1 /* 2 * Copyright 2007 Sun Microsystems, Inc. All rights reserved. 3 * Use is subject to license terms. 4 * 5 * Copyright (c) 1983, 1988, 1993 6 * The Regents of the University of California. All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. All advertising materials mentioning features or use of this software 17 * must display the following acknowledgment: 18 * This product includes software developed by the University of 19 * California, Berkeley and its contributors. 20 * 4. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * $FreeBSD: src/sbin/routed/table.c,v 1.15 2000/08/11 08:24:38 sheldonh Exp $ 37 */ 38 39 #pragma ident "%Z%%M% %I% %E% SMI" 40 41 #include "defs.h" 42 #include <fcntl.h> 43 #include <stropts.h> 44 #include <sys/tihdr.h> 45 #include <inet/mib2.h> 46 #include <inet/ip.h> 47 48 /* This structure is used to store a disassembled routing socket message. */ 49 struct rt_addrinfo { 50 int rti_addrs; 51 struct sockaddr_storage *rti_info[RTAX_MAX]; 52 }; 53 54 static struct rt_spare *rts_better(struct rt_entry *); 55 static struct rt_spare rts_empty = EMPTY_RT_SPARE; 56 static void set_need_flash(void); 57 static void rtbad(struct rt_entry *, struct interface *); 58 static int rt_xaddrs(struct rt_addrinfo *, struct sockaddr_storage *, 59 char *, int); 60 static struct interface *gwkludge_iflookup(in_addr_t, in_addr_t, in_addr_t); 61 62 struct radix_node_head *rhead; /* root of the radix tree */ 63 64 /* Flash update needed. _B_TRUE to suppress the 1st. */ 65 boolean_t need_flash = _B_TRUE; 66 67 struct timeval age_timer; /* next check of old routes */ 68 struct timeval need_kern = { /* need to update kernel table */ 69 EPOCH+MIN_WAITTIME-1, 0 70 }; 71 72 static uint32_t total_routes; 73 74 #define ROUNDUP_LONG(a) \ 75 ((a) > 0 ? (1 + (((a) - 1) | (sizeof (long) - 1))) : sizeof (long)) 76 77 /* 78 * It is desirable to "aggregate" routes, to combine differing routes of 79 * the same metric and next hop into a common route with a smaller netmask 80 * or to suppress redundant routes, routes that add no information to 81 * routes with smaller netmasks. 82 * 83 * A route is redundant if and only if any and all routes with smaller 84 * but matching netmasks and nets are the same. Since routes are 85 * kept sorted in the radix tree, redundant routes always come second. 86 * 87 * There are two kinds of aggregations. First, two routes of the same bit 88 * mask and differing only in the least significant bit of the network 89 * number can be combined into a single route with a coarser mask. 90 * 91 * Second, a route can be suppressed in favor of another route with a more 92 * coarse mask provided no incompatible routes with intermediate masks 93 * are present. The second kind of aggregation involves suppressing routes. 94 * A route must not be suppressed if an incompatible route exists with 95 * an intermediate mask, since the suppressed route would be covered 96 * by the intermediate. 97 * 98 * This code relies on the radix tree walk encountering routes 99 * sorted first by address, with the smallest address first. 100 */ 101 102 static struct ag_info ag_slots[NUM_AG_SLOTS], *ag_avail, *ag_corsest, 103 *ag_finest; 104 105 #ifdef DEBUG_AG 106 #define CHECK_AG() do { int acnt = 0; struct ag_info *cag; \ 107 for (cag = ag_avail; cag != NULL; cag = cag->ag_fine) \ 108 acnt++; \ 109 for (cag = ag_corsest; cag != NULL; cag = cag->ag_fine) \ 110 acnt++; \ 111 if (acnt != NUM_AG_SLOTS) \ 112 abort(); \ 113 } while (_B_FALSE) 114 #else 115 #define CHECK_AG() (void)0 116 #endif 117 118 119 /* 120 * Output the contents of an aggregation table slot. 121 * This function must always be immediately followed with the deletion 122 * of the target slot. 123 */ 124 static void 125 ag_out(struct ag_info *ag, void (*out)(struct ag_info *)) 126 { 127 struct ag_info *ag_cors; 128 uint32_t bit; 129 130 131 /* Forget it if this route should not be output for split-horizon. */ 132 if (ag->ag_state & AGS_SPLIT_HZ) 133 return; 134 135 /* 136 * If we output both the even and odd twins, then the immediate parent, 137 * if it is present, is redundant, unless the parent manages to 138 * aggregate into something coarser. 139 * On successive calls, this code detects the even and odd twins, 140 * and marks the parent. 141 * 142 * Note that the order in which the radix tree code emits routes 143 * ensures that the twins are seen before the parent is emitted. 144 */ 145 ag_cors = ag->ag_cors; 146 if (ag_cors != NULL && 147 ag_cors->ag_mask == (ag->ag_mask << 1) && 148 ag_cors->ag_dst_h == (ag->ag_dst_h & ag_cors->ag_mask)) { 149 ag_cors->ag_state |= ((ag_cors->ag_dst_h == ag->ag_dst_h) ? 150 AGS_REDUN0 : AGS_REDUN1); 151 } 152 153 /* 154 * Skip it if this route is itself redundant. 155 * 156 * It is ok to change the contents of the slot here, since it is 157 * always deleted next. 158 */ 159 if (ag->ag_state & AGS_REDUN0) { 160 if (ag->ag_state & AGS_REDUN1) 161 return; /* quit if fully redundant */ 162 /* make it finer if it is half-redundant */ 163 bit = (-ag->ag_mask) >> 1; 164 ag->ag_dst_h |= bit; 165 ag->ag_mask |= bit; 166 167 } else if (ag->ag_state & AGS_REDUN1) { 168 /* make it finer if it is half-redundant */ 169 bit = (-ag->ag_mask) >> 1; 170 ag->ag_mask |= bit; 171 } 172 out(ag); 173 } 174 175 176 static void 177 ag_del(struct ag_info *ag) 178 { 179 CHECK_AG(); 180 181 if (ag->ag_cors == NULL) 182 ag_corsest = ag->ag_fine; 183 else 184 ag->ag_cors->ag_fine = ag->ag_fine; 185 186 if (ag->ag_fine == NULL) 187 ag_finest = ag->ag_cors; 188 else 189 ag->ag_fine->ag_cors = ag->ag_cors; 190 191 ag->ag_fine = ag_avail; 192 ag_avail = ag; 193 194 CHECK_AG(); 195 } 196 197 198 /* Look for a route that can suppress the given route. */ 199 static struct ag_info * 200 ag_find_suppressor(struct ag_info *ag) 201 { 202 struct ag_info *ag_cors; 203 in_addr_t dst_h = ag->ag_dst_h; 204 205 for (ag_cors = ag->ag_cors; ag_cors != NULL; 206 ag_cors = ag_cors->ag_cors) { 207 208 if ((dst_h & ag_cors->ag_mask) == ag_cors->ag_dst_h) { 209 /* 210 * We found a route with a coarser mask that covers 211 * the given target. It can suppress the target 212 * only if it has a good enough metric and it 213 * either has the same (gateway, ifp), or if its state 214 * includes AGS_CORS_GATE or the target's state 215 * includes AGS_FINE_GATE. 216 */ 217 if (ag_cors->ag_pref <= ag->ag_pref && 218 (((ag->ag_nhop == ag_cors->ag_nhop) && 219 (ag->ag_ifp == ag_cors->ag_ifp)) || 220 ag_cors->ag_state & AGS_CORS_GATE || 221 ag->ag_state & AGS_FINE_GATE)) { 222 return (ag_cors); 223 } 224 } 225 } 226 227 return (NULL); 228 } 229 230 231 /* 232 * Flush routes waiting for aggregation. 233 * This must not suppress a route unless it is known that among all routes 234 * with coarser masks that match it, the one with the longest mask is 235 * appropriate. This is ensured by scanning the routes in lexical order, 236 * and with the most restrictive mask first among routes to the same 237 * destination. 238 */ 239 void 240 ag_flush(in_addr_t lim_dst_h, /* flush routes to here */ 241 in_addr_t lim_mask, /* matching this mask */ 242 void (*out)(struct ag_info *)) 243 { 244 struct ag_info *ag, *ag_cors, *ag_supr; 245 in_addr_t dst_h; 246 247 248 for (ag = ag_finest; ag != NULL && ag->ag_mask >= lim_mask; 249 ag = ag_cors) { 250 /* Get the next route now, before we delete ag. */ 251 ag_cors = ag->ag_cors; 252 253 /* Work on only the specified routes. */ 254 dst_h = ag->ag_dst_h; 255 if ((dst_h & lim_mask) != lim_dst_h) 256 continue; 257 258 /* 259 * Don't try to suppress the route if its state doesn't 260 * include AGS_SUPPRESS. 261 */ 262 if (!(ag->ag_state & AGS_SUPPRESS)) { 263 ag_out(ag, out); 264 ag_del(ag); 265 continue; 266 } 267 268 ag_supr = ag_find_suppressor(ag); 269 if (ag_supr == NULL) { 270 /* 271 * We didn't find a route which suppresses the 272 * target, so the target can go out. 273 */ 274 ag_out(ag, out); 275 } else { 276 /* 277 * We found a route which suppresses the target, so 278 * don't output the target. 279 */ 280 if (TRACEACTIONS) { 281 trace_misc("aggregated away %s", 282 rtname(htonl(ag->ag_dst_h), ag->ag_mask, 283 ag->ag_nhop)); 284 trace_misc("on coarser route %s", 285 rtname(htonl(ag_supr->ag_dst_h), 286 ag_supr->ag_mask, ag_supr->ag_nhop)); 287 } 288 /* 289 * If the suppressed target was redundant, then 290 * mark the suppressor as redundant. 291 */ 292 if (AG_IS_REDUN(ag->ag_state) && 293 ag_supr->ag_mask == (ag->ag_mask<<1)) { 294 if (ag_supr->ag_dst_h == dst_h) 295 ag_supr->ag_state |= AGS_REDUN0; 296 else 297 ag_supr->ag_state |= AGS_REDUN1; 298 } 299 if (ag->ag_tag != ag_supr->ag_tag) 300 ag_supr->ag_tag = 0; 301 if (ag->ag_nhop != ag_supr->ag_nhop) 302 ag_supr->ag_nhop = 0; 303 } 304 305 /* The route has either been output or suppressed */ 306 ag_del(ag); 307 } 308 309 CHECK_AG(); 310 } 311 312 313 /* Try to aggregate a route with previous routes. */ 314 void 315 ag_check(in_addr_t dst, 316 in_addr_t mask, 317 in_addr_t gate, 318 struct interface *ifp, 319 in_addr_t nhop, 320 uint8_t metric, 321 uint8_t pref, 322 uint32_t seqno, 323 uint16_t tag, 324 uint16_t state, 325 void (*out)(struct ag_info *)) /* output using this */ 326 { 327 struct ag_info *ag, *nag, *ag_cors; 328 in_addr_t xaddr; 329 int tmp; 330 struct interface *xifp; 331 332 dst = ntohl(dst); 333 334 /* 335 * Don't bother trying to aggregate routes with non-contiguous 336 * subnet masks. 337 * 338 * (X & -X) contains a single bit if and only if X is a power of 2. 339 * (X + (X & -X)) == 0 if and only if X is a power of 2. 340 */ 341 if ((mask & -mask) + mask != 0) { 342 struct ag_info nc_ag; 343 344 nc_ag.ag_dst_h = dst; 345 nc_ag.ag_mask = mask; 346 nc_ag.ag_gate = gate; 347 nc_ag.ag_ifp = ifp; 348 nc_ag.ag_nhop = nhop; 349 nc_ag.ag_metric = metric; 350 nc_ag.ag_pref = pref; 351 nc_ag.ag_tag = tag; 352 nc_ag.ag_state = state; 353 nc_ag.ag_seqno = seqno; 354 out(&nc_ag); 355 return; 356 } 357 358 /* Search for the right slot in the aggregation table. */ 359 ag_cors = NULL; 360 ag = ag_corsest; 361 while (ag != NULL) { 362 if (ag->ag_mask >= mask) 363 break; 364 365 /* 366 * Suppress old routes (i.e. combine with compatible routes 367 * with coarser masks) as we look for the right slot in the 368 * aggregation table for the new route. 369 * A route to an address less than the current destination 370 * will not be affected by the current route or any route 371 * seen hereafter. That means it is safe to suppress it. 372 * This check keeps poor routes (e.g. with large hop counts) 373 * from preventing suppression of finer routes. 374 */ 375 if (ag_cors != NULL && ag->ag_dst_h < dst && 376 (ag->ag_state & AGS_SUPPRESS) && 377 ag_cors->ag_pref <= ag->ag_pref && 378 (ag->ag_dst_h & ag_cors->ag_mask) == ag_cors->ag_dst_h && 379 ((ag_cors->ag_nhop == ag->ag_nhop && 380 (ag_cors->ag_ifp == ag->ag_ifp))|| 381 (ag->ag_state & AGS_FINE_GATE) || 382 (ag_cors->ag_state & AGS_CORS_GATE))) { 383 /* 384 * If the suppressed target was redundant, 385 * then mark the suppressor redundant. 386 */ 387 if (AG_IS_REDUN(ag->ag_state) && 388 ag_cors->ag_mask == (ag->ag_mask << 1)) { 389 if (ag_cors->ag_dst_h == dst) 390 ag_cors->ag_state |= AGS_REDUN0; 391 else 392 ag_cors->ag_state |= AGS_REDUN1; 393 } 394 if (ag->ag_tag != ag_cors->ag_tag) 395 ag_cors->ag_tag = 0; 396 if (ag->ag_nhop != ag_cors->ag_nhop) 397 ag_cors->ag_nhop = 0; 398 ag_del(ag); 399 CHECK_AG(); 400 } else { 401 ag_cors = ag; 402 } 403 ag = ag_cors->ag_fine; 404 } 405 406 /* 407 * If we find the even/odd twin of the new route, and if the 408 * masks and so forth are equal, we can aggregate them. 409 * We can probably promote one of the pair. 410 * 411 * Since the routes are encountered in lexical order, 412 * the new route must be odd. However, the second or later 413 * times around this loop, it could be the even twin promoted 414 * from the even/odd pair of twins of the finer route. 415 */ 416 while (ag != NULL && ag->ag_mask == mask && 417 ((ag->ag_dst_h ^ dst) & (mask<<1)) == 0) { 418 419 /* 420 * Here we know the target route and the route in the current 421 * slot have the same netmasks and differ by at most the 422 * last bit. They are either for the same destination, or 423 * for an even/odd pair of destinations. 424 */ 425 if (ag->ag_dst_h == dst) { 426 if (ag->ag_nhop == nhop && ag->ag_ifp == ifp) { 427 /* 428 * We have two routes to the same destination, 429 * with the same nexthop and interface. 430 * Routes are encountered in lexical order, 431 * so a route is never promoted until the 432 * parent route is already present. So we 433 * know that the new route is a promoted (or 434 * aggregated) pair and the route already in 435 * the slot is the explicit route. 436 * 437 * Prefer the best route if their metrics 438 * differ, or the aggregated one if not, 439 * following a sort of longest-match rule. 440 */ 441 if (pref <= ag->ag_pref) { 442 ag->ag_gate = gate; 443 ag->ag_ifp = ifp; 444 ag->ag_nhop = nhop; 445 ag->ag_tag = tag; 446 ag->ag_metric = metric; 447 ag->ag_pref = pref; 448 if (seqno > ag->ag_seqno) 449 ag->ag_seqno = seqno; 450 tmp = ag->ag_state; 451 ag->ag_state = state; 452 state = tmp; 453 } 454 455 /* 456 * Some bits are set if they are set on 457 * either route, except when the route is 458 * for an interface. 459 */ 460 if (!(ag->ag_state & AGS_IF)) 461 ag->ag_state |= 462 (state & (AGS_AGGREGATE_EITHER | 463 AGS_REDUN0 | AGS_REDUN1)); 464 465 return; 466 } else { 467 /* 468 * multiple routes to same dest/mask with 469 * differing gate nexthop/or ifp. Flush 470 * both out. 471 */ 472 break; 473 } 474 } 475 476 /* 477 * If one of the routes can be promoted and the other can 478 * be suppressed, it may be possible to combine them or 479 * worthwhile to promote one. 480 * 481 * Any route that can be promoted is always 482 * marked to be eligible to be suppressed. 483 */ 484 if (!((state & AGS_AGGREGATE) && 485 (ag->ag_state & AGS_SUPPRESS)) && 486 !((ag->ag_state & AGS_AGGREGATE) && (state & AGS_SUPPRESS))) 487 break; 488 489 /* 490 * A pair of even/odd twin routes can be combined 491 * if either is redundant, or if they are via the 492 * same gateway and have the same metric. 493 */ 494 if (AG_IS_REDUN(ag->ag_state) || AG_IS_REDUN(state) || 495 (ag->ag_nhop == nhop && ag->ag_ifp == ifp && 496 ag->ag_pref == pref && 497 (state & ag->ag_state & AGS_AGGREGATE) != 0)) { 498 499 /* 500 * We have both the even and odd pairs. 501 * Since the routes are encountered in order, 502 * the route in the slot must be the even twin. 503 * 504 * Combine and promote (aggregate) the pair of routes. 505 */ 506 if (seqno < ag->ag_seqno) 507 seqno = ag->ag_seqno; 508 if (!AG_IS_REDUN(state)) 509 state &= ~AGS_REDUN1; 510 if (AG_IS_REDUN(ag->ag_state)) 511 state |= AGS_REDUN0; 512 else 513 state &= ~AGS_REDUN0; 514 state |= (ag->ag_state & AGS_AGGREGATE_EITHER); 515 if (ag->ag_tag != tag) 516 tag = 0; 517 if (ag->ag_nhop != nhop) 518 nhop = 0; 519 520 /* 521 * Get rid of the even twin that was already 522 * in the slot. 523 */ 524 ag_del(ag); 525 526 } else if (ag->ag_pref >= pref && 527 (ag->ag_state & AGS_AGGREGATE)) { 528 /* 529 * If we cannot combine the pair, maybe the route 530 * with the worse metric can be promoted. 531 * 532 * Promote the old, even twin, by giving its slot 533 * in the table to the new, odd twin. 534 */ 535 ag->ag_dst_h = dst; 536 537 xaddr = ag->ag_gate; 538 ag->ag_gate = gate; 539 gate = xaddr; 540 541 xifp = ag->ag_ifp; 542 ag->ag_ifp = ifp; 543 ifp = xifp; 544 545 xaddr = ag->ag_nhop; 546 ag->ag_nhop = nhop; 547 nhop = xaddr; 548 549 tmp = ag->ag_tag; 550 ag->ag_tag = tag; 551 tag = tmp; 552 553 /* 554 * The promoted route is even-redundant only if the 555 * even twin was fully redundant. It is not 556 * odd-redundant because the odd-twin will still be 557 * in the table. 558 */ 559 tmp = ag->ag_state; 560 if (!AG_IS_REDUN(tmp)) 561 tmp &= ~AGS_REDUN0; 562 tmp &= ~AGS_REDUN1; 563 ag->ag_state = state; 564 state = tmp; 565 566 tmp = ag->ag_metric; 567 ag->ag_metric = metric; 568 metric = tmp; 569 570 tmp = ag->ag_pref; 571 ag->ag_pref = pref; 572 pref = tmp; 573 574 /* take the newest sequence number */ 575 if (seqno <= ag->ag_seqno) 576 seqno = ag->ag_seqno; 577 else 578 ag->ag_seqno = seqno; 579 580 } else { 581 if (!(state & AGS_AGGREGATE)) 582 break; /* cannot promote either twin */ 583 584 /* 585 * Promote the new, odd twin by shaving its 586 * mask and address. 587 * The promoted route is odd-redundant only if the 588 * odd twin was fully redundant. It is not 589 * even-redundant because the even twin is still in 590 * the table. 591 */ 592 if (!AG_IS_REDUN(state)) 593 state &= ~AGS_REDUN1; 594 state &= ~AGS_REDUN0; 595 if (seqno < ag->ag_seqno) 596 seqno = ag->ag_seqno; 597 else 598 ag->ag_seqno = seqno; 599 } 600 601 mask <<= 1; 602 dst &= mask; 603 604 if (ag_cors == NULL) { 605 ag = ag_corsest; 606 break; 607 } 608 ag = ag_cors; 609 ag_cors = ag->ag_cors; 610 } 611 612 /* 613 * When we can no longer promote and combine routes, 614 * flush the old route in the target slot. Also flush 615 * any finer routes that we know will never be aggregated by 616 * the new route. 617 * 618 * In case we moved toward coarser masks, 619 * get back where we belong 620 */ 621 if (ag != NULL && ag->ag_mask < mask) { 622 ag_cors = ag; 623 ag = ag->ag_fine; 624 } 625 626 /* Empty the target slot */ 627 if (ag != NULL && ag->ag_mask == mask) { 628 ag_flush(ag->ag_dst_h, ag->ag_mask, out); 629 ag = (ag_cors == NULL) ? ag_corsest : ag_cors->ag_fine; 630 } 631 632 #ifdef DEBUG_AG 633 if (ag == NULL && ag_cors != ag_finest) 634 abort(); 635 if (ag_cors == NULL && ag != ag_corsest) 636 abort(); 637 if (ag != NULL && ag->ag_cors != ag_cors) 638 abort(); 639 if (ag_cors != NULL && ag_cors->ag_fine != ag) 640 abort(); 641 CHECK_AG(); 642 #endif 643 644 /* Save the new route on the end of the table. */ 645 nag = ag_avail; 646 ag_avail = nag->ag_fine; 647 648 nag->ag_dst_h = dst; 649 nag->ag_mask = mask; 650 nag->ag_ifp = ifp; 651 nag->ag_gate = gate; 652 nag->ag_nhop = nhop; 653 nag->ag_metric = metric; 654 nag->ag_pref = pref; 655 nag->ag_tag = tag; 656 nag->ag_state = state; 657 nag->ag_seqno = seqno; 658 659 nag->ag_fine = ag; 660 if (ag != NULL) 661 ag->ag_cors = nag; 662 else 663 ag_finest = nag; 664 nag->ag_cors = ag_cors; 665 if (ag_cors == NULL) 666 ag_corsest = nag; 667 else 668 ag_cors->ag_fine = nag; 669 CHECK_AG(); 670 } 671 672 673 static const char * 674 rtm_type_name(uchar_t type) 675 { 676 static const char *rtm_types[] = { 677 "RTM_ADD", 678 "RTM_DELETE", 679 "RTM_CHANGE", 680 "RTM_GET", 681 "RTM_LOSING", 682 "RTM_REDIRECT", 683 "RTM_MISS", 684 "RTM_LOCK", 685 "RTM_OLDADD", 686 "RTM_OLDDEL", 687 "RTM_RESOLVE", 688 "RTM_NEWADDR", 689 "RTM_DELADDR", 690 "RTM_IFINFO", 691 "RTM_NEWMADDR", 692 "RTM_DELMADDR" 693 }; 694 #define NEW_RTM_PAT "RTM type %#x" 695 static char name0[sizeof (NEW_RTM_PAT) + 2]; 696 697 if (type > sizeof (rtm_types) / sizeof (rtm_types[0]) || type == 0) { 698 (void) snprintf(name0, sizeof (name0), NEW_RTM_PAT, type); 699 return (name0); 700 } else { 701 return (rtm_types[type-1]); 702 } 703 #undef NEW_RTM_PAT 704 } 705 706 707 static void 708 dump_rt_msg(const char *act, struct rt_msghdr *rtm, int mlen) 709 { 710 const char *mtype; 711 uchar_t *cp; 712 int i, j; 713 char buffer[16*3 + 1], *ibs; 714 struct ifa_msghdr *ifam; 715 struct if_msghdr *ifm; 716 717 switch (rtm->rtm_type) { 718 case RTM_NEWADDR: 719 case RTM_DELADDR: 720 mtype = "ifam"; 721 break; 722 case RTM_IFINFO: 723 mtype = "ifm"; 724 break; 725 default: 726 mtype = "rtm"; 727 break; 728 } 729 trace_misc("%s %s %d bytes", act, mtype, mlen); 730 if (mlen > rtm->rtm_msglen) { 731 trace_misc("%s: extra %d bytes ignored", mtype, 732 mlen - rtm->rtm_msglen); 733 mlen = rtm->rtm_msglen; 734 } else if (mlen < rtm->rtm_msglen) { 735 trace_misc("%s: truncated by %d bytes", mtype, 736 rtm->rtm_msglen - mlen); 737 } 738 switch (rtm->rtm_type) { 739 case RTM_NEWADDR: 740 case RTM_DELADDR: 741 ifam = (struct ifa_msghdr *)rtm; 742 trace_misc("ifam: msglen %d version %d type %d addrs %X", 743 ifam->ifam_msglen, ifam->ifam_version, ifam->ifam_type, 744 ifam->ifam_addrs); 745 trace_misc("ifam: flags %X index %d metric %d", 746 ifam->ifam_flags, ifam->ifam_index, ifam->ifam_metric); 747 cp = (uchar_t *)(ifam + 1); 748 break; 749 case RTM_IFINFO: 750 ifm = (struct if_msghdr *)rtm; 751 trace_misc("ifm: msglen %d version %d type %d addrs %X", 752 ifm->ifm_msglen, ifm->ifm_version, ifm->ifm_type, 753 ifm->ifm_addrs); 754 ibs = if_bit_string(ifm->ifm_flags, _B_TRUE); 755 if (ibs == NULL) { 756 trace_misc("ifm: flags %#x index %d", ifm->ifm_flags, 757 ifm->ifm_index); 758 } else { 759 trace_misc("ifm: flags %s index %d", ibs, 760 ifm->ifm_index); 761 free(ibs); 762 } 763 cp = (uchar_t *)(ifm + 1); 764 break; 765 default: 766 trace_misc("rtm: msglen %d version %d type %d index %d", 767 rtm->rtm_msglen, rtm->rtm_version, rtm->rtm_type, 768 rtm->rtm_index); 769 trace_misc("rtm: flags %X addrs %X pid %d seq %d", 770 rtm->rtm_flags, rtm->rtm_addrs, rtm->rtm_pid, rtm->rtm_seq); 771 trace_misc("rtm: errno %d use %d inits %X", rtm->rtm_errno, 772 rtm->rtm_use, rtm->rtm_inits); 773 cp = (uchar_t *)(rtm + 1); 774 break; 775 } 776 i = mlen - (cp - (uint8_t *)rtm); 777 while (i > 0) { 778 buffer[0] = '\0'; 779 ibs = buffer; 780 for (j = 0; j < 16 && i > 0; j++, i--) 781 ibs += sprintf(ibs, " %02X", *cp++); 782 trace_misc("addr%s", buffer); 783 } 784 } 785 786 /* 787 * Tell the kernel to add, delete or change a route 788 * Pass k_state from khash in for diagnostic info. 789 */ 790 static void 791 rtioctl(int action, /* RTM_DELETE, etc */ 792 in_addr_t dst, 793 in_addr_t gate, 794 in_addr_t mask, 795 struct interface *ifp, 796 uint8_t metric, 797 int flags) 798 { 799 static int rt_sock_seqno = 0; 800 struct { 801 struct rt_msghdr w_rtm; 802 struct sockaddr_in w_dst; 803 struct sockaddr_in w_gate; 804 uint8_t w_space[512]; 805 } w; 806 struct sockaddr_in w_mask; 807 struct sockaddr_dl w_ifp; 808 uint8_t *cp; 809 long cc; 810 #define PAT " %-10s %s metric=%d flags=%#x" 811 #define ARGS rtm_type_name(action), rtname(dst, mask, gate), metric, flags 812 813 again: 814 (void) memset(&w, 0, sizeof (w)); 815 (void) memset(&w_mask, 0, sizeof (w_mask)); 816 (void) memset(&w_ifp, 0, sizeof (w_ifp)); 817 cp = w.w_space; 818 w.w_rtm.rtm_msglen = sizeof (struct rt_msghdr) + 819 2 * ROUNDUP_LONG(sizeof (struct sockaddr_in)); 820 w.w_rtm.rtm_version = RTM_VERSION; 821 w.w_rtm.rtm_type = action; 822 w.w_rtm.rtm_flags = flags; 823 w.w_rtm.rtm_seq = ++rt_sock_seqno; 824 w.w_rtm.rtm_addrs = RTA_DST|RTA_GATEWAY; 825 if (metric != 0 || action == RTM_CHANGE) { 826 w.w_rtm.rtm_rmx.rmx_hopcount = metric; 827 w.w_rtm.rtm_inits |= RTV_HOPCOUNT; 828 } 829 w.w_dst.sin_family = AF_INET; 830 w.w_dst.sin_addr.s_addr = dst; 831 w.w_gate.sin_family = AF_INET; 832 w.w_gate.sin_addr.s_addr = gate; 833 if (mask == HOST_MASK) { 834 w.w_rtm.rtm_flags |= RTF_HOST; 835 } else { 836 w.w_rtm.rtm_addrs |= RTA_NETMASK; 837 w_mask.sin_family = AF_INET; 838 w_mask.sin_addr.s_addr = htonl(mask); 839 (void) memmove(cp, &w_mask, sizeof (w_mask)); 840 cp += ROUNDUP_LONG(sizeof (struct sockaddr_in)); 841 w.w_rtm.rtm_msglen += ROUNDUP_LONG(sizeof (struct sockaddr_in)); 842 } 843 if (ifp == NULL) 844 ifp = iflookup(gate); 845 846 if ((ifp == NULL) || (ifp->int_phys == NULL)) { 847 trace_misc("no ifp for" PAT, ARGS); 848 } else { 849 if (ifp->int_phys->phyi_index > UINT16_MAX) { 850 trace_misc("ifindex %d is too big for sdl_index", 851 ifp->int_phys->phyi_index); 852 } else { 853 w_ifp.sdl_family = AF_LINK; 854 w.w_rtm.rtm_addrs |= RTA_IFP; 855 w_ifp.sdl_index = ifp->int_phys->phyi_index; 856 (void) memmove(cp, &w_ifp, sizeof (w_ifp)); 857 w.w_rtm.rtm_msglen += 858 ROUNDUP_LONG(sizeof (struct sockaddr_dl)); 859 } 860 } 861 862 863 if (!no_install) { 864 if (TRACERTS) 865 dump_rt_msg("write", &w.w_rtm, w.w_rtm.rtm_msglen); 866 cc = write(rt_sock, &w, w.w_rtm.rtm_msglen); 867 if (cc < 0) { 868 if (errno == ESRCH && (action == RTM_CHANGE || 869 action == RTM_DELETE)) { 870 trace_act("route disappeared before" PAT, ARGS); 871 if (action == RTM_CHANGE) { 872 action = RTM_ADD; 873 goto again; 874 } 875 return; 876 } 877 writelog(LOG_WARNING, "write(rt_sock)" PAT ": %s ", 878 ARGS, rip_strerror(errno)); 879 return; 880 } else if (cc != w.w_rtm.rtm_msglen) { 881 msglog("write(rt_sock) wrote %ld instead of %d for" PAT, 882 cc, w.w_rtm.rtm_msglen, ARGS); 883 return; 884 } 885 } 886 if (TRACEKERNEL) 887 trace_misc("write kernel" PAT, ARGS); 888 #undef PAT 889 #undef ARGS 890 } 891 892 893 /* Hash table containing our image of the kernel forwarding table. */ 894 #define KHASH_SIZE 71 /* should be prime */ 895 #define KHASH(a, m) khash_bins[((a) ^ (m)) % KHASH_SIZE] 896 static struct khash *khash_bins[KHASH_SIZE]; 897 898 #define K_KEEP_LIM 30 /* k_keep */ 899 900 static struct khash * 901 kern_find(in_addr_t dst, in_addr_t mask, in_addr_t gate, 902 struct interface *ifp, struct khash ***ppk) 903 { 904 struct khash *k, **pk; 905 906 if (ifp != NULL && ifp->int_phys != NULL) { 907 ifp = ifwithname(ifp->int_phys->phyi_name); 908 } 909 910 for (pk = &KHASH(dst, mask); (k = *pk) != NULL; pk = &k->k_next) { 911 if (k->k_dst == dst && k->k_mask == mask && 912 (gate == 0 || k->k_gate == gate) && 913 (ifp == NULL || k->k_ifp == ifp)) { 914 break; 915 } 916 } 917 if (ppk != NULL) 918 *ppk = pk; 919 return (k); 920 } 921 922 923 /* 924 * Find out if there is an alternate route to a given destination 925 * off of a given interface. 926 */ 927 static struct khash * 928 kern_alternate(in_addr_t dst, in_addr_t mask, in_addr_t gate, 929 struct interface *ifp, struct khash ***ppk) 930 { 931 struct khash *k, **pk; 932 933 if (ifp != NULL && ifp->int_phys != NULL) { 934 ifp = ifwithname(ifp->int_phys->phyi_name); 935 } 936 for (pk = &KHASH(dst, mask); (k = *pk) != NULL; pk = &k->k_next) { 937 if (k->k_dst == dst && k->k_mask == mask && 938 (k->k_gate != gate) && 939 (k->k_ifp == ifp)) { 940 break; 941 } 942 } 943 if (ppk != NULL) 944 *ppk = pk; 945 return (k); 946 } 947 948 static struct khash * 949 kern_add(in_addr_t dst, uint32_t mask, in_addr_t gate, struct interface *ifp) 950 { 951 struct khash *k, **pk; 952 953 if (ifp != NULL && ifp->int_phys != NULL) { 954 ifp = ifwithname(ifp->int_phys->phyi_name); 955 } 956 k = kern_find(dst, mask, gate, ifp, &pk); 957 if (k != NULL) 958 return (k); 959 960 k = rtmalloc(sizeof (*k), "kern_add"); 961 962 (void) memset(k, 0, sizeof (*k)); 963 k->k_dst = dst; 964 k->k_mask = mask; 965 k->k_state = KS_NEW; 966 k->k_keep = now.tv_sec; 967 k->k_gate = gate; 968 k->k_ifp = ifp; 969 *pk = k; 970 971 return (k); 972 } 973 974 /* delete all khash entries that are wired through the interface ifp */ 975 void 976 kern_flush_ifp(struct interface *ifp) 977 { 978 struct khash *k, *kprev, *knext; 979 int i; 980 981 if (ifp != NULL && ifp->int_phys != NULL) { 982 ifp = ifwithname(ifp->int_phys->phyi_name); 983 } 984 for (i = 0; i < KHASH_SIZE; i++) { 985 kprev = NULL; 986 for (k = khash_bins[i]; k != NULL; k = knext) { 987 knext = k->k_next; 988 if (k->k_ifp == ifp) { 989 if (kprev != NULL) 990 kprev->k_next = k->k_next; 991 else 992 khash_bins[i] = k->k_next; 993 free(k); 994 continue; 995 } 996 kprev = k; 997 } 998 } 999 } 1000 1001 /* 1002 * rewire khash entries that currently go through oldifp to 1003 * go through newifp. 1004 */ 1005 void 1006 kern_rewire_ifp(struct interface *oldifp, struct interface *newifp) 1007 { 1008 struct khash *k; 1009 int i; 1010 1011 if (oldifp != NULL && oldifp->int_phys != NULL) { 1012 oldifp = ifwithname(oldifp->int_phys->phyi_name); 1013 } 1014 if (newifp != NULL && newifp->int_phys != NULL) { 1015 newifp = ifwithname(newifp->int_phys->phyi_name); 1016 } 1017 for (i = 0; i < KHASH_SIZE; i++) { 1018 for (k = khash_bins[i]; k; k = k->k_next) { 1019 if (k->k_ifp == oldifp) { 1020 k->k_ifp = newifp; 1021 trace_misc("kern_rewire_ifp k 0x%lx " 1022 "from %s to %s", k, oldifp->int_name, 1023 newifp->int_name); 1024 } 1025 } 1026 } 1027 } 1028 1029 1030 /* 1031 * Check that a static route it is still in the daemon table, and not 1032 * deleted by interfaces coming and going. This is also the routine 1033 * responsible for adding new static routes to the daemon table. 1034 */ 1035 static void 1036 kern_check_static(struct khash *k, struct interface *ifp) 1037 { 1038 struct rt_entry *rt; 1039 struct rt_spare new; 1040 uint16_t rt_state = RS_STATIC; 1041 1042 if (ifp != NULL && ifp->int_phys != NULL) { 1043 ifp = ifwithname(ifp->int_phys->phyi_name); 1044 } 1045 (void) memset(&new, 0, sizeof (new)); 1046 new.rts_ifp = ifp; 1047 new.rts_gate = k->k_gate; 1048 new.rts_router = (ifp != NULL) ? ifp->int_addr : loopaddr; 1049 new.rts_metric = k->k_metric; 1050 new.rts_time = now.tv_sec; 1051 new.rts_origin = RO_STATIC; 1052 1053 rt = rtget(k->k_dst, k->k_mask); 1054 if ((ifp != NULL && !IS_IFF_ROUTING(ifp->int_if_flags)) || 1055 (k->k_state & KS_PRIVATE)) 1056 rt_state |= RS_NOPROPAGATE; 1057 1058 if (rt != NULL) { 1059 if ((rt->rt_state & RS_STATIC) == 0) { 1060 /* 1061 * We are already tracking this dest/mask 1062 * via RIP/RDISC. Ignore the static route, 1063 * because we don't currently have a good 1064 * way to compare metrics on static routes 1065 * with rip metrics, and therefore cannot 1066 * mix and match the two. 1067 */ 1068 return; 1069 } 1070 rt_state |= rt->rt_state; 1071 if (rt->rt_state != rt_state) 1072 rtchange(rt, rt_state, &new, 0); 1073 } else { 1074 rtadd(k->k_dst, k->k_mask, rt_state, &new); 1075 } 1076 } 1077 1078 1079 /* operate on a kernel entry */ 1080 static void 1081 kern_ioctl(struct khash *k, 1082 int action, /* RTM_DELETE, etc */ 1083 int flags) 1084 { 1085 if (((k->k_state & (KS_IF|KS_PASSIVE)) == KS_IF) || 1086 (k->k_state & KS_DEPRE_IF)) { 1087 /* 1088 * Prevent execution of RTM_DELETE, RTM_ADD or 1089 * RTM_CHANGE of interface routes 1090 */ 1091 trace_act("Blocking execution of %s %s --> %s ", 1092 rtm_type_name(action), 1093 addrname(k->k_dst, k->k_mask, 0), naddr_ntoa(k->k_gate)); 1094 return; 1095 } 1096 1097 switch (action) { 1098 case RTM_DELETE: 1099 k->k_state &= ~KS_DYNAMIC; 1100 if (k->k_state & KS_DELETED) 1101 return; 1102 k->k_state |= KS_DELETED; 1103 break; 1104 case RTM_ADD: 1105 k->k_state &= ~KS_DELETED; 1106 break; 1107 case RTM_CHANGE: 1108 if (k->k_state & KS_DELETED) { 1109 action = RTM_ADD; 1110 k->k_state &= ~KS_DELETED; 1111 } 1112 break; 1113 } 1114 1115 rtioctl(action, k->k_dst, k->k_gate, k->k_mask, k->k_ifp, 1116 k->k_metric, flags); 1117 } 1118 1119 1120 /* add a route the kernel told us */ 1121 static void 1122 rtm_add(struct rt_msghdr *rtm, 1123 struct rt_addrinfo *info, 1124 time_t keep, 1125 boolean_t interf_route, 1126 struct interface *ifptr) 1127 { 1128 struct khash *k; 1129 struct interface *ifp = ifptr; 1130 in_addr_t mask, gate = 0; 1131 static struct msg_limit msg_no_ifp; 1132 1133 if (rtm->rtm_flags & RTF_HOST) { 1134 mask = HOST_MASK; 1135 } else if (INFO_MASK(info) != 0) { 1136 mask = ntohl(S_ADDR(INFO_MASK(info))); 1137 } else { 1138 writelog(LOG_WARNING, 1139 "ignore %s without mask", rtm_type_name(rtm->rtm_type)); 1140 return; 1141 } 1142 1143 /* 1144 * Find the interface toward the gateway. 1145 */ 1146 if (INFO_GATE(info) != NULL) 1147 gate = S_ADDR(INFO_GATE(info)); 1148 1149 if (ifp == NULL) { 1150 if (INFO_GATE(info) != NULL) 1151 ifp = iflookup(gate); 1152 if (ifp == NULL) { 1153 msglim(&msg_no_ifp, gate, 1154 "route %s --> %s nexthop is not directly connected", 1155 addrname(S_ADDR(INFO_DST(info)), mask, 0), 1156 naddr_ntoa(gate)); 1157 } else { 1158 if (ifp->int_phys != NULL) { 1159 ifp = ifwithname(ifp->int_phys->phyi_name); 1160 } 1161 } 1162 } 1163 1164 k = kern_add(S_ADDR(INFO_DST(info)), mask, gate, ifp); 1165 1166 if (k->k_state & KS_NEW) 1167 k->k_keep = now.tv_sec+keep; 1168 if (INFO_GATE(info) == 0) { 1169 trace_act("note %s without gateway", 1170 rtm_type_name(rtm->rtm_type)); 1171 k->k_metric = HOPCNT_INFINITY; 1172 } else if (INFO_GATE(info)->ss_family != AF_INET) { 1173 trace_act("note %s with gateway AF=%d", 1174 rtm_type_name(rtm->rtm_type), 1175 INFO_GATE(info)->ss_family); 1176 k->k_metric = HOPCNT_INFINITY; 1177 } else { 1178 k->k_gate = S_ADDR(INFO_GATE(info)); 1179 k->k_metric = rtm->rtm_rmx.rmx_hopcount; 1180 if (k->k_metric < 0) 1181 k->k_metric = 0; 1182 else if (k->k_metric > HOPCNT_INFINITY-1) 1183 k->k_metric = HOPCNT_INFINITY-1; 1184 } 1185 1186 if ((k->k_state & KS_NEW) && interf_route) { 1187 if (k->k_gate != 0 && findifaddr(k->k_gate) == NULL) 1188 k->k_state |= KS_DEPRE_IF; 1189 else 1190 k->k_state |= KS_IF; 1191 } 1192 1193 k->k_state &= ~(KS_NEW | KS_DELETE | KS_ADD | KS_CHANGE | KS_DEL_ADD | 1194 KS_STATIC | KS_GATEWAY | KS_DELETED | KS_PRIVATE | KS_CHECK); 1195 if (rtm->rtm_flags & RTF_GATEWAY) 1196 k->k_state |= KS_GATEWAY; 1197 if (rtm->rtm_flags & RTF_STATIC) 1198 k->k_state |= KS_STATIC; 1199 if (rtm->rtm_flags & RTF_PRIVATE) 1200 k->k_state |= KS_PRIVATE; 1201 1202 1203 if (rtm->rtm_flags & (RTF_DYNAMIC | RTF_MODIFIED)) { 1204 if (INFO_AUTHOR(info) != 0 && 1205 INFO_AUTHOR(info)->ss_family == AF_INET) 1206 ifp = iflookup(S_ADDR(INFO_AUTHOR(info))); 1207 else 1208 ifp = NULL; 1209 if (should_supply(ifp) && (ifp == NULL || 1210 !(ifp->int_state & IS_REDIRECT_OK))) { 1211 /* 1212 * Routers are not supposed to listen to redirects, 1213 * so delete it if it came via an unknown interface 1214 * or the interface does not have special permission. 1215 */ 1216 k->k_state &= ~KS_DYNAMIC; 1217 k->k_state |= KS_DELETE; 1218 LIM_SEC(need_kern, 0); 1219 trace_act("mark for deletion redirected %s --> %s" 1220 " via %s", 1221 addrname(k->k_dst, k->k_mask, 0), 1222 naddr_ntoa(k->k_gate), 1223 ifp ? ifp->int_name : "unknown interface"); 1224 } else { 1225 k->k_state |= KS_DYNAMIC; 1226 k->k_redirect_time = now.tv_sec; 1227 trace_act("accept redirected %s --> %s via %s", 1228 addrname(k->k_dst, k->k_mask, 0), 1229 naddr_ntoa(k->k_gate), 1230 ifp ? ifp->int_name : "unknown interface"); 1231 } 1232 return; 1233 } 1234 1235 /* 1236 * If it is not a static route, quit until the next comparison 1237 * between the kernel and daemon tables, when it will be deleted. 1238 */ 1239 if (!(k->k_state & KS_STATIC)) { 1240 if (!(k->k_state & (KS_IF|KS_DEPRE_IF|KS_FILE))) 1241 k->k_state |= KS_DELETE; 1242 LIM_SEC(need_kern, k->k_keep); 1243 return; 1244 } 1245 1246 /* 1247 * Put static routes with real metrics into the daemon table so 1248 * they can be advertised. 1249 */ 1250 1251 kern_check_static(k, ifp); 1252 } 1253 1254 1255 /* deal with packet loss */ 1256 static void 1257 rtm_lose(struct rt_msghdr *rtm, struct rt_addrinfo *info) 1258 { 1259 if (INFO_GATE(info) == NULL || INFO_GATE(info)->ss_family != AF_INET) { 1260 trace_act("ignore %s without gateway", 1261 rtm_type_name(rtm->rtm_type)); 1262 age(0); 1263 return; 1264 } 1265 1266 if (rdisc_ok) 1267 rdisc_age(S_ADDR(INFO_GATE(info))); 1268 age(S_ADDR(INFO_GATE(info))); 1269 } 1270 1271 1272 /* 1273 * Make the gateway slot of an info structure point to something 1274 * useful. If it is not already useful, but it specifies an interface, 1275 * then fill in the sockaddr_in provided and point it there. 1276 */ 1277 static int 1278 get_info_gate(struct sockaddr_storage **ssp, struct sockaddr_in *sin) 1279 { 1280 struct sockaddr_dl *sdl = (struct sockaddr_dl *)*ssp; 1281 struct interface *ifp; 1282 1283 if (sdl == NULL) 1284 return (0); 1285 if ((sdl)->sdl_family == AF_INET) 1286 return (1); 1287 if ((sdl)->sdl_family != AF_LINK) 1288 return (0); 1289 1290 ifp = ifwithindex(sdl->sdl_index, _B_TRUE); 1291 if (ifp == NULL) 1292 return (0); 1293 1294 sin->sin_addr.s_addr = ifp->int_addr; 1295 sin->sin_family = AF_INET; 1296 /* LINTED */ 1297 *ssp = (struct sockaddr_storage *)sin; 1298 1299 return (1); 1300 } 1301 1302 1303 /* 1304 * Clean the kernel table by copying it to the daemon image. 1305 * Eventually the daemon will delete any extra routes. 1306 */ 1307 void 1308 sync_kern(void) 1309 { 1310 int i; 1311 struct khash *k; 1312 struct { 1313 struct T_optmgmt_req req; 1314 struct opthdr hdr; 1315 } req; 1316 union { 1317 struct T_optmgmt_ack ack; 1318 unsigned char space[64]; 1319 } ack; 1320 struct opthdr *rh; 1321 struct strbuf cbuf, dbuf; 1322 int ipfd, nroutes, flags, r; 1323 mib2_ipRouteEntry_t routes[8]; 1324 mib2_ipRouteEntry_t *rp; 1325 struct rt_msghdr rtm; 1326 struct rt_addrinfo info; 1327 struct sockaddr_in sin_dst; 1328 struct sockaddr_in sin_gate; 1329 struct sockaddr_in sin_mask; 1330 struct sockaddr_in sin_author; 1331 struct interface *ifp; 1332 char ifname[LIFNAMSIZ + 1]; 1333 1334 for (i = 0; i < KHASH_SIZE; i++) { 1335 for (k = khash_bins[i]; k != NULL; k = k->k_next) { 1336 if (!(k->k_state & (KS_IF|KS_DEPRE_IF))) 1337 k->k_state |= KS_CHECK; 1338 } 1339 } 1340 1341 ipfd = open(IP_DEV_NAME, O_RDWR); 1342 if (ipfd == -1) { 1343 msglog("open " IP_DEV_NAME ": %s", rip_strerror(errno)); 1344 goto hash_clean; 1345 } 1346 1347 req.req.PRIM_type = T_OPTMGMT_REQ; 1348 req.req.OPT_offset = (caddr_t)&req.hdr - (caddr_t)&req; 1349 req.req.OPT_length = sizeof (req.hdr); 1350 req.req.MGMT_flags = T_CURRENT; 1351 1352 req.hdr.level = MIB2_IP; 1353 req.hdr.name = 0; 1354 req.hdr.len = 0; 1355 1356 cbuf.buf = (caddr_t)&req; 1357 cbuf.len = sizeof (req); 1358 1359 if (putmsg(ipfd, &cbuf, NULL, 0) == -1) { 1360 msglog("T_OPTMGMT_REQ putmsg: %s", rip_strerror(errno)); 1361 goto hash_clean; 1362 } 1363 1364 for (;;) { 1365 cbuf.buf = (caddr_t)&ack; 1366 cbuf.maxlen = sizeof (ack); 1367 dbuf.buf = (caddr_t)routes; 1368 dbuf.maxlen = sizeof (routes); 1369 flags = 0; 1370 r = getmsg(ipfd, &cbuf, &dbuf, &flags); 1371 if (r == -1) { 1372 msglog("T_OPTMGMT_REQ getmsg: %s", rip_strerror(errno)); 1373 goto hash_clean; 1374 } 1375 1376 if (cbuf.len < sizeof (struct T_optmgmt_ack) || 1377 ack.ack.PRIM_type != T_OPTMGMT_ACK || 1378 ack.ack.MGMT_flags != T_SUCCESS || 1379 ack.ack.OPT_length < sizeof (struct opthdr)) { 1380 msglog("bad T_OPTMGMT response; len=%d prim=%d " 1381 "flags=%d optlen=%d", cbuf.len, ack.ack.PRIM_type, 1382 ack.ack.MGMT_flags, ack.ack.OPT_length); 1383 goto hash_clean; 1384 } 1385 /* LINTED */ 1386 rh = (struct opthdr *)((caddr_t)&ack + ack.ack.OPT_offset); 1387 if (rh->level == 0 && rh->name == 0) { 1388 break; 1389 } 1390 if (rh->level != MIB2_IP || rh->name != MIB2_IP_21) { 1391 while (r == MOREDATA) { 1392 r = getmsg(ipfd, NULL, &dbuf, &flags); 1393 } 1394 continue; 1395 } 1396 break; 1397 } 1398 1399 (void) memset(&rtm, 0, sizeof (rtm)); 1400 (void) memset(&info, 0, sizeof (info)); 1401 (void) memset(&sin_dst, 0, sizeof (sin_dst)); 1402 (void) memset(&sin_gate, 0, sizeof (sin_gate)); 1403 (void) memset(&sin_mask, 0, sizeof (sin_mask)); 1404 (void) memset(&sin_author, 0, sizeof (sin_author)); 1405 sin_dst.sin_family = AF_INET; 1406 /* LINTED */ 1407 info.rti_info[RTAX_DST] = (struct sockaddr_storage *)&sin_dst; 1408 sin_gate.sin_family = AF_INET; 1409 /* LINTED */ 1410 info.rti_info[RTAX_GATEWAY] = (struct sockaddr_storage *)&sin_gate; 1411 sin_mask.sin_family = AF_INET; 1412 /* LINTED */ 1413 info.rti_info[RTAX_NETMASK] = (struct sockaddr_storage *)&sin_mask; 1414 sin_dst.sin_family = AF_INET; 1415 /* LINTED */ 1416 info.rti_info[RTAX_AUTHOR] = (struct sockaddr_storage *)&sin_author; 1417 1418 for (;;) { 1419 nroutes = dbuf.len / sizeof (mib2_ipRouteEntry_t); 1420 for (rp = routes; nroutes > 0; ++rp, nroutes--) { 1421 1422 /* 1423 * Ignore IRE cache, broadcast, and local address 1424 * entries; they're not subject to routing socket 1425 * control. 1426 */ 1427 if (rp->ipRouteInfo.re_ire_type & 1428 (IRE_BROADCAST | IRE_CACHE | IRE_LOCAL)) 1429 continue; 1430 1431 /* ignore multicast and link local addresses */ 1432 if (IN_MULTICAST(ntohl(rp->ipRouteDest)) || 1433 IN_LINKLOCAL(ntohl(rp->ipRouteDest))) { 1434 continue; 1435 } 1436 1437 1438 #ifdef DEBUG_KERNEL_ROUTE_READ 1439 (void) fprintf(stderr, "route type %d, ire type %08X, " 1440 "flags %08X: %s", rp->ipRouteType, 1441 rp->ipRouteInfo.re_ire_type, 1442 rp->ipRouteInfo.re_flags, 1443 naddr_ntoa(rp->ipRouteDest)); 1444 (void) fprintf(stderr, " %s", 1445 naddr_ntoa(rp->ipRouteMask)); 1446 (void) fprintf(stderr, " %s\n", 1447 naddr_ntoa(rp->ipRouteNextHop)); 1448 #endif 1449 1450 /* Fake up the needed entries */ 1451 rtm.rtm_flags = rp->ipRouteInfo.re_flags; 1452 rtm.rtm_type = RTM_GET; 1453 rtm.rtm_rmx.rmx_hopcount = rp->ipRouteMetric1; 1454 1455 (void) memset(ifname, 0, sizeof (ifname)); 1456 if (rp->ipRouteIfIndex.o_length < 1457 sizeof (rp->ipRouteIfIndex.o_bytes)) 1458 rp->ipRouteIfIndex.o_bytes[ 1459 rp->ipRouteIfIndex.o_length] = '\0'; 1460 (void) strncpy(ifname, 1461 rp->ipRouteIfIndex.o_bytes, 1462 sizeof (ifname)); 1463 1464 /* 1465 * First try to match up on gwkludge entries 1466 * before trying to match ifp by name. 1467 */ 1468 if ((ifp = gwkludge_iflookup(rp->ipRouteDest, 1469 rp->ipRouteNextHop, 1470 ntohl(rp->ipRouteMask))) == NULL) { 1471 ifp = ifwithname(ifname); 1472 if (ifp != NULL && ifp->int_phys != NULL) { 1473 ifp = ifwithname( 1474 ifp->int_phys->phyi_name); 1475 } 1476 } 1477 1478 info.rti_addrs = RTA_DST | RTA_GATEWAY | RTA_NETMASK; 1479 if (rp->ipRouteInfo.re_ire_type & IRE_HOST_REDIRECT) 1480 info.rti_addrs |= RTA_AUTHOR; 1481 sin_dst.sin_addr.s_addr = rp->ipRouteDest; 1482 sin_gate.sin_addr.s_addr = rp->ipRouteNextHop; 1483 sin_mask.sin_addr.s_addr = rp->ipRouteMask; 1484 sin_author.sin_addr.s_addr = 1485 rp->ipRouteInfo.re_src_addr; 1486 1487 /* 1488 * Note static routes and interface routes, and also 1489 * preload the image of the kernel table so that 1490 * we can later clean it, as well as avoid making 1491 * unneeded changes. Keep the old kernel routes for a 1492 * few seconds to allow a RIP or router-discovery 1493 * response to be heard. 1494 */ 1495 rtm_add(&rtm, &info, MAX_WAITTIME, 1496 ((rp->ipRouteInfo.re_ire_type & 1497 (IRE_INTERFACE|IRE_LOOPBACK)) != 0), ifp); 1498 } 1499 if (r == 0) { 1500 break; 1501 } 1502 r = getmsg(ipfd, NULL, &dbuf, &flags); 1503 } 1504 1505 hash_clean: 1506 if (ipfd != -1) 1507 (void) close(ipfd); 1508 for (i = 0; i < KHASH_SIZE; i++) { 1509 for (k = khash_bins[i]; k != NULL; k = k->k_next) { 1510 1511 /* 1512 * KS_DELETED routes have been removed from the 1513 * kernel, but we keep them around for reasons 1514 * stated in del_static(), so we skip the check 1515 * for KS_DELETED routes here. 1516 */ 1517 if ((k->k_state & (KS_CHECK|KS_DELETED)) == KS_CHECK) { 1518 1519 if (!(k->k_state & KS_DYNAMIC)) { 1520 writelog(LOG_WARNING, 1521 "%s --> %s disappeared from kernel", 1522 addrname(k->k_dst, k->k_mask, 0), 1523 naddr_ntoa(k->k_gate)); 1524 } 1525 del_static(k->k_dst, k->k_mask, k->k_gate, 1526 k->k_ifp, 1); 1527 1528 } 1529 } 1530 } 1531 } 1532 1533 1534 /* Listen to announcements from the kernel */ 1535 void 1536 read_rt(void) 1537 { 1538 long cc; 1539 struct interface *ifp; 1540 struct sockaddr_in gate_sin; 1541 in_addr_t mask, gate; 1542 union { 1543 struct { 1544 struct rt_msghdr rtm; 1545 struct sockaddr_storage addrs[RTA_NUMBITS]; 1546 } r; 1547 struct if_msghdr ifm; 1548 } m; 1549 char str[100], *strp; 1550 struct rt_addrinfo info; 1551 1552 1553 for (;;) { 1554 cc = read(rt_sock, &m, sizeof (m)); 1555 if (cc <= 0) { 1556 if (cc < 0 && errno != EWOULDBLOCK) 1557 LOGERR("read(rt_sock)"); 1558 return; 1559 } 1560 1561 if (TRACERTS) 1562 dump_rt_msg("read", &m.r.rtm, cc); 1563 1564 if (cc < m.r.rtm.rtm_msglen) { 1565 msglog("routing message truncated (%d < %d)", 1566 cc, m.r.rtm.rtm_msglen); 1567 } 1568 1569 if (m.r.rtm.rtm_version != RTM_VERSION) { 1570 msglog("bogus routing message version %d", 1571 m.r.rtm.rtm_version); 1572 continue; 1573 } 1574 1575 ifp = NULL; 1576 1577 if (m.r.rtm.rtm_type == RTM_IFINFO || 1578 m.r.rtm.rtm_type == RTM_NEWADDR || 1579 m.r.rtm.rtm_type == RTM_DELADDR) { 1580 strp = if_bit_string(m.ifm.ifm_flags, _B_TRUE); 1581 if (strp == NULL) { 1582 strp = str; 1583 (void) sprintf(str, "%#x", m.ifm.ifm_flags); 1584 } 1585 ifp = ifwithindex(m.ifm.ifm_index, 1586 m.r.rtm.rtm_type != RTM_DELADDR); 1587 if (ifp == NULL) { 1588 char ifname[LIFNAMSIZ], *ifnamep; 1589 1590 ifnamep = if_indextoname(m.ifm.ifm_index, 1591 ifname); 1592 if (ifnamep == NULL) { 1593 trace_act("note %s with flags %s" 1594 " for unknown interface index #%d", 1595 rtm_type_name(m.r.rtm.rtm_type), 1596 strp, m.ifm.ifm_index); 1597 } else { 1598 trace_act("note %s with flags %s" 1599 " for unknown interface %s", 1600 rtm_type_name(m.r.rtm.rtm_type), 1601 strp, ifnamep); 1602 } 1603 } else { 1604 trace_act("note %s with flags %s for %s", 1605 rtm_type_name(m.r.rtm.rtm_type), 1606 strp, ifp->int_name); 1607 } 1608 if (strp != str) 1609 free(strp); 1610 1611 /* 1612 * After being informed of a change to an interface, 1613 * check them all now if the check would otherwise 1614 * be a long time from now, if the interface is 1615 * not known, or if the interface has been turned 1616 * off or on. 1617 */ 1618 if (ifscan_timer.tv_sec-now.tv_sec >= 1619 CHECK_BAD_INTERVAL || ifp == NULL || 1620 ((ifp->int_if_flags ^ m.ifm.ifm_flags) & 1621 IFF_UP) != 0) 1622 ifscan_timer.tv_sec = now.tv_sec; 1623 continue; 1624 } else { 1625 if (m.r.rtm.rtm_index != 0) 1626 ifp = ifwithindex(m.r.rtm.rtm_index, 1); 1627 } 1628 1629 (void) strlcpy(str, rtm_type_name(m.r.rtm.rtm_type), 1630 sizeof (str)); 1631 strp = &str[strlen(str)]; 1632 if (m.r.rtm.rtm_type <= RTM_CHANGE) 1633 strp += snprintf(strp, sizeof (str) - (strp - str), 1634 " from pid %d", (int)m.r.rtm.rtm_pid); 1635 1636 /* LINTED */ 1637 (void) rt_xaddrs(&info, (struct sockaddr_storage *)(&m.r.rtm + 1638 1), (char *)&m + cc, m.r.rtm.rtm_addrs); 1639 1640 if (INFO_DST(&info) == 0) { 1641 trace_act("ignore %s without dst", str); 1642 continue; 1643 } 1644 1645 if (INFO_DST(&info)->ss_family != AF_INET) { 1646 trace_act("ignore %s for AF %d", str, 1647 INFO_DST(&info)->ss_family); 1648 continue; 1649 } 1650 1651 mask = ((INFO_MASK(&info) != 0) ? 1652 ntohl(S_ADDR(INFO_MASK(&info))) : 1653 (m.r.rtm.rtm_flags & RTF_HOST) ? 1654 HOST_MASK : std_mask(S_ADDR(INFO_DST(&info)))); 1655 1656 strp += snprintf(strp, sizeof (str) - (strp - str), ": %s", 1657 addrname(S_ADDR(INFO_DST(&info)), mask, 0)); 1658 1659 if (IN_MULTICAST(ntohl(S_ADDR(INFO_DST(&info)))) || 1660 IN_LINKLOCAL(ntohl(S_ADDR(INFO_DST(&info))))) { 1661 trace_act("ignore multicast/link local %s", str); 1662 continue; 1663 } 1664 1665 if (m.r.rtm.rtm_flags & RTF_LLINFO) { 1666 trace_act("ignore ARP %s", str); 1667 continue; 1668 } 1669 1670 if (get_info_gate(&INFO_GATE(&info), &gate_sin)) { 1671 gate = S_ADDR(INFO_GATE(&info)); 1672 strp += snprintf(strp, sizeof (str) - (strp - str), 1673 " --> %s", naddr_ntoa(gate)); 1674 } else { 1675 gate = 0; 1676 } 1677 1678 if (INFO_AUTHOR(&info) != 0) 1679 strp += snprintf(strp, sizeof (str) - (strp - str), 1680 " by authority of %s", 1681 saddr_ntoa(INFO_AUTHOR(&info))); 1682 1683 switch (m.r.rtm.rtm_type) { 1684 case RTM_ADD: 1685 case RTM_CHANGE: 1686 case RTM_REDIRECT: 1687 if (m.r.rtm.rtm_errno != 0) { 1688 trace_act("ignore %s with \"%s\" error", 1689 str, rip_strerror(m.r.rtm.rtm_errno)); 1690 } else { 1691 trace_act("%s", str); 1692 rtm_add(&m.r.rtm, &info, 0, 1693 !(m.r.rtm.rtm_flags & RTF_GATEWAY) && 1694 m.r.rtm.rtm_type != RTM_REDIRECT, ifp); 1695 1696 } 1697 break; 1698 1699 case RTM_DELETE: 1700 if (m.r.rtm.rtm_errno != 0 && 1701 m.r.rtm.rtm_errno != ESRCH) { 1702 trace_act("ignore %s with \"%s\" error", 1703 str, rip_strerror(m.r.rtm.rtm_errno)); 1704 } else { 1705 trace_act("%s", str); 1706 del_static(S_ADDR(INFO_DST(&info)), mask, 1707 gate, ifp, 1); 1708 } 1709 break; 1710 1711 case RTM_LOSING: 1712 trace_act("%s", str); 1713 rtm_lose(&m.r.rtm, &info); 1714 break; 1715 1716 default: 1717 trace_act("ignore %s", str); 1718 break; 1719 } 1720 } 1721 } 1722 1723 1724 /* 1725 * Disassemble a routing message. The result is an array of pointers 1726 * to sockaddr_storage structures stored in the info argument. 1727 * 1728 * ss is a pointer to the beginning of the data following the 1729 * rt_msghdr contained in the routing socket message, which consists 1730 * of a string of concatenated sockaddr structure of different types. 1731 * 1732 * Extended attributes can be appended at the end of the list. 1733 */ 1734 static int 1735 rt_xaddrs(struct rt_addrinfo *info, 1736 struct sockaddr_storage *ss, 1737 char *lim, 1738 int addrs) 1739 { 1740 int retv = 0; 1741 int i; 1742 int abit; 1743 int complaints; 1744 static int prev_complaints; 1745 1746 #define XBAD_AF 0x1 1747 #define XBAD_SHORT 0x2 1748 #define XBAD_LONG 0x4 1749 1750 (void) memset(info, 0, sizeof (*info)); 1751 info->rti_addrs = addrs; 1752 complaints = 0; 1753 for (i = 0, abit = 1; i < RTAX_MAX && (char *)ss < lim; 1754 i++, abit <<= 1) { 1755 if ((addrs & abit) == 0) 1756 continue; 1757 info->rti_info[i] = ss; 1758 /* Horrible interface here */ 1759 switch (ss->ss_family) { 1760 case AF_UNIX: 1761 /* LINTED */ 1762 ss = (struct sockaddr_storage *)( 1763 (struct sockaddr_un *)ss + 1); 1764 break; 1765 case AF_INET: 1766 /* LINTED */ 1767 ss = (struct sockaddr_storage *)( 1768 (struct sockaddr_in *)ss + 1); 1769 break; 1770 case AF_LINK: 1771 /* LINTED */ 1772 ss = (struct sockaddr_storage *)( 1773 (struct sockaddr_dl *)ss + 1); 1774 break; 1775 case AF_INET6: 1776 /* LINTED */ 1777 ss = (struct sockaddr_storage *)( 1778 (struct sockaddr_in6 *)ss + 1); 1779 break; 1780 default: 1781 if (!(prev_complaints & XBAD_AF)) 1782 writelog(LOG_WARNING, 1783 "unknown address family %d " 1784 "encountered", ss->ss_family); 1785 if (complaints & XBAD_AF) 1786 goto xaddr_done; 1787 /* LINTED */ 1788 ss = (struct sockaddr_storage *)( 1789 (struct sockaddr *)ss + 1); 1790 complaints |= XBAD_AF; 1791 info->rti_addrs &= abit - 1; 1792 addrs = info->rti_addrs; 1793 retv = -1; 1794 break; 1795 } 1796 if ((char *)ss > lim) { 1797 if (!(prev_complaints & XBAD_SHORT)) 1798 msglog("sockaddr %d too short by %d " 1799 "bytes", i + 1, (char *)ss - lim); 1800 complaints |= XBAD_SHORT; 1801 info->rti_info[i] = NULL; 1802 info->rti_addrs &= abit - 1; 1803 retv = -1; 1804 goto xaddr_done; 1805 } 1806 } 1807 1808 while (((char *)ss + sizeof (rtm_ext_t)) <= lim) { 1809 rtm_ext_t *tp; 1810 char *nxt; 1811 1812 /* LINTED: alignment */ 1813 tp = (rtm_ext_t *)ss; 1814 nxt = (char *)(tp + 1) + tp->rtmex_len; 1815 1816 if (!IS_P2ALIGNED(tp->rtmex_len, sizeof (uint32_t)) || 1817 nxt > lim) { 1818 break; 1819 } 1820 1821 /* LINTED: alignment */ 1822 ss = (struct sockaddr_storage *)nxt; 1823 } 1824 1825 if ((char *)ss != lim) { 1826 if ((char *)ss > lim) { 1827 if (!(prev_complaints & XBAD_SHORT)) 1828 msglog("routing message too short by %d bytes", 1829 (char *)ss - lim); 1830 complaints |= XBAD_SHORT; 1831 } else if (!(prev_complaints & XBAD_LONG)) { 1832 msglog("%d bytes of routing message left over", 1833 lim - (char *)ss); 1834 complaints |= XBAD_LONG; 1835 } 1836 retv = -1; 1837 } 1838 xaddr_done: 1839 prev_complaints = complaints; 1840 return (retv); 1841 } 1842 1843 1844 /* after aggregating, note routes that belong in the kernel */ 1845 static void 1846 kern_out(struct ag_info *ag) 1847 { 1848 struct khash *k; 1849 struct interface *ifp; 1850 1851 ifp = ag->ag_ifp; 1852 1853 if (ifp != NULL && ifp->int_phys != NULL) { 1854 ifp = ifwithname(ifp->int_phys->phyi_name); 1855 } 1856 1857 /* 1858 * Do not install bad routes if they are not already present. 1859 * This includes routes that had RS_NET_SYN for interfaces that 1860 * recently died. 1861 */ 1862 if (ag->ag_metric == HOPCNT_INFINITY) { 1863 k = kern_find(htonl(ag->ag_dst_h), ag->ag_mask, 1864 ag->ag_nhop, ag->ag_ifp, NULL); 1865 if (k == NULL) 1866 return; 1867 } else { 1868 k = kern_add(htonl(ag->ag_dst_h), ag->ag_mask, ag->ag_nhop, 1869 ifp); 1870 } 1871 1872 if (k->k_state & KS_NEW) { 1873 /* will need to add new entry to the kernel table */ 1874 k->k_state = KS_ADD; 1875 if (ag->ag_state & AGS_GATEWAY) 1876 k->k_state |= KS_GATEWAY; 1877 if (ag->ag_state & AGS_IF) 1878 k->k_state |= KS_IF; 1879 if (ag->ag_state & AGS_PASSIVE) 1880 k->k_state |= KS_PASSIVE; 1881 if (ag->ag_state & AGS_FILE) 1882 k->k_state |= KS_FILE; 1883 k->k_gate = ag->ag_nhop; 1884 k->k_ifp = ifp; 1885 k->k_metric = ag->ag_metric; 1886 return; 1887 } 1888 1889 if ((k->k_state & (KS_STATIC|KS_DEPRE_IF)) || 1890 ((k->k_state & (KS_IF|KS_PASSIVE)) == KS_IF)) { 1891 return; 1892 } 1893 1894 /* modify existing kernel entry if necessary */ 1895 if (k->k_gate == ag->ag_nhop && k->k_ifp == ag->ag_ifp && 1896 k->k_metric != ag->ag_metric) { 1897 /* 1898 * Must delete bad interface routes etc. 1899 * to change them. 1900 */ 1901 if (k->k_metric == HOPCNT_INFINITY) 1902 k->k_state |= KS_DEL_ADD; 1903 k->k_gate = ag->ag_nhop; 1904 k->k_metric = ag->ag_metric; 1905 k->k_state |= KS_CHANGE; 1906 } 1907 1908 /* 1909 * If the daemon thinks the route should exist, forget 1910 * about any redirections. 1911 * If the daemon thinks the route should exist, eventually 1912 * override manual intervention by the operator. 1913 */ 1914 if ((k->k_state & (KS_DYNAMIC | KS_DELETED)) != 0) { 1915 k->k_state &= ~KS_DYNAMIC; 1916 k->k_state |= (KS_ADD | KS_DEL_ADD); 1917 } 1918 1919 if ((k->k_state & KS_GATEWAY) && !(ag->ag_state & AGS_GATEWAY)) { 1920 k->k_state &= ~KS_GATEWAY; 1921 k->k_state |= (KS_ADD | KS_DEL_ADD); 1922 } else if (!(k->k_state & KS_GATEWAY) && (ag->ag_state & AGS_GATEWAY)) { 1923 k->k_state |= KS_GATEWAY; 1924 k->k_state |= (KS_ADD | KS_DEL_ADD); 1925 } 1926 1927 /* 1928 * Deleting-and-adding is necessary to change aspects of a route. 1929 * Just delete instead of deleting and then adding a bad route. 1930 * Otherwise, we want to keep the route in the kernel. 1931 */ 1932 if (k->k_metric == HOPCNT_INFINITY && (k->k_state & KS_DEL_ADD)) 1933 k->k_state |= KS_DELETE; 1934 else 1935 k->k_state &= ~KS_DELETE; 1936 #undef RT 1937 } 1938 1939 /* 1940 * Update our image of the kernel forwarding table using the given 1941 * route from our internal routing table. 1942 */ 1943 1944 /*ARGSUSED1*/ 1945 static int 1946 walk_kern(struct radix_node *rn, void *argp) 1947 { 1948 #define RT ((struct rt_entry *)rn) 1949 uint8_t metric, pref; 1950 uint_t ags = 0; 1951 int i; 1952 struct rt_spare *rts; 1953 1954 /* Do not install synthetic routes */ 1955 if (RT->rt_state & RS_NET_SYN) 1956 return (0); 1957 1958 /* 1959 * Do not install static routes here. Only 1960 * read_rt->rtm_add->kern_add should install those 1961 */ 1962 if ((RT->rt_state & RS_STATIC) && 1963 (RT->rt_spares[0].rts_origin != RO_FILE)) 1964 return (0); 1965 1966 /* Do not clobber kernel if this is a route for a dead interface */ 1967 if (RT->rt_state & RS_BADIF) 1968 return (0); 1969 1970 if (!(RT->rt_state & RS_IF)) { 1971 /* This is an ordinary route, not for an interface. */ 1972 1973 /* 1974 * aggregate, ordinary good routes without regard to 1975 * their metric 1976 */ 1977 pref = 1; 1978 ags |= (AGS_GATEWAY | AGS_SUPPRESS | AGS_AGGREGATE); 1979 1980 /* 1981 * Do not install host routes directly to hosts, to avoid 1982 * interfering with ARP entries in the kernel table. 1983 */ 1984 if (RT_ISHOST(RT) && ntohl(RT->rt_dst) == RT->rt_gate) 1985 return (0); 1986 1987 } else { 1988 /* 1989 * This is an interface route. 1990 * Do not install routes for "external" remote interfaces. 1991 */ 1992 if (RT->rt_ifp != NULL && (RT->rt_ifp->int_state & IS_EXTERNAL)) 1993 return (0); 1994 1995 /* Interfaces should override received routes. */ 1996 pref = 0; 1997 ags |= (AGS_IF | AGS_CORS_GATE); 1998 if (RT->rt_ifp != NULL && 1999 !(RT->rt_ifp->int_if_flags & IFF_LOOPBACK) && 2000 (RT->rt_ifp->int_state & (IS_PASSIVE|IS_ALIAS)) == 2001 IS_PASSIVE) { 2002 ags |= AGS_PASSIVE; 2003 } 2004 2005 /* 2006 * If it is not an interface, or an alias for an interface, 2007 * it must be a "gateway." 2008 * 2009 * If it is a "remote" interface, it is also a "gateway" to 2010 * the kernel if is not a alias. 2011 */ 2012 if (RT->rt_ifp == NULL || (RT->rt_ifp->int_state & IS_REMOTE)) { 2013 2014 ags |= (AGS_GATEWAY | AGS_SUPPRESS); 2015 2016 /* 2017 * Do not aggregate IS_PASSIVE routes. 2018 */ 2019 if (!(RT->rt_ifp->int_state & IS_PASSIVE)) 2020 ags |= AGS_AGGREGATE; 2021 } 2022 } 2023 2024 metric = RT->rt_metric; 2025 if (metric == HOPCNT_INFINITY) { 2026 /* If the route is dead, try hard to aggregate. */ 2027 pref = HOPCNT_INFINITY; 2028 ags |= (AGS_FINE_GATE | AGS_SUPPRESS); 2029 ags &= ~(AGS_IF | AGS_CORS_GATE); 2030 } 2031 2032 /* 2033 * dump all routes that have the same metric as rt_spares[0] 2034 * into the kern_table, to be added to the kernel. 2035 */ 2036 for (i = 0; i < RT->rt_num_spares; i++) { 2037 rts = &RT->rt_spares[i]; 2038 2039 /* Do not install external routes */ 2040 if (rts->rts_flags & RTS_EXTERNAL) 2041 continue; 2042 2043 if (rts->rts_metric == metric) { 2044 ag_check(RT->rt_dst, RT->rt_mask, 2045 rts->rts_router, rts->rts_ifp, rts->rts_gate, 2046 metric, pref, 0, 0, 2047 (rts->rts_origin & RO_FILE) ? (ags|AGS_FILE) : ags, 2048 kern_out); 2049 } 2050 } 2051 return (0); 2052 #undef RT 2053 } 2054 2055 2056 /* Update the kernel table to match the daemon table. */ 2057 static void 2058 fix_kern(void) 2059 { 2060 int i; 2061 struct khash *k, *pk, *knext; 2062 2063 2064 need_kern = age_timer; 2065 2066 /* Walk daemon table, updating the copy of the kernel table. */ 2067 (void) rn_walktree(rhead, walk_kern, NULL); 2068 ag_flush(0, 0, kern_out); 2069 2070 for (i = 0; i < KHASH_SIZE; i++) { 2071 pk = NULL; 2072 for (k = khash_bins[i]; k != NULL; k = knext) { 2073 knext = k->k_next; 2074 2075 /* Do not touch local interface routes */ 2076 if ((k->k_state & KS_DEPRE_IF) || 2077 (k->k_state & (KS_IF|KS_PASSIVE)) == KS_IF) { 2078 pk = k; 2079 continue; 2080 } 2081 2082 /* Do not touch static routes */ 2083 if (k->k_state & KS_STATIC) { 2084 kern_check_static(k, 0); 2085 pk = k; 2086 continue; 2087 } 2088 2089 /* check hold on routes deleted by the operator */ 2090 if (k->k_keep > now.tv_sec) { 2091 /* ensure we check when the hold is over */ 2092 LIM_SEC(need_kern, k->k_keep); 2093 pk = k; 2094 continue; 2095 } 2096 2097 if ((k->k_state & KS_DELETE) && 2098 !(k->k_state & KS_DYNAMIC)) { 2099 if ((k->k_dst == RIP_DEFAULT) && 2100 (k->k_ifp != NULL) && 2101 (kern_alternate(RIP_DEFAULT, 2102 k->k_mask, k->k_gate, k->k_ifp, 2103 NULL) == NULL)) 2104 rdisc_restore(k->k_ifp); 2105 kern_ioctl(k, RTM_DELETE, 0); 2106 if (pk != NULL) 2107 pk->k_next = knext; 2108 else 2109 khash_bins[i] = knext; 2110 free(k); 2111 continue; 2112 } 2113 2114 if (k->k_state & KS_DEL_ADD) 2115 kern_ioctl(k, RTM_DELETE, 0); 2116 2117 if (k->k_state & KS_ADD) { 2118 if ((k->k_dst == RIP_DEFAULT) && 2119 (k->k_ifp != NULL)) 2120 rdisc_suppress(k->k_ifp); 2121 kern_ioctl(k, RTM_ADD, 2122 ((0 != (k->k_state & (KS_GATEWAY | 2123 KS_DYNAMIC))) ? RTF_GATEWAY : 0)); 2124 } else if (k->k_state & KS_CHANGE) { 2125 /* 2126 * Should be using RTM_CHANGE here, but 2127 * since RTM_CHANGE is currently 2128 * not multipath-aware, and assumes 2129 * that RTF_GATEWAY implies the gateway 2130 * of the route for dst has to be 2131 * changed, we play safe, and do a del + add. 2132 */ 2133 kern_ioctl(k, RTM_DELETE, 0); 2134 kern_ioctl(k, RTM_ADD, 2135 ((0 != (k->k_state & (KS_GATEWAY | 2136 KS_DYNAMIC))) ? RTF_GATEWAY : 0)); 2137 } 2138 k->k_state &= ~(KS_ADD|KS_CHANGE|KS_DEL_ADD); 2139 2140 /* 2141 * Mark this route to be deleted in the next cycle. 2142 * This deletes routes that disappear from the 2143 * daemon table, since the normal aging code 2144 * will clear the bit for routes that have not 2145 * disappeared from the daemon table. 2146 */ 2147 k->k_state |= KS_DELETE; 2148 pk = k; 2149 } 2150 } 2151 } 2152 2153 2154 /* Delete a static route in the image of the kernel table. */ 2155 void 2156 del_static(in_addr_t dst, in_addr_t mask, in_addr_t gate, 2157 struct interface *ifp, int gone) 2158 { 2159 struct khash *k; 2160 struct rt_entry *rt; 2161 2162 /* 2163 * Just mark it in the table to be deleted next time the kernel 2164 * table is updated. 2165 * If it has already been deleted, mark it as such, and set its 2166 * keep-timer so that it will not be deleted again for a while. 2167 * This lets the operator delete a route added by the daemon 2168 * and add a replacement. 2169 */ 2170 k = kern_find(dst, mask, gate, ifp, NULL); 2171 if (k != NULL && (gate == 0 || k->k_gate == gate)) { 2172 k->k_state &= ~(KS_STATIC | KS_DYNAMIC | KS_CHECK); 2173 k->k_state |= KS_DELETE; 2174 if (gone) { 2175 k->k_state |= KS_DELETED; 2176 k->k_keep = now.tv_sec + K_KEEP_LIM; 2177 } 2178 } 2179 2180 rt = rtget(dst, mask); 2181 if (rt != NULL && (rt->rt_state & RS_STATIC)) 2182 rtbad(rt, NULL); 2183 } 2184 2185 2186 /* 2187 * Delete all routes generated from ICMP Redirects that use a given gateway, 2188 * as well as old redirected routes. 2189 */ 2190 void 2191 del_redirects(in_addr_t bad_gate, time_t old) 2192 { 2193 int i; 2194 struct khash *k; 2195 boolean_t dosupply = should_supply(NULL); 2196 2197 for (i = 0; i < KHASH_SIZE; i++) { 2198 for (k = khash_bins[i]; k != NULL; k = k->k_next) { 2199 if (!(k->k_state & KS_DYNAMIC) || 2200 (k->k_state & (KS_STATIC|KS_IF|KS_DEPRE_IF))) 2201 continue; 2202 2203 if (k->k_gate != bad_gate && k->k_redirect_time > old && 2204 !dosupply) 2205 continue; 2206 2207 k->k_state |= KS_DELETE; 2208 k->k_state &= ~KS_DYNAMIC; 2209 need_kern.tv_sec = now.tv_sec; 2210 trace_act("mark redirected %s --> %s for deletion", 2211 addrname(k->k_dst, k->k_mask, 0), 2212 naddr_ntoa(k->k_gate)); 2213 } 2214 } 2215 } 2216 2217 /* Start the daemon tables. */ 2218 void 2219 rtinit(void) 2220 { 2221 int i; 2222 struct ag_info *ag; 2223 2224 /* Initialize the radix trees */ 2225 rn_init(); 2226 (void) rn_inithead((void**)&rhead, 32); 2227 2228 /* mark all of the slots in the table free */ 2229 ag_avail = ag_slots; 2230 for (ag = ag_slots, i = 1; i < NUM_AG_SLOTS; i++) { 2231 ag->ag_fine = ag+1; 2232 ag++; 2233 } 2234 } 2235 2236 2237 static struct sockaddr_in dst_sock = {AF_INET}; 2238 static struct sockaddr_in mask_sock = {AF_INET}; 2239 2240 2241 static void 2242 set_need_flash(void) 2243 { 2244 if (!need_flash) { 2245 need_flash = _B_TRUE; 2246 /* 2247 * Do not send the flash update immediately. Wait a little 2248 * while to hear from other routers. 2249 */ 2250 no_flash.tv_sec = now.tv_sec + MIN_WAITTIME; 2251 } 2252 } 2253 2254 2255 /* Get a particular routing table entry */ 2256 struct rt_entry * 2257 rtget(in_addr_t dst, in_addr_t mask) 2258 { 2259 struct rt_entry *rt; 2260 2261 dst_sock.sin_addr.s_addr = dst; 2262 mask_sock.sin_addr.s_addr = htonl(mask); 2263 rt = (struct rt_entry *)rhead->rnh_lookup(&dst_sock, &mask_sock, rhead); 2264 if (rt == NULL || rt->rt_dst != dst || rt->rt_mask != mask) 2265 return (NULL); 2266 2267 return (rt); 2268 } 2269 2270 2271 /* Find a route to dst as the kernel would. */ 2272 struct rt_entry * 2273 rtfind(in_addr_t dst) 2274 { 2275 dst_sock.sin_addr.s_addr = dst; 2276 return ((struct rt_entry *)rhead->rnh_matchaddr(&dst_sock, rhead)); 2277 } 2278 2279 2280 /* add a route to the table */ 2281 void 2282 rtadd(in_addr_t dst, 2283 in_addr_t mask, 2284 uint16_t state, /* rt_state for the entry */ 2285 struct rt_spare *new) 2286 { 2287 struct rt_entry *rt; 2288 in_addr_t smask; 2289 int i; 2290 struct rt_spare *rts; 2291 2292 /* This is the only function that increments total_routes. */ 2293 if (total_routes == MAX_ROUTES) { 2294 msglog("have maximum (%d) routes", total_routes); 2295 return; 2296 } 2297 2298 rt = rtmalloc(sizeof (*rt), "rtadd"); 2299 (void) memset(rt, 0, sizeof (*rt)); 2300 rt->rt_spares = rtmalloc(SPARE_INC * sizeof (struct rt_spare), 2301 "rtadd"); 2302 rt->rt_num_spares = SPARE_INC; 2303 (void) memset(rt->rt_spares, 0, SPARE_INC * sizeof (struct rt_spare)); 2304 for (rts = rt->rt_spares, i = rt->rt_num_spares; i != 0; i--, rts++) 2305 rts->rts_metric = HOPCNT_INFINITY; 2306 2307 rt->rt_nodes->rn_key = (uint8_t *)&rt->rt_dst_sock; 2308 rt->rt_dst = dst; 2309 rt->rt_dst_sock.sin_family = AF_INET; 2310 if (mask != HOST_MASK) { 2311 smask = std_mask(dst); 2312 if ((smask & ~mask) == 0 && mask > smask) 2313 state |= RS_SUBNET; 2314 } 2315 mask_sock.sin_addr.s_addr = htonl(mask); 2316 rt->rt_mask = mask; 2317 rt->rt_spares[0] = *new; 2318 rt->rt_state = state; 2319 rt->rt_time = now.tv_sec; 2320 rt->rt_poison_metric = HOPCNT_INFINITY; 2321 rt->rt_seqno = update_seqno; 2322 2323 if (TRACEACTIONS) 2324 trace_add_del("Add", rt); 2325 2326 need_kern.tv_sec = now.tv_sec; 2327 set_need_flash(); 2328 2329 if (NULL == rhead->rnh_addaddr(&rt->rt_dst_sock, &mask_sock, rhead, 2330 rt->rt_nodes)) { 2331 msglog("rnh_addaddr() failed for %s mask=%s", 2332 naddr_ntoa(dst), naddr_ntoa(htonl(mask))); 2333 free(rt); 2334 } 2335 2336 total_routes++; 2337 } 2338 2339 2340 /* notice a changed route */ 2341 void 2342 rtchange(struct rt_entry *rt, 2343 uint16_t state, /* new state bits */ 2344 struct rt_spare *new, 2345 char *label) 2346 { 2347 if (rt->rt_metric != new->rts_metric) { 2348 /* 2349 * Fix the kernel immediately if it seems the route 2350 * has gone bad, since there may be a working route that 2351 * aggregates this route. 2352 */ 2353 if (new->rts_metric == HOPCNT_INFINITY) { 2354 need_kern.tv_sec = now.tv_sec; 2355 if (new->rts_time >= now.tv_sec - EXPIRE_TIME) 2356 new->rts_time = now.tv_sec - EXPIRE_TIME; 2357 } 2358 rt->rt_seqno = update_seqno; 2359 set_need_flash(); 2360 } 2361 2362 if (rt->rt_gate != new->rts_gate) { 2363 need_kern.tv_sec = now.tv_sec; 2364 rt->rt_seqno = update_seqno; 2365 set_need_flash(); 2366 } 2367 2368 state |= (rt->rt_state & RS_SUBNET); 2369 2370 /* Keep various things from deciding ageless routes are stale. */ 2371 if (!AGE_RT(state, rt->rt_spares[0].rts_origin, new->rts_ifp)) 2372 new->rts_time = now.tv_sec; 2373 2374 if (TRACEACTIONS) 2375 trace_change(rt, state, new, 2376 label ? label : "Chg "); 2377 2378 rt->rt_state = state; 2379 /* 2380 * If the interface state of the new primary route is good, 2381 * turn off RS_BADIF flag 2382 */ 2383 if ((rt->rt_state & RS_BADIF) && 2384 IS_IFF_UP(new->rts_ifp->int_if_flags) && 2385 !(new->rts_ifp->int_state & (IS_BROKE | IS_SICK))) 2386 rt->rt_state &= ~(RS_BADIF); 2387 2388 rt->rt_spares[0] = *new; 2389 } 2390 2391 2392 /* check for a better route among the spares */ 2393 static struct rt_spare * 2394 rts_better(struct rt_entry *rt) 2395 { 2396 struct rt_spare *rts, *rts1; 2397 int i; 2398 2399 /* find the best alternative among the spares */ 2400 rts = rt->rt_spares+1; 2401 for (i = rt->rt_num_spares, rts1 = rts+1; i > 2; i--, rts1++) { 2402 if (BETTER_LINK(rt, rts1, rts)) 2403 rts = rts1; 2404 } 2405 2406 return (rts); 2407 } 2408 2409 2410 /* switch to a backup route */ 2411 void 2412 rtswitch(struct rt_entry *rt, 2413 struct rt_spare *rts) 2414 { 2415 struct rt_spare swap; 2416 char label[10]; 2417 2418 /* Do not change permanent routes */ 2419 if (0 != (rt->rt_state & (RS_MHOME | RS_STATIC | 2420 RS_NET_SYN | RS_IF))) 2421 return; 2422 2423 /* find the best alternative among the spares */ 2424 if (rts == NULL) 2425 rts = rts_better(rt); 2426 2427 /* Do not bother if it is not worthwhile. */ 2428 if (!BETTER_LINK(rt, rts, rt->rt_spares)) 2429 return; 2430 2431 swap = rt->rt_spares[0]; 2432 (void) snprintf(label, sizeof (label), "Use #%d", 2433 (int)(rts - rt->rt_spares)); 2434 rtchange(rt, rt->rt_state & ~(RS_NET_SYN), rts, label); 2435 2436 if (swap.rts_metric == HOPCNT_INFINITY) { 2437 *rts = rts_empty; 2438 } else { 2439 *rts = swap; 2440 } 2441 2442 } 2443 2444 2445 void 2446 rtdelete(struct rt_entry *rt) 2447 { 2448 struct rt_entry *deleted_rt; 2449 struct rt_spare *rts; 2450 int i; 2451 in_addr_t gate = rt->rt_gate; /* for debugging */ 2452 2453 if (TRACEACTIONS) 2454 trace_add_del("Del", rt); 2455 2456 for (i = 0; i < rt->rt_num_spares; i++) { 2457 rts = &rt->rt_spares[i]; 2458 rts_delete(rt, rts); 2459 } 2460 2461 dst_sock.sin_addr.s_addr = rt->rt_dst; 2462 mask_sock.sin_addr.s_addr = htonl(rt->rt_mask); 2463 if (rt != (deleted_rt = 2464 ((struct rt_entry *)rhead->rnh_deladdr(&dst_sock, &mask_sock, 2465 rhead)))) { 2466 msglog("rnh_deladdr(%s) failed; found rt 0x%lx", 2467 rtname(rt->rt_dst, rt->rt_mask, gate), deleted_rt); 2468 if (deleted_rt != NULL) 2469 free(deleted_rt); 2470 } 2471 total_routes--; 2472 free(rt->rt_spares); 2473 free(rt); 2474 2475 if (dst_sock.sin_addr.s_addr == RIP_DEFAULT) { 2476 /* 2477 * we just deleted the default route. Trigger rdisc_sort 2478 * so that we can recover from any rdisc information that 2479 * is valid 2480 */ 2481 rdisc_timer.tv_sec = 0; 2482 } 2483 } 2484 2485 void 2486 rts_delete(struct rt_entry *rt, struct rt_spare *rts) 2487 { 2488 struct khash *k; 2489 2490 trace_upslot(rt, rts, &rts_empty); 2491 k = kern_find(rt->rt_dst, rt->rt_mask, 2492 rts->rts_gate, rts->rts_ifp, NULL); 2493 if (k != NULL && 2494 !(k->k_state & KS_DEPRE_IF) && 2495 ((k->k_state & (KS_IF|KS_PASSIVE)) != KS_IF)) { 2496 k->k_state |= KS_DELETE; 2497 need_kern.tv_sec = now.tv_sec; 2498 } 2499 2500 *rts = rts_empty; 2501 } 2502 2503 /* 2504 * Get rid of a bad route, and try to switch to a replacement. 2505 * If the route has gone bad because of a bad interface, 2506 * the information about the dead interface is available in badifp 2507 * for the purpose of sanity checks, if_flags checks etc. 2508 */ 2509 static void 2510 rtbad(struct rt_entry *rt, struct interface *badifp) 2511 { 2512 struct rt_spare new; 2513 uint16_t rt_state; 2514 2515 2516 if (badifp == NULL || (rt->rt_spares[0].rts_ifp == badifp)) { 2517 /* Poison the route */ 2518 new = rt->rt_spares[0]; 2519 new.rts_metric = HOPCNT_INFINITY; 2520 rt_state = rt->rt_state & ~(RS_IF | RS_LOCAL | RS_STATIC); 2521 } 2522 2523 if (badifp != NULL) { 2524 /* 2525 * Dont mark the rtentry bad unless the ifp for the primary 2526 * route is the bad ifp 2527 */ 2528 if (rt->rt_spares[0].rts_ifp != badifp) 2529 return; 2530 /* 2531 * badifp has just gone bad. We want to keep this 2532 * rt_entry around so that we tell our rip-neighbors 2533 * about the bad route, but we can't do anything 2534 * to the kernel itself, so mark it as RS_BADIF 2535 */ 2536 trace_misc("rtbad:Setting RS_BADIF (%s)", badifp->int_name); 2537 rt_state |= RS_BADIF; 2538 new.rts_ifp = &dummy_ifp; 2539 } 2540 rtchange(rt, rt_state, &new, 0); 2541 rtswitch(rt, 0); 2542 } 2543 2544 2545 /* 2546 * Junk a RS_NET_SYN or RS_LOCAL route, 2547 * unless it is needed by another interface. 2548 */ 2549 void 2550 rtbad_sub(struct rt_entry *rt, struct interface *badifp) 2551 { 2552 struct interface *ifp, *ifp1; 2553 struct intnet *intnetp; 2554 uint_t state; 2555 2556 2557 ifp1 = NULL; 2558 state = 0; 2559 2560 if (rt->rt_state & RS_LOCAL) { 2561 /* 2562 * Is this the route through loopback for the interface? 2563 * If so, see if it is used by any other interfaces, such 2564 * as a point-to-point interface with the same local address. 2565 */ 2566 for (ifp = ifnet; ifp != NULL; ifp = ifp->int_next) { 2567 /* Retain it if another interface needs it. */ 2568 if (ifp->int_addr == rt->rt_ifp->int_addr) { 2569 state |= RS_LOCAL; 2570 ifp1 = ifp; 2571 break; 2572 } 2573 } 2574 2575 } 2576 2577 if (!(state & RS_LOCAL)) { 2578 /* 2579 * Retain RIPv1 logical network route if there is another 2580 * interface that justifies it. 2581 */ 2582 if (rt->rt_state & RS_NET_SYN) { 2583 for (ifp = ifnet; ifp != NULL; ifp = ifp->int_next) { 2584 if ((ifp->int_state & IS_NEED_NET_SYN) && 2585 rt->rt_mask == ifp->int_std_mask && 2586 rt->rt_dst == ifp->int_std_addr) { 2587 state |= RS_NET_SYN; 2588 ifp1 = ifp; 2589 break; 2590 } 2591 } 2592 } 2593 2594 /* or if there is an authority route that needs it. */ 2595 for (intnetp = intnets; intnetp != NULL; 2596 intnetp = intnetp->intnet_next) { 2597 if (intnetp->intnet_addr == rt->rt_dst && 2598 intnetp->intnet_mask == rt->rt_mask) { 2599 state |= (RS_NET_SYN | RS_NET_INT); 2600 break; 2601 } 2602 } 2603 } 2604 2605 if (ifp1 != NULL || (state & RS_NET_SYN)) { 2606 struct rt_spare new = rt->rt_spares[0]; 2607 new.rts_ifp = ifp1; 2608 rtchange(rt, ((rt->rt_state & ~(RS_NET_SYN|RS_LOCAL)) | state), 2609 &new, 0); 2610 } else { 2611 rtbad(rt, badifp); 2612 } 2613 } 2614 2615 /* 2616 * Called while walking the table looking for sick interfaces 2617 * or after a time change. 2618 */ 2619 int 2620 walk_bad(struct radix_node *rn, 2621 void *argp) 2622 { 2623 #define RT ((struct rt_entry *)rn) 2624 struct rt_spare *rts; 2625 int i, j = -1; 2626 2627 /* fix any spare routes through the interface */ 2628 for (i = 1; i < RT->rt_num_spares; i++) { 2629 rts = &((struct rt_entry *)rn)->rt_spares[i]; 2630 2631 if (rts->rts_metric < HOPCNT_INFINITY && 2632 (rts->rts_ifp == NULL || 2633 (rts->rts_ifp->int_state & IS_BROKE))) 2634 rts_delete(RT, rts); 2635 else { 2636 if (rts->rts_origin != RO_NONE) 2637 j = i; 2638 } 2639 } 2640 2641 /* 2642 * Deal with the main route 2643 * finished if it has been handled before or if its interface is ok 2644 */ 2645 if (RT->rt_ifp == NULL || !(RT->rt_ifp->int_state & IS_BROKE)) 2646 return (0); 2647 2648 /* Bad routes for other than interfaces are easy. */ 2649 if (!(RT->rt_state & (RS_IF | RS_NET_SYN | RS_LOCAL))) { 2650 if (j > 0) { 2651 RT->rt_spares[0].rts_metric = HOPCNT_INFINITY; 2652 rtswitch(RT, NULL); 2653 } else { 2654 rtbad(RT, (struct interface *)argp); 2655 } 2656 return (0); 2657 } 2658 2659 rtbad_sub(RT, (struct interface *)argp); 2660 return (0); 2661 #undef RT 2662 } 2663 2664 /* 2665 * Called while walking the table to replace a duplicate interface 2666 * with a backup. 2667 */ 2668 int 2669 walk_rewire(struct radix_node *rn, void *argp) 2670 { 2671 struct rt_entry *RT = (struct rt_entry *)rn; 2672 struct rewire_data *wire = (struct rewire_data *)argp; 2673 struct rt_spare *rts; 2674 int i; 2675 2676 /* fix any spare routes through the interface */ 2677 rts = RT->rt_spares; 2678 for (i = RT->rt_num_spares; i > 0; i--, rts++) { 2679 if (rts->rts_ifp == wire->if_old) { 2680 rts->rts_ifp = wire->if_new; 2681 if ((RT->rt_dst == RIP_DEFAULT) && 2682 (wire->if_old->int_state & IS_SUPPRESS_RDISC)) 2683 rdisc_suppress(rts->rts_ifp); 2684 if ((rts->rts_metric += wire->metric_delta) > 2685 HOPCNT_INFINITY) 2686 rts->rts_metric = HOPCNT_INFINITY; 2687 2688 /* 2689 * If the main route is getting a worse metric, 2690 * then it may be time to switch to a backup. 2691 */ 2692 if (i == RT->rt_num_spares && wire->metric_delta > 0) { 2693 rtswitch(RT, NULL); 2694 } 2695 } 2696 } 2697 2698 return (0); 2699 } 2700 2701 /* Check the age of an individual route. */ 2702 static int 2703 walk_age(struct radix_node *rn, void *argp) 2704 { 2705 #define RT ((struct rt_entry *)rn) 2706 struct interface *ifp; 2707 struct rt_spare *rts; 2708 int i; 2709 in_addr_t age_bad_gate = *(in_addr_t *)argp; 2710 2711 2712 /* 2713 * age all of the spare routes, including the primary route 2714 * currently in use 2715 */ 2716 rts = RT->rt_spares; 2717 for (i = RT->rt_num_spares; i != 0; i--, rts++) { 2718 2719 ifp = rts->rts_ifp; 2720 if (i == RT->rt_num_spares) { 2721 if (!AGE_RT(RT->rt_state, rts->rts_origin, ifp)) { 2722 /* 2723 * Keep various things from deciding ageless 2724 * routes are stale 2725 */ 2726 rts->rts_time = now.tv_sec; 2727 continue; 2728 } 2729 2730 /* forget RIP routes after RIP has been turned off. */ 2731 if (rip_sock < 0) { 2732 rts->rts_time = now_stale + 1; 2733 } 2734 } 2735 2736 /* age failing routes */ 2737 if (age_bad_gate == rts->rts_gate && 2738 rts->rts_time >= now_stale) { 2739 rts->rts_time -= SUPPLY_INTERVAL; 2740 } 2741 2742 /* trash the spare routes when they go bad */ 2743 if (rts->rts_origin == RO_RIP && 2744 ((rip_sock < 0) || 2745 (rts->rts_metric < HOPCNT_INFINITY && 2746 now_garbage > rts->rts_time)) && 2747 i != RT->rt_num_spares) { 2748 rts_delete(RT, rts); 2749 } 2750 } 2751 2752 2753 /* finished if the active route is still fresh */ 2754 if (now_stale <= RT->rt_time) 2755 return (0); 2756 2757 /* try to switch to an alternative */ 2758 rtswitch(RT, NULL); 2759 2760 /* Delete a dead route after it has been publically mourned. */ 2761 if (now_garbage > RT->rt_time) { 2762 rtdelete(RT); 2763 return (0); 2764 } 2765 2766 /* Start poisoning a bad route before deleting it. */ 2767 if (now.tv_sec - RT->rt_time > EXPIRE_TIME) { 2768 struct rt_spare new = RT->rt_spares[0]; 2769 2770 new.rts_metric = HOPCNT_INFINITY; 2771 rtchange(RT, RT->rt_state, &new, 0); 2772 } 2773 return (0); 2774 } 2775 2776 2777 /* Watch for dead routes and interfaces. */ 2778 void 2779 age(in_addr_t bad_gate) 2780 { 2781 struct interface *ifp; 2782 int need_query = 0; 2783 2784 /* 2785 * If not listening to RIP, there is no need to age the routes in 2786 * the table. 2787 */ 2788 age_timer.tv_sec = (now.tv_sec 2789 + ((rip_sock < 0) ? NEVER : SUPPLY_INTERVAL)); 2790 2791 /* 2792 * Check for dead IS_REMOTE interfaces by timing their 2793 * transmissions. 2794 */ 2795 for (ifp = ifnet; ifp; ifp = ifp->int_next) { 2796 if (!(ifp->int_state & IS_REMOTE)) 2797 continue; 2798 2799 /* ignore unreachable remote interfaces */ 2800 if (!check_remote(ifp)) 2801 continue; 2802 2803 /* Restore remote interface that has become reachable */ 2804 if (ifp->int_state & IS_BROKE) 2805 if_ok(ifp, "remote ", _B_FALSE); 2806 2807 if (ifp->int_act_time != NEVER && 2808 now.tv_sec - ifp->int_act_time > EXPIRE_TIME) { 2809 writelog(LOG_NOTICE, 2810 "remote interface %s to %s timed out after" 2811 " %ld:%ld", 2812 ifp->int_name, 2813 naddr_ntoa(ifp->int_dstaddr), 2814 (now.tv_sec - ifp->int_act_time)/60, 2815 (now.tv_sec - ifp->int_act_time)%60); 2816 if_sick(ifp, _B_FALSE); 2817 } 2818 2819 /* 2820 * If we have not heard from the other router 2821 * recently, ask it. 2822 */ 2823 if (now.tv_sec >= ifp->int_query_time) { 2824 ifp->int_query_time = NEVER; 2825 need_query = 1; 2826 } 2827 } 2828 2829 /* Age routes. */ 2830 (void) rn_walktree(rhead, walk_age, &bad_gate); 2831 2832 /* 2833 * delete old redirected routes to keep the kernel table small 2834 * and prevent blackholes 2835 */ 2836 del_redirects(bad_gate, now.tv_sec-STALE_TIME); 2837 2838 /* Update the kernel routing table. */ 2839 fix_kern(); 2840 2841 /* poke reticent remote gateways */ 2842 if (need_query) 2843 rip_query(); 2844 } 2845 2846 void 2847 kern_dump(void) 2848 { 2849 int i; 2850 struct khash *k; 2851 2852 for (i = 0; i < KHASH_SIZE; i++) { 2853 for (k = khash_bins[i]; k != NULL; k = k->k_next) 2854 trace_khash(k); 2855 } 2856 } 2857 2858 2859 static struct interface * 2860 gwkludge_iflookup(in_addr_t dstaddr, in_addr_t addr, in_addr_t mask) 2861 { 2862 uint32_t int_state; 2863 struct interface *ifp; 2864 2865 for (ifp = ifnet; ifp != NULL; ifp = ifp->int_next) { 2866 int_state = ifp->int_state; 2867 2868 if (!(int_state & IS_REMOTE)) 2869 continue; 2870 2871 if (ifp->int_dstaddr == dstaddr && ifp->int_addr == addr && 2872 ifp->int_mask == mask) 2873 return (ifp); 2874 } 2875 return (NULL); 2876 } 2877