1 /* $OpenBSD: route.c,v 1.388 2020/01/08 10:02:55 claudio 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 KASSERT(rdomain == rtable_l2(rdomain)); 1105 1106 memset(&info, 0, sizeof(info)); 1107 info.rti_ifa = ifa; 1108 info.rti_flags = flags; 1109 info.rti_info[RTAX_DST] = dst; 1110 if (flags & RTF_LLINFO) 1111 info.rti_info[RTAX_GATEWAY] = sdltosa(ifp->if_sadl); 1112 else 1113 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr; 1114 info.rti_info[RTAX_LABEL] = rtlabel_id2sa(ifp->if_rtlabelid, &sa_rl); 1115 1116 #ifdef MPLS 1117 if ((flags & RTF_MPLS) == RTF_MPLS) 1118 info.rti_mpls = MPLS_OP_POP; 1119 #endif /* MPLS */ 1120 1121 if ((flags & RTF_HOST) == 0) 1122 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask; 1123 1124 if (flags & (RTF_LOCAL|RTF_BROADCAST)) 1125 prio = RTP_LOCAL; 1126 1127 if (flags & RTF_CONNECTED) 1128 prio = ifp->if_priority + RTP_CONNECTED; 1129 1130 error = rtrequest(RTM_ADD, &info, prio, &rt, rdomain); 1131 if (error == 0) { 1132 /* 1133 * A local route is created for every address configured 1134 * on an interface, so use this information to notify 1135 * userland that a new address has been added. 1136 */ 1137 if (flags & RTF_LOCAL) 1138 rtm_addr(RTM_NEWADDR, ifa); 1139 rtm_send(rt, RTM_ADD, 0, rdomain); 1140 rtfree(rt); 1141 } 1142 return (error); 1143 } 1144 1145 int 1146 rt_ifa_del(struct ifaddr *ifa, int flags, struct sockaddr *dst, 1147 unsigned int rdomain) 1148 { 1149 struct ifnet *ifp = ifa->ifa_ifp; 1150 struct rtentry *rt; 1151 struct mbuf *m = NULL; 1152 struct sockaddr *deldst; 1153 struct rt_addrinfo info; 1154 struct sockaddr_rtlabel sa_rl; 1155 uint8_t prio = ifp->if_priority + RTP_STATIC; 1156 int error; 1157 1158 KASSERT(rdomain == rtable_l2(rdomain)); 1159 1160 if ((flags & RTF_HOST) == 0 && ifa->ifa_netmask) { 1161 m = m_get(M_DONTWAIT, MT_SONAME); 1162 if (m == NULL) 1163 return (ENOBUFS); 1164 deldst = mtod(m, struct sockaddr *); 1165 rt_maskedcopy(dst, deldst, ifa->ifa_netmask); 1166 dst = deldst; 1167 } 1168 1169 memset(&info, 0, sizeof(info)); 1170 info.rti_ifa = ifa; 1171 info.rti_flags = flags; 1172 info.rti_info[RTAX_DST] = dst; 1173 if ((flags & RTF_LLINFO) == 0) 1174 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr; 1175 info.rti_info[RTAX_LABEL] = rtlabel_id2sa(ifp->if_rtlabelid, &sa_rl); 1176 1177 if ((flags & RTF_HOST) == 0) 1178 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask; 1179 1180 if (flags & (RTF_LOCAL|RTF_BROADCAST)) 1181 prio = RTP_LOCAL; 1182 1183 if (flags & RTF_CONNECTED) 1184 prio = ifp->if_priority + RTP_CONNECTED; 1185 1186 error = rtrequest_delete(&info, prio, ifp, &rt, rdomain); 1187 if (error == 0) { 1188 rtm_send(rt, RTM_DELETE, 0, rdomain); 1189 if (flags & RTF_LOCAL) 1190 rtm_addr(RTM_DELADDR, ifa); 1191 rtfree(rt); 1192 } 1193 m_free(m); 1194 1195 return (error); 1196 } 1197 1198 /* 1199 * Add ifa's address as a local rtentry. 1200 */ 1201 int 1202 rt_ifa_addlocal(struct ifaddr *ifa) 1203 { 1204 struct ifnet *ifp = ifa->ifa_ifp; 1205 struct rtentry *rt; 1206 u_int flags = RTF_HOST|RTF_LOCAL; 1207 int error = 0; 1208 1209 /* 1210 * If the configured address correspond to the magical "any" 1211 * address do not add a local route entry because that might 1212 * corrupt the routing tree which uses this value for the 1213 * default routes. 1214 */ 1215 switch (ifa->ifa_addr->sa_family) { 1216 case AF_INET: 1217 if (satosin(ifa->ifa_addr)->sin_addr.s_addr == INADDR_ANY) 1218 return (0); 1219 break; 1220 #ifdef INET6 1221 case AF_INET6: 1222 if (IN6_ARE_ADDR_EQUAL(&satosin6(ifa->ifa_addr)->sin6_addr, 1223 &in6addr_any)) 1224 return (0); 1225 break; 1226 #endif 1227 default: 1228 break; 1229 } 1230 1231 if (!ISSET(ifp->if_flags, (IFF_LOOPBACK|IFF_POINTOPOINT))) 1232 flags |= RTF_LLINFO; 1233 1234 /* If there is no local entry, allocate one. */ 1235 rt = rtalloc(ifa->ifa_addr, 0, ifp->if_rdomain); 1236 if (rt == NULL || ISSET(rt->rt_flags, flags) != flags) { 1237 error = rt_ifa_add(ifa, flags | RTF_MPATH, ifa->ifa_addr, 1238 ifp->if_rdomain); 1239 } 1240 rtfree(rt); 1241 1242 return (error); 1243 } 1244 1245 /* 1246 * Remove local rtentry of ifa's addresss if it exists. 1247 */ 1248 int 1249 rt_ifa_dellocal(struct ifaddr *ifa) 1250 { 1251 struct ifnet *ifp = ifa->ifa_ifp; 1252 struct rtentry *rt; 1253 u_int flags = RTF_HOST|RTF_LOCAL; 1254 int error = 0; 1255 1256 /* 1257 * We do not add local routes for such address, so do not bother 1258 * removing them. 1259 */ 1260 switch (ifa->ifa_addr->sa_family) { 1261 case AF_INET: 1262 if (satosin(ifa->ifa_addr)->sin_addr.s_addr == INADDR_ANY) 1263 return (0); 1264 break; 1265 #ifdef INET6 1266 case AF_INET6: 1267 if (IN6_ARE_ADDR_EQUAL(&satosin6(ifa->ifa_addr)->sin6_addr, 1268 &in6addr_any)) 1269 return (0); 1270 break; 1271 #endif 1272 default: 1273 break; 1274 } 1275 1276 if (!ISSET(ifp->if_flags, (IFF_LOOPBACK|IFF_POINTOPOINT))) 1277 flags |= RTF_LLINFO; 1278 1279 /* 1280 * Before deleting, check if a corresponding local host 1281 * route surely exists. With this check, we can avoid to 1282 * delete an interface direct route whose destination is same 1283 * as the address being removed. This can happen when removing 1284 * a subnet-router anycast address on an interface attached 1285 * to a shared medium. 1286 */ 1287 rt = rtalloc(ifa->ifa_addr, 0, ifp->if_rdomain); 1288 if (rt != NULL && ISSET(rt->rt_flags, flags) == flags) { 1289 error = rt_ifa_del(ifa, flags, ifa->ifa_addr, 1290 ifp->if_rdomain); 1291 } 1292 rtfree(rt); 1293 1294 return (error); 1295 } 1296 1297 /* 1298 * Remove all addresses attached to ``ifa''. 1299 */ 1300 int 1301 rt_ifa_purge(struct ifaddr *ifa) 1302 { 1303 struct ifnet *ifp = ifa->ifa_ifp; 1304 struct rtentry *rt = NULL; 1305 unsigned int rtableid; 1306 int error, af = ifa->ifa_addr->sa_family; 1307 1308 KASSERT(ifp != NULL); 1309 1310 for (rtableid = 0; rtableid < rtmap_limit; rtableid++) { 1311 /* skip rtables that are not in the rdomain of the ifp */ 1312 if (rtable_l2(rtableid) != ifp->if_rdomain) 1313 continue; 1314 1315 do { 1316 error = rtable_walk(rtableid, af, &rt, 1317 rt_ifa_purge_walker, ifa); 1318 if (rt != NULL && error == EEXIST) { 1319 error = rtdeletemsg(rt, ifp, rtableid); 1320 if (error == 0) 1321 error = EAGAIN; 1322 } 1323 rtfree(rt); 1324 rt = NULL; 1325 } while (error == EAGAIN); 1326 1327 if (error == EAFNOSUPPORT) 1328 error = 0; 1329 1330 if (error) 1331 break; 1332 } 1333 1334 return error; 1335 } 1336 1337 int 1338 rt_ifa_purge_walker(struct rtentry *rt, void *vifa, unsigned int rtableid) 1339 { 1340 struct ifaddr *ifa = vifa; 1341 1342 if (rt->rt_ifa == ifa) 1343 return EEXIST; 1344 1345 return 0; 1346 } 1347 1348 /* 1349 * Route timer routines. These routes allow functions to be called 1350 * for various routes at any time. This is useful in supporting 1351 * path MTU discovery and redirect route deletion. 1352 * 1353 * This is similar to some BSDI internal functions, but it provides 1354 * for multiple queues for efficiency's sake... 1355 */ 1356 1357 LIST_HEAD(, rttimer_queue) rttimer_queue_head; 1358 static int rt_init_done = 0; 1359 1360 #define RTTIMER_CALLOUT(r) { \ 1361 if (r->rtt_func != NULL) { \ 1362 (*r->rtt_func)(r->rtt_rt, r); \ 1363 } else { \ 1364 struct ifnet *ifp; \ 1365 \ 1366 ifp = if_get(r->rtt_rt->rt_ifidx); \ 1367 if (ifp != NULL) \ 1368 rtdeletemsg(r->rtt_rt, ifp, r->rtt_tableid); \ 1369 if_put(ifp); \ 1370 } \ 1371 } 1372 1373 /* 1374 * Some subtle order problems with domain initialization mean that 1375 * we cannot count on this being run from rt_init before various 1376 * protocol initializations are done. Therefore, we make sure 1377 * that this is run when the first queue is added... 1378 */ 1379 1380 void 1381 rt_timer_init(void) 1382 { 1383 static struct timeout rt_timer_timeout; 1384 1385 if (rt_init_done) 1386 panic("rt_timer_init: already initialized"); 1387 1388 pool_init(&rttimer_pool, sizeof(struct rttimer), 0, IPL_SOFTNET, 0, 1389 "rttmr", NULL); 1390 1391 LIST_INIT(&rttimer_queue_head); 1392 timeout_set_proc(&rt_timer_timeout, rt_timer_timer, &rt_timer_timeout); 1393 timeout_add_sec(&rt_timer_timeout, 1); 1394 rt_init_done = 1; 1395 } 1396 1397 struct rttimer_queue * 1398 rt_timer_queue_create(u_int timeout) 1399 { 1400 struct rttimer_queue *rtq; 1401 1402 if (rt_init_done == 0) 1403 rt_timer_init(); 1404 1405 if ((rtq = malloc(sizeof(*rtq), M_RTABLE, M_NOWAIT|M_ZERO)) == NULL) 1406 return (NULL); 1407 1408 rtq->rtq_timeout = timeout; 1409 rtq->rtq_count = 0; 1410 TAILQ_INIT(&rtq->rtq_head); 1411 LIST_INSERT_HEAD(&rttimer_queue_head, rtq, rtq_link); 1412 1413 return (rtq); 1414 } 1415 1416 void 1417 rt_timer_queue_change(struct rttimer_queue *rtq, long timeout) 1418 { 1419 rtq->rtq_timeout = timeout; 1420 } 1421 1422 void 1423 rt_timer_queue_destroy(struct rttimer_queue *rtq) 1424 { 1425 struct rttimer *r; 1426 1427 NET_ASSERT_LOCKED(); 1428 1429 while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL) { 1430 LIST_REMOVE(r, rtt_link); 1431 TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next); 1432 RTTIMER_CALLOUT(r); 1433 pool_put(&rttimer_pool, r); 1434 if (rtq->rtq_count > 0) 1435 rtq->rtq_count--; 1436 else 1437 printf("rt_timer_queue_destroy: rtq_count reached 0\n"); 1438 } 1439 1440 LIST_REMOVE(rtq, rtq_link); 1441 free(rtq, M_RTABLE, sizeof(*rtq)); 1442 } 1443 1444 unsigned long 1445 rt_timer_queue_count(struct rttimer_queue *rtq) 1446 { 1447 return (rtq->rtq_count); 1448 } 1449 1450 void 1451 rt_timer_remove_all(struct rtentry *rt) 1452 { 1453 struct rttimer *r; 1454 1455 while ((r = LIST_FIRST(&rt->rt_timer)) != NULL) { 1456 LIST_REMOVE(r, rtt_link); 1457 TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next); 1458 if (r->rtt_queue->rtq_count > 0) 1459 r->rtt_queue->rtq_count--; 1460 else 1461 printf("rt_timer_remove_all: rtq_count reached 0\n"); 1462 pool_put(&rttimer_pool, r); 1463 } 1464 } 1465 1466 int 1467 rt_timer_add(struct rtentry *rt, void (*func)(struct rtentry *, 1468 struct rttimer *), struct rttimer_queue *queue, u_int rtableid) 1469 { 1470 struct rttimer *r; 1471 long current_time; 1472 1473 current_time = time_uptime; 1474 rt->rt_expire = time_uptime + queue->rtq_timeout; 1475 1476 /* 1477 * If there's already a timer with this action, destroy it before 1478 * we add a new one. 1479 */ 1480 LIST_FOREACH(r, &rt->rt_timer, rtt_link) { 1481 if (r->rtt_func == func) { 1482 LIST_REMOVE(r, rtt_link); 1483 TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next); 1484 if (r->rtt_queue->rtq_count > 0) 1485 r->rtt_queue->rtq_count--; 1486 else 1487 printf("rt_timer_add: rtq_count reached 0\n"); 1488 pool_put(&rttimer_pool, r); 1489 break; /* only one per list, so we can quit... */ 1490 } 1491 } 1492 1493 r = pool_get(&rttimer_pool, PR_NOWAIT | PR_ZERO); 1494 if (r == NULL) 1495 return (ENOBUFS); 1496 1497 r->rtt_rt = rt; 1498 r->rtt_time = current_time; 1499 r->rtt_func = func; 1500 r->rtt_queue = queue; 1501 r->rtt_tableid = rtableid; 1502 LIST_INSERT_HEAD(&rt->rt_timer, r, rtt_link); 1503 TAILQ_INSERT_TAIL(&queue->rtq_head, r, rtt_next); 1504 r->rtt_queue->rtq_count++; 1505 1506 return (0); 1507 } 1508 1509 void 1510 rt_timer_timer(void *arg) 1511 { 1512 struct timeout *to = (struct timeout *)arg; 1513 struct rttimer_queue *rtq; 1514 struct rttimer *r; 1515 long current_time; 1516 1517 current_time = time_uptime; 1518 1519 NET_LOCK(); 1520 LIST_FOREACH(rtq, &rttimer_queue_head, rtq_link) { 1521 while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL && 1522 (r->rtt_time + rtq->rtq_timeout) < current_time) { 1523 LIST_REMOVE(r, rtt_link); 1524 TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next); 1525 RTTIMER_CALLOUT(r); 1526 pool_put(&rttimer_pool, r); 1527 if (rtq->rtq_count > 0) 1528 rtq->rtq_count--; 1529 else 1530 printf("rt_timer_timer: rtq_count reached 0\n"); 1531 } 1532 } 1533 NET_UNLOCK(); 1534 1535 timeout_add_sec(to, 1); 1536 } 1537 1538 #ifdef MPLS 1539 int 1540 rt_mpls_set(struct rtentry *rt, struct sockaddr *src, uint8_t op) 1541 { 1542 struct sockaddr_mpls *psa_mpls = (struct sockaddr_mpls *)src; 1543 struct rt_mpls *rt_mpls; 1544 1545 if (psa_mpls == NULL && op != MPLS_OP_POP) 1546 return (EOPNOTSUPP); 1547 if (psa_mpls != NULL && psa_mpls->smpls_len != sizeof(*psa_mpls)) 1548 return (EINVAL); 1549 if (psa_mpls != NULL && psa_mpls->smpls_family != AF_MPLS) 1550 return (EAFNOSUPPORT); 1551 1552 rt->rt_llinfo = malloc(sizeof(struct rt_mpls), M_TEMP, M_NOWAIT|M_ZERO); 1553 if (rt->rt_llinfo == NULL) 1554 return (ENOMEM); 1555 1556 rt_mpls = (struct rt_mpls *)rt->rt_llinfo; 1557 if (psa_mpls != NULL) 1558 rt_mpls->mpls_label = psa_mpls->smpls_label; 1559 rt_mpls->mpls_operation = op; 1560 /* XXX: set experimental bits */ 1561 rt->rt_flags |= RTF_MPLS; 1562 1563 return (0); 1564 } 1565 1566 void 1567 rt_mpls_clear(struct rtentry *rt) 1568 { 1569 if (rt->rt_llinfo != NULL && rt->rt_flags & RTF_MPLS) { 1570 free(rt->rt_llinfo, M_TEMP, sizeof(struct rt_mpls)); 1571 rt->rt_llinfo = NULL; 1572 } 1573 rt->rt_flags &= ~RTF_MPLS; 1574 } 1575 #endif 1576 1577 u_int16_t 1578 rtlabel_name2id(char *name) 1579 { 1580 struct rt_label *label, *p; 1581 u_int16_t new_id = 1; 1582 1583 if (!name[0]) 1584 return (0); 1585 1586 TAILQ_FOREACH(label, &rt_labels, rtl_entry) 1587 if (strcmp(name, label->rtl_name) == 0) { 1588 label->rtl_ref++; 1589 return (label->rtl_id); 1590 } 1591 1592 /* 1593 * to avoid fragmentation, we do a linear search from the beginning 1594 * and take the first free slot we find. if there is none or the list 1595 * is empty, append a new entry at the end. 1596 */ 1597 TAILQ_FOREACH(p, &rt_labels, rtl_entry) { 1598 if (p->rtl_id != new_id) 1599 break; 1600 new_id = p->rtl_id + 1; 1601 } 1602 if (new_id > LABELID_MAX) 1603 return (0); 1604 1605 label = malloc(sizeof(*label), M_RTABLE, M_NOWAIT|M_ZERO); 1606 if (label == NULL) 1607 return (0); 1608 strlcpy(label->rtl_name, name, sizeof(label->rtl_name)); 1609 label->rtl_id = new_id; 1610 label->rtl_ref++; 1611 1612 if (p != NULL) /* insert new entry before p */ 1613 TAILQ_INSERT_BEFORE(p, label, rtl_entry); 1614 else /* either list empty or no free slot in between */ 1615 TAILQ_INSERT_TAIL(&rt_labels, label, rtl_entry); 1616 1617 return (label->rtl_id); 1618 } 1619 1620 const char * 1621 rtlabel_id2name(u_int16_t id) 1622 { 1623 struct rt_label *label; 1624 1625 TAILQ_FOREACH(label, &rt_labels, rtl_entry) 1626 if (label->rtl_id == id) 1627 return (label->rtl_name); 1628 1629 return (NULL); 1630 } 1631 1632 struct sockaddr * 1633 rtlabel_id2sa(u_int16_t labelid, struct sockaddr_rtlabel *sa_rl) 1634 { 1635 const char *label; 1636 1637 if (labelid == 0 || (label = rtlabel_id2name(labelid)) == NULL) 1638 return (NULL); 1639 1640 bzero(sa_rl, sizeof(*sa_rl)); 1641 sa_rl->sr_len = sizeof(*sa_rl); 1642 sa_rl->sr_family = AF_UNSPEC; 1643 strlcpy(sa_rl->sr_label, label, sizeof(sa_rl->sr_label)); 1644 1645 return ((struct sockaddr *)sa_rl); 1646 } 1647 1648 void 1649 rtlabel_unref(u_int16_t id) 1650 { 1651 struct rt_label *p, *next; 1652 1653 if (id == 0) 1654 return; 1655 1656 TAILQ_FOREACH_SAFE(p, &rt_labels, rtl_entry, next) { 1657 if (id == p->rtl_id) { 1658 if (--p->rtl_ref == 0) { 1659 TAILQ_REMOVE(&rt_labels, p, rtl_entry); 1660 free(p, M_RTABLE, sizeof(*p)); 1661 } 1662 break; 1663 } 1664 } 1665 } 1666 1667 int 1668 rt_if_track(struct ifnet *ifp) 1669 { 1670 unsigned int rtableid; 1671 struct rtentry *rt = NULL; 1672 int i, error; 1673 1674 for (rtableid = 0; rtableid < rtmap_limit; rtableid++) { 1675 /* skip rtables that are not in the rdomain of the ifp */ 1676 if (rtable_l2(rtableid) != ifp->if_rdomain) 1677 continue; 1678 for (i = 1; i <= AF_MAX; i++) { 1679 if (!rtable_mpath_capable(rtableid, i)) 1680 continue; 1681 1682 do { 1683 error = rtable_walk(rtableid, i, &rt, 1684 rt_if_linkstate_change, ifp); 1685 if (rt != NULL && error == EEXIST) { 1686 error = rtdeletemsg(rt, ifp, rtableid); 1687 if (error == 0) 1688 error = EAGAIN; 1689 } 1690 rtfree(rt); 1691 rt = NULL; 1692 } while (error == EAGAIN); 1693 1694 if (error == EAFNOSUPPORT) 1695 error = 0; 1696 1697 if (error) 1698 break; 1699 } 1700 } 1701 1702 return (error); 1703 } 1704 1705 int 1706 rt_if_linkstate_change(struct rtentry *rt, void *arg, u_int id) 1707 { 1708 struct ifnet *ifp = arg; 1709 struct sockaddr_in6 sa_mask; 1710 int error; 1711 1712 if (rt->rt_ifidx != ifp->if_index) 1713 return (0); 1714 1715 /* Local routes are always usable. */ 1716 if (rt->rt_flags & RTF_LOCAL) { 1717 rt->rt_flags |= RTF_UP; 1718 return (0); 1719 } 1720 1721 if (LINK_STATE_IS_UP(ifp->if_link_state) && ifp->if_flags & IFF_UP) { 1722 if (ISSET(rt->rt_flags, RTF_UP)) 1723 return (0); 1724 1725 /* bring route up */ 1726 rt->rt_flags |= RTF_UP; 1727 error = rtable_mpath_reprio(id, rt_key(rt), rt_plen(rt), 1728 rt->rt_priority & RTP_MASK, rt); 1729 } else { 1730 /* 1731 * Remove redirected and cloned routes (mainly ARP) 1732 * from down interfaces so we have a chance to get 1733 * new routes from a better source. 1734 */ 1735 if (ISSET(rt->rt_flags, RTF_CLONED|RTF_DYNAMIC) && 1736 !ISSET(rt->rt_flags, RTF_CACHED|RTF_BFD)) { 1737 return (EEXIST); 1738 } 1739 1740 if (!ISSET(rt->rt_flags, RTF_UP)) 1741 return (0); 1742 1743 /* take route down */ 1744 rt->rt_flags &= ~RTF_UP; 1745 error = rtable_mpath_reprio(id, rt_key(rt), rt_plen(rt), 1746 rt->rt_priority | RTP_DOWN, rt); 1747 } 1748 if_group_routechange(rt_key(rt), rt_plen2mask(rt, &sa_mask)); 1749 1750 return (error); 1751 } 1752 1753 struct sockaddr * 1754 rt_plentosa(sa_family_t af, int plen, struct sockaddr_in6 *sa_mask) 1755 { 1756 struct sockaddr_in *sin = (struct sockaddr_in *)sa_mask; 1757 #ifdef INET6 1758 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sa_mask; 1759 #endif 1760 1761 KASSERT(plen >= 0 || plen == -1); 1762 1763 if (plen == -1) 1764 return (NULL); 1765 1766 memset(sa_mask, 0, sizeof(*sa_mask)); 1767 1768 switch (af) { 1769 case AF_INET: 1770 sin->sin_family = AF_INET; 1771 sin->sin_len = sizeof(struct sockaddr_in); 1772 in_prefixlen2mask(&sin->sin_addr, plen); 1773 break; 1774 #ifdef INET6 1775 case AF_INET6: 1776 sin6->sin6_family = AF_INET6; 1777 sin6->sin6_len = sizeof(struct sockaddr_in6); 1778 in6_prefixlen2mask(&sin6->sin6_addr, plen); 1779 break; 1780 #endif /* INET6 */ 1781 default: 1782 return (NULL); 1783 } 1784 1785 return ((struct sockaddr *)sa_mask); 1786 } 1787 1788 struct sockaddr * 1789 rt_plen2mask(struct rtentry *rt, struct sockaddr_in6 *sa_mask) 1790 { 1791 return (rt_plentosa(rt_key(rt)->sa_family, rt_plen(rt), sa_mask)); 1792 } 1793 1794 #ifdef DDB 1795 #include <machine/db_machdep.h> 1796 #include <ddb/db_output.h> 1797 1798 void 1799 db_print_sa(struct sockaddr *sa) 1800 { 1801 int len; 1802 u_char *p; 1803 1804 if (sa == NULL) { 1805 db_printf("[NULL]"); 1806 return; 1807 } 1808 1809 p = (u_char *)sa; 1810 len = sa->sa_len; 1811 db_printf("["); 1812 while (len > 0) { 1813 db_printf("%d", *p); 1814 p++; 1815 len--; 1816 if (len) 1817 db_printf(","); 1818 } 1819 db_printf("]\n"); 1820 } 1821 1822 void 1823 db_print_ifa(struct ifaddr *ifa) 1824 { 1825 if (ifa == NULL) 1826 return; 1827 db_printf(" ifa_addr="); 1828 db_print_sa(ifa->ifa_addr); 1829 db_printf(" ifa_dsta="); 1830 db_print_sa(ifa->ifa_dstaddr); 1831 db_printf(" ifa_mask="); 1832 db_print_sa(ifa->ifa_netmask); 1833 db_printf(" flags=0x%x, refcnt=%d, metric=%d\n", 1834 ifa->ifa_flags, ifa->ifa_refcnt, ifa->ifa_metric); 1835 } 1836 1837 /* 1838 * Function to pass to rtalble_walk(). 1839 * Return non-zero error to abort walk. 1840 */ 1841 int 1842 db_show_rtentry(struct rtentry *rt, void *w, unsigned int id) 1843 { 1844 db_printf("rtentry=%p", rt); 1845 1846 db_printf(" flags=0x%x refcnt=%d use=%llu expire=%lld rtableid=%u\n", 1847 rt->rt_flags, rt->rt_refcnt, rt->rt_use, rt->rt_expire, id); 1848 1849 db_printf(" key="); db_print_sa(rt_key(rt)); 1850 db_printf(" plen=%d", rt_plen(rt)); 1851 db_printf(" gw="); db_print_sa(rt->rt_gateway); 1852 db_printf(" ifidx=%u ", rt->rt_ifidx); 1853 db_printf(" ifa=%p\n", rt->rt_ifa); 1854 db_print_ifa(rt->rt_ifa); 1855 1856 db_printf(" gwroute=%p llinfo=%p\n", rt->rt_gwroute, rt->rt_llinfo); 1857 return (0); 1858 } 1859 1860 /* 1861 * Function to print all the route trees. 1862 * Use this from ddb: "call db_show_arptab" 1863 */ 1864 int 1865 db_show_arptab(void) 1866 { 1867 db_printf("Route tree for AF_INET\n"); 1868 rtable_walk(0, AF_INET, NULL, db_show_rtentry, NULL); 1869 return (0); 1870 } 1871 #endif /* DDB */ 1872