1 /* 2 * Copyright (c) 1980, 1986, 1991, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)route.c 8.3 (Berkeley) 1/9/95 34 * $FreeBSD: src/sys/net/route.c,v 1.59.2.10 2003/01/17 08:04:00 ru Exp $ 35 * $DragonFly: src/sys/net/route.c,v 1.9 2004/12/15 00:11:04 hsu Exp $ 36 */ 37 38 #include "opt_inet.h" 39 40 #include <sys/param.h> 41 #include <sys/systm.h> 42 #include <sys/malloc.h> 43 #include <sys/mbuf.h> 44 #include <sys/socket.h> 45 #include <sys/domain.h> 46 #include <sys/kernel.h> 47 48 #include <net/if.h> 49 #include <net/route.h> 50 51 #include <netinet/in.h> 52 #include <net/ip_mroute/ip_mroute.h> 53 54 #define SA(p) ((struct sockaddr *)(p)) 55 56 struct route_cb route_cb; 57 static struct rtstat rtstat; 58 struct radix_node_head *rt_tables[AF_MAX+1]; 59 60 static int rttrash; /* routes not in table but not freed */ 61 62 static void rt_maskedcopy (struct sockaddr *, struct sockaddr *, 63 struct sockaddr *); 64 static void rtable_init (void **); 65 66 static void 67 rtable_init(void **table) 68 { 69 struct domain *dom; 70 71 for (dom = domains; dom; dom = dom->dom_next) 72 if (dom->dom_rtattach) 73 dom->dom_rtattach(&table[dom->dom_family], 74 dom->dom_rtoffset); 75 } 76 77 void 78 route_init() 79 { 80 rn_init(); /* initialize all zeroes, all ones, mask table */ 81 rtable_init((void **)rt_tables); 82 } 83 84 /* 85 * Packet routing routines. 86 */ 87 void 88 rtalloc(struct route *ro) 89 { 90 rtalloc_ign(ro, 0UL); 91 } 92 93 void 94 rtalloc_ign(struct route *ro, u_long ignore) 95 { 96 struct rtentry *rt; 97 int s; 98 99 if ((rt = ro->ro_rt) != NULL) { 100 if (rt->rt_ifp != NULL && rt->rt_flags & RTF_UP) 101 return; 102 /* XXX - We are probably always at splnet here already. */ 103 s = splnet(); 104 RTFREE(rt); 105 ro->ro_rt = NULL; 106 splx(s); 107 } 108 ro->ro_rt = rtalloc1(&ro->ro_dst, 1, ignore); 109 } 110 111 /* 112 * Look up the route that matches the address given 113 * Or, at least try.. Create a cloned route if needed. 114 */ 115 struct rtentry * 116 rtalloc1(struct sockaddr *dst, int report, u_long ignflags) 117 { 118 struct radix_node_head *rnh = rt_tables[dst->sa_family]; 119 struct rtentry *rt; 120 struct radix_node *rn; 121 struct rtentry *newrt = NULL; 122 struct rt_addrinfo info; 123 u_long nflags; 124 int s = splnet(), err = 0, msgtype = RTM_MISS; 125 126 /* 127 * Look up the address in the table for that Address Family 128 */ 129 if (rnh != NULL && (rn = rnh->rnh_matchaddr((char *)dst, rnh)) && 130 !(rn->rn_flags & RNF_ROOT)) { 131 /* 132 * If we find it and it's not the root node, then 133 * get a refernce on the rtentry associated. 134 */ 135 newrt = rt = (struct rtentry *)rn; 136 nflags = rt->rt_flags & ~ignflags; 137 if (report && (nflags & (RTF_CLONING | RTF_PRCLONING))) { 138 /* 139 * We are apparently adding (report = 0 in delete). 140 * If it requires that it be cloned, do so. 141 * (This implies it wasn't a HOST route.) 142 */ 143 err = rtrequest(RTM_RESOLVE, dst, SA(0), 144 SA(0), 0, &newrt); 145 if (err) { 146 /* 147 * If the cloning didn't succeed, maybe 148 * what we have will do. Return that. 149 */ 150 newrt = rt; 151 rt->rt_refcnt++; 152 goto miss; 153 } 154 if ((rt = newrt) && (rt->rt_flags & RTF_XRESOLVE)) { 155 /* 156 * If the new route specifies it be 157 * externally resolved, then go do that. 158 */ 159 msgtype = RTM_RESOLVE; 160 goto miss; 161 } 162 /* Inform listeners of the new route. */ 163 bzero(&info, sizeof(info)); 164 info.rti_info[RTAX_DST] = rt_key(rt); 165 info.rti_info[RTAX_NETMASK] = rt_mask(rt); 166 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 167 if (rt->rt_ifp != NULL) { 168 info.rti_info[RTAX_IFP] = 169 TAILQ_FIRST(&rt->rt_ifp->if_addrhead)-> 170 ifa_addr; 171 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr; 172 } 173 rt_missmsg(RTM_ADD, &info, rt->rt_flags, 0); 174 } else 175 rt->rt_refcnt++; 176 } else { 177 /* 178 * Either we hit the root or couldn't find any match, 179 * Which basically means 180 * "caint get there frm here" 181 */ 182 rtstat.rts_unreach++; 183 miss: 184 if (report) { 185 /* 186 * If required, report the failure to the supervising 187 * Authorities. 188 * For a delete, this is not an error. (report == 0) 189 */ 190 bzero(&info, sizeof(info)); 191 info.rti_info[RTAX_DST] = dst; 192 rt_missmsg(msgtype, &info, 0, err); 193 } 194 } 195 splx(s); 196 return (newrt); 197 } 198 199 /* 200 * Remove a reference count from an rtentry. 201 * If the count gets low enough, take it out of the routing table 202 */ 203 void 204 rtfree(struct rtentry *rt) 205 { 206 /* 207 * find the tree for that address family 208 */ 209 struct radix_node_head *rnh = rt_tables[rt_key(rt)->sa_family]; 210 struct ifaddr *ifa; 211 212 if (rt == NULL || rnh == NULL) 213 panic("rtfree"); 214 215 /* 216 * decrement the reference count by one and if it reaches 0, 217 * and there is a close function defined, call the close function 218 */ 219 rt->rt_refcnt--; 220 if(rnh->rnh_close && rt->rt_refcnt == 0) { 221 rnh->rnh_close((struct radix_node *)rt, rnh); 222 } 223 224 /* 225 * If we are no longer "up" (and ref == 0) 226 * then we can free the resources associated 227 * with the route. 228 */ 229 if (rt->rt_refcnt <= 0 && !(rt->rt_flags & RTF_UP)) { 230 if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT)) 231 panic ("rtfree 2"); 232 /* 233 * the rtentry must have been removed from the routing table 234 * so it is represented in rttrash.. remove that now. 235 */ 236 rttrash--; 237 238 #ifdef DIAGNOSTIC 239 if (rt->rt_refcnt < 0) { 240 printf("rtfree: %p not freed (neg refs)\n", rt); 241 return; 242 } 243 #endif 244 245 /* 246 * release references on items we hold them on.. 247 * e.g other routes and ifaddrs. 248 */ 249 if((ifa = rt->rt_ifa)) 250 IFAFREE(ifa); 251 if (rt->rt_parent) { 252 RTFREE(rt->rt_parent); 253 } 254 255 /* 256 * The key is separatly alloc'd so free it (see rt_setgate()). 257 * This also frees the gateway, as they are always malloc'd 258 * together. 259 */ 260 Free(rt_key(rt)); 261 262 /* 263 * and the rtentry itself of course 264 */ 265 Free(rt); 266 } 267 } 268 269 void 270 ifafree(struct ifaddr *ifa) 271 { 272 if (ifa == NULL) 273 panic("ifafree"); 274 if (ifa->ifa_refcnt == 0) 275 free(ifa, M_IFADDR); 276 else 277 ifa->ifa_refcnt--; 278 } 279 280 #define sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0) 281 282 /* 283 * Force a routing table entry to the specified 284 * destination to go through the given gateway. 285 * Normally called as a result of a routing redirect 286 * message from the network layer. 287 * 288 * N.B.: must be called at splnet 289 * 290 */ 291 void 292 rtredirect( 293 struct sockaddr *dst, 294 struct sockaddr *gateway, 295 struct sockaddr *netmask, 296 int flags, 297 struct sockaddr *src, 298 struct rtentry **rtp) 299 { 300 struct rtentry *rt; 301 int error = 0; 302 short *stat = NULL; 303 struct rt_addrinfo info; 304 struct ifaddr *ifa; 305 306 /* verify the gateway is directly reachable */ 307 if ((ifa = ifa_ifwithnet(gateway)) == NULL) { 308 error = ENETUNREACH; 309 goto out; 310 } 311 rt = rtalloc1(dst, 0, 0UL); 312 /* 313 * If the redirect isn't from our current router for this dst, 314 * it's either old or wrong. If it redirects us to ourselves, 315 * we have a routing loop, perhaps as a result of an interface 316 * going down recently. 317 */ 318 if (!(flags & RTF_DONE) && rt && 319 (!sa_equal(src, rt->rt_gateway) || rt->rt_ifa != ifa)) 320 error = EINVAL; 321 else if (ifa_ifwithaddr(gateway)) 322 error = EHOSTUNREACH; 323 if (error) 324 goto done; 325 /* 326 * Create a new entry if we just got back a wildcard entry 327 * or the the lookup failed. This is necessary for hosts 328 * which use routing redirects generated by smart gateways 329 * to dynamically build the routing tables. 330 */ 331 if ((rt == NULL) || (rt_mask(rt) != NULL && rt_mask(rt)->sa_len < 2)) 332 goto create; 333 /* 334 * Don't listen to the redirect if it's 335 * for a route to an interface. 336 */ 337 if (rt->rt_flags & RTF_GATEWAY) { 338 if ((!(rt->rt_flags & RTF_HOST)) && (flags & RTF_HOST)) { 339 /* 340 * Changing from route to net => route to host. 341 * Create new route, rather than smashing route to net. 342 */ 343 create: 344 if (rt) 345 rtfree(rt); 346 flags |= RTF_GATEWAY | RTF_DYNAMIC; 347 bzero(&info, sizeof(info)); 348 info.rti_info[RTAX_DST] = dst; 349 info.rti_info[RTAX_GATEWAY] = gateway; 350 info.rti_info[RTAX_NETMASK] = netmask; 351 info.rti_ifa = ifa; 352 info.rti_flags = flags; 353 rt = NULL; 354 error = rtrequest1(RTM_ADD, &info, &rt); 355 if (rt != NULL) 356 flags = rt->rt_flags; 357 stat = &rtstat.rts_dynamic; 358 } else { 359 /* 360 * Smash the current notion of the gateway to 361 * this destination. Should check about netmask!!! 362 */ 363 rt->rt_flags |= RTF_MODIFIED; 364 flags |= RTF_MODIFIED; 365 stat = &rtstat.rts_newgateway; 366 /* 367 * add the key and gateway (in one malloc'd chunk). 368 */ 369 rt_setgate(rt, rt_key(rt), gateway); 370 } 371 } else 372 error = EHOSTUNREACH; 373 done: 374 if (rt) { 375 if (rtp != NULL && !error) 376 *rtp = rt; 377 else 378 rtfree(rt); 379 } 380 out: 381 if (error) 382 rtstat.rts_badredirect++; 383 else if (stat != NULL) 384 (*stat)++; 385 bzero(&info, sizeof(info)); 386 info.rti_info[RTAX_DST] = dst; 387 info.rti_info[RTAX_GATEWAY] = gateway; 388 info.rti_info[RTAX_NETMASK] = netmask; 389 info.rti_info[RTAX_AUTHOR] = src; 390 rt_missmsg(RTM_REDIRECT, &info, flags, error); 391 } 392 393 /* 394 * Routing table ioctl interface. 395 */ 396 int 397 rtioctl(u_long req, caddr_t data, struct thread *td) 398 { 399 #ifdef INET 400 /* Multicast goop, grrr... */ 401 return mrt_ioctl ? mrt_ioctl(req, data) : EOPNOTSUPP; 402 #else 403 return ENXIO; 404 #endif 405 } 406 407 struct ifaddr * 408 ifa_ifwithroute(int flags, struct sockaddr *dst, struct sockaddr *gateway) 409 { 410 struct ifaddr *ifa; 411 412 if (!(flags & RTF_GATEWAY)) { 413 /* 414 * If we are adding a route to an interface, 415 * and the interface is a pt to pt link 416 * we should search for the destination 417 * as our clue to the interface. Otherwise 418 * we can use the local address. 419 */ 420 ifa = NULL; 421 if (flags & RTF_HOST) { 422 ifa = ifa_ifwithdstaddr(dst); 423 } 424 if (ifa == NULL) 425 ifa = ifa_ifwithaddr(gateway); 426 } else { 427 /* 428 * If we are adding a route to a remote net 429 * or host, the gateway may still be on the 430 * other end of a pt to pt link. 431 */ 432 ifa = ifa_ifwithdstaddr(gateway); 433 } 434 if (ifa == NULL) 435 ifa = ifa_ifwithnet(gateway); 436 if (ifa == NULL) { 437 struct rtentry *rt = rtalloc1(gateway, 0, 0UL); 438 if (rt == NULL) 439 return (NULL); 440 rt->rt_refcnt--; 441 if ((ifa = rt->rt_ifa) == NULL) 442 return (NULL); 443 } 444 if (ifa->ifa_addr->sa_family != dst->sa_family) { 445 struct ifaddr *oifa = ifa; 446 ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp); 447 if (ifa == NULL) 448 ifa = oifa; 449 } 450 return (ifa); 451 } 452 453 static int rt_fixdelete (struct radix_node *, void *); 454 static int rt_fixchange (struct radix_node *, void *); 455 456 struct rtfc_arg { 457 struct rtentry *rt0; 458 struct radix_node_head *rnh; 459 }; 460 461 /* 462 * Do appropriate manipulations of a routing tree given 463 * all the bits of info needed 464 */ 465 int 466 rtrequest( 467 int req, 468 struct sockaddr *dst, 469 struct sockaddr *gateway, 470 struct sockaddr *netmask, 471 int flags, 472 struct rtentry **ret_nrt) 473 { 474 struct rt_addrinfo info; 475 476 bzero(&info, sizeof(info)); 477 info.rti_flags = flags; 478 info.rti_info[RTAX_DST] = dst; 479 info.rti_info[RTAX_GATEWAY] = gateway; 480 info.rti_info[RTAX_NETMASK] = netmask; 481 return rtrequest1(req, &info, ret_nrt); 482 } 483 484 int 485 rt_getifa(struct rt_addrinfo *info) 486 { 487 struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY]; 488 struct sockaddr *dst = info->rti_info[RTAX_DST]; 489 struct sockaddr *ifaaddr = info->rti_info[RTAX_IFA]; 490 struct sockaddr *ifpaddr = info->rti_info[RTAX_IFP]; 491 int flags = info->rti_flags; 492 struct ifaddr *ifa; 493 int error = 0; 494 495 /* 496 * ifp may be specified by sockaddr_dl 497 * when protocol address is ambiguous. 498 */ 499 if (info->rti_ifp == NULL && ifpaddr != NULL && 500 ifpaddr->sa_family == AF_LINK && 501 (ifa = ifa_ifwithnet(ifpaddr)) != NULL) 502 info->rti_ifp = ifa->ifa_ifp; 503 if (info->rti_ifa == NULL && ifaaddr != NULL) 504 info->rti_ifa = ifa_ifwithaddr(ifaaddr); 505 if (info->rti_ifa == NULL) { 506 struct sockaddr *sa; 507 508 sa = ifaaddr != NULL ? ifaaddr : 509 (gateway != NULL ? gateway : dst); 510 if (sa != NULL && info->rti_ifp != NULL) 511 info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp); 512 else if (dst != NULL && gateway != NULL) 513 info->rti_ifa = ifa_ifwithroute(flags, dst, gateway); 514 else if (sa != NULL) 515 info->rti_ifa = ifa_ifwithroute(flags, sa, sa); 516 } 517 if ((ifa = info->rti_ifa) != NULL) { 518 if (info->rti_ifp == NULL) 519 info->rti_ifp = ifa->ifa_ifp; 520 } else 521 error = ENETUNREACH; 522 return (error); 523 } 524 525 int 526 rtrequest1(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt) 527 { 528 struct sockaddr *dst = info->rti_info[RTAX_DST]; 529 struct rtentry *rt; 530 struct radix_node *rn; 531 struct radix_node_head *rnh; 532 struct ifaddr *ifa; 533 struct sockaddr *ndst; 534 int s = splnet(); 535 int error = 0; 536 537 #define gotoerr(x) { error = x ; goto bad; } 538 539 /* 540 * Find the correct routing tree to use for this Address Family 541 */ 542 if ((rnh = rt_tables[dst->sa_family]) == NULL) 543 gotoerr(EAFNOSUPPORT); 544 /* 545 * If we are adding a host route then we don't want to put 546 * a netmask in the tree, nor do we want to clone it. 547 */ 548 if (info->rti_flags & RTF_HOST) { 549 info->rti_info[RTAX_NETMASK] = NULL; 550 info->rti_flags &= ~(RTF_CLONING | RTF_PRCLONING); 551 } 552 switch (req) { 553 case RTM_DELETE: 554 /* 555 * Remove the item from the tree and return it. 556 * Complain if it is not there and do no more processing. 557 */ 558 if ((rn = rnh->rnh_deladdr((char *)dst, 559 (char *)info->rti_info[RTAX_NETMASK], rnh)) == NULL) 560 gotoerr(ESRCH); 561 if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT)) 562 panic ("rtrequest delete"); 563 rt = (struct rtentry *)rn; 564 565 /* 566 * Now search what's left of the subtree for any cloned 567 * routes which might have been formed from this node. 568 */ 569 if ((rt->rt_flags & (RTF_CLONING | RTF_PRCLONING)) && 570 rt_mask(rt) != NULL) { 571 rnh->rnh_walktree_from(rnh, (char *)dst, 572 (char *)rt_mask(rt), 573 rt_fixdelete, rt); 574 } 575 576 /* 577 * Remove any external references we may have. 578 * This might result in another rtentry being freed if 579 * we held its last reference. 580 */ 581 if (rt->rt_gwroute) { 582 rt = rt->rt_gwroute; 583 RTFREE(rt); 584 (rt = (struct rtentry *)rn)->rt_gwroute = NULL; 585 } 586 587 /* 588 * NB: RTF_UP must be set during the search above, 589 * because we might delete the last ref, causing 590 * rt to get freed prematurely. 591 * eh? then why not just add a reference? 592 * I'm not sure how RTF_UP helps matters. (JRE) 593 */ 594 rt->rt_flags &= ~RTF_UP; 595 596 /* 597 * give the protocol a chance to keep things in sync. 598 */ 599 if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest) 600 ifa->ifa_rtrequest(RTM_DELETE, rt, info); 601 602 /* 603 * one more rtentry floating around that is not 604 * linked to the routing table. 605 */ 606 rttrash++; 607 608 /* 609 * If the caller wants it, then it can have it, 610 * but it's up to it to free the rtentry as we won't be 611 * doing it. 612 */ 613 if (ret_nrt != NULL) 614 *ret_nrt = rt; 615 else if (rt->rt_refcnt <= 0) { 616 rt->rt_refcnt++; /* make a 1->0 transition */ 617 rtfree(rt); 618 } 619 break; 620 621 case RTM_RESOLVE: 622 if (ret_nrt == NULL || (rt = *ret_nrt) == NULL) 623 gotoerr(EINVAL); 624 ifa = rt->rt_ifa; 625 info->rti_flags = rt->rt_flags & 626 ~(RTF_CLONING | RTF_PRCLONING | RTF_STATIC); 627 info->rti_flags |= RTF_WASCLONED; 628 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway; 629 if ((info->rti_info[RTAX_NETMASK] = rt->rt_genmask) == NULL) 630 info->rti_flags |= RTF_HOST; 631 goto makeroute; 632 633 case RTM_ADD: 634 if ((info->rti_flags & RTF_GATEWAY) && 635 !info->rti_info[RTAX_GATEWAY]) 636 panic("rtrequest: GATEWAY but no gateway"); 637 638 if (info->rti_ifa == NULL && (error = rt_getifa(info))) 639 gotoerr(error); 640 ifa = info->rti_ifa; 641 642 makeroute: 643 R_Malloc(rt, struct rtentry *, sizeof(*rt)); 644 if (rt == NULL) 645 gotoerr(ENOBUFS); 646 bzero(rt, sizeof(*rt)); 647 rt->rt_flags = RTF_UP | info->rti_flags; 648 /* 649 * Add the gateway. Possibly re-malloc-ing the storage for it 650 * also add the rt_gwroute if possible. 651 */ 652 if ((error = rt_setgate(rt, dst, info->rti_info[RTAX_GATEWAY])) 653 != 0) { 654 Free(rt); 655 gotoerr(error); 656 } 657 658 /* 659 * point to the (possibly newly malloc'd) dest address. 660 */ 661 ndst = rt_key(rt); 662 663 /* 664 * make sure it contains the value we want (masked if needed). 665 */ 666 if (info->rti_info[RTAX_NETMASK] != NULL) { 667 rt_maskedcopy(dst, ndst, info->rti_info[RTAX_NETMASK]); 668 } else 669 bcopy(dst, ndst, dst->sa_len); 670 671 /* 672 * Note that we now have a reference to the ifa. 673 * This moved from below so that rnh->rnh_addaddr() can 674 * examine the ifa and ifa->ifa_ifp if it so desires. 675 */ 676 ifa->ifa_refcnt++; 677 rt->rt_ifa = ifa; 678 rt->rt_ifp = ifa->ifa_ifp; 679 /* XXX mtu manipulation will be done in rnh_addaddr -- itojun */ 680 681 rn = rnh->rnh_addaddr((char *)ndst, 682 (char *)info->rti_info[RTAX_NETMASK], 683 rnh, rt->rt_nodes); 684 if (rn == NULL) { 685 struct rtentry *rt2; 686 /* 687 * Uh-oh, we already have one of these in the tree. 688 * We do a special hack: if the route that's already 689 * there was generated by the protocol-cloning 690 * mechanism, then we just blow it away and retry 691 * the insertion of the new one. 692 */ 693 rt2 = rtalloc1(dst, 0, RTF_PRCLONING); 694 if (rt2 != NULL && rt2->rt_parent) { 695 rtrequest(RTM_DELETE, 696 (struct sockaddr *)rt_key(rt2), 697 rt2->rt_gateway, 698 rt_mask(rt2), rt2->rt_flags, 0); 699 RTFREE(rt2); 700 rn = rnh->rnh_addaddr((char *)ndst, 701 (char *)info->rti_info[RTAX_NETMASK], 702 rnh, rt->rt_nodes); 703 } else if (rt2 != NULL) { 704 /* undo the extra ref we got */ 705 RTFREE(rt2); 706 } 707 } 708 709 /* 710 * If it still failed to go into the tree, 711 * then un-make it (this should be a function) 712 */ 713 if (rn == NULL) { 714 if (rt->rt_gwroute) 715 rtfree(rt->rt_gwroute); 716 if (rt->rt_ifa) { 717 IFAFREE(rt->rt_ifa); 718 } 719 Free(rt_key(rt)); 720 Free(rt); 721 gotoerr(EEXIST); 722 } 723 724 rt->rt_parent = 0; 725 726 /* 727 * If we got here from RESOLVE, then we are cloning 728 * so clone the rest, and note that we 729 * are a clone (and increment the parent's references) 730 */ 731 if (req == RTM_RESOLVE) { 732 rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */ 733 rt->rt_rmx.rmx_pksent = 0; /* reset packet counter */ 734 if ((*ret_nrt)->rt_flags & 735 (RTF_CLONING | RTF_PRCLONING)) { 736 rt->rt_parent = *ret_nrt; 737 (*ret_nrt)->rt_refcnt++; 738 } 739 } 740 741 /* 742 * if this protocol has something to add to this then 743 * allow it to do that as well. 744 */ 745 if (ifa->ifa_rtrequest) 746 ifa->ifa_rtrequest(req, rt, info); 747 748 /* 749 * We repeat the same procedure from rt_setgate() here because 750 * it doesn't fire when we call it there because the node 751 * hasn't been added to the tree yet. 752 */ 753 if (req == RTM_ADD && !(rt->rt_flags & RTF_HOST) && 754 rt_mask(rt) != NULL) { 755 struct rtfc_arg arg; 756 arg.rnh = rnh; 757 arg.rt0 = rt; 758 rnh->rnh_walktree_from(rnh, (char *)rt_key(rt), 759 (char *)rt_mask(rt), 760 rt_fixchange, &arg); 761 } 762 763 /* 764 * actually return a resultant rtentry and 765 * give the caller a single reference. 766 */ 767 if (ret_nrt != NULL) { 768 *ret_nrt = rt; 769 rt->rt_refcnt++; 770 } 771 break; 772 default: 773 error = EOPNOTSUPP; 774 } 775 bad: 776 splx(s); 777 return (error); 778 } 779 780 /* 781 * Called from rtrequest(RTM_DELETE, ...) to fix up the route's ``family'' 782 * (i.e., the routes related to it by the operation of cloning). This 783 * routine is iterated over all potential former-child-routes by way of 784 * rnh->rnh_walktree_from() above, and those that actually are children of 785 * the late parent (passed in as VP here) are themselves deleted. 786 */ 787 static int 788 rt_fixdelete(struct radix_node *rn, void *vp) 789 { 790 struct rtentry *rt = (struct rtentry *)rn; 791 struct rtentry *rt0 = vp; 792 793 if (rt->rt_parent == rt0 && 794 !(rt->rt_flags & (RTF_PINNED | RTF_CLONING | RTF_PRCLONING))) { 795 return rtrequest(RTM_DELETE, rt_key(rt), 796 (struct sockaddr *)0, rt_mask(rt), 797 rt->rt_flags, (struct rtentry **)0); 798 } 799 return 0; 800 } 801 802 /* 803 * This routine is called from rt_setgate() to do the analogous thing for 804 * adds and changes. There is the added complication in this case of a 805 * middle insert; i.e., insertion of a new network route between an older 806 * network route and (cloned) host routes. For this reason, a simple check 807 * of rt->rt_parent is insufficient; each candidate route must be tested 808 * against the (mask, value) of the new route (passed as before in vp) 809 * to see if the new route matches it. 810 * 811 * XXX - it may be possible to do fixdelete() for changes and reserve this 812 * routine just for adds. I'm not sure why I thought it was necessary to do 813 * changes this way. 814 */ 815 #ifdef DEBUG 816 static int rtfcdebug = 0; 817 #endif 818 819 static int 820 rt_fixchange(struct radix_node *rn, void *vp) 821 { 822 struct rtentry *rt = (struct rtentry *)rn; 823 struct rtfc_arg *ap = vp; 824 struct rtentry *rt0 = ap->rt0; 825 struct radix_node_head *rnh = ap->rnh; 826 u_char *xk1, *xm1, *xk2, *xmp; 827 int i, len, mlen; 828 829 #ifdef DEBUG 830 if (rtfcdebug) 831 printf("rt_fixchange: rt %p, rt0 %p\n", rt, rt0); 832 #endif 833 834 if (!rt->rt_parent || 835 (rt->rt_flags & (RTF_PINNED | RTF_CLONING | RTF_PRCLONING))) { 836 #ifdef DEBUG 837 if(rtfcdebug) printf("no parent, pinned or cloning\n"); 838 #endif 839 return 0; 840 } 841 842 if (rt->rt_parent == rt0) { 843 #ifdef DEBUG 844 if(rtfcdebug) printf("parent match\n"); 845 #endif 846 return rtrequest(RTM_DELETE, rt_key(rt), 847 (struct sockaddr *)0, rt_mask(rt), 848 rt->rt_flags, (struct rtentry **)0); 849 } 850 851 /* 852 * There probably is a function somewhere which does this... 853 * if not, there should be. 854 */ 855 len = imin(((struct sockaddr *)rt_key(rt0))->sa_len, 856 ((struct sockaddr *)rt_key(rt))->sa_len); 857 858 xk1 = (u_char *)rt_key(rt0); 859 xm1 = (u_char *)rt_mask(rt0); 860 xk2 = (u_char *)rt_key(rt); 861 862 /* avoid applying a less specific route */ 863 xmp = (u_char *)rt_mask(rt->rt_parent); 864 mlen = ((struct sockaddr *)rt_key(rt->rt_parent))->sa_len; 865 if (mlen > ((struct sockaddr *)rt_key(rt0))->sa_len) { 866 #ifdef DEBUG 867 if (rtfcdebug) 868 printf("rt_fixchange: inserting a less " 869 "specific route\n"); 870 #endif 871 return 0; 872 } 873 for (i = rnh->rnh_treetop->rn_offset; i < mlen; i++) { 874 if ((xmp[i] & ~(xmp[i] ^ xm1[i])) != xmp[i]) { 875 #ifdef DEBUG 876 if (rtfcdebug) 877 printf("rt_fixchange: inserting a less " 878 "specific route\n"); 879 #endif 880 return 0; 881 } 882 } 883 884 for (i = rnh->rnh_treetop->rn_offset; i < len; i++) { 885 if ((xk2[i] & xm1[i]) != xk1[i]) { 886 #ifdef DEBUG 887 if(rtfcdebug) printf("no match\n"); 888 #endif 889 return 0; 890 } 891 } 892 893 /* 894 * OK, this node is a clone, and matches the node currently being 895 * changed/added under the node's mask. So, get rid of it. 896 */ 897 #ifdef DEBUG 898 if(rtfcdebug) printf("deleting\n"); 899 #endif 900 return rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0, 901 rt_mask(rt), rt->rt_flags, (struct rtentry **)0); 902 } 903 904 #define ROUNDUP(a) (a>0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long)) 905 906 int 907 rt_setgate(struct rtentry *rt0, struct sockaddr *dst, struct sockaddr *gate) 908 { 909 caddr_t newkey, oldkey; 910 int dlen = ROUNDUP(dst->sa_len), glen = ROUNDUP(gate->sa_len); 911 struct rtentry *rt = rt0; 912 struct radix_node_head *rnh = rt_tables[dst->sa_family]; 913 914 /* 915 * A host route with the destination equal to the gateway 916 * will interfere with keeping LLINFO in the routing 917 * table, so disallow it. 918 */ 919 if (((rt0->rt_flags & (RTF_HOST|RTF_GATEWAY|RTF_LLINFO)) == 920 (RTF_HOST|RTF_GATEWAY)) && 921 (dst->sa_len == gate->sa_len) && 922 (bcmp(dst, gate, dst->sa_len) == 0)) { 923 /* 924 * The route might already exist if this is an RTM_CHANGE 925 * or a routing redirect, so try to delete it. 926 */ 927 if (rt_key(rt0) != NULL) 928 rtrequest(RTM_DELETE, rt_key(rt0), rt0->rt_gateway, 929 rt_mask(rt0), rt0->rt_flags, 0); 930 return EADDRNOTAVAIL; 931 } 932 933 /* 934 * Both dst and gateway are stored in the same malloc'd chunk 935 * (If I ever get my hands on....) 936 * if we need to malloc a new chunk, then keep the old one around 937 * till we don't need it any more. 938 */ 939 if (rt->rt_gateway == NULL || glen > ROUNDUP(rt->rt_gateway->sa_len)) { 940 oldkey = (caddr_t)rt_key(rt); 941 R_Malloc(newkey, caddr_t, dlen + glen); 942 if (newkey == NULL) 943 return ENOBUFS; 944 rt->rt_nodes->rn_key = newkey; 945 } else { 946 /* 947 * otherwise just overwrite the old one 948 */ 949 newkey = rt->rt_nodes->rn_key; 950 oldkey = NULL; 951 } 952 953 /* 954 * copy the new gateway value into the memory chunk 955 */ 956 rt->rt_gateway = (struct sockaddr *)(newkey + dlen); 957 bcopy(gate, rt->rt_gateway, glen); 958 959 /* 960 * if we are replacing the chunk (or it's new) we need to 961 * replace the dst as well 962 */ 963 if (oldkey != NULL) { 964 bcopy(dst, newkey, dlen); 965 Free(oldkey); 966 } 967 968 /* 969 * If there is already a gwroute, it's now almost definitly wrong 970 * so drop it. 971 */ 972 if (rt->rt_gwroute != NULL) { 973 RTFREE(rt->rt_gwroute); 974 rt->rt_gwroute = NULL; 975 } 976 /* 977 * Cloning loop avoidance: 978 * In the presence of protocol-cloning and bad configuration, 979 * it is possible to get stuck in bottomless mutual recursion 980 * (rtrequest rt_setgate rtalloc1). We avoid this by not allowing 981 * protocol-cloning to operate for gateways (which is probably the 982 * correct choice anyway), and avoid the resulting reference loops 983 * by disallowing any route to run through itself as a gateway. 984 * This is obviously mandatory when we get rt->rt_output(). 985 */ 986 if (rt->rt_flags & RTF_GATEWAY) { 987 rt->rt_gwroute = rtalloc1(gate, 1, RTF_PRCLONING); 988 if (rt->rt_gwroute == rt) { 989 RTFREE(rt->rt_gwroute); 990 rt->rt_gwroute = 0; 991 return EDQUOT; /* failure */ 992 } 993 } 994 995 /* 996 * This isn't going to do anything useful for host routes, so 997 * don't bother. Also make sure we have a reasonable mask 998 * (we don't yet have one during adds). 999 */ 1000 if (!(rt->rt_flags & RTF_HOST) && rt_mask(rt) != NULL) { 1001 struct rtfc_arg arg; 1002 arg.rnh = rnh; 1003 arg.rt0 = rt; 1004 rnh->rnh_walktree_from(rnh, (char*)rt_key(rt), 1005 (char *)rt_mask(rt), 1006 rt_fixchange, &arg); 1007 } 1008 1009 return 0; 1010 } 1011 1012 static void 1013 rt_maskedcopy(struct sockaddr *src, struct sockaddr *dst, 1014 struct sockaddr *netmask) 1015 { 1016 u_char *cp1 = (u_char *)src; 1017 u_char *cp2 = (u_char *)dst; 1018 u_char *cp3 = (u_char *)netmask; 1019 u_char *cplim = cp2 + *cp3; 1020 u_char *cplim2 = cp2 + *cp1; 1021 1022 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */ 1023 cp3 += 2; 1024 if (cplim > cplim2) 1025 cplim = cplim2; 1026 while (cp2 < cplim) 1027 *cp2++ = *cp1++ & *cp3++; 1028 if (cp2 < cplim2) 1029 bzero(cp2, (unsigned)(cplim2 - cp2)); 1030 } 1031 1032 /* 1033 * Set up a routing table entry, normally 1034 * for an interface. 1035 */ 1036 int 1037 rtinit(struct ifaddr *ifa, int cmd, int flags) 1038 { 1039 struct rtentry *rt; 1040 struct sockaddr *dst; 1041 struct sockaddr *deldst; 1042 struct sockaddr *netmask; 1043 struct mbuf *m = NULL; 1044 struct rtentry *nrt = NULL; 1045 struct radix_node_head *rnh; 1046 struct radix_node *rn; 1047 int error; 1048 struct rt_addrinfo info; 1049 1050 if (flags & RTF_HOST) { 1051 dst = ifa->ifa_dstaddr; 1052 netmask = NULL; 1053 } else { 1054 dst = ifa->ifa_addr; 1055 netmask = ifa->ifa_netmask; 1056 } 1057 /* 1058 * If it's a delete, check that if it exists, it's on the correct 1059 * interface or we might scrub a route to another ifa which would 1060 * be confusing at best and possibly worse. 1061 */ 1062 if (cmd == RTM_DELETE) { 1063 /* 1064 * It's a delete, so it should already exist.. 1065 * If it's a net, mask off the host bits 1066 * (Assuming we have a mask) 1067 */ 1068 if (netmask != NULL) { 1069 m = m_get(MB_DONTWAIT, MT_SONAME); 1070 if (m == NULL) 1071 return(ENOBUFS); 1072 deldst = mtod(m, struct sockaddr *); 1073 rt_maskedcopy(dst, deldst, netmask); 1074 dst = deldst; 1075 } 1076 /* 1077 * Look up an rtentry that is in the routing tree and 1078 * contains the correct info. 1079 */ 1080 if ((rnh = rt_tables[dst->sa_family]) == NULL || 1081 (rn = rnh->rnh_lookup((char *)dst, (char *)netmask, 1082 rnh)) == NULL || 1083 (rn->rn_flags & RNF_ROOT) || 1084 ((struct rtentry *)rn)->rt_ifa != ifa || 1085 !sa_equal(SA(rn->rn_key), dst)) { 1086 if (m != NULL) 1087 (void) m_free(m); 1088 return (flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH); 1089 } 1090 /* XXX */ 1091 #if 0 1092 else { 1093 /* 1094 * One would think that as we are deleting, and we know 1095 * it doesn't exist, we could just return at this point 1096 * with an "ELSE" clause, but apparently not.. 1097 */ 1098 return (flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH); 1099 } 1100 #endif 1101 } 1102 /* 1103 * Do the actual request 1104 */ 1105 bzero(&info, sizeof(info)); 1106 info.rti_ifa = ifa; 1107 info.rti_flags = flags | ifa->ifa_flags; 1108 info.rti_info[RTAX_DST] = dst; 1109 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr; 1110 info.rti_info[RTAX_NETMASK] = netmask; 1111 error = rtrequest1(cmd, &info, &nrt); 1112 if (error == 0 && (rt = nrt) != NULL) { 1113 /* 1114 * notify any listening routing agents of the change 1115 */ 1116 rt_newaddrmsg(cmd, ifa, error, rt); 1117 if (cmd == RTM_DELETE) { 1118 /* 1119 * If we are deleting, and we found an entry, then 1120 * it's been removed from the tree.. now throw it away. 1121 */ 1122 if (rt->rt_refcnt <= 0) { 1123 rt->rt_refcnt++; /* make a 1->0 transition */ 1124 rtfree(rt); 1125 } 1126 } else if (cmd == RTM_ADD) { 1127 /* 1128 * We just wanted to add it.. we don't actually 1129 * need a reference. 1130 */ 1131 rt->rt_refcnt--; 1132 } 1133 } 1134 if (m != NULL) 1135 (void) m_free(m); 1136 return (error); 1137 } 1138 1139 /* This must be before ip6_init2(), which is now SI_ORDER_MIDDLE */ 1140 SYSINIT(route, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, route_init, 0); 1141