1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1988, 1991, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * @(#)rtsock.c 8.7 (Berkeley) 10/12/95 32 * $FreeBSD$ 33 */ 34 #include "opt_ddb.h" 35 #include "opt_mpath.h" 36 #include "opt_inet.h" 37 #include "opt_inet6.h" 38 39 #include <sys/param.h> 40 #include <sys/jail.h> 41 #include <sys/kernel.h> 42 #include <sys/domain.h> 43 #include <sys/lock.h> 44 #include <sys/malloc.h> 45 #include <sys/mbuf.h> 46 #include <sys/priv.h> 47 #include <sys/proc.h> 48 #include <sys/protosw.h> 49 #include <sys/rmlock.h> 50 #include <sys/rwlock.h> 51 #include <sys/signalvar.h> 52 #include <sys/socket.h> 53 #include <sys/socketvar.h> 54 #include <sys/sysctl.h> 55 #include <sys/systm.h> 56 57 #ifdef DDB 58 #include <ddb/ddb.h> 59 #include <ddb/db_lex.h> 60 #endif 61 62 #include <net/if.h> 63 #include <net/if_var.h> 64 #include <net/if_dl.h> 65 #include <net/if_llatbl.h> 66 #include <net/if_types.h> 67 #include <net/netisr.h> 68 #include <net/raw_cb.h> 69 #include <net/route.h> 70 #include <net/route_var.h> 71 #include <net/vnet.h> 72 73 #include <netinet/in.h> 74 #include <netinet/if_ether.h> 75 #include <netinet/ip_carp.h> 76 #ifdef INET6 77 #include <netinet6/ip6_var.h> 78 #include <netinet6/scope6_var.h> 79 #endif 80 81 #ifdef COMPAT_FREEBSD32 82 #include <sys/mount.h> 83 #include <compat/freebsd32/freebsd32.h> 84 85 struct if_msghdr32 { 86 uint16_t ifm_msglen; 87 uint8_t ifm_version; 88 uint8_t ifm_type; 89 int32_t ifm_addrs; 90 int32_t ifm_flags; 91 uint16_t ifm_index; 92 uint16_t _ifm_spare1; 93 struct if_data ifm_data; 94 }; 95 96 struct if_msghdrl32 { 97 uint16_t ifm_msglen; 98 uint8_t ifm_version; 99 uint8_t ifm_type; 100 int32_t ifm_addrs; 101 int32_t ifm_flags; 102 uint16_t ifm_index; 103 uint16_t _ifm_spare1; 104 uint16_t ifm_len; 105 uint16_t ifm_data_off; 106 uint32_t _ifm_spare2; 107 struct if_data ifm_data; 108 }; 109 110 struct ifa_msghdrl32 { 111 uint16_t ifam_msglen; 112 uint8_t ifam_version; 113 uint8_t ifam_type; 114 int32_t ifam_addrs; 115 int32_t ifam_flags; 116 uint16_t ifam_index; 117 uint16_t _ifam_spare1; 118 uint16_t ifam_len; 119 uint16_t ifam_data_off; 120 int32_t ifam_metric; 121 struct if_data ifam_data; 122 }; 123 124 #define SA_SIZE32(sa) \ 125 ( (((struct sockaddr *)(sa))->sa_len == 0) ? \ 126 sizeof(int) : \ 127 1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) ) 128 129 #endif /* COMPAT_FREEBSD32 */ 130 131 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables"); 132 133 /* NB: these are not modified */ 134 static struct sockaddr route_src = { 2, PF_ROUTE, }; 135 static struct sockaddr sa_zero = { sizeof(sa_zero), AF_INET, }; 136 137 /* These are external hooks for CARP. */ 138 int (*carp_get_vhid_p)(struct ifaddr *); 139 140 /* 141 * Used by rtsock/raw_input callback code to decide whether to filter the update 142 * notification to a socket bound to a particular FIB. 143 */ 144 #define RTS_FILTER_FIB M_PROTO8 145 146 typedef struct { 147 int ip_count; /* attached w/ AF_INET */ 148 int ip6_count; /* attached w/ AF_INET6 */ 149 int any_count; /* total attached */ 150 } route_cb_t; 151 VNET_DEFINE_STATIC(route_cb_t, route_cb); 152 #define V_route_cb VNET(route_cb) 153 154 struct mtx rtsock_mtx; 155 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF); 156 157 #define RTSOCK_LOCK() mtx_lock(&rtsock_mtx) 158 #define RTSOCK_UNLOCK() mtx_unlock(&rtsock_mtx) 159 #define RTSOCK_LOCK_ASSERT() mtx_assert(&rtsock_mtx, MA_OWNED) 160 161 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 162 ""); 163 164 struct walkarg { 165 int w_tmemsize; 166 int w_op, w_arg; 167 caddr_t w_tmem; 168 struct sysctl_req *w_req; 169 }; 170 171 static void rts_input(struct mbuf *m); 172 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo); 173 static int rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, 174 struct walkarg *w, int *plen); 175 static int rt_xaddrs(caddr_t cp, caddr_t cplim, 176 struct rt_addrinfo *rtinfo); 177 static int sysctl_dumpentry(struct radix_node *rn, void *vw); 178 static int sysctl_iflist(int af, struct walkarg *w); 179 static int sysctl_ifmalist(int af, struct walkarg *w); 180 static int route_output(struct mbuf *m, struct socket *so, ...); 181 static void rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out); 182 static void rt_dispatch(struct mbuf *, sa_family_t); 183 static struct sockaddr *rtsock_fix_netmask(struct sockaddr *dst, 184 struct sockaddr *smask, struct sockaddr_storage *dmask); 185 static int handle_rtm_get(struct rt_addrinfo *info, u_int fibnum, 186 struct rt_msghdr *rtm, struct rtentry **ret_nrt); 187 static int update_rtm_from_rte(struct rt_addrinfo *info, 188 struct rt_msghdr **prtm, int alloc_len, 189 struct rtentry *rt); 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 int can_export_rte(struct ucred *td_ucred, const struct rtentry *rt); 194 195 static struct netisr_handler rtsock_nh = { 196 .nh_name = "rtsock", 197 .nh_handler = rts_input, 198 .nh_proto = NETISR_ROUTE, 199 .nh_policy = NETISR_POLICY_SOURCE, 200 }; 201 202 static int 203 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS) 204 { 205 int error, qlimit; 206 207 netisr_getqlimit(&rtsock_nh, &qlimit); 208 error = sysctl_handle_int(oidp, &qlimit, 0, req); 209 if (error || !req->newptr) 210 return (error); 211 if (qlimit < 1) 212 return (EINVAL); 213 return (netisr_setqlimit(&rtsock_nh, qlimit)); 214 } 215 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, 216 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 217 0, 0, sysctl_route_netisr_maxqlen, "I", 218 "maximum routing socket dispatch queue length"); 219 220 static void 221 vnet_rts_init(void) 222 { 223 int tmp; 224 225 if (IS_DEFAULT_VNET(curvnet)) { 226 if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp)) 227 rtsock_nh.nh_qlimit = tmp; 228 netisr_register(&rtsock_nh); 229 } 230 #ifdef VIMAGE 231 else 232 netisr_register_vnet(&rtsock_nh); 233 #endif 234 } 235 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 236 vnet_rts_init, 0); 237 238 #ifdef VIMAGE 239 static void 240 vnet_rts_uninit(void) 241 { 242 243 netisr_unregister_vnet(&rtsock_nh); 244 } 245 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 246 vnet_rts_uninit, 0); 247 #endif 248 249 static int 250 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src, 251 struct rawcb *rp) 252 { 253 int fibnum; 254 255 KASSERT(m != NULL, ("%s: m is NULL", __func__)); 256 KASSERT(proto != NULL, ("%s: proto is NULL", __func__)); 257 KASSERT(rp != NULL, ("%s: rp is NULL", __func__)); 258 259 /* No filtering requested. */ 260 if ((m->m_flags & RTS_FILTER_FIB) == 0) 261 return (0); 262 263 /* Check if it is a rts and the fib matches the one of the socket. */ 264 fibnum = M_GETFIB(m); 265 if (proto->sp_family != PF_ROUTE || 266 rp->rcb_socket == NULL || 267 rp->rcb_socket->so_fibnum == fibnum) 268 return (0); 269 270 /* Filtering requested and no match, the socket shall be skipped. */ 271 return (1); 272 } 273 274 static void 275 rts_input(struct mbuf *m) 276 { 277 struct sockproto route_proto; 278 unsigned short *family; 279 struct m_tag *tag; 280 281 route_proto.sp_family = PF_ROUTE; 282 tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL); 283 if (tag != NULL) { 284 family = (unsigned short *)(tag + 1); 285 route_proto.sp_protocol = *family; 286 m_tag_delete(m, tag); 287 } else 288 route_proto.sp_protocol = 0; 289 290 raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb); 291 } 292 293 /* 294 * It really doesn't make any sense at all for this code to share much 295 * with raw_usrreq.c, since its functionality is so restricted. XXX 296 */ 297 static void 298 rts_abort(struct socket *so) 299 { 300 301 raw_usrreqs.pru_abort(so); 302 } 303 304 static void 305 rts_close(struct socket *so) 306 { 307 308 raw_usrreqs.pru_close(so); 309 } 310 311 /* pru_accept is EOPNOTSUPP */ 312 313 static int 314 rts_attach(struct socket *so, int proto, struct thread *td) 315 { 316 struct rawcb *rp; 317 int error; 318 319 KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL")); 320 321 /* XXX */ 322 rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO); 323 324 so->so_pcb = (caddr_t)rp; 325 so->so_fibnum = td->td_proc->p_fibnum; 326 error = raw_attach(so, proto); 327 rp = sotorawcb(so); 328 if (error) { 329 so->so_pcb = NULL; 330 free(rp, M_PCB); 331 return error; 332 } 333 RTSOCK_LOCK(); 334 switch(rp->rcb_proto.sp_protocol) { 335 case AF_INET: 336 V_route_cb.ip_count++; 337 break; 338 case AF_INET6: 339 V_route_cb.ip6_count++; 340 break; 341 } 342 V_route_cb.any_count++; 343 RTSOCK_UNLOCK(); 344 soisconnected(so); 345 so->so_options |= SO_USELOOPBACK; 346 return 0; 347 } 348 349 static int 350 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 351 { 352 353 return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */ 354 } 355 356 static int 357 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 358 { 359 360 return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */ 361 } 362 363 /* pru_connect2 is EOPNOTSUPP */ 364 /* pru_control is EOPNOTSUPP */ 365 366 static void 367 rts_detach(struct socket *so) 368 { 369 struct rawcb *rp = sotorawcb(so); 370 371 KASSERT(rp != NULL, ("rts_detach: rp == NULL")); 372 373 RTSOCK_LOCK(); 374 switch(rp->rcb_proto.sp_protocol) { 375 case AF_INET: 376 V_route_cb.ip_count--; 377 break; 378 case AF_INET6: 379 V_route_cb.ip6_count--; 380 break; 381 } 382 V_route_cb.any_count--; 383 RTSOCK_UNLOCK(); 384 raw_usrreqs.pru_detach(so); 385 } 386 387 static int 388 rts_disconnect(struct socket *so) 389 { 390 391 return (raw_usrreqs.pru_disconnect(so)); 392 } 393 394 /* pru_listen is EOPNOTSUPP */ 395 396 static int 397 rts_peeraddr(struct socket *so, struct sockaddr **nam) 398 { 399 400 return (raw_usrreqs.pru_peeraddr(so, nam)); 401 } 402 403 /* pru_rcvd is EOPNOTSUPP */ 404 /* pru_rcvoob is EOPNOTSUPP */ 405 406 static int 407 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 408 struct mbuf *control, struct thread *td) 409 { 410 411 return (raw_usrreqs.pru_send(so, flags, m, nam, control, td)); 412 } 413 414 /* pru_sense is null */ 415 416 static int 417 rts_shutdown(struct socket *so) 418 { 419 420 return (raw_usrreqs.pru_shutdown(so)); 421 } 422 423 static int 424 rts_sockaddr(struct socket *so, struct sockaddr **nam) 425 { 426 427 return (raw_usrreqs.pru_sockaddr(so, nam)); 428 } 429 430 static struct pr_usrreqs route_usrreqs = { 431 .pru_abort = rts_abort, 432 .pru_attach = rts_attach, 433 .pru_bind = rts_bind, 434 .pru_connect = rts_connect, 435 .pru_detach = rts_detach, 436 .pru_disconnect = rts_disconnect, 437 .pru_peeraddr = rts_peeraddr, 438 .pru_send = rts_send, 439 .pru_shutdown = rts_shutdown, 440 .pru_sockaddr = rts_sockaddr, 441 .pru_close = rts_close, 442 }; 443 444 #ifndef _SOCKADDR_UNION_DEFINED 445 #define _SOCKADDR_UNION_DEFINED 446 /* 447 * The union of all possible address formats we handle. 448 */ 449 union sockaddr_union { 450 struct sockaddr sa; 451 struct sockaddr_in sin; 452 struct sockaddr_in6 sin6; 453 }; 454 #endif /* _SOCKADDR_UNION_DEFINED */ 455 456 static int 457 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp, 458 struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred) 459 { 460 #if defined(INET) || defined(INET6) 461 struct epoch_tracker et; 462 #endif 463 464 /* First, see if the returned address is part of the jail. */ 465 if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) { 466 info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr; 467 return (0); 468 } 469 470 switch (info->rti_info[RTAX_DST]->sa_family) { 471 #ifdef INET 472 case AF_INET: 473 { 474 struct in_addr ia; 475 struct ifaddr *ifa; 476 int found; 477 478 found = 0; 479 /* 480 * Try to find an address on the given outgoing interface 481 * that belongs to the jail. 482 */ 483 NET_EPOCH_ENTER(et); 484 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 485 struct sockaddr *sa; 486 sa = ifa->ifa_addr; 487 if (sa->sa_family != AF_INET) 488 continue; 489 ia = ((struct sockaddr_in *)sa)->sin_addr; 490 if (prison_check_ip4(cred, &ia) == 0) { 491 found = 1; 492 break; 493 } 494 } 495 NET_EPOCH_EXIT(et); 496 if (!found) { 497 /* 498 * As a last resort return the 'default' jail address. 499 */ 500 ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)-> 501 sin_addr; 502 if (prison_get_ip4(cred, &ia) != 0) 503 return (ESRCH); 504 } 505 bzero(&saun->sin, sizeof(struct sockaddr_in)); 506 saun->sin.sin_len = sizeof(struct sockaddr_in); 507 saun->sin.sin_family = AF_INET; 508 saun->sin.sin_addr.s_addr = ia.s_addr; 509 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin; 510 break; 511 } 512 #endif 513 #ifdef INET6 514 case AF_INET6: 515 { 516 struct in6_addr ia6; 517 struct ifaddr *ifa; 518 int found; 519 520 found = 0; 521 /* 522 * Try to find an address on the given outgoing interface 523 * that belongs to the jail. 524 */ 525 NET_EPOCH_ENTER(et); 526 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 527 struct sockaddr *sa; 528 sa = ifa->ifa_addr; 529 if (sa->sa_family != AF_INET6) 530 continue; 531 bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr, 532 &ia6, sizeof(struct in6_addr)); 533 if (prison_check_ip6(cred, &ia6) == 0) { 534 found = 1; 535 break; 536 } 537 } 538 NET_EPOCH_EXIT(et); 539 if (!found) { 540 /* 541 * As a last resort return the 'default' jail address. 542 */ 543 ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)-> 544 sin6_addr; 545 if (prison_get_ip6(cred, &ia6) != 0) 546 return (ESRCH); 547 } 548 bzero(&saun->sin6, sizeof(struct sockaddr_in6)); 549 saun->sin6.sin6_len = sizeof(struct sockaddr_in6); 550 saun->sin6.sin6_family = AF_INET6; 551 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr)); 552 if (sa6_recoverscope(&saun->sin6) != 0) 553 return (ESRCH); 554 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6; 555 break; 556 } 557 #endif 558 default: 559 return (ESRCH); 560 } 561 return (0); 562 } 563 564 /* 565 * Fills in @info based on userland-provided @rtm message. 566 * 567 * Returns 0 on success. 568 */ 569 static int 570 fill_addrinfo(struct rt_msghdr *rtm, int len, u_int fibnum, struct rt_addrinfo *info) 571 { 572 int error; 573 sa_family_t saf; 574 575 rtm->rtm_pid = curproc->p_pid; 576 info->rti_addrs = rtm->rtm_addrs; 577 578 info->rti_mflags = rtm->rtm_inits; 579 info->rti_rmx = &rtm->rtm_rmx; 580 581 /* 582 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6 583 * link-local address because rtrequest requires addresses with 584 * embedded scope id. 585 */ 586 if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info)) 587 return (EINVAL); 588 589 if (rtm->rtm_flags & RTF_RNH_LOCKED) 590 return (EINVAL); 591 info->rti_flags = rtm->rtm_flags; 592 if (info->rti_info[RTAX_DST] == NULL || 593 info->rti_info[RTAX_DST]->sa_family >= AF_MAX || 594 (info->rti_info[RTAX_GATEWAY] != NULL && 595 info->rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) 596 return (EINVAL); 597 saf = info->rti_info[RTAX_DST]->sa_family; 598 /* 599 * Verify that the caller has the appropriate privilege; RTM_GET 600 * is the only operation the non-superuser is allowed. 601 */ 602 if (rtm->rtm_type != RTM_GET) { 603 error = priv_check(curthread, PRIV_NET_ROUTE); 604 if (error != 0) 605 return (error); 606 } 607 608 /* 609 * The given gateway address may be an interface address. 610 * For example, issuing a "route change" command on a route 611 * entry that was created from a tunnel, and the gateway 612 * address given is the local end point. In this case the 613 * RTF_GATEWAY flag must be cleared or the destination will 614 * not be reachable even though there is no error message. 615 */ 616 if (info->rti_info[RTAX_GATEWAY] != NULL && 617 info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) { 618 struct rt_addrinfo ginfo; 619 struct sockaddr *gdst; 620 struct sockaddr_storage ss; 621 622 bzero(&ginfo, sizeof(ginfo)); 623 bzero(&ss, sizeof(ss)); 624 ss.ss_len = sizeof(ss); 625 626 ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss; 627 gdst = info->rti_info[RTAX_GATEWAY]; 628 629 /* 630 * A host route through the loopback interface is 631 * installed for each interface adddress. In pre 8.0 632 * releases the interface address of a PPP link type 633 * is not reachable locally. This behavior is fixed as 634 * part of the new L2/L3 redesign and rewrite work. The 635 * signature of this interface address route is the 636 * AF_LINK sa_family type of the rt_gateway, and the 637 * rt_ifp has the IFF_LOOPBACK flag set. 638 */ 639 if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) { 640 if (ss.ss_family == AF_LINK && 641 ginfo.rti_ifp->if_flags & IFF_LOOPBACK) { 642 info->rti_flags &= ~RTF_GATEWAY; 643 info->rti_flags |= RTF_GWFLAG_COMPAT; 644 } 645 rib_free_info(&ginfo); 646 } 647 } 648 649 return (0); 650 } 651 652 /* 653 * Handles RTM_GET message from routing socket, returning matching rt. 654 * 655 * Returns: 656 * 0 on success, with locked and referenced matching rt in @rt_nrt 657 * errno of failure 658 */ 659 static int 660 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum, 661 struct rt_msghdr *rtm, struct rtentry **ret_nrt) 662 { 663 RIB_RLOCK_TRACKER; 664 struct rtentry *rt; 665 struct rib_head *rnh; 666 sa_family_t saf; 667 668 saf = info->rti_info[RTAX_DST]->sa_family; 669 670 rnh = rt_tables_get_rnh(fibnum, saf); 671 if (rnh == NULL) 672 return (EAFNOSUPPORT); 673 674 RIB_RLOCK(rnh); 675 676 if (info->rti_info[RTAX_NETMASK] == NULL) { 677 /* 678 * Provide longest prefix match for 679 * address lookup (no mask). 680 * 'route -n get addr' 681 */ 682 rt = (struct rtentry *) rnh->rnh_matchaddr( 683 info->rti_info[RTAX_DST], &rnh->head); 684 } else 685 rt = (struct rtentry *) rnh->rnh_lookup( 686 info->rti_info[RTAX_DST], 687 info->rti_info[RTAX_NETMASK], &rnh->head); 688 689 if (rt == NULL) { 690 RIB_RUNLOCK(rnh); 691 return (ESRCH); 692 } 693 #ifdef RADIX_MPATH 694 /* 695 * for RTM_GET, gate is optional even with multipath. 696 * if gate == NULL the first match is returned. 697 * (no need to call rt_mpath_matchgate if gate == NULL) 698 */ 699 if (rt_mpath_capable(rnh) && info->rti_info[RTAX_GATEWAY]) { 700 rt = rt_mpath_matchgate(rt, info->rti_info[RTAX_GATEWAY]); 701 if (!rt) { 702 RIB_RUNLOCK(rnh); 703 return (ESRCH); 704 } 705 } 706 #endif 707 /* 708 * If performing proxied L2 entry insertion, and 709 * the actual PPP host entry is found, perform 710 * another search to retrieve the prefix route of 711 * the local end point of the PPP link. 712 */ 713 if (rtm->rtm_flags & RTF_ANNOUNCE) { 714 struct sockaddr laddr; 715 716 if (rt->rt_ifp != NULL && 717 rt->rt_ifp->if_type == IFT_PROPVIRTUAL) { 718 struct epoch_tracker et; 719 struct ifaddr *ifa; 720 721 NET_EPOCH_ENTER(et); 722 ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1, 723 RT_ALL_FIBS); 724 NET_EPOCH_EXIT(et); 725 if (ifa != NULL) 726 rt_maskedcopy(ifa->ifa_addr, 727 &laddr, 728 ifa->ifa_netmask); 729 } else 730 rt_maskedcopy(rt->rt_ifa->ifa_addr, 731 &laddr, 732 rt->rt_ifa->ifa_netmask); 733 /* 734 * refactor rt and no lock operation necessary 735 */ 736 rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr, 737 &rnh->head); 738 if (rt == NULL) { 739 RIB_RUNLOCK(rnh); 740 return (ESRCH); 741 } 742 } 743 RT_LOCK(rt); 744 RT_ADDREF(rt); 745 RIB_RUNLOCK(rnh); 746 747 *ret_nrt = rt; 748 749 return (0); 750 } 751 752 /* 753 * Update sockaddrs, flags, etc in @prtm based on @rt data. 754 * Assumes @rt is locked. 755 * rtm can be reallocated. 756 * 757 * Returns 0 on success, along with pointer to (potentially reallocated) 758 * rtm. 759 * 760 */ 761 static int 762 update_rtm_from_rte(struct rt_addrinfo *info, struct rt_msghdr **prtm, 763 int alloc_len, struct rtentry *rt) 764 { 765 struct sockaddr_storage netmask_ss; 766 struct walkarg w; 767 union sockaddr_union saun; 768 struct rt_msghdr *rtm, *orig_rtm = NULL; 769 struct ifnet *ifp; 770 int error, len; 771 772 RT_LOCK_ASSERT(rt); 773 774 rtm = *prtm; 775 776 info->rti_info[RTAX_DST] = rt_key(rt); 777 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway; 778 info->rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt), 779 rt_mask(rt), &netmask_ss); 780 info->rti_info[RTAX_GENMASK] = 0; 781 ifp = rt->rt_ifp; 782 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) { 783 if (ifp) { 784 info->rti_info[RTAX_IFP] = 785 ifp->if_addr->ifa_addr; 786 error = rtm_get_jailed(info, ifp, rt, 787 &saun, curthread->td_ucred); 788 if (error != 0) 789 return (error); 790 if (ifp->if_flags & IFF_POINTOPOINT) 791 info->rti_info[RTAX_BRD] = 792 rt->rt_ifa->ifa_dstaddr; 793 rtm->rtm_index = ifp->if_index; 794 } else { 795 info->rti_info[RTAX_IFP] = NULL; 796 info->rti_info[RTAX_IFA] = NULL; 797 } 798 } else if (ifp != NULL) 799 rtm->rtm_index = ifp->if_index; 800 801 /* Check if we need to realloc storage */ 802 rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len); 803 if (len > alloc_len) { 804 struct rt_msghdr *tmp_rtm; 805 806 tmp_rtm = malloc(len, M_TEMP, M_NOWAIT); 807 if (tmp_rtm == NULL) 808 return (ENOBUFS); 809 bcopy(rtm, tmp_rtm, rtm->rtm_msglen); 810 orig_rtm = rtm; 811 rtm = tmp_rtm; 812 alloc_len = len; 813 814 /* 815 * Delay freeing original rtm as info contains 816 * data referencing it. 817 */ 818 } 819 820 w.w_tmem = (caddr_t)rtm; 821 w.w_tmemsize = alloc_len; 822 rtsock_msg_buffer(rtm->rtm_type, info, &w, &len); 823 824 if (rt->rt_flags & RTF_GWFLAG_COMPAT) 825 rtm->rtm_flags = RTF_GATEWAY | 826 (rt->rt_flags & ~RTF_GWFLAG_COMPAT); 827 else 828 rtm->rtm_flags = rt->rt_flags; 829 rt_getmetrics(rt, &rtm->rtm_rmx); 830 rtm->rtm_addrs = info->rti_addrs; 831 832 if (orig_rtm != NULL) 833 free(orig_rtm, M_TEMP); 834 *prtm = rtm; 835 836 return (0); 837 } 838 839 /*ARGSUSED*/ 840 static int 841 route_output(struct mbuf *m, struct socket *so, ...) 842 { 843 struct rt_msghdr *rtm = NULL; 844 struct rtentry *rt = NULL; 845 struct rt_addrinfo info; 846 struct epoch_tracker et; 847 #ifdef INET6 848 struct sockaddr_storage ss; 849 struct sockaddr_in6 *sin6; 850 int i, rti_need_deembed = 0; 851 #endif 852 int alloc_len = 0, len, error = 0, fibnum; 853 sa_family_t saf = AF_UNSPEC; 854 struct walkarg w; 855 856 fibnum = so->so_fibnum; 857 858 #define senderr(e) { error = e; goto flush;} 859 if (m == NULL || ((m->m_len < sizeof(long)) && 860 (m = m_pullup(m, sizeof(long))) == NULL)) 861 return (ENOBUFS); 862 if ((m->m_flags & M_PKTHDR) == 0) 863 panic("route_output"); 864 NET_EPOCH_ENTER(et); 865 len = m->m_pkthdr.len; 866 if (len < sizeof(*rtm) || 867 len != mtod(m, struct rt_msghdr *)->rtm_msglen) 868 senderr(EINVAL); 869 870 /* 871 * Most of current messages are in range 200-240 bytes, 872 * minimize possible re-allocation on reply using larger size 873 * buffer aligned on 1k boundaty. 874 */ 875 alloc_len = roundup2(len, 1024); 876 if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL) 877 senderr(ENOBUFS); 878 879 m_copydata(m, 0, len, (caddr_t)rtm); 880 bzero(&info, sizeof(info)); 881 bzero(&w, sizeof(w)); 882 883 if (rtm->rtm_version != RTM_VERSION) { 884 /* Do not touch message since format is unknown */ 885 free(rtm, M_TEMP); 886 rtm = NULL; 887 senderr(EPROTONOSUPPORT); 888 } 889 890 /* 891 * Starting from here, it is possible 892 * to alter original message and insert 893 * caller PID and error value. 894 */ 895 896 if ((error = fill_addrinfo(rtm, len, fibnum, &info)) != 0) { 897 senderr(error); 898 } 899 900 saf = info.rti_info[RTAX_DST]->sa_family; 901 902 /* support for new ARP code */ 903 if (rtm->rtm_flags & RTF_LLDATA) { 904 error = lla_rt_output(rtm, &info); 905 #ifdef INET6 906 if (error == 0) 907 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0; 908 #endif 909 goto flush; 910 } 911 912 switch (rtm->rtm_type) { 913 struct rtentry *saved_nrt; 914 915 case RTM_ADD: 916 case RTM_CHANGE: 917 if (rtm->rtm_type == RTM_ADD) { 918 if (info.rti_info[RTAX_GATEWAY] == NULL) 919 senderr(EINVAL); 920 } 921 saved_nrt = NULL; 922 error = rtrequest1_fib(rtm->rtm_type, &info, &saved_nrt, 923 fibnum); 924 if (error == 0 && saved_nrt != NULL) { 925 #ifdef INET6 926 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0; 927 #endif 928 RT_LOCK(saved_nrt); 929 rtm->rtm_index = saved_nrt->rt_ifp->if_index; 930 RT_REMREF(saved_nrt); 931 RT_UNLOCK(saved_nrt); 932 } 933 break; 934 935 case RTM_DELETE: 936 saved_nrt = NULL; 937 error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt, fibnum); 938 if (error == 0) { 939 RT_LOCK(saved_nrt); 940 rt = saved_nrt; 941 goto report; 942 } 943 #ifdef INET6 944 /* rt_msg2() will not be used when RTM_DELETE fails. */ 945 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0; 946 #endif 947 break; 948 949 case RTM_GET: 950 error = handle_rtm_get(&info, fibnum, rtm, &rt); 951 if (error != 0) 952 senderr(error); 953 954 report: 955 RT_LOCK_ASSERT(rt); 956 if (!can_export_rte(curthread->td_ucred, rt)) { 957 RT_UNLOCK(rt); 958 senderr(ESRCH); 959 } 960 error = update_rtm_from_rte(&info, &rtm, alloc_len, rt); 961 /* 962 * Note that some sockaddr pointers may have changed to 963 * point to memory outsize @rtm. Some may be pointing 964 * to the on-stack variables. 965 * Given that, any pointer in @info CANNOT BE USED. 966 */ 967 968 /* 969 * scopeid deembedding has been performed while 970 * writing updated rtm in rtsock_msg_buffer(). 971 * With that in mind, skip deembedding procedure below. 972 */ 973 #ifdef INET6 974 rti_need_deembed = 0; 975 #endif 976 RT_UNLOCK(rt); 977 if (error != 0) 978 senderr(error); 979 break; 980 981 default: 982 senderr(EOPNOTSUPP); 983 } 984 985 flush: 986 NET_EPOCH_EXIT(et); 987 if (rt != NULL) 988 RTFREE(rt); 989 990 #ifdef INET6 991 if (rtm != NULL) { 992 if (rti_need_deembed) { 993 /* sin6_scope_id is recovered before sending rtm. */ 994 sin6 = (struct sockaddr_in6 *)&ss; 995 for (i = 0; i < RTAX_MAX; i++) { 996 if (info.rti_info[i] == NULL) 997 continue; 998 if (info.rti_info[i]->sa_family != AF_INET6) 999 continue; 1000 bcopy(info.rti_info[i], sin6, sizeof(*sin6)); 1001 if (sa6_recoverscope(sin6) == 0) 1002 bcopy(sin6, info.rti_info[i], 1003 sizeof(*sin6)); 1004 } 1005 } 1006 } 1007 #endif 1008 send_rtm_reply(so, rtm, m, saf, fibnum, error); 1009 1010 return (error); 1011 } 1012 1013 /* 1014 * Sends the prepared reply message in @rtm to all rtsock clients. 1015 * Frees @m and @rtm. 1016 * 1017 */ 1018 static void 1019 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m, 1020 sa_family_t saf, u_int fibnum, int rtm_errno) 1021 { 1022 struct rawcb *rp = NULL; 1023 1024 /* 1025 * Check to see if we don't want our own messages. 1026 */ 1027 if ((so->so_options & SO_USELOOPBACK) == 0) { 1028 if (V_route_cb.any_count <= 1) { 1029 if (rtm != NULL) 1030 free(rtm, M_TEMP); 1031 m_freem(m); 1032 return; 1033 } 1034 /* There is another listener, so construct message */ 1035 rp = sotorawcb(so); 1036 } 1037 1038 if (rtm != NULL) { 1039 if (rtm_errno!= 0) 1040 rtm->rtm_errno = rtm_errno; 1041 else 1042 rtm->rtm_flags |= RTF_DONE; 1043 1044 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm); 1045 if (m->m_pkthdr.len < rtm->rtm_msglen) { 1046 m_freem(m); 1047 m = NULL; 1048 } else if (m->m_pkthdr.len > rtm->rtm_msglen) 1049 m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len); 1050 1051 free(rtm, M_TEMP); 1052 } 1053 if (m != NULL) { 1054 M_SETFIB(m, fibnum); 1055 m->m_flags |= RTS_FILTER_FIB; 1056 if (rp) { 1057 /* 1058 * XXX insure we don't get a copy by 1059 * invalidating our protocol 1060 */ 1061 unsigned short family = rp->rcb_proto.sp_family; 1062 rp->rcb_proto.sp_family = 0; 1063 rt_dispatch(m, saf); 1064 rp->rcb_proto.sp_family = family; 1065 } else 1066 rt_dispatch(m, saf); 1067 } 1068 } 1069 1070 1071 static void 1072 rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out) 1073 { 1074 1075 bzero(out, sizeof(*out)); 1076 out->rmx_mtu = rt->rt_mtu; 1077 out->rmx_weight = rt->rt_weight; 1078 out->rmx_pksent = counter_u64_fetch(rt->rt_pksent); 1079 /* Kernel -> userland timebase conversion. */ 1080 out->rmx_expire = rt->rt_expire ? 1081 rt->rt_expire - time_uptime + time_second : 0; 1082 } 1083 1084 /* 1085 * Extract the addresses of the passed sockaddrs. 1086 * Do a little sanity checking so as to avoid bad memory references. 1087 * This data is derived straight from userland. 1088 */ 1089 static int 1090 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo) 1091 { 1092 struct sockaddr *sa; 1093 int i; 1094 1095 for (i = 0; i < RTAX_MAX && cp < cplim; i++) { 1096 if ((rtinfo->rti_addrs & (1 << i)) == 0) 1097 continue; 1098 sa = (struct sockaddr *)cp; 1099 /* 1100 * It won't fit. 1101 */ 1102 if (cp + sa->sa_len > cplim) 1103 return (EINVAL); 1104 /* 1105 * there are no more.. quit now 1106 * If there are more bits, they are in error. 1107 * I've seen this. route(1) can evidently generate these. 1108 * This causes kernel to core dump. 1109 * for compatibility, If we see this, point to a safe address. 1110 */ 1111 if (sa->sa_len == 0) { 1112 rtinfo->rti_info[i] = &sa_zero; 1113 return (0); /* should be EINVAL but for compat */ 1114 } 1115 /* accept it */ 1116 #ifdef INET6 1117 if (sa->sa_family == AF_INET6) 1118 sa6_embedscope((struct sockaddr_in6 *)sa, 1119 V_ip6_use_defzone); 1120 #endif 1121 rtinfo->rti_info[i] = sa; 1122 cp += SA_SIZE(sa); 1123 } 1124 return (0); 1125 } 1126 1127 /* 1128 * Fill in @dmask with valid netmask leaving original @smask 1129 * intact. Mostly used with radix netmasks. 1130 */ 1131 static struct sockaddr * 1132 rtsock_fix_netmask(struct sockaddr *dst, struct sockaddr *smask, 1133 struct sockaddr_storage *dmask) 1134 { 1135 if (dst == NULL || smask == NULL) 1136 return (NULL); 1137 1138 memset(dmask, 0, dst->sa_len); 1139 memcpy(dmask, smask, smask->sa_len); 1140 dmask->ss_len = dst->sa_len; 1141 dmask->ss_family = dst->sa_family; 1142 1143 return ((struct sockaddr *)dmask); 1144 } 1145 1146 /* 1147 * Writes information related to @rtinfo object to newly-allocated mbuf. 1148 * Assumes MCLBYTES is enough to construct any message. 1149 * Used for OS notifications of vaious events (if/ifa announces,etc) 1150 * 1151 * Returns allocated mbuf or NULL on failure. 1152 */ 1153 static struct mbuf * 1154 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo) 1155 { 1156 struct rt_msghdr *rtm; 1157 struct mbuf *m; 1158 int i; 1159 struct sockaddr *sa; 1160 #ifdef INET6 1161 struct sockaddr_storage ss; 1162 struct sockaddr_in6 *sin6; 1163 #endif 1164 int len, dlen; 1165 1166 switch (type) { 1167 1168 case RTM_DELADDR: 1169 case RTM_NEWADDR: 1170 len = sizeof(struct ifa_msghdr); 1171 break; 1172 1173 case RTM_DELMADDR: 1174 case RTM_NEWMADDR: 1175 len = sizeof(struct ifma_msghdr); 1176 break; 1177 1178 case RTM_IFINFO: 1179 len = sizeof(struct if_msghdr); 1180 break; 1181 1182 case RTM_IFANNOUNCE: 1183 case RTM_IEEE80211: 1184 len = sizeof(struct if_announcemsghdr); 1185 break; 1186 1187 default: 1188 len = sizeof(struct rt_msghdr); 1189 } 1190 1191 /* XXXGL: can we use MJUMPAGESIZE cluster here? */ 1192 KASSERT(len <= MCLBYTES, ("%s: message too big", __func__)); 1193 if (len > MHLEN) 1194 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 1195 else 1196 m = m_gethdr(M_NOWAIT, MT_DATA); 1197 if (m == NULL) 1198 return (m); 1199 1200 m->m_pkthdr.len = m->m_len = len; 1201 rtm = mtod(m, struct rt_msghdr *); 1202 bzero((caddr_t)rtm, len); 1203 for (i = 0; i < RTAX_MAX; i++) { 1204 if ((sa = rtinfo->rti_info[i]) == NULL) 1205 continue; 1206 rtinfo->rti_addrs |= (1 << i); 1207 dlen = SA_SIZE(sa); 1208 #ifdef INET6 1209 if (V_deembed_scopeid && sa->sa_family == AF_INET6) { 1210 sin6 = (struct sockaddr_in6 *)&ss; 1211 bcopy(sa, sin6, sizeof(*sin6)); 1212 if (sa6_recoverscope(sin6) == 0) 1213 sa = (struct sockaddr *)sin6; 1214 } 1215 #endif 1216 m_copyback(m, len, dlen, (caddr_t)sa); 1217 len += dlen; 1218 } 1219 if (m->m_pkthdr.len != len) { 1220 m_freem(m); 1221 return (NULL); 1222 } 1223 rtm->rtm_msglen = len; 1224 rtm->rtm_version = RTM_VERSION; 1225 rtm->rtm_type = type; 1226 return (m); 1227 } 1228 1229 /* 1230 * Writes information related to @rtinfo object to preallocated buffer. 1231 * Stores needed size in @plen. If @w is NULL, calculates size without 1232 * writing. 1233 * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation. 1234 * 1235 * Returns 0 on success. 1236 * 1237 */ 1238 static int 1239 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen) 1240 { 1241 int i; 1242 int len, buflen = 0, dlen; 1243 caddr_t cp = NULL; 1244 struct rt_msghdr *rtm = NULL; 1245 #ifdef INET6 1246 struct sockaddr_storage ss; 1247 struct sockaddr_in6 *sin6; 1248 #endif 1249 #ifdef COMPAT_FREEBSD32 1250 bool compat32 = false; 1251 #endif 1252 1253 switch (type) { 1254 1255 case RTM_DELADDR: 1256 case RTM_NEWADDR: 1257 if (w != NULL && w->w_op == NET_RT_IFLISTL) { 1258 #ifdef COMPAT_FREEBSD32 1259 if (w->w_req->flags & SCTL_MASK32) { 1260 len = sizeof(struct ifa_msghdrl32); 1261 compat32 = true; 1262 } else 1263 #endif 1264 len = sizeof(struct ifa_msghdrl); 1265 } else 1266 len = sizeof(struct ifa_msghdr); 1267 break; 1268 1269 case RTM_IFINFO: 1270 #ifdef COMPAT_FREEBSD32 1271 if (w != NULL && w->w_req->flags & SCTL_MASK32) { 1272 if (w->w_op == NET_RT_IFLISTL) 1273 len = sizeof(struct if_msghdrl32); 1274 else 1275 len = sizeof(struct if_msghdr32); 1276 compat32 = true; 1277 break; 1278 } 1279 #endif 1280 if (w != NULL && w->w_op == NET_RT_IFLISTL) 1281 len = sizeof(struct if_msghdrl); 1282 else 1283 len = sizeof(struct if_msghdr); 1284 break; 1285 1286 case RTM_NEWMADDR: 1287 len = sizeof(struct ifma_msghdr); 1288 break; 1289 1290 default: 1291 len = sizeof(struct rt_msghdr); 1292 } 1293 1294 if (w != NULL) { 1295 rtm = (struct rt_msghdr *)w->w_tmem; 1296 buflen = w->w_tmemsize - len; 1297 cp = (caddr_t)w->w_tmem + len; 1298 } 1299 1300 rtinfo->rti_addrs = 0; 1301 for (i = 0; i < RTAX_MAX; i++) { 1302 struct sockaddr *sa; 1303 1304 if ((sa = rtinfo->rti_info[i]) == NULL) 1305 continue; 1306 rtinfo->rti_addrs |= (1 << i); 1307 #ifdef COMPAT_FREEBSD32 1308 if (compat32) 1309 dlen = SA_SIZE32(sa); 1310 else 1311 #endif 1312 dlen = SA_SIZE(sa); 1313 if (cp != NULL && buflen >= dlen) { 1314 #ifdef INET6 1315 if (V_deembed_scopeid && sa->sa_family == AF_INET6) { 1316 sin6 = (struct sockaddr_in6 *)&ss; 1317 bcopy(sa, sin6, sizeof(*sin6)); 1318 if (sa6_recoverscope(sin6) == 0) 1319 sa = (struct sockaddr *)sin6; 1320 } 1321 #endif 1322 bcopy((caddr_t)sa, cp, (unsigned)dlen); 1323 cp += dlen; 1324 buflen -= dlen; 1325 } else if (cp != NULL) { 1326 /* 1327 * Buffer too small. Count needed size 1328 * and return with error. 1329 */ 1330 cp = NULL; 1331 } 1332 1333 len += dlen; 1334 } 1335 1336 if (cp != NULL) { 1337 dlen = ALIGN(len) - len; 1338 if (buflen < dlen) 1339 cp = NULL; 1340 else { 1341 bzero(cp, dlen); 1342 cp += dlen; 1343 buflen -= dlen; 1344 } 1345 } 1346 len = ALIGN(len); 1347 1348 if (cp != NULL) { 1349 /* fill header iff buffer is large enough */ 1350 rtm->rtm_version = RTM_VERSION; 1351 rtm->rtm_type = type; 1352 rtm->rtm_msglen = len; 1353 } 1354 1355 *plen = len; 1356 1357 if (w != NULL && cp == NULL) 1358 return (ENOBUFS); 1359 1360 return (0); 1361 } 1362 1363 /* 1364 * This routine is called to generate a message from the routing 1365 * socket indicating that a redirect has occurred, a routing lookup 1366 * has failed, or that a protocol has detected timeouts to a particular 1367 * destination. 1368 */ 1369 void 1370 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error, 1371 int fibnum) 1372 { 1373 struct rt_msghdr *rtm; 1374 struct mbuf *m; 1375 struct sockaddr *sa = rtinfo->rti_info[RTAX_DST]; 1376 1377 if (V_route_cb.any_count == 0) 1378 return; 1379 m = rtsock_msg_mbuf(type, rtinfo); 1380 if (m == NULL) 1381 return; 1382 1383 if (fibnum != RT_ALL_FIBS) { 1384 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out " 1385 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs)); 1386 M_SETFIB(m, fibnum); 1387 m->m_flags |= RTS_FILTER_FIB; 1388 } 1389 1390 rtm = mtod(m, struct rt_msghdr *); 1391 rtm->rtm_flags = RTF_DONE | flags; 1392 rtm->rtm_errno = error; 1393 rtm->rtm_addrs = rtinfo->rti_addrs; 1394 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1395 } 1396 1397 void 1398 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error) 1399 { 1400 1401 rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS); 1402 } 1403 1404 /* 1405 * This routine is called to generate a message from the routing 1406 * socket indicating that the status of a network interface has changed. 1407 */ 1408 void 1409 rt_ifmsg(struct ifnet *ifp) 1410 { 1411 struct if_msghdr *ifm; 1412 struct mbuf *m; 1413 struct rt_addrinfo info; 1414 1415 if (V_route_cb.any_count == 0) 1416 return; 1417 bzero((caddr_t)&info, sizeof(info)); 1418 m = rtsock_msg_mbuf(RTM_IFINFO, &info); 1419 if (m == NULL) 1420 return; 1421 ifm = mtod(m, struct if_msghdr *); 1422 ifm->ifm_index = ifp->if_index; 1423 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1424 if_data_copy(ifp, &ifm->ifm_data); 1425 ifm->ifm_addrs = 0; 1426 rt_dispatch(m, AF_UNSPEC); 1427 } 1428 1429 /* 1430 * Announce interface address arrival/withdraw. 1431 * Please do not call directly, use rt_addrmsg(). 1432 * Assume input data to be valid. 1433 * Returns 0 on success. 1434 */ 1435 int 1436 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum) 1437 { 1438 struct rt_addrinfo info; 1439 struct sockaddr *sa; 1440 int ncmd; 1441 struct mbuf *m; 1442 struct ifa_msghdr *ifam; 1443 struct ifnet *ifp = ifa->ifa_ifp; 1444 struct sockaddr_storage ss; 1445 1446 if (V_route_cb.any_count == 0) 1447 return (0); 1448 1449 ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR; 1450 1451 bzero((caddr_t)&info, sizeof(info)); 1452 info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr; 1453 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr; 1454 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask( 1455 info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss); 1456 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr; 1457 if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL) 1458 return (ENOBUFS); 1459 ifam = mtod(m, struct ifa_msghdr *); 1460 ifam->ifam_index = ifp->if_index; 1461 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 1462 ifam->ifam_flags = ifa->ifa_flags; 1463 ifam->ifam_addrs = info.rti_addrs; 1464 1465 if (fibnum != RT_ALL_FIBS) { 1466 M_SETFIB(m, fibnum); 1467 m->m_flags |= RTS_FILTER_FIB; 1468 } 1469 1470 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1471 1472 return (0); 1473 } 1474 1475 /* 1476 * Announce route addition/removal to rtsock based on @rt data. 1477 * Callers are advives to use rt_routemsg() instead of using this 1478 * function directly. 1479 * Assume @rt data is consistent. 1480 * 1481 * Returns 0 on success. 1482 */ 1483 int 1484 rtsock_routemsg(int cmd, struct rtentry *rt, struct ifnet *ifp, int rti_addrs, 1485 int fibnum) 1486 { 1487 struct sockaddr_storage ss; 1488 struct rt_addrinfo info; 1489 1490 if (V_route_cb.any_count == 0) 1491 return (0); 1492 1493 bzero((caddr_t)&info, sizeof(info)); 1494 info.rti_info[RTAX_DST] = rt_key(rt); 1495 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt), rt_mask(rt), &ss); 1496 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 1497 info.rti_flags = rt->rt_flags; 1498 info.rti_ifp = ifp; 1499 1500 return (rtsock_routemsg_info(cmd, &info, fibnum)); 1501 } 1502 1503 int 1504 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum) 1505 { 1506 struct rt_msghdr *rtm; 1507 struct sockaddr *sa; 1508 struct mbuf *m; 1509 1510 if (V_route_cb.any_count == 0) 1511 return (0); 1512 1513 if (info->rti_flags & RTF_HOST) 1514 info->rti_info[RTAX_NETMASK] = NULL; 1515 1516 m = rtsock_msg_mbuf(cmd, info); 1517 if (m == NULL) 1518 return (ENOBUFS); 1519 1520 if (fibnum != RT_ALL_FIBS) { 1521 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out " 1522 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs)); 1523 M_SETFIB(m, fibnum); 1524 m->m_flags |= RTS_FILTER_FIB; 1525 } 1526 1527 rtm = mtod(m, struct rt_msghdr *); 1528 rtm->rtm_addrs = info->rti_addrs; 1529 if (info->rti_ifp != NULL) 1530 rtm->rtm_index = info->rti_ifp->if_index; 1531 /* Add RTF_DONE to indicate command 'completion' required by API */ 1532 info->rti_flags |= RTF_DONE; 1533 /* Reported routes has to be up */ 1534 if (cmd == RTM_ADD || cmd == RTM_CHANGE) 1535 info->rti_flags |= RTF_UP; 1536 rtm->rtm_flags = info->rti_flags; 1537 1538 sa = info->rti_info[RTAX_DST]; 1539 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1540 1541 return (0); 1542 } 1543 1544 /* 1545 * This is the analogue to the rt_newaddrmsg which performs the same 1546 * function but for multicast group memberhips. This is easier since 1547 * there is no route state to worry about. 1548 */ 1549 void 1550 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma) 1551 { 1552 struct rt_addrinfo info; 1553 struct mbuf *m = NULL; 1554 struct ifnet *ifp = ifma->ifma_ifp; 1555 struct ifma_msghdr *ifmam; 1556 1557 if (V_route_cb.any_count == 0) 1558 return; 1559 1560 bzero((caddr_t)&info, sizeof(info)); 1561 info.rti_info[RTAX_IFA] = ifma->ifma_addr; 1562 if (ifp && ifp->if_addr) 1563 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr; 1564 else 1565 info.rti_info[RTAX_IFP] = NULL; 1566 /* 1567 * If a link-layer address is present, present it as a ``gateway'' 1568 * (similarly to how ARP entries, e.g., are presented). 1569 */ 1570 info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr; 1571 m = rtsock_msg_mbuf(cmd, &info); 1572 if (m == NULL) 1573 return; 1574 ifmam = mtod(m, struct ifma_msghdr *); 1575 KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n", 1576 __func__)); 1577 ifmam->ifmam_index = ifp->if_index; 1578 ifmam->ifmam_addrs = info.rti_addrs; 1579 rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC); 1580 } 1581 1582 static struct mbuf * 1583 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what, 1584 struct rt_addrinfo *info) 1585 { 1586 struct if_announcemsghdr *ifan; 1587 struct mbuf *m; 1588 1589 if (V_route_cb.any_count == 0) 1590 return NULL; 1591 bzero((caddr_t)info, sizeof(*info)); 1592 m = rtsock_msg_mbuf(type, info); 1593 if (m != NULL) { 1594 ifan = mtod(m, struct if_announcemsghdr *); 1595 ifan->ifan_index = ifp->if_index; 1596 strlcpy(ifan->ifan_name, ifp->if_xname, 1597 sizeof(ifan->ifan_name)); 1598 ifan->ifan_what = what; 1599 } 1600 return m; 1601 } 1602 1603 /* 1604 * This is called to generate routing socket messages indicating 1605 * IEEE80211 wireless events. 1606 * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way. 1607 */ 1608 void 1609 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len) 1610 { 1611 struct mbuf *m; 1612 struct rt_addrinfo info; 1613 1614 m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info); 1615 if (m != NULL) { 1616 /* 1617 * Append the ieee80211 data. Try to stick it in the 1618 * mbuf containing the ifannounce msg; otherwise allocate 1619 * a new mbuf and append. 1620 * 1621 * NB: we assume m is a single mbuf. 1622 */ 1623 if (data_len > M_TRAILINGSPACE(m)) { 1624 struct mbuf *n = m_get(M_NOWAIT, MT_DATA); 1625 if (n == NULL) { 1626 m_freem(m); 1627 return; 1628 } 1629 bcopy(data, mtod(n, void *), data_len); 1630 n->m_len = data_len; 1631 m->m_next = n; 1632 } else if (data_len > 0) { 1633 bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len); 1634 m->m_len += data_len; 1635 } 1636 if (m->m_flags & M_PKTHDR) 1637 m->m_pkthdr.len += data_len; 1638 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len; 1639 rt_dispatch(m, AF_UNSPEC); 1640 } 1641 } 1642 1643 /* 1644 * This is called to generate routing socket messages indicating 1645 * network interface arrival and departure. 1646 */ 1647 void 1648 rt_ifannouncemsg(struct ifnet *ifp, int what) 1649 { 1650 struct mbuf *m; 1651 struct rt_addrinfo info; 1652 1653 m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info); 1654 if (m != NULL) 1655 rt_dispatch(m, AF_UNSPEC); 1656 } 1657 1658 static void 1659 rt_dispatch(struct mbuf *m, sa_family_t saf) 1660 { 1661 struct m_tag *tag; 1662 1663 /* 1664 * Preserve the family from the sockaddr, if any, in an m_tag for 1665 * use when injecting the mbuf into the routing socket buffer from 1666 * the netisr. 1667 */ 1668 if (saf != AF_UNSPEC) { 1669 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short), 1670 M_NOWAIT); 1671 if (tag == NULL) { 1672 m_freem(m); 1673 return; 1674 } 1675 *(unsigned short *)(tag + 1) = saf; 1676 m_tag_prepend(m, tag); 1677 } 1678 #ifdef VIMAGE 1679 if (V_loif) 1680 m->m_pkthdr.rcvif = V_loif; 1681 else { 1682 m_freem(m); 1683 return; 1684 } 1685 #endif 1686 netisr_queue(NETISR_ROUTE, m); /* mbuf is free'd on failure. */ 1687 } 1688 1689 /* 1690 * Checks if rte can be exported v.r.t jails/vnets. 1691 * 1692 * Returns 1 if it can, 0 otherwise. 1693 */ 1694 static int 1695 can_export_rte(struct ucred *td_ucred, const struct rtentry *rt) 1696 { 1697 1698 if ((rt->rt_flags & RTF_HOST) == 0 1699 ? jailed_without_vnet(td_ucred) 1700 : prison_if(td_ucred, rt_key_const(rt)) != 0) 1701 return (0); 1702 return (1); 1703 } 1704 1705 /* 1706 * This is used in dumping the kernel table via sysctl(). 1707 */ 1708 static int 1709 sysctl_dumpentry(struct radix_node *rn, void *vw) 1710 { 1711 struct walkarg *w = vw; 1712 struct rtentry *rt = (struct rtentry *)rn; 1713 int error = 0, size; 1714 struct rt_addrinfo info; 1715 struct sockaddr_storage ss; 1716 1717 NET_EPOCH_ASSERT(); 1718 1719 if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg)) 1720 return 0; 1721 if (!can_export_rte(w->w_req->td->td_ucred, rt)) 1722 return (0); 1723 bzero((caddr_t)&info, sizeof(info)); 1724 info.rti_info[RTAX_DST] = rt_key(rt); 1725 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 1726 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt), 1727 rt_mask(rt), &ss); 1728 info.rti_info[RTAX_GENMASK] = 0; 1729 if (rt->rt_ifp && !(rt->rt_ifp->if_flags & IFF_DYING)) { 1730 info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr; 1731 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr; 1732 if (rt->rt_ifp->if_flags & IFF_POINTOPOINT) 1733 info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr; 1734 } 1735 if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0) 1736 return (error); 1737 if (w->w_req && w->w_tmem) { 1738 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem; 1739 1740 bzero(&rtm->rtm_index, 1741 sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index)); 1742 if (rt->rt_flags & RTF_GWFLAG_COMPAT) 1743 rtm->rtm_flags = RTF_GATEWAY | 1744 (rt->rt_flags & ~RTF_GWFLAG_COMPAT); 1745 else 1746 rtm->rtm_flags = rt->rt_flags; 1747 rt_getmetrics(rt, &rtm->rtm_rmx); 1748 rtm->rtm_index = rt->rt_ifp->if_index; 1749 rtm->rtm_addrs = info.rti_addrs; 1750 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size); 1751 return (error); 1752 } 1753 return (error); 1754 } 1755 1756 static int 1757 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd, 1758 struct rt_addrinfo *info, struct walkarg *w, int len) 1759 { 1760 struct if_msghdrl *ifm; 1761 struct if_data *ifd; 1762 1763 ifm = (struct if_msghdrl *)w->w_tmem; 1764 1765 #ifdef COMPAT_FREEBSD32 1766 if (w->w_req->flags & SCTL_MASK32) { 1767 struct if_msghdrl32 *ifm32; 1768 1769 ifm32 = (struct if_msghdrl32 *)ifm; 1770 ifm32->ifm_addrs = info->rti_addrs; 1771 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1772 ifm32->ifm_index = ifp->if_index; 1773 ifm32->_ifm_spare1 = 0; 1774 ifm32->ifm_len = sizeof(*ifm32); 1775 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data); 1776 ifm32->_ifm_spare2 = 0; 1777 ifd = &ifm32->ifm_data; 1778 } else 1779 #endif 1780 { 1781 ifm->ifm_addrs = info->rti_addrs; 1782 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1783 ifm->ifm_index = ifp->if_index; 1784 ifm->_ifm_spare1 = 0; 1785 ifm->ifm_len = sizeof(*ifm); 1786 ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data); 1787 ifm->_ifm_spare2 = 0; 1788 ifd = &ifm->ifm_data; 1789 } 1790 1791 memcpy(ifd, src_ifd, sizeof(*ifd)); 1792 1793 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len)); 1794 } 1795 1796 static int 1797 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd, 1798 struct rt_addrinfo *info, struct walkarg *w, int len) 1799 { 1800 struct if_msghdr *ifm; 1801 struct if_data *ifd; 1802 1803 ifm = (struct if_msghdr *)w->w_tmem; 1804 1805 #ifdef COMPAT_FREEBSD32 1806 if (w->w_req->flags & SCTL_MASK32) { 1807 struct if_msghdr32 *ifm32; 1808 1809 ifm32 = (struct if_msghdr32 *)ifm; 1810 ifm32->ifm_addrs = info->rti_addrs; 1811 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1812 ifm32->ifm_index = ifp->if_index; 1813 ifm32->_ifm_spare1 = 0; 1814 ifd = &ifm32->ifm_data; 1815 } else 1816 #endif 1817 { 1818 ifm->ifm_addrs = info->rti_addrs; 1819 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1820 ifm->ifm_index = ifp->if_index; 1821 ifm->_ifm_spare1 = 0; 1822 ifd = &ifm->ifm_data; 1823 } 1824 1825 memcpy(ifd, src_ifd, sizeof(*ifd)); 1826 1827 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len)); 1828 } 1829 1830 static int 1831 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info, 1832 struct walkarg *w, int len) 1833 { 1834 struct ifa_msghdrl *ifam; 1835 struct if_data *ifd; 1836 1837 ifam = (struct ifa_msghdrl *)w->w_tmem; 1838 1839 #ifdef COMPAT_FREEBSD32 1840 if (w->w_req->flags & SCTL_MASK32) { 1841 struct ifa_msghdrl32 *ifam32; 1842 1843 ifam32 = (struct ifa_msghdrl32 *)ifam; 1844 ifam32->ifam_addrs = info->rti_addrs; 1845 ifam32->ifam_flags = ifa->ifa_flags; 1846 ifam32->ifam_index = ifa->ifa_ifp->if_index; 1847 ifam32->_ifam_spare1 = 0; 1848 ifam32->ifam_len = sizeof(*ifam32); 1849 ifam32->ifam_data_off = 1850 offsetof(struct ifa_msghdrl32, ifam_data); 1851 ifam32->ifam_metric = ifa->ifa_ifp->if_metric; 1852 ifd = &ifam32->ifam_data; 1853 } else 1854 #endif 1855 { 1856 ifam->ifam_addrs = info->rti_addrs; 1857 ifam->ifam_flags = ifa->ifa_flags; 1858 ifam->ifam_index = ifa->ifa_ifp->if_index; 1859 ifam->_ifam_spare1 = 0; 1860 ifam->ifam_len = sizeof(*ifam); 1861 ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data); 1862 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 1863 ifd = &ifam->ifam_data; 1864 } 1865 1866 bzero(ifd, sizeof(*ifd)); 1867 ifd->ifi_datalen = sizeof(struct if_data); 1868 ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets); 1869 ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets); 1870 ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes); 1871 ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes); 1872 1873 /* Fixup if_data carp(4) vhid. */ 1874 if (carp_get_vhid_p != NULL) 1875 ifd->ifi_vhid = (*carp_get_vhid_p)(ifa); 1876 1877 return (SYSCTL_OUT(w->w_req, w->w_tmem, len)); 1878 } 1879 1880 static int 1881 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info, 1882 struct walkarg *w, int len) 1883 { 1884 struct ifa_msghdr *ifam; 1885 1886 ifam = (struct ifa_msghdr *)w->w_tmem; 1887 ifam->ifam_addrs = info->rti_addrs; 1888 ifam->ifam_flags = ifa->ifa_flags; 1889 ifam->ifam_index = ifa->ifa_ifp->if_index; 1890 ifam->_ifam_spare1 = 0; 1891 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 1892 1893 return (SYSCTL_OUT(w->w_req, w->w_tmem, len)); 1894 } 1895 1896 static int 1897 sysctl_iflist(int af, struct walkarg *w) 1898 { 1899 struct ifnet *ifp; 1900 struct ifaddr *ifa; 1901 struct if_data ifd; 1902 struct rt_addrinfo info; 1903 int len, error = 0; 1904 struct sockaddr_storage ss; 1905 1906 bzero((caddr_t)&info, sizeof(info)); 1907 bzero(&ifd, sizeof(ifd)); 1908 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1909 if (w->w_arg && w->w_arg != ifp->if_index) 1910 continue; 1911 if_data_copy(ifp, &ifd); 1912 ifa = ifp->if_addr; 1913 info.rti_info[RTAX_IFP] = ifa->ifa_addr; 1914 error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len); 1915 if (error != 0) 1916 goto done; 1917 info.rti_info[RTAX_IFP] = NULL; 1918 if (w->w_req && w->w_tmem) { 1919 if (w->w_op == NET_RT_IFLISTL) 1920 error = sysctl_iflist_ifml(ifp, &ifd, &info, w, 1921 len); 1922 else 1923 error = sysctl_iflist_ifm(ifp, &ifd, &info, w, 1924 len); 1925 if (error) 1926 goto done; 1927 } 1928 while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) { 1929 if (af && af != ifa->ifa_addr->sa_family) 1930 continue; 1931 if (prison_if(w->w_req->td->td_ucred, 1932 ifa->ifa_addr) != 0) 1933 continue; 1934 info.rti_info[RTAX_IFA] = ifa->ifa_addr; 1935 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask( 1936 ifa->ifa_addr, ifa->ifa_netmask, &ss); 1937 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr; 1938 error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len); 1939 if (error != 0) 1940 goto done; 1941 if (w->w_req && w->w_tmem) { 1942 if (w->w_op == NET_RT_IFLISTL) 1943 error = sysctl_iflist_ifaml(ifa, &info, 1944 w, len); 1945 else 1946 error = sysctl_iflist_ifam(ifa, &info, 1947 w, len); 1948 if (error) 1949 goto done; 1950 } 1951 } 1952 info.rti_info[RTAX_IFA] = NULL; 1953 info.rti_info[RTAX_NETMASK] = NULL; 1954 info.rti_info[RTAX_BRD] = NULL; 1955 } 1956 done: 1957 return (error); 1958 } 1959 1960 static int 1961 sysctl_ifmalist(int af, struct walkarg *w) 1962 { 1963 struct rt_addrinfo info; 1964 struct ifaddr *ifa; 1965 struct ifmultiaddr *ifma; 1966 struct ifnet *ifp; 1967 int error, len; 1968 1969 NET_EPOCH_ASSERT(); 1970 1971 error = 0; 1972 bzero((caddr_t)&info, sizeof(info)); 1973 1974 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1975 if (w->w_arg && w->w_arg != ifp->if_index) 1976 continue; 1977 ifa = ifp->if_addr; 1978 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL; 1979 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 1980 if (af && af != ifma->ifma_addr->sa_family) 1981 continue; 1982 if (prison_if(w->w_req->td->td_ucred, 1983 ifma->ifma_addr) != 0) 1984 continue; 1985 info.rti_info[RTAX_IFA] = ifma->ifma_addr; 1986 info.rti_info[RTAX_GATEWAY] = 1987 (ifma->ifma_addr->sa_family != AF_LINK) ? 1988 ifma->ifma_lladdr : NULL; 1989 error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len); 1990 if (error != 0) 1991 break; 1992 if (w->w_req && w->w_tmem) { 1993 struct ifma_msghdr *ifmam; 1994 1995 ifmam = (struct ifma_msghdr *)w->w_tmem; 1996 ifmam->ifmam_index = ifma->ifma_ifp->if_index; 1997 ifmam->ifmam_flags = 0; 1998 ifmam->ifmam_addrs = info.rti_addrs; 1999 ifmam->_ifmam_spare1 = 0; 2000 error = SYSCTL_OUT(w->w_req, w->w_tmem, len); 2001 if (error != 0) 2002 break; 2003 } 2004 } 2005 if (error != 0) 2006 break; 2007 } 2008 return (error); 2009 } 2010 2011 static int 2012 sysctl_rtsock(SYSCTL_HANDLER_ARGS) 2013 { 2014 RIB_RLOCK_TRACKER; 2015 struct epoch_tracker et; 2016 int *name = (int *)arg1; 2017 u_int namelen = arg2; 2018 struct rib_head *rnh = NULL; /* silence compiler. */ 2019 int i, lim, error = EINVAL; 2020 int fib = 0; 2021 u_char af; 2022 struct walkarg w; 2023 2024 name ++; 2025 namelen--; 2026 if (req->newptr) 2027 return (EPERM); 2028 if (name[1] == NET_RT_DUMP) { 2029 if (namelen == 3) 2030 fib = req->td->td_proc->p_fibnum; 2031 else if (namelen == 4) 2032 fib = (name[3] == RT_ALL_FIBS) ? 2033 req->td->td_proc->p_fibnum : name[3]; 2034 else 2035 return ((namelen < 3) ? EISDIR : ENOTDIR); 2036 if (fib < 0 || fib >= rt_numfibs) 2037 return (EINVAL); 2038 } else if (namelen != 3) 2039 return ((namelen < 3) ? EISDIR : ENOTDIR); 2040 af = name[0]; 2041 if (af > AF_MAX) 2042 return (EINVAL); 2043 bzero(&w, sizeof(w)); 2044 w.w_op = name[1]; 2045 w.w_arg = name[2]; 2046 w.w_req = req; 2047 2048 error = sysctl_wire_old_buffer(req, 0); 2049 if (error) 2050 return (error); 2051 2052 /* 2053 * Allocate reply buffer in advance. 2054 * All rtsock messages has maximum length of u_short. 2055 */ 2056 w.w_tmemsize = 65536; 2057 w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK); 2058 2059 NET_EPOCH_ENTER(et); 2060 switch (w.w_op) { 2061 case NET_RT_DUMP: 2062 case NET_RT_FLAGS: 2063 if (af == 0) { /* dump all tables */ 2064 i = 1; 2065 lim = AF_MAX; 2066 } else /* dump only one table */ 2067 i = lim = af; 2068 2069 /* 2070 * take care of llinfo entries, the caller must 2071 * specify an AF 2072 */ 2073 if (w.w_op == NET_RT_FLAGS && 2074 (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) { 2075 if (af != 0) 2076 error = lltable_sysctl_dumparp(af, w.w_req); 2077 else 2078 error = EINVAL; 2079 break; 2080 } 2081 /* 2082 * take care of routing entries 2083 */ 2084 for (error = 0; error == 0 && i <= lim; i++) { 2085 rnh = rt_tables_get_rnh(fib, i); 2086 if (rnh != NULL) { 2087 RIB_RLOCK(rnh); 2088 error = rnh->rnh_walktree(&rnh->head, 2089 sysctl_dumpentry, &w); 2090 RIB_RUNLOCK(rnh); 2091 } else if (af != 0) 2092 error = EAFNOSUPPORT; 2093 } 2094 break; 2095 2096 case NET_RT_IFLIST: 2097 case NET_RT_IFLISTL: 2098 error = sysctl_iflist(af, &w); 2099 break; 2100 2101 case NET_RT_IFMALIST: 2102 error = sysctl_ifmalist(af, &w); 2103 break; 2104 } 2105 NET_EPOCH_EXIT(et); 2106 2107 free(w.w_tmem, M_TEMP); 2108 return (error); 2109 } 2110 2111 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE, 2112 sysctl_rtsock, "Return route tables and interface/address lists"); 2113 2114 /* 2115 * Definitions of protocols supported in the ROUTE domain. 2116 */ 2117 2118 static struct domain routedomain; /* or at least forward */ 2119 2120 static struct protosw routesw[] = { 2121 { 2122 .pr_type = SOCK_RAW, 2123 .pr_domain = &routedomain, 2124 .pr_flags = PR_ATOMIC|PR_ADDR, 2125 .pr_output = route_output, 2126 .pr_ctlinput = raw_ctlinput, 2127 .pr_init = raw_init, 2128 .pr_usrreqs = &route_usrreqs 2129 } 2130 }; 2131 2132 static struct domain routedomain = { 2133 .dom_family = PF_ROUTE, 2134 .dom_name = "route", 2135 .dom_protosw = routesw, 2136 .dom_protoswNPROTOSW = &routesw[nitems(routesw)] 2137 }; 2138 2139 VNET_DOMAIN_SET(route); 2140 2141 #ifdef DDB 2142 /* 2143 * Unfortunately, RTF_ values are expressed as raw masks rather than powers of 2144 * 2, so we cannot use them as nice C99 initializer indices below. 2145 */ 2146 static const char * const rtf_flag_strings[] = { 2147 "UP", 2148 "GATEWAY", 2149 "HOST", 2150 "REJECT", 2151 "DYNAMIC", 2152 "MODIFIED", 2153 "DONE", 2154 "UNUSED_0x80", 2155 "UNUSED_0x100", 2156 "XRESOLVE", 2157 "LLDATA", 2158 "STATIC", 2159 "BLACKHOLE", 2160 "UNUSED_0x2000", 2161 "PROTO2", 2162 "PROTO1", 2163 "UNUSED_0x10000", 2164 "UNUSED_0x20000", 2165 "PROTO3", 2166 "FIXEDMTU", 2167 "PINNED", 2168 "LOCAL", 2169 "BROADCAST", 2170 "MULTICAST", 2171 /* Big gap. */ 2172 [28] = "STICKY", 2173 [30] = "RNH_LOCKED", 2174 [31] = "GWFLAG_COMPAT", 2175 }; 2176 2177 static const char * __pure 2178 rt_flag_name(unsigned idx) 2179 { 2180 if (idx >= nitems(rtf_flag_strings)) 2181 return ("INVALID_FLAG"); 2182 if (rtf_flag_strings[idx] == NULL) 2183 return ("UNKNOWN"); 2184 return (rtf_flag_strings[idx]); 2185 } 2186 2187 static void 2188 rt_dumpaddr_ddb(const char *name, const struct sockaddr *sa) 2189 { 2190 char buf[INET6_ADDRSTRLEN], *res; 2191 2192 res = NULL; 2193 if (sa == NULL) 2194 res = "NULL"; 2195 else if (sa->sa_family == AF_INET) { 2196 res = inet_ntop(AF_INET, 2197 &((const struct sockaddr_in *)sa)->sin_addr, 2198 buf, sizeof(buf)); 2199 } else if (sa->sa_family == AF_INET6) { 2200 res = inet_ntop(AF_INET6, 2201 &((const struct sockaddr_in6 *)sa)->sin6_addr, 2202 buf, sizeof(buf)); 2203 } else if (sa->sa_family == AF_LINK) { 2204 res = "on link"; 2205 } 2206 2207 if (res != NULL) { 2208 db_printf("%s <%s> ", name, res); 2209 return; 2210 } 2211 2212 db_printf("%s <af:%d> ", name, sa->sa_family); 2213 } 2214 2215 static int 2216 rt_dumpentry_ddb(struct radix_node *rn, void *arg __unused) 2217 { 2218 struct sockaddr_storage ss; 2219 struct rtentry *rt; 2220 int flags, idx; 2221 2222 /* If RNTORT is important, put it in a header. */ 2223 rt = (void *)rn; 2224 2225 rt_dumpaddr_ddb("dst", rt_key(rt)); 2226 rt_dumpaddr_ddb("gateway", rt->rt_gateway); 2227 rt_dumpaddr_ddb("netmask", rtsock_fix_netmask(rt_key(rt), rt_mask(rt), 2228 &ss)); 2229 if (rt->rt_ifp != NULL && (rt->rt_ifp->if_flags & IFF_DYING) == 0) { 2230 rt_dumpaddr_ddb("ifp", rt->rt_ifp->if_addr->ifa_addr); 2231 rt_dumpaddr_ddb("ifa", rt->rt_ifa->ifa_addr); 2232 } 2233 2234 db_printf("flags "); 2235 flags = rt->rt_flags; 2236 if (flags == 0) 2237 db_printf("none"); 2238 2239 while ((idx = ffs(flags)) > 0) { 2240 idx--; 2241 2242 if (flags != rt->rt_flags) 2243 db_printf(","); 2244 db_printf("%s", rt_flag_name(idx)); 2245 2246 flags &= ~(1ul << idx); 2247 } 2248 2249 db_printf("\n"); 2250 return (0); 2251 } 2252 2253 DB_SHOW_COMMAND(routetable, db_show_routetable_cmd) 2254 { 2255 struct rib_head *rnh; 2256 int error, i, lim; 2257 2258 if (have_addr) 2259 i = lim = addr; 2260 else { 2261 i = 1; 2262 lim = AF_MAX; 2263 } 2264 2265 for (; i <= lim; i++) { 2266 rnh = rt_tables_get_rnh(0, i); 2267 if (rnh == NULL) { 2268 if (have_addr) { 2269 db_printf("%s: AF %d not supported?\n", 2270 __func__, i); 2271 break; 2272 } 2273 continue; 2274 } 2275 2276 if (!have_addr && i > 1) 2277 db_printf("\n"); 2278 2279 db_printf("Route table for AF %d%s%s%s:\n", i, 2280 (i == AF_INET || i == AF_INET6) ? " (" : "", 2281 (i == AF_INET) ? "INET" : (i == AF_INET6) ? "INET6" : "", 2282 (i == AF_INET || i == AF_INET6) ? ")" : ""); 2283 2284 error = rnh->rnh_walktree(&rnh->head, rt_dumpentry_ddb, NULL); 2285 if (error != 0) 2286 db_printf("%s: walktree(%d): %d\n", __func__, i, 2287 error); 2288 } 2289 } 2290 2291 _DB_FUNC(_show, route, db_show_route_cmd, db_show_table, CS_OWN, NULL) 2292 { 2293 char buf[INET6_ADDRSTRLEN], *bp; 2294 const void *dst_addrp; 2295 struct sockaddr *dstp; 2296 struct rtentry *rt; 2297 union { 2298 struct sockaddr_in dest_sin; 2299 struct sockaddr_in6 dest_sin6; 2300 } u; 2301 uint16_t hextets[8]; 2302 unsigned i, tets; 2303 int t, af, exp, tokflags; 2304 2305 /* 2306 * Undecoded address family. No double-colon expansion seen yet. 2307 */ 2308 af = -1; 2309 exp = -1; 2310 /* Assume INET6 to start; we can work back if guess was wrong. */ 2311 tokflags = DRT_WSPACE | DRT_HEX | DRT_HEXADECIMAL; 2312 2313 /* 2314 * db_command has lexed 'show route' for us. 2315 */ 2316 t = db_read_token_flags(tokflags); 2317 if (t == tWSPACE) 2318 t = db_read_token_flags(tokflags); 2319 2320 /* 2321 * tEOL: Just 'show route' isn't a valid mode. 2322 * tMINUS: It's either '-h' or some invalid option. Regardless, usage. 2323 */ 2324 if (t == tEOL || t == tMINUS) 2325 goto usage; 2326 2327 db_unread_token(t); 2328 2329 tets = nitems(hextets); 2330 2331 /* 2332 * Each loop iteration, we expect to read one octet (v4) or hextet 2333 * (v6), followed by an appropriate field separator ('.' or ':' or 2334 * '::'). 2335 * 2336 * At the start of each loop, we're looking for a number (octet or 2337 * hextet). 2338 * 2339 * INET6 addresses have a special case where they may begin with '::'. 2340 */ 2341 for (i = 0; i < tets; i++) { 2342 t = db_read_token_flags(tokflags); 2343 2344 if (t == tCOLONCOLON) { 2345 /* INET6 with leading '::' or invalid. */ 2346 if (i != 0) { 2347 db_printf("Parse error: unexpected extra " 2348 "colons.\n"); 2349 goto exit; 2350 } 2351 2352 af = AF_INET6; 2353 exp = i; 2354 hextets[i] = 0; 2355 continue; 2356 } else if (t == tNUMBER) { 2357 /* 2358 * Lexer separates out '-' as tMINUS, but make the 2359 * assumption explicit here. 2360 */ 2361 MPASS(db_tok_number >= 0); 2362 2363 if (af == AF_INET && db_tok_number > UINT8_MAX) { 2364 db_printf("Not a valid v4 octet: %ld\n", 2365 (long)db_tok_number); 2366 goto exit; 2367 } 2368 hextets[i] = db_tok_number; 2369 } else if (t == tEOL) { 2370 /* 2371 * We can only detect the end of an IPv6 address in 2372 * compact representation with EOL. 2373 */ 2374 if (af != AF_INET6 || exp < 0) { 2375 db_printf("Parse failed. Got unexpected EOF " 2376 "when the address is not a compact-" 2377 "representation IPv6 address.\n"); 2378 goto exit; 2379 } 2380 break; 2381 } else { 2382 db_printf("Parse failed. Unexpected token %d.\n", t); 2383 goto exit; 2384 } 2385 2386 /* Next, look for a separator, if appropriate. */ 2387 if (i == tets - 1) 2388 continue; 2389 2390 t = db_read_token_flags(tokflags); 2391 if (af < 0) { 2392 if (t == tCOLON) { 2393 af = AF_INET6; 2394 continue; 2395 } 2396 if (t == tCOLONCOLON) { 2397 af = AF_INET6; 2398 i++; 2399 hextets[i] = 0; 2400 exp = i; 2401 continue; 2402 } 2403 if (t == tDOT) { 2404 unsigned hn, dn; 2405 2406 af = AF_INET; 2407 /* Need to fixup the first parsed number. */ 2408 if (hextets[0] > 0x255 || 2409 (hextets[0] & 0xf0) > 0x90 || 2410 (hextets[0] & 0xf) > 9) { 2411 db_printf("Not a valid v4 octet: %x\n", 2412 hextets[0]); 2413 goto exit; 2414 } 2415 2416 hn = hextets[0]; 2417 dn = (hn >> 8) * 100 + 2418 ((hn >> 4) & 0xf) * 10 + 2419 (hn & 0xf); 2420 2421 hextets[0] = dn; 2422 2423 /* Switch to decimal for remaining octets. */ 2424 tokflags &= ~DRT_RADIX_MASK; 2425 tokflags |= DRT_DECIMAL; 2426 2427 tets = 4; 2428 continue; 2429 } 2430 2431 db_printf("Parse error. Unexpected token %d.\n", t); 2432 goto exit; 2433 } else if (af == AF_INET) { 2434 if (t == tDOT) 2435 continue; 2436 db_printf("Expected '.' (%d) between octets but got " 2437 "(%d).\n", tDOT, t); 2438 goto exit; 2439 2440 } else if (af == AF_INET6) { 2441 if (t == tCOLON) 2442 continue; 2443 if (t == tCOLONCOLON) { 2444 if (exp < 0) { 2445 i++; 2446 hextets[i] = 0; 2447 exp = i; 2448 continue; 2449 } 2450 db_printf("Got bogus second '::' in v6 " 2451 "address.\n"); 2452 goto exit; 2453 } 2454 if (t == tEOL) { 2455 /* 2456 * Handle in the earlier part of the loop 2457 * because we need to handle trailing :: too. 2458 */ 2459 db_unread_token(t); 2460 continue; 2461 } 2462 2463 db_printf("Expected ':' (%d) or '::' (%d) between " 2464 "hextets but got (%d).\n", tCOLON, tCOLONCOLON, t); 2465 goto exit; 2466 } 2467 } 2468 2469 /* Check for trailing garbage. */ 2470 if (i == tets) { 2471 t = db_read_token_flags(tokflags); 2472 if (t != tEOL) { 2473 db_printf("Got unexpected garbage after address " 2474 "(%d).\n", t); 2475 goto exit; 2476 } 2477 } 2478 2479 /* 2480 * Need to expand compact INET6 addresses. 2481 * 2482 * Technically '::' for a single ':0:' is MUST NOT but just in case, 2483 * don't bother expanding that form (exp >= 0 && i == tets case). 2484 */ 2485 if (af == AF_INET6 && exp >= 0 && i < tets) { 2486 if (exp + 1 < i) { 2487 memmove(&hextets[exp + 1 + (nitems(hextets) - i)], 2488 &hextets[exp + 1], 2489 (i - (exp + 1)) * sizeof(hextets[0])); 2490 } 2491 memset(&hextets[exp + 1], 0, (nitems(hextets) - i) * 2492 sizeof(hextets[0])); 2493 } 2494 2495 memset(&u, 0, sizeof(u)); 2496 if (af == AF_INET) { 2497 u.dest_sin.sin_family = AF_INET; 2498 u.dest_sin.sin_len = sizeof(u.dest_sin); 2499 u.dest_sin.sin_addr.s_addr = htonl( 2500 ((uint32_t)hextets[0] << 24) | 2501 ((uint32_t)hextets[1] << 16) | 2502 ((uint32_t)hextets[2] << 8) | 2503 (uint32_t)hextets[3]); 2504 dstp = (void *)&u.dest_sin; 2505 dst_addrp = &u.dest_sin.sin_addr; 2506 } else if (af == AF_INET6) { 2507 u.dest_sin6.sin6_family = AF_INET6; 2508 u.dest_sin6.sin6_len = sizeof(u.dest_sin6); 2509 for (i = 0; i < nitems(hextets); i++) 2510 u.dest_sin6.sin6_addr.s6_addr16[i] = htons(hextets[i]); 2511 dstp = (void *)&u.dest_sin6; 2512 dst_addrp = &u.dest_sin6.sin6_addr; 2513 } else { 2514 MPASS(false); 2515 /* UNREACHABLE */ 2516 /* Appease Clang false positive: */ 2517 dstp = NULL; 2518 } 2519 2520 bp = inet_ntop(af, dst_addrp, buf, sizeof(buf)); 2521 if (bp != NULL) 2522 db_printf("Looking up route to destination '%s'\n", bp); 2523 2524 CURVNET_SET(vnet0); 2525 rt = rtalloc1(dstp, 0, RTF_RNH_LOCKED); 2526 CURVNET_RESTORE(); 2527 2528 if (rt == NULL) { 2529 db_printf("Could not get route for that server.\n"); 2530 return; 2531 } 2532 2533 rt_dumpentry_ddb((void *)rt, NULL); 2534 RTFREE_LOCKED(rt); 2535 2536 return; 2537 usage: 2538 db_printf("Usage: 'show route <address>'\n" 2539 " Currently accepts only dotted-decimal INET or colon-separated\n" 2540 " hextet INET6 addresses.\n"); 2541 exit: 2542 db_skip_to_eol(); 2543 } 2544 #endif 2545