1 /* 2 * Copyright (c) 2004, 2005 The DragonFly Project. All rights reserved. 3 * 4 * This code is derived from software contributed to The DragonFly Project 5 * by Jeffrey M. Hsu. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of The DragonFly Project nor the names of its 16 * contributors may be used to endorse or promote products derived 17 * from this software without specific, prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 21 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 22 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 23 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 24 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING, 25 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 26 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED 27 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 28 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT 29 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33 /* 34 * Copyright (c) 2004, 2005 Jeffrey M. Hsu. All rights reserved. 35 * 36 * License terms: all terms for the DragonFly license above plus the following: 37 * 38 * 4. All advertising materials mentioning features or use of this software 39 * must display the following acknowledgement: 40 * 41 * This product includes software developed by Jeffrey M. Hsu 42 * for the DragonFly Project. 43 * 44 * This requirement may be waived with permission from Jeffrey Hsu. 45 * Permission will be granted to any DragonFly user for free. 46 * This requirement will sunset and may be removed on Jan 31, 2006, 47 * after which the standard DragonFly license (as shown above) will 48 * apply. 49 */ 50 51 /* 52 * Copyright (c) 1980, 1986, 1991, 1993 53 * The Regents of the University of California. All rights reserved. 54 * 55 * Redistribution and use in source and binary forms, with or without 56 * modification, are permitted provided that the following conditions 57 * are met: 58 * 1. Redistributions of source code must retain the above copyright 59 * notice, this list of conditions and the following disclaimer. 60 * 2. Redistributions in binary form must reproduce the above copyright 61 * notice, this list of conditions and the following disclaimer in the 62 * documentation and/or other materials provided with the distribution. 63 * 3. All advertising materials mentioning features or use of this software 64 * must display the following acknowledgement: 65 * This product includes software developed by the University of 66 * California, Berkeley and its contributors. 67 * 4. Neither the name of the University nor the names of its contributors 68 * may be used to endorse or promote products derived from this software 69 * without specific prior written permission. 70 * 71 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 72 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 73 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 74 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 75 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 76 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 77 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 78 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 79 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 80 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 81 * SUCH DAMAGE. 82 * 83 * @(#)route.c 8.3 (Berkeley) 1/9/95 84 * $FreeBSD: src/sys/net/route.c,v 1.59.2.10 2003/01/17 08:04:00 ru Exp $ 85 * $DragonFly: src/sys/net/route.c,v 1.16 2005/02/28 11:39:33 hsu Exp $ 86 */ 87 88 #include "opt_inet.h" 89 90 #include <sys/param.h> 91 #include <sys/systm.h> 92 #include <sys/malloc.h> 93 #include <sys/mbuf.h> 94 #include <sys/socket.h> 95 #include <sys/domain.h> 96 #include <sys/kernel.h> 97 98 #include <net/if.h> 99 #include <net/route.h> 100 101 #include <netinet/in.h> 102 #include <net/ip_mroute/ip_mroute.h> 103 104 static struct rtstat rtstat; 105 struct radix_node_head *rt_tables[AF_MAX+1]; 106 107 static void rt_maskedcopy (struct sockaddr *, struct sockaddr *, 108 struct sockaddr *); 109 static void rtable_init (void **); 110 111 static void 112 rtable_init(void **table) 113 { 114 struct domain *dom; 115 116 for (dom = domains; dom; dom = dom->dom_next) 117 if (dom->dom_rtattach) 118 dom->dom_rtattach(&table[dom->dom_family], 119 dom->dom_rtoffset); 120 } 121 122 void 123 route_init() 124 { 125 rn_init(); /* initialize all zeroes, all ones, mask table */ 126 rtable_init((void **)rt_tables); 127 } 128 129 /* 130 * Packet routing routines. 131 */ 132 133 /* 134 * Look up and fill in the "ro_rt" rtentry field in a route structure given 135 * an address in the "ro_dst" field. Always send a report on a miss and 136 * always clone routes. 137 */ 138 void 139 rtalloc(struct route *ro) 140 { 141 rtalloc_ign(ro, 0UL); 142 } 143 144 /* 145 * Look up and fill in the "ro_rt" rtentry field in a route structure given 146 * an address in the "ro_dst" field. Always send a report on a miss and 147 * optionally clone routes when RTF_CLONING or RTF_PRCLONING are not being 148 * ignored. 149 */ 150 void 151 rtalloc_ign(struct route *ro, u_long ignoreflags) 152 { 153 if (ro->ro_rt != NULL) { 154 if (ro->ro_rt->rt_ifp != NULL && ro->ro_rt->rt_flags & RTF_UP) 155 return; 156 rtfree(ro->ro_rt); 157 ro->ro_rt = NULL; 158 } 159 ro->ro_rt = _rtlookup(&ro->ro_dst, RTL_REPORTMSG, ignoreflags); 160 } 161 162 /* 163 * Look up the route that matches the given "dst" address. 164 * 165 * Route lookup can have the side-effect of creating and returning 166 * a cloned route instead if "dst" matches a cloning route and the 167 * RTF_CLONING and RTF_PRCLONING flags are not being ignored. 168 * 169 * Any route returned has its reference count incremented. 170 */ 171 struct rtentry * 172 _rtlookup(struct sockaddr *dst, boolean_t report, u_long ignore) 173 { 174 struct radix_node_head *rnh = rt_tables[dst->sa_family]; 175 struct rtentry *rt; 176 177 if (rnh == NULL) 178 goto unreach; 179 180 /* 181 * Look up route in the radix tree. 182 */ 183 rt = (struct rtentry *) rnh->rnh_matchaddr((char *)dst, rnh); 184 if (rt == NULL) 185 goto unreach; 186 187 /* 188 * Handle cloning routes. 189 */ 190 if ((rt->rt_flags & ~ignore & (RTF_CLONING | RTF_PRCLONING)) != 0) { 191 struct rtentry *clonedroute; 192 int error; 193 194 clonedroute = rt; /* copy in/copy out parameter */ 195 error = rtrequest(RTM_RESOLVE, dst, NULL, NULL, 0, 196 &clonedroute); /* clone the route */ 197 if (error != 0) { /* cloning failed */ 198 if (report) 199 rt_dstmsg(RTM_MISS, dst, error); 200 rt->rt_refcnt++; 201 return (rt); /* return the uncloned route */ 202 } 203 if (report) { 204 if (clonedroute->rt_flags & RTF_XRESOLVE) 205 rt_dstmsg(RTM_RESOLVE, dst, 0); 206 else 207 rt_rtmsg(RTM_ADD, clonedroute, 208 clonedroute->rt_ifp, 0); 209 } 210 return (clonedroute); /* return cloned route */ 211 } 212 213 /* 214 * Increment the reference count of the matched route and return. 215 */ 216 rt->rt_refcnt++; 217 return (rt); 218 219 unreach: 220 rtstat.rts_unreach++; 221 if (report) 222 rt_dstmsg(RTM_MISS, dst, 0); 223 return (NULL); 224 } 225 226 void 227 rtfree(struct rtentry *rt) 228 { 229 KASSERT(rt->rt_refcnt > 0, ("rtfree: rt_refcnt %ld", rt->rt_refcnt)); 230 231 --rt->rt_refcnt; 232 if (rt->rt_refcnt == 0) { 233 struct radix_node_head *rnh = rt_tables[rt_key(rt)->sa_family]; 234 235 if (rnh->rnh_close) 236 rnh->rnh_close((struct radix_node *)rt, rnh); 237 if (!(rt->rt_flags & RTF_UP)) { 238 /* deallocate route */ 239 if (rt->rt_ifa != NULL) 240 IFAFREE(rt->rt_ifa); 241 if (rt->rt_parent != NULL) 242 RTFREE(rt->rt_parent); /* recursive call! */ 243 Free(rt_key(rt)); 244 Free(rt); 245 } 246 } 247 } 248 249 /* 250 * Force a routing table entry to the specified 251 * destination to go through the given gateway. 252 * Normally called as a result of a routing redirect 253 * message from the network layer. 254 * 255 * N.B.: must be called at splnet 256 */ 257 void 258 rtredirect( 259 struct sockaddr *dst, 260 struct sockaddr *gateway, 261 struct sockaddr *netmask, 262 int flags, 263 struct sockaddr *src, 264 struct rtentry **rtp) 265 { 266 struct rtentry *rt; 267 struct rt_addrinfo info; 268 struct ifaddr *ifa; 269 short *stat = NULL; 270 int error; 271 272 /* verify the gateway is directly reachable */ 273 if ((ifa = ifa_ifwithnet(gateway)) == NULL) { 274 error = ENETUNREACH; 275 goto out; 276 } 277 278 /* 279 * If the redirect isn't from our current router for this dst, 280 * it's either old or wrong. If it redirects us to ourselves, 281 * we have a routing loop, perhaps as a result of an interface 282 * going down recently. 283 */ 284 if (!(flags & RTF_DONE) && 285 (rt = rtpurelookup(dst)) != NULL && 286 (!sa_equal(src, rt->rt_gateway) || rt->rt_ifa != ifa)) { 287 error = EINVAL; 288 goto done; 289 } else if (ifa_ifwithaddr(gateway)) { 290 error = EHOSTUNREACH; 291 goto done; 292 } 293 294 /* 295 * Create a new entry if we just got back a wildcard entry 296 * or the the lookup failed. This is necessary for hosts 297 * which use routing redirects generated by smart gateways 298 * to dynamically build the routing tables. 299 */ 300 if (rt == NULL || (rt_mask(rt) != NULL && rt_mask(rt)->sa_len < 2)) 301 goto create; 302 303 /* 304 * Don't listen to the redirect if it's for a route to an interface. 305 */ 306 if (rt->rt_flags & RTF_GATEWAY) { 307 if ((!(rt->rt_flags & RTF_HOST)) && (flags & RTF_HOST)) { 308 /* 309 * Changing from route to net => route to host. 310 * Create new route, rather than smashing route to net. 311 */ 312 create: 313 if (rt != NULL) 314 rtfree(rt); 315 flags |= RTF_GATEWAY | RTF_DYNAMIC; 316 bzero(&info, sizeof info); 317 info.rti_info[RTAX_DST] = dst; 318 info.rti_info[RTAX_GATEWAY] = gateway; 319 info.rti_info[RTAX_NETMASK] = netmask; 320 info.rti_ifa = ifa; 321 info.rti_flags = flags; 322 rt = NULL; 323 error = rtrequest1(RTM_ADD, &info, &rt); 324 if (rt != NULL) 325 flags = rt->rt_flags; 326 stat = &rtstat.rts_dynamic; 327 } else { 328 /* 329 * Smash the current notion of the gateway to 330 * this destination. Should check about netmask!!! 331 */ 332 rt->rt_flags |= RTF_MODIFIED; 333 flags |= RTF_MODIFIED; 334 stat = &rtstat.rts_newgateway; 335 /* Add the key and gateway (in one malloc'ed chunk). */ 336 rt_setgate(rt, rt_key(rt), gateway); 337 error = 0; 338 } 339 } else { 340 error = EHOSTUNREACH; 341 } 342 343 done: 344 if (rt != NULL) { 345 if (rtp != NULL && error == 0) 346 *rtp = rt; 347 else 348 rtfree(rt); 349 } 350 351 out: 352 if (error != 0) 353 rtstat.rts_badredirect++; 354 else if (stat != NULL) 355 (*stat)++; 356 357 bzero(&info, sizeof info); 358 info.rti_info[RTAX_DST] = dst; 359 info.rti_info[RTAX_GATEWAY] = gateway; 360 info.rti_info[RTAX_NETMASK] = netmask; 361 info.rti_info[RTAX_AUTHOR] = src; 362 rt_missmsg(RTM_REDIRECT, &info, flags, error); 363 } 364 365 /* 366 * Routing table ioctl interface. 367 */ 368 int 369 rtioctl(u_long req, caddr_t data, struct thread *td) 370 { 371 #ifdef INET 372 /* Multicast goop, grrr... */ 373 return mrt_ioctl ? mrt_ioctl(req, data) : EOPNOTSUPP; 374 #else 375 return ENXIO; 376 #endif 377 } 378 379 struct ifaddr * 380 ifa_ifwithroute(int flags, struct sockaddr *dst, struct sockaddr *gateway) 381 { 382 struct ifaddr *ifa; 383 384 if (!(flags & RTF_GATEWAY)) { 385 /* 386 * If we are adding a route to an interface, 387 * and the interface is a point-to-point link, 388 * we should search for the destination 389 * as our clue to the interface. Otherwise 390 * we can use the local address. 391 */ 392 ifa = NULL; 393 if (flags & RTF_HOST) { 394 ifa = ifa_ifwithdstaddr(dst); 395 } 396 if (ifa == NULL) 397 ifa = ifa_ifwithaddr(gateway); 398 } else { 399 /* 400 * If we are adding a route to a remote net 401 * or host, the gateway may still be on the 402 * other end of a pt to pt link. 403 */ 404 ifa = ifa_ifwithdstaddr(gateway); 405 } 406 if (ifa == NULL) 407 ifa = ifa_ifwithnet(gateway); 408 if (ifa == NULL) { 409 struct rtentry *rt; 410 411 rt = rtpurelookup(gateway); 412 if (rt == NULL) 413 return (NULL); 414 rt->rt_refcnt--; 415 if ((ifa = rt->rt_ifa) == NULL) 416 return (NULL); 417 } 418 if (ifa->ifa_addr->sa_family != dst->sa_family) { 419 struct ifaddr *oifa = ifa; 420 421 ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp); 422 if (ifa == NULL) 423 ifa = oifa; 424 } 425 return (ifa); 426 } 427 428 static int rt_fixdelete (struct radix_node *, void *); 429 static int rt_fixchange (struct radix_node *, void *); 430 431 struct rtfc_arg { 432 struct rtentry *rt0; 433 struct radix_node_head *rnh; 434 }; 435 436 /* 437 * Set rtinfo->rti_ifa and rtinfo->rti_ifp. 438 */ 439 int 440 rt_getifa(struct rt_addrinfo *rtinfo) 441 { 442 struct sockaddr *gateway = rtinfo->rti_info[RTAX_GATEWAY]; 443 struct sockaddr *dst = rtinfo->rti_info[RTAX_DST]; 444 struct sockaddr *ifaaddr = rtinfo->rti_info[RTAX_IFA]; 445 int flags = rtinfo->rti_flags; 446 447 /* 448 * ifp may be specified by sockaddr_dl 449 * when protocol address is ambiguous. 450 */ 451 if (rtinfo->rti_ifp == NULL) { 452 struct sockaddr *ifpaddr; 453 454 ifpaddr = rtinfo->rti_info[RTAX_IFP]; 455 if (ifpaddr != NULL && ifpaddr->sa_family == AF_LINK) { 456 struct ifaddr *ifa; 457 458 ifa = ifa_ifwithnet(ifpaddr); 459 if (ifa != NULL) 460 rtinfo->rti_ifp = ifa->ifa_ifp; 461 } 462 } 463 464 if (rtinfo->rti_ifa == NULL && ifaaddr != NULL) 465 rtinfo->rti_ifa = ifa_ifwithaddr(ifaaddr); 466 if (rtinfo->rti_ifa == NULL) { 467 struct sockaddr *sa; 468 469 sa = ifaaddr != NULL ? ifaaddr : 470 (gateway != NULL ? gateway : dst); 471 if (sa != NULL && rtinfo->rti_ifp != NULL) 472 rtinfo->rti_ifa = ifaof_ifpforaddr(sa, rtinfo->rti_ifp); 473 else if (dst != NULL && gateway != NULL) 474 rtinfo->rti_ifa = ifa_ifwithroute(flags, dst, gateway); 475 else if (sa != NULL) 476 rtinfo->rti_ifa = ifa_ifwithroute(flags, sa, sa); 477 } 478 if (rtinfo->rti_ifa == NULL) 479 return (ENETUNREACH); 480 481 if (rtinfo->rti_ifp == NULL) 482 rtinfo->rti_ifp = rtinfo->rti_ifa->ifa_ifp; 483 return (0); 484 } 485 486 /* 487 * Do appropriate manipulations of a routing tree given 488 * all the bits of info needed 489 */ 490 int 491 rtrequest( 492 int req, 493 struct sockaddr *dst, 494 struct sockaddr *gateway, 495 struct sockaddr *netmask, 496 int flags, 497 struct rtentry **ret_nrt) 498 { 499 struct rt_addrinfo info; 500 501 bzero(&info, sizeof info); 502 info.rti_flags = flags; 503 info.rti_info[RTAX_DST] = dst; 504 info.rti_info[RTAX_GATEWAY] = gateway; 505 info.rti_info[RTAX_NETMASK] = netmask; 506 return rtrequest1(req, &info, ret_nrt); 507 } 508 509 int 510 rtrequest1(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt) 511 { 512 struct sockaddr *dst = info->rti_info[RTAX_DST]; 513 struct rtentry *rt; 514 struct radix_node *rn; 515 struct radix_node_head *rnh; 516 struct ifaddr *ifa; 517 struct sockaddr *ndst; 518 int error = 0; 519 int s; 520 521 #define gotoerr(x) { error = x ; goto bad; } 522 523 s = splnet(); 524 /* 525 * Find the correct routing tree to use for this Address Family 526 */ 527 if ((rnh = rt_tables[dst->sa_family]) == NULL) 528 gotoerr(EAFNOSUPPORT); 529 530 /* 531 * If we are adding a host route then we don't want to put 532 * a netmask in the tree, nor do we want to clone it. 533 */ 534 if (info->rti_flags & RTF_HOST) { 535 info->rti_info[RTAX_NETMASK] = NULL; 536 info->rti_flags &= ~(RTF_CLONING | RTF_PRCLONING); 537 } 538 539 switch (req) { 540 case RTM_DELETE: 541 /* Remove the item from the tree. */ 542 rn = rnh->rnh_deladdr((char *)info->rti_info[RTAX_DST], 543 (char *)info->rti_info[RTAX_NETMASK], 544 rnh); 545 if (rn == NULL) 546 gotoerr(ESRCH); 547 KASSERT(!(rn->rn_flags & (RNF_ACTIVE | RNF_ROOT)), 548 ("rnh_deladdr returned flags 0x%x", rn->rn_flags)); 549 rt = (struct rtentry *)rn; 550 551 /* Free any routes cloned from this one. */ 552 if ((rt->rt_flags & (RTF_CLONING | RTF_PRCLONING)) && 553 rt_mask(rt) != NULL) { 554 rnh->rnh_walktree_from(rnh, (char *)rt_key(rt), 555 (char *)rt_mask(rt), 556 rt_fixdelete, rt); 557 } 558 559 if (rt->rt_gwroute != NULL) { 560 RTFREE(rt->rt_gwroute); 561 rt->rt_gwroute = NULL; 562 } 563 564 /* 565 * NB: RTF_UP must be set during the search above, 566 * because we might delete the last ref, causing 567 * rt to get freed prematurely. 568 */ 569 rt->rt_flags &= ~RTF_UP; 570 571 /* Give the protocol a chance to keep things in sync. */ 572 if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest) 573 ifa->ifa_rtrequest(RTM_DELETE, rt, info); 574 575 /* 576 * If the caller wants it, then it can have it, 577 * but it's up to it to free the rtentry as we won't be 578 * doing it. 579 */ 580 KASSERT(rt->rt_refcnt >= 0, 581 ("rtrequest1(DELETE): refcnt %ld", rt->rt_refcnt)); 582 if (ret_nrt != NULL) { 583 *ret_nrt = rt; 584 } else if (rt->rt_refcnt == 0) { 585 rt->rt_refcnt++; /* refcnt > 0 required for rtfree() */ 586 rtfree(rt); 587 } 588 break; 589 590 case RTM_RESOLVE: 591 if (ret_nrt == NULL || (rt = *ret_nrt) == NULL) 592 gotoerr(EINVAL); 593 ifa = rt->rt_ifa; 594 info->rti_flags = 595 rt->rt_flags & ~(RTF_CLONING | RTF_PRCLONING | RTF_STATIC); 596 info->rti_flags |= RTF_WASCLONED; 597 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway; 598 if ((info->rti_info[RTAX_NETMASK] = rt->rt_genmask) == NULL) 599 info->rti_flags |= RTF_HOST; 600 goto makeroute; 601 602 case RTM_ADD: 603 KASSERT(!(info->rti_flags & RTF_GATEWAY) || 604 info->rti_info[RTAX_GATEWAY] != NULL, 605 ("rtrequest: GATEWAY but no gateway")); 606 607 if (info->rti_ifa == NULL && (error = rt_getifa(info))) 608 gotoerr(error); 609 ifa = info->rti_ifa; 610 makeroute: 611 R_Malloc(rt, struct rtentry *, sizeof *rt); 612 if (rt == NULL) 613 gotoerr(ENOBUFS); 614 bzero(rt, sizeof *rt); 615 rt->rt_flags = RTF_UP | info->rti_flags; 616 error = rt_setgate(rt, dst, info->rti_info[RTAX_GATEWAY]); 617 if (error != 0) { 618 Free(rt); 619 gotoerr(error); 620 } 621 622 ndst = rt_key(rt); 623 if (info->rti_info[RTAX_NETMASK] != NULL) 624 rt_maskedcopy(dst, ndst, info->rti_info[RTAX_NETMASK]); 625 else 626 bcopy(dst, ndst, dst->sa_len); 627 628 /* 629 * Note that we now have a reference to the ifa. 630 * This moved from below so that rnh->rnh_addaddr() can 631 * examine the ifa and ifa->ifa_ifp if it so desires. 632 */ 633 IFAREF(ifa); 634 rt->rt_ifa = ifa; 635 rt->rt_ifp = ifa->ifa_ifp; 636 /* XXX mtu manipulation will be done in rnh_addaddr -- itojun */ 637 638 rn = rnh->rnh_addaddr((char *)ndst, 639 (char *)info->rti_info[RTAX_NETMASK], 640 rnh, rt->rt_nodes); 641 if (rn == NULL) { 642 struct rtentry *oldrt; 643 644 /* 645 * We already have one of these in the tree. 646 * We do a special hack: if the old route was 647 * cloned, then we blow it away and try 648 * re-inserting the new one. 649 */ 650 oldrt = rtpurelookup(ndst); 651 if (oldrt != NULL) { 652 --oldrt->rt_refcnt; 653 if (oldrt->rt_flags & RTF_WASCLONED) { 654 rtrequest(RTM_DELETE, rt_key(oldrt), 655 oldrt->rt_gateway, 656 rt_mask(oldrt), 657 oldrt->rt_flags, NULL); 658 rn = rnh->rnh_addaddr((char *)ndst, 659 (char *) 660 info->rti_info[RTAX_NETMASK], 661 rnh, rt->rt_nodes); 662 } 663 } 664 } 665 666 /* 667 * If it still failed to go into the tree, 668 * then un-make it (this should be a function). 669 */ 670 if (rn == NULL) { 671 if (rt->rt_gwroute != NULL) 672 rtfree(rt->rt_gwroute); 673 IFAFREE(ifa); 674 Free(rt_key(rt)); 675 Free(rt); 676 gotoerr(EEXIST); 677 } 678 679 /* 680 * If we got here from RESOLVE, then we are cloning 681 * so clone the rest, and note that we 682 * are a clone (and increment the parent's references) 683 */ 684 if (req == RTM_RESOLVE) { 685 rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */ 686 rt->rt_rmx.rmx_pksent = 0; /* reset packet counter */ 687 if ((*ret_nrt)->rt_flags & 688 (RTF_CLONING | RTF_PRCLONING)) { 689 rt->rt_parent = *ret_nrt; 690 (*ret_nrt)->rt_refcnt++; 691 } 692 } 693 694 /* 695 * if this protocol has something to add to this then 696 * allow it to do that as well. 697 */ 698 if (ifa->ifa_rtrequest != NULL) 699 ifa->ifa_rtrequest(req, rt, info); 700 701 /* 702 * We repeat the same procedure from rt_setgate() here because 703 * it doesn't fire when we call it there because the node 704 * hasn't been added to the tree yet. 705 */ 706 if (req == RTM_ADD && !(rt->rt_flags & RTF_HOST) && 707 rt_mask(rt) != NULL) { 708 struct rtfc_arg arg = { rt, rnh }; 709 710 rnh->rnh_walktree_from(rnh, (char *)rt_key(rt), 711 (char *)rt_mask(rt), 712 rt_fixchange, &arg); 713 } 714 715 /* 716 * Return the resulting rtentry, 717 * increasing the number of references by one. 718 */ 719 if (ret_nrt != NULL) { 720 rt->rt_refcnt++; 721 *ret_nrt = rt; 722 } 723 break; 724 default: 725 error = EOPNOTSUPP; 726 } 727 bad: 728 splx(s); 729 return (error); 730 } 731 732 /* 733 * Called from rtrequest(RTM_DELETE, ...) to fix up the route's ``family'' 734 * (i.e., the routes related to it by the operation of cloning). This 735 * routine is iterated over all potential former-child-routes by way of 736 * rnh->rnh_walktree_from() above, and those that actually are children of 737 * the late parent (passed in as VP here) are themselves deleted. 738 */ 739 static int 740 rt_fixdelete(struct radix_node *rn, void *vp) 741 { 742 struct rtentry *rt = (struct rtentry *)rn; 743 struct rtentry *rt0 = vp; 744 745 if (rt->rt_parent == rt0 && 746 !(rt->rt_flags & (RTF_PINNED | RTF_CLONING | RTF_PRCLONING))) { 747 return rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 748 rt->rt_flags, NULL); 749 } 750 return 0; 751 } 752 753 /* 754 * This routine is called from rt_setgate() to do the analogous thing for 755 * adds and changes. There is the added complication in this case of a 756 * middle insert; i.e., insertion of a new network route between an older 757 * network route and (cloned) host routes. For this reason, a simple check 758 * of rt->rt_parent is insufficient; each candidate route must be tested 759 * against the (mask, value) of the new route (passed as before in vp) 760 * to see if the new route matches it. 761 * 762 * XXX - it may be possible to do fixdelete() for changes and reserve this 763 * routine just for adds. I'm not sure why I thought it was necessary to do 764 * changes this way. 765 */ 766 #ifdef DEBUG 767 static int rtfcdebug = 0; 768 #endif 769 770 static int 771 rt_fixchange(struct radix_node *rn, void *vp) 772 { 773 struct rtentry *rt = (struct rtentry *)rn; 774 struct rtfc_arg *ap = vp; 775 struct rtentry *rt0 = ap->rt0; 776 struct radix_node_head *rnh = ap->rnh; 777 u_char *xk1, *xm1, *xk2, *xmp; 778 int i, len, mlen; 779 780 #ifdef DEBUG 781 if (rtfcdebug) 782 printf("rt_fixchange: rt %p, rt0 %p\n", rt, rt0); 783 #endif 784 785 if (rt->rt_parent == NULL || 786 (rt->rt_flags & (RTF_PINNED | RTF_CLONING | RTF_PRCLONING))) { 787 #ifdef DEBUG 788 if (rtfcdebug) printf("no parent, pinned or cloning\n"); 789 #endif 790 return 0; 791 } 792 793 if (rt->rt_parent == rt0) { 794 #ifdef DEBUG 795 if (rtfcdebug) printf("parent match\n"); 796 #endif 797 return rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 798 rt->rt_flags, NULL); 799 } 800 801 /* 802 * There probably is a function somewhere which does this... 803 * if not, there should be. 804 */ 805 len = imin(rt_key(rt0)->sa_len, rt_key(rt)->sa_len); 806 807 xk1 = (u_char *)rt_key(rt0); 808 xm1 = (u_char *)rt_mask(rt0); 809 xk2 = (u_char *)rt_key(rt); 810 811 /* avoid applying a less specific route */ 812 xmp = (u_char *)rt_mask(rt->rt_parent); 813 mlen = rt_key(rt->rt_parent)->sa_len; 814 if (mlen > rt_key(rt0)->sa_len) { 815 #ifdef DEBUG 816 if (rtfcdebug) 817 printf("rt_fixchange: inserting a less " 818 "specific route\n"); 819 #endif 820 return 0; 821 } 822 for (i = rnh->rnh_treetop->rn_offset; i < mlen; i++) { 823 if ((xmp[i] & ~(xmp[i] ^ xm1[i])) != xmp[i]) { 824 #ifdef DEBUG 825 if (rtfcdebug) 826 printf("rt_fixchange: inserting a less " 827 "specific route\n"); 828 #endif 829 return 0; 830 } 831 } 832 833 for (i = rnh->rnh_treetop->rn_offset; i < len; i++) { 834 if ((xk2[i] & xm1[i]) != xk1[i]) { 835 #ifdef DEBUG 836 if (rtfcdebug) printf("no match\n"); 837 #endif 838 return 0; 839 } 840 } 841 842 /* 843 * OK, this node is a clone, and matches the node currently being 844 * changed/added under the node's mask. So, get rid of it. 845 */ 846 #ifdef DEBUG 847 if (rtfcdebug) printf("deleting\n"); 848 #endif 849 return rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 850 rt->rt_flags, NULL); 851 } 852 853 #define ROUNDUP(a) (a>0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long)) 854 855 int 856 rt_setgate(struct rtentry *rt0, struct sockaddr *dst, struct sockaddr *gate) 857 { 858 char *space, *oldspace; 859 int dlen = ROUNDUP(dst->sa_len), glen = ROUNDUP(gate->sa_len); 860 struct rtentry *rt = rt0; 861 struct radix_node_head *rnh = rt_tables[dst->sa_family]; 862 863 /* 864 * A host route with the destination equal to the gateway 865 * will interfere with keeping LLINFO in the routing 866 * table, so disallow it. 867 */ 868 if (((rt0->rt_flags & (RTF_HOST | RTF_GATEWAY | RTF_LLINFO)) == 869 (RTF_HOST | RTF_GATEWAY)) && 870 dst->sa_len == gate->sa_len && 871 sa_equal(dst, gate)) { 872 /* 873 * The route might already exist if this is an RTM_CHANGE 874 * or a routing redirect, so try to delete it. 875 */ 876 if (rt_key(rt0) != NULL) 877 rtrequest(RTM_DELETE, rt_key(rt0), rt0->rt_gateway, 878 rt_mask(rt0), rt0->rt_flags, NULL); 879 return EADDRNOTAVAIL; 880 } 881 882 /* 883 * Both dst and gateway are stored in the same malloc'ed chunk 884 * (If I ever get my hands on....) 885 * if we need to malloc a new chunk, then keep the old one around 886 * till we don't need it any more. 887 */ 888 if (rt->rt_gateway == NULL || glen > ROUNDUP(rt->rt_gateway->sa_len)) { 889 oldspace = (char *)rt_key(rt); 890 R_Malloc(space, char *, dlen + glen); 891 if (space == NULL) 892 return ENOBUFS; 893 rt->rt_nodes->rn_key = space; 894 } else { 895 space = (char *)rt_key(rt); /* Just use the old space. */ 896 oldspace = NULL; 897 } 898 899 /* Set the gateway value. */ 900 rt->rt_gateway = (struct sockaddr *)(space + dlen); 901 bcopy(gate, rt->rt_gateway, glen); 902 903 if (oldspace != NULL) { 904 /* 905 * If we allocated a new chunk, preserve the original dst. 906 * This way, rt_setgate() really just sets the gate 907 * and leaves the dst field alone. 908 */ 909 bcopy(dst, space, dlen); 910 Free(oldspace); 911 } 912 913 /* 914 * If there is already a gwroute, it's now almost definitely wrong 915 * so drop it. 916 */ 917 if (rt->rt_gwroute != NULL) { 918 RTFREE(rt->rt_gwroute); 919 rt->rt_gwroute = NULL; 920 } 921 if (rt->rt_flags & RTF_GATEWAY) { 922 /* 923 * Cloning loop avoidance: In the presence of 924 * protocol-cloning and bad configuration, it is 925 * possible to get stuck in bottomless mutual recursion 926 * (rtrequest rt_setgate rtlookup). We avoid this 927 * by not allowing protocol-cloning to operate for 928 * gateways (which is probably the correct choice 929 * anyway), and avoid the resulting reference loops 930 * by disallowing any route to run through itself as 931 * a gateway. This is obviously mandatory when we 932 * get rt->rt_output(). 933 * 934 * This breaks TTCP for hosts outside the gateway! XXX JH 935 */ 936 rt->rt_gwroute = _rtlookup(gate, RTL_REPORTMSG, RTF_PRCLONING); 937 if (rt->rt_gwroute == rt) { 938 rt->rt_gwroute = NULL; 939 --rt->rt_refcnt; 940 return EDQUOT; /* failure */ 941 } 942 } 943 944 /* 945 * This isn't going to do anything useful for host routes, so 946 * don't bother. Also make sure we have a reasonable mask 947 * (we don't yet have one during adds). 948 */ 949 if (!(rt->rt_flags & RTF_HOST) && rt_mask(rt) != NULL) { 950 struct rtfc_arg arg = { rt, rnh }; 951 952 rnh->rnh_walktree_from(rnh, (char *)rt_key(rt), 953 (char *)rt_mask(rt), 954 rt_fixchange, &arg); 955 } 956 957 return 0; 958 } 959 960 static void 961 rt_maskedcopy( 962 struct sockaddr *src, 963 struct sockaddr *dst, 964 struct sockaddr *netmask) 965 { 966 u_char *cp1 = (u_char *)src; 967 u_char *cp2 = (u_char *)dst; 968 u_char *cp3 = (u_char *)netmask; 969 u_char *cplim = cp2 + *cp3; 970 u_char *cplim2 = cp2 + *cp1; 971 972 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */ 973 cp3 += 2; 974 if (cplim > cplim2) 975 cplim = cplim2; 976 while (cp2 < cplim) 977 *cp2++ = *cp1++ & *cp3++; 978 if (cp2 < cplim2) 979 bzero(cp2, cplim2 - cp2); 980 } 981 982 int 983 rt_llroute(struct sockaddr *dst, struct rtentry *rt0, struct rtentry **drt) 984 { 985 struct rtentry *up_rt, *rt; 986 987 if (!(rt0->rt_flags & RTF_UP)) { 988 up_rt = rtlookup(dst); 989 if (up_rt == NULL) 990 return (EHOSTUNREACH); 991 up_rt->rt_refcnt--; 992 } else 993 up_rt = rt0; 994 if (up_rt->rt_flags & RTF_GATEWAY) { 995 if (up_rt->rt_gwroute == NULL) { 996 up_rt->rt_gwroute = rtlookup(up_rt->rt_gateway); 997 if (up_rt->rt_gwroute == NULL) 998 return (EHOSTUNREACH); 999 } else if (!(up_rt->rt_gwroute->rt_flags & RTF_UP)) { 1000 rtfree(up_rt->rt_gwroute); 1001 up_rt->rt_gwroute = rtlookup(up_rt->rt_gateway); 1002 if (up_rt->rt_gwroute == NULL) 1003 return (EHOSTUNREACH); 1004 } 1005 rt = up_rt->rt_gwroute; 1006 } else 1007 rt = up_rt; 1008 if (rt->rt_flags & RTF_REJECT && 1009 (rt->rt_rmx.rmx_expire == 0 || /* rt doesn't expire */ 1010 time_second < rt->rt_rmx.rmx_expire)) /* rt not expired */ 1011 return (rt->rt_flags & RTF_HOST ? EHOSTDOWN : EHOSTUNREACH); 1012 *drt = rt; 1013 return 0; 1014 } 1015 1016 /* 1017 * Set up a routing table entry, normally 1018 * for an interface. 1019 */ 1020 int 1021 rtinit(struct ifaddr *ifa, int cmd, int flags) 1022 { 1023 struct sockaddr *dst, *deldst, *netmask; 1024 struct rtentry *rt; 1025 struct rtentry *nrt = NULL; 1026 struct mbuf *m = NULL; 1027 struct radix_node_head *rnh; 1028 struct radix_node *rn; 1029 struct rt_addrinfo info; 1030 int error; 1031 1032 if (flags & RTF_HOST) { 1033 dst = ifa->ifa_dstaddr; 1034 netmask = NULL; 1035 } else { 1036 dst = ifa->ifa_addr; 1037 netmask = ifa->ifa_netmask; 1038 } 1039 /* 1040 * If it's a delete, check that if it exists, it's on the correct 1041 * interface or we might scrub a route to another ifa which would 1042 * be confusing at best and possibly worse. 1043 */ 1044 if (cmd == RTM_DELETE) { 1045 /* 1046 * It's a delete, so it should already exist.. 1047 * If it's a net, mask off the host bits 1048 * (Assuming we have a mask) 1049 */ 1050 if (netmask != NULL) { 1051 m = m_get(MB_DONTWAIT, MT_SONAME); 1052 if (m == NULL) 1053 return (ENOBUFS); 1054 deldst = mtod(m, struct sockaddr *); 1055 rt_maskedcopy(dst, deldst, netmask); 1056 dst = deldst; 1057 } 1058 /* 1059 * Look up an rtentry that is in the routing tree and 1060 * contains the correct info. 1061 */ 1062 if ((rnh = rt_tables[dst->sa_family]) == NULL || 1063 (rn = rnh->rnh_lookup((char *)dst, 1064 (char *)netmask, rnh)) == NULL || 1065 ((struct rtentry *)rn)->rt_ifa != ifa || 1066 !sa_equal((struct sockaddr *)rn->rn_key, dst)) { 1067 if (m != NULL) 1068 m_free(m); 1069 return (flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH); 1070 } 1071 /* XXX */ 1072 #if 0 1073 else { 1074 /* 1075 * One would think that as we are deleting, and we know 1076 * it doesn't exist, we could just return at this point 1077 * with an "ELSE" clause, but apparently not.. 1078 */ 1079 return (flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH); 1080 } 1081 #endif 1082 } 1083 /* 1084 * Do the actual request 1085 */ 1086 bzero(&info, sizeof info); 1087 info.rti_ifa = ifa; 1088 info.rti_flags = flags | ifa->ifa_flags; 1089 info.rti_info[RTAX_DST] = dst; 1090 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr; 1091 info.rti_info[RTAX_NETMASK] = netmask; 1092 error = rtrequest1(cmd, &info, &nrt); 1093 if (error == 0 && (rt = nrt) != NULL) { 1094 /* 1095 * notify any listening routing agents of the change 1096 */ 1097 rt_newaddrmsg(cmd, ifa, error, rt); 1098 if (cmd == RTM_DELETE) { 1099 /* 1100 * If we are deleting, and we found an entry, then 1101 * it's been removed from the tree.. now throw it away. 1102 */ 1103 if (rt->rt_refcnt == 0) { 1104 rt->rt_refcnt++; /* make a 1->0 transition */ 1105 rtfree(rt); 1106 } 1107 } else if (cmd == RTM_ADD) { 1108 /* 1109 * We just wanted to add it.. we don't actually 1110 * need a reference. 1111 */ 1112 rt->rt_refcnt--; 1113 } 1114 } 1115 if (m != NULL) 1116 m_free(m); 1117 return (error); 1118 } 1119 1120 /* This must be before ip6_init2(), which is now SI_ORDER_MIDDLE */ 1121 SYSINIT(route, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, route_init, 0); 1122