1 /*- 2 * Copyright (c) 1980, 1986, 1991, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)route.c 8.3.1.1 (Berkeley) 2/23/95 30 * $FreeBSD$ 31 */ 32 /************************************************************************ 33 * Note: In this file a 'fib' is a "forwarding information base" * 34 * Which is the new name for an in kernel routing (next hop) table. * 35 ***********************************************************************/ 36 37 #include "opt_inet.h" 38 #include "opt_inet6.h" 39 #include "opt_route.h" 40 #include "opt_sctp.h" 41 #include "opt_mrouting.h" 42 #include "opt_mpath.h" 43 44 #include <sys/param.h> 45 #include <sys/systm.h> 46 #include <sys/malloc.h> 47 #include <sys/mbuf.h> 48 #include <sys/socket.h> 49 #include <sys/sysctl.h> 50 #include <sys/syslog.h> 51 #include <sys/sysproto.h> 52 #include <sys/proc.h> 53 #include <sys/domain.h> 54 #include <sys/kernel.h> 55 56 #include <net/if.h> 57 #include <net/if_var.h> 58 #include <net/if_dl.h> 59 #include <net/route.h> 60 #include <net/route_var.h> 61 #include <net/vnet.h> 62 #include <net/flowtable.h> 63 64 #ifdef RADIX_MPATH 65 #include <net/radix_mpath.h> 66 #endif 67 68 #include <netinet/in.h> 69 #include <netinet/ip_mroute.h> 70 71 #include <vm/uma.h> 72 73 #define RT_MAXFIBS UINT16_MAX 74 75 /* Kernel config default option. */ 76 #ifdef ROUTETABLES 77 #if ROUTETABLES <= 0 78 #error "ROUTETABLES defined too low" 79 #endif 80 #if ROUTETABLES > RT_MAXFIBS 81 #error "ROUTETABLES defined too big" 82 #endif 83 #define RT_NUMFIBS ROUTETABLES 84 #endif /* ROUTETABLES */ 85 /* Initialize to default if not otherwise set. */ 86 #ifndef RT_NUMFIBS 87 #define RT_NUMFIBS 1 88 #endif 89 90 #if defined(INET) || defined(INET6) 91 #ifdef SCTP 92 extern void sctp_addr_change(struct ifaddr *ifa, int cmd); 93 #endif /* SCTP */ 94 #endif 95 96 97 /* This is read-only.. */ 98 u_int rt_numfibs = RT_NUMFIBS; 99 SYSCTL_UINT(_net, OID_AUTO, fibs, CTLFLAG_RDTUN, &rt_numfibs, 0, ""); 100 101 /* 102 * By default add routes to all fibs for new interfaces. 103 * Once this is set to 0 then only allocate routes on interface 104 * changes for the FIB of the caller when adding a new set of addresses 105 * to an interface. XXX this is a shotgun aproach to a problem that needs 106 * a more fine grained solution.. that will come. 107 * XXX also has the problems getting the FIB from curthread which will not 108 * always work given the fib can be overridden and prefixes can be added 109 * from the network stack context. 110 */ 111 VNET_DEFINE(u_int, rt_add_addr_allfibs) = 1; 112 SYSCTL_UINT(_net, OID_AUTO, add_addr_allfibs, CTLFLAG_RWTUN | CTLFLAG_VNET, 113 &VNET_NAME(rt_add_addr_allfibs), 0, ""); 114 115 VNET_DEFINE(struct rtstat, rtstat); 116 #define V_rtstat VNET(rtstat) 117 118 VNET_DEFINE(struct rib_head *, rt_tables); 119 #define V_rt_tables VNET(rt_tables) 120 121 VNET_DEFINE(int, rttrash); /* routes not in table but not freed */ 122 #define V_rttrash VNET(rttrash) 123 124 125 /* 126 * Convert a 'struct radix_node *' to a 'struct rtentry *'. 127 * The operation can be done safely (in this code) because a 128 * 'struct rtentry' starts with two 'struct radix_node''s, the first 129 * one representing leaf nodes in the routing tree, which is 130 * what the code in radix.c passes us as a 'struct radix_node'. 131 * 132 * But because there are a lot of assumptions in this conversion, 133 * do not cast explicitly, but always use the macro below. 134 */ 135 #define RNTORT(p) ((struct rtentry *)(p)) 136 137 static VNET_DEFINE(uma_zone_t, rtzone); /* Routing table UMA zone. */ 138 #define V_rtzone VNET(rtzone) 139 140 static int rtrequest1_fib_change(struct rib_head *, struct rt_addrinfo *, 141 struct rtentry **, u_int); 142 static void rt_setmetrics(const struct rt_addrinfo *, struct rtentry *); 143 static int rt_ifdelroute(const struct rtentry *rt, void *arg); 144 static struct rtentry *rt_unlinkrte(struct rib_head *rnh, 145 struct rt_addrinfo *info, int *perror); 146 static void rt_notifydelete(struct rtentry *rt, struct rt_addrinfo *info); 147 #ifdef RADIX_MPATH 148 static struct radix_node *rt_mpath_unlink(struct rib_head *rnh, 149 struct rt_addrinfo *info, struct rtentry *rto, int *perror); 150 #endif 151 static int rt_exportinfo(struct rtentry *rt, struct rt_addrinfo *info, 152 int flags); 153 154 struct if_mtuinfo 155 { 156 struct ifnet *ifp; 157 int mtu; 158 }; 159 160 static int if_updatemtu_cb(struct radix_node *, void *); 161 162 /* 163 * handler for net.my_fibnum 164 */ 165 static int 166 sysctl_my_fibnum(SYSCTL_HANDLER_ARGS) 167 { 168 int fibnum; 169 int error; 170 171 fibnum = curthread->td_proc->p_fibnum; 172 error = sysctl_handle_int(oidp, &fibnum, 0, req); 173 return (error); 174 } 175 176 SYSCTL_PROC(_net, OID_AUTO, my_fibnum, CTLTYPE_INT|CTLFLAG_RD, 177 NULL, 0, &sysctl_my_fibnum, "I", "default FIB of caller"); 178 179 static __inline struct rib_head ** 180 rt_tables_get_rnh_ptr(int table, int fam) 181 { 182 struct rib_head **rnh; 183 184 KASSERT(table >= 0 && table < rt_numfibs, ("%s: table out of bounds.", 185 __func__)); 186 KASSERT(fam >= 0 && fam < (AF_MAX+1), ("%s: fam out of bounds.", 187 __func__)); 188 189 /* rnh is [fib=0][af=0]. */ 190 rnh = (struct rib_head **)V_rt_tables; 191 /* Get the offset to the requested table and fam. */ 192 rnh += table * (AF_MAX+1) + fam; 193 194 return (rnh); 195 } 196 197 struct rib_head * 198 rt_tables_get_rnh(int table, int fam) 199 { 200 201 return (*rt_tables_get_rnh_ptr(table, fam)); 202 } 203 204 u_int 205 rt_tables_get_gen(int table, int fam) 206 { 207 struct rib_head *rnh; 208 209 rnh = *rt_tables_get_rnh_ptr(table, fam); 210 KASSERT(rnh != NULL, ("%s: NULL rib_head pointer table %d fam %d", 211 __func__, table, fam)); 212 return (rnh->rnh_gen); 213 } 214 215 216 /* 217 * route initialization must occur before ip6_init2(), which happenas at 218 * SI_ORDER_MIDDLE. 219 */ 220 static void 221 route_init(void) 222 { 223 224 /* whack the tunable ints into line. */ 225 if (rt_numfibs > RT_MAXFIBS) 226 rt_numfibs = RT_MAXFIBS; 227 if (rt_numfibs == 0) 228 rt_numfibs = 1; 229 } 230 SYSINIT(route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, route_init, 0); 231 232 static int 233 rtentry_zinit(void *mem, int size, int how) 234 { 235 struct rtentry *rt = mem; 236 237 rt->rt_pksent = counter_u64_alloc(how); 238 if (rt->rt_pksent == NULL) 239 return (ENOMEM); 240 241 RT_LOCK_INIT(rt); 242 243 return (0); 244 } 245 246 static void 247 rtentry_zfini(void *mem, int size) 248 { 249 struct rtentry *rt = mem; 250 251 RT_LOCK_DESTROY(rt); 252 counter_u64_free(rt->rt_pksent); 253 } 254 255 static int 256 rtentry_ctor(void *mem, int size, void *arg, int how) 257 { 258 struct rtentry *rt = mem; 259 260 bzero(rt, offsetof(struct rtentry, rt_endzero)); 261 counter_u64_zero(rt->rt_pksent); 262 rt->rt_chain = NULL; 263 264 return (0); 265 } 266 267 static void 268 rtentry_dtor(void *mem, int size, void *arg) 269 { 270 struct rtentry *rt = mem; 271 272 RT_UNLOCK_COND(rt); 273 } 274 275 static void 276 vnet_route_init(const void *unused __unused) 277 { 278 struct domain *dom; 279 struct rib_head **rnh; 280 int table; 281 int fam; 282 283 V_rt_tables = malloc(rt_numfibs * (AF_MAX+1) * 284 sizeof(struct rib_head *), M_RTABLE, M_WAITOK|M_ZERO); 285 286 V_rtzone = uma_zcreate("rtentry", sizeof(struct rtentry), 287 rtentry_ctor, rtentry_dtor, 288 rtentry_zinit, rtentry_zfini, UMA_ALIGN_PTR, 0); 289 for (dom = domains; dom; dom = dom->dom_next) { 290 if (dom->dom_rtattach == NULL) 291 continue; 292 293 for (table = 0; table < rt_numfibs; table++) { 294 fam = dom->dom_family; 295 if (table != 0 && fam != AF_INET6 && fam != AF_INET) 296 break; 297 298 rnh = rt_tables_get_rnh_ptr(table, fam); 299 if (rnh == NULL) 300 panic("%s: rnh NULL", __func__); 301 dom->dom_rtattach((void **)rnh, 0); 302 } 303 } 304 } 305 VNET_SYSINIT(vnet_route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, 306 vnet_route_init, 0); 307 308 #ifdef VIMAGE 309 static void 310 vnet_route_uninit(const void *unused __unused) 311 { 312 int table; 313 int fam; 314 struct domain *dom; 315 struct rib_head **rnh; 316 317 for (dom = domains; dom; dom = dom->dom_next) { 318 if (dom->dom_rtdetach == NULL) 319 continue; 320 321 for (table = 0; table < rt_numfibs; table++) { 322 fam = dom->dom_family; 323 324 if (table != 0 && fam != AF_INET6 && fam != AF_INET) 325 break; 326 327 rnh = rt_tables_get_rnh_ptr(table, fam); 328 if (rnh == NULL) 329 panic("%s: rnh NULL", __func__); 330 dom->dom_rtdetach((void **)rnh, 0); 331 } 332 } 333 334 free(V_rt_tables, M_RTABLE); 335 uma_zdestroy(V_rtzone); 336 } 337 VNET_SYSUNINIT(vnet_route_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST, 338 vnet_route_uninit, 0); 339 #endif 340 341 struct rib_head * 342 rt_table_init(int offset) 343 { 344 struct rib_head *rh; 345 346 rh = malloc(sizeof(struct rib_head), M_RTABLE, M_WAITOK | M_ZERO); 347 348 /* TODO: These details should be hidded inside radix.c */ 349 /* Init masks tree */ 350 rn_inithead_internal(&rh->head, rh->rnh_nodes, offset); 351 rn_inithead_internal(&rh->rmhead.head, rh->rmhead.mask_nodes, 0); 352 rh->head.rnh_masks = &rh->rmhead; 353 354 /* Init locks */ 355 RIB_LOCK_INIT(rh); 356 357 /* Finally, set base callbacks */ 358 rh->rnh_addaddr = rn_addroute; 359 rh->rnh_deladdr = rn_delete; 360 rh->rnh_matchaddr = rn_match; 361 rh->rnh_lookup = rn_lookup; 362 rh->rnh_walktree = rn_walktree; 363 rh->rnh_walktree_from = rn_walktree_from; 364 365 return (rh); 366 } 367 368 static int 369 rt_freeentry(struct radix_node *rn, void *arg) 370 { 371 struct radix_head * const rnh = arg; 372 struct radix_node *x; 373 374 x = (struct radix_node *)rn_delete(rn + 2, NULL, rnh); 375 if (x != NULL) 376 R_Free(x); 377 return (0); 378 } 379 380 void 381 rt_table_destroy(struct rib_head *rh) 382 { 383 384 rn_walktree(&rh->rmhead.head, rt_freeentry, &rh->rmhead.head); 385 386 /* Assume table is already empty */ 387 RIB_LOCK_DESTROY(rh); 388 free(rh, M_RTABLE); 389 } 390 391 392 #ifndef _SYS_SYSPROTO_H_ 393 struct setfib_args { 394 int fibnum; 395 }; 396 #endif 397 int 398 sys_setfib(struct thread *td, struct setfib_args *uap) 399 { 400 if (uap->fibnum < 0 || uap->fibnum >= rt_numfibs) 401 return EINVAL; 402 td->td_proc->p_fibnum = uap->fibnum; 403 return (0); 404 } 405 406 /* 407 * Packet routing routines. 408 */ 409 void 410 rtalloc_ign_fib(struct route *ro, u_long ignore, u_int fibnum) 411 { 412 struct rtentry *rt; 413 414 if ((rt = ro->ro_rt) != NULL) { 415 if (rt->rt_ifp != NULL && rt->rt_flags & RTF_UP) 416 return; 417 RTFREE(rt); 418 ro->ro_rt = NULL; 419 } 420 ro->ro_rt = rtalloc1_fib(&ro->ro_dst, 1, ignore, fibnum); 421 if (ro->ro_rt) 422 RT_UNLOCK(ro->ro_rt); 423 } 424 425 /* 426 * Look up the route that matches the address given 427 * Or, at least try.. Create a cloned route if needed. 428 * 429 * The returned route, if any, is locked. 430 */ 431 struct rtentry * 432 rtalloc1(struct sockaddr *dst, int report, u_long ignflags) 433 { 434 435 return (rtalloc1_fib(dst, report, ignflags, RT_DEFAULT_FIB)); 436 } 437 438 struct rtentry * 439 rtalloc1_fib(struct sockaddr *dst, int report, u_long ignflags, 440 u_int fibnum) 441 { 442 struct rib_head *rh; 443 struct radix_node *rn; 444 struct rtentry *newrt; 445 struct rt_addrinfo info; 446 int err = 0, msgtype = RTM_MISS; 447 448 KASSERT((fibnum < rt_numfibs), ("rtalloc1_fib: bad fibnum")); 449 rh = rt_tables_get_rnh(fibnum, dst->sa_family); 450 newrt = NULL; 451 if (rh == NULL) 452 goto miss; 453 454 /* 455 * Look up the address in the table for that Address Family 456 */ 457 RIB_RLOCK(rh); 458 rn = rh->rnh_matchaddr(dst, &rh->head); 459 if (rn && ((rn->rn_flags & RNF_ROOT) == 0)) { 460 newrt = RNTORT(rn); 461 RT_LOCK(newrt); 462 RT_ADDREF(newrt); 463 RIB_RUNLOCK(rh); 464 return (newrt); 465 466 } else 467 RIB_RUNLOCK(rh); 468 469 /* 470 * Either we hit the root or could not find any match, 471 * which basically means: "cannot get there from here". 472 */ 473 miss: 474 V_rtstat.rts_unreach++; 475 476 if (report) { 477 /* 478 * If required, report the failure to the supervising 479 * Authorities. 480 * For a delete, this is not an error. (report == 0) 481 */ 482 bzero(&info, sizeof(info)); 483 info.rti_info[RTAX_DST] = dst; 484 rt_missmsg_fib(msgtype, &info, 0, err, fibnum); 485 } 486 return (newrt); 487 } 488 489 /* 490 * Remove a reference count from an rtentry. 491 * If the count gets low enough, take it out of the routing table 492 */ 493 void 494 rtfree(struct rtentry *rt) 495 { 496 struct rib_head *rnh; 497 498 KASSERT(rt != NULL,("%s: NULL rt", __func__)); 499 rnh = rt_tables_get_rnh(rt->rt_fibnum, rt_key(rt)->sa_family); 500 KASSERT(rnh != NULL,("%s: NULL rnh", __func__)); 501 502 RT_LOCK_ASSERT(rt); 503 504 /* 505 * The callers should use RTFREE_LOCKED() or RTFREE(), so 506 * we should come here exactly with the last reference. 507 */ 508 RT_REMREF(rt); 509 if (rt->rt_refcnt > 0) { 510 log(LOG_DEBUG, "%s: %p has %d refs\n", __func__, rt, rt->rt_refcnt); 511 goto done; 512 } 513 514 /* 515 * On last reference give the "close method" a chance 516 * to cleanup private state. This also permits (for 517 * IPv4 and IPv6) a chance to decide if the routing table 518 * entry should be purged immediately or at a later time. 519 * When an immediate purge is to happen the close routine 520 * typically calls rtexpunge which clears the RTF_UP flag 521 * on the entry so that the code below reclaims the storage. 522 */ 523 if (rt->rt_refcnt == 0 && rnh->rnh_close) 524 rnh->rnh_close((struct radix_node *)rt, &rnh->head); 525 526 /* 527 * If we are no longer "up" (and ref == 0) 528 * then we can free the resources associated 529 * with the route. 530 */ 531 if ((rt->rt_flags & RTF_UP) == 0) { 532 if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT)) 533 panic("rtfree 2"); 534 /* 535 * the rtentry must have been removed from the routing table 536 * so it is represented in rttrash.. remove that now. 537 */ 538 V_rttrash--; 539 #ifdef DIAGNOSTIC 540 if (rt->rt_refcnt < 0) { 541 printf("rtfree: %p not freed (neg refs)\n", rt); 542 goto done; 543 } 544 #endif 545 /* 546 * release references on items we hold them on.. 547 * e.g other routes and ifaddrs. 548 */ 549 if (rt->rt_ifa) 550 ifa_free(rt->rt_ifa); 551 /* 552 * The key is separatly alloc'd so free it (see rt_setgate()). 553 * This also frees the gateway, as they are always malloc'd 554 * together. 555 */ 556 R_Free(rt_key(rt)); 557 558 /* 559 * and the rtentry itself of course 560 */ 561 uma_zfree(V_rtzone, rt); 562 return; 563 } 564 done: 565 RT_UNLOCK(rt); 566 } 567 568 569 /* 570 * Force a routing table entry to the specified 571 * destination to go through the given gateway. 572 * Normally called as a result of a routing redirect 573 * message from the network layer. 574 */ 575 void 576 rtredirect_fib(struct sockaddr *dst, 577 struct sockaddr *gateway, 578 struct sockaddr *netmask, 579 int flags, 580 struct sockaddr *src, 581 u_int fibnum) 582 { 583 struct rtentry *rt; 584 int error = 0; 585 short *stat = NULL; 586 struct rt_addrinfo info; 587 struct ifaddr *ifa; 588 struct rib_head *rnh; 589 590 ifa = NULL; 591 rnh = rt_tables_get_rnh(fibnum, dst->sa_family); 592 if (rnh == NULL) { 593 error = EAFNOSUPPORT; 594 goto out; 595 } 596 597 /* verify the gateway is directly reachable */ 598 if ((ifa = ifa_ifwithnet(gateway, 0, fibnum)) == NULL) { 599 error = ENETUNREACH; 600 goto out; 601 } 602 rt = rtalloc1_fib(dst, 0, 0UL, fibnum); /* NB: rt is locked */ 603 /* 604 * If the redirect isn't from our current router for this dst, 605 * it's either old or wrong. If it redirects us to ourselves, 606 * we have a routing loop, perhaps as a result of an interface 607 * going down recently. 608 */ 609 if (!(flags & RTF_DONE) && rt) { 610 if (!sa_equal(src, rt->rt_gateway)) { 611 error = EINVAL; 612 goto done; 613 } 614 if (rt->rt_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK) { 615 error = EINVAL; 616 goto done; 617 } 618 } 619 if ((flags & RTF_GATEWAY) && ifa_ifwithaddr_check(gateway)) { 620 error = EHOSTUNREACH; 621 goto done; 622 } 623 /* 624 * Create a new entry if we just got back a wildcard entry 625 * or the lookup failed. This is necessary for hosts 626 * which use routing redirects generated by smart gateways 627 * to dynamically build the routing tables. 628 */ 629 if (rt == NULL || (rt_mask(rt) && rt_mask(rt)->sa_len < 2)) 630 goto create; 631 /* 632 * Don't listen to the redirect if it's 633 * for a route to an interface. 634 */ 635 if (rt->rt_flags & RTF_GATEWAY) { 636 if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) { 637 /* 638 * Changing from route to net => route to host. 639 * Create new route, rather than smashing route to net. 640 */ 641 create: 642 if (rt != NULL) 643 RTFREE_LOCKED(rt); 644 645 flags |= RTF_DYNAMIC; 646 bzero((caddr_t)&info, sizeof(info)); 647 info.rti_info[RTAX_DST] = dst; 648 info.rti_info[RTAX_GATEWAY] = gateway; 649 info.rti_info[RTAX_NETMASK] = netmask; 650 info.rti_ifa = ifa; 651 info.rti_flags = flags; 652 error = rtrequest1_fib(RTM_ADD, &info, &rt, fibnum); 653 if (rt != NULL) { 654 RT_LOCK(rt); 655 flags = rt->rt_flags; 656 } 657 658 stat = &V_rtstat.rts_dynamic; 659 } else { 660 661 /* 662 * Smash the current notion of the gateway to 663 * this destination. Should check about netmask!!! 664 */ 665 if ((flags & RTF_GATEWAY) == 0) 666 rt->rt_flags &= ~RTF_GATEWAY; 667 rt->rt_flags |= RTF_MODIFIED; 668 flags |= RTF_MODIFIED; 669 stat = &V_rtstat.rts_newgateway; 670 /* 671 * add the key and gateway (in one malloc'd chunk). 672 */ 673 RT_UNLOCK(rt); 674 RIB_WLOCK(rnh); 675 RT_LOCK(rt); 676 rt_setgate(rt, rt_key(rt), gateway); 677 RIB_WUNLOCK(rnh); 678 } 679 } else 680 error = EHOSTUNREACH; 681 done: 682 if (rt) 683 RTFREE_LOCKED(rt); 684 out: 685 if (error) 686 V_rtstat.rts_badredirect++; 687 else if (stat != NULL) 688 (*stat)++; 689 bzero((caddr_t)&info, sizeof(info)); 690 info.rti_info[RTAX_DST] = dst; 691 info.rti_info[RTAX_GATEWAY] = gateway; 692 info.rti_info[RTAX_NETMASK] = netmask; 693 info.rti_info[RTAX_AUTHOR] = src; 694 rt_missmsg_fib(RTM_REDIRECT, &info, flags, error, fibnum); 695 if (ifa != NULL) 696 ifa_free(ifa); 697 } 698 699 /* 700 * Routing table ioctl interface. 701 */ 702 int 703 rtioctl_fib(u_long req, caddr_t data, u_int fibnum) 704 { 705 706 /* 707 * If more ioctl commands are added here, make sure the proper 708 * super-user checks are being performed because it is possible for 709 * prison-root to make it this far if raw sockets have been enabled 710 * in jails. 711 */ 712 #ifdef INET 713 /* Multicast goop, grrr... */ 714 return mrt_ioctl ? mrt_ioctl(req, data, fibnum) : EOPNOTSUPP; 715 #else /* INET */ 716 return ENXIO; 717 #endif /* INET */ 718 } 719 720 struct ifaddr * 721 ifa_ifwithroute(int flags, const struct sockaddr *dst, struct sockaddr *gateway, 722 u_int fibnum) 723 { 724 struct ifaddr *ifa; 725 int not_found = 0; 726 727 if ((flags & RTF_GATEWAY) == 0) { 728 /* 729 * If we are adding a route to an interface, 730 * and the interface is a pt to pt link 731 * we should search for the destination 732 * as our clue to the interface. Otherwise 733 * we can use the local address. 734 */ 735 ifa = NULL; 736 if (flags & RTF_HOST) 737 ifa = ifa_ifwithdstaddr(dst, fibnum); 738 if (ifa == NULL) 739 ifa = ifa_ifwithaddr(gateway); 740 } else { 741 /* 742 * If we are adding a route to a remote net 743 * or host, the gateway may still be on the 744 * other end of a pt to pt link. 745 */ 746 ifa = ifa_ifwithdstaddr(gateway, fibnum); 747 } 748 if (ifa == NULL) 749 ifa = ifa_ifwithnet(gateway, 0, fibnum); 750 if (ifa == NULL) { 751 struct rtentry *rt = rtalloc1_fib(gateway, 0, 0, fibnum); 752 if (rt == NULL) 753 return (NULL); 754 /* 755 * dismiss a gateway that is reachable only 756 * through the default router 757 */ 758 switch (gateway->sa_family) { 759 case AF_INET: 760 if (satosin(rt_key(rt))->sin_addr.s_addr == INADDR_ANY) 761 not_found = 1; 762 break; 763 case AF_INET6: 764 if (IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(rt))->sin6_addr)) 765 not_found = 1; 766 break; 767 default: 768 break; 769 } 770 if (!not_found && rt->rt_ifa != NULL) { 771 ifa = rt->rt_ifa; 772 ifa_ref(ifa); 773 } 774 RT_REMREF(rt); 775 RT_UNLOCK(rt); 776 if (not_found || ifa == NULL) 777 return (NULL); 778 } 779 if (ifa->ifa_addr->sa_family != dst->sa_family) { 780 struct ifaddr *oifa = ifa; 781 ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp); 782 if (ifa == NULL) 783 ifa = oifa; 784 else 785 ifa_free(oifa); 786 } 787 return (ifa); 788 } 789 790 /* 791 * Do appropriate manipulations of a routing tree given 792 * all the bits of info needed 793 */ 794 int 795 rtrequest_fib(int req, 796 struct sockaddr *dst, 797 struct sockaddr *gateway, 798 struct sockaddr *netmask, 799 int flags, 800 struct rtentry **ret_nrt, 801 u_int fibnum) 802 { 803 struct rt_addrinfo info; 804 805 if (dst->sa_len == 0) 806 return(EINVAL); 807 808 bzero((caddr_t)&info, sizeof(info)); 809 info.rti_flags = flags; 810 info.rti_info[RTAX_DST] = dst; 811 info.rti_info[RTAX_GATEWAY] = gateway; 812 info.rti_info[RTAX_NETMASK] = netmask; 813 return rtrequest1_fib(req, &info, ret_nrt, fibnum); 814 } 815 816 817 /* 818 * Copy most of @rt data into @info. 819 * 820 * If @flags contains NHR_COPY, copies dst,netmask and gw to the 821 * pointers specified by @info structure. Assume such pointers 822 * are zeroed sockaddr-like structures with sa_len field initialized 823 * to reflect size of the provided buffer. if no NHR_COPY is specified, 824 * point dst,netmask and gw @info fields to appropriate @rt values. 825 * 826 * if @flags contains NHR_REF, do refcouting on rt_ifp. 827 * 828 * Returns 0 on success. 829 */ 830 int 831 rt_exportinfo(struct rtentry *rt, struct rt_addrinfo *info, int flags) 832 { 833 struct rt_metrics *rmx; 834 struct sockaddr *src, *dst; 835 int sa_len; 836 837 if (flags & NHR_COPY) { 838 /* Copy destination if dst is non-zero */ 839 src = rt_key(rt); 840 dst = info->rti_info[RTAX_DST]; 841 sa_len = src->sa_len; 842 if (dst != NULL) { 843 if (src->sa_len > dst->sa_len) 844 return (ENOMEM); 845 memcpy(dst, src, src->sa_len); 846 info->rti_addrs |= RTA_DST; 847 } 848 849 /* Copy mask if set && dst is non-zero */ 850 src = rt_mask(rt); 851 dst = info->rti_info[RTAX_NETMASK]; 852 if (src != NULL && dst != NULL) { 853 854 /* 855 * Radix stores different value in sa_len, 856 * assume rt_mask() to have the same length 857 * as rt_key() 858 */ 859 if (sa_len > dst->sa_len) 860 return (ENOMEM); 861 memcpy(dst, src, src->sa_len); 862 info->rti_addrs |= RTA_NETMASK; 863 } 864 865 /* Copy gateway is set && dst is non-zero */ 866 src = rt->rt_gateway; 867 dst = info->rti_info[RTAX_GATEWAY]; 868 if ((rt->rt_flags & RTF_GATEWAY) && src != NULL && dst != NULL){ 869 if (src->sa_len > dst->sa_len) 870 return (ENOMEM); 871 memcpy(dst, src, src->sa_len); 872 info->rti_addrs |= RTA_GATEWAY; 873 } 874 } else { 875 info->rti_info[RTAX_DST] = rt_key(rt); 876 info->rti_addrs |= RTA_DST; 877 if (rt_mask(rt) != NULL) { 878 info->rti_info[RTAX_NETMASK] = rt_mask(rt); 879 info->rti_addrs |= RTA_NETMASK; 880 } 881 if (rt->rt_flags & RTF_GATEWAY) { 882 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway; 883 info->rti_addrs |= RTA_GATEWAY; 884 } 885 } 886 887 rmx = info->rti_rmx; 888 if (rmx != NULL) { 889 info->rti_mflags |= RTV_MTU; 890 rmx->rmx_mtu = rt->rt_mtu; 891 } 892 893 info->rti_flags = rt->rt_flags; 894 info->rti_ifp = rt->rt_ifp; 895 info->rti_ifa = rt->rt_ifa; 896 897 if (flags & NHR_REF) { 898 /* Do 'traditional' refcouting */ 899 if_ref(info->rti_ifp); 900 } 901 902 return (0); 903 } 904 905 /* 906 * Lookups up route entry for @dst in RIB database for fib @fibnum. 907 * Exports entry data to @info using rt_exportinfo(). 908 * 909 * if @flags contains NHR_REF, refcouting is performed on rt_ifp. 910 * All references can be released later by calling rib_free_info() 911 * 912 * Returns 0 on success. 913 * Returns ENOENT for lookup failure, ENOMEM for export failure. 914 */ 915 int 916 rib_lookup_info(uint32_t fibnum, const struct sockaddr *dst, uint32_t flags, 917 uint32_t flowid, struct rt_addrinfo *info) 918 { 919 struct rib_head *rh; 920 struct radix_node *rn; 921 struct rtentry *rt; 922 int error; 923 924 KASSERT((fibnum < rt_numfibs), ("rib_lookup_rte: bad fibnum")); 925 rh = rt_tables_get_rnh(fibnum, dst->sa_family); 926 if (rh == NULL) 927 return (ENOENT); 928 929 RIB_RLOCK(rh); 930 rn = rh->rnh_matchaddr(__DECONST(void *, dst), &rh->head); 931 if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) { 932 rt = RNTORT(rn); 933 /* Ensure route & ifp is UP */ 934 if (RT_LINK_IS_UP(rt->rt_ifp)) { 935 flags = (flags & NHR_REF) | NHR_COPY; 936 error = rt_exportinfo(rt, info, flags); 937 RIB_RUNLOCK(rh); 938 939 return (error); 940 } 941 } 942 RIB_RUNLOCK(rh); 943 944 return (ENOENT); 945 } 946 947 /* 948 * Releases all references acquired by rib_lookup_info() when 949 * called with NHR_REF flags. 950 */ 951 void 952 rib_free_info(struct rt_addrinfo *info) 953 { 954 955 if_rele(info->rti_ifp); 956 } 957 958 /* 959 * Iterates over all existing fibs in system calling 960 * @setwa_f function prior to traversing each fib. 961 * Calls @wa_f function for each element in current fib. 962 * If af is not AF_UNSPEC, iterates over fibs in particular 963 * address family. 964 */ 965 void 966 rt_foreach_fib_walk(int af, rt_setwarg_t *setwa_f, rt_walktree_f_t *wa_f, 967 void *arg) 968 { 969 struct rib_head *rnh; 970 uint32_t fibnum; 971 int i; 972 973 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) { 974 /* Do we want some specific family? */ 975 if (af != AF_UNSPEC) { 976 rnh = rt_tables_get_rnh(fibnum, af); 977 if (rnh == NULL) 978 continue; 979 if (setwa_f != NULL) 980 setwa_f(rnh, fibnum, af, arg); 981 982 RIB_WLOCK(rnh); 983 rnh->rnh_walktree(&rnh->head, (walktree_f_t *)wa_f,arg); 984 RIB_WUNLOCK(rnh); 985 continue; 986 } 987 988 for (i = 1; i <= AF_MAX; i++) { 989 rnh = rt_tables_get_rnh(fibnum, i); 990 if (rnh == NULL) 991 continue; 992 if (setwa_f != NULL) 993 setwa_f(rnh, fibnum, i, arg); 994 995 RIB_WLOCK(rnh); 996 rnh->rnh_walktree(&rnh->head, (walktree_f_t *)wa_f,arg); 997 RIB_WUNLOCK(rnh); 998 } 999 } 1000 } 1001 1002 struct rt_delinfo 1003 { 1004 struct rt_addrinfo info; 1005 struct rib_head *rnh; 1006 struct rtentry *head; 1007 }; 1008 1009 /* 1010 * Conditionally unlinks @rn from radix tree based 1011 * on info data passed in @arg. 1012 */ 1013 static int 1014 rt_checkdelroute(struct radix_node *rn, void *arg) 1015 { 1016 struct rt_delinfo *di; 1017 struct rt_addrinfo *info; 1018 struct rtentry *rt; 1019 int error; 1020 1021 di = (struct rt_delinfo *)arg; 1022 rt = (struct rtentry *)rn; 1023 info = &di->info; 1024 error = 0; 1025 1026 info->rti_info[RTAX_DST] = rt_key(rt); 1027 info->rti_info[RTAX_NETMASK] = rt_mask(rt); 1028 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway; 1029 1030 rt = rt_unlinkrte(di->rnh, info, &error); 1031 if (rt == NULL) { 1032 /* Either not allowed or not matched. Skip entry */ 1033 return (0); 1034 } 1035 1036 /* Entry was unlinked. Add to the list and return */ 1037 rt->rt_chain = di->head; 1038 di->head = rt; 1039 1040 return (0); 1041 } 1042 1043 /* 1044 * Iterates over all existing fibs in system. 1045 * Deletes each element for which @filter_f function returned 1046 * non-zero value. 1047 * If @af is not AF_UNSPEC, iterates over fibs in particular 1048 * address family. 1049 */ 1050 void 1051 rt_foreach_fib_walk_del(int af, rt_filter_f_t *filter_f, void *arg) 1052 { 1053 struct rib_head *rnh; 1054 struct rt_delinfo di; 1055 struct rtentry *rt; 1056 uint32_t fibnum; 1057 int i, start, end; 1058 1059 bzero(&di, sizeof(di)); 1060 di.info.rti_filter = filter_f; 1061 di.info.rti_filterdata = arg; 1062 1063 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) { 1064 /* Do we want some specific family? */ 1065 if (af != AF_UNSPEC) { 1066 start = af; 1067 end = af; 1068 } else { 1069 start = 1; 1070 end = AF_MAX; 1071 } 1072 1073 for (i = start; i <= end; i++) { 1074 rnh = rt_tables_get_rnh(fibnum, i); 1075 if (rnh == NULL) 1076 continue; 1077 di.rnh = rnh; 1078 1079 RIB_WLOCK(rnh); 1080 rnh->rnh_walktree(&rnh->head, rt_checkdelroute, &di); 1081 RIB_WUNLOCK(rnh); 1082 1083 if (di.head == NULL) 1084 continue; 1085 1086 /* We might have something to reclaim */ 1087 while (di.head != NULL) { 1088 rt = di.head; 1089 di.head = rt->rt_chain; 1090 rt->rt_chain = NULL; 1091 1092 /* TODO std rt -> rt_addrinfo export */ 1093 di.info.rti_info[RTAX_DST] = rt_key(rt); 1094 di.info.rti_info[RTAX_NETMASK] = rt_mask(rt); 1095 1096 rt_notifydelete(rt, &di.info); 1097 RTFREE_LOCKED(rt); 1098 } 1099 1100 } 1101 } 1102 } 1103 1104 /* 1105 * Delete Routes for a Network Interface 1106 * 1107 * Called for each routing entry via the rnh->rnh_walktree() call above 1108 * to delete all route entries referencing a detaching network interface. 1109 * 1110 * Arguments: 1111 * rt pointer to rtentry 1112 * arg argument passed to rnh->rnh_walktree() - detaching interface 1113 * 1114 * Returns: 1115 * 0 successful 1116 * errno failed - reason indicated 1117 */ 1118 static int 1119 rt_ifdelroute(const struct rtentry *rt, void *arg) 1120 { 1121 struct ifnet *ifp = arg; 1122 1123 if (rt->rt_ifp != ifp) 1124 return (0); 1125 1126 /* 1127 * Protect (sorta) against walktree recursion problems 1128 * with cloned routes 1129 */ 1130 if ((rt->rt_flags & RTF_UP) == 0) 1131 return (0); 1132 1133 return (1); 1134 } 1135 1136 /* 1137 * Delete all remaining routes using this interface 1138 * Unfortuneatly the only way to do this is to slog through 1139 * the entire routing table looking for routes which point 1140 * to this interface...oh well... 1141 */ 1142 void 1143 rt_flushifroutes_af(struct ifnet *ifp, int af) 1144 { 1145 KASSERT((af >= 1 && af <= AF_MAX), ("%s: af %d not >= 1 and <= %d", 1146 __func__, af, AF_MAX)); 1147 1148 rt_foreach_fib_walk_del(af, rt_ifdelroute, ifp); 1149 } 1150 1151 void 1152 rt_flushifroutes(struct ifnet *ifp) 1153 { 1154 1155 rt_foreach_fib_walk_del(AF_UNSPEC, rt_ifdelroute, ifp); 1156 } 1157 1158 /* 1159 * Conditionally unlinks rtentry matching data inside @info from @rnh. 1160 * Returns unlinked, locked and referenced @rtentry on success, 1161 * Returns NULL and sets @perror to: 1162 * ESRCH - if prefix was not found, 1163 * EADDRINUSE - if trying to delete PINNED route without appropriate flag. 1164 * ENOENT - if supplied filter function returned 0 (not matched). 1165 */ 1166 static struct rtentry * 1167 rt_unlinkrte(struct rib_head *rnh, struct rt_addrinfo *info, int *perror) 1168 { 1169 struct sockaddr *dst, *netmask; 1170 struct rtentry *rt; 1171 struct radix_node *rn; 1172 1173 dst = info->rti_info[RTAX_DST]; 1174 netmask = info->rti_info[RTAX_NETMASK]; 1175 1176 rt = (struct rtentry *)rnh->rnh_lookup(dst, netmask, &rnh->head); 1177 if (rt == NULL) { 1178 *perror = ESRCH; 1179 return (NULL); 1180 } 1181 1182 if ((info->rti_flags & RTF_PINNED) == 0) { 1183 /* Check if target route can be deleted */ 1184 if (rt->rt_flags & RTF_PINNED) { 1185 *perror = EADDRINUSE; 1186 return (NULL); 1187 } 1188 } 1189 1190 if (info->rti_filter != NULL) { 1191 if (info->rti_filter(rt, info->rti_filterdata) == 0) { 1192 /* Not matched */ 1193 *perror = ENOENT; 1194 return (NULL); 1195 } 1196 1197 /* 1198 * Filter function requested rte deletion. 1199 * Ease the caller work by filling in remaining info 1200 * from that particular entry. 1201 */ 1202 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway; 1203 } 1204 1205 /* 1206 * Remove the item from the tree and return it. 1207 * Complain if it is not there and do no more processing. 1208 */ 1209 *perror = ESRCH; 1210 #ifdef RADIX_MPATH 1211 if (rt_mpath_capable(rnh)) 1212 rn = rt_mpath_unlink(rnh, info, rt, perror); 1213 else 1214 #endif 1215 rn = rnh->rnh_deladdr(dst, netmask, &rnh->head); 1216 if (rn == NULL) 1217 return (NULL); 1218 1219 if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT)) 1220 panic ("rtrequest delete"); 1221 1222 rt = RNTORT(rn); 1223 RT_LOCK(rt); 1224 RT_ADDREF(rt); 1225 rt->rt_flags &= ~RTF_UP; 1226 1227 *perror = 0; 1228 1229 return (rt); 1230 } 1231 1232 static void 1233 rt_notifydelete(struct rtentry *rt, struct rt_addrinfo *info) 1234 { 1235 struct ifaddr *ifa; 1236 1237 /* 1238 * give the protocol a chance to keep things in sync. 1239 */ 1240 ifa = rt->rt_ifa; 1241 if (ifa != NULL && ifa->ifa_rtrequest != NULL) 1242 ifa->ifa_rtrequest(RTM_DELETE, rt, info); 1243 1244 /* 1245 * One more rtentry floating around that is not 1246 * linked to the routing table. rttrash will be decremented 1247 * when RTFREE(rt) is eventually called. 1248 */ 1249 V_rttrash++; 1250 } 1251 1252 1253 /* 1254 * These (questionable) definitions of apparent local variables apply 1255 * to the next two functions. XXXXXX!!! 1256 */ 1257 #define dst info->rti_info[RTAX_DST] 1258 #define gateway info->rti_info[RTAX_GATEWAY] 1259 #define netmask info->rti_info[RTAX_NETMASK] 1260 #define ifaaddr info->rti_info[RTAX_IFA] 1261 #define ifpaddr info->rti_info[RTAX_IFP] 1262 #define flags info->rti_flags 1263 1264 /* 1265 * Look up rt_addrinfo for a specific fib. Note that if rti_ifa is defined, 1266 * it will be referenced so the caller must free it. 1267 */ 1268 int 1269 rt_getifa_fib(struct rt_addrinfo *info, u_int fibnum) 1270 { 1271 struct ifaddr *ifa; 1272 int error = 0; 1273 1274 /* 1275 * ifp may be specified by sockaddr_dl 1276 * when protocol address is ambiguous. 1277 */ 1278 if (info->rti_ifp == NULL && ifpaddr != NULL && 1279 ifpaddr->sa_family == AF_LINK && 1280 (ifa = ifa_ifwithnet(ifpaddr, 0, fibnum)) != NULL) { 1281 info->rti_ifp = ifa->ifa_ifp; 1282 ifa_free(ifa); 1283 } 1284 if (info->rti_ifa == NULL && ifaaddr != NULL) 1285 info->rti_ifa = ifa_ifwithaddr(ifaaddr); 1286 if (info->rti_ifa == NULL) { 1287 struct sockaddr *sa; 1288 1289 sa = ifaaddr != NULL ? ifaaddr : 1290 (gateway != NULL ? gateway : dst); 1291 if (sa != NULL && info->rti_ifp != NULL) 1292 info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp); 1293 else if (dst != NULL && gateway != NULL) 1294 info->rti_ifa = ifa_ifwithroute(flags, dst, gateway, 1295 fibnum); 1296 else if (sa != NULL) 1297 info->rti_ifa = ifa_ifwithroute(flags, sa, sa, 1298 fibnum); 1299 } 1300 if ((ifa = info->rti_ifa) != NULL) { 1301 if (info->rti_ifp == NULL) 1302 info->rti_ifp = ifa->ifa_ifp; 1303 } else 1304 error = ENETUNREACH; 1305 return (error); 1306 } 1307 1308 static int 1309 if_updatemtu_cb(struct radix_node *rn, void *arg) 1310 { 1311 struct rtentry *rt; 1312 struct if_mtuinfo *ifmtu; 1313 1314 rt = (struct rtentry *)rn; 1315 ifmtu = (struct if_mtuinfo *)arg; 1316 1317 if (rt->rt_ifp != ifmtu->ifp) 1318 return (0); 1319 1320 if (rt->rt_mtu >= ifmtu->mtu) { 1321 /* We have to decrease mtu regardless of flags */ 1322 rt->rt_mtu = ifmtu->mtu; 1323 return (0); 1324 } 1325 1326 /* 1327 * New MTU is bigger. Check if are allowed to alter it 1328 */ 1329 if ((rt->rt_flags & (RTF_FIXEDMTU | RTF_GATEWAY | RTF_HOST)) != 0) { 1330 1331 /* 1332 * Skip routes with user-supplied MTU and 1333 * non-interface routes 1334 */ 1335 return (0); 1336 } 1337 1338 /* We are safe to update route MTU */ 1339 rt->rt_mtu = ifmtu->mtu; 1340 1341 return (0); 1342 } 1343 1344 void 1345 rt_updatemtu(struct ifnet *ifp) 1346 { 1347 struct if_mtuinfo ifmtu; 1348 struct rib_head *rnh; 1349 int i, j; 1350 1351 ifmtu.ifp = ifp; 1352 1353 /* 1354 * Try to update rt_mtu for all routes using this interface 1355 * Unfortunately the only way to do this is to traverse all 1356 * routing tables in all fibs/domains. 1357 */ 1358 for (i = 1; i <= AF_MAX; i++) { 1359 ifmtu.mtu = if_getmtu_family(ifp, i); 1360 for (j = 0; j < rt_numfibs; j++) { 1361 rnh = rt_tables_get_rnh(j, i); 1362 if (rnh == NULL) 1363 continue; 1364 RIB_WLOCK(rnh); 1365 rnh->rnh_walktree(&rnh->head, if_updatemtu_cb, &ifmtu); 1366 RIB_WUNLOCK(rnh); 1367 } 1368 } 1369 } 1370 1371 1372 #if 0 1373 int p_sockaddr(char *buf, int buflen, struct sockaddr *s); 1374 int rt_print(char *buf, int buflen, struct rtentry *rt); 1375 1376 int 1377 p_sockaddr(char *buf, int buflen, struct sockaddr *s) 1378 { 1379 void *paddr = NULL; 1380 1381 switch (s->sa_family) { 1382 case AF_INET: 1383 paddr = &((struct sockaddr_in *)s)->sin_addr; 1384 break; 1385 case AF_INET6: 1386 paddr = &((struct sockaddr_in6 *)s)->sin6_addr; 1387 break; 1388 } 1389 1390 if (paddr == NULL) 1391 return (0); 1392 1393 if (inet_ntop(s->sa_family, paddr, buf, buflen) == NULL) 1394 return (0); 1395 1396 return (strlen(buf)); 1397 } 1398 1399 int 1400 rt_print(char *buf, int buflen, struct rtentry *rt) 1401 { 1402 struct sockaddr *addr, *mask; 1403 int i = 0; 1404 1405 addr = rt_key(rt); 1406 mask = rt_mask(rt); 1407 1408 i = p_sockaddr(buf, buflen, addr); 1409 if (!(rt->rt_flags & RTF_HOST)) { 1410 buf[i++] = '/'; 1411 i += p_sockaddr(buf + i, buflen - i, mask); 1412 } 1413 1414 if (rt->rt_flags & RTF_GATEWAY) { 1415 buf[i++] = '>'; 1416 i += p_sockaddr(buf + i, buflen - i, rt->rt_gateway); 1417 } 1418 1419 return (i); 1420 } 1421 #endif 1422 1423 #ifdef RADIX_MPATH 1424 /* 1425 * Deletes key for single-path routes, unlinks rtentry with 1426 * gateway specified in @info from multi-path routes. 1427 * 1428 * Returnes unlinked entry. In case of failure, returns NULL 1429 * and sets @perror to ESRCH. 1430 */ 1431 static struct radix_node * 1432 rt_mpath_unlink(struct rib_head *rnh, struct rt_addrinfo *info, 1433 struct rtentry *rto, int *perror) 1434 { 1435 /* 1436 * if we got multipath routes, we require users to specify 1437 * a matching RTAX_GATEWAY. 1438 */ 1439 struct rtentry *rt; // *rto = NULL; 1440 struct radix_node *rn; 1441 struct sockaddr *gw; 1442 1443 gw = info->rti_info[RTAX_GATEWAY]; 1444 rt = rt_mpath_matchgate(rto, gw); 1445 if (rt == NULL) { 1446 *perror = ESRCH; 1447 return (NULL); 1448 } 1449 1450 /* 1451 * this is the first entry in the chain 1452 */ 1453 if (rto == rt) { 1454 rn = rn_mpath_next((struct radix_node *)rt); 1455 /* 1456 * there is another entry, now it's active 1457 */ 1458 if (rn) { 1459 rto = RNTORT(rn); 1460 RT_LOCK(rto); 1461 rto->rt_flags |= RTF_UP; 1462 RT_UNLOCK(rto); 1463 } else if (rt->rt_flags & RTF_GATEWAY) { 1464 /* 1465 * For gateway routes, we need to 1466 * make sure that we we are deleting 1467 * the correct gateway. 1468 * rt_mpath_matchgate() does not 1469 * check the case when there is only 1470 * one route in the chain. 1471 */ 1472 if (gw && 1473 (rt->rt_gateway->sa_len != gw->sa_len || 1474 memcmp(rt->rt_gateway, gw, gw->sa_len))) { 1475 *perror = ESRCH; 1476 return (NULL); 1477 } 1478 } 1479 1480 /* 1481 * use the normal delete code to remove 1482 * the first entry 1483 */ 1484 rn = rnh->rnh_deladdr(dst, netmask, &rnh->head); 1485 *perror = 0; 1486 return (rn); 1487 } 1488 1489 /* 1490 * if the entry is 2nd and on up 1491 */ 1492 if (rt_mpath_deldup(rto, rt) == 0) 1493 panic ("rtrequest1: rt_mpath_deldup"); 1494 *perror = 0; 1495 rn = (struct radix_node *)rt; 1496 return (rn); 1497 } 1498 #endif 1499 1500 #ifdef FLOWTABLE 1501 static struct rtentry * 1502 rt_flowtable_check_route(struct rib_head *rnh, struct rt_addrinfo *info) 1503 { 1504 #if defined(INET6) || defined(INET) 1505 struct radix_node *rn; 1506 #endif 1507 struct rtentry *rt0; 1508 1509 rt0 = NULL; 1510 /* "flow-table" only supports IPv6 and IPv4 at the moment. */ 1511 switch (dst->sa_family) { 1512 #ifdef INET6 1513 case AF_INET6: 1514 #endif 1515 #ifdef INET 1516 case AF_INET: 1517 #endif 1518 #if defined(INET6) || defined(INET) 1519 rn = rnh->rnh_matchaddr(dst, &rnh->head); 1520 if (rn && ((rn->rn_flags & RNF_ROOT) == 0)) { 1521 struct sockaddr *mask; 1522 u_char *m, *n; 1523 int len; 1524 1525 /* 1526 * compare mask to see if the new route is 1527 * more specific than the existing one 1528 */ 1529 rt0 = RNTORT(rn); 1530 RT_LOCK(rt0); 1531 RT_ADDREF(rt0); 1532 RT_UNLOCK(rt0); 1533 /* 1534 * A host route is already present, so 1535 * leave the flow-table entries as is. 1536 */ 1537 if (rt0->rt_flags & RTF_HOST) { 1538 RTFREE(rt0); 1539 rt0 = NULL; 1540 } else if (!(flags & RTF_HOST) && netmask) { 1541 mask = rt_mask(rt0); 1542 len = mask->sa_len; 1543 m = (u_char *)mask; 1544 n = (u_char *)netmask; 1545 while (len-- > 0) { 1546 if (*n != *m) 1547 break; 1548 n++; 1549 m++; 1550 } 1551 if (len == 0 || (*n < *m)) { 1552 RTFREE(rt0); 1553 rt0 = NULL; 1554 } 1555 } 1556 } 1557 #endif/* INET6 || INET */ 1558 } 1559 1560 return (rt0); 1561 } 1562 #endif 1563 1564 int 1565 rtrequest1_fib(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt, 1566 u_int fibnum) 1567 { 1568 int error = 0; 1569 struct rtentry *rt, *rt_old; 1570 #ifdef FLOWTABLE 1571 struct rtentry *rt0; 1572 #endif 1573 struct radix_node *rn; 1574 struct rib_head *rnh; 1575 struct ifaddr *ifa; 1576 struct sockaddr *ndst; 1577 struct sockaddr_storage mdst; 1578 1579 KASSERT((fibnum < rt_numfibs), ("rtrequest1_fib: bad fibnum")); 1580 KASSERT((flags & RTF_RNH_LOCKED) == 0, ("rtrequest1_fib: locked")); 1581 switch (dst->sa_family) { 1582 case AF_INET6: 1583 case AF_INET: 1584 /* We support multiple FIBs. */ 1585 break; 1586 default: 1587 fibnum = RT_DEFAULT_FIB; 1588 break; 1589 } 1590 1591 /* 1592 * Find the correct routing tree to use for this Address Family 1593 */ 1594 rnh = rt_tables_get_rnh(fibnum, dst->sa_family); 1595 if (rnh == NULL) 1596 return (EAFNOSUPPORT); 1597 1598 /* 1599 * If we are adding a host route then we don't want to put 1600 * a netmask in the tree, nor do we want to clone it. 1601 */ 1602 if (flags & RTF_HOST) 1603 netmask = NULL; 1604 1605 switch (req) { 1606 case RTM_DELETE: 1607 if (netmask) { 1608 rt_maskedcopy(dst, (struct sockaddr *)&mdst, netmask); 1609 dst = (struct sockaddr *)&mdst; 1610 } 1611 1612 RIB_WLOCK(rnh); 1613 rt = rt_unlinkrte(rnh, info, &error); 1614 RIB_WUNLOCK(rnh); 1615 if (error != 0) 1616 return (error); 1617 1618 rt_notifydelete(rt, info); 1619 1620 /* 1621 * If the caller wants it, then it can have it, 1622 * but it's up to it to free the rtentry as we won't be 1623 * doing it. 1624 */ 1625 if (ret_nrt) { 1626 *ret_nrt = rt; 1627 RT_UNLOCK(rt); 1628 } else 1629 RTFREE_LOCKED(rt); 1630 break; 1631 case RTM_RESOLVE: 1632 /* 1633 * resolve was only used for route cloning 1634 * here for compat 1635 */ 1636 break; 1637 case RTM_ADD: 1638 if ((flags & RTF_GATEWAY) && !gateway) 1639 return (EINVAL); 1640 if (dst && gateway && (dst->sa_family != gateway->sa_family) && 1641 (gateway->sa_family != AF_UNSPEC) && (gateway->sa_family != AF_LINK)) 1642 return (EINVAL); 1643 1644 if (info->rti_ifa == NULL) { 1645 error = rt_getifa_fib(info, fibnum); 1646 if (error) 1647 return (error); 1648 } else 1649 ifa_ref(info->rti_ifa); 1650 ifa = info->rti_ifa; 1651 rt = uma_zalloc(V_rtzone, M_NOWAIT); 1652 if (rt == NULL) { 1653 ifa_free(ifa); 1654 return (ENOBUFS); 1655 } 1656 rt->rt_flags = RTF_UP | flags; 1657 rt->rt_fibnum = fibnum; 1658 /* 1659 * Add the gateway. Possibly re-malloc-ing the storage for it. 1660 */ 1661 if ((error = rt_setgate(rt, dst, gateway)) != 0) { 1662 ifa_free(ifa); 1663 uma_zfree(V_rtzone, rt); 1664 return (error); 1665 } 1666 1667 /* 1668 * point to the (possibly newly malloc'd) dest address. 1669 */ 1670 ndst = (struct sockaddr *)rt_key(rt); 1671 1672 /* 1673 * make sure it contains the value we want (masked if needed). 1674 */ 1675 if (netmask) { 1676 rt_maskedcopy(dst, ndst, netmask); 1677 } else 1678 bcopy(dst, ndst, dst->sa_len); 1679 1680 /* 1681 * We use the ifa reference returned by rt_getifa_fib(). 1682 * This moved from below so that rnh->rnh_addaddr() can 1683 * examine the ifa and ifa->ifa_ifp if it so desires. 1684 */ 1685 rt->rt_ifa = ifa; 1686 rt->rt_ifp = ifa->ifa_ifp; 1687 rt->rt_weight = 1; 1688 1689 rt_setmetrics(info, rt); 1690 1691 RIB_WLOCK(rnh); 1692 RT_LOCK(rt); 1693 #ifdef RADIX_MPATH 1694 /* do not permit exactly the same dst/mask/gw pair */ 1695 if (rt_mpath_capable(rnh) && 1696 rt_mpath_conflict(rnh, rt, netmask)) { 1697 RIB_WUNLOCK(rnh); 1698 1699 ifa_free(rt->rt_ifa); 1700 R_Free(rt_key(rt)); 1701 uma_zfree(V_rtzone, rt); 1702 return (EEXIST); 1703 } 1704 #endif 1705 1706 #ifdef FLOWTABLE 1707 rt0 = rt_flowtable_check_route(rnh, info); 1708 #endif /* FLOWTABLE */ 1709 1710 /* XXX mtu manipulation will be done in rnh_addaddr -- itojun */ 1711 rn = rnh->rnh_addaddr(ndst, netmask, &rnh->head, rt->rt_nodes); 1712 1713 rt_old = NULL; 1714 if (rn == NULL && (info->rti_flags & RTF_PINNED) != 0) { 1715 1716 /* 1717 * Force removal and re-try addition 1718 * TODO: better multipath&pinned support 1719 */ 1720 struct sockaddr *info_dst = info->rti_info[RTAX_DST]; 1721 info->rti_info[RTAX_DST] = ndst; 1722 /* Do not delete existing PINNED(interface) routes */ 1723 info->rti_flags &= ~RTF_PINNED; 1724 rt_old = rt_unlinkrte(rnh, info, &error); 1725 info->rti_flags |= RTF_PINNED; 1726 info->rti_info[RTAX_DST] = info_dst; 1727 if (rt_old != NULL) 1728 rn = rnh->rnh_addaddr(ndst, netmask, &rnh->head, 1729 rt->rt_nodes); 1730 } 1731 RIB_WUNLOCK(rnh); 1732 1733 if (rt_old != NULL) 1734 RT_UNLOCK(rt_old); 1735 1736 /* 1737 * If it still failed to go into the tree, 1738 * then un-make it (this should be a function) 1739 */ 1740 if (rn == NULL) { 1741 ifa_free(rt->rt_ifa); 1742 R_Free(rt_key(rt)); 1743 uma_zfree(V_rtzone, rt); 1744 #ifdef FLOWTABLE 1745 if (rt0 != NULL) 1746 RTFREE(rt0); 1747 #endif 1748 return (EEXIST); 1749 } 1750 #ifdef FLOWTABLE 1751 else if (rt0 != NULL) { 1752 flowtable_route_flush(dst->sa_family, rt0); 1753 RTFREE(rt0); 1754 } 1755 #endif 1756 1757 if (rt_old != NULL) { 1758 rt_notifydelete(rt_old, info); 1759 RTFREE(rt_old); 1760 } 1761 1762 /* 1763 * If this protocol has something to add to this then 1764 * allow it to do that as well. 1765 */ 1766 if (ifa->ifa_rtrequest) 1767 ifa->ifa_rtrequest(req, rt, info); 1768 1769 /* 1770 * actually return a resultant rtentry and 1771 * give the caller a single reference. 1772 */ 1773 if (ret_nrt) { 1774 *ret_nrt = rt; 1775 RT_ADDREF(rt); 1776 } 1777 rnh->rnh_gen++; /* Routing table updated */ 1778 RT_UNLOCK(rt); 1779 break; 1780 case RTM_CHANGE: 1781 RIB_WLOCK(rnh); 1782 error = rtrequest1_fib_change(rnh, info, ret_nrt, fibnum); 1783 RIB_WUNLOCK(rnh); 1784 break; 1785 default: 1786 error = EOPNOTSUPP; 1787 } 1788 1789 return (error); 1790 } 1791 1792 #undef dst 1793 #undef gateway 1794 #undef netmask 1795 #undef ifaaddr 1796 #undef ifpaddr 1797 #undef flags 1798 1799 static int 1800 rtrequest1_fib_change(struct rib_head *rnh, struct rt_addrinfo *info, 1801 struct rtentry **ret_nrt, u_int fibnum) 1802 { 1803 struct rtentry *rt = NULL; 1804 int error = 0; 1805 int free_ifa = 0; 1806 int family, mtu; 1807 struct if_mtuinfo ifmtu; 1808 1809 rt = (struct rtentry *)rnh->rnh_lookup(info->rti_info[RTAX_DST], 1810 info->rti_info[RTAX_NETMASK], &rnh->head); 1811 1812 if (rt == NULL) 1813 return (ESRCH); 1814 1815 #ifdef RADIX_MPATH 1816 /* 1817 * If we got multipath routes, 1818 * we require users to specify a matching RTAX_GATEWAY. 1819 */ 1820 if (rt_mpath_capable(rnh)) { 1821 rt = rt_mpath_matchgate(rt, info->rti_info[RTAX_GATEWAY]); 1822 if (rt == NULL) 1823 return (ESRCH); 1824 } 1825 #endif 1826 1827 RT_LOCK(rt); 1828 1829 rt_setmetrics(info, rt); 1830 1831 /* 1832 * New gateway could require new ifaddr, ifp; 1833 * flags may also be different; ifp may be specified 1834 * by ll sockaddr when protocol address is ambiguous 1835 */ 1836 if (((rt->rt_flags & RTF_GATEWAY) && 1837 info->rti_info[RTAX_GATEWAY] != NULL) || 1838 info->rti_info[RTAX_IFP] != NULL || 1839 (info->rti_info[RTAX_IFA] != NULL && 1840 !sa_equal(info->rti_info[RTAX_IFA], rt->rt_ifa->ifa_addr))) { 1841 1842 error = rt_getifa_fib(info, fibnum); 1843 if (info->rti_ifa != NULL) 1844 free_ifa = 1; 1845 1846 if (error != 0) 1847 goto bad; 1848 } 1849 1850 /* Check if outgoing interface has changed */ 1851 if (info->rti_ifa != NULL && info->rti_ifa != rt->rt_ifa && 1852 rt->rt_ifa != NULL && rt->rt_ifa->ifa_rtrequest != NULL) { 1853 rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt, info); 1854 ifa_free(rt->rt_ifa); 1855 } 1856 /* Update gateway address */ 1857 if (info->rti_info[RTAX_GATEWAY] != NULL) { 1858 error = rt_setgate(rt, rt_key(rt), info->rti_info[RTAX_GATEWAY]); 1859 if (error != 0) 1860 goto bad; 1861 1862 rt->rt_flags &= ~RTF_GATEWAY; 1863 rt->rt_flags |= (RTF_GATEWAY & info->rti_flags); 1864 } 1865 1866 if (info->rti_ifa != NULL && info->rti_ifa != rt->rt_ifa) { 1867 ifa_ref(info->rti_ifa); 1868 rt->rt_ifa = info->rti_ifa; 1869 rt->rt_ifp = info->rti_ifp; 1870 } 1871 /* Allow some flags to be toggled on change. */ 1872 rt->rt_flags &= ~RTF_FMASK; 1873 rt->rt_flags |= info->rti_flags & RTF_FMASK; 1874 1875 if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest != NULL) 1876 rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, info); 1877 1878 /* Alter route MTU if necessary */ 1879 if (rt->rt_ifp != NULL) { 1880 family = info->rti_info[RTAX_DST]->sa_family; 1881 mtu = if_getmtu_family(rt->rt_ifp, family); 1882 /* Set default MTU */ 1883 if (rt->rt_mtu == 0) 1884 rt->rt_mtu = mtu; 1885 if (rt->rt_mtu != mtu) { 1886 /* Check if we really need to update */ 1887 ifmtu.ifp = rt->rt_ifp; 1888 ifmtu.mtu = mtu; 1889 if_updatemtu_cb(rt->rt_nodes, &ifmtu); 1890 } 1891 } 1892 1893 if (ret_nrt) { 1894 *ret_nrt = rt; 1895 RT_ADDREF(rt); 1896 } 1897 bad: 1898 RT_UNLOCK(rt); 1899 if (free_ifa != 0) 1900 ifa_free(info->rti_ifa); 1901 return (error); 1902 } 1903 1904 static void 1905 rt_setmetrics(const struct rt_addrinfo *info, struct rtentry *rt) 1906 { 1907 1908 if (info->rti_mflags & RTV_MTU) { 1909 if (info->rti_rmx->rmx_mtu != 0) { 1910 1911 /* 1912 * MTU was explicitly provided by user. 1913 * Keep it. 1914 */ 1915 rt->rt_flags |= RTF_FIXEDMTU; 1916 } else { 1917 1918 /* 1919 * User explicitly sets MTU to 0. 1920 * Assume rollback to default. 1921 */ 1922 rt->rt_flags &= ~RTF_FIXEDMTU; 1923 } 1924 rt->rt_mtu = info->rti_rmx->rmx_mtu; 1925 } 1926 if (info->rti_mflags & RTV_WEIGHT) 1927 rt->rt_weight = info->rti_rmx->rmx_weight; 1928 /* Kernel -> userland timebase conversion. */ 1929 if (info->rti_mflags & RTV_EXPIRE) 1930 rt->rt_expire = info->rti_rmx->rmx_expire ? 1931 info->rti_rmx->rmx_expire - time_second + time_uptime : 0; 1932 } 1933 1934 int 1935 rt_setgate(struct rtentry *rt, struct sockaddr *dst, struct sockaddr *gate) 1936 { 1937 /* XXX dst may be overwritten, can we move this to below */ 1938 int dlen = SA_SIZE(dst), glen = SA_SIZE(gate); 1939 1940 /* 1941 * Prepare to store the gateway in rt->rt_gateway. 1942 * Both dst and gateway are stored one after the other in the same 1943 * malloc'd chunk. If we have room, we can reuse the old buffer, 1944 * rt_gateway already points to the right place. 1945 * Otherwise, malloc a new block and update the 'dst' address. 1946 */ 1947 if (rt->rt_gateway == NULL || glen > SA_SIZE(rt->rt_gateway)) { 1948 caddr_t new; 1949 1950 R_Malloc(new, caddr_t, dlen + glen); 1951 if (new == NULL) 1952 return ENOBUFS; 1953 /* 1954 * XXX note, we copy from *dst and not *rt_key(rt) because 1955 * rt_setgate() can be called to initialize a newly 1956 * allocated route entry, in which case rt_key(rt) == NULL 1957 * (and also rt->rt_gateway == NULL). 1958 * Free()/free() handle a NULL argument just fine. 1959 */ 1960 bcopy(dst, new, dlen); 1961 R_Free(rt_key(rt)); /* free old block, if any */ 1962 rt_key(rt) = (struct sockaddr *)new; 1963 rt->rt_gateway = (struct sockaddr *)(new + dlen); 1964 } 1965 1966 /* 1967 * Copy the new gateway value into the memory chunk. 1968 */ 1969 bcopy(gate, rt->rt_gateway, glen); 1970 1971 return (0); 1972 } 1973 1974 void 1975 rt_maskedcopy(struct sockaddr *src, struct sockaddr *dst, struct sockaddr *netmask) 1976 { 1977 u_char *cp1 = (u_char *)src; 1978 u_char *cp2 = (u_char *)dst; 1979 u_char *cp3 = (u_char *)netmask; 1980 u_char *cplim = cp2 + *cp3; 1981 u_char *cplim2 = cp2 + *cp1; 1982 1983 *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */ 1984 cp3 += 2; 1985 if (cplim > cplim2) 1986 cplim = cplim2; 1987 while (cp2 < cplim) 1988 *cp2++ = *cp1++ & *cp3++; 1989 if (cp2 < cplim2) 1990 bzero((caddr_t)cp2, (unsigned)(cplim2 - cp2)); 1991 } 1992 1993 /* 1994 * Set up a routing table entry, normally 1995 * for an interface. 1996 */ 1997 #define _SOCKADDR_TMPSIZE 128 /* Not too big.. kernel stack size is limited */ 1998 static inline int 1999 rtinit1(struct ifaddr *ifa, int cmd, int flags, int fibnum) 2000 { 2001 struct sockaddr *dst; 2002 struct sockaddr *netmask; 2003 struct rtentry *rt = NULL; 2004 struct rt_addrinfo info; 2005 int error = 0; 2006 int startfib, endfib; 2007 char tempbuf[_SOCKADDR_TMPSIZE]; 2008 int didwork = 0; 2009 int a_failure = 0; 2010 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; 2011 struct rib_head *rnh; 2012 2013 if (flags & RTF_HOST) { 2014 dst = ifa->ifa_dstaddr; 2015 netmask = NULL; 2016 } else { 2017 dst = ifa->ifa_addr; 2018 netmask = ifa->ifa_netmask; 2019 } 2020 if (dst->sa_len == 0) 2021 return(EINVAL); 2022 switch (dst->sa_family) { 2023 case AF_INET6: 2024 case AF_INET: 2025 /* We support multiple FIBs. */ 2026 break; 2027 default: 2028 fibnum = RT_DEFAULT_FIB; 2029 break; 2030 } 2031 if (fibnum == RT_ALL_FIBS) { 2032 if (V_rt_add_addr_allfibs == 0 && cmd == (int)RTM_ADD) 2033 startfib = endfib = ifa->ifa_ifp->if_fib; 2034 else { 2035 startfib = 0; 2036 endfib = rt_numfibs - 1; 2037 } 2038 } else { 2039 KASSERT((fibnum < rt_numfibs), ("rtinit1: bad fibnum")); 2040 startfib = fibnum; 2041 endfib = fibnum; 2042 } 2043 2044 /* 2045 * If it's a delete, check that if it exists, 2046 * it's on the correct interface or we might scrub 2047 * a route to another ifa which would 2048 * be confusing at best and possibly worse. 2049 */ 2050 if (cmd == RTM_DELETE) { 2051 /* 2052 * It's a delete, so it should already exist.. 2053 * If it's a net, mask off the host bits 2054 * (Assuming we have a mask) 2055 * XXX this is kinda inet specific.. 2056 */ 2057 if (netmask != NULL) { 2058 rt_maskedcopy(dst, (struct sockaddr *)tempbuf, netmask); 2059 dst = (struct sockaddr *)tempbuf; 2060 } 2061 } 2062 /* 2063 * Now go through all the requested tables (fibs) and do the 2064 * requested action. Realistically, this will either be fib 0 2065 * for protocols that don't do multiple tables or all the 2066 * tables for those that do. 2067 */ 2068 for ( fibnum = startfib; fibnum <= endfib; fibnum++) { 2069 if (cmd == RTM_DELETE) { 2070 struct radix_node *rn; 2071 /* 2072 * Look up an rtentry that is in the routing tree and 2073 * contains the correct info. 2074 */ 2075 rnh = rt_tables_get_rnh(fibnum, dst->sa_family); 2076 if (rnh == NULL) 2077 /* this table doesn't exist but others might */ 2078 continue; 2079 RIB_RLOCK(rnh); 2080 rn = rnh->rnh_lookup(dst, netmask, &rnh->head); 2081 #ifdef RADIX_MPATH 2082 if (rt_mpath_capable(rnh)) { 2083 2084 if (rn == NULL) 2085 error = ESRCH; 2086 else { 2087 rt = RNTORT(rn); 2088 /* 2089 * for interface route the 2090 * rt->rt_gateway is sockaddr_intf 2091 * for cloning ARP entries, so 2092 * rt_mpath_matchgate must use the 2093 * interface address 2094 */ 2095 rt = rt_mpath_matchgate(rt, 2096 ifa->ifa_addr); 2097 if (rt == NULL) 2098 error = ESRCH; 2099 } 2100 } 2101 #endif 2102 error = (rn == NULL || 2103 (rn->rn_flags & RNF_ROOT) || 2104 RNTORT(rn)->rt_ifa != ifa); 2105 RIB_RUNLOCK(rnh); 2106 if (error) { 2107 /* this is only an error if bad on ALL tables */ 2108 continue; 2109 } 2110 } 2111 /* 2112 * Do the actual request 2113 */ 2114 bzero((caddr_t)&info, sizeof(info)); 2115 info.rti_ifa = ifa; 2116 info.rti_flags = flags | 2117 (ifa->ifa_flags & ~IFA_RTSELF) | RTF_PINNED; 2118 info.rti_info[RTAX_DST] = dst; 2119 /* 2120 * doing this for compatibility reasons 2121 */ 2122 if (cmd == RTM_ADD) 2123 info.rti_info[RTAX_GATEWAY] = 2124 (struct sockaddr *)&null_sdl; 2125 else 2126 info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr; 2127 info.rti_info[RTAX_NETMASK] = netmask; 2128 error = rtrequest1_fib(cmd, &info, &rt, fibnum); 2129 2130 if (error == 0 && rt != NULL) { 2131 /* 2132 * notify any listening routing agents of the change 2133 */ 2134 RT_LOCK(rt); 2135 #ifdef RADIX_MPATH 2136 /* 2137 * in case address alias finds the first address 2138 * e.g. ifconfig bge0 192.0.2.246/24 2139 * e.g. ifconfig bge0 192.0.2.247/24 2140 * the address set in the route is 192.0.2.246 2141 * so we need to replace it with 192.0.2.247 2142 */ 2143 if (memcmp(rt->rt_ifa->ifa_addr, 2144 ifa->ifa_addr, ifa->ifa_addr->sa_len)) { 2145 ifa_free(rt->rt_ifa); 2146 ifa_ref(ifa); 2147 rt->rt_ifp = ifa->ifa_ifp; 2148 rt->rt_ifa = ifa; 2149 } 2150 #endif 2151 /* 2152 * doing this for compatibility reasons 2153 */ 2154 if (cmd == RTM_ADD) { 2155 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type = 2156 rt->rt_ifp->if_type; 2157 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index = 2158 rt->rt_ifp->if_index; 2159 } 2160 RT_ADDREF(rt); 2161 RT_UNLOCK(rt); 2162 rt_newaddrmsg_fib(cmd, ifa, error, rt, fibnum); 2163 RT_LOCK(rt); 2164 RT_REMREF(rt); 2165 if (cmd == RTM_DELETE) { 2166 /* 2167 * If we are deleting, and we found an entry, 2168 * then it's been removed from the tree.. 2169 * now throw it away. 2170 */ 2171 RTFREE_LOCKED(rt); 2172 } else { 2173 if (cmd == RTM_ADD) { 2174 /* 2175 * We just wanted to add it.. 2176 * we don't actually need a reference. 2177 */ 2178 RT_REMREF(rt); 2179 } 2180 RT_UNLOCK(rt); 2181 } 2182 didwork = 1; 2183 } 2184 if (error) 2185 a_failure = error; 2186 } 2187 if (cmd == RTM_DELETE) { 2188 if (didwork) { 2189 error = 0; 2190 } else { 2191 /* we only give an error if it wasn't in any table */ 2192 error = ((flags & RTF_HOST) ? 2193 EHOSTUNREACH : ENETUNREACH); 2194 } 2195 } else { 2196 if (a_failure) { 2197 /* return an error if any of them failed */ 2198 error = a_failure; 2199 } 2200 } 2201 return (error); 2202 } 2203 2204 /* 2205 * Set up a routing table entry, normally 2206 * for an interface. 2207 */ 2208 int 2209 rtinit(struct ifaddr *ifa, int cmd, int flags) 2210 { 2211 struct sockaddr *dst; 2212 int fib = RT_DEFAULT_FIB; 2213 2214 if (flags & RTF_HOST) { 2215 dst = ifa->ifa_dstaddr; 2216 } else { 2217 dst = ifa->ifa_addr; 2218 } 2219 2220 switch (dst->sa_family) { 2221 case AF_INET6: 2222 case AF_INET: 2223 /* We do support multiple FIBs. */ 2224 fib = RT_ALL_FIBS; 2225 break; 2226 } 2227 return (rtinit1(ifa, cmd, flags, fib)); 2228 } 2229 2230 /* 2231 * Announce interface address arrival/withdraw 2232 * Returns 0 on success. 2233 */ 2234 int 2235 rt_addrmsg(int cmd, struct ifaddr *ifa, int fibnum) 2236 { 2237 2238 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE, 2239 ("unexpected cmd %d", cmd)); 2240 2241 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs), 2242 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs)); 2243 2244 #if defined(INET) || defined(INET6) 2245 #ifdef SCTP 2246 /* 2247 * notify the SCTP stack 2248 * this will only get called when an address is added/deleted 2249 * XXX pass the ifaddr struct instead if ifa->ifa_addr... 2250 */ 2251 sctp_addr_change(ifa, cmd); 2252 #endif /* SCTP */ 2253 #endif 2254 return (rtsock_addrmsg(cmd, ifa, fibnum)); 2255 } 2256 2257 /* 2258 * Announce route addition/removal. 2259 * Users of this function MUST validate input data BEFORE calling. 2260 * However we have to be able to handle invalid data: 2261 * if some userland app sends us "invalid" route message (invalid mask, 2262 * no dst, wrong address families, etc...) we need to pass it back 2263 * to app (and any other rtsock consumers) with rtm_errno field set to 2264 * non-zero value. 2265 * Returns 0 on success. 2266 */ 2267 int 2268 rt_routemsg(int cmd, struct ifnet *ifp, int error, struct rtentry *rt, 2269 int fibnum) 2270 { 2271 2272 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE, 2273 ("unexpected cmd %d", cmd)); 2274 2275 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs), 2276 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs)); 2277 2278 KASSERT(rt_key(rt) != NULL, (":%s: rt_key must be supplied", __func__)); 2279 2280 return (rtsock_routemsg(cmd, ifp, error, rt, fibnum)); 2281 } 2282 2283 void 2284 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt) 2285 { 2286 2287 rt_newaddrmsg_fib(cmd, ifa, error, rt, RT_ALL_FIBS); 2288 } 2289 2290 /* 2291 * This is called to generate messages from the routing socket 2292 * indicating a network interface has had addresses associated with it. 2293 */ 2294 void 2295 rt_newaddrmsg_fib(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt, 2296 int fibnum) 2297 { 2298 2299 KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE, 2300 ("unexpected cmd %u", cmd)); 2301 KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs), 2302 ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs)); 2303 2304 if (cmd == RTM_ADD) { 2305 rt_addrmsg(cmd, ifa, fibnum); 2306 if (rt != NULL) 2307 rt_routemsg(cmd, ifa->ifa_ifp, error, rt, fibnum); 2308 } else { 2309 if (rt != NULL) 2310 rt_routemsg(cmd, ifa->ifa_ifp, error, rt, fibnum); 2311 rt_addrmsg(cmd, ifa, fibnum); 2312 } 2313 } 2314 2315