1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1988, 1991, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * @(#)rtsock.c 8.7 (Berkeley) 10/12/95 32 * $FreeBSD$ 33 */ 34 #include "opt_ddb.h" 35 #include "opt_route.h" 36 #include "opt_inet.h" 37 #include "opt_inet6.h" 38 39 #include <sys/param.h> 40 #include <sys/jail.h> 41 #include <sys/kernel.h> 42 #include <sys/domain.h> 43 #include <sys/lock.h> 44 #include <sys/malloc.h> 45 #include <sys/mbuf.h> 46 #include <sys/priv.h> 47 #include <sys/proc.h> 48 #include <sys/protosw.h> 49 #include <sys/rmlock.h> 50 #include <sys/rwlock.h> 51 #include <sys/signalvar.h> 52 #include <sys/socket.h> 53 #include <sys/socketvar.h> 54 #include <sys/sysctl.h> 55 #include <sys/systm.h> 56 57 #include <net/if.h> 58 #include <net/if_var.h> 59 #include <net/if_dl.h> 60 #include <net/if_llatbl.h> 61 #include <net/if_types.h> 62 #include <net/netisr.h> 63 #include <net/raw_cb.h> 64 #include <net/route.h> 65 #include <net/route/route_ctl.h> 66 #include <net/route/route_var.h> 67 #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/ip6_var.h> 74 #include <netinet6/scope6_var.h> 75 #endif 76 #include <net/route/nhop.h> 77 78 #ifdef COMPAT_FREEBSD32 79 #include <sys/mount.h> 80 #include <compat/freebsd32/freebsd32.h> 81 82 struct if_msghdr32 { 83 uint16_t ifm_msglen; 84 uint8_t ifm_version; 85 uint8_t ifm_type; 86 int32_t ifm_addrs; 87 int32_t ifm_flags; 88 uint16_t ifm_index; 89 uint16_t _ifm_spare1; 90 struct if_data ifm_data; 91 }; 92 93 struct if_msghdrl32 { 94 uint16_t ifm_msglen; 95 uint8_t ifm_version; 96 uint8_t ifm_type; 97 int32_t ifm_addrs; 98 int32_t ifm_flags; 99 uint16_t ifm_index; 100 uint16_t _ifm_spare1; 101 uint16_t ifm_len; 102 uint16_t ifm_data_off; 103 uint32_t _ifm_spare2; 104 struct if_data ifm_data; 105 }; 106 107 struct ifa_msghdrl32 { 108 uint16_t ifam_msglen; 109 uint8_t ifam_version; 110 uint8_t ifam_type; 111 int32_t ifam_addrs; 112 int32_t ifam_flags; 113 uint16_t ifam_index; 114 uint16_t _ifam_spare1; 115 uint16_t ifam_len; 116 uint16_t ifam_data_off; 117 int32_t ifam_metric; 118 struct if_data ifam_data; 119 }; 120 121 #define SA_SIZE32(sa) \ 122 ( (((struct sockaddr *)(sa))->sa_len == 0) ? \ 123 sizeof(int) : \ 124 1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) ) 125 126 #endif /* COMPAT_FREEBSD32 */ 127 128 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables"); 129 130 /* NB: these are not modified */ 131 static struct sockaddr route_src = { 2, PF_ROUTE, }; 132 static struct sockaddr sa_zero = { sizeof(sa_zero), AF_INET, }; 133 134 /* These are external hooks for CARP. */ 135 int (*carp_get_vhid_p)(struct ifaddr *); 136 137 /* 138 * Used by rtsock/raw_input callback code to decide whether to filter the update 139 * notification to a socket bound to a particular FIB. 140 */ 141 #define RTS_FILTER_FIB M_PROTO8 142 143 typedef struct { 144 int ip_count; /* attached w/ AF_INET */ 145 int ip6_count; /* attached w/ AF_INET6 */ 146 int any_count; /* total attached */ 147 } route_cb_t; 148 VNET_DEFINE_STATIC(route_cb_t, route_cb); 149 #define V_route_cb VNET(route_cb) 150 151 struct mtx rtsock_mtx; 152 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF); 153 154 #define RTSOCK_LOCK() mtx_lock(&rtsock_mtx) 155 #define RTSOCK_UNLOCK() mtx_unlock(&rtsock_mtx) 156 #define RTSOCK_LOCK_ASSERT() mtx_assert(&rtsock_mtx, MA_OWNED) 157 158 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, ""); 159 160 struct walkarg { 161 int family; 162 int w_tmemsize; 163 int w_op, w_arg; 164 caddr_t w_tmem; 165 struct sysctl_req *w_req; 166 struct sockaddr *dst; 167 struct sockaddr *mask; 168 }; 169 170 static void rts_input(struct mbuf *m); 171 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo); 172 static int rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, 173 struct walkarg *w, int *plen); 174 static int rt_xaddrs(caddr_t cp, caddr_t cplim, 175 struct rt_addrinfo *rtinfo); 176 static int sysctl_dumpentry(struct rtentry *rt, void *vw); 177 static int sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, 178 uint32_t weight, struct walkarg *w); 179 static int sysctl_iflist(int af, struct walkarg *w); 180 static int sysctl_ifmalist(int af, struct walkarg *w); 181 static int route_output(struct mbuf *m, struct socket *so, ...); 182 static void rt_getmetrics(const struct rtentry *rt, 183 const struct nhop_object *nh, struct rt_metrics *out); 184 static void rt_dispatch(struct mbuf *, sa_family_t); 185 static int handle_rtm_get(struct rt_addrinfo *info, u_int fibnum, 186 struct rt_msghdr *rtm, struct rib_cmd_info *rc); 187 static int update_rtm_from_rc(struct rt_addrinfo *info, 188 struct rt_msghdr **prtm, int alloc_len, 189 struct rib_cmd_info *rc, struct nhop_object *nh); 190 static void send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, 191 struct mbuf *m, sa_family_t saf, u_int fibnum, 192 int rtm_errno); 193 static bool can_export_rte(struct ucred *td_ucred, bool rt_is_host, 194 const struct sockaddr *rt_dst); 195 196 static struct netisr_handler rtsock_nh = { 197 .nh_name = "rtsock", 198 .nh_handler = rts_input, 199 .nh_proto = NETISR_ROUTE, 200 .nh_policy = NETISR_POLICY_SOURCE, 201 }; 202 203 static int 204 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS) 205 { 206 int error, qlimit; 207 208 netisr_getqlimit(&rtsock_nh, &qlimit); 209 error = sysctl_handle_int(oidp, &qlimit, 0, req); 210 if (error || !req->newptr) 211 return (error); 212 if (qlimit < 1) 213 return (EINVAL); 214 return (netisr_setqlimit(&rtsock_nh, qlimit)); 215 } 216 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, 217 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 218 0, 0, sysctl_route_netisr_maxqlen, "I", 219 "maximum routing socket dispatch queue length"); 220 221 static void 222 vnet_rts_init(void) 223 { 224 int tmp; 225 226 if (IS_DEFAULT_VNET(curvnet)) { 227 if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp)) 228 rtsock_nh.nh_qlimit = tmp; 229 netisr_register(&rtsock_nh); 230 } 231 #ifdef VIMAGE 232 else 233 netisr_register_vnet(&rtsock_nh); 234 #endif 235 } 236 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 237 vnet_rts_init, 0); 238 239 #ifdef VIMAGE 240 static void 241 vnet_rts_uninit(void) 242 { 243 244 netisr_unregister_vnet(&rtsock_nh); 245 } 246 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 247 vnet_rts_uninit, 0); 248 #endif 249 250 static int 251 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src, 252 struct rawcb *rp) 253 { 254 int fibnum; 255 256 KASSERT(m != NULL, ("%s: m is NULL", __func__)); 257 KASSERT(proto != NULL, ("%s: proto is NULL", __func__)); 258 KASSERT(rp != NULL, ("%s: rp is NULL", __func__)); 259 260 /* No filtering requested. */ 261 if ((m->m_flags & RTS_FILTER_FIB) == 0) 262 return (0); 263 264 /* Check if it is a rts and the fib matches the one of the socket. */ 265 fibnum = M_GETFIB(m); 266 if (proto->sp_family != PF_ROUTE || 267 rp->rcb_socket == NULL || 268 rp->rcb_socket->so_fibnum == fibnum) 269 return (0); 270 271 /* Filtering requested and no match, the socket shall be skipped. */ 272 return (1); 273 } 274 275 static void 276 rts_input(struct mbuf *m) 277 { 278 struct sockproto route_proto; 279 unsigned short *family; 280 struct m_tag *tag; 281 282 route_proto.sp_family = PF_ROUTE; 283 tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL); 284 if (tag != NULL) { 285 family = (unsigned short *)(tag + 1); 286 route_proto.sp_protocol = *family; 287 m_tag_delete(m, tag); 288 } else 289 route_proto.sp_protocol = 0; 290 291 raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb); 292 } 293 294 /* 295 * It really doesn't make any sense at all for this code to share much 296 * with raw_usrreq.c, since its functionality is so restricted. XXX 297 */ 298 static void 299 rts_abort(struct socket *so) 300 { 301 302 raw_usrreqs.pru_abort(so); 303 } 304 305 static void 306 rts_close(struct socket *so) 307 { 308 309 raw_usrreqs.pru_close(so); 310 } 311 312 /* pru_accept is EOPNOTSUPP */ 313 314 static int 315 rts_attach(struct socket *so, int proto, struct thread *td) 316 { 317 struct rawcb *rp; 318 int error; 319 320 KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL")); 321 322 /* XXX */ 323 rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO); 324 325 so->so_pcb = (caddr_t)rp; 326 so->so_fibnum = td->td_proc->p_fibnum; 327 error = raw_attach(so, proto); 328 rp = sotorawcb(so); 329 if (error) { 330 so->so_pcb = NULL; 331 free(rp, M_PCB); 332 return error; 333 } 334 RTSOCK_LOCK(); 335 switch(rp->rcb_proto.sp_protocol) { 336 case AF_INET: 337 V_route_cb.ip_count++; 338 break; 339 case AF_INET6: 340 V_route_cb.ip6_count++; 341 break; 342 } 343 V_route_cb.any_count++; 344 RTSOCK_UNLOCK(); 345 soisconnected(so); 346 so->so_options |= SO_USELOOPBACK; 347 return 0; 348 } 349 350 static int 351 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 352 { 353 354 return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */ 355 } 356 357 static int 358 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 359 { 360 361 return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */ 362 } 363 364 /* pru_connect2 is EOPNOTSUPP */ 365 /* pru_control is EOPNOTSUPP */ 366 367 static void 368 rts_detach(struct socket *so) 369 { 370 struct rawcb *rp = sotorawcb(so); 371 372 KASSERT(rp != NULL, ("rts_detach: rp == NULL")); 373 374 RTSOCK_LOCK(); 375 switch(rp->rcb_proto.sp_protocol) { 376 case AF_INET: 377 V_route_cb.ip_count--; 378 break; 379 case AF_INET6: 380 V_route_cb.ip6_count--; 381 break; 382 } 383 V_route_cb.any_count--; 384 RTSOCK_UNLOCK(); 385 raw_usrreqs.pru_detach(so); 386 } 387 388 static int 389 rts_disconnect(struct socket *so) 390 { 391 392 return (raw_usrreqs.pru_disconnect(so)); 393 } 394 395 /* pru_listen is EOPNOTSUPP */ 396 397 static int 398 rts_peeraddr(struct socket *so, struct sockaddr **nam) 399 { 400 401 return (raw_usrreqs.pru_peeraddr(so, nam)); 402 } 403 404 /* pru_rcvd is EOPNOTSUPP */ 405 /* pru_rcvoob is EOPNOTSUPP */ 406 407 static int 408 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 409 struct mbuf *control, struct thread *td) 410 { 411 412 return (raw_usrreqs.pru_send(so, flags, m, nam, control, td)); 413 } 414 415 /* pru_sense is null */ 416 417 static int 418 rts_shutdown(struct socket *so) 419 { 420 421 return (raw_usrreqs.pru_shutdown(so)); 422 } 423 424 static int 425 rts_sockaddr(struct socket *so, struct sockaddr **nam) 426 { 427 428 return (raw_usrreqs.pru_sockaddr(so, nam)); 429 } 430 431 static struct pr_usrreqs route_usrreqs = { 432 .pru_abort = rts_abort, 433 .pru_attach = rts_attach, 434 .pru_bind = rts_bind, 435 .pru_connect = rts_connect, 436 .pru_detach = rts_detach, 437 .pru_disconnect = rts_disconnect, 438 .pru_peeraddr = rts_peeraddr, 439 .pru_send = rts_send, 440 .pru_shutdown = rts_shutdown, 441 .pru_sockaddr = rts_sockaddr, 442 .pru_close = rts_close, 443 }; 444 445 #ifndef _SOCKADDR_UNION_DEFINED 446 #define _SOCKADDR_UNION_DEFINED 447 /* 448 * The union of all possible address formats we handle. 449 */ 450 union sockaddr_union { 451 struct sockaddr sa; 452 struct sockaddr_in sin; 453 struct sockaddr_in6 sin6; 454 }; 455 #endif /* _SOCKADDR_UNION_DEFINED */ 456 457 static int 458 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp, 459 struct nhop_object *nh, union sockaddr_union *saun, struct ucred *cred) 460 { 461 #if defined(INET) || defined(INET6) 462 struct epoch_tracker et; 463 #endif 464 465 /* First, see if the returned address is part of the jail. */ 466 if (prison_if(cred, nh->nh_ifa->ifa_addr) == 0) { 467 info->rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr; 468 return (0); 469 } 470 471 switch (info->rti_info[RTAX_DST]->sa_family) { 472 #ifdef INET 473 case AF_INET: 474 { 475 struct in_addr ia; 476 struct ifaddr *ifa; 477 int found; 478 479 found = 0; 480 /* 481 * Try to find an address on the given outgoing interface 482 * that belongs to the jail. 483 */ 484 NET_EPOCH_ENTER(et); 485 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 486 struct sockaddr *sa; 487 sa = ifa->ifa_addr; 488 if (sa->sa_family != AF_INET) 489 continue; 490 ia = ((struct sockaddr_in *)sa)->sin_addr; 491 if (prison_check_ip4(cred, &ia) == 0) { 492 found = 1; 493 break; 494 } 495 } 496 NET_EPOCH_EXIT(et); 497 if (!found) { 498 /* 499 * As a last resort return the 'default' jail address. 500 */ 501 ia = ((struct sockaddr_in *)nh->nh_ifa->ifa_addr)-> 502 sin_addr; 503 if (prison_get_ip4(cred, &ia) != 0) 504 return (ESRCH); 505 } 506 bzero(&saun->sin, sizeof(struct sockaddr_in)); 507 saun->sin.sin_len = sizeof(struct sockaddr_in); 508 saun->sin.sin_family = AF_INET; 509 saun->sin.sin_addr.s_addr = ia.s_addr; 510 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin; 511 break; 512 } 513 #endif 514 #ifdef INET6 515 case AF_INET6: 516 { 517 struct in6_addr ia6; 518 struct ifaddr *ifa; 519 int found; 520 521 found = 0; 522 /* 523 * Try to find an address on the given outgoing interface 524 * that belongs to the jail. 525 */ 526 NET_EPOCH_ENTER(et); 527 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 528 struct sockaddr *sa; 529 sa = ifa->ifa_addr; 530 if (sa->sa_family != AF_INET6) 531 continue; 532 bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr, 533 &ia6, sizeof(struct in6_addr)); 534 if (prison_check_ip6(cred, &ia6) == 0) { 535 found = 1; 536 break; 537 } 538 } 539 NET_EPOCH_EXIT(et); 540 if (!found) { 541 /* 542 * As a last resort return the 'default' jail address. 543 */ 544 ia6 = ((struct sockaddr_in6 *)nh->nh_ifa->ifa_addr)-> 545 sin6_addr; 546 if (prison_get_ip6(cred, &ia6) != 0) 547 return (ESRCH); 548 } 549 bzero(&saun->sin6, sizeof(struct sockaddr_in6)); 550 saun->sin6.sin6_len = sizeof(struct sockaddr_in6); 551 saun->sin6.sin6_family = AF_INET6; 552 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr)); 553 if (sa6_recoverscope(&saun->sin6) != 0) 554 return (ESRCH); 555 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6; 556 break; 557 } 558 #endif 559 default: 560 return (ESRCH); 561 } 562 return (0); 563 } 564 565 static int 566 fill_blackholeinfo(struct rt_addrinfo *info, union sockaddr_union *saun) 567 { 568 struct ifaddr *ifa; 569 sa_family_t saf; 570 571 if (V_loif == NULL) { 572 printf("Unable to add blackhole/reject nhop without loopback"); 573 return (ENOTSUP); 574 } 575 info->rti_ifp = V_loif; 576 577 saf = info->rti_info[RTAX_DST]->sa_family; 578 579 CK_STAILQ_FOREACH(ifa, &info->rti_ifp->if_addrhead, ifa_link) { 580 if (ifa->ifa_addr->sa_family == saf) { 581 info->rti_ifa = ifa; 582 break; 583 } 584 } 585 if (info->rti_ifa == NULL) 586 return (ENOTSUP); 587 588 bzero(saun, sizeof(union sockaddr_union)); 589 switch (saf) { 590 case AF_INET: 591 saun->sin.sin_family = AF_INET; 592 saun->sin.sin_len = sizeof(struct sockaddr_in); 593 saun->sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK); 594 break; 595 case AF_INET6: 596 saun->sin6.sin6_family = AF_INET6; 597 saun->sin6.sin6_len = sizeof(struct sockaddr_in6); 598 saun->sin6.sin6_addr = in6addr_loopback; 599 break; 600 default: 601 return (ENOTSUP); 602 } 603 info->rti_info[RTAX_GATEWAY] = &saun->sa; 604 info->rti_flags |= RTF_GATEWAY; 605 606 return (0); 607 } 608 609 /* 610 * Fills in @info based on userland-provided @rtm message. 611 * 612 * Returns 0 on success. 613 */ 614 static int 615 fill_addrinfo(struct rt_msghdr *rtm, int len, u_int fibnum, struct rt_addrinfo *info) 616 { 617 int error; 618 sa_family_t saf; 619 620 rtm->rtm_pid = curproc->p_pid; 621 info->rti_addrs = rtm->rtm_addrs; 622 623 info->rti_mflags = rtm->rtm_inits; 624 info->rti_rmx = &rtm->rtm_rmx; 625 626 /* 627 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6 628 * link-local address because rtrequest requires addresses with 629 * embedded scope id. 630 */ 631 if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info)) 632 return (EINVAL); 633 634 if (rtm->rtm_flags & RTF_RNH_LOCKED) 635 return (EINVAL); 636 info->rti_flags = rtm->rtm_flags; 637 if (info->rti_info[RTAX_DST] == NULL || 638 info->rti_info[RTAX_DST]->sa_family >= AF_MAX || 639 (info->rti_info[RTAX_GATEWAY] != NULL && 640 info->rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) 641 return (EINVAL); 642 saf = info->rti_info[RTAX_DST]->sa_family; 643 /* 644 * Verify that the caller has the appropriate privilege; RTM_GET 645 * is the only operation the non-superuser is allowed. 646 */ 647 if (rtm->rtm_type != RTM_GET) { 648 error = priv_check(curthread, PRIV_NET_ROUTE); 649 if (error != 0) 650 return (error); 651 } 652 653 /* 654 * The given gateway address may be an interface address. 655 * For example, issuing a "route change" command on a route 656 * entry that was created from a tunnel, and the gateway 657 * address given is the local end point. In this case the 658 * RTF_GATEWAY flag must be cleared or the destination will 659 * not be reachable even though there is no error message. 660 */ 661 if (info->rti_info[RTAX_GATEWAY] != NULL && 662 info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) { 663 struct rt_addrinfo ginfo; 664 struct sockaddr *gdst; 665 struct sockaddr_storage ss; 666 667 bzero(&ginfo, sizeof(ginfo)); 668 bzero(&ss, sizeof(ss)); 669 ss.ss_len = sizeof(ss); 670 671 ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss; 672 gdst = info->rti_info[RTAX_GATEWAY]; 673 674 /* 675 * A host route through the loopback interface is 676 * installed for each interface adddress. In pre 8.0 677 * releases the interface address of a PPP link type 678 * is not reachable locally. This behavior is fixed as 679 * part of the new L2/L3 redesign and rewrite work. The 680 * signature of this interface address route is the 681 * AF_LINK sa_family type of the gateway, and the 682 * rt_ifp has the IFF_LOOPBACK flag set. 683 */ 684 if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) { 685 if (ss.ss_family == AF_LINK && 686 ginfo.rti_ifp->if_flags & IFF_LOOPBACK) { 687 info->rti_flags &= ~RTF_GATEWAY; 688 info->rti_flags |= RTF_GWFLAG_COMPAT; 689 } 690 rib_free_info(&ginfo); 691 } 692 } 693 694 return (0); 695 } 696 697 static struct nhop_object * 698 select_nhop(struct nhop_object *nh, const struct sockaddr *gw) 699 { 700 if (!NH_IS_NHGRP(nh)) 701 return (nh); 702 #ifdef ROUTE_MPATH 703 struct weightened_nhop *wn; 704 uint32_t num_nhops; 705 wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops); 706 if (gw == NULL) 707 return (wn[0].nh); 708 for (int i = 0; i < num_nhops; i++) { 709 if (match_nhop_gw(wn[i].nh, gw)) 710 return (wn[i].nh); 711 } 712 #endif 713 return (NULL); 714 } 715 716 /* 717 * Handles RTM_GET message from routing socket, returning matching rt. 718 * 719 * Returns: 720 * 0 on success, with locked and referenced matching rt in @rt_nrt 721 * errno of failure 722 */ 723 static int 724 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum, 725 struct rt_msghdr *rtm, struct rib_cmd_info *rc) 726 { 727 RIB_RLOCK_TRACKER; 728 struct rib_head *rnh; 729 struct nhop_object *nh; 730 sa_family_t saf; 731 732 saf = info->rti_info[RTAX_DST]->sa_family; 733 734 rnh = rt_tables_get_rnh(fibnum, saf); 735 if (rnh == NULL) 736 return (EAFNOSUPPORT); 737 738 RIB_RLOCK(rnh); 739 740 if (info->rti_info[RTAX_NETMASK] == NULL) { 741 /* 742 * Provide longest prefix match for 743 * address lookup (no mask). 744 * 'route -n get addr' 745 */ 746 rc->rc_rt = (struct rtentry *) rnh->rnh_matchaddr( 747 info->rti_info[RTAX_DST], &rnh->head); 748 } else 749 rc->rc_rt = (struct rtentry *) rnh->rnh_lookup( 750 info->rti_info[RTAX_DST], 751 info->rti_info[RTAX_NETMASK], &rnh->head); 752 753 if (rc->rc_rt == NULL) { 754 RIB_RUNLOCK(rnh); 755 return (ESRCH); 756 } 757 758 nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]); 759 if (nh == NULL) { 760 RIB_RUNLOCK(rnh); 761 return (ESRCH); 762 } 763 /* 764 * If performing proxied L2 entry insertion, and 765 * the actual PPP host entry is found, perform 766 * another search to retrieve the prefix route of 767 * the local end point of the PPP link. 768 * TODO: move this logic to userland. 769 */ 770 if (rtm->rtm_flags & RTF_ANNOUNCE) { 771 struct sockaddr laddr; 772 773 if (nh->nh_ifp != NULL && 774 nh->nh_ifp->if_type == IFT_PROPVIRTUAL) { 775 struct ifaddr *ifa; 776 777 ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1, 778 RT_ALL_FIBS); 779 if (ifa != NULL) 780 rt_maskedcopy(ifa->ifa_addr, 781 &laddr, 782 ifa->ifa_netmask); 783 } else 784 rt_maskedcopy(nh->nh_ifa->ifa_addr, 785 &laddr, 786 nh->nh_ifa->ifa_netmask); 787 /* 788 * refactor rt and no lock operation necessary 789 */ 790 rc->rc_rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr, 791 &rnh->head); 792 if (rc->rc_rt == NULL) { 793 RIB_RUNLOCK(rnh); 794 return (ESRCH); 795 } 796 nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]); 797 if (nh == NULL) { 798 RIB_RUNLOCK(rnh); 799 return (ESRCH); 800 } 801 } 802 rc->rc_nh_new = nh; 803 rc->rc_nh_weight = rc->rc_rt->rt_weight; 804 RIB_RUNLOCK(rnh); 805 806 return (0); 807 } 808 809 static void 810 init_sockaddrs_family(int family, struct sockaddr *dst, struct sockaddr *mask) 811 { 812 #ifdef INET 813 if (family == AF_INET) { 814 struct sockaddr_in *dst4 = (struct sockaddr_in *)dst; 815 struct sockaddr_in *mask4 = (struct sockaddr_in *)mask; 816 817 bzero(dst4, sizeof(struct sockaddr_in)); 818 bzero(mask4, sizeof(struct sockaddr_in)); 819 820 dst4->sin_family = AF_INET; 821 dst4->sin_len = sizeof(struct sockaddr_in); 822 mask4->sin_family = AF_INET; 823 mask4->sin_len = sizeof(struct sockaddr_in); 824 } 825 #endif 826 #ifdef INET6 827 if (family == AF_INET6) { 828 struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst; 829 struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask; 830 831 bzero(dst6, sizeof(struct sockaddr_in6)); 832 bzero(mask6, sizeof(struct sockaddr_in6)); 833 834 dst6->sin6_family = AF_INET6; 835 dst6->sin6_len = sizeof(struct sockaddr_in6); 836 mask6->sin6_family = AF_INET6; 837 mask6->sin6_len = sizeof(struct sockaddr_in6); 838 } 839 #endif 840 } 841 842 static void 843 export_rtaddrs(const struct rtentry *rt, struct sockaddr *dst, 844 struct sockaddr *mask) 845 { 846 #ifdef INET 847 if (dst->sa_family == AF_INET) { 848 struct sockaddr_in *dst4 = (struct sockaddr_in *)dst; 849 struct sockaddr_in *mask4 = (struct sockaddr_in *)mask; 850 uint32_t scopeid = 0; 851 rt_get_inet_prefix_pmask(rt, &dst4->sin_addr, &mask4->sin_addr, 852 &scopeid); 853 return; 854 } 855 #endif 856 #ifdef INET6 857 if (dst->sa_family == AF_INET6) { 858 struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst; 859 struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask; 860 uint32_t scopeid = 0; 861 rt_get_inet6_prefix_pmask(rt, &dst6->sin6_addr, 862 &mask6->sin6_addr, &scopeid); 863 dst6->sin6_scope_id = scopeid; 864 return; 865 } 866 #endif 867 } 868 869 870 /* 871 * Update sockaddrs, flags, etc in @prtm based on @rc data. 872 * rtm can be reallocated. 873 * 874 * Returns 0 on success, along with pointer to (potentially reallocated) 875 * rtm. 876 * 877 */ 878 static int 879 update_rtm_from_rc(struct rt_addrinfo *info, struct rt_msghdr **prtm, 880 int alloc_len, struct rib_cmd_info *rc, struct nhop_object *nh) 881 { 882 struct walkarg w; 883 union sockaddr_union saun; 884 struct rt_msghdr *rtm, *orig_rtm = NULL; 885 struct ifnet *ifp; 886 int error, len; 887 888 rtm = *prtm; 889 union sockaddr_union sa_dst, sa_mask; 890 int family = info->rti_info[RTAX_DST]->sa_family; 891 init_sockaddrs_family(family, &sa_dst.sa, &sa_mask.sa); 892 export_rtaddrs(rc->rc_rt, &sa_dst.sa, &sa_mask.sa); 893 894 info->rti_info[RTAX_DST] = &sa_dst.sa; 895 info->rti_info[RTAX_NETMASK] = rt_is_host(rc->rc_rt) ? NULL : &sa_mask.sa; 896 info->rti_info[RTAX_GATEWAY] = &nh->gw_sa; 897 info->rti_info[RTAX_GENMASK] = 0; 898 ifp = nh->nh_ifp; 899 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) { 900 if (ifp) { 901 info->rti_info[RTAX_IFP] = 902 ifp->if_addr->ifa_addr; 903 error = rtm_get_jailed(info, ifp, nh, 904 &saun, curthread->td_ucred); 905 if (error != 0) 906 return (error); 907 if (ifp->if_flags & IFF_POINTOPOINT) 908 info->rti_info[RTAX_BRD] = 909 nh->nh_ifa->ifa_dstaddr; 910 rtm->rtm_index = ifp->if_index; 911 } else { 912 info->rti_info[RTAX_IFP] = NULL; 913 info->rti_info[RTAX_IFA] = NULL; 914 } 915 } else if (ifp != NULL) 916 rtm->rtm_index = ifp->if_index; 917 918 /* Check if we need to realloc storage */ 919 rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len); 920 if (len > alloc_len) { 921 struct rt_msghdr *tmp_rtm; 922 923 tmp_rtm = malloc(len, M_TEMP, M_NOWAIT); 924 if (tmp_rtm == NULL) 925 return (ENOBUFS); 926 bcopy(rtm, tmp_rtm, rtm->rtm_msglen); 927 orig_rtm = rtm; 928 rtm = tmp_rtm; 929 alloc_len = len; 930 931 /* 932 * Delay freeing original rtm as info contains 933 * data referencing it. 934 */ 935 } 936 937 w.w_tmem = (caddr_t)rtm; 938 w.w_tmemsize = alloc_len; 939 rtsock_msg_buffer(rtm->rtm_type, info, &w, &len); 940 941 rtm->rtm_flags = rc->rc_rt->rte_flags | nhop_get_rtflags(nh); 942 if (rtm->rtm_flags & RTF_GWFLAG_COMPAT) 943 rtm->rtm_flags = RTF_GATEWAY | 944 (rtm->rtm_flags & ~RTF_GWFLAG_COMPAT); 945 rt_getmetrics(rc->rc_rt, nh, &rtm->rtm_rmx); 946 rtm->rtm_rmx.rmx_weight = rc->rc_nh_weight; 947 rtm->rtm_addrs = info->rti_addrs; 948 949 if (orig_rtm != NULL) 950 free(orig_rtm, M_TEMP); 951 *prtm = rtm; 952 953 return (0); 954 } 955 956 #ifdef ROUTE_MPATH 957 static void 958 save_del_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_DELETE) 963 *rc_new = *rc; 964 } 965 966 static void 967 save_add_notification(struct rib_cmd_info *rc, void *_cbdata) 968 { 969 struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata; 970 971 if (rc->rc_cmd == RTM_ADD) 972 *rc_new = *rc; 973 } 974 #endif 975 976 /*ARGSUSED*/ 977 static int 978 route_output(struct mbuf *m, struct socket *so, ...) 979 { 980 struct rt_msghdr *rtm = NULL; 981 struct rtentry *rt = NULL; 982 struct rt_addrinfo info; 983 struct epoch_tracker et; 984 #ifdef INET6 985 struct sockaddr_storage ss; 986 struct sockaddr_in6 *sin6; 987 int i, rti_need_deembed = 0; 988 #endif 989 int alloc_len = 0, len, error = 0, fibnum; 990 sa_family_t saf = AF_UNSPEC; 991 struct rib_cmd_info rc; 992 struct nhop_object *nh; 993 994 fibnum = so->so_fibnum; 995 #define senderr(e) { error = e; goto flush;} 996 if (m == NULL || ((m->m_len < sizeof(long)) && 997 (m = m_pullup(m, sizeof(long))) == NULL)) 998 return (ENOBUFS); 999 if ((m->m_flags & M_PKTHDR) == 0) 1000 panic("route_output"); 1001 NET_EPOCH_ENTER(et); 1002 len = m->m_pkthdr.len; 1003 if (len < sizeof(*rtm) || 1004 len != mtod(m, struct rt_msghdr *)->rtm_msglen) 1005 senderr(EINVAL); 1006 1007 /* 1008 * Most of current messages are in range 200-240 bytes, 1009 * minimize possible re-allocation on reply using larger size 1010 * buffer aligned on 1k boundaty. 1011 */ 1012 alloc_len = roundup2(len, 1024); 1013 if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL) 1014 senderr(ENOBUFS); 1015 1016 m_copydata(m, 0, len, (caddr_t)rtm); 1017 bzero(&info, sizeof(info)); 1018 nh = NULL; 1019 1020 if (rtm->rtm_version != RTM_VERSION) { 1021 /* Do not touch message since format is unknown */ 1022 free(rtm, M_TEMP); 1023 rtm = NULL; 1024 senderr(EPROTONOSUPPORT); 1025 } 1026 1027 /* 1028 * Starting from here, it is possible 1029 * to alter original message and insert 1030 * caller PID and error value. 1031 */ 1032 1033 if ((error = fill_addrinfo(rtm, len, fibnum, &info)) != 0) { 1034 senderr(error); 1035 } 1036 1037 saf = info.rti_info[RTAX_DST]->sa_family; 1038 1039 /* support for new ARP code */ 1040 if (rtm->rtm_flags & RTF_LLDATA) { 1041 error = lla_rt_output(rtm, &info); 1042 #ifdef INET6 1043 if (error == 0) 1044 rti_need_deembed = 1; 1045 #endif 1046 goto flush; 1047 } 1048 1049 union sockaddr_union gw_saun; 1050 int blackhole_flags = rtm->rtm_flags & (RTF_BLACKHOLE|RTF_REJECT); 1051 if (blackhole_flags != 0) { 1052 if (blackhole_flags != (RTF_BLACKHOLE | RTF_REJECT)) 1053 error = fill_blackholeinfo(&info, &gw_saun); 1054 else 1055 error = EINVAL; 1056 if (error != 0) 1057 senderr(error); 1058 /* TODO: rebuild rtm from scratch */ 1059 } 1060 1061 switch (rtm->rtm_type) { 1062 case RTM_ADD: 1063 case RTM_CHANGE: 1064 if (rtm->rtm_type == RTM_ADD) { 1065 if (info.rti_info[RTAX_GATEWAY] == NULL) 1066 senderr(EINVAL); 1067 } 1068 error = rib_action(fibnum, rtm->rtm_type, &info, &rc); 1069 if (error == 0) { 1070 #ifdef INET6 1071 rti_need_deembed = 1; 1072 #endif 1073 #ifdef ROUTE_MPATH 1074 if (NH_IS_NHGRP(rc.rc_nh_new) || 1075 (rc.rc_nh_old && NH_IS_NHGRP(rc.rc_nh_old))) { 1076 struct rib_cmd_info rc_simple = {}; 1077 rib_decompose_notification(&rc, 1078 save_add_notification, (void *)&rc_simple); 1079 rc = rc_simple; 1080 } 1081 #endif 1082 nh = rc.rc_nh_new; 1083 rtm->rtm_index = nh->nh_ifp->if_index; 1084 rtm->rtm_flags = rc.rc_rt->rte_flags | nhop_get_rtflags(nh); 1085 } 1086 break; 1087 1088 case RTM_DELETE: 1089 error = rib_action(fibnum, RTM_DELETE, &info, &rc); 1090 if (error == 0) { 1091 #ifdef ROUTE_MPATH 1092 if (NH_IS_NHGRP(rc.rc_nh_old) || 1093 (rc.rc_nh_new && NH_IS_NHGRP(rc.rc_nh_new))) { 1094 struct rib_cmd_info rc_simple = {}; 1095 rib_decompose_notification(&rc, 1096 save_del_notification, (void *)&rc_simple); 1097 rc = rc_simple; 1098 } 1099 #endif 1100 nh = rc.rc_nh_old; 1101 goto report; 1102 } 1103 #ifdef INET6 1104 /* rt_msg2() will not be used when RTM_DELETE fails. */ 1105 rti_need_deembed = 1; 1106 #endif 1107 break; 1108 1109 case RTM_GET: 1110 error = handle_rtm_get(&info, fibnum, rtm, &rc); 1111 if (error != 0) 1112 senderr(error); 1113 nh = rc.rc_nh_new; 1114 1115 report: 1116 if (!can_export_rte(curthread->td_ucred, 1117 info.rti_info[RTAX_NETMASK] == NULL, 1118 info.rti_info[RTAX_DST])) { 1119 senderr(ESRCH); 1120 } 1121 1122 error = update_rtm_from_rc(&info, &rtm, alloc_len, &rc, nh); 1123 /* 1124 * Note that some sockaddr pointers may have changed to 1125 * point to memory outsize @rtm. Some may be pointing 1126 * to the on-stack variables. 1127 * Given that, any pointer in @info CANNOT BE USED. 1128 */ 1129 1130 /* 1131 * scopeid deembedding has been performed while 1132 * writing updated rtm in rtsock_msg_buffer(). 1133 * With that in mind, skip deembedding procedure below. 1134 */ 1135 #ifdef INET6 1136 rti_need_deembed = 0; 1137 #endif 1138 if (error != 0) 1139 senderr(error); 1140 break; 1141 1142 default: 1143 senderr(EOPNOTSUPP); 1144 } 1145 1146 flush: 1147 NET_EPOCH_EXIT(et); 1148 rt = NULL; 1149 1150 #ifdef INET6 1151 if (rtm != NULL) { 1152 if (rti_need_deembed) { 1153 /* sin6_scope_id is recovered before sending rtm. */ 1154 sin6 = (struct sockaddr_in6 *)&ss; 1155 for (i = 0; i < RTAX_MAX; i++) { 1156 if (info.rti_info[i] == NULL) 1157 continue; 1158 if (info.rti_info[i]->sa_family != AF_INET6) 1159 continue; 1160 bcopy(info.rti_info[i], sin6, sizeof(*sin6)); 1161 if (sa6_recoverscope(sin6) == 0) 1162 bcopy(sin6, info.rti_info[i], 1163 sizeof(*sin6)); 1164 } 1165 } 1166 } 1167 #endif 1168 send_rtm_reply(so, rtm, m, saf, fibnum, error); 1169 1170 return (error); 1171 } 1172 1173 /* 1174 * Sends the prepared reply message in @rtm to all rtsock clients. 1175 * Frees @m and @rtm. 1176 * 1177 */ 1178 static void 1179 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m, 1180 sa_family_t saf, u_int fibnum, int rtm_errno) 1181 { 1182 struct rawcb *rp = NULL; 1183 1184 /* 1185 * Check to see if we don't want our own messages. 1186 */ 1187 if ((so->so_options & SO_USELOOPBACK) == 0) { 1188 if (V_route_cb.any_count <= 1) { 1189 if (rtm != NULL) 1190 free(rtm, M_TEMP); 1191 m_freem(m); 1192 return; 1193 } 1194 /* There is another listener, so construct message */ 1195 rp = sotorawcb(so); 1196 } 1197 1198 if (rtm != NULL) { 1199 if (rtm_errno!= 0) 1200 rtm->rtm_errno = rtm_errno; 1201 else 1202 rtm->rtm_flags |= RTF_DONE; 1203 1204 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm); 1205 if (m->m_pkthdr.len < rtm->rtm_msglen) { 1206 m_freem(m); 1207 m = NULL; 1208 } else if (m->m_pkthdr.len > rtm->rtm_msglen) 1209 m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len); 1210 1211 free(rtm, M_TEMP); 1212 } 1213 if (m != NULL) { 1214 M_SETFIB(m, fibnum); 1215 m->m_flags |= RTS_FILTER_FIB; 1216 if (rp) { 1217 /* 1218 * XXX insure we don't get a copy by 1219 * invalidating our protocol 1220 */ 1221 unsigned short family = rp->rcb_proto.sp_family; 1222 rp->rcb_proto.sp_family = 0; 1223 rt_dispatch(m, saf); 1224 rp->rcb_proto.sp_family = family; 1225 } else 1226 rt_dispatch(m, saf); 1227 } 1228 } 1229 1230 static void 1231 rt_getmetrics(const struct rtentry *rt, const struct nhop_object *nh, 1232 struct rt_metrics *out) 1233 { 1234 1235 bzero(out, sizeof(*out)); 1236 out->rmx_mtu = nh->nh_mtu; 1237 out->rmx_weight = rt->rt_weight; 1238 out->rmx_nhidx = nhop_get_idx(nh); 1239 /* Kernel -> userland timebase conversion. */ 1240 out->rmx_expire = rt->rt_expire ? 1241 rt->rt_expire - time_uptime + time_second : 0; 1242 } 1243 1244 /* 1245 * Extract the addresses of the passed sockaddrs. 1246 * Do a little sanity checking so as to avoid bad memory references. 1247 * This data is derived straight from userland. 1248 */ 1249 static int 1250 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo) 1251 { 1252 struct sockaddr *sa; 1253 int i; 1254 1255 for (i = 0; i < RTAX_MAX && cp < cplim; i++) { 1256 if ((rtinfo->rti_addrs & (1 << i)) == 0) 1257 continue; 1258 sa = (struct sockaddr *)cp; 1259 /* 1260 * It won't fit. 1261 */ 1262 if (cp + sa->sa_len > cplim) 1263 return (EINVAL); 1264 /* 1265 * there are no more.. quit now 1266 * If there are more bits, they are in error. 1267 * I've seen this. route(1) can evidently generate these. 1268 * This causes kernel to core dump. 1269 * for compatibility, If we see this, point to a safe address. 1270 */ 1271 if (sa->sa_len == 0) { 1272 rtinfo->rti_info[i] = &sa_zero; 1273 return (0); /* should be EINVAL but for compat */ 1274 } 1275 /* accept it */ 1276 #ifdef INET6 1277 if (sa->sa_family == AF_INET6) 1278 sa6_embedscope((struct sockaddr_in6 *)sa, 1279 V_ip6_use_defzone); 1280 #endif 1281 rtinfo->rti_info[i] = sa; 1282 cp += SA_SIZE(sa); 1283 } 1284 return (0); 1285 } 1286 1287 /* 1288 * Fill in @dmask with valid netmask leaving original @smask 1289 * intact. Mostly used with radix netmasks. 1290 */ 1291 struct sockaddr * 1292 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask, 1293 struct sockaddr_storage *dmask) 1294 { 1295 if (dst == NULL || smask == NULL) 1296 return (NULL); 1297 1298 memset(dmask, 0, dst->sa_len); 1299 memcpy(dmask, smask, smask->sa_len); 1300 dmask->ss_len = dst->sa_len; 1301 dmask->ss_family = dst->sa_family; 1302 1303 return ((struct sockaddr *)dmask); 1304 } 1305 1306 /* 1307 * Writes information related to @rtinfo object to newly-allocated mbuf. 1308 * Assumes MCLBYTES is enough to construct any message. 1309 * Used for OS notifications of vaious events (if/ifa announces,etc) 1310 * 1311 * Returns allocated mbuf or NULL on failure. 1312 */ 1313 static struct mbuf * 1314 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo) 1315 { 1316 struct sockaddr_storage ss; 1317 struct rt_msghdr *rtm; 1318 struct mbuf *m; 1319 int i; 1320 struct sockaddr *sa; 1321 #ifdef INET6 1322 struct sockaddr_in6 *sin6; 1323 #endif 1324 int len, dlen; 1325 1326 switch (type) { 1327 case RTM_DELADDR: 1328 case RTM_NEWADDR: 1329 len = sizeof(struct ifa_msghdr); 1330 break; 1331 1332 case RTM_DELMADDR: 1333 case RTM_NEWMADDR: 1334 len = sizeof(struct ifma_msghdr); 1335 break; 1336 1337 case RTM_IFINFO: 1338 len = sizeof(struct if_msghdr); 1339 break; 1340 1341 case RTM_IFANNOUNCE: 1342 case RTM_IEEE80211: 1343 len = sizeof(struct if_announcemsghdr); 1344 break; 1345 1346 default: 1347 len = sizeof(struct rt_msghdr); 1348 } 1349 1350 /* XXXGL: can we use MJUMPAGESIZE cluster here? */ 1351 KASSERT(len <= MCLBYTES, ("%s: message too big", __func__)); 1352 if (len > MHLEN) 1353 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 1354 else 1355 m = m_gethdr(M_NOWAIT, MT_DATA); 1356 if (m == NULL) 1357 return (m); 1358 1359 m->m_pkthdr.len = m->m_len = len; 1360 rtm = mtod(m, struct rt_msghdr *); 1361 bzero((caddr_t)rtm, len); 1362 for (i = 0; i < RTAX_MAX; i++) { 1363 if ((sa = rtinfo->rti_info[i]) == NULL) 1364 continue; 1365 rtinfo->rti_addrs |= (1 << i); 1366 1367 dlen = SA_SIZE(sa); 1368 KASSERT(dlen <= sizeof(ss), 1369 ("%s: sockaddr size overflow", __func__)); 1370 bzero(&ss, sizeof(ss)); 1371 bcopy(sa, &ss, sa->sa_len); 1372 sa = (struct sockaddr *)&ss; 1373 #ifdef INET6 1374 if (sa->sa_family == AF_INET6) { 1375 sin6 = (struct sockaddr_in6 *)sa; 1376 (void)sa6_recoverscope(sin6); 1377 } 1378 #endif 1379 m_copyback(m, len, dlen, (caddr_t)sa); 1380 len += dlen; 1381 } 1382 if (m->m_pkthdr.len != len) { 1383 m_freem(m); 1384 return (NULL); 1385 } 1386 rtm->rtm_msglen = len; 1387 rtm->rtm_version = RTM_VERSION; 1388 rtm->rtm_type = type; 1389 return (m); 1390 } 1391 1392 /* 1393 * Writes information related to @rtinfo object to preallocated buffer. 1394 * Stores needed size in @plen. If @w is NULL, calculates size without 1395 * writing. 1396 * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation. 1397 * 1398 * Returns 0 on success. 1399 * 1400 */ 1401 static int 1402 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen) 1403 { 1404 struct sockaddr_storage ss; 1405 int len, buflen = 0, dlen, i; 1406 caddr_t cp = NULL; 1407 struct rt_msghdr *rtm = NULL; 1408 #ifdef INET6 1409 struct sockaddr_in6 *sin6; 1410 #endif 1411 #ifdef COMPAT_FREEBSD32 1412 bool compat32 = false; 1413 #endif 1414 1415 switch (type) { 1416 case RTM_DELADDR: 1417 case RTM_NEWADDR: 1418 if (w != NULL && w->w_op == NET_RT_IFLISTL) { 1419 #ifdef COMPAT_FREEBSD32 1420 if (w->w_req->flags & SCTL_MASK32) { 1421 len = sizeof(struct ifa_msghdrl32); 1422 compat32 = true; 1423 } else 1424 #endif 1425 len = sizeof(struct ifa_msghdrl); 1426 } else 1427 len = sizeof(struct ifa_msghdr); 1428 break; 1429 1430 case RTM_IFINFO: 1431 #ifdef COMPAT_FREEBSD32 1432 if (w != NULL && w->w_req->flags & SCTL_MASK32) { 1433 if (w->w_op == NET_RT_IFLISTL) 1434 len = sizeof(struct if_msghdrl32); 1435 else 1436 len = sizeof(struct if_msghdr32); 1437 compat32 = true; 1438 break; 1439 } 1440 #endif 1441 if (w != NULL && w->w_op == NET_RT_IFLISTL) 1442 len = sizeof(struct if_msghdrl); 1443 else 1444 len = sizeof(struct if_msghdr); 1445 break; 1446 1447 case RTM_NEWMADDR: 1448 len = sizeof(struct ifma_msghdr); 1449 break; 1450 1451 default: 1452 len = sizeof(struct rt_msghdr); 1453 } 1454 1455 if (w != NULL) { 1456 rtm = (struct rt_msghdr *)w->w_tmem; 1457 buflen = w->w_tmemsize - len; 1458 cp = (caddr_t)w->w_tmem + len; 1459 } 1460 1461 rtinfo->rti_addrs = 0; 1462 for (i = 0; i < RTAX_MAX; i++) { 1463 struct sockaddr *sa; 1464 1465 if ((sa = rtinfo->rti_info[i]) == NULL) 1466 continue; 1467 rtinfo->rti_addrs |= (1 << i); 1468 #ifdef COMPAT_FREEBSD32 1469 if (compat32) 1470 dlen = SA_SIZE32(sa); 1471 else 1472 #endif 1473 dlen = SA_SIZE(sa); 1474 if (cp != NULL && buflen >= dlen) { 1475 KASSERT(dlen <= sizeof(ss), 1476 ("%s: sockaddr size overflow", __func__)); 1477 bzero(&ss, sizeof(ss)); 1478 bcopy(sa, &ss, sa->sa_len); 1479 sa = (struct sockaddr *)&ss; 1480 #ifdef INET6 1481 if (sa->sa_family == AF_INET6) { 1482 sin6 = (struct sockaddr_in6 *)sa; 1483 (void)sa6_recoverscope(sin6); 1484 } 1485 #endif 1486 bcopy((caddr_t)sa, cp, (unsigned)dlen); 1487 cp += dlen; 1488 buflen -= dlen; 1489 } else if (cp != NULL) { 1490 /* 1491 * Buffer too small. Count needed size 1492 * and return with error. 1493 */ 1494 cp = NULL; 1495 } 1496 1497 len += dlen; 1498 } 1499 1500 if (cp != NULL) { 1501 dlen = ALIGN(len) - len; 1502 if (buflen < dlen) 1503 cp = NULL; 1504 else { 1505 bzero(cp, dlen); 1506 cp += dlen; 1507 buflen -= dlen; 1508 } 1509 } 1510 len = ALIGN(len); 1511 1512 if (cp != NULL) { 1513 /* fill header iff buffer is large enough */ 1514 rtm->rtm_version = RTM_VERSION; 1515 rtm->rtm_type = type; 1516 rtm->rtm_msglen = len; 1517 } 1518 1519 *plen = len; 1520 1521 if (w != NULL && cp == NULL) 1522 return (ENOBUFS); 1523 1524 return (0); 1525 } 1526 1527 /* 1528 * This routine is called to generate a message from the routing 1529 * socket indicating that a redirect has occurred, a routing lookup 1530 * has failed, or that a protocol has detected timeouts to a particular 1531 * destination. 1532 */ 1533 void 1534 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error, 1535 int fibnum) 1536 { 1537 struct rt_msghdr *rtm; 1538 struct mbuf *m; 1539 struct sockaddr *sa = rtinfo->rti_info[RTAX_DST]; 1540 1541 if (V_route_cb.any_count == 0) 1542 return; 1543 m = rtsock_msg_mbuf(type, rtinfo); 1544 if (m == NULL) 1545 return; 1546 1547 if (fibnum != RT_ALL_FIBS) { 1548 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out " 1549 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs)); 1550 M_SETFIB(m, fibnum); 1551 m->m_flags |= RTS_FILTER_FIB; 1552 } 1553 1554 rtm = mtod(m, struct rt_msghdr *); 1555 rtm->rtm_flags = RTF_DONE | flags; 1556 rtm->rtm_errno = error; 1557 rtm->rtm_addrs = rtinfo->rti_addrs; 1558 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1559 } 1560 1561 void 1562 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error) 1563 { 1564 1565 rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS); 1566 } 1567 1568 /* 1569 * This routine is called to generate a message from the routing 1570 * socket indicating that the status of a network interface has changed. 1571 */ 1572 void 1573 rt_ifmsg(struct ifnet *ifp) 1574 { 1575 struct if_msghdr *ifm; 1576 struct mbuf *m; 1577 struct rt_addrinfo info; 1578 1579 if (V_route_cb.any_count == 0) 1580 return; 1581 bzero((caddr_t)&info, sizeof(info)); 1582 m = rtsock_msg_mbuf(RTM_IFINFO, &info); 1583 if (m == NULL) 1584 return; 1585 ifm = mtod(m, struct if_msghdr *); 1586 ifm->ifm_index = ifp->if_index; 1587 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1588 if_data_copy(ifp, &ifm->ifm_data); 1589 ifm->ifm_addrs = 0; 1590 rt_dispatch(m, AF_UNSPEC); 1591 } 1592 1593 /* 1594 * Announce interface address arrival/withdraw. 1595 * Please do not call directly, use rt_addrmsg(). 1596 * Assume input data to be valid. 1597 * Returns 0 on success. 1598 */ 1599 int 1600 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum) 1601 { 1602 struct rt_addrinfo info; 1603 struct sockaddr *sa; 1604 int ncmd; 1605 struct mbuf *m; 1606 struct ifa_msghdr *ifam; 1607 struct ifnet *ifp = ifa->ifa_ifp; 1608 struct sockaddr_storage ss; 1609 1610 if (V_route_cb.any_count == 0) 1611 return (0); 1612 1613 ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR; 1614 1615 bzero((caddr_t)&info, sizeof(info)); 1616 info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr; 1617 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr; 1618 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask( 1619 info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss); 1620 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr; 1621 if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL) 1622 return (ENOBUFS); 1623 ifam = mtod(m, struct ifa_msghdr *); 1624 ifam->ifam_index = ifp->if_index; 1625 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 1626 ifam->ifam_flags = ifa->ifa_flags; 1627 ifam->ifam_addrs = info.rti_addrs; 1628 1629 if (fibnum != RT_ALL_FIBS) { 1630 M_SETFIB(m, fibnum); 1631 m->m_flags |= RTS_FILTER_FIB; 1632 } 1633 1634 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1635 1636 return (0); 1637 } 1638 1639 /* 1640 * Announce route addition/removal to rtsock based on @rt data. 1641 * Callers are advives to use rt_routemsg() instead of using this 1642 * function directly. 1643 * Assume @rt data is consistent. 1644 * 1645 * Returns 0 on success. 1646 */ 1647 int 1648 rtsock_routemsg(int cmd, struct rtentry *rt, struct nhop_object *nh, 1649 int fibnum) 1650 { 1651 union sockaddr_union dst, mask; 1652 struct rt_addrinfo info; 1653 1654 if (V_route_cb.any_count == 0) 1655 return (0); 1656 1657 int family = rt_get_family(rt); 1658 init_sockaddrs_family(family, &dst.sa, &mask.sa); 1659 export_rtaddrs(rt, &dst.sa, &mask.sa); 1660 1661 bzero((caddr_t)&info, sizeof(info)); 1662 info.rti_info[RTAX_DST] = &dst.sa; 1663 info.rti_info[RTAX_NETMASK] = &mask.sa; 1664 info.rti_info[RTAX_GATEWAY] = &nh->gw_sa; 1665 info.rti_flags = rt->rte_flags | nhop_get_rtflags(nh); 1666 info.rti_ifp = nh->nh_ifp; 1667 1668 return (rtsock_routemsg_info(cmd, &info, fibnum)); 1669 } 1670 1671 int 1672 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum) 1673 { 1674 struct rt_msghdr *rtm; 1675 struct sockaddr *sa; 1676 struct mbuf *m; 1677 1678 if (V_route_cb.any_count == 0) 1679 return (0); 1680 1681 if (info->rti_flags & RTF_HOST) 1682 info->rti_info[RTAX_NETMASK] = NULL; 1683 1684 m = rtsock_msg_mbuf(cmd, info); 1685 if (m == NULL) 1686 return (ENOBUFS); 1687 1688 if (fibnum != RT_ALL_FIBS) { 1689 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out " 1690 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs)); 1691 M_SETFIB(m, fibnum); 1692 m->m_flags |= RTS_FILTER_FIB; 1693 } 1694 1695 rtm = mtod(m, struct rt_msghdr *); 1696 rtm->rtm_addrs = info->rti_addrs; 1697 if (info->rti_ifp != NULL) 1698 rtm->rtm_index = info->rti_ifp->if_index; 1699 /* Add RTF_DONE to indicate command 'completion' required by API */ 1700 info->rti_flags |= RTF_DONE; 1701 /* Reported routes has to be up */ 1702 if (cmd == RTM_ADD || cmd == RTM_CHANGE) 1703 info->rti_flags |= RTF_UP; 1704 rtm->rtm_flags = info->rti_flags; 1705 1706 sa = info->rti_info[RTAX_DST]; 1707 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1708 1709 return (0); 1710 } 1711 1712 /* 1713 * This is the analogue to the rt_newaddrmsg which performs the same 1714 * function but for multicast group memberhips. This is easier since 1715 * there is no route state to worry about. 1716 */ 1717 void 1718 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma) 1719 { 1720 struct rt_addrinfo info; 1721 struct mbuf *m = NULL; 1722 struct ifnet *ifp = ifma->ifma_ifp; 1723 struct ifma_msghdr *ifmam; 1724 1725 if (V_route_cb.any_count == 0) 1726 return; 1727 1728 bzero((caddr_t)&info, sizeof(info)); 1729 info.rti_info[RTAX_IFA] = ifma->ifma_addr; 1730 if (ifp && ifp->if_addr) 1731 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr; 1732 else 1733 info.rti_info[RTAX_IFP] = NULL; 1734 /* 1735 * If a link-layer address is present, present it as a ``gateway'' 1736 * (similarly to how ARP entries, e.g., are presented). 1737 */ 1738 info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr; 1739 m = rtsock_msg_mbuf(cmd, &info); 1740 if (m == NULL) 1741 return; 1742 ifmam = mtod(m, struct ifma_msghdr *); 1743 KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n", 1744 __func__)); 1745 ifmam->ifmam_index = ifp->if_index; 1746 ifmam->ifmam_addrs = info.rti_addrs; 1747 rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC); 1748 } 1749 1750 static struct mbuf * 1751 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what, 1752 struct rt_addrinfo *info) 1753 { 1754 struct if_announcemsghdr *ifan; 1755 struct mbuf *m; 1756 1757 if (V_route_cb.any_count == 0) 1758 return NULL; 1759 bzero((caddr_t)info, sizeof(*info)); 1760 m = rtsock_msg_mbuf(type, info); 1761 if (m != NULL) { 1762 ifan = mtod(m, struct if_announcemsghdr *); 1763 ifan->ifan_index = ifp->if_index; 1764 strlcpy(ifan->ifan_name, ifp->if_xname, 1765 sizeof(ifan->ifan_name)); 1766 ifan->ifan_what = what; 1767 } 1768 return m; 1769 } 1770 1771 /* 1772 * This is called to generate routing socket messages indicating 1773 * IEEE80211 wireless events. 1774 * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way. 1775 */ 1776 void 1777 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len) 1778 { 1779 struct mbuf *m; 1780 struct rt_addrinfo info; 1781 1782 m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info); 1783 if (m != NULL) { 1784 /* 1785 * Append the ieee80211 data. Try to stick it in the 1786 * mbuf containing the ifannounce msg; otherwise allocate 1787 * a new mbuf and append. 1788 * 1789 * NB: we assume m is a single mbuf. 1790 */ 1791 if (data_len > M_TRAILINGSPACE(m)) { 1792 struct mbuf *n = m_get(M_NOWAIT, MT_DATA); 1793 if (n == NULL) { 1794 m_freem(m); 1795 return; 1796 } 1797 bcopy(data, mtod(n, void *), data_len); 1798 n->m_len = data_len; 1799 m->m_next = n; 1800 } else if (data_len > 0) { 1801 bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len); 1802 m->m_len += data_len; 1803 } 1804 if (m->m_flags & M_PKTHDR) 1805 m->m_pkthdr.len += data_len; 1806 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len; 1807 rt_dispatch(m, AF_UNSPEC); 1808 } 1809 } 1810 1811 /* 1812 * This is called to generate routing socket messages indicating 1813 * network interface arrival and departure. 1814 */ 1815 void 1816 rt_ifannouncemsg(struct ifnet *ifp, int what) 1817 { 1818 struct mbuf *m; 1819 struct rt_addrinfo info; 1820 1821 m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info); 1822 if (m != NULL) 1823 rt_dispatch(m, AF_UNSPEC); 1824 } 1825 1826 static void 1827 rt_dispatch(struct mbuf *m, sa_family_t saf) 1828 { 1829 struct m_tag *tag; 1830 1831 /* 1832 * Preserve the family from the sockaddr, if any, in an m_tag for 1833 * use when injecting the mbuf into the routing socket buffer from 1834 * the netisr. 1835 */ 1836 if (saf != AF_UNSPEC) { 1837 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short), 1838 M_NOWAIT); 1839 if (tag == NULL) { 1840 m_freem(m); 1841 return; 1842 } 1843 *(unsigned short *)(tag + 1) = saf; 1844 m_tag_prepend(m, tag); 1845 } 1846 #ifdef VIMAGE 1847 if (V_loif) 1848 m->m_pkthdr.rcvif = V_loif; 1849 else { 1850 m_freem(m); 1851 return; 1852 } 1853 #endif 1854 netisr_queue(NETISR_ROUTE, m); /* mbuf is free'd on failure. */ 1855 } 1856 1857 /* 1858 * Checks if rte can be exported v.r.t jails/vnets. 1859 * 1860 * Returns 1 if it can, 0 otherwise. 1861 */ 1862 static bool 1863 can_export_rte(struct ucred *td_ucred, bool rt_is_host, 1864 const struct sockaddr *rt_dst) 1865 { 1866 1867 if ((!rt_is_host) ? jailed_without_vnet(td_ucred) 1868 : prison_if(td_ucred, rt_dst) != 0) 1869 return (false); 1870 return (true); 1871 } 1872 1873 1874 /* 1875 * This is used in dumping the kernel table via sysctl(). 1876 */ 1877 static int 1878 sysctl_dumpentry(struct rtentry *rt, void *vw) 1879 { 1880 struct walkarg *w = vw; 1881 struct nhop_object *nh; 1882 int error = 0; 1883 1884 NET_EPOCH_ASSERT(); 1885 1886 export_rtaddrs(rt, w->dst, w->mask); 1887 if (!can_export_rte(w->w_req->td->td_ucred, rt_is_host(rt), w->dst)) 1888 return (0); 1889 nh = rt_get_raw_nhop(rt); 1890 #ifdef ROUTE_MPATH 1891 if (NH_IS_NHGRP(nh)) { 1892 struct weightened_nhop *wn; 1893 uint32_t num_nhops; 1894 wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops); 1895 for (int i = 0; i < num_nhops; i++) { 1896 error = sysctl_dumpnhop(rt, wn[i].nh, wn[i].weight, w); 1897 if (error != 0) 1898 return (error); 1899 } 1900 } else 1901 #endif 1902 error = sysctl_dumpnhop(rt, nh, rt->rt_weight, w); 1903 1904 return (0); 1905 } 1906 1907 1908 static int 1909 sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, uint32_t weight, 1910 struct walkarg *w) 1911 { 1912 struct rt_addrinfo info; 1913 int error = 0, size; 1914 uint32_t rtflags; 1915 1916 rtflags = nhop_get_rtflags(nh); 1917 1918 if (w->w_op == NET_RT_FLAGS && !(rtflags & w->w_arg)) 1919 return (0); 1920 1921 bzero((caddr_t)&info, sizeof(info)); 1922 info.rti_info[RTAX_DST] = w->dst; 1923 info.rti_info[RTAX_GATEWAY] = &nh->gw_sa; 1924 info.rti_info[RTAX_NETMASK] = (rtflags & RTF_HOST) ? NULL : w->mask; 1925 info.rti_info[RTAX_GENMASK] = 0; 1926 if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) { 1927 info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr; 1928 info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr; 1929 if (nh->nh_ifp->if_flags & IFF_POINTOPOINT) 1930 info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr; 1931 } 1932 if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0) 1933 return (error); 1934 if (w->w_req && w->w_tmem) { 1935 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem; 1936 1937 bzero(&rtm->rtm_index, 1938 sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index)); 1939 1940 /* 1941 * rte flags may consist of RTF_HOST (duplicated in nhop rtflags) 1942 * and RTF_UP (if entry is linked, which is always true here). 1943 * Given that, use nhop rtflags & add RTF_UP. 1944 */ 1945 rtm->rtm_flags = rtflags | RTF_UP; 1946 if (rtm->rtm_flags & RTF_GWFLAG_COMPAT) 1947 rtm->rtm_flags = RTF_GATEWAY | 1948 (rtm->rtm_flags & ~RTF_GWFLAG_COMPAT); 1949 rt_getmetrics(rt, nh, &rtm->rtm_rmx); 1950 rtm->rtm_rmx.rmx_weight = weight; 1951 rtm->rtm_index = nh->nh_ifp->if_index; 1952 rtm->rtm_addrs = info.rti_addrs; 1953 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size); 1954 return (error); 1955 } 1956 return (error); 1957 } 1958 1959 static int 1960 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd, 1961 struct rt_addrinfo *info, struct walkarg *w, int len) 1962 { 1963 struct if_msghdrl *ifm; 1964 struct if_data *ifd; 1965 1966 ifm = (struct if_msghdrl *)w->w_tmem; 1967 1968 #ifdef COMPAT_FREEBSD32 1969 if (w->w_req->flags & SCTL_MASK32) { 1970 struct if_msghdrl32 *ifm32; 1971 1972 ifm32 = (struct if_msghdrl32 *)ifm; 1973 ifm32->ifm_addrs = info->rti_addrs; 1974 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1975 ifm32->ifm_index = ifp->if_index; 1976 ifm32->_ifm_spare1 = 0; 1977 ifm32->ifm_len = sizeof(*ifm32); 1978 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data); 1979 ifm32->_ifm_spare2 = 0; 1980 ifd = &ifm32->ifm_data; 1981 } else 1982 #endif 1983 { 1984 ifm->ifm_addrs = info->rti_addrs; 1985 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1986 ifm->ifm_index = ifp->if_index; 1987 ifm->_ifm_spare1 = 0; 1988 ifm->ifm_len = sizeof(*ifm); 1989 ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data); 1990 ifm->_ifm_spare2 = 0; 1991 ifd = &ifm->ifm_data; 1992 } 1993 1994 memcpy(ifd, src_ifd, sizeof(*ifd)); 1995 1996 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len)); 1997 } 1998 1999 static int 2000 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd, 2001 struct rt_addrinfo *info, struct walkarg *w, int len) 2002 { 2003 struct if_msghdr *ifm; 2004 struct if_data *ifd; 2005 2006 ifm = (struct if_msghdr *)w->w_tmem; 2007 2008 #ifdef COMPAT_FREEBSD32 2009 if (w->w_req->flags & SCTL_MASK32) { 2010 struct if_msghdr32 *ifm32; 2011 2012 ifm32 = (struct if_msghdr32 *)ifm; 2013 ifm32->ifm_addrs = info->rti_addrs; 2014 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 2015 ifm32->ifm_index = ifp->if_index; 2016 ifm32->_ifm_spare1 = 0; 2017 ifd = &ifm32->ifm_data; 2018 } else 2019 #endif 2020 { 2021 ifm->ifm_addrs = info->rti_addrs; 2022 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 2023 ifm->ifm_index = ifp->if_index; 2024 ifm->_ifm_spare1 = 0; 2025 ifd = &ifm->ifm_data; 2026 } 2027 2028 memcpy(ifd, src_ifd, sizeof(*ifd)); 2029 2030 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len)); 2031 } 2032 2033 static int 2034 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info, 2035 struct walkarg *w, int len) 2036 { 2037 struct ifa_msghdrl *ifam; 2038 struct if_data *ifd; 2039 2040 ifam = (struct ifa_msghdrl *)w->w_tmem; 2041 2042 #ifdef COMPAT_FREEBSD32 2043 if (w->w_req->flags & SCTL_MASK32) { 2044 struct ifa_msghdrl32 *ifam32; 2045 2046 ifam32 = (struct ifa_msghdrl32 *)ifam; 2047 ifam32->ifam_addrs = info->rti_addrs; 2048 ifam32->ifam_flags = ifa->ifa_flags; 2049 ifam32->ifam_index = ifa->ifa_ifp->if_index; 2050 ifam32->_ifam_spare1 = 0; 2051 ifam32->ifam_len = sizeof(*ifam32); 2052 ifam32->ifam_data_off = 2053 offsetof(struct ifa_msghdrl32, ifam_data); 2054 ifam32->ifam_metric = ifa->ifa_ifp->if_metric; 2055 ifd = &ifam32->ifam_data; 2056 } else 2057 #endif 2058 { 2059 ifam->ifam_addrs = info->rti_addrs; 2060 ifam->ifam_flags = ifa->ifa_flags; 2061 ifam->ifam_index = ifa->ifa_ifp->if_index; 2062 ifam->_ifam_spare1 = 0; 2063 ifam->ifam_len = sizeof(*ifam); 2064 ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data); 2065 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 2066 ifd = &ifam->ifam_data; 2067 } 2068 2069 bzero(ifd, sizeof(*ifd)); 2070 ifd->ifi_datalen = sizeof(struct if_data); 2071 ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets); 2072 ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets); 2073 ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes); 2074 ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes); 2075 2076 /* Fixup if_data carp(4) vhid. */ 2077 if (carp_get_vhid_p != NULL) 2078 ifd->ifi_vhid = (*carp_get_vhid_p)(ifa); 2079 2080 return (SYSCTL_OUT(w->w_req, w->w_tmem, len)); 2081 } 2082 2083 static int 2084 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info, 2085 struct walkarg *w, int len) 2086 { 2087 struct ifa_msghdr *ifam; 2088 2089 ifam = (struct ifa_msghdr *)w->w_tmem; 2090 ifam->ifam_addrs = info->rti_addrs; 2091 ifam->ifam_flags = ifa->ifa_flags; 2092 ifam->ifam_index = ifa->ifa_ifp->if_index; 2093 ifam->_ifam_spare1 = 0; 2094 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 2095 2096 return (SYSCTL_OUT(w->w_req, w->w_tmem, len)); 2097 } 2098 2099 static int 2100 sysctl_iflist(int af, struct walkarg *w) 2101 { 2102 struct ifnet *ifp; 2103 struct ifaddr *ifa; 2104 struct if_data ifd; 2105 struct rt_addrinfo info; 2106 int len, error = 0; 2107 struct sockaddr_storage ss; 2108 2109 bzero((caddr_t)&info, sizeof(info)); 2110 bzero(&ifd, sizeof(ifd)); 2111 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2112 if (w->w_arg && w->w_arg != ifp->if_index) 2113 continue; 2114 if_data_copy(ifp, &ifd); 2115 ifa = ifp->if_addr; 2116 info.rti_info[RTAX_IFP] = ifa->ifa_addr; 2117 error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len); 2118 if (error != 0) 2119 goto done; 2120 info.rti_info[RTAX_IFP] = NULL; 2121 if (w->w_req && w->w_tmem) { 2122 if (w->w_op == NET_RT_IFLISTL) 2123 error = sysctl_iflist_ifml(ifp, &ifd, &info, w, 2124 len); 2125 else 2126 error = sysctl_iflist_ifm(ifp, &ifd, &info, w, 2127 len); 2128 if (error) 2129 goto done; 2130 } 2131 while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) { 2132 if (af && af != ifa->ifa_addr->sa_family) 2133 continue; 2134 if (prison_if(w->w_req->td->td_ucred, 2135 ifa->ifa_addr) != 0) 2136 continue; 2137 info.rti_info[RTAX_IFA] = ifa->ifa_addr; 2138 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask( 2139 ifa->ifa_addr, ifa->ifa_netmask, &ss); 2140 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr; 2141 error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len); 2142 if (error != 0) 2143 goto done; 2144 if (w->w_req && w->w_tmem) { 2145 if (w->w_op == NET_RT_IFLISTL) 2146 error = sysctl_iflist_ifaml(ifa, &info, 2147 w, len); 2148 else 2149 error = sysctl_iflist_ifam(ifa, &info, 2150 w, len); 2151 if (error) 2152 goto done; 2153 } 2154 } 2155 info.rti_info[RTAX_IFA] = NULL; 2156 info.rti_info[RTAX_NETMASK] = NULL; 2157 info.rti_info[RTAX_BRD] = NULL; 2158 } 2159 done: 2160 return (error); 2161 } 2162 2163 static int 2164 sysctl_ifmalist(int af, struct walkarg *w) 2165 { 2166 struct rt_addrinfo info; 2167 struct ifaddr *ifa; 2168 struct ifmultiaddr *ifma; 2169 struct ifnet *ifp; 2170 int error, len; 2171 2172 NET_EPOCH_ASSERT(); 2173 2174 error = 0; 2175 bzero((caddr_t)&info, sizeof(info)); 2176 2177 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2178 if (w->w_arg && w->w_arg != ifp->if_index) 2179 continue; 2180 ifa = ifp->if_addr; 2181 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL; 2182 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 2183 if (af && af != ifma->ifma_addr->sa_family) 2184 continue; 2185 if (prison_if(w->w_req->td->td_ucred, 2186 ifma->ifma_addr) != 0) 2187 continue; 2188 info.rti_info[RTAX_IFA] = ifma->ifma_addr; 2189 info.rti_info[RTAX_GATEWAY] = 2190 (ifma->ifma_addr->sa_family != AF_LINK) ? 2191 ifma->ifma_lladdr : NULL; 2192 error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len); 2193 if (error != 0) 2194 break; 2195 if (w->w_req && w->w_tmem) { 2196 struct ifma_msghdr *ifmam; 2197 2198 ifmam = (struct ifma_msghdr *)w->w_tmem; 2199 ifmam->ifmam_index = ifma->ifma_ifp->if_index; 2200 ifmam->ifmam_flags = 0; 2201 ifmam->ifmam_addrs = info.rti_addrs; 2202 ifmam->_ifmam_spare1 = 0; 2203 error = SYSCTL_OUT(w->w_req, w->w_tmem, len); 2204 if (error != 0) 2205 break; 2206 } 2207 } 2208 if (error != 0) 2209 break; 2210 } 2211 return (error); 2212 } 2213 2214 static void 2215 rtable_sysctl_dump(uint32_t fibnum, int family, struct walkarg *w) 2216 { 2217 union sockaddr_union sa_dst, sa_mask; 2218 2219 w->family = family; 2220 w->dst = (struct sockaddr *)&sa_dst; 2221 w->mask = (struct sockaddr *)&sa_mask; 2222 2223 init_sockaddrs_family(family, w->dst, w->mask); 2224 2225 rib_walk(fibnum, family, false, sysctl_dumpentry, w); 2226 } 2227 2228 static int 2229 sysctl_rtsock(SYSCTL_HANDLER_ARGS) 2230 { 2231 struct epoch_tracker et; 2232 int *name = (int *)arg1; 2233 u_int namelen = arg2; 2234 struct rib_head *rnh = NULL; /* silence compiler. */ 2235 int i, lim, error = EINVAL; 2236 int fib = 0; 2237 u_char af; 2238 struct walkarg w; 2239 2240 name ++; 2241 namelen--; 2242 if (req->newptr) 2243 return (EPERM); 2244 if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP || name[1] == NET_RT_NHGRP) { 2245 if (namelen == 3) 2246 fib = req->td->td_proc->p_fibnum; 2247 else if (namelen == 4) 2248 fib = (name[3] == RT_ALL_FIBS) ? 2249 req->td->td_proc->p_fibnum : name[3]; 2250 else 2251 return ((namelen < 3) ? EISDIR : ENOTDIR); 2252 if (fib < 0 || fib >= rt_numfibs) 2253 return (EINVAL); 2254 } else if (namelen != 3) 2255 return ((namelen < 3) ? EISDIR : ENOTDIR); 2256 af = name[0]; 2257 if (af > AF_MAX) 2258 return (EINVAL); 2259 bzero(&w, sizeof(w)); 2260 w.w_op = name[1]; 2261 w.w_arg = name[2]; 2262 w.w_req = req; 2263 2264 error = sysctl_wire_old_buffer(req, 0); 2265 if (error) 2266 return (error); 2267 2268 /* 2269 * Allocate reply buffer in advance. 2270 * All rtsock messages has maximum length of u_short. 2271 */ 2272 w.w_tmemsize = 65536; 2273 w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK); 2274 2275 NET_EPOCH_ENTER(et); 2276 switch (w.w_op) { 2277 case NET_RT_DUMP: 2278 case NET_RT_FLAGS: 2279 if (af == 0) { /* dump all tables */ 2280 i = 1; 2281 lim = AF_MAX; 2282 } else /* dump only one table */ 2283 i = lim = af; 2284 2285 /* 2286 * take care of llinfo entries, the caller must 2287 * specify an AF 2288 */ 2289 if (w.w_op == NET_RT_FLAGS && 2290 (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) { 2291 if (af != 0) 2292 error = lltable_sysctl_dumparp(af, w.w_req); 2293 else 2294 error = EINVAL; 2295 break; 2296 } 2297 /* 2298 * take care of routing entries 2299 */ 2300 for (error = 0; error == 0 && i <= lim; i++) { 2301 rnh = rt_tables_get_rnh(fib, i); 2302 if (rnh != NULL) { 2303 rtable_sysctl_dump(fib, i, &w); 2304 } else if (af != 0) 2305 error = EAFNOSUPPORT; 2306 } 2307 break; 2308 case NET_RT_NHOP: 2309 case NET_RT_NHGRP: 2310 /* Allow dumping one specific af/fib at a time */ 2311 if (namelen < 4) { 2312 error = EINVAL; 2313 break; 2314 } 2315 fib = name[3]; 2316 if (fib < 0 || fib > rt_numfibs) { 2317 error = EINVAL; 2318 break; 2319 } 2320 rnh = rt_tables_get_rnh(fib, af); 2321 if (rnh == NULL) { 2322 error = EAFNOSUPPORT; 2323 break; 2324 } 2325 if (w.w_op == NET_RT_NHOP) 2326 error = nhops_dump_sysctl(rnh, w.w_req); 2327 else 2328 #ifdef ROUTE_MPATH 2329 error = nhgrp_dump_sysctl(rnh, w.w_req); 2330 #else 2331 error = ENOTSUP; 2332 #endif 2333 break; 2334 case NET_RT_IFLIST: 2335 case NET_RT_IFLISTL: 2336 error = sysctl_iflist(af, &w); 2337 break; 2338 2339 case NET_RT_IFMALIST: 2340 error = sysctl_ifmalist(af, &w); 2341 break; 2342 } 2343 NET_EPOCH_EXIT(et); 2344 2345 free(w.w_tmem, M_TEMP); 2346 return (error); 2347 } 2348 2349 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE, 2350 sysctl_rtsock, "Return route tables and interface/address lists"); 2351 2352 /* 2353 * Definitions of protocols supported in the ROUTE domain. 2354 */ 2355 2356 static struct domain routedomain; /* or at least forward */ 2357 2358 static struct protosw routesw[] = { 2359 { 2360 .pr_type = SOCK_RAW, 2361 .pr_domain = &routedomain, 2362 .pr_flags = PR_ATOMIC|PR_ADDR, 2363 .pr_output = route_output, 2364 .pr_ctlinput = raw_ctlinput, 2365 .pr_init = raw_init, 2366 .pr_usrreqs = &route_usrreqs 2367 } 2368 }; 2369 2370 static struct domain routedomain = { 2371 .dom_family = PF_ROUTE, 2372 .dom_name = "route", 2373 .dom_protosw = routesw, 2374 .dom_protoswNPROTOSW = &routesw[nitems(routesw)] 2375 }; 2376 2377 VNET_DOMAIN_SET(route); 2378