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 uint32_t scopeid = 0; 803 #ifdef INET 804 if (dst->sa_family == AF_INET) { 805 struct sockaddr_in *dst4 = (struct sockaddr_in *)dst; 806 struct sockaddr_in *mask4 = (struct sockaddr_in *)mask; 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 rt_get_inet6_prefix_pmask(rt, &dst6->sin6_addr, &mask6->sin6_addr, 817 &scopeid); 818 dst6->sin6_scope_id = scopeid; 819 return; 820 } 821 #endif 822 } 823 824 825 /* 826 * Update sockaddrs, flags, etc in @prtm based on @rc data. 827 * rtm can be reallocated. 828 * 829 * Returns 0 on success, along with pointer to (potentially reallocated) 830 * rtm. 831 * 832 */ 833 static int 834 update_rtm_from_rc(struct rt_addrinfo *info, struct rt_msghdr **prtm, 835 int alloc_len, struct rib_cmd_info *rc, struct nhop_object *nh) 836 { 837 struct walkarg w; 838 union sockaddr_union saun; 839 struct rt_msghdr *rtm, *orig_rtm = NULL; 840 struct ifnet *ifp; 841 int error, len; 842 843 rtm = *prtm; 844 union sockaddr_union sa_dst, sa_mask; 845 int family = info->rti_info[RTAX_DST]->sa_family; 846 init_sockaddrs_family(family, &sa_dst.sa, &sa_mask.sa); 847 export_rtaddrs(rc->rc_rt, &sa_dst.sa, &sa_mask.sa); 848 849 info->rti_info[RTAX_DST] = &sa_dst.sa; 850 info->rti_info[RTAX_NETMASK] = rt_is_host(rc->rc_rt) ? NULL : &sa_mask.sa; 851 info->rti_info[RTAX_GATEWAY] = &nh->gw_sa; 852 info->rti_info[RTAX_GENMASK] = 0; 853 ifp = nh->nh_ifp; 854 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) { 855 if (ifp) { 856 info->rti_info[RTAX_IFP] = 857 ifp->if_addr->ifa_addr; 858 error = rtm_get_jailed(info, ifp, nh, 859 &saun, curthread->td_ucred); 860 if (error != 0) 861 return (error); 862 if (ifp->if_flags & IFF_POINTOPOINT) 863 info->rti_info[RTAX_BRD] = 864 nh->nh_ifa->ifa_dstaddr; 865 rtm->rtm_index = ifp->if_index; 866 } else { 867 info->rti_info[RTAX_IFP] = NULL; 868 info->rti_info[RTAX_IFA] = NULL; 869 } 870 } else if (ifp != NULL) 871 rtm->rtm_index = ifp->if_index; 872 873 /* Check if we need to realloc storage */ 874 rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len); 875 if (len > alloc_len) { 876 struct rt_msghdr *tmp_rtm; 877 878 tmp_rtm = malloc(len, M_TEMP, M_NOWAIT); 879 if (tmp_rtm == NULL) 880 return (ENOBUFS); 881 bcopy(rtm, tmp_rtm, rtm->rtm_msglen); 882 orig_rtm = rtm; 883 rtm = tmp_rtm; 884 alloc_len = len; 885 886 /* 887 * Delay freeing original rtm as info contains 888 * data referencing it. 889 */ 890 } 891 892 w.w_tmem = (caddr_t)rtm; 893 w.w_tmemsize = alloc_len; 894 rtsock_msg_buffer(rtm->rtm_type, info, &w, &len); 895 896 rtm->rtm_flags = rc->rc_rt->rte_flags | nhop_get_rtflags(nh); 897 if (rtm->rtm_flags & RTF_GWFLAG_COMPAT) 898 rtm->rtm_flags = RTF_GATEWAY | 899 (rtm->rtm_flags & ~RTF_GWFLAG_COMPAT); 900 rt_getmetrics(rc->rc_rt, nh, &rtm->rtm_rmx); 901 rtm->rtm_rmx.rmx_weight = rc->rc_nh_weight; 902 rtm->rtm_addrs = info->rti_addrs; 903 904 if (orig_rtm != NULL) 905 free(orig_rtm, M_TEMP); 906 *prtm = rtm; 907 908 return (0); 909 } 910 911 #ifdef ROUTE_MPATH 912 static void 913 save_del_notification(struct rib_cmd_info *rc, void *_cbdata) 914 { 915 struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata; 916 917 if (rc->rc_cmd == RTM_DELETE) 918 *rc_new = *rc; 919 } 920 921 static void 922 save_add_notification(struct rib_cmd_info *rc, void *_cbdata) 923 { 924 struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata; 925 926 if (rc->rc_cmd == RTM_ADD) 927 *rc_new = *rc; 928 } 929 #endif 930 931 /*ARGSUSED*/ 932 static int 933 route_output(struct mbuf *m, struct socket *so, ...) 934 { 935 struct rt_msghdr *rtm = NULL; 936 struct rtentry *rt = NULL; 937 struct rt_addrinfo info; 938 struct epoch_tracker et; 939 #ifdef INET6 940 struct sockaddr_storage ss; 941 struct sockaddr_in6 *sin6; 942 int i, rti_need_deembed = 0; 943 #endif 944 int alloc_len = 0, len, error = 0, fibnum; 945 sa_family_t saf = AF_UNSPEC; 946 struct walkarg w; 947 struct rib_cmd_info rc; 948 struct nhop_object *nh; 949 950 fibnum = so->so_fibnum; 951 #define senderr(e) { error = e; goto flush;} 952 if (m == NULL || ((m->m_len < sizeof(long)) && 953 (m = m_pullup(m, sizeof(long))) == NULL)) 954 return (ENOBUFS); 955 if ((m->m_flags & M_PKTHDR) == 0) 956 panic("route_output"); 957 NET_EPOCH_ENTER(et); 958 len = m->m_pkthdr.len; 959 if (len < sizeof(*rtm) || 960 len != mtod(m, struct rt_msghdr *)->rtm_msglen) 961 senderr(EINVAL); 962 963 /* 964 * Most of current messages are in range 200-240 bytes, 965 * minimize possible re-allocation on reply using larger size 966 * buffer aligned on 1k boundaty. 967 */ 968 alloc_len = roundup2(len, 1024); 969 if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL) 970 senderr(ENOBUFS); 971 972 m_copydata(m, 0, len, (caddr_t)rtm); 973 bzero(&info, sizeof(info)); 974 bzero(&w, sizeof(w)); 975 nh = NULL; 976 977 if (rtm->rtm_version != RTM_VERSION) { 978 /* Do not touch message since format is unknown */ 979 free(rtm, M_TEMP); 980 rtm = NULL; 981 senderr(EPROTONOSUPPORT); 982 } 983 984 /* 985 * Starting from here, it is possible 986 * to alter original message and insert 987 * caller PID and error value. 988 */ 989 990 if ((error = fill_addrinfo(rtm, len, fibnum, &info)) != 0) { 991 senderr(error); 992 } 993 994 saf = info.rti_info[RTAX_DST]->sa_family; 995 996 /* support for new ARP code */ 997 if (rtm->rtm_flags & RTF_LLDATA) { 998 error = lla_rt_output(rtm, &info); 999 #ifdef INET6 1000 if (error == 0) 1001 rti_need_deembed = 1; 1002 #endif 1003 goto flush; 1004 } 1005 1006 switch (rtm->rtm_type) { 1007 case RTM_ADD: 1008 case RTM_CHANGE: 1009 if (rtm->rtm_type == RTM_ADD) { 1010 if (info.rti_info[RTAX_GATEWAY] == NULL) 1011 senderr(EINVAL); 1012 } 1013 error = rib_action(fibnum, rtm->rtm_type, &info, &rc); 1014 if (error == 0) { 1015 #ifdef INET6 1016 rti_need_deembed = 1; 1017 #endif 1018 #ifdef ROUTE_MPATH 1019 if (NH_IS_NHGRP(rc.rc_nh_new) || 1020 (rc.rc_nh_old && NH_IS_NHGRP(rc.rc_nh_old))) { 1021 struct rib_cmd_info rc_simple = {}; 1022 rib_decompose_notification(&rc, 1023 save_add_notification, (void *)&rc_simple); 1024 rc = rc_simple; 1025 } 1026 #endif 1027 nh = rc.rc_nh_new; 1028 rtm->rtm_index = nh->nh_ifp->if_index; 1029 rtm->rtm_flags = rc.rc_rt->rte_flags | nhop_get_rtflags(nh); 1030 } 1031 break; 1032 1033 case RTM_DELETE: 1034 error = rib_action(fibnum, RTM_DELETE, &info, &rc); 1035 if (error == 0) { 1036 #ifdef ROUTE_MPATH 1037 if (NH_IS_NHGRP(rc.rc_nh_old) || 1038 (rc.rc_nh_new && NH_IS_NHGRP(rc.rc_nh_new))) { 1039 struct rib_cmd_info rc_simple = {}; 1040 rib_decompose_notification(&rc, 1041 save_del_notification, (void *)&rc_simple); 1042 rc = rc_simple; 1043 } 1044 #endif 1045 nh = rc.rc_nh_old; 1046 goto report; 1047 } 1048 #ifdef INET6 1049 /* rt_msg2() will not be used when RTM_DELETE fails. */ 1050 rti_need_deembed = 1; 1051 #endif 1052 break; 1053 1054 case RTM_GET: 1055 error = handle_rtm_get(&info, fibnum, rtm, &rc); 1056 if (error != 0) 1057 senderr(error); 1058 nh = rc.rc_nh_new; 1059 1060 report: 1061 if (!can_export_rte(curthread->td_ucred, 1062 info.rti_info[RTAX_NETMASK] == NULL, 1063 info.rti_info[RTAX_DST])) { 1064 senderr(ESRCH); 1065 } 1066 1067 error = update_rtm_from_rc(&info, &rtm, alloc_len, &rc, nh); 1068 /* 1069 * Note that some sockaddr pointers may have changed to 1070 * point to memory outsize @rtm. Some may be pointing 1071 * to the on-stack variables. 1072 * Given that, any pointer in @info CANNOT BE USED. 1073 */ 1074 1075 /* 1076 * scopeid deembedding has been performed while 1077 * writing updated rtm in rtsock_msg_buffer(). 1078 * With that in mind, skip deembedding procedure below. 1079 */ 1080 #ifdef INET6 1081 rti_need_deembed = 0; 1082 #endif 1083 if (error != 0) 1084 senderr(error); 1085 break; 1086 1087 default: 1088 senderr(EOPNOTSUPP); 1089 } 1090 1091 flush: 1092 NET_EPOCH_EXIT(et); 1093 rt = NULL; 1094 1095 #ifdef INET6 1096 if (rtm != NULL) { 1097 if (rti_need_deembed) { 1098 /* sin6_scope_id is recovered before sending rtm. */ 1099 sin6 = (struct sockaddr_in6 *)&ss; 1100 for (i = 0; i < RTAX_MAX; i++) { 1101 if (info.rti_info[i] == NULL) 1102 continue; 1103 if (info.rti_info[i]->sa_family != AF_INET6) 1104 continue; 1105 bcopy(info.rti_info[i], sin6, sizeof(*sin6)); 1106 if (sa6_recoverscope(sin6) == 0) 1107 bcopy(sin6, info.rti_info[i], 1108 sizeof(*sin6)); 1109 } 1110 } 1111 } 1112 #endif 1113 send_rtm_reply(so, rtm, m, saf, fibnum, error); 1114 1115 return (error); 1116 } 1117 1118 /* 1119 * Sends the prepared reply message in @rtm to all rtsock clients. 1120 * Frees @m and @rtm. 1121 * 1122 */ 1123 static void 1124 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m, 1125 sa_family_t saf, u_int fibnum, int rtm_errno) 1126 { 1127 struct rawcb *rp = NULL; 1128 1129 /* 1130 * Check to see if we don't want our own messages. 1131 */ 1132 if ((so->so_options & SO_USELOOPBACK) == 0) { 1133 if (V_route_cb.any_count <= 1) { 1134 if (rtm != NULL) 1135 free(rtm, M_TEMP); 1136 m_freem(m); 1137 return; 1138 } 1139 /* There is another listener, so construct message */ 1140 rp = sotorawcb(so); 1141 } 1142 1143 if (rtm != NULL) { 1144 if (rtm_errno!= 0) 1145 rtm->rtm_errno = rtm_errno; 1146 else 1147 rtm->rtm_flags |= RTF_DONE; 1148 1149 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm); 1150 if (m->m_pkthdr.len < rtm->rtm_msglen) { 1151 m_freem(m); 1152 m = NULL; 1153 } else if (m->m_pkthdr.len > rtm->rtm_msglen) 1154 m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len); 1155 1156 free(rtm, M_TEMP); 1157 } 1158 if (m != NULL) { 1159 M_SETFIB(m, fibnum); 1160 m->m_flags |= RTS_FILTER_FIB; 1161 if (rp) { 1162 /* 1163 * XXX insure we don't get a copy by 1164 * invalidating our protocol 1165 */ 1166 unsigned short family = rp->rcb_proto.sp_family; 1167 rp->rcb_proto.sp_family = 0; 1168 rt_dispatch(m, saf); 1169 rp->rcb_proto.sp_family = family; 1170 } else 1171 rt_dispatch(m, saf); 1172 } 1173 } 1174 1175 static void 1176 rt_getmetrics(const struct rtentry *rt, const struct nhop_object *nh, 1177 struct rt_metrics *out) 1178 { 1179 1180 bzero(out, sizeof(*out)); 1181 out->rmx_mtu = nh->nh_mtu; 1182 out->rmx_weight = rt->rt_weight; 1183 out->rmx_nhidx = nhop_get_idx(nh); 1184 /* Kernel -> userland timebase conversion. */ 1185 out->rmx_expire = rt->rt_expire ? 1186 rt->rt_expire - time_uptime + time_second : 0; 1187 } 1188 1189 /* 1190 * Extract the addresses of the passed sockaddrs. 1191 * Do a little sanity checking so as to avoid bad memory references. 1192 * This data is derived straight from userland. 1193 */ 1194 static int 1195 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo) 1196 { 1197 struct sockaddr *sa; 1198 int i; 1199 1200 for (i = 0; i < RTAX_MAX && cp < cplim; i++) { 1201 if ((rtinfo->rti_addrs & (1 << i)) == 0) 1202 continue; 1203 sa = (struct sockaddr *)cp; 1204 /* 1205 * It won't fit. 1206 */ 1207 if (cp + sa->sa_len > cplim) 1208 return (EINVAL); 1209 /* 1210 * there are no more.. quit now 1211 * If there are more bits, they are in error. 1212 * I've seen this. route(1) can evidently generate these. 1213 * This causes kernel to core dump. 1214 * for compatibility, If we see this, point to a safe address. 1215 */ 1216 if (sa->sa_len == 0) { 1217 rtinfo->rti_info[i] = &sa_zero; 1218 return (0); /* should be EINVAL but for compat */ 1219 } 1220 /* accept it */ 1221 #ifdef INET6 1222 if (sa->sa_family == AF_INET6) 1223 sa6_embedscope((struct sockaddr_in6 *)sa, 1224 V_ip6_use_defzone); 1225 #endif 1226 rtinfo->rti_info[i] = sa; 1227 cp += SA_SIZE(sa); 1228 } 1229 return (0); 1230 } 1231 1232 /* 1233 * Fill in @dmask with valid netmask leaving original @smask 1234 * intact. Mostly used with radix netmasks. 1235 */ 1236 struct sockaddr * 1237 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask, 1238 struct sockaddr_storage *dmask) 1239 { 1240 if (dst == NULL || smask == NULL) 1241 return (NULL); 1242 1243 memset(dmask, 0, dst->sa_len); 1244 memcpy(dmask, smask, smask->sa_len); 1245 dmask->ss_len = dst->sa_len; 1246 dmask->ss_family = dst->sa_family; 1247 1248 return ((struct sockaddr *)dmask); 1249 } 1250 1251 /* 1252 * Writes information related to @rtinfo object to newly-allocated mbuf. 1253 * Assumes MCLBYTES is enough to construct any message. 1254 * Used for OS notifications of vaious events (if/ifa announces,etc) 1255 * 1256 * Returns allocated mbuf or NULL on failure. 1257 */ 1258 static struct mbuf * 1259 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo) 1260 { 1261 struct sockaddr_storage ss; 1262 struct rt_msghdr *rtm; 1263 struct mbuf *m; 1264 int i; 1265 struct sockaddr *sa; 1266 #ifdef INET6 1267 struct sockaddr_in6 *sin6; 1268 #endif 1269 int len, dlen; 1270 1271 switch (type) { 1272 case RTM_DELADDR: 1273 case RTM_NEWADDR: 1274 len = sizeof(struct ifa_msghdr); 1275 break; 1276 1277 case RTM_DELMADDR: 1278 case RTM_NEWMADDR: 1279 len = sizeof(struct ifma_msghdr); 1280 break; 1281 1282 case RTM_IFINFO: 1283 len = sizeof(struct if_msghdr); 1284 break; 1285 1286 case RTM_IFANNOUNCE: 1287 case RTM_IEEE80211: 1288 len = sizeof(struct if_announcemsghdr); 1289 break; 1290 1291 default: 1292 len = sizeof(struct rt_msghdr); 1293 } 1294 1295 /* XXXGL: can we use MJUMPAGESIZE cluster here? */ 1296 KASSERT(len <= MCLBYTES, ("%s: message too big", __func__)); 1297 if (len > MHLEN) 1298 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 1299 else 1300 m = m_gethdr(M_NOWAIT, MT_DATA); 1301 if (m == NULL) 1302 return (m); 1303 1304 m->m_pkthdr.len = m->m_len = len; 1305 rtm = mtod(m, struct rt_msghdr *); 1306 bzero((caddr_t)rtm, len); 1307 for (i = 0; i < RTAX_MAX; i++) { 1308 if ((sa = rtinfo->rti_info[i]) == NULL) 1309 continue; 1310 rtinfo->rti_addrs |= (1 << i); 1311 1312 dlen = SA_SIZE(sa); 1313 KASSERT(dlen <= sizeof(ss), 1314 ("%s: sockaddr size overflow", __func__)); 1315 bzero(&ss, sizeof(ss)); 1316 bcopy(sa, &ss, sa->sa_len); 1317 sa = (struct sockaddr *)&ss; 1318 #ifdef INET6 1319 if (sa->sa_family == AF_INET6) { 1320 sin6 = (struct sockaddr_in6 *)sa; 1321 (void)sa6_recoverscope(sin6); 1322 } 1323 #endif 1324 m_copyback(m, len, dlen, (caddr_t)sa); 1325 len += dlen; 1326 } 1327 if (m->m_pkthdr.len != len) { 1328 m_freem(m); 1329 return (NULL); 1330 } 1331 rtm->rtm_msglen = len; 1332 rtm->rtm_version = RTM_VERSION; 1333 rtm->rtm_type = type; 1334 return (m); 1335 } 1336 1337 /* 1338 * Writes information related to @rtinfo object to preallocated buffer. 1339 * Stores needed size in @plen. If @w is NULL, calculates size without 1340 * writing. 1341 * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation. 1342 * 1343 * Returns 0 on success. 1344 * 1345 */ 1346 static int 1347 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen) 1348 { 1349 struct sockaddr_storage ss; 1350 int len, buflen = 0, dlen, i; 1351 caddr_t cp = NULL; 1352 struct rt_msghdr *rtm = NULL; 1353 #ifdef INET6 1354 struct sockaddr_in6 *sin6; 1355 #endif 1356 #ifdef COMPAT_FREEBSD32 1357 bool compat32 = false; 1358 #endif 1359 1360 switch (type) { 1361 case RTM_DELADDR: 1362 case RTM_NEWADDR: 1363 if (w != NULL && w->w_op == NET_RT_IFLISTL) { 1364 #ifdef COMPAT_FREEBSD32 1365 if (w->w_req->flags & SCTL_MASK32) { 1366 len = sizeof(struct ifa_msghdrl32); 1367 compat32 = true; 1368 } else 1369 #endif 1370 len = sizeof(struct ifa_msghdrl); 1371 } else 1372 len = sizeof(struct ifa_msghdr); 1373 break; 1374 1375 case RTM_IFINFO: 1376 #ifdef COMPAT_FREEBSD32 1377 if (w != NULL && w->w_req->flags & SCTL_MASK32) { 1378 if (w->w_op == NET_RT_IFLISTL) 1379 len = sizeof(struct if_msghdrl32); 1380 else 1381 len = sizeof(struct if_msghdr32); 1382 compat32 = true; 1383 break; 1384 } 1385 #endif 1386 if (w != NULL && w->w_op == NET_RT_IFLISTL) 1387 len = sizeof(struct if_msghdrl); 1388 else 1389 len = sizeof(struct if_msghdr); 1390 break; 1391 1392 case RTM_NEWMADDR: 1393 len = sizeof(struct ifma_msghdr); 1394 break; 1395 1396 default: 1397 len = sizeof(struct rt_msghdr); 1398 } 1399 1400 if (w != NULL) { 1401 rtm = (struct rt_msghdr *)w->w_tmem; 1402 buflen = w->w_tmemsize - len; 1403 cp = (caddr_t)w->w_tmem + len; 1404 } 1405 1406 rtinfo->rti_addrs = 0; 1407 for (i = 0; i < RTAX_MAX; i++) { 1408 struct sockaddr *sa; 1409 1410 if ((sa = rtinfo->rti_info[i]) == NULL) 1411 continue; 1412 rtinfo->rti_addrs |= (1 << i); 1413 #ifdef COMPAT_FREEBSD32 1414 if (compat32) 1415 dlen = SA_SIZE32(sa); 1416 else 1417 #endif 1418 dlen = SA_SIZE(sa); 1419 if (cp != NULL && buflen >= dlen) { 1420 KASSERT(dlen <= sizeof(ss), 1421 ("%s: sockaddr size overflow", __func__)); 1422 bzero(&ss, sizeof(ss)); 1423 bcopy(sa, &ss, sa->sa_len); 1424 sa = (struct sockaddr *)&ss; 1425 #ifdef INET6 1426 if (sa->sa_family == AF_INET6) { 1427 sin6 = (struct sockaddr_in6 *)sa; 1428 (void)sa6_recoverscope(sin6); 1429 } 1430 #endif 1431 bcopy((caddr_t)sa, cp, (unsigned)dlen); 1432 cp += dlen; 1433 buflen -= dlen; 1434 } else if (cp != NULL) { 1435 /* 1436 * Buffer too small. Count needed size 1437 * and return with error. 1438 */ 1439 cp = NULL; 1440 } 1441 1442 len += dlen; 1443 } 1444 1445 if (cp != NULL) { 1446 dlen = ALIGN(len) - len; 1447 if (buflen < dlen) 1448 cp = NULL; 1449 else { 1450 bzero(cp, dlen); 1451 cp += dlen; 1452 buflen -= dlen; 1453 } 1454 } 1455 len = ALIGN(len); 1456 1457 if (cp != NULL) { 1458 /* fill header iff buffer is large enough */ 1459 rtm->rtm_version = RTM_VERSION; 1460 rtm->rtm_type = type; 1461 rtm->rtm_msglen = len; 1462 } 1463 1464 *plen = len; 1465 1466 if (w != NULL && cp == NULL) 1467 return (ENOBUFS); 1468 1469 return (0); 1470 } 1471 1472 /* 1473 * This routine is called to generate a message from the routing 1474 * socket indicating that a redirect has occurred, a routing lookup 1475 * has failed, or that a protocol has detected timeouts to a particular 1476 * destination. 1477 */ 1478 void 1479 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error, 1480 int fibnum) 1481 { 1482 struct rt_msghdr *rtm; 1483 struct mbuf *m; 1484 struct sockaddr *sa = rtinfo->rti_info[RTAX_DST]; 1485 1486 if (V_route_cb.any_count == 0) 1487 return; 1488 m = rtsock_msg_mbuf(type, rtinfo); 1489 if (m == NULL) 1490 return; 1491 1492 if (fibnum != RT_ALL_FIBS) { 1493 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out " 1494 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs)); 1495 M_SETFIB(m, fibnum); 1496 m->m_flags |= RTS_FILTER_FIB; 1497 } 1498 1499 rtm = mtod(m, struct rt_msghdr *); 1500 rtm->rtm_flags = RTF_DONE | flags; 1501 rtm->rtm_errno = error; 1502 rtm->rtm_addrs = rtinfo->rti_addrs; 1503 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1504 } 1505 1506 void 1507 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error) 1508 { 1509 1510 rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS); 1511 } 1512 1513 /* 1514 * This routine is called to generate a message from the routing 1515 * socket indicating that the status of a network interface has changed. 1516 */ 1517 void 1518 rt_ifmsg(struct ifnet *ifp) 1519 { 1520 struct if_msghdr *ifm; 1521 struct mbuf *m; 1522 struct rt_addrinfo info; 1523 1524 if (V_route_cb.any_count == 0) 1525 return; 1526 bzero((caddr_t)&info, sizeof(info)); 1527 m = rtsock_msg_mbuf(RTM_IFINFO, &info); 1528 if (m == NULL) 1529 return; 1530 ifm = mtod(m, struct if_msghdr *); 1531 ifm->ifm_index = ifp->if_index; 1532 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1533 if_data_copy(ifp, &ifm->ifm_data); 1534 ifm->ifm_addrs = 0; 1535 rt_dispatch(m, AF_UNSPEC); 1536 } 1537 1538 /* 1539 * Announce interface address arrival/withdraw. 1540 * Please do not call directly, use rt_addrmsg(). 1541 * Assume input data to be valid. 1542 * Returns 0 on success. 1543 */ 1544 int 1545 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum) 1546 { 1547 struct rt_addrinfo info; 1548 struct sockaddr *sa; 1549 int ncmd; 1550 struct mbuf *m; 1551 struct ifa_msghdr *ifam; 1552 struct ifnet *ifp = ifa->ifa_ifp; 1553 struct sockaddr_storage ss; 1554 1555 if (V_route_cb.any_count == 0) 1556 return (0); 1557 1558 ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR; 1559 1560 bzero((caddr_t)&info, sizeof(info)); 1561 info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr; 1562 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr; 1563 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask( 1564 info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss); 1565 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr; 1566 if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL) 1567 return (ENOBUFS); 1568 ifam = mtod(m, struct ifa_msghdr *); 1569 ifam->ifam_index = ifp->if_index; 1570 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 1571 ifam->ifam_flags = ifa->ifa_flags; 1572 ifam->ifam_addrs = info.rti_addrs; 1573 1574 if (fibnum != RT_ALL_FIBS) { 1575 M_SETFIB(m, fibnum); 1576 m->m_flags |= RTS_FILTER_FIB; 1577 } 1578 1579 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1580 1581 return (0); 1582 } 1583 1584 /* 1585 * Announce route addition/removal to rtsock based on @rt data. 1586 * Callers are advives to use rt_routemsg() instead of using this 1587 * function directly. 1588 * Assume @rt data is consistent. 1589 * 1590 * Returns 0 on success. 1591 */ 1592 int 1593 rtsock_routemsg(int cmd, struct rtentry *rt, struct ifnet *ifp, int rti_addrs, 1594 int fibnum) 1595 { 1596 struct sockaddr_storage ss; 1597 struct rt_addrinfo info; 1598 struct nhop_object *nh; 1599 1600 if (V_route_cb.any_count == 0) 1601 return (0); 1602 1603 nh = rt->rt_nhop; 1604 bzero((caddr_t)&info, sizeof(info)); 1605 info.rti_info[RTAX_DST] = rt_key(rt); 1606 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt), rt_mask(rt), &ss); 1607 info.rti_info[RTAX_GATEWAY] = &nh->gw_sa; 1608 info.rti_flags = rt->rte_flags | nhop_get_rtflags(nh); 1609 info.rti_ifp = ifp; 1610 1611 return (rtsock_routemsg_info(cmd, &info, fibnum)); 1612 } 1613 1614 int 1615 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum) 1616 { 1617 struct rt_msghdr *rtm; 1618 struct sockaddr *sa; 1619 struct mbuf *m; 1620 1621 if (V_route_cb.any_count == 0) 1622 return (0); 1623 1624 if (info->rti_flags & RTF_HOST) 1625 info->rti_info[RTAX_NETMASK] = NULL; 1626 1627 m = rtsock_msg_mbuf(cmd, info); 1628 if (m == NULL) 1629 return (ENOBUFS); 1630 1631 if (fibnum != RT_ALL_FIBS) { 1632 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out " 1633 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs)); 1634 M_SETFIB(m, fibnum); 1635 m->m_flags |= RTS_FILTER_FIB; 1636 } 1637 1638 rtm = mtod(m, struct rt_msghdr *); 1639 rtm->rtm_addrs = info->rti_addrs; 1640 if (info->rti_ifp != NULL) 1641 rtm->rtm_index = info->rti_ifp->if_index; 1642 /* Add RTF_DONE to indicate command 'completion' required by API */ 1643 info->rti_flags |= RTF_DONE; 1644 /* Reported routes has to be up */ 1645 if (cmd == RTM_ADD || cmd == RTM_CHANGE) 1646 info->rti_flags |= RTF_UP; 1647 rtm->rtm_flags = info->rti_flags; 1648 1649 sa = info->rti_info[RTAX_DST]; 1650 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC); 1651 1652 return (0); 1653 } 1654 1655 /* 1656 * This is the analogue to the rt_newaddrmsg which performs the same 1657 * function but for multicast group memberhips. This is easier since 1658 * there is no route state to worry about. 1659 */ 1660 void 1661 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma) 1662 { 1663 struct rt_addrinfo info; 1664 struct mbuf *m = NULL; 1665 struct ifnet *ifp = ifma->ifma_ifp; 1666 struct ifma_msghdr *ifmam; 1667 1668 if (V_route_cb.any_count == 0) 1669 return; 1670 1671 bzero((caddr_t)&info, sizeof(info)); 1672 info.rti_info[RTAX_IFA] = ifma->ifma_addr; 1673 if (ifp && ifp->if_addr) 1674 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr; 1675 else 1676 info.rti_info[RTAX_IFP] = NULL; 1677 /* 1678 * If a link-layer address is present, present it as a ``gateway'' 1679 * (similarly to how ARP entries, e.g., are presented). 1680 */ 1681 info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr; 1682 m = rtsock_msg_mbuf(cmd, &info); 1683 if (m == NULL) 1684 return; 1685 ifmam = mtod(m, struct ifma_msghdr *); 1686 KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n", 1687 __func__)); 1688 ifmam->ifmam_index = ifp->if_index; 1689 ifmam->ifmam_addrs = info.rti_addrs; 1690 rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC); 1691 } 1692 1693 static struct mbuf * 1694 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what, 1695 struct rt_addrinfo *info) 1696 { 1697 struct if_announcemsghdr *ifan; 1698 struct mbuf *m; 1699 1700 if (V_route_cb.any_count == 0) 1701 return NULL; 1702 bzero((caddr_t)info, sizeof(*info)); 1703 m = rtsock_msg_mbuf(type, info); 1704 if (m != NULL) { 1705 ifan = mtod(m, struct if_announcemsghdr *); 1706 ifan->ifan_index = ifp->if_index; 1707 strlcpy(ifan->ifan_name, ifp->if_xname, 1708 sizeof(ifan->ifan_name)); 1709 ifan->ifan_what = what; 1710 } 1711 return m; 1712 } 1713 1714 /* 1715 * This is called to generate routing socket messages indicating 1716 * IEEE80211 wireless events. 1717 * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way. 1718 */ 1719 void 1720 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len) 1721 { 1722 struct mbuf *m; 1723 struct rt_addrinfo info; 1724 1725 m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info); 1726 if (m != NULL) { 1727 /* 1728 * Append the ieee80211 data. Try to stick it in the 1729 * mbuf containing the ifannounce msg; otherwise allocate 1730 * a new mbuf and append. 1731 * 1732 * NB: we assume m is a single mbuf. 1733 */ 1734 if (data_len > M_TRAILINGSPACE(m)) { 1735 struct mbuf *n = m_get(M_NOWAIT, MT_DATA); 1736 if (n == NULL) { 1737 m_freem(m); 1738 return; 1739 } 1740 bcopy(data, mtod(n, void *), data_len); 1741 n->m_len = data_len; 1742 m->m_next = n; 1743 } else if (data_len > 0) { 1744 bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len); 1745 m->m_len += data_len; 1746 } 1747 if (m->m_flags & M_PKTHDR) 1748 m->m_pkthdr.len += data_len; 1749 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len; 1750 rt_dispatch(m, AF_UNSPEC); 1751 } 1752 } 1753 1754 /* 1755 * This is called to generate routing socket messages indicating 1756 * network interface arrival and departure. 1757 */ 1758 void 1759 rt_ifannouncemsg(struct ifnet *ifp, int what) 1760 { 1761 struct mbuf *m; 1762 struct rt_addrinfo info; 1763 1764 m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info); 1765 if (m != NULL) 1766 rt_dispatch(m, AF_UNSPEC); 1767 } 1768 1769 static void 1770 rt_dispatch(struct mbuf *m, sa_family_t saf) 1771 { 1772 struct m_tag *tag; 1773 1774 /* 1775 * Preserve the family from the sockaddr, if any, in an m_tag for 1776 * use when injecting the mbuf into the routing socket buffer from 1777 * the netisr. 1778 */ 1779 if (saf != AF_UNSPEC) { 1780 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short), 1781 M_NOWAIT); 1782 if (tag == NULL) { 1783 m_freem(m); 1784 return; 1785 } 1786 *(unsigned short *)(tag + 1) = saf; 1787 m_tag_prepend(m, tag); 1788 } 1789 #ifdef VIMAGE 1790 if (V_loif) 1791 m->m_pkthdr.rcvif = V_loif; 1792 else { 1793 m_freem(m); 1794 return; 1795 } 1796 #endif 1797 netisr_queue(NETISR_ROUTE, m); /* mbuf is free'd on failure. */ 1798 } 1799 1800 /* 1801 * Checks if rte can be exported v.r.t jails/vnets. 1802 * 1803 * Returns 1 if it can, 0 otherwise. 1804 */ 1805 static bool 1806 can_export_rte(struct ucred *td_ucred, bool rt_is_host, 1807 const struct sockaddr *rt_dst) 1808 { 1809 1810 if ((!rt_is_host) ? jailed_without_vnet(td_ucred) 1811 : prison_if(td_ucred, rt_dst) != 0) 1812 return (false); 1813 return (true); 1814 } 1815 1816 1817 /* 1818 * This is used in dumping the kernel table via sysctl(). 1819 */ 1820 static int 1821 sysctl_dumpentry(struct rtentry *rt, void *vw) 1822 { 1823 struct walkarg *w = vw; 1824 struct nhop_object *nh; 1825 int error = 0; 1826 1827 NET_EPOCH_ASSERT(); 1828 1829 export_rtaddrs(rt, w->dst, w->mask); 1830 if (!can_export_rte(w->w_req->td->td_ucred, rt_is_host(rt), w->dst)) 1831 return (0); 1832 nh = rt_get_raw_nhop(rt); 1833 #ifdef ROUTE_MPATH 1834 if (NH_IS_NHGRP(nh)) { 1835 struct weightened_nhop *wn; 1836 uint32_t num_nhops; 1837 wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops); 1838 for (int i = 0; i < num_nhops; i++) { 1839 error = sysctl_dumpnhop(rt, wn[i].nh, wn[i].weight, w); 1840 if (error != 0) 1841 return (error); 1842 } 1843 } else 1844 #endif 1845 error = sysctl_dumpnhop(rt, nh, rt->rt_weight, w); 1846 1847 return (0); 1848 } 1849 1850 1851 static int 1852 sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, uint32_t weight, 1853 struct walkarg *w) 1854 { 1855 struct rt_addrinfo info; 1856 int error = 0, size; 1857 uint32_t rtflags; 1858 1859 rtflags = nhop_get_rtflags(nh); 1860 1861 if (w->w_op == NET_RT_FLAGS && !(rtflags & w->w_arg)) 1862 return (0); 1863 1864 bzero((caddr_t)&info, sizeof(info)); 1865 info.rti_info[RTAX_DST] = w->dst; 1866 info.rti_info[RTAX_GATEWAY] = &nh->gw_sa; 1867 info.rti_info[RTAX_NETMASK] = (rtflags & RTF_HOST) ? NULL : w->mask; 1868 info.rti_info[RTAX_GENMASK] = 0; 1869 if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) { 1870 info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr; 1871 info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr; 1872 if (nh->nh_ifp->if_flags & IFF_POINTOPOINT) 1873 info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr; 1874 } 1875 if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0) 1876 return (error); 1877 if (w->w_req && w->w_tmem) { 1878 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem; 1879 1880 bzero(&rtm->rtm_index, 1881 sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index)); 1882 1883 /* 1884 * rte flags may consist of RTF_HOST (duplicated in nhop rtflags) 1885 * and RTF_UP (if entry is linked, which is always true here). 1886 * Given that, use nhop rtflags & add RTF_UP. 1887 */ 1888 rtm->rtm_flags = rtflags | RTF_UP; 1889 if (rtm->rtm_flags & RTF_GWFLAG_COMPAT) 1890 rtm->rtm_flags = RTF_GATEWAY | 1891 (rtm->rtm_flags & ~RTF_GWFLAG_COMPAT); 1892 rt_getmetrics(rt, nh, &rtm->rtm_rmx); 1893 rtm->rtm_rmx.rmx_weight = weight; 1894 rtm->rtm_index = nh->nh_ifp->if_index; 1895 rtm->rtm_addrs = info.rti_addrs; 1896 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size); 1897 return (error); 1898 } 1899 return (error); 1900 } 1901 1902 static int 1903 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd, 1904 struct rt_addrinfo *info, struct walkarg *w, int len) 1905 { 1906 struct if_msghdrl *ifm; 1907 struct if_data *ifd; 1908 1909 ifm = (struct if_msghdrl *)w->w_tmem; 1910 1911 #ifdef COMPAT_FREEBSD32 1912 if (w->w_req->flags & SCTL_MASK32) { 1913 struct if_msghdrl32 *ifm32; 1914 1915 ifm32 = (struct if_msghdrl32 *)ifm; 1916 ifm32->ifm_addrs = info->rti_addrs; 1917 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1918 ifm32->ifm_index = ifp->if_index; 1919 ifm32->_ifm_spare1 = 0; 1920 ifm32->ifm_len = sizeof(*ifm32); 1921 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data); 1922 ifm32->_ifm_spare2 = 0; 1923 ifd = &ifm32->ifm_data; 1924 } else 1925 #endif 1926 { 1927 ifm->ifm_addrs = info->rti_addrs; 1928 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1929 ifm->ifm_index = ifp->if_index; 1930 ifm->_ifm_spare1 = 0; 1931 ifm->ifm_len = sizeof(*ifm); 1932 ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data); 1933 ifm->_ifm_spare2 = 0; 1934 ifd = &ifm->ifm_data; 1935 } 1936 1937 memcpy(ifd, src_ifd, sizeof(*ifd)); 1938 1939 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len)); 1940 } 1941 1942 static int 1943 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd, 1944 struct rt_addrinfo *info, struct walkarg *w, int len) 1945 { 1946 struct if_msghdr *ifm; 1947 struct if_data *ifd; 1948 1949 ifm = (struct if_msghdr *)w->w_tmem; 1950 1951 #ifdef COMPAT_FREEBSD32 1952 if (w->w_req->flags & SCTL_MASK32) { 1953 struct if_msghdr32 *ifm32; 1954 1955 ifm32 = (struct if_msghdr32 *)ifm; 1956 ifm32->ifm_addrs = info->rti_addrs; 1957 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1958 ifm32->ifm_index = ifp->if_index; 1959 ifm32->_ifm_spare1 = 0; 1960 ifd = &ifm32->ifm_data; 1961 } else 1962 #endif 1963 { 1964 ifm->ifm_addrs = info->rti_addrs; 1965 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags; 1966 ifm->ifm_index = ifp->if_index; 1967 ifm->_ifm_spare1 = 0; 1968 ifd = &ifm->ifm_data; 1969 } 1970 1971 memcpy(ifd, src_ifd, sizeof(*ifd)); 1972 1973 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len)); 1974 } 1975 1976 static int 1977 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info, 1978 struct walkarg *w, int len) 1979 { 1980 struct ifa_msghdrl *ifam; 1981 struct if_data *ifd; 1982 1983 ifam = (struct ifa_msghdrl *)w->w_tmem; 1984 1985 #ifdef COMPAT_FREEBSD32 1986 if (w->w_req->flags & SCTL_MASK32) { 1987 struct ifa_msghdrl32 *ifam32; 1988 1989 ifam32 = (struct ifa_msghdrl32 *)ifam; 1990 ifam32->ifam_addrs = info->rti_addrs; 1991 ifam32->ifam_flags = ifa->ifa_flags; 1992 ifam32->ifam_index = ifa->ifa_ifp->if_index; 1993 ifam32->_ifam_spare1 = 0; 1994 ifam32->ifam_len = sizeof(*ifam32); 1995 ifam32->ifam_data_off = 1996 offsetof(struct ifa_msghdrl32, ifam_data); 1997 ifam32->ifam_metric = ifa->ifa_ifp->if_metric; 1998 ifd = &ifam32->ifam_data; 1999 } else 2000 #endif 2001 { 2002 ifam->ifam_addrs = info->rti_addrs; 2003 ifam->ifam_flags = ifa->ifa_flags; 2004 ifam->ifam_index = ifa->ifa_ifp->if_index; 2005 ifam->_ifam_spare1 = 0; 2006 ifam->ifam_len = sizeof(*ifam); 2007 ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data); 2008 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 2009 ifd = &ifam->ifam_data; 2010 } 2011 2012 bzero(ifd, sizeof(*ifd)); 2013 ifd->ifi_datalen = sizeof(struct if_data); 2014 ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets); 2015 ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets); 2016 ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes); 2017 ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes); 2018 2019 /* Fixup if_data carp(4) vhid. */ 2020 if (carp_get_vhid_p != NULL) 2021 ifd->ifi_vhid = (*carp_get_vhid_p)(ifa); 2022 2023 return (SYSCTL_OUT(w->w_req, w->w_tmem, len)); 2024 } 2025 2026 static int 2027 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info, 2028 struct walkarg *w, int len) 2029 { 2030 struct ifa_msghdr *ifam; 2031 2032 ifam = (struct ifa_msghdr *)w->w_tmem; 2033 ifam->ifam_addrs = info->rti_addrs; 2034 ifam->ifam_flags = ifa->ifa_flags; 2035 ifam->ifam_index = ifa->ifa_ifp->if_index; 2036 ifam->_ifam_spare1 = 0; 2037 ifam->ifam_metric = ifa->ifa_ifp->if_metric; 2038 2039 return (SYSCTL_OUT(w->w_req, w->w_tmem, len)); 2040 } 2041 2042 static int 2043 sysctl_iflist(int af, struct walkarg *w) 2044 { 2045 struct ifnet *ifp; 2046 struct ifaddr *ifa; 2047 struct if_data ifd; 2048 struct rt_addrinfo info; 2049 int len, error = 0; 2050 struct sockaddr_storage ss; 2051 2052 bzero((caddr_t)&info, sizeof(info)); 2053 bzero(&ifd, sizeof(ifd)); 2054 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2055 if (w->w_arg && w->w_arg != ifp->if_index) 2056 continue; 2057 if_data_copy(ifp, &ifd); 2058 ifa = ifp->if_addr; 2059 info.rti_info[RTAX_IFP] = ifa->ifa_addr; 2060 error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len); 2061 if (error != 0) 2062 goto done; 2063 info.rti_info[RTAX_IFP] = NULL; 2064 if (w->w_req && w->w_tmem) { 2065 if (w->w_op == NET_RT_IFLISTL) 2066 error = sysctl_iflist_ifml(ifp, &ifd, &info, w, 2067 len); 2068 else 2069 error = sysctl_iflist_ifm(ifp, &ifd, &info, w, 2070 len); 2071 if (error) 2072 goto done; 2073 } 2074 while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) { 2075 if (af && af != ifa->ifa_addr->sa_family) 2076 continue; 2077 if (prison_if(w->w_req->td->td_ucred, 2078 ifa->ifa_addr) != 0) 2079 continue; 2080 info.rti_info[RTAX_IFA] = ifa->ifa_addr; 2081 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask( 2082 ifa->ifa_addr, ifa->ifa_netmask, &ss); 2083 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr; 2084 error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len); 2085 if (error != 0) 2086 goto done; 2087 if (w->w_req && w->w_tmem) { 2088 if (w->w_op == NET_RT_IFLISTL) 2089 error = sysctl_iflist_ifaml(ifa, &info, 2090 w, len); 2091 else 2092 error = sysctl_iflist_ifam(ifa, &info, 2093 w, len); 2094 if (error) 2095 goto done; 2096 } 2097 } 2098 info.rti_info[RTAX_IFA] = NULL; 2099 info.rti_info[RTAX_NETMASK] = NULL; 2100 info.rti_info[RTAX_BRD] = NULL; 2101 } 2102 done: 2103 return (error); 2104 } 2105 2106 static int 2107 sysctl_ifmalist(int af, struct walkarg *w) 2108 { 2109 struct rt_addrinfo info; 2110 struct ifaddr *ifa; 2111 struct ifmultiaddr *ifma; 2112 struct ifnet *ifp; 2113 int error, len; 2114 2115 NET_EPOCH_ASSERT(); 2116 2117 error = 0; 2118 bzero((caddr_t)&info, sizeof(info)); 2119 2120 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2121 if (w->w_arg && w->w_arg != ifp->if_index) 2122 continue; 2123 ifa = ifp->if_addr; 2124 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL; 2125 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 2126 if (af && af != ifma->ifma_addr->sa_family) 2127 continue; 2128 if (prison_if(w->w_req->td->td_ucred, 2129 ifma->ifma_addr) != 0) 2130 continue; 2131 info.rti_info[RTAX_IFA] = ifma->ifma_addr; 2132 info.rti_info[RTAX_GATEWAY] = 2133 (ifma->ifma_addr->sa_family != AF_LINK) ? 2134 ifma->ifma_lladdr : NULL; 2135 error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len); 2136 if (error != 0) 2137 break; 2138 if (w->w_req && w->w_tmem) { 2139 struct ifma_msghdr *ifmam; 2140 2141 ifmam = (struct ifma_msghdr *)w->w_tmem; 2142 ifmam->ifmam_index = ifma->ifma_ifp->if_index; 2143 ifmam->ifmam_flags = 0; 2144 ifmam->ifmam_addrs = info.rti_addrs; 2145 ifmam->_ifmam_spare1 = 0; 2146 error = SYSCTL_OUT(w->w_req, w->w_tmem, len); 2147 if (error != 0) 2148 break; 2149 } 2150 } 2151 if (error != 0) 2152 break; 2153 } 2154 return (error); 2155 } 2156 2157 static void 2158 rtable_sysctl_dump(uint32_t fibnum, int family, struct walkarg *w) 2159 { 2160 union sockaddr_union sa_dst, sa_mask; 2161 2162 w->family = family; 2163 w->dst = (struct sockaddr *)&sa_dst; 2164 w->mask = (struct sockaddr *)&sa_mask; 2165 2166 init_sockaddrs_family(family, w->dst, w->mask); 2167 2168 rib_walk(fibnum, family, false, sysctl_dumpentry, w); 2169 } 2170 2171 static int 2172 sysctl_rtsock(SYSCTL_HANDLER_ARGS) 2173 { 2174 struct epoch_tracker et; 2175 int *name = (int *)arg1; 2176 u_int namelen = arg2; 2177 struct rib_head *rnh = NULL; /* silence compiler. */ 2178 int i, lim, error = EINVAL; 2179 int fib = 0; 2180 u_char af; 2181 struct walkarg w; 2182 2183 name ++; 2184 namelen--; 2185 if (req->newptr) 2186 return (EPERM); 2187 if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP || name[1] == NET_RT_NHGRP) { 2188 if (namelen == 3) 2189 fib = req->td->td_proc->p_fibnum; 2190 else if (namelen == 4) 2191 fib = (name[3] == RT_ALL_FIBS) ? 2192 req->td->td_proc->p_fibnum : name[3]; 2193 else 2194 return ((namelen < 3) ? EISDIR : ENOTDIR); 2195 if (fib < 0 || fib >= rt_numfibs) 2196 return (EINVAL); 2197 } else if (namelen != 3) 2198 return ((namelen < 3) ? EISDIR : ENOTDIR); 2199 af = name[0]; 2200 if (af > AF_MAX) 2201 return (EINVAL); 2202 bzero(&w, sizeof(w)); 2203 w.w_op = name[1]; 2204 w.w_arg = name[2]; 2205 w.w_req = req; 2206 2207 error = sysctl_wire_old_buffer(req, 0); 2208 if (error) 2209 return (error); 2210 2211 /* 2212 * Allocate reply buffer in advance. 2213 * All rtsock messages has maximum length of u_short. 2214 */ 2215 w.w_tmemsize = 65536; 2216 w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK); 2217 2218 NET_EPOCH_ENTER(et); 2219 switch (w.w_op) { 2220 case NET_RT_DUMP: 2221 case NET_RT_FLAGS: 2222 if (af == 0) { /* dump all tables */ 2223 i = 1; 2224 lim = AF_MAX; 2225 } else /* dump only one table */ 2226 i = lim = af; 2227 2228 /* 2229 * take care of llinfo entries, the caller must 2230 * specify an AF 2231 */ 2232 if (w.w_op == NET_RT_FLAGS && 2233 (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) { 2234 if (af != 0) 2235 error = lltable_sysctl_dumparp(af, w.w_req); 2236 else 2237 error = EINVAL; 2238 break; 2239 } 2240 /* 2241 * take care of routing entries 2242 */ 2243 for (error = 0; error == 0 && i <= lim; i++) { 2244 rnh = rt_tables_get_rnh(fib, i); 2245 if (rnh != NULL) { 2246 rtable_sysctl_dump(fib, i, &w); 2247 } else if (af != 0) 2248 error = EAFNOSUPPORT; 2249 } 2250 break; 2251 case NET_RT_NHOP: 2252 case NET_RT_NHGRP: 2253 /* Allow dumping one specific af/fib at a time */ 2254 if (namelen < 4) { 2255 error = EINVAL; 2256 break; 2257 } 2258 fib = name[3]; 2259 if (fib < 0 || fib > rt_numfibs) { 2260 error = EINVAL; 2261 break; 2262 } 2263 rnh = rt_tables_get_rnh(fib, af); 2264 if (rnh == NULL) { 2265 error = EAFNOSUPPORT; 2266 break; 2267 } 2268 if (w.w_op == NET_RT_NHOP) 2269 error = nhops_dump_sysctl(rnh, w.w_req); 2270 else 2271 #ifdef ROUTE_MPATH 2272 error = nhgrp_dump_sysctl(rnh, w.w_req); 2273 #else 2274 error = ENOTSUP; 2275 #endif 2276 break; 2277 case NET_RT_IFLIST: 2278 case NET_RT_IFLISTL: 2279 error = sysctl_iflist(af, &w); 2280 break; 2281 2282 case NET_RT_IFMALIST: 2283 error = sysctl_ifmalist(af, &w); 2284 break; 2285 } 2286 NET_EPOCH_EXIT(et); 2287 2288 free(w.w_tmem, M_TEMP); 2289 return (error); 2290 } 2291 2292 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE, 2293 sysctl_rtsock, "Return route tables and interface/address lists"); 2294 2295 /* 2296 * Definitions of protocols supported in the ROUTE domain. 2297 */ 2298 2299 static struct domain routedomain; /* or at least forward */ 2300 2301 static struct protosw routesw[] = { 2302 { 2303 .pr_type = SOCK_RAW, 2304 .pr_domain = &routedomain, 2305 .pr_flags = PR_ATOMIC|PR_ADDR, 2306 .pr_output = route_output, 2307 .pr_ctlinput = raw_ctlinput, 2308 .pr_init = raw_init, 2309 .pr_usrreqs = &route_usrreqs 2310 } 2311 }; 2312 2313 static struct domain routedomain = { 2314 .dom_family = PF_ROUTE, 2315 .dom_name = "route", 2316 .dom_protosw = routesw, 2317 .dom_protoswNPROTOSW = &routesw[nitems(routesw)] 2318 }; 2319 2320 VNET_DOMAIN_SET(route); 2321