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) 1980, 1986, 1991, 1993 35 * The Regents of the University of California. All rights reserved. 36 * 37 * Redistribution and use in source and binary forms, with or without 38 * modification, are permitted provided that the following conditions 39 * are met: 40 * 1. Redistributions of source code must retain the above copyright 41 * notice, this list of conditions and the following disclaimer. 42 * 2. Redistributions in binary form must reproduce the above copyright 43 * notice, this list of conditions and the following disclaimer in the 44 * documentation and/or other materials provided with the distribution. 45 * 3. All advertising materials mentioning features or use of this software 46 * must display the following acknowledgement: 47 * This product includes software developed by the University of 48 * California, Berkeley and its contributors. 49 * 4. Neither the name of the University nor the names of its contributors 50 * may be used to endorse or promote products derived from this software 51 * without specific prior written permission. 52 * 53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 63 * SUCH DAMAGE. 64 * 65 * @(#)route.c 8.3 (Berkeley) 1/9/95 66 * $FreeBSD: src/sys/net/route.c,v 1.59.2.10 2003/01/17 08:04:00 ru Exp $ 67 * $DragonFly: src/sys/net/route.c,v 1.41 2008/11/09 10:50:15 sephe Exp $ 68 */ 69 70 #include "opt_inet.h" 71 #include "opt_mpls.h" 72 73 #include <sys/param.h> 74 #include <sys/systm.h> 75 #include <sys/malloc.h> 76 #include <sys/mbuf.h> 77 #include <sys/socket.h> 78 #include <sys/domain.h> 79 #include <sys/kernel.h> 80 #include <sys/sysctl.h> 81 #include <sys/globaldata.h> 82 #include <sys/thread.h> 83 84 #include <net/if.h> 85 #include <net/route.h> 86 #include <net/netisr.h> 87 88 #include <netinet/in.h> 89 #include <net/ip_mroute/ip_mroute.h> 90 91 #include <sys/thread2.h> 92 #include <sys/msgport2.h> 93 #include <net/netmsg2.h> 94 95 #ifdef MPLS 96 #include <netproto/mpls/mpls.h> 97 #endif 98 99 static struct rtstatistics rtstatistics_percpu[MAXCPU]; 100 #ifdef SMP 101 #define rtstat rtstatistics_percpu[mycpuid] 102 #else 103 #define rtstat rtstatistics_percpu[0] 104 #endif 105 106 struct radix_node_head *rt_tables[MAXCPU][AF_MAX+1]; 107 struct lwkt_port *rt_ports[MAXCPU]; 108 109 static void rt_maskedcopy (struct sockaddr *, struct sockaddr *, 110 struct sockaddr *); 111 static void rtable_init(void); 112 static void rtable_service_loop(void *dummy); 113 static void rtinit_rtrequest_callback(int, int, struct rt_addrinfo *, 114 struct rtentry *, void *); 115 116 #ifdef SMP 117 static void rtredirect_msghandler(struct netmsg *netmsg); 118 static void rtrequest1_msghandler(struct netmsg *netmsg); 119 #endif 120 121 static int rt_setshims(struct rtentry *, struct sockaddr **); 122 123 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RW, 0, "Routing"); 124 125 #ifdef ROUTE_DEBUG 126 static int route_debug = 1; 127 SYSCTL_INT(_net_route, OID_AUTO, route_debug, CTLFLAG_RW, 128 &route_debug, 0, ""); 129 #endif 130 131 int route_assert_owner_access = 0; 132 SYSCTL_INT(_net_route, OID_AUTO, assert_owner_access, CTLFLAG_RW, 133 &route_assert_owner_access, 0, ""); 134 SYSCTL_INT(_net_route, OID_AUTO, remote_free_panic, CTLFLAG_RW, 135 &route_assert_owner_access, 0, ""); /* alias */ 136 extern void db_print_backtrace(void); 137 138 /* 139 * Initialize the route table(s) for protocol domains and 140 * create a helper thread which will be responsible for updating 141 * route table entries on each cpu. 142 */ 143 void 144 route_init(void) 145 { 146 int cpu; 147 thread_t rtd; 148 149 for (cpu = 0; cpu < ncpus; ++cpu) 150 bzero(&rtstatistics_percpu[cpu], sizeof(struct rtstatistics)); 151 rn_init(); /* initialize all zeroes, all ones, mask table */ 152 rtable_init(); /* call dom_rtattach() on each cpu */ 153 154 for (cpu = 0; cpu < ncpus; cpu++) { 155 lwkt_create(rtable_service_loop, NULL, &rtd, NULL, 156 0, cpu, "rtable_cpu %d", cpu); 157 rt_ports[cpu] = &rtd->td_msgport; 158 } 159 } 160 161 static void 162 rtable_init_oncpu(struct netmsg *nmsg) 163 { 164 struct domain *dom; 165 int cpu = mycpuid; 166 167 SLIST_FOREACH(dom, &domains, dom_next) { 168 if (dom->dom_rtattach) { 169 dom->dom_rtattach( 170 (void **)&rt_tables[cpu][dom->dom_family], 171 dom->dom_rtoffset); 172 } 173 } 174 ifnet_forwardmsg(&nmsg->nm_lmsg, cpu + 1); 175 } 176 177 static void 178 rtable_init(void) 179 { 180 struct netmsg nmsg; 181 182 netmsg_init(&nmsg, &curthread->td_msgport, 0, rtable_init_oncpu); 183 ifnet_domsg(&nmsg.nm_lmsg, 0); 184 } 185 186 /* 187 * Our per-cpu table management protocol thread. All route table operations 188 * are sequentially chained through all cpus starting at cpu #0 in order to 189 * maintain duplicate route tables on each cpu. Having a spearate route 190 * table management thread allows the protocol and interrupt threads to 191 * issue route table changes. 192 */ 193 static void 194 rtable_service_loop(void *dummy __unused) 195 { 196 struct netmsg *netmsg; 197 thread_t td = curthread; 198 199 while ((netmsg = lwkt_waitport(&td->td_msgport, 0)) != NULL) { 200 netmsg->nm_dispatch(netmsg); 201 } 202 } 203 204 /* 205 * Routing statistics. 206 */ 207 #ifdef SMP 208 static int 209 sysctl_rtstatistics(SYSCTL_HANDLER_ARGS) 210 { 211 int cpu, error = 0; 212 213 for (cpu = 0; cpu < ncpus; ++cpu) { 214 if ((error = SYSCTL_OUT(req, &rtstatistics_percpu[cpu], 215 sizeof(struct rtstatistics)))) 216 break; 217 if ((error = SYSCTL_IN(req, &rtstatistics_percpu[cpu], 218 sizeof(struct rtstatistics)))) 219 break; 220 } 221 222 return (error); 223 } 224 SYSCTL_PROC(_net_route, OID_AUTO, stats, (CTLTYPE_OPAQUE|CTLFLAG_RW), 225 0, 0, sysctl_rtstatistics, "S,rtstatistics", "Routing statistics"); 226 #else 227 SYSCTL_STRUCT(_net_route, OID_AUTO, stats, CTLFLAG_RW, &rtstat, rtstatistics, 228 "Routing statistics"); 229 #endif 230 231 /* 232 * Packet routing routines. 233 */ 234 235 /* 236 * Look up and fill in the "ro_rt" rtentry field in a route structure given 237 * an address in the "ro_dst" field. Always send a report on a miss and 238 * always clone routes. 239 */ 240 void 241 rtalloc(struct route *ro) 242 { 243 rtalloc_ign(ro, 0UL); 244 } 245 246 /* 247 * Look up and fill in the "ro_rt" rtentry field in a route structure given 248 * an address in the "ro_dst" field. Always send a report on a miss and 249 * optionally clone routes when RTF_CLONING or RTF_PRCLONING are not being 250 * ignored. 251 */ 252 void 253 rtalloc_ign(struct route *ro, u_long ignoreflags) 254 { 255 if (ro->ro_rt != NULL) { 256 if (ro->ro_rt->rt_ifp != NULL && ro->ro_rt->rt_flags & RTF_UP) 257 return; 258 rtfree(ro->ro_rt); 259 ro->ro_rt = NULL; 260 } 261 ro->ro_rt = _rtlookup(&ro->ro_dst, RTL_REPORTMSG, ignoreflags); 262 } 263 264 /* 265 * Look up the route that matches the given "dst" address. 266 * 267 * Route lookup can have the side-effect of creating and returning 268 * a cloned route instead when "dst" matches a cloning route and the 269 * RTF_CLONING and RTF_PRCLONING flags are not being ignored. 270 * 271 * Any route returned has its reference count incremented. 272 */ 273 struct rtentry * 274 _rtlookup(struct sockaddr *dst, boolean_t generate_report, u_long ignore) 275 { 276 struct radix_node_head *rnh = rt_tables[mycpuid][dst->sa_family]; 277 struct rtentry *rt; 278 279 if (rnh == NULL) 280 goto unreach; 281 282 /* 283 * Look up route in the radix tree. 284 */ 285 rt = (struct rtentry *) rnh->rnh_matchaddr((char *)dst, rnh); 286 if (rt == NULL) 287 goto unreach; 288 289 /* 290 * Handle cloning routes. 291 */ 292 if ((rt->rt_flags & ~ignore & (RTF_CLONING | RTF_PRCLONING)) != 0) { 293 struct rtentry *clonedroute; 294 int error; 295 296 clonedroute = rt; /* copy in/copy out parameter */ 297 error = rtrequest(RTM_RESOLVE, dst, NULL, NULL, 0, 298 &clonedroute); /* clone the route */ 299 if (error != 0) { /* cloning failed */ 300 if (generate_report) 301 rt_dstmsg(RTM_MISS, dst, error); 302 rt->rt_refcnt++; 303 return (rt); /* return the uncloned route */ 304 } 305 if (generate_report) { 306 if (clonedroute->rt_flags & RTF_XRESOLVE) 307 rt_dstmsg(RTM_RESOLVE, dst, 0); 308 else 309 rt_rtmsg(RTM_ADD, clonedroute, 310 clonedroute->rt_ifp, 0); 311 } 312 return (clonedroute); /* return cloned route */ 313 } 314 315 /* 316 * Increment the reference count of the matched route and return. 317 */ 318 rt->rt_refcnt++; 319 return (rt); 320 321 unreach: 322 rtstat.rts_unreach++; 323 if (generate_report) 324 rt_dstmsg(RTM_MISS, dst, 0); 325 return (NULL); 326 } 327 328 void 329 rtfree(struct rtentry *rt) 330 { 331 if (rt->rt_cpuid == mycpuid) 332 rtfree_oncpu(rt); 333 else 334 rtfree_remote(rt, 1); 335 } 336 337 void 338 rtfree_oncpu(struct rtentry *rt) 339 { 340 KKASSERT(rt->rt_cpuid == mycpuid); 341 KASSERT(rt->rt_refcnt > 0, ("rtfree: rt_refcnt %ld", rt->rt_refcnt)); 342 343 --rt->rt_refcnt; 344 if (rt->rt_refcnt == 0) { 345 struct radix_node_head *rnh = 346 rt_tables[mycpuid][rt_key(rt)->sa_family]; 347 348 if (rnh->rnh_close) 349 rnh->rnh_close((struct radix_node *)rt, rnh); 350 if (!(rt->rt_flags & RTF_UP)) { 351 /* deallocate route */ 352 if (rt->rt_ifa != NULL) 353 IFAFREE(rt->rt_ifa); 354 if (rt->rt_parent != NULL) 355 RTFREE(rt->rt_parent); /* recursive call! */ 356 Free(rt_key(rt)); 357 Free(rt); 358 } 359 } 360 } 361 362 static void 363 rtfree_remote_dispatch(struct netmsg *nmsg) 364 { 365 struct lwkt_msg *lmsg = &nmsg->nm_lmsg; 366 struct rtentry *rt = lmsg->u.ms_resultp; 367 368 rtfree_oncpu(rt); 369 lwkt_replymsg(lmsg, 0); 370 } 371 372 void 373 rtfree_remote(struct rtentry *rt, int allow_panic) 374 { 375 struct netmsg nmsg; 376 struct lwkt_msg *lmsg; 377 378 KKASSERT(rt->rt_cpuid != mycpuid); 379 380 if (route_assert_owner_access && allow_panic) { 381 panic("rt remote free rt_cpuid %d, mycpuid %d\n", 382 rt->rt_cpuid, mycpuid); 383 } else { 384 kprintf("rt remote free rt_cpuid %d, mycpuid %d\n", 385 rt->rt_cpuid, mycpuid); 386 db_print_backtrace(); 387 } 388 389 netmsg_init(&nmsg, &curthread->td_msgport, 0, rtfree_remote_dispatch); 390 lmsg = &nmsg.nm_lmsg; 391 lmsg->u.ms_resultp = rt; 392 393 lwkt_domsg(rtable_portfn(rt->rt_cpuid), lmsg, 0); 394 } 395 396 static int 397 rtredirect_oncpu(struct sockaddr *dst, struct sockaddr *gateway, 398 struct sockaddr *netmask, int flags, struct sockaddr *src) 399 { 400 struct rtentry *rt = NULL; 401 struct rt_addrinfo rtinfo; 402 struct ifaddr *ifa; 403 u_long *stat = NULL; 404 int error; 405 406 /* verify the gateway is directly reachable */ 407 if ((ifa = ifa_ifwithnet(gateway)) == NULL) { 408 error = ENETUNREACH; 409 goto out; 410 } 411 412 /* 413 * If the redirect isn't from our current router for this destination, 414 * it's either old or wrong. 415 */ 416 if (!(flags & RTF_DONE) && /* XXX JH */ 417 (rt = rtpurelookup(dst)) != NULL && 418 (!sa_equal(src, rt->rt_gateway) || rt->rt_ifa != ifa)) { 419 error = EINVAL; 420 goto done; 421 } 422 423 /* 424 * If it redirects us to ourselves, we have a routing loop, 425 * perhaps as a result of an interface going down recently. 426 */ 427 if (ifa_ifwithaddr(gateway)) { 428 error = EHOSTUNREACH; 429 goto done; 430 } 431 432 /* 433 * Create a new entry if the lookup failed or if we got back 434 * a wildcard entry for the default route. This is necessary 435 * for hosts which use routing redirects generated by smart 436 * gateways to dynamically build the routing tables. 437 */ 438 if (rt == NULL) 439 goto create; 440 if ((rt_mask(rt) != NULL && rt_mask(rt)->sa_len < 2)) { 441 rtfree(rt); 442 goto create; 443 } 444 445 /* Ignore redirects for directly connected hosts. */ 446 if (!(rt->rt_flags & RTF_GATEWAY)) { 447 error = EHOSTUNREACH; 448 goto done; 449 } 450 451 if (!(rt->rt_flags & RTF_HOST) && (flags & RTF_HOST)) { 452 /* 453 * Changing from a network route to a host route. 454 * Create a new host route rather than smashing the 455 * network route. 456 */ 457 create: 458 flags |= RTF_GATEWAY | RTF_DYNAMIC; 459 bzero(&rtinfo, sizeof(struct rt_addrinfo)); 460 rtinfo.rti_info[RTAX_DST] = dst; 461 rtinfo.rti_info[RTAX_GATEWAY] = gateway; 462 rtinfo.rti_info[RTAX_NETMASK] = netmask; 463 rtinfo.rti_flags = flags; 464 rtinfo.rti_ifa = ifa; 465 rt = NULL; /* copy-in/copy-out parameter */ 466 error = rtrequest1(RTM_ADD, &rtinfo, &rt); 467 if (rt != NULL) 468 flags = rt->rt_flags; 469 stat = &rtstat.rts_dynamic; 470 } else { 471 /* 472 * Smash the current notion of the gateway to this destination. 473 * Should check about netmask!!! 474 */ 475 rt->rt_flags |= RTF_MODIFIED; 476 flags |= RTF_MODIFIED; 477 rt_setgate(rt, rt_key(rt), gateway); 478 error = 0; 479 stat = &rtstat.rts_newgateway; 480 } 481 482 done: 483 if (rt != NULL) 484 rtfree(rt); 485 out: 486 if (error != 0) 487 rtstat.rts_badredirect++; 488 else if (stat != NULL) 489 (*stat)++; 490 491 return error; 492 } 493 494 #ifdef SMP 495 496 struct netmsg_rtredirect { 497 struct netmsg netmsg; 498 struct sockaddr *dst; 499 struct sockaddr *gateway; 500 struct sockaddr *netmask; 501 int flags; 502 struct sockaddr *src; 503 }; 504 505 #endif 506 507 /* 508 * Force a routing table entry to the specified 509 * destination to go through the given gateway. 510 * Normally called as a result of a routing redirect 511 * message from the network layer. 512 * 513 * N.B.: must be called at splnet 514 */ 515 void 516 rtredirect(struct sockaddr *dst, struct sockaddr *gateway, 517 struct sockaddr *netmask, int flags, struct sockaddr *src) 518 { 519 struct rt_addrinfo rtinfo; 520 int error; 521 #ifdef SMP 522 struct netmsg_rtredirect msg; 523 524 netmsg_init(&msg.netmsg, &curthread->td_msgport, 0, 525 rtredirect_msghandler); 526 msg.dst = dst; 527 msg.gateway = gateway; 528 msg.netmask = netmask; 529 msg.flags = flags; 530 msg.src = src; 531 error = lwkt_domsg(rtable_portfn(0), &msg.netmsg.nm_lmsg, 0); 532 #else 533 error = rtredirect_oncpu(dst, gateway, netmask, flags, src); 534 #endif 535 bzero(&rtinfo, sizeof(struct rt_addrinfo)); 536 rtinfo.rti_info[RTAX_DST] = dst; 537 rtinfo.rti_info[RTAX_GATEWAY] = gateway; 538 rtinfo.rti_info[RTAX_NETMASK] = netmask; 539 rtinfo.rti_info[RTAX_AUTHOR] = src; 540 rt_missmsg(RTM_REDIRECT, &rtinfo, flags, error); 541 } 542 543 #ifdef SMP 544 545 static void 546 rtredirect_msghandler(struct netmsg *netmsg) 547 { 548 struct netmsg_rtredirect *msg = (void *)netmsg; 549 int nextcpu; 550 551 rtredirect_oncpu(msg->dst, msg->gateway, msg->netmask, 552 msg->flags, msg->src); 553 nextcpu = mycpuid + 1; 554 if (nextcpu < ncpus) 555 lwkt_forwardmsg(rtable_portfn(nextcpu), &netmsg->nm_lmsg); 556 else 557 lwkt_replymsg(&netmsg->nm_lmsg, 0); 558 } 559 560 #endif 561 562 /* 563 * Routing table ioctl interface. 564 */ 565 int 566 rtioctl(u_long req, caddr_t data, struct ucred *cred) 567 { 568 #ifdef INET 569 /* Multicast goop, grrr... */ 570 return mrt_ioctl ? mrt_ioctl(req, data) : EOPNOTSUPP; 571 #else 572 return ENXIO; 573 #endif 574 } 575 576 struct ifaddr * 577 ifa_ifwithroute(int flags, struct sockaddr *dst, struct sockaddr *gateway) 578 { 579 struct ifaddr *ifa; 580 581 if (!(flags & RTF_GATEWAY)) { 582 /* 583 * If we are adding a route to an interface, 584 * and the interface is a point-to-point link, 585 * we should search for the destination 586 * as our clue to the interface. Otherwise 587 * we can use the local address. 588 */ 589 ifa = NULL; 590 if (flags & RTF_HOST) { 591 ifa = ifa_ifwithdstaddr(dst); 592 } 593 if (ifa == NULL) 594 ifa = ifa_ifwithaddr(gateway); 595 } else { 596 /* 597 * If we are adding a route to a remote net 598 * or host, the gateway may still be on the 599 * other end of a pt to pt link. 600 */ 601 ifa = ifa_ifwithdstaddr(gateway); 602 } 603 if (ifa == NULL) 604 ifa = ifa_ifwithnet(gateway); 605 if (ifa == NULL) { 606 struct rtentry *rt; 607 608 rt = rtpurelookup(gateway); 609 if (rt == NULL) 610 return (NULL); 611 rt->rt_refcnt--; 612 if ((ifa = rt->rt_ifa) == NULL) 613 return (NULL); 614 } 615 if (ifa->ifa_addr->sa_family != dst->sa_family) { 616 struct ifaddr *oldifa = ifa; 617 618 ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp); 619 if (ifa == NULL) 620 ifa = oldifa; 621 } 622 return (ifa); 623 } 624 625 static int rt_fixdelete (struct radix_node *, void *); 626 static int rt_fixchange (struct radix_node *, void *); 627 628 struct rtfc_arg { 629 struct rtentry *rt0; 630 struct radix_node_head *rnh; 631 }; 632 633 /* 634 * Set rtinfo->rti_ifa and rtinfo->rti_ifp. 635 */ 636 int 637 rt_getifa(struct rt_addrinfo *rtinfo) 638 { 639 struct sockaddr *gateway = rtinfo->rti_info[RTAX_GATEWAY]; 640 struct sockaddr *dst = rtinfo->rti_info[RTAX_DST]; 641 struct sockaddr *ifaaddr = rtinfo->rti_info[RTAX_IFA]; 642 int flags = rtinfo->rti_flags; 643 644 /* 645 * ifp may be specified by sockaddr_dl 646 * when protocol address is ambiguous. 647 */ 648 if (rtinfo->rti_ifp == NULL) { 649 struct sockaddr *ifpaddr; 650 651 ifpaddr = rtinfo->rti_info[RTAX_IFP]; 652 if (ifpaddr != NULL && ifpaddr->sa_family == AF_LINK) { 653 struct ifaddr *ifa; 654 655 ifa = ifa_ifwithnet(ifpaddr); 656 if (ifa != NULL) 657 rtinfo->rti_ifp = ifa->ifa_ifp; 658 } 659 } 660 661 if (rtinfo->rti_ifa == NULL && ifaaddr != NULL) 662 rtinfo->rti_ifa = ifa_ifwithaddr(ifaaddr); 663 if (rtinfo->rti_ifa == NULL) { 664 struct sockaddr *sa; 665 666 sa = ifaaddr != NULL ? ifaaddr : 667 (gateway != NULL ? gateway : dst); 668 if (sa != NULL && rtinfo->rti_ifp != NULL) 669 rtinfo->rti_ifa = ifaof_ifpforaddr(sa, rtinfo->rti_ifp); 670 else if (dst != NULL && gateway != NULL) 671 rtinfo->rti_ifa = ifa_ifwithroute(flags, dst, gateway); 672 else if (sa != NULL) 673 rtinfo->rti_ifa = ifa_ifwithroute(flags, sa, sa); 674 } 675 if (rtinfo->rti_ifa == NULL) 676 return (ENETUNREACH); 677 678 if (rtinfo->rti_ifp == NULL) 679 rtinfo->rti_ifp = rtinfo->rti_ifa->ifa_ifp; 680 return (0); 681 } 682 683 /* 684 * Do appropriate manipulations of a routing tree given 685 * all the bits of info needed 686 */ 687 int 688 rtrequest( 689 int req, 690 struct sockaddr *dst, 691 struct sockaddr *gateway, 692 struct sockaddr *netmask, 693 int flags, 694 struct rtentry **ret_nrt) 695 { 696 struct rt_addrinfo rtinfo; 697 698 bzero(&rtinfo, sizeof(struct rt_addrinfo)); 699 rtinfo.rti_info[RTAX_DST] = dst; 700 rtinfo.rti_info[RTAX_GATEWAY] = gateway; 701 rtinfo.rti_info[RTAX_NETMASK] = netmask; 702 rtinfo.rti_flags = flags; 703 return rtrequest1(req, &rtinfo, ret_nrt); 704 } 705 706 int 707 rtrequest_global( 708 int req, 709 struct sockaddr *dst, 710 struct sockaddr *gateway, 711 struct sockaddr *netmask, 712 int flags) 713 { 714 struct rt_addrinfo rtinfo; 715 716 bzero(&rtinfo, sizeof(struct rt_addrinfo)); 717 rtinfo.rti_info[RTAX_DST] = dst; 718 rtinfo.rti_info[RTAX_GATEWAY] = gateway; 719 rtinfo.rti_info[RTAX_NETMASK] = netmask; 720 rtinfo.rti_flags = flags; 721 return rtrequest1_global(req, &rtinfo, NULL, NULL); 722 } 723 724 #ifdef SMP 725 726 struct netmsg_rtq { 727 struct netmsg netmsg; 728 int req; 729 struct rt_addrinfo *rtinfo; 730 rtrequest1_callback_func_t callback; 731 void *arg; 732 }; 733 734 #endif 735 736 int 737 rtrequest1_global(int req, struct rt_addrinfo *rtinfo, 738 rtrequest1_callback_func_t callback, void *arg) 739 { 740 int error; 741 #ifdef SMP 742 struct netmsg_rtq msg; 743 744 netmsg_init(&msg.netmsg, &curthread->td_msgport, 0, 745 rtrequest1_msghandler); 746 msg.netmsg.nm_lmsg.ms_error = -1; 747 msg.req = req; 748 msg.rtinfo = rtinfo; 749 msg.callback = callback; 750 msg.arg = arg; 751 error = lwkt_domsg(rtable_portfn(0), &msg.netmsg.nm_lmsg, 0); 752 #else 753 struct rtentry *rt = NULL; 754 755 error = rtrequest1(req, rtinfo, &rt); 756 if (rt) 757 --rt->rt_refcnt; 758 if (callback) 759 callback(req, error, rtinfo, rt, arg); 760 #endif 761 return (error); 762 } 763 764 /* 765 * Handle a route table request on the current cpu. Since the route table's 766 * are supposed to be identical on each cpu, an error occuring later in the 767 * message chain is considered system-fatal. 768 */ 769 #ifdef SMP 770 771 static void 772 rtrequest1_msghandler(struct netmsg *netmsg) 773 { 774 struct netmsg_rtq *msg = (void *)netmsg; 775 struct rtentry *rt = NULL; 776 int nextcpu; 777 int error; 778 779 error = rtrequest1(msg->req, msg->rtinfo, &rt); 780 if (rt) 781 --rt->rt_refcnt; 782 if (msg->callback) 783 msg->callback(msg->req, error, msg->rtinfo, rt, msg->arg); 784 785 /* 786 * RTM_DELETE's are propogated even if an error occurs, since a 787 * cloned route might be undergoing deletion and cloned routes 788 * are not necessarily replicated. An overall error is returned 789 * only if no cpus have the route in question. 790 */ 791 if (msg->netmsg.nm_lmsg.ms_error < 0 || error == 0) 792 msg->netmsg.nm_lmsg.ms_error = error; 793 794 nextcpu = mycpuid + 1; 795 if (error && msg->req != RTM_DELETE) { 796 if (mycpuid != 0) { 797 panic("rtrequest1_msghandler: rtrequest table " 798 "error was not on cpu #0: %p", msg->rtinfo); 799 } 800 lwkt_replymsg(&msg->netmsg.nm_lmsg, error); 801 } else if (nextcpu < ncpus) { 802 lwkt_forwardmsg(rtable_portfn(nextcpu), &msg->netmsg.nm_lmsg); 803 } else { 804 lwkt_replymsg(&msg->netmsg.nm_lmsg, 805 msg->netmsg.nm_lmsg.ms_error); 806 } 807 } 808 809 #endif 810 811 int 812 rtrequest1(int req, struct rt_addrinfo *rtinfo, struct rtentry **ret_nrt) 813 { 814 struct sockaddr *dst = rtinfo->rti_info[RTAX_DST]; 815 struct rtentry *rt; 816 struct radix_node *rn; 817 struct radix_node_head *rnh; 818 struct ifaddr *ifa; 819 struct sockaddr *ndst; 820 int error = 0; 821 822 #define gotoerr(x) { error = x ; goto bad; } 823 824 #ifdef ROUTE_DEBUG 825 if (route_debug) 826 rt_addrinfo_print(req, rtinfo); 827 #endif 828 829 crit_enter(); 830 /* 831 * Find the correct routing tree to use for this Address Family 832 */ 833 if ((rnh = rt_tables[mycpuid][dst->sa_family]) == NULL) 834 gotoerr(EAFNOSUPPORT); 835 836 /* 837 * If we are adding a host route then we don't want to put 838 * a netmask in the tree, nor do we want to clone it. 839 */ 840 if (rtinfo->rti_flags & RTF_HOST) { 841 rtinfo->rti_info[RTAX_NETMASK] = NULL; 842 rtinfo->rti_flags &= ~(RTF_CLONING | RTF_PRCLONING); 843 } 844 845 switch (req) { 846 case RTM_DELETE: 847 /* Remove the item from the tree. */ 848 rn = rnh->rnh_deladdr((char *)rtinfo->rti_info[RTAX_DST], 849 (char *)rtinfo->rti_info[RTAX_NETMASK], 850 rnh); 851 if (rn == NULL) 852 gotoerr(ESRCH); 853 KASSERT(!(rn->rn_flags & (RNF_ACTIVE | RNF_ROOT)), 854 ("rnh_deladdr returned flags 0x%x", rn->rn_flags)); 855 rt = (struct rtentry *)rn; 856 857 /* ref to prevent a deletion race */ 858 ++rt->rt_refcnt; 859 860 /* Free any routes cloned from this one. */ 861 if ((rt->rt_flags & (RTF_CLONING | RTF_PRCLONING)) && 862 rt_mask(rt) != NULL) { 863 rnh->rnh_walktree_from(rnh, (char *)rt_key(rt), 864 (char *)rt_mask(rt), 865 rt_fixdelete, rt); 866 } 867 868 if (rt->rt_gwroute != NULL) { 869 RTFREE(rt->rt_gwroute); 870 rt->rt_gwroute = NULL; 871 } 872 873 /* 874 * NB: RTF_UP must be set during the search above, 875 * because we might delete the last ref, causing 876 * rt to get freed prematurely. 877 */ 878 rt->rt_flags &= ~RTF_UP; 879 880 #ifdef ROUTE_DEBUG 881 if (route_debug) 882 rt_print(rtinfo, rt); 883 #endif 884 885 /* Give the protocol a chance to keep things in sync. */ 886 if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest) 887 ifa->ifa_rtrequest(RTM_DELETE, rt, rtinfo); 888 889 /* 890 * If the caller wants it, then it can have it, 891 * but it's up to it to free the rtentry as we won't be 892 * doing it. 893 */ 894 KASSERT(rt->rt_refcnt >= 0, 895 ("rtrequest1(DELETE): refcnt %ld", rt->rt_refcnt)); 896 if (ret_nrt != NULL) { 897 /* leave ref intact for return */ 898 *ret_nrt = rt; 899 } else { 900 /* deref / attempt to destroy */ 901 rtfree(rt); 902 } 903 break; 904 905 case RTM_RESOLVE: 906 if (ret_nrt == NULL || (rt = *ret_nrt) == NULL) 907 gotoerr(EINVAL); 908 ifa = rt->rt_ifa; 909 rtinfo->rti_flags = 910 rt->rt_flags & ~(RTF_CLONING | RTF_PRCLONING | RTF_STATIC); 911 rtinfo->rti_flags |= RTF_WASCLONED; 912 rtinfo->rti_info[RTAX_GATEWAY] = rt->rt_gateway; 913 if ((rtinfo->rti_info[RTAX_NETMASK] = rt->rt_genmask) == NULL) 914 rtinfo->rti_flags |= RTF_HOST; 915 rtinfo->rti_info[RTAX_MPLS1] = rt->rt_shim[0]; 916 rtinfo->rti_info[RTAX_MPLS2] = rt->rt_shim[1]; 917 rtinfo->rti_info[RTAX_MPLS3] = rt->rt_shim[2]; 918 goto makeroute; 919 920 case RTM_ADD: 921 KASSERT(!(rtinfo->rti_flags & RTF_GATEWAY) || 922 rtinfo->rti_info[RTAX_GATEWAY] != NULL, 923 ("rtrequest: GATEWAY but no gateway")); 924 925 if (rtinfo->rti_ifa == NULL && (error = rt_getifa(rtinfo))) 926 gotoerr(error); 927 ifa = rtinfo->rti_ifa; 928 makeroute: 929 R_Malloc(rt, struct rtentry *, sizeof(struct rtentry)); 930 if (rt == NULL) 931 gotoerr(ENOBUFS); 932 bzero(rt, sizeof(struct rtentry)); 933 rt->rt_flags = RTF_UP | rtinfo->rti_flags; 934 rt->rt_cpuid = mycpuid; 935 error = rt_setgate(rt, dst, rtinfo->rti_info[RTAX_GATEWAY]); 936 if (error != 0) { 937 Free(rt); 938 gotoerr(error); 939 } 940 941 ndst = rt_key(rt); 942 if (rtinfo->rti_info[RTAX_NETMASK] != NULL) 943 rt_maskedcopy(dst, ndst, 944 rtinfo->rti_info[RTAX_NETMASK]); 945 else 946 bcopy(dst, ndst, dst->sa_len); 947 948 if (rtinfo->rti_info[RTAX_MPLS1] != NULL) 949 rt_setshims(rt, rtinfo->rti_info); 950 951 /* 952 * Note that we now have a reference to the ifa. 953 * This moved from below so that rnh->rnh_addaddr() can 954 * examine the ifa and ifa->ifa_ifp if it so desires. 955 */ 956 IFAREF(ifa); 957 rt->rt_ifa = ifa; 958 rt->rt_ifp = ifa->ifa_ifp; 959 /* XXX mtu manipulation will be done in rnh_addaddr -- itojun */ 960 961 rn = rnh->rnh_addaddr((char *)ndst, 962 (char *)rtinfo->rti_info[RTAX_NETMASK], 963 rnh, rt->rt_nodes); 964 if (rn == NULL) { 965 struct rtentry *oldrt; 966 967 /* 968 * We already have one of these in the tree. 969 * We do a special hack: if the old route was 970 * cloned, then we blow it away and try 971 * re-inserting the new one. 972 */ 973 oldrt = rtpurelookup(ndst); 974 if (oldrt != NULL) { 975 --oldrt->rt_refcnt; 976 if (oldrt->rt_flags & RTF_WASCLONED) { 977 rtrequest(RTM_DELETE, rt_key(oldrt), 978 oldrt->rt_gateway, 979 rt_mask(oldrt), 980 oldrt->rt_flags, NULL); 981 rn = rnh->rnh_addaddr((char *)ndst, 982 (char *) 983 rtinfo->rti_info[RTAX_NETMASK], 984 rnh, rt->rt_nodes); 985 } 986 } 987 } 988 989 /* 990 * If it still failed to go into the tree, 991 * then un-make it (this should be a function). 992 */ 993 if (rn == NULL) { 994 if (rt->rt_gwroute != NULL) 995 rtfree(rt->rt_gwroute); 996 IFAFREE(ifa); 997 Free(rt_key(rt)); 998 Free(rt); 999 gotoerr(EEXIST); 1000 } 1001 1002 /* 1003 * If we got here from RESOLVE, then we are cloning 1004 * so clone the rest, and note that we 1005 * are a clone (and increment the parent's references) 1006 */ 1007 if (req == RTM_RESOLVE) { 1008 rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */ 1009 rt->rt_rmx.rmx_pksent = 0; /* reset packet counter */ 1010 if ((*ret_nrt)->rt_flags & 1011 (RTF_CLONING | RTF_PRCLONING)) { 1012 rt->rt_parent = *ret_nrt; 1013 (*ret_nrt)->rt_refcnt++; 1014 } 1015 } 1016 1017 /* 1018 * if this protocol has something to add to this then 1019 * allow it to do that as well. 1020 */ 1021 if (ifa->ifa_rtrequest != NULL) 1022 ifa->ifa_rtrequest(req, rt, rtinfo); 1023 1024 /* 1025 * We repeat the same procedure from rt_setgate() here because 1026 * it doesn't fire when we call it there because the node 1027 * hasn't been added to the tree yet. 1028 */ 1029 if (req == RTM_ADD && !(rt->rt_flags & RTF_HOST) && 1030 rt_mask(rt) != NULL) { 1031 struct rtfc_arg arg = { rt, rnh }; 1032 1033 rnh->rnh_walktree_from(rnh, (char *)rt_key(rt), 1034 (char *)rt_mask(rt), 1035 rt_fixchange, &arg); 1036 } 1037 1038 #ifdef ROUTE_DEBUG 1039 if (route_debug) 1040 rt_print(rtinfo, rt); 1041 #endif 1042 /* 1043 * Return the resulting rtentry, 1044 * increasing the number of references by one. 1045 */ 1046 if (ret_nrt != NULL) { 1047 rt->rt_refcnt++; 1048 *ret_nrt = rt; 1049 } 1050 break; 1051 default: 1052 error = EOPNOTSUPP; 1053 } 1054 bad: 1055 #ifdef ROUTE_DEBUG 1056 if (route_debug) { 1057 if (error) 1058 kprintf("rti %p failed error %d\n", rtinfo, error); 1059 else 1060 kprintf("rti %p succeeded\n", rtinfo); 1061 } 1062 #endif 1063 crit_exit(); 1064 return (error); 1065 } 1066 1067 /* 1068 * Called from rtrequest(RTM_DELETE, ...) to fix up the route's ``family'' 1069 * (i.e., the routes related to it by the operation of cloning). This 1070 * routine is iterated over all potential former-child-routes by way of 1071 * rnh->rnh_walktree_from() above, and those that actually are children of 1072 * the late parent (passed in as VP here) are themselves deleted. 1073 */ 1074 static int 1075 rt_fixdelete(struct radix_node *rn, void *vp) 1076 { 1077 struct rtentry *rt = (struct rtentry *)rn; 1078 struct rtentry *rt0 = vp; 1079 1080 if (rt->rt_parent == rt0 && 1081 !(rt->rt_flags & (RTF_PINNED | RTF_CLONING | RTF_PRCLONING))) { 1082 return rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 1083 rt->rt_flags, NULL); 1084 } 1085 return 0; 1086 } 1087 1088 /* 1089 * This routine is called from rt_setgate() to do the analogous thing for 1090 * adds and changes. There is the added complication in this case of a 1091 * middle insert; i.e., insertion of a new network route between an older 1092 * network route and (cloned) host routes. For this reason, a simple check 1093 * of rt->rt_parent is insufficient; each candidate route must be tested 1094 * against the (mask, value) of the new route (passed as before in vp) 1095 * to see if the new route matches it. 1096 * 1097 * XXX - it may be possible to do fixdelete() for changes and reserve this 1098 * routine just for adds. I'm not sure why I thought it was necessary to do 1099 * changes this way. 1100 */ 1101 #ifdef DEBUG 1102 static int rtfcdebug = 0; 1103 #endif 1104 1105 static int 1106 rt_fixchange(struct radix_node *rn, void *vp) 1107 { 1108 struct rtentry *rt = (struct rtentry *)rn; 1109 struct rtfc_arg *ap = vp; 1110 struct rtentry *rt0 = ap->rt0; 1111 struct radix_node_head *rnh = ap->rnh; 1112 u_char *xk1, *xm1, *xk2, *xmp; 1113 int i, len, mlen; 1114 1115 #ifdef DEBUG 1116 if (rtfcdebug) 1117 kprintf("rt_fixchange: rt %p, rt0 %p\n", rt, rt0); 1118 #endif 1119 1120 if (rt->rt_parent == NULL || 1121 (rt->rt_flags & (RTF_PINNED | RTF_CLONING | RTF_PRCLONING))) { 1122 #ifdef DEBUG 1123 if (rtfcdebug) kprintf("no parent, pinned or cloning\n"); 1124 #endif 1125 return 0; 1126 } 1127 1128 if (rt->rt_parent == rt0) { 1129 #ifdef DEBUG 1130 if (rtfcdebug) kprintf("parent match\n"); 1131 #endif 1132 return rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 1133 rt->rt_flags, NULL); 1134 } 1135 1136 /* 1137 * There probably is a function somewhere which does this... 1138 * if not, there should be. 1139 */ 1140 len = imin(rt_key(rt0)->sa_len, rt_key(rt)->sa_len); 1141 1142 xk1 = (u_char *)rt_key(rt0); 1143 xm1 = (u_char *)rt_mask(rt0); 1144 xk2 = (u_char *)rt_key(rt); 1145 1146 /* avoid applying a less specific route */ 1147 xmp = (u_char *)rt_mask(rt->rt_parent); 1148 mlen = rt_key(rt->rt_parent)->sa_len; 1149 if (mlen > rt_key(rt0)->sa_len) { 1150 #ifdef DEBUG 1151 if (rtfcdebug) 1152 kprintf("rt_fixchange: inserting a less " 1153 "specific route\n"); 1154 #endif 1155 return 0; 1156 } 1157 for (i = rnh->rnh_treetop->rn_offset; i < mlen; i++) { 1158 if ((xmp[i] & ~(xmp[i] ^ xm1[i])) != xmp[i]) { 1159 #ifdef DEBUG 1160 if (rtfcdebug) 1161 kprintf("rt_fixchange: inserting a less " 1162 "specific route\n"); 1163 #endif 1164 return 0; 1165 } 1166 } 1167 1168 for (i = rnh->rnh_treetop->rn_offset; i < len; i++) { 1169 if ((xk2[i] & xm1[i]) != xk1[i]) { 1170 #ifdef DEBUG 1171 if (rtfcdebug) kprintf("no match\n"); 1172 #endif 1173 return 0; 1174 } 1175 } 1176 1177 /* 1178 * OK, this node is a clone, and matches the node currently being 1179 * changed/added under the node's mask. So, get rid of it. 1180 */ 1181 #ifdef DEBUG 1182 if (rtfcdebug) kprintf("deleting\n"); 1183 #endif 1184 return rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 1185 rt->rt_flags, NULL); 1186 } 1187 1188 #define ROUNDUP(a) (a>0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long)) 1189 1190 int 1191 rt_setgate(struct rtentry *rt0, struct sockaddr *dst, struct sockaddr *gate) 1192 { 1193 char *space, *oldspace; 1194 int dlen = ROUNDUP(dst->sa_len), glen = ROUNDUP(gate->sa_len); 1195 struct rtentry *rt = rt0; 1196 struct radix_node_head *rnh = rt_tables[mycpuid][dst->sa_family]; 1197 1198 /* 1199 * A host route with the destination equal to the gateway 1200 * will interfere with keeping LLINFO in the routing 1201 * table, so disallow it. 1202 */ 1203 if (((rt0->rt_flags & (RTF_HOST | RTF_GATEWAY | RTF_LLINFO)) == 1204 (RTF_HOST | RTF_GATEWAY)) && 1205 dst->sa_len == gate->sa_len && 1206 sa_equal(dst, gate)) { 1207 /* 1208 * The route might already exist if this is an RTM_CHANGE 1209 * or a routing redirect, so try to delete it. 1210 */ 1211 if (rt_key(rt0) != NULL) 1212 rtrequest(RTM_DELETE, rt_key(rt0), rt0->rt_gateway, 1213 rt_mask(rt0), rt0->rt_flags, NULL); 1214 return EADDRNOTAVAIL; 1215 } 1216 1217 /* 1218 * Both dst and gateway are stored in the same malloc'ed chunk 1219 * (If I ever get my hands on....) 1220 * if we need to malloc a new chunk, then keep the old one around 1221 * till we don't need it any more. 1222 */ 1223 if (rt->rt_gateway == NULL || glen > ROUNDUP(rt->rt_gateway->sa_len)) { 1224 oldspace = (char *)rt_key(rt); 1225 R_Malloc(space, char *, dlen + glen); 1226 if (space == NULL) 1227 return ENOBUFS; 1228 rt->rt_nodes->rn_key = space; 1229 } else { 1230 space = (char *)rt_key(rt); /* Just use the old space. */ 1231 oldspace = NULL; 1232 } 1233 1234 /* Set the gateway value. */ 1235 rt->rt_gateway = (struct sockaddr *)(space + dlen); 1236 bcopy(gate, rt->rt_gateway, glen); 1237 1238 if (oldspace != NULL) { 1239 /* 1240 * If we allocated a new chunk, preserve the original dst. 1241 * This way, rt_setgate() really just sets the gate 1242 * and leaves the dst field alone. 1243 */ 1244 bcopy(dst, space, dlen); 1245 Free(oldspace); 1246 } 1247 1248 /* 1249 * If there is already a gwroute, it's now almost definitely wrong 1250 * so drop it. 1251 */ 1252 if (rt->rt_gwroute != NULL) { 1253 RTFREE(rt->rt_gwroute); 1254 rt->rt_gwroute = NULL; 1255 } 1256 if (rt->rt_flags & RTF_GATEWAY) { 1257 /* 1258 * Cloning loop avoidance: In the presence of 1259 * protocol-cloning and bad configuration, it is 1260 * possible to get stuck in bottomless mutual recursion 1261 * (rtrequest rt_setgate rtlookup). We avoid this 1262 * by not allowing protocol-cloning to operate for 1263 * gateways (which is probably the correct choice 1264 * anyway), and avoid the resulting reference loops 1265 * by disallowing any route to run through itself as 1266 * a gateway. This is obviously mandatory when we 1267 * get rt->rt_output(). 1268 * 1269 * This breaks TTCP for hosts outside the gateway! XXX JH 1270 */ 1271 rt->rt_gwroute = _rtlookup(gate, RTL_REPORTMSG, RTF_PRCLONING); 1272 if (rt->rt_gwroute == rt) { 1273 rt->rt_gwroute = NULL; 1274 --rt->rt_refcnt; 1275 return EDQUOT; /* failure */ 1276 } 1277 } 1278 1279 /* 1280 * This isn't going to do anything useful for host routes, so 1281 * don't bother. Also make sure we have a reasonable mask 1282 * (we don't yet have one during adds). 1283 */ 1284 if (!(rt->rt_flags & RTF_HOST) && rt_mask(rt) != NULL) { 1285 struct rtfc_arg arg = { rt, rnh }; 1286 1287 rnh->rnh_walktree_from(rnh, (char *)rt_key(rt), 1288 (char *)rt_mask(rt), 1289 rt_fixchange, &arg); 1290 } 1291 1292 return 0; 1293 } 1294 1295 static void 1296 rt_maskedcopy( 1297 struct sockaddr *src, 1298 struct sockaddr *dst, 1299 struct sockaddr *netmask) 1300 { 1301 u_char *cp1 = (u_char *)src; 1302 u_char *cp2 = (u_char *)dst; 1303 u_char *cp3 = (u_char *)netmask; 1304 u_char *cplim = cp2 + *cp3; 1305 u_char *cplim2 = cp2 + *cp1; 1306 1307 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */ 1308 cp3 += 2; 1309 if (cplim > cplim2) 1310 cplim = cplim2; 1311 while (cp2 < cplim) 1312 *cp2++ = *cp1++ & *cp3++; 1313 if (cp2 < cplim2) 1314 bzero(cp2, cplim2 - cp2); 1315 } 1316 1317 int 1318 rt_llroute(struct sockaddr *dst, struct rtentry *rt0, struct rtentry **drt) 1319 { 1320 struct rtentry *up_rt, *rt; 1321 1322 if (!(rt0->rt_flags & RTF_UP)) { 1323 up_rt = rtlookup(dst); 1324 if (up_rt == NULL) 1325 return (EHOSTUNREACH); 1326 up_rt->rt_refcnt--; 1327 } else 1328 up_rt = rt0; 1329 if (up_rt->rt_flags & RTF_GATEWAY) { 1330 if (up_rt->rt_gwroute == NULL) { 1331 up_rt->rt_gwroute = rtlookup(up_rt->rt_gateway); 1332 if (up_rt->rt_gwroute == NULL) 1333 return (EHOSTUNREACH); 1334 } else if (!(up_rt->rt_gwroute->rt_flags & RTF_UP)) { 1335 rtfree(up_rt->rt_gwroute); 1336 up_rt->rt_gwroute = rtlookup(up_rt->rt_gateway); 1337 if (up_rt->rt_gwroute == NULL) 1338 return (EHOSTUNREACH); 1339 } 1340 rt = up_rt->rt_gwroute; 1341 } else 1342 rt = up_rt; 1343 if (rt->rt_flags & RTF_REJECT && 1344 (rt->rt_rmx.rmx_expire == 0 || /* rt doesn't expire */ 1345 time_second < rt->rt_rmx.rmx_expire)) /* rt not expired */ 1346 return (rt->rt_flags & RTF_HOST ? EHOSTDOWN : EHOSTUNREACH); 1347 *drt = rt; 1348 return 0; 1349 } 1350 1351 static int 1352 rt_setshims(struct rtentry *rt, struct sockaddr **rt_shim){ 1353 int i; 1354 1355 for (i=0; i<3; i++) { 1356 struct sockaddr *shim = rt_shim[RTAX_MPLS1 + i]; 1357 int shimlen; 1358 1359 if (shim == NULL) 1360 break; 1361 1362 shimlen = ROUNDUP(shim->sa_len); 1363 R_Malloc(rt->rt_shim[i], struct sockaddr *, shimlen); 1364 bcopy(shim, rt->rt_shim[i], shimlen); 1365 } 1366 1367 return 0; 1368 } 1369 1370 #ifdef ROUTE_DEBUG 1371 1372 /* 1373 * Print out a route table entry 1374 */ 1375 void 1376 rt_print(struct rt_addrinfo *rtinfo, struct rtentry *rn) 1377 { 1378 kprintf("rti %p cpu %d route %p flags %08lx: ", 1379 rtinfo, mycpuid, rn, rn->rt_flags); 1380 sockaddr_print(rt_key(rn)); 1381 kprintf(" mask "); 1382 sockaddr_print(rt_mask(rn)); 1383 kprintf(" gw "); 1384 sockaddr_print(rn->rt_gateway); 1385 kprintf(" ifc \"%s\"", rn->rt_ifp ? rn->rt_ifp->if_dname : "?"); 1386 kprintf(" ifa %p\n", rn->rt_ifa); 1387 } 1388 1389 void 1390 rt_addrinfo_print(int cmd, struct rt_addrinfo *rti) 1391 { 1392 int didit = 0; 1393 int i; 1394 1395 #ifdef ROUTE_DEBUG 1396 if (cmd == RTM_DELETE && route_debug > 1) 1397 db_print_backtrace(); 1398 #endif 1399 1400 switch(cmd) { 1401 case RTM_ADD: 1402 kprintf("ADD "); 1403 break; 1404 case RTM_RESOLVE: 1405 kprintf("RES "); 1406 break; 1407 case RTM_DELETE: 1408 kprintf("DEL "); 1409 break; 1410 default: 1411 kprintf("C%02d ", cmd); 1412 break; 1413 } 1414 kprintf("rti %p cpu %d ", rti, mycpuid); 1415 for (i = 0; i < rti->rti_addrs; ++i) { 1416 if (rti->rti_info[i] == NULL) 1417 continue; 1418 if (didit) 1419 kprintf(" ,"); 1420 switch(i) { 1421 case RTAX_DST: 1422 kprintf("(DST "); 1423 break; 1424 case RTAX_GATEWAY: 1425 kprintf("(GWY "); 1426 break; 1427 case RTAX_NETMASK: 1428 kprintf("(MSK "); 1429 break; 1430 case RTAX_GENMASK: 1431 kprintf("(GEN "); 1432 break; 1433 case RTAX_IFP: 1434 kprintf("(IFP "); 1435 break; 1436 case RTAX_IFA: 1437 kprintf("(IFA "); 1438 break; 1439 case RTAX_AUTHOR: 1440 kprintf("(AUT "); 1441 break; 1442 case RTAX_BRD: 1443 kprintf("(BRD "); 1444 break; 1445 default: 1446 kprintf("(?%02d ", i); 1447 break; 1448 } 1449 sockaddr_print(rti->rti_info[i]); 1450 kprintf(")"); 1451 didit = 1; 1452 } 1453 kprintf("\n"); 1454 } 1455 1456 void 1457 sockaddr_print(struct sockaddr *sa) 1458 { 1459 struct sockaddr_in *sa4; 1460 struct sockaddr_in6 *sa6; 1461 int len; 1462 int i; 1463 1464 if (sa == NULL) { 1465 kprintf("NULL"); 1466 return; 1467 } 1468 1469 len = sa->sa_len - offsetof(struct sockaddr, sa_data[0]); 1470 1471 switch(sa->sa_family) { 1472 case AF_INET: 1473 case AF_INET6: 1474 default: 1475 switch(sa->sa_family) { 1476 case AF_INET: 1477 sa4 = (struct sockaddr_in *)sa; 1478 kprintf("INET %d %d.%d.%d.%d", 1479 ntohs(sa4->sin_port), 1480 (ntohl(sa4->sin_addr.s_addr) >> 24) & 255, 1481 (ntohl(sa4->sin_addr.s_addr) >> 16) & 255, 1482 (ntohl(sa4->sin_addr.s_addr) >> 8) & 255, 1483 (ntohl(sa4->sin_addr.s_addr) >> 0) & 255 1484 ); 1485 break; 1486 case AF_INET6: 1487 sa6 = (struct sockaddr_in6 *)sa; 1488 kprintf("INET6 %d %04x:%04x%04x:%04x:%04x:%04x:%04x:%04x", 1489 ntohs(sa6->sin6_port), 1490 sa6->sin6_addr.s6_addr16[0], 1491 sa6->sin6_addr.s6_addr16[1], 1492 sa6->sin6_addr.s6_addr16[2], 1493 sa6->sin6_addr.s6_addr16[3], 1494 sa6->sin6_addr.s6_addr16[4], 1495 sa6->sin6_addr.s6_addr16[5], 1496 sa6->sin6_addr.s6_addr16[6], 1497 sa6->sin6_addr.s6_addr16[7] 1498 ); 1499 break; 1500 default: 1501 kprintf("AF%d ", sa->sa_family); 1502 while (len > 0 && sa->sa_data[len-1] == 0) 1503 --len; 1504 1505 for (i = 0; i < len; ++i) { 1506 if (i) 1507 kprintf("."); 1508 kprintf("%d", (unsigned char)sa->sa_data[i]); 1509 } 1510 break; 1511 } 1512 } 1513 } 1514 1515 #endif 1516 1517 /* 1518 * Set up a routing table entry, normally for an interface. 1519 */ 1520 int 1521 rtinit(struct ifaddr *ifa, int cmd, int flags) 1522 { 1523 struct sockaddr *dst, *deldst, *netmask; 1524 struct mbuf *m = NULL; 1525 struct radix_node_head *rnh; 1526 struct radix_node *rn; 1527 struct rt_addrinfo rtinfo; 1528 int error; 1529 1530 if (flags & RTF_HOST) { 1531 dst = ifa->ifa_dstaddr; 1532 netmask = NULL; 1533 } else { 1534 dst = ifa->ifa_addr; 1535 netmask = ifa->ifa_netmask; 1536 } 1537 /* 1538 * If it's a delete, check that if it exists, it's on the correct 1539 * interface or we might scrub a route to another ifa which would 1540 * be confusing at best and possibly worse. 1541 */ 1542 if (cmd == RTM_DELETE) { 1543 /* 1544 * It's a delete, so it should already exist.. 1545 * If it's a net, mask off the host bits 1546 * (Assuming we have a mask) 1547 */ 1548 if (netmask != NULL) { 1549 m = m_get(MB_DONTWAIT, MT_SONAME); 1550 if (m == NULL) 1551 return (ENOBUFS); 1552 mbuftrackid(m, 34); 1553 deldst = mtod(m, struct sockaddr *); 1554 rt_maskedcopy(dst, deldst, netmask); 1555 dst = deldst; 1556 } 1557 /* 1558 * Look up an rtentry that is in the routing tree and 1559 * contains the correct info. 1560 */ 1561 if ((rnh = rt_tables[mycpuid][dst->sa_family]) == NULL || 1562 (rn = rnh->rnh_lookup((char *)dst, 1563 (char *)netmask, rnh)) == NULL || 1564 ((struct rtentry *)rn)->rt_ifa != ifa || 1565 !sa_equal((struct sockaddr *)rn->rn_key, dst)) { 1566 if (m != NULL) 1567 m_free(m); 1568 return (flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH); 1569 } 1570 /* XXX */ 1571 #if 0 1572 else { 1573 /* 1574 * One would think that as we are deleting, and we know 1575 * it doesn't exist, we could just return at this point 1576 * with an "ELSE" clause, but apparently not.. 1577 */ 1578 return (flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH); 1579 } 1580 #endif 1581 } 1582 /* 1583 * Do the actual request 1584 */ 1585 bzero(&rtinfo, sizeof(struct rt_addrinfo)); 1586 rtinfo.rti_info[RTAX_DST] = dst; 1587 rtinfo.rti_info[RTAX_GATEWAY] = ifa->ifa_addr; 1588 rtinfo.rti_info[RTAX_NETMASK] = netmask; 1589 rtinfo.rti_flags = flags | ifa->ifa_flags; 1590 rtinfo.rti_ifa = ifa; 1591 error = rtrequest1_global(cmd, &rtinfo, rtinit_rtrequest_callback, ifa); 1592 if (m != NULL) 1593 m_free(m); 1594 return (error); 1595 } 1596 1597 static void 1598 rtinit_rtrequest_callback(int cmd, int error, 1599 struct rt_addrinfo *rtinfo, struct rtentry *rt, 1600 void *arg) 1601 { 1602 struct ifaddr *ifa = arg; 1603 1604 if (error == 0 && rt) { 1605 if (mycpuid == 0) { 1606 ++rt->rt_refcnt; 1607 rt_newaddrmsg(cmd, ifa, error, rt); 1608 --rt->rt_refcnt; 1609 } 1610 if (cmd == RTM_DELETE) { 1611 if (rt->rt_refcnt == 0) { 1612 ++rt->rt_refcnt; 1613 rtfree(rt); 1614 } 1615 } 1616 } 1617 } 1618 1619 /* This must be before ip6_init2(), which is now SI_ORDER_MIDDLE */ 1620 SYSINIT(route, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, route_init, 0); 1621