1 /* $OpenBSD: route.c,v 1.387 2019/06/24 22:26:25 bluhm 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 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 rtentry *parent, unsigned int rtableid) 730 { 731 struct rtentry *rt = NULL; 732 struct ifnet *ifp; 733 int error; 734 735 #ifdef DIAGNOSTIC 736 if (!parent || (parent->rt_flags & RTF_CLONING) == 0) 737 panic("rtflushclone: called with a non-cloning route"); 738 #endif 739 740 do { 741 error = rtable_walk(rtableid, rt_key(parent)->sa_family, &rt, 742 rtflushclone1, parent); 743 if (rt != NULL && error == EEXIST) { 744 ifp = if_get(rt->rt_ifidx); 745 if (ifp == NULL) { 746 error = EAGAIN; 747 } else { 748 error = rtdeletemsg(rt, ifp, rtableid); 749 if (error == 0) 750 error = EAGAIN; 751 if_put(ifp); 752 } 753 } 754 rtfree(rt); 755 rt = NULL; 756 } while (error == EAGAIN); 757 758 return error; 759 760 } 761 762 int 763 rtrequest_delete(struct rt_addrinfo *info, u_int8_t prio, struct ifnet *ifp, 764 struct rtentry **ret_nrt, u_int tableid) 765 { 766 struct rtentry *rt; 767 int error; 768 769 NET_ASSERT_LOCKED(); 770 771 if (!rtable_exists(tableid)) 772 return (EAFNOSUPPORT); 773 rt = rtable_lookup(tableid, info->rti_info[RTAX_DST], 774 info->rti_info[RTAX_NETMASK], info->rti_info[RTAX_GATEWAY], prio); 775 if (rt == NULL) 776 return (ESRCH); 777 778 /* Make sure that's the route the caller want to delete. */ 779 if (ifp != NULL && ifp->if_index != rt->rt_ifidx) { 780 rtfree(rt); 781 return (ESRCH); 782 } 783 784 #ifdef BFD 785 if (ISSET(rt->rt_flags, RTF_BFD)) 786 bfdclear(rt); 787 #endif 788 789 error = rtable_delete(tableid, info->rti_info[RTAX_DST], 790 info->rti_info[RTAX_NETMASK], rt); 791 if (error != 0) { 792 rtfree(rt); 793 return (ESRCH); 794 } 795 796 /* Release next hop cache before flushing cloned entries. */ 797 rt_putgwroute(rt); 798 799 /* Clean up any cloned children. */ 800 if (ISSET(rt->rt_flags, RTF_CLONING)) 801 rtflushclone(rt, tableid); 802 803 rtfree(rt->rt_parent); 804 rt->rt_parent = NULL; 805 806 rt->rt_flags &= ~RTF_UP; 807 808 KASSERT(ifp->if_index == rt->rt_ifidx); 809 ifp->if_rtrequest(ifp, RTM_DELETE, rt); 810 811 atomic_inc_int(&rttrash); 812 813 if (ret_nrt != NULL) 814 *ret_nrt = rt; 815 else 816 rtfree(rt); 817 818 return (0); 819 } 820 821 int 822 rtrequest(int req, struct rt_addrinfo *info, u_int8_t prio, 823 struct rtentry **ret_nrt, u_int tableid) 824 { 825 struct ifnet *ifp; 826 struct rtentry *rt, *crt; 827 struct ifaddr *ifa; 828 struct sockaddr *ndst; 829 struct sockaddr_rtlabel *sa_rl, sa_rl2; 830 struct sockaddr_dl sa_dl = { sizeof(sa_dl), AF_LINK }; 831 int error; 832 833 NET_ASSERT_LOCKED(); 834 835 if (!rtable_exists(tableid)) 836 return (EAFNOSUPPORT); 837 if (info->rti_flags & RTF_HOST) 838 info->rti_info[RTAX_NETMASK] = NULL; 839 switch (req) { 840 case RTM_DELETE: 841 return (EINVAL); 842 843 case RTM_RESOLVE: 844 if (ret_nrt == NULL || (rt = *ret_nrt) == NULL) 845 return (EINVAL); 846 if ((rt->rt_flags & RTF_CLONING) == 0) 847 return (EINVAL); 848 KASSERT(rt->rt_ifa->ifa_ifp != NULL); 849 info->rti_ifa = rt->rt_ifa; 850 info->rti_flags = rt->rt_flags | (RTF_CLONED|RTF_HOST); 851 info->rti_flags &= ~(RTF_CLONING|RTF_CONNECTED|RTF_STATIC); 852 info->rti_info[RTAX_GATEWAY] = sdltosa(&sa_dl); 853 info->rti_info[RTAX_LABEL] = 854 rtlabel_id2sa(rt->rt_labelid, &sa_rl2); 855 /* FALLTHROUGH */ 856 857 case RTM_ADD: 858 if (info->rti_ifa == NULL) 859 return (EINVAL); 860 ifa = info->rti_ifa; 861 ifp = ifa->ifa_ifp; 862 if (prio == 0) 863 prio = ifp->if_priority + RTP_STATIC; 864 865 error = rt_copysa(info->rti_info[RTAX_DST], 866 info->rti_info[RTAX_NETMASK], &ndst); 867 if (error) 868 return (error); 869 870 rt = pool_get(&rtentry_pool, PR_NOWAIT | PR_ZERO); 871 if (rt == NULL) { 872 free(ndst, M_RTABLE, ndst->sa_len); 873 return (ENOBUFS); 874 } 875 876 rt->rt_refcnt = 1; 877 rt->rt_flags = info->rti_flags | RTF_UP; 878 rt->rt_priority = prio; /* init routing priority */ 879 LIST_INIT(&rt->rt_timer); 880 881 /* Check the link state if the table supports it. */ 882 if (rtable_mpath_capable(tableid, ndst->sa_family) && 883 !ISSET(rt->rt_flags, RTF_LOCAL) && 884 (!LINK_STATE_IS_UP(ifp->if_link_state) || 885 !ISSET(ifp->if_flags, IFF_UP))) { 886 rt->rt_flags &= ~RTF_UP; 887 rt->rt_priority |= RTP_DOWN; 888 } 889 890 if (info->rti_info[RTAX_LABEL] != NULL) { 891 sa_rl = (struct sockaddr_rtlabel *) 892 info->rti_info[RTAX_LABEL]; 893 rt->rt_labelid = rtlabel_name2id(sa_rl->sr_label); 894 } 895 896 #ifdef MPLS 897 /* We have to allocate additional space for MPLS infos */ 898 if (info->rti_flags & RTF_MPLS && 899 (info->rti_info[RTAX_SRC] != NULL || 900 info->rti_info[RTAX_DST]->sa_family == AF_MPLS)) { 901 error = rt_mpls_set(rt, info->rti_info[RTAX_SRC], 902 info->rti_mpls); 903 if (error) { 904 free(ndst, M_RTABLE, ndst->sa_len); 905 pool_put(&rtentry_pool, rt); 906 return (error); 907 } 908 } else 909 rt_mpls_clear(rt); 910 #endif 911 912 ifa->ifa_refcnt++; 913 rt->rt_ifa = ifa; 914 rt->rt_ifidx = ifp->if_index; 915 /* 916 * Copy metrics and a back pointer from the cloned 917 * route's parent. 918 */ 919 if (ISSET(rt->rt_flags, RTF_CLONED)) { 920 rtref(*ret_nrt); 921 rt->rt_parent = *ret_nrt; 922 rt->rt_rmx = (*ret_nrt)->rt_rmx; 923 } 924 925 /* 926 * We must set rt->rt_gateway before adding ``rt'' to 927 * the routing table because the radix MPATH code use 928 * it to (re)order routes. 929 */ 930 if ((error = rt_setgate(rt, info->rti_info[RTAX_GATEWAY], 931 tableid))) { 932 ifafree(ifa); 933 rtfree(rt->rt_parent); 934 rt_putgwroute(rt); 935 free(rt->rt_gateway, M_RTABLE, 0); 936 free(ndst, M_RTABLE, ndst->sa_len); 937 pool_put(&rtentry_pool, rt); 938 return (error); 939 } 940 941 error = rtable_insert(tableid, ndst, 942 info->rti_info[RTAX_NETMASK], info->rti_info[RTAX_GATEWAY], 943 rt->rt_priority, rt); 944 if (error != 0 && 945 (crt = rtable_match(tableid, ndst, NULL)) != NULL) { 946 /* overwrite cloned route */ 947 if (ISSET(crt->rt_flags, RTF_CLONED)) { 948 struct ifnet *cifp; 949 950 cifp = if_get(crt->rt_ifidx); 951 KASSERT(cifp != NULL); 952 rtdeletemsg(crt, cifp, tableid); 953 if_put(cifp); 954 955 error = rtable_insert(tableid, ndst, 956 info->rti_info[RTAX_NETMASK], 957 info->rti_info[RTAX_GATEWAY], 958 rt->rt_priority, rt); 959 } 960 rtfree(crt); 961 } 962 if (error != 0) { 963 ifafree(ifa); 964 rtfree(rt->rt_parent); 965 rt_putgwroute(rt); 966 free(rt->rt_gateway, M_RTABLE, 0); 967 free(ndst, M_RTABLE, ndst->sa_len); 968 pool_put(&rtentry_pool, rt); 969 return (EEXIST); 970 } 971 ifp->if_rtrequest(ifp, req, rt); 972 973 if_group_routechange(info->rti_info[RTAX_DST], 974 info->rti_info[RTAX_NETMASK]); 975 976 if (ret_nrt != NULL) 977 *ret_nrt = rt; 978 else 979 rtfree(rt); 980 break; 981 } 982 983 return (0); 984 } 985 986 int 987 rt_setgate(struct rtentry *rt, struct sockaddr *gate, u_int rtableid) 988 { 989 int glen = ROUNDUP(gate->sa_len); 990 struct sockaddr *sa; 991 992 if (rt->rt_gateway == NULL || glen != ROUNDUP(rt->rt_gateway->sa_len)) { 993 sa = malloc(glen, M_RTABLE, M_NOWAIT); 994 if (sa == NULL) 995 return (ENOBUFS); 996 if (rt->rt_gateway != NULL) { 997 free(rt->rt_gateway, M_RTABLE, 998 ROUNDUP(rt->rt_gateway->sa_len)); 999 } 1000 rt->rt_gateway = sa; 1001 } 1002 memmove(rt->rt_gateway, gate, glen); 1003 1004 if (ISSET(rt->rt_flags, RTF_GATEWAY)) 1005 return (rt_setgwroute(rt, rtableid)); 1006 1007 return (0); 1008 } 1009 1010 /* 1011 * Return the route entry containing the next hop link-layer 1012 * address corresponding to ``rt''. 1013 */ 1014 struct rtentry * 1015 rt_getll(struct rtentry *rt) 1016 { 1017 if (ISSET(rt->rt_flags, RTF_GATEWAY)) { 1018 KASSERT(rt->rt_gwroute != NULL); 1019 return (rt->rt_gwroute); 1020 } 1021 1022 return (rt); 1023 } 1024 1025 void 1026 rt_maskedcopy(struct sockaddr *src, struct sockaddr *dst, 1027 struct sockaddr *netmask) 1028 { 1029 u_char *cp1 = (u_char *)src; 1030 u_char *cp2 = (u_char *)dst; 1031 u_char *cp3 = (u_char *)netmask; 1032 u_char *cplim = cp2 + *cp3; 1033 u_char *cplim2 = cp2 + *cp1; 1034 1035 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */ 1036 cp3 += 2; 1037 if (cplim > cplim2) 1038 cplim = cplim2; 1039 while (cp2 < cplim) 1040 *cp2++ = *cp1++ & *cp3++; 1041 if (cp2 < cplim2) 1042 bzero(cp2, cplim2 - cp2); 1043 } 1044 1045 /* 1046 * allocate new sockaddr structure based on the user supplied src and mask 1047 * that is useable for the routing table. 1048 */ 1049 static int 1050 rt_copysa(struct sockaddr *src, struct sockaddr *mask, struct sockaddr **dst) 1051 { 1052 static const u_char maskarray[] = { 1053 0x0, 0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe }; 1054 struct sockaddr *ndst; 1055 struct domain *dp; 1056 u_char *csrc, *cdst; 1057 int i, plen; 1058 1059 for (i = 0; (dp = domains[i]) != NULL; i++) { 1060 if (dp->dom_rtoffset == 0) 1061 continue; 1062 if (src->sa_family == dp->dom_family) 1063 break; 1064 } 1065 if (dp == NULL) 1066 return (EAFNOSUPPORT); 1067 1068 if (src->sa_len < dp->dom_sasize) 1069 return (EINVAL); 1070 1071 plen = rtable_satoplen(src->sa_family, mask); 1072 if (plen == -1) 1073 return (EINVAL); 1074 1075 ndst = malloc(dp->dom_sasize, M_RTABLE, M_NOWAIT|M_ZERO); 1076 if (ndst == NULL) 1077 return (ENOBUFS); 1078 1079 ndst->sa_family = src->sa_family; 1080 ndst->sa_len = dp->dom_sasize; 1081 1082 csrc = (u_char *)src + dp->dom_rtoffset; 1083 cdst = (u_char *)ndst + dp->dom_rtoffset; 1084 1085 memcpy(cdst, csrc, plen / 8); 1086 if (plen % 8 != 0) 1087 cdst[plen / 8] = csrc[plen / 8] & maskarray[plen % 8]; 1088 1089 *dst = ndst; 1090 return (0); 1091 } 1092 1093 int 1094 rt_ifa_add(struct ifaddr *ifa, int flags, struct sockaddr *dst, 1095 unsigned int rdomain) 1096 { 1097 struct ifnet *ifp = ifa->ifa_ifp; 1098 struct rtentry *rt; 1099 struct sockaddr_rtlabel sa_rl; 1100 struct rt_addrinfo info; 1101 uint8_t prio = ifp->if_priority + RTP_STATIC; 1102 int error; 1103 1104 memset(&info, 0, sizeof(info)); 1105 info.rti_ifa = ifa; 1106 info.rti_flags = flags; 1107 info.rti_info[RTAX_DST] = dst; 1108 if (flags & RTF_LLINFO) 1109 info.rti_info[RTAX_GATEWAY] = sdltosa(ifp->if_sadl); 1110 else 1111 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr; 1112 1113 KASSERT(rdomain == rtable_l2(rdomain)); 1114 if (rdomain == rtable_l2(ifp->if_rtlabelid)) { 1115 info.rti_info[RTAX_LABEL] = 1116 rtlabel_id2sa(ifp->if_rtlabelid, &sa_rl); 1117 } 1118 1119 #ifdef MPLS 1120 if ((flags & RTF_MPLS) == RTF_MPLS) 1121 info.rti_mpls = MPLS_OP_POP; 1122 #endif /* MPLS */ 1123 1124 if ((flags & RTF_HOST) == 0) 1125 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask; 1126 1127 if (flags & (RTF_LOCAL|RTF_BROADCAST)) 1128 prio = RTP_LOCAL; 1129 1130 if (flags & RTF_CONNECTED) 1131 prio = ifp->if_priority + RTP_CONNECTED; 1132 1133 error = rtrequest(RTM_ADD, &info, prio, &rt, rdomain); 1134 if (error == 0) { 1135 /* 1136 * A local route is created for every address configured 1137 * on an interface, so use this information to notify 1138 * userland that a new address has been added. 1139 */ 1140 if (flags & RTF_LOCAL) 1141 rtm_addr(RTM_NEWADDR, ifa); 1142 rtm_send(rt, RTM_ADD, 0, rdomain); 1143 rtfree(rt); 1144 } 1145 return (error); 1146 } 1147 1148 int 1149 rt_ifa_del(struct ifaddr *ifa, int flags, struct sockaddr *dst, 1150 unsigned int rdomain) 1151 { 1152 struct ifnet *ifp = ifa->ifa_ifp; 1153 struct rtentry *rt; 1154 struct mbuf *m = NULL; 1155 struct sockaddr *deldst; 1156 struct rt_addrinfo info; 1157 struct sockaddr_rtlabel sa_rl; 1158 uint8_t prio = ifp->if_priority + RTP_STATIC; 1159 int error; 1160 1161 if ((flags & RTF_HOST) == 0 && ifa->ifa_netmask) { 1162 m = m_get(M_DONTWAIT, MT_SONAME); 1163 if (m == NULL) 1164 return (ENOBUFS); 1165 deldst = mtod(m, struct sockaddr *); 1166 rt_maskedcopy(dst, deldst, ifa->ifa_netmask); 1167 dst = deldst; 1168 } 1169 1170 memset(&info, 0, sizeof(info)); 1171 info.rti_ifa = ifa; 1172 info.rti_flags = flags; 1173 info.rti_info[RTAX_DST] = dst; 1174 if ((flags & RTF_LLINFO) == 0) 1175 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr; 1176 1177 if (rdomain == rtable_l2(ifp->if_rtlabelid)) { 1178 info.rti_info[RTAX_LABEL] = 1179 rtlabel_id2sa(ifp->if_rtlabelid, &sa_rl); 1180 } 1181 1182 if ((flags & RTF_HOST) == 0) 1183 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask; 1184 1185 if (flags & (RTF_LOCAL|RTF_BROADCAST)) 1186 prio = RTP_LOCAL; 1187 1188 if (flags & RTF_CONNECTED) 1189 prio = ifp->if_priority + RTP_CONNECTED; 1190 1191 error = rtrequest_delete(&info, prio, ifp, &rt, rdomain); 1192 if (error == 0) { 1193 rtm_send(rt, RTM_DELETE, 0, rdomain); 1194 if (flags & RTF_LOCAL) 1195 rtm_addr(RTM_DELADDR, ifa); 1196 rtfree(rt); 1197 } 1198 m_free(m); 1199 1200 return (error); 1201 } 1202 1203 /* 1204 * Add ifa's address as a local rtentry. 1205 */ 1206 int 1207 rt_ifa_addlocal(struct ifaddr *ifa) 1208 { 1209 struct ifnet *ifp = ifa->ifa_ifp; 1210 struct rtentry *rt; 1211 u_int flags = RTF_HOST|RTF_LOCAL; 1212 int error = 0; 1213 1214 /* 1215 * If the configured address correspond to the magical "any" 1216 * address do not add a local route entry because that might 1217 * corrupt the routing tree which uses this value for the 1218 * default routes. 1219 */ 1220 switch (ifa->ifa_addr->sa_family) { 1221 case AF_INET: 1222 if (satosin(ifa->ifa_addr)->sin_addr.s_addr == INADDR_ANY) 1223 return (0); 1224 break; 1225 #ifdef INET6 1226 case AF_INET6: 1227 if (IN6_ARE_ADDR_EQUAL(&satosin6(ifa->ifa_addr)->sin6_addr, 1228 &in6addr_any)) 1229 return (0); 1230 break; 1231 #endif 1232 default: 1233 break; 1234 } 1235 1236 if (!ISSET(ifp->if_flags, (IFF_LOOPBACK|IFF_POINTOPOINT))) 1237 flags |= RTF_LLINFO; 1238 1239 /* If there is no local entry, allocate one. */ 1240 rt = rtalloc(ifa->ifa_addr, 0, ifp->if_rdomain); 1241 if (rt == NULL || ISSET(rt->rt_flags, flags) != flags) { 1242 error = rt_ifa_add(ifa, flags | RTF_MPATH, ifa->ifa_addr, 1243 ifp->if_rdomain); 1244 } 1245 rtfree(rt); 1246 1247 return (error); 1248 } 1249 1250 /* 1251 * Remove local rtentry of ifa's addresss if it exists. 1252 */ 1253 int 1254 rt_ifa_dellocal(struct ifaddr *ifa) 1255 { 1256 struct ifnet *ifp = ifa->ifa_ifp; 1257 struct rtentry *rt; 1258 u_int flags = RTF_HOST|RTF_LOCAL; 1259 int error = 0; 1260 1261 /* 1262 * We do not add local routes for such address, so do not bother 1263 * removing them. 1264 */ 1265 switch (ifa->ifa_addr->sa_family) { 1266 case AF_INET: 1267 if (satosin(ifa->ifa_addr)->sin_addr.s_addr == INADDR_ANY) 1268 return (0); 1269 break; 1270 #ifdef INET6 1271 case AF_INET6: 1272 if (IN6_ARE_ADDR_EQUAL(&satosin6(ifa->ifa_addr)->sin6_addr, 1273 &in6addr_any)) 1274 return (0); 1275 break; 1276 #endif 1277 default: 1278 break; 1279 } 1280 1281 if (!ISSET(ifp->if_flags, (IFF_LOOPBACK|IFF_POINTOPOINT))) 1282 flags |= RTF_LLINFO; 1283 1284 /* 1285 * Before deleting, check if a corresponding local host 1286 * route surely exists. With this check, we can avoid to 1287 * delete an interface direct route whose destination is same 1288 * as the address being removed. This can happen when removing 1289 * a subnet-router anycast address on an interface attached 1290 * to a shared medium. 1291 */ 1292 rt = rtalloc(ifa->ifa_addr, 0, ifp->if_rdomain); 1293 if (rt != NULL && ISSET(rt->rt_flags, flags) == flags) { 1294 error = rt_ifa_del(ifa, flags, ifa->ifa_addr, 1295 ifp->if_rdomain); 1296 } 1297 rtfree(rt); 1298 1299 return (error); 1300 } 1301 1302 /* 1303 * Remove all addresses attached to ``ifa''. 1304 */ 1305 int 1306 rt_ifa_purge(struct ifaddr *ifa) 1307 { 1308 struct ifnet *ifp = ifa->ifa_ifp; 1309 struct rtentry *rt = NULL; 1310 unsigned int rtableid; 1311 int error, af = ifa->ifa_addr->sa_family; 1312 1313 KASSERT(ifp != NULL); 1314 1315 for (rtableid = 0; rtableid < rtmap_limit; rtableid++) { 1316 /* skip rtables that are not in the rdomain of the ifp */ 1317 if (rtable_l2(rtableid) != ifp->if_rdomain) 1318 continue; 1319 1320 do { 1321 error = rtable_walk(rtableid, af, &rt, 1322 rt_ifa_purge_walker, ifa); 1323 if (rt != NULL && error == EEXIST) { 1324 error = rtdeletemsg(rt, ifp, rtableid); 1325 if (error == 0) 1326 error = EAGAIN; 1327 } 1328 rtfree(rt); 1329 rt = NULL; 1330 } while (error == EAGAIN); 1331 1332 if (error == EAFNOSUPPORT) 1333 error = 0; 1334 1335 if (error) 1336 break; 1337 } 1338 1339 return error; 1340 } 1341 1342 int 1343 rt_ifa_purge_walker(struct rtentry *rt, void *vifa, unsigned int rtableid) 1344 { 1345 struct ifaddr *ifa = vifa; 1346 1347 if (rt->rt_ifa == ifa) 1348 return EEXIST; 1349 1350 return 0; 1351 } 1352 1353 /* 1354 * Route timer routines. These routes allow functions to be called 1355 * for various routes at any time. This is useful in supporting 1356 * path MTU discovery and redirect route deletion. 1357 * 1358 * This is similar to some BSDI internal functions, but it provides 1359 * for multiple queues for efficiency's sake... 1360 */ 1361 1362 LIST_HEAD(, rttimer_queue) rttimer_queue_head; 1363 static int rt_init_done = 0; 1364 1365 #define RTTIMER_CALLOUT(r) { \ 1366 if (r->rtt_func != NULL) { \ 1367 (*r->rtt_func)(r->rtt_rt, r); \ 1368 } else { \ 1369 struct ifnet *ifp; \ 1370 \ 1371 ifp = if_get(r->rtt_rt->rt_ifidx); \ 1372 if (ifp != NULL) \ 1373 rtdeletemsg(r->rtt_rt, ifp, r->rtt_tableid); \ 1374 if_put(ifp); \ 1375 } \ 1376 } 1377 1378 /* 1379 * Some subtle order problems with domain initialization mean that 1380 * we cannot count on this being run from rt_init before various 1381 * protocol initializations are done. Therefore, we make sure 1382 * that this is run when the first queue is added... 1383 */ 1384 1385 void 1386 rt_timer_init(void) 1387 { 1388 static struct timeout rt_timer_timeout; 1389 1390 if (rt_init_done) 1391 panic("rt_timer_init: already initialized"); 1392 1393 pool_init(&rttimer_pool, sizeof(struct rttimer), 0, IPL_SOFTNET, 0, 1394 "rttmr", NULL); 1395 1396 LIST_INIT(&rttimer_queue_head); 1397 timeout_set_proc(&rt_timer_timeout, rt_timer_timer, &rt_timer_timeout); 1398 timeout_add_sec(&rt_timer_timeout, 1); 1399 rt_init_done = 1; 1400 } 1401 1402 struct rttimer_queue * 1403 rt_timer_queue_create(u_int timeout) 1404 { 1405 struct rttimer_queue *rtq; 1406 1407 if (rt_init_done == 0) 1408 rt_timer_init(); 1409 1410 if ((rtq = malloc(sizeof(*rtq), M_RTABLE, M_NOWAIT|M_ZERO)) == NULL) 1411 return (NULL); 1412 1413 rtq->rtq_timeout = timeout; 1414 rtq->rtq_count = 0; 1415 TAILQ_INIT(&rtq->rtq_head); 1416 LIST_INSERT_HEAD(&rttimer_queue_head, rtq, rtq_link); 1417 1418 return (rtq); 1419 } 1420 1421 void 1422 rt_timer_queue_change(struct rttimer_queue *rtq, long timeout) 1423 { 1424 rtq->rtq_timeout = timeout; 1425 } 1426 1427 void 1428 rt_timer_queue_destroy(struct rttimer_queue *rtq) 1429 { 1430 struct rttimer *r; 1431 1432 NET_ASSERT_LOCKED(); 1433 1434 while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL) { 1435 LIST_REMOVE(r, rtt_link); 1436 TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next); 1437 RTTIMER_CALLOUT(r); 1438 pool_put(&rttimer_pool, r); 1439 if (rtq->rtq_count > 0) 1440 rtq->rtq_count--; 1441 else 1442 printf("rt_timer_queue_destroy: rtq_count reached 0\n"); 1443 } 1444 1445 LIST_REMOVE(rtq, rtq_link); 1446 free(rtq, M_RTABLE, sizeof(*rtq)); 1447 } 1448 1449 unsigned long 1450 rt_timer_queue_count(struct rttimer_queue *rtq) 1451 { 1452 return (rtq->rtq_count); 1453 } 1454 1455 void 1456 rt_timer_remove_all(struct rtentry *rt) 1457 { 1458 struct rttimer *r; 1459 1460 while ((r = LIST_FIRST(&rt->rt_timer)) != NULL) { 1461 LIST_REMOVE(r, rtt_link); 1462 TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next); 1463 if (r->rtt_queue->rtq_count > 0) 1464 r->rtt_queue->rtq_count--; 1465 else 1466 printf("rt_timer_remove_all: rtq_count reached 0\n"); 1467 pool_put(&rttimer_pool, r); 1468 } 1469 } 1470 1471 int 1472 rt_timer_add(struct rtentry *rt, void (*func)(struct rtentry *, 1473 struct rttimer *), struct rttimer_queue *queue, u_int rtableid) 1474 { 1475 struct rttimer *r; 1476 long current_time; 1477 1478 current_time = time_uptime; 1479 rt->rt_expire = time_uptime + queue->rtq_timeout; 1480 1481 /* 1482 * If there's already a timer with this action, destroy it before 1483 * we add a new one. 1484 */ 1485 LIST_FOREACH(r, &rt->rt_timer, rtt_link) { 1486 if (r->rtt_func == func) { 1487 LIST_REMOVE(r, rtt_link); 1488 TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next); 1489 if (r->rtt_queue->rtq_count > 0) 1490 r->rtt_queue->rtq_count--; 1491 else 1492 printf("rt_timer_add: rtq_count reached 0\n"); 1493 pool_put(&rttimer_pool, r); 1494 break; /* only one per list, so we can quit... */ 1495 } 1496 } 1497 1498 r = pool_get(&rttimer_pool, PR_NOWAIT | PR_ZERO); 1499 if (r == NULL) 1500 return (ENOBUFS); 1501 1502 r->rtt_rt = rt; 1503 r->rtt_time = current_time; 1504 r->rtt_func = func; 1505 r->rtt_queue = queue; 1506 r->rtt_tableid = rtableid; 1507 LIST_INSERT_HEAD(&rt->rt_timer, r, rtt_link); 1508 TAILQ_INSERT_TAIL(&queue->rtq_head, r, rtt_next); 1509 r->rtt_queue->rtq_count++; 1510 1511 return (0); 1512 } 1513 1514 void 1515 rt_timer_timer(void *arg) 1516 { 1517 struct timeout *to = (struct timeout *)arg; 1518 struct rttimer_queue *rtq; 1519 struct rttimer *r; 1520 long current_time; 1521 1522 current_time = time_uptime; 1523 1524 NET_LOCK(); 1525 LIST_FOREACH(rtq, &rttimer_queue_head, rtq_link) { 1526 while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL && 1527 (r->rtt_time + rtq->rtq_timeout) < current_time) { 1528 LIST_REMOVE(r, rtt_link); 1529 TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next); 1530 RTTIMER_CALLOUT(r); 1531 pool_put(&rttimer_pool, r); 1532 if (rtq->rtq_count > 0) 1533 rtq->rtq_count--; 1534 else 1535 printf("rt_timer_timer: rtq_count reached 0\n"); 1536 } 1537 } 1538 NET_UNLOCK(); 1539 1540 timeout_add_sec(to, 1); 1541 } 1542 1543 #ifdef MPLS 1544 int 1545 rt_mpls_set(struct rtentry *rt, struct sockaddr *src, uint8_t op) 1546 { 1547 struct sockaddr_mpls *psa_mpls = (struct sockaddr_mpls *)src; 1548 struct rt_mpls *rt_mpls; 1549 1550 if (psa_mpls == NULL && op != MPLS_OP_POP) 1551 return (EOPNOTSUPP); 1552 if (psa_mpls != NULL && psa_mpls->smpls_len != sizeof(*psa_mpls)) 1553 return (EINVAL); 1554 if (psa_mpls != NULL && psa_mpls->smpls_family != AF_MPLS) 1555 return (EAFNOSUPPORT); 1556 1557 rt->rt_llinfo = malloc(sizeof(struct rt_mpls), M_TEMP, M_NOWAIT|M_ZERO); 1558 if (rt->rt_llinfo == NULL) 1559 return (ENOMEM); 1560 1561 rt_mpls = (struct rt_mpls *)rt->rt_llinfo; 1562 if (psa_mpls != NULL) 1563 rt_mpls->mpls_label = psa_mpls->smpls_label; 1564 rt_mpls->mpls_operation = op; 1565 /* XXX: set experimental bits */ 1566 rt->rt_flags |= RTF_MPLS; 1567 1568 return (0); 1569 } 1570 1571 void 1572 rt_mpls_clear(struct rtentry *rt) 1573 { 1574 if (rt->rt_llinfo != NULL && rt->rt_flags & RTF_MPLS) { 1575 free(rt->rt_llinfo, M_TEMP, sizeof(struct rt_mpls)); 1576 rt->rt_llinfo = NULL; 1577 } 1578 rt->rt_flags &= ~RTF_MPLS; 1579 } 1580 #endif 1581 1582 u_int16_t 1583 rtlabel_name2id(char *name) 1584 { 1585 struct rt_label *label, *p; 1586 u_int16_t new_id = 1; 1587 1588 if (!name[0]) 1589 return (0); 1590 1591 TAILQ_FOREACH(label, &rt_labels, rtl_entry) 1592 if (strcmp(name, label->rtl_name) == 0) { 1593 label->rtl_ref++; 1594 return (label->rtl_id); 1595 } 1596 1597 /* 1598 * to avoid fragmentation, we do a linear search from the beginning 1599 * and take the first free slot we find. if there is none or the list 1600 * is empty, append a new entry at the end. 1601 */ 1602 TAILQ_FOREACH(p, &rt_labels, rtl_entry) { 1603 if (p->rtl_id != new_id) 1604 break; 1605 new_id = p->rtl_id + 1; 1606 } 1607 if (new_id > LABELID_MAX) 1608 return (0); 1609 1610 label = malloc(sizeof(*label), M_RTABLE, M_NOWAIT|M_ZERO); 1611 if (label == NULL) 1612 return (0); 1613 strlcpy(label->rtl_name, name, sizeof(label->rtl_name)); 1614 label->rtl_id = new_id; 1615 label->rtl_ref++; 1616 1617 if (p != NULL) /* insert new entry before p */ 1618 TAILQ_INSERT_BEFORE(p, label, rtl_entry); 1619 else /* either list empty or no free slot in between */ 1620 TAILQ_INSERT_TAIL(&rt_labels, label, rtl_entry); 1621 1622 return (label->rtl_id); 1623 } 1624 1625 const char * 1626 rtlabel_id2name(u_int16_t id) 1627 { 1628 struct rt_label *label; 1629 1630 TAILQ_FOREACH(label, &rt_labels, rtl_entry) 1631 if (label->rtl_id == id) 1632 return (label->rtl_name); 1633 1634 return (NULL); 1635 } 1636 1637 struct sockaddr * 1638 rtlabel_id2sa(u_int16_t labelid, struct sockaddr_rtlabel *sa_rl) 1639 { 1640 const char *label; 1641 1642 if (labelid == 0 || (label = rtlabel_id2name(labelid)) == NULL) 1643 return (NULL); 1644 1645 bzero(sa_rl, sizeof(*sa_rl)); 1646 sa_rl->sr_len = sizeof(*sa_rl); 1647 sa_rl->sr_family = AF_UNSPEC; 1648 strlcpy(sa_rl->sr_label, label, sizeof(sa_rl->sr_label)); 1649 1650 return ((struct sockaddr *)sa_rl); 1651 } 1652 1653 void 1654 rtlabel_unref(u_int16_t id) 1655 { 1656 struct rt_label *p, *next; 1657 1658 if (id == 0) 1659 return; 1660 1661 TAILQ_FOREACH_SAFE(p, &rt_labels, rtl_entry, next) { 1662 if (id == p->rtl_id) { 1663 if (--p->rtl_ref == 0) { 1664 TAILQ_REMOVE(&rt_labels, p, rtl_entry); 1665 free(p, M_RTABLE, sizeof(*p)); 1666 } 1667 break; 1668 } 1669 } 1670 } 1671 1672 int 1673 rt_if_track(struct ifnet *ifp) 1674 { 1675 unsigned int rtableid; 1676 struct rtentry *rt = NULL; 1677 int i, error; 1678 1679 for (rtableid = 0; rtableid < rtmap_limit; rtableid++) { 1680 /* skip rtables that are not in the rdomain of the ifp */ 1681 if (rtable_l2(rtableid) != ifp->if_rdomain) 1682 continue; 1683 for (i = 1; i <= AF_MAX; i++) { 1684 if (!rtable_mpath_capable(rtableid, i)) 1685 continue; 1686 1687 do { 1688 error = rtable_walk(rtableid, i, &rt, 1689 rt_if_linkstate_change, ifp); 1690 if (rt != NULL && error == EEXIST) { 1691 error = rtdeletemsg(rt, ifp, rtableid); 1692 if (error == 0) 1693 error = EAGAIN; 1694 } 1695 rtfree(rt); 1696 rt = NULL; 1697 } while (error == EAGAIN); 1698 1699 if (error == EAFNOSUPPORT) 1700 error = 0; 1701 1702 if (error) 1703 break; 1704 } 1705 } 1706 1707 return (error); 1708 } 1709 1710 int 1711 rt_if_linkstate_change(struct rtentry *rt, void *arg, u_int id) 1712 { 1713 struct ifnet *ifp = arg; 1714 struct sockaddr_in6 sa_mask; 1715 int error; 1716 1717 if (rt->rt_ifidx != ifp->if_index) 1718 return (0); 1719 1720 /* Local routes are always usable. */ 1721 if (rt->rt_flags & RTF_LOCAL) { 1722 rt->rt_flags |= RTF_UP; 1723 return (0); 1724 } 1725 1726 if (LINK_STATE_IS_UP(ifp->if_link_state) && ifp->if_flags & IFF_UP) { 1727 if (ISSET(rt->rt_flags, RTF_UP)) 1728 return (0); 1729 1730 /* bring route up */ 1731 rt->rt_flags |= RTF_UP; 1732 error = rtable_mpath_reprio(id, rt_key(rt), rt_plen(rt), 1733 rt->rt_priority & RTP_MASK, rt); 1734 } else { 1735 /* 1736 * Remove redirected and cloned routes (mainly ARP) 1737 * from down interfaces so we have a chance to get 1738 * new routes from a better source. 1739 */ 1740 if (ISSET(rt->rt_flags, RTF_CLONED|RTF_DYNAMIC) && 1741 !ISSET(rt->rt_flags, RTF_CACHED|RTF_BFD)) { 1742 return (EEXIST); 1743 } 1744 1745 if (!ISSET(rt->rt_flags, RTF_UP)) 1746 return (0); 1747 1748 /* take route down */ 1749 rt->rt_flags &= ~RTF_UP; 1750 error = rtable_mpath_reprio(id, rt_key(rt), rt_plen(rt), 1751 rt->rt_priority | RTP_DOWN, rt); 1752 } 1753 if_group_routechange(rt_key(rt), rt_plen2mask(rt, &sa_mask)); 1754 1755 return (error); 1756 } 1757 1758 struct sockaddr * 1759 rt_plentosa(sa_family_t af, int plen, struct sockaddr_in6 *sa_mask) 1760 { 1761 struct sockaddr_in *sin = (struct sockaddr_in *)sa_mask; 1762 #ifdef INET6 1763 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sa_mask; 1764 #endif 1765 1766 KASSERT(plen >= 0 || plen == -1); 1767 1768 if (plen == -1) 1769 return (NULL); 1770 1771 memset(sa_mask, 0, sizeof(*sa_mask)); 1772 1773 switch (af) { 1774 case AF_INET: 1775 sin->sin_family = AF_INET; 1776 sin->sin_len = sizeof(struct sockaddr_in); 1777 in_prefixlen2mask(&sin->sin_addr, plen); 1778 break; 1779 #ifdef INET6 1780 case AF_INET6: 1781 sin6->sin6_family = AF_INET6; 1782 sin6->sin6_len = sizeof(struct sockaddr_in6); 1783 in6_prefixlen2mask(&sin6->sin6_addr, plen); 1784 break; 1785 #endif /* INET6 */ 1786 default: 1787 return (NULL); 1788 } 1789 1790 return ((struct sockaddr *)sa_mask); 1791 } 1792 1793 struct sockaddr * 1794 rt_plen2mask(struct rtentry *rt, struct sockaddr_in6 *sa_mask) 1795 { 1796 return (rt_plentosa(rt_key(rt)->sa_family, rt_plen(rt), sa_mask)); 1797 } 1798 1799 #ifdef DDB 1800 #include <machine/db_machdep.h> 1801 #include <ddb/db_output.h> 1802 1803 void 1804 db_print_sa(struct sockaddr *sa) 1805 { 1806 int len; 1807 u_char *p; 1808 1809 if (sa == NULL) { 1810 db_printf("[NULL]"); 1811 return; 1812 } 1813 1814 p = (u_char *)sa; 1815 len = sa->sa_len; 1816 db_printf("["); 1817 while (len > 0) { 1818 db_printf("%d", *p); 1819 p++; 1820 len--; 1821 if (len) 1822 db_printf(","); 1823 } 1824 db_printf("]\n"); 1825 } 1826 1827 void 1828 db_print_ifa(struct ifaddr *ifa) 1829 { 1830 if (ifa == NULL) 1831 return; 1832 db_printf(" ifa_addr="); 1833 db_print_sa(ifa->ifa_addr); 1834 db_printf(" ifa_dsta="); 1835 db_print_sa(ifa->ifa_dstaddr); 1836 db_printf(" ifa_mask="); 1837 db_print_sa(ifa->ifa_netmask); 1838 db_printf(" flags=0x%x, refcnt=%d, metric=%d\n", 1839 ifa->ifa_flags, ifa->ifa_refcnt, ifa->ifa_metric); 1840 } 1841 1842 /* 1843 * Function to pass to rtalble_walk(). 1844 * Return non-zero error to abort walk. 1845 */ 1846 int 1847 db_show_rtentry(struct rtentry *rt, void *w, unsigned int id) 1848 { 1849 db_printf("rtentry=%p", rt); 1850 1851 db_printf(" flags=0x%x refcnt=%d use=%llu expire=%lld rtableid=%u\n", 1852 rt->rt_flags, rt->rt_refcnt, rt->rt_use, rt->rt_expire, id); 1853 1854 db_printf(" key="); db_print_sa(rt_key(rt)); 1855 db_printf(" plen=%d", rt_plen(rt)); 1856 db_printf(" gw="); db_print_sa(rt->rt_gateway); 1857 db_printf(" ifidx=%u ", rt->rt_ifidx); 1858 db_printf(" ifa=%p\n", rt->rt_ifa); 1859 db_print_ifa(rt->rt_ifa); 1860 1861 db_printf(" gwroute=%p llinfo=%p\n", rt->rt_gwroute, rt->rt_llinfo); 1862 return (0); 1863 } 1864 1865 /* 1866 * Function to print all the route trees. 1867 * Use this from ddb: "call db_show_arptab" 1868 */ 1869 int 1870 db_show_arptab(void) 1871 { 1872 db_printf("Route tree for AF_INET\n"); 1873 rtable_walk(0, AF_INET, NULL, db_show_rtentry, NULL); 1874 return (0); 1875 } 1876 #endif /* DDB */ 1877