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 #include <net/if.h> 58 #include <net/if_var.h> 59 #include <net/if_dl.h> 60 #include <net/if_llatbl.h> 61 #include <net/if_types.h> 62 #include <net/netisr.h> 63 #include <net/raw_cb.h> 64 #include <net/route.h> 65 #include <net/route/route_ctl.h> 66 #include <net/route/route_var.h> 67 #ifdef RADIX_MPATH 68 #include <net/radix_mpath.h> 69 #endif 70 #include <net/vnet.h> 71 72 #include <netinet/in.h> 73 #include <netinet/if_ether.h> 74 #include <netinet/ip_carp.h> 75 #ifdef INET6 76 #include <netinet6/ip6_var.h> 77 #include <netinet6/scope6_var.h> 78 #endif 79 #include <net/route/nhop.h> 80 #include <net/route/shared.h> 81 82 #ifdef COMPAT_FREEBSD32 83 #include <sys/mount.h> 84 #include <compat/freebsd32/freebsd32.h> 85 86 struct if_msghdr32 { 87 uint16_t ifm_msglen; 88 uint8_t ifm_version; 89 uint8_t ifm_type; 90 int32_t ifm_addrs; 91 int32_t ifm_flags; 92 uint16_t ifm_index; 93 uint16_t _ifm_spare1; 94 struct if_data ifm_data; 95 }; 96 97 struct if_msghdrl32 { 98 uint16_t ifm_msglen; 99 uint8_t ifm_version; 100 uint8_t ifm_type; 101 int32_t ifm_addrs; 102 int32_t ifm_flags; 103 uint16_t ifm_index; 104 uint16_t _ifm_spare1; 105 uint16_t ifm_len; 106 uint16_t ifm_data_off; 107 uint32_t _ifm_spare2; 108 struct if_data ifm_data; 109 }; 110 111 struct ifa_msghdrl32 { 112 uint16_t ifam_msglen; 113 uint8_t ifam_version; 114 uint8_t ifam_type; 115 int32_t ifam_addrs; 116 int32_t ifam_flags; 117 uint16_t ifam_index; 118 uint16_t _ifam_spare1; 119 uint16_t ifam_len; 120 uint16_t ifam_data_off; 121 int32_t ifam_metric; 122 struct if_data ifam_data; 123 }; 124 125 #define SA_SIZE32(sa) \ 126 ( (((struct sockaddr *)(sa))->sa_len == 0) ? \ 127 sizeof(int) : \ 128 1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) ) 129 130 #endif /* COMPAT_FREEBSD32 */ 131 132 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables"); 133 134 /* NB: these are not modified */ 135 static struct sockaddr route_src = { 2, PF_ROUTE, }; 136 static struct sockaddr sa_zero = { sizeof(sa_zero), AF_INET, }; 137 138 /* These are external hooks for CARP. */ 139 int (*carp_get_vhid_p)(struct ifaddr *); 140 141 /* 142 * Used by rtsock/raw_input callback code to decide whether to filter the update 143 * notification to a socket bound to a particular FIB. 144 */ 145 #define RTS_FILTER_FIB M_PROTO8 146 147 typedef struct { 148 int ip_count; /* attached w/ AF_INET */ 149 int ip6_count; /* attached w/ AF_INET6 */ 150 int any_count; /* total attached */ 151 } route_cb_t; 152 VNET_DEFINE_STATIC(route_cb_t, route_cb); 153 #define V_route_cb VNET(route_cb) 154 155 struct mtx rtsock_mtx; 156 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF); 157 158 #define RTSOCK_LOCK() mtx_lock(&rtsock_mtx) 159 #define RTSOCK_UNLOCK() mtx_unlock(&rtsock_mtx) 160 #define RTSOCK_LOCK_ASSERT() mtx_assert(&rtsock_mtx, MA_OWNED) 161 162 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 163 ""); 164 165 struct walkarg { 166 int w_tmemsize; 167 int w_op, w_arg; 168 caddr_t w_tmem; 169 struct sysctl_req *w_req; 170 }; 171 172 static void rts_input(struct mbuf *m); 173 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo); 174 static int rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, 175 struct walkarg *w, int *plen); 176 static int rt_xaddrs(caddr_t cp, caddr_t cplim, 177 struct rt_addrinfo *rtinfo); 178 static int sysctl_dumpentry(struct radix_node *rn, void *vw); 179 static int sysctl_iflist(int af, struct walkarg *w); 180 static int sysctl_ifmalist(int af, struct walkarg *w); 181 static int route_output(struct mbuf *m, struct socket *so, ...); 182 static void rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out); 183 static void rt_dispatch(struct mbuf *, sa_family_t); 184 static int handle_rtm_get(struct rt_addrinfo *info, u_int fibnum, 185 struct rt_msghdr *rtm, struct rib_cmd_info *rc); 186 static int update_rtm_from_rte(struct rt_addrinfo *info, 187 struct rt_msghdr **prtm, int alloc_len, 188 struct rtentry *rt, struct nhop_object *nh); 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, 215 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 216 0, 0, sysctl_route_netisr_maxqlen, "I", 217 "maximum routing socket dispatch queue length"); 218 219 static void 220 vnet_rts_init(void) 221 { 222 int tmp; 223 224 if (IS_DEFAULT_VNET(curvnet)) { 225 if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp)) 226 rtsock_nh.nh_qlimit = tmp; 227 netisr_register(&rtsock_nh); 228 } 229 #ifdef VIMAGE 230 else 231 netisr_register_vnet(&rtsock_nh); 232 #endif 233 } 234 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 235 vnet_rts_init, 0); 236 237 #ifdef VIMAGE 238 static void 239 vnet_rts_uninit(void) 240 { 241 242 netisr_unregister_vnet(&rtsock_nh); 243 } 244 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 245 vnet_rts_uninit, 0); 246 #endif 247 248 static int 249 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src, 250 struct rawcb *rp) 251 { 252 int fibnum; 253 254 KASSERT(m != NULL, ("%s: m is NULL", __func__)); 255 KASSERT(proto != NULL, ("%s: proto is NULL", __func__)); 256 KASSERT(rp != NULL, ("%s: rp is NULL", __func__)); 257 258 /* No filtering requested. */ 259 if ((m->m_flags & RTS_FILTER_FIB) == 0) 260 return (0); 261 262 /* Check if it is a rts and the fib matches the one of the socket. */ 263 fibnum = M_GETFIB(m); 264 if (proto->sp_family != PF_ROUTE || 265 rp->rcb_socket == NULL || 266 rp->rcb_socket->so_fibnum == fibnum) 267 return (0); 268 269 /* Filtering requested and no match, the socket shall be skipped. */ 270 return (1); 271 } 272 273 static void 274 rts_input(struct mbuf *m) 275 { 276 struct sockproto route_proto; 277 unsigned short *family; 278 struct m_tag *tag; 279 280 route_proto.sp_family = PF_ROUTE; 281 tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL); 282 if (tag != NULL) { 283 family = (unsigned short *)(tag + 1); 284 route_proto.sp_protocol = *family; 285 m_tag_delete(m, tag); 286 } else 287 route_proto.sp_protocol = 0; 288 289 raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb); 290 } 291 292 /* 293 * It really doesn't make any sense at all for this code to share much 294 * with raw_usrreq.c, since its functionality is so restricted. XXX 295 */ 296 static void 297 rts_abort(struct socket *so) 298 { 299 300 raw_usrreqs.pru_abort(so); 301 } 302 303 static void 304 rts_close(struct socket *so) 305 { 306 307 raw_usrreqs.pru_close(so); 308 } 309 310 /* pru_accept is EOPNOTSUPP */ 311 312 static int 313 rts_attach(struct socket *so, int proto, struct thread *td) 314 { 315 struct rawcb *rp; 316 int error; 317 318 KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL")); 319 320 /* XXX */ 321 rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO); 322 323 so->so_pcb = (caddr_t)rp; 324 so->so_fibnum = td->td_proc->p_fibnum; 325 error = raw_attach(so, proto); 326 rp = sotorawcb(so); 327 if (error) { 328 so->so_pcb = NULL; 329 free(rp, M_PCB); 330 return error; 331 } 332 RTSOCK_LOCK(); 333 switch(rp->rcb_proto.sp_protocol) { 334 case AF_INET: 335 V_route_cb.ip_count++; 336 break; 337 case AF_INET6: 338 V_route_cb.ip6_count++; 339 break; 340 } 341 V_route_cb.any_count++; 342 RTSOCK_UNLOCK(); 343 soisconnected(so); 344 so->so_options |= SO_USELOOPBACK; 345 return 0; 346 } 347 348 static int 349 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 350 { 351 352 return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */ 353 } 354 355 static int 356 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 357 { 358 359 return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */ 360 } 361 362 /* pru_connect2 is EOPNOTSUPP */ 363 /* pru_control is EOPNOTSUPP */ 364 365 static void 366 rts_detach(struct socket *so) 367 { 368 struct rawcb *rp = sotorawcb(so); 369 370 KASSERT(rp != NULL, ("rts_detach: rp == NULL")); 371 372 RTSOCK_LOCK(); 373 switch(rp->rcb_proto.sp_protocol) { 374 case AF_INET: 375 V_route_cb.ip_count--; 376 break; 377 case AF_INET6: 378 V_route_cb.ip6_count--; 379 break; 380 } 381 V_route_cb.any_count--; 382 RTSOCK_UNLOCK(); 383 raw_usrreqs.pru_detach(so); 384 } 385 386 static int 387 rts_disconnect(struct socket *so) 388 { 389 390 return (raw_usrreqs.pru_disconnect(so)); 391 } 392 393 /* pru_listen is EOPNOTSUPP */ 394 395 static int 396 rts_peeraddr(struct socket *so, struct sockaddr **nam) 397 { 398 399 return (raw_usrreqs.pru_peeraddr(so, nam)); 400 } 401 402 /* pru_rcvd is EOPNOTSUPP */ 403 /* pru_rcvoob is EOPNOTSUPP */ 404 405 static int 406 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 407 struct mbuf *control, struct thread *td) 408 { 409 410 return (raw_usrreqs.pru_send(so, flags, m, nam, control, td)); 411 } 412 413 /* pru_sense is null */ 414 415 static int 416 rts_shutdown(struct socket *so) 417 { 418 419 return (raw_usrreqs.pru_shutdown(so)); 420 } 421 422 static int 423 rts_sockaddr(struct socket *so, struct sockaddr **nam) 424 { 425 426 return (raw_usrreqs.pru_sockaddr(so, nam)); 427 } 428 429 static struct pr_usrreqs route_usrreqs = { 430 .pru_abort = rts_abort, 431 .pru_attach = rts_attach, 432 .pru_bind = rts_bind, 433 .pru_connect = rts_connect, 434 .pru_detach = rts_detach, 435 .pru_disconnect = rts_disconnect, 436 .pru_peeraddr = rts_peeraddr, 437 .pru_send = rts_send, 438 .pru_shutdown = rts_shutdown, 439 .pru_sockaddr = rts_sockaddr, 440 .pru_close = rts_close, 441 }; 442 443 #ifndef _SOCKADDR_UNION_DEFINED 444 #define _SOCKADDR_UNION_DEFINED 445 /* 446 * The union of all possible address formats we handle. 447 */ 448 union sockaddr_union { 449 struct sockaddr sa; 450 struct sockaddr_in sin; 451 struct sockaddr_in6 sin6; 452 }; 453 #endif /* _SOCKADDR_UNION_DEFINED */ 454 455 static int 456 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp, 457 struct nhop_object *nh, union sockaddr_union *saun, struct ucred *cred) 458 { 459 #if defined(INET) || defined(INET6) 460 struct epoch_tracker et; 461 #endif 462 463 /* First, see if the returned address is part of the jail. */ 464 if (prison_if(cred, nh->nh_ifa->ifa_addr) == 0) { 465 info->rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr; 466 return (0); 467 } 468 469 switch (info->rti_info[RTAX_DST]->sa_family) { 470 #ifdef INET 471 case AF_INET: 472 { 473 struct in_addr ia; 474 struct ifaddr *ifa; 475 int found; 476 477 found = 0; 478 /* 479 * Try to find an address on the given outgoing interface 480 * that belongs to the jail. 481 */ 482 NET_EPOCH_ENTER(et); 483 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 484 struct sockaddr *sa; 485 sa = ifa->ifa_addr; 486 if (sa->sa_family != AF_INET) 487 continue; 488 ia = ((struct sockaddr_in *)sa)->sin_addr; 489 if (prison_check_ip4(cred, &ia) == 0) { 490 found = 1; 491 break; 492 } 493 } 494 NET_EPOCH_EXIT(et); 495 if (!found) { 496 /* 497 * As a last resort return the 'default' jail address. 498 */ 499 ia = ((struct sockaddr_in *)nh->nh_ifa->ifa_addr)-> 500 sin_addr; 501 if (prison_get_ip4(cred, &ia) != 0) 502 return (ESRCH); 503 } 504 bzero(&saun->sin, sizeof(struct sockaddr_in)); 505 saun->sin.sin_len = sizeof(struct sockaddr_in); 506 saun->sin.sin_family = AF_INET; 507 saun->sin.sin_addr.s_addr = ia.s_addr; 508 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin; 509 break; 510 } 511 #endif 512 #ifdef INET6 513 case AF_INET6: 514 { 515 struct in6_addr ia6; 516 struct ifaddr *ifa; 517 int found; 518 519 found = 0; 520 /* 521 * Try to find an address on the given outgoing interface 522 * that belongs to the jail. 523 */ 524 NET_EPOCH_ENTER(et); 525 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 526 struct sockaddr *sa; 527 sa = ifa->ifa_addr; 528 if (sa->sa_family != AF_INET6) 529 continue; 530 bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr, 531 &ia6, sizeof(struct in6_addr)); 532 if (prison_check_ip6(cred, &ia6) == 0) { 533 found = 1; 534 break; 535 } 536 } 537 NET_EPOCH_EXIT(et); 538 if (!found) { 539 /* 540 * As a last resort return the 'default' jail address. 541 */ 542 ia6 = ((struct sockaddr_in6 *)nh->nh_ifa->ifa_addr)-> 543 sin6_addr; 544 if (prison_get_ip6(cred, &ia6) != 0) 545 return (ESRCH); 546 } 547 bzero(&saun->sin6, sizeof(struct sockaddr_in6)); 548 saun->sin6.sin6_len = sizeof(struct sockaddr_in6); 549 saun->sin6.sin6_family = AF_INET6; 550 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr)); 551 if (sa6_recoverscope(&saun->sin6) != 0) 552 return (ESRCH); 553 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6; 554 break; 555 } 556 #endif 557 default: 558 return (ESRCH); 559 } 560 return (0); 561 } 562 563 /* 564 * Fills in @info based on userland-provided @rtm message. 565 * 566 * Returns 0 on success. 567 */ 568 static int 569 fill_addrinfo(struct rt_msghdr *rtm, int len, u_int fibnum, struct rt_addrinfo *info) 570 { 571 int error; 572 sa_family_t saf; 573 574 rtm->rtm_pid = curproc->p_pid; 575 info->rti_addrs = rtm->rtm_addrs; 576 577 info->rti_mflags = rtm->rtm_inits; 578 info->rti_rmx = &rtm->rtm_rmx; 579 580 /* 581 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6 582 * link-local address because rtrequest requires addresses with 583 * embedded scope id. 584 */ 585 if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info)) 586 return (EINVAL); 587 588 if (rtm->rtm_flags & RTF_RNH_LOCKED) 589 return (EINVAL); 590 info->rti_flags = rtm->rtm_flags; 591 if (info->rti_info[RTAX_DST] == NULL || 592 info->rti_info[RTAX_DST]->sa_family >= AF_MAX || 593 (info->rti_info[RTAX_GATEWAY] != NULL && 594 info->rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) 595 return (EINVAL); 596 saf = info->rti_info[RTAX_DST]->sa_family; 597 /* 598 * Verify that the caller has the appropriate privilege; RTM_GET 599 * is the only operation the non-superuser is allowed. 600 */ 601 if (rtm->rtm_type != RTM_GET) { 602 error = priv_check(curthread, PRIV_NET_ROUTE); 603 if (error != 0) 604 return (error); 605 } 606 607 /* 608 * The given gateway address may be an interface address. 609 * For example, issuing a "route change" command on a route 610 * entry that was created from a tunnel, and the gateway 611 * address given is the local end point. In this case the 612 * RTF_GATEWAY flag must be cleared or the destination will 613 * not be reachable even though there is no error message. 614 */ 615 if (info->rti_info[RTAX_GATEWAY] != NULL && 616 info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) { 617 struct rt_addrinfo ginfo; 618 struct sockaddr *gdst; 619 struct sockaddr_storage ss; 620 621 bzero(&ginfo, sizeof(ginfo)); 622 bzero(&ss, sizeof(ss)); 623 ss.ss_len = sizeof(ss); 624 625 ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss; 626 gdst = info->rti_info[RTAX_GATEWAY]; 627 628 /* 629 * A host route through the loopback interface is 630 * installed for each interface adddress. In pre 8.0 631 * releases the interface address of a PPP link type 632 * is not reachable locally. This behavior is fixed as 633 * part of the new L2/L3 redesign and rewrite work. The 634 * signature of this interface address route is the 635 * AF_LINK sa_family type of the gateway, and the 636 * rt_ifp has the IFF_LOOPBACK flag set. 637 */ 638 if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) { 639 if (ss.ss_family == AF_LINK && 640 ginfo.rti_ifp->if_flags & IFF_LOOPBACK) { 641 info->rti_flags &= ~RTF_GATEWAY; 642 info->rti_flags |= RTF_GWFLAG_COMPAT; 643 } 644 rib_free_info(&ginfo); 645 } 646 } 647 648 return (0); 649 } 650 651 /* 652 * Handles RTM_GET message from routing socket, returning matching rt. 653 * 654 * Returns: 655 * 0 on success, with locked and referenced matching rt in @rt_nrt 656 * errno of failure 657 */ 658 static int 659 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum, 660 struct rt_msghdr *rtm, struct rib_cmd_info *rc) 661 { 662 RIB_RLOCK_TRACKER; 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 rc->rc_rt = (struct rtentry *) rnh->rnh_matchaddr( 681 info->rti_info[RTAX_DST], &rnh->head); 682 } else 683 rc->rc_rt = (struct rtentry *) rnh->rnh_lookup( 684 info->rti_info[RTAX_DST], 685 info->rti_info[RTAX_NETMASK], &rnh->head); 686 687 if (rc->rc_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 rc->rc_rt = rt_mpath_matchgate(rc->rc_rt, 699 info->rti_info[RTAX_GATEWAY]); 700 if (rc->rc_rt == NULL) { 701 RIB_RUNLOCK(rnh); 702 return (ESRCH); 703 } 704 } 705 #endif 706 /* 707 * If performing proxied L2 entry insertion, and 708 * the actual PPP host entry is found, perform 709 * another search to retrieve the prefix route of 710 * the local end point of the PPP link. 711 * TODO: move this logic to userland. 712 */ 713 if (rtm->rtm_flags & RTF_ANNOUNCE) { 714 struct sockaddr laddr; 715 struct nhop_object *nh; 716 717 nh = rc->rc_rt->rt_nhop; 718 if (nh->nh_ifp != NULL && 719 nh->nh_ifp->if_type == IFT_PROPVIRTUAL) { 720 struct ifaddr *ifa; 721 722 ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1, 723 RT_ALL_FIBS); 724 if (ifa != NULL) 725 rt_maskedcopy(ifa->ifa_addr, 726 &laddr, 727 ifa->ifa_netmask); 728 } else 729 rt_maskedcopy(nh->nh_ifa->ifa_addr, 730 &laddr, 731 nh->nh_ifa->ifa_netmask); 732 /* 733 * refactor rt and no lock operation necessary 734 */ 735 rc->rc_rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr, 736 &rnh->head); 737 if (rc->rc_rt == NULL) { 738 RIB_RUNLOCK(rnh); 739 return (ESRCH); 740 } 741 } 742 rc->rc_nh_new = rc->rc_rt->rt_nhop; 743 RIB_RUNLOCK(rnh); 744 745 return (0); 746 } 747 748 /* 749 * Update sockaddrs, flags, etc in @prtm based on @rt data. 750 * rtm can be reallocated. 751 * 752 * Returns 0 on success, along with pointer to (potentially reallocated) 753 * rtm. 754 * 755 */ 756 static int 757 update_rtm_from_rte(struct rt_addrinfo *info, struct rt_msghdr **prtm, 758 int alloc_len, struct rtentry *rt, struct nhop_object *nh) 759 { 760 struct sockaddr_storage netmask_ss; 761 struct walkarg w; 762 union sockaddr_union saun; 763 struct rt_msghdr *rtm, *orig_rtm = NULL; 764 struct ifnet *ifp; 765 int error, len; 766 767 rtm = *prtm; 768 769 info->rti_info[RTAX_DST] = rt_key(rt); 770 info->rti_info[RTAX_GATEWAY] = &nh->gw_sa; 771 info->rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt), 772 rt_mask(rt), &netmask_ss); 773 info->rti_info[RTAX_GENMASK] = 0; 774 ifp = nh->nh_ifp; 775 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) { 776 if (ifp) { 777 info->rti_info[RTAX_IFP] = 778 ifp->if_addr->ifa_addr; 779 error = rtm_get_jailed(info, ifp, nh, 780 &saun, curthread->td_ucred); 781 if (error != 0) 782 return (error); 783 if (ifp->if_flags & IFF_POINTOPOINT) 784 info->rti_info[RTAX_BRD] = 785 nh->nh_ifa->ifa_dstaddr; 786 rtm->rtm_index = ifp->if_index; 787 } else { 788 info->rti_info[RTAX_IFP] = NULL; 789 info->rti_info[RTAX_IFA] = NULL; 790 } 791 } else if (ifp != NULL) 792 rtm->rtm_index = ifp->if_index; 793 794 /* Check if we need to realloc storage */ 795 rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len); 796 if (len > alloc_len) { 797 struct rt_msghdr *tmp_rtm; 798 799 tmp_rtm = malloc(len, M_TEMP, M_NOWAIT); 800 if (tmp_rtm == NULL) 801 return (ENOBUFS); 802 bcopy(rtm, tmp_rtm, rtm->rtm_msglen); 803 orig_rtm = rtm; 804 rtm = tmp_rtm; 805 alloc_len = len; 806 807 /* 808 * Delay freeing original rtm as info contains 809 * data referencing it. 810 */ 811 } 812 813 w.w_tmem = (caddr_t)rtm; 814 w.w_tmemsize = alloc_len; 815 rtsock_msg_buffer(rtm->rtm_type, info, &w, &len); 816 817 if (rt->rt_flags & RTF_GWFLAG_COMPAT) 818 rtm->rtm_flags = RTF_GATEWAY | 819 (rt->rt_flags & ~RTF_GWFLAG_COMPAT); 820 else 821 rtm->rtm_flags = rt->rt_flags; 822 rt_getmetrics(rt, &rtm->rtm_rmx); 823 rtm->rtm_addrs = info->rti_addrs; 824 825 if (orig_rtm != NULL) 826 free(orig_rtm, M_TEMP); 827 *prtm = rtm; 828 829 return (0); 830 } 831 832 /*ARGSUSED*/ 833 static int 834 route_output(struct mbuf *m, struct socket *so, ...) 835 { 836 struct rt_msghdr *rtm = NULL; 837 struct rtentry *rt = NULL; 838 struct rt_addrinfo info; 839 struct epoch_tracker et; 840 #ifdef INET6 841 struct sockaddr_storage ss; 842 struct sockaddr_in6 *sin6; 843 int i, rti_need_deembed = 0; 844 #endif 845 int alloc_len = 0, len, error = 0, fibnum; 846 sa_family_t saf = AF_UNSPEC; 847 struct walkarg w; 848 struct rib_cmd_info rc; 849 struct nhop_object *nh; 850 851 fibnum = so->so_fibnum; 852 853 #define senderr(e) { error = e; goto flush;} 854 if (m == NULL || ((m->m_len < sizeof(long)) && 855 (m = m_pullup(m, sizeof(long))) == NULL)) 856 return (ENOBUFS); 857 if ((m->m_flags & M_PKTHDR) == 0) 858 panic("route_output"); 859 NET_EPOCH_ENTER(et); 860 len = m->m_pkthdr.len; 861 if (len < sizeof(*rtm) || 862 len != mtod(m, struct rt_msghdr *)->rtm_msglen) 863 senderr(EINVAL); 864 865 /* 866 * Most of current messages are in range 200-240 bytes, 867 * minimize possible re-allocation on reply using larger size 868 * buffer aligned on 1k boundaty. 869 */ 870 alloc_len = roundup2(len, 1024); 871 if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL) 872 senderr(ENOBUFS); 873 874 m_copydata(m, 0, len, (caddr_t)rtm); 875 bzero(&info, sizeof(info)); 876 bzero(&w, sizeof(w)); 877 nh = NULL; 878 879 if (rtm->rtm_version != RTM_VERSION) { 880 /* Do not touch message since format is unknown */ 881 free(rtm, M_TEMP); 882 rtm = NULL; 883 senderr(EPROTONOSUPPORT); 884 } 885 886 /* 887 * Starting from here, it is possible 888 * to alter original message and insert 889 * caller PID and error value. 890 */ 891 892 if ((error = fill_addrinfo(rtm, len, fibnum, &info)) != 0) { 893 senderr(error); 894 } 895 896 saf = info.rti_info[RTAX_DST]->sa_family; 897 898 /* support for new ARP code */ 899 if (rtm->rtm_flags & RTF_LLDATA) { 900 error = lla_rt_output(rtm, &info); 901 #ifdef INET6 902 if (error == 0) 903 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0; 904 #endif 905 goto flush; 906 } 907 908 switch (rtm->rtm_type) { 909 case RTM_ADD: 910 case RTM_CHANGE: 911 if (rtm->rtm_type == RTM_ADD) { 912 if (info.rti_info[RTAX_GATEWAY] == NULL) 913 senderr(EINVAL); 914 } 915 error = rib_action(fibnum, rtm->rtm_type, &info, &rc); 916 if (error == 0) { 917 #ifdef INET6 918 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0; 919 #endif 920 rtm->rtm_index = rc.rc_nh_new->nh_ifp->if_index; 921 nh = rc.rc_nh_new; 922 } 923 break; 924 925 case RTM_DELETE: 926 error = rib_action(fibnum, RTM_DELETE, &info, &rc); 927 if (error == 0) { 928 nh = rc.rc_nh_old; 929 goto report; 930 } 931 #ifdef INET6 932 /* rt_msg2() will not be used when RTM_DELETE fails. */ 933 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0; 934 #endif 935 break; 936 937 case RTM_GET: 938 error = handle_rtm_get(&info, fibnum, rtm, &rc); 939 if (error != 0) 940 senderr(error); 941 nh = rc.rc_nh_new; 942 943 report: 944 if (!can_export_rte(curthread->td_ucred, rc.rc_rt)) { 945 senderr(ESRCH); 946 } 947 948 error = update_rtm_from_rte(&info, &rtm, alloc_len, rc.rc_rt, nh); 949 /* 950 * Note that some sockaddr pointers may have changed to 951 * point to memory outsize @rtm. Some may be pointing 952 * to the on-stack variables. 953 * Given that, any pointer in @info CANNOT BE USED. 954 */ 955 956 /* 957 * scopeid deembedding has been performed while 958 * writing updated rtm in rtsock_msg_buffer(). 959 * With that in mind, skip deembedding procedure below. 960 */ 961 #ifdef INET6 962 rti_need_deembed = 0; 963 #endif 964 if (error != 0) 965 senderr(error); 966 break; 967 968 default: 969 senderr(EOPNOTSUPP); 970 } 971 972 flush: 973 NET_EPOCH_EXIT(et); 974 rt = NULL; 975 976 #ifdef INET6 977 if (rtm != NULL) { 978 if (rti_need_deembed) { 979 /* sin6_scope_id is recovered before sending rtm. */ 980 sin6 = (struct sockaddr_in6 *)&ss; 981 for (i = 0; i < RTAX_MAX; i++) { 982 if (info.rti_info[i] == NULL) 983 continue; 984 if (info.rti_info[i]->sa_family != AF_INET6) 985 continue; 986 bcopy(info.rti_info[i], sin6, sizeof(*sin6)); 987 if (sa6_recoverscope(sin6) == 0) 988 bcopy(sin6, info.rti_info[i], 989 sizeof(*sin6)); 990 } 991 } 992 } 993 #endif 994 send_rtm_reply(so, rtm, m, saf, fibnum, error); 995 996 return (error); 997 } 998 999 /* 1000 * Sends the prepared reply message in @rtm to all rtsock clients. 1001 * Frees @m and @rtm. 1002 * 1003 */ 1004 static void 1005 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m, 1006 sa_family_t saf, u_int fibnum, int rtm_errno) 1007 { 1008 struct rawcb *rp = NULL; 1009 1010 /* 1011 * Check to see if we don't want our own messages. 1012 */ 1013 if ((so->so_options & SO_USELOOPBACK) == 0) { 1014 if (V_route_cb.any_count <= 1) { 1015 if (rtm != NULL) 1016 free(rtm, M_TEMP); 1017 m_freem(m); 1018 return; 1019 } 1020 /* There is another listener, so construct message */ 1021 rp = sotorawcb(so); 1022 } 1023 1024 if (rtm != NULL) { 1025 if (rtm_errno!= 0) 1026 rtm->rtm_errno = rtm_errno; 1027 else 1028 rtm->rtm_flags |= RTF_DONE; 1029 1030 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm); 1031 if (m->m_pkthdr.len < rtm->rtm_msglen) { 1032 m_freem(m); 1033 m = NULL; 1034 } else if (m->m_pkthdr.len > rtm->rtm_msglen) 1035 m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len); 1036 1037 free(rtm, M_TEMP); 1038 } 1039 if (m != NULL) { 1040 M_SETFIB(m, fibnum); 1041 m->m_flags |= RTS_FILTER_FIB; 1042 if (rp) { 1043 /* 1044 * XXX insure we don't get a copy by 1045 * invalidating our protocol 1046 */ 1047 unsigned short family = rp->rcb_proto.sp_family; 1048 rp->rcb_proto.sp_family = 0; 1049 rt_dispatch(m, saf); 1050 rp->rcb_proto.sp_family = family; 1051 } else 1052 rt_dispatch(m, saf); 1053 } 1054 } 1055 1056 1057 static void 1058 rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out) 1059 { 1060 1061 bzero(out, sizeof(*out)); 1062 out->rmx_mtu = rt->rt_nhop->nh_mtu; 1063 out->rmx_weight = rt->rt_weight; 1064 out->rmx_nhidx = nhop_get_idx(rt->rt_nhop); 1065 /* Kernel -> userland timebase conversion. */ 1066 out->rmx_expire = rt->rt_expire ? 1067 rt->rt_expire - time_uptime + time_second : 0; 1068 } 1069 1070 /* 1071 * Extract the addresses of the passed sockaddrs. 1072 * Do a little sanity checking so as to avoid bad memory references. 1073 * This data is derived straight from userland. 1074 */ 1075 static int 1076 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo) 1077 { 1078 struct sockaddr *sa; 1079 int i; 1080 1081 for (i = 0; i < RTAX_MAX && cp < cplim; i++) { 1082 if ((rtinfo->rti_addrs & (1 << i)) == 0) 1083 continue; 1084 sa = (struct sockaddr *)cp; 1085 /* 1086 * It won't fit. 1087 */ 1088 if (cp + sa->sa_len > cplim) 1089 return (EINVAL); 1090 /* 1091 * there are no more.. quit now 1092 * If there are more bits, they are in error. 1093 * I've seen this. route(1) can evidently generate these. 1094 * This causes kernel to core dump. 1095 * for compatibility, If we see this, point to a safe address. 1096 */ 1097 if (sa->sa_len == 0) { 1098 rtinfo->rti_info[i] = &sa_zero; 1099 return (0); /* should be EINVAL but for compat */ 1100 } 1101 /* accept it */ 1102 #ifdef INET6 1103 if (sa->sa_family == AF_INET6) 1104 sa6_embedscope((struct sockaddr_in6 *)sa, 1105 V_ip6_use_defzone); 1106 #endif 1107 rtinfo->rti_info[i] = sa; 1108 cp += SA_SIZE(sa); 1109 } 1110 return (0); 1111 } 1112 1113 /* 1114 * Fill in @dmask with valid netmask leaving original @smask 1115 * intact. Mostly used with radix netmasks. 1116 */ 1117 struct sockaddr * 1118 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask, 1119 struct sockaddr_storage *dmask) 1120 { 1121 if (dst == NULL || smask == NULL) 1122 return (NULL); 1123 1124 memset(dmask, 0, dst->sa_len); 1125 memcpy(dmask, smask, smask->sa_len); 1126 dmask->ss_len = dst->sa_len; 1127 dmask->ss_family = dst->sa_family; 1128 1129 return ((struct sockaddr *)dmask); 1130 } 1131 1132 /* 1133 * Writes information related to @rtinfo object to newly-allocated mbuf. 1134 * Assumes MCLBYTES is enough to construct any message. 1135 * Used for OS notifications of vaious events (if/ifa announces,etc) 1136 * 1137 * Returns allocated mbuf or NULL on failure. 1138 */ 1139 static struct mbuf * 1140 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo) 1141 { 1142 struct rt_msghdr *rtm; 1143 struct mbuf *m; 1144 int i; 1145 struct sockaddr *sa; 1146 #ifdef INET6 1147 struct sockaddr_storage ss; 1148 struct sockaddr_in6 *sin6; 1149 #endif 1150 int len, dlen; 1151 1152 switch (type) { 1153 1154 case RTM_DELADDR: 1155 case RTM_NEWADDR: 1156 len = sizeof(struct ifa_msghdr); 1157 break; 1158 1159 case RTM_DELMADDR: 1160 case RTM_NEWMADDR: 1161 len = sizeof(struct ifma_msghdr); 1162 break; 1163 1164 case RTM_IFINFO: 1165 len = sizeof(struct if_msghdr); 1166 break; 1167 1168 case RTM_IFANNOUNCE: 1169 case RTM_IEEE80211: 1170 len = sizeof(struct if_announcemsghdr); 1171 break; 1172 1173 default: 1174 len = sizeof(struct rt_msghdr); 1175 } 1176 1177 /* XXXGL: can we use MJUMPAGESIZE cluster here? */ 1178 KASSERT(len <= MCLBYTES, ("%s: message too big", __func__)); 1179 if (len > MHLEN) 1180 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 1181 else 1182 m = m_gethdr(M_NOWAIT, MT_DATA); 1183 if (m == NULL) 1184 return (m); 1185 1186 m->m_pkthdr.len = m->m_len = len; 1187 rtm = mtod(m, struct rt_msghdr *); 1188 bzero((caddr_t)rtm, len); 1189 for (i = 0; i < RTAX_MAX; i++) { 1190 if ((sa = rtinfo->rti_info[i]) == NULL) 1191 continue; 1192 rtinfo->rti_addrs |= (1 << i); 1193 dlen = SA_SIZE(sa); 1194 #ifdef INET6 1195 if (V_deembed_scopeid && sa->sa_family == AF_INET6) { 1196 sin6 = (struct sockaddr_in6 *)&ss; 1197 bcopy(sa, sin6, sizeof(*sin6)); 1198 if (sa6_recoverscope(sin6) == 0) 1199 sa = (struct sockaddr *)sin6; 1200 } 1201 #endif 1202 m_copyback(m, len, dlen, (caddr_t)sa); 1203 len += dlen; 1204 } 1205 if (m->m_pkthdr.len != len) { 1206 m_freem(m); 1207 return (NULL); 1208 } 1209 rtm->rtm_msglen = len; 1210 rtm->rtm_version = RTM_VERSION; 1211 rtm->rtm_type = type; 1212 return (m); 1213 } 1214 1215 /* 1216 * Writes information related to @rtinfo object to preallocated buffer. 1217 * Stores needed size in @plen. If @w is NULL, calculates size without 1218 * writing. 1219 * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation. 1220 * 1221 * Returns 0 on success. 1222 * 1223 */ 1224 static int 1225 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen) 1226 { 1227 int i; 1228 int len, buflen = 0, dlen; 1229 caddr_t cp = NULL; 1230 struct rt_msghdr *rtm = NULL; 1231 #ifdef INET6 1232 struct sockaddr_storage ss; 1233 struct sockaddr_in6 *sin6; 1234 #endif 1235 #ifdef COMPAT_FREEBSD32 1236 bool compat32 = false; 1237 #endif 1238 1239 switch (type) { 1240 1241 case RTM_DELADDR: 1242 case RTM_NEWADDR: 1243 if (w != NULL && w->w_op == NET_RT_IFLISTL) { 1244 #ifdef COMPAT_FREEBSD32 1245 if (w->w_req->flags & SCTL_MASK32) { 1246 len = sizeof(struct ifa_msghdrl32); 1247 compat32 = true; 1248 } else 1249 #endif 1250 len = sizeof(struct ifa_msghdrl); 1251 } else 1252 len = sizeof(struct ifa_msghdr); 1253 break; 1254 1255 case RTM_IFINFO: 1256 #ifdef COMPAT_FREEBSD32 1257 if (w != NULL && w->w_req->flags & SCTL_MASK32) { 1258 if (w->w_op == NET_RT_IFLISTL) 1259 len = sizeof(struct if_msghdrl32); 1260 else 1261 len = sizeof(struct if_msghdr32); 1262 compat32 = true; 1263 break; 1264 } 1265 #endif 1266 if (w != NULL && w->w_op == NET_RT_IFLISTL) 1267 len = sizeof(struct if_msghdrl); 1268 else 1269 len = sizeof(struct if_msghdr); 1270 break; 1271 1272 case RTM_NEWMADDR: 1273 len = sizeof(struct ifma_msghdr); 1274 break; 1275 1276 default: 1277 len = sizeof(struct rt_msghdr); 1278 } 1279 1280 if (w != NULL) { 1281 rtm = (struct rt_msghdr *)w->w_tmem; 1282 buflen = w->w_tmemsize - len; 1283 cp = (caddr_t)w->w_tmem + len; 1284 } 1285 1286 rtinfo->rti_addrs = 0; 1287 for (i = 0; i < RTAX_MAX; i++) { 1288 struct sockaddr *sa; 1289 1290 if ((sa = rtinfo->rti_info[i]) == NULL) 1291 continue; 1292 rtinfo->rti_addrs |= (1 << i); 1293 #ifdef COMPAT_FREEBSD32 1294 if (compat32) 1295 dlen = SA_SIZE32(sa); 1296 else 1297 #endif 1298 dlen = SA_SIZE(sa); 1299 if (cp != NULL && buflen >= dlen) { 1300 #ifdef INET6 1301 if (V_deembed_scopeid && sa->sa_family == AF_INET6) { 1302 sin6 = (struct sockaddr_in6 *)&ss; 1303 bcopy(sa, sin6, sizeof(*sin6)); 1304 if (sa6_recoverscope(sin6) == 0) 1305 sa = (struct sockaddr *)sin6; 1306 } 1307 #endif 1308 bcopy((caddr_t)sa, cp, (unsigned)dlen); 1309 cp += dlen; 1310 buflen -= dlen; 1311 } else if (cp != NULL) { 1312 /* 1313 * Buffer too small. Count needed size 1314 * and return with error. 1315 */ 1316 cp = NULL; 1317 } 1318 1319 len += dlen; 1320 } 1321 1322 if (cp != NULL) { 1323 dlen = ALIGN(len) - len; 1324 if (buflen < dlen) 1325 cp = NULL; 1326 else { 1327 bzero(cp, dlen); 1328 cp += dlen; 1329 buflen -= dlen; 1330 } 1331 } 1332 len = ALIGN(len); 1333 1334 if (cp != NULL) { 1335 /* fill header iff buffer is large enough */ 1336 rtm->rtm_version = RTM_VERSION; 1337 rtm->rtm_type = type; 1338 rtm->rtm_msglen = len; 1339 } 1340 1341 *plen = len; 1342 1343 if (w != NULL && cp == NULL) 1344 return (ENOBUFS); 1345 1346 return (0); 1347 } 1348 1349 /* 1350 * This routine is called to generate a message from the routing 1351 * socket indicating that a redirect has occurred, a routing lookup 1352 * has failed, or that a protocol has detected timeouts to a particular 1353 * destination. 1354 */ 1355 void 1356 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error, 1357 int fibnum) 1358 { 1359 struct rt_msghdr *rtm; 1360 struct mbuf *m; 1361 struct sockaddr *sa = rtinfo->rti_info[RTAX_DST]; 1362 1363 if (V_route_cb.any_count == 0) 1364 return; 1365 m = rtsock_msg_mbuf(type, rtinfo); 1366 if (m == NULL) 1367 return; 1368 1369 if (fibnum != RT_ALL_FIBS) { 1370 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out " 1371 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs)); 1372 M_SETFIB(m, fibnum); 1373 m->m_flags |= RTS_FILTER_FIB; 1374 } 1375 1376 rtm = mtod(m, struct rt_msghdr *); 1377 rtm->rtm_flags = RTF_DONE | flags; 1378 rtm->rtm_errno = error; 1379 rtm->rtm_addrs = rtinfo->rti_addrs; 1380 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1381 } 1382 1383 void 1384 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error) 1385 { 1386 1387 rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS); 1388 } 1389 1390 /* 1391 * This routine is called to generate a message from the routing 1392 * socket indicating that the status of a network interface has changed. 1393 */ 1394 void 1395 rt_ifmsg(struct ifnet *ifp) 1396 { 1397 struct if_msghdr *ifm; 1398 struct mbuf *m; 1399 struct rt_addrinfo info; 1400 1401 if (V_route_cb.any_count == 0) 1402 return; 1403 bzero((caddr_t)&info, sizeof(info)); 1404 m = rtsock_msg_mbuf(RTM_IFINFO, &info); 1405 if (m == NULL) 1406 return; 1407 ifm = mtod(m, struct if_msghdr *); 1408 ifm->ifm_index = ifp->if_index; 1409 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1410 if_data_copy(ifp, &ifm->ifm_data); 1411 ifm->ifm_addrs = 0; 1412 rt_dispatch(m, AF_UNSPEC); 1413 } 1414 1415 /* 1416 * Announce interface address arrival/withdraw. 1417 * Please do not call directly, use rt_addrmsg(). 1418 * Assume input data to be valid. 1419 * Returns 0 on success. 1420 */ 1421 int 1422 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum) 1423 { 1424 struct rt_addrinfo info; 1425 struct sockaddr *sa; 1426 int ncmd; 1427 struct mbuf *m; 1428 struct ifa_msghdr *ifam; 1429 struct ifnet *ifp = ifa->ifa_ifp; 1430 struct sockaddr_storage ss; 1431 1432 if (V_route_cb.any_count == 0) 1433 return (0); 1434 1435 ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR; 1436 1437 bzero((caddr_t)&info, sizeof(info)); 1438 info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr; 1439 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr; 1440 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask( 1441 info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss); 1442 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr; 1443 if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL) 1444 return (ENOBUFS); 1445 ifam = mtod(m, struct ifa_msghdr *); 1446 ifam->ifam_index = ifp->if_index; 1447 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 1448 ifam->ifam_flags = ifa->ifa_flags; 1449 ifam->ifam_addrs = info.rti_addrs; 1450 1451 if (fibnum != RT_ALL_FIBS) { 1452 M_SETFIB(m, fibnum); 1453 m->m_flags |= RTS_FILTER_FIB; 1454 } 1455 1456 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1457 1458 return (0); 1459 } 1460 1461 /* 1462 * Announce route addition/removal to rtsock based on @rt data. 1463 * Callers are advives to use rt_routemsg() instead of using this 1464 * function directly. 1465 * Assume @rt data is consistent. 1466 * 1467 * Returns 0 on success. 1468 */ 1469 int 1470 rtsock_routemsg(int cmd, struct rtentry *rt, struct ifnet *ifp, int rti_addrs, 1471 int fibnum) 1472 { 1473 struct sockaddr_storage ss; 1474 struct rt_addrinfo info; 1475 1476 if (V_route_cb.any_count == 0) 1477 return (0); 1478 1479 bzero((caddr_t)&info, sizeof(info)); 1480 info.rti_info[RTAX_DST] = rt_key(rt); 1481 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt), rt_mask(rt), &ss); 1482 info.rti_info[RTAX_GATEWAY] = &rt->rt_nhop->gw_sa; 1483 info.rti_flags = rt->rt_flags; 1484 info.rti_ifp = ifp; 1485 1486 return (rtsock_routemsg_info(cmd, &info, fibnum)); 1487 } 1488 1489 int 1490 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum) 1491 { 1492 struct rt_msghdr *rtm; 1493 struct sockaddr *sa; 1494 struct mbuf *m; 1495 1496 if (V_route_cb.any_count == 0) 1497 return (0); 1498 1499 if (info->rti_flags & RTF_HOST) 1500 info->rti_info[RTAX_NETMASK] = NULL; 1501 1502 m = rtsock_msg_mbuf(cmd, info); 1503 if (m == NULL) 1504 return (ENOBUFS); 1505 1506 if (fibnum != RT_ALL_FIBS) { 1507 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out " 1508 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs)); 1509 M_SETFIB(m, fibnum); 1510 m->m_flags |= RTS_FILTER_FIB; 1511 } 1512 1513 rtm = mtod(m, struct rt_msghdr *); 1514 rtm->rtm_addrs = info->rti_addrs; 1515 if (info->rti_ifp != NULL) 1516 rtm->rtm_index = info->rti_ifp->if_index; 1517 /* Add RTF_DONE to indicate command 'completion' required by API */ 1518 info->rti_flags |= RTF_DONE; 1519 /* Reported routes has to be up */ 1520 if (cmd == RTM_ADD || cmd == RTM_CHANGE) 1521 info->rti_flags |= RTF_UP; 1522 rtm->rtm_flags = info->rti_flags; 1523 1524 sa = info->rti_info[RTAX_DST]; 1525 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1526 1527 return (0); 1528 } 1529 1530 /* 1531 * This is the analogue to the rt_newaddrmsg which performs the same 1532 * function but for multicast group memberhips. This is easier since 1533 * there is no route state to worry about. 1534 */ 1535 void 1536 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma) 1537 { 1538 struct rt_addrinfo info; 1539 struct mbuf *m = NULL; 1540 struct ifnet *ifp = ifma->ifma_ifp; 1541 struct ifma_msghdr *ifmam; 1542 1543 if (V_route_cb.any_count == 0) 1544 return; 1545 1546 bzero((caddr_t)&info, sizeof(info)); 1547 info.rti_info[RTAX_IFA] = ifma->ifma_addr; 1548 if (ifp && ifp->if_addr) 1549 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr; 1550 else 1551 info.rti_info[RTAX_IFP] = NULL; 1552 /* 1553 * If a link-layer address is present, present it as a ``gateway'' 1554 * (similarly to how ARP entries, e.g., are presented). 1555 */ 1556 info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr; 1557 m = rtsock_msg_mbuf(cmd, &info); 1558 if (m == NULL) 1559 return; 1560 ifmam = mtod(m, struct ifma_msghdr *); 1561 KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n", 1562 __func__)); 1563 ifmam->ifmam_index = ifp->if_index; 1564 ifmam->ifmam_addrs = info.rti_addrs; 1565 rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC); 1566 } 1567 1568 static struct mbuf * 1569 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what, 1570 struct rt_addrinfo *info) 1571 { 1572 struct if_announcemsghdr *ifan; 1573 struct mbuf *m; 1574 1575 if (V_route_cb.any_count == 0) 1576 return NULL; 1577 bzero((caddr_t)info, sizeof(*info)); 1578 m = rtsock_msg_mbuf(type, info); 1579 if (m != NULL) { 1580 ifan = mtod(m, struct if_announcemsghdr *); 1581 ifan->ifan_index = ifp->if_index; 1582 strlcpy(ifan->ifan_name, ifp->if_xname, 1583 sizeof(ifan->ifan_name)); 1584 ifan->ifan_what = what; 1585 } 1586 return m; 1587 } 1588 1589 /* 1590 * This is called to generate routing socket messages indicating 1591 * IEEE80211 wireless events. 1592 * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way. 1593 */ 1594 void 1595 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len) 1596 { 1597 struct mbuf *m; 1598 struct rt_addrinfo info; 1599 1600 m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info); 1601 if (m != NULL) { 1602 /* 1603 * Append the ieee80211 data. Try to stick it in the 1604 * mbuf containing the ifannounce msg; otherwise allocate 1605 * a new mbuf and append. 1606 * 1607 * NB: we assume m is a single mbuf. 1608 */ 1609 if (data_len > M_TRAILINGSPACE(m)) { 1610 struct mbuf *n = m_get(M_NOWAIT, MT_DATA); 1611 if (n == NULL) { 1612 m_freem(m); 1613 return; 1614 } 1615 bcopy(data, mtod(n, void *), data_len); 1616 n->m_len = data_len; 1617 m->m_next = n; 1618 } else if (data_len > 0) { 1619 bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len); 1620 m->m_len += data_len; 1621 } 1622 if (m->m_flags & M_PKTHDR) 1623 m->m_pkthdr.len += data_len; 1624 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len; 1625 rt_dispatch(m, AF_UNSPEC); 1626 } 1627 } 1628 1629 /* 1630 * This is called to generate routing socket messages indicating 1631 * network interface arrival and departure. 1632 */ 1633 void 1634 rt_ifannouncemsg(struct ifnet *ifp, int what) 1635 { 1636 struct mbuf *m; 1637 struct rt_addrinfo info; 1638 1639 m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info); 1640 if (m != NULL) 1641 rt_dispatch(m, AF_UNSPEC); 1642 } 1643 1644 static void 1645 rt_dispatch(struct mbuf *m, sa_family_t saf) 1646 { 1647 struct m_tag *tag; 1648 1649 /* 1650 * Preserve the family from the sockaddr, if any, in an m_tag for 1651 * use when injecting the mbuf into the routing socket buffer from 1652 * the netisr. 1653 */ 1654 if (saf != AF_UNSPEC) { 1655 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short), 1656 M_NOWAIT); 1657 if (tag == NULL) { 1658 m_freem(m); 1659 return; 1660 } 1661 *(unsigned short *)(tag + 1) = saf; 1662 m_tag_prepend(m, tag); 1663 } 1664 #ifdef VIMAGE 1665 if (V_loif) 1666 m->m_pkthdr.rcvif = V_loif; 1667 else { 1668 m_freem(m); 1669 return; 1670 } 1671 #endif 1672 netisr_queue(NETISR_ROUTE, m); /* mbuf is free'd on failure. */ 1673 } 1674 1675 /* 1676 * Checks if rte can be exported v.r.t jails/vnets. 1677 * 1678 * Returns 1 if it can, 0 otherwise. 1679 */ 1680 static int 1681 can_export_rte(struct ucred *td_ucred, const struct rtentry *rt) 1682 { 1683 1684 if ((rt->rt_flags & RTF_HOST) == 0 1685 ? jailed_without_vnet(td_ucred) 1686 : prison_if(td_ucred, rt_key_const(rt)) != 0) 1687 return (0); 1688 return (1); 1689 } 1690 1691 /* 1692 * This is used in dumping the kernel table via sysctl(). 1693 */ 1694 static int 1695 sysctl_dumpentry(struct radix_node *rn, void *vw) 1696 { 1697 struct walkarg *w = vw; 1698 struct rtentry *rt = (struct rtentry *)rn; 1699 struct nhop_object *nh; 1700 int error = 0, size; 1701 struct rt_addrinfo info; 1702 struct sockaddr_storage ss; 1703 1704 NET_EPOCH_ASSERT(); 1705 1706 if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg)) 1707 return 0; 1708 if (!can_export_rte(w->w_req->td->td_ucred, rt)) 1709 return (0); 1710 bzero((caddr_t)&info, sizeof(info)); 1711 info.rti_info[RTAX_DST] = rt_key(rt); 1712 info.rti_info[RTAX_GATEWAY] = &rt->rt_nhop->gw_sa; 1713 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt), 1714 rt_mask(rt), &ss); 1715 info.rti_info[RTAX_GENMASK] = 0; 1716 nh = rt->rt_nhop; 1717 if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) { 1718 info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr; 1719 info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr; 1720 if (nh->nh_ifp->if_flags & IFF_POINTOPOINT) 1721 info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr; 1722 } 1723 if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0) 1724 return (error); 1725 if (w->w_req && w->w_tmem) { 1726 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem; 1727 1728 bzero(&rtm->rtm_index, 1729 sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index)); 1730 if (rt->rt_flags & RTF_GWFLAG_COMPAT) 1731 rtm->rtm_flags = RTF_GATEWAY | 1732 (rt->rt_flags & ~RTF_GWFLAG_COMPAT); 1733 else 1734 rtm->rtm_flags = rt->rt_flags; 1735 rtm->rtm_flags |= nhop_get_rtflags(nh); 1736 rt_getmetrics(rt, &rtm->rtm_rmx); 1737 rtm->rtm_index = nh->nh_ifp->if_index; 1738 rtm->rtm_addrs = info.rti_addrs; 1739 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size); 1740 return (error); 1741 } 1742 return (error); 1743 } 1744 1745 static int 1746 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd, 1747 struct rt_addrinfo *info, struct walkarg *w, int len) 1748 { 1749 struct if_msghdrl *ifm; 1750 struct if_data *ifd; 1751 1752 ifm = (struct if_msghdrl *)w->w_tmem; 1753 1754 #ifdef COMPAT_FREEBSD32 1755 if (w->w_req->flags & SCTL_MASK32) { 1756 struct if_msghdrl32 *ifm32; 1757 1758 ifm32 = (struct if_msghdrl32 *)ifm; 1759 ifm32->ifm_addrs = info->rti_addrs; 1760 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1761 ifm32->ifm_index = ifp->if_index; 1762 ifm32->_ifm_spare1 = 0; 1763 ifm32->ifm_len = sizeof(*ifm32); 1764 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data); 1765 ifm32->_ifm_spare2 = 0; 1766 ifd = &ifm32->ifm_data; 1767 } else 1768 #endif 1769 { 1770 ifm->ifm_addrs = info->rti_addrs; 1771 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1772 ifm->ifm_index = ifp->if_index; 1773 ifm->_ifm_spare1 = 0; 1774 ifm->ifm_len = sizeof(*ifm); 1775 ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data); 1776 ifm->_ifm_spare2 = 0; 1777 ifd = &ifm->ifm_data; 1778 } 1779 1780 memcpy(ifd, src_ifd, sizeof(*ifd)); 1781 1782 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len)); 1783 } 1784 1785 static int 1786 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd, 1787 struct rt_addrinfo *info, struct walkarg *w, int len) 1788 { 1789 struct if_msghdr *ifm; 1790 struct if_data *ifd; 1791 1792 ifm = (struct if_msghdr *)w->w_tmem; 1793 1794 #ifdef COMPAT_FREEBSD32 1795 if (w->w_req->flags & SCTL_MASK32) { 1796 struct if_msghdr32 *ifm32; 1797 1798 ifm32 = (struct if_msghdr32 *)ifm; 1799 ifm32->ifm_addrs = info->rti_addrs; 1800 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1801 ifm32->ifm_index = ifp->if_index; 1802 ifm32->_ifm_spare1 = 0; 1803 ifd = &ifm32->ifm_data; 1804 } else 1805 #endif 1806 { 1807 ifm->ifm_addrs = info->rti_addrs; 1808 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1809 ifm->ifm_index = ifp->if_index; 1810 ifm->_ifm_spare1 = 0; 1811 ifd = &ifm->ifm_data; 1812 } 1813 1814 memcpy(ifd, src_ifd, sizeof(*ifd)); 1815 1816 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len)); 1817 } 1818 1819 static int 1820 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info, 1821 struct walkarg *w, int len) 1822 { 1823 struct ifa_msghdrl *ifam; 1824 struct if_data *ifd; 1825 1826 ifam = (struct ifa_msghdrl *)w->w_tmem; 1827 1828 #ifdef COMPAT_FREEBSD32 1829 if (w->w_req->flags & SCTL_MASK32) { 1830 struct ifa_msghdrl32 *ifam32; 1831 1832 ifam32 = (struct ifa_msghdrl32 *)ifam; 1833 ifam32->ifam_addrs = info->rti_addrs; 1834 ifam32->ifam_flags = ifa->ifa_flags; 1835 ifam32->ifam_index = ifa->ifa_ifp->if_index; 1836 ifam32->_ifam_spare1 = 0; 1837 ifam32->ifam_len = sizeof(*ifam32); 1838 ifam32->ifam_data_off = 1839 offsetof(struct ifa_msghdrl32, ifam_data); 1840 ifam32->ifam_metric = ifa->ifa_ifp->if_metric; 1841 ifd = &ifam32->ifam_data; 1842 } else 1843 #endif 1844 { 1845 ifam->ifam_addrs = info->rti_addrs; 1846 ifam->ifam_flags = ifa->ifa_flags; 1847 ifam->ifam_index = ifa->ifa_ifp->if_index; 1848 ifam->_ifam_spare1 = 0; 1849 ifam->ifam_len = sizeof(*ifam); 1850 ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data); 1851 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 1852 ifd = &ifam->ifam_data; 1853 } 1854 1855 bzero(ifd, sizeof(*ifd)); 1856 ifd->ifi_datalen = sizeof(struct if_data); 1857 ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets); 1858 ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets); 1859 ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes); 1860 ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes); 1861 1862 /* Fixup if_data carp(4) vhid. */ 1863 if (carp_get_vhid_p != NULL) 1864 ifd->ifi_vhid = (*carp_get_vhid_p)(ifa); 1865 1866 return (SYSCTL_OUT(w->w_req, w->w_tmem, len)); 1867 } 1868 1869 static int 1870 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info, 1871 struct walkarg *w, int len) 1872 { 1873 struct ifa_msghdr *ifam; 1874 1875 ifam = (struct ifa_msghdr *)w->w_tmem; 1876 ifam->ifam_addrs = info->rti_addrs; 1877 ifam->ifam_flags = ifa->ifa_flags; 1878 ifam->ifam_index = ifa->ifa_ifp->if_index; 1879 ifam->_ifam_spare1 = 0; 1880 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 1881 1882 return (SYSCTL_OUT(w->w_req, w->w_tmem, len)); 1883 } 1884 1885 static int 1886 sysctl_iflist(int af, struct walkarg *w) 1887 { 1888 struct ifnet *ifp; 1889 struct ifaddr *ifa; 1890 struct if_data ifd; 1891 struct rt_addrinfo info; 1892 int len, error = 0; 1893 struct sockaddr_storage ss; 1894 1895 bzero((caddr_t)&info, sizeof(info)); 1896 bzero(&ifd, sizeof(ifd)); 1897 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1898 if (w->w_arg && w->w_arg != ifp->if_index) 1899 continue; 1900 if_data_copy(ifp, &ifd); 1901 ifa = ifp->if_addr; 1902 info.rti_info[RTAX_IFP] = ifa->ifa_addr; 1903 error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len); 1904 if (error != 0) 1905 goto done; 1906 info.rti_info[RTAX_IFP] = NULL; 1907 if (w->w_req && w->w_tmem) { 1908 if (w->w_op == NET_RT_IFLISTL) 1909 error = sysctl_iflist_ifml(ifp, &ifd, &info, w, 1910 len); 1911 else 1912 error = sysctl_iflist_ifm(ifp, &ifd, &info, w, 1913 len); 1914 if (error) 1915 goto done; 1916 } 1917 while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) { 1918 if (af && af != ifa->ifa_addr->sa_family) 1919 continue; 1920 if (prison_if(w->w_req->td->td_ucred, 1921 ifa->ifa_addr) != 0) 1922 continue; 1923 info.rti_info[RTAX_IFA] = ifa->ifa_addr; 1924 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask( 1925 ifa->ifa_addr, ifa->ifa_netmask, &ss); 1926 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr; 1927 error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len); 1928 if (error != 0) 1929 goto done; 1930 if (w->w_req && w->w_tmem) { 1931 if (w->w_op == NET_RT_IFLISTL) 1932 error = sysctl_iflist_ifaml(ifa, &info, 1933 w, len); 1934 else 1935 error = sysctl_iflist_ifam(ifa, &info, 1936 w, len); 1937 if (error) 1938 goto done; 1939 } 1940 } 1941 info.rti_info[RTAX_IFA] = NULL; 1942 info.rti_info[RTAX_NETMASK] = NULL; 1943 info.rti_info[RTAX_BRD] = NULL; 1944 } 1945 done: 1946 return (error); 1947 } 1948 1949 static int 1950 sysctl_ifmalist(int af, struct walkarg *w) 1951 { 1952 struct rt_addrinfo info; 1953 struct ifaddr *ifa; 1954 struct ifmultiaddr *ifma; 1955 struct ifnet *ifp; 1956 int error, len; 1957 1958 NET_EPOCH_ASSERT(); 1959 1960 error = 0; 1961 bzero((caddr_t)&info, sizeof(info)); 1962 1963 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1964 if (w->w_arg && w->w_arg != ifp->if_index) 1965 continue; 1966 ifa = ifp->if_addr; 1967 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL; 1968 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 1969 if (af && af != ifma->ifma_addr->sa_family) 1970 continue; 1971 if (prison_if(w->w_req->td->td_ucred, 1972 ifma->ifma_addr) != 0) 1973 continue; 1974 info.rti_info[RTAX_IFA] = ifma->ifma_addr; 1975 info.rti_info[RTAX_GATEWAY] = 1976 (ifma->ifma_addr->sa_family != AF_LINK) ? 1977 ifma->ifma_lladdr : NULL; 1978 error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len); 1979 if (error != 0) 1980 break; 1981 if (w->w_req && w->w_tmem) { 1982 struct ifma_msghdr *ifmam; 1983 1984 ifmam = (struct ifma_msghdr *)w->w_tmem; 1985 ifmam->ifmam_index = ifma->ifma_ifp->if_index; 1986 ifmam->ifmam_flags = 0; 1987 ifmam->ifmam_addrs = info.rti_addrs; 1988 ifmam->_ifmam_spare1 = 0; 1989 error = SYSCTL_OUT(w->w_req, w->w_tmem, len); 1990 if (error != 0) 1991 break; 1992 } 1993 } 1994 if (error != 0) 1995 break; 1996 } 1997 return (error); 1998 } 1999 2000 static int 2001 sysctl_rtsock(SYSCTL_HANDLER_ARGS) 2002 { 2003 RIB_RLOCK_TRACKER; 2004 struct epoch_tracker et; 2005 int *name = (int *)arg1; 2006 u_int namelen = arg2; 2007 struct rib_head *rnh = NULL; /* silence compiler. */ 2008 int i, lim, error = EINVAL; 2009 int fib = 0; 2010 u_char af; 2011 struct walkarg w; 2012 2013 name ++; 2014 namelen--; 2015 if (req->newptr) 2016 return (EPERM); 2017 if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP) { 2018 if (namelen == 3) 2019 fib = req->td->td_proc->p_fibnum; 2020 else if (namelen == 4) 2021 fib = (name[3] == RT_ALL_FIBS) ? 2022 req->td->td_proc->p_fibnum : name[3]; 2023 else 2024 return ((namelen < 3) ? EISDIR : ENOTDIR); 2025 if (fib < 0 || fib >= rt_numfibs) 2026 return (EINVAL); 2027 } else if (namelen != 3) 2028 return ((namelen < 3) ? EISDIR : ENOTDIR); 2029 af = name[0]; 2030 if (af > AF_MAX) 2031 return (EINVAL); 2032 bzero(&w, sizeof(w)); 2033 w.w_op = name[1]; 2034 w.w_arg = name[2]; 2035 w.w_req = req; 2036 2037 error = sysctl_wire_old_buffer(req, 0); 2038 if (error) 2039 return (error); 2040 2041 /* 2042 * Allocate reply buffer in advance. 2043 * All rtsock messages has maximum length of u_short. 2044 */ 2045 w.w_tmemsize = 65536; 2046 w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK); 2047 2048 NET_EPOCH_ENTER(et); 2049 switch (w.w_op) { 2050 case NET_RT_DUMP: 2051 case NET_RT_FLAGS: 2052 if (af == 0) { /* dump all tables */ 2053 i = 1; 2054 lim = AF_MAX; 2055 } else /* dump only one table */ 2056 i = lim = af; 2057 2058 /* 2059 * take care of llinfo entries, the caller must 2060 * specify an AF 2061 */ 2062 if (w.w_op == NET_RT_FLAGS && 2063 (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) { 2064 if (af != 0) 2065 error = lltable_sysctl_dumparp(af, w.w_req); 2066 else 2067 error = EINVAL; 2068 break; 2069 } 2070 /* 2071 * take care of routing entries 2072 */ 2073 for (error = 0; error == 0 && i <= lim; i++) { 2074 rnh = rt_tables_get_rnh(fib, i); 2075 if (rnh != NULL) { 2076 RIB_RLOCK(rnh); 2077 error = rnh->rnh_walktree(&rnh->head, 2078 sysctl_dumpentry, &w); 2079 RIB_RUNLOCK(rnh); 2080 } else if (af != 0) 2081 error = EAFNOSUPPORT; 2082 } 2083 break; 2084 case NET_RT_NHOP: 2085 /* Allow dumping one specific af/fib at a time */ 2086 if (namelen < 4) { 2087 error = EINVAL; 2088 break; 2089 } 2090 fib = name[3]; 2091 if (fib < 0 || fib > rt_numfibs) { 2092 error = EINVAL; 2093 break; 2094 } 2095 rnh = rt_tables_get_rnh(fib, af); 2096 if (rnh == NULL) { 2097 error = EAFNOSUPPORT; 2098 break; 2099 } 2100 if (w.w_op == NET_RT_NHOP) 2101 error = nhops_dump_sysctl(rnh, w.w_req); 2102 break; 2103 case NET_RT_IFLIST: 2104 case NET_RT_IFLISTL: 2105 error = sysctl_iflist(af, &w); 2106 break; 2107 2108 case NET_RT_IFMALIST: 2109 error = sysctl_ifmalist(af, &w); 2110 break; 2111 } 2112 NET_EPOCH_EXIT(et); 2113 2114 free(w.w_tmem, M_TEMP); 2115 return (error); 2116 } 2117 2118 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE, 2119 sysctl_rtsock, "Return route tables and interface/address lists"); 2120 2121 /* 2122 * Definitions of protocols supported in the ROUTE domain. 2123 */ 2124 2125 static struct domain routedomain; /* or at least forward */ 2126 2127 static struct protosw routesw[] = { 2128 { 2129 .pr_type = SOCK_RAW, 2130 .pr_domain = &routedomain, 2131 .pr_flags = PR_ATOMIC|PR_ADDR, 2132 .pr_output = route_output, 2133 .pr_ctlinput = raw_ctlinput, 2134 .pr_init = raw_init, 2135 .pr_usrreqs = &route_usrreqs 2136 } 2137 }; 2138 2139 static struct domain routedomain = { 2140 .dom_family = PF_ROUTE, 2141 .dom_name = "route", 2142 .dom_protosw = routesw, 2143 .dom_protoswNPROTOSW = &routesw[nitems(routesw)] 2144 }; 2145 2146 VNET_DOMAIN_SET(route); 2147 2148