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