) 1385 { 1386 struct if_announcemsghdr *ifan; 1387 struct mbuf *m; 1388 1389 if (V_route_cb.any_count == 0) 1390 return NULL; 1391 bzero((caddr_t)info, sizeof(*info)); 1392 m = rt_msg1(type, info); 1393 if (m != NULL) { 1394 ifan = mtod(m, struct if_announcemsghdr *); 1395 ifan->ifan_index = ifp->if_index; 1396 strlcpy(ifan->ifan_name, ifp->if_xname, 1397 sizeof(ifan->ifan_name)); 1398 ifan->ifan_what = what; 1399 } 1400 return m; 1401 } 1402 1403 /* 1404 * This is called to generate routing socket messages indicating 1405 * IEEE80211 wireless events. 1406 * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way. 1407 */ 1408 void 1409 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len) 1410 { 1411 struct mbuf *m; 1412 struct rt_addrinfo info; 1413 1414 m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info); 1415 if (m != NULL) { 1416 /* 1417 * Append the ieee80211 data. Try to stick it in the 1418 * mbuf containing the ifannounce msg; otherwise allocate 1419 * a new mbuf and append. 1420 * 1421 * NB: we assume m is a single mbuf. 1422 */ 1423 if (data_len > M_TRAILINGSPACE(m)) { 1424 struct mbuf *n = m_get(M_NOWAIT, MT_DATA); 1425 if (n == NULL) { 1426 m_freem(m); 1427 return; 1428 } 1429 bcopy(data, mtod(n, void *), data_len); 1430 n->m_len = data_len; 1431 m->m_next = n; 1432 } else if (data_len > 0) { 1433 bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len); 1434 m->m_len += data_len; 1435 } 1436 if (m->m_flags & M_PKTHDR) 1437 m->m_pkthdr.len += data_len; 1438 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len; 1439 rt_dispatch(m, AF_UNSPEC); 1440 } 1441 } 1442 1443 /* 1444 * This is called to generate routing socket messages indicating 1445 * network interface arrival and departure. 1446 */ 1447 void 1448 rt_ifannouncemsg(struct ifnet *ifp, int what) 1449 { 1450 struct mbuf *m; 1451 struct rt_addrinfo info; 1452 1453 m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info); 1454 if (m != NULL) 1455 rt_dispatch(m, AF_UNSPEC); 1456 } 1457 1458 static void 1459 rt_dispatch(struct mbuf *m, sa_family_t saf) 1460 { 1461 struct m_tag *tag; 1462 1463 /* 1464 * Preserve the family from the sockaddr, if any, in an m_tag for 1465 * use when injecting the mbuf into the routing socket buffer from 1466 * the netisr. 1467 */ 1468 if (saf != AF_UNSPEC) { 1469 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short), 1470 M_NOWAIT); 1471 if (tag == NULL) { 1472 m_freem(m); 1473 return; 1474 } 1475 *(unsigned short *)(tag + 1) = saf; 1476 m_tag_prepend(m, tag); 1477 } 1478 #ifdef VIMAGE 1479 if (V_loif) 1480 m->m_pkthdr.rcvif = V_loif; 1481 else { 1482 m_freem(m); 1483 return; 1484 } 1485 #endif 1486 netisr_queue(NETISR_ROUTE, m); /* mbuf is free'd on failure. */ 1487 } 1488 1489 /* 1490 * This is used in dumping the kernel table via sysctl(). 1491 */ 1492 static int 1493 sysctl_dumpentry(struct radix_node *rn, void *vw) 1494 { 1495 struct walkarg *w = vw; 1496 struct rtentry *rt = (struct rtentry *)rn; 1497 int error = 0, size; 1498 struct rt_addrinfo info; 1499 1500 if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg)) 1501 return 0; 1502 if ((rt->rt_flags & RTF_HOST) == 0 1503 ? jailed_without_vnet(w->w_req->td->td_ucred) 1504 : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0) 1505 return (0); 1506 bzero((caddr_t)&info, sizeof(info)); 1507 info.rti_info[RTAX_DST] = rt_key(rt); 1508 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 1509 info.rti_info[RTAX_NETMASK] = rt_mask(rt); 1510 info.rti_info[RTAX_GENMASK] = 0; 1511 if (rt->rt_ifp) { 1512 info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr; 1513 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr; 1514 if (rt->rt_ifp->if_flags & IFF_POINTOPOINT) 1515 info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr; 1516 } 1517 if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0) 1518 return (error); 1519 if (w->w_req && w->w_tmem) { 1520 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem; 1521 1522 if (rt->rt_flags & RTF_GWFLAG_COMPAT) 1523 rtm->rtm_flags = RTF_GATEWAY | 1524 (rt->rt_flags & ~RTF_GWFLAG_COMPAT); 1525 else 1526 rtm->rtm_flags = rt->rt_flags; 1527 rt_getmetrics(rt, &rtm->rtm_rmx); 1528 rtm->rtm_index = rt->rt_ifp->if_index; 1529 rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0; 1530 rtm->rtm_addrs = info.rti_addrs; 1531 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size); 1532 return (error); 1533 } 1534 return (error); 1535 } 1536 1537 static int 1538 sysctl_iflist_ifml(struct ifnet *ifp, struct rt_addrinfo *info, 1539 struct walkarg *w, int len) 1540 { 1541 struct if_msghdrl *ifm; 1542 struct if_data *ifd; 1543 1544 ifm = (struct if_msghdrl *)w->w_tmem; 1545 1546 #ifdef COMPAT_FREEBSD32 1547 if (w->w_req->flags & SCTL_MASK32) { 1548 struct if_msghdrl32 *ifm32; 1549 1550 ifm32 = (struct if_msghdrl32 *)ifm; 1551 ifm32->ifm_addrs = info->rti_addrs; 1552 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1553 ifm32->ifm_index = ifp->if_index; 1554 ifm32->_ifm_spare1 = 0; 1555 ifm32->ifm_len = sizeof(*ifm32); 1556 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data); 1557 ifd = &ifm32->ifm_data; 1558 } else 1559 #endif 1560 { 1561 ifm->ifm_addrs = info->rti_addrs; 1562 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1563 ifm->ifm_index = ifp->if_index; 1564 ifm->_ifm_spare1 = 0; 1565 ifm->ifm_len = sizeof(*ifm); 1566 ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data); 1567 ifd = &ifm->ifm_data; 1568 } 1569 1570 *ifd = ifp->if_data; 1571 1572 /* Some drivers still use ifqueue(9), add its stats. */ 1573 ifd->ifi_oqdrops += ifp->if_snd.ifq_drops; 1574 1575 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len)); 1576 } 1577 1578 static int 1579 sysctl_iflist_ifm(struct ifnet *ifp, struct rt_addrinfo *info, 1580 struct walkarg *w, int len) 1581 { 1582 struct if_msghdr *ifm; 1583 struct if_data *ifd; 1584 1585 ifm = (struct if_msghdr *)w->w_tmem; 1586 1587 #ifdef COMPAT_FREEBSD32 1588 if (w->w_req->flags & SCTL_MASK32) { 1589 struct if_msghdr32 *ifm32; 1590 1591 ifm32 = (struct if_msghdr32 *)ifm; 1592 ifm32->ifm_addrs = info->rti_addrs; 1593 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1594 ifm32->ifm_index = ifp->if_index; 1595 ifd = &ifm32->ifm_data; 1596 } else 1597 #endif 1598 { 1599 ifm->ifm_addrs = info->rti_addrs; 1600 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1601 ifm->ifm_index = ifp->if_index; 1602 ifd = &ifm->ifm_data; 1603 } 1604 1605 *ifd = ifp->if_data; 1606 1607 /* Some drivers still use ifqueue(9), add its stats. */ 1608 ifd->ifi_oqdrops += ifp->if_snd.ifq_drops; 1609 1610 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len)); 1611 } 1612 1613 static int 1614 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info, 1615 struct walkarg *w, int len) 1616 { 1617 struct ifa_msghdrl *ifam; 1618 struct if_data *ifd; 1619 1620 ifam = (struct ifa_msghdrl *)w->w_tmem; 1621 1622 #ifdef COMPAT_FREEBSD32 1623 if (w->w_req->flags & SCTL_MASK32) { 1624 struct ifa_msghdrl32 *ifam32; 1625 1626 ifam32 = (struct ifa_msghdrl32 *)ifam; 1627 ifam32->ifam_addrs = info->rti_addrs; 1628 ifam32->ifam_flags = ifa->ifa_flags; 1629 ifam32->ifam_index = ifa->ifa_ifp->if_index; 1630 ifam32->_ifam_spare1 = 0; 1631 ifam32->ifam_len = sizeof(*ifam32); 1632 ifam32->ifam_data_off = 1633 offsetof(struct ifa_msghdrl32, ifam_data); 1634 ifam32->ifam_metric = ifa->ifa_metric; 1635 ifd = &ifam32->ifam_data; 1636 } else 1637 #endif 1638 { 1639 ifam->ifam_addrs = info->rti_addrs; 1640 ifam->ifam_flags = ifa->ifa_flags; 1641 ifam->ifam_index = ifa->ifa_ifp->if_index; 1642 ifam->_ifam_spare1 = 0; 1643 ifam->ifam_len = sizeof(*ifam); 1644 ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data); 1645 ifam->ifam_metric = ifa->ifa_metric; 1646 ifd = &ifam->ifam_data; 1647 } 1648 1649 bzero(ifd, sizeof(*ifd)); 1650 ifd->ifi_datalen = sizeof(struct if_data); 1651 ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets); 1652 ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets); 1653 ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes); 1654 ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes); 1655 1656 /* Fixup if_data carp(4) vhid. */ 1657 if (carp_get_vhid_p != NULL) 1658 ifd->ifi_vhid = (*carp_get_vhid_p)(ifa); 1659 1660 return (SYSCTL_OUT(w->w_req, w->w_tmem, len)); 1661 } 1662 1663 static int 1664 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info, 1665 struct walkarg *w, int len) 1666 { 1667 struct ifa_msghdr *ifam; 1668 1669 ifam = (struct ifa_msghdr *)w->w_tmem; 1670 ifam->ifam_addrs = info->rti_addrs; 1671 ifam->ifam_flags = ifa->ifa_flags; 1672 ifam->ifam_index = ifa->ifa_ifp->if_index; 1673 ifam->ifam_metric = ifa->ifa_metric; 1674 1675 return (SYSCTL_OUT(w->w_req, w->w_tmem, len)); 1676 } 1677 1678 static int 1679 sysctl_iflist(int af, struct walkarg *w) 1680 { 1681 struct ifnet *ifp; 1682 struct ifaddr *ifa; 1683 struct rt_addrinfo info; 1684 int len, error = 0; 1685 1686 bzero((caddr_t)&info, sizeof(info)); 1687 IFNET_RLOCK_NOSLEEP(); 1688 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1689 if (w->w_arg && w->w_arg != ifp->if_index) 1690 continue; 1691 IF_ADDR_RLOCK(ifp); 1692 ifa = ifp->if_addr; 1693 info.rti_info[RTAX_IFP] = ifa->ifa_addr; 1694 error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len); 1695 if (error != 0) 1696 goto done; 1697 info.rti_info[RTAX_IFP] = NULL; 1698 if (w->w_req && w->w_tmem) { 1699 if (w->w_op == NET_RT_IFLISTL) 1700 error = sysctl_iflist_ifml(ifp, &info, w, len); 1701 else 1702 error = sysctl_iflist_ifm(ifp, &info, w, len); 1703 if (error) 1704 goto done; 1705 } 1706 while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) { 1707 if (af && af != ifa->ifa_addr->sa_family) 1708 continue; 1709 if (prison_if(w->w_req->td->td_ucred, 1710 ifa->ifa_addr) != 0) 1711 continue; 1712 info.rti_info[RTAX_IFA] = ifa->ifa_addr; 1713 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask; 1714 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr; 1715 error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len); 1716 if (error != 0) 1717 goto done; 1718 if (w->w_req && w->w_tmem) { 1719 if (w->w_op == NET_RT_IFLISTL) 1720 error = sysctl_iflist_ifaml(ifa, &info, 1721 w, len); 1722 else 1723 error = sysctl_iflist_ifam(ifa, &info, 1724 w, len); 1725 if (error) 1726 goto done; 1727 } 1728 } 1729 IF_ADDR_RUNLOCK(ifp); 1730 info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] = 1731 info.rti_info[RTAX_BRD] = NULL; 1732 } 1733 done: 1734 if (ifp != NULL) 1735 IF_ADDR_RUNLOCK(ifp); 1736 IFNET_RUNLOCK_NOSLEEP(); 1737 return (error); 1738 } 1739 1740 static int 1741 sysctl_ifmalist(int af, struct walkarg *w) 1742 { 1743 struct ifnet *ifp; 1744 struct ifmultiaddr *ifma; 1745 struct rt_addrinfo info; 1746 int len, error = 0; 1747 struct ifaddr *ifa; 1748 1749 bzero((caddr_t)&info, sizeof(info)); 1750 IFNET_RLOCK_NOSLEEP(); 1751 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1752 if (w->w_arg && w->w_arg != ifp->if_index) 1753 continue; 1754 ifa = ifp->if_addr; 1755 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL; 1756 IF_ADDR_RLOCK(ifp); 1757 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 1758 if (af && af != ifma->ifma_addr->sa_family) 1759 continue; 1760 if (prison_if(w->w_req->td->td_ucred, 1761 ifma->ifma_addr) != 0) 1762 continue; 1763 info.rti_info[RTAX_IFA] = ifma->ifma_addr; 1764 info.rti_info[RTAX_GATEWAY] = 1765 (ifma->ifma_addr->sa_family != AF_LINK) ? 1766 ifma->ifma_lladdr : NULL; 1767 error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len); 1768 if (error != 0) 1769 goto done; 1770 if (w->w_req && w->w_tmem) { 1771 struct ifma_msghdr *ifmam; 1772 1773 ifmam = (struct ifma_msghdr *)w->w_tmem; 1774 ifmam->ifmam_index = ifma->ifma_ifp->if_index; 1775 ifmam->ifmam_flags = 0; 1776 ifmam->ifmam_addrs = info.rti_addrs; 1777 error = SYSCTL_OUT(w->w_req, w->w_tmem, len); 1778 if (error) { 1779 IF_ADDR_RUNLOCK(ifp); 1780 goto done; 1781 } 1782 } 1783 } 1784 IF_ADDR_RUNLOCK(ifp); 1785 } 1786 done: 1787 IFNET_RUNLOCK_NOSLEEP(); 1788 return (error); 1789 } 1790 1791 static int 1792 sysctl_rtsock(SYSCTL_HANDLER_ARGS) 1793 { 1794 int *name = (int *)arg1; 1795 u_int namelen = arg2; 1796 struct radix_node_head *rnh = NULL; /* silence compiler. */ 1797 int i, lim, error = EINVAL; 1798 int fib = 0; 1799 u_char af; 1800 struct walkarg w; 1801 1802 name ++; 1803 namelen--; 1804 if (req->newptr) 1805 return (EPERM); 1806 if (name[1] == NET_RT_DUMP) { 1807 if (namelen == 3) 1808 fib = req->td->td_proc->p_fibnum; 1809 else if (namelen == 4) 1810 fib = (name[3] == RT_ALL_FIBS) ? 1811 req->td->td_proc->p_fibnum : name[3]; 1812 else 1813 return ((namelen < 3) ? EISDIR : ENOTDIR); 1814 if (fib < 0 || fib >= rt_numfibs) 1815 return (EINVAL); 1816 } else if (namelen != 3) 1817 return ((namelen < 3) ? EISDIR : ENOTDIR); 1818 af = name[0]; 1819 if (af > AF_MAX) 1820 return (EINVAL); 1821 bzero(&w, sizeof(w)); 1822 w.w_op = name[1]; 1823 w.w_arg = name[2]; 1824 w.w_req = req; 1825 1826 error = sysctl_wire_old_buffer(req, 0); 1827 if (error) 1828 return (error); 1829 1830 /* 1831 * Allocate reply buffer in advance. 1832 * All rtsock messages has maximum length of u_short. 1833 */ 1834 w.w_tmemsize = 65536; 1835 w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK); 1836 1837 switch (w.w_op) { 1838 1839 case NET_RT_DUMP: 1840 case NET_RT_FLAGS: 1841 if (af == 0) { /* dump all tables */ 1842 i = 1; 1843 lim = AF_MAX; 1844 } else /* dump only one table */ 1845 i = lim = af; 1846 1847 /* 1848 * take care of llinfo entries, the caller must 1849 * specify an AF 1850 */ 1851 if (w.w_op == NET_RT_FLAGS && 1852 (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) { 1853 if (af != 0) 1854 error = lltable_sysctl_dumparp(af, w.w_req); 1855 else 1856 error = EINVAL; 1857 break; 1858 } 1859 /* 1860 * take care of routing entries 1861 */ 1862 for (error = 0; error == 0 && i <= lim; i++) { 1863 rnh = rt_tables_get_rnh(fib, i); 1864 if (rnh != NULL) { 1865 RADIX_NODE_HEAD_RLOCK(rnh); 1866 error = rnh->rnh_walktree(rnh, 1867 sysctl_dumpentry, &w); 1868 RADIX_NODE_HEAD_RUNLOCK(rnh); 1869 } else if (af != 0) 1870 error = EAFNOSUPPORT; 1871 } 1872 break; 1873 1874 case NET_RT_IFLIST: 1875 case NET_RT_IFLISTL: 1876 error = sysctl_iflist(af, &w); 1877 break; 1878 1879 case NET_RT_IFMALIST: 1880 error = sysctl_ifmalist(af, &w); 1881 break; 1882 } 1883 1884 free(w.w_tmem, M_TEMP); 1885 return (error); 1886 } 1887 1888 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, ""); 1889 1890 /* 1891 * Definitions of protocols supported in the ROUTE domain. 1892 */ 1893 1894 static struct domain routedomain; /* or at least forward */ 1895 1896 static struct protosw routesw[] = { 1897 { 1898 .pr_type = SOCK_RAW, 1899 .pr_domain = &routedomain, 1900 .pr_flags = PR_ATOMIC|PR_ADDR, 1901 .pr_output = route_output, 1902 .pr_ctlinput = raw_ctlinput, 1903 .pr_init = raw_init, 1904 .pr_usrreqs = &route_usrreqs 1905 } 1906 }; 1907 1908 static struct domain routedomain = { 1909 .dom_family = PF_ROUTE, 1910 .dom_name = "route", 1911 .dom_protosw = routesw, 1912 .dom_protoswNPROTOSW = &routesw[sizeof(routesw)/sizeof(routesw[0])] 1913 }; 1914 1915 VNET_DOMAIN_SET(route); 1916