1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1988, 1991, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * @(#)rtsock.c 8.7 (Berkeley) 10/12/95 32 * $FreeBSD$ 33 */ 34 #include "opt_ddb.h" 35 #include "opt_route.h" 36 #include "opt_inet.h" 37 #include "opt_inet6.h" 38 39 #include <sys/param.h> 40 #include <sys/jail.h> 41 #include <sys/kernel.h> 42 #include <sys/eventhandler.h> 43 #include <sys/domain.h> 44 #include <sys/lock.h> 45 #include <sys/malloc.h> 46 #include <sys/mbuf.h> 47 #include <sys/priv.h> 48 #include <sys/proc.h> 49 #include <sys/protosw.h> 50 #include <sys/rmlock.h> 51 #include <sys/rwlock.h> 52 #include <sys/signalvar.h> 53 #include <sys/socket.h> 54 #include <sys/socketvar.h> 55 #include <sys/sysctl.h> 56 #include <sys/systm.h> 57 58 #include <net/if.h> 59 #include <net/if_var.h> 60 #include <net/if_dl.h> 61 #include <net/if_llatbl.h> 62 #include <net/if_types.h> 63 #include <net/netisr.h> 64 #include <net/route.h> 65 #include <net/route/route_ctl.h> 66 #include <net/route/route_var.h> 67 #include <net/vnet.h> 68 69 #include <netinet/in.h> 70 #include <netinet/if_ether.h> 71 #include <netinet/ip_carp.h> 72 #ifdef INET6 73 #include <netinet6/in6_var.h> 74 #include <netinet6/ip6_var.h> 75 #include <netinet6/scope6_var.h> 76 #endif 77 #include <net/route/nhop.h> 78 79 #define DEBUG_MOD_NAME rtsock 80 #define DEBUG_MAX_LEVEL LOG_DEBUG 81 #include <net/route/route_debug.h> 82 _DECLARE_DEBUG(LOG_INFO); 83 84 #ifdef COMPAT_FREEBSD32 85 #include <sys/mount.h> 86 #include <compat/freebsd32/freebsd32.h> 87 88 struct if_msghdr32 { 89 uint16_t ifm_msglen; 90 uint8_t ifm_version; 91 uint8_t ifm_type; 92 int32_t ifm_addrs; 93 int32_t ifm_flags; 94 uint16_t ifm_index; 95 uint16_t _ifm_spare1; 96 struct if_data ifm_data; 97 }; 98 99 struct if_msghdrl32 { 100 uint16_t ifm_msglen; 101 uint8_t ifm_version; 102 uint8_t ifm_type; 103 int32_t ifm_addrs; 104 int32_t ifm_flags; 105 uint16_t ifm_index; 106 uint16_t _ifm_spare1; 107 uint16_t ifm_len; 108 uint16_t ifm_data_off; 109 uint32_t _ifm_spare2; 110 struct if_data ifm_data; 111 }; 112 113 struct ifa_msghdrl32 { 114 uint16_t ifam_msglen; 115 uint8_t ifam_version; 116 uint8_t ifam_type; 117 int32_t ifam_addrs; 118 int32_t ifam_flags; 119 uint16_t ifam_index; 120 uint16_t _ifam_spare1; 121 uint16_t ifam_len; 122 uint16_t ifam_data_off; 123 int32_t ifam_metric; 124 struct if_data ifam_data; 125 }; 126 127 #define SA_SIZE32(sa) \ 128 ( (((struct sockaddr *)(sa))->sa_len == 0) ? \ 129 sizeof(int) : \ 130 1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) ) 131 132 #endif /* COMPAT_FREEBSD32 */ 133 134 struct linear_buffer { 135 char *base; /* Base allocated memory pointer */ 136 uint32_t offset; /* Currently used offset */ 137 uint32_t size; /* Total buffer size */ 138 }; 139 #define SCRATCH_BUFFER_SIZE 1024 140 141 #define RTS_PID_LOG(_l, _fmt, ...) RT_LOG_##_l(_l, "PID %d: " _fmt, curproc ? curproc->p_pid : 0, ## __VA_ARGS__) 142 143 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables"); 144 145 /* NB: these are not modified */ 146 static struct sockaddr route_src = { 2, PF_ROUTE, }; 147 static struct sockaddr sa_zero = { sizeof(sa_zero), AF_INET, }; 148 149 /* These are external hooks for CARP. */ 150 int (*carp_get_vhid_p)(struct ifaddr *); 151 152 /* 153 * Used by rtsock callback code to decide whether to filter the update 154 * notification to a socket bound to a particular FIB. 155 */ 156 #define RTS_FILTER_FIB M_PROTO8 157 /* 158 * Used to store address family of the notification. 159 */ 160 #define m_rtsock_family m_pkthdr.PH_loc.eight[0] 161 162 struct rcb { 163 LIST_ENTRY(rcb) list; 164 struct socket *rcb_socket; 165 sa_family_t rcb_family; 166 }; 167 168 typedef struct { 169 LIST_HEAD(, rcb) cblist; 170 int ip_count; /* attached w/ AF_INET */ 171 int ip6_count; /* attached w/ AF_INET6 */ 172 int any_count; /* total attached */ 173 } route_cb_t; 174 VNET_DEFINE_STATIC(route_cb_t, route_cb); 175 #define V_route_cb VNET(route_cb) 176 177 struct mtx rtsock_mtx; 178 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF); 179 180 #define RTSOCK_LOCK() mtx_lock(&rtsock_mtx) 181 #define RTSOCK_UNLOCK() mtx_unlock(&rtsock_mtx) 182 #define RTSOCK_LOCK_ASSERT() mtx_assert(&rtsock_mtx, MA_OWNED) 183 184 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, ""); 185 186 struct walkarg { 187 int family; 188 int w_tmemsize; 189 int w_op, w_arg; 190 caddr_t w_tmem; 191 struct sysctl_req *w_req; 192 struct sockaddr *dst; 193 struct sockaddr *mask; 194 }; 195 196 static void rts_input(struct mbuf *m); 197 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo); 198 static int rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, 199 struct walkarg *w, int *plen); 200 static int rt_xaddrs(caddr_t cp, caddr_t cplim, 201 struct rt_addrinfo *rtinfo); 202 static int cleanup_xaddrs(struct rt_addrinfo *info, struct linear_buffer *lb); 203 static int sysctl_dumpentry(struct rtentry *rt, void *vw); 204 static int sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, 205 uint32_t weight, struct walkarg *w); 206 static int sysctl_iflist(int af, struct walkarg *w); 207 static int sysctl_ifmalist(int af, struct walkarg *w); 208 static void rt_getmetrics(const struct rtentry *rt, 209 const struct nhop_object *nh, struct rt_metrics *out); 210 static void rt_dispatch(struct mbuf *, sa_family_t); 211 static void rt_ifannouncemsg(struct ifnet *ifp, int what); 212 static int handle_rtm_get(struct rt_addrinfo *info, u_int fibnum, 213 struct rt_msghdr *rtm, struct rib_cmd_info *rc); 214 static int update_rtm_from_rc(struct rt_addrinfo *info, 215 struct rt_msghdr **prtm, int alloc_len, 216 struct rib_cmd_info *rc, struct nhop_object *nh); 217 static void send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, 218 struct mbuf *m, sa_family_t saf, u_int fibnum, 219 int rtm_errno); 220 static bool can_export_rte(struct ucred *td_ucred, bool rt_is_host, 221 const struct sockaddr *rt_dst); 222 223 static struct netisr_handler rtsock_nh = { 224 .nh_name = "rtsock", 225 .nh_handler = rts_input, 226 .nh_proto = NETISR_ROUTE, 227 .nh_policy = NETISR_POLICY_SOURCE, 228 }; 229 230 static int 231 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS) 232 { 233 int error, qlimit; 234 235 netisr_getqlimit(&rtsock_nh, &qlimit); 236 error = sysctl_handle_int(oidp, &qlimit, 0, req); 237 if (error || !req->newptr) 238 return (error); 239 if (qlimit < 1) 240 return (EINVAL); 241 return (netisr_setqlimit(&rtsock_nh, qlimit)); 242 } 243 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, 244 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 245 0, 0, sysctl_route_netisr_maxqlen, "I", 246 "maximum routing socket dispatch queue length"); 247 248 static void 249 vnet_rts_init(void) 250 { 251 int tmp; 252 253 if (IS_DEFAULT_VNET(curvnet)) { 254 if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp)) 255 rtsock_nh.nh_qlimit = tmp; 256 netisr_register(&rtsock_nh); 257 } 258 #ifdef VIMAGE 259 else 260 netisr_register_vnet(&rtsock_nh); 261 #endif 262 } 263 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 264 vnet_rts_init, 0); 265 266 #ifdef VIMAGE 267 static void 268 vnet_rts_uninit(void) 269 { 270 271 netisr_unregister_vnet(&rtsock_nh); 272 } 273 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 274 vnet_rts_uninit, 0); 275 #endif 276 277 static void 278 rts_handle_ifnet_arrival(void *arg __unused, struct ifnet *ifp) 279 { 280 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 281 } 282 EVENTHANDLER_DEFINE(ifnet_arrival_event, rts_handle_ifnet_arrival, NULL, 0); 283 284 static void 285 rts_handle_ifnet_departure(void *arg __unused, struct ifnet *ifp) 286 { 287 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 288 } 289 EVENTHANDLER_DEFINE(ifnet_departure_event, rts_handle_ifnet_departure, NULL, 0); 290 291 static void 292 rts_append_data(struct socket *so, struct mbuf *m) 293 { 294 295 if (sbappendaddr(&so->so_rcv, &route_src, m, NULL) == 0) { 296 soroverflow(so); 297 m_freem(m); 298 } else 299 sorwakeup(so); 300 } 301 302 static void 303 rts_input(struct mbuf *m) 304 { 305 struct rcb *rcb; 306 struct socket *last; 307 308 last = NULL; 309 RTSOCK_LOCK(); 310 LIST_FOREACH(rcb, &V_route_cb.cblist, list) { 311 if (rcb->rcb_family != AF_UNSPEC && 312 rcb->rcb_family != m->m_rtsock_family) 313 continue; 314 if ((m->m_flags & RTS_FILTER_FIB) && 315 M_GETFIB(m) != rcb->rcb_socket->so_fibnum) 316 continue; 317 if (last != NULL) { 318 struct mbuf *n; 319 320 n = m_copym(m, 0, M_COPYALL, M_NOWAIT); 321 if (n != NULL) 322 rts_append_data(last, n); 323 } 324 last = rcb->rcb_socket; 325 } 326 if (last != NULL) 327 rts_append_data(last, m); 328 else 329 m_freem(m); 330 RTSOCK_UNLOCK(); 331 } 332 333 static void 334 rts_close(struct socket *so) 335 { 336 337 soisdisconnected(so); 338 } 339 340 static SYSCTL_NODE(_net, OID_AUTO, rtsock, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 341 "Routing socket infrastructure"); 342 static u_long rts_sendspace = 8192; 343 SYSCTL_ULONG(_net_rtsock, OID_AUTO, sendspace, CTLFLAG_RW, &rts_sendspace, 0, 344 "Default routing socket send space"); 345 static u_long rts_recvspace = 8192; 346 SYSCTL_ULONG(_net_rtsock, OID_AUTO, recvspace, CTLFLAG_RW, &rts_recvspace, 0, 347 "Default routing socket receive space"); 348 349 static int 350 rts_attach(struct socket *so, int proto, struct thread *td) 351 { 352 struct rcb *rcb; 353 int error; 354 355 error = soreserve(so, rts_sendspace, rts_recvspace); 356 if (error) 357 return (error); 358 359 rcb = malloc(sizeof(*rcb), M_PCB, M_WAITOK); 360 rcb->rcb_socket = so; 361 rcb->rcb_family = proto; 362 363 so->so_pcb = rcb; 364 so->so_fibnum = td->td_proc->p_fibnum; 365 so->so_options |= SO_USELOOPBACK; 366 367 RTSOCK_LOCK(); 368 LIST_INSERT_HEAD(&V_route_cb.cblist, rcb, list); 369 switch (proto) { 370 case AF_INET: 371 V_route_cb.ip_count++; 372 break; 373 case AF_INET6: 374 V_route_cb.ip6_count++; 375 break; 376 } 377 V_route_cb.any_count++; 378 RTSOCK_UNLOCK(); 379 soisconnected(so); 380 381 return (0); 382 } 383 384 static void 385 rts_detach(struct socket *so) 386 { 387 struct rcb *rcb = so->so_pcb; 388 389 RTSOCK_LOCK(); 390 LIST_REMOVE(rcb, list); 391 switch(rcb->rcb_family) { 392 case AF_INET: 393 V_route_cb.ip_count--; 394 break; 395 case AF_INET6: 396 V_route_cb.ip6_count--; 397 break; 398 } 399 V_route_cb.any_count--; 400 RTSOCK_UNLOCK(); 401 free(rcb, M_PCB); 402 so->so_pcb = NULL; 403 } 404 405 static int 406 rts_shutdown(struct socket *so) 407 { 408 409 socantsendmore(so); 410 return (0); 411 } 412 413 #ifndef _SOCKADDR_UNION_DEFINED 414 #define _SOCKADDR_UNION_DEFINED 415 /* 416 * The union of all possible address formats we handle. 417 */ 418 union sockaddr_union { 419 struct sockaddr sa; 420 struct sockaddr_in sin; 421 struct sockaddr_in6 sin6; 422 }; 423 #endif /* _SOCKADDR_UNION_DEFINED */ 424 425 static int 426 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp, 427 struct nhop_object *nh, union sockaddr_union *saun, struct ucred *cred) 428 { 429 #if defined(INET) || defined(INET6) 430 struct epoch_tracker et; 431 #endif 432 433 /* First, see if the returned address is part of the jail. */ 434 if (prison_if(cred, nh->nh_ifa->ifa_addr) == 0) { 435 info->rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr; 436 return (0); 437 } 438 439 switch (info->rti_info[RTAX_DST]->sa_family) { 440 #ifdef INET 441 case AF_INET: 442 { 443 struct in_addr ia; 444 struct ifaddr *ifa; 445 int found; 446 447 found = 0; 448 /* 449 * Try to find an address on the given outgoing interface 450 * that belongs to the jail. 451 */ 452 NET_EPOCH_ENTER(et); 453 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 454 struct sockaddr *sa; 455 sa = ifa->ifa_addr; 456 if (sa->sa_family != AF_INET) 457 continue; 458 ia = ((struct sockaddr_in *)sa)->sin_addr; 459 if (prison_check_ip4(cred, &ia) == 0) { 460 found = 1; 461 break; 462 } 463 } 464 NET_EPOCH_EXIT(et); 465 if (!found) { 466 /* 467 * As a last resort return the 'default' jail address. 468 */ 469 ia = ((struct sockaddr_in *)nh->nh_ifa->ifa_addr)-> 470 sin_addr; 471 if (prison_get_ip4(cred, &ia) != 0) 472 return (ESRCH); 473 } 474 bzero(&saun->sin, sizeof(struct sockaddr_in)); 475 saun->sin.sin_len = sizeof(struct sockaddr_in); 476 saun->sin.sin_family = AF_INET; 477 saun->sin.sin_addr.s_addr = ia.s_addr; 478 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin; 479 break; 480 } 481 #endif 482 #ifdef INET6 483 case AF_INET6: 484 { 485 struct in6_addr ia6; 486 struct ifaddr *ifa; 487 int found; 488 489 found = 0; 490 /* 491 * Try to find an address on the given outgoing interface 492 * that belongs to the jail. 493 */ 494 NET_EPOCH_ENTER(et); 495 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 496 struct sockaddr *sa; 497 sa = ifa->ifa_addr; 498 if (sa->sa_family != AF_INET6) 499 continue; 500 bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr, 501 &ia6, sizeof(struct in6_addr)); 502 if (prison_check_ip6(cred, &ia6) == 0) { 503 found = 1; 504 break; 505 } 506 } 507 NET_EPOCH_EXIT(et); 508 if (!found) { 509 /* 510 * As a last resort return the 'default' jail address. 511 */ 512 ia6 = ((struct sockaddr_in6 *)nh->nh_ifa->ifa_addr)-> 513 sin6_addr; 514 if (prison_get_ip6(cred, &ia6) != 0) 515 return (ESRCH); 516 } 517 bzero(&saun->sin6, sizeof(struct sockaddr_in6)); 518 saun->sin6.sin6_len = sizeof(struct sockaddr_in6); 519 saun->sin6.sin6_family = AF_INET6; 520 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr)); 521 if (sa6_recoverscope(&saun->sin6) != 0) 522 return (ESRCH); 523 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6; 524 break; 525 } 526 #endif 527 default: 528 return (ESRCH); 529 } 530 return (0); 531 } 532 533 static int 534 fill_blackholeinfo(struct rt_addrinfo *info, union sockaddr_union *saun) 535 { 536 struct ifaddr *ifa; 537 sa_family_t saf; 538 539 if (V_loif == NULL) { 540 RTS_PID_LOG(LOG_INFO, "Unable to add blackhole/reject nhop without loopback"); 541 return (ENOTSUP); 542 } 543 info->rti_ifp = V_loif; 544 545 saf = info->rti_info[RTAX_DST]->sa_family; 546 547 CK_STAILQ_FOREACH(ifa, &info->rti_ifp->if_addrhead, ifa_link) { 548 if (ifa->ifa_addr->sa_family == saf) { 549 info->rti_ifa = ifa; 550 break; 551 } 552 } 553 if (info->rti_ifa == NULL) { 554 RTS_PID_LOG(LOG_INFO, "Unable to find ifa for blackhole/reject nhop"); 555 return (ENOTSUP); 556 } 557 558 bzero(saun, sizeof(union sockaddr_union)); 559 switch (saf) { 560 #ifdef INET 561 case AF_INET: 562 saun->sin.sin_family = AF_INET; 563 saun->sin.sin_len = sizeof(struct sockaddr_in); 564 saun->sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK); 565 break; 566 #endif 567 #ifdef INET6 568 case AF_INET6: 569 saun->sin6.sin6_family = AF_INET6; 570 saun->sin6.sin6_len = sizeof(struct sockaddr_in6); 571 saun->sin6.sin6_addr = in6addr_loopback; 572 break; 573 #endif 574 default: 575 RTS_PID_LOG(LOG_INFO, "unsupported family: %d", saf); 576 return (ENOTSUP); 577 } 578 info->rti_info[RTAX_GATEWAY] = &saun->sa; 579 info->rti_flags |= RTF_GATEWAY; 580 581 return (0); 582 } 583 584 /* 585 * Fills in @info based on userland-provided @rtm message. 586 * 587 * Returns 0 on success. 588 */ 589 static int 590 fill_addrinfo(struct rt_msghdr *rtm, int len, struct linear_buffer *lb, u_int fibnum, 591 struct rt_addrinfo *info) 592 { 593 int error; 594 595 rtm->rtm_pid = curproc->p_pid; 596 info->rti_addrs = rtm->rtm_addrs; 597 598 info->rti_mflags = rtm->rtm_inits; 599 info->rti_rmx = &rtm->rtm_rmx; 600 601 /* 602 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6 603 * link-local address because rtrequest requires addresses with 604 * embedded scope id. 605 */ 606 if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info)) 607 return (EINVAL); 608 609 info->rti_flags = rtm->rtm_flags; 610 error = cleanup_xaddrs(info, lb); 611 if (error != 0) 612 return (error); 613 /* 614 * Verify that the caller has the appropriate privilege; RTM_GET 615 * is the only operation the non-superuser is allowed. 616 */ 617 if (rtm->rtm_type != RTM_GET) { 618 error = priv_check(curthread, PRIV_NET_ROUTE); 619 if (error != 0) 620 return (error); 621 } 622 623 /* 624 * The given gateway address may be an interface address. 625 * For example, issuing a "route change" command on a route 626 * entry that was created from a tunnel, and the gateway 627 * address given is the local end point. In this case the 628 * RTF_GATEWAY flag must be cleared or the destination will 629 * not be reachable even though there is no error message. 630 */ 631 if (info->rti_info[RTAX_GATEWAY] != NULL && 632 info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) { 633 struct rt_addrinfo ginfo; 634 struct sockaddr *gdst; 635 struct sockaddr_storage ss; 636 637 bzero(&ginfo, sizeof(ginfo)); 638 bzero(&ss, sizeof(ss)); 639 ss.ss_len = sizeof(ss); 640 641 ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss; 642 gdst = info->rti_info[RTAX_GATEWAY]; 643 644 /* 645 * A host route through the loopback interface is 646 * installed for each interface adddress. In pre 8.0 647 * releases the interface address of a PPP link type 648 * is not reachable locally. This behavior is fixed as 649 * part of the new L2/L3 redesign and rewrite work. The 650 * signature of this interface address route is the 651 * AF_LINK sa_family type of the gateway, and the 652 * rt_ifp has the IFF_LOOPBACK flag set. 653 */ 654 if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) { 655 if (ss.ss_family == AF_LINK && 656 ginfo.rti_ifp->if_flags & IFF_LOOPBACK) { 657 info->rti_flags &= ~RTF_GATEWAY; 658 info->rti_flags |= RTF_GWFLAG_COMPAT; 659 } 660 rib_free_info(&ginfo); 661 } 662 } 663 664 return (0); 665 } 666 667 static struct nhop_object * 668 select_nhop(struct nhop_object *nh, const struct sockaddr *gw) 669 { 670 if (!NH_IS_NHGRP(nh)) 671 return (nh); 672 #ifdef ROUTE_MPATH 673 const struct weightened_nhop *wn; 674 uint32_t num_nhops; 675 wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops); 676 if (gw == NULL) 677 return (wn[0].nh); 678 for (int i = 0; i < num_nhops; i++) { 679 if (match_nhop_gw(wn[i].nh, gw)) 680 return (wn[i].nh); 681 } 682 #endif 683 return (NULL); 684 } 685 686 /* 687 * Handles RTM_GET message from routing socket, returning matching rt. 688 * 689 * Returns: 690 * 0 on success, with locked and referenced matching rt in @rt_nrt 691 * errno of failure 692 */ 693 static int 694 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum, 695 struct rt_msghdr *rtm, struct rib_cmd_info *rc) 696 { 697 RIB_RLOCK_TRACKER; 698 struct rib_head *rnh; 699 struct nhop_object *nh; 700 sa_family_t saf; 701 702 saf = info->rti_info[RTAX_DST]->sa_family; 703 704 rnh = rt_tables_get_rnh(fibnum, saf); 705 if (rnh == NULL) 706 return (EAFNOSUPPORT); 707 708 RIB_RLOCK(rnh); 709 710 /* 711 * By (implicit) convention host route (one without netmask) 712 * means longest-prefix-match request and the route with netmask 713 * means exact-match lookup. 714 * As cleanup_xaddrs() cleans up info flags&addrs for the /32,/128 715 * prefixes, use original data to check for the netmask presence. 716 */ 717 if ((rtm->rtm_addrs & RTA_NETMASK) == 0) { 718 /* 719 * Provide longest prefix match for 720 * address lookup (no mask). 721 * 'route -n get addr' 722 */ 723 rc->rc_rt = (struct rtentry *) rnh->rnh_matchaddr( 724 info->rti_info[RTAX_DST], &rnh->head); 725 } else 726 rc->rc_rt = (struct rtentry *) rnh->rnh_lookup( 727 info->rti_info[RTAX_DST], 728 info->rti_info[RTAX_NETMASK], &rnh->head); 729 730 if (rc->rc_rt == NULL) { 731 RIB_RUNLOCK(rnh); 732 return (ESRCH); 733 } 734 735 nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]); 736 if (nh == NULL) { 737 RIB_RUNLOCK(rnh); 738 return (ESRCH); 739 } 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 * TODO: move this logic to userland. 746 */ 747 if (rtm->rtm_flags & RTF_ANNOUNCE) { 748 struct sockaddr_storage laddr; 749 750 if (nh->nh_ifp != NULL && 751 nh->nh_ifp->if_type == IFT_PROPVIRTUAL) { 752 struct ifaddr *ifa; 753 754 ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1, 755 RT_ALL_FIBS); 756 if (ifa != NULL) 757 rt_maskedcopy(ifa->ifa_addr, 758 (struct sockaddr *)&laddr, 759 ifa->ifa_netmask); 760 } else 761 rt_maskedcopy(nh->nh_ifa->ifa_addr, 762 (struct sockaddr *)&laddr, 763 nh->nh_ifa->ifa_netmask); 764 /* 765 * refactor rt and no lock operation necessary 766 */ 767 rc->rc_rt = (struct rtentry *)rnh->rnh_matchaddr( 768 (struct sockaddr *)&laddr, &rnh->head); 769 if (rc->rc_rt == NULL) { 770 RIB_RUNLOCK(rnh); 771 return (ESRCH); 772 } 773 nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]); 774 if (nh == NULL) { 775 RIB_RUNLOCK(rnh); 776 return (ESRCH); 777 } 778 } 779 rc->rc_nh_new = nh; 780 rc->rc_nh_weight = rc->rc_rt->rt_weight; 781 RIB_RUNLOCK(rnh); 782 783 return (0); 784 } 785 786 static void 787 init_sockaddrs_family(int family, struct sockaddr *dst, struct sockaddr *mask) 788 { 789 #ifdef INET 790 if (family == AF_INET) { 791 struct sockaddr_in *dst4 = (struct sockaddr_in *)dst; 792 struct sockaddr_in *mask4 = (struct sockaddr_in *)mask; 793 794 bzero(dst4, sizeof(struct sockaddr_in)); 795 bzero(mask4, sizeof(struct sockaddr_in)); 796 797 dst4->sin_family = AF_INET; 798 dst4->sin_len = sizeof(struct sockaddr_in); 799 mask4->sin_family = AF_INET; 800 mask4->sin_len = sizeof(struct sockaddr_in); 801 } 802 #endif 803 #ifdef INET6 804 if (family == AF_INET6) { 805 struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst; 806 struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask; 807 808 bzero(dst6, sizeof(struct sockaddr_in6)); 809 bzero(mask6, sizeof(struct sockaddr_in6)); 810 811 dst6->sin6_family = AF_INET6; 812 dst6->sin6_len = sizeof(struct sockaddr_in6); 813 mask6->sin6_family = AF_INET6; 814 mask6->sin6_len = sizeof(struct sockaddr_in6); 815 } 816 #endif 817 } 818 819 static void 820 export_rtaddrs(const struct rtentry *rt, struct sockaddr *dst, 821 struct sockaddr *mask) 822 { 823 #ifdef INET 824 if (dst->sa_family == AF_INET) { 825 struct sockaddr_in *dst4 = (struct sockaddr_in *)dst; 826 struct sockaddr_in *mask4 = (struct sockaddr_in *)mask; 827 uint32_t scopeid = 0; 828 rt_get_inet_prefix_pmask(rt, &dst4->sin_addr, &mask4->sin_addr, 829 &scopeid); 830 return; 831 } 832 #endif 833 #ifdef INET6 834 if (dst->sa_family == AF_INET6) { 835 struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst; 836 struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask; 837 uint32_t scopeid = 0; 838 rt_get_inet6_prefix_pmask(rt, &dst6->sin6_addr, 839 &mask6->sin6_addr, &scopeid); 840 dst6->sin6_scope_id = scopeid; 841 return; 842 } 843 #endif 844 } 845 846 static int 847 update_rtm_from_info(struct rt_addrinfo *info, struct rt_msghdr **prtm, 848 int alloc_len) 849 { 850 struct rt_msghdr *rtm, *orig_rtm = NULL; 851 struct walkarg w; 852 int len; 853 854 rtm = *prtm; 855 /* Check if we need to realloc storage */ 856 rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len); 857 if (len > alloc_len) { 858 struct rt_msghdr *tmp_rtm; 859 860 tmp_rtm = malloc(len, M_TEMP, M_NOWAIT); 861 if (tmp_rtm == NULL) 862 return (ENOBUFS); 863 bcopy(rtm, tmp_rtm, rtm->rtm_msglen); 864 orig_rtm = rtm; 865 rtm = tmp_rtm; 866 alloc_len = len; 867 868 /* 869 * Delay freeing original rtm as info contains 870 * data referencing it. 871 */ 872 } 873 874 w.w_tmem = (caddr_t)rtm; 875 w.w_tmemsize = alloc_len; 876 rtsock_msg_buffer(rtm->rtm_type, info, &w, &len); 877 rtm->rtm_addrs = info->rti_addrs; 878 879 if (orig_rtm != NULL) 880 free(orig_rtm, M_TEMP); 881 *prtm = rtm; 882 return (0); 883 } 884 885 886 /* 887 * Update sockaddrs, flags, etc in @prtm based on @rc data. 888 * rtm can be reallocated. 889 * 890 * Returns 0 on success, along with pointer to (potentially reallocated) 891 * rtm. 892 * 893 */ 894 static int 895 update_rtm_from_rc(struct rt_addrinfo *info, struct rt_msghdr **prtm, 896 int alloc_len, struct rib_cmd_info *rc, struct nhop_object *nh) 897 { 898 union sockaddr_union saun; 899 struct rt_msghdr *rtm; 900 struct ifnet *ifp; 901 int error; 902 903 rtm = *prtm; 904 union sockaddr_union sa_dst, sa_mask; 905 int family = info->rti_info[RTAX_DST]->sa_family; 906 init_sockaddrs_family(family, &sa_dst.sa, &sa_mask.sa); 907 export_rtaddrs(rc->rc_rt, &sa_dst.sa, &sa_mask.sa); 908 909 info->rti_info[RTAX_DST] = &sa_dst.sa; 910 info->rti_info[RTAX_NETMASK] = rt_is_host(rc->rc_rt) ? NULL : &sa_mask.sa; 911 info->rti_info[RTAX_GATEWAY] = &nh->gw_sa; 912 info->rti_info[RTAX_GENMASK] = 0; 913 ifp = nh->nh_ifp; 914 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) { 915 if (ifp) { 916 info->rti_info[RTAX_IFP] = 917 ifp->if_addr->ifa_addr; 918 error = rtm_get_jailed(info, ifp, nh, 919 &saun, curthread->td_ucred); 920 if (error != 0) 921 return (error); 922 if (ifp->if_flags & IFF_POINTOPOINT) 923 info->rti_info[RTAX_BRD] = 924 nh->nh_ifa->ifa_dstaddr; 925 rtm->rtm_index = ifp->if_index; 926 } else { 927 info->rti_info[RTAX_IFP] = NULL; 928 info->rti_info[RTAX_IFA] = NULL; 929 } 930 } else if (ifp != NULL) 931 rtm->rtm_index = ifp->if_index; 932 933 if ((error = update_rtm_from_info(info, prtm, alloc_len)) != 0) 934 return (error); 935 936 rtm = *prtm; 937 rtm->rtm_flags = rc->rc_rt->rte_flags | nhop_get_rtflags(nh); 938 if (rtm->rtm_flags & RTF_GWFLAG_COMPAT) 939 rtm->rtm_flags = RTF_GATEWAY | 940 (rtm->rtm_flags & ~RTF_GWFLAG_COMPAT); 941 rt_getmetrics(rc->rc_rt, nh, &rtm->rtm_rmx); 942 rtm->rtm_rmx.rmx_weight = rc->rc_nh_weight; 943 944 return (0); 945 } 946 947 #ifdef ROUTE_MPATH 948 static void 949 save_del_notification(struct rib_cmd_info *rc, void *_cbdata) 950 { 951 struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata; 952 953 if (rc->rc_cmd == RTM_DELETE) 954 *rc_new = *rc; 955 } 956 957 static void 958 save_add_notification(struct rib_cmd_info *rc, void *_cbdata) 959 { 960 struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata; 961 962 if (rc->rc_cmd == RTM_ADD) 963 *rc_new = *rc; 964 } 965 #endif 966 967 #if defined(INET6) || defined(INET) 968 static struct sockaddr * 969 alloc_sockaddr_aligned(struct linear_buffer *lb, int len) 970 { 971 len = roundup2(len, sizeof(uint64_t)); 972 if (lb->offset + len > lb->size) 973 return (NULL); 974 struct sockaddr *sa = (struct sockaddr *)(lb->base + lb->offset); 975 lb->offset += len; 976 return (sa); 977 } 978 #endif 979 980 static int 981 rts_send(struct socket *so, int flags, struct mbuf *m, 982 struct sockaddr *nam, struct mbuf *control, struct thread *td) 983 { 984 struct rt_msghdr *rtm = NULL; 985 struct rt_addrinfo info; 986 struct epoch_tracker et; 987 #ifdef INET6 988 struct sockaddr_storage ss; 989 struct sockaddr_in6 *sin6; 990 int i, rti_need_deembed = 0; 991 #endif 992 int alloc_len = 0, len, error = 0, fibnum; 993 sa_family_t saf = AF_UNSPEC; 994 struct rib_cmd_info rc; 995 struct nhop_object *nh; 996 997 if ((flags & PRUS_OOB) || control != NULL) { 998 m_freem(m); 999 if (control != NULL) 1000 m_freem(control); 1001 return (EOPNOTSUPP); 1002 } 1003 1004 fibnum = so->so_fibnum; 1005 #define senderr(e) { error = e; goto flush;} 1006 if (m == NULL || ((m->m_len < sizeof(long)) && 1007 (m = m_pullup(m, sizeof(long))) == NULL)) 1008 return (ENOBUFS); 1009 if ((m->m_flags & M_PKTHDR) == 0) 1010 panic("route_output"); 1011 NET_EPOCH_ENTER(et); 1012 len = m->m_pkthdr.len; 1013 if (len < sizeof(*rtm) || 1014 len != mtod(m, struct rt_msghdr *)->rtm_msglen) 1015 senderr(EINVAL); 1016 1017 /* 1018 * Most of current messages are in range 200-240 bytes, 1019 * minimize possible re-allocation on reply using larger size 1020 * buffer aligned on 1k boundaty. 1021 */ 1022 alloc_len = roundup2(len, 1024); 1023 int total_len = alloc_len + SCRATCH_BUFFER_SIZE; 1024 if ((rtm = malloc(total_len, M_TEMP, M_NOWAIT)) == NULL) 1025 senderr(ENOBUFS); 1026 1027 m_copydata(m, 0, len, (caddr_t)rtm); 1028 bzero(&info, sizeof(info)); 1029 nh = NULL; 1030 struct linear_buffer lb = { 1031 .base = (char *)rtm + alloc_len, 1032 .size = SCRATCH_BUFFER_SIZE, 1033 }; 1034 1035 if (rtm->rtm_version != RTM_VERSION) { 1036 /* Do not touch message since format is unknown */ 1037 free(rtm, M_TEMP); 1038 rtm = NULL; 1039 senderr(EPROTONOSUPPORT); 1040 } 1041 1042 /* 1043 * Starting from here, it is possible 1044 * to alter original message and insert 1045 * caller PID and error value. 1046 */ 1047 1048 if ((error = fill_addrinfo(rtm, len, &lb, fibnum, &info)) != 0) { 1049 senderr(error); 1050 } 1051 /* fill_addringo() embeds scope into IPv6 addresses */ 1052 #ifdef INET6 1053 rti_need_deembed = 1; 1054 #endif 1055 1056 saf = info.rti_info[RTAX_DST]->sa_family; 1057 1058 /* support for new ARP code */ 1059 if (rtm->rtm_flags & RTF_LLDATA) { 1060 error = lla_rt_output(rtm, &info); 1061 goto flush; 1062 } 1063 1064 union sockaddr_union gw_saun; 1065 int blackhole_flags = rtm->rtm_flags & (RTF_BLACKHOLE|RTF_REJECT); 1066 if (blackhole_flags != 0) { 1067 if (blackhole_flags != (RTF_BLACKHOLE | RTF_REJECT)) 1068 error = fill_blackholeinfo(&info, &gw_saun); 1069 else { 1070 RTS_PID_LOG(LOG_DEBUG, "both BLACKHOLE and REJECT flags specifiied"); 1071 error = EINVAL; 1072 } 1073 if (error != 0) 1074 senderr(error); 1075 } 1076 1077 switch (rtm->rtm_type) { 1078 case RTM_ADD: 1079 case RTM_CHANGE: 1080 if (rtm->rtm_type == RTM_ADD) { 1081 if (info.rti_info[RTAX_GATEWAY] == NULL) { 1082 RTS_PID_LOG(LOG_DEBUG, "RTM_ADD w/o gateway"); 1083 senderr(EINVAL); 1084 } 1085 } 1086 error = rib_action(fibnum, rtm->rtm_type, &info, &rc); 1087 if (error == 0) { 1088 #ifdef ROUTE_MPATH 1089 if (NH_IS_NHGRP(rc.rc_nh_new) || 1090 (rc.rc_nh_old && NH_IS_NHGRP(rc.rc_nh_old))) { 1091 struct rib_cmd_info rc_simple = {}; 1092 rib_decompose_notification(&rc, 1093 save_add_notification, (void *)&rc_simple); 1094 rc = rc_simple; 1095 } 1096 #endif 1097 /* nh MAY be empty if RTM_CHANGE request is no-op */ 1098 nh = rc.rc_nh_new; 1099 if (nh != NULL) { 1100 rtm->rtm_index = nh->nh_ifp->if_index; 1101 rtm->rtm_flags = rc.rc_rt->rte_flags | nhop_get_rtflags(nh); 1102 } 1103 } 1104 break; 1105 1106 case RTM_DELETE: 1107 error = rib_action(fibnum, RTM_DELETE, &info, &rc); 1108 if (error == 0) { 1109 #ifdef ROUTE_MPATH 1110 if (NH_IS_NHGRP(rc.rc_nh_old) || 1111 (rc.rc_nh_new && NH_IS_NHGRP(rc.rc_nh_new))) { 1112 struct rib_cmd_info rc_simple = {}; 1113 rib_decompose_notification(&rc, 1114 save_del_notification, (void *)&rc_simple); 1115 rc = rc_simple; 1116 } 1117 #endif 1118 nh = rc.rc_nh_old; 1119 } 1120 break; 1121 1122 case RTM_GET: 1123 error = handle_rtm_get(&info, fibnum, rtm, &rc); 1124 if (error != 0) 1125 senderr(error); 1126 nh = rc.rc_nh_new; 1127 1128 if (!can_export_rte(curthread->td_ucred, 1129 info.rti_info[RTAX_NETMASK] == NULL, 1130 info.rti_info[RTAX_DST])) { 1131 senderr(ESRCH); 1132 } 1133 break; 1134 1135 default: 1136 senderr(EOPNOTSUPP); 1137 } 1138 1139 if (error == 0 && nh != NULL) { 1140 error = update_rtm_from_rc(&info, &rtm, alloc_len, &rc, nh); 1141 /* 1142 * Note that some sockaddr pointers may have changed to 1143 * point to memory outsize @rtm. Some may be pointing 1144 * to the on-stack variables. 1145 * Given that, any pointer in @info CANNOT BE USED. 1146 */ 1147 1148 /* 1149 * scopeid deembedding has been performed while 1150 * writing updated rtm in rtsock_msg_buffer(). 1151 * With that in mind, skip deembedding procedure below. 1152 */ 1153 #ifdef INET6 1154 rti_need_deembed = 0; 1155 #endif 1156 } 1157 1158 flush: 1159 NET_EPOCH_EXIT(et); 1160 1161 #ifdef INET6 1162 if (rtm != NULL) { 1163 if (rti_need_deembed) { 1164 /* sin6_scope_id is recovered before sending rtm. */ 1165 sin6 = (struct sockaddr_in6 *)&ss; 1166 for (i = 0; i < RTAX_MAX; i++) { 1167 if (info.rti_info[i] == NULL) 1168 continue; 1169 if (info.rti_info[i]->sa_family != AF_INET6) 1170 continue; 1171 bcopy(info.rti_info[i], sin6, sizeof(*sin6)); 1172 if (sa6_recoverscope(sin6) == 0) 1173 bcopy(sin6, info.rti_info[i], 1174 sizeof(*sin6)); 1175 } 1176 if (update_rtm_from_info(&info, &rtm, alloc_len) != 0) { 1177 if (error != 0) 1178 error = ENOBUFS; 1179 } 1180 } 1181 } 1182 #endif 1183 send_rtm_reply(so, rtm, m, saf, fibnum, error); 1184 1185 return (error); 1186 } 1187 1188 /* 1189 * Sends the prepared reply message in @rtm to all rtsock clients. 1190 * Frees @m and @rtm. 1191 * 1192 */ 1193 static void 1194 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m, 1195 sa_family_t saf, u_int fibnum, int rtm_errno) 1196 { 1197 struct rcb *rcb = NULL; 1198 1199 /* 1200 * Check to see if we don't want our own messages. 1201 */ 1202 if ((so->so_options & SO_USELOOPBACK) == 0) { 1203 if (V_route_cb.any_count <= 1) { 1204 if (rtm != NULL) 1205 free(rtm, M_TEMP); 1206 m_freem(m); 1207 return; 1208 } 1209 /* There is another listener, so construct message */ 1210 rcb = so->so_pcb; 1211 } 1212 1213 if (rtm != NULL) { 1214 if (rtm_errno!= 0) 1215 rtm->rtm_errno = rtm_errno; 1216 else 1217 rtm->rtm_flags |= RTF_DONE; 1218 1219 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm); 1220 if (m->m_pkthdr.len < rtm->rtm_msglen) { 1221 m_freem(m); 1222 m = NULL; 1223 } else if (m->m_pkthdr.len > rtm->rtm_msglen) 1224 m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len); 1225 1226 free(rtm, M_TEMP); 1227 } 1228 if (m != NULL) { 1229 M_SETFIB(m, fibnum); 1230 m->m_flags |= RTS_FILTER_FIB; 1231 if (rcb) { 1232 /* 1233 * XXX insure we don't get a copy by 1234 * invalidating our protocol 1235 */ 1236 sa_family_t family = rcb->rcb_family; 1237 rcb->rcb_family = AF_UNSPEC; 1238 rt_dispatch(m, saf); 1239 rcb->rcb_family = family; 1240 } else 1241 rt_dispatch(m, saf); 1242 } 1243 } 1244 1245 static void 1246 rt_getmetrics(const struct rtentry *rt, const struct nhop_object *nh, 1247 struct rt_metrics *out) 1248 { 1249 1250 bzero(out, sizeof(*out)); 1251 out->rmx_mtu = nh->nh_mtu; 1252 out->rmx_weight = rt->rt_weight; 1253 out->rmx_nhidx = nhop_get_idx(nh); 1254 /* Kernel -> userland timebase conversion. */ 1255 out->rmx_expire = nhop_get_expire(nh) ? 1256 nhop_get_expire(nh) - time_uptime + time_second : 0; 1257 } 1258 1259 /* 1260 * Extract the addresses of the passed sockaddrs. 1261 * Do a little sanity checking so as to avoid bad memory references. 1262 * This data is derived straight from userland. 1263 */ 1264 static int 1265 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo) 1266 { 1267 struct sockaddr *sa; 1268 int i; 1269 1270 for (i = 0; i < RTAX_MAX && cp < cplim; i++) { 1271 if ((rtinfo->rti_addrs & (1 << i)) == 0) 1272 continue; 1273 sa = (struct sockaddr *)cp; 1274 /* 1275 * It won't fit. 1276 */ 1277 if (cp + sa->sa_len > cplim) { 1278 RTS_PID_LOG(LOG_DEBUG, "sa_len too big for sa type %d", i); 1279 return (EINVAL); 1280 } 1281 /* 1282 * there are no more.. quit now 1283 * If there are more bits, they are in error. 1284 * I've seen this. route(1) can evidently generate these. 1285 * This causes kernel to core dump. 1286 * for compatibility, If we see this, point to a safe address. 1287 */ 1288 if (sa->sa_len == 0) { 1289 rtinfo->rti_info[i] = &sa_zero; 1290 return (0); /* should be EINVAL but for compat */ 1291 } 1292 /* accept it */ 1293 #ifdef INET6 1294 if (sa->sa_family == AF_INET6) 1295 sa6_embedscope((struct sockaddr_in6 *)sa, 1296 V_ip6_use_defzone); 1297 #endif 1298 rtinfo->rti_info[i] = sa; 1299 cp += SA_SIZE(sa); 1300 } 1301 return (0); 1302 } 1303 1304 #ifdef INET 1305 static inline void 1306 fill_sockaddr_inet(struct sockaddr_in *sin, struct in_addr addr) 1307 { 1308 1309 const struct sockaddr_in nsin = { 1310 .sin_family = AF_INET, 1311 .sin_len = sizeof(struct sockaddr_in), 1312 .sin_addr = addr, 1313 }; 1314 *sin = nsin; 1315 } 1316 #endif 1317 1318 #ifdef INET6 1319 static inline void 1320 fill_sockaddr_inet6(struct sockaddr_in6 *sin6, const struct in6_addr *addr6, 1321 uint32_t scopeid) 1322 { 1323 1324 const struct sockaddr_in6 nsin6 = { 1325 .sin6_family = AF_INET6, 1326 .sin6_len = sizeof(struct sockaddr_in6), 1327 .sin6_addr = *addr6, 1328 .sin6_scope_id = scopeid, 1329 }; 1330 *sin6 = nsin6; 1331 } 1332 #endif 1333 1334 #if defined(INET6) || defined(INET) 1335 /* 1336 * Checks if gateway is suitable for lltable operations. 1337 * Lltable code requires AF_LINK gateway with ifindex 1338 * and mac address specified. 1339 * Returns 0 on success. 1340 */ 1341 static int 1342 cleanup_xaddrs_lladdr(struct rt_addrinfo *info) 1343 { 1344 struct sockaddr_dl *sdl = (struct sockaddr_dl *)info->rti_info[RTAX_GATEWAY]; 1345 1346 if (sdl->sdl_family != AF_LINK) 1347 return (EINVAL); 1348 1349 if (sdl->sdl_index == 0) { 1350 RTS_PID_LOG(LOG_DEBUG, "AF_LINK gateway w/o ifindex"); 1351 return (EINVAL); 1352 } 1353 1354 if (offsetof(struct sockaddr_dl, sdl_data) + sdl->sdl_nlen + sdl->sdl_alen > sdl->sdl_len) { 1355 RTS_PID_LOG(LOG_DEBUG, "AF_LINK gw: sdl_nlen/sdl_alen too large"); 1356 return (EINVAL); 1357 } 1358 1359 return (0); 1360 } 1361 1362 static int 1363 cleanup_xaddrs_gateway(struct rt_addrinfo *info, struct linear_buffer *lb) 1364 { 1365 struct sockaddr *gw = info->rti_info[RTAX_GATEWAY]; 1366 struct sockaddr *sa; 1367 1368 if (info->rti_flags & RTF_LLDATA) 1369 return (cleanup_xaddrs_lladdr(info)); 1370 1371 switch (gw->sa_family) { 1372 #ifdef INET 1373 case AF_INET: 1374 { 1375 struct sockaddr_in *gw_sin = (struct sockaddr_in *)gw; 1376 1377 /* Ensure reads do not go beyoud SA boundary */ 1378 if (SA_SIZE(gw) < offsetof(struct sockaddr_in, sin_zero)) { 1379 RTS_PID_LOG(LOG_DEBUG, "gateway sin_len too small: %d", 1380 gw->sa_len); 1381 return (EINVAL); 1382 } 1383 sa = alloc_sockaddr_aligned(lb, sizeof(struct sockaddr_in)); 1384 if (sa == NULL) 1385 return (ENOBUFS); 1386 fill_sockaddr_inet((struct sockaddr_in *)sa, gw_sin->sin_addr); 1387 info->rti_info[RTAX_GATEWAY] = sa; 1388 } 1389 break; 1390 #endif 1391 #ifdef INET6 1392 case AF_INET6: 1393 { 1394 struct sockaddr_in6 *gw_sin6 = (struct sockaddr_in6 *)gw; 1395 if (gw_sin6->sin6_len < sizeof(struct sockaddr_in6)) { 1396 RTS_PID_LOG(LOG_DEBUG, "gateway sin6_len too small: %d", 1397 gw->sa_len); 1398 return (EINVAL); 1399 } 1400 fill_sockaddr_inet6(gw_sin6, &gw_sin6->sin6_addr, 0); 1401 break; 1402 } 1403 #endif 1404 case AF_LINK: 1405 { 1406 struct sockaddr_dl *gw_sdl; 1407 1408 size_t sdl_min_len = offsetof(struct sockaddr_dl, sdl_data); 1409 gw_sdl = (struct sockaddr_dl *)gw; 1410 if (gw_sdl->sdl_len < sdl_min_len) { 1411 RTS_PID_LOG(LOG_DEBUG, "gateway sdl_len too small: %d", 1412 gw_sdl->sdl_len); 1413 return (EINVAL); 1414 } 1415 sa = alloc_sockaddr_aligned(lb, sizeof(struct sockaddr_dl_short)); 1416 if (sa == NULL) 1417 return (ENOBUFS); 1418 1419 const struct sockaddr_dl_short sdl = { 1420 .sdl_family = AF_LINK, 1421 .sdl_len = sizeof(struct sockaddr_dl_short), 1422 .sdl_index = gw_sdl->sdl_index, 1423 }; 1424 *((struct sockaddr_dl_short *)sa) = sdl; 1425 info->rti_info[RTAX_GATEWAY] = sa; 1426 break; 1427 } 1428 } 1429 1430 return (0); 1431 } 1432 #endif 1433 1434 static void 1435 remove_netmask(struct rt_addrinfo *info) 1436 { 1437 info->rti_info[RTAX_NETMASK] = NULL; 1438 info->rti_flags |= RTF_HOST; 1439 info->rti_addrs &= ~RTA_NETMASK; 1440 } 1441 1442 #ifdef INET 1443 static int 1444 cleanup_xaddrs_inet(struct rt_addrinfo *info, struct linear_buffer *lb) 1445 { 1446 struct sockaddr_in *dst_sa, *mask_sa; 1447 const int sa_len = sizeof(struct sockaddr_in); 1448 struct in_addr dst, mask; 1449 1450 /* Check & fixup dst/netmask combination first */ 1451 dst_sa = (struct sockaddr_in *)info->rti_info[RTAX_DST]; 1452 mask_sa = (struct sockaddr_in *)info->rti_info[RTAX_NETMASK]; 1453 1454 /* Ensure reads do not go beyound the buffer size */ 1455 if (SA_SIZE(dst_sa) < offsetof(struct sockaddr_in, sin_zero)) { 1456 RTS_PID_LOG(LOG_DEBUG, "prefix dst sin_len too small: %d", 1457 dst_sa->sin_len); 1458 return (EINVAL); 1459 } 1460 1461 if ((mask_sa != NULL) && mask_sa->sin_len < sizeof(struct sockaddr_in)) { 1462 /* 1463 * Some older routing software encode mask length into the 1464 * sin_len, thus resulting in "truncated" sockaddr. 1465 */ 1466 int len = mask_sa->sin_len - offsetof(struct sockaddr_in, sin_addr); 1467 if (len >= 0) { 1468 mask.s_addr = 0; 1469 if (len > sizeof(struct in_addr)) 1470 len = sizeof(struct in_addr); 1471 memcpy(&mask, &mask_sa->sin_addr, len); 1472 } else { 1473 RTS_PID_LOG(LOG_DEBUG, "prefix mask sin_len too small: %d", 1474 mask_sa->sin_len); 1475 return (EINVAL); 1476 } 1477 } else 1478 mask.s_addr = mask_sa ? mask_sa->sin_addr.s_addr : INADDR_BROADCAST; 1479 1480 dst.s_addr = htonl(ntohl(dst_sa->sin_addr.s_addr) & ntohl(mask.s_addr)); 1481 1482 /* Construct new "clean" dst/mask sockaddresses */ 1483 if ((dst_sa = (struct sockaddr_in *)alloc_sockaddr_aligned(lb, sa_len)) == NULL) 1484 return (ENOBUFS); 1485 fill_sockaddr_inet(dst_sa, dst); 1486 info->rti_info[RTAX_DST] = (struct sockaddr *)dst_sa; 1487 1488 if (mask.s_addr != INADDR_BROADCAST) { 1489 if ((mask_sa = (struct sockaddr_in *)alloc_sockaddr_aligned(lb, sa_len)) == NULL) 1490 return (ENOBUFS); 1491 fill_sockaddr_inet(mask_sa, mask); 1492 info->rti_info[RTAX_NETMASK] = (struct sockaddr *)mask_sa; 1493 info->rti_flags &= ~RTF_HOST; 1494 } else 1495 remove_netmask(info); 1496 1497 /* Check gateway */ 1498 if (info->rti_info[RTAX_GATEWAY] != NULL) 1499 return (cleanup_xaddrs_gateway(info, lb)); 1500 1501 return (0); 1502 } 1503 #endif 1504 1505 #ifdef INET6 1506 static int 1507 cleanup_xaddrs_inet6(struct rt_addrinfo *info, struct linear_buffer *lb) 1508 { 1509 struct sockaddr *sa; 1510 struct sockaddr_in6 *dst_sa, *mask_sa; 1511 struct in6_addr mask, *dst; 1512 const int sa_len = sizeof(struct sockaddr_in6); 1513 1514 /* Check & fixup dst/netmask combination first */ 1515 dst_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_DST]; 1516 mask_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_NETMASK]; 1517 1518 if (dst_sa->sin6_len < sizeof(struct sockaddr_in6)) { 1519 RTS_PID_LOG(LOG_DEBUG, "prefix dst sin6_len too small: %d", 1520 dst_sa->sin6_len); 1521 return (EINVAL); 1522 } 1523 1524 if (mask_sa && mask_sa->sin6_len < sizeof(struct sockaddr_in6)) { 1525 /* 1526 * Some older routing software encode mask length into the 1527 * sin6_len, thus resulting in "truncated" sockaddr. 1528 */ 1529 int len = mask_sa->sin6_len - offsetof(struct sockaddr_in6, sin6_addr); 1530 if (len >= 0) { 1531 bzero(&mask, sizeof(mask)); 1532 if (len > sizeof(struct in6_addr)) 1533 len = sizeof(struct in6_addr); 1534 memcpy(&mask, &mask_sa->sin6_addr, len); 1535 } else { 1536 RTS_PID_LOG(LOG_DEBUG, "rtsock: prefix mask sin6_len too small: %d", 1537 mask_sa->sin6_len); 1538 return (EINVAL); 1539 } 1540 } else 1541 mask = mask_sa ? mask_sa->sin6_addr : in6mask128; 1542 1543 dst = &dst_sa->sin6_addr; 1544 IN6_MASK_ADDR(dst, &mask); 1545 1546 if ((sa = alloc_sockaddr_aligned(lb, sa_len)) == NULL) 1547 return (ENOBUFS); 1548 fill_sockaddr_inet6((struct sockaddr_in6 *)sa, dst, 0); 1549 info->rti_info[RTAX_DST] = sa; 1550 1551 if (!IN6_ARE_ADDR_EQUAL(&mask, &in6mask128)) { 1552 if ((sa = alloc_sockaddr_aligned(lb, sa_len)) == NULL) 1553 return (ENOBUFS); 1554 fill_sockaddr_inet6((struct sockaddr_in6 *)sa, &mask, 0); 1555 info->rti_info[RTAX_NETMASK] = sa; 1556 info->rti_flags &= ~RTF_HOST; 1557 } else 1558 remove_netmask(info); 1559 1560 /* Check gateway */ 1561 if (info->rti_info[RTAX_GATEWAY] != NULL) 1562 return (cleanup_xaddrs_gateway(info, lb)); 1563 1564 return (0); 1565 } 1566 #endif 1567 1568 static int 1569 cleanup_xaddrs(struct rt_addrinfo *info, struct linear_buffer *lb) 1570 { 1571 int error = EAFNOSUPPORT; 1572 1573 if (info->rti_info[RTAX_DST] == NULL) { 1574 RTS_PID_LOG(LOG_DEBUG, "prefix dst is not set"); 1575 return (EINVAL); 1576 } 1577 1578 if (info->rti_flags & RTF_LLDATA) { 1579 /* 1580 * arp(8)/ndp(8) sends RTA_NETMASK for the associated 1581 * prefix along with the actual address in RTA_DST. 1582 * Remove netmask to avoid unnecessary address masking. 1583 */ 1584 remove_netmask(info); 1585 } 1586 1587 switch (info->rti_info[RTAX_DST]->sa_family) { 1588 #ifdef INET 1589 case AF_INET: 1590 error = cleanup_xaddrs_inet(info, lb); 1591 break; 1592 #endif 1593 #ifdef INET6 1594 case AF_INET6: 1595 error = cleanup_xaddrs_inet6(info, lb); 1596 break; 1597 #endif 1598 } 1599 1600 return (error); 1601 } 1602 1603 /* 1604 * Fill in @dmask with valid netmask leaving original @smask 1605 * intact. Mostly used with radix netmasks. 1606 */ 1607 struct sockaddr * 1608 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask, 1609 struct sockaddr_storage *dmask) 1610 { 1611 if (dst == NULL || smask == NULL) 1612 return (NULL); 1613 1614 memset(dmask, 0, dst->sa_len); 1615 memcpy(dmask, smask, smask->sa_len); 1616 dmask->ss_len = dst->sa_len; 1617 dmask->ss_family = dst->sa_family; 1618 1619 return ((struct sockaddr *)dmask); 1620 } 1621 1622 /* 1623 * Writes information related to @rtinfo object to newly-allocated mbuf. 1624 * Assumes MCLBYTES is enough to construct any message. 1625 * Used for OS notifications of vaious events (if/ifa announces,etc) 1626 * 1627 * Returns allocated mbuf or NULL on failure. 1628 */ 1629 static struct mbuf * 1630 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo) 1631 { 1632 struct sockaddr_storage ss; 1633 struct rt_msghdr *rtm; 1634 struct mbuf *m; 1635 int i; 1636 struct sockaddr *sa; 1637 #ifdef INET6 1638 struct sockaddr_in6 *sin6; 1639 #endif 1640 int len, dlen; 1641 1642 switch (type) { 1643 case RTM_DELADDR: 1644 case RTM_NEWADDR: 1645 len = sizeof(struct ifa_msghdr); 1646 break; 1647 1648 case RTM_DELMADDR: 1649 case RTM_NEWMADDR: 1650 len = sizeof(struct ifma_msghdr); 1651 break; 1652 1653 case RTM_IFINFO: 1654 len = sizeof(struct if_msghdr); 1655 break; 1656 1657 case RTM_IFANNOUNCE: 1658 case RTM_IEEE80211: 1659 len = sizeof(struct if_announcemsghdr); 1660 break; 1661 1662 default: 1663 len = sizeof(struct rt_msghdr); 1664 } 1665 1666 /* XXXGL: can we use MJUMPAGESIZE cluster here? */ 1667 KASSERT(len <= MCLBYTES, ("%s: message too big", __func__)); 1668 if (len > MHLEN) 1669 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 1670 else 1671 m = m_gethdr(M_NOWAIT, MT_DATA); 1672 if (m == NULL) 1673 return (m); 1674 1675 m->m_pkthdr.len = m->m_len = len; 1676 rtm = mtod(m, struct rt_msghdr *); 1677 bzero((caddr_t)rtm, len); 1678 for (i = 0; i < RTAX_MAX; i++) { 1679 if ((sa = rtinfo->rti_info[i]) == NULL) 1680 continue; 1681 rtinfo->rti_addrs |= (1 << i); 1682 1683 dlen = SA_SIZE(sa); 1684 KASSERT(dlen <= sizeof(ss), 1685 ("%s: sockaddr size overflow", __func__)); 1686 bzero(&ss, sizeof(ss)); 1687 bcopy(sa, &ss, sa->sa_len); 1688 sa = (struct sockaddr *)&ss; 1689 #ifdef INET6 1690 if (sa->sa_family == AF_INET6) { 1691 sin6 = (struct sockaddr_in6 *)sa; 1692 (void)sa6_recoverscope(sin6); 1693 } 1694 #endif 1695 m_copyback(m, len, dlen, (caddr_t)sa); 1696 len += dlen; 1697 } 1698 if (m->m_pkthdr.len != len) { 1699 m_freem(m); 1700 return (NULL); 1701 } 1702 rtm->rtm_msglen = len; 1703 rtm->rtm_version = RTM_VERSION; 1704 rtm->rtm_type = type; 1705 return (m); 1706 } 1707 1708 /* 1709 * Writes information related to @rtinfo object to preallocated buffer. 1710 * Stores needed size in @plen. If @w is NULL, calculates size without 1711 * writing. 1712 * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation. 1713 * 1714 * Returns 0 on success. 1715 * 1716 */ 1717 static int 1718 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen) 1719 { 1720 struct sockaddr_storage ss; 1721 int len, buflen = 0, dlen, i; 1722 caddr_t cp = NULL; 1723 struct rt_msghdr *rtm = NULL; 1724 #ifdef INET6 1725 struct sockaddr_in6 *sin6; 1726 #endif 1727 #ifdef COMPAT_FREEBSD32 1728 bool compat32 = false; 1729 #endif 1730 1731 switch (type) { 1732 case RTM_DELADDR: 1733 case RTM_NEWADDR: 1734 if (w != NULL && w->w_op == NET_RT_IFLISTL) { 1735 #ifdef COMPAT_FREEBSD32 1736 if (w->w_req->flags & SCTL_MASK32) { 1737 len = sizeof(struct ifa_msghdrl32); 1738 compat32 = true; 1739 } else 1740 #endif 1741 len = sizeof(struct ifa_msghdrl); 1742 } else 1743 len = sizeof(struct ifa_msghdr); 1744 break; 1745 1746 case RTM_IFINFO: 1747 #ifdef COMPAT_FREEBSD32 1748 if (w != NULL && w->w_req->flags & SCTL_MASK32) { 1749 if (w->w_op == NET_RT_IFLISTL) 1750 len = sizeof(struct if_msghdrl32); 1751 else 1752 len = sizeof(struct if_msghdr32); 1753 compat32 = true; 1754 break; 1755 } 1756 #endif 1757 if (w != NULL && w->w_op == NET_RT_IFLISTL) 1758 len = sizeof(struct if_msghdrl); 1759 else 1760 len = sizeof(struct if_msghdr); 1761 break; 1762 1763 case RTM_NEWMADDR: 1764 len = sizeof(struct ifma_msghdr); 1765 break; 1766 1767 default: 1768 len = sizeof(struct rt_msghdr); 1769 } 1770 1771 if (w != NULL) { 1772 rtm = (struct rt_msghdr *)w->w_tmem; 1773 buflen = w->w_tmemsize - len; 1774 cp = (caddr_t)w->w_tmem + len; 1775 } 1776 1777 rtinfo->rti_addrs = 0; 1778 for (i = 0; i < RTAX_MAX; i++) { 1779 struct sockaddr *sa; 1780 1781 if ((sa = rtinfo->rti_info[i]) == NULL) 1782 continue; 1783 rtinfo->rti_addrs |= (1 << i); 1784 #ifdef COMPAT_FREEBSD32 1785 if (compat32) 1786 dlen = SA_SIZE32(sa); 1787 else 1788 #endif 1789 dlen = SA_SIZE(sa); 1790 if (cp != NULL && buflen >= dlen) { 1791 KASSERT(dlen <= sizeof(ss), 1792 ("%s: sockaddr size overflow", __func__)); 1793 bzero(&ss, sizeof(ss)); 1794 bcopy(sa, &ss, sa->sa_len); 1795 sa = (struct sockaddr *)&ss; 1796 #ifdef INET6 1797 if (sa->sa_family == AF_INET6) { 1798 sin6 = (struct sockaddr_in6 *)sa; 1799 (void)sa6_recoverscope(sin6); 1800 } 1801 #endif 1802 bcopy((caddr_t)sa, cp, (unsigned)dlen); 1803 cp += dlen; 1804 buflen -= dlen; 1805 } else if (cp != NULL) { 1806 /* 1807 * Buffer too small. Count needed size 1808 * and return with error. 1809 */ 1810 cp = NULL; 1811 } 1812 1813 len += dlen; 1814 } 1815 1816 if (cp != NULL) { 1817 dlen = ALIGN(len) - len; 1818 if (buflen < dlen) 1819 cp = NULL; 1820 else { 1821 bzero(cp, dlen); 1822 cp += dlen; 1823 buflen -= dlen; 1824 } 1825 } 1826 len = ALIGN(len); 1827 1828 if (cp != NULL) { 1829 /* fill header iff buffer is large enough */ 1830 rtm->rtm_version = RTM_VERSION; 1831 rtm->rtm_type = type; 1832 rtm->rtm_msglen = len; 1833 } 1834 1835 *plen = len; 1836 1837 if (w != NULL && cp == NULL) 1838 return (ENOBUFS); 1839 1840 return (0); 1841 } 1842 1843 /* 1844 * This routine is called to generate a message from the routing 1845 * socket indicating that a redirect has occurred, a routing lookup 1846 * has failed, or that a protocol has detected timeouts to a particular 1847 * destination. 1848 */ 1849 void 1850 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error, 1851 int fibnum) 1852 { 1853 struct rt_msghdr *rtm; 1854 struct mbuf *m; 1855 struct sockaddr *sa = rtinfo->rti_info[RTAX_DST]; 1856 1857 if (V_route_cb.any_count == 0) 1858 return; 1859 m = rtsock_msg_mbuf(type, rtinfo); 1860 if (m == NULL) 1861 return; 1862 1863 if (fibnum != RT_ALL_FIBS) { 1864 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out " 1865 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs)); 1866 M_SETFIB(m, fibnum); 1867 m->m_flags |= RTS_FILTER_FIB; 1868 } 1869 1870 rtm = mtod(m, struct rt_msghdr *); 1871 rtm->rtm_flags = RTF_DONE | flags; 1872 rtm->rtm_errno = error; 1873 rtm->rtm_addrs = rtinfo->rti_addrs; 1874 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1875 } 1876 1877 void 1878 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error) 1879 { 1880 1881 rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS); 1882 } 1883 1884 /* 1885 * This routine is called to generate a message from the routing 1886 * socket indicating that the status of a network interface has changed. 1887 */ 1888 void 1889 rt_ifmsg(struct ifnet *ifp) 1890 { 1891 struct if_msghdr *ifm; 1892 struct mbuf *m; 1893 struct rt_addrinfo info; 1894 1895 if (V_route_cb.any_count == 0) 1896 return; 1897 bzero((caddr_t)&info, sizeof(info)); 1898 m = rtsock_msg_mbuf(RTM_IFINFO, &info); 1899 if (m == NULL) 1900 return; 1901 ifm = mtod(m, struct if_msghdr *); 1902 ifm->ifm_index = ifp->if_index; 1903 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1904 if_data_copy(ifp, &ifm->ifm_data); 1905 ifm->ifm_addrs = 0; 1906 rt_dispatch(m, AF_UNSPEC); 1907 } 1908 1909 /* 1910 * Announce interface address arrival/withdraw. 1911 * Please do not call directly, use rt_addrmsg(). 1912 * Assume input data to be valid. 1913 * Returns 0 on success. 1914 */ 1915 int 1916 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum) 1917 { 1918 struct rt_addrinfo info; 1919 struct sockaddr *sa; 1920 int ncmd; 1921 struct mbuf *m; 1922 struct ifa_msghdr *ifam; 1923 struct ifnet *ifp = ifa->ifa_ifp; 1924 struct sockaddr_storage ss; 1925 1926 if (V_route_cb.any_count == 0) 1927 return (0); 1928 1929 ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR; 1930 1931 bzero((caddr_t)&info, sizeof(info)); 1932 info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr; 1933 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr; 1934 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask( 1935 info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss); 1936 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr; 1937 if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL) 1938 return (ENOBUFS); 1939 ifam = mtod(m, struct ifa_msghdr *); 1940 ifam->ifam_index = ifp->if_index; 1941 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 1942 ifam->ifam_flags = ifa->ifa_flags; 1943 ifam->ifam_addrs = info.rti_addrs; 1944 1945 if (fibnum != RT_ALL_FIBS) { 1946 M_SETFIB(m, fibnum); 1947 m->m_flags |= RTS_FILTER_FIB; 1948 } 1949 1950 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1951 1952 return (0); 1953 } 1954 1955 /* 1956 * Announce route addition/removal to rtsock based on @rt data. 1957 * Callers are advives to use rt_routemsg() instead of using this 1958 * function directly. 1959 * Assume @rt data is consistent. 1960 * 1961 * Returns 0 on success. 1962 */ 1963 int 1964 rtsock_routemsg(int cmd, struct rtentry *rt, struct nhop_object *nh, 1965 int fibnum) 1966 { 1967 union sockaddr_union dst, mask; 1968 struct rt_addrinfo info; 1969 1970 if (V_route_cb.any_count == 0) 1971 return (0); 1972 1973 int family = rt_get_family(rt); 1974 init_sockaddrs_family(family, &dst.sa, &mask.sa); 1975 export_rtaddrs(rt, &dst.sa, &mask.sa); 1976 1977 bzero((caddr_t)&info, sizeof(info)); 1978 info.rti_info[RTAX_DST] = &dst.sa; 1979 info.rti_info[RTAX_NETMASK] = &mask.sa; 1980 info.rti_info[RTAX_GATEWAY] = &nh->gw_sa; 1981 info.rti_flags = rt->rte_flags | nhop_get_rtflags(nh); 1982 info.rti_ifp = nh->nh_ifp; 1983 1984 return (rtsock_routemsg_info(cmd, &info, fibnum)); 1985 } 1986 1987 int 1988 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum) 1989 { 1990 struct rt_msghdr *rtm; 1991 struct sockaddr *sa; 1992 struct mbuf *m; 1993 1994 if (V_route_cb.any_count == 0) 1995 return (0); 1996 1997 if (info->rti_flags & RTF_HOST) 1998 info->rti_info[RTAX_NETMASK] = NULL; 1999 2000 m = rtsock_msg_mbuf(cmd, info); 2001 if (m == NULL) 2002 return (ENOBUFS); 2003 2004 if (fibnum != RT_ALL_FIBS) { 2005 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out " 2006 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs)); 2007 M_SETFIB(m, fibnum); 2008 m->m_flags |= RTS_FILTER_FIB; 2009 } 2010 2011 rtm = mtod(m, struct rt_msghdr *); 2012 rtm->rtm_addrs = info->rti_addrs; 2013 if (info->rti_ifp != NULL) 2014 rtm->rtm_index = info->rti_ifp->if_index; 2015 /* Add RTF_DONE to indicate command 'completion' required by API */ 2016 info->rti_flags |= RTF_DONE; 2017 /* Reported routes has to be up */ 2018 if (cmd == RTM_ADD || cmd == RTM_CHANGE) 2019 info->rti_flags |= RTF_UP; 2020 rtm->rtm_flags = info->rti_flags; 2021 2022 sa = info->rti_info[RTAX_DST]; 2023 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 2024 2025 return (0); 2026 } 2027 2028 /* 2029 * This is the analogue to the rt_newaddrmsg which performs the same 2030 * function but for multicast group memberhips. This is easier since 2031 * there is no route state to worry about. 2032 */ 2033 void 2034 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma) 2035 { 2036 struct rt_addrinfo info; 2037 struct mbuf *m = NULL; 2038 struct ifnet *ifp = ifma->ifma_ifp; 2039 struct ifma_msghdr *ifmam; 2040 2041 if (V_route_cb.any_count == 0) 2042 return; 2043 2044 bzero((caddr_t)&info, sizeof(info)); 2045 info.rti_info[RTAX_IFA] = ifma->ifma_addr; 2046 if (ifp && ifp->if_addr) 2047 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr; 2048 else 2049 info.rti_info[RTAX_IFP] = NULL; 2050 /* 2051 * If a link-layer address is present, present it as a ``gateway'' 2052 * (similarly to how ARP entries, e.g., are presented). 2053 */ 2054 info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr; 2055 m = rtsock_msg_mbuf(cmd, &info); 2056 if (m == NULL) 2057 return; 2058 ifmam = mtod(m, struct ifma_msghdr *); 2059 KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n", 2060 __func__)); 2061 ifmam->ifmam_index = ifp->if_index; 2062 ifmam->ifmam_addrs = info.rti_addrs; 2063 rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC); 2064 } 2065 2066 static struct mbuf * 2067 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what, 2068 struct rt_addrinfo *info) 2069 { 2070 struct if_announcemsghdr *ifan; 2071 struct mbuf *m; 2072 2073 if (V_route_cb.any_count == 0) 2074 return NULL; 2075 bzero((caddr_t)info, sizeof(*info)); 2076 m = rtsock_msg_mbuf(type, info); 2077 if (m != NULL) { 2078 ifan = mtod(m, struct if_announcemsghdr *); 2079 ifan->ifan_index = ifp->if_index; 2080 strlcpy(ifan->ifan_name, ifp->if_xname, 2081 sizeof(ifan->ifan_name)); 2082 ifan->ifan_what = what; 2083 } 2084 return m; 2085 } 2086 2087 /* 2088 * This is called to generate routing socket messages indicating 2089 * IEEE80211 wireless events. 2090 * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way. 2091 */ 2092 void 2093 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len) 2094 { 2095 struct mbuf *m; 2096 struct rt_addrinfo info; 2097 2098 m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info); 2099 if (m != NULL) { 2100 /* 2101 * Append the ieee80211 data. Try to stick it in the 2102 * mbuf containing the ifannounce msg; otherwise allocate 2103 * a new mbuf and append. 2104 * 2105 * NB: we assume m is a single mbuf. 2106 */ 2107 if (data_len > M_TRAILINGSPACE(m)) { 2108 struct mbuf *n = m_get(M_NOWAIT, MT_DATA); 2109 if (n == NULL) { 2110 m_freem(m); 2111 return; 2112 } 2113 bcopy(data, mtod(n, void *), data_len); 2114 n->m_len = data_len; 2115 m->m_next = n; 2116 } else if (data_len > 0) { 2117 bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len); 2118 m->m_len += data_len; 2119 } 2120 if (m->m_flags & M_PKTHDR) 2121 m->m_pkthdr.len += data_len; 2122 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len; 2123 rt_dispatch(m, AF_UNSPEC); 2124 } 2125 } 2126 2127 /* 2128 * This is called to generate routing socket messages indicating 2129 * network interface arrival and departure. 2130 */ 2131 static void 2132 rt_ifannouncemsg(struct ifnet *ifp, int what) 2133 { 2134 struct mbuf *m; 2135 struct rt_addrinfo info; 2136 2137 m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info); 2138 if (m != NULL) 2139 rt_dispatch(m, AF_UNSPEC); 2140 } 2141 2142 static void 2143 rt_dispatch(struct mbuf *m, sa_family_t saf) 2144 { 2145 2146 M_ASSERTPKTHDR(m); 2147 2148 m->m_rtsock_family = saf; 2149 if (V_loif) 2150 m->m_pkthdr.rcvif = V_loif; 2151 else { 2152 m_freem(m); 2153 return; 2154 } 2155 netisr_queue(NETISR_ROUTE, m); /* mbuf is free'd on failure. */ 2156 } 2157 2158 /* 2159 * Checks if rte can be exported w.r.t jails/vnets. 2160 * 2161 * Returns true if it can, false otherwise. 2162 */ 2163 static bool 2164 can_export_rte(struct ucred *td_ucred, bool rt_is_host, 2165 const struct sockaddr *rt_dst) 2166 { 2167 2168 if ((!rt_is_host) ? jailed_without_vnet(td_ucred) 2169 : prison_if(td_ucred, rt_dst) != 0) 2170 return (false); 2171 return (true); 2172 } 2173 2174 2175 /* 2176 * This is used in dumping the kernel table via sysctl(). 2177 */ 2178 static int 2179 sysctl_dumpentry(struct rtentry *rt, void *vw) 2180 { 2181 struct walkarg *w = vw; 2182 struct nhop_object *nh; 2183 2184 NET_EPOCH_ASSERT(); 2185 2186 export_rtaddrs(rt, w->dst, w->mask); 2187 if (!can_export_rte(w->w_req->td->td_ucred, rt_is_host(rt), w->dst)) 2188 return (0); 2189 nh = rt_get_raw_nhop(rt); 2190 #ifdef ROUTE_MPATH 2191 if (NH_IS_NHGRP(nh)) { 2192 const struct weightened_nhop *wn; 2193 uint32_t num_nhops; 2194 int error; 2195 wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops); 2196 for (int i = 0; i < num_nhops; i++) { 2197 error = sysctl_dumpnhop(rt, wn[i].nh, wn[i].weight, w); 2198 if (error != 0) 2199 return (error); 2200 } 2201 } else 2202 #endif 2203 sysctl_dumpnhop(rt, nh, rt->rt_weight, w); 2204 2205 return (0); 2206 } 2207 2208 2209 static int 2210 sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, uint32_t weight, 2211 struct walkarg *w) 2212 { 2213 struct rt_addrinfo info; 2214 int error = 0, size; 2215 uint32_t rtflags; 2216 2217 rtflags = nhop_get_rtflags(nh); 2218 2219 if (w->w_op == NET_RT_FLAGS && !(rtflags & w->w_arg)) 2220 return (0); 2221 2222 bzero((caddr_t)&info, sizeof(info)); 2223 info.rti_info[RTAX_DST] = w->dst; 2224 info.rti_info[RTAX_GATEWAY] = &nh->gw_sa; 2225 info.rti_info[RTAX_NETMASK] = (rtflags & RTF_HOST) ? NULL : w->mask; 2226 info.rti_info[RTAX_GENMASK] = 0; 2227 if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) { 2228 info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr; 2229 info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr; 2230 if (nh->nh_ifp->if_flags & IFF_POINTOPOINT) 2231 info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr; 2232 } 2233 if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0) 2234 return (error); 2235 if (w->w_req && w->w_tmem) { 2236 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem; 2237 2238 bzero(&rtm->rtm_index, 2239 sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index)); 2240 2241 /* 2242 * rte flags may consist of RTF_HOST (duplicated in nhop rtflags) 2243 * and RTF_UP (if entry is linked, which is always true here). 2244 * Given that, use nhop rtflags & add RTF_UP. 2245 */ 2246 rtm->rtm_flags = rtflags | RTF_UP; 2247 if (rtm->rtm_flags & RTF_GWFLAG_COMPAT) 2248 rtm->rtm_flags = RTF_GATEWAY | 2249 (rtm->rtm_flags & ~RTF_GWFLAG_COMPAT); 2250 rt_getmetrics(rt, nh, &rtm->rtm_rmx); 2251 rtm->rtm_rmx.rmx_weight = weight; 2252 rtm->rtm_index = nh->nh_ifp->if_index; 2253 rtm->rtm_addrs = info.rti_addrs; 2254 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size); 2255 return (error); 2256 } 2257 return (error); 2258 } 2259 2260 static int 2261 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd, 2262 struct rt_addrinfo *info, struct walkarg *w, int len) 2263 { 2264 struct if_msghdrl *ifm; 2265 struct if_data *ifd; 2266 2267 ifm = (struct if_msghdrl *)w->w_tmem; 2268 2269 #ifdef COMPAT_FREEBSD32 2270 if (w->w_req->flags & SCTL_MASK32) { 2271 struct if_msghdrl32 *ifm32; 2272 2273 ifm32 = (struct if_msghdrl32 *)ifm; 2274 ifm32->ifm_addrs = info->rti_addrs; 2275 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 2276 ifm32->ifm_index = ifp->if_index; 2277 ifm32->_ifm_spare1 = 0; 2278 ifm32->ifm_len = sizeof(*ifm32); 2279 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data); 2280 ifm32->_ifm_spare2 = 0; 2281 ifd = &ifm32->ifm_data; 2282 } else 2283 #endif 2284 { 2285 ifm->ifm_addrs = info->rti_addrs; 2286 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 2287 ifm->ifm_index = ifp->if_index; 2288 ifm->_ifm_spare1 = 0; 2289 ifm->ifm_len = sizeof(*ifm); 2290 ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data); 2291 ifm->_ifm_spare2 = 0; 2292 ifd = &ifm->ifm_data; 2293 } 2294 2295 memcpy(ifd, src_ifd, sizeof(*ifd)); 2296 2297 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len)); 2298 } 2299 2300 static int 2301 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd, 2302 struct rt_addrinfo *info, struct walkarg *w, int len) 2303 { 2304 struct if_msghdr *ifm; 2305 struct if_data *ifd; 2306 2307 ifm = (struct if_msghdr *)w->w_tmem; 2308 2309 #ifdef COMPAT_FREEBSD32 2310 if (w->w_req->flags & SCTL_MASK32) { 2311 struct if_msghdr32 *ifm32; 2312 2313 ifm32 = (struct if_msghdr32 *)ifm; 2314 ifm32->ifm_addrs = info->rti_addrs; 2315 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 2316 ifm32->ifm_index = ifp->if_index; 2317 ifm32->_ifm_spare1 = 0; 2318 ifd = &ifm32->ifm_data; 2319 } else 2320 #endif 2321 { 2322 ifm->ifm_addrs = info->rti_addrs; 2323 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 2324 ifm->ifm_index = ifp->if_index; 2325 ifm->_ifm_spare1 = 0; 2326 ifd = &ifm->ifm_data; 2327 } 2328 2329 memcpy(ifd, src_ifd, sizeof(*ifd)); 2330 2331 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len)); 2332 } 2333 2334 static int 2335 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info, 2336 struct walkarg *w, int len) 2337 { 2338 struct ifa_msghdrl *ifam; 2339 struct if_data *ifd; 2340 2341 ifam = (struct ifa_msghdrl *)w->w_tmem; 2342 2343 #ifdef COMPAT_FREEBSD32 2344 if (w->w_req->flags & SCTL_MASK32) { 2345 struct ifa_msghdrl32 *ifam32; 2346 2347 ifam32 = (struct ifa_msghdrl32 *)ifam; 2348 ifam32->ifam_addrs = info->rti_addrs; 2349 ifam32->ifam_flags = ifa->ifa_flags; 2350 ifam32->ifam_index = ifa->ifa_ifp->if_index; 2351 ifam32->_ifam_spare1 = 0; 2352 ifam32->ifam_len = sizeof(*ifam32); 2353 ifam32->ifam_data_off = 2354 offsetof(struct ifa_msghdrl32, ifam_data); 2355 ifam32->ifam_metric = ifa->ifa_ifp->if_metric; 2356 ifd = &ifam32->ifam_data; 2357 } else 2358 #endif 2359 { 2360 ifam->ifam_addrs = info->rti_addrs; 2361 ifam->ifam_flags = ifa->ifa_flags; 2362 ifam->ifam_index = ifa->ifa_ifp->if_index; 2363 ifam->_ifam_spare1 = 0; 2364 ifam->ifam_len = sizeof(*ifam); 2365 ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data); 2366 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 2367 ifd = &ifam->ifam_data; 2368 } 2369 2370 bzero(ifd, sizeof(*ifd)); 2371 ifd->ifi_datalen = sizeof(struct if_data); 2372 ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets); 2373 ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets); 2374 ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes); 2375 ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes); 2376 2377 /* Fixup if_data carp(4) vhid. */ 2378 if (carp_get_vhid_p != NULL) 2379 ifd->ifi_vhid = (*carp_get_vhid_p)(ifa); 2380 2381 return (SYSCTL_OUT(w->w_req, w->w_tmem, len)); 2382 } 2383 2384 static int 2385 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info, 2386 struct walkarg *w, int len) 2387 { 2388 struct ifa_msghdr *ifam; 2389 2390 ifam = (struct ifa_msghdr *)w->w_tmem; 2391 ifam->ifam_addrs = info->rti_addrs; 2392 ifam->ifam_flags = ifa->ifa_flags; 2393 ifam->ifam_index = ifa->ifa_ifp->if_index; 2394 ifam->_ifam_spare1 = 0; 2395 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 2396 2397 return (SYSCTL_OUT(w->w_req, w->w_tmem, len)); 2398 } 2399 2400 static int 2401 sysctl_iflist(int af, struct walkarg *w) 2402 { 2403 struct ifnet *ifp; 2404 struct ifaddr *ifa; 2405 struct if_data ifd; 2406 struct rt_addrinfo info; 2407 int len, error = 0; 2408 struct sockaddr_storage ss; 2409 2410 bzero((caddr_t)&info, sizeof(info)); 2411 bzero(&ifd, sizeof(ifd)); 2412 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2413 if (w->w_arg && w->w_arg != ifp->if_index) 2414 continue; 2415 if_data_copy(ifp, &ifd); 2416 ifa = ifp->if_addr; 2417 info.rti_info[RTAX_IFP] = ifa->ifa_addr; 2418 error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len); 2419 if (error != 0) 2420 goto done; 2421 info.rti_info[RTAX_IFP] = NULL; 2422 if (w->w_req && w->w_tmem) { 2423 if (w->w_op == NET_RT_IFLISTL) 2424 error = sysctl_iflist_ifml(ifp, &ifd, &info, w, 2425 len); 2426 else 2427 error = sysctl_iflist_ifm(ifp, &ifd, &info, w, 2428 len); 2429 if (error) 2430 goto done; 2431 } 2432 while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) { 2433 if (af && af != ifa->ifa_addr->sa_family) 2434 continue; 2435 if (prison_if(w->w_req->td->td_ucred, 2436 ifa->ifa_addr) != 0) 2437 continue; 2438 info.rti_info[RTAX_IFA] = ifa->ifa_addr; 2439 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask( 2440 ifa->ifa_addr, ifa->ifa_netmask, &ss); 2441 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr; 2442 error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len); 2443 if (error != 0) 2444 goto done; 2445 if (w->w_req && w->w_tmem) { 2446 if (w->w_op == NET_RT_IFLISTL) 2447 error = sysctl_iflist_ifaml(ifa, &info, 2448 w, len); 2449 else 2450 error = sysctl_iflist_ifam(ifa, &info, 2451 w, len); 2452 if (error) 2453 goto done; 2454 } 2455 } 2456 info.rti_info[RTAX_IFA] = NULL; 2457 info.rti_info[RTAX_NETMASK] = NULL; 2458 info.rti_info[RTAX_BRD] = NULL; 2459 } 2460 done: 2461 return (error); 2462 } 2463 2464 static int 2465 sysctl_ifmalist(int af, struct walkarg *w) 2466 { 2467 struct rt_addrinfo info; 2468 struct ifaddr *ifa; 2469 struct ifmultiaddr *ifma; 2470 struct ifnet *ifp; 2471 int error, len; 2472 2473 NET_EPOCH_ASSERT(); 2474 2475 error = 0; 2476 bzero((caddr_t)&info, sizeof(info)); 2477 2478 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2479 if (w->w_arg && w->w_arg != ifp->if_index) 2480 continue; 2481 ifa = ifp->if_addr; 2482 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL; 2483 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 2484 if (af && af != ifma->ifma_addr->sa_family) 2485 continue; 2486 if (prison_if(w->w_req->td->td_ucred, 2487 ifma->ifma_addr) != 0) 2488 continue; 2489 info.rti_info[RTAX_IFA] = ifma->ifma_addr; 2490 info.rti_info[RTAX_GATEWAY] = 2491 (ifma->ifma_addr->sa_family != AF_LINK) ? 2492 ifma->ifma_lladdr : NULL; 2493 error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len); 2494 if (error != 0) 2495 break; 2496 if (w->w_req && w->w_tmem) { 2497 struct ifma_msghdr *ifmam; 2498 2499 ifmam = (struct ifma_msghdr *)w->w_tmem; 2500 ifmam->ifmam_index = ifma->ifma_ifp->if_index; 2501 ifmam->ifmam_flags = 0; 2502 ifmam->ifmam_addrs = info.rti_addrs; 2503 ifmam->_ifmam_spare1 = 0; 2504 error = SYSCTL_OUT(w->w_req, w->w_tmem, len); 2505 if (error != 0) 2506 break; 2507 } 2508 } 2509 if (error != 0) 2510 break; 2511 } 2512 return (error); 2513 } 2514 2515 static void 2516 rtable_sysctl_dump(uint32_t fibnum, int family, struct walkarg *w) 2517 { 2518 union sockaddr_union sa_dst, sa_mask; 2519 2520 w->family = family; 2521 w->dst = (struct sockaddr *)&sa_dst; 2522 w->mask = (struct sockaddr *)&sa_mask; 2523 2524 init_sockaddrs_family(family, w->dst, w->mask); 2525 2526 rib_walk(fibnum, family, false, sysctl_dumpentry, w); 2527 } 2528 2529 static int 2530 sysctl_rtsock(SYSCTL_HANDLER_ARGS) 2531 { 2532 struct epoch_tracker et; 2533 int *name = (int *)arg1; 2534 u_int namelen = arg2; 2535 struct rib_head *rnh = NULL; /* silence compiler. */ 2536 int i, lim, error = EINVAL; 2537 int fib = 0; 2538 u_char af; 2539 struct walkarg w; 2540 2541 if (namelen < 3) 2542 return (EINVAL); 2543 2544 name++; 2545 namelen--; 2546 if (req->newptr) 2547 return (EPERM); 2548 if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP || name[1] == NET_RT_NHGRP) { 2549 if (namelen == 3) 2550 fib = req->td->td_proc->p_fibnum; 2551 else if (namelen == 4) 2552 fib = (name[3] == RT_ALL_FIBS) ? 2553 req->td->td_proc->p_fibnum : name[3]; 2554 else 2555 return ((namelen < 3) ? EISDIR : ENOTDIR); 2556 if (fib < 0 || fib >= rt_numfibs) 2557 return (EINVAL); 2558 } else if (namelen != 3) 2559 return ((namelen < 3) ? EISDIR : ENOTDIR); 2560 af = name[0]; 2561 if (af > AF_MAX) 2562 return (EINVAL); 2563 bzero(&w, sizeof(w)); 2564 w.w_op = name[1]; 2565 w.w_arg = name[2]; 2566 w.w_req = req; 2567 2568 error = sysctl_wire_old_buffer(req, 0); 2569 if (error) 2570 return (error); 2571 2572 /* 2573 * Allocate reply buffer in advance. 2574 * All rtsock messages has maximum length of u_short. 2575 */ 2576 w.w_tmemsize = 65536; 2577 w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK); 2578 2579 NET_EPOCH_ENTER(et); 2580 switch (w.w_op) { 2581 case NET_RT_DUMP: 2582 case NET_RT_FLAGS: 2583 if (af == 0) { /* dump all tables */ 2584 i = 1; 2585 lim = AF_MAX; 2586 } else /* dump only one table */ 2587 i = lim = af; 2588 2589 /* 2590 * take care of llinfo entries, the caller must 2591 * specify an AF 2592 */ 2593 if (w.w_op == NET_RT_FLAGS && 2594 (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) { 2595 if (af != 0) 2596 error = lltable_sysctl_dumparp(af, w.w_req); 2597 else 2598 error = EINVAL; 2599 break; 2600 } 2601 /* 2602 * take care of routing entries 2603 */ 2604 for (error = 0; error == 0 && i <= lim; i++) { 2605 rnh = rt_tables_get_rnh(fib, i); 2606 if (rnh != NULL) { 2607 rtable_sysctl_dump(fib, i, &w); 2608 } else if (af != 0) 2609 error = EAFNOSUPPORT; 2610 } 2611 break; 2612 case NET_RT_NHOP: 2613 case NET_RT_NHGRP: 2614 /* Allow dumping one specific af/fib at a time */ 2615 if (namelen < 4) { 2616 error = EINVAL; 2617 break; 2618 } 2619 fib = name[3]; 2620 if (fib < 0 || fib > rt_numfibs) { 2621 error = EINVAL; 2622 break; 2623 } 2624 rnh = rt_tables_get_rnh(fib, af); 2625 if (rnh == NULL) { 2626 error = EAFNOSUPPORT; 2627 break; 2628 } 2629 if (w.w_op == NET_RT_NHOP) 2630 error = nhops_dump_sysctl(rnh, w.w_req); 2631 else 2632 #ifdef ROUTE_MPATH 2633 error = nhgrp_dump_sysctl(rnh, w.w_req); 2634 #else 2635 error = ENOTSUP; 2636 #endif 2637 break; 2638 case NET_RT_IFLIST: 2639 case NET_RT_IFLISTL: 2640 error = sysctl_iflist(af, &w); 2641 break; 2642 2643 case NET_RT_IFMALIST: 2644 error = sysctl_ifmalist(af, &w); 2645 break; 2646 } 2647 NET_EPOCH_EXIT(et); 2648 2649 free(w.w_tmem, M_TEMP); 2650 return (error); 2651 } 2652 2653 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE, 2654 sysctl_rtsock, "Return route tables and interface/address lists"); 2655 2656 /* 2657 * Definitions of protocols supported in the ROUTE domain. 2658 */ 2659 2660 static struct domain routedomain; /* or at least forward */ 2661 2662 static struct pr_usrreqs route_usrreqs = { 2663 .pru_abort = rts_close, 2664 .pru_attach = rts_attach, 2665 .pru_detach = rts_detach, 2666 .pru_send = rts_send, 2667 .pru_shutdown = rts_shutdown, 2668 .pru_close = rts_close, 2669 }; 2670 2671 static struct protosw routesw[] = { 2672 { 2673 .pr_type = SOCK_RAW, 2674 .pr_domain = &routedomain, 2675 .pr_flags = PR_ATOMIC|PR_ADDR, 2676 .pr_usrreqs = &route_usrreqs 2677 } 2678 }; 2679 2680 static struct domain routedomain = { 2681 .dom_family = PF_ROUTE, 2682 .dom_name = "route", 2683 .dom_protosw = routesw, 2684 .dom_protoswNPROTOSW = &routesw[nitems(routesw)] 2685 }; 2686 2687 DOMAIN_SET(route); 2688