1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1988, 1991, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * @(#)rtsock.c 8.7 (Berkeley) 10/12/95 32 * $FreeBSD$ 33 */ 34 #include "opt_ddb.h" 35 #include "opt_route.h" 36 #include "opt_inet.h" 37 #include "opt_inet6.h" 38 39 #include <sys/param.h> 40 #include <sys/jail.h> 41 #include <sys/kernel.h> 42 #include <sys/domain.h> 43 #include <sys/lock.h> 44 #include <sys/malloc.h> 45 #include <sys/mbuf.h> 46 #include <sys/priv.h> 47 #include <sys/proc.h> 48 #include <sys/protosw.h> 49 #include <sys/rmlock.h> 50 #include <sys/rwlock.h> 51 #include <sys/signalvar.h> 52 #include <sys/socket.h> 53 #include <sys/socketvar.h> 54 #include <sys/sysctl.h> 55 #include <sys/systm.h> 56 57 #include <net/if.h> 58 #include <net/if_var.h> 59 #include <net/if_dl.h> 60 #include <net/if_llatbl.h> 61 #include <net/if_types.h> 62 #include <net/netisr.h> 63 #include <net/raw_cb.h> 64 #include <net/route.h> 65 #include <net/route/route_ctl.h> 66 #include <net/route/route_var.h> 67 #include <net/vnet.h> 68 69 #include <netinet/in.h> 70 #include <netinet/if_ether.h> 71 #include <netinet/ip_carp.h> 72 #ifdef INET6 73 #include <netinet6/ip6_var.h> 74 #include <netinet6/scope6_var.h> 75 #endif 76 #include <net/route/nhop.h> 77 78 #ifdef COMPAT_FREEBSD32 79 #include <sys/mount.h> 80 #include <compat/freebsd32/freebsd32.h> 81 82 struct if_msghdr32 { 83 uint16_t ifm_msglen; 84 uint8_t ifm_version; 85 uint8_t ifm_type; 86 int32_t ifm_addrs; 87 int32_t ifm_flags; 88 uint16_t ifm_index; 89 uint16_t _ifm_spare1; 90 struct if_data ifm_data; 91 }; 92 93 struct if_msghdrl32 { 94 uint16_t ifm_msglen; 95 uint8_t ifm_version; 96 uint8_t ifm_type; 97 int32_t ifm_addrs; 98 int32_t ifm_flags; 99 uint16_t ifm_index; 100 uint16_t _ifm_spare1; 101 uint16_t ifm_len; 102 uint16_t ifm_data_off; 103 uint32_t _ifm_spare2; 104 struct if_data ifm_data; 105 }; 106 107 struct ifa_msghdrl32 { 108 uint16_t ifam_msglen; 109 uint8_t ifam_version; 110 uint8_t ifam_type; 111 int32_t ifam_addrs; 112 int32_t ifam_flags; 113 uint16_t ifam_index; 114 uint16_t _ifam_spare1; 115 uint16_t ifam_len; 116 uint16_t ifam_data_off; 117 int32_t ifam_metric; 118 struct if_data ifam_data; 119 }; 120 121 #define SA_SIZE32(sa) \ 122 ( (((struct sockaddr *)(sa))->sa_len == 0) ? \ 123 sizeof(int) : \ 124 1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) ) 125 126 #endif /* COMPAT_FREEBSD32 */ 127 128 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables"); 129 130 /* NB: these are not modified */ 131 static struct sockaddr route_src = { 2, PF_ROUTE, }; 132 static struct sockaddr sa_zero = { sizeof(sa_zero), AF_INET, }; 133 134 /* These are external hooks for CARP. */ 135 int (*carp_get_vhid_p)(struct ifaddr *); 136 137 /* 138 * Used by rtsock/raw_input callback code to decide whether to filter the update 139 * notification to a socket bound to a particular FIB. 140 */ 141 #define RTS_FILTER_FIB M_PROTO8 142 143 typedef struct { 144 int ip_count; /* attached w/ AF_INET */ 145 int ip6_count; /* attached w/ AF_INET6 */ 146 int any_count; /* total attached */ 147 } route_cb_t; 148 VNET_DEFINE_STATIC(route_cb_t, route_cb); 149 #define V_route_cb VNET(route_cb) 150 151 struct mtx rtsock_mtx; 152 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF); 153 154 #define RTSOCK_LOCK() mtx_lock(&rtsock_mtx) 155 #define RTSOCK_UNLOCK() mtx_unlock(&rtsock_mtx) 156 #define RTSOCK_LOCK_ASSERT() mtx_assert(&rtsock_mtx, MA_OWNED) 157 158 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, ""); 159 160 struct walkarg { 161 int family; 162 int w_tmemsize; 163 int w_op, w_arg; 164 caddr_t w_tmem; 165 struct sysctl_req *w_req; 166 struct sockaddr *dst; 167 struct sockaddr *mask; 168 }; 169 170 static void rts_input(struct mbuf *m); 171 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo); 172 static int rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, 173 struct walkarg *w, int *plen); 174 static int rt_xaddrs(caddr_t cp, caddr_t cplim, 175 struct rt_addrinfo *rtinfo); 176 static int sysctl_dumpentry(struct rtentry *rt, void *vw); 177 static int sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, 178 uint32_t weight, struct walkarg *w); 179 static int sysctl_iflist(int af, struct walkarg *w); 180 static int sysctl_ifmalist(int af, struct walkarg *w); 181 static int route_output(struct mbuf *m, struct socket *so, ...); 182 static void rt_getmetrics(const struct rtentry *rt, 183 const struct nhop_object *nh, struct rt_metrics *out); 184 static void rt_dispatch(struct mbuf *, sa_family_t); 185 static int handle_rtm_get(struct rt_addrinfo *info, u_int fibnum, 186 struct rt_msghdr *rtm, struct rib_cmd_info *rc); 187 static int update_rtm_from_rc(struct rt_addrinfo *info, 188 struct rt_msghdr **prtm, int alloc_len, 189 struct rib_cmd_info *rc, struct nhop_object *nh); 190 static void send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, 191 struct mbuf *m, sa_family_t saf, u_int fibnum, 192 int rtm_errno); 193 static bool can_export_rte(struct ucred *td_ucred, bool rt_is_host, 194 const struct sockaddr *rt_dst); 195 196 static struct netisr_handler rtsock_nh = { 197 .nh_name = "rtsock", 198 .nh_handler = rts_input, 199 .nh_proto = NETISR_ROUTE, 200 .nh_policy = NETISR_POLICY_SOURCE, 201 }; 202 203 static int 204 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS) 205 { 206 int error, qlimit; 207 208 netisr_getqlimit(&rtsock_nh, &qlimit); 209 error = sysctl_handle_int(oidp, &qlimit, 0, req); 210 if (error || !req->newptr) 211 return (error); 212 if (qlimit < 1) 213 return (EINVAL); 214 return (netisr_setqlimit(&rtsock_nh, qlimit)); 215 } 216 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, 217 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 218 0, 0, sysctl_route_netisr_maxqlen, "I", 219 "maximum routing socket dispatch queue length"); 220 221 static void 222 vnet_rts_init(void) 223 { 224 int tmp; 225 226 if (IS_DEFAULT_VNET(curvnet)) { 227 if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp)) 228 rtsock_nh.nh_qlimit = tmp; 229 netisr_register(&rtsock_nh); 230 } 231 #ifdef VIMAGE 232 else 233 netisr_register_vnet(&rtsock_nh); 234 #endif 235 } 236 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 237 vnet_rts_init, 0); 238 239 #ifdef VIMAGE 240 static void 241 vnet_rts_uninit(void) 242 { 243 244 netisr_unregister_vnet(&rtsock_nh); 245 } 246 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 247 vnet_rts_uninit, 0); 248 #endif 249 250 static int 251 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src, 252 struct rawcb *rp) 253 { 254 int fibnum; 255 256 KASSERT(m != NULL, ("%s: m is NULL", __func__)); 257 KASSERT(proto != NULL, ("%s: proto is NULL", __func__)); 258 KASSERT(rp != NULL, ("%s: rp is NULL", __func__)); 259 260 /* No filtering requested. */ 261 if ((m->m_flags & RTS_FILTER_FIB) == 0) 262 return (0); 263 264 /* Check if it is a rts and the fib matches the one of the socket. */ 265 fibnum = M_GETFIB(m); 266 if (proto->sp_family != PF_ROUTE || 267 rp->rcb_socket == NULL || 268 rp->rcb_socket->so_fibnum == fibnum) 269 return (0); 270 271 /* Filtering requested and no match, the socket shall be skipped. */ 272 return (1); 273 } 274 275 static void 276 rts_input(struct mbuf *m) 277 { 278 struct sockproto route_proto; 279 unsigned short *family; 280 struct m_tag *tag; 281 282 route_proto.sp_family = PF_ROUTE; 283 tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL); 284 if (tag != NULL) { 285 family = (unsigned short *)(tag + 1); 286 route_proto.sp_protocol = *family; 287 m_tag_delete(m, tag); 288 } else 289 route_proto.sp_protocol = 0; 290 291 raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb); 292 } 293 294 /* 295 * It really doesn't make any sense at all for this code to share much 296 * with raw_usrreq.c, since its functionality is so restricted. XXX 297 */ 298 static void 299 rts_abort(struct socket *so) 300 { 301 302 raw_usrreqs.pru_abort(so); 303 } 304 305 static void 306 rts_close(struct socket *so) 307 { 308 309 raw_usrreqs.pru_close(so); 310 } 311 312 /* pru_accept is EOPNOTSUPP */ 313 314 static int 315 rts_attach(struct socket *so, int proto, struct thread *td) 316 { 317 struct rawcb *rp; 318 int error; 319 320 KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL")); 321 322 /* XXX */ 323 rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO); 324 325 so->so_pcb = (caddr_t)rp; 326 so->so_fibnum = td->td_proc->p_fibnum; 327 error = raw_attach(so, proto); 328 rp = sotorawcb(so); 329 if (error) { 330 so->so_pcb = NULL; 331 free(rp, M_PCB); 332 return error; 333 } 334 RTSOCK_LOCK(); 335 switch(rp->rcb_proto.sp_protocol) { 336 case AF_INET: 337 V_route_cb.ip_count++; 338 break; 339 case AF_INET6: 340 V_route_cb.ip6_count++; 341 break; 342 } 343 V_route_cb.any_count++; 344 RTSOCK_UNLOCK(); 345 soisconnected(so); 346 so->so_options |= SO_USELOOPBACK; 347 return 0; 348 } 349 350 static int 351 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 352 { 353 354 return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */ 355 } 356 357 static int 358 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 359 { 360 361 return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */ 362 } 363 364 /* pru_connect2 is EOPNOTSUPP */ 365 /* pru_control is EOPNOTSUPP */ 366 367 static void 368 rts_detach(struct socket *so) 369 { 370 struct rawcb *rp = sotorawcb(so); 371 372 KASSERT(rp != NULL, ("rts_detach: rp == NULL")); 373 374 RTSOCK_LOCK(); 375 switch(rp->rcb_proto.sp_protocol) { 376 case AF_INET: 377 V_route_cb.ip_count--; 378 break; 379 case AF_INET6: 380 V_route_cb.ip6_count--; 381 break; 382 } 383 V_route_cb.any_count--; 384 RTSOCK_UNLOCK(); 385 raw_usrreqs.pru_detach(so); 386 } 387 388 static int 389 rts_disconnect(struct socket *so) 390 { 391 392 return (raw_usrreqs.pru_disconnect(so)); 393 } 394 395 /* pru_listen is EOPNOTSUPP */ 396 397 static int 398 rts_peeraddr(struct socket *so, struct sockaddr **nam) 399 { 400 401 return (raw_usrreqs.pru_peeraddr(so, nam)); 402 } 403 404 /* pru_rcvd is EOPNOTSUPP */ 405 /* pru_rcvoob is EOPNOTSUPP */ 406 407 static int 408 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 409 struct mbuf *control, struct thread *td) 410 { 411 412 return (raw_usrreqs.pru_send(so, flags, m, nam, control, td)); 413 } 414 415 /* pru_sense is null */ 416 417 static int 418 rts_shutdown(struct socket *so) 419 { 420 421 return (raw_usrreqs.pru_shutdown(so)); 422 } 423 424 static int 425 rts_sockaddr(struct socket *so, struct sockaddr **nam) 426 { 427 428 return (raw_usrreqs.pru_sockaddr(so, nam)); 429 } 430 431 static struct pr_usrreqs route_usrreqs = { 432 .pru_abort = rts_abort, 433 .pru_attach = rts_attach, 434 .pru_bind = rts_bind, 435 .pru_connect = rts_connect, 436 .pru_detach = rts_detach, 437 .pru_disconnect = rts_disconnect, 438 .pru_peeraddr = rts_peeraddr, 439 .pru_send = rts_send, 440 .pru_shutdown = rts_shutdown, 441 .pru_sockaddr = rts_sockaddr, 442 .pru_close = rts_close, 443 }; 444 445 #ifndef _SOCKADDR_UNION_DEFINED 446 #define _SOCKADDR_UNION_DEFINED 447 /* 448 * The union of all possible address formats we handle. 449 */ 450 union sockaddr_union { 451 struct sockaddr sa; 452 struct sockaddr_in sin; 453 struct sockaddr_in6 sin6; 454 }; 455 #endif /* _SOCKADDR_UNION_DEFINED */ 456 457 static int 458 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp, 459 struct nhop_object *nh, union sockaddr_union *saun, struct ucred *cred) 460 { 461 #if defined(INET) || defined(INET6) 462 struct epoch_tracker et; 463 #endif 464 465 /* First, see if the returned address is part of the jail. */ 466 if (prison_if(cred, nh->nh_ifa->ifa_addr) == 0) { 467 info->rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr; 468 return (0); 469 } 470 471 switch (info->rti_info[RTAX_DST]->sa_family) { 472 #ifdef INET 473 case AF_INET: 474 { 475 struct in_addr ia; 476 struct ifaddr *ifa; 477 int found; 478 479 found = 0; 480 /* 481 * Try to find an address on the given outgoing interface 482 * that belongs to the jail. 483 */ 484 NET_EPOCH_ENTER(et); 485 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 486 struct sockaddr *sa; 487 sa = ifa->ifa_addr; 488 if (sa->sa_family != AF_INET) 489 continue; 490 ia = ((struct sockaddr_in *)sa)->sin_addr; 491 if (prison_check_ip4(cred, &ia) == 0) { 492 found = 1; 493 break; 494 } 495 } 496 NET_EPOCH_EXIT(et); 497 if (!found) { 498 /* 499 * As a last resort return the 'default' jail address. 500 */ 501 ia = ((struct sockaddr_in *)nh->nh_ifa->ifa_addr)-> 502 sin_addr; 503 if (prison_get_ip4(cred, &ia) != 0) 504 return (ESRCH); 505 } 506 bzero(&saun->sin, sizeof(struct sockaddr_in)); 507 saun->sin.sin_len = sizeof(struct sockaddr_in); 508 saun->sin.sin_family = AF_INET; 509 saun->sin.sin_addr.s_addr = ia.s_addr; 510 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin; 511 break; 512 } 513 #endif 514 #ifdef INET6 515 case AF_INET6: 516 { 517 struct in6_addr ia6; 518 struct ifaddr *ifa; 519 int found; 520 521 found = 0; 522 /* 523 * Try to find an address on the given outgoing interface 524 * that belongs to the jail. 525 */ 526 NET_EPOCH_ENTER(et); 527 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 528 struct sockaddr *sa; 529 sa = ifa->ifa_addr; 530 if (sa->sa_family != AF_INET6) 531 continue; 532 bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr, 533 &ia6, sizeof(struct in6_addr)); 534 if (prison_check_ip6(cred, &ia6) == 0) { 535 found = 1; 536 break; 537 } 538 } 539 NET_EPOCH_EXIT(et); 540 if (!found) { 541 /* 542 * As a last resort return the 'default' jail address. 543 */ 544 ia6 = ((struct sockaddr_in6 *)nh->nh_ifa->ifa_addr)-> 545 sin6_addr; 546 if (prison_get_ip6(cred, &ia6) != 0) 547 return (ESRCH); 548 } 549 bzero(&saun->sin6, sizeof(struct sockaddr_in6)); 550 saun->sin6.sin6_len = sizeof(struct sockaddr_in6); 551 saun->sin6.sin6_family = AF_INET6; 552 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr)); 553 if (sa6_recoverscope(&saun->sin6) != 0) 554 return (ESRCH); 555 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6; 556 break; 557 } 558 #endif 559 default: 560 return (ESRCH); 561 } 562 return (0); 563 } 564 565 /* 566 * Fills in @info based on userland-provided @rtm message. 567 * 568 * Returns 0 on success. 569 */ 570 static int 571 fill_addrinfo(struct rt_msghdr *rtm, int len, u_int fibnum, struct rt_addrinfo *info) 572 { 573 int error; 574 sa_family_t saf; 575 576 rtm->rtm_pid = curproc->p_pid; 577 info->rti_addrs = rtm->rtm_addrs; 578 579 info->rti_mflags = rtm->rtm_inits; 580 info->rti_rmx = &rtm->rtm_rmx; 581 582 /* 583 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6 584 * link-local address because rtrequest requires addresses with 585 * embedded scope id. 586 */ 587 if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info)) 588 return (EINVAL); 589 590 if (rtm->rtm_flags & RTF_RNH_LOCKED) 591 return (EINVAL); 592 info->rti_flags = rtm->rtm_flags; 593 if (info->rti_info[RTAX_DST] == NULL || 594 info->rti_info[RTAX_DST]->sa_family >= AF_MAX || 595 (info->rti_info[RTAX_GATEWAY] != NULL && 596 info->rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) 597 return (EINVAL); 598 saf = info->rti_info[RTAX_DST]->sa_family; 599 /* 600 * Verify that the caller has the appropriate privilege; RTM_GET 601 * is the only operation the non-superuser is allowed. 602 */ 603 if (rtm->rtm_type != RTM_GET) { 604 error = priv_check(curthread, PRIV_NET_ROUTE); 605 if (error != 0) 606 return (error); 607 } 608 609 /* 610 * The given gateway address may be an interface address. 611 * For example, issuing a "route change" command on a route 612 * entry that was created from a tunnel, and the gateway 613 * address given is the local end point. In this case the 614 * RTF_GATEWAY flag must be cleared or the destination will 615 * not be reachable even though there is no error message. 616 */ 617 if (info->rti_info[RTAX_GATEWAY] != NULL && 618 info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) { 619 struct rt_addrinfo ginfo; 620 struct sockaddr *gdst; 621 struct sockaddr_storage ss; 622 623 bzero(&ginfo, sizeof(ginfo)); 624 bzero(&ss, sizeof(ss)); 625 ss.ss_len = sizeof(ss); 626 627 ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss; 628 gdst = info->rti_info[RTAX_GATEWAY]; 629 630 /* 631 * A host route through the loopback interface is 632 * installed for each interface adddress. In pre 8.0 633 * releases the interface address of a PPP link type 634 * is not reachable locally. This behavior is fixed as 635 * part of the new L2/L3 redesign and rewrite work. The 636 * signature of this interface address route is the 637 * AF_LINK sa_family type of the gateway, and the 638 * rt_ifp has the IFF_LOOPBACK flag set. 639 */ 640 if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) { 641 if (ss.ss_family == AF_LINK && 642 ginfo.rti_ifp->if_flags & IFF_LOOPBACK) { 643 info->rti_flags &= ~RTF_GATEWAY; 644 info->rti_flags |= RTF_GWFLAG_COMPAT; 645 } 646 rib_free_info(&ginfo); 647 } 648 } 649 650 return (0); 651 } 652 653 static struct nhop_object * 654 select_nhop(struct nhop_object *nh, const struct sockaddr *gw) 655 { 656 if (!NH_IS_NHGRP(nh)) 657 return (nh); 658 #ifdef ROUTE_MPATH 659 struct weightened_nhop *wn; 660 uint32_t num_nhops; 661 wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops); 662 if (gw == NULL) 663 return (wn[0].nh); 664 for (int i = 0; i < num_nhops; i++) { 665 if (match_nhop_gw(wn[i].nh, gw)) 666 return (wn[i].nh); 667 } 668 #endif 669 return (NULL); 670 } 671 672 /* 673 * Handles RTM_GET message from routing socket, returning matching rt. 674 * 675 * Returns: 676 * 0 on success, with locked and referenced matching rt in @rt_nrt 677 * errno of failure 678 */ 679 static int 680 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum, 681 struct rt_msghdr *rtm, struct rib_cmd_info *rc) 682 { 683 RIB_RLOCK_TRACKER; 684 struct rib_head *rnh; 685 struct nhop_object *nh; 686 sa_family_t saf; 687 688 saf = info->rti_info[RTAX_DST]->sa_family; 689 690 rnh = rt_tables_get_rnh(fibnum, saf); 691 if (rnh == NULL) 692 return (EAFNOSUPPORT); 693 694 RIB_RLOCK(rnh); 695 696 if (info->rti_info[RTAX_NETMASK] == NULL) { 697 /* 698 * Provide longest prefix match for 699 * address lookup (no mask). 700 * 'route -n get addr' 701 */ 702 rc->rc_rt = (struct rtentry *) rnh->rnh_matchaddr( 703 info->rti_info[RTAX_DST], &rnh->head); 704 } else 705 rc->rc_rt = (struct rtentry *) rnh->rnh_lookup( 706 info->rti_info[RTAX_DST], 707 info->rti_info[RTAX_NETMASK], &rnh->head); 708 709 if (rc->rc_rt == NULL) { 710 RIB_RUNLOCK(rnh); 711 return (ESRCH); 712 } 713 714 nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]); 715 if (nh == NULL) { 716 RIB_RUNLOCK(rnh); 717 return (ESRCH); 718 } 719 /* 720 * If performing proxied L2 entry insertion, and 721 * the actual PPP host entry is found, perform 722 * another search to retrieve the prefix route of 723 * the local end point of the PPP link. 724 * TODO: move this logic to userland. 725 */ 726 if (rtm->rtm_flags & RTF_ANNOUNCE) { 727 struct sockaddr laddr; 728 729 if (nh->nh_ifp != NULL && 730 nh->nh_ifp->if_type == IFT_PROPVIRTUAL) { 731 struct ifaddr *ifa; 732 733 ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1, 734 RT_ALL_FIBS); 735 if (ifa != NULL) 736 rt_maskedcopy(ifa->ifa_addr, 737 &laddr, 738 ifa->ifa_netmask); 739 } else 740 rt_maskedcopy(nh->nh_ifa->ifa_addr, 741 &laddr, 742 nh->nh_ifa->ifa_netmask); 743 /* 744 * refactor rt and no lock operation necessary 745 */ 746 rc->rc_rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr, 747 &rnh->head); 748 if (rc->rc_rt == NULL) { 749 RIB_RUNLOCK(rnh); 750 return (ESRCH); 751 } 752 nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]); 753 if (nh == NULL) { 754 RIB_RUNLOCK(rnh); 755 return (ESRCH); 756 } 757 } 758 rc->rc_nh_new = nh; 759 rc->rc_nh_weight = rc->rc_rt->rt_weight; 760 RIB_RUNLOCK(rnh); 761 762 return (0); 763 } 764 765 static void 766 init_sockaddrs_family(int family, struct sockaddr *dst, struct sockaddr *mask) 767 { 768 #ifdef INET 769 if (family == AF_INET) { 770 struct sockaddr_in *dst4 = (struct sockaddr_in *)dst; 771 struct sockaddr_in *mask4 = (struct sockaddr_in *)mask; 772 773 bzero(dst4, sizeof(struct sockaddr_in)); 774 bzero(mask4, sizeof(struct sockaddr_in)); 775 776 dst4->sin_family = AF_INET; 777 dst4->sin_len = sizeof(struct sockaddr_in); 778 mask4->sin_family = AF_INET; 779 mask4->sin_len = sizeof(struct sockaddr_in); 780 } 781 #endif 782 #ifdef INET6 783 if (family == AF_INET6) { 784 struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst; 785 struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask; 786 787 bzero(dst6, sizeof(struct sockaddr_in6)); 788 bzero(mask6, sizeof(struct sockaddr_in6)); 789 790 dst6->sin6_family = AF_INET6; 791 dst6->sin6_len = sizeof(struct sockaddr_in6); 792 mask6->sin6_family = AF_INET6; 793 mask6->sin6_len = sizeof(struct sockaddr_in6); 794 } 795 #endif 796 } 797 798 static void 799 export_rtaddrs(const struct rtentry *rt, struct sockaddr *dst, 800 struct sockaddr *mask) 801 { 802 #ifdef INET 803 if (dst->sa_family == AF_INET) { 804 struct sockaddr_in *dst4 = (struct sockaddr_in *)dst; 805 struct sockaddr_in *mask4 = (struct sockaddr_in *)mask; 806 uint32_t scopeid = 0; 807 rt_get_inet_prefix_pmask(rt, &dst4->sin_addr, &mask4->sin_addr, 808 &scopeid); 809 return; 810 } 811 #endif 812 #ifdef INET6 813 if (dst->sa_family == AF_INET6) { 814 struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst; 815 struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask; 816 uint32_t scopeid = 0; 817 rt_get_inet6_prefix_pmask(rt, &dst6->sin6_addr, 818 &mask6->sin6_addr, &scopeid); 819 dst6->sin6_scope_id = scopeid; 820 return; 821 } 822 #endif 823 } 824 825 826 /* 827 * Update sockaddrs, flags, etc in @prtm based on @rc data. 828 * rtm can be reallocated. 829 * 830 * Returns 0 on success, along with pointer to (potentially reallocated) 831 * rtm. 832 * 833 */ 834 static int 835 update_rtm_from_rc(struct rt_addrinfo *info, struct rt_msghdr **prtm, 836 int alloc_len, struct rib_cmd_info *rc, struct nhop_object *nh) 837 { 838 struct walkarg w; 839 union sockaddr_union saun; 840 struct rt_msghdr *rtm, *orig_rtm = NULL; 841 struct ifnet *ifp; 842 int error, len; 843 844 rtm = *prtm; 845 union sockaddr_union sa_dst, sa_mask; 846 int family = info->rti_info[RTAX_DST]->sa_family; 847 init_sockaddrs_family(family, &sa_dst.sa, &sa_mask.sa); 848 export_rtaddrs(rc->rc_rt, &sa_dst.sa, &sa_mask.sa); 849 850 info->rti_info[RTAX_DST] = &sa_dst.sa; 851 info->rti_info[RTAX_NETMASK] = rt_is_host(rc->rc_rt) ? NULL : &sa_mask.sa; 852 info->rti_info[RTAX_GATEWAY] = &nh->gw_sa; 853 info->rti_info[RTAX_GENMASK] = 0; 854 ifp = nh->nh_ifp; 855 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) { 856 if (ifp) { 857 info->rti_info[RTAX_IFP] = 858 ifp->if_addr->ifa_addr; 859 error = rtm_get_jailed(info, ifp, nh, 860 &saun, curthread->td_ucred); 861 if (error != 0) 862 return (error); 863 if (ifp->if_flags & IFF_POINTOPOINT) 864 info->rti_info[RTAX_BRD] = 865 nh->nh_ifa->ifa_dstaddr; 866 rtm->rtm_index = ifp->if_index; 867 } else { 868 info->rti_info[RTAX_IFP] = NULL; 869 info->rti_info[RTAX_IFA] = NULL; 870 } 871 } else if (ifp != NULL) 872 rtm->rtm_index = ifp->if_index; 873 874 /* Check if we need to realloc storage */ 875 rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len); 876 if (len > alloc_len) { 877 struct rt_msghdr *tmp_rtm; 878 879 tmp_rtm = malloc(len, M_TEMP, M_NOWAIT); 880 if (tmp_rtm == NULL) 881 return (ENOBUFS); 882 bcopy(rtm, tmp_rtm, rtm->rtm_msglen); 883 orig_rtm = rtm; 884 rtm = tmp_rtm; 885 alloc_len = len; 886 887 /* 888 * Delay freeing original rtm as info contains 889 * data referencing it. 890 */ 891 } 892 893 w.w_tmem = (caddr_t)rtm; 894 w.w_tmemsize = alloc_len; 895 rtsock_msg_buffer(rtm->rtm_type, info, &w, &len); 896 897 rtm->rtm_flags = rc->rc_rt->rte_flags | nhop_get_rtflags(nh); 898 if (rtm->rtm_flags & RTF_GWFLAG_COMPAT) 899 rtm->rtm_flags = RTF_GATEWAY | 900 (rtm->rtm_flags & ~RTF_GWFLAG_COMPAT); 901 rt_getmetrics(rc->rc_rt, nh, &rtm->rtm_rmx); 902 rtm->rtm_rmx.rmx_weight = rc->rc_nh_weight; 903 rtm->rtm_addrs = info->rti_addrs; 904 905 if (orig_rtm != NULL) 906 free(orig_rtm, M_TEMP); 907 *prtm = rtm; 908 909 return (0); 910 } 911 912 #ifdef ROUTE_MPATH 913 static void 914 save_del_notification(struct rib_cmd_info *rc, void *_cbdata) 915 { 916 struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata; 917 918 if (rc->rc_cmd == RTM_DELETE) 919 *rc_new = *rc; 920 } 921 922 static void 923 save_add_notification(struct rib_cmd_info *rc, void *_cbdata) 924 { 925 struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata; 926 927 if (rc->rc_cmd == RTM_ADD) 928 *rc_new = *rc; 929 } 930 #endif 931 932 /*ARGSUSED*/ 933 static int 934 route_output(struct mbuf *m, struct socket *so, ...) 935 { 936 struct rt_msghdr *rtm = NULL; 937 struct rtentry *rt = NULL; 938 struct rt_addrinfo info; 939 struct epoch_tracker et; 940 #ifdef INET6 941 struct sockaddr_storage ss; 942 struct sockaddr_in6 *sin6; 943 int i, rti_need_deembed = 0; 944 #endif 945 int alloc_len = 0, len, error = 0, fibnum; 946 sa_family_t saf = AF_UNSPEC; 947 struct walkarg w; 948 struct rib_cmd_info rc; 949 struct nhop_object *nh; 950 951 fibnum = so->so_fibnum; 952 #define senderr(e) { error = e; goto flush;} 953 if (m == NULL || ((m->m_len < sizeof(long)) && 954 (m = m_pullup(m, sizeof(long))) == NULL)) 955 return (ENOBUFS); 956 if ((m->m_flags & M_PKTHDR) == 0) 957 panic("route_output"); 958 NET_EPOCH_ENTER(et); 959 len = m->m_pkthdr.len; 960 if (len < sizeof(*rtm) || 961 len != mtod(m, struct rt_msghdr *)->rtm_msglen) 962 senderr(EINVAL); 963 964 /* 965 * Most of current messages are in range 200-240 bytes, 966 * minimize possible re-allocation on reply using larger size 967 * buffer aligned on 1k boundaty. 968 */ 969 alloc_len = roundup2(len, 1024); 970 if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL) 971 senderr(ENOBUFS); 972 973 m_copydata(m, 0, len, (caddr_t)rtm); 974 bzero(&info, sizeof(info)); 975 bzero(&w, sizeof(w)); 976 nh = NULL; 977 978 if (rtm->rtm_version != RTM_VERSION) { 979 /* Do not touch message since format is unknown */ 980 free(rtm, M_TEMP); 981 rtm = NULL; 982 senderr(EPROTONOSUPPORT); 983 } 984 985 /* 986 * Starting from here, it is possible 987 * to alter original message and insert 988 * caller PID and error value. 989 */ 990 991 if ((error = fill_addrinfo(rtm, len, fibnum, &info)) != 0) { 992 senderr(error); 993 } 994 995 saf = info.rti_info[RTAX_DST]->sa_family; 996 997 /* support for new ARP code */ 998 if (rtm->rtm_flags & RTF_LLDATA) { 999 error = lla_rt_output(rtm, &info); 1000 #ifdef INET6 1001 if (error == 0) 1002 rti_need_deembed = 1; 1003 #endif 1004 goto flush; 1005 } 1006 1007 switch (rtm->rtm_type) { 1008 case RTM_ADD: 1009 case RTM_CHANGE: 1010 if (rtm->rtm_type == RTM_ADD) { 1011 if (info.rti_info[RTAX_GATEWAY] == NULL) 1012 senderr(EINVAL); 1013 } 1014 error = rib_action(fibnum, rtm->rtm_type, &info, &rc); 1015 if (error == 0) { 1016 #ifdef INET6 1017 rti_need_deembed = 1; 1018 #endif 1019 #ifdef ROUTE_MPATH 1020 if (NH_IS_NHGRP(rc.rc_nh_new) || 1021 (rc.rc_nh_old && NH_IS_NHGRP(rc.rc_nh_old))) { 1022 struct rib_cmd_info rc_simple = {}; 1023 rib_decompose_notification(&rc, 1024 save_add_notification, (void *)&rc_simple); 1025 rc = rc_simple; 1026 } 1027 #endif 1028 nh = rc.rc_nh_new; 1029 rtm->rtm_index = nh->nh_ifp->if_index; 1030 rtm->rtm_flags = rc.rc_rt->rte_flags | nhop_get_rtflags(nh); 1031 } 1032 break; 1033 1034 case RTM_DELETE: 1035 error = rib_action(fibnum, RTM_DELETE, &info, &rc); 1036 if (error == 0) { 1037 #ifdef ROUTE_MPATH 1038 if (NH_IS_NHGRP(rc.rc_nh_old) || 1039 (rc.rc_nh_new && NH_IS_NHGRP(rc.rc_nh_new))) { 1040 struct rib_cmd_info rc_simple = {}; 1041 rib_decompose_notification(&rc, 1042 save_del_notification, (void *)&rc_simple); 1043 rc = rc_simple; 1044 } 1045 #endif 1046 nh = rc.rc_nh_old; 1047 goto report; 1048 } 1049 #ifdef INET6 1050 /* rt_msg2() will not be used when RTM_DELETE fails. */ 1051 rti_need_deembed = 1; 1052 #endif 1053 break; 1054 1055 case RTM_GET: 1056 error = handle_rtm_get(&info, fibnum, rtm, &rc); 1057 if (error != 0) 1058 senderr(error); 1059 nh = rc.rc_nh_new; 1060 1061 report: 1062 if (!can_export_rte(curthread->td_ucred, 1063 info.rti_info[RTAX_NETMASK] == NULL, 1064 info.rti_info[RTAX_DST])) { 1065 senderr(ESRCH); 1066 } 1067 1068 error = update_rtm_from_rc(&info, &rtm, alloc_len, &rc, nh); 1069 /* 1070 * Note that some sockaddr pointers may have changed to 1071 * point to memory outsize @rtm. Some may be pointing 1072 * to the on-stack variables. 1073 * Given that, any pointer in @info CANNOT BE USED. 1074 */ 1075 1076 /* 1077 * scopeid deembedding has been performed while 1078 * writing updated rtm in rtsock_msg_buffer(). 1079 * With that in mind, skip deembedding procedure below. 1080 */ 1081 #ifdef INET6 1082 rti_need_deembed = 0; 1083 #endif 1084 if (error != 0) 1085 senderr(error); 1086 break; 1087 1088 default: 1089 senderr(EOPNOTSUPP); 1090 } 1091 1092 flush: 1093 NET_EPOCH_EXIT(et); 1094 rt = NULL; 1095 1096 #ifdef INET6 1097 if (rtm != NULL) { 1098 if (rti_need_deembed) { 1099 /* sin6_scope_id is recovered before sending rtm. */ 1100 sin6 = (struct sockaddr_in6 *)&ss; 1101 for (i = 0; i < RTAX_MAX; i++) { 1102 if (info.rti_info[i] == NULL) 1103 continue; 1104 if (info.rti_info[i]->sa_family != AF_INET6) 1105 continue; 1106 bcopy(info.rti_info[i], sin6, sizeof(*sin6)); 1107 if (sa6_recoverscope(sin6) == 0) 1108 bcopy(sin6, info.rti_info[i], 1109 sizeof(*sin6)); 1110 } 1111 } 1112 } 1113 #endif 1114 send_rtm_reply(so, rtm, m, saf, fibnum, error); 1115 1116 return (error); 1117 } 1118 1119 /* 1120 * Sends the prepared reply message in @rtm to all rtsock clients. 1121 * Frees @m and @rtm. 1122 * 1123 */ 1124 static void 1125 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m, 1126 sa_family_t saf, u_int fibnum, int rtm_errno) 1127 { 1128 struct rawcb *rp = NULL; 1129 1130 /* 1131 * Check to see if we don't want our own messages. 1132 */ 1133 if ((so->so_options & SO_USELOOPBACK) == 0) { 1134 if (V_route_cb.any_count <= 1) { 1135 if (rtm != NULL) 1136 free(rtm, M_TEMP); 1137 m_freem(m); 1138 return; 1139 } 1140 /* There is another listener, so construct message */ 1141 rp = sotorawcb(so); 1142 } 1143 1144 if (rtm != NULL) { 1145 if (rtm_errno!= 0) 1146 rtm->rtm_errno = rtm_errno; 1147 else 1148 rtm->rtm_flags |= RTF_DONE; 1149 1150 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm); 1151 if (m->m_pkthdr.len < rtm->rtm_msglen) { 1152 m_freem(m); 1153 m = NULL; 1154 } else if (m->m_pkthdr.len > rtm->rtm_msglen) 1155 m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len); 1156 1157 free(rtm, M_TEMP); 1158 } 1159 if (m != NULL) { 1160 M_SETFIB(m, fibnum); 1161 m->m_flags |= RTS_FILTER_FIB; 1162 if (rp) { 1163 /* 1164 * XXX insure we don't get a copy by 1165 * invalidating our protocol 1166 */ 1167 unsigned short family = rp->rcb_proto.sp_family; 1168 rp->rcb_proto.sp_family = 0; 1169 rt_dispatch(m, saf); 1170 rp->rcb_proto.sp_family = family; 1171 } else 1172 rt_dispatch(m, saf); 1173 } 1174 } 1175 1176 static void 1177 rt_getmetrics(const struct rtentry *rt, const struct nhop_object *nh, 1178 struct rt_metrics *out) 1179 { 1180 1181 bzero(out, sizeof(*out)); 1182 out->rmx_mtu = nh->nh_mtu; 1183 out->rmx_weight = rt->rt_weight; 1184 out->rmx_nhidx = nhop_get_idx(nh); 1185 /* Kernel -> userland timebase conversion. */ 1186 out->rmx_expire = rt->rt_expire ? 1187 rt->rt_expire - time_uptime + time_second : 0; 1188 } 1189 1190 /* 1191 * Extract the addresses of the passed sockaddrs. 1192 * Do a little sanity checking so as to avoid bad memory references. 1193 * This data is derived straight from userland. 1194 */ 1195 static int 1196 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo) 1197 { 1198 struct sockaddr *sa; 1199 int i; 1200 1201 for (i = 0; i < RTAX_MAX && cp < cplim; i++) { 1202 if ((rtinfo->rti_addrs & (1 << i)) == 0) 1203 continue; 1204 sa = (struct sockaddr *)cp; 1205 /* 1206 * It won't fit. 1207 */ 1208 if (cp + sa->sa_len > cplim) 1209 return (EINVAL); 1210 /* 1211 * there are no more.. quit now 1212 * If there are more bits, they are in error. 1213 * I've seen this. route(1) can evidently generate these. 1214 * This causes kernel to core dump. 1215 * for compatibility, If we see this, point to a safe address. 1216 */ 1217 if (sa->sa_len == 0) { 1218 rtinfo->rti_info[i] = &sa_zero; 1219 return (0); /* should be EINVAL but for compat */ 1220 } 1221 /* accept it */ 1222 #ifdef INET6 1223 if (sa->sa_family == AF_INET6) 1224 sa6_embedscope((struct sockaddr_in6 *)sa, 1225 V_ip6_use_defzone); 1226 #endif 1227 rtinfo->rti_info[i] = sa; 1228 cp += SA_SIZE(sa); 1229 } 1230 return (0); 1231 } 1232 1233 /* 1234 * Fill in @dmask with valid netmask leaving original @smask 1235 * intact. Mostly used with radix netmasks. 1236 */ 1237 struct sockaddr * 1238 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask, 1239 struct sockaddr_storage *dmask) 1240 { 1241 if (dst == NULL || smask == NULL) 1242 return (NULL); 1243 1244 memset(dmask, 0, dst->sa_len); 1245 memcpy(dmask, smask, smask->sa_len); 1246 dmask->ss_len = dst->sa_len; 1247 dmask->ss_family = dst->sa_family; 1248 1249 return ((struct sockaddr *)dmask); 1250 } 1251 1252 /* 1253 * Writes information related to @rtinfo object to newly-allocated mbuf. 1254 * Assumes MCLBYTES is enough to construct any message. 1255 * Used for OS notifications of vaious events (if/ifa announces,etc) 1256 * 1257 * Returns allocated mbuf or NULL on failure. 1258 */ 1259 static struct mbuf * 1260 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo) 1261 { 1262 struct sockaddr_storage ss; 1263 struct rt_msghdr *rtm; 1264 struct mbuf *m; 1265 int i; 1266 struct sockaddr *sa; 1267 #ifdef INET6 1268 struct sockaddr_in6 *sin6; 1269 #endif 1270 int len, dlen; 1271 1272 switch (type) { 1273 case RTM_DELADDR: 1274 case RTM_NEWADDR: 1275 len = sizeof(struct ifa_msghdr); 1276 break; 1277 1278 case RTM_DELMADDR: 1279 case RTM_NEWMADDR: 1280 len = sizeof(struct ifma_msghdr); 1281 break; 1282 1283 case RTM_IFINFO: 1284 len = sizeof(struct if_msghdr); 1285 break; 1286 1287 case RTM_IFANNOUNCE: 1288 case RTM_IEEE80211: 1289 len = sizeof(struct if_announcemsghdr); 1290 break; 1291 1292 default: 1293 len = sizeof(struct rt_msghdr); 1294 } 1295 1296 /* XXXGL: can we use MJUMPAGESIZE cluster here? */ 1297 KASSERT(len <= MCLBYTES, ("%s: message too big", __func__)); 1298 if (len > MHLEN) 1299 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 1300 else 1301 m = m_gethdr(M_NOWAIT, MT_DATA); 1302 if (m == NULL) 1303 return (m); 1304 1305 m->m_pkthdr.len = m->m_len = len; 1306 rtm = mtod(m, struct rt_msghdr *); 1307 bzero((caddr_t)rtm, len); 1308 for (i = 0; i < RTAX_MAX; i++) { 1309 if ((sa = rtinfo->rti_info[i]) == NULL) 1310 continue; 1311 rtinfo->rti_addrs |= (1 << i); 1312 1313 dlen = SA_SIZE(sa); 1314 KASSERT(dlen <= sizeof(ss), 1315 ("%s: sockaddr size overflow", __func__)); 1316 bzero(&ss, sizeof(ss)); 1317 bcopy(sa, &ss, sa->sa_len); 1318 sa = (struct sockaddr *)&ss; 1319 #ifdef INET6 1320 if (sa->sa_family == AF_INET6) { 1321 sin6 = (struct sockaddr_in6 *)sa; 1322 (void)sa6_recoverscope(sin6); 1323 } 1324 #endif 1325 m_copyback(m, len, dlen, (caddr_t)sa); 1326 len += dlen; 1327 } 1328 if (m->m_pkthdr.len != len) { 1329 m_freem(m); 1330 return (NULL); 1331 } 1332 rtm->rtm_msglen = len; 1333 rtm->rtm_version = RTM_VERSION; 1334 rtm->rtm_type = type; 1335 return (m); 1336 } 1337 1338 /* 1339 * Writes information related to @rtinfo object to preallocated buffer. 1340 * Stores needed size in @plen. If @w is NULL, calculates size without 1341 * writing. 1342 * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation. 1343 * 1344 * Returns 0 on success. 1345 * 1346 */ 1347 static int 1348 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen) 1349 { 1350 struct sockaddr_storage ss; 1351 int len, buflen = 0, dlen, i; 1352 caddr_t cp = NULL; 1353 struct rt_msghdr *rtm = NULL; 1354 #ifdef INET6 1355 struct sockaddr_in6 *sin6; 1356 #endif 1357 #ifdef COMPAT_FREEBSD32 1358 bool compat32 = false; 1359 #endif 1360 1361 switch (type) { 1362 case RTM_DELADDR: 1363 case RTM_NEWADDR: 1364 if (w != NULL && w->w_op == NET_RT_IFLISTL) { 1365 #ifdef COMPAT_FREEBSD32 1366 if (w->w_req->flags & SCTL_MASK32) { 1367 len = sizeof(struct ifa_msghdrl32); 1368 compat32 = true; 1369 } else 1370 #endif 1371 len = sizeof(struct ifa_msghdrl); 1372 } else 1373 len = sizeof(struct ifa_msghdr); 1374 break; 1375 1376 case RTM_IFINFO: 1377 #ifdef COMPAT_FREEBSD32 1378 if (w != NULL && w->w_req->flags & SCTL_MASK32) { 1379 if (w->w_op == NET_RT_IFLISTL) 1380 len = sizeof(struct if_msghdrl32); 1381 else 1382 len = sizeof(struct if_msghdr32); 1383 compat32 = true; 1384 break; 1385 } 1386 #endif 1387 if (w != NULL && w->w_op == NET_RT_IFLISTL) 1388 len = sizeof(struct if_msghdrl); 1389 else 1390 len = sizeof(struct if_msghdr); 1391 break; 1392 1393 case RTM_NEWMADDR: 1394 len = sizeof(struct ifma_msghdr); 1395 break; 1396 1397 default: 1398 len = sizeof(struct rt_msghdr); 1399 } 1400 1401 if (w != NULL) { 1402 rtm = (struct rt_msghdr *)w->w_tmem; 1403 buflen = w->w_tmemsize - len; 1404 cp = (caddr_t)w->w_tmem + len; 1405 } 1406 1407 rtinfo->rti_addrs = 0; 1408 for (i = 0; i < RTAX_MAX; i++) { 1409 struct sockaddr *sa; 1410 1411 if ((sa = rtinfo->rti_info[i]) == NULL) 1412 continue; 1413 rtinfo->rti_addrs |= (1 << i); 1414 #ifdef COMPAT_FREEBSD32 1415 if (compat32) 1416 dlen = SA_SIZE32(sa); 1417 else 1418 #endif 1419 dlen = SA_SIZE(sa); 1420 if (cp != NULL && buflen >= dlen) { 1421 KASSERT(dlen <= sizeof(ss), 1422 ("%s: sockaddr size overflow", __func__)); 1423 bzero(&ss, sizeof(ss)); 1424 bcopy(sa, &ss, sa->sa_len); 1425 sa = (struct sockaddr *)&ss; 1426 #ifdef INET6 1427 if (sa->sa_family == AF_INET6) { 1428 sin6 = (struct sockaddr_in6 *)sa; 1429 (void)sa6_recoverscope(sin6); 1430 } 1431 #endif 1432 bcopy((caddr_t)sa, cp, (unsigned)dlen); 1433 cp += dlen; 1434 buflen -= dlen; 1435 } else if (cp != NULL) { 1436 /* 1437 * Buffer too small. Count needed size 1438 * and return with error. 1439 */ 1440 cp = NULL; 1441 } 1442 1443 len += dlen; 1444 } 1445 1446 if (cp != NULL) { 1447 dlen = ALIGN(len) - len; 1448 if (buflen < dlen) 1449 cp = NULL; 1450 else { 1451 bzero(cp, dlen); 1452 cp += dlen; 1453 buflen -= dlen; 1454 } 1455 } 1456 len = ALIGN(len); 1457 1458 if (cp != NULL) { 1459 /* fill header iff buffer is large enough */ 1460 rtm->rtm_version = RTM_VERSION; 1461 rtm->rtm_type = type; 1462 rtm->rtm_msglen = len; 1463 } 1464 1465 *plen = len; 1466 1467 if (w != NULL && cp == NULL) 1468 return (ENOBUFS); 1469 1470 return (0); 1471 } 1472 1473 /* 1474 * This routine is called to generate a message from the routing 1475 * socket indicating that a redirect has occurred, a routing lookup 1476 * has failed, or that a protocol has detected timeouts to a particular 1477 * destination. 1478 */ 1479 void 1480 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error, 1481 int fibnum) 1482 { 1483 struct rt_msghdr *rtm; 1484 struct mbuf *m; 1485 struct sockaddr *sa = rtinfo->rti_info[RTAX_DST]; 1486 1487 if (V_route_cb.any_count == 0) 1488 return; 1489 m = rtsock_msg_mbuf(type, rtinfo); 1490 if (m == NULL) 1491 return; 1492 1493 if (fibnum != RT_ALL_FIBS) { 1494 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out " 1495 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs)); 1496 M_SETFIB(m, fibnum); 1497 m->m_flags |= RTS_FILTER_FIB; 1498 } 1499 1500 rtm = mtod(m, struct rt_msghdr *); 1501 rtm->rtm_flags = RTF_DONE | flags; 1502 rtm->rtm_errno = error; 1503 rtm->rtm_addrs = rtinfo->rti_addrs; 1504 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1505 } 1506 1507 void 1508 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error) 1509 { 1510 1511 rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS); 1512 } 1513 1514 /* 1515 * This routine is called to generate a message from the routing 1516 * socket indicating that the status of a network interface has changed. 1517 */ 1518 void 1519 rt_ifmsg(struct ifnet *ifp) 1520 { 1521 struct if_msghdr *ifm; 1522 struct mbuf *m; 1523 struct rt_addrinfo info; 1524 1525 if (V_route_cb.any_count == 0) 1526 return; 1527 bzero((caddr_t)&info, sizeof(info)); 1528 m = rtsock_msg_mbuf(RTM_IFINFO, &info); 1529 if (m == NULL) 1530 return; 1531 ifm = mtod(m, struct if_msghdr *); 1532 ifm->ifm_index = ifp->if_index; 1533 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1534 if_data_copy(ifp, &ifm->ifm_data); 1535 ifm->ifm_addrs = 0; 1536 rt_dispatch(m, AF_UNSPEC); 1537 } 1538 1539 /* 1540 * Announce interface address arrival/withdraw. 1541 * Please do not call directly, use rt_addrmsg(). 1542 * Assume input data to be valid. 1543 * Returns 0 on success. 1544 */ 1545 int 1546 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum) 1547 { 1548 struct rt_addrinfo info; 1549 struct sockaddr *sa; 1550 int ncmd; 1551 struct mbuf *m; 1552 struct ifa_msghdr *ifam; 1553 struct ifnet *ifp = ifa->ifa_ifp; 1554 struct sockaddr_storage ss; 1555 1556 if (V_route_cb.any_count == 0) 1557 return (0); 1558 1559 ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR; 1560 1561 bzero((caddr_t)&info, sizeof(info)); 1562 info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr; 1563 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr; 1564 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask( 1565 info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss); 1566 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr; 1567 if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL) 1568 return (ENOBUFS); 1569 ifam = mtod(m, struct ifa_msghdr *); 1570 ifam->ifam_index = ifp->if_index; 1571 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 1572 ifam->ifam_flags = ifa->ifa_flags; 1573 ifam->ifam_addrs = info.rti_addrs; 1574 1575 if (fibnum != RT_ALL_FIBS) { 1576 M_SETFIB(m, fibnum); 1577 m->m_flags |= RTS_FILTER_FIB; 1578 } 1579 1580 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1581 1582 return (0); 1583 } 1584 1585 /* 1586 * Announce route addition/removal to rtsock based on @rt data. 1587 * Callers are advives to use rt_routemsg() instead of using this 1588 * function directly. 1589 * Assume @rt data is consistent. 1590 * 1591 * Returns 0 on success. 1592 */ 1593 int 1594 rtsock_routemsg(int cmd, struct rtentry *rt, struct nhop_object *nh, 1595 int fibnum) 1596 { 1597 union sockaddr_union dst, mask; 1598 struct rt_addrinfo info; 1599 1600 if (V_route_cb.any_count == 0) 1601 return (0); 1602 1603 int family = rt_get_family(rt); 1604 init_sockaddrs_family(family, &dst.sa, &mask.sa); 1605 export_rtaddrs(rt, &dst.sa, &mask.sa); 1606 1607 bzero((caddr_t)&info, sizeof(info)); 1608 info.rti_info[RTAX_DST] = &dst.sa; 1609 info.rti_info[RTAX_NETMASK] = &mask.sa; 1610 info.rti_info[RTAX_GATEWAY] = &nh->gw_sa; 1611 info.rti_flags = rt->rte_flags | nhop_get_rtflags(nh); 1612 info.rti_ifp = nh->nh_ifp; 1613 1614 return (rtsock_routemsg_info(cmd, &info, fibnum)); 1615 } 1616 1617 int 1618 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum) 1619 { 1620 struct rt_msghdr *rtm; 1621 struct sockaddr *sa; 1622 struct mbuf *m; 1623 1624 if (V_route_cb.any_count == 0) 1625 return (0); 1626 1627 if (info->rti_flags & RTF_HOST) 1628 info->rti_info[RTAX_NETMASK] = NULL; 1629 1630 m = rtsock_msg_mbuf(cmd, info); 1631 if (m == NULL) 1632 return (ENOBUFS); 1633 1634 if (fibnum != RT_ALL_FIBS) { 1635 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out " 1636 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs)); 1637 M_SETFIB(m, fibnum); 1638 m->m_flags |= RTS_FILTER_FIB; 1639 } 1640 1641 rtm = mtod(m, struct rt_msghdr *); 1642 rtm->rtm_addrs = info->rti_addrs; 1643 if (info->rti_ifp != NULL) 1644 rtm->rtm_index = info->rti_ifp->if_index; 1645 /* Add RTF_DONE to indicate command 'completion' required by API */ 1646 info->rti_flags |= RTF_DONE; 1647 /* Reported routes has to be up */ 1648 if (cmd == RTM_ADD || cmd == RTM_CHANGE) 1649 info->rti_flags |= RTF_UP; 1650 rtm->rtm_flags = info->rti_flags; 1651 1652 sa = info->rti_info[RTAX_DST]; 1653 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1654 1655 return (0); 1656 } 1657 1658 /* 1659 * This is the analogue to the rt_newaddrmsg which performs the same 1660 * function but for multicast group memberhips. This is easier since 1661 * there is no route state to worry about. 1662 */ 1663 void 1664 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma) 1665 { 1666 struct rt_addrinfo info; 1667 struct mbuf *m = NULL; 1668 struct ifnet *ifp = ifma->ifma_ifp; 1669 struct ifma_msghdr *ifmam; 1670 1671 if (V_route_cb.any_count == 0) 1672 return; 1673 1674 bzero((caddr_t)&info, sizeof(info)); 1675 info.rti_info[RTAX_IFA] = ifma->ifma_addr; 1676 if (ifp && ifp->if_addr) 1677 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr; 1678 else 1679 info.rti_info[RTAX_IFP] = NULL; 1680 /* 1681 * If a link-layer address is present, present it as a ``gateway'' 1682 * (similarly to how ARP entries, e.g., are presented). 1683 */ 1684 info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr; 1685 m = rtsock_msg_mbuf(cmd, &info); 1686 if (m == NULL) 1687 return; 1688 ifmam = mtod(m, struct ifma_msghdr *); 1689 KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n", 1690 __func__)); 1691 ifmam->ifmam_index = ifp->if_index; 1692 ifmam->ifmam_addrs = info.rti_addrs; 1693 rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC); 1694 } 1695 1696 static struct mbuf * 1697 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what, 1698 struct rt_addrinfo *info) 1699 { 1700 struct if_announcemsghdr *ifan; 1701 struct mbuf *m; 1702 1703 if (V_route_cb.any_count == 0) 1704 return NULL; 1705 bzero((caddr_t)info, sizeof(*info)); 1706 m = rtsock_msg_mbuf(type, info); 1707 if (m != NULL) { 1708 ifan = mtod(m, struct if_announcemsghdr *); 1709 ifan->ifan_index = ifp->if_index; 1710 strlcpy(ifan->ifan_name, ifp->if_xname, 1711 sizeof(ifan->ifan_name)); 1712 ifan->ifan_what = what; 1713 } 1714 return m; 1715 } 1716 1717 /* 1718 * This is called to generate routing socket messages indicating 1719 * IEEE80211 wireless events. 1720 * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way. 1721 */ 1722 void 1723 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len) 1724 { 1725 struct mbuf *m; 1726 struct rt_addrinfo info; 1727 1728 m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info); 1729 if (m != NULL) { 1730 /* 1731 * Append the ieee80211 data. Try to stick it in the 1732 * mbuf containing the ifannounce msg; otherwise allocate 1733 * a new mbuf and append. 1734 * 1735 * NB: we assume m is a single mbuf. 1736 */ 1737 if (data_len > M_TRAILINGSPACE(m)) { 1738 struct mbuf *n = m_get(M_NOWAIT, MT_DATA); 1739 if (n == NULL) { 1740 m_freem(m); 1741 return; 1742 } 1743 bcopy(data, mtod(n, void *), data_len); 1744 n->m_len = data_len; 1745 m->m_next = n; 1746 } else if (data_len > 0) { 1747 bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len); 1748 m->m_len += data_len; 1749 } 1750 if (m->m_flags & M_PKTHDR) 1751 m->m_pkthdr.len += data_len; 1752 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len; 1753 rt_dispatch(m, AF_UNSPEC); 1754 } 1755 } 1756 1757 /* 1758 * This is called to generate routing socket messages indicating 1759 * network interface arrival and departure. 1760 */ 1761 void 1762 rt_ifannouncemsg(struct ifnet *ifp, int what) 1763 { 1764 struct mbuf *m; 1765 struct rt_addrinfo info; 1766 1767 m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info); 1768 if (m != NULL) 1769 rt_dispatch(m, AF_UNSPEC); 1770 } 1771 1772 static void 1773 rt_dispatch(struct mbuf *m, sa_family_t saf) 1774 { 1775 struct m_tag *tag; 1776 1777 /* 1778 * Preserve the family from the sockaddr, if any, in an m_tag for 1779 * use when injecting the mbuf into the routing socket buffer from 1780 * the netisr. 1781 */ 1782 if (saf != AF_UNSPEC) { 1783 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short), 1784 M_NOWAIT); 1785 if (tag == NULL) { 1786 m_freem(m); 1787 return; 1788 } 1789 *(unsigned short *)(tag + 1) = saf; 1790 m_tag_prepend(m, tag); 1791 } 1792 #ifdef VIMAGE 1793 if (V_loif) 1794 m->m_pkthdr.rcvif = V_loif; 1795 else { 1796 m_freem(m); 1797 return; 1798 } 1799 #endif 1800 netisr_queue(NETISR_ROUTE, m); /* mbuf is free'd on failure. */ 1801 } 1802 1803 /* 1804 * Checks if rte can be exported v.r.t jails/vnets. 1805 * 1806 * Returns 1 if it can, 0 otherwise. 1807 */ 1808 static bool 1809 can_export_rte(struct ucred *td_ucred, bool rt_is_host, 1810 const struct sockaddr *rt_dst) 1811 { 1812 1813 if ((!rt_is_host) ? jailed_without_vnet(td_ucred) 1814 : prison_if(td_ucred, rt_dst) != 0) 1815 return (false); 1816 return (true); 1817 } 1818 1819 1820 /* 1821 * This is used in dumping the kernel table via sysctl(). 1822 */ 1823 static int 1824 sysctl_dumpentry(struct rtentry *rt, void *vw) 1825 { 1826 struct walkarg *w = vw; 1827 struct nhop_object *nh; 1828 int error = 0; 1829 1830 NET_EPOCH_ASSERT(); 1831 1832 export_rtaddrs(rt, w->dst, w->mask); 1833 if (!can_export_rte(w->w_req->td->td_ucred, rt_is_host(rt), w->dst)) 1834 return (0); 1835 nh = rt_get_raw_nhop(rt); 1836 #ifdef ROUTE_MPATH 1837 if (NH_IS_NHGRP(nh)) { 1838 struct weightened_nhop *wn; 1839 uint32_t num_nhops; 1840 wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops); 1841 for (int i = 0; i < num_nhops; i++) { 1842 error = sysctl_dumpnhop(rt, wn[i].nh, wn[i].weight, w); 1843 if (error != 0) 1844 return (error); 1845 } 1846 } else 1847 #endif 1848 error = sysctl_dumpnhop(rt, nh, rt->rt_weight, w); 1849 1850 return (0); 1851 } 1852 1853 1854 static int 1855 sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, uint32_t weight, 1856 struct walkarg *w) 1857 { 1858 struct rt_addrinfo info; 1859 int error = 0, size; 1860 uint32_t rtflags; 1861 1862 rtflags = nhop_get_rtflags(nh); 1863 1864 if (w->w_op == NET_RT_FLAGS && !(rtflags & w->w_arg)) 1865 return (0); 1866 1867 bzero((caddr_t)&info, sizeof(info)); 1868 info.rti_info[RTAX_DST] = w->dst; 1869 info.rti_info[RTAX_GATEWAY] = &nh->gw_sa; 1870 info.rti_info[RTAX_NETMASK] = (rtflags & RTF_HOST) ? NULL : w->mask; 1871 info.rti_info[RTAX_GENMASK] = 0; 1872 if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) { 1873 info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr; 1874 info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr; 1875 if (nh->nh_ifp->if_flags & IFF_POINTOPOINT) 1876 info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr; 1877 } 1878 if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0) 1879 return (error); 1880 if (w->w_req && w->w_tmem) { 1881 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem; 1882 1883 bzero(&rtm->rtm_index, 1884 sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index)); 1885 1886 /* 1887 * rte flags may consist of RTF_HOST (duplicated in nhop rtflags) 1888 * and RTF_UP (if entry is linked, which is always true here). 1889 * Given that, use nhop rtflags & add RTF_UP. 1890 */ 1891 rtm->rtm_flags = rtflags | RTF_UP; 1892 if (rtm->rtm_flags & RTF_GWFLAG_COMPAT) 1893 rtm->rtm_flags = RTF_GATEWAY | 1894 (rtm->rtm_flags & ~RTF_GWFLAG_COMPAT); 1895 rt_getmetrics(rt, nh, &rtm->rtm_rmx); 1896 rtm->rtm_rmx.rmx_weight = weight; 1897 rtm->rtm_index = nh->nh_ifp->if_index; 1898 rtm->rtm_addrs = info.rti_addrs; 1899 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size); 1900 return (error); 1901 } 1902 return (error); 1903 } 1904 1905 static int 1906 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd, 1907 struct rt_addrinfo *info, struct walkarg *w, int len) 1908 { 1909 struct if_msghdrl *ifm; 1910 struct if_data *ifd; 1911 1912 ifm = (struct if_msghdrl *)w->w_tmem; 1913 1914 #ifdef COMPAT_FREEBSD32 1915 if (w->w_req->flags & SCTL_MASK32) { 1916 struct if_msghdrl32 *ifm32; 1917 1918 ifm32 = (struct if_msghdrl32 *)ifm; 1919 ifm32->ifm_addrs = info->rti_addrs; 1920 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1921 ifm32->ifm_index = ifp->if_index; 1922 ifm32->_ifm_spare1 = 0; 1923 ifm32->ifm_len = sizeof(*ifm32); 1924 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data); 1925 ifm32->_ifm_spare2 = 0; 1926 ifd = &ifm32->ifm_data; 1927 } else 1928 #endif 1929 { 1930 ifm->ifm_addrs = info->rti_addrs; 1931 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1932 ifm->ifm_index = ifp->if_index; 1933 ifm->_ifm_spare1 = 0; 1934 ifm->ifm_len = sizeof(*ifm); 1935 ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data); 1936 ifm->_ifm_spare2 = 0; 1937 ifd = &ifm->ifm_data; 1938 } 1939 1940 memcpy(ifd, src_ifd, sizeof(*ifd)); 1941 1942 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len)); 1943 } 1944 1945 static int 1946 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd, 1947 struct rt_addrinfo *info, struct walkarg *w, int len) 1948 { 1949 struct if_msghdr *ifm; 1950 struct if_data *ifd; 1951 1952 ifm = (struct if_msghdr *)w->w_tmem; 1953 1954 #ifdef COMPAT_FREEBSD32 1955 if (w->w_req->flags & SCTL_MASK32) { 1956 struct if_msghdr32 *ifm32; 1957 1958 ifm32 = (struct if_msghdr32 *)ifm; 1959 ifm32->ifm_addrs = info->rti_addrs; 1960 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1961 ifm32->ifm_index = ifp->if_index; 1962 ifm32->_ifm_spare1 = 0; 1963 ifd = &ifm32->ifm_data; 1964 } else 1965 #endif 1966 { 1967 ifm->ifm_addrs = info->rti_addrs; 1968 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1969 ifm->ifm_index = ifp->if_index; 1970 ifm->_ifm_spare1 = 0; 1971 ifd = &ifm->ifm_data; 1972 } 1973 1974 memcpy(ifd, src_ifd, sizeof(*ifd)); 1975 1976 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len)); 1977 } 1978 1979 static int 1980 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info, 1981 struct walkarg *w, int len) 1982 { 1983 struct ifa_msghdrl *ifam; 1984 struct if_data *ifd; 1985 1986 ifam = (struct ifa_msghdrl *)w->w_tmem; 1987 1988 #ifdef COMPAT_FREEBSD32 1989 if (w->w_req->flags & SCTL_MASK32) { 1990 struct ifa_msghdrl32 *ifam32; 1991 1992 ifam32 = (struct ifa_msghdrl32 *)ifam; 1993 ifam32->ifam_addrs = info->rti_addrs; 1994 ifam32->ifam_flags = ifa->ifa_flags; 1995 ifam32->ifam_index = ifa->ifa_ifp->if_index; 1996 ifam32->_ifam_spare1 = 0; 1997 ifam32->ifam_len = sizeof(*ifam32); 1998 ifam32->ifam_data_off = 1999 offsetof(struct ifa_msghdrl32, ifam_data); 2000 ifam32->ifam_metric = ifa->ifa_ifp->if_metric; 2001 ifd = &ifam32->ifam_data; 2002 } else 2003 #endif 2004 { 2005 ifam->ifam_addrs = info->rti_addrs; 2006 ifam->ifam_flags = ifa->ifa_flags; 2007 ifam->ifam_index = ifa->ifa_ifp->if_index; 2008 ifam->_ifam_spare1 = 0; 2009 ifam->ifam_len = sizeof(*ifam); 2010 ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data); 2011 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 2012 ifd = &ifam->ifam_data; 2013 } 2014 2015 bzero(ifd, sizeof(*ifd)); 2016 ifd->ifi_datalen = sizeof(struct if_data); 2017 ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets); 2018 ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets); 2019 ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes); 2020 ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes); 2021 2022 /* Fixup if_data carp(4) vhid. */ 2023 if (carp_get_vhid_p != NULL) 2024 ifd->ifi_vhid = (*carp_get_vhid_p)(ifa); 2025 2026 return (SYSCTL_OUT(w->w_req, w->w_tmem, len)); 2027 } 2028 2029 static int 2030 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info, 2031 struct walkarg *w, int len) 2032 { 2033 struct ifa_msghdr *ifam; 2034 2035 ifam = (struct ifa_msghdr *)w->w_tmem; 2036 ifam->ifam_addrs = info->rti_addrs; 2037 ifam->ifam_flags = ifa->ifa_flags; 2038 ifam->ifam_index = ifa->ifa_ifp->if_index; 2039 ifam->_ifam_spare1 = 0; 2040 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 2041 2042 return (SYSCTL_OUT(w->w_req, w->w_tmem, len)); 2043 } 2044 2045 static int 2046 sysctl_iflist(int af, struct walkarg *w) 2047 { 2048 struct ifnet *ifp; 2049 struct ifaddr *ifa; 2050 struct if_data ifd; 2051 struct rt_addrinfo info; 2052 int len, error = 0; 2053 struct sockaddr_storage ss; 2054 2055 bzero((caddr_t)&info, sizeof(info)); 2056 bzero(&ifd, sizeof(ifd)); 2057 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2058 if (w->w_arg && w->w_arg != ifp->if_index) 2059 continue; 2060 if_data_copy(ifp, &ifd); 2061 ifa = ifp->if_addr; 2062 info.rti_info[RTAX_IFP] = ifa->ifa_addr; 2063 error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len); 2064 if (error != 0) 2065 goto done; 2066 info.rti_info[RTAX_IFP] = NULL; 2067 if (w->w_req && w->w_tmem) { 2068 if (w->w_op == NET_RT_IFLISTL) 2069 error = sysctl_iflist_ifml(ifp, &ifd, &info, w, 2070 len); 2071 else 2072 error = sysctl_iflist_ifm(ifp, &ifd, &info, w, 2073 len); 2074 if (error) 2075 goto done; 2076 } 2077 while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) { 2078 if (af && af != ifa->ifa_addr->sa_family) 2079 continue; 2080 if (prison_if(w->w_req->td->td_ucred, 2081 ifa->ifa_addr) != 0) 2082 continue; 2083 info.rti_info[RTAX_IFA] = ifa->ifa_addr; 2084 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask( 2085 ifa->ifa_addr, ifa->ifa_netmask, &ss); 2086 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr; 2087 error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len); 2088 if (error != 0) 2089 goto done; 2090 if (w->w_req && w->w_tmem) { 2091 if (w->w_op == NET_RT_IFLISTL) 2092 error = sysctl_iflist_ifaml(ifa, &info, 2093 w, len); 2094 else 2095 error = sysctl_iflist_ifam(ifa, &info, 2096 w, len); 2097 if (error) 2098 goto done; 2099 } 2100 } 2101 info.rti_info[RTAX_IFA] = NULL; 2102 info.rti_info[RTAX_NETMASK] = NULL; 2103 info.rti_info[RTAX_BRD] = NULL; 2104 } 2105 done: 2106 return (error); 2107 } 2108 2109 static int 2110 sysctl_ifmalist(int af, struct walkarg *w) 2111 { 2112 struct rt_addrinfo info; 2113 struct ifaddr *ifa; 2114 struct ifmultiaddr *ifma; 2115 struct ifnet *ifp; 2116 int error, len; 2117 2118 NET_EPOCH_ASSERT(); 2119 2120 error = 0; 2121 bzero((caddr_t)&info, sizeof(info)); 2122 2123 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2124 if (w->w_arg && w->w_arg != ifp->if_index) 2125 continue; 2126 ifa = ifp->if_addr; 2127 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL; 2128 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 2129 if (af && af != ifma->ifma_addr->sa_family) 2130 continue; 2131 if (prison_if(w->w_req->td->td_ucred, 2132 ifma->ifma_addr) != 0) 2133 continue; 2134 info.rti_info[RTAX_IFA] = ifma->ifma_addr; 2135 info.rti_info[RTAX_GATEWAY] = 2136 (ifma->ifma_addr->sa_family != AF_LINK) ? 2137 ifma->ifma_lladdr : NULL; 2138 error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len); 2139 if (error != 0) 2140 break; 2141 if (w->w_req && w->w_tmem) { 2142 struct ifma_msghdr *ifmam; 2143 2144 ifmam = (struct ifma_msghdr *)w->w_tmem; 2145 ifmam->ifmam_index = ifma->ifma_ifp->if_index; 2146 ifmam->ifmam_flags = 0; 2147 ifmam->ifmam_addrs = info.rti_addrs; 2148 ifmam->_ifmam_spare1 = 0; 2149 error = SYSCTL_OUT(w->w_req, w->w_tmem, len); 2150 if (error != 0) 2151 break; 2152 } 2153 } 2154 if (error != 0) 2155 break; 2156 } 2157 return (error); 2158 } 2159 2160 static void 2161 rtable_sysctl_dump(uint32_t fibnum, int family, struct walkarg *w) 2162 { 2163 union sockaddr_union sa_dst, sa_mask; 2164 2165 w->family = family; 2166 w->dst = (struct sockaddr *)&sa_dst; 2167 w->mask = (struct sockaddr *)&sa_mask; 2168 2169 init_sockaddrs_family(family, w->dst, w->mask); 2170 2171 rib_walk(fibnum, family, false, sysctl_dumpentry, w); 2172 } 2173 2174 static int 2175 sysctl_rtsock(SYSCTL_HANDLER_ARGS) 2176 { 2177 struct epoch_tracker et; 2178 int *name = (int *)arg1; 2179 u_int namelen = arg2; 2180 struct rib_head *rnh = NULL; /* silence compiler. */ 2181 int i, lim, error = EINVAL; 2182 int fib = 0; 2183 u_char af; 2184 struct walkarg w; 2185 2186 name ++; 2187 namelen--; 2188 if (req->newptr) 2189 return (EPERM); 2190 if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP || name[1] == NET_RT_NHGRP) { 2191 if (namelen == 3) 2192 fib = req->td->td_proc->p_fibnum; 2193 else if (namelen == 4) 2194 fib = (name[3] == RT_ALL_FIBS) ? 2195 req->td->td_proc->p_fibnum : name[3]; 2196 else 2197 return ((namelen < 3) ? EISDIR : ENOTDIR); 2198 if (fib < 0 || fib >= rt_numfibs) 2199 return (EINVAL); 2200 } else if (namelen != 3) 2201 return ((namelen < 3) ? EISDIR : ENOTDIR); 2202 af = name[0]; 2203 if (af > AF_MAX) 2204 return (EINVAL); 2205 bzero(&w, sizeof(w)); 2206 w.w_op = name[1]; 2207 w.w_arg = name[2]; 2208 w.w_req = req; 2209 2210 error = sysctl_wire_old_buffer(req, 0); 2211 if (error) 2212 return (error); 2213 2214 /* 2215 * Allocate reply buffer in advance. 2216 * All rtsock messages has maximum length of u_short. 2217 */ 2218 w.w_tmemsize = 65536; 2219 w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK); 2220 2221 NET_EPOCH_ENTER(et); 2222 switch (w.w_op) { 2223 case NET_RT_DUMP: 2224 case NET_RT_FLAGS: 2225 if (af == 0) { /* dump all tables */ 2226 i = 1; 2227 lim = AF_MAX; 2228 } else /* dump only one table */ 2229 i = lim = af; 2230 2231 /* 2232 * take care of llinfo entries, the caller must 2233 * specify an AF 2234 */ 2235 if (w.w_op == NET_RT_FLAGS && 2236 (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) { 2237 if (af != 0) 2238 error = lltable_sysctl_dumparp(af, w.w_req); 2239 else 2240 error = EINVAL; 2241 break; 2242 } 2243 /* 2244 * take care of routing entries 2245 */ 2246 for (error = 0; error == 0 && i <= lim; i++) { 2247 rnh = rt_tables_get_rnh(fib, i); 2248 if (rnh != NULL) { 2249 rtable_sysctl_dump(fib, i, &w); 2250 } else if (af != 0) 2251 error = EAFNOSUPPORT; 2252 } 2253 break; 2254 case NET_RT_NHOP: 2255 case NET_RT_NHGRP: 2256 /* Allow dumping one specific af/fib at a time */ 2257 if (namelen < 4) { 2258 error = EINVAL; 2259 break; 2260 } 2261 fib = name[3]; 2262 if (fib < 0 || fib > rt_numfibs) { 2263 error = EINVAL; 2264 break; 2265 } 2266 rnh = rt_tables_get_rnh(fib, af); 2267 if (rnh == NULL) { 2268 error = EAFNOSUPPORT; 2269 break; 2270 } 2271 if (w.w_op == NET_RT_NHOP) 2272 error = nhops_dump_sysctl(rnh, w.w_req); 2273 else 2274 #ifdef ROUTE_MPATH 2275 error = nhgrp_dump_sysctl(rnh, w.w_req); 2276 #else 2277 error = ENOTSUP; 2278 #endif 2279 break; 2280 case NET_RT_IFLIST: 2281 case NET_RT_IFLISTL: 2282 error = sysctl_iflist(af, &w); 2283 break; 2284 2285 case NET_RT_IFMALIST: 2286 error = sysctl_ifmalist(af, &w); 2287 break; 2288 } 2289 NET_EPOCH_EXIT(et); 2290 2291 free(w.w_tmem, M_TEMP); 2292 return (error); 2293 } 2294 2295 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE, 2296 sysctl_rtsock, "Return route tables and interface/address lists"); 2297 2298 /* 2299 * Definitions of protocols supported in the ROUTE domain. 2300 */ 2301 2302 static struct domain routedomain; /* or at least forward */ 2303 2304 static struct protosw routesw[] = { 2305 { 2306 .pr_type = SOCK_RAW, 2307 .pr_domain = &routedomain, 2308 .pr_flags = PR_ATOMIC|PR_ADDR, 2309 .pr_output = route_output, 2310 .pr_ctlinput = raw_ctlinput, 2311 .pr_init = raw_init, 2312 .pr_usrreqs = &route_usrreqs 2313 } 2314 }; 2315 2316 static struct domain routedomain = { 2317 .dom_family = PF_ROUTE, 2318 .dom_name = "route", 2319 .dom_protosw = routesw, 2320 .dom_protoswNPROTOSW = &routesw[nitems(routesw)] 2321 }; 2322 2323 VNET_DOMAIN_SET(route); 2324