1 /* $OpenBSD: route.c,v 1.415 2023/01/21 17:35:01 mvs 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/ioctl.h> 112 #include <sys/kernel.h> 113 #include <sys/queue.h> 114 #include <sys/pool.h> 115 #include <sys/atomic.h> 116 #include <sys/rwlock.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 struct rwlock rt_lock = RWLOCK_INITIALIZER("rtlck"); 144 145 /* Give some jitter to hash, to avoid synchronization between routers. */ 146 static uint32_t rt_hashjitter; 147 148 extern unsigned int rtmap_limit; 149 150 struct cpumem * rtcounters; 151 int rttrash; /* routes not in table but not freed */ 152 153 struct pool rtentry_pool; /* pool for rtentry structures */ 154 struct pool rttimer_pool; /* pool for rttimer structures */ 155 156 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 #define LABELID_MAX 50000 167 168 struct rt_label { 169 TAILQ_ENTRY(rt_label) rtl_entry; 170 char rtl_name[RTLABEL_LEN]; 171 u_int16_t rtl_id; 172 int rtl_ref; 173 }; 174 175 TAILQ_HEAD(rt_labels, rt_label) rt_labels = TAILQ_HEAD_INITIALIZER(rt_labels); 176 177 void 178 route_init(void) 179 { 180 rtcounters = counters_alloc(rts_ncounters); 181 182 pool_init(&rtentry_pool, sizeof(struct rtentry), 0, IPL_MPFLOOR, 0, 183 "rtentry", NULL); 184 185 while (rt_hashjitter == 0) 186 rt_hashjitter = arc4random(); 187 188 #ifdef BFD 189 bfdinit(); 190 #endif 191 } 192 193 /* 194 * Returns 1 if the (cached) ``rt'' entry is still valid, 0 otherwise. 195 */ 196 int 197 rtisvalid(struct rtentry *rt) 198 { 199 if (rt == NULL) 200 return (0); 201 202 if (!ISSET(rt->rt_flags, RTF_UP)) 203 return (0); 204 205 if (ISSET(rt->rt_flags, RTF_GATEWAY)) { 206 KASSERT(rt->rt_gwroute != NULL); 207 KASSERT(!ISSET(rt->rt_gwroute->rt_flags, RTF_GATEWAY)); 208 if (!ISSET(rt->rt_gwroute->rt_flags, RTF_UP)) 209 return (0); 210 } 211 212 return (1); 213 } 214 215 /* 216 * Do the actual lookup for rtalloc(9), do not use directly! 217 * 218 * Return the best matching entry for the destination ``dst''. 219 * 220 * "RT_RESOLVE" means that a corresponding L2 entry should 221 * be added to the routing table and resolved (via ARP or 222 * NDP), if it does not exist. 223 */ 224 struct rtentry * 225 rt_match(struct sockaddr *dst, uint32_t *src, int flags, unsigned int tableid) 226 { 227 struct rtentry *rt = NULL; 228 229 rt = rtable_match(tableid, dst, src); 230 if (rt == NULL) { 231 rtstat_inc(rts_unreach); 232 return (NULL); 233 } 234 235 if (ISSET(rt->rt_flags, RTF_CLONING) && ISSET(flags, RT_RESOLVE)) 236 rt_clone(&rt, dst, tableid); 237 238 rt->rt_use++; 239 return (rt); 240 } 241 242 int 243 rt_clone(struct rtentry **rtp, struct sockaddr *dst, unsigned int rtableid) 244 { 245 struct rt_addrinfo info; 246 struct rtentry *rt = *rtp; 247 int error = 0; 248 249 memset(&info, 0, sizeof(info)); 250 info.rti_info[RTAX_DST] = dst; 251 252 /* 253 * The priority of cloned route should be different 254 * to avoid conflict with /32 cloning routes. 255 * 256 * It should also be higher to let the ARP layer find 257 * cloned routes instead of the cloning one. 258 */ 259 RT_LOCK(); 260 error = rtrequest(RTM_RESOLVE, &info, rt->rt_priority - 1, &rt, 261 rtableid); 262 RT_UNLOCK(); 263 if (error) { 264 rtm_miss(RTM_MISS, &info, 0, RTP_NONE, 0, error, rtableid); 265 } else { 266 /* Inform listeners of the new route */ 267 rtm_send(rt, RTM_ADD, 0, rtableid); 268 rtfree(*rtp); 269 *rtp = rt; 270 } 271 return (error); 272 } 273 274 /* 275 * Originated from bridge_hash() in if_bridge.c 276 */ 277 #define mix(a, b, c) do { \ 278 a -= b; a -= c; a ^= (c >> 13); \ 279 b -= c; b -= a; b ^= (a << 8); \ 280 c -= a; c -= b; c ^= (b >> 13); \ 281 a -= b; a -= c; a ^= (c >> 12); \ 282 b -= c; b -= a; b ^= (a << 16); \ 283 c -= a; c -= b; c ^= (b >> 5); \ 284 a -= b; a -= c; a ^= (c >> 3); \ 285 b -= c; b -= a; b ^= (a << 10); \ 286 c -= a; c -= b; c ^= (b >> 15); \ 287 } while (0) 288 289 int 290 rt_hash(struct rtentry *rt, struct sockaddr *dst, uint32_t *src) 291 { 292 uint32_t a, b, c; 293 294 if (src == NULL || !rtisvalid(rt) || !ISSET(rt->rt_flags, RTF_MPATH)) 295 return (-1); 296 297 a = b = 0x9e3779b9; 298 c = rt_hashjitter; 299 300 switch (dst->sa_family) { 301 case AF_INET: 302 { 303 struct sockaddr_in *sin; 304 305 if (!ipmultipath) 306 return (-1); 307 308 sin = satosin(dst); 309 a += sin->sin_addr.s_addr; 310 b += src[0]; 311 mix(a, b, c); 312 break; 313 } 314 #ifdef INET6 315 case AF_INET6: 316 { 317 struct sockaddr_in6 *sin6; 318 319 if (!ip6_multipath) 320 return (-1); 321 322 sin6 = satosin6(dst); 323 a += sin6->sin6_addr.s6_addr32[0]; 324 b += sin6->sin6_addr.s6_addr32[2]; 325 c += src[0]; 326 mix(a, b, c); 327 a += sin6->sin6_addr.s6_addr32[1]; 328 b += sin6->sin6_addr.s6_addr32[3]; 329 c += src[1]; 330 mix(a, b, c); 331 a += sin6->sin6_addr.s6_addr32[2]; 332 b += sin6->sin6_addr.s6_addr32[1]; 333 c += src[2]; 334 mix(a, b, c); 335 a += sin6->sin6_addr.s6_addr32[3]; 336 b += sin6->sin6_addr.s6_addr32[0]; 337 c += src[3]; 338 mix(a, b, c); 339 break; 340 } 341 #endif /* INET6 */ 342 } 343 344 return (c & 0xffff); 345 } 346 347 /* 348 * Allocate a route, potentially using multipath to select the peer. 349 */ 350 struct rtentry * 351 rtalloc_mpath(struct sockaddr *dst, uint32_t *src, unsigned int rtableid) 352 { 353 return (rt_match(dst, src, RT_RESOLVE, rtableid)); 354 } 355 356 /* 357 * Look in the routing table for the best matching entry for 358 * ``dst''. 359 * 360 * If a route with a gateway is found and its next hop is no 361 * longer valid, try to cache it. 362 */ 363 struct rtentry * 364 rtalloc(struct sockaddr *dst, int flags, unsigned int rtableid) 365 { 366 return (rt_match(dst, NULL, flags, rtableid)); 367 } 368 369 /* 370 * Cache the route entry corresponding to a reachable next hop in 371 * the gateway entry ``rt''. 372 */ 373 int 374 rt_setgwroute(struct rtentry *rt, u_int rtableid) 375 { 376 struct rtentry *prt, *nhrt; 377 unsigned int rdomain = rtable_l2(rtableid); 378 int error; 379 380 NET_ASSERT_LOCKED(); 381 382 KASSERT(ISSET(rt->rt_flags, RTF_GATEWAY)); 383 384 /* If we cannot find a valid next hop bail. */ 385 nhrt = rt_match(rt->rt_gateway, NULL, RT_RESOLVE, rdomain); 386 if (nhrt == NULL) 387 return (ENOENT); 388 389 /* Next hop entry must be on the same interface. */ 390 if (nhrt->rt_ifidx != rt->rt_ifidx) { 391 struct sockaddr_in6 sa_mask; 392 393 if (!ISSET(nhrt->rt_flags, RTF_LLINFO) || 394 !ISSET(nhrt->rt_flags, RTF_CLONED)) { 395 rtfree(nhrt); 396 return (EHOSTUNREACH); 397 } 398 399 /* 400 * We found a L2 entry, so we might have multiple 401 * RTF_CLONING routes for the same subnet. Query 402 * the first route of the multipath chain and iterate 403 * until we find the correct one. 404 */ 405 prt = rtable_lookup(rdomain, rt_key(nhrt->rt_parent), 406 rt_plen2mask(nhrt->rt_parent, &sa_mask), NULL, RTP_ANY); 407 rtfree(nhrt); 408 409 while (prt != NULL && prt->rt_ifidx != rt->rt_ifidx) 410 prt = rtable_iterate(prt); 411 412 /* We found nothing or a non-cloning MPATH route. */ 413 if (prt == NULL || !ISSET(prt->rt_flags, RTF_CLONING)) { 414 rtfree(prt); 415 return (EHOSTUNREACH); 416 } 417 418 error = rt_clone(&prt, rt->rt_gateway, rdomain); 419 if (error) { 420 rtfree(prt); 421 return (error); 422 } 423 nhrt = prt; 424 } 425 426 /* 427 * Next hop must be reachable, this also prevents rtentry 428 * loops for example when rt->rt_gwroute points to rt. 429 */ 430 if (ISSET(nhrt->rt_flags, RTF_CLONING|RTF_GATEWAY)) { 431 rtfree(nhrt); 432 return (ENETUNREACH); 433 } 434 435 /* Next hop is valid so remove possible old cache. */ 436 rt_putgwroute(rt); 437 KASSERT(rt->rt_gwroute == NULL); 438 439 /* 440 * If the MTU of next hop is 0, this will reset the MTU of the 441 * route to run PMTUD again from scratch. 442 */ 443 if (!ISSET(rt->rt_locks, RTV_MTU) && (rt->rt_mtu > nhrt->rt_mtu)) 444 rt->rt_mtu = nhrt->rt_mtu; 445 446 /* 447 * To avoid reference counting problems when writing link-layer 448 * addresses in an outgoing packet, we ensure that the lifetime 449 * of a cached entry is greater than the bigger lifetime of the 450 * gateway entries it is pointed by. 451 */ 452 nhrt->rt_flags |= RTF_CACHED; 453 nhrt->rt_cachecnt++; 454 455 rt->rt_gwroute = nhrt; 456 457 return (0); 458 } 459 460 /* 461 * Invalidate the cached route entry of the gateway entry ``rt''. 462 */ 463 void 464 rt_putgwroute(struct rtentry *rt) 465 { 466 struct rtentry *nhrt = rt->rt_gwroute; 467 468 NET_ASSERT_LOCKED(); 469 470 if (!ISSET(rt->rt_flags, RTF_GATEWAY) || nhrt == NULL) 471 return; 472 473 KASSERT(ISSET(nhrt->rt_flags, RTF_CACHED)); 474 KASSERT(nhrt->rt_cachecnt > 0); 475 476 --nhrt->rt_cachecnt; 477 if (nhrt->rt_cachecnt == 0) 478 nhrt->rt_flags &= ~RTF_CACHED; 479 480 rtfree(rt->rt_gwroute); 481 rt->rt_gwroute = NULL; 482 } 483 484 void 485 rtref(struct rtentry *rt) 486 { 487 refcnt_take(&rt->rt_refcnt); 488 } 489 490 void 491 rtfree(struct rtentry *rt) 492 { 493 if (rt == NULL) 494 return; 495 496 if (refcnt_rele(&rt->rt_refcnt) == 0) 497 return; 498 499 KASSERT(!ISSET(rt->rt_flags, RTF_UP)); 500 KASSERT(!RT_ROOT(rt)); 501 atomic_dec_int(&rttrash); 502 503 KERNEL_LOCK(); 504 rt_timer_remove_all(rt); 505 ifafree(rt->rt_ifa); 506 rtlabel_unref(rt->rt_labelid); 507 #ifdef MPLS 508 rt_mpls_clear(rt); 509 #endif 510 free(rt->rt_gateway, M_RTABLE, ROUNDUP(rt->rt_gateway->sa_len)); 511 free(rt_key(rt), M_RTABLE, rt_key(rt)->sa_len); 512 KERNEL_UNLOCK(); 513 514 pool_put(&rtentry_pool, rt); 515 } 516 517 struct ifaddr * 518 ifaref(struct ifaddr *ifa) 519 { 520 refcnt_take(&ifa->ifa_refcnt); 521 return ifa; 522 } 523 524 void 525 ifafree(struct ifaddr *ifa) 526 { 527 if (refcnt_rele(&ifa->ifa_refcnt) == 0) 528 return; 529 free(ifa, M_IFADDR, 0); 530 } 531 532 /* 533 * Force a routing table entry to the specified 534 * destination to go through the given gateway. 535 * Normally called as a result of a routing redirect 536 * message from the network layer. 537 */ 538 void 539 rtredirect(struct sockaddr *dst, struct sockaddr *gateway, 540 struct sockaddr *src, struct rtentry **rtp, unsigned int rdomain) 541 { 542 struct rtentry *rt; 543 int error = 0; 544 enum rtstat_counters stat = rts_ncounters; 545 struct rt_addrinfo info; 546 struct ifaddr *ifa; 547 unsigned int ifidx = 0; 548 int flags = RTF_GATEWAY|RTF_HOST; 549 uint8_t prio = RTP_NONE; 550 551 NET_ASSERT_LOCKED(); 552 553 /* verify the gateway is directly reachable */ 554 rt = rtalloc(gateway, 0, rdomain); 555 if (!rtisvalid(rt) || ISSET(rt->rt_flags, RTF_GATEWAY)) { 556 rtfree(rt); 557 error = ENETUNREACH; 558 goto out; 559 } 560 ifidx = rt->rt_ifidx; 561 ifa = rt->rt_ifa; 562 rtfree(rt); 563 rt = NULL; 564 565 rt = rtable_lookup(rdomain, dst, NULL, NULL, RTP_ANY); 566 /* 567 * If the redirect isn't from our current router for this dst, 568 * it's either old or wrong. If it redirects us to ourselves, 569 * we have a routing loop, perhaps as a result of an interface 570 * going down recently. 571 */ 572 #define equal(a1, a2) \ 573 ((a1)->sa_len == (a2)->sa_len && \ 574 bcmp((caddr_t)(a1), (caddr_t)(a2), (a1)->sa_len) == 0) 575 if (rt != NULL && (!equal(src, rt->rt_gateway) || rt->rt_ifa != ifa)) 576 error = EINVAL; 577 else if (ifa_ifwithaddr(gateway, rdomain) != NULL || 578 (gateway->sa_family == AF_INET && 579 in_broadcast(satosin(gateway)->sin_addr, rdomain))) 580 error = EHOSTUNREACH; 581 if (error) 582 goto done; 583 /* 584 * Create a new entry if we just got back a wildcard entry 585 * or the lookup failed. This is necessary for hosts 586 * which use routing redirects generated by smart gateways 587 * to dynamically build the routing tables. 588 */ 589 if (rt == NULL) 590 goto create; 591 /* 592 * Don't listen to the redirect if it's 593 * for a route to an interface. 594 */ 595 if (ISSET(rt->rt_flags, RTF_GATEWAY)) { 596 if (!ISSET(rt->rt_flags, RTF_HOST)) { 597 /* 598 * Changing from route to net => route to host. 599 * Create new route, rather than smashing route to net. 600 */ 601 create: 602 rtfree(rt); 603 flags |= RTF_DYNAMIC; 604 bzero(&info, sizeof(info)); 605 info.rti_info[RTAX_DST] = dst; 606 info.rti_info[RTAX_GATEWAY] = gateway; 607 info.rti_ifa = ifa; 608 info.rti_flags = flags; 609 rt = NULL; 610 error = rtrequest(RTM_ADD, &info, RTP_DEFAULT, &rt, 611 rdomain); 612 if (error == 0) { 613 flags = rt->rt_flags; 614 prio = rt->rt_priority; 615 } 616 stat = rts_dynamic; 617 } else { 618 /* 619 * Smash the current notion of the gateway to 620 * this destination. Should check about netmask!!! 621 */ 622 rt->rt_flags |= RTF_MODIFIED; 623 flags |= RTF_MODIFIED; 624 prio = rt->rt_priority; 625 stat = rts_newgateway; 626 rt_setgate(rt, gateway, rdomain); 627 } 628 } else 629 error = EHOSTUNREACH; 630 done: 631 if (rt) { 632 if (rtp && !error) 633 *rtp = rt; 634 else 635 rtfree(rt); 636 } 637 out: 638 if (error) 639 rtstat_inc(rts_badredirect); 640 else if (stat != rts_ncounters) 641 rtstat_inc(stat); 642 bzero((caddr_t)&info, sizeof(info)); 643 info.rti_info[RTAX_DST] = dst; 644 info.rti_info[RTAX_GATEWAY] = gateway; 645 info.rti_info[RTAX_AUTHOR] = src; 646 rtm_miss(RTM_REDIRECT, &info, flags, prio, ifidx, error, rdomain); 647 } 648 649 /* 650 * Delete a route and generate a message 651 */ 652 int 653 rtdeletemsg(struct rtentry *rt, struct ifnet *ifp, u_int tableid) 654 { 655 int error; 656 struct rt_addrinfo info; 657 struct sockaddr_rtlabel sa_rl; 658 struct sockaddr_in6 sa_mask; 659 660 KASSERT(rt->rt_ifidx == ifp->if_index); 661 662 /* 663 * Request the new route so that the entry is not actually 664 * deleted. That will allow the information being reported to 665 * be accurate (and consistent with route_output()). 666 */ 667 memset(&info, 0, sizeof(info)); 668 info.rti_info[RTAX_DST] = rt_key(rt); 669 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 670 if (!ISSET(rt->rt_flags, RTF_HOST)) 671 info.rti_info[RTAX_NETMASK] = rt_plen2mask(rt, &sa_mask); 672 info.rti_info[RTAX_LABEL] = rtlabel_id2sa(rt->rt_labelid, &sa_rl); 673 info.rti_flags = rt->rt_flags; 674 info.rti_info[RTAX_IFP] = sdltosa(ifp->if_sadl); 675 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr; 676 error = rtrequest_delete(&info, rt->rt_priority, ifp, &rt, tableid); 677 rtm_miss(RTM_DELETE, &info, info.rti_flags, rt->rt_priority, 678 rt->rt_ifidx, error, tableid); 679 if (error == 0) 680 rtfree(rt); 681 return (error); 682 } 683 684 static inline int 685 rtequal(struct rtentry *a, struct rtentry *b) 686 { 687 if (a == b) 688 return 1; 689 690 if (memcmp(rt_key(a), rt_key(b), rt_key(a)->sa_len) == 0 && 691 rt_plen(a) == rt_plen(b)) 692 return 1; 693 else 694 return 0; 695 } 696 697 int 698 rtflushclone1(struct rtentry *rt, void *arg, u_int id) 699 { 700 struct rtentry *cloningrt = arg; 701 struct ifnet *ifp; 702 703 if (!ISSET(rt->rt_flags, RTF_CLONED)) 704 return 0; 705 706 /* Cached route must stay alive as long as their parent are alive. */ 707 if (ISSET(rt->rt_flags, RTF_CACHED) && (rt->rt_parent != cloningrt)) 708 return 0; 709 710 if (!rtequal(rt->rt_parent, cloningrt)) 711 return 0; 712 /* 713 * This happens when an interface with a RTF_CLONING route is 714 * being detached. In this case it's safe to bail because all 715 * the routes are being purged by rt_ifa_purge(). 716 */ 717 ifp = if_get(rt->rt_ifidx); 718 if (ifp == NULL) 719 return 0; 720 721 if_put(ifp); 722 return EEXIST; 723 } 724 725 int 726 rtflushclone(struct rtentry *parent, unsigned int rtableid) 727 { 728 struct rtentry *rt = NULL; 729 struct ifnet *ifp; 730 int error; 731 732 #ifdef DIAGNOSTIC 733 if (!parent || (parent->rt_flags & RTF_CLONING) == 0) 734 panic("rtflushclone: called with a non-cloning route"); 735 #endif 736 737 do { 738 error = rtable_walk(rtableid, rt_key(parent)->sa_family, &rt, 739 rtflushclone1, parent); 740 if (rt != NULL && error == EEXIST) { 741 ifp = if_get(rt->rt_ifidx); 742 if (ifp == NULL) { 743 error = EAGAIN; 744 } else { 745 error = rtdeletemsg(rt, ifp, rtableid); 746 if (error == 0) 747 error = EAGAIN; 748 if_put(ifp); 749 } 750 } 751 rtfree(rt); 752 rt = NULL; 753 } while (error == EAGAIN); 754 755 return error; 756 757 } 758 759 int 760 rtrequest_delete(struct rt_addrinfo *info, u_int8_t prio, struct ifnet *ifp, 761 struct rtentry **ret_nrt, u_int tableid) 762 { 763 struct rtentry *rt; 764 int error; 765 766 NET_ASSERT_LOCKED(); 767 768 if (!rtable_exists(tableid)) 769 return (EAFNOSUPPORT); 770 rt = rtable_lookup(tableid, info->rti_info[RTAX_DST], 771 info->rti_info[RTAX_NETMASK], info->rti_info[RTAX_GATEWAY], prio); 772 if (rt == NULL) 773 return (ESRCH); 774 775 /* Make sure that's the route the caller want to delete. */ 776 if (ifp != NULL && ifp->if_index != rt->rt_ifidx) { 777 rtfree(rt); 778 return (ESRCH); 779 } 780 781 #ifdef BFD 782 if (ISSET(rt->rt_flags, RTF_BFD)) 783 bfdclear(rt); 784 #endif 785 786 error = rtable_delete(tableid, info->rti_info[RTAX_DST], 787 info->rti_info[RTAX_NETMASK], rt); 788 if (error != 0) { 789 rtfree(rt); 790 return (ESRCH); 791 } 792 793 /* Release next hop cache before flushing cloned entries. */ 794 rt_putgwroute(rt); 795 796 /* Clean up any cloned children. */ 797 if (ISSET(rt->rt_flags, RTF_CLONING)) 798 rtflushclone(rt, tableid); 799 800 rtfree(rt->rt_parent); 801 rt->rt_parent = NULL; 802 803 rt->rt_flags &= ~RTF_UP; 804 805 KASSERT(ifp->if_index == rt->rt_ifidx); 806 ifp->if_rtrequest(ifp, RTM_DELETE, rt); 807 808 atomic_inc_int(&rttrash); 809 810 if (ret_nrt != NULL) 811 *ret_nrt = rt; 812 else 813 rtfree(rt); 814 815 return (0); 816 } 817 818 int 819 rtrequest(int req, struct rt_addrinfo *info, u_int8_t prio, 820 struct rtentry **ret_nrt, u_int tableid) 821 { 822 struct ifnet *ifp; 823 struct rtentry *rt, *crt; 824 struct ifaddr *ifa; 825 struct sockaddr *ndst; 826 struct sockaddr_rtlabel *sa_rl, sa_rl2; 827 struct sockaddr_dl sa_dl = { sizeof(sa_dl), AF_LINK }; 828 int error; 829 830 NET_ASSERT_LOCKED(); 831 832 if (!rtable_exists(tableid)) 833 return (EAFNOSUPPORT); 834 if (info->rti_flags & RTF_HOST) 835 info->rti_info[RTAX_NETMASK] = NULL; 836 switch (req) { 837 case RTM_DELETE: 838 return (EINVAL); 839 840 case RTM_RESOLVE: 841 if (ret_nrt == NULL || (rt = *ret_nrt) == NULL) 842 return (EINVAL); 843 if ((rt->rt_flags & RTF_CLONING) == 0) 844 return (EINVAL); 845 KASSERT(rt->rt_ifa->ifa_ifp != NULL); 846 info->rti_ifa = rt->rt_ifa; 847 info->rti_flags = rt->rt_flags | (RTF_CLONED|RTF_HOST); 848 info->rti_flags &= ~(RTF_CLONING|RTF_CONNECTED|RTF_STATIC); 849 info->rti_info[RTAX_GATEWAY] = sdltosa(&sa_dl); 850 info->rti_info[RTAX_LABEL] = 851 rtlabel_id2sa(rt->rt_labelid, &sa_rl2); 852 /* FALLTHROUGH */ 853 854 case RTM_ADD: 855 if (info->rti_ifa == NULL) 856 return (EINVAL); 857 ifa = info->rti_ifa; 858 ifp = ifa->ifa_ifp; 859 if (prio == 0) 860 prio = ifp->if_priority + RTP_STATIC; 861 862 error = rt_copysa(info->rti_info[RTAX_DST], 863 info->rti_info[RTAX_NETMASK], &ndst); 864 if (error) 865 return (error); 866 867 rt = pool_get(&rtentry_pool, PR_NOWAIT | PR_ZERO); 868 if (rt == NULL) { 869 free(ndst, M_RTABLE, ndst->sa_len); 870 return (ENOBUFS); 871 } 872 873 refcnt_init(&rt->rt_refcnt); 874 rt->rt_flags = info->rti_flags | RTF_UP; 875 rt->rt_priority = prio; /* init routing priority */ 876 LIST_INIT(&rt->rt_timer); 877 878 /* Check the link state if the table supports it. */ 879 if (rtable_mpath_capable(tableid, ndst->sa_family) && 880 !ISSET(rt->rt_flags, RTF_LOCAL) && 881 (!LINK_STATE_IS_UP(ifp->if_link_state) || 882 !ISSET(ifp->if_flags, IFF_UP))) { 883 rt->rt_flags &= ~RTF_UP; 884 rt->rt_priority |= RTP_DOWN; 885 } 886 887 if (info->rti_info[RTAX_LABEL] != NULL) { 888 sa_rl = (struct sockaddr_rtlabel *) 889 info->rti_info[RTAX_LABEL]; 890 rt->rt_labelid = rtlabel_name2id(sa_rl->sr_label); 891 } 892 893 #ifdef MPLS 894 /* We have to allocate additional space for MPLS infos */ 895 if (info->rti_flags & RTF_MPLS && 896 (info->rti_info[RTAX_SRC] != NULL || 897 info->rti_info[RTAX_DST]->sa_family == AF_MPLS)) { 898 error = rt_mpls_set(rt, info->rti_info[RTAX_SRC], 899 info->rti_mpls); 900 if (error) { 901 free(ndst, M_RTABLE, ndst->sa_len); 902 pool_put(&rtentry_pool, rt); 903 return (error); 904 } 905 } else 906 rt_mpls_clear(rt); 907 #endif 908 909 rt->rt_ifa = ifaref(ifa); 910 rt->rt_ifidx = ifp->if_index; 911 /* 912 * Copy metrics and a back pointer from the cloned 913 * route's parent. 914 */ 915 if (ISSET(rt->rt_flags, RTF_CLONED)) { 916 rtref(*ret_nrt); 917 rt->rt_parent = *ret_nrt; 918 rt->rt_rmx = (*ret_nrt)->rt_rmx; 919 } 920 921 /* 922 * We must set rt->rt_gateway before adding ``rt'' to 923 * the routing table because the radix MPATH code use 924 * it to (re)order routes. 925 */ 926 if ((error = rt_setgate(rt, info->rti_info[RTAX_GATEWAY], 927 tableid))) { 928 ifafree(ifa); 929 rtfree(rt->rt_parent); 930 rt_putgwroute(rt); 931 free(rt->rt_gateway, M_RTABLE, 932 ROUNDUP(rt->rt_gateway->sa_len)); 933 free(ndst, M_RTABLE, ndst->sa_len); 934 pool_put(&rtentry_pool, rt); 935 return (error); 936 } 937 938 error = rtable_insert(tableid, ndst, 939 info->rti_info[RTAX_NETMASK], info->rti_info[RTAX_GATEWAY], 940 rt->rt_priority, rt); 941 if (error != 0 && 942 (crt = rtable_match(tableid, ndst, NULL)) != NULL) { 943 /* overwrite cloned route */ 944 if (ISSET(crt->rt_flags, RTF_CLONED) && 945 !ISSET(crt->rt_flags, RTF_CACHED)) { 946 struct ifnet *cifp; 947 948 cifp = if_get(crt->rt_ifidx); 949 KASSERT(cifp != NULL); 950 rtdeletemsg(crt, cifp, tableid); 951 if_put(cifp); 952 953 error = rtable_insert(tableid, ndst, 954 info->rti_info[RTAX_NETMASK], 955 info->rti_info[RTAX_GATEWAY], 956 rt->rt_priority, rt); 957 } 958 rtfree(crt); 959 } 960 if (error != 0) { 961 ifafree(ifa); 962 rtfree(rt->rt_parent); 963 rt_putgwroute(rt); 964 free(rt->rt_gateway, M_RTABLE, 965 ROUNDUP(rt->rt_gateway->sa_len)); 966 free(ndst, M_RTABLE, ndst->sa_len); 967 pool_put(&rtentry_pool, rt); 968 return (EEXIST); 969 } 970 ifp->if_rtrequest(ifp, req, rt); 971 972 if_group_routechange(info->rti_info[RTAX_DST], 973 info->rti_info[RTAX_NETMASK]); 974 975 if (ret_nrt != NULL) 976 *ret_nrt = rt; 977 else 978 rtfree(rt); 979 break; 980 } 981 982 return (0); 983 } 984 985 int 986 rt_setgate(struct rtentry *rt, struct sockaddr *gate, u_int rtableid) 987 { 988 int glen = ROUNDUP(gate->sa_len); 989 struct sockaddr *sa; 990 991 if (rt->rt_gateway == NULL || glen != ROUNDUP(rt->rt_gateway->sa_len)) { 992 sa = malloc(glen, M_RTABLE, M_NOWAIT); 993 if (sa == NULL) 994 return (ENOBUFS); 995 if (rt->rt_gateway != NULL) { 996 free(rt->rt_gateway, M_RTABLE, 997 ROUNDUP(rt->rt_gateway->sa_len)); 998 } 999 rt->rt_gateway = sa; 1000 } 1001 memmove(rt->rt_gateway, gate, glen); 1002 1003 if (ISSET(rt->rt_flags, RTF_GATEWAY)) 1004 return (rt_setgwroute(rt, rtableid)); 1005 1006 return (0); 1007 } 1008 1009 /* 1010 * Return the route entry containing the next hop link-layer 1011 * address corresponding to ``rt''. 1012 */ 1013 struct rtentry * 1014 rt_getll(struct rtentry *rt) 1015 { 1016 if (ISSET(rt->rt_flags, RTF_GATEWAY)) { 1017 KASSERT(rt->rt_gwroute != NULL); 1018 return (rt->rt_gwroute); 1019 } 1020 1021 return (rt); 1022 } 1023 1024 void 1025 rt_maskedcopy(struct sockaddr *src, struct sockaddr *dst, 1026 struct sockaddr *netmask) 1027 { 1028 u_char *cp1 = (u_char *)src; 1029 u_char *cp2 = (u_char *)dst; 1030 u_char *cp3 = (u_char *)netmask; 1031 u_char *cplim = cp2 + *cp3; 1032 u_char *cplim2 = cp2 + *cp1; 1033 1034 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */ 1035 cp3 += 2; 1036 if (cplim > cplim2) 1037 cplim = cplim2; 1038 while (cp2 < cplim) 1039 *cp2++ = *cp1++ & *cp3++; 1040 if (cp2 < cplim2) 1041 bzero(cp2, cplim2 - cp2); 1042 } 1043 1044 /* 1045 * allocate new sockaddr structure based on the user supplied src and mask 1046 * that is useable for the routing table. 1047 */ 1048 static int 1049 rt_copysa(struct sockaddr *src, struct sockaddr *mask, struct sockaddr **dst) 1050 { 1051 static const u_char maskarray[] = { 1052 0x0, 0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe }; 1053 struct sockaddr *ndst; 1054 const struct domain *dp; 1055 u_char *csrc, *cdst; 1056 int i, plen; 1057 1058 for (i = 0; (dp = domains[i]) != NULL; i++) { 1059 if (dp->dom_rtoffset == 0) 1060 continue; 1061 if (src->sa_family == dp->dom_family) 1062 break; 1063 } 1064 if (dp == NULL) 1065 return (EAFNOSUPPORT); 1066 1067 if (src->sa_len < dp->dom_sasize) 1068 return (EINVAL); 1069 1070 plen = rtable_satoplen(src->sa_family, mask); 1071 if (plen == -1) 1072 return (EINVAL); 1073 1074 ndst = malloc(dp->dom_sasize, M_RTABLE, M_NOWAIT|M_ZERO); 1075 if (ndst == NULL) 1076 return (ENOBUFS); 1077 1078 ndst->sa_family = src->sa_family; 1079 ndst->sa_len = dp->dom_sasize; 1080 1081 csrc = (u_char *)src + dp->dom_rtoffset; 1082 cdst = (u_char *)ndst + dp->dom_rtoffset; 1083 1084 memcpy(cdst, csrc, plen / 8); 1085 if (plen % 8 != 0) 1086 cdst[plen / 8] = csrc[plen / 8] & maskarray[plen % 8]; 1087 1088 *dst = ndst; 1089 return (0); 1090 } 1091 1092 int 1093 rt_ifa_add(struct ifaddr *ifa, int flags, struct sockaddr *dst, 1094 unsigned int rdomain) 1095 { 1096 struct ifnet *ifp = ifa->ifa_ifp; 1097 struct rtentry *rt; 1098 struct sockaddr_rtlabel sa_rl; 1099 struct rt_addrinfo info; 1100 uint8_t prio = ifp->if_priority + RTP_STATIC; 1101 int error; 1102 1103 KASSERT(rdomain == rtable_l2(rdomain)); 1104 1105 memset(&info, 0, sizeof(info)); 1106 info.rti_ifa = ifa; 1107 info.rti_flags = flags; 1108 info.rti_info[RTAX_DST] = dst; 1109 if (flags & RTF_LLINFO) 1110 info.rti_info[RTAX_GATEWAY] = sdltosa(ifp->if_sadl); 1111 else 1112 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr; 1113 info.rti_info[RTAX_LABEL] = rtlabel_id2sa(ifp->if_rtlabelid, &sa_rl); 1114 1115 #ifdef MPLS 1116 if ((flags & RTF_MPLS) == RTF_MPLS) 1117 info.rti_mpls = MPLS_OP_POP; 1118 #endif /* MPLS */ 1119 1120 if ((flags & RTF_HOST) == 0) 1121 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask; 1122 1123 if (flags & (RTF_LOCAL|RTF_BROADCAST)) 1124 prio = RTP_LOCAL; 1125 1126 if (flags & RTF_CONNECTED) 1127 prio = ifp->if_priority + RTP_CONNECTED; 1128 1129 error = rtrequest(RTM_ADD, &info, prio, &rt, rdomain); 1130 if (error == 0) { 1131 /* 1132 * A local route is created for every address configured 1133 * on an interface, so use this information to notify 1134 * userland that a new address has been added. 1135 */ 1136 if (flags & RTF_LOCAL) 1137 rtm_addr(RTM_NEWADDR, ifa); 1138 rtm_send(rt, RTM_ADD, 0, rdomain); 1139 rtfree(rt); 1140 } 1141 return (error); 1142 } 1143 1144 int 1145 rt_ifa_del(struct ifaddr *ifa, int flags, struct sockaddr *dst, 1146 unsigned int rdomain) 1147 { 1148 struct ifnet *ifp = ifa->ifa_ifp; 1149 struct rtentry *rt; 1150 struct mbuf *m = NULL; 1151 struct sockaddr *deldst; 1152 struct rt_addrinfo info; 1153 struct sockaddr_rtlabel sa_rl; 1154 uint8_t prio = ifp->if_priority + RTP_STATIC; 1155 int error; 1156 1157 KASSERT(rdomain == rtable_l2(rdomain)); 1158 1159 if ((flags & RTF_HOST) == 0 && ifa->ifa_netmask) { 1160 m = m_get(M_DONTWAIT, MT_SONAME); 1161 if (m == NULL) 1162 return (ENOBUFS); 1163 deldst = mtod(m, struct sockaddr *); 1164 rt_maskedcopy(dst, deldst, ifa->ifa_netmask); 1165 dst = deldst; 1166 } 1167 1168 memset(&info, 0, sizeof(info)); 1169 info.rti_ifa = ifa; 1170 info.rti_flags = flags; 1171 info.rti_info[RTAX_DST] = dst; 1172 if ((flags & RTF_LLINFO) == 0) 1173 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr; 1174 info.rti_info[RTAX_LABEL] = rtlabel_id2sa(ifp->if_rtlabelid, &sa_rl); 1175 1176 if ((flags & RTF_HOST) == 0) 1177 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask; 1178 1179 if (flags & (RTF_LOCAL|RTF_BROADCAST)) 1180 prio = RTP_LOCAL; 1181 1182 if (flags & RTF_CONNECTED) 1183 prio = ifp->if_priority + RTP_CONNECTED; 1184 1185 rtable_clearsource(rdomain, ifa->ifa_addr); 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 address 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 void 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 1335 int 1336 rt_ifa_purge_walker(struct rtentry *rt, void *vifa, unsigned int rtableid) 1337 { 1338 struct ifaddr *ifa = vifa; 1339 1340 if (rt->rt_ifa == ifa) 1341 return EEXIST; 1342 1343 return 0; 1344 } 1345 1346 /* 1347 * Route timer routines. These routes allow functions to be called 1348 * for various routes at any time. This is useful in supporting 1349 * path MTU discovery and redirect route deletion. 1350 * 1351 * This is similar to some BSDI internal functions, but it provides 1352 * for multiple queues for efficiency's sake... 1353 */ 1354 1355 struct mutex rttimer_mtx; 1356 1357 struct rttimer { 1358 TAILQ_ENTRY(rttimer) rtt_next; /* [T] entry on timer queue */ 1359 LIST_ENTRY(rttimer) rtt_link; /* [T] timers per rtentry */ 1360 struct timeout rtt_timeout; /* [I] timeout for this entry */ 1361 struct rttimer_queue *rtt_queue; /* [I] back pointer to queue */ 1362 struct rtentry *rtt_rt; /* [T] back pointer to route */ 1363 time_t rtt_expire; /* [I] rt expire time */ 1364 u_int rtt_tableid; /* [I] rtable id of rtt_rt */ 1365 }; 1366 1367 #define RTTIMER_CALLOUT(r) { \ 1368 if (r->rtt_queue->rtq_func != NULL) { \ 1369 (*r->rtt_queue->rtq_func)(r->rtt_rt, r->rtt_tableid); \ 1370 } else { \ 1371 struct ifnet *ifp; \ 1372 \ 1373 ifp = if_get(r->rtt_rt->rt_ifidx); \ 1374 if (ifp != NULL && \ 1375 (r->rtt_rt->rt_flags & (RTF_DYNAMIC|RTF_HOST)) == \ 1376 (RTF_DYNAMIC|RTF_HOST)) \ 1377 rtdeletemsg(r->rtt_rt, ifp, r->rtt_tableid); \ 1378 if_put(ifp); \ 1379 } \ 1380 } 1381 1382 /* 1383 * Some subtle order problems with domain initialization mean that 1384 * we cannot count on this being run from rt_init before various 1385 * protocol initializations are done. Therefore, we make sure 1386 * that this is run when the first queue is added... 1387 */ 1388 1389 void 1390 rt_timer_init(void) 1391 { 1392 pool_init(&rttimer_pool, sizeof(struct rttimer), 0, 1393 IPL_MPFLOOR, 0, "rttmr", NULL); 1394 mtx_init(&rttimer_mtx, IPL_MPFLOOR); 1395 } 1396 1397 void 1398 rt_timer_queue_init(struct rttimer_queue *rtq, int timeout, 1399 void (*func)(struct rtentry *, u_int)) 1400 { 1401 rtq->rtq_timeout = timeout; 1402 rtq->rtq_count = 0; 1403 rtq->rtq_func = func; 1404 TAILQ_INIT(&rtq->rtq_head); 1405 } 1406 1407 void 1408 rt_timer_queue_change(struct rttimer_queue *rtq, int timeout) 1409 { 1410 mtx_enter(&rttimer_mtx); 1411 rtq->rtq_timeout = timeout; 1412 mtx_leave(&rttimer_mtx); 1413 } 1414 1415 void 1416 rt_timer_queue_flush(struct rttimer_queue *rtq) 1417 { 1418 struct rttimer *r; 1419 TAILQ_HEAD(, rttimer) rttlist; 1420 1421 NET_ASSERT_LOCKED(); 1422 1423 TAILQ_INIT(&rttlist); 1424 mtx_enter(&rttimer_mtx); 1425 while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL) { 1426 LIST_REMOVE(r, rtt_link); 1427 TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next); 1428 TAILQ_INSERT_TAIL(&rttlist, r, rtt_next); 1429 KASSERT(rtq->rtq_count > 0); 1430 rtq->rtq_count--; 1431 } 1432 mtx_leave(&rttimer_mtx); 1433 1434 while ((r = TAILQ_FIRST(&rttlist)) != NULL) { 1435 TAILQ_REMOVE(&rttlist, r, rtt_next); 1436 RTTIMER_CALLOUT(r); 1437 pool_put(&rttimer_pool, r); 1438 } 1439 } 1440 1441 unsigned long 1442 rt_timer_queue_count(struct rttimer_queue *rtq) 1443 { 1444 return (rtq->rtq_count); 1445 } 1446 1447 static inline struct rttimer * 1448 rt_timer_unlink(struct rttimer *r) 1449 { 1450 MUTEX_ASSERT_LOCKED(&rttimer_mtx); 1451 1452 LIST_REMOVE(r, rtt_link); 1453 r->rtt_rt = NULL; 1454 1455 if (timeout_del(&r->rtt_timeout) == 0) { 1456 /* timeout fired, so rt_timer_timer will do the cleanup */ 1457 return NULL; 1458 } 1459 1460 TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next); 1461 KASSERT(r->rtt_queue->rtq_count > 0); 1462 r->rtt_queue->rtq_count--; 1463 return r; 1464 } 1465 1466 void 1467 rt_timer_remove_all(struct rtentry *rt) 1468 { 1469 struct rttimer *r; 1470 TAILQ_HEAD(, rttimer) rttlist; 1471 1472 TAILQ_INIT(&rttlist); 1473 mtx_enter(&rttimer_mtx); 1474 while ((r = LIST_FIRST(&rt->rt_timer)) != NULL) { 1475 r = rt_timer_unlink(r); 1476 if (r != NULL) 1477 TAILQ_INSERT_TAIL(&rttlist, r, rtt_next); 1478 } 1479 mtx_leave(&rttimer_mtx); 1480 1481 while ((r = TAILQ_FIRST(&rttlist)) != NULL) { 1482 TAILQ_REMOVE(&rttlist, r, rtt_next); 1483 pool_put(&rttimer_pool, r); 1484 } 1485 } 1486 1487 time_t 1488 rt_timer_get_expire(const struct rtentry *rt) 1489 { 1490 const struct rttimer *r; 1491 time_t expire = 0; 1492 1493 mtx_enter(&rttimer_mtx); 1494 LIST_FOREACH(r, &rt->rt_timer, rtt_link) { 1495 if (expire == 0 || expire > r->rtt_expire) 1496 expire = r->rtt_expire; 1497 } 1498 mtx_leave(&rttimer_mtx); 1499 1500 return expire; 1501 } 1502 1503 int 1504 rt_timer_add(struct rtentry *rt, struct rttimer_queue *queue, u_int rtableid) 1505 { 1506 struct rttimer *r, *rnew; 1507 1508 rnew = pool_get(&rttimer_pool, PR_NOWAIT | PR_ZERO); 1509 if (rnew == NULL) 1510 return (ENOBUFS); 1511 1512 rnew->rtt_rt = rt; 1513 rnew->rtt_queue = queue; 1514 rnew->rtt_tableid = rtableid; 1515 rnew->rtt_expire = getuptime() + queue->rtq_timeout; 1516 timeout_set_proc(&rnew->rtt_timeout, rt_timer_timer, rnew); 1517 1518 mtx_enter(&rttimer_mtx); 1519 /* 1520 * If there's already a timer with this action, destroy it before 1521 * we add a new one. 1522 */ 1523 LIST_FOREACH(r, &rt->rt_timer, rtt_link) { 1524 if (r->rtt_queue == queue) { 1525 r = rt_timer_unlink(r); 1526 break; /* only one per list, so we can quit... */ 1527 } 1528 } 1529 1530 LIST_INSERT_HEAD(&rt->rt_timer, rnew, rtt_link); 1531 TAILQ_INSERT_TAIL(&queue->rtq_head, rnew, rtt_next); 1532 timeout_add_sec(&rnew->rtt_timeout, queue->rtq_timeout); 1533 rnew->rtt_queue->rtq_count++; 1534 mtx_leave(&rttimer_mtx); 1535 1536 if (r != NULL) 1537 pool_put(&rttimer_pool, r); 1538 1539 return (0); 1540 } 1541 1542 void 1543 rt_timer_timer(void *arg) 1544 { 1545 struct rttimer *r = arg; 1546 struct rttimer_queue *rtq = r->rtt_queue; 1547 1548 NET_LOCK(); 1549 mtx_enter(&rttimer_mtx); 1550 1551 if (r->rtt_rt != NULL) 1552 LIST_REMOVE(r, rtt_link); 1553 TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next); 1554 KASSERT(rtq->rtq_count > 0); 1555 rtq->rtq_count--; 1556 1557 mtx_leave(&rttimer_mtx); 1558 1559 if (r->rtt_rt != NULL) 1560 RTTIMER_CALLOUT(r); 1561 NET_UNLOCK(); 1562 1563 pool_put(&rttimer_pool, r); 1564 } 1565 1566 #ifdef MPLS 1567 int 1568 rt_mpls_set(struct rtentry *rt, struct sockaddr *src, uint8_t op) 1569 { 1570 struct sockaddr_mpls *psa_mpls = (struct sockaddr_mpls *)src; 1571 struct rt_mpls *rt_mpls; 1572 1573 if (psa_mpls == NULL && op != MPLS_OP_POP) 1574 return (EOPNOTSUPP); 1575 if (psa_mpls != NULL && psa_mpls->smpls_len != sizeof(*psa_mpls)) 1576 return (EINVAL); 1577 if (psa_mpls != NULL && psa_mpls->smpls_family != AF_MPLS) 1578 return (EAFNOSUPPORT); 1579 1580 rt->rt_llinfo = malloc(sizeof(struct rt_mpls), M_TEMP, M_NOWAIT|M_ZERO); 1581 if (rt->rt_llinfo == NULL) 1582 return (ENOMEM); 1583 1584 rt_mpls = (struct rt_mpls *)rt->rt_llinfo; 1585 if (psa_mpls != NULL) 1586 rt_mpls->mpls_label = psa_mpls->smpls_label; 1587 rt_mpls->mpls_operation = op; 1588 /* XXX: set experimental bits */ 1589 rt->rt_flags |= RTF_MPLS; 1590 1591 return (0); 1592 } 1593 1594 void 1595 rt_mpls_clear(struct rtentry *rt) 1596 { 1597 if (rt->rt_llinfo != NULL && rt->rt_flags & RTF_MPLS) { 1598 free(rt->rt_llinfo, M_TEMP, sizeof(struct rt_mpls)); 1599 rt->rt_llinfo = NULL; 1600 } 1601 rt->rt_flags &= ~RTF_MPLS; 1602 } 1603 #endif 1604 1605 u_int16_t 1606 rtlabel_name2id(char *name) 1607 { 1608 struct rt_label *label, *p; 1609 u_int16_t new_id = 1; 1610 1611 if (!name[0]) 1612 return (0); 1613 1614 TAILQ_FOREACH(label, &rt_labels, rtl_entry) 1615 if (strcmp(name, label->rtl_name) == 0) { 1616 label->rtl_ref++; 1617 return (label->rtl_id); 1618 } 1619 1620 /* 1621 * to avoid fragmentation, we do a linear search from the beginning 1622 * and take the first free slot we find. if there is none or the list 1623 * is empty, append a new entry at the end. 1624 */ 1625 TAILQ_FOREACH(p, &rt_labels, rtl_entry) { 1626 if (p->rtl_id != new_id) 1627 break; 1628 new_id = p->rtl_id + 1; 1629 } 1630 if (new_id > LABELID_MAX) 1631 return (0); 1632 1633 label = malloc(sizeof(*label), M_RTABLE, M_NOWAIT|M_ZERO); 1634 if (label == NULL) 1635 return (0); 1636 strlcpy(label->rtl_name, name, sizeof(label->rtl_name)); 1637 label->rtl_id = new_id; 1638 label->rtl_ref++; 1639 1640 if (p != NULL) /* insert new entry before p */ 1641 TAILQ_INSERT_BEFORE(p, label, rtl_entry); 1642 else /* either list empty or no free slot in between */ 1643 TAILQ_INSERT_TAIL(&rt_labels, label, rtl_entry); 1644 1645 return (label->rtl_id); 1646 } 1647 1648 const char * 1649 rtlabel_id2name(u_int16_t id) 1650 { 1651 struct rt_label *label; 1652 1653 TAILQ_FOREACH(label, &rt_labels, rtl_entry) 1654 if (label->rtl_id == id) 1655 return (label->rtl_name); 1656 1657 return (NULL); 1658 } 1659 1660 struct sockaddr * 1661 rtlabel_id2sa(u_int16_t labelid, struct sockaddr_rtlabel *sa_rl) 1662 { 1663 const char *label; 1664 1665 if (labelid == 0 || (label = rtlabel_id2name(labelid)) == NULL) 1666 return (NULL); 1667 1668 bzero(sa_rl, sizeof(*sa_rl)); 1669 sa_rl->sr_len = sizeof(*sa_rl); 1670 sa_rl->sr_family = AF_UNSPEC; 1671 strlcpy(sa_rl->sr_label, label, sizeof(sa_rl->sr_label)); 1672 1673 return ((struct sockaddr *)sa_rl); 1674 } 1675 1676 void 1677 rtlabel_unref(u_int16_t id) 1678 { 1679 struct rt_label *p, *next; 1680 1681 if (id == 0) 1682 return; 1683 1684 TAILQ_FOREACH_SAFE(p, &rt_labels, rtl_entry, next) { 1685 if (id == p->rtl_id) { 1686 if (--p->rtl_ref == 0) { 1687 TAILQ_REMOVE(&rt_labels, p, rtl_entry); 1688 free(p, M_RTABLE, sizeof(*p)); 1689 } 1690 break; 1691 } 1692 } 1693 } 1694 1695 int 1696 rt_if_track(struct ifnet *ifp) 1697 { 1698 unsigned int rtableid; 1699 struct rtentry *rt = NULL; 1700 int i, error = 0; 1701 1702 for (rtableid = 0; rtableid < rtmap_limit; rtableid++) { 1703 /* skip rtables that are not in the rdomain of the ifp */ 1704 if (rtable_l2(rtableid) != ifp->if_rdomain) 1705 continue; 1706 for (i = 1; i <= AF_MAX; i++) { 1707 if (!rtable_mpath_capable(rtableid, i)) 1708 continue; 1709 1710 do { 1711 error = rtable_walk(rtableid, i, &rt, 1712 rt_if_linkstate_change, ifp); 1713 if (rt != NULL && error == EEXIST) { 1714 error = rtdeletemsg(rt, ifp, rtableid); 1715 if (error == 0) 1716 error = EAGAIN; 1717 } 1718 rtfree(rt); 1719 rt = NULL; 1720 } while (error == EAGAIN); 1721 1722 if (error == EAFNOSUPPORT) 1723 error = 0; 1724 1725 if (error) 1726 break; 1727 } 1728 } 1729 1730 return (error); 1731 } 1732 1733 int 1734 rt_if_linkstate_change(struct rtentry *rt, void *arg, u_int id) 1735 { 1736 struct ifnet *ifp = arg; 1737 struct sockaddr_in6 sa_mask; 1738 int error; 1739 1740 if (rt->rt_ifidx != ifp->if_index) 1741 return (0); 1742 1743 /* Local routes are always usable. */ 1744 if (rt->rt_flags & RTF_LOCAL) { 1745 rt->rt_flags |= RTF_UP; 1746 return (0); 1747 } 1748 1749 if (LINK_STATE_IS_UP(ifp->if_link_state) && ifp->if_flags & IFF_UP) { 1750 if (ISSET(rt->rt_flags, RTF_UP)) 1751 return (0); 1752 1753 /* bring route up */ 1754 rt->rt_flags |= RTF_UP; 1755 error = rtable_mpath_reprio(id, rt_key(rt), rt_plen(rt), 1756 rt->rt_priority & RTP_MASK, rt); 1757 } else { 1758 /* 1759 * Remove redirected and cloned routes (mainly ARP) 1760 * from down interfaces so we have a chance to get 1761 * new routes from a better source. 1762 */ 1763 if (ISSET(rt->rt_flags, RTF_CLONED|RTF_DYNAMIC) && 1764 !ISSET(rt->rt_flags, RTF_CACHED|RTF_BFD)) { 1765 return (EEXIST); 1766 } 1767 1768 if (!ISSET(rt->rt_flags, RTF_UP)) 1769 return (0); 1770 1771 /* take route down */ 1772 rt->rt_flags &= ~RTF_UP; 1773 error = rtable_mpath_reprio(id, rt_key(rt), rt_plen(rt), 1774 rt->rt_priority | RTP_DOWN, rt); 1775 } 1776 if_group_routechange(rt_key(rt), rt_plen2mask(rt, &sa_mask)); 1777 1778 return (error); 1779 } 1780 1781 struct sockaddr * 1782 rt_plentosa(sa_family_t af, int plen, struct sockaddr_in6 *sa_mask) 1783 { 1784 struct sockaddr_in *sin = (struct sockaddr_in *)sa_mask; 1785 #ifdef INET6 1786 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sa_mask; 1787 #endif 1788 1789 KASSERT(plen >= 0 || plen == -1); 1790 1791 if (plen == -1) 1792 return (NULL); 1793 1794 memset(sa_mask, 0, sizeof(*sa_mask)); 1795 1796 switch (af) { 1797 case AF_INET: 1798 sin->sin_family = AF_INET; 1799 sin->sin_len = sizeof(struct sockaddr_in); 1800 in_prefixlen2mask(&sin->sin_addr, plen); 1801 break; 1802 #ifdef INET6 1803 case AF_INET6: 1804 sin6->sin6_family = AF_INET6; 1805 sin6->sin6_len = sizeof(struct sockaddr_in6); 1806 in6_prefixlen2mask(&sin6->sin6_addr, plen); 1807 break; 1808 #endif /* INET6 */ 1809 default: 1810 return (NULL); 1811 } 1812 1813 return ((struct sockaddr *)sa_mask); 1814 } 1815 1816 struct sockaddr * 1817 rt_plen2mask(struct rtentry *rt, struct sockaddr_in6 *sa_mask) 1818 { 1819 return (rt_plentosa(rt_key(rt)->sa_family, rt_plen(rt), sa_mask)); 1820 } 1821 1822 #ifdef DDB 1823 #include <machine/db_machdep.h> 1824 #include <ddb/db_output.h> 1825 1826 void db_print_sa(struct sockaddr *); 1827 void db_print_ifa(struct ifaddr *); 1828 1829 void 1830 db_print_sa(struct sockaddr *sa) 1831 { 1832 int len; 1833 u_char *p; 1834 1835 if (sa == NULL) { 1836 db_printf("[NULL]"); 1837 return; 1838 } 1839 1840 p = (u_char *)sa; 1841 len = sa->sa_len; 1842 db_printf("["); 1843 while (len > 0) { 1844 db_printf("%d", *p); 1845 p++; 1846 len--; 1847 if (len) 1848 db_printf(","); 1849 } 1850 db_printf("]\n"); 1851 } 1852 1853 void 1854 db_print_ifa(struct ifaddr *ifa) 1855 { 1856 if (ifa == NULL) 1857 return; 1858 db_printf(" ifa_addr="); 1859 db_print_sa(ifa->ifa_addr); 1860 db_printf(" ifa_dsta="); 1861 db_print_sa(ifa->ifa_dstaddr); 1862 db_printf(" ifa_mask="); 1863 db_print_sa(ifa->ifa_netmask); 1864 db_printf(" flags=0x%x, refcnt=%u, metric=%d\n", 1865 ifa->ifa_flags, ifa->ifa_refcnt.r_refs, ifa->ifa_metric); 1866 } 1867 1868 /* 1869 * Function to pass to rtable_walk(). 1870 * Return non-zero error to abort walk. 1871 */ 1872 int 1873 db_show_rtentry(struct rtentry *rt, void *w, unsigned int id) 1874 { 1875 db_printf("rtentry=%p", rt); 1876 1877 db_printf(" flags=0x%x refcnt=%u use=%llu expire=%lld\n", 1878 rt->rt_flags, rt->rt_refcnt.r_refs, rt->rt_use, rt->rt_expire); 1879 1880 db_printf(" key="); db_print_sa(rt_key(rt)); 1881 db_printf(" plen=%d", rt_plen(rt)); 1882 db_printf(" gw="); db_print_sa(rt->rt_gateway); 1883 db_printf(" ifidx=%u ", rt->rt_ifidx); 1884 db_printf(" ifa=%p\n", rt->rt_ifa); 1885 db_print_ifa(rt->rt_ifa); 1886 1887 db_printf(" gwroute=%p llinfo=%p priority=%d\n", 1888 rt->rt_gwroute, rt->rt_llinfo, rt->rt_priority); 1889 return (0); 1890 } 1891 1892 /* 1893 * Function to print all the route trees. 1894 */ 1895 int 1896 db_show_rtable(int af, unsigned int rtableid) 1897 { 1898 db_printf("Route tree for af %d, rtableid %u\n", af, rtableid); 1899 rtable_walk(rtableid, af, NULL, db_show_rtentry, NULL); 1900 return (0); 1901 } 1902 #endif /* DDB */ 1903