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 #include <net/route/nhop.h> 81 #include <net/route/shared.h> 82 83 #ifdef COMPAT_FREEBSD32 84 #include <sys/mount.h> 85 #include <compat/freebsd32/freebsd32.h> 86 87 struct if_msghdr32 { 88 uint16_t ifm_msglen; 89 uint8_t ifm_version; 90 uint8_t ifm_type; 91 int32_t ifm_addrs; 92 int32_t ifm_flags; 93 uint16_t ifm_index; 94 uint16_t _ifm_spare1; 95 struct if_data ifm_data; 96 }; 97 98 struct if_msghdrl32 { 99 uint16_t ifm_msglen; 100 uint8_t ifm_version; 101 uint8_t ifm_type; 102 int32_t ifm_addrs; 103 int32_t ifm_flags; 104 uint16_t ifm_index; 105 uint16_t _ifm_spare1; 106 uint16_t ifm_len; 107 uint16_t ifm_data_off; 108 uint32_t _ifm_spare2; 109 struct if_data ifm_data; 110 }; 111 112 struct ifa_msghdrl32 { 113 uint16_t ifam_msglen; 114 uint8_t ifam_version; 115 uint8_t ifam_type; 116 int32_t ifam_addrs; 117 int32_t ifam_flags; 118 uint16_t ifam_index; 119 uint16_t _ifam_spare1; 120 uint16_t ifam_len; 121 uint16_t ifam_data_off; 122 int32_t ifam_metric; 123 struct if_data ifam_data; 124 }; 125 126 #define SA_SIZE32(sa) \ 127 ( (((struct sockaddr *)(sa))->sa_len == 0) ? \ 128 sizeof(int) : \ 129 1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) ) 130 131 #endif /* COMPAT_FREEBSD32 */ 132 133 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables"); 134 135 /* NB: these are not modified */ 136 static struct sockaddr route_src = { 2, PF_ROUTE, }; 137 static struct sockaddr sa_zero = { sizeof(sa_zero), AF_INET, }; 138 139 /* These are external hooks for CARP. */ 140 int (*carp_get_vhid_p)(struct ifaddr *); 141 142 /* 143 * Used by rtsock/raw_input callback code to decide whether to filter the update 144 * notification to a socket bound to a particular FIB. 145 */ 146 #define RTS_FILTER_FIB M_PROTO8 147 148 typedef struct { 149 int ip_count; /* attached w/ AF_INET */ 150 int ip6_count; /* attached w/ AF_INET6 */ 151 int any_count; /* total attached */ 152 } route_cb_t; 153 VNET_DEFINE_STATIC(route_cb_t, route_cb); 154 #define V_route_cb VNET(route_cb) 155 156 struct mtx rtsock_mtx; 157 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF); 158 159 #define RTSOCK_LOCK() mtx_lock(&rtsock_mtx) 160 #define RTSOCK_UNLOCK() mtx_unlock(&rtsock_mtx) 161 #define RTSOCK_LOCK_ASSERT() mtx_assert(&rtsock_mtx, MA_OWNED) 162 163 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 164 ""); 165 166 struct walkarg { 167 int w_tmemsize; 168 int w_op, w_arg; 169 caddr_t w_tmem; 170 struct sysctl_req *w_req; 171 }; 172 173 static void rts_input(struct mbuf *m); 174 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo); 175 static int rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, 176 struct walkarg *w, int *plen); 177 static int rt_xaddrs(caddr_t cp, caddr_t cplim, 178 struct rt_addrinfo *rtinfo); 179 static int sysctl_dumpentry(struct radix_node *rn, void *vw); 180 static int sysctl_iflist(int af, struct walkarg *w); 181 static int sysctl_ifmalist(int af, struct walkarg *w); 182 static int route_output(struct mbuf *m, struct socket *so, ...); 183 static void rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out); 184 static void rt_dispatch(struct mbuf *, sa_family_t); 185 static struct sockaddr *rtsock_fix_netmask(struct sockaddr *dst, 186 struct sockaddr *smask, struct sockaddr_storage *dmask); 187 static int handle_rtm_get(struct rt_addrinfo *info, u_int fibnum, 188 struct rt_msghdr *rtm, struct rtentry **ret_nrt); 189 static int update_rtm_from_rte(struct rt_addrinfo *info, 190 struct rt_msghdr **prtm, int alloc_len, 191 struct rtentry *rt); 192 static void send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, 193 struct mbuf *m, sa_family_t saf, u_int fibnum, 194 int rtm_errno); 195 static int can_export_rte(struct ucred *td_ucred, const struct rtentry *rt); 196 197 static struct netisr_handler rtsock_nh = { 198 .nh_name = "rtsock", 199 .nh_handler = rts_input, 200 .nh_proto = NETISR_ROUTE, 201 .nh_policy = NETISR_POLICY_SOURCE, 202 }; 203 204 static int 205 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS) 206 { 207 int error, qlimit; 208 209 netisr_getqlimit(&rtsock_nh, &qlimit); 210 error = sysctl_handle_int(oidp, &qlimit, 0, req); 211 if (error || !req->newptr) 212 return (error); 213 if (qlimit < 1) 214 return (EINVAL); 215 return (netisr_setqlimit(&rtsock_nh, qlimit)); 216 } 217 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, 218 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 219 0, 0, sysctl_route_netisr_maxqlen, "I", 220 "maximum routing socket dispatch queue length"); 221 222 static void 223 vnet_rts_init(void) 224 { 225 int tmp; 226 227 if (IS_DEFAULT_VNET(curvnet)) { 228 if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp)) 229 rtsock_nh.nh_qlimit = tmp; 230 netisr_register(&rtsock_nh); 231 } 232 #ifdef VIMAGE 233 else 234 netisr_register_vnet(&rtsock_nh); 235 #endif 236 } 237 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 238 vnet_rts_init, 0); 239 240 #ifdef VIMAGE 241 static void 242 vnet_rts_uninit(void) 243 { 244 245 netisr_unregister_vnet(&rtsock_nh); 246 } 247 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 248 vnet_rts_uninit, 0); 249 #endif 250 251 static int 252 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src, 253 struct rawcb *rp) 254 { 255 int fibnum; 256 257 KASSERT(m != NULL, ("%s: m is NULL", __func__)); 258 KASSERT(proto != NULL, ("%s: proto is NULL", __func__)); 259 KASSERT(rp != NULL, ("%s: rp is NULL", __func__)); 260 261 /* No filtering requested. */ 262 if ((m->m_flags & RTS_FILTER_FIB) == 0) 263 return (0); 264 265 /* Check if it is a rts and the fib matches the one of the socket. */ 266 fibnum = M_GETFIB(m); 267 if (proto->sp_family != PF_ROUTE || 268 rp->rcb_socket == NULL || 269 rp->rcb_socket->so_fibnum == fibnum) 270 return (0); 271 272 /* Filtering requested and no match, the socket shall be skipped. */ 273 return (1); 274 } 275 276 static void 277 rts_input(struct mbuf *m) 278 { 279 struct sockproto route_proto; 280 unsigned short *family; 281 struct m_tag *tag; 282 283 route_proto.sp_family = PF_ROUTE; 284 tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL); 285 if (tag != NULL) { 286 family = (unsigned short *)(tag + 1); 287 route_proto.sp_protocol = *family; 288 m_tag_delete(m, tag); 289 } else 290 route_proto.sp_protocol = 0; 291 292 raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb); 293 } 294 295 /* 296 * It really doesn't make any sense at all for this code to share much 297 * with raw_usrreq.c, since its functionality is so restricted. XXX 298 */ 299 static void 300 rts_abort(struct socket *so) 301 { 302 303 raw_usrreqs.pru_abort(so); 304 } 305 306 static void 307 rts_close(struct socket *so) 308 { 309 310 raw_usrreqs.pru_close(so); 311 } 312 313 /* pru_accept is EOPNOTSUPP */ 314 315 static int 316 rts_attach(struct socket *so, int proto, struct thread *td) 317 { 318 struct rawcb *rp; 319 int error; 320 321 KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL")); 322 323 /* XXX */ 324 rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO); 325 326 so->so_pcb = (caddr_t)rp; 327 so->so_fibnum = td->td_proc->p_fibnum; 328 error = raw_attach(so, proto); 329 rp = sotorawcb(so); 330 if (error) { 331 so->so_pcb = NULL; 332 free(rp, M_PCB); 333 return error; 334 } 335 RTSOCK_LOCK(); 336 switch(rp->rcb_proto.sp_protocol) { 337 case AF_INET: 338 V_route_cb.ip_count++; 339 break; 340 case AF_INET6: 341 V_route_cb.ip6_count++; 342 break; 343 } 344 V_route_cb.any_count++; 345 RTSOCK_UNLOCK(); 346 soisconnected(so); 347 so->so_options |= SO_USELOOPBACK; 348 return 0; 349 } 350 351 static int 352 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 353 { 354 355 return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */ 356 } 357 358 static int 359 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 360 { 361 362 return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */ 363 } 364 365 /* pru_connect2 is EOPNOTSUPP */ 366 /* pru_control is EOPNOTSUPP */ 367 368 static void 369 rts_detach(struct socket *so) 370 { 371 struct rawcb *rp = sotorawcb(so); 372 373 KASSERT(rp != NULL, ("rts_detach: rp == NULL")); 374 375 RTSOCK_LOCK(); 376 switch(rp->rcb_proto.sp_protocol) { 377 case AF_INET: 378 V_route_cb.ip_count--; 379 break; 380 case AF_INET6: 381 V_route_cb.ip6_count--; 382 break; 383 } 384 V_route_cb.any_count--; 385 RTSOCK_UNLOCK(); 386 raw_usrreqs.pru_detach(so); 387 } 388 389 static int 390 rts_disconnect(struct socket *so) 391 { 392 393 return (raw_usrreqs.pru_disconnect(so)); 394 } 395 396 /* pru_listen is EOPNOTSUPP */ 397 398 static int 399 rts_peeraddr(struct socket *so, struct sockaddr **nam) 400 { 401 402 return (raw_usrreqs.pru_peeraddr(so, nam)); 403 } 404 405 /* pru_rcvd is EOPNOTSUPP */ 406 /* pru_rcvoob is EOPNOTSUPP */ 407 408 static int 409 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 410 struct mbuf *control, struct thread *td) 411 { 412 413 return (raw_usrreqs.pru_send(so, flags, m, nam, control, td)); 414 } 415 416 /* pru_sense is null */ 417 418 static int 419 rts_shutdown(struct socket *so) 420 { 421 422 return (raw_usrreqs.pru_shutdown(so)); 423 } 424 425 static int 426 rts_sockaddr(struct socket *so, struct sockaddr **nam) 427 { 428 429 return (raw_usrreqs.pru_sockaddr(so, nam)); 430 } 431 432 static struct pr_usrreqs route_usrreqs = { 433 .pru_abort = rts_abort, 434 .pru_attach = rts_attach, 435 .pru_bind = rts_bind, 436 .pru_connect = rts_connect, 437 .pru_detach = rts_detach, 438 .pru_disconnect = rts_disconnect, 439 .pru_peeraddr = rts_peeraddr, 440 .pru_send = rts_send, 441 .pru_shutdown = rts_shutdown, 442 .pru_sockaddr = rts_sockaddr, 443 .pru_close = rts_close, 444 }; 445 446 #ifndef _SOCKADDR_UNION_DEFINED 447 #define _SOCKADDR_UNION_DEFINED 448 /* 449 * The union of all possible address formats we handle. 450 */ 451 union sockaddr_union { 452 struct sockaddr sa; 453 struct sockaddr_in sin; 454 struct sockaddr_in6 sin6; 455 }; 456 #endif /* _SOCKADDR_UNION_DEFINED */ 457 458 static int 459 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp, 460 struct nhop_object *nh, union sockaddr_union *saun, struct ucred *cred) 461 { 462 #if defined(INET) || defined(INET6) 463 struct epoch_tracker et; 464 #endif 465 466 /* First, see if the returned address is part of the jail. */ 467 if (prison_if(cred, nh->nh_ifa->ifa_addr) == 0) { 468 info->rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr; 469 return (0); 470 } 471 472 switch (info->rti_info[RTAX_DST]->sa_family) { 473 #ifdef INET 474 case AF_INET: 475 { 476 struct in_addr ia; 477 struct ifaddr *ifa; 478 int found; 479 480 found = 0; 481 /* 482 * Try to find an address on the given outgoing interface 483 * that belongs to the jail. 484 */ 485 NET_EPOCH_ENTER(et); 486 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 487 struct sockaddr *sa; 488 sa = ifa->ifa_addr; 489 if (sa->sa_family != AF_INET) 490 continue; 491 ia = ((struct sockaddr_in *)sa)->sin_addr; 492 if (prison_check_ip4(cred, &ia) == 0) { 493 found = 1; 494 break; 495 } 496 } 497 NET_EPOCH_EXIT(et); 498 if (!found) { 499 /* 500 * As a last resort return the 'default' jail address. 501 */ 502 ia = ((struct sockaddr_in *)nh->nh_ifa->ifa_addr)-> 503 sin_addr; 504 if (prison_get_ip4(cred, &ia) != 0) 505 return (ESRCH); 506 } 507 bzero(&saun->sin, sizeof(struct sockaddr_in)); 508 saun->sin.sin_len = sizeof(struct sockaddr_in); 509 saun->sin.sin_family = AF_INET; 510 saun->sin.sin_addr.s_addr = ia.s_addr; 511 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin; 512 break; 513 } 514 #endif 515 #ifdef INET6 516 case AF_INET6: 517 { 518 struct in6_addr ia6; 519 struct ifaddr *ifa; 520 int found; 521 522 found = 0; 523 /* 524 * Try to find an address on the given outgoing interface 525 * that belongs to the jail. 526 */ 527 NET_EPOCH_ENTER(et); 528 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 529 struct sockaddr *sa; 530 sa = ifa->ifa_addr; 531 if (sa->sa_family != AF_INET6) 532 continue; 533 bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr, 534 &ia6, sizeof(struct in6_addr)); 535 if (prison_check_ip6(cred, &ia6) == 0) { 536 found = 1; 537 break; 538 } 539 } 540 NET_EPOCH_EXIT(et); 541 if (!found) { 542 /* 543 * As a last resort return the 'default' jail address. 544 */ 545 ia6 = ((struct sockaddr_in6 *)nh->nh_ifa->ifa_addr)-> 546 sin6_addr; 547 if (prison_get_ip6(cred, &ia6) != 0) 548 return (ESRCH); 549 } 550 bzero(&saun->sin6, sizeof(struct sockaddr_in6)); 551 saun->sin6.sin6_len = sizeof(struct sockaddr_in6); 552 saun->sin6.sin6_family = AF_INET6; 553 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr)); 554 if (sa6_recoverscope(&saun->sin6) != 0) 555 return (ESRCH); 556 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6; 557 break; 558 } 559 #endif 560 default: 561 return (ESRCH); 562 } 563 return (0); 564 } 565 566 /* 567 * Fills in @info based on userland-provided @rtm message. 568 * 569 * Returns 0 on success. 570 */ 571 static int 572 fill_addrinfo(struct rt_msghdr *rtm, int len, u_int fibnum, struct rt_addrinfo *info) 573 { 574 int error; 575 sa_family_t saf; 576 577 rtm->rtm_pid = curproc->p_pid; 578 info->rti_addrs = rtm->rtm_addrs; 579 580 info->rti_mflags = rtm->rtm_inits; 581 info->rti_rmx = &rtm->rtm_rmx; 582 583 /* 584 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6 585 * link-local address because rtrequest requires addresses with 586 * embedded scope id. 587 */ 588 if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info)) 589 return (EINVAL); 590 591 if (rtm->rtm_flags & RTF_RNH_LOCKED) 592 return (EINVAL); 593 info->rti_flags = rtm->rtm_flags; 594 if (info->rti_info[RTAX_DST] == NULL || 595 info->rti_info[RTAX_DST]->sa_family >= AF_MAX || 596 (info->rti_info[RTAX_GATEWAY] != NULL && 597 info->rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) 598 return (EINVAL); 599 saf = info->rti_info[RTAX_DST]->sa_family; 600 /* 601 * Verify that the caller has the appropriate privilege; RTM_GET 602 * is the only operation the non-superuser is allowed. 603 */ 604 if (rtm->rtm_type != RTM_GET) { 605 error = priv_check(curthread, PRIV_NET_ROUTE); 606 if (error != 0) 607 return (error); 608 } 609 610 /* 611 * The given gateway address may be an interface address. 612 * For example, issuing a "route change" command on a route 613 * entry that was created from a tunnel, and the gateway 614 * address given is the local end point. In this case the 615 * RTF_GATEWAY flag must be cleared or the destination will 616 * not be reachable even though there is no error message. 617 */ 618 if (info->rti_info[RTAX_GATEWAY] != NULL && 619 info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) { 620 struct rt_addrinfo ginfo; 621 struct sockaddr *gdst; 622 struct sockaddr_storage ss; 623 624 bzero(&ginfo, sizeof(ginfo)); 625 bzero(&ss, sizeof(ss)); 626 ss.ss_len = sizeof(ss); 627 628 ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss; 629 gdst = info->rti_info[RTAX_GATEWAY]; 630 631 /* 632 * A host route through the loopback interface is 633 * installed for each interface adddress. In pre 8.0 634 * releases the interface address of a PPP link type 635 * is not reachable locally. This behavior is fixed as 636 * part of the new L2/L3 redesign and rewrite work. The 637 * signature of this interface address route is the 638 * AF_LINK sa_family type of the gateway, and the 639 * rt_ifp has the IFF_LOOPBACK flag set. 640 */ 641 if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) { 642 if (ss.ss_family == AF_LINK && 643 ginfo.rti_ifp->if_flags & IFF_LOOPBACK) { 644 info->rti_flags &= ~RTF_GATEWAY; 645 info->rti_flags |= RTF_GWFLAG_COMPAT; 646 } 647 rib_free_info(&ginfo); 648 } 649 } 650 651 return (0); 652 } 653 654 /* 655 * Handles RTM_GET message from routing socket, returning matching rt. 656 * 657 * Returns: 658 * 0 on success, with locked and referenced matching rt in @rt_nrt 659 * errno of failure 660 */ 661 static int 662 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum, 663 struct rt_msghdr *rtm, struct rtentry **ret_nrt) 664 { 665 RIB_RLOCK_TRACKER; 666 struct rtentry *rt; 667 struct rib_head *rnh; 668 sa_family_t saf; 669 670 saf = info->rti_info[RTAX_DST]->sa_family; 671 672 rnh = rt_tables_get_rnh(fibnum, saf); 673 if (rnh == NULL) 674 return (EAFNOSUPPORT); 675 676 RIB_RLOCK(rnh); 677 678 if (info->rti_info[RTAX_NETMASK] == NULL) { 679 /* 680 * Provide longest prefix match for 681 * address lookup (no mask). 682 * 'route -n get addr' 683 */ 684 rt = (struct rtentry *) rnh->rnh_matchaddr( 685 info->rti_info[RTAX_DST], &rnh->head); 686 } else 687 rt = (struct rtentry *) rnh->rnh_lookup( 688 info->rti_info[RTAX_DST], 689 info->rti_info[RTAX_NETMASK], &rnh->head); 690 691 if (rt == NULL) { 692 RIB_RUNLOCK(rnh); 693 return (ESRCH); 694 } 695 #ifdef RADIX_MPATH 696 /* 697 * for RTM_GET, gate is optional even with multipath. 698 * if gate == NULL the first match is returned. 699 * (no need to call rt_mpath_matchgate if gate == NULL) 700 */ 701 if (rt_mpath_capable(rnh) && info->rti_info[RTAX_GATEWAY]) { 702 rt = rt_mpath_matchgate(rt, info->rti_info[RTAX_GATEWAY]); 703 if (!rt) { 704 RIB_RUNLOCK(rnh); 705 return (ESRCH); 706 } 707 } 708 #endif 709 /* 710 * If performing proxied L2 entry insertion, and 711 * the actual PPP host entry is found, perform 712 * another search to retrieve the prefix route of 713 * the local end point of the PPP link. 714 * TODO: move this logic to userland. 715 */ 716 if (rtm->rtm_flags & RTF_ANNOUNCE) { 717 struct sockaddr laddr; 718 struct nhop_object *nh; 719 720 nh = rt->rt_nhop; 721 if (nh->nh_ifp != NULL && 722 nh->nh_ifp->if_type == IFT_PROPVIRTUAL) { 723 struct epoch_tracker et; 724 struct ifaddr *ifa; 725 726 NET_EPOCH_ENTER(et); 727 ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1, 728 RT_ALL_FIBS); 729 NET_EPOCH_EXIT(et); 730 if (ifa != NULL) 731 rt_maskedcopy(ifa->ifa_addr, 732 &laddr, 733 ifa->ifa_netmask); 734 } else 735 rt_maskedcopy(nh->nh_ifa->ifa_addr, 736 &laddr, 737 nh->nh_ifa->ifa_netmask); 738 /* 739 * refactor rt and no lock operation necessary 740 */ 741 rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr, 742 &rnh->head); 743 if (rt == NULL) { 744 RIB_RUNLOCK(rnh); 745 return (ESRCH); 746 } 747 } 748 RT_LOCK(rt); 749 RT_ADDREF(rt); 750 RIB_RUNLOCK(rnh); 751 752 *ret_nrt = rt; 753 754 return (0); 755 } 756 757 /* 758 * Update sockaddrs, flags, etc in @prtm based on @rt data. 759 * Assumes @rt is locked. 760 * rtm can be reallocated. 761 * 762 * Returns 0 on success, along with pointer to (potentially reallocated) 763 * rtm. 764 * 765 */ 766 static int 767 update_rtm_from_rte(struct rt_addrinfo *info, struct rt_msghdr **prtm, 768 int alloc_len, struct rtentry *rt) 769 { 770 struct sockaddr_storage netmask_ss; 771 struct walkarg w; 772 union sockaddr_union saun; 773 struct rt_msghdr *rtm, *orig_rtm = NULL; 774 struct nhop_object *nh; 775 struct ifnet *ifp; 776 int error, len; 777 778 RT_LOCK_ASSERT(rt); 779 780 rtm = *prtm; 781 782 nh = rt->rt_nhop; 783 info->rti_info[RTAX_DST] = rt_key(rt); 784 info->rti_info[RTAX_GATEWAY] = &nh->gw_sa; 785 info->rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt), 786 rt_mask(rt), &netmask_ss); 787 info->rti_info[RTAX_GENMASK] = 0; 788 ifp = nh->nh_ifp; 789 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) { 790 if (ifp) { 791 info->rti_info[RTAX_IFP] = 792 ifp->if_addr->ifa_addr; 793 error = rtm_get_jailed(info, ifp, nh, 794 &saun, curthread->td_ucred); 795 if (error != 0) 796 return (error); 797 if (ifp->if_flags & IFF_POINTOPOINT) 798 info->rti_info[RTAX_BRD] = 799 nh->nh_ifa->ifa_dstaddr; 800 rtm->rtm_index = ifp->if_index; 801 } else { 802 info->rti_info[RTAX_IFP] = NULL; 803 info->rti_info[RTAX_IFA] = NULL; 804 } 805 } else if (ifp != NULL) 806 rtm->rtm_index = ifp->if_index; 807 808 /* Check if we need to realloc storage */ 809 rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len); 810 if (len > alloc_len) { 811 struct rt_msghdr *tmp_rtm; 812 813 tmp_rtm = malloc(len, M_TEMP, M_NOWAIT); 814 if (tmp_rtm == NULL) 815 return (ENOBUFS); 816 bcopy(rtm, tmp_rtm, rtm->rtm_msglen); 817 orig_rtm = rtm; 818 rtm = tmp_rtm; 819 alloc_len = len; 820 821 /* 822 * Delay freeing original rtm as info contains 823 * data referencing it. 824 */ 825 } 826 827 w.w_tmem = (caddr_t)rtm; 828 w.w_tmemsize = alloc_len; 829 rtsock_msg_buffer(rtm->rtm_type, info, &w, &len); 830 831 if (rt->rt_flags & RTF_GWFLAG_COMPAT) 832 rtm->rtm_flags = RTF_GATEWAY | 833 (rt->rt_flags & ~RTF_GWFLAG_COMPAT); 834 else 835 rtm->rtm_flags = rt->rt_flags; 836 rt_getmetrics(rt, &rtm->rtm_rmx); 837 rtm->rtm_addrs = info->rti_addrs; 838 839 if (orig_rtm != NULL) 840 free(orig_rtm, M_TEMP); 841 *prtm = rtm; 842 843 return (0); 844 } 845 846 /*ARGSUSED*/ 847 static int 848 route_output(struct mbuf *m, struct socket *so, ...) 849 { 850 struct rt_msghdr *rtm = NULL; 851 struct rtentry *rt = NULL; 852 struct rt_addrinfo info; 853 struct epoch_tracker et; 854 #ifdef INET6 855 struct sockaddr_storage ss; 856 struct sockaddr_in6 *sin6; 857 int i, rti_need_deembed = 0; 858 #endif 859 int alloc_len = 0, len, error = 0, fibnum; 860 sa_family_t saf = AF_UNSPEC; 861 struct walkarg w; 862 863 fibnum = so->so_fibnum; 864 865 #define senderr(e) { error = e; goto flush;} 866 if (m == NULL || ((m->m_len < sizeof(long)) && 867 (m = m_pullup(m, sizeof(long))) == NULL)) 868 return (ENOBUFS); 869 if ((m->m_flags & M_PKTHDR) == 0) 870 panic("route_output"); 871 NET_EPOCH_ENTER(et); 872 len = m->m_pkthdr.len; 873 if (len < sizeof(*rtm) || 874 len != mtod(m, struct rt_msghdr *)->rtm_msglen) 875 senderr(EINVAL); 876 877 /* 878 * Most of current messages are in range 200-240 bytes, 879 * minimize possible re-allocation on reply using larger size 880 * buffer aligned on 1k boundaty. 881 */ 882 alloc_len = roundup2(len, 1024); 883 if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL) 884 senderr(ENOBUFS); 885 886 m_copydata(m, 0, len, (caddr_t)rtm); 887 bzero(&info, sizeof(info)); 888 bzero(&w, sizeof(w)); 889 890 if (rtm->rtm_version != RTM_VERSION) { 891 /* Do not touch message since format is unknown */ 892 free(rtm, M_TEMP); 893 rtm = NULL; 894 senderr(EPROTONOSUPPORT); 895 } 896 897 /* 898 * Starting from here, it is possible 899 * to alter original message and insert 900 * caller PID and error value. 901 */ 902 903 if ((error = fill_addrinfo(rtm, len, fibnum, &info)) != 0) { 904 senderr(error); 905 } 906 907 saf = info.rti_info[RTAX_DST]->sa_family; 908 909 /* support for new ARP code */ 910 if (rtm->rtm_flags & RTF_LLDATA) { 911 error = lla_rt_output(rtm, &info); 912 #ifdef INET6 913 if (error == 0) 914 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0; 915 #endif 916 goto flush; 917 } 918 919 switch (rtm->rtm_type) { 920 struct rtentry *saved_nrt; 921 922 case RTM_ADD: 923 case RTM_CHANGE: 924 if (rtm->rtm_type == RTM_ADD) { 925 if (info.rti_info[RTAX_GATEWAY] == NULL) 926 senderr(EINVAL); 927 } 928 saved_nrt = NULL; 929 error = rtrequest1_fib(rtm->rtm_type, &info, &saved_nrt, 930 fibnum); 931 if (error == 0 && saved_nrt != NULL) { 932 #ifdef INET6 933 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0; 934 #endif 935 RT_LOCK(saved_nrt); 936 rtm->rtm_index = saved_nrt->rt_ifp->if_index; 937 RT_REMREF(saved_nrt); 938 RT_UNLOCK(saved_nrt); 939 } 940 break; 941 942 case RTM_DELETE: 943 saved_nrt = NULL; 944 error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt, fibnum); 945 if (error == 0) { 946 RT_LOCK(saved_nrt); 947 rt = saved_nrt; 948 goto report; 949 } 950 #ifdef INET6 951 /* rt_msg2() will not be used when RTM_DELETE fails. */ 952 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0; 953 #endif 954 break; 955 956 case RTM_GET: 957 error = handle_rtm_get(&info, fibnum, rtm, &rt); 958 if (error != 0) 959 senderr(error); 960 961 report: 962 RT_LOCK_ASSERT(rt); 963 if (!can_export_rte(curthread->td_ucred, rt)) { 964 RT_UNLOCK(rt); 965 senderr(ESRCH); 966 } 967 error = update_rtm_from_rte(&info, &rtm, alloc_len, rt); 968 /* 969 * Note that some sockaddr pointers may have changed to 970 * point to memory outsize @rtm. Some may be pointing 971 * to the on-stack variables. 972 * Given that, any pointer in @info CANNOT BE USED. 973 */ 974 975 /* 976 * scopeid deembedding has been performed while 977 * writing updated rtm in rtsock_msg_buffer(). 978 * With that in mind, skip deembedding procedure below. 979 */ 980 #ifdef INET6 981 rti_need_deembed = 0; 982 #endif 983 RT_UNLOCK(rt); 984 if (error != 0) 985 senderr(error); 986 break; 987 988 default: 989 senderr(EOPNOTSUPP); 990 } 991 992 flush: 993 NET_EPOCH_EXIT(et); 994 if (rt != NULL) 995 RTFREE(rt); 996 997 #ifdef INET6 998 if (rtm != NULL) { 999 if (rti_need_deembed) { 1000 /* sin6_scope_id is recovered before sending rtm. */ 1001 sin6 = (struct sockaddr_in6 *)&ss; 1002 for (i = 0; i < RTAX_MAX; i++) { 1003 if (info.rti_info[i] == NULL) 1004 continue; 1005 if (info.rti_info[i]->sa_family != AF_INET6) 1006 continue; 1007 bcopy(info.rti_info[i], sin6, sizeof(*sin6)); 1008 if (sa6_recoverscope(sin6) == 0) 1009 bcopy(sin6, info.rti_info[i], 1010 sizeof(*sin6)); 1011 } 1012 } 1013 } 1014 #endif 1015 send_rtm_reply(so, rtm, m, saf, fibnum, error); 1016 1017 return (error); 1018 } 1019 1020 /* 1021 * Sends the prepared reply message in @rtm to all rtsock clients. 1022 * Frees @m and @rtm. 1023 * 1024 */ 1025 static void 1026 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m, 1027 sa_family_t saf, u_int fibnum, int rtm_errno) 1028 { 1029 struct rawcb *rp = NULL; 1030 1031 /* 1032 * Check to see if we don't want our own messages. 1033 */ 1034 if ((so->so_options & SO_USELOOPBACK) == 0) { 1035 if (V_route_cb.any_count <= 1) { 1036 if (rtm != NULL) 1037 free(rtm, M_TEMP); 1038 m_freem(m); 1039 return; 1040 } 1041 /* There is another listener, so construct message */ 1042 rp = sotorawcb(so); 1043 } 1044 1045 if (rtm != NULL) { 1046 if (rtm_errno!= 0) 1047 rtm->rtm_errno = rtm_errno; 1048 else 1049 rtm->rtm_flags |= RTF_DONE; 1050 1051 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm); 1052 if (m->m_pkthdr.len < rtm->rtm_msglen) { 1053 m_freem(m); 1054 m = NULL; 1055 } else if (m->m_pkthdr.len > rtm->rtm_msglen) 1056 m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len); 1057 1058 free(rtm, M_TEMP); 1059 } 1060 if (m != NULL) { 1061 M_SETFIB(m, fibnum); 1062 m->m_flags |= RTS_FILTER_FIB; 1063 if (rp) { 1064 /* 1065 * XXX insure we don't get a copy by 1066 * invalidating our protocol 1067 */ 1068 unsigned short family = rp->rcb_proto.sp_family; 1069 rp->rcb_proto.sp_family = 0; 1070 rt_dispatch(m, saf); 1071 rp->rcb_proto.sp_family = family; 1072 } else 1073 rt_dispatch(m, saf); 1074 } 1075 } 1076 1077 1078 static void 1079 rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out) 1080 { 1081 1082 bzero(out, sizeof(*out)); 1083 out->rmx_mtu = rt->rt_nhop->nh_mtu; 1084 out->rmx_weight = rt->rt_weight; 1085 out->rmx_pksent = counter_u64_fetch(rt->rt_pksent); 1086 out->rmx_nhidx = nhop_get_idx(rt->rt_nhop); 1087 /* Kernel -> userland timebase conversion. */ 1088 out->rmx_expire = rt->rt_expire ? 1089 rt->rt_expire - time_uptime + time_second : 0; 1090 } 1091 1092 /* 1093 * Extract the addresses of the passed sockaddrs. 1094 * Do a little sanity checking so as to avoid bad memory references. 1095 * This data is derived straight from userland. 1096 */ 1097 static int 1098 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo) 1099 { 1100 struct sockaddr *sa; 1101 int i; 1102 1103 for (i = 0; i < RTAX_MAX && cp < cplim; i++) { 1104 if ((rtinfo->rti_addrs & (1 << i)) == 0) 1105 continue; 1106 sa = (struct sockaddr *)cp; 1107 /* 1108 * It won't fit. 1109 */ 1110 if (cp + sa->sa_len > cplim) 1111 return (EINVAL); 1112 /* 1113 * there are no more.. quit now 1114 * If there are more bits, they are in error. 1115 * I've seen this. route(1) can evidently generate these. 1116 * This causes kernel to core dump. 1117 * for compatibility, If we see this, point to a safe address. 1118 */ 1119 if (sa->sa_len == 0) { 1120 rtinfo->rti_info[i] = &sa_zero; 1121 return (0); /* should be EINVAL but for compat */ 1122 } 1123 /* accept it */ 1124 #ifdef INET6 1125 if (sa->sa_family == AF_INET6) 1126 sa6_embedscope((struct sockaddr_in6 *)sa, 1127 V_ip6_use_defzone); 1128 #endif 1129 rtinfo->rti_info[i] = sa; 1130 cp += SA_SIZE(sa); 1131 } 1132 return (0); 1133 } 1134 1135 /* 1136 * Fill in @dmask with valid netmask leaving original @smask 1137 * intact. Mostly used with radix netmasks. 1138 */ 1139 static struct sockaddr * 1140 rtsock_fix_netmask(struct sockaddr *dst, struct sockaddr *smask, 1141 struct sockaddr_storage *dmask) 1142 { 1143 if (dst == NULL || smask == NULL) 1144 return (NULL); 1145 1146 memset(dmask, 0, dst->sa_len); 1147 memcpy(dmask, smask, smask->sa_len); 1148 dmask->ss_len = dst->sa_len; 1149 dmask->ss_family = dst->sa_family; 1150 1151 return ((struct sockaddr *)dmask); 1152 } 1153 1154 /* 1155 * Writes information related to @rtinfo object to newly-allocated mbuf. 1156 * Assumes MCLBYTES is enough to construct any message. 1157 * Used for OS notifications of vaious events (if/ifa announces,etc) 1158 * 1159 * Returns allocated mbuf or NULL on failure. 1160 */ 1161 static struct mbuf * 1162 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo) 1163 { 1164 struct rt_msghdr *rtm; 1165 struct mbuf *m; 1166 int i; 1167 struct sockaddr *sa; 1168 #ifdef INET6 1169 struct sockaddr_storage ss; 1170 struct sockaddr_in6 *sin6; 1171 #endif 1172 int len, dlen; 1173 1174 switch (type) { 1175 1176 case RTM_DELADDR: 1177 case RTM_NEWADDR: 1178 len = sizeof(struct ifa_msghdr); 1179 break; 1180 1181 case RTM_DELMADDR: 1182 case RTM_NEWMADDR: 1183 len = sizeof(struct ifma_msghdr); 1184 break; 1185 1186 case RTM_IFINFO: 1187 len = sizeof(struct if_msghdr); 1188 break; 1189 1190 case RTM_IFANNOUNCE: 1191 case RTM_IEEE80211: 1192 len = sizeof(struct if_announcemsghdr); 1193 break; 1194 1195 default: 1196 len = sizeof(struct rt_msghdr); 1197 } 1198 1199 /* XXXGL: can we use MJUMPAGESIZE cluster here? */ 1200 KASSERT(len <= MCLBYTES, ("%s: message too big", __func__)); 1201 if (len > MHLEN) 1202 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 1203 else 1204 m = m_gethdr(M_NOWAIT, MT_DATA); 1205 if (m == NULL) 1206 return (m); 1207 1208 m->m_pkthdr.len = m->m_len = len; 1209 rtm = mtod(m, struct rt_msghdr *); 1210 bzero((caddr_t)rtm, len); 1211 for (i = 0; i < RTAX_MAX; i++) { 1212 if ((sa = rtinfo->rti_info[i]) == NULL) 1213 continue; 1214 rtinfo->rti_addrs |= (1 << i); 1215 dlen = SA_SIZE(sa); 1216 #ifdef INET6 1217 if (V_deembed_scopeid && sa->sa_family == AF_INET6) { 1218 sin6 = (struct sockaddr_in6 *)&ss; 1219 bcopy(sa, sin6, sizeof(*sin6)); 1220 if (sa6_recoverscope(sin6) == 0) 1221 sa = (struct sockaddr *)sin6; 1222 } 1223 #endif 1224 m_copyback(m, len, dlen, (caddr_t)sa); 1225 len += dlen; 1226 } 1227 if (m->m_pkthdr.len != len) { 1228 m_freem(m); 1229 return (NULL); 1230 } 1231 rtm->rtm_msglen = len; 1232 rtm->rtm_version = RTM_VERSION; 1233 rtm->rtm_type = type; 1234 return (m); 1235 } 1236 1237 /* 1238 * Writes information related to @rtinfo object to preallocated buffer. 1239 * Stores needed size in @plen. If @w is NULL, calculates size without 1240 * writing. 1241 * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation. 1242 * 1243 * Returns 0 on success. 1244 * 1245 */ 1246 static int 1247 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen) 1248 { 1249 int i; 1250 int len, buflen = 0, dlen; 1251 caddr_t cp = NULL; 1252 struct rt_msghdr *rtm = NULL; 1253 #ifdef INET6 1254 struct sockaddr_storage ss; 1255 struct sockaddr_in6 *sin6; 1256 #endif 1257 #ifdef COMPAT_FREEBSD32 1258 bool compat32 = false; 1259 #endif 1260 1261 switch (type) { 1262 1263 case RTM_DELADDR: 1264 case RTM_NEWADDR: 1265 if (w != NULL && w->w_op == NET_RT_IFLISTL) { 1266 #ifdef COMPAT_FREEBSD32 1267 if (w->w_req->flags & SCTL_MASK32) { 1268 len = sizeof(struct ifa_msghdrl32); 1269 compat32 = true; 1270 } else 1271 #endif 1272 len = sizeof(struct ifa_msghdrl); 1273 } else 1274 len = sizeof(struct ifa_msghdr); 1275 break; 1276 1277 case RTM_IFINFO: 1278 #ifdef COMPAT_FREEBSD32 1279 if (w != NULL && w->w_req->flags & SCTL_MASK32) { 1280 if (w->w_op == NET_RT_IFLISTL) 1281 len = sizeof(struct if_msghdrl32); 1282 else 1283 len = sizeof(struct if_msghdr32); 1284 compat32 = true; 1285 break; 1286 } 1287 #endif 1288 if (w != NULL && w->w_op == NET_RT_IFLISTL) 1289 len = sizeof(struct if_msghdrl); 1290 else 1291 len = sizeof(struct if_msghdr); 1292 break; 1293 1294 case RTM_NEWMADDR: 1295 len = sizeof(struct ifma_msghdr); 1296 break; 1297 1298 default: 1299 len = sizeof(struct rt_msghdr); 1300 } 1301 1302 if (w != NULL) { 1303 rtm = (struct rt_msghdr *)w->w_tmem; 1304 buflen = w->w_tmemsize - len; 1305 cp = (caddr_t)w->w_tmem + len; 1306 } 1307 1308 rtinfo->rti_addrs = 0; 1309 for (i = 0; i < RTAX_MAX; i++) { 1310 struct sockaddr *sa; 1311 1312 if ((sa = rtinfo->rti_info[i]) == NULL) 1313 continue; 1314 rtinfo->rti_addrs |= (1 << i); 1315 #ifdef COMPAT_FREEBSD32 1316 if (compat32) 1317 dlen = SA_SIZE32(sa); 1318 else 1319 #endif 1320 dlen = SA_SIZE(sa); 1321 if (cp != NULL && buflen >= dlen) { 1322 #ifdef INET6 1323 if (V_deembed_scopeid && sa->sa_family == AF_INET6) { 1324 sin6 = (struct sockaddr_in6 *)&ss; 1325 bcopy(sa, sin6, sizeof(*sin6)); 1326 if (sa6_recoverscope(sin6) == 0) 1327 sa = (struct sockaddr *)sin6; 1328 } 1329 #endif 1330 bcopy((caddr_t)sa, cp, (unsigned)dlen); 1331 cp += dlen; 1332 buflen -= dlen; 1333 } else if (cp != NULL) { 1334 /* 1335 * Buffer too small. Count needed size 1336 * and return with error. 1337 */ 1338 cp = NULL; 1339 } 1340 1341 len += dlen; 1342 } 1343 1344 if (cp != NULL) { 1345 dlen = ALIGN(len) - len; 1346 if (buflen < dlen) 1347 cp = NULL; 1348 else { 1349 bzero(cp, dlen); 1350 cp += dlen; 1351 buflen -= dlen; 1352 } 1353 } 1354 len = ALIGN(len); 1355 1356 if (cp != NULL) { 1357 /* fill header iff buffer is large enough */ 1358 rtm->rtm_version = RTM_VERSION; 1359 rtm->rtm_type = type; 1360 rtm->rtm_msglen = len; 1361 } 1362 1363 *plen = len; 1364 1365 if (w != NULL && cp == NULL) 1366 return (ENOBUFS); 1367 1368 return (0); 1369 } 1370 1371 /* 1372 * This routine is called to generate a message from the routing 1373 * socket indicating that a redirect has occurred, a routing lookup 1374 * has failed, or that a protocol has detected timeouts to a particular 1375 * destination. 1376 */ 1377 void 1378 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error, 1379 int fibnum) 1380 { 1381 struct rt_msghdr *rtm; 1382 struct mbuf *m; 1383 struct sockaddr *sa = rtinfo->rti_info[RTAX_DST]; 1384 1385 if (V_route_cb.any_count == 0) 1386 return; 1387 m = rtsock_msg_mbuf(type, rtinfo); 1388 if (m == NULL) 1389 return; 1390 1391 if (fibnum != RT_ALL_FIBS) { 1392 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out " 1393 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs)); 1394 M_SETFIB(m, fibnum); 1395 m->m_flags |= RTS_FILTER_FIB; 1396 } 1397 1398 rtm = mtod(m, struct rt_msghdr *); 1399 rtm->rtm_flags = RTF_DONE | flags; 1400 rtm->rtm_errno = error; 1401 rtm->rtm_addrs = rtinfo->rti_addrs; 1402 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1403 } 1404 1405 void 1406 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error) 1407 { 1408 1409 rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS); 1410 } 1411 1412 /* 1413 * This routine is called to generate a message from the routing 1414 * socket indicating that the status of a network interface has changed. 1415 */ 1416 void 1417 rt_ifmsg(struct ifnet *ifp) 1418 { 1419 struct if_msghdr *ifm; 1420 struct mbuf *m; 1421 struct rt_addrinfo info; 1422 1423 if (V_route_cb.any_count == 0) 1424 return; 1425 bzero((caddr_t)&info, sizeof(info)); 1426 m = rtsock_msg_mbuf(RTM_IFINFO, &info); 1427 if (m == NULL) 1428 return; 1429 ifm = mtod(m, struct if_msghdr *); 1430 ifm->ifm_index = ifp->if_index; 1431 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1432 if_data_copy(ifp, &ifm->ifm_data); 1433 ifm->ifm_addrs = 0; 1434 rt_dispatch(m, AF_UNSPEC); 1435 } 1436 1437 /* 1438 * Announce interface address arrival/withdraw. 1439 * Please do not call directly, use rt_addrmsg(). 1440 * Assume input data to be valid. 1441 * Returns 0 on success. 1442 */ 1443 int 1444 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum) 1445 { 1446 struct rt_addrinfo info; 1447 struct sockaddr *sa; 1448 int ncmd; 1449 struct mbuf *m; 1450 struct ifa_msghdr *ifam; 1451 struct ifnet *ifp = ifa->ifa_ifp; 1452 struct sockaddr_storage ss; 1453 1454 if (V_route_cb.any_count == 0) 1455 return (0); 1456 1457 ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR; 1458 1459 bzero((caddr_t)&info, sizeof(info)); 1460 info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr; 1461 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr; 1462 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask( 1463 info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss); 1464 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr; 1465 if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL) 1466 return (ENOBUFS); 1467 ifam = mtod(m, struct ifa_msghdr *); 1468 ifam->ifam_index = ifp->if_index; 1469 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 1470 ifam->ifam_flags = ifa->ifa_flags; 1471 ifam->ifam_addrs = info.rti_addrs; 1472 1473 if (fibnum != RT_ALL_FIBS) { 1474 M_SETFIB(m, fibnum); 1475 m->m_flags |= RTS_FILTER_FIB; 1476 } 1477 1478 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1479 1480 return (0); 1481 } 1482 1483 /* 1484 * Announce route addition/removal to rtsock based on @rt data. 1485 * Callers are advives to use rt_routemsg() instead of using this 1486 * function directly. 1487 * Assume @rt data is consistent. 1488 * 1489 * Returns 0 on success. 1490 */ 1491 int 1492 rtsock_routemsg(int cmd, struct rtentry *rt, struct ifnet *ifp, int rti_addrs, 1493 int fibnum) 1494 { 1495 struct sockaddr_storage ss; 1496 struct rt_addrinfo info; 1497 1498 if (V_route_cb.any_count == 0) 1499 return (0); 1500 1501 bzero((caddr_t)&info, sizeof(info)); 1502 info.rti_info[RTAX_DST] = rt_key(rt); 1503 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt), rt_mask(rt), &ss); 1504 info.rti_info[RTAX_GATEWAY] = &rt->rt_nhop->gw_sa; 1505 info.rti_flags = rt->rt_flags; 1506 info.rti_ifp = ifp; 1507 1508 return (rtsock_routemsg_info(cmd, &info, fibnum)); 1509 } 1510 1511 int 1512 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum) 1513 { 1514 struct rt_msghdr *rtm; 1515 struct sockaddr *sa; 1516 struct mbuf *m; 1517 1518 if (V_route_cb.any_count == 0) 1519 return (0); 1520 1521 if (info->rti_flags & RTF_HOST) 1522 info->rti_info[RTAX_NETMASK] = NULL; 1523 1524 m = rtsock_msg_mbuf(cmd, info); 1525 if (m == NULL) 1526 return (ENOBUFS); 1527 1528 if (fibnum != RT_ALL_FIBS) { 1529 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out " 1530 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs)); 1531 M_SETFIB(m, fibnum); 1532 m->m_flags |= RTS_FILTER_FIB; 1533 } 1534 1535 rtm = mtod(m, struct rt_msghdr *); 1536 rtm->rtm_addrs = info->rti_addrs; 1537 if (info->rti_ifp != NULL) 1538 rtm->rtm_index = info->rti_ifp->if_index; 1539 /* Add RTF_DONE to indicate command 'completion' required by API */ 1540 info->rti_flags |= RTF_DONE; 1541 /* Reported routes has to be up */ 1542 if (cmd == RTM_ADD || cmd == RTM_CHANGE) 1543 info->rti_flags |= RTF_UP; 1544 rtm->rtm_flags = info->rti_flags; 1545 1546 sa = info->rti_info[RTAX_DST]; 1547 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1548 1549 return (0); 1550 } 1551 1552 /* 1553 * This is the analogue to the rt_newaddrmsg which performs the same 1554 * function but for multicast group memberhips. This is easier since 1555 * there is no route state to worry about. 1556 */ 1557 void 1558 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma) 1559 { 1560 struct rt_addrinfo info; 1561 struct mbuf *m = NULL; 1562 struct ifnet *ifp = ifma->ifma_ifp; 1563 struct ifma_msghdr *ifmam; 1564 1565 if (V_route_cb.any_count == 0) 1566 return; 1567 1568 bzero((caddr_t)&info, sizeof(info)); 1569 info.rti_info[RTAX_IFA] = ifma->ifma_addr; 1570 if (ifp && ifp->if_addr) 1571 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr; 1572 else 1573 info.rti_info[RTAX_IFP] = NULL; 1574 /* 1575 * If a link-layer address is present, present it as a ``gateway'' 1576 * (similarly to how ARP entries, e.g., are presented). 1577 */ 1578 info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr; 1579 m = rtsock_msg_mbuf(cmd, &info); 1580 if (m == NULL) 1581 return; 1582 ifmam = mtod(m, struct ifma_msghdr *); 1583 KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n", 1584 __func__)); 1585 ifmam->ifmam_index = ifp->if_index; 1586 ifmam->ifmam_addrs = info.rti_addrs; 1587 rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC); 1588 } 1589 1590 static struct mbuf * 1591 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what, 1592 struct rt_addrinfo *info) 1593 { 1594 struct if_announcemsghdr *ifan; 1595 struct mbuf *m; 1596 1597 if (V_route_cb.any_count == 0) 1598 return NULL; 1599 bzero((caddr_t)info, sizeof(*info)); 1600 m = rtsock_msg_mbuf(type, info); 1601 if (m != NULL) { 1602 ifan = mtod(m, struct if_announcemsghdr *); 1603 ifan->ifan_index = ifp->if_index; 1604 strlcpy(ifan->ifan_name, ifp->if_xname, 1605 sizeof(ifan->ifan_name)); 1606 ifan->ifan_what = what; 1607 } 1608 return m; 1609 } 1610 1611 /* 1612 * This is called to generate routing socket messages indicating 1613 * IEEE80211 wireless events. 1614 * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way. 1615 */ 1616 void 1617 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len) 1618 { 1619 struct mbuf *m; 1620 struct rt_addrinfo info; 1621 1622 m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info); 1623 if (m != NULL) { 1624 /* 1625 * Append the ieee80211 data. Try to stick it in the 1626 * mbuf containing the ifannounce msg; otherwise allocate 1627 * a new mbuf and append. 1628 * 1629 * NB: we assume m is a single mbuf. 1630 */ 1631 if (data_len > M_TRAILINGSPACE(m)) { 1632 struct mbuf *n = m_get(M_NOWAIT, MT_DATA); 1633 if (n == NULL) { 1634 m_freem(m); 1635 return; 1636 } 1637 bcopy(data, mtod(n, void *), data_len); 1638 n->m_len = data_len; 1639 m->m_next = n; 1640 } else if (data_len > 0) { 1641 bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len); 1642 m->m_len += data_len; 1643 } 1644 if (m->m_flags & M_PKTHDR) 1645 m->m_pkthdr.len += data_len; 1646 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len; 1647 rt_dispatch(m, AF_UNSPEC); 1648 } 1649 } 1650 1651 /* 1652 * This is called to generate routing socket messages indicating 1653 * network interface arrival and departure. 1654 */ 1655 void 1656 rt_ifannouncemsg(struct ifnet *ifp, int what) 1657 { 1658 struct mbuf *m; 1659 struct rt_addrinfo info; 1660 1661 m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info); 1662 if (m != NULL) 1663 rt_dispatch(m, AF_UNSPEC); 1664 } 1665 1666 static void 1667 rt_dispatch(struct mbuf *m, sa_family_t saf) 1668 { 1669 struct m_tag *tag; 1670 1671 /* 1672 * Preserve the family from the sockaddr, if any, in an m_tag for 1673 * use when injecting the mbuf into the routing socket buffer from 1674 * the netisr. 1675 */ 1676 if (saf != AF_UNSPEC) { 1677 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short), 1678 M_NOWAIT); 1679 if (tag == NULL) { 1680 m_freem(m); 1681 return; 1682 } 1683 *(unsigned short *)(tag + 1) = saf; 1684 m_tag_prepend(m, tag); 1685 } 1686 #ifdef VIMAGE 1687 if (V_loif) 1688 m->m_pkthdr.rcvif = V_loif; 1689 else { 1690 m_freem(m); 1691 return; 1692 } 1693 #endif 1694 netisr_queue(NETISR_ROUTE, m); /* mbuf is free'd on failure. */ 1695 } 1696 1697 /* 1698 * Checks if rte can be exported v.r.t jails/vnets. 1699 * 1700 * Returns 1 if it can, 0 otherwise. 1701 */ 1702 static int 1703 can_export_rte(struct ucred *td_ucred, const struct rtentry *rt) 1704 { 1705 1706 if ((rt->rt_flags & RTF_HOST) == 0 1707 ? jailed_without_vnet(td_ucred) 1708 : prison_if(td_ucred, rt_key_const(rt)) != 0) 1709 return (0); 1710 return (1); 1711 } 1712 1713 /* 1714 * This is used in dumping the kernel table via sysctl(). 1715 */ 1716 static int 1717 sysctl_dumpentry(struct radix_node *rn, void *vw) 1718 { 1719 struct walkarg *w = vw; 1720 struct rtentry *rt = (struct rtentry *)rn; 1721 int error = 0, size; 1722 struct rt_addrinfo info; 1723 struct sockaddr_storage ss; 1724 1725 NET_EPOCH_ASSERT(); 1726 1727 if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg)) 1728 return 0; 1729 if (!can_export_rte(w->w_req->td->td_ucred, rt)) 1730 return (0); 1731 bzero((caddr_t)&info, sizeof(info)); 1732 info.rti_info[RTAX_DST] = rt_key(rt); 1733 info.rti_info[RTAX_GATEWAY] = &rt->rt_nhop->gw_sa; 1734 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt), 1735 rt_mask(rt), &ss); 1736 info.rti_info[RTAX_GENMASK] = 0; 1737 if (rt->rt_ifp && !(rt->rt_ifp->if_flags & IFF_DYING)) { 1738 info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr; 1739 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr; 1740 if (rt->rt_ifp->if_flags & IFF_POINTOPOINT) 1741 info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr; 1742 } 1743 if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0) 1744 return (error); 1745 if (w->w_req && w->w_tmem) { 1746 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem; 1747 1748 bzero(&rtm->rtm_index, 1749 sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index)); 1750 if (rt->rt_flags & RTF_GWFLAG_COMPAT) 1751 rtm->rtm_flags = RTF_GATEWAY | 1752 (rt->rt_flags & ~RTF_GWFLAG_COMPAT); 1753 else 1754 rtm->rtm_flags = rt->rt_flags; 1755 rt_getmetrics(rt, &rtm->rtm_rmx); 1756 rtm->rtm_index = rt->rt_ifp->if_index; 1757 rtm->rtm_addrs = info.rti_addrs; 1758 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size); 1759 return (error); 1760 } 1761 return (error); 1762 } 1763 1764 static int 1765 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd, 1766 struct rt_addrinfo *info, struct walkarg *w, int len) 1767 { 1768 struct if_msghdrl *ifm; 1769 struct if_data *ifd; 1770 1771 ifm = (struct if_msghdrl *)w->w_tmem; 1772 1773 #ifdef COMPAT_FREEBSD32 1774 if (w->w_req->flags & SCTL_MASK32) { 1775 struct if_msghdrl32 *ifm32; 1776 1777 ifm32 = (struct if_msghdrl32 *)ifm; 1778 ifm32->ifm_addrs = info->rti_addrs; 1779 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1780 ifm32->ifm_index = ifp->if_index; 1781 ifm32->_ifm_spare1 = 0; 1782 ifm32->ifm_len = sizeof(*ifm32); 1783 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data); 1784 ifm32->_ifm_spare2 = 0; 1785 ifd = &ifm32->ifm_data; 1786 } else 1787 #endif 1788 { 1789 ifm->ifm_addrs = info->rti_addrs; 1790 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1791 ifm->ifm_index = ifp->if_index; 1792 ifm->_ifm_spare1 = 0; 1793 ifm->ifm_len = sizeof(*ifm); 1794 ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data); 1795 ifm->_ifm_spare2 = 0; 1796 ifd = &ifm->ifm_data; 1797 } 1798 1799 memcpy(ifd, src_ifd, sizeof(*ifd)); 1800 1801 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len)); 1802 } 1803 1804 static int 1805 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd, 1806 struct rt_addrinfo *info, struct walkarg *w, int len) 1807 { 1808 struct if_msghdr *ifm; 1809 struct if_data *ifd; 1810 1811 ifm = (struct if_msghdr *)w->w_tmem; 1812 1813 #ifdef COMPAT_FREEBSD32 1814 if (w->w_req->flags & SCTL_MASK32) { 1815 struct if_msghdr32 *ifm32; 1816 1817 ifm32 = (struct if_msghdr32 *)ifm; 1818 ifm32->ifm_addrs = info->rti_addrs; 1819 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1820 ifm32->ifm_index = ifp->if_index; 1821 ifm32->_ifm_spare1 = 0; 1822 ifd = &ifm32->ifm_data; 1823 } else 1824 #endif 1825 { 1826 ifm->ifm_addrs = info->rti_addrs; 1827 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1828 ifm->ifm_index = ifp->if_index; 1829 ifm->_ifm_spare1 = 0; 1830 ifd = &ifm->ifm_data; 1831 } 1832 1833 memcpy(ifd, src_ifd, sizeof(*ifd)); 1834 1835 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len)); 1836 } 1837 1838 static int 1839 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info, 1840 struct walkarg *w, int len) 1841 { 1842 struct ifa_msghdrl *ifam; 1843 struct if_data *ifd; 1844 1845 ifam = (struct ifa_msghdrl *)w->w_tmem; 1846 1847 #ifdef COMPAT_FREEBSD32 1848 if (w->w_req->flags & SCTL_MASK32) { 1849 struct ifa_msghdrl32 *ifam32; 1850 1851 ifam32 = (struct ifa_msghdrl32 *)ifam; 1852 ifam32->ifam_addrs = info->rti_addrs; 1853 ifam32->ifam_flags = ifa->ifa_flags; 1854 ifam32->ifam_index = ifa->ifa_ifp->if_index; 1855 ifam32->_ifam_spare1 = 0; 1856 ifam32->ifam_len = sizeof(*ifam32); 1857 ifam32->ifam_data_off = 1858 offsetof(struct ifa_msghdrl32, ifam_data); 1859 ifam32->ifam_metric = ifa->ifa_ifp->if_metric; 1860 ifd = &ifam32->ifam_data; 1861 } else 1862 #endif 1863 { 1864 ifam->ifam_addrs = info->rti_addrs; 1865 ifam->ifam_flags = ifa->ifa_flags; 1866 ifam->ifam_index = ifa->ifa_ifp->if_index; 1867 ifam->_ifam_spare1 = 0; 1868 ifam->ifam_len = sizeof(*ifam); 1869 ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data); 1870 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 1871 ifd = &ifam->ifam_data; 1872 } 1873 1874 bzero(ifd, sizeof(*ifd)); 1875 ifd->ifi_datalen = sizeof(struct if_data); 1876 ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets); 1877 ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets); 1878 ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes); 1879 ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes); 1880 1881 /* Fixup if_data carp(4) vhid. */ 1882 if (carp_get_vhid_p != NULL) 1883 ifd->ifi_vhid = (*carp_get_vhid_p)(ifa); 1884 1885 return (SYSCTL_OUT(w->w_req, w->w_tmem, len)); 1886 } 1887 1888 static int 1889 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info, 1890 struct walkarg *w, int len) 1891 { 1892 struct ifa_msghdr *ifam; 1893 1894 ifam = (struct ifa_msghdr *)w->w_tmem; 1895 ifam->ifam_addrs = info->rti_addrs; 1896 ifam->ifam_flags = ifa->ifa_flags; 1897 ifam->ifam_index = ifa->ifa_ifp->if_index; 1898 ifam->_ifam_spare1 = 0; 1899 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 1900 1901 return (SYSCTL_OUT(w->w_req, w->w_tmem, len)); 1902 } 1903 1904 static int 1905 sysctl_iflist(int af, struct walkarg *w) 1906 { 1907 struct ifnet *ifp; 1908 struct ifaddr *ifa; 1909 struct if_data ifd; 1910 struct rt_addrinfo info; 1911 int len, error = 0; 1912 struct sockaddr_storage ss; 1913 1914 bzero((caddr_t)&info, sizeof(info)); 1915 bzero(&ifd, sizeof(ifd)); 1916 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1917 if (w->w_arg && w->w_arg != ifp->if_index) 1918 continue; 1919 if_data_copy(ifp, &ifd); 1920 ifa = ifp->if_addr; 1921 info.rti_info[RTAX_IFP] = ifa->ifa_addr; 1922 error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len); 1923 if (error != 0) 1924 goto done; 1925 info.rti_info[RTAX_IFP] = NULL; 1926 if (w->w_req && w->w_tmem) { 1927 if (w->w_op == NET_RT_IFLISTL) 1928 error = sysctl_iflist_ifml(ifp, &ifd, &info, w, 1929 len); 1930 else 1931 error = sysctl_iflist_ifm(ifp, &ifd, &info, w, 1932 len); 1933 if (error) 1934 goto done; 1935 } 1936 while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) { 1937 if (af && af != ifa->ifa_addr->sa_family) 1938 continue; 1939 if (prison_if(w->w_req->td->td_ucred, 1940 ifa->ifa_addr) != 0) 1941 continue; 1942 info.rti_info[RTAX_IFA] = ifa->ifa_addr; 1943 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask( 1944 ifa->ifa_addr, ifa->ifa_netmask, &ss); 1945 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr; 1946 error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len); 1947 if (error != 0) 1948 goto done; 1949 if (w->w_req && w->w_tmem) { 1950 if (w->w_op == NET_RT_IFLISTL) 1951 error = sysctl_iflist_ifaml(ifa, &info, 1952 w, len); 1953 else 1954 error = sysctl_iflist_ifam(ifa, &info, 1955 w, len); 1956 if (error) 1957 goto done; 1958 } 1959 } 1960 info.rti_info[RTAX_IFA] = NULL; 1961 info.rti_info[RTAX_NETMASK] = NULL; 1962 info.rti_info[RTAX_BRD] = NULL; 1963 } 1964 done: 1965 return (error); 1966 } 1967 1968 static int 1969 sysctl_ifmalist(int af, struct walkarg *w) 1970 { 1971 struct rt_addrinfo info; 1972 struct ifaddr *ifa; 1973 struct ifmultiaddr *ifma; 1974 struct ifnet *ifp; 1975 int error, len; 1976 1977 NET_EPOCH_ASSERT(); 1978 1979 error = 0; 1980 bzero((caddr_t)&info, sizeof(info)); 1981 1982 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1983 if (w->w_arg && w->w_arg != ifp->if_index) 1984 continue; 1985 ifa = ifp->if_addr; 1986 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL; 1987 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 1988 if (af && af != ifma->ifma_addr->sa_family) 1989 continue; 1990 if (prison_if(w->w_req->td->td_ucred, 1991 ifma->ifma_addr) != 0) 1992 continue; 1993 info.rti_info[RTAX_IFA] = ifma->ifma_addr; 1994 info.rti_info[RTAX_GATEWAY] = 1995 (ifma->ifma_addr->sa_family != AF_LINK) ? 1996 ifma->ifma_lladdr : NULL; 1997 error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len); 1998 if (error != 0) 1999 break; 2000 if (w->w_req && w->w_tmem) { 2001 struct ifma_msghdr *ifmam; 2002 2003 ifmam = (struct ifma_msghdr *)w->w_tmem; 2004 ifmam->ifmam_index = ifma->ifma_ifp->if_index; 2005 ifmam->ifmam_flags = 0; 2006 ifmam->ifmam_addrs = info.rti_addrs; 2007 ifmam->_ifmam_spare1 = 0; 2008 error = SYSCTL_OUT(w->w_req, w->w_tmem, len); 2009 if (error != 0) 2010 break; 2011 } 2012 } 2013 if (error != 0) 2014 break; 2015 } 2016 return (error); 2017 } 2018 2019 static int 2020 sysctl_rtsock(SYSCTL_HANDLER_ARGS) 2021 { 2022 RIB_RLOCK_TRACKER; 2023 struct epoch_tracker et; 2024 int *name = (int *)arg1; 2025 u_int namelen = arg2; 2026 struct rib_head *rnh = NULL; /* silence compiler. */ 2027 int i, lim, error = EINVAL; 2028 int fib = 0; 2029 u_char af; 2030 struct walkarg w; 2031 2032 name ++; 2033 namelen--; 2034 if (req->newptr) 2035 return (EPERM); 2036 if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP) { 2037 if (namelen == 3) 2038 fib = req->td->td_proc->p_fibnum; 2039 else if (namelen == 4) 2040 fib = (name[3] == RT_ALL_FIBS) ? 2041 req->td->td_proc->p_fibnum : name[3]; 2042 else 2043 return ((namelen < 3) ? EISDIR : ENOTDIR); 2044 if (fib < 0 || fib >= rt_numfibs) 2045 return (EINVAL); 2046 } else if (namelen != 3) 2047 return ((namelen < 3) ? EISDIR : ENOTDIR); 2048 af = name[0]; 2049 if (af > AF_MAX) 2050 return (EINVAL); 2051 bzero(&w, sizeof(w)); 2052 w.w_op = name[1]; 2053 w.w_arg = name[2]; 2054 w.w_req = req; 2055 2056 error = sysctl_wire_old_buffer(req, 0); 2057 if (error) 2058 return (error); 2059 2060 /* 2061 * Allocate reply buffer in advance. 2062 * All rtsock messages has maximum length of u_short. 2063 */ 2064 w.w_tmemsize = 65536; 2065 w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK); 2066 2067 NET_EPOCH_ENTER(et); 2068 switch (w.w_op) { 2069 case NET_RT_DUMP: 2070 case NET_RT_FLAGS: 2071 if (af == 0) { /* dump all tables */ 2072 i = 1; 2073 lim = AF_MAX; 2074 } else /* dump only one table */ 2075 i = lim = af; 2076 2077 /* 2078 * take care of llinfo entries, the caller must 2079 * specify an AF 2080 */ 2081 if (w.w_op == NET_RT_FLAGS && 2082 (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) { 2083 if (af != 0) 2084 error = lltable_sysctl_dumparp(af, w.w_req); 2085 else 2086 error = EINVAL; 2087 break; 2088 } 2089 /* 2090 * take care of routing entries 2091 */ 2092 for (error = 0; error == 0 && i <= lim; i++) { 2093 rnh = rt_tables_get_rnh(fib, i); 2094 if (rnh != NULL) { 2095 RIB_RLOCK(rnh); 2096 error = rnh->rnh_walktree(&rnh->head, 2097 sysctl_dumpentry, &w); 2098 RIB_RUNLOCK(rnh); 2099 } else if (af != 0) 2100 error = EAFNOSUPPORT; 2101 } 2102 break; 2103 case NET_RT_NHOP: 2104 /* Allow dumping one specific af/fib at a time */ 2105 if (namelen < 4) { 2106 error = EINVAL; 2107 break; 2108 } 2109 fib = name[3]; 2110 if (fib < 0 || fib > rt_numfibs) { 2111 error = EINVAL; 2112 break; 2113 } 2114 rnh = rt_tables_get_rnh(fib, af); 2115 if (rnh == NULL) { 2116 error = EAFNOSUPPORT; 2117 break; 2118 } 2119 if (w.w_op == NET_RT_NHOP) 2120 error = nhops_dump_sysctl(rnh, w.w_req); 2121 break; 2122 case NET_RT_IFLIST: 2123 case NET_RT_IFLISTL: 2124 error = sysctl_iflist(af, &w); 2125 break; 2126 2127 case NET_RT_IFMALIST: 2128 error = sysctl_ifmalist(af, &w); 2129 break; 2130 } 2131 NET_EPOCH_EXIT(et); 2132 2133 free(w.w_tmem, M_TEMP); 2134 return (error); 2135 } 2136 2137 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE, 2138 sysctl_rtsock, "Return route tables and interface/address lists"); 2139 2140 /* 2141 * Definitions of protocols supported in the ROUTE domain. 2142 */ 2143 2144 static struct domain routedomain; /* or at least forward */ 2145 2146 static struct protosw routesw[] = { 2147 { 2148 .pr_type = SOCK_RAW, 2149 .pr_domain = &routedomain, 2150 .pr_flags = PR_ATOMIC|PR_ADDR, 2151 .pr_output = route_output, 2152 .pr_ctlinput = raw_ctlinput, 2153 .pr_init = raw_init, 2154 .pr_usrreqs = &route_usrreqs 2155 } 2156 }; 2157 2158 static struct domain routedomain = { 2159 .dom_family = PF_ROUTE, 2160 .dom_name = "route", 2161 .dom_protosw = routesw, 2162 .dom_protoswNPROTOSW = &routesw[nitems(routesw)] 2163 }; 2164 2165 VNET_DOMAIN_SET(route); 2166 2167 #ifdef DDB 2168 /* 2169 * Unfortunately, RTF_ values are expressed as raw masks rather than powers of 2170 * 2, so we cannot use them as nice C99 initializer indices below. 2171 */ 2172 static const char * const rtf_flag_strings[] = { 2173 "UP", 2174 "GATEWAY", 2175 "HOST", 2176 "REJECT", 2177 "DYNAMIC", 2178 "MODIFIED", 2179 "DONE", 2180 "UNUSED_0x80", 2181 "UNUSED_0x100", 2182 "XRESOLVE", 2183 "LLDATA", 2184 "STATIC", 2185 "BLACKHOLE", 2186 "UNUSED_0x2000", 2187 "PROTO2", 2188 "PROTO1", 2189 "UNUSED_0x10000", 2190 "UNUSED_0x20000", 2191 "PROTO3", 2192 "FIXEDMTU", 2193 "PINNED", 2194 "LOCAL", 2195 "BROADCAST", 2196 "MULTICAST", 2197 /* Big gap. */ 2198 [28] = "STICKY", 2199 [30] = "RNH_LOCKED", 2200 [31] = "GWFLAG_COMPAT", 2201 }; 2202 2203 static const char * __pure 2204 rt_flag_name(unsigned idx) 2205 { 2206 if (idx >= nitems(rtf_flag_strings)) 2207 return ("INVALID_FLAG"); 2208 if (rtf_flag_strings[idx] == NULL) 2209 return ("UNKNOWN"); 2210 return (rtf_flag_strings[idx]); 2211 } 2212 2213 static void 2214 rt_dumpaddr_ddb(const char *name, const struct sockaddr *sa) 2215 { 2216 char buf[INET6_ADDRSTRLEN], *res; 2217 2218 res = NULL; 2219 if (sa == NULL) 2220 res = "NULL"; 2221 else if (sa->sa_family == AF_INET) { 2222 res = inet_ntop(AF_INET, 2223 &((const struct sockaddr_in *)sa)->sin_addr, 2224 buf, sizeof(buf)); 2225 } else if (sa->sa_family == AF_INET6) { 2226 res = inet_ntop(AF_INET6, 2227 &((const struct sockaddr_in6 *)sa)->sin6_addr, 2228 buf, sizeof(buf)); 2229 } else if (sa->sa_family == AF_LINK) { 2230 res = "on link"; 2231 } 2232 2233 if (res != NULL) { 2234 db_printf("%s <%s> ", name, res); 2235 return; 2236 } 2237 2238 db_printf("%s <af:%d> ", name, sa->sa_family); 2239 } 2240 2241 static int 2242 rt_dumpentry_ddb(struct radix_node *rn, void *arg __unused) 2243 { 2244 struct sockaddr_storage ss; 2245 struct rtentry *rt; 2246 int flags, idx; 2247 2248 /* If RNTORT is important, put it in a header. */ 2249 rt = (void *)rn; 2250 2251 rt_dumpaddr_ddb("dst", rt_key(rt)); 2252 rt_dumpaddr_ddb("gateway", &rt->rt_nhop->gw_sa); 2253 rt_dumpaddr_ddb("netmask", rtsock_fix_netmask(rt_key(rt), rt_mask(rt), 2254 &ss)); 2255 if (rt->rt_ifp != NULL && (rt->rt_ifp->if_flags & IFF_DYING) == 0) { 2256 rt_dumpaddr_ddb("ifp", rt->rt_ifp->if_addr->ifa_addr); 2257 rt_dumpaddr_ddb("ifa", rt->rt_ifa->ifa_addr); 2258 } 2259 2260 db_printf("flags "); 2261 flags = rt->rt_flags; 2262 if (flags == 0) 2263 db_printf("none"); 2264 2265 while ((idx = ffs(flags)) > 0) { 2266 idx--; 2267 2268 if (flags != rt->rt_flags) 2269 db_printf(","); 2270 db_printf("%s", rt_flag_name(idx)); 2271 2272 flags &= ~(1ul << idx); 2273 } 2274 2275 db_printf("\n"); 2276 return (0); 2277 } 2278 2279 DB_SHOW_COMMAND(routetable, db_show_routetable_cmd) 2280 { 2281 struct rib_head *rnh; 2282 int error, i, lim; 2283 2284 if (have_addr) 2285 i = lim = addr; 2286 else { 2287 i = 1; 2288 lim = AF_MAX; 2289 } 2290 2291 for (; i <= lim; i++) { 2292 rnh = rt_tables_get_rnh(0, i); 2293 if (rnh == NULL) { 2294 if (have_addr) { 2295 db_printf("%s: AF %d not supported?\n", 2296 __func__, i); 2297 break; 2298 } 2299 continue; 2300 } 2301 2302 if (!have_addr && i > 1) 2303 db_printf("\n"); 2304 2305 db_printf("Route table for AF %d%s%s%s:\n", i, 2306 (i == AF_INET || i == AF_INET6) ? " (" : "", 2307 (i == AF_INET) ? "INET" : (i == AF_INET6) ? "INET6" : "", 2308 (i == AF_INET || i == AF_INET6) ? ")" : ""); 2309 2310 error = rnh->rnh_walktree(&rnh->head, rt_dumpentry_ddb, NULL); 2311 if (error != 0) 2312 db_printf("%s: walktree(%d): %d\n", __func__, i, 2313 error); 2314 } 2315 } 2316 2317 _DB_FUNC(_show, route, db_show_route_cmd, db_show_table, CS_OWN, NULL) 2318 { 2319 char buf[INET6_ADDRSTRLEN], *bp; 2320 const void *dst_addrp; 2321 struct sockaddr *dstp; 2322 struct rtentry *rt; 2323 union { 2324 struct sockaddr_in dest_sin; 2325 struct sockaddr_in6 dest_sin6; 2326 } u; 2327 uint16_t hextets[8]; 2328 unsigned i, tets; 2329 int t, af, exp, tokflags; 2330 2331 /* 2332 * Undecoded address family. No double-colon expansion seen yet. 2333 */ 2334 af = -1; 2335 exp = -1; 2336 /* Assume INET6 to start; we can work back if guess was wrong. */ 2337 tokflags = DRT_WSPACE | DRT_HEX | DRT_HEXADECIMAL; 2338 2339 /* 2340 * db_command has lexed 'show route' for us. 2341 */ 2342 t = db_read_token_flags(tokflags); 2343 if (t == tWSPACE) 2344 t = db_read_token_flags(tokflags); 2345 2346 /* 2347 * tEOL: Just 'show route' isn't a valid mode. 2348 * tMINUS: It's either '-h' or some invalid option. Regardless, usage. 2349 */ 2350 if (t == tEOL || t == tMINUS) 2351 goto usage; 2352 2353 db_unread_token(t); 2354 2355 tets = nitems(hextets); 2356 2357 /* 2358 * Each loop iteration, we expect to read one octet (v4) or hextet 2359 * (v6), followed by an appropriate field separator ('.' or ':' or 2360 * '::'). 2361 * 2362 * At the start of each loop, we're looking for a number (octet or 2363 * hextet). 2364 * 2365 * INET6 addresses have a special case where they may begin with '::'. 2366 */ 2367 for (i = 0; i < tets; i++) { 2368 t = db_read_token_flags(tokflags); 2369 2370 if (t == tCOLONCOLON) { 2371 /* INET6 with leading '::' or invalid. */ 2372 if (i != 0) { 2373 db_printf("Parse error: unexpected extra " 2374 "colons.\n"); 2375 goto exit; 2376 } 2377 2378 af = AF_INET6; 2379 exp = i; 2380 hextets[i] = 0; 2381 continue; 2382 } else if (t == tNUMBER) { 2383 /* 2384 * Lexer separates out '-' as tMINUS, but make the 2385 * assumption explicit here. 2386 */ 2387 MPASS(db_tok_number >= 0); 2388 2389 if (af == AF_INET && db_tok_number > UINT8_MAX) { 2390 db_printf("Not a valid v4 octet: %ld\n", 2391 (long)db_tok_number); 2392 goto exit; 2393 } 2394 hextets[i] = db_tok_number; 2395 } else if (t == tEOL) { 2396 /* 2397 * We can only detect the end of an IPv6 address in 2398 * compact representation with EOL. 2399 */ 2400 if (af != AF_INET6 || exp < 0) { 2401 db_printf("Parse failed. Got unexpected EOF " 2402 "when the address is not a compact-" 2403 "representation IPv6 address.\n"); 2404 goto exit; 2405 } 2406 break; 2407 } else { 2408 db_printf("Parse failed. Unexpected token %d.\n", t); 2409 goto exit; 2410 } 2411 2412 /* Next, look for a separator, if appropriate. */ 2413 if (i == tets - 1) 2414 continue; 2415 2416 t = db_read_token_flags(tokflags); 2417 if (af < 0) { 2418 if (t == tCOLON) { 2419 af = AF_INET6; 2420 continue; 2421 } 2422 if (t == tCOLONCOLON) { 2423 af = AF_INET6; 2424 i++; 2425 hextets[i] = 0; 2426 exp = i; 2427 continue; 2428 } 2429 if (t == tDOT) { 2430 unsigned hn, dn; 2431 2432 af = AF_INET; 2433 /* Need to fixup the first parsed number. */ 2434 if (hextets[0] > 0x255 || 2435 (hextets[0] & 0xf0) > 0x90 || 2436 (hextets[0] & 0xf) > 9) { 2437 db_printf("Not a valid v4 octet: %x\n", 2438 hextets[0]); 2439 goto exit; 2440 } 2441 2442 hn = hextets[0]; 2443 dn = (hn >> 8) * 100 + 2444 ((hn >> 4) & 0xf) * 10 + 2445 (hn & 0xf); 2446 2447 hextets[0] = dn; 2448 2449 /* Switch to decimal for remaining octets. */ 2450 tokflags &= ~DRT_RADIX_MASK; 2451 tokflags |= DRT_DECIMAL; 2452 2453 tets = 4; 2454 continue; 2455 } 2456 2457 db_printf("Parse error. Unexpected token %d.\n", t); 2458 goto exit; 2459 } else if (af == AF_INET) { 2460 if (t == tDOT) 2461 continue; 2462 db_printf("Expected '.' (%d) between octets but got " 2463 "(%d).\n", tDOT, t); 2464 goto exit; 2465 2466 } else if (af == AF_INET6) { 2467 if (t == tCOLON) 2468 continue; 2469 if (t == tCOLONCOLON) { 2470 if (exp < 0) { 2471 i++; 2472 hextets[i] = 0; 2473 exp = i; 2474 continue; 2475 } 2476 db_printf("Got bogus second '::' in v6 " 2477 "address.\n"); 2478 goto exit; 2479 } 2480 if (t == tEOL) { 2481 /* 2482 * Handle in the earlier part of the loop 2483 * because we need to handle trailing :: too. 2484 */ 2485 db_unread_token(t); 2486 continue; 2487 } 2488 2489 db_printf("Expected ':' (%d) or '::' (%d) between " 2490 "hextets but got (%d).\n", tCOLON, tCOLONCOLON, t); 2491 goto exit; 2492 } 2493 } 2494 2495 /* Check for trailing garbage. */ 2496 if (i == tets) { 2497 t = db_read_token_flags(tokflags); 2498 if (t != tEOL) { 2499 db_printf("Got unexpected garbage after address " 2500 "(%d).\n", t); 2501 goto exit; 2502 } 2503 } 2504 2505 /* 2506 * Need to expand compact INET6 addresses. 2507 * 2508 * Technically '::' for a single ':0:' is MUST NOT but just in case, 2509 * don't bother expanding that form (exp >= 0 && i == tets case). 2510 */ 2511 if (af == AF_INET6 && exp >= 0 && i < tets) { 2512 if (exp + 1 < i) { 2513 memmove(&hextets[exp + 1 + (nitems(hextets) - i)], 2514 &hextets[exp + 1], 2515 (i - (exp + 1)) * sizeof(hextets[0])); 2516 } 2517 memset(&hextets[exp + 1], 0, (nitems(hextets) - i) * 2518 sizeof(hextets[0])); 2519 } 2520 2521 memset(&u, 0, sizeof(u)); 2522 if (af == AF_INET) { 2523 u.dest_sin.sin_family = AF_INET; 2524 u.dest_sin.sin_len = sizeof(u.dest_sin); 2525 u.dest_sin.sin_addr.s_addr = htonl( 2526 ((uint32_t)hextets[0] << 24) | 2527 ((uint32_t)hextets[1] << 16) | 2528 ((uint32_t)hextets[2] << 8) | 2529 (uint32_t)hextets[3]); 2530 dstp = (void *)&u.dest_sin; 2531 dst_addrp = &u.dest_sin.sin_addr; 2532 } else if (af == AF_INET6) { 2533 u.dest_sin6.sin6_family = AF_INET6; 2534 u.dest_sin6.sin6_len = sizeof(u.dest_sin6); 2535 for (i = 0; i < nitems(hextets); i++) 2536 u.dest_sin6.sin6_addr.s6_addr16[i] = htons(hextets[i]); 2537 dstp = (void *)&u.dest_sin6; 2538 dst_addrp = &u.dest_sin6.sin6_addr; 2539 } else { 2540 MPASS(false); 2541 /* UNREACHABLE */ 2542 /* Appease Clang false positive: */ 2543 dstp = NULL; 2544 } 2545 2546 bp = inet_ntop(af, dst_addrp, buf, sizeof(buf)); 2547 if (bp != NULL) 2548 db_printf("Looking up route to destination '%s'\n", bp); 2549 2550 CURVNET_SET(vnet0); 2551 rt = rtalloc1(dstp, 0, RTF_RNH_LOCKED); 2552 CURVNET_RESTORE(); 2553 2554 if (rt == NULL) { 2555 db_printf("Could not get route for that server.\n"); 2556 return; 2557 } 2558 2559 rt_dumpentry_ddb((void *)rt, NULL); 2560 RTFREE_LOCKED(rt); 2561 2562 return; 2563 usage: 2564 db_printf("Usage: 'show route <address>'\n" 2565 " Currently accepts only dotted-decimal INET or colon-separated\n" 2566 " hextet INET6 addresses.\n"); 2567 exit: 2568 db_skip_to_eol(); 2569 } 2570 #endif 2571