1 /* $NetBSD: route.c,v 1.67 2005/12/11 12:24:52 christos Exp $ */ 2 3 /*- 4 * Copyright (c) 1998 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Kevin M. Lahey of the Numerical Aerospace Simulation Facility, 9 * NASA Ames Research Center. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 3. All advertising materials mentioning features or use of this software 20 * must display the following acknowledgement: 21 * This product includes software developed by the NetBSD 22 * Foundation, Inc. and its contributors. 23 * 4. Neither the name of The NetBSD Foundation nor the names of its 24 * contributors may be used to endorse or promote products derived 25 * from this software without specific prior written permission. 26 * 27 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 28 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 29 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 30 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 31 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 32 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 33 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 34 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 35 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 36 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 37 * POSSIBILITY OF SUCH DAMAGE. 38 */ 39 40 /* 41 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 42 * All rights reserved. 43 * 44 * Redistribution and use in source and binary forms, with or without 45 * modification, are permitted provided that the following conditions 46 * are met: 47 * 1. Redistributions of source code must retain the above copyright 48 * notice, this list of conditions and the following disclaimer. 49 * 2. Redistributions in binary form must reproduce the above copyright 50 * notice, this list of conditions and the following disclaimer in the 51 * documentation and/or other materials provided with the distribution. 52 * 3. Neither the name of the project nor the names of its contributors 53 * may be used to endorse or promote products derived from this software 54 * without specific prior written permission. 55 * 56 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 57 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 58 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 59 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 60 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 61 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 62 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 63 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 64 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 65 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 66 * SUCH DAMAGE. 67 */ 68 69 /* 70 * Copyright (c) 1980, 1986, 1991, 1993 71 * The Regents of the University of California. All rights reserved. 72 * 73 * Redistribution and use in source and binary forms, with or without 74 * modification, are permitted provided that the following conditions 75 * are met: 76 * 1. Redistributions of source code must retain the above copyright 77 * notice, this list of conditions and the following disclaimer. 78 * 2. Redistributions in binary form must reproduce the above copyright 79 * notice, this list of conditions and the following disclaimer in the 80 * documentation and/or other materials provided with the distribution. 81 * 3. Neither the name of the University nor the names of its contributors 82 * may be used to endorse or promote products derived from this software 83 * without specific prior written permission. 84 * 85 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 86 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 87 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 88 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 89 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 90 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 91 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 92 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 93 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 94 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 95 * SUCH DAMAGE. 96 * 97 * @(#)route.c 8.3 (Berkeley) 1/9/95 98 */ 99 100 #include <sys/cdefs.h> 101 __KERNEL_RCSID(0, "$NetBSD: route.c,v 1.67 2005/12/11 12:24:52 christos Exp $"); 102 103 #include "opt_ns.h" 104 105 #include <sys/param.h> 106 #include <sys/systm.h> 107 #include <sys/callout.h> 108 #include <sys/proc.h> 109 #include <sys/mbuf.h> 110 #include <sys/socket.h> 111 #include <sys/socketvar.h> 112 #include <sys/domain.h> 113 #include <sys/protosw.h> 114 #include <sys/kernel.h> 115 #include <sys/ioctl.h> 116 #include <sys/pool.h> 117 118 #include <net/if.h> 119 #include <net/route.h> 120 #include <net/raw_cb.h> 121 122 #include <netinet/in.h> 123 #include <netinet/in_var.h> 124 125 #ifdef NS 126 #include <netns/ns.h> 127 #endif 128 129 #define SA(p) ((struct sockaddr *)(p)) 130 131 struct route_cb route_cb; 132 struct rtstat rtstat; 133 struct radix_node_head *rt_tables[AF_MAX+1]; 134 135 int rttrash; /* routes not in table but not freed */ 136 struct sockaddr wildcard; /* zero valued cookie for wildcard searches */ 137 138 POOL_INIT(rtentry_pool, sizeof(struct rtentry), 0, 0, 0, "rtentpl", NULL); 139 POOL_INIT(rttimer_pool, sizeof(struct rttimer), 0, 0, 0, "rttmrpl", NULL); 140 141 struct callout rt_timer_ch; /* callout for rt_timer_timer() */ 142 143 static int rtdeletemsg(struct rtentry *); 144 static int rtflushclone1(struct radix_node *, void *); 145 static void rtflushclone(struct radix_node_head *, struct rtentry *); 146 147 void 148 rtable_init(void **table) 149 { 150 struct domain *dom; 151 DOMAIN_FOREACH(dom) 152 if (dom->dom_rtattach) 153 dom->dom_rtattach(&table[dom->dom_family], 154 dom->dom_rtoffset); 155 } 156 157 void 158 route_init(void) 159 { 160 161 rn_init(); /* initialize all zeroes, all ones, mask table */ 162 rtable_init((void **)rt_tables); 163 } 164 165 /* 166 * Packet routing routines. 167 */ 168 void 169 rtalloc(struct route *ro) 170 { 171 if (ro->ro_rt && ro->ro_rt->rt_ifp && (ro->ro_rt->rt_flags & RTF_UP)) 172 return; /* XXX */ 173 ro->ro_rt = rtalloc1(&ro->ro_dst, 1); 174 } 175 176 struct rtentry * 177 rtalloc1(const struct sockaddr *dst, int report) 178 { 179 struct radix_node_head *rnh = rt_tables[dst->sa_family]; 180 struct rtentry *rt; 181 struct radix_node *rn; 182 struct rtentry *newrt = 0; 183 struct rt_addrinfo info; 184 int s = splsoftnet(), err = 0, msgtype = RTM_MISS; 185 186 if (rnh && (rn = rnh->rnh_matchaddr(dst, rnh)) && 187 ((rn->rn_flags & RNF_ROOT) == 0)) { 188 newrt = rt = (struct rtentry *)rn; 189 if (report && (rt->rt_flags & RTF_CLONING)) { 190 err = rtrequest(RTM_RESOLVE, dst, SA(0), 191 SA(0), 0, &newrt); 192 if (err) { 193 newrt = rt; 194 rt->rt_refcnt++; 195 goto miss; 196 } 197 if ((rt = newrt) && (rt->rt_flags & RTF_XRESOLVE)) { 198 msgtype = RTM_RESOLVE; 199 goto miss; 200 } 201 /* Inform listeners of the new route */ 202 memset(&info, 0, sizeof(info)); 203 info.rti_info[RTAX_DST] = rt_key(rt); 204 info.rti_info[RTAX_NETMASK] = rt_mask(rt); 205 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 206 if (rt->rt_ifp != NULL) { 207 info.rti_info[RTAX_IFP] = 208 TAILQ_FIRST(&rt->rt_ifp->if_addrlist)->ifa_addr; 209 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr; 210 } 211 rt_missmsg(RTM_ADD, &info, rt->rt_flags, 0); 212 } else 213 rt->rt_refcnt++; 214 } else { 215 rtstat.rts_unreach++; 216 miss: if (report) { 217 memset((caddr_t)&info, 0, sizeof(info)); 218 info.rti_info[RTAX_DST] = dst; 219 rt_missmsg(msgtype, &info, 0, err); 220 } 221 } 222 splx(s); 223 return (newrt); 224 } 225 226 void 227 rtfree(struct rtentry *rt) 228 { 229 struct ifaddr *ifa; 230 231 if (rt == 0) 232 panic("rtfree"); 233 rt->rt_refcnt--; 234 if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_UP) == 0) { 235 if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT)) 236 panic ("rtfree 2"); 237 rttrash--; 238 if (rt->rt_refcnt < 0) { 239 printf("rtfree: %p not freed (neg refs)\n", rt); 240 return; 241 } 242 rt_timer_remove_all(rt, 0); 243 ifa = rt->rt_ifa; 244 IFAFREE(ifa); 245 Free(rt_key(rt)); 246 pool_put(&rtentry_pool, rt); 247 } 248 } 249 250 void 251 ifafree(struct ifaddr *ifa) 252 { 253 254 #ifdef DIAGNOSTIC 255 if (ifa == NULL) 256 panic("ifafree: null ifa"); 257 if (ifa->ifa_refcnt != 0) 258 panic("ifafree: ifa_refcnt != 0 (%d)", ifa->ifa_refcnt); 259 #endif 260 #ifdef IFAREF_DEBUG 261 printf("ifafree: freeing ifaddr %p\n", ifa); 262 #endif 263 free(ifa, M_IFADDR); 264 } 265 266 /* 267 * Force a routing table entry to the specified 268 * destination to go through the given gateway. 269 * Normally called as a result of a routing redirect 270 * message from the network layer. 271 * 272 * N.B.: must be called at splsoftnet 273 */ 274 void 275 rtredirect(const struct sockaddr *dst, const struct sockaddr *gateway, 276 const struct sockaddr *netmask, int flags, const struct sockaddr *src, 277 struct rtentry **rtp) 278 { 279 struct rtentry *rt; 280 int error = 0; 281 u_quad_t *stat = 0; 282 struct rt_addrinfo info; 283 struct ifaddr *ifa; 284 285 /* verify the gateway is directly reachable */ 286 if ((ifa = ifa_ifwithnet(gateway)) == 0) { 287 error = ENETUNREACH; 288 goto out; 289 } 290 rt = rtalloc1(dst, 0); 291 /* 292 * If the redirect isn't from our current router for this dst, 293 * it's either old or wrong. If it redirects us to ourselves, 294 * we have a routing loop, perhaps as a result of an interface 295 * going down recently. 296 */ 297 #define equal(a1, a2) \ 298 ((a1)->sa_len == (a2)->sa_len && \ 299 memcmp((a1), (a2), (a1)->sa_len) == 0) 300 if (!(flags & RTF_DONE) && rt && 301 (!equal(src, rt->rt_gateway) || rt->rt_ifa != ifa)) 302 error = EINVAL; 303 else if (ifa_ifwithaddr(gateway)) 304 error = EHOSTUNREACH; 305 if (error) 306 goto done; 307 /* 308 * Create a new entry if we just got back a wildcard entry 309 * or the lookup failed. This is necessary for hosts 310 * which use routing redirects generated by smart gateways 311 * to dynamically build the routing tables. 312 */ 313 if ((rt == 0) || (rt_mask(rt) && rt_mask(rt)->sa_len < 2)) 314 goto create; 315 /* 316 * Don't listen to the redirect if it's 317 * for a route to an interface. 318 */ 319 if (rt->rt_flags & RTF_GATEWAY) { 320 if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) { 321 /* 322 * Changing from route to net => route to host. 323 * Create new route, rather than smashing route to net. 324 */ 325 create: 326 if (rt) 327 rtfree(rt); 328 flags |= RTF_GATEWAY | RTF_DYNAMIC; 329 info.rti_info[RTAX_DST] = dst; 330 info.rti_info[RTAX_GATEWAY] = gateway; 331 info.rti_info[RTAX_NETMASK] = netmask; 332 info.rti_ifa = ifa; 333 info.rti_flags = flags; 334 rt = NULL; 335 error = rtrequest1(RTM_ADD, &info, &rt); 336 if (rt != NULL) 337 flags = rt->rt_flags; 338 stat = &rtstat.rts_dynamic; 339 } else { 340 /* 341 * Smash the current notion of the gateway to 342 * this destination. Should check about netmask!!! 343 */ 344 rt->rt_flags |= RTF_MODIFIED; 345 flags |= RTF_MODIFIED; 346 stat = &rtstat.rts_newgateway; 347 rt_setgate(rt, rt_key(rt), gateway); 348 } 349 } else 350 error = EHOSTUNREACH; 351 done: 352 if (rt) { 353 if (rtp && !error) 354 *rtp = rt; 355 else 356 rtfree(rt); 357 } 358 out: 359 if (error) 360 rtstat.rts_badredirect++; 361 else if (stat != NULL) 362 (*stat)++; 363 memset((caddr_t)&info, 0, sizeof(info)); 364 info.rti_info[RTAX_DST] = dst; 365 info.rti_info[RTAX_GATEWAY] = gateway; 366 info.rti_info[RTAX_NETMASK] = netmask; 367 info.rti_info[RTAX_AUTHOR] = src; 368 rt_missmsg(RTM_REDIRECT, &info, flags, error); 369 } 370 371 /* 372 * Delete a route and generate a message 373 */ 374 static int 375 rtdeletemsg(struct rtentry *rt) 376 { 377 int error; 378 struct rt_addrinfo info; 379 380 /* 381 * Request the new route so that the entry is not actually 382 * deleted. That will allow the information being reported to 383 * be accurate (and consistent with route_output()). 384 */ 385 memset((caddr_t)&info, 0, sizeof(info)); 386 info.rti_info[RTAX_DST] = rt_key(rt); 387 info.rti_info[RTAX_NETMASK] = rt_mask(rt); 388 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 389 info.rti_flags = rt->rt_flags; 390 error = rtrequest1(RTM_DELETE, &info, &rt); 391 392 rt_missmsg(RTM_DELETE, &info, info.rti_flags, error); 393 394 /* Adjust the refcount */ 395 if (error == 0 && rt->rt_refcnt <= 0) { 396 rt->rt_refcnt++; 397 rtfree(rt); 398 } 399 return (error); 400 } 401 402 static int 403 rtflushclone1(struct radix_node *rn, void *arg) 404 { 405 struct rtentry *rt, *parent; 406 407 rt = (struct rtentry *)rn; 408 parent = (struct rtentry *)arg; 409 if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent == parent) 410 rtdeletemsg(rt); 411 return 0; 412 } 413 414 static void 415 rtflushclone(struct radix_node_head *rnh, struct rtentry *parent) 416 { 417 418 #ifdef DIAGNOSTIC 419 if (!parent || (parent->rt_flags & RTF_CLONING) == 0) 420 panic("rtflushclone: called with a non-cloning route"); 421 if (!rnh->rnh_walktree) 422 panic("rtflushclone: no rnh_walktree"); 423 #endif 424 rnh->rnh_walktree(rnh, rtflushclone1, (void *)parent); 425 } 426 427 /* 428 * Routing table ioctl interface. 429 */ 430 int 431 rtioctl(u_long req, caddr_t data, struct lwp *l) 432 { 433 return (EOPNOTSUPP); 434 } 435 436 struct ifaddr * 437 ifa_ifwithroute(int flags, const struct sockaddr *dst, 438 const struct sockaddr *gateway) 439 { 440 struct ifaddr *ifa; 441 if ((flags & RTF_GATEWAY) == 0) { 442 /* 443 * If we are adding a route to an interface, 444 * and the interface is a pt to pt link 445 * we should search for the destination 446 * as our clue to the interface. Otherwise 447 * we can use the local address. 448 */ 449 ifa = 0; 450 if (flags & RTF_HOST) 451 ifa = ifa_ifwithdstaddr(dst); 452 if (ifa == 0) 453 ifa = ifa_ifwithaddr(gateway); 454 } else { 455 /* 456 * If we are adding a route to a remote net 457 * or host, the gateway may still be on the 458 * other end of a pt to pt link. 459 */ 460 ifa = ifa_ifwithdstaddr(gateway); 461 } 462 if (ifa == 0) 463 ifa = ifa_ifwithnet(gateway); 464 if (ifa == 0) { 465 struct rtentry *rt = rtalloc1(dst, 0); 466 if (rt == 0) 467 return (0); 468 rt->rt_refcnt--; 469 if ((ifa = rt->rt_ifa) == 0) 470 return (0); 471 } 472 if (ifa->ifa_addr->sa_family != dst->sa_family) { 473 struct ifaddr *oifa = ifa; 474 ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp); 475 if (ifa == 0) 476 ifa = oifa; 477 } 478 return (ifa); 479 } 480 481 #define ROUNDUP(a) (a>0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long)) 482 483 int 484 rtrequest(int req, const struct sockaddr *dst, const struct sockaddr *gateway, 485 const struct sockaddr *netmask, int flags, struct rtentry **ret_nrt) 486 { 487 struct rt_addrinfo info; 488 489 memset(&info, 0, sizeof(info)); 490 info.rti_flags = flags; 491 info.rti_info[RTAX_DST] = dst; 492 info.rti_info[RTAX_GATEWAY] = gateway; 493 info.rti_info[RTAX_NETMASK] = netmask; 494 return rtrequest1(req, &info, ret_nrt); 495 } 496 497 /* 498 * These (questionable) definitions of apparent local variables apply 499 * to the next function. XXXXXX!!! 500 */ 501 #define dst info->rti_info[RTAX_DST] 502 #define gateway info->rti_info[RTAX_GATEWAY] 503 #define netmask info->rti_info[RTAX_NETMASK] 504 #define ifaaddr info->rti_info[RTAX_IFA] 505 #define ifpaddr info->rti_info[RTAX_IFP] 506 #define flags info->rti_flags 507 508 int 509 rt_getifa(struct rt_addrinfo *info) 510 { 511 struct ifaddr *ifa; 512 int error = 0; 513 514 /* 515 * ifp may be specified by sockaddr_dl when protocol address 516 * is ambiguous 517 */ 518 if (info->rti_ifp == NULL && ifpaddr != NULL 519 && ifpaddr->sa_family == AF_LINK && 520 (ifa = ifa_ifwithnet((const struct sockaddr *)ifpaddr)) != NULL) 521 info->rti_ifp = ifa->ifa_ifp; 522 if (info->rti_ifa == NULL && ifaaddr != NULL) 523 info->rti_ifa = ifa_ifwithaddr(ifaaddr); 524 if (info->rti_ifa == NULL) { 525 const struct sockaddr *sa; 526 527 sa = ifaaddr != NULL ? ifaaddr : 528 (gateway != NULL ? gateway : dst); 529 if (sa != NULL && info->rti_ifp != NULL) 530 info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp); 531 else if (dst != NULL && gateway != NULL) 532 info->rti_ifa = ifa_ifwithroute(flags, dst, gateway); 533 else if (sa != NULL) 534 info->rti_ifa = ifa_ifwithroute(flags, sa, sa); 535 } 536 if ((ifa = info->rti_ifa) != NULL) { 537 if (info->rti_ifp == NULL) 538 info->rti_ifp = ifa->ifa_ifp; 539 } else 540 error = ENETUNREACH; 541 return (error); 542 } 543 544 int 545 rtrequest1(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt) 546 { 547 int s = splsoftnet(); 548 int error = 0; 549 struct rtentry *rt, *crt; 550 struct radix_node *rn; 551 struct radix_node_head *rnh; 552 struct ifaddr *ifa; 553 struct sockaddr *ndst; 554 struct sockaddr_storage deldst; 555 #define senderr(x) { error = x ; goto bad; } 556 557 if ((rnh = rt_tables[dst->sa_family]) == 0) 558 senderr(ESRCH); 559 if (flags & RTF_HOST) 560 netmask = 0; 561 switch (req) { 562 case RTM_DELETE: 563 if (netmask) { 564 rt_maskedcopy(dst, (struct sockaddr *)&deldst, netmask); 565 dst = (struct sockaddr *)&deldst; 566 } 567 if ((rn = rnh->rnh_lookup(dst, netmask, rnh)) == 0) 568 senderr(ESRCH); 569 rt = (struct rtentry *)rn; 570 if ((rt->rt_flags & RTF_CLONING) != 0) { 571 /* clean up any cloned children */ 572 rtflushclone(rnh, rt); 573 } 574 if ((rn = rnh->rnh_deladdr(dst, netmask, rnh)) == 0) 575 senderr(ESRCH); 576 if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT)) 577 panic ("rtrequest delete"); 578 rt = (struct rtentry *)rn; 579 if (rt->rt_gwroute) { 580 rt = rt->rt_gwroute; RTFREE(rt); 581 (rt = (struct rtentry *)rn)->rt_gwroute = 0; 582 } 583 if (rt->rt_parent) { 584 rt->rt_parent->rt_refcnt--; 585 rt->rt_parent = NULL; 586 } 587 rt->rt_flags &= ~RTF_UP; 588 if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest) 589 ifa->ifa_rtrequest(RTM_DELETE, rt, info); 590 rttrash++; 591 if (ret_nrt) 592 *ret_nrt = rt; 593 else if (rt->rt_refcnt <= 0) { 594 rt->rt_refcnt++; 595 rtfree(rt); 596 } 597 break; 598 599 case RTM_RESOLVE: 600 if (ret_nrt == 0 || (rt = *ret_nrt) == 0) 601 senderr(EINVAL); 602 if ((rt->rt_flags & RTF_CLONING) == 0) 603 senderr(EINVAL); 604 ifa = rt->rt_ifa; 605 flags = rt->rt_flags & ~(RTF_CLONING | RTF_STATIC); 606 flags |= RTF_CLONED; 607 gateway = rt->rt_gateway; 608 if ((netmask = rt->rt_genmask) == 0) 609 flags |= RTF_HOST; 610 goto makeroute; 611 612 case RTM_ADD: 613 if (info->rti_ifa == 0 && (error = rt_getifa(info))) 614 senderr(error); 615 ifa = info->rti_ifa; 616 makeroute: 617 rt = pool_get(&rtentry_pool, PR_NOWAIT); 618 if (rt == 0) 619 senderr(ENOBUFS); 620 Bzero(rt, sizeof(*rt)); 621 rt->rt_flags = RTF_UP | flags; 622 LIST_INIT(&rt->rt_timer); 623 if (rt_setgate(rt, dst, gateway)) { 624 pool_put(&rtentry_pool, rt); 625 senderr(ENOBUFS); 626 } 627 ndst = rt_key(rt); 628 if (netmask) { 629 rt_maskedcopy(dst, ndst, netmask); 630 } else 631 Bcopy(dst, ndst, dst->sa_len); 632 IFAREF(ifa); 633 rt->rt_ifa = ifa; 634 rt->rt_ifp = ifa->ifa_ifp; 635 if (req == RTM_RESOLVE) { 636 rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */ 637 rt->rt_parent = *ret_nrt; 638 rt->rt_parent->rt_refcnt++; 639 } 640 rn = rnh->rnh_addaddr(ndst, netmask, rnh, rt->rt_nodes); 641 if (rn == NULL && (crt = rtalloc1(ndst, 0)) != NULL) { 642 /* overwrite cloned route */ 643 if ((crt->rt_flags & RTF_CLONED) != 0) { 644 rtdeletemsg(crt); 645 rn = rnh->rnh_addaddr(ndst, 646 netmask, rnh, rt->rt_nodes); 647 } 648 RTFREE(crt); 649 } 650 if (rn == 0) { 651 IFAFREE(ifa); 652 if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent) 653 rtfree(rt->rt_parent); 654 if (rt->rt_gwroute) 655 rtfree(rt->rt_gwroute); 656 Free(rt_key(rt)); 657 pool_put(&rtentry_pool, rt); 658 senderr(EEXIST); 659 } 660 if (ifa->ifa_rtrequest) 661 ifa->ifa_rtrequest(req, rt, info); 662 if (ret_nrt) { 663 *ret_nrt = rt; 664 rt->rt_refcnt++; 665 } 666 if ((rt->rt_flags & RTF_CLONING) != 0) { 667 /* clean up any cloned children */ 668 rtflushclone(rnh, rt); 669 } 670 break; 671 } 672 bad: 673 splx(s); 674 return (error); 675 } 676 677 #undef dst 678 #undef gateway 679 #undef netmask 680 #undef ifaaddr 681 #undef ifpaddr 682 #undef flags 683 684 int 685 rt_setgate( struct rtentry *rt0, const struct sockaddr *dst, 686 const struct sockaddr *gate) 687 { 688 char *new, *old; 689 u_int dlen = ROUNDUP(dst->sa_len), glen = ROUNDUP(gate->sa_len); 690 struct rtentry *rt = rt0; 691 692 if (rt->rt_gateway == 0 || glen > ROUNDUP(rt->rt_gateway->sa_len)) { 693 old = (caddr_t)rt_key(rt); 694 R_Malloc(new, caddr_t, dlen + glen); 695 if (new == 0) 696 return 1; 697 Bzero(new, dlen + glen); 698 rt->rt_nodes->rn_key = new; 699 } else { 700 new = __UNCONST(rt->rt_nodes->rn_key); /*XXXUNCONST*/ 701 old = 0; 702 } 703 Bcopy(gate, (rt->rt_gateway = (struct sockaddr *)(new + dlen)), glen); 704 if (old) { 705 Bcopy(dst, new, dlen); 706 Free(old); 707 } 708 if (rt->rt_gwroute) { 709 rt = rt->rt_gwroute; RTFREE(rt); 710 rt = rt0; rt->rt_gwroute = 0; 711 } 712 if (rt->rt_flags & RTF_GATEWAY) { 713 rt->rt_gwroute = rtalloc1(gate, 1); 714 /* 715 * If we switched gateways, grab the MTU from the new 716 * gateway route if the current MTU, if the current MTU is 717 * greater than the MTU of gateway. 718 * Note that, if the MTU of gateway is 0, we will reset the 719 * MTU of the route to run PMTUD again from scratch. XXX 720 */ 721 if (rt->rt_gwroute 722 && !(rt->rt_rmx.rmx_locks & RTV_MTU) 723 && rt->rt_rmx.rmx_mtu 724 && rt->rt_rmx.rmx_mtu > rt->rt_gwroute->rt_rmx.rmx_mtu) { 725 rt->rt_rmx.rmx_mtu = rt->rt_gwroute->rt_rmx.rmx_mtu; 726 } 727 } 728 return 0; 729 } 730 731 void 732 rt_maskedcopy(const struct sockaddr *src, struct sockaddr *dst, 733 const struct sockaddr *netmask) 734 { 735 const u_char *cp1 = (const u_char *)src; 736 u_char *cp2 = (u_char *)dst; 737 const u_char *cp3 = (const u_char *)netmask; 738 u_char *cplim = cp2 + *cp3; 739 u_char *cplim2 = cp2 + *cp1; 740 741 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */ 742 cp3 += 2; 743 if (cplim > cplim2) 744 cplim = cplim2; 745 while (cp2 < cplim) 746 *cp2++ = *cp1++ & *cp3++; 747 if (cp2 < cplim2) 748 memset(cp2, 0, (unsigned)(cplim2 - cp2)); 749 } 750 751 /* 752 * Set up or tear down a routing table entry, normally 753 * for an interface. 754 */ 755 int 756 rtinit(struct ifaddr *ifa, int cmd, int flags) 757 { 758 struct rtentry *rt; 759 struct sockaddr *dst, *odst; 760 struct sockaddr_storage deldst; 761 struct rtentry *nrt = 0; 762 int error; 763 struct rt_addrinfo info; 764 765 dst = flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr; 766 if (cmd == RTM_DELETE) { 767 if ((flags & RTF_HOST) == 0 && ifa->ifa_netmask) { 768 /* Delete subnet route for this interface */ 769 odst = dst; 770 dst = (struct sockaddr *)&deldst; 771 rt_maskedcopy(odst, dst, ifa->ifa_netmask); 772 } 773 if ((rt = rtalloc1(dst, 0)) != NULL) { 774 rt->rt_refcnt--; 775 if (rt->rt_ifa != ifa) 776 return (flags & RTF_HOST ? EHOSTUNREACH 777 : ENETUNREACH); 778 } 779 } 780 memset(&info, 0, sizeof(info)); 781 info.rti_ifa = ifa; 782 info.rti_flags = flags | ifa->ifa_flags; 783 info.rti_info[RTAX_DST] = dst; 784 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr; 785 /* 786 * XXX here, it seems that we are assuming that ifa_netmask is NULL 787 * for RTF_HOST. bsdi4 passes NULL explicitly (via intermediate 788 * variable) when RTF_HOST is 1. still not sure if i can safely 789 * change it to meet bsdi4 behavior. 790 */ 791 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask; 792 error = rtrequest1(cmd, &info, &nrt); 793 if (cmd == RTM_DELETE && error == 0 && (rt = nrt)) { 794 rt_newaddrmsg(cmd, ifa, error, nrt); 795 if (rt->rt_refcnt <= 0) { 796 rt->rt_refcnt++; 797 rtfree(rt); 798 } 799 } 800 if (cmd == RTM_ADD && error == 0 && (rt = nrt)) { 801 rt->rt_refcnt--; 802 if (rt->rt_ifa != ifa) { 803 printf("rtinit: wrong ifa (%p) was (%p)\n", ifa, 804 rt->rt_ifa); 805 if (rt->rt_ifa->ifa_rtrequest) 806 rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt, NULL); 807 IFAFREE(rt->rt_ifa); 808 rt->rt_ifa = ifa; 809 rt->rt_ifp = ifa->ifa_ifp; 810 IFAREF(ifa); 811 if (ifa->ifa_rtrequest) 812 ifa->ifa_rtrequest(RTM_ADD, rt, NULL); 813 } 814 rt_newaddrmsg(cmd, ifa, error, nrt); 815 } 816 return (error); 817 } 818 819 /* 820 * Route timer routines. These routes allow functions to be called 821 * for various routes at any time. This is useful in supporting 822 * path MTU discovery and redirect route deletion. 823 * 824 * This is similar to some BSDI internal functions, but it provides 825 * for multiple queues for efficiency's sake... 826 */ 827 828 LIST_HEAD(, rttimer_queue) rttimer_queue_head; 829 static int rt_init_done = 0; 830 831 #define RTTIMER_CALLOUT(r) do { \ 832 if (r->rtt_func != NULL) { \ 833 (*r->rtt_func)(r->rtt_rt, r); \ 834 } else { \ 835 rtrequest((int) RTM_DELETE, \ 836 (struct sockaddr *)rt_key(r->rtt_rt), \ 837 0, 0, 0, 0); \ 838 } \ 839 } while (/*CONSTCOND*/0) 840 841 /* 842 * Some subtle order problems with domain initialization mean that 843 * we cannot count on this being run from rt_init before various 844 * protocol initializations are done. Therefore, we make sure 845 * that this is run when the first queue is added... 846 */ 847 848 void 849 rt_timer_init(void) 850 { 851 assert(rt_init_done == 0); 852 853 LIST_INIT(&rttimer_queue_head); 854 callout_init(&rt_timer_ch); 855 callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL); 856 rt_init_done = 1; 857 } 858 859 struct rttimer_queue * 860 rt_timer_queue_create(u_int timeout) 861 { 862 struct rttimer_queue *rtq; 863 864 if (rt_init_done == 0) 865 rt_timer_init(); 866 867 R_Malloc(rtq, struct rttimer_queue *, sizeof *rtq); 868 if (rtq == NULL) 869 return (NULL); 870 Bzero(rtq, sizeof *rtq); 871 872 rtq->rtq_timeout = timeout; 873 rtq->rtq_count = 0; 874 TAILQ_INIT(&rtq->rtq_head); 875 LIST_INSERT_HEAD(&rttimer_queue_head, rtq, rtq_link); 876 877 return (rtq); 878 } 879 880 void 881 rt_timer_queue_change(struct rttimer_queue *rtq, long timeout) 882 { 883 884 rtq->rtq_timeout = timeout; 885 } 886 887 void 888 rt_timer_queue_remove_all(struct rttimer_queue *rtq, int destroy) 889 { 890 struct rttimer *r; 891 892 while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL) { 893 LIST_REMOVE(r, rtt_link); 894 TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next); 895 if (destroy) 896 RTTIMER_CALLOUT(r); 897 pool_put(&rttimer_pool, r); 898 if (rtq->rtq_count > 0) 899 rtq->rtq_count--; 900 else 901 printf("rt_timer_queue_remove_all: " 902 "rtq_count reached 0\n"); 903 } 904 } 905 906 void 907 rt_timer_queue_destroy(struct rttimer_queue *rtq, int destroy) 908 { 909 910 rt_timer_queue_remove_all(rtq, destroy); 911 912 LIST_REMOVE(rtq, rtq_link); 913 914 /* 915 * Caller is responsible for freeing the rttimer_queue structure. 916 */ 917 } 918 919 unsigned long 920 rt_timer_count(struct rttimer_queue *rtq) 921 { 922 return rtq->rtq_count; 923 } 924 925 void 926 rt_timer_remove_all(struct rtentry *rt, int destroy) 927 { 928 struct rttimer *r; 929 930 while ((r = LIST_FIRST(&rt->rt_timer)) != NULL) { 931 LIST_REMOVE(r, rtt_link); 932 TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next); 933 if (destroy) 934 RTTIMER_CALLOUT(r); 935 if (r->rtt_queue->rtq_count > 0) 936 r->rtt_queue->rtq_count--; 937 else 938 printf("rt_timer_remove_all: rtq_count reached 0\n"); 939 pool_put(&rttimer_pool, r); 940 } 941 } 942 943 int 944 rt_timer_add(struct rtentry *rt, 945 void (*func)(struct rtentry *, struct rttimer *), 946 struct rttimer_queue *queue) 947 { 948 struct rttimer *r; 949 long current_time; 950 int s; 951 952 s = splclock(); 953 current_time = mono_time.tv_sec; 954 splx(s); 955 956 /* 957 * If there's already a timer with this action, destroy it before 958 * we add a new one. 959 */ 960 for (r = LIST_FIRST(&rt->rt_timer); r != NULL; 961 r = LIST_NEXT(r, rtt_link)) { 962 if (r->rtt_func == func) { 963 LIST_REMOVE(r, rtt_link); 964 TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next); 965 if (r->rtt_queue->rtq_count > 0) 966 r->rtt_queue->rtq_count--; 967 else 968 printf("rt_timer_add: rtq_count reached 0\n"); 969 pool_put(&rttimer_pool, r); 970 break; /* only one per list, so we can quit... */ 971 } 972 } 973 974 r = pool_get(&rttimer_pool, PR_NOWAIT); 975 if (r == NULL) 976 return (ENOBUFS); 977 Bzero(r, sizeof(*r)); 978 979 r->rtt_rt = rt; 980 r->rtt_time = current_time; 981 r->rtt_func = func; 982 r->rtt_queue = queue; 983 LIST_INSERT_HEAD(&rt->rt_timer, r, rtt_link); 984 TAILQ_INSERT_TAIL(&queue->rtq_head, r, rtt_next); 985 r->rtt_queue->rtq_count++; 986 987 return (0); 988 } 989 990 /* ARGSUSED */ 991 void 992 rt_timer_timer(void *arg) 993 { 994 struct rttimer_queue *rtq; 995 struct rttimer *r; 996 long current_time; 997 int s; 998 999 s = splclock(); 1000 current_time = mono_time.tv_sec; 1001 splx(s); 1002 1003 s = splsoftnet(); 1004 for (rtq = LIST_FIRST(&rttimer_queue_head); rtq != NULL; 1005 rtq = LIST_NEXT(rtq, rtq_link)) { 1006 while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL && 1007 (r->rtt_time + rtq->rtq_timeout) < current_time) { 1008 LIST_REMOVE(r, rtt_link); 1009 TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next); 1010 RTTIMER_CALLOUT(r); 1011 pool_put(&rttimer_pool, r); 1012 if (rtq->rtq_count > 0) 1013 rtq->rtq_count--; 1014 else 1015 printf("rt_timer_timer: rtq_count reached 0\n"); 1016 } 1017 } 1018 splx(s); 1019 1020 callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL); 1021 } 1022