1 /* $OpenBSD: route.c,v 1.386 2019/06/21 17:11:42 mpi Exp $ */ 2 /* $NetBSD: route.c,v 1.14 1996/02/13 22:00:46 christos Exp $ */ 3 4 /* 5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the project nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33 /* 34 * Copyright (c) 1980, 1986, 1991, 1993 35 * The Regents of the University of California. All rights reserved. 36 * 37 * Redistribution and use in source and binary forms, with or without 38 * modification, are permitted provided that the following conditions 39 * are met: 40 * 1. Redistributions of source code must retain the above copyright 41 * notice, this list of conditions and the following disclaimer. 42 * 2. Redistributions in binary form must reproduce the above copyright 43 * notice, this list of conditions and the following disclaimer in the 44 * documentation and/or other materials provided with the distribution. 45 * 3. Neither the name of the University nor the names of its contributors 46 * may be used to endorse or promote products derived from this software 47 * without specific prior written permission. 48 * 49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 59 * SUCH DAMAGE. 60 * 61 * @(#)route.c 8.2 (Berkeley) 11/15/93 62 */ 63 64 /* 65 * @(#)COPYRIGHT 1.1 (NRL) 17 January 1995 66 * 67 * NRL grants permission for redistribution and use in source and binary 68 * forms, with or without modification, of the software and documentation 69 * created at NRL provided that the following conditions are met: 70 * 71 * 1. Redistributions of source code must retain the above copyright 72 * notice, this list of conditions and the following disclaimer. 73 * 2. Redistributions in binary form must reproduce the above copyright 74 * notice, this list of conditions and the following disclaimer in the 75 * documentation and/or other materials provided with the distribution. 76 * 3. All advertising materials mentioning features or use of this software 77 * must display the following acknowledgements: 78 * This product includes software developed by the University of 79 * California, Berkeley and its contributors. 80 * This product includes software developed at the Information 81 * Technology Division, US Naval Research Laboratory. 82 * 4. Neither the name of the NRL nor the names of its contributors 83 * may be used to endorse or promote products derived from this software 84 * without specific prior written permission. 85 * 86 * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS 87 * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 88 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A 89 * PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL NRL OR 90 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, 91 * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 92 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 93 * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF 94 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING 95 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 96 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 97 * 98 * The views and conclusions contained in the software and documentation 99 * are those of the authors and should not be interpreted as representing 100 * official policies, either expressed or implied, of the US Naval 101 * Research Laboratory (NRL). 102 */ 103 104 #include <sys/param.h> 105 #include <sys/systm.h> 106 #include <sys/mbuf.h> 107 #include <sys/socket.h> 108 #include <sys/socketvar.h> 109 #include <sys/timeout.h> 110 #include <sys/domain.h> 111 #include <sys/protosw.h> 112 #include <sys/ioctl.h> 113 #include <sys/kernel.h> 114 #include <sys/queue.h> 115 #include <sys/pool.h> 116 #include <sys/atomic.h> 117 118 #include <net/if.h> 119 #include <net/if_var.h> 120 #include <net/if_dl.h> 121 #include <net/route.h> 122 123 #include <netinet/in.h> 124 #include <netinet/ip_var.h> 125 #include <netinet/in_var.h> 126 127 #ifdef INET6 128 #include <netinet/ip6.h> 129 #include <netinet6/ip6_var.h> 130 #include <netinet6/in6_var.h> 131 #endif 132 133 #ifdef MPLS 134 #include <netmpls/mpls.h> 135 #endif 136 137 #ifdef BFD 138 #include <net/bfd.h> 139 #endif 140 141 #define ROUNDUP(a) (a>0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long)) 142 143 /* Give some jitter to hash, to avoid synchronization between routers. */ 144 static uint32_t rt_hashjitter; 145 146 extern unsigned int rtmap_limit; 147 148 struct cpumem * rtcounters; 149 int rttrash; /* routes not in table but not freed */ 150 int ifatrash; /* ifas not in ifp list but not free */ 151 152 struct pool rtentry_pool; /* pool for rtentry structures */ 153 struct pool rttimer_pool; /* pool for rttimer structures */ 154 155 void rt_timer_init(void); 156 int rt_setgwroute(struct rtentry *, u_int); 157 void rt_putgwroute(struct rtentry *); 158 int rtflushclone1(struct rtentry *, void *, u_int); 159 int rtflushclone(struct ifnet *ifp, struct rtentry *, unsigned int); 160 int rt_ifa_purge_walker(struct rtentry *, void *, unsigned int); 161 struct rtentry *rt_match(struct sockaddr *, uint32_t *, int, unsigned int); 162 int rt_clone(struct rtentry **, struct sockaddr *, unsigned int); 163 struct sockaddr *rt_plentosa(sa_family_t, int, struct sockaddr_in6 *); 164 static int rt_copysa(struct sockaddr *, struct sockaddr *, struct sockaddr **); 165 166 #ifdef DDB 167 void db_print_sa(struct sockaddr *); 168 void db_print_ifa(struct ifaddr *); 169 int db_show_rtentry(struct rtentry *, void *, unsigned int); 170 #endif 171 172 #define LABELID_MAX 50000 173 174 struct rt_label { 175 TAILQ_ENTRY(rt_label) rtl_entry; 176 char rtl_name[RTLABEL_LEN]; 177 u_int16_t rtl_id; 178 int rtl_ref; 179 }; 180 181 TAILQ_HEAD(rt_labels, rt_label) rt_labels = TAILQ_HEAD_INITIALIZER(rt_labels); 182 183 void 184 route_init(void) 185 { 186 rtcounters = counters_alloc(rts_ncounters); 187 188 pool_init(&rtentry_pool, sizeof(struct rtentry), 0, IPL_SOFTNET, 0, 189 "rtentry", NULL); 190 191 while (rt_hashjitter == 0) 192 rt_hashjitter = arc4random(); 193 194 #ifdef BFD 195 bfdinit(); 196 #endif 197 } 198 199 /* 200 * Returns 1 if the (cached) ``rt'' entry is still valid, 0 otherwise. 201 */ 202 int 203 rtisvalid(struct rtentry *rt) 204 { 205 if (rt == NULL) 206 return (0); 207 208 if (!ISSET(rt->rt_flags, RTF_UP)) 209 return (0); 210 211 if (ISSET(rt->rt_flags, RTF_GATEWAY)) { 212 KASSERT(rt->rt_gwroute != NULL); 213 KASSERT(!ISSET(rt->rt_gwroute->rt_flags, RTF_GATEWAY)); 214 if (!ISSET(rt->rt_gwroute->rt_flags, RTF_UP)) 215 return (0); 216 } 217 218 return (1); 219 } 220 221 /* 222 * Do the actual lookup for rtalloc(9), do not use directly! 223 * 224 * Return the best matching entry for the destination ``dst''. 225 * 226 * "RT_RESOLVE" means that a corresponding L2 entry should 227 * be added to the routing table and resolved (via ARP or 228 * NDP), if it does not exist. 229 */ 230 struct rtentry * 231 rt_match(struct sockaddr *dst, uint32_t *src, int flags, unsigned int tableid) 232 { 233 struct rtentry *rt = NULL; 234 235 rt = rtable_match(tableid, dst, src); 236 if (rt == NULL) { 237 rtstat_inc(rts_unreach); 238 return (NULL); 239 } 240 241 if (ISSET(rt->rt_flags, RTF_CLONING) && ISSET(flags, RT_RESOLVE)) 242 rt_clone(&rt, dst, tableid); 243 244 rt->rt_use++; 245 return (rt); 246 } 247 248 int 249 rt_clone(struct rtentry **rtp, struct sockaddr *dst, unsigned int rtableid) 250 { 251 struct rt_addrinfo info; 252 struct rtentry *rt = *rtp; 253 int error = 0; 254 255 memset(&info, 0, sizeof(info)); 256 info.rti_info[RTAX_DST] = dst; 257 258 /* 259 * The priority of cloned route should be different 260 * to avoid conflict with /32 cloning routes. 261 * 262 * It should also be higher to let the ARP layer find 263 * cloned routes instead of the cloning one. 264 */ 265 KERNEL_LOCK(); 266 error = rtrequest(RTM_RESOLVE, &info, rt->rt_priority - 1, &rt, 267 rtableid); 268 KERNEL_UNLOCK(); 269 if (error) { 270 rtm_miss(RTM_MISS, &info, 0, RTP_NONE, 0, error, rtableid); 271 } else { 272 /* Inform listeners of the new route */ 273 rtm_send(rt, RTM_ADD, 0, rtableid); 274 rtfree(*rtp); 275 *rtp = rt; 276 } 277 return (error); 278 } 279 280 /* 281 * Originated from bridge_hash() in if_bridge.c 282 */ 283 #define mix(a, b, c) do { \ 284 a -= b; a -= c; a ^= (c >> 13); \ 285 b -= c; b -= a; b ^= (a << 8); \ 286 c -= a; c -= b; c ^= (b >> 13); \ 287 a -= b; a -= c; a ^= (c >> 12); \ 288 b -= c; b -= a; b ^= (a << 16); \ 289 c -= a; c -= b; c ^= (b >> 5); \ 290 a -= b; a -= c; a ^= (c >> 3); \ 291 b -= c; b -= a; b ^= (a << 10); \ 292 c -= a; c -= b; c ^= (b >> 15); \ 293 } while (0) 294 295 int 296 rt_hash(struct rtentry *rt, struct sockaddr *dst, uint32_t *src) 297 { 298 uint32_t a, b, c; 299 300 if (src == NULL || !rtisvalid(rt) || !ISSET(rt->rt_flags, RTF_MPATH)) 301 return (-1); 302 303 a = b = 0x9e3779b9; 304 c = rt_hashjitter; 305 306 switch (dst->sa_family) { 307 case AF_INET: 308 { 309 struct sockaddr_in *sin; 310 311 if (!ipmultipath) 312 return (-1); 313 314 sin = satosin(dst); 315 a += sin->sin_addr.s_addr; 316 b += (src != NULL) ? src[0] : 0; 317 mix(a, b, c); 318 break; 319 } 320 #ifdef INET6 321 case AF_INET6: 322 { 323 struct sockaddr_in6 *sin6; 324 325 if (!ip6_multipath) 326 return (-1); 327 328 sin6 = satosin6(dst); 329 a += sin6->sin6_addr.s6_addr32[0]; 330 b += sin6->sin6_addr.s6_addr32[2]; 331 c += (src != NULL) ? src[0] : 0; 332 mix(a, b, c); 333 a += sin6->sin6_addr.s6_addr32[1]; 334 b += sin6->sin6_addr.s6_addr32[3]; 335 c += (src != NULL) ? src[1] : 0; 336 mix(a, b, c); 337 a += sin6->sin6_addr.s6_addr32[2]; 338 b += sin6->sin6_addr.s6_addr32[1]; 339 c += (src != NULL) ? src[2] : 0; 340 mix(a, b, c); 341 a += sin6->sin6_addr.s6_addr32[3]; 342 b += sin6->sin6_addr.s6_addr32[0]; 343 c += (src != NULL) ? src[3] : 0; 344 mix(a, b, c); 345 break; 346 } 347 #endif /* INET6 */ 348 } 349 350 return (c & 0xffff); 351 } 352 353 /* 354 * Allocate a route, potentially using multipath to select the peer. 355 */ 356 struct rtentry * 357 rtalloc_mpath(struct sockaddr *dst, uint32_t *src, unsigned int rtableid) 358 { 359 return (rt_match(dst, src, RT_RESOLVE, rtableid)); 360 } 361 362 /* 363 * Look in the routing table for the best matching entry for 364 * ``dst''. 365 * 366 * If a route with a gateway is found and its next hop is no 367 * longer valid, try to cache it. 368 */ 369 struct rtentry * 370 rtalloc(struct sockaddr *dst, int flags, unsigned int rtableid) 371 { 372 return (rt_match(dst, NULL, flags, rtableid)); 373 } 374 375 /* 376 * Cache the route entry corresponding to a reachable next hop in 377 * the gateway entry ``rt''. 378 */ 379 int 380 rt_setgwroute(struct rtentry *rt, u_int rtableid) 381 { 382 struct rtentry *prt, *nhrt; 383 unsigned int rdomain = rtable_l2(rtableid); 384 int error; 385 386 NET_ASSERT_LOCKED(); 387 388 KASSERT(ISSET(rt->rt_flags, RTF_GATEWAY)); 389 390 /* If we cannot find a valid next hop bail. */ 391 nhrt = rt_match(rt->rt_gateway, NULL, RT_RESOLVE, rdomain); 392 if (nhrt == NULL) 393 return (ENOENT); 394 395 /* Next hop entry must be on the same interface. */ 396 if (nhrt->rt_ifidx != rt->rt_ifidx) { 397 struct sockaddr_in6 sa_mask; 398 399 if (!ISSET(nhrt->rt_flags, RTF_LLINFO) || 400 !ISSET(nhrt->rt_flags, RTF_CLONED)) { 401 rtfree(nhrt); 402 return (EHOSTUNREACH); 403 } 404 405 /* 406 * We found a L2 entry, so we might have multiple 407 * RTF_CLONING routes for the same subnet. Query 408 * the first route of the multipath chain and iterate 409 * until we find the correct one. 410 */ 411 prt = rtable_lookup(rdomain, rt_key(nhrt->rt_parent), 412 rt_plen2mask(nhrt->rt_parent, &sa_mask), NULL, RTP_ANY); 413 rtfree(nhrt); 414 415 while (prt != NULL && prt->rt_ifidx != rt->rt_ifidx) 416 prt = rtable_iterate(prt); 417 418 /* We found nothing or a non-cloning MPATH route. */ 419 if (prt == NULL || !ISSET(prt->rt_flags, RTF_CLONING)) { 420 rtfree(prt); 421 return (EHOSTUNREACH); 422 } 423 424 error = rt_clone(&prt, rt->rt_gateway, rdomain); 425 if (error) { 426 rtfree(prt); 427 return (error); 428 } 429 nhrt = prt; 430 } 431 432 /* 433 * Next hop must be reachable, this also prevents rtentry 434 * loops for example when rt->rt_gwroute points to rt. 435 */ 436 if (ISSET(nhrt->rt_flags, RTF_CLONING|RTF_GATEWAY)) { 437 rtfree(nhrt); 438 return (ENETUNREACH); 439 } 440 441 /* Next hop is valid so remove possible old cache. */ 442 rt_putgwroute(rt); 443 KASSERT(rt->rt_gwroute == NULL); 444 445 /* 446 * If the MTU of next hop is 0, this will reset the MTU of the 447 * route to run PMTUD again from scratch. 448 */ 449 if (!ISSET(rt->rt_locks, RTV_MTU) && (rt->rt_mtu > nhrt->rt_mtu)) 450 rt->rt_mtu = nhrt->rt_mtu; 451 452 /* 453 * To avoid reference counting problems when writting link-layer 454 * addresses in an outgoing packet, we ensure that the lifetime 455 * of a cached entry is greater that the bigger lifetime of the 456 * gateway entries it is pointed by. 457 */ 458 nhrt->rt_flags |= RTF_CACHED; 459 nhrt->rt_cachecnt++; 460 461 rt->rt_gwroute = nhrt; 462 463 return (0); 464 } 465 466 /* 467 * Invalidate the cached route entry of the gateway entry ``rt''. 468 */ 469 void 470 rt_putgwroute(struct rtentry *rt) 471 { 472 struct rtentry *nhrt = rt->rt_gwroute; 473 474 NET_ASSERT_LOCKED(); 475 476 if (!ISSET(rt->rt_flags, RTF_GATEWAY) || nhrt == NULL) 477 return; 478 479 KASSERT(ISSET(nhrt->rt_flags, RTF_CACHED)); 480 KASSERT(nhrt->rt_cachecnt > 0); 481 482 --nhrt->rt_cachecnt; 483 if (nhrt->rt_cachecnt == 0) 484 nhrt->rt_flags &= ~RTF_CACHED; 485 486 rtfree(rt->rt_gwroute); 487 rt->rt_gwroute = NULL; 488 } 489 490 void 491 rtref(struct rtentry *rt) 492 { 493 atomic_inc_int(&rt->rt_refcnt); 494 } 495 496 void 497 rtfree(struct rtentry *rt) 498 { 499 int refcnt; 500 501 if (rt == NULL) 502 return; 503 504 refcnt = (int)atomic_dec_int_nv(&rt->rt_refcnt); 505 if (refcnt <= 0) { 506 KASSERT(!ISSET(rt->rt_flags, RTF_UP)); 507 KASSERT(!RT_ROOT(rt)); 508 atomic_dec_int(&rttrash); 509 if (refcnt < 0) { 510 printf("rtfree: %p not freed (neg refs)\n", rt); 511 return; 512 } 513 514 KERNEL_LOCK(); 515 rt_timer_remove_all(rt); 516 ifafree(rt->rt_ifa); 517 rtlabel_unref(rt->rt_labelid); 518 #ifdef MPLS 519 rt_mpls_clear(rt); 520 #endif 521 free(rt->rt_gateway, M_RTABLE, ROUNDUP(rt->rt_gateway->sa_len)); 522 free(rt_key(rt), M_RTABLE, rt_key(rt)->sa_len); 523 KERNEL_UNLOCK(); 524 525 pool_put(&rtentry_pool, rt); 526 } 527 } 528 529 void 530 ifafree(struct ifaddr *ifa) 531 { 532 if (ifa == NULL) 533 panic("ifafree"); 534 if (ifa->ifa_refcnt == 0) { 535 ifatrash--; 536 free(ifa, M_IFADDR, 0); 537 } else 538 ifa->ifa_refcnt--; 539 } 540 541 /* 542 * Force a routing table entry to the specified 543 * destination to go through the given gateway. 544 * Normally called as a result of a routing redirect 545 * message from the network layer. 546 */ 547 void 548 rtredirect(struct sockaddr *dst, struct sockaddr *gateway, 549 struct sockaddr *src, struct rtentry **rtp, unsigned int rdomain) 550 { 551 struct rtentry *rt; 552 int error = 0; 553 enum rtstat_counters stat = rts_ncounters; 554 struct rt_addrinfo info; 555 struct ifaddr *ifa; 556 unsigned int ifidx = 0; 557 int flags = RTF_GATEWAY|RTF_HOST; 558 uint8_t prio = RTP_NONE; 559 560 NET_ASSERT_LOCKED(); 561 562 /* verify the gateway is directly reachable */ 563 rt = rtalloc(gateway, 0, rdomain); 564 if (!rtisvalid(rt) || ISSET(rt->rt_flags, RTF_GATEWAY)) { 565 rtfree(rt); 566 error = ENETUNREACH; 567 goto out; 568 } 569 ifidx = rt->rt_ifidx; 570 ifa = rt->rt_ifa; 571 rtfree(rt); 572 rt = NULL; 573 574 rt = rtable_lookup(rdomain, dst, NULL, NULL, RTP_ANY); 575 /* 576 * If the redirect isn't from our current router for this dst, 577 * it's either old or wrong. If it redirects us to ourselves, 578 * we have a routing loop, perhaps as a result of an interface 579 * going down recently. 580 */ 581 #define equal(a1, a2) \ 582 ((a1)->sa_len == (a2)->sa_len && \ 583 bcmp((caddr_t)(a1), (caddr_t)(a2), (a1)->sa_len) == 0) 584 if (rt != NULL && (!equal(src, rt->rt_gateway) || rt->rt_ifa != ifa)) 585 error = EINVAL; 586 else if (ifa_ifwithaddr(gateway, rdomain) != NULL || 587 (gateway->sa_family = AF_INET && 588 in_broadcast(satosin(gateway)->sin_addr, rdomain))) 589 error = EHOSTUNREACH; 590 if (error) 591 goto done; 592 /* 593 * Create a new entry if we just got back a wildcard entry 594 * or the lookup failed. This is necessary for hosts 595 * which use routing redirects generated by smart gateways 596 * to dynamically build the routing tables. 597 */ 598 if (rt == NULL) 599 goto create; 600 /* 601 * Don't listen to the redirect if it's 602 * for a route to an interface. 603 */ 604 if (ISSET(rt->rt_flags, RTF_GATEWAY)) { 605 if (!ISSET(rt->rt_flags, RTF_HOST)) { 606 /* 607 * Changing from route to net => route to host. 608 * Create new route, rather than smashing route to net. 609 */ 610 create: 611 rtfree(rt); 612 flags |= RTF_DYNAMIC; 613 bzero(&info, sizeof(info)); 614 info.rti_info[RTAX_DST] = dst; 615 info.rti_info[RTAX_GATEWAY] = gateway; 616 info.rti_ifa = ifa; 617 info.rti_flags = flags; 618 rt = NULL; 619 error = rtrequest(RTM_ADD, &info, RTP_DEFAULT, &rt, 620 rdomain); 621 if (error == 0) { 622 flags = rt->rt_flags; 623 prio = rt->rt_priority; 624 } 625 stat = rts_dynamic; 626 } else { 627 /* 628 * Smash the current notion of the gateway to 629 * this destination. Should check about netmask!!! 630 */ 631 rt->rt_flags |= RTF_MODIFIED; 632 flags |= RTF_MODIFIED; 633 prio = rt->rt_priority; 634 stat = rts_newgateway; 635 rt_setgate(rt, gateway, rdomain); 636 } 637 } else 638 error = EHOSTUNREACH; 639 done: 640 if (rt) { 641 if (rtp && !error) 642 *rtp = rt; 643 else 644 rtfree(rt); 645 } 646 out: 647 if (error) 648 rtstat_inc(rts_badredirect); 649 else if (stat != rts_ncounters) 650 rtstat_inc(stat); 651 bzero((caddr_t)&info, sizeof(info)); 652 info.rti_info[RTAX_DST] = dst; 653 info.rti_info[RTAX_GATEWAY] = gateway; 654 info.rti_info[RTAX_AUTHOR] = src; 655 rtm_miss(RTM_REDIRECT, &info, flags, prio, ifidx, error, rdomain); 656 } 657 658 /* 659 * Delete a route and generate a message 660 */ 661 int 662 rtdeletemsg(struct rtentry *rt, struct ifnet *ifp, u_int tableid) 663 { 664 int error; 665 struct rt_addrinfo info; 666 struct sockaddr_in6 sa_mask; 667 668 KASSERT(rt->rt_ifidx == ifp->if_index); 669 670 /* 671 * Request the new route so that the entry is not actually 672 * deleted. That will allow the information being reported to 673 * be accurate (and consistent with route_output()). 674 */ 675 memset(&info, 0, sizeof(info)); 676 info.rti_info[RTAX_DST] = rt_key(rt); 677 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 678 if (!ISSET(rt->rt_flags, RTF_HOST)) 679 info.rti_info[RTAX_NETMASK] = rt_plen2mask(rt, &sa_mask); 680 error = rtrequest_delete(&info, rt->rt_priority, ifp, &rt, tableid); 681 rtm_send(rt, RTM_DELETE, error, tableid); 682 if (error == 0) 683 rtfree(rt); 684 return (error); 685 } 686 687 static inline int 688 rtequal(struct rtentry *a, struct rtentry *b) 689 { 690 if (a == b) 691 return 1; 692 693 if (memcmp(rt_key(a), rt_key(b), rt_key(a)->sa_len) == 0 && 694 rt_plen(a) == rt_plen(b)) 695 return 1; 696 else 697 return 0; 698 } 699 700 int 701 rtflushclone1(struct rtentry *rt, void *arg, u_int id) 702 { 703 struct rtentry *cloningrt = arg; 704 struct ifnet *ifp; 705 706 if (!ISSET(rt->rt_flags, RTF_CLONED)) 707 return 0; 708 709 /* Cached route must stay alive as long as their parent are alive. */ 710 if (ISSET(rt->rt_flags, RTF_CACHED) && (rt->rt_parent != cloningrt)) 711 return 0; 712 713 if (!rtequal(rt->rt_parent, cloningrt)) 714 return 0; 715 /* 716 * This happens when an interface with a RTF_CLONING route is 717 * being detached. In this case it's safe to bail because all 718 * the routes are being purged by rt_ifa_purge(). 719 */ 720 ifp = if_get(rt->rt_ifidx); 721 if (ifp == NULL) 722 return 0; 723 724 if_put(ifp); 725 return EEXIST; 726 } 727 728 int 729 rtflushclone(struct ifnet *ifp, struct rtentry *parent, unsigned int rtableid) 730 { 731 struct rtentry *rt = NULL; 732 int error; 733 734 #ifdef DIAGNOSTIC 735 if (!parent || (parent->rt_flags & RTF_CLONING) == 0) 736 panic("rtflushclone: called with a non-cloning route"); 737 #endif 738 739 do { 740 error = rtable_walk(rtableid, rt_key(parent)->sa_family, &rt, 741 rtflushclone1, parent); 742 if (rt != NULL && error == EEXIST) { 743 error = rtdeletemsg(rt, ifp, rtableid); 744 if (error == 0) 745 error = EAGAIN; 746 } 747 rtfree(rt); 748 rt = NULL; 749 } while (error == EAGAIN); 750 751 return error; 752 753 } 754 755 int 756 rtrequest_delete(struct rt_addrinfo *info, u_int8_t prio, struct ifnet *ifp, 757 struct rtentry **ret_nrt, u_int tableid) 758 { 759 struct rtentry *rt; 760 int error; 761 762 NET_ASSERT_LOCKED(); 763 764 if (!rtable_exists(tableid)) 765 return (EAFNOSUPPORT); 766 rt = rtable_lookup(tableid, info->rti_info[RTAX_DST], 767 info->rti_info[RTAX_NETMASK], info->rti_info[RTAX_GATEWAY], prio); 768 if (rt == NULL) 769 return (ESRCH); 770 771 /* Make sure that's the route the caller want to delete. */ 772 if (ifp != NULL && ifp->if_index != rt->rt_ifidx) { 773 rtfree(rt); 774 return (ESRCH); 775 } 776 777 #ifdef BFD 778 if (ISSET(rt->rt_flags, RTF_BFD)) 779 bfdclear(rt); 780 #endif 781 782 error = rtable_delete(tableid, info->rti_info[RTAX_DST], 783 info->rti_info[RTAX_NETMASK], rt); 784 if (error != 0) { 785 rtfree(rt); 786 return (ESRCH); 787 } 788 789 /* Release next hop cache before flushing cloned entries. */ 790 rt_putgwroute(rt); 791 792 /* Clean up any cloned children. */ 793 if (ISSET(rt->rt_flags, RTF_CLONING)) 794 rtflushclone(ifp, rt, tableid); 795 796 rtfree(rt->rt_parent); 797 rt->rt_parent = NULL; 798 799 rt->rt_flags &= ~RTF_UP; 800 801 KASSERT(ifp->if_index == rt->rt_ifidx); 802 ifp->if_rtrequest(ifp, RTM_DELETE, rt); 803 804 atomic_inc_int(&rttrash); 805 806 if (ret_nrt != NULL) 807 *ret_nrt = rt; 808 else 809 rtfree(rt); 810 811 return (0); 812 } 813 814 int 815 rtrequest(int req, struct rt_addrinfo *info, u_int8_t prio, 816 struct rtentry **ret_nrt, u_int tableid) 817 { 818 struct ifnet *ifp; 819 struct rtentry *rt, *crt; 820 struct ifaddr *ifa; 821 struct sockaddr *ndst; 822 struct sockaddr_rtlabel *sa_rl, sa_rl2; 823 struct sockaddr_dl sa_dl = { sizeof(sa_dl), AF_LINK }; 824 int error; 825 826 NET_ASSERT_LOCKED(); 827 828 if (!rtable_exists(tableid)) 829 return (EAFNOSUPPORT); 830 if (info->rti_flags & RTF_HOST) 831 info->rti_info[RTAX_NETMASK] = NULL; 832 switch (req) { 833 case RTM_DELETE: 834 return (EINVAL); 835 836 case RTM_RESOLVE: 837 if (ret_nrt == NULL || (rt = *ret_nrt) == NULL) 838 return (EINVAL); 839 if ((rt->rt_flags & RTF_CLONING) == 0) 840 return (EINVAL); 841 KASSERT(rt->rt_ifa->ifa_ifp != NULL); 842 info->rti_ifa = rt->rt_ifa; 843 info->rti_flags = rt->rt_flags | (RTF_CLONED|RTF_HOST); 844 info->rti_flags &= ~(RTF_CLONING|RTF_CONNECTED|RTF_STATIC); 845 info->rti_info[RTAX_GATEWAY] = sdltosa(&sa_dl); 846 info->rti_info[RTAX_LABEL] = 847 rtlabel_id2sa(rt->rt_labelid, &sa_rl2); 848 /* FALLTHROUGH */ 849 850 case RTM_ADD: 851 if (info->rti_ifa == NULL) 852 return (EINVAL); 853 ifa = info->rti_ifa; 854 ifp = ifa->ifa_ifp; 855 if (prio == 0) 856 prio = ifp->if_priority + RTP_STATIC; 857 858 error = rt_copysa(info->rti_info[RTAX_DST], 859 info->rti_info[RTAX_NETMASK], &ndst); 860 if (error) 861 return (error); 862 863 rt = pool_get(&rtentry_pool, PR_NOWAIT | PR_ZERO); 864 if (rt == NULL) { 865 free(ndst, M_RTABLE, ndst->sa_len); 866 return (ENOBUFS); 867 } 868 869 rt->rt_refcnt = 1; 870 rt->rt_flags = info->rti_flags | RTF_UP; 871 rt->rt_priority = prio; /* init routing priority */ 872 LIST_INIT(&rt->rt_timer); 873 874 /* Check the link state if the table supports it. */ 875 if (rtable_mpath_capable(tableid, ndst->sa_family) && 876 !ISSET(rt->rt_flags, RTF_LOCAL) && 877 (!LINK_STATE_IS_UP(ifp->if_link_state) || 878 !ISSET(ifp->if_flags, IFF_UP))) { 879 rt->rt_flags &= ~RTF_UP; 880 rt->rt_priority |= RTP_DOWN; 881 } 882 883 if (info->rti_info[RTAX_LABEL] != NULL) { 884 sa_rl = (struct sockaddr_rtlabel *) 885 info->rti_info[RTAX_LABEL]; 886 rt->rt_labelid = rtlabel_name2id(sa_rl->sr_label); 887 } 888 889 #ifdef MPLS 890 /* We have to allocate additional space for MPLS infos */ 891 if (info->rti_flags & RTF_MPLS && 892 (info->rti_info[RTAX_SRC] != NULL || 893 info->rti_info[RTAX_DST]->sa_family == AF_MPLS)) { 894 error = rt_mpls_set(rt, info->rti_info[RTAX_SRC], 895 info->rti_mpls); 896 if (error) { 897 free(ndst, M_RTABLE, ndst->sa_len); 898 pool_put(&rtentry_pool, rt); 899 return (error); 900 } 901 } else 902 rt_mpls_clear(rt); 903 #endif 904 905 ifa->ifa_refcnt++; 906 rt->rt_ifa = ifa; 907 rt->rt_ifidx = ifp->if_index; 908 /* 909 * Copy metrics and a back pointer from the cloned 910 * route's parent. 911 */ 912 if (ISSET(rt->rt_flags, RTF_CLONED)) { 913 rtref(*ret_nrt); 914 rt->rt_parent = *ret_nrt; 915 rt->rt_rmx = (*ret_nrt)->rt_rmx; 916 } 917 918 /* 919 * We must set rt->rt_gateway before adding ``rt'' to 920 * the routing table because the radix MPATH code use 921 * it to (re)order routes. 922 */ 923 if ((error = rt_setgate(rt, info->rti_info[RTAX_GATEWAY], 924 tableid))) { 925 ifafree(ifa); 926 rtfree(rt->rt_parent); 927 rt_putgwroute(rt); 928 free(rt->rt_gateway, M_RTABLE, 0); 929 free(ndst, M_RTABLE, ndst->sa_len); 930 pool_put(&rtentry_pool, rt); 931 return (error); 932 } 933 934 error = rtable_insert(tableid, ndst, 935 info->rti_info[RTAX_NETMASK], info->rti_info[RTAX_GATEWAY], 936 rt->rt_priority, rt); 937 if (error != 0 && 938 (crt = rtable_match(tableid, ndst, NULL)) != NULL) { 939 /* overwrite cloned route */ 940 if (ISSET(crt->rt_flags, RTF_CLONED)) { 941 struct ifnet *cifp; 942 943 cifp = if_get(crt->rt_ifidx); 944 KASSERT(cifp != NULL); 945 rtdeletemsg(crt, cifp, tableid); 946 if_put(cifp); 947 948 error = rtable_insert(tableid, ndst, 949 info->rti_info[RTAX_NETMASK], 950 info->rti_info[RTAX_GATEWAY], 951 rt->rt_priority, rt); 952 } 953 rtfree(crt); 954 } 955 if (error != 0) { 956 ifafree(ifa); 957 rtfree(rt->rt_parent); 958 rt_putgwroute(rt); 959 free(rt->rt_gateway, M_RTABLE, 0); 960 free(ndst, M_RTABLE, ndst->sa_len); 961 pool_put(&rtentry_pool, rt); 962 return (EEXIST); 963 } 964 ifp->if_rtrequest(ifp, req, rt); 965 966 if_group_routechange(info->rti_info[RTAX_DST], 967 info->rti_info[RTAX_NETMASK]); 968 969 if (ret_nrt != NULL) 970 *ret_nrt = rt; 971 else 972 rtfree(rt); 973 break; 974 } 975 976 return (0); 977 } 978 979 int 980 rt_setgate(struct rtentry *rt, struct sockaddr *gate, u_int rtableid) 981 { 982 int glen = ROUNDUP(gate->sa_len); 983 struct sockaddr *sa; 984 985 if (rt->rt_gateway == NULL || glen != ROUNDUP(rt->rt_gateway->sa_len)) { 986 sa = malloc(glen, M_RTABLE, M_NOWAIT); 987 if (sa == NULL) 988 return (ENOBUFS); 989 if (rt->rt_gateway != NULL) { 990 free(rt->rt_gateway, M_RTABLE, 991 ROUNDUP(rt->rt_gateway->sa_len)); 992 } 993 rt->rt_gateway = sa; 994 } 995 memmove(rt->rt_gateway, gate, glen); 996 997 if (ISSET(rt->rt_flags, RTF_GATEWAY)) 998 return (rt_setgwroute(rt, rtableid)); 999 1000 return (0); 1001 } 1002 1003 /* 1004 * Return the route entry containing the next hop link-layer 1005 * address corresponding to ``rt''. 1006 */ 1007 struct rtentry * 1008 rt_getll(struct rtentry *rt) 1009 { 1010 if (ISSET(rt->rt_flags, RTF_GATEWAY)) { 1011 KASSERT(rt->rt_gwroute != NULL); 1012 return (rt->rt_gwroute); 1013 } 1014 1015 return (rt); 1016 } 1017 1018 void 1019 rt_maskedcopy(struct sockaddr *src, struct sockaddr *dst, 1020 struct sockaddr *netmask) 1021 { 1022 u_char *cp1 = (u_char *)src; 1023 u_char *cp2 = (u_char *)dst; 1024 u_char *cp3 = (u_char *)netmask; 1025 u_char *cplim = cp2 + *cp3; 1026 u_char *cplim2 = cp2 + *cp1; 1027 1028 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */ 1029 cp3 += 2; 1030 if (cplim > cplim2) 1031 cplim = cplim2; 1032 while (cp2 < cplim) 1033 *cp2++ = *cp1++ & *cp3++; 1034 if (cp2 < cplim2) 1035 bzero(cp2, cplim2 - cp2); 1036 } 1037 1038 /* 1039 * allocate new sockaddr structure based on the user supplied src and mask 1040 * that is useable for the routing table. 1041 */ 1042 static int 1043 rt_copysa(struct sockaddr *src, struct sockaddr *mask, struct sockaddr **dst) 1044 { 1045 static const u_char maskarray[] = { 1046 0x0, 0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe }; 1047 struct sockaddr *ndst; 1048 struct domain *dp; 1049 u_char *csrc, *cdst; 1050 int i, plen; 1051 1052 for (i = 0; (dp = domains[i]) != NULL; i++) { 1053 if (dp->dom_rtoffset == 0) 1054 continue; 1055 if (src->sa_family == dp->dom_family) 1056 break; 1057 } 1058 if (dp == NULL) 1059 return (EAFNOSUPPORT); 1060 1061 if (src->sa_len < dp->dom_sasize) 1062 return (EINVAL); 1063 1064 plen = rtable_satoplen(src->sa_family, mask); 1065 if (plen == -1) 1066 return (EINVAL); 1067 1068 ndst = malloc(dp->dom_sasize, M_RTABLE, M_NOWAIT|M_ZERO); 1069 if (ndst == NULL) 1070 return (ENOBUFS); 1071 1072 ndst->sa_family = src->sa_family; 1073 ndst->sa_len = dp->dom_sasize; 1074 1075 csrc = (u_char *)src + dp->dom_rtoffset; 1076 cdst = (u_char *)ndst + dp->dom_rtoffset; 1077 1078 memcpy(cdst, csrc, plen / 8); 1079 if (plen % 8 != 0) 1080 cdst[plen / 8] = csrc[plen / 8] & maskarray[plen % 8]; 1081 1082 *dst = ndst; 1083 return (0); 1084 } 1085 1086 int 1087 rt_ifa_add(struct ifaddr *ifa, int flags, struct sockaddr *dst, 1088 unsigned int rdomain) 1089 { 1090 struct ifnet *ifp = ifa->ifa_ifp; 1091 struct rtentry *rt; 1092 struct sockaddr_rtlabel sa_rl; 1093 struct rt_addrinfo info; 1094 uint8_t prio = ifp->if_priority + RTP_STATIC; 1095 int error; 1096 1097 memset(&info, 0, sizeof(info)); 1098 info.rti_ifa = ifa; 1099 info.rti_flags = flags; 1100 info.rti_info[RTAX_DST] = dst; 1101 if (flags & RTF_LLINFO) 1102 info.rti_info[RTAX_GATEWAY] = sdltosa(ifp->if_sadl); 1103 else 1104 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr; 1105 1106 KASSERT(rdomain == rtable_l2(rdomain)); 1107 if (rdomain == rtable_l2(ifp->if_rtlabelid)) { 1108 info.rti_info[RTAX_LABEL] = 1109 rtlabel_id2sa(ifp->if_rtlabelid, &sa_rl); 1110 } 1111 1112 #ifdef MPLS 1113 if ((flags & RTF_MPLS) == RTF_MPLS) 1114 info.rti_mpls = MPLS_OP_POP; 1115 #endif /* MPLS */ 1116 1117 if ((flags & RTF_HOST) == 0) 1118 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask; 1119 1120 if (flags & (RTF_LOCAL|RTF_BROADCAST)) 1121 prio = RTP_LOCAL; 1122 1123 if (flags & RTF_CONNECTED) 1124 prio = ifp->if_priority + RTP_CONNECTED; 1125 1126 error = rtrequest(RTM_ADD, &info, prio, &rt, rdomain); 1127 if (error == 0) { 1128 /* 1129 * A local route is created for every address configured 1130 * on an interface, so use this information to notify 1131 * userland that a new address has been added. 1132 */ 1133 if (flags & RTF_LOCAL) 1134 rtm_addr(RTM_NEWADDR, ifa); 1135 rtm_send(rt, RTM_ADD, 0, rdomain); 1136 rtfree(rt); 1137 } 1138 return (error); 1139 } 1140 1141 int 1142 rt_ifa_del(struct ifaddr *ifa, int flags, struct sockaddr *dst, 1143 unsigned int rdomain) 1144 { 1145 struct ifnet *ifp = ifa->ifa_ifp; 1146 struct rtentry *rt; 1147 struct mbuf *m = NULL; 1148 struct sockaddr *deldst; 1149 struct rt_addrinfo info; 1150 struct sockaddr_rtlabel sa_rl; 1151 uint8_t prio = ifp->if_priority + RTP_STATIC; 1152 int error; 1153 1154 if ((flags & RTF_HOST) == 0 && ifa->ifa_netmask) { 1155 m = m_get(M_DONTWAIT, MT_SONAME); 1156 if (m == NULL) 1157 return (ENOBUFS); 1158 deldst = mtod(m, struct sockaddr *); 1159 rt_maskedcopy(dst, deldst, ifa->ifa_netmask); 1160 dst = deldst; 1161 } 1162 1163 memset(&info, 0, sizeof(info)); 1164 info.rti_ifa = ifa; 1165 info.rti_flags = flags; 1166 info.rti_info[RTAX_DST] = dst; 1167 if ((flags & RTF_LLINFO) == 0) 1168 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr; 1169 1170 if (rdomain == rtable_l2(ifp->if_rtlabelid)) { 1171 info.rti_info[RTAX_LABEL] = 1172 rtlabel_id2sa(ifp->if_rtlabelid, &sa_rl); 1173 } 1174 1175 if ((flags & RTF_HOST) == 0) 1176 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask; 1177 1178 if (flags & (RTF_LOCAL|RTF_BROADCAST)) 1179 prio = RTP_LOCAL; 1180 1181 if (flags & RTF_CONNECTED) 1182 prio = ifp->if_priority + RTP_CONNECTED; 1183 1184 error = rtrequest_delete(&info, prio, ifp, &rt, rdomain); 1185 if (error == 0) { 1186 rtm_send(rt, RTM_DELETE, 0, rdomain); 1187 if (flags & RTF_LOCAL) 1188 rtm_addr(RTM_DELADDR, ifa); 1189 rtfree(rt); 1190 } 1191 m_free(m); 1192 1193 return (error); 1194 } 1195 1196 /* 1197 * Add ifa's address as a local rtentry. 1198 */ 1199 int 1200 rt_ifa_addlocal(struct ifaddr *ifa) 1201 { 1202 struct ifnet *ifp = ifa->ifa_ifp; 1203 struct rtentry *rt; 1204 u_int flags = RTF_HOST|RTF_LOCAL; 1205 int error = 0; 1206 1207 /* 1208 * If the configured address correspond to the magical "any" 1209 * address do not add a local route entry because that might 1210 * corrupt the routing tree which uses this value for the 1211 * default routes. 1212 */ 1213 switch (ifa->ifa_addr->sa_family) { 1214 case AF_INET: 1215 if (satosin(ifa->ifa_addr)->sin_addr.s_addr == INADDR_ANY) 1216 return (0); 1217 break; 1218 #ifdef INET6 1219 case AF_INET6: 1220 if (IN6_ARE_ADDR_EQUAL(&satosin6(ifa->ifa_addr)->sin6_addr, 1221 &in6addr_any)) 1222 return (0); 1223 break; 1224 #endif 1225 default: 1226 break; 1227 } 1228 1229 if (!ISSET(ifp->if_flags, (IFF_LOOPBACK|IFF_POINTOPOINT))) 1230 flags |= RTF_LLINFO; 1231 1232 /* If there is no local entry, allocate one. */ 1233 rt = rtalloc(ifa->ifa_addr, 0, ifp->if_rdomain); 1234 if (rt == NULL || ISSET(rt->rt_flags, flags) != flags) { 1235 error = rt_ifa_add(ifa, flags | RTF_MPATH, ifa->ifa_addr, 1236 ifp->if_rdomain); 1237 } 1238 rtfree(rt); 1239 1240 return (error); 1241 } 1242 1243 /* 1244 * Remove local rtentry of ifa's addresss if it exists. 1245 */ 1246 int 1247 rt_ifa_dellocal(struct ifaddr *ifa) 1248 { 1249 struct ifnet *ifp = ifa->ifa_ifp; 1250 struct rtentry *rt; 1251 u_int flags = RTF_HOST|RTF_LOCAL; 1252 int error = 0; 1253 1254 /* 1255 * We do not add local routes for such address, so do not bother 1256 * removing them. 1257 */ 1258 switch (ifa->ifa_addr->sa_family) { 1259 case AF_INET: 1260 if (satosin(ifa->ifa_addr)->sin_addr.s_addr == INADDR_ANY) 1261 return (0); 1262 break; 1263 #ifdef INET6 1264 case AF_INET6: 1265 if (IN6_ARE_ADDR_EQUAL(&satosin6(ifa->ifa_addr)->sin6_addr, 1266 &in6addr_any)) 1267 return (0); 1268 break; 1269 #endif 1270 default: 1271 break; 1272 } 1273 1274 if (!ISSET(ifp->if_flags, (IFF_LOOPBACK|IFF_POINTOPOINT))) 1275 flags |= RTF_LLINFO; 1276 1277 /* 1278 * Before deleting, check if a corresponding local host 1279 * route surely exists. With this check, we can avoid to 1280 * delete an interface direct route whose destination is same 1281 * as the address being removed. This can happen when removing 1282 * a subnet-router anycast address on an interface attached 1283 * to a shared medium. 1284 */ 1285 rt = rtalloc(ifa->ifa_addr, 0, ifp->if_rdomain); 1286 if (rt != NULL && ISSET(rt->rt_flags, flags) == flags) { 1287 error = rt_ifa_del(ifa, flags, ifa->ifa_addr, 1288 ifp->if_rdomain); 1289 } 1290 rtfree(rt); 1291 1292 return (error); 1293 } 1294 1295 /* 1296 * Remove all addresses attached to ``ifa''. 1297 */ 1298 int 1299 rt_ifa_purge(struct ifaddr *ifa) 1300 { 1301 struct ifnet *ifp = ifa->ifa_ifp; 1302 struct rtentry *rt = NULL; 1303 unsigned int rtableid; 1304 int error, af = ifa->ifa_addr->sa_family; 1305 1306 KASSERT(ifp != NULL); 1307 1308 for (rtableid = 0; rtableid < rtmap_limit; rtableid++) { 1309 /* skip rtables that are not in the rdomain of the ifp */ 1310 if (rtable_l2(rtableid) != ifp->if_rdomain) 1311 continue; 1312 1313 do { 1314 error = rtable_walk(rtableid, af, &rt, 1315 rt_ifa_purge_walker, ifa); 1316 if (rt != NULL && error == EEXIST) { 1317 error = rtdeletemsg(rt, ifp, rtableid); 1318 if (error == 0) 1319 error = EAGAIN; 1320 } 1321 rtfree(rt); 1322 rt = NULL; 1323 } while (error == EAGAIN); 1324 1325 if (error == EAFNOSUPPORT) 1326 error = 0; 1327 1328 if (error) 1329 break; 1330 } 1331 1332 return error; 1333 } 1334 1335 int 1336 rt_ifa_purge_walker(struct rtentry *rt, void *vifa, unsigned int rtableid) 1337 { 1338 struct ifaddr *ifa = vifa; 1339 1340 if (rt->rt_ifa == ifa) 1341 return EEXIST; 1342 1343 return 0; 1344 } 1345 1346 /* 1347 * Route timer routines. These routes allow functions to be called 1348 * for various routes at any time. This is useful in supporting 1349 * path MTU discovery and redirect route deletion. 1350 * 1351 * This is similar to some BSDI internal functions, but it provides 1352 * for multiple queues for efficiency's sake... 1353 */ 1354 1355 LIST_HEAD(, rttimer_queue) rttimer_queue_head; 1356 static int rt_init_done = 0; 1357 1358 #define RTTIMER_CALLOUT(r) { \ 1359 if (r->rtt_func != NULL) { \ 1360 (*r->rtt_func)(r->rtt_rt, r); \ 1361 } else { \ 1362 struct ifnet *ifp; \ 1363 \ 1364 ifp = if_get(r->rtt_rt->rt_ifidx); \ 1365 if (ifp != NULL) \ 1366 rtdeletemsg(r->rtt_rt, ifp, r->rtt_tableid); \ 1367 if_put(ifp); \ 1368 } \ 1369 } 1370 1371 /* 1372 * Some subtle order problems with domain initialization mean that 1373 * we cannot count on this being run from rt_init before various 1374 * protocol initializations are done. Therefore, we make sure 1375 * that this is run when the first queue is added... 1376 */ 1377 1378 void 1379 rt_timer_init(void) 1380 { 1381 static struct timeout rt_timer_timeout; 1382 1383 if (rt_init_done) 1384 panic("rt_timer_init: already initialized"); 1385 1386 pool_init(&rttimer_pool, sizeof(struct rttimer), 0, IPL_SOFTNET, 0, 1387 "rttmr", NULL); 1388 1389 LIST_INIT(&rttimer_queue_head); 1390 timeout_set_proc(&rt_timer_timeout, rt_timer_timer, &rt_timer_timeout); 1391 timeout_add_sec(&rt_timer_timeout, 1); 1392 rt_init_done = 1; 1393 } 1394 1395 struct rttimer_queue * 1396 rt_timer_queue_create(u_int timeout) 1397 { 1398 struct rttimer_queue *rtq; 1399 1400 if (rt_init_done == 0) 1401 rt_timer_init(); 1402 1403 if ((rtq = malloc(sizeof(*rtq), M_RTABLE, M_NOWAIT|M_ZERO)) == NULL) 1404 return (NULL); 1405 1406 rtq->rtq_timeout = timeout; 1407 rtq->rtq_count = 0; 1408 TAILQ_INIT(&rtq->rtq_head); 1409 LIST_INSERT_HEAD(&rttimer_queue_head, rtq, rtq_link); 1410 1411 return (rtq); 1412 } 1413 1414 void 1415 rt_timer_queue_change(struct rttimer_queue *rtq, long timeout) 1416 { 1417 rtq->rtq_timeout = timeout; 1418 } 1419 1420 void 1421 rt_timer_queue_destroy(struct rttimer_queue *rtq) 1422 { 1423 struct rttimer *r; 1424 1425 NET_ASSERT_LOCKED(); 1426 1427 while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL) { 1428 LIST_REMOVE(r, rtt_link); 1429 TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next); 1430 RTTIMER_CALLOUT(r); 1431 pool_put(&rttimer_pool, r); 1432 if (rtq->rtq_count > 0) 1433 rtq->rtq_count--; 1434 else 1435 printf("rt_timer_queue_destroy: rtq_count reached 0\n"); 1436 } 1437 1438 LIST_REMOVE(rtq, rtq_link); 1439 free(rtq, M_RTABLE, sizeof(*rtq)); 1440 } 1441 1442 unsigned long 1443 rt_timer_queue_count(struct rttimer_queue *rtq) 1444 { 1445 return (rtq->rtq_count); 1446 } 1447 1448 void 1449 rt_timer_remove_all(struct rtentry *rt) 1450 { 1451 struct rttimer *r; 1452 1453 while ((r = LIST_FIRST(&rt->rt_timer)) != NULL) { 1454 LIST_REMOVE(r, rtt_link); 1455 TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next); 1456 if (r->rtt_queue->rtq_count > 0) 1457 r->rtt_queue->rtq_count--; 1458 else 1459 printf("rt_timer_remove_all: rtq_count reached 0\n"); 1460 pool_put(&rttimer_pool, r); 1461 } 1462 } 1463 1464 int 1465 rt_timer_add(struct rtentry *rt, void (*func)(struct rtentry *, 1466 struct rttimer *), struct rttimer_queue *queue, u_int rtableid) 1467 { 1468 struct rttimer *r; 1469 long current_time; 1470 1471 current_time = time_uptime; 1472 rt->rt_expire = time_uptime + queue->rtq_timeout; 1473 1474 /* 1475 * If there's already a timer with this action, destroy it before 1476 * we add a new one. 1477 */ 1478 LIST_FOREACH(r, &rt->rt_timer, rtt_link) { 1479 if (r->rtt_func == func) { 1480 LIST_REMOVE(r, rtt_link); 1481 TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next); 1482 if (r->rtt_queue->rtq_count > 0) 1483 r->rtt_queue->rtq_count--; 1484 else 1485 printf("rt_timer_add: rtq_count reached 0\n"); 1486 pool_put(&rttimer_pool, r); 1487 break; /* only one per list, so we can quit... */ 1488 } 1489 } 1490 1491 r = pool_get(&rttimer_pool, PR_NOWAIT | PR_ZERO); 1492 if (r == NULL) 1493 return (ENOBUFS); 1494 1495 r->rtt_rt = rt; 1496 r->rtt_time = current_time; 1497 r->rtt_func = func; 1498 r->rtt_queue = queue; 1499 r->rtt_tableid = rtableid; 1500 LIST_INSERT_HEAD(&rt->rt_timer, r, rtt_link); 1501 TAILQ_INSERT_TAIL(&queue->rtq_head, r, rtt_next); 1502 r->rtt_queue->rtq_count++; 1503 1504 return (0); 1505 } 1506 1507 void 1508 rt_timer_timer(void *arg) 1509 { 1510 struct timeout *to = (struct timeout *)arg; 1511 struct rttimer_queue *rtq; 1512 struct rttimer *r; 1513 long current_time; 1514 1515 current_time = time_uptime; 1516 1517 NET_LOCK(); 1518 LIST_FOREACH(rtq, &rttimer_queue_head, rtq_link) { 1519 while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL && 1520 (r->rtt_time + rtq->rtq_timeout) < current_time) { 1521 LIST_REMOVE(r, rtt_link); 1522 TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next); 1523 RTTIMER_CALLOUT(r); 1524 pool_put(&rttimer_pool, r); 1525 if (rtq->rtq_count > 0) 1526 rtq->rtq_count--; 1527 else 1528 printf("rt_timer_timer: rtq_count reached 0\n"); 1529 } 1530 } 1531 NET_UNLOCK(); 1532 1533 timeout_add_sec(to, 1); 1534 } 1535 1536 #ifdef MPLS 1537 int 1538 rt_mpls_set(struct rtentry *rt, struct sockaddr *src, uint8_t op) 1539 { 1540 struct sockaddr_mpls *psa_mpls = (struct sockaddr_mpls *)src; 1541 struct rt_mpls *rt_mpls; 1542 1543 if (psa_mpls == NULL && op != MPLS_OP_POP) 1544 return (EOPNOTSUPP); 1545 if (psa_mpls != NULL && psa_mpls->smpls_len != sizeof(*psa_mpls)) 1546 return (EINVAL); 1547 if (psa_mpls != NULL && psa_mpls->smpls_family != AF_MPLS) 1548 return (EAFNOSUPPORT); 1549 1550 rt->rt_llinfo = malloc(sizeof(struct rt_mpls), M_TEMP, M_NOWAIT|M_ZERO); 1551 if (rt->rt_llinfo == NULL) 1552 return (ENOMEM); 1553 1554 rt_mpls = (struct rt_mpls *)rt->rt_llinfo; 1555 if (psa_mpls != NULL) 1556 rt_mpls->mpls_label = psa_mpls->smpls_label; 1557 rt_mpls->mpls_operation = op; 1558 /* XXX: set experimental bits */ 1559 rt->rt_flags |= RTF_MPLS; 1560 1561 return (0); 1562 } 1563 1564 void 1565 rt_mpls_clear(struct rtentry *rt) 1566 { 1567 if (rt->rt_llinfo != NULL && rt->rt_flags & RTF_MPLS) { 1568 free(rt->rt_llinfo, M_TEMP, sizeof(struct rt_mpls)); 1569 rt->rt_llinfo = NULL; 1570 } 1571 rt->rt_flags &= ~RTF_MPLS; 1572 } 1573 #endif 1574 1575 u_int16_t 1576 rtlabel_name2id(char *name) 1577 { 1578 struct rt_label *label, *p; 1579 u_int16_t new_id = 1; 1580 1581 if (!name[0]) 1582 return (0); 1583 1584 TAILQ_FOREACH(label, &rt_labels, rtl_entry) 1585 if (strcmp(name, label->rtl_name) == 0) { 1586 label->rtl_ref++; 1587 return (label->rtl_id); 1588 } 1589 1590 /* 1591 * to avoid fragmentation, we do a linear search from the beginning 1592 * and take the first free slot we find. if there is none or the list 1593 * is empty, append a new entry at the end. 1594 */ 1595 TAILQ_FOREACH(p, &rt_labels, rtl_entry) { 1596 if (p->rtl_id != new_id) 1597 break; 1598 new_id = p->rtl_id + 1; 1599 } 1600 if (new_id > LABELID_MAX) 1601 return (0); 1602 1603 label = malloc(sizeof(*label), M_RTABLE, M_NOWAIT|M_ZERO); 1604 if (label == NULL) 1605 return (0); 1606 strlcpy(label->rtl_name, name, sizeof(label->rtl_name)); 1607 label->rtl_id = new_id; 1608 label->rtl_ref++; 1609 1610 if (p != NULL) /* insert new entry before p */ 1611 TAILQ_INSERT_BEFORE(p, label, rtl_entry); 1612 else /* either list empty or no free slot in between */ 1613 TAILQ_INSERT_TAIL(&rt_labels, label, rtl_entry); 1614 1615 return (label->rtl_id); 1616 } 1617 1618 const char * 1619 rtlabel_id2name(u_int16_t id) 1620 { 1621 struct rt_label *label; 1622 1623 TAILQ_FOREACH(label, &rt_labels, rtl_entry) 1624 if (label->rtl_id == id) 1625 return (label->rtl_name); 1626 1627 return (NULL); 1628 } 1629 1630 struct sockaddr * 1631 rtlabel_id2sa(u_int16_t labelid, struct sockaddr_rtlabel *sa_rl) 1632 { 1633 const char *label; 1634 1635 if (labelid == 0 || (label = rtlabel_id2name(labelid)) == NULL) 1636 return (NULL); 1637 1638 bzero(sa_rl, sizeof(*sa_rl)); 1639 sa_rl->sr_len = sizeof(*sa_rl); 1640 sa_rl->sr_family = AF_UNSPEC; 1641 strlcpy(sa_rl->sr_label, label, sizeof(sa_rl->sr_label)); 1642 1643 return ((struct sockaddr *)sa_rl); 1644 } 1645 1646 void 1647 rtlabel_unref(u_int16_t id) 1648 { 1649 struct rt_label *p, *next; 1650 1651 if (id == 0) 1652 return; 1653 1654 TAILQ_FOREACH_SAFE(p, &rt_labels, rtl_entry, next) { 1655 if (id == p->rtl_id) { 1656 if (--p->rtl_ref == 0) { 1657 TAILQ_REMOVE(&rt_labels, p, rtl_entry); 1658 free(p, M_RTABLE, sizeof(*p)); 1659 } 1660 break; 1661 } 1662 } 1663 } 1664 1665 int 1666 rt_if_track(struct ifnet *ifp) 1667 { 1668 unsigned int rtableid; 1669 struct rtentry *rt = NULL; 1670 int i, error; 1671 1672 for (rtableid = 0; rtableid < rtmap_limit; rtableid++) { 1673 /* skip rtables that are not in the rdomain of the ifp */ 1674 if (rtable_l2(rtableid) != ifp->if_rdomain) 1675 continue; 1676 for (i = 1; i <= AF_MAX; i++) { 1677 if (!rtable_mpath_capable(rtableid, i)) 1678 continue; 1679 1680 do { 1681 error = rtable_walk(rtableid, i, &rt, 1682 rt_if_linkstate_change, ifp); 1683 if (rt != NULL && error == EEXIST) { 1684 error = rtdeletemsg(rt, ifp, rtableid); 1685 if (error == 0) 1686 error = EAGAIN; 1687 } 1688 rtfree(rt); 1689 rt = NULL; 1690 } while (error == EAGAIN); 1691 1692 if (error == EAFNOSUPPORT) 1693 error = 0; 1694 1695 if (error) 1696 break; 1697 } 1698 } 1699 1700 return (error); 1701 } 1702 1703 int 1704 rt_if_linkstate_change(struct rtentry *rt, void *arg, u_int id) 1705 { 1706 struct ifnet *ifp = arg; 1707 struct sockaddr_in6 sa_mask; 1708 int error; 1709 1710 if (rt->rt_ifidx != ifp->if_index) 1711 return (0); 1712 1713 /* Local routes are always usable. */ 1714 if (rt->rt_flags & RTF_LOCAL) { 1715 rt->rt_flags |= RTF_UP; 1716 return (0); 1717 } 1718 1719 if (LINK_STATE_IS_UP(ifp->if_link_state) && ifp->if_flags & IFF_UP) { 1720 if (ISSET(rt->rt_flags, RTF_UP)) 1721 return (0); 1722 1723 /* bring route up */ 1724 rt->rt_flags |= RTF_UP; 1725 error = rtable_mpath_reprio(id, rt_key(rt), rt_plen(rt), 1726 rt->rt_priority & RTP_MASK, rt); 1727 } else { 1728 /* 1729 * Remove redirected and cloned routes (mainly ARP) 1730 * from down interfaces so we have a chance to get 1731 * new routes from a better source. 1732 */ 1733 if (ISSET(rt->rt_flags, RTF_CLONED|RTF_DYNAMIC) && 1734 !ISSET(rt->rt_flags, RTF_CACHED|RTF_BFD)) { 1735 return (EEXIST); 1736 } 1737 1738 if (!ISSET(rt->rt_flags, RTF_UP)) 1739 return (0); 1740 1741 /* take route down */ 1742 rt->rt_flags &= ~RTF_UP; 1743 error = rtable_mpath_reprio(id, rt_key(rt), rt_plen(rt), 1744 rt->rt_priority | RTP_DOWN, rt); 1745 } 1746 if_group_routechange(rt_key(rt), rt_plen2mask(rt, &sa_mask)); 1747 1748 return (error); 1749 } 1750 1751 struct sockaddr * 1752 rt_plentosa(sa_family_t af, int plen, struct sockaddr_in6 *sa_mask) 1753 { 1754 struct sockaddr_in *sin = (struct sockaddr_in *)sa_mask; 1755 #ifdef INET6 1756 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sa_mask; 1757 #endif 1758 1759 KASSERT(plen >= 0 || plen == -1); 1760 1761 if (plen == -1) 1762 return (NULL); 1763 1764 memset(sa_mask, 0, sizeof(*sa_mask)); 1765 1766 switch (af) { 1767 case AF_INET: 1768 sin->sin_family = AF_INET; 1769 sin->sin_len = sizeof(struct sockaddr_in); 1770 in_prefixlen2mask(&sin->sin_addr, plen); 1771 break; 1772 #ifdef INET6 1773 case AF_INET6: 1774 sin6->sin6_family = AF_INET6; 1775 sin6->sin6_len = sizeof(struct sockaddr_in6); 1776 in6_prefixlen2mask(&sin6->sin6_addr, plen); 1777 break; 1778 #endif /* INET6 */ 1779 default: 1780 return (NULL); 1781 } 1782 1783 return ((struct sockaddr *)sa_mask); 1784 } 1785 1786 struct sockaddr * 1787 rt_plen2mask(struct rtentry *rt, struct sockaddr_in6 *sa_mask) 1788 { 1789 return (rt_plentosa(rt_key(rt)->sa_family, rt_plen(rt), sa_mask)); 1790 } 1791 1792 #ifdef DDB 1793 #include <machine/db_machdep.h> 1794 #include <ddb/db_output.h> 1795 1796 void 1797 db_print_sa(struct sockaddr *sa) 1798 { 1799 int len; 1800 u_char *p; 1801 1802 if (sa == NULL) { 1803 db_printf("[NULL]"); 1804 return; 1805 } 1806 1807 p = (u_char *)sa; 1808 len = sa->sa_len; 1809 db_printf("["); 1810 while (len > 0) { 1811 db_printf("%d", *p); 1812 p++; 1813 len--; 1814 if (len) 1815 db_printf(","); 1816 } 1817 db_printf("]\n"); 1818 } 1819 1820 void 1821 db_print_ifa(struct ifaddr *ifa) 1822 { 1823 if (ifa == NULL) 1824 return; 1825 db_printf(" ifa_addr="); 1826 db_print_sa(ifa->ifa_addr); 1827 db_printf(" ifa_dsta="); 1828 db_print_sa(ifa->ifa_dstaddr); 1829 db_printf(" ifa_mask="); 1830 db_print_sa(ifa->ifa_netmask); 1831 db_printf(" flags=0x%x, refcnt=%d, metric=%d\n", 1832 ifa->ifa_flags, ifa->ifa_refcnt, ifa->ifa_metric); 1833 } 1834 1835 /* 1836 * Function to pass to rtalble_walk(). 1837 * Return non-zero error to abort walk. 1838 */ 1839 int 1840 db_show_rtentry(struct rtentry *rt, void *w, unsigned int id) 1841 { 1842 db_printf("rtentry=%p", rt); 1843 1844 db_printf(" flags=0x%x refcnt=%d use=%llu expire=%lld rtableid=%u\n", 1845 rt->rt_flags, rt->rt_refcnt, rt->rt_use, rt->rt_expire, id); 1846 1847 db_printf(" key="); db_print_sa(rt_key(rt)); 1848 db_printf(" plen=%d", rt_plen(rt)); 1849 db_printf(" gw="); db_print_sa(rt->rt_gateway); 1850 db_printf(" ifidx=%u ", rt->rt_ifidx); 1851 db_printf(" ifa=%p\n", rt->rt_ifa); 1852 db_print_ifa(rt->rt_ifa); 1853 1854 db_printf(" gwroute=%p llinfo=%p\n", rt->rt_gwroute, rt->rt_llinfo); 1855 return (0); 1856 } 1857 1858 /* 1859 * Function to print all the route trees. 1860 * Use this from ddb: "call db_show_arptab" 1861 */ 1862 int 1863 db_show_arptab(void) 1864 { 1865 db_printf("Route tree for AF_INET\n"); 1866 rtable_walk(0, AF_INET, NULL, db_show_rtentry, NULL); 1867 return (0); 1868 } 1869 #endif /* DDB */ 1870