1 /* $NetBSD: route.c,v 1.43 2001/02/21 05:45:11 itojun 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. All advertising materials mentioning features or use of this software 82 * must display the following acknowledgement: 83 * This product includes software developed by the University of 84 * California, Berkeley and its contributors. 85 * 4. Neither the name of the University nor the names of its contributors 86 * may be used to endorse or promote products derived from this software 87 * without specific prior written permission. 88 * 89 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 90 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 91 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 92 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 93 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 94 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 95 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 96 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 97 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 98 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 99 * SUCH DAMAGE. 100 * 101 * @(#)route.c 8.3 (Berkeley) 1/9/95 102 */ 103 104 #include "opt_ns.h" 105 106 #include <sys/param.h> 107 #include <sys/systm.h> 108 #include <sys/callout.h> 109 #include <sys/proc.h> 110 #include <sys/mbuf.h> 111 #include <sys/socket.h> 112 #include <sys/socketvar.h> 113 #include <sys/domain.h> 114 #include <sys/protosw.h> 115 #include <sys/kernel.h> 116 #include <sys/ioctl.h> 117 #include <sys/pool.h> 118 119 #include <net/if.h> 120 #include <net/route.h> 121 #include <net/raw_cb.h> 122 123 #include <netinet/in.h> 124 #include <netinet/in_var.h> 125 126 #ifdef NS 127 #include <netns/ns.h> 128 #endif 129 130 #define SA(p) ((struct sockaddr *)(p)) 131 132 int rttrash; /* routes not in table but not freed */ 133 struct sockaddr wildcard; /* zero valued cookie for wildcard searches */ 134 135 struct pool rtentry_pool; /* pool for rtentry structures */ 136 struct pool rttimer_pool; /* pool for rttimer structures */ 137 138 struct callout rt_timer_ch; /* callout for rt_timer_timer() */ 139 140 static int rtdeletemsg __P((struct rtentry *)); 141 static int rtflushclone1 __P((struct radix_node *, void *)); 142 static void rtflushclone __P((struct radix_node_head *, struct rtentry *)); 143 144 void 145 rtable_init(table) 146 void **table; 147 { 148 struct domain *dom; 149 for (dom = domains; dom; dom = dom->dom_next) 150 if (dom->dom_rtattach) 151 dom->dom_rtattach(&table[dom->dom_family], 152 dom->dom_rtoffset); 153 } 154 155 void 156 route_init() 157 { 158 159 pool_init(&rtentry_pool, sizeof(struct rtentry), 0, 0, 0, "rtentpl", 160 0, NULL, NULL, M_RTABLE); 161 162 rn_init(); /* initialize all zeroes, all ones, mask table */ 163 rtable_init((void **)rt_tables); 164 } 165 166 /* 167 * Packet routing routines. 168 */ 169 void 170 rtalloc(ro) 171 struct route *ro; 172 { 173 if (ro->ro_rt && ro->ro_rt->rt_ifp && (ro->ro_rt->rt_flags & RTF_UP)) 174 return; /* XXX */ 175 ro->ro_rt = rtalloc1(&ro->ro_dst, 1); 176 } 177 178 struct rtentry * 179 rtalloc1(dst, report) 180 struct sockaddr *dst; 181 int report; 182 { 183 struct radix_node_head *rnh = rt_tables[dst->sa_family]; 184 struct rtentry *rt; 185 struct radix_node *rn; 186 struct rtentry *newrt = 0; 187 struct rt_addrinfo info; 188 int s = splsoftnet(), err = 0, msgtype = RTM_MISS; 189 190 if (rnh && (rn = rnh->rnh_matchaddr((caddr_t)dst, rnh)) && 191 ((rn->rn_flags & RNF_ROOT) == 0)) { 192 newrt = rt = (struct rtentry *)rn; 193 if (report && (rt->rt_flags & RTF_CLONING)) { 194 err = rtrequest(RTM_RESOLVE, dst, SA(0), 195 SA(0), 0, &newrt); 196 if (err) { 197 newrt = rt; 198 rt->rt_refcnt++; 199 goto miss; 200 } 201 if ((rt = newrt) && (rt->rt_flags & RTF_XRESOLVE)) { 202 msgtype = RTM_RESOLVE; 203 goto miss; 204 } 205 /* Inform listeners of the new route */ 206 bzero(&info, sizeof(info)); 207 info.rti_info[RTAX_DST] = rt_key(rt); 208 info.rti_info[RTAX_NETMASK] = rt_mask(rt); 209 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 210 if (rt->rt_ifp != NULL) { 211 info.rti_info[RTAX_IFP] = 212 rt->rt_ifp->if_addrlist.tqh_first->ifa_addr; 213 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr; 214 } 215 rt_missmsg(RTM_ADD, &info, rt->rt_flags, 0); 216 } else 217 rt->rt_refcnt++; 218 } else { 219 rtstat.rts_unreach++; 220 miss: if (report) { 221 bzero((caddr_t)&info, sizeof(info)); 222 info.rti_info[RTAX_DST] = dst; 223 rt_missmsg(msgtype, &info, 0, err); 224 } 225 } 226 splx(s); 227 return (newrt); 228 } 229 230 void 231 rtfree(rt) 232 struct rtentry *rt; 233 { 234 struct ifaddr *ifa; 235 236 if (rt == 0) 237 panic("rtfree"); 238 rt->rt_refcnt--; 239 if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_UP) == 0) { 240 if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT)) 241 panic ("rtfree 2"); 242 rttrash--; 243 if (rt->rt_refcnt < 0) { 244 printf("rtfree: %p not freed (neg refs)\n", rt); 245 return; 246 } 247 rt_timer_remove_all(rt); 248 ifa = rt->rt_ifa; 249 IFAFREE(ifa); 250 Free(rt_key(rt)); 251 pool_put(&rtentry_pool, rt); 252 } 253 } 254 255 void 256 ifafree(ifa) 257 struct ifaddr *ifa; 258 { 259 260 #ifdef DIAGNOSTIC 261 if (ifa == NULL) 262 panic("ifafree: null ifa"); 263 if (ifa->ifa_refcnt != 0) 264 panic("ifafree: ifa_refcnt != 0 (%d)", ifa->ifa_refcnt); 265 #endif 266 #ifdef IFAREF_DEBUG 267 printf("ifafree: freeing ifaddr %p\n", ifa); 268 #endif 269 free(ifa, M_IFADDR); 270 } 271 272 /* 273 * Force a routing table entry to the specified 274 * destination to go through the given gateway. 275 * Normally called as a result of a routing redirect 276 * message from the network layer. 277 * 278 * N.B.: must be called at splsoftnet 279 */ 280 void 281 rtredirect(dst, gateway, netmask, flags, src, rtp) 282 struct sockaddr *dst, *gateway, *netmask, *src; 283 int flags; 284 struct rtentry **rtp; 285 { 286 struct rtentry *rt; 287 int error = 0; 288 u_quad_t *stat = 0; 289 struct rt_addrinfo info; 290 struct ifaddr *ifa; 291 292 /* verify the gateway is directly reachable */ 293 if ((ifa = ifa_ifwithnet(gateway)) == 0) { 294 error = ENETUNREACH; 295 goto out; 296 } 297 rt = rtalloc1(dst, 0); 298 /* 299 * If the redirect isn't from our current router for this dst, 300 * it's either old or wrong. If it redirects us to ourselves, 301 * we have a routing loop, perhaps as a result of an interface 302 * going down recently. 303 */ 304 #define equal(a1, a2) (bcmp((caddr_t)(a1), (caddr_t)(a2), (a1)->sa_len) == 0) 305 if (!(flags & RTF_DONE) && rt && 306 (!equal(src, rt->rt_gateway) || rt->rt_ifa != ifa)) 307 error = EINVAL; 308 else if (ifa_ifwithaddr(gateway)) 309 error = EHOSTUNREACH; 310 if (error) 311 goto done; 312 /* 313 * Create a new entry if we just got back a wildcard entry 314 * or the lookup failed. This is necessary for hosts 315 * which use routing redirects generated by smart gateways 316 * to dynamically build the routing tables. 317 */ 318 if ((rt == 0) || (rt_mask(rt) && rt_mask(rt)->sa_len < 2)) 319 goto create; 320 /* 321 * Don't listen to the redirect if it's 322 * for a route to an interface. 323 */ 324 if (rt->rt_flags & RTF_GATEWAY) { 325 if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) { 326 /* 327 * Changing from route to net => route to host. 328 * Create new route, rather than smashing route to net. 329 */ 330 create: 331 if (rt) 332 rtfree(rt); 333 flags |= RTF_GATEWAY | RTF_DYNAMIC; 334 info.rti_info[RTAX_DST] = dst; 335 info.rti_info[RTAX_GATEWAY] = gateway; 336 info.rti_info[RTAX_NETMASK] = netmask; 337 info.rti_ifa = ifa; 338 info.rti_flags = flags; 339 rt = NULL; 340 error = rtrequest1(RTM_ADD, &info, &rt); 341 if (rt != NULL) 342 flags = rt->rt_flags; 343 stat = &rtstat.rts_dynamic; 344 } else { 345 /* 346 * Smash the current notion of the gateway to 347 * this destination. Should check about netmask!!! 348 */ 349 rt->rt_flags |= RTF_MODIFIED; 350 flags |= RTF_MODIFIED; 351 stat = &rtstat.rts_newgateway; 352 rt_setgate(rt, rt_key(rt), gateway); 353 } 354 } else 355 error = EHOSTUNREACH; 356 done: 357 if (rt) { 358 if (rtp && !error) 359 *rtp = rt; 360 else 361 rtfree(rt); 362 } 363 out: 364 if (error) 365 rtstat.rts_badredirect++; 366 else if (stat != NULL) 367 (*stat)++; 368 bzero((caddr_t)&info, sizeof(info)); 369 info.rti_info[RTAX_DST] = dst; 370 info.rti_info[RTAX_GATEWAY] = gateway; 371 info.rti_info[RTAX_NETMASK] = netmask; 372 info.rti_info[RTAX_AUTHOR] = src; 373 rt_missmsg(RTM_REDIRECT, &info, flags, error); 374 } 375 376 /* 377 * Delete a route and generate a message 378 */ 379 static int 380 rtdeletemsg(rt) 381 struct rtentry *rt; 382 { 383 int error; 384 struct rt_addrinfo info; 385 386 /* 387 * Request the new route so that the entry is not actually 388 * deleted. That will allow the information being reported to 389 * be accurate (and consistent with route_output()). 390 */ 391 bzero((caddr_t)&info, sizeof(info)); 392 info.rti_info[RTAX_DST] = rt_key(rt); 393 info.rti_info[RTAX_NETMASK] = rt_mask(rt); 394 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 395 info.rti_flags = rt->rt_flags; 396 error = rtrequest1(RTM_DELETE, &info, &rt); 397 398 rt_missmsg(RTM_DELETE, &info, info.rti_flags, error); 399 400 /* Adjust the refcount */ 401 if (error == 0 && rt->rt_refcnt <= 0) { 402 rt->rt_refcnt++; 403 rtfree(rt); 404 } 405 return (error); 406 } 407 408 static int 409 rtflushclone1(rn, arg) 410 struct radix_node *rn; 411 void *arg; 412 { 413 struct rtentry *rt, *parent; 414 415 rt = (struct rtentry *)rn; 416 parent = (struct rtentry *)arg; 417 if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent == parent) 418 rtdeletemsg(rt); 419 return 0; 420 } 421 422 static void 423 rtflushclone(rnh, parent) 424 struct radix_node_head *rnh; 425 struct rtentry *parent; 426 { 427 428 #ifdef DIAGNOSTIC 429 if (!parent || (parent->rt_flags & RTF_CLONING) == 0) 430 panic("rtflushclone: called with a non-cloning route"); 431 if (!rnh->rnh_walktree) 432 panic("rtflushclone: no rnh_walktree"); 433 #endif 434 rnh->rnh_walktree(rnh, rtflushclone1, (void *)parent); 435 } 436 437 /* 438 * Routing table ioctl interface. 439 */ 440 int 441 rtioctl(req, data, p) 442 u_long req; 443 caddr_t data; 444 struct proc *p; 445 { 446 return (EOPNOTSUPP); 447 } 448 449 struct ifaddr * 450 ifa_ifwithroute(flags, dst, gateway) 451 int flags; 452 struct sockaddr *dst, *gateway; 453 { 454 struct ifaddr *ifa; 455 if ((flags & RTF_GATEWAY) == 0) { 456 /* 457 * If we are adding a route to an interface, 458 * and the interface is a pt to pt link 459 * we should search for the destination 460 * as our clue to the interface. Otherwise 461 * we can use the local address. 462 */ 463 ifa = 0; 464 if (flags & RTF_HOST) 465 ifa = ifa_ifwithdstaddr(dst); 466 if (ifa == 0) 467 ifa = ifa_ifwithaddr(gateway); 468 } else { 469 /* 470 * If we are adding a route to a remote net 471 * or host, the gateway may still be on the 472 * other end of a pt to pt link. 473 */ 474 ifa = ifa_ifwithdstaddr(gateway); 475 } 476 if (ifa == 0) 477 ifa = ifa_ifwithnet(gateway); 478 if (ifa == 0) { 479 struct rtentry *rt = rtalloc1(dst, 0); 480 if (rt == 0) 481 return (0); 482 rt->rt_refcnt--; 483 if ((ifa = rt->rt_ifa) == 0) 484 return (0); 485 } 486 if (ifa->ifa_addr->sa_family != dst->sa_family) { 487 struct ifaddr *oifa = ifa; 488 ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp); 489 if (ifa == 0) 490 ifa = oifa; 491 } 492 return (ifa); 493 } 494 495 #define ROUNDUP(a) (a>0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long)) 496 497 int 498 rtrequest(req, dst, gateway, netmask, flags, ret_nrt) 499 int req, flags; 500 struct sockaddr *dst, *gateway, *netmask; 501 struct rtentry **ret_nrt; 502 { 503 struct rt_addrinfo info; 504 505 bzero(&info, sizeof(info)); 506 info.rti_flags = flags; 507 info.rti_info[RTAX_DST] = dst; 508 info.rti_info[RTAX_GATEWAY] = gateway; 509 info.rti_info[RTAX_NETMASK] = netmask; 510 return rtrequest1(req, &info, ret_nrt); 511 } 512 513 /* 514 * These (questionable) definitions of apparent local variables apply 515 * to the next function. XXXXXX!!! 516 */ 517 #define dst info->rti_info[RTAX_DST] 518 #define gateway info->rti_info[RTAX_GATEWAY] 519 #define netmask info->rti_info[RTAX_NETMASK] 520 #define ifaaddr info->rti_info[RTAX_IFA] 521 #define ifpaddr info->rti_info[RTAX_IFP] 522 #define flags info->rti_flags 523 524 int 525 rt_getifa(info) 526 struct rt_addrinfo *info; 527 { 528 struct ifaddr *ifa; 529 int error = 0; 530 531 /* 532 * ifp may be specified by sockaddr_dl when protocol address 533 * is ambiguous 534 */ 535 if (info->rti_ifp == NULL && ifpaddr != NULL 536 && ifpaddr->sa_family == AF_LINK && 537 (ifa = ifa_ifwithnet((struct sockaddr *)ifpaddr)) != NULL) 538 info->rti_ifp = ifa->ifa_ifp; 539 if (info->rti_ifa == NULL && ifaaddr != NULL) 540 info->rti_ifa = ifa_ifwithaddr(ifaaddr); 541 if (info->rti_ifa == NULL) { 542 struct sockaddr *sa; 543 544 sa = ifaaddr != NULL ? ifaaddr : 545 (gateway != NULL ? gateway : dst); 546 if (sa != NULL && info->rti_ifp != NULL) 547 info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp); 548 else if (dst != NULL && gateway != NULL) 549 info->rti_ifa = ifa_ifwithroute(flags, dst, gateway); 550 else if (sa != NULL) 551 info->rti_ifa = ifa_ifwithroute(flags, sa, sa); 552 } 553 if ((ifa = info->rti_ifa) != NULL) { 554 if (info->rti_ifp == NULL) 555 info->rti_ifp = ifa->ifa_ifp; 556 } else 557 error = ENETUNREACH; 558 return (error); 559 } 560 561 int 562 rtrequest1(req, info, ret_nrt) 563 int req; 564 struct rt_addrinfo *info; 565 struct rtentry **ret_nrt; 566 { 567 int s = splsoftnet(); int error = 0; 568 struct rtentry *rt, *crt; 569 struct radix_node *rn; 570 struct radix_node_head *rnh; 571 struct ifaddr *ifa; 572 struct sockaddr *ndst; 573 #define senderr(x) { error = x ; goto bad; } 574 575 if ((rnh = rt_tables[dst->sa_family]) == 0) 576 senderr(ESRCH); 577 if (flags & RTF_HOST) 578 netmask = 0; 579 switch (req) { 580 case RTM_DELETE: 581 if ((rn = rnh->rnh_lookup(dst, netmask, rnh)) == 0) 582 senderr(ESRCH); 583 rt = (struct rtentry *)rn; 584 if ((rt->rt_flags & RTF_CLONING) != 0) { 585 /* clean up any cloned children */ 586 rtflushclone(rnh, rt); 587 } 588 if ((rn = rnh->rnh_deladdr(dst, netmask, rnh)) == 0) 589 senderr(ESRCH); 590 if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT)) 591 panic ("rtrequest delete"); 592 rt = (struct rtentry *)rn; 593 if (rt->rt_gwroute) { 594 rt = rt->rt_gwroute; RTFREE(rt); 595 (rt = (struct rtentry *)rn)->rt_gwroute = 0; 596 } 597 rt->rt_flags &= ~RTF_UP; 598 if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest) 599 ifa->ifa_rtrequest(RTM_DELETE, rt, info); 600 rttrash++; 601 if (ret_nrt) 602 *ret_nrt = rt; 603 else if (rt->rt_refcnt <= 0) { 604 rt->rt_refcnt++; 605 rtfree(rt); 606 } 607 break; 608 609 case RTM_RESOLVE: 610 if (ret_nrt == 0 || (rt = *ret_nrt) == 0) 611 senderr(EINVAL); 612 if ((rt->rt_flags & RTF_CLONING) == 0) 613 senderr(EINVAL); 614 ifa = rt->rt_ifa; 615 flags = rt->rt_flags & ~(RTF_CLONING | RTF_STATIC); 616 flags |= RTF_CLONED; 617 gateway = rt->rt_gateway; 618 if ((netmask = rt->rt_genmask) == 0) 619 flags |= RTF_HOST; 620 goto makeroute; 621 622 case RTM_ADD: 623 if (info->rti_ifa == 0 && (error = rt_getifa(info))) 624 senderr(error); 625 ifa = info->rti_ifa; 626 makeroute: 627 rt = pool_get(&rtentry_pool, PR_NOWAIT); 628 if (rt == 0) 629 senderr(ENOBUFS); 630 Bzero(rt, sizeof(*rt)); 631 rt->rt_flags = RTF_UP | flags; 632 LIST_INIT(&rt->rt_timer); 633 if (rt_setgate(rt, dst, gateway)) { 634 pool_put(&rtentry_pool, rt); 635 senderr(ENOBUFS); 636 } 637 ndst = rt_key(rt); 638 if (netmask) { 639 rt_maskedcopy(dst, ndst, netmask); 640 } else 641 Bcopy(dst, ndst, dst->sa_len); 642 IFAREF(ifa); 643 rt->rt_ifa = ifa; 644 rt->rt_ifp = ifa->ifa_ifp; 645 if (req == RTM_RESOLVE) { 646 rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */ 647 rt->rt_parent = *ret_nrt; 648 rt->rt_parent->rt_refcnt++; 649 } else if (rt->rt_rmx.rmx_mtu == 0 650 && !(rt->rt_rmx.rmx_locks & RTV_MTU)) { /* XXX */ 651 if (rt->rt_gwroute != NULL) { 652 rt->rt_rmx.rmx_mtu = rt->rt_gwroute->rt_rmx.rmx_mtu; 653 } else { 654 rt->rt_rmx.rmx_mtu = ifa->ifa_ifp->if_mtu; 655 } 656 } 657 rn = rnh->rnh_addaddr((caddr_t)ndst, (caddr_t)netmask, 658 rnh, rt->rt_nodes); 659 if (rn == NULL && (crt = rtalloc1(ndst, 0)) != NULL) { 660 /* overwrite cloned route */ 661 if ((crt->rt_flags & RTF_CLONED) != 0) { 662 rtdeletemsg(crt); 663 rn = rnh->rnh_addaddr((caddr_t)ndst, 664 (caddr_t)netmask, rnh, rt->rt_nodes); 665 } 666 RTFREE(crt); 667 } 668 if (rn == 0) { 669 IFAFREE(ifa); 670 if ((rt->rt_flags & RTF_CLONED) != 0 && rt->rt_parent) 671 rtfree(rt->rt_parent); 672 if (rt->rt_gwroute) 673 rtfree(rt->rt_gwroute); 674 Free(rt_key(rt)); 675 pool_put(&rtentry_pool, rt); 676 senderr(EEXIST); 677 } 678 if (ifa->ifa_rtrequest) 679 ifa->ifa_rtrequest(req, rt, info); 680 if (ret_nrt) { 681 *ret_nrt = rt; 682 rt->rt_refcnt++; 683 } 684 if ((rt->rt_flags & RTF_CLONING) != 0) { 685 /* clean up any cloned children */ 686 rtflushclone(rnh, rt); 687 } 688 break; 689 } 690 bad: 691 splx(s); 692 return (error); 693 } 694 695 #undef dst 696 #undef gateway 697 #undef netmask 698 #undef ifaaddr 699 #undef ifpaddr 700 #undef flags 701 702 int 703 rt_setgate(rt0, dst, gate) 704 struct rtentry *rt0; 705 struct sockaddr *dst, *gate; 706 { 707 caddr_t new, old; 708 int dlen = ROUNDUP(dst->sa_len), glen = ROUNDUP(gate->sa_len); 709 struct rtentry *rt = rt0; 710 711 if (rt->rt_gateway == 0 || glen > ROUNDUP(rt->rt_gateway->sa_len)) { 712 old = (caddr_t)rt_key(rt); 713 R_Malloc(new, caddr_t, dlen + glen); 714 if (new == 0) 715 return 1; 716 Bzero(new, dlen + glen); 717 rt->rt_nodes->rn_key = new; 718 } else { 719 new = rt->rt_nodes->rn_key; 720 old = 0; 721 } 722 Bcopy(gate, (rt->rt_gateway = (struct sockaddr *)(new + dlen)), glen); 723 if (old) { 724 Bcopy(dst, new, dlen); 725 Free(old); 726 } 727 if (rt->rt_gwroute) { 728 rt = rt->rt_gwroute; RTFREE(rt); 729 rt = rt0; rt->rt_gwroute = 0; 730 } 731 if (rt->rt_flags & RTF_GATEWAY) { 732 rt->rt_gwroute = rtalloc1(gate, 1); 733 /* 734 * If we switched gateways, grab the MTU from the new 735 * gateway route if the current MTU is 0 or greater 736 * than the MTU of gateway. 737 */ 738 if (rt->rt_gwroute 739 && !(rt->rt_rmx.rmx_locks & RTV_MTU) 740 && (rt->rt_rmx.rmx_mtu == 0 741 || rt->rt_rmx.rmx_mtu > rt->rt_gwroute->rt_rmx.rmx_mtu)) { /* XXX */ 742 rt->rt_rmx.rmx_mtu = rt->rt_gwroute->rt_rmx.rmx_mtu; 743 } 744 } 745 return 0; 746 } 747 748 void 749 rt_maskedcopy(src, dst, netmask) 750 struct sockaddr *src, *dst, *netmask; 751 { 752 u_char *cp1 = (u_char *)src; 753 u_char *cp2 = (u_char *)dst; 754 u_char *cp3 = (u_char *)netmask; 755 u_char *cplim = cp2 + *cp3; 756 u_char *cplim2 = cp2 + *cp1; 757 758 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */ 759 cp3 += 2; 760 if (cplim > cplim2) 761 cplim = cplim2; 762 while (cp2 < cplim) 763 *cp2++ = *cp1++ & *cp3++; 764 if (cp2 < cplim2) 765 bzero((caddr_t)cp2, (unsigned)(cplim2 - cp2)); 766 } 767 768 /* 769 * Set up or tear down a routing table entry, normally 770 * for an interface. 771 */ 772 int 773 rtinit(ifa, cmd, flags) 774 struct ifaddr *ifa; 775 int cmd, flags; 776 { 777 struct rtentry *rt; 778 struct sockaddr *dst, *odst; 779 struct sockaddr_storage deldst; 780 struct rtentry *nrt = 0; 781 int error; 782 struct rt_addrinfo info; 783 784 dst = flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr; 785 if (cmd == RTM_DELETE) { 786 if ((flags & RTF_HOST) == 0 && ifa->ifa_netmask) { 787 /* Delete subnet route for this interface */ 788 odst = dst; 789 dst = (struct sockaddr *)&deldst; 790 rt_maskedcopy(odst, dst, ifa->ifa_netmask); 791 } 792 if ((rt = rtalloc1(dst, 0)) != NULL) { 793 rt->rt_refcnt--; 794 if (rt->rt_ifa != ifa) 795 return (flags & RTF_HOST ? EHOSTUNREACH 796 : ENETUNREACH); 797 } 798 } 799 bzero(&info, sizeof(info)); 800 info.rti_ifa = ifa; 801 info.rti_flags = flags | ifa->ifa_flags; 802 info.rti_info[RTAX_DST] = dst; 803 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr; 804 /* 805 * XXX here, it seems that we are assuming that ifa_netmask is NULL 806 * for RTF_HOST. bsdi4 passes NULL explicitly (via intermediate 807 * variable) when RTF_HOST is 1. still not sure if i can safely 808 * change it to meet bsdi4 behavior. 809 */ 810 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask; 811 error = rtrequest1(cmd, &info, &nrt); 812 if (cmd == RTM_DELETE && error == 0 && (rt = nrt)) { 813 rt_newaddrmsg(cmd, ifa, error, nrt); 814 if (rt->rt_refcnt <= 0) { 815 rt->rt_refcnt++; 816 rtfree(rt); 817 } 818 } 819 if (cmd == RTM_ADD && error == 0 && (rt = nrt)) { 820 rt->rt_refcnt--; 821 if (rt->rt_ifa != ifa) { 822 printf("rtinit: wrong ifa (%p) was (%p)\n", ifa, 823 rt->rt_ifa); 824 if (rt->rt_ifa->ifa_rtrequest) 825 rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt, NULL); 826 IFAFREE(rt->rt_ifa); 827 rt->rt_ifa = ifa; 828 rt->rt_ifp = ifa->ifa_ifp; 829 rt->rt_rmx.rmx_mtu = ifa->ifa_ifp->if_mtu; /*XXX*/ 830 IFAREF(ifa); 831 if (ifa->ifa_rtrequest) 832 ifa->ifa_rtrequest(RTM_ADD, rt, NULL); 833 } 834 rt_newaddrmsg(cmd, ifa, error, nrt); 835 } 836 return (error); 837 } 838 839 /* 840 * Route timer routines. These routes allow functions to be called 841 * for various routes at any time. This is useful in supporting 842 * path MTU discovery and redirect route deletion. 843 * 844 * This is similar to some BSDI internal functions, but it provides 845 * for multiple queues for efficiency's sake... 846 */ 847 848 LIST_HEAD(, rttimer_queue) rttimer_queue_head; 849 static int rt_init_done = 0; 850 851 #define RTTIMER_CALLOUT(r) { \ 852 if (r->rtt_func != NULL) { \ 853 (*r->rtt_func)(r->rtt_rt, r); \ 854 } else { \ 855 rtrequest((int) RTM_DELETE, \ 856 (struct sockaddr *)rt_key(r->rtt_rt), \ 857 0, 0, 0, 0); \ 858 } \ 859 } 860 861 /* 862 * Some subtle order problems with domain initialization mean that 863 * we cannot count on this being run from rt_init before various 864 * protocol initializations are done. Therefore, we make sure 865 * that this is run when the first queue is added... 866 */ 867 868 void 869 rt_timer_init() 870 { 871 assert(rt_init_done == 0); 872 873 pool_init(&rttimer_pool, sizeof(struct rttimer), 0, 0, 0, "rttmrpl", 874 0, NULL, NULL, M_RTABLE); 875 876 LIST_INIT(&rttimer_queue_head); 877 callout_init(&rt_timer_ch); 878 callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL); 879 rt_init_done = 1; 880 } 881 882 struct rttimer_queue * 883 rt_timer_queue_create(timeout) 884 u_int timeout; 885 { 886 struct rttimer_queue *rtq; 887 888 if (rt_init_done == 0) 889 rt_timer_init(); 890 891 R_Malloc(rtq, struct rttimer_queue *, sizeof *rtq); 892 if (rtq == NULL) 893 return (NULL); 894 Bzero(rtq, sizeof *rtq); 895 896 rtq->rtq_timeout = timeout; 897 rtq->rtq_count = 0; 898 TAILQ_INIT(&rtq->rtq_head); 899 LIST_INSERT_HEAD(&rttimer_queue_head, rtq, rtq_link); 900 901 return (rtq); 902 } 903 904 void 905 rt_timer_queue_change(rtq, timeout) 906 struct rttimer_queue *rtq; 907 long timeout; 908 { 909 910 rtq->rtq_timeout = timeout; 911 } 912 913 void 914 rt_timer_queue_destroy(rtq, destroy) 915 struct rttimer_queue *rtq; 916 int destroy; 917 { 918 struct rttimer *r; 919 920 while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL) { 921 LIST_REMOVE(r, rtt_link); 922 TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next); 923 if (destroy) 924 RTTIMER_CALLOUT(r); 925 pool_put(&rttimer_pool, r); 926 if (rtq->rtq_count > 0) 927 rtq->rtq_count--; 928 else 929 printf("rt_timer_queue_destroy: rtq_count reached 0\n"); 930 } 931 932 LIST_REMOVE(rtq, rtq_link); 933 934 /* 935 * Caller is responsible for freeing the rttimer_queue structure. 936 */ 937 } 938 939 unsigned long 940 rt_timer_count(rtq) 941 struct rttimer_queue *rtq; 942 { 943 944 return rtq->rtq_count; 945 } 946 947 void 948 rt_timer_remove_all(rt) 949 struct rtentry *rt; 950 { 951 struct rttimer *r; 952 953 while ((r = LIST_FIRST(&rt->rt_timer)) != NULL) { 954 LIST_REMOVE(r, rtt_link); 955 TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next); 956 if (r->rtt_queue->rtq_count > 0) 957 r->rtt_queue->rtq_count--; 958 else 959 printf("rt_timer_remove_all: rtq_count reached 0\n"); 960 pool_put(&rttimer_pool, r); 961 } 962 } 963 964 int 965 rt_timer_add(rt, func, queue) 966 struct rtentry *rt; 967 void(*func) __P((struct rtentry *, struct rttimer *)); 968 struct rttimer_queue *queue; 969 { 970 struct rttimer *r; 971 long current_time; 972 int s; 973 974 s = splclock(); 975 current_time = mono_time.tv_sec; 976 splx(s); 977 978 /* 979 * If there's already a timer with this action, destroy it before 980 * we add a new one. 981 */ 982 for (r = LIST_FIRST(&rt->rt_timer); r != NULL; 983 r = LIST_NEXT(r, rtt_link)) { 984 if (r->rtt_func == func) { 985 LIST_REMOVE(r, rtt_link); 986 TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next); 987 if (r->rtt_queue->rtq_count > 0) 988 r->rtt_queue->rtq_count--; 989 else 990 printf("rt_timer_add: rtq_count reached 0\n"); 991 pool_put(&rttimer_pool, r); 992 break; /* only one per list, so we can quit... */ 993 } 994 } 995 996 r = pool_get(&rttimer_pool, PR_NOWAIT); 997 if (r == NULL) 998 return (ENOBUFS); 999 Bzero(r, sizeof(*r)); 1000 1001 r->rtt_rt = rt; 1002 r->rtt_time = current_time; 1003 r->rtt_func = func; 1004 r->rtt_queue = queue; 1005 LIST_INSERT_HEAD(&rt->rt_timer, r, rtt_link); 1006 TAILQ_INSERT_TAIL(&queue->rtq_head, r, rtt_next); 1007 r->rtt_queue->rtq_count++; 1008 1009 return (0); 1010 } 1011 1012 /* ARGSUSED */ 1013 void 1014 rt_timer_timer(arg) 1015 void *arg; 1016 { 1017 struct rttimer_queue *rtq; 1018 struct rttimer *r; 1019 long current_time; 1020 int s; 1021 1022 s = splclock(); 1023 current_time = mono_time.tv_sec; 1024 splx(s); 1025 1026 s = splsoftnet(); 1027 for (rtq = LIST_FIRST(&rttimer_queue_head); rtq != NULL; 1028 rtq = LIST_NEXT(rtq, rtq_link)) { 1029 while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL && 1030 (r->rtt_time + rtq->rtq_timeout) < current_time) { 1031 LIST_REMOVE(r, rtt_link); 1032 TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next); 1033 RTTIMER_CALLOUT(r); 1034 pool_put(&rttimer_pool, r); 1035 if (rtq->rtq_count > 0) 1036 rtq->rtq_count--; 1037 else 1038 printf("rt_timer_timer: rtq_count reached 0\n"); 1039 } 1040 } 1041 splx(s); 1042 1043 callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL); 1044 } 1045