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, 0, ""); 162 163 struct walkarg { 164 int w_tmemsize; 165 int w_op, w_arg; 166 caddr_t w_tmem; 167 struct sysctl_req *w_req; 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 radix_node *rn, void *vw); 177 static int sysctl_iflist(int af, struct walkarg *w); 178 static int sysctl_ifmalist(int af, struct walkarg *w); 179 static int route_output(struct mbuf *m, struct socket *so, ...); 180 static void rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out); 181 static void rt_dispatch(struct mbuf *, sa_family_t); 182 static struct sockaddr *rtsock_fix_netmask(struct sockaddr *dst, 183 struct sockaddr *smask, struct sockaddr_storage *dmask); 184 static int handle_rtm_get(struct rt_addrinfo *info, u_int fibnum, 185 struct rt_msghdr *rtm, struct rtentry **ret_nrt); 186 static int update_rtm_from_rte(struct rt_addrinfo *info, 187 struct rt_msghdr **prtm, int alloc_len, 188 struct rtentry *rt); 189 static void send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, 190 struct mbuf *m, sa_family_t saf, u_int fibnum, 191 int rtm_errno); 192 static int can_export_rte(struct ucred *td_ucred, const struct rtentry *rt); 193 194 static struct netisr_handler rtsock_nh = { 195 .nh_name = "rtsock", 196 .nh_handler = rts_input, 197 .nh_proto = NETISR_ROUTE, 198 .nh_policy = NETISR_POLICY_SOURCE, 199 }; 200 201 static int 202 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS) 203 { 204 int error, qlimit; 205 206 netisr_getqlimit(&rtsock_nh, &qlimit); 207 error = sysctl_handle_int(oidp, &qlimit, 0, req); 208 if (error || !req->newptr) 209 return (error); 210 if (qlimit < 1) 211 return (EINVAL); 212 return (netisr_setqlimit(&rtsock_nh, qlimit)); 213 } 214 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW, 215 0, 0, sysctl_route_netisr_maxqlen, "I", 216 "maximum routing socket dispatch queue length"); 217 218 static void 219 vnet_rts_init(void) 220 { 221 int tmp; 222 223 if (IS_DEFAULT_VNET(curvnet)) { 224 if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp)) 225 rtsock_nh.nh_qlimit = tmp; 226 netisr_register(&rtsock_nh); 227 } 228 #ifdef VIMAGE 229 else 230 netisr_register_vnet(&rtsock_nh); 231 #endif 232 } 233 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 234 vnet_rts_init, 0); 235 236 #ifdef VIMAGE 237 static void 238 vnet_rts_uninit(void) 239 { 240 241 netisr_unregister_vnet(&rtsock_nh); 242 } 243 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 244 vnet_rts_uninit, 0); 245 #endif 246 247 static int 248 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src, 249 struct rawcb *rp) 250 { 251 int fibnum; 252 253 KASSERT(m != NULL, ("%s: m is NULL", __func__)); 254 KASSERT(proto != NULL, ("%s: proto is NULL", __func__)); 255 KASSERT(rp != NULL, ("%s: rp is NULL", __func__)); 256 257 /* No filtering requested. */ 258 if ((m->m_flags & RTS_FILTER_FIB) == 0) 259 return (0); 260 261 /* Check if it is a rts and the fib matches the one of the socket. */ 262 fibnum = M_GETFIB(m); 263 if (proto->sp_family != PF_ROUTE || 264 rp->rcb_socket == NULL || 265 rp->rcb_socket->so_fibnum == fibnum) 266 return (0); 267 268 /* Filtering requested and no match, the socket shall be skipped. */ 269 return (1); 270 } 271 272 static void 273 rts_input(struct mbuf *m) 274 { 275 struct sockproto route_proto; 276 unsigned short *family; 277 struct m_tag *tag; 278 279 route_proto.sp_family = PF_ROUTE; 280 tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL); 281 if (tag != NULL) { 282 family = (unsigned short *)(tag + 1); 283 route_proto.sp_protocol = *family; 284 m_tag_delete(m, tag); 285 } else 286 route_proto.sp_protocol = 0; 287 288 raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb); 289 } 290 291 /* 292 * It really doesn't make any sense at all for this code to share much 293 * with raw_usrreq.c, since its functionality is so restricted. XXX 294 */ 295 static void 296 rts_abort(struct socket *so) 297 { 298 299 raw_usrreqs.pru_abort(so); 300 } 301 302 static void 303 rts_close(struct socket *so) 304 { 305 306 raw_usrreqs.pru_close(so); 307 } 308 309 /* pru_accept is EOPNOTSUPP */ 310 311 static int 312 rts_attach(struct socket *so, int proto, struct thread *td) 313 { 314 struct rawcb *rp; 315 int error; 316 317 KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL")); 318 319 /* XXX */ 320 rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO); 321 322 so->so_pcb = (caddr_t)rp; 323 so->so_fibnum = td->td_proc->p_fibnum; 324 error = raw_attach(so, proto); 325 rp = sotorawcb(so); 326 if (error) { 327 so->so_pcb = NULL; 328 free(rp, M_PCB); 329 return error; 330 } 331 RTSOCK_LOCK(); 332 switch(rp->rcb_proto.sp_protocol) { 333 case AF_INET: 334 V_route_cb.ip_count++; 335 break; 336 case AF_INET6: 337 V_route_cb.ip6_count++; 338 break; 339 } 340 V_route_cb.any_count++; 341 RTSOCK_UNLOCK(); 342 soisconnected(so); 343 so->so_options |= SO_USELOOPBACK; 344 return 0; 345 } 346 347 static int 348 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 349 { 350 351 return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */ 352 } 353 354 static int 355 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 356 { 357 358 return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */ 359 } 360 361 /* pru_connect2 is EOPNOTSUPP */ 362 /* pru_control is EOPNOTSUPP */ 363 364 static void 365 rts_detach(struct socket *so) 366 { 367 struct rawcb *rp = sotorawcb(so); 368 369 KASSERT(rp != NULL, ("rts_detach: rp == NULL")); 370 371 RTSOCK_LOCK(); 372 switch(rp->rcb_proto.sp_protocol) { 373 case AF_INET: 374 V_route_cb.ip_count--; 375 break; 376 case AF_INET6: 377 V_route_cb.ip6_count--; 378 break; 379 } 380 V_route_cb.any_count--; 381 RTSOCK_UNLOCK(); 382 raw_usrreqs.pru_detach(so); 383 } 384 385 static int 386 rts_disconnect(struct socket *so) 387 { 388 389 return (raw_usrreqs.pru_disconnect(so)); 390 } 391 392 /* pru_listen is EOPNOTSUPP */ 393 394 static int 395 rts_peeraddr(struct socket *so, struct sockaddr **nam) 396 { 397 398 return (raw_usrreqs.pru_peeraddr(so, nam)); 399 } 400 401 /* pru_rcvd is EOPNOTSUPP */ 402 /* pru_rcvoob is EOPNOTSUPP */ 403 404 static int 405 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 406 struct mbuf *control, struct thread *td) 407 { 408 409 return (raw_usrreqs.pru_send(so, flags, m, nam, control, td)); 410 } 411 412 /* pru_sense is null */ 413 414 static int 415 rts_shutdown(struct socket *so) 416 { 417 418 return (raw_usrreqs.pru_shutdown(so)); 419 } 420 421 static int 422 rts_sockaddr(struct socket *so, struct sockaddr **nam) 423 { 424 425 return (raw_usrreqs.pru_sockaddr(so, nam)); 426 } 427 428 static struct pr_usrreqs route_usrreqs = { 429 .pru_abort = rts_abort, 430 .pru_attach = rts_attach, 431 .pru_bind = rts_bind, 432 .pru_connect = rts_connect, 433 .pru_detach = rts_detach, 434 .pru_disconnect = rts_disconnect, 435 .pru_peeraddr = rts_peeraddr, 436 .pru_send = rts_send, 437 .pru_shutdown = rts_shutdown, 438 .pru_sockaddr = rts_sockaddr, 439 .pru_close = rts_close, 440 }; 441 442 #ifndef _SOCKADDR_UNION_DEFINED 443 #define _SOCKADDR_UNION_DEFINED 444 /* 445 * The union of all possible address formats we handle. 446 */ 447 union sockaddr_union { 448 struct sockaddr sa; 449 struct sockaddr_in sin; 450 struct sockaddr_in6 sin6; 451 }; 452 #endif /* _SOCKADDR_UNION_DEFINED */ 453 454 static int 455 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp, 456 struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred) 457 { 458 #if defined(INET) || defined(INET6) 459 struct epoch_tracker et; 460 #endif 461 462 /* First, see if the returned address is part of the jail. */ 463 if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) { 464 info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr; 465 return (0); 466 } 467 468 switch (info->rti_info[RTAX_DST]->sa_family) { 469 #ifdef INET 470 case AF_INET: 471 { 472 struct in_addr ia; 473 struct ifaddr *ifa; 474 int found; 475 476 found = 0; 477 /* 478 * Try to find an address on the given outgoing interface 479 * that belongs to the jail. 480 */ 481 NET_EPOCH_ENTER(et); 482 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 483 struct sockaddr *sa; 484 sa = ifa->ifa_addr; 485 if (sa->sa_family != AF_INET) 486 continue; 487 ia = ((struct sockaddr_in *)sa)->sin_addr; 488 if (prison_check_ip4(cred, &ia) == 0) { 489 found = 1; 490 break; 491 } 492 } 493 NET_EPOCH_EXIT(et); 494 if (!found) { 495 /* 496 * As a last resort return the 'default' jail address. 497 */ 498 ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)-> 499 sin_addr; 500 if (prison_get_ip4(cred, &ia) != 0) 501 return (ESRCH); 502 } 503 bzero(&saun->sin, sizeof(struct sockaddr_in)); 504 saun->sin.sin_len = sizeof(struct sockaddr_in); 505 saun->sin.sin_family = AF_INET; 506 saun->sin.sin_addr.s_addr = ia.s_addr; 507 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin; 508 break; 509 } 510 #endif 511 #ifdef INET6 512 case AF_INET6: 513 { 514 struct in6_addr ia6; 515 struct ifaddr *ifa; 516 int found; 517 518 found = 0; 519 /* 520 * Try to find an address on the given outgoing interface 521 * that belongs to the jail. 522 */ 523 NET_EPOCH_ENTER(et); 524 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 525 struct sockaddr *sa; 526 sa = ifa->ifa_addr; 527 if (sa->sa_family != AF_INET6) 528 continue; 529 bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr, 530 &ia6, sizeof(struct in6_addr)); 531 if (prison_check_ip6(cred, &ia6) == 0) { 532 found = 1; 533 break; 534 } 535 } 536 NET_EPOCH_EXIT(et); 537 if (!found) { 538 /* 539 * As a last resort return the 'default' jail address. 540 */ 541 ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)-> 542 sin6_addr; 543 if (prison_get_ip6(cred, &ia6) != 0) 544 return (ESRCH); 545 } 546 bzero(&saun->sin6, sizeof(struct sockaddr_in6)); 547 saun->sin6.sin6_len = sizeof(struct sockaddr_in6); 548 saun->sin6.sin6_family = AF_INET6; 549 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr)); 550 if (sa6_recoverscope(&saun->sin6) != 0) 551 return (ESRCH); 552 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6; 553 break; 554 } 555 #endif 556 default: 557 return (ESRCH); 558 } 559 return (0); 560 } 561 562 /* 563 * Fills in @info based on userland-provided @rtm message. 564 * 565 * Returns 0 on success. 566 */ 567 static int 568 fill_addrinfo(struct rt_msghdr *rtm, int len, u_int fibnum, struct rt_addrinfo *info) 569 { 570 int error; 571 sa_family_t saf; 572 573 rtm->rtm_pid = curproc->p_pid; 574 info->rti_addrs = rtm->rtm_addrs; 575 576 info->rti_mflags = rtm->rtm_inits; 577 info->rti_rmx = &rtm->rtm_rmx; 578 579 /* 580 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6 581 * link-local address because rtrequest requires addresses with 582 * embedded scope id. 583 */ 584 if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info)) 585 return (EINVAL); 586 587 if (rtm->rtm_flags & RTF_RNH_LOCKED) 588 return (EINVAL); 589 info->rti_flags = rtm->rtm_flags; 590 if (info->rti_info[RTAX_DST] == NULL || 591 info->rti_info[RTAX_DST]->sa_family >= AF_MAX || 592 (info->rti_info[RTAX_GATEWAY] != NULL && 593 info->rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) 594 return (EINVAL); 595 saf = info->rti_info[RTAX_DST]->sa_family; 596 /* 597 * Verify that the caller has the appropriate privilege; RTM_GET 598 * is the only operation the non-superuser is allowed. 599 */ 600 if (rtm->rtm_type != RTM_GET) { 601 error = priv_check(curthread, PRIV_NET_ROUTE); 602 if (error != 0) 603 return (error); 604 } 605 606 /* 607 * The given gateway address may be an interface address. 608 * For example, issuing a "route change" command on a route 609 * entry that was created from a tunnel, and the gateway 610 * address given is the local end point. In this case the 611 * RTF_GATEWAY flag must be cleared or the destination will 612 * not be reachable even though there is no error message. 613 */ 614 if (info->rti_info[RTAX_GATEWAY] != NULL && 615 info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) { 616 struct rt_addrinfo ginfo; 617 struct sockaddr *gdst; 618 struct sockaddr_storage ss; 619 620 bzero(&ginfo, sizeof(ginfo)); 621 bzero(&ss, sizeof(ss)); 622 ss.ss_len = sizeof(ss); 623 624 ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss; 625 gdst = info->rti_info[RTAX_GATEWAY]; 626 627 /* 628 * A host route through the loopback interface is 629 * installed for each interface adddress. In pre 8.0 630 * releases the interface address of a PPP link type 631 * is not reachable locally. This behavior is fixed as 632 * part of the new L2/L3 redesign and rewrite work. The 633 * signature of this interface address route is the 634 * AF_LINK sa_family type of the rt_gateway, and the 635 * rt_ifp has the IFF_LOOPBACK flag set. 636 */ 637 if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) { 638 if (ss.ss_family == AF_LINK && 639 ginfo.rti_ifp->if_flags & IFF_LOOPBACK) { 640 info->rti_flags &= ~RTF_GATEWAY; 641 info->rti_flags |= RTF_GWFLAG_COMPAT; 642 } 643 rib_free_info(&ginfo); 644 } 645 } 646 647 return (0); 648 } 649 650 /* 651 * Handles RTM_GET message from routing socket, returning matching rt. 652 * 653 * Returns: 654 * 0 on success, with locked and referenced matching rt in @rt_nrt 655 * errno of failure 656 */ 657 static int 658 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum, 659 struct rt_msghdr *rtm, struct rtentry **ret_nrt) 660 { 661 RIB_RLOCK_TRACKER; 662 struct rtentry *rt; 663 struct rib_head *rnh; 664 sa_family_t saf; 665 666 saf = info->rti_info[RTAX_DST]->sa_family; 667 668 rnh = rt_tables_get_rnh(fibnum, saf); 669 if (rnh == NULL) 670 return (EAFNOSUPPORT); 671 672 RIB_RLOCK(rnh); 673 674 if (info->rti_info[RTAX_NETMASK] == NULL) { 675 /* 676 * Provide longest prefix match for 677 * address lookup (no mask). 678 * 'route -n get addr' 679 */ 680 rt = (struct rtentry *) rnh->rnh_matchaddr( 681 info->rti_info[RTAX_DST], &rnh->head); 682 } else 683 rt = (struct rtentry *) rnh->rnh_lookup( 684 info->rti_info[RTAX_DST], 685 info->rti_info[RTAX_NETMASK], &rnh->head); 686 687 if (rt == NULL) { 688 RIB_RUNLOCK(rnh); 689 return (ESRCH); 690 } 691 #ifdef RADIX_MPATH 692 /* 693 * for RTM_GET, gate is optional even with multipath. 694 * if gate == NULL the first match is returned. 695 * (no need to call rt_mpath_matchgate if gate == NULL) 696 */ 697 if (rt_mpath_capable(rnh) && info->rti_info[RTAX_GATEWAY]) { 698 rt = rt_mpath_matchgate(rt, info->rti_info[RTAX_GATEWAY]); 699 if (!rt) { 700 RIB_RUNLOCK(rnh); 701 return (ESRCH); 702 } 703 } 704 #endif 705 /* 706 * If performing proxied L2 entry insertion, and 707 * the actual PPP host entry is found, perform 708 * another search to retrieve the prefix route of 709 * the local end point of the PPP link. 710 */ 711 if (rtm->rtm_flags & RTF_ANNOUNCE) { 712 struct sockaddr laddr; 713 714 if (rt->rt_ifp != NULL && 715 rt->rt_ifp->if_type == IFT_PROPVIRTUAL) { 716 struct epoch_tracker et; 717 struct ifaddr *ifa; 718 719 NET_EPOCH_ENTER(et); 720 ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1, 721 RT_ALL_FIBS); 722 NET_EPOCH_EXIT(et); 723 if (ifa != NULL) 724 rt_maskedcopy(ifa->ifa_addr, 725 &laddr, 726 ifa->ifa_netmask); 727 } else 728 rt_maskedcopy(rt->rt_ifa->ifa_addr, 729 &laddr, 730 rt->rt_ifa->ifa_netmask); 731 /* 732 * refactor rt and no lock operation necessary 733 */ 734 rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr, 735 &rnh->head); 736 if (rt == NULL) { 737 RIB_RUNLOCK(rnh); 738 return (ESRCH); 739 } 740 } 741 RT_LOCK(rt); 742 RT_ADDREF(rt); 743 RIB_RUNLOCK(rnh); 744 745 *ret_nrt = rt; 746 747 return (0); 748 } 749 750 /* 751 * Update sockaddrs, flags, etc in @prtm based on @rt data. 752 * Assumes @rt is locked. 753 * rtm can be reallocated. 754 * 755 * Returns 0 on success, along with pointer to (potentially reallocated) 756 * rtm. 757 * 758 */ 759 static int 760 update_rtm_from_rte(struct rt_addrinfo *info, struct rt_msghdr **prtm, 761 int alloc_len, struct rtentry *rt) 762 { 763 struct sockaddr_storage netmask_ss; 764 struct walkarg w; 765 union sockaddr_union saun; 766 struct rt_msghdr *rtm, *orig_rtm = NULL; 767 struct ifnet *ifp; 768 int error, len; 769 770 RT_LOCK_ASSERT(rt); 771 772 rtm = *prtm; 773 774 info->rti_info[RTAX_DST] = rt_key(rt); 775 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway; 776 info->rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt), 777 rt_mask(rt), &netmask_ss); 778 info->rti_info[RTAX_GENMASK] = 0; 779 ifp = rt->rt_ifp; 780 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) { 781 if (ifp) { 782 info->rti_info[RTAX_IFP] = 783 ifp->if_addr->ifa_addr; 784 error = rtm_get_jailed(info, ifp, rt, 785 &saun, curthread->td_ucred); 786 if (error != 0) 787 return (error); 788 if (ifp->if_flags & IFF_POINTOPOINT) 789 info->rti_info[RTAX_BRD] = 790 rt->rt_ifa->ifa_dstaddr; 791 rtm->rtm_index = ifp->if_index; 792 } else { 793 info->rti_info[RTAX_IFP] = NULL; 794 info->rti_info[RTAX_IFA] = NULL; 795 } 796 } else if (ifp != NULL) 797 rtm->rtm_index = ifp->if_index; 798 799 /* Check if we need to realloc storage */ 800 rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len); 801 if (len > alloc_len) { 802 struct rt_msghdr *tmp_rtm; 803 804 tmp_rtm = malloc(len, M_TEMP, M_NOWAIT); 805 if (tmp_rtm == NULL) 806 return (ENOBUFS); 807 bcopy(rtm, tmp_rtm, rtm->rtm_msglen); 808 orig_rtm = rtm; 809 rtm = tmp_rtm; 810 alloc_len = len; 811 812 /* 813 * Delay freeing original rtm as info contains 814 * data referencing it. 815 */ 816 } 817 818 w.w_tmem = (caddr_t)rtm; 819 w.w_tmemsize = alloc_len; 820 rtsock_msg_buffer(rtm->rtm_type, info, &w, &len); 821 822 if (rt->rt_flags & RTF_GWFLAG_COMPAT) 823 rtm->rtm_flags = RTF_GATEWAY | 824 (rt->rt_flags & ~RTF_GWFLAG_COMPAT); 825 else 826 rtm->rtm_flags = rt->rt_flags; 827 rt_getmetrics(rt, &rtm->rtm_rmx); 828 rtm->rtm_addrs = info->rti_addrs; 829 830 if (orig_rtm != NULL) 831 free(orig_rtm, M_TEMP); 832 *prtm = rtm; 833 834 return (0); 835 } 836 837 /*ARGSUSED*/ 838 static int 839 route_output(struct mbuf *m, struct socket *so, ...) 840 { 841 struct rt_msghdr *rtm = NULL; 842 struct rtentry *rt = NULL; 843 struct rt_addrinfo info; 844 struct epoch_tracker et; 845 #ifdef INET6 846 struct sockaddr_storage ss; 847 struct sockaddr_in6 *sin6; 848 int i, rti_need_deembed = 0; 849 #endif 850 int alloc_len = 0, len, error = 0, fibnum; 851 sa_family_t saf = AF_UNSPEC; 852 struct walkarg w; 853 854 fibnum = so->so_fibnum; 855 856 #define senderr(e) { error = e; goto flush;} 857 if (m == NULL || ((m->m_len < sizeof(long)) && 858 (m = m_pullup(m, sizeof(long))) == NULL)) 859 return (ENOBUFS); 860 if ((m->m_flags & M_PKTHDR) == 0) 861 panic("route_output"); 862 NET_EPOCH_ENTER(et); 863 len = m->m_pkthdr.len; 864 if (len < sizeof(*rtm) || 865 len != mtod(m, struct rt_msghdr *)->rtm_msglen) 866 senderr(EINVAL); 867 868 /* 869 * Most of current messages are in range 200-240 bytes, 870 * minimize possible re-allocation on reply using larger size 871 * buffer aligned on 1k boundaty. 872 */ 873 alloc_len = roundup2(len, 1024); 874 if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL) 875 senderr(ENOBUFS); 876 877 m_copydata(m, 0, len, (caddr_t)rtm); 878 bzero(&info, sizeof(info)); 879 bzero(&w, sizeof(w)); 880 881 if (rtm->rtm_version != RTM_VERSION) { 882 /* Do not touch message since format is unknown */ 883 free(rtm, M_TEMP); 884 rtm = NULL; 885 senderr(EPROTONOSUPPORT); 886 } 887 888 /* 889 * Starting from here, it is possible 890 * to alter original message and insert 891 * caller PID and error value. 892 */ 893 894 if ((error = fill_addrinfo(rtm, len, fibnum, &info)) != 0) { 895 senderr(error); 896 } 897 898 saf = info.rti_info[RTAX_DST]->sa_family; 899 900 /* support for new ARP code */ 901 if (rtm->rtm_flags & RTF_LLDATA) { 902 error = lla_rt_output(rtm, &info); 903 #ifdef INET6 904 if (error == 0) 905 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0; 906 #endif 907 goto flush; 908 } 909 910 switch (rtm->rtm_type) { 911 struct rtentry *saved_nrt; 912 913 case RTM_ADD: 914 case RTM_CHANGE: 915 if (rtm->rtm_type == RTM_ADD) { 916 if (info.rti_info[RTAX_GATEWAY] == NULL) 917 senderr(EINVAL); 918 } 919 saved_nrt = NULL; 920 error = rtrequest1_fib(rtm->rtm_type, &info, &saved_nrt, 921 fibnum); 922 if (error == 0 && saved_nrt != NULL) { 923 #ifdef INET6 924 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0; 925 #endif 926 RT_LOCK(saved_nrt); 927 rtm->rtm_index = saved_nrt->rt_ifp->if_index; 928 RT_REMREF(saved_nrt); 929 RT_UNLOCK(saved_nrt); 930 } 931 break; 932 933 case RTM_DELETE: 934 saved_nrt = NULL; 935 error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt, fibnum); 936 if (error == 0) { 937 RT_LOCK(saved_nrt); 938 rt = saved_nrt; 939 goto report; 940 } 941 #ifdef INET6 942 /* rt_msg2() will not be used when RTM_DELETE fails. */ 943 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0; 944 #endif 945 break; 946 947 case RTM_GET: 948 error = handle_rtm_get(&info, fibnum, rtm, &rt); 949 if (error != 0) 950 senderr(error); 951 952 report: 953 RT_LOCK_ASSERT(rt); 954 if (!can_export_rte(curthread->td_ucred, rt)) { 955 RT_UNLOCK(rt); 956 senderr(ESRCH); 957 } 958 error = update_rtm_from_rte(&info, &rtm, alloc_len, rt); 959 /* 960 * Note that some sockaddr pointers may have changed to 961 * point to memory outsize @rtm. Some may be pointing 962 * to the on-stack variables. 963 * Given that, any pointer in @info CANNOT BE USED. 964 */ 965 966 /* 967 * scopeid deembedding has been performed while 968 * writing updated rtm in rtsock_msg_buffer(). 969 * With that in mind, skip deembedding procedure below. 970 */ 971 #ifdef INET6 972 rti_need_deembed = 0; 973 #endif 974 RT_UNLOCK(rt); 975 if (error != 0) 976 senderr(error); 977 break; 978 979 default: 980 senderr(EOPNOTSUPP); 981 } 982 983 flush: 984 NET_EPOCH_EXIT(et); 985 if (rt != NULL) 986 RTFREE(rt); 987 988 #ifdef INET6 989 if (rtm != NULL) { 990 if (rti_need_deembed) { 991 /* sin6_scope_id is recovered before sending rtm. */ 992 sin6 = (struct sockaddr_in6 *)&ss; 993 for (i = 0; i < RTAX_MAX; i++) { 994 if (info.rti_info[i] == NULL) 995 continue; 996 if (info.rti_info[i]->sa_family != AF_INET6) 997 continue; 998 bcopy(info.rti_info[i], sin6, sizeof(*sin6)); 999 if (sa6_recoverscope(sin6) == 0) 1000 bcopy(sin6, info.rti_info[i], 1001 sizeof(*sin6)); 1002 } 1003 } 1004 } 1005 #endif 1006 send_rtm_reply(so, rtm, m, saf, fibnum, error); 1007 1008 return (error); 1009 } 1010 1011 /* 1012 * Sends the prepared reply message in @rtm to all rtsock clients. 1013 * Frees @m and @rtm. 1014 * 1015 */ 1016 static void 1017 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m, 1018 sa_family_t saf, u_int fibnum, int rtm_errno) 1019 { 1020 struct rawcb *rp = NULL; 1021 1022 /* 1023 * Check to see if we don't want our own messages. 1024 */ 1025 if ((so->so_options & SO_USELOOPBACK) == 0) { 1026 if (V_route_cb.any_count <= 1) { 1027 if (rtm != NULL) 1028 free(rtm, M_TEMP); 1029 m_freem(m); 1030 return; 1031 } 1032 /* There is another listener, so construct message */ 1033 rp = sotorawcb(so); 1034 } 1035 1036 if (rtm != NULL) { 1037 if (rtm_errno!= 0) 1038 rtm->rtm_errno = rtm_errno; 1039 else 1040 rtm->rtm_flags |= RTF_DONE; 1041 1042 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm); 1043 if (m->m_pkthdr.len < rtm->rtm_msglen) { 1044 m_freem(m); 1045 m = NULL; 1046 } else if (m->m_pkthdr.len > rtm->rtm_msglen) 1047 m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len); 1048 1049 free(rtm, M_TEMP); 1050 } 1051 if (m != NULL) { 1052 M_SETFIB(m, fibnum); 1053 m->m_flags |= RTS_FILTER_FIB; 1054 if (rp) { 1055 /* 1056 * XXX insure we don't get a copy by 1057 * invalidating our protocol 1058 */ 1059 unsigned short family = rp->rcb_proto.sp_family; 1060 rp->rcb_proto.sp_family = 0; 1061 rt_dispatch(m, saf); 1062 rp->rcb_proto.sp_family = family; 1063 } else 1064 rt_dispatch(m, saf); 1065 } 1066 } 1067 1068 1069 static void 1070 rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out) 1071 { 1072 1073 bzero(out, sizeof(*out)); 1074 out->rmx_mtu = rt->rt_mtu; 1075 out->rmx_weight = rt->rt_weight; 1076 out->rmx_pksent = counter_u64_fetch(rt->rt_pksent); 1077 /* Kernel -> userland timebase conversion. */ 1078 out->rmx_expire = rt->rt_expire ? 1079 rt->rt_expire - time_uptime + time_second : 0; 1080 } 1081 1082 /* 1083 * Extract the addresses of the passed sockaddrs. 1084 * Do a little sanity checking so as to avoid bad memory references. 1085 * This data is derived straight from userland. 1086 */ 1087 static int 1088 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo) 1089 { 1090 struct sockaddr *sa; 1091 int i; 1092 1093 for (i = 0; i < RTAX_MAX && cp < cplim; i++) { 1094 if ((rtinfo->rti_addrs & (1 << i)) == 0) 1095 continue; 1096 sa = (struct sockaddr *)cp; 1097 /* 1098 * It won't fit. 1099 */ 1100 if (cp + sa->sa_len > cplim) 1101 return (EINVAL); 1102 /* 1103 * there are no more.. quit now 1104 * If there are more bits, they are in error. 1105 * I've seen this. route(1) can evidently generate these. 1106 * This causes kernel to core dump. 1107 * for compatibility, If we see this, point to a safe address. 1108 */ 1109 if (sa->sa_len == 0) { 1110 rtinfo->rti_info[i] = &sa_zero; 1111 return (0); /* should be EINVAL but for compat */ 1112 } 1113 /* accept it */ 1114 #ifdef INET6 1115 if (sa->sa_family == AF_INET6) 1116 sa6_embedscope((struct sockaddr_in6 *)sa, 1117 V_ip6_use_defzone); 1118 #endif 1119 rtinfo->rti_info[i] = sa; 1120 cp += SA_SIZE(sa); 1121 } 1122 return (0); 1123 } 1124 1125 /* 1126 * Fill in @dmask with valid netmask leaving original @smask 1127 * intact. Mostly used with radix netmasks. 1128 */ 1129 static struct sockaddr * 1130 rtsock_fix_netmask(struct sockaddr *dst, struct sockaddr *smask, 1131 struct sockaddr_storage *dmask) 1132 { 1133 if (dst == NULL || smask == NULL) 1134 return (NULL); 1135 1136 memset(dmask, 0, dst->sa_len); 1137 memcpy(dmask, smask, smask->sa_len); 1138 dmask->ss_len = dst->sa_len; 1139 dmask->ss_family = dst->sa_family; 1140 1141 return ((struct sockaddr *)dmask); 1142 } 1143 1144 /* 1145 * Writes information related to @rtinfo object to newly-allocated mbuf. 1146 * Assumes MCLBYTES is enough to construct any message. 1147 * Used for OS notifications of vaious events (if/ifa announces,etc) 1148 * 1149 * Returns allocated mbuf or NULL on failure. 1150 */ 1151 static struct mbuf * 1152 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo) 1153 { 1154 struct rt_msghdr *rtm; 1155 struct mbuf *m; 1156 int i; 1157 struct sockaddr *sa; 1158 #ifdef INET6 1159 struct sockaddr_storage ss; 1160 struct sockaddr_in6 *sin6; 1161 #endif 1162 int len, dlen; 1163 1164 switch (type) { 1165 1166 case RTM_DELADDR: 1167 case RTM_NEWADDR: 1168 len = sizeof(struct ifa_msghdr); 1169 break; 1170 1171 case RTM_DELMADDR: 1172 case RTM_NEWMADDR: 1173 len = sizeof(struct ifma_msghdr); 1174 break; 1175 1176 case RTM_IFINFO: 1177 len = sizeof(struct if_msghdr); 1178 break; 1179 1180 case RTM_IFANNOUNCE: 1181 case RTM_IEEE80211: 1182 len = sizeof(struct if_announcemsghdr); 1183 break; 1184 1185 default: 1186 len = sizeof(struct rt_msghdr); 1187 } 1188 1189 /* XXXGL: can we use MJUMPAGESIZE cluster here? */ 1190 KASSERT(len <= MCLBYTES, ("%s: message too big", __func__)); 1191 if (len > MHLEN) 1192 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 1193 else 1194 m = m_gethdr(M_NOWAIT, MT_DATA); 1195 if (m == NULL) 1196 return (m); 1197 1198 m->m_pkthdr.len = m->m_len = len; 1199 rtm = mtod(m, struct rt_msghdr *); 1200 bzero((caddr_t)rtm, len); 1201 for (i = 0; i < RTAX_MAX; i++) { 1202 if ((sa = rtinfo->rti_info[i]) == NULL) 1203 continue; 1204 rtinfo->rti_addrs |= (1 << i); 1205 dlen = SA_SIZE(sa); 1206 #ifdef INET6 1207 if (V_deembed_scopeid && sa->sa_family == AF_INET6) { 1208 sin6 = (struct sockaddr_in6 *)&ss; 1209 bcopy(sa, sin6, sizeof(*sin6)); 1210 if (sa6_recoverscope(sin6) == 0) 1211 sa = (struct sockaddr *)sin6; 1212 } 1213 #endif 1214 m_copyback(m, len, dlen, (caddr_t)sa); 1215 len += dlen; 1216 } 1217 if (m->m_pkthdr.len != len) { 1218 m_freem(m); 1219 return (NULL); 1220 } 1221 rtm->rtm_msglen = len; 1222 rtm->rtm_version = RTM_VERSION; 1223 rtm->rtm_type = type; 1224 return (m); 1225 } 1226 1227 /* 1228 * Writes information related to @rtinfo object to preallocated buffer. 1229 * Stores needed size in @plen. If @w is NULL, calculates size without 1230 * writing. 1231 * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation. 1232 * 1233 * Returns 0 on success. 1234 * 1235 */ 1236 static int 1237 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen) 1238 { 1239 int i; 1240 int len, buflen = 0, dlen; 1241 caddr_t cp = NULL; 1242 struct rt_msghdr *rtm = NULL; 1243 #ifdef INET6 1244 struct sockaddr_storage ss; 1245 struct sockaddr_in6 *sin6; 1246 #endif 1247 #ifdef COMPAT_FREEBSD32 1248 bool compat32 = false; 1249 #endif 1250 1251 switch (type) { 1252 1253 case RTM_DELADDR: 1254 case RTM_NEWADDR: 1255 if (w != NULL && w->w_op == NET_RT_IFLISTL) { 1256 #ifdef COMPAT_FREEBSD32 1257 if (w->w_req->flags & SCTL_MASK32) { 1258 len = sizeof(struct ifa_msghdrl32); 1259 compat32 = true; 1260 } else 1261 #endif 1262 len = sizeof(struct ifa_msghdrl); 1263 } else 1264 len = sizeof(struct ifa_msghdr); 1265 break; 1266 1267 case RTM_IFINFO: 1268 #ifdef COMPAT_FREEBSD32 1269 if (w != NULL && w->w_req->flags & SCTL_MASK32) { 1270 if (w->w_op == NET_RT_IFLISTL) 1271 len = sizeof(struct if_msghdrl32); 1272 else 1273 len = sizeof(struct if_msghdr32); 1274 compat32 = true; 1275 break; 1276 } 1277 #endif 1278 if (w != NULL && w->w_op == NET_RT_IFLISTL) 1279 len = sizeof(struct if_msghdrl); 1280 else 1281 len = sizeof(struct if_msghdr); 1282 break; 1283 1284 case RTM_NEWMADDR: 1285 len = sizeof(struct ifma_msghdr); 1286 break; 1287 1288 default: 1289 len = sizeof(struct rt_msghdr); 1290 } 1291 1292 if (w != NULL) { 1293 rtm = (struct rt_msghdr *)w->w_tmem; 1294 buflen = w->w_tmemsize - len; 1295 cp = (caddr_t)w->w_tmem + len; 1296 } 1297 1298 rtinfo->rti_addrs = 0; 1299 for (i = 0; i < RTAX_MAX; i++) { 1300 struct sockaddr *sa; 1301 1302 if ((sa = rtinfo->rti_info[i]) == NULL) 1303 continue; 1304 rtinfo->rti_addrs |= (1 << i); 1305 #ifdef COMPAT_FREEBSD32 1306 if (compat32) 1307 dlen = SA_SIZE32(sa); 1308 else 1309 #endif 1310 dlen = SA_SIZE(sa); 1311 if (cp != NULL && buflen >= dlen) { 1312 #ifdef INET6 1313 if (V_deembed_scopeid && sa->sa_family == AF_INET6) { 1314 sin6 = (struct sockaddr_in6 *)&ss; 1315 bcopy(sa, sin6, sizeof(*sin6)); 1316 if (sa6_recoverscope(sin6) == 0) 1317 sa = (struct sockaddr *)sin6; 1318 } 1319 #endif 1320 bcopy((caddr_t)sa, cp, (unsigned)dlen); 1321 cp += dlen; 1322 buflen -= dlen; 1323 } else if (cp != NULL) { 1324 /* 1325 * Buffer too small. Count needed size 1326 * and return with error. 1327 */ 1328 cp = NULL; 1329 } 1330 1331 len += dlen; 1332 } 1333 1334 if (cp != NULL) { 1335 dlen = ALIGN(len) - len; 1336 if (buflen < dlen) 1337 cp = NULL; 1338 else { 1339 bzero(cp, dlen); 1340 cp += dlen; 1341 buflen -= dlen; 1342 } 1343 } 1344 len = ALIGN(len); 1345 1346 if (cp != NULL) { 1347 /* fill header iff buffer is large enough */ 1348 rtm->rtm_version = RTM_VERSION; 1349 rtm->rtm_type = type; 1350 rtm->rtm_msglen = len; 1351 } 1352 1353 *plen = len; 1354 1355 if (w != NULL && cp == NULL) 1356 return (ENOBUFS); 1357 1358 return (0); 1359 } 1360 1361 /* 1362 * This routine is called to generate a message from the routing 1363 * socket indicating that a redirect has occurred, a routing lookup 1364 * has failed, or that a protocol has detected timeouts to a particular 1365 * destination. 1366 */ 1367 void 1368 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error, 1369 int fibnum) 1370 { 1371 struct rt_msghdr *rtm; 1372 struct mbuf *m; 1373 struct sockaddr *sa = rtinfo->rti_info[RTAX_DST]; 1374 1375 if (V_route_cb.any_count == 0) 1376 return; 1377 m = rtsock_msg_mbuf(type, rtinfo); 1378 if (m == NULL) 1379 return; 1380 1381 if (fibnum != RT_ALL_FIBS) { 1382 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out " 1383 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs)); 1384 M_SETFIB(m, fibnum); 1385 m->m_flags |= RTS_FILTER_FIB; 1386 } 1387 1388 rtm = mtod(m, struct rt_msghdr *); 1389 rtm->rtm_flags = RTF_DONE | flags; 1390 rtm->rtm_errno = error; 1391 rtm->rtm_addrs = rtinfo->rti_addrs; 1392 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1393 } 1394 1395 void 1396 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error) 1397 { 1398 1399 rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS); 1400 } 1401 1402 /* 1403 * This routine is called to generate a message from the routing 1404 * socket indicating that the status of a network interface has changed. 1405 */ 1406 void 1407 rt_ifmsg(struct ifnet *ifp) 1408 { 1409 struct if_msghdr *ifm; 1410 struct mbuf *m; 1411 struct rt_addrinfo info; 1412 1413 if (V_route_cb.any_count == 0) 1414 return; 1415 bzero((caddr_t)&info, sizeof(info)); 1416 m = rtsock_msg_mbuf(RTM_IFINFO, &info); 1417 if (m == NULL) 1418 return; 1419 ifm = mtod(m, struct if_msghdr *); 1420 ifm->ifm_index = ifp->if_index; 1421 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1422 if_data_copy(ifp, &ifm->ifm_data); 1423 ifm->ifm_addrs = 0; 1424 rt_dispatch(m, AF_UNSPEC); 1425 } 1426 1427 /* 1428 * Announce interface address arrival/withdraw. 1429 * Please do not call directly, use rt_addrmsg(). 1430 * Assume input data to be valid. 1431 * Returns 0 on success. 1432 */ 1433 int 1434 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum) 1435 { 1436 struct rt_addrinfo info; 1437 struct sockaddr *sa; 1438 int ncmd; 1439 struct mbuf *m; 1440 struct ifa_msghdr *ifam; 1441 struct ifnet *ifp = ifa->ifa_ifp; 1442 struct sockaddr_storage ss; 1443 1444 if (V_route_cb.any_count == 0) 1445 return (0); 1446 1447 ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR; 1448 1449 bzero((caddr_t)&info, sizeof(info)); 1450 info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr; 1451 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr; 1452 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask( 1453 info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss); 1454 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr; 1455 if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL) 1456 return (ENOBUFS); 1457 ifam = mtod(m, struct ifa_msghdr *); 1458 ifam->ifam_index = ifp->if_index; 1459 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 1460 ifam->ifam_flags = ifa->ifa_flags; 1461 ifam->ifam_addrs = info.rti_addrs; 1462 1463 if (fibnum != RT_ALL_FIBS) { 1464 M_SETFIB(m, fibnum); 1465 m->m_flags |= RTS_FILTER_FIB; 1466 } 1467 1468 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1469 1470 return (0); 1471 } 1472 1473 /* 1474 * Announce route addition/removal to rtsock based on @rt data. 1475 * Callers are advives to use rt_routemsg() instead of using this 1476 * function directly. 1477 * Assume @rt data is consistent. 1478 * 1479 * Returns 0 on success. 1480 */ 1481 int 1482 rtsock_routemsg(int cmd, struct rtentry *rt, struct ifnet *ifp, int rti_addrs, 1483 int fibnum) 1484 { 1485 struct sockaddr_storage ss; 1486 struct rt_addrinfo info; 1487 1488 if (V_route_cb.any_count == 0) 1489 return (0); 1490 1491 bzero((caddr_t)&info, sizeof(info)); 1492 info.rti_info[RTAX_DST] = rt_key(rt); 1493 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt), rt_mask(rt), &ss); 1494 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 1495 info.rti_flags = rt->rt_flags; 1496 info.rti_ifp = ifp; 1497 1498 return (rtsock_routemsg_info(cmd, &info, fibnum)); 1499 } 1500 1501 int 1502 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum) 1503 { 1504 struct rt_msghdr *rtm; 1505 struct sockaddr *sa; 1506 struct mbuf *m; 1507 1508 if (V_route_cb.any_count == 0) 1509 return (0); 1510 1511 if (info->rti_flags & RTF_HOST) 1512 info->rti_info[RTAX_NETMASK] = NULL; 1513 1514 m = rtsock_msg_mbuf(cmd, info); 1515 if (m == NULL) 1516 return (ENOBUFS); 1517 1518 if (fibnum != RT_ALL_FIBS) { 1519 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out " 1520 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs)); 1521 M_SETFIB(m, fibnum); 1522 m->m_flags |= RTS_FILTER_FIB; 1523 } 1524 1525 rtm = mtod(m, struct rt_msghdr *); 1526 rtm->rtm_addrs = info->rti_addrs; 1527 if (info->rti_ifp != NULL) 1528 rtm->rtm_index = info->rti_ifp->if_index; 1529 /* Add RTF_DONE to indicate command 'completion' required by API */ 1530 info->rti_flags |= RTF_DONE; 1531 /* Reported routes has to be up */ 1532 if (cmd == RTM_ADD || cmd == RTM_CHANGE) 1533 info->rti_flags |= RTF_UP; 1534 rtm->rtm_flags = info->rti_flags; 1535 1536 sa = info->rti_info[RTAX_DST]; 1537 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1538 1539 return (0); 1540 } 1541 1542 /* 1543 * This is the analogue to the rt_newaddrmsg which performs the same 1544 * function but for multicast group memberhips. This is easier since 1545 * there is no route state to worry about. 1546 */ 1547 void 1548 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma) 1549 { 1550 struct rt_addrinfo info; 1551 struct mbuf *m = NULL; 1552 struct ifnet *ifp = ifma->ifma_ifp; 1553 struct ifma_msghdr *ifmam; 1554 1555 if (V_route_cb.any_count == 0) 1556 return; 1557 1558 bzero((caddr_t)&info, sizeof(info)); 1559 info.rti_info[RTAX_IFA] = ifma->ifma_addr; 1560 if (ifp && ifp->if_addr) 1561 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr; 1562 else 1563 info.rti_info[RTAX_IFP] = NULL; 1564 /* 1565 * If a link-layer address is present, present it as a ``gateway'' 1566 * (similarly to how ARP entries, e.g., are presented). 1567 */ 1568 info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr; 1569 m = rtsock_msg_mbuf(cmd, &info); 1570 if (m == NULL) 1571 return; 1572 ifmam = mtod(m, struct ifma_msghdr *); 1573 KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n", 1574 __func__)); 1575 ifmam->ifmam_index = ifp->if_index; 1576 ifmam->ifmam_addrs = info.rti_addrs; 1577 rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC); 1578 } 1579 1580 static struct mbuf * 1581 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what, 1582 struct rt_addrinfo *info) 1583 { 1584 struct if_announcemsghdr *ifan; 1585 struct mbuf *m; 1586 1587 if (V_route_cb.any_count == 0) 1588 return NULL; 1589 bzero((caddr_t)info, sizeof(*info)); 1590 m = rtsock_msg_mbuf(type, info); 1591 if (m != NULL) { 1592 ifan = mtod(m, struct if_announcemsghdr *); 1593 ifan->ifan_index = ifp->if_index; 1594 strlcpy(ifan->ifan_name, ifp->if_xname, 1595 sizeof(ifan->ifan_name)); 1596 ifan->ifan_what = what; 1597 } 1598 return m; 1599 } 1600 1601 /* 1602 * This is called to generate routing socket messages indicating 1603 * IEEE80211 wireless events. 1604 * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way. 1605 */ 1606 void 1607 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len) 1608 { 1609 struct mbuf *m; 1610 struct rt_addrinfo info; 1611 1612 m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info); 1613 if (m != NULL) { 1614 /* 1615 * Append the ieee80211 data. Try to stick it in the 1616 * mbuf containing the ifannounce msg; otherwise allocate 1617 * a new mbuf and append. 1618 * 1619 * NB: we assume m is a single mbuf. 1620 */ 1621 if (data_len > M_TRAILINGSPACE(m)) { 1622 struct mbuf *n = m_get(M_NOWAIT, MT_DATA); 1623 if (n == NULL) { 1624 m_freem(m); 1625 return; 1626 } 1627 bcopy(data, mtod(n, void *), data_len); 1628 n->m_len = data_len; 1629 m->m_next = n; 1630 } else if (data_len > 0) { 1631 bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len); 1632 m->m_len += data_len; 1633 } 1634 if (m->m_flags & M_PKTHDR) 1635 m->m_pkthdr.len += data_len; 1636 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len; 1637 rt_dispatch(m, AF_UNSPEC); 1638 } 1639 } 1640 1641 /* 1642 * This is called to generate routing socket messages indicating 1643 * network interface arrival and departure. 1644 */ 1645 void 1646 rt_ifannouncemsg(struct ifnet *ifp, int what) 1647 { 1648 struct mbuf *m; 1649 struct rt_addrinfo info; 1650 1651 m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info); 1652 if (m != NULL) 1653 rt_dispatch(m, AF_UNSPEC); 1654 } 1655 1656 static void 1657 rt_dispatch(struct mbuf *m, sa_family_t saf) 1658 { 1659 struct m_tag *tag; 1660 1661 /* 1662 * Preserve the family from the sockaddr, if any, in an m_tag for 1663 * use when injecting the mbuf into the routing socket buffer from 1664 * the netisr. 1665 */ 1666 if (saf != AF_UNSPEC) { 1667 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short), 1668 M_NOWAIT); 1669 if (tag == NULL) { 1670 m_freem(m); 1671 return; 1672 } 1673 *(unsigned short *)(tag + 1) = saf; 1674 m_tag_prepend(m, tag); 1675 } 1676 #ifdef VIMAGE 1677 if (V_loif) 1678 m->m_pkthdr.rcvif = V_loif; 1679 else { 1680 m_freem(m); 1681 return; 1682 } 1683 #endif 1684 netisr_queue(NETISR_ROUTE, m); /* mbuf is free'd on failure. */ 1685 } 1686 1687 /* 1688 * Checks if rte can be exported v.r.t jails/vnets. 1689 * 1690 * Returns 1 if it can, 0 otherwise. 1691 */ 1692 static int 1693 can_export_rte(struct ucred *td_ucred, const struct rtentry *rt) 1694 { 1695 1696 if ((rt->rt_flags & RTF_HOST) == 0 1697 ? jailed_without_vnet(td_ucred) 1698 : prison_if(td_ucred, rt_key_const(rt)) != 0) 1699 return (0); 1700 return (1); 1701 } 1702 1703 /* 1704 * This is used in dumping the kernel table via sysctl(). 1705 */ 1706 static int 1707 sysctl_dumpentry(struct radix_node *rn, void *vw) 1708 { 1709 struct walkarg *w = vw; 1710 struct rtentry *rt = (struct rtentry *)rn; 1711 int error = 0, size; 1712 struct rt_addrinfo info; 1713 struct sockaddr_storage ss; 1714 1715 NET_EPOCH_ASSERT(); 1716 1717 if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg)) 1718 return 0; 1719 if (!can_export_rte(w->w_req->td->td_ucred, rt)) 1720 return (0); 1721 bzero((caddr_t)&info, sizeof(info)); 1722 info.rti_info[RTAX_DST] = rt_key(rt); 1723 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 1724 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt), 1725 rt_mask(rt), &ss); 1726 info.rti_info[RTAX_GENMASK] = 0; 1727 if (rt->rt_ifp && !(rt->rt_ifp->if_flags & IFF_DYING)) { 1728 info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr; 1729 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr; 1730 if (rt->rt_ifp->if_flags & IFF_POINTOPOINT) 1731 info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr; 1732 } 1733 if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0) 1734 return (error); 1735 if (w->w_req && w->w_tmem) { 1736 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem; 1737 1738 bzero(&rtm->rtm_index, 1739 sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index)); 1740 if (rt->rt_flags & RTF_GWFLAG_COMPAT) 1741 rtm->rtm_flags = RTF_GATEWAY | 1742 (rt->rt_flags & ~RTF_GWFLAG_COMPAT); 1743 else 1744 rtm->rtm_flags = rt->rt_flags; 1745 rt_getmetrics(rt, &rtm->rtm_rmx); 1746 rtm->rtm_index = rt->rt_ifp->if_index; 1747 rtm->rtm_addrs = info.rti_addrs; 1748 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size); 1749 return (error); 1750 } 1751 return (error); 1752 } 1753 1754 static int 1755 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd, 1756 struct rt_addrinfo *info, struct walkarg *w, int len) 1757 { 1758 struct if_msghdrl *ifm; 1759 struct if_data *ifd; 1760 1761 ifm = (struct if_msghdrl *)w->w_tmem; 1762 1763 #ifdef COMPAT_FREEBSD32 1764 if (w->w_req->flags & SCTL_MASK32) { 1765 struct if_msghdrl32 *ifm32; 1766 1767 ifm32 = (struct if_msghdrl32 *)ifm; 1768 ifm32->ifm_addrs = info->rti_addrs; 1769 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1770 ifm32->ifm_index = ifp->if_index; 1771 ifm32->_ifm_spare1 = 0; 1772 ifm32->ifm_len = sizeof(*ifm32); 1773 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data); 1774 ifm32->_ifm_spare2 = 0; 1775 ifd = &ifm32->ifm_data; 1776 } else 1777 #endif 1778 { 1779 ifm->ifm_addrs = info->rti_addrs; 1780 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1781 ifm->ifm_index = ifp->if_index; 1782 ifm->_ifm_spare1 = 0; 1783 ifm->ifm_len = sizeof(*ifm); 1784 ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data); 1785 ifm->_ifm_spare2 = 0; 1786 ifd = &ifm->ifm_data; 1787 } 1788 1789 memcpy(ifd, src_ifd, sizeof(*ifd)); 1790 1791 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len)); 1792 } 1793 1794 static int 1795 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd, 1796 struct rt_addrinfo *info, struct walkarg *w, int len) 1797 { 1798 struct if_msghdr *ifm; 1799 struct if_data *ifd; 1800 1801 ifm = (struct if_msghdr *)w->w_tmem; 1802 1803 #ifdef COMPAT_FREEBSD32 1804 if (w->w_req->flags & SCTL_MASK32) { 1805 struct if_msghdr32 *ifm32; 1806 1807 ifm32 = (struct if_msghdr32 *)ifm; 1808 ifm32->ifm_addrs = info->rti_addrs; 1809 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1810 ifm32->ifm_index = ifp->if_index; 1811 ifm32->_ifm_spare1 = 0; 1812 ifd = &ifm32->ifm_data; 1813 } else 1814 #endif 1815 { 1816 ifm->ifm_addrs = info->rti_addrs; 1817 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1818 ifm->ifm_index = ifp->if_index; 1819 ifm->_ifm_spare1 = 0; 1820 ifd = &ifm->ifm_data; 1821 } 1822 1823 memcpy(ifd, src_ifd, sizeof(*ifd)); 1824 1825 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len)); 1826 } 1827 1828 static int 1829 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info, 1830 struct walkarg *w, int len) 1831 { 1832 struct ifa_msghdrl *ifam; 1833 struct if_data *ifd; 1834 1835 ifam = (struct ifa_msghdrl *)w->w_tmem; 1836 1837 #ifdef COMPAT_FREEBSD32 1838 if (w->w_req->flags & SCTL_MASK32) { 1839 struct ifa_msghdrl32 *ifam32; 1840 1841 ifam32 = (struct ifa_msghdrl32 *)ifam; 1842 ifam32->ifam_addrs = info->rti_addrs; 1843 ifam32->ifam_flags = ifa->ifa_flags; 1844 ifam32->ifam_index = ifa->ifa_ifp->if_index; 1845 ifam32->_ifam_spare1 = 0; 1846 ifam32->ifam_len = sizeof(*ifam32); 1847 ifam32->ifam_data_off = 1848 offsetof(struct ifa_msghdrl32, ifam_data); 1849 ifam32->ifam_metric = ifa->ifa_ifp->if_metric; 1850 ifd = &ifam32->ifam_data; 1851 } else 1852 #endif 1853 { 1854 ifam->ifam_addrs = info->rti_addrs; 1855 ifam->ifam_flags = ifa->ifa_flags; 1856 ifam->ifam_index = ifa->ifa_ifp->if_index; 1857 ifam->_ifam_spare1 = 0; 1858 ifam->ifam_len = sizeof(*ifam); 1859 ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data); 1860 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 1861 ifd = &ifam->ifam_data; 1862 } 1863 1864 bzero(ifd, sizeof(*ifd)); 1865 ifd->ifi_datalen = sizeof(struct if_data); 1866 ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets); 1867 ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets); 1868 ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes); 1869 ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes); 1870 1871 /* Fixup if_data carp(4) vhid. */ 1872 if (carp_get_vhid_p != NULL) 1873 ifd->ifi_vhid = (*carp_get_vhid_p)(ifa); 1874 1875 return (SYSCTL_OUT(w->w_req, w->w_tmem, len)); 1876 } 1877 1878 static int 1879 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info, 1880 struct walkarg *w, int len) 1881 { 1882 struct ifa_msghdr *ifam; 1883 1884 ifam = (struct ifa_msghdr *)w->w_tmem; 1885 ifam->ifam_addrs = info->rti_addrs; 1886 ifam->ifam_flags = ifa->ifa_flags; 1887 ifam->ifam_index = ifa->ifa_ifp->if_index; 1888 ifam->_ifam_spare1 = 0; 1889 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 1890 1891 return (SYSCTL_OUT(w->w_req, w->w_tmem, len)); 1892 } 1893 1894 static int 1895 sysctl_iflist(int af, struct walkarg *w) 1896 { 1897 struct ifnet *ifp; 1898 struct ifaddr *ifa; 1899 struct if_data ifd; 1900 struct rt_addrinfo info; 1901 int len, error = 0; 1902 struct sockaddr_storage ss; 1903 1904 bzero((caddr_t)&info, sizeof(info)); 1905 bzero(&ifd, sizeof(ifd)); 1906 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1907 if (w->w_arg && w->w_arg != ifp->if_index) 1908 continue; 1909 if_data_copy(ifp, &ifd); 1910 ifa = ifp->if_addr; 1911 info.rti_info[RTAX_IFP] = ifa->ifa_addr; 1912 error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len); 1913 if (error != 0) 1914 goto done; 1915 info.rti_info[RTAX_IFP] = NULL; 1916 if (w->w_req && w->w_tmem) { 1917 if (w->w_op == NET_RT_IFLISTL) 1918 error = sysctl_iflist_ifml(ifp, &ifd, &info, w, 1919 len); 1920 else 1921 error = sysctl_iflist_ifm(ifp, &ifd, &info, w, 1922 len); 1923 if (error) 1924 goto done; 1925 } 1926 while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) { 1927 if (af && af != ifa->ifa_addr->sa_family) 1928 continue; 1929 if (prison_if(w->w_req->td->td_ucred, 1930 ifa->ifa_addr) != 0) 1931 continue; 1932 info.rti_info[RTAX_IFA] = ifa->ifa_addr; 1933 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask( 1934 ifa->ifa_addr, ifa->ifa_netmask, &ss); 1935 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr; 1936 error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len); 1937 if (error != 0) 1938 goto done; 1939 if (w->w_req && w->w_tmem) { 1940 if (w->w_op == NET_RT_IFLISTL) 1941 error = sysctl_iflist_ifaml(ifa, &info, 1942 w, len); 1943 else 1944 error = sysctl_iflist_ifam(ifa, &info, 1945 w, len); 1946 if (error) 1947 goto done; 1948 } 1949 } 1950 info.rti_info[RTAX_IFA] = NULL; 1951 info.rti_info[RTAX_NETMASK] = NULL; 1952 info.rti_info[RTAX_BRD] = NULL; 1953 } 1954 done: 1955 return (error); 1956 } 1957 1958 static int 1959 sysctl_ifmalist(int af, struct walkarg *w) 1960 { 1961 struct rt_addrinfo info; 1962 struct ifaddr *ifa; 1963 struct ifmultiaddr *ifma; 1964 struct ifnet *ifp; 1965 int error, len; 1966 1967 NET_EPOCH_ASSERT(); 1968 1969 error = 0; 1970 bzero((caddr_t)&info, sizeof(info)); 1971 1972 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1973 if (w->w_arg && w->w_arg != ifp->if_index) 1974 continue; 1975 ifa = ifp->if_addr; 1976 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL; 1977 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 1978 if (af && af != ifma->ifma_addr->sa_family) 1979 continue; 1980 if (prison_if(w->w_req->td->td_ucred, 1981 ifma->ifma_addr) != 0) 1982 continue; 1983 info.rti_info[RTAX_IFA] = ifma->ifma_addr; 1984 info.rti_info[RTAX_GATEWAY] = 1985 (ifma->ifma_addr->sa_family != AF_LINK) ? 1986 ifma->ifma_lladdr : NULL; 1987 error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len); 1988 if (error != 0) 1989 break; 1990 if (w->w_req && w->w_tmem) { 1991 struct ifma_msghdr *ifmam; 1992 1993 ifmam = (struct ifma_msghdr *)w->w_tmem; 1994 ifmam->ifmam_index = ifma->ifma_ifp->if_index; 1995 ifmam->ifmam_flags = 0; 1996 ifmam->ifmam_addrs = info.rti_addrs; 1997 ifmam->_ifmam_spare1 = 0; 1998 error = SYSCTL_OUT(w->w_req, w->w_tmem, len); 1999 if (error != 0) 2000 break; 2001 } 2002 } 2003 if (error != 0) 2004 break; 2005 } 2006 return (error); 2007 } 2008 2009 static int 2010 sysctl_rtsock(SYSCTL_HANDLER_ARGS) 2011 { 2012 RIB_RLOCK_TRACKER; 2013 struct epoch_tracker et; 2014 int *name = (int *)arg1; 2015 u_int namelen = arg2; 2016 struct rib_head *rnh = NULL; /* silence compiler. */ 2017 int i, lim, error = EINVAL; 2018 int fib = 0; 2019 u_char af; 2020 struct walkarg w; 2021 2022 name ++; 2023 namelen--; 2024 if (req->newptr) 2025 return (EPERM); 2026 if (name[1] == NET_RT_DUMP) { 2027 if (namelen == 3) 2028 fib = req->td->td_proc->p_fibnum; 2029 else if (namelen == 4) 2030 fib = (name[3] == RT_ALL_FIBS) ? 2031 req->td->td_proc->p_fibnum : name[3]; 2032 else 2033 return ((namelen < 3) ? EISDIR : ENOTDIR); 2034 if (fib < 0 || fib >= rt_numfibs) 2035 return (EINVAL); 2036 } else if (namelen != 3) 2037 return ((namelen < 3) ? EISDIR : ENOTDIR); 2038 af = name[0]; 2039 if (af > AF_MAX) 2040 return (EINVAL); 2041 bzero(&w, sizeof(w)); 2042 w.w_op = name[1]; 2043 w.w_arg = name[2]; 2044 w.w_req = req; 2045 2046 error = sysctl_wire_old_buffer(req, 0); 2047 if (error) 2048 return (error); 2049 2050 /* 2051 * Allocate reply buffer in advance. 2052 * All rtsock messages has maximum length of u_short. 2053 */ 2054 w.w_tmemsize = 65536; 2055 w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK); 2056 2057 NET_EPOCH_ENTER(et); 2058 switch (w.w_op) { 2059 case NET_RT_DUMP: 2060 case NET_RT_FLAGS: 2061 if (af == 0) { /* dump all tables */ 2062 i = 1; 2063 lim = AF_MAX; 2064 } else /* dump only one table */ 2065 i = lim = af; 2066 2067 /* 2068 * take care of llinfo entries, the caller must 2069 * specify an AF 2070 */ 2071 if (w.w_op == NET_RT_FLAGS && 2072 (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) { 2073 if (af != 0) 2074 error = lltable_sysctl_dumparp(af, w.w_req); 2075 else 2076 error = EINVAL; 2077 break; 2078 } 2079 /* 2080 * take care of routing entries 2081 */ 2082 for (error = 0; error == 0 && i <= lim; i++) { 2083 rnh = rt_tables_get_rnh(fib, i); 2084 if (rnh != NULL) { 2085 RIB_RLOCK(rnh); 2086 error = rnh->rnh_walktree(&rnh->head, 2087 sysctl_dumpentry, &w); 2088 RIB_RUNLOCK(rnh); 2089 } else if (af != 0) 2090 error = EAFNOSUPPORT; 2091 } 2092 break; 2093 2094 case NET_RT_IFLIST: 2095 case NET_RT_IFLISTL: 2096 error = sysctl_iflist(af, &w); 2097 break; 2098 2099 case NET_RT_IFMALIST: 2100 error = sysctl_ifmalist(af, &w); 2101 break; 2102 } 2103 NET_EPOCH_EXIT(et); 2104 2105 free(w.w_tmem, M_TEMP); 2106 return (error); 2107 } 2108 2109 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE, 2110 sysctl_rtsock, "Return route tables and interface/address lists"); 2111 2112 /* 2113 * Definitions of protocols supported in the ROUTE domain. 2114 */ 2115 2116 static struct domain routedomain; /* or at least forward */ 2117 2118 static struct protosw routesw[] = { 2119 { 2120 .pr_type = SOCK_RAW, 2121 .pr_domain = &routedomain, 2122 .pr_flags = PR_ATOMIC|PR_ADDR, 2123 .pr_output = route_output, 2124 .pr_ctlinput = raw_ctlinput, 2125 .pr_init = raw_init, 2126 .pr_usrreqs = &route_usrreqs 2127 } 2128 }; 2129 2130 static struct domain routedomain = { 2131 .dom_family = PF_ROUTE, 2132 .dom_name = "route", 2133 .dom_protosw = routesw, 2134 .dom_protoswNPROTOSW = &routesw[nitems(routesw)] 2135 }; 2136 2137 VNET_DOMAIN_SET(route); 2138 2139 #ifdef DDB 2140 /* 2141 * Unfortunately, RTF_ values are expressed as raw masks rather than powers of 2142 * 2, so we cannot use them as nice C99 initializer indices below. 2143 */ 2144 static const char * const rtf_flag_strings[] = { 2145 "UP", 2146 "GATEWAY", 2147 "HOST", 2148 "REJECT", 2149 "DYNAMIC", 2150 "MODIFIED", 2151 "DONE", 2152 "UNUSED_0x80", 2153 "UNUSED_0x100", 2154 "XRESOLVE", 2155 "LLDATA", 2156 "STATIC", 2157 "BLACKHOLE", 2158 "UNUSED_0x2000", 2159 "PROTO2", 2160 "PROTO1", 2161 "UNUSED_0x10000", 2162 "UNUSED_0x20000", 2163 "PROTO3", 2164 "FIXEDMTU", 2165 "PINNED", 2166 "LOCAL", 2167 "BROADCAST", 2168 "MULTICAST", 2169 /* Big gap. */ 2170 [28] = "STICKY", 2171 [30] = "RNH_LOCKED", 2172 [31] = "GWFLAG_COMPAT", 2173 }; 2174 2175 static const char * __pure 2176 rt_flag_name(unsigned idx) 2177 { 2178 if (idx >= nitems(rtf_flag_strings)) 2179 return ("INVALID_FLAG"); 2180 if (rtf_flag_strings[idx] == NULL) 2181 return ("UNKNOWN"); 2182 return (rtf_flag_strings[idx]); 2183 } 2184 2185 static void 2186 rt_dumpaddr_ddb(const char *name, const struct sockaddr *sa) 2187 { 2188 char buf[INET6_ADDRSTRLEN], *res; 2189 2190 res = NULL; 2191 if (sa == NULL) 2192 res = "NULL"; 2193 else if (sa->sa_family == AF_INET) { 2194 res = inet_ntop(AF_INET, 2195 &((const struct sockaddr_in *)sa)->sin_addr, 2196 buf, sizeof(buf)); 2197 } else if (sa->sa_family == AF_INET6) { 2198 res = inet_ntop(AF_INET6, 2199 &((const struct sockaddr_in6 *)sa)->sin6_addr, 2200 buf, sizeof(buf)); 2201 } else if (sa->sa_family == AF_LINK) { 2202 res = "on link"; 2203 } 2204 2205 if (res != NULL) { 2206 db_printf("%s <%s> ", name, res); 2207 return; 2208 } 2209 2210 db_printf("%s <af:%d> ", name, sa->sa_family); 2211 } 2212 2213 static int 2214 rt_dumpentry_ddb(struct radix_node *rn, void *arg __unused) 2215 { 2216 struct sockaddr_storage ss; 2217 struct rtentry *rt; 2218 int flags, idx; 2219 2220 /* If RNTORT is important, put it in a header. */ 2221 rt = (void *)rn; 2222 2223 rt_dumpaddr_ddb("dst", rt_key(rt)); 2224 rt_dumpaddr_ddb("gateway", rt->rt_gateway); 2225 rt_dumpaddr_ddb("netmask", rtsock_fix_netmask(rt_key(rt), rt_mask(rt), 2226 &ss)); 2227 if (rt->rt_ifp != NULL && (rt->rt_ifp->if_flags & IFF_DYING) == 0) { 2228 rt_dumpaddr_ddb("ifp", rt->rt_ifp->if_addr->ifa_addr); 2229 rt_dumpaddr_ddb("ifa", rt->rt_ifa->ifa_addr); 2230 } 2231 2232 db_printf("flags "); 2233 flags = rt->rt_flags; 2234 if (flags == 0) 2235 db_printf("none"); 2236 2237 while ((idx = ffs(flags)) > 0) { 2238 idx--; 2239 2240 if (flags != rt->rt_flags) 2241 db_printf(","); 2242 db_printf("%s", rt_flag_name(idx)); 2243 2244 flags &= ~(1ul << idx); 2245 } 2246 2247 db_printf("\n"); 2248 return (0); 2249 } 2250 2251 DB_SHOW_COMMAND(routetable, db_show_routetable_cmd) 2252 { 2253 struct rib_head *rnh; 2254 int error, i, lim; 2255 2256 if (have_addr) 2257 i = lim = addr; 2258 else { 2259 i = 1; 2260 lim = AF_MAX; 2261 } 2262 2263 for (; i <= lim; i++) { 2264 rnh = rt_tables_get_rnh(0, i); 2265 if (rnh == NULL) { 2266 if (have_addr) { 2267 db_printf("%s: AF %d not supported?\n", 2268 __func__, i); 2269 break; 2270 } 2271 continue; 2272 } 2273 2274 if (!have_addr && i > 1) 2275 db_printf("\n"); 2276 2277 db_printf("Route table for AF %d%s%s%s:\n", i, 2278 (i == AF_INET || i == AF_INET6) ? " (" : "", 2279 (i == AF_INET) ? "INET" : (i == AF_INET6) ? "INET6" : "", 2280 (i == AF_INET || i == AF_INET6) ? ")" : ""); 2281 2282 error = rnh->rnh_walktree(&rnh->head, rt_dumpentry_ddb, NULL); 2283 if (error != 0) 2284 db_printf("%s: walktree(%d): %d\n", __func__, i, 2285 error); 2286 } 2287 } 2288 2289 _DB_FUNC(_show, route, db_show_route_cmd, db_show_table, CS_OWN, NULL) 2290 { 2291 char buf[INET6_ADDRSTRLEN], *bp; 2292 const void *dst_addrp; 2293 struct sockaddr *dstp; 2294 struct rtentry *rt; 2295 union { 2296 struct sockaddr_in dest_sin; 2297 struct sockaddr_in6 dest_sin6; 2298 } u; 2299 uint16_t hextets[8]; 2300 unsigned i, tets; 2301 int t, af, exp, tokflags; 2302 2303 /* 2304 * Undecoded address family. No double-colon expansion seen yet. 2305 */ 2306 af = -1; 2307 exp = -1; 2308 /* Assume INET6 to start; we can work back if guess was wrong. */ 2309 tokflags = DRT_WSPACE | DRT_HEX | DRT_HEXADECIMAL; 2310 2311 /* 2312 * db_command has lexed 'show route' for us. 2313 */ 2314 t = db_read_token_flags(tokflags); 2315 if (t == tWSPACE) 2316 t = db_read_token_flags(tokflags); 2317 2318 /* 2319 * tEOL: Just 'show route' isn't a valid mode. 2320 * tMINUS: It's either '-h' or some invalid option. Regardless, usage. 2321 */ 2322 if (t == tEOL || t == tMINUS) 2323 goto usage; 2324 2325 db_unread_token(t); 2326 2327 tets = nitems(hextets); 2328 2329 /* 2330 * Each loop iteration, we expect to read one octet (v4) or hextet 2331 * (v6), followed by an appropriate field separator ('.' or ':' or 2332 * '::'). 2333 * 2334 * At the start of each loop, we're looking for a number (octet or 2335 * hextet). 2336 * 2337 * INET6 addresses have a special case where they may begin with '::'. 2338 */ 2339 for (i = 0; i < tets; i++) { 2340 t = db_read_token_flags(tokflags); 2341 2342 if (t == tCOLONCOLON) { 2343 /* INET6 with leading '::' or invalid. */ 2344 if (i != 0) { 2345 db_printf("Parse error: unexpected extra " 2346 "colons.\n"); 2347 goto exit; 2348 } 2349 2350 af = AF_INET6; 2351 exp = i; 2352 hextets[i] = 0; 2353 continue; 2354 } else if (t == tNUMBER) { 2355 /* 2356 * Lexer separates out '-' as tMINUS, but make the 2357 * assumption explicit here. 2358 */ 2359 MPASS(db_tok_number >= 0); 2360 2361 if (af == AF_INET && db_tok_number > UINT8_MAX) { 2362 db_printf("Not a valid v4 octet: %ld\n", 2363 (long)db_tok_number); 2364 goto exit; 2365 } 2366 hextets[i] = db_tok_number; 2367 } else if (t == tEOL) { 2368 /* 2369 * We can only detect the end of an IPv6 address in 2370 * compact representation with EOL. 2371 */ 2372 if (af != AF_INET6 || exp < 0) { 2373 db_printf("Parse failed. Got unexpected EOF " 2374 "when the address is not a compact-" 2375 "representation IPv6 address.\n"); 2376 goto exit; 2377 } 2378 break; 2379 } else { 2380 db_printf("Parse failed. Unexpected token %d.\n", t); 2381 goto exit; 2382 } 2383 2384 /* Next, look for a separator, if appropriate. */ 2385 if (i == tets - 1) 2386 continue; 2387 2388 t = db_read_token_flags(tokflags); 2389 if (af < 0) { 2390 if (t == tCOLON) { 2391 af = AF_INET6; 2392 continue; 2393 } 2394 if (t == tCOLONCOLON) { 2395 af = AF_INET6; 2396 i++; 2397 hextets[i] = 0; 2398 exp = i; 2399 continue; 2400 } 2401 if (t == tDOT) { 2402 unsigned hn, dn; 2403 2404 af = AF_INET; 2405 /* Need to fixup the first parsed number. */ 2406 if (hextets[0] > 0x255 || 2407 (hextets[0] & 0xf0) > 0x90 || 2408 (hextets[0] & 0xf) > 9) { 2409 db_printf("Not a valid v4 octet: %x\n", 2410 hextets[0]); 2411 goto exit; 2412 } 2413 2414 hn = hextets[0]; 2415 dn = (hn >> 8) * 100 + 2416 ((hn >> 4) & 0xf) * 10 + 2417 (hn & 0xf); 2418 2419 hextets[0] = dn; 2420 2421 /* Switch to decimal for remaining octets. */ 2422 tokflags &= ~DRT_RADIX_MASK; 2423 tokflags |= DRT_DECIMAL; 2424 2425 tets = 4; 2426 continue; 2427 } 2428 2429 db_printf("Parse error. Unexpected token %d.\n", t); 2430 goto exit; 2431 } else if (af == AF_INET) { 2432 if (t == tDOT) 2433 continue; 2434 db_printf("Expected '.' (%d) between octets but got " 2435 "(%d).\n", tDOT, t); 2436 goto exit; 2437 2438 } else if (af == AF_INET6) { 2439 if (t == tCOLON) 2440 continue; 2441 if (t == tCOLONCOLON) { 2442 if (exp < 0) { 2443 i++; 2444 hextets[i] = 0; 2445 exp = i; 2446 continue; 2447 } 2448 db_printf("Got bogus second '::' in v6 " 2449 "address.\n"); 2450 goto exit; 2451 } 2452 if (t == tEOL) { 2453 /* 2454 * Handle in the earlier part of the loop 2455 * because we need to handle trailing :: too. 2456 */ 2457 db_unread_token(t); 2458 continue; 2459 } 2460 2461 db_printf("Expected ':' (%d) or '::' (%d) between " 2462 "hextets but got (%d).\n", tCOLON, tCOLONCOLON, t); 2463 goto exit; 2464 } 2465 } 2466 2467 /* Check for trailing garbage. */ 2468 if (i == tets) { 2469 t = db_read_token_flags(tokflags); 2470 if (t != tEOL) { 2471 db_printf("Got unexpected garbage after address " 2472 "(%d).\n", t); 2473 goto exit; 2474 } 2475 } 2476 2477 /* 2478 * Need to expand compact INET6 addresses. 2479 * 2480 * Technically '::' for a single ':0:' is MUST NOT but just in case, 2481 * don't bother expanding that form (exp >= 0 && i == tets case). 2482 */ 2483 if (af == AF_INET6 && exp >= 0 && i < tets) { 2484 if (exp + 1 < i) { 2485 memmove(&hextets[exp + 1 + (nitems(hextets) - i)], 2486 &hextets[exp + 1], 2487 (i - (exp + 1)) * sizeof(hextets[0])); 2488 } 2489 memset(&hextets[exp + 1], 0, (nitems(hextets) - i) * 2490 sizeof(hextets[0])); 2491 } 2492 2493 memset(&u, 0, sizeof(u)); 2494 if (af == AF_INET) { 2495 u.dest_sin.sin_family = AF_INET; 2496 u.dest_sin.sin_len = sizeof(u.dest_sin); 2497 u.dest_sin.sin_addr.s_addr = htonl( 2498 ((uint32_t)hextets[0] << 24) | 2499 ((uint32_t)hextets[1] << 16) | 2500 ((uint32_t)hextets[2] << 8) | 2501 (uint32_t)hextets[3]); 2502 dstp = (void *)&u.dest_sin; 2503 dst_addrp = &u.dest_sin.sin_addr; 2504 } else if (af == AF_INET6) { 2505 u.dest_sin6.sin6_family = AF_INET6; 2506 u.dest_sin6.sin6_len = sizeof(u.dest_sin6); 2507 for (i = 0; i < nitems(hextets); i++) 2508 u.dest_sin6.sin6_addr.s6_addr16[i] = htons(hextets[i]); 2509 dstp = (void *)&u.dest_sin6; 2510 dst_addrp = &u.dest_sin6.sin6_addr; 2511 } else { 2512 MPASS(false); 2513 /* UNREACHABLE */ 2514 /* Appease Clang false positive: */ 2515 dstp = NULL; 2516 } 2517 2518 bp = inet_ntop(af, dst_addrp, buf, sizeof(buf)); 2519 if (bp != NULL) 2520 db_printf("Looking up route to destination '%s'\n", bp); 2521 2522 CURVNET_SET(vnet0); 2523 rt = rtalloc1(dstp, 0, RTF_RNH_LOCKED); 2524 CURVNET_RESTORE(); 2525 2526 if (rt == NULL) { 2527 db_printf("Could not get route for that server.\n"); 2528 return; 2529 } 2530 2531 rt_dumpentry_ddb((void *)rt, NULL); 2532 RTFREE_LOCKED(rt); 2533 2534 return; 2535 usage: 2536 db_printf("Usage: 'show route <address>'\n" 2537 " Currently accepts only dotted-decimal INET or colon-separated\n" 2538 " hextet INET6 addresses.\n"); 2539 exit: 2540 db_skip_to_eol(); 2541 } 2542 #endif 2543