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 addresses */ 1432 if (IN_MULTICAST(ntohl(rp->ipRouteDest))) 1433 continue; 1434 1435 1436 #ifdef DEBUG_KERNEL_ROUTE_READ 1437 (void) fprintf(stderr, "route type %d, ire type %08X, " 1438 "flags %08X: %s", rp->ipRouteType, 1439 rp->ipRouteInfo.re_ire_type, 1440 rp->ipRouteInfo.re_flags, 1441 naddr_ntoa(rp->ipRouteDest)); 1442 (void) fprintf(stderr, " %s", 1443 naddr_ntoa(rp->ipRouteMask)); 1444 (void) fprintf(stderr, " %s\n", 1445 naddr_ntoa(rp->ipRouteNextHop)); 1446 #endif 1447 1448 /* Fake up the needed entries */ 1449 rtm.rtm_flags = rp->ipRouteInfo.re_flags; 1450 rtm.rtm_type = RTM_GET; 1451 rtm.rtm_rmx.rmx_hopcount = rp->ipRouteMetric1; 1452 1453 (void) memset(ifname, 0, sizeof (ifname)); 1454 if (rp->ipRouteIfIndex.o_length < 1455 sizeof (rp->ipRouteIfIndex.o_bytes)) 1456 rp->ipRouteIfIndex.o_bytes[ 1457 rp->ipRouteIfIndex.o_length] = '\0'; 1458 (void) strncpy(ifname, 1459 rp->ipRouteIfIndex.o_bytes, 1460 sizeof (ifname)); 1461 1462 /* 1463 * First try to match up on gwkludge entries 1464 * before trying to match ifp by name. 1465 */ 1466 if ((ifp = gwkludge_iflookup(rp->ipRouteDest, 1467 rp->ipRouteNextHop, 1468 ntohl(rp->ipRouteMask))) == NULL) { 1469 ifp = ifwithname(ifname); 1470 if (ifp != NULL && ifp->int_phys != NULL) { 1471 ifp = ifwithname( 1472 ifp->int_phys->phyi_name); 1473 } 1474 } 1475 1476 info.rti_addrs = RTA_DST | RTA_GATEWAY | RTA_NETMASK; 1477 if (rp->ipRouteInfo.re_ire_type & IRE_HOST_REDIRECT) 1478 info.rti_addrs |= RTA_AUTHOR; 1479 sin_dst.sin_addr.s_addr = rp->ipRouteDest; 1480 sin_gate.sin_addr.s_addr = rp->ipRouteNextHop; 1481 sin_mask.sin_addr.s_addr = rp->ipRouteMask; 1482 sin_author.sin_addr.s_addr = 1483 rp->ipRouteInfo.re_src_addr; 1484 1485 /* 1486 * Note static routes and interface routes, and also 1487 * preload the image of the kernel table so that 1488 * we can later clean it, as well as avoid making 1489 * unneeded changes. Keep the old kernel routes for a 1490 * few seconds to allow a RIP or router-discovery 1491 * response to be heard. 1492 */ 1493 rtm_add(&rtm, &info, MAX_WAITTIME, 1494 ((rp->ipRouteInfo.re_ire_type & 1495 (IRE_INTERFACE|IRE_LOOPBACK)) != 0), ifp); 1496 } 1497 if (r == 0) { 1498 break; 1499 } 1500 r = getmsg(ipfd, NULL, &dbuf, &flags); 1501 } 1502 1503 hash_clean: 1504 if (ipfd != -1) 1505 (void) close(ipfd); 1506 for (i = 0; i < KHASH_SIZE; i++) { 1507 for (k = khash_bins[i]; k != NULL; k = k->k_next) { 1508 1509 /* 1510 * KS_DELETED routes have been removed from the 1511 * kernel, but we keep them around for reasons 1512 * stated in del_static(), so we skip the check 1513 * for KS_DELETED routes here. 1514 */ 1515 if ((k->k_state & (KS_CHECK|KS_DELETED)) == KS_CHECK) { 1516 1517 if (!(k->k_state & KS_DYNAMIC)) 1518 writelog(LOG_WARNING, 1519 "%s --> %s disappeared from kernel", 1520 addrname(k->k_dst, k->k_mask, 0), 1521 naddr_ntoa(k->k_gate)); 1522 del_static(k->k_dst, k->k_mask, k->k_gate, 1523 k->k_ifp, 1); 1524 1525 } 1526 } 1527 } 1528 } 1529 1530 1531 /* Listen to announcements from the kernel */ 1532 void 1533 read_rt(void) 1534 { 1535 long cc; 1536 struct interface *ifp; 1537 struct sockaddr_in gate_sin; 1538 in_addr_t mask, gate; 1539 union { 1540 struct { 1541 struct rt_msghdr rtm; 1542 struct sockaddr_storage addrs[RTA_NUMBITS]; 1543 } r; 1544 struct if_msghdr ifm; 1545 } m; 1546 char str[100], *strp; 1547 struct rt_addrinfo info; 1548 1549 1550 for (;;) { 1551 cc = read(rt_sock, &m, sizeof (m)); 1552 if (cc <= 0) { 1553 if (cc < 0 && errno != EWOULDBLOCK) 1554 LOGERR("read(rt_sock)"); 1555 return; 1556 } 1557 1558 if (TRACERTS) 1559 dump_rt_msg("read", &m.r.rtm, cc); 1560 1561 if (cc < m.r.rtm.rtm_msglen) { 1562 msglog("routing message truncated (%d < %d)", 1563 cc, m.r.rtm.rtm_msglen); 1564 } 1565 1566 if (m.r.rtm.rtm_version != RTM_VERSION) { 1567 msglog("bogus routing message version %d", 1568 m.r.rtm.rtm_version); 1569 continue; 1570 } 1571 1572 ifp = NULL; 1573 1574 if (m.r.rtm.rtm_type == RTM_IFINFO || 1575 m.r.rtm.rtm_type == RTM_NEWADDR || 1576 m.r.rtm.rtm_type == RTM_DELADDR) { 1577 strp = if_bit_string(m.ifm.ifm_flags, _B_TRUE); 1578 if (strp == NULL) { 1579 strp = str; 1580 (void) sprintf(str, "%#x", m.ifm.ifm_flags); 1581 } 1582 ifp = ifwithindex(m.ifm.ifm_index, 1583 m.r.rtm.rtm_type != RTM_DELADDR); 1584 if (ifp == NULL) { 1585 char ifname[LIFNAMSIZ], *ifnamep; 1586 1587 ifnamep = if_indextoname(m.ifm.ifm_index, 1588 ifname); 1589 if (ifnamep == NULL) { 1590 trace_act("note %s with flags %s" 1591 " for unknown interface index #%d", 1592 rtm_type_name(m.r.rtm.rtm_type), 1593 strp, m.ifm.ifm_index); 1594 } else { 1595 trace_act("note %s with flags %s" 1596 " for unknown interface %s", 1597 rtm_type_name(m.r.rtm.rtm_type), 1598 strp, ifnamep); 1599 } 1600 } else { 1601 trace_act("note %s with flags %s for %s", 1602 rtm_type_name(m.r.rtm.rtm_type), 1603 strp, ifp->int_name); 1604 } 1605 if (strp != str) 1606 free(strp); 1607 1608 /* 1609 * After being informed of a change to an interface, 1610 * check them all now if the check would otherwise 1611 * be a long time from now, if the interface is 1612 * not known, or if the interface has been turned 1613 * off or on. 1614 */ 1615 if (ifscan_timer.tv_sec-now.tv_sec >= 1616 CHECK_BAD_INTERVAL || ifp == NULL || 1617 ((ifp->int_if_flags ^ m.ifm.ifm_flags) & 1618 IFF_UP) != 0) 1619 ifscan_timer.tv_sec = now.tv_sec; 1620 continue; 1621 } else { 1622 if (m.r.rtm.rtm_index != 0) 1623 ifp = ifwithindex(m.r.rtm.rtm_index, 1); 1624 } 1625 1626 (void) strlcpy(str, rtm_type_name(m.r.rtm.rtm_type), 1627 sizeof (str)); 1628 strp = &str[strlen(str)]; 1629 if (m.r.rtm.rtm_type <= RTM_CHANGE) 1630 strp += snprintf(strp, sizeof (str) - (strp - str), 1631 " from pid %d", (int)m.r.rtm.rtm_pid); 1632 1633 /* LINTED */ 1634 (void) rt_xaddrs(&info, (struct sockaddr_storage *)(&m.r.rtm + 1635 1), (char *)&m + cc, m.r.rtm.rtm_addrs); 1636 1637 if (INFO_DST(&info) == 0) { 1638 trace_act("ignore %s without dst", str); 1639 continue; 1640 } 1641 1642 if (INFO_DST(&info)->ss_family != AF_INET) { 1643 trace_act("ignore %s for AF %d", str, 1644 INFO_DST(&info)->ss_family); 1645 continue; 1646 } 1647 1648 mask = ((INFO_MASK(&info) != 0) ? 1649 ntohl(S_ADDR(INFO_MASK(&info))) : 1650 (m.r.rtm.rtm_flags & RTF_HOST) ? 1651 HOST_MASK : std_mask(S_ADDR(INFO_DST(&info)))); 1652 1653 strp += snprintf(strp, sizeof (str) - (strp - str), ": %s", 1654 addrname(S_ADDR(INFO_DST(&info)), mask, 0)); 1655 1656 if (IN_MULTICAST(ntohl(S_ADDR(INFO_DST(&info))))) { 1657 trace_act("ignore multicast %s", str); 1658 continue; 1659 } 1660 1661 if (m.r.rtm.rtm_flags & RTF_LLINFO) { 1662 trace_act("ignore ARP %s", str); 1663 continue; 1664 } 1665 1666 if (get_info_gate(&INFO_GATE(&info), &gate_sin)) { 1667 gate = S_ADDR(INFO_GATE(&info)); 1668 strp += snprintf(strp, sizeof (str) - (strp - str), 1669 " --> %s", naddr_ntoa(gate)); 1670 } else { 1671 gate = 0; 1672 } 1673 1674 if (INFO_AUTHOR(&info) != 0) 1675 strp += snprintf(strp, sizeof (str) - (strp - str), 1676 " by authority of %s", 1677 saddr_ntoa(INFO_AUTHOR(&info))); 1678 1679 switch (m.r.rtm.rtm_type) { 1680 case RTM_ADD: 1681 case RTM_CHANGE: 1682 case RTM_REDIRECT: 1683 if (m.r.rtm.rtm_errno != 0) { 1684 trace_act("ignore %s with \"%s\" error", 1685 str, rip_strerror(m.r.rtm.rtm_errno)); 1686 } else { 1687 trace_act("%s", str); 1688 rtm_add(&m.r.rtm, &info, 0, 1689 !(m.r.rtm.rtm_flags & RTF_GATEWAY) && 1690 m.r.rtm.rtm_type != RTM_REDIRECT, ifp); 1691 1692 } 1693 break; 1694 1695 case RTM_DELETE: 1696 if (m.r.rtm.rtm_errno != 0 && 1697 m.r.rtm.rtm_errno != ESRCH) { 1698 trace_act("ignore %s with \"%s\" error", 1699 str, rip_strerror(m.r.rtm.rtm_errno)); 1700 } else { 1701 trace_act("%s", str); 1702 del_static(S_ADDR(INFO_DST(&info)), mask, 1703 gate, ifp, 1); 1704 } 1705 break; 1706 1707 case RTM_LOSING: 1708 trace_act("%s", str); 1709 rtm_lose(&m.r.rtm, &info); 1710 break; 1711 1712 default: 1713 trace_act("ignore %s", str); 1714 break; 1715 } 1716 } 1717 } 1718 1719 1720 /* 1721 * Disassemble a routing message. The result is an array of pointers 1722 * to sockaddr_storage structures stored in the info argument. 1723 * 1724 * ss is a pointer to the beginning of the data following the 1725 * rt_msghdr contained in the routing socket message, which consists 1726 * of a string of concatenated sockaddr structure of different types. 1727 * 1728 * Extended attributes can be appended at the end of the list. 1729 */ 1730 static int 1731 rt_xaddrs(struct rt_addrinfo *info, 1732 struct sockaddr_storage *ss, 1733 char *lim, 1734 int addrs) 1735 { 1736 int retv = 0; 1737 int i; 1738 int abit; 1739 int complaints; 1740 static int prev_complaints; 1741 1742 #define XBAD_AF 0x1 1743 #define XBAD_SHORT 0x2 1744 #define XBAD_LONG 0x4 1745 1746 (void) memset(info, 0, sizeof (*info)); 1747 info->rti_addrs = addrs; 1748 complaints = 0; 1749 for (i = 0, abit = 1; i < RTAX_MAX && (char *)ss < lim; 1750 i++, abit <<= 1) { 1751 if ((addrs & abit) == 0) 1752 continue; 1753 info->rti_info[i] = ss; 1754 /* Horrible interface here */ 1755 switch (ss->ss_family) { 1756 case AF_UNIX: 1757 /* LINTED */ 1758 ss = (struct sockaddr_storage *)( 1759 (struct sockaddr_un *)ss + 1); 1760 break; 1761 case AF_INET: 1762 /* LINTED */ 1763 ss = (struct sockaddr_storage *)( 1764 (struct sockaddr_in *)ss + 1); 1765 break; 1766 case AF_LINK: 1767 /* LINTED */ 1768 ss = (struct sockaddr_storage *)( 1769 (struct sockaddr_dl *)ss + 1); 1770 break; 1771 case AF_INET6: 1772 /* LINTED */ 1773 ss = (struct sockaddr_storage *)( 1774 (struct sockaddr_in6 *)ss + 1); 1775 break; 1776 default: 1777 if (!(prev_complaints & XBAD_AF)) 1778 writelog(LOG_WARNING, 1779 "unknown address family %d " 1780 "encountered", ss->ss_family); 1781 if (complaints & XBAD_AF) 1782 goto xaddr_done; 1783 /* LINTED */ 1784 ss = (struct sockaddr_storage *)( 1785 (struct sockaddr *)ss + 1); 1786 complaints |= XBAD_AF; 1787 info->rti_addrs &= abit - 1; 1788 addrs = info->rti_addrs; 1789 retv = -1; 1790 break; 1791 } 1792 if ((char *)ss > lim) { 1793 if (!(prev_complaints & XBAD_SHORT)) 1794 msglog("sockaddr %d too short by %d " 1795 "bytes", i + 1, (char *)ss - lim); 1796 complaints |= XBAD_SHORT; 1797 info->rti_info[i] = NULL; 1798 info->rti_addrs &= abit - 1; 1799 retv = -1; 1800 goto xaddr_done; 1801 } 1802 } 1803 1804 while (((char *)ss + sizeof (rtm_ext_t)) <= lim) { 1805 rtm_ext_t *tp; 1806 char *nxt; 1807 1808 /* LINTED: alignment */ 1809 tp = (rtm_ext_t *)ss; 1810 nxt = (char *)(tp + 1) + tp->rtmex_len; 1811 1812 if (!IS_P2ALIGNED(tp->rtmex_len, sizeof (uint32_t)) || 1813 nxt > lim) { 1814 break; 1815 } 1816 1817 /* LINTED: alignment */ 1818 ss = (struct sockaddr_storage *)nxt; 1819 } 1820 1821 if ((char *)ss != lim) { 1822 if ((char *)ss > lim) { 1823 if (!(prev_complaints & XBAD_SHORT)) 1824 msglog("routing message too short by %d bytes", 1825 (char *)ss - lim); 1826 complaints |= XBAD_SHORT; 1827 } else if (!(prev_complaints & XBAD_LONG)) { 1828 msglog("%d bytes of routing message left over", 1829 lim - (char *)ss); 1830 complaints |= XBAD_LONG; 1831 } 1832 retv = -1; 1833 } 1834 xaddr_done: 1835 prev_complaints = complaints; 1836 return (retv); 1837 } 1838 1839 1840 /* after aggregating, note routes that belong in the kernel */ 1841 static void 1842 kern_out(struct ag_info *ag) 1843 { 1844 struct khash *k; 1845 struct interface *ifp; 1846 1847 ifp = ag->ag_ifp; 1848 1849 if (ifp != NULL && ifp->int_phys != NULL) { 1850 ifp = ifwithname(ifp->int_phys->phyi_name); 1851 } 1852 1853 /* 1854 * Do not install bad routes if they are not already present. 1855 * This includes routes that had RS_NET_SYN for interfaces that 1856 * recently died. 1857 */ 1858 if (ag->ag_metric == HOPCNT_INFINITY) { 1859 k = kern_find(htonl(ag->ag_dst_h), ag->ag_mask, 1860 ag->ag_nhop, ag->ag_ifp, NULL); 1861 if (k == NULL) 1862 return; 1863 } else { 1864 k = kern_add(htonl(ag->ag_dst_h), ag->ag_mask, ag->ag_nhop, 1865 ifp); 1866 } 1867 1868 if (k->k_state & KS_NEW) { 1869 /* will need to add new entry to the kernel table */ 1870 k->k_state = KS_ADD; 1871 if (ag->ag_state & AGS_GATEWAY) 1872 k->k_state |= KS_GATEWAY; 1873 if (ag->ag_state & AGS_IF) 1874 k->k_state |= KS_IF; 1875 if (ag->ag_state & AGS_PASSIVE) 1876 k->k_state |= KS_PASSIVE; 1877 if (ag->ag_state & AGS_FILE) 1878 k->k_state |= KS_FILE; 1879 k->k_gate = ag->ag_nhop; 1880 k->k_ifp = ifp; 1881 k->k_metric = ag->ag_metric; 1882 return; 1883 } 1884 1885 if ((k->k_state & (KS_STATIC|KS_DEPRE_IF)) || 1886 ((k->k_state & (KS_IF|KS_PASSIVE)) == KS_IF)) { 1887 return; 1888 } 1889 1890 /* modify existing kernel entry if necessary */ 1891 if (k->k_gate == ag->ag_nhop && k->k_ifp == ag->ag_ifp && 1892 k->k_metric != ag->ag_metric) { 1893 /* 1894 * Must delete bad interface routes etc. 1895 * to change them. 1896 */ 1897 if (k->k_metric == HOPCNT_INFINITY) 1898 k->k_state |= KS_DEL_ADD; 1899 k->k_gate = ag->ag_nhop; 1900 k->k_metric = ag->ag_metric; 1901 k->k_state |= KS_CHANGE; 1902 } 1903 1904 /* 1905 * If the daemon thinks the route should exist, forget 1906 * about any redirections. 1907 * If the daemon thinks the route should exist, eventually 1908 * override manual intervention by the operator. 1909 */ 1910 if ((k->k_state & (KS_DYNAMIC | KS_DELETED)) != 0) { 1911 k->k_state &= ~KS_DYNAMIC; 1912 k->k_state |= (KS_ADD | KS_DEL_ADD); 1913 } 1914 1915 if ((k->k_state & KS_GATEWAY) && !(ag->ag_state & AGS_GATEWAY)) { 1916 k->k_state &= ~KS_GATEWAY; 1917 k->k_state |= (KS_ADD | KS_DEL_ADD); 1918 } else if (!(k->k_state & KS_GATEWAY) && (ag->ag_state & AGS_GATEWAY)) { 1919 k->k_state |= KS_GATEWAY; 1920 k->k_state |= (KS_ADD | KS_DEL_ADD); 1921 } 1922 1923 /* 1924 * Deleting-and-adding is necessary to change aspects of a route. 1925 * Just delete instead of deleting and then adding a bad route. 1926 * Otherwise, we want to keep the route in the kernel. 1927 */ 1928 if (k->k_metric == HOPCNT_INFINITY && (k->k_state & KS_DEL_ADD)) 1929 k->k_state |= KS_DELETE; 1930 else 1931 k->k_state &= ~KS_DELETE; 1932 #undef RT 1933 } 1934 1935 /* 1936 * Update our image of the kernel forwarding table using the given 1937 * route from our internal routing table. 1938 */ 1939 1940 /*ARGSUSED1*/ 1941 static int 1942 walk_kern(struct radix_node *rn, void *argp) 1943 { 1944 #define RT ((struct rt_entry *)rn) 1945 uint8_t metric, pref; 1946 uint_t ags = 0; 1947 int i; 1948 struct rt_spare *rts; 1949 1950 /* Do not install synthetic routes */ 1951 if (RT->rt_state & RS_NET_SYN) 1952 return (0); 1953 1954 /* 1955 * Do not install static routes here. Only 1956 * read_rt->rtm_add->kern_add should install those 1957 */ 1958 if ((RT->rt_state & RS_STATIC) && 1959 (RT->rt_spares[0].rts_origin != RO_FILE)) 1960 return (0); 1961 1962 /* Do not clobber kernel if this is a route for a dead interface */ 1963 if (RT->rt_state & RS_BADIF) 1964 return (0); 1965 1966 if (!(RT->rt_state & RS_IF)) { 1967 /* This is an ordinary route, not for an interface. */ 1968 1969 /* 1970 * aggregate, ordinary good routes without regard to 1971 * their metric 1972 */ 1973 pref = 1; 1974 ags |= (AGS_GATEWAY | AGS_SUPPRESS | AGS_AGGREGATE); 1975 1976 /* 1977 * Do not install host routes directly to hosts, to avoid 1978 * interfering with ARP entries in the kernel table. 1979 */ 1980 if (RT_ISHOST(RT) && ntohl(RT->rt_dst) == RT->rt_gate) 1981 return (0); 1982 1983 } else { 1984 /* 1985 * This is an interface route. 1986 * Do not install routes for "external" remote interfaces. 1987 */ 1988 if (RT->rt_ifp != NULL && (RT->rt_ifp->int_state & IS_EXTERNAL)) 1989 return (0); 1990 1991 /* Interfaces should override received routes. */ 1992 pref = 0; 1993 ags |= (AGS_IF | AGS_CORS_GATE); 1994 if (RT->rt_ifp != NULL && 1995 !(RT->rt_ifp->int_if_flags & IFF_LOOPBACK) && 1996 (RT->rt_ifp->int_state & (IS_PASSIVE|IS_ALIAS)) == 1997 IS_PASSIVE) { 1998 ags |= AGS_PASSIVE; 1999 } 2000 2001 /* 2002 * If it is not an interface, or an alias for an interface, 2003 * it must be a "gateway." 2004 * 2005 * If it is a "remote" interface, it is also a "gateway" to 2006 * the kernel if is not a alias. 2007 */ 2008 if (RT->rt_ifp == NULL || (RT->rt_ifp->int_state & IS_REMOTE)) { 2009 2010 ags |= (AGS_GATEWAY | AGS_SUPPRESS); 2011 2012 /* 2013 * Do not aggregate IS_PASSIVE routes. 2014 */ 2015 if (!(RT->rt_ifp->int_state & IS_PASSIVE)) 2016 ags |= AGS_AGGREGATE; 2017 } 2018 } 2019 2020 metric = RT->rt_metric; 2021 if (metric == HOPCNT_INFINITY) { 2022 /* If the route is dead, try hard to aggregate. */ 2023 pref = HOPCNT_INFINITY; 2024 ags |= (AGS_FINE_GATE | AGS_SUPPRESS); 2025 ags &= ~(AGS_IF | AGS_CORS_GATE); 2026 } 2027 2028 /* 2029 * dump all routes that have the same metric as rt_spares[0] 2030 * into the kern_table, to be added to the kernel. 2031 */ 2032 for (i = 0; i < RT->rt_num_spares; i++) { 2033 rts = &RT->rt_spares[i]; 2034 2035 /* Do not install external routes */ 2036 if (rts->rts_flags & RTS_EXTERNAL) 2037 continue; 2038 2039 if (rts->rts_metric == metric) { 2040 ag_check(RT->rt_dst, RT->rt_mask, 2041 rts->rts_router, rts->rts_ifp, rts->rts_gate, 2042 metric, pref, 0, 0, 2043 (rts->rts_origin & RO_FILE) ? (ags|AGS_FILE) : ags, 2044 kern_out); 2045 } 2046 } 2047 return (0); 2048 #undef RT 2049 } 2050 2051 2052 /* Update the kernel table to match the daemon table. */ 2053 static void 2054 fix_kern(void) 2055 { 2056 int i; 2057 struct khash *k, *pk, *knext; 2058 2059 2060 need_kern = age_timer; 2061 2062 /* Walk daemon table, updating the copy of the kernel table. */ 2063 (void) rn_walktree(rhead, walk_kern, NULL); 2064 ag_flush(0, 0, kern_out); 2065 2066 for (i = 0; i < KHASH_SIZE; i++) { 2067 pk = NULL; 2068 for (k = khash_bins[i]; k != NULL; k = knext) { 2069 knext = k->k_next; 2070 2071 /* Do not touch local interface routes */ 2072 if ((k->k_state & KS_DEPRE_IF) || 2073 (k->k_state & (KS_IF|KS_PASSIVE)) == KS_IF) { 2074 pk = k; 2075 continue; 2076 } 2077 2078 /* Do not touch static routes */ 2079 if (k->k_state & KS_STATIC) { 2080 kern_check_static(k, 0); 2081 pk = k; 2082 continue; 2083 } 2084 2085 /* check hold on routes deleted by the operator */ 2086 if (k->k_keep > now.tv_sec) { 2087 /* ensure we check when the hold is over */ 2088 LIM_SEC(need_kern, k->k_keep); 2089 pk = k; 2090 continue; 2091 } 2092 2093 if ((k->k_state & KS_DELETE) && 2094 !(k->k_state & KS_DYNAMIC)) { 2095 if ((k->k_dst == RIP_DEFAULT) && 2096 (k->k_ifp != NULL) && 2097 (kern_alternate(RIP_DEFAULT, 2098 k->k_mask, k->k_gate, k->k_ifp, 2099 NULL) == NULL)) 2100 rdisc_restore(k->k_ifp); 2101 kern_ioctl(k, RTM_DELETE, 0); 2102 if (pk != NULL) 2103 pk->k_next = knext; 2104 else 2105 khash_bins[i] = knext; 2106 free(k); 2107 continue; 2108 } 2109 2110 if (k->k_state & KS_DEL_ADD) 2111 kern_ioctl(k, RTM_DELETE, 0); 2112 2113 if (k->k_state & KS_ADD) { 2114 if ((k->k_dst == RIP_DEFAULT) && 2115 (k->k_ifp != NULL)) 2116 rdisc_suppress(k->k_ifp); 2117 kern_ioctl(k, RTM_ADD, 2118 ((0 != (k->k_state & (KS_GATEWAY | 2119 KS_DYNAMIC))) ? RTF_GATEWAY : 0)); 2120 } else if (k->k_state & KS_CHANGE) { 2121 /* 2122 * Should be using RTM_CHANGE here, but 2123 * since RTM_CHANGE is currently 2124 * not multipath-aware, and assumes 2125 * that RTF_GATEWAY implies the gateway 2126 * of the route for dst has to be 2127 * changed, we play safe, and do a del + add. 2128 */ 2129 kern_ioctl(k, RTM_DELETE, 0); 2130 kern_ioctl(k, RTM_ADD, 2131 ((0 != (k->k_state & (KS_GATEWAY | 2132 KS_DYNAMIC))) ? RTF_GATEWAY : 0)); 2133 } 2134 k->k_state &= ~(KS_ADD|KS_CHANGE|KS_DEL_ADD); 2135 2136 /* 2137 * Mark this route to be deleted in the next cycle. 2138 * This deletes routes that disappear from the 2139 * daemon table, since the normal aging code 2140 * will clear the bit for routes that have not 2141 * disappeared from the daemon table. 2142 */ 2143 k->k_state |= KS_DELETE; 2144 pk = k; 2145 } 2146 } 2147 } 2148 2149 2150 /* Delete a static route in the image of the kernel table. */ 2151 void 2152 del_static(in_addr_t dst, in_addr_t mask, in_addr_t gate, 2153 struct interface *ifp, int gone) 2154 { 2155 struct khash *k; 2156 struct rt_entry *rt; 2157 2158 /* 2159 * Just mark it in the table to be deleted next time the kernel 2160 * table is updated. 2161 * If it has already been deleted, mark it as such, and set its 2162 * keep-timer so that it will not be deleted again for a while. 2163 * This lets the operator delete a route added by the daemon 2164 * and add a replacement. 2165 */ 2166 k = kern_find(dst, mask, gate, ifp, NULL); 2167 if (k != NULL && (gate == 0 || k->k_gate == gate)) { 2168 k->k_state &= ~(KS_STATIC | KS_DYNAMIC | KS_CHECK); 2169 k->k_state |= KS_DELETE; 2170 if (gone) { 2171 k->k_state |= KS_DELETED; 2172 k->k_keep = now.tv_sec + K_KEEP_LIM; 2173 } 2174 } 2175 2176 rt = rtget(dst, mask); 2177 if (rt != NULL && (rt->rt_state & RS_STATIC)) 2178 rtbad(rt, NULL); 2179 } 2180 2181 2182 /* 2183 * Delete all routes generated from ICMP Redirects that use a given gateway, 2184 * as well as old redirected routes. 2185 */ 2186 void 2187 del_redirects(in_addr_t bad_gate, time_t old) 2188 { 2189 int i; 2190 struct khash *k; 2191 boolean_t dosupply = should_supply(NULL); 2192 2193 for (i = 0; i < KHASH_SIZE; i++) { 2194 for (k = khash_bins[i]; k != NULL; k = k->k_next) { 2195 if (!(k->k_state & KS_DYNAMIC) || 2196 (k->k_state & (KS_STATIC|KS_IF|KS_DEPRE_IF))) 2197 continue; 2198 2199 if (k->k_gate != bad_gate && k->k_redirect_time > old && 2200 !dosupply) 2201 continue; 2202 2203 k->k_state |= KS_DELETE; 2204 k->k_state &= ~KS_DYNAMIC; 2205 need_kern.tv_sec = now.tv_sec; 2206 trace_act("mark redirected %s --> %s for deletion", 2207 addrname(k->k_dst, k->k_mask, 0), 2208 naddr_ntoa(k->k_gate)); 2209 } 2210 } 2211 } 2212 2213 /* Start the daemon tables. */ 2214 void 2215 rtinit(void) 2216 { 2217 int i; 2218 struct ag_info *ag; 2219 2220 /* Initialize the radix trees */ 2221 rn_init(); 2222 (void) rn_inithead((void**)&rhead, 32); 2223 2224 /* mark all of the slots in the table free */ 2225 ag_avail = ag_slots; 2226 for (ag = ag_slots, i = 1; i < NUM_AG_SLOTS; i++) { 2227 ag->ag_fine = ag+1; 2228 ag++; 2229 } 2230 } 2231 2232 2233 static struct sockaddr_in dst_sock = {AF_INET}; 2234 static struct sockaddr_in mask_sock = {AF_INET}; 2235 2236 2237 static void 2238 set_need_flash(void) 2239 { 2240 if (!need_flash) { 2241 need_flash = _B_TRUE; 2242 /* 2243 * Do not send the flash update immediately. Wait a little 2244 * while to hear from other routers. 2245 */ 2246 no_flash.tv_sec = now.tv_sec + MIN_WAITTIME; 2247 } 2248 } 2249 2250 2251 /* Get a particular routing table entry */ 2252 struct rt_entry * 2253 rtget(in_addr_t dst, in_addr_t mask) 2254 { 2255 struct rt_entry *rt; 2256 2257 dst_sock.sin_addr.s_addr = dst; 2258 mask_sock.sin_addr.s_addr = htonl(mask); 2259 rt = (struct rt_entry *)rhead->rnh_lookup(&dst_sock, &mask_sock, rhead); 2260 if (rt == NULL || rt->rt_dst != dst || rt->rt_mask != mask) 2261 return (NULL); 2262 2263 return (rt); 2264 } 2265 2266 2267 /* Find a route to dst as the kernel would. */ 2268 struct rt_entry * 2269 rtfind(in_addr_t dst) 2270 { 2271 dst_sock.sin_addr.s_addr = dst; 2272 return ((struct rt_entry *)rhead->rnh_matchaddr(&dst_sock, rhead)); 2273 } 2274 2275 2276 /* add a route to the table */ 2277 void 2278 rtadd(in_addr_t dst, 2279 in_addr_t mask, 2280 uint16_t state, /* rt_state for the entry */ 2281 struct rt_spare *new) 2282 { 2283 struct rt_entry *rt; 2284 in_addr_t smask; 2285 int i; 2286 struct rt_spare *rts; 2287 2288 /* This is the only function that increments total_routes. */ 2289 if (total_routes == MAX_ROUTES) { 2290 msglog("have maximum (%d) routes", total_routes); 2291 return; 2292 } 2293 2294 rt = rtmalloc(sizeof (*rt), "rtadd"); 2295 (void) memset(rt, 0, sizeof (*rt)); 2296 rt->rt_spares = rtmalloc(SPARE_INC * sizeof (struct rt_spare), 2297 "rtadd"); 2298 rt->rt_num_spares = SPARE_INC; 2299 (void) memset(rt->rt_spares, 0, SPARE_INC * sizeof (struct rt_spare)); 2300 for (rts = rt->rt_spares, i = rt->rt_num_spares; i != 0; i--, rts++) 2301 rts->rts_metric = HOPCNT_INFINITY; 2302 2303 rt->rt_nodes->rn_key = (uint8_t *)&rt->rt_dst_sock; 2304 rt->rt_dst = dst; 2305 rt->rt_dst_sock.sin_family = AF_INET; 2306 if (mask != HOST_MASK) { 2307 smask = std_mask(dst); 2308 if ((smask & ~mask) == 0 && mask > smask) 2309 state |= RS_SUBNET; 2310 } 2311 mask_sock.sin_addr.s_addr = htonl(mask); 2312 rt->rt_mask = mask; 2313 rt->rt_spares[0] = *new; 2314 rt->rt_state = state; 2315 rt->rt_time = now.tv_sec; 2316 rt->rt_poison_metric = HOPCNT_INFINITY; 2317 rt->rt_seqno = update_seqno; 2318 2319 if (TRACEACTIONS) 2320 trace_add_del("Add", rt); 2321 2322 need_kern.tv_sec = now.tv_sec; 2323 set_need_flash(); 2324 2325 if (NULL == rhead->rnh_addaddr(&rt->rt_dst_sock, &mask_sock, rhead, 2326 rt->rt_nodes)) { 2327 msglog("rnh_addaddr() failed for %s mask=%s", 2328 naddr_ntoa(dst), naddr_ntoa(htonl(mask))); 2329 free(rt); 2330 } 2331 2332 total_routes++; 2333 } 2334 2335 2336 /* notice a changed route */ 2337 void 2338 rtchange(struct rt_entry *rt, 2339 uint16_t state, /* new state bits */ 2340 struct rt_spare *new, 2341 char *label) 2342 { 2343 if (rt->rt_metric != new->rts_metric) { 2344 /* 2345 * Fix the kernel immediately if it seems the route 2346 * has gone bad, since there may be a working route that 2347 * aggregates this route. 2348 */ 2349 if (new->rts_metric == HOPCNT_INFINITY) { 2350 need_kern.tv_sec = now.tv_sec; 2351 if (new->rts_time >= now.tv_sec - EXPIRE_TIME) 2352 new->rts_time = now.tv_sec - EXPIRE_TIME; 2353 } 2354 rt->rt_seqno = update_seqno; 2355 set_need_flash(); 2356 } 2357 2358 if (rt->rt_gate != new->rts_gate) { 2359 need_kern.tv_sec = now.tv_sec; 2360 rt->rt_seqno = update_seqno; 2361 set_need_flash(); 2362 } 2363 2364 state |= (rt->rt_state & RS_SUBNET); 2365 2366 /* Keep various things from deciding ageless routes are stale. */ 2367 if (!AGE_RT(state, rt->rt_spares[0].rts_origin, new->rts_ifp)) 2368 new->rts_time = now.tv_sec; 2369 2370 if (TRACEACTIONS) 2371 trace_change(rt, state, new, 2372 label ? label : "Chg "); 2373 2374 rt->rt_state = state; 2375 /* 2376 * If the interface state of the new primary route is good, 2377 * turn off RS_BADIF flag 2378 */ 2379 if ((rt->rt_state & RS_BADIF) && 2380 IS_IFF_UP(new->rts_ifp->int_if_flags) && 2381 !(new->rts_ifp->int_state & (IS_BROKE | IS_SICK))) 2382 rt->rt_state &= ~(RS_BADIF); 2383 2384 rt->rt_spares[0] = *new; 2385 } 2386 2387 2388 /* check for a better route among the spares */ 2389 static struct rt_spare * 2390 rts_better(struct rt_entry *rt) 2391 { 2392 struct rt_spare *rts, *rts1; 2393 int i; 2394 2395 /* find the best alternative among the spares */ 2396 rts = rt->rt_spares+1; 2397 for (i = rt->rt_num_spares, rts1 = rts+1; i > 2; i--, rts1++) { 2398 if (BETTER_LINK(rt, rts1, rts)) 2399 rts = rts1; 2400 } 2401 2402 return (rts); 2403 } 2404 2405 2406 /* switch to a backup route */ 2407 void 2408 rtswitch(struct rt_entry *rt, 2409 struct rt_spare *rts) 2410 { 2411 struct rt_spare swap; 2412 char label[10]; 2413 2414 /* Do not change permanent routes */ 2415 if (0 != (rt->rt_state & (RS_MHOME | RS_STATIC | 2416 RS_NET_SYN | RS_IF))) 2417 return; 2418 2419 /* find the best alternative among the spares */ 2420 if (rts == NULL) 2421 rts = rts_better(rt); 2422 2423 /* Do not bother if it is not worthwhile. */ 2424 if (!BETTER_LINK(rt, rts, rt->rt_spares)) 2425 return; 2426 2427 swap = rt->rt_spares[0]; 2428 (void) snprintf(label, sizeof (label), "Use #%d", 2429 (int)(rts - rt->rt_spares)); 2430 rtchange(rt, rt->rt_state & ~(RS_NET_SYN), rts, label); 2431 2432 if (swap.rts_metric == HOPCNT_INFINITY) { 2433 *rts = rts_empty; 2434 } else { 2435 *rts = swap; 2436 } 2437 2438 } 2439 2440 2441 void 2442 rtdelete(struct rt_entry *rt) 2443 { 2444 struct rt_entry *deleted_rt; 2445 struct rt_spare *rts; 2446 int i; 2447 in_addr_t gate = rt->rt_gate; /* for debugging */ 2448 2449 if (TRACEACTIONS) 2450 trace_add_del("Del", rt); 2451 2452 for (i = 0; i < rt->rt_num_spares; i++) { 2453 rts = &rt->rt_spares[i]; 2454 rts_delete(rt, rts); 2455 } 2456 2457 dst_sock.sin_addr.s_addr = rt->rt_dst; 2458 mask_sock.sin_addr.s_addr = htonl(rt->rt_mask); 2459 if (rt != (deleted_rt = 2460 ((struct rt_entry *)rhead->rnh_deladdr(&dst_sock, &mask_sock, 2461 rhead)))) { 2462 msglog("rnh_deladdr(%s) failed; found rt 0x%lx", 2463 rtname(rt->rt_dst, rt->rt_mask, gate), deleted_rt); 2464 if (deleted_rt != NULL) 2465 free(deleted_rt); 2466 } 2467 total_routes--; 2468 free(rt->rt_spares); 2469 free(rt); 2470 2471 if (dst_sock.sin_addr.s_addr == RIP_DEFAULT) { 2472 /* 2473 * we just deleted the default route. Trigger rdisc_sort 2474 * so that we can recover from any rdisc information that 2475 * is valid 2476 */ 2477 rdisc_timer.tv_sec = 0; 2478 } 2479 } 2480 2481 void 2482 rts_delete(struct rt_entry *rt, struct rt_spare *rts) 2483 { 2484 struct khash *k; 2485 2486 trace_upslot(rt, rts, &rts_empty); 2487 k = kern_find(rt->rt_dst, rt->rt_mask, 2488 rts->rts_gate, rts->rts_ifp, NULL); 2489 if (k != NULL && 2490 !(k->k_state & KS_DEPRE_IF) && 2491 ((k->k_state & (KS_IF|KS_PASSIVE)) != KS_IF)) { 2492 k->k_state |= KS_DELETE; 2493 need_kern.tv_sec = now.tv_sec; 2494 } 2495 2496 *rts = rts_empty; 2497 } 2498 2499 /* 2500 * Get rid of a bad route, and try to switch to a replacement. 2501 * If the route has gone bad because of a bad interface, 2502 * the information about the dead interface is available in badifp 2503 * for the purpose of sanity checks, if_flags checks etc. 2504 */ 2505 static void 2506 rtbad(struct rt_entry *rt, struct interface *badifp) 2507 { 2508 struct rt_spare new; 2509 uint16_t rt_state; 2510 2511 2512 if (badifp == NULL || (rt->rt_spares[0].rts_ifp == badifp)) { 2513 /* Poison the route */ 2514 new = rt->rt_spares[0]; 2515 new.rts_metric = HOPCNT_INFINITY; 2516 rt_state = rt->rt_state & ~(RS_IF | RS_LOCAL | RS_STATIC); 2517 } 2518 2519 if (badifp != NULL) { 2520 /* 2521 * Dont mark the rtentry bad unless the ifp for the primary 2522 * route is the bad ifp 2523 */ 2524 if (rt->rt_spares[0].rts_ifp != badifp) 2525 return; 2526 /* 2527 * badifp has just gone bad. We want to keep this 2528 * rt_entry around so that we tell our rip-neighbors 2529 * about the bad route, but we can't do anything 2530 * to the kernel itself, so mark it as RS_BADIF 2531 */ 2532 trace_misc("rtbad:Setting RS_BADIF (%s)", badifp->int_name); 2533 rt_state |= RS_BADIF; 2534 new.rts_ifp = &dummy_ifp; 2535 } 2536 rtchange(rt, rt_state, &new, 0); 2537 rtswitch(rt, 0); 2538 } 2539 2540 2541 /* 2542 * Junk a RS_NET_SYN or RS_LOCAL route, 2543 * unless it is needed by another interface. 2544 */ 2545 void 2546 rtbad_sub(struct rt_entry *rt, struct interface *badifp) 2547 { 2548 struct interface *ifp, *ifp1; 2549 struct intnet *intnetp; 2550 uint_t state; 2551 2552 2553 ifp1 = NULL; 2554 state = 0; 2555 2556 if (rt->rt_state & RS_LOCAL) { 2557 /* 2558 * Is this the route through loopback for the interface? 2559 * If so, see if it is used by any other interfaces, such 2560 * as a point-to-point interface with the same local address. 2561 */ 2562 for (ifp = ifnet; ifp != NULL; ifp = ifp->int_next) { 2563 /* Retain it if another interface needs it. */ 2564 if (ifp->int_addr == rt->rt_ifp->int_addr) { 2565 state |= RS_LOCAL; 2566 ifp1 = ifp; 2567 break; 2568 } 2569 } 2570 2571 } 2572 2573 if (!(state & RS_LOCAL)) { 2574 /* 2575 * Retain RIPv1 logical network route if there is another 2576 * interface that justifies it. 2577 */ 2578 if (rt->rt_state & RS_NET_SYN) { 2579 for (ifp = ifnet; ifp != NULL; ifp = ifp->int_next) { 2580 if ((ifp->int_state & IS_NEED_NET_SYN) && 2581 rt->rt_mask == ifp->int_std_mask && 2582 rt->rt_dst == ifp->int_std_addr) { 2583 state |= RS_NET_SYN; 2584 ifp1 = ifp; 2585 break; 2586 } 2587 } 2588 } 2589 2590 /* or if there is an authority route that needs it. */ 2591 for (intnetp = intnets; intnetp != NULL; 2592 intnetp = intnetp->intnet_next) { 2593 if (intnetp->intnet_addr == rt->rt_dst && 2594 intnetp->intnet_mask == rt->rt_mask) { 2595 state |= (RS_NET_SYN | RS_NET_INT); 2596 break; 2597 } 2598 } 2599 } 2600 2601 if (ifp1 != NULL || (state & RS_NET_SYN)) { 2602 struct rt_spare new = rt->rt_spares[0]; 2603 new.rts_ifp = ifp1; 2604 rtchange(rt, ((rt->rt_state & ~(RS_NET_SYN|RS_LOCAL)) | state), 2605 &new, 0); 2606 } else { 2607 rtbad(rt, badifp); 2608 } 2609 } 2610 2611 /* 2612 * Called while walking the table looking for sick interfaces 2613 * or after a time change. 2614 */ 2615 int 2616 walk_bad(struct radix_node *rn, 2617 void *argp) 2618 { 2619 #define RT ((struct rt_entry *)rn) 2620 struct rt_spare *rts; 2621 int i, j = -1; 2622 2623 /* fix any spare routes through the interface */ 2624 for (i = 1; i < RT->rt_num_spares; i++) { 2625 rts = &((struct rt_entry *)rn)->rt_spares[i]; 2626 2627 if (rts->rts_metric < HOPCNT_INFINITY && 2628 (rts->rts_ifp == NULL || 2629 (rts->rts_ifp->int_state & IS_BROKE))) 2630 rts_delete(RT, rts); 2631 else { 2632 if (rts->rts_origin != RO_NONE) 2633 j = i; 2634 } 2635 } 2636 2637 /* 2638 * Deal with the main route 2639 * finished if it has been handled before or if its interface is ok 2640 */ 2641 if (RT->rt_ifp == NULL || !(RT->rt_ifp->int_state & IS_BROKE)) 2642 return (0); 2643 2644 /* Bad routes for other than interfaces are easy. */ 2645 if (!(RT->rt_state & (RS_IF | RS_NET_SYN | RS_LOCAL))) { 2646 if (j > 0) { 2647 RT->rt_spares[0].rts_metric = HOPCNT_INFINITY; 2648 rtswitch(RT, NULL); 2649 } else { 2650 rtbad(RT, (struct interface *)argp); 2651 } 2652 return (0); 2653 } 2654 2655 rtbad_sub(RT, (struct interface *)argp); 2656 return (0); 2657 #undef RT 2658 } 2659 2660 /* 2661 * Called while walking the table to replace a duplicate interface 2662 * with a backup. 2663 */ 2664 int 2665 walk_rewire(struct radix_node *rn, void *argp) 2666 { 2667 struct rt_entry *RT = (struct rt_entry *)rn; 2668 struct rewire_data *wire = (struct rewire_data *)argp; 2669 struct rt_spare *rts; 2670 int i; 2671 2672 /* fix any spare routes through the interface */ 2673 rts = RT->rt_spares; 2674 for (i = RT->rt_num_spares; i > 0; i--, rts++) { 2675 if (rts->rts_ifp == wire->if_old) { 2676 rts->rts_ifp = wire->if_new; 2677 if ((RT->rt_dst == RIP_DEFAULT) && 2678 (wire->if_old->int_state & IS_SUPPRESS_RDISC)) 2679 rdisc_suppress(rts->rts_ifp); 2680 if ((rts->rts_metric += wire->metric_delta) > 2681 HOPCNT_INFINITY) 2682 rts->rts_metric = HOPCNT_INFINITY; 2683 2684 /* 2685 * If the main route is getting a worse metric, 2686 * then it may be time to switch to a backup. 2687 */ 2688 if (i == RT->rt_num_spares && wire->metric_delta > 0) { 2689 rtswitch(RT, NULL); 2690 } 2691 } 2692 } 2693 2694 return (0); 2695 } 2696 2697 /* Check the age of an individual route. */ 2698 static int 2699 walk_age(struct radix_node *rn, void *argp) 2700 { 2701 #define RT ((struct rt_entry *)rn) 2702 struct interface *ifp; 2703 struct rt_spare *rts; 2704 int i; 2705 in_addr_t age_bad_gate = *(in_addr_t *)argp; 2706 2707 2708 /* 2709 * age all of the spare routes, including the primary route 2710 * currently in use 2711 */ 2712 rts = RT->rt_spares; 2713 for (i = RT->rt_num_spares; i != 0; i--, rts++) { 2714 2715 ifp = rts->rts_ifp; 2716 if (i == RT->rt_num_spares) { 2717 if (!AGE_RT(RT->rt_state, rts->rts_origin, ifp)) { 2718 /* 2719 * Keep various things from deciding ageless 2720 * routes are stale 2721 */ 2722 rts->rts_time = now.tv_sec; 2723 continue; 2724 } 2725 2726 /* forget RIP routes after RIP has been turned off. */ 2727 if (rip_sock < 0) { 2728 rts->rts_time = now_stale + 1; 2729 } 2730 } 2731 2732 /* age failing routes */ 2733 if (age_bad_gate == rts->rts_gate && 2734 rts->rts_time >= now_stale) { 2735 rts->rts_time -= SUPPLY_INTERVAL; 2736 } 2737 2738 /* trash the spare routes when they go bad */ 2739 if (rts->rts_origin == RO_RIP && 2740 ((rip_sock < 0) || 2741 (rts->rts_metric < HOPCNT_INFINITY && 2742 now_garbage > rts->rts_time)) && 2743 i != RT->rt_num_spares) { 2744 rts_delete(RT, rts); 2745 } 2746 } 2747 2748 2749 /* finished if the active route is still fresh */ 2750 if (now_stale <= RT->rt_time) 2751 return (0); 2752 2753 /* try to switch to an alternative */ 2754 rtswitch(RT, NULL); 2755 2756 /* Delete a dead route after it has been publically mourned. */ 2757 if (now_garbage > RT->rt_time) { 2758 rtdelete(RT); 2759 return (0); 2760 } 2761 2762 /* Start poisoning a bad route before deleting it. */ 2763 if (now.tv_sec - RT->rt_time > EXPIRE_TIME) { 2764 struct rt_spare new = RT->rt_spares[0]; 2765 2766 new.rts_metric = HOPCNT_INFINITY; 2767 rtchange(RT, RT->rt_state, &new, 0); 2768 } 2769 return (0); 2770 } 2771 2772 2773 /* Watch for dead routes and interfaces. */ 2774 void 2775 age(in_addr_t bad_gate) 2776 { 2777 struct interface *ifp; 2778 int need_query = 0; 2779 2780 /* 2781 * If not listening to RIP, there is no need to age the routes in 2782 * the table. 2783 */ 2784 age_timer.tv_sec = (now.tv_sec 2785 + ((rip_sock < 0) ? NEVER : SUPPLY_INTERVAL)); 2786 2787 /* 2788 * Check for dead IS_REMOTE interfaces by timing their 2789 * transmissions. 2790 */ 2791 for (ifp = ifnet; ifp; ifp = ifp->int_next) { 2792 if (!(ifp->int_state & IS_REMOTE)) 2793 continue; 2794 2795 /* ignore unreachable remote interfaces */ 2796 if (!check_remote(ifp)) 2797 continue; 2798 2799 /* Restore remote interface that has become reachable */ 2800 if (ifp->int_state & IS_BROKE) 2801 if_ok(ifp, "remote ", _B_FALSE); 2802 2803 if (ifp->int_act_time != NEVER && 2804 now.tv_sec - ifp->int_act_time > EXPIRE_TIME) { 2805 writelog(LOG_NOTICE, 2806 "remote interface %s to %s timed out after" 2807 " %ld:%ld", 2808 ifp->int_name, 2809 naddr_ntoa(ifp->int_dstaddr), 2810 (now.tv_sec - ifp->int_act_time)/60, 2811 (now.tv_sec - ifp->int_act_time)%60); 2812 if_sick(ifp, _B_FALSE); 2813 } 2814 2815 /* 2816 * If we have not heard from the other router 2817 * recently, ask it. 2818 */ 2819 if (now.tv_sec >= ifp->int_query_time) { 2820 ifp->int_query_time = NEVER; 2821 need_query = 1; 2822 } 2823 } 2824 2825 /* Age routes. */ 2826 (void) rn_walktree(rhead, walk_age, &bad_gate); 2827 2828 /* 2829 * delete old redirected routes to keep the kernel table small 2830 * and prevent blackholes 2831 */ 2832 del_redirects(bad_gate, now.tv_sec-STALE_TIME); 2833 2834 /* Update the kernel routing table. */ 2835 fix_kern(); 2836 2837 /* poke reticent remote gateways */ 2838 if (need_query) 2839 rip_query(); 2840 } 2841 2842 void 2843 kern_dump(void) 2844 { 2845 int i; 2846 struct khash *k; 2847 2848 for (i = 0; i < KHASH_SIZE; i++) { 2849 for (k = khash_bins[i]; k != NULL; k = k->k_next) 2850 trace_khash(k); 2851 } 2852 } 2853 2854 2855 static struct interface * 2856 gwkludge_iflookup(in_addr_t dstaddr, in_addr_t addr, in_addr_t mask) 2857 { 2858 uint32_t int_state; 2859 struct interface *ifp; 2860 2861 for (ifp = ifnet; ifp != NULL; ifp = ifp->int_next) { 2862 int_state = ifp->int_state; 2863 2864 if (!(int_state & IS_REMOTE)) 2865 continue; 2866 2867 if (ifp->int_dstaddr == dstaddr && ifp->int_addr == addr && 2868 ifp->int_mask == mask) 2869 return (ifp); 2870 } 2871 return (NULL); 2872 } 2873