1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD 3 * 4 * Copyright (c) 2004 Luigi Rizzo, Alessandro Cerri. All rights reserved. 5 * Copyright (c) 2004-2008 Qing Li. All rights reserved. 6 * Copyright (c) 2008 Kip Macy. 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 * 17 * THIS SOFTWARE IS PROVIDED BY AUTHOR AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL AUTHOR OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 */ 29 #include <sys/cdefs.h> 30 __FBSDID("$FreeBSD$"); 31 32 #include "opt_ddb.h" 33 #include "opt_inet.h" 34 #include "opt_inet6.h" 35 36 #include <sys/param.h> 37 #include <sys/systm.h> 38 #include <sys/eventhandler.h> 39 #include <sys/malloc.h> 40 #include <sys/mbuf.h> 41 #include <sys/syslog.h> 42 #include <sys/sysctl.h> 43 #include <sys/socket.h> 44 #include <sys/kernel.h> 45 #include <sys/lock.h> 46 #include <sys/mutex.h> 47 #include <sys/rwlock.h> 48 49 #ifdef DDB 50 #include <ddb/ddb.h> 51 #endif 52 53 #include <vm/uma.h> 54 55 #include <netinet/in.h> 56 #include <net/if_llatbl.h> 57 #include <net/if.h> 58 #include <net/if_dl.h> 59 #include <net/if_var.h> 60 #include <net/route.h> 61 #include <net/route/route_ctl.h> 62 #include <net/route/route_debug.h> 63 #include <net/vnet.h> 64 #include <netinet/if_ether.h> 65 #include <netinet6/in6_var.h> 66 #include <netinet6/nd6.h> 67 68 MALLOC_DEFINE(M_LLTABLE, "lltable", "link level address tables"); 69 70 VNET_DEFINE_STATIC(SLIST_HEAD(, lltable), lltables) = 71 SLIST_HEAD_INITIALIZER(lltables); 72 #define V_lltables VNET(lltables) 73 74 static struct rwlock lltable_list_lock; 75 RW_SYSINIT(lltable_list_lock, &lltable_list_lock, "lltable_list_lock"); 76 #define LLTABLE_LIST_RLOCK() rw_rlock(&lltable_list_lock) 77 #define LLTABLE_LIST_RUNLOCK() rw_runlock(&lltable_list_lock) 78 #define LLTABLE_LIST_WLOCK() rw_wlock(&lltable_list_lock) 79 #define LLTABLE_LIST_WUNLOCK() rw_wunlock(&lltable_list_lock) 80 #define LLTABLE_LIST_LOCK_ASSERT() rw_assert(&lltable_list_lock, RA_LOCKED) 81 82 static void lltable_unlink(struct lltable *llt); 83 static void llentries_unlink(struct lltable *llt, struct llentries *head); 84 85 /* 86 * Dump lle state for a specific address family. 87 */ 88 static int 89 lltable_dump_af(struct lltable *llt, struct sysctl_req *wr) 90 { 91 struct epoch_tracker et; 92 int error; 93 94 LLTABLE_LIST_LOCK_ASSERT(); 95 96 if (llt->llt_ifp->if_flags & IFF_LOOPBACK) 97 return (0); 98 error = 0; 99 100 NET_EPOCH_ENTER(et); 101 error = lltable_foreach_lle(llt, 102 (llt_foreach_cb_t *)llt->llt_dump_entry, wr); 103 NET_EPOCH_EXIT(et); 104 105 return (error); 106 } 107 108 /* 109 * Dump arp state for a specific address family. 110 */ 111 int 112 lltable_sysctl_dumparp(int af, struct sysctl_req *wr) 113 { 114 struct lltable *llt; 115 int error = 0; 116 117 LLTABLE_LIST_RLOCK(); 118 SLIST_FOREACH(llt, &V_lltables, llt_link) { 119 if (llt->llt_af == af) { 120 error = lltable_dump_af(llt, wr); 121 if (error != 0) 122 goto done; 123 } 124 } 125 done: 126 LLTABLE_LIST_RUNLOCK(); 127 return (error); 128 } 129 130 /* 131 * Common function helpers for chained hash table. 132 */ 133 134 /* 135 * Runs specified callback for each entry in @llt. 136 * Caller does the locking. 137 * 138 */ 139 static int 140 htable_foreach_lle(struct lltable *llt, llt_foreach_cb_t *f, void *farg) 141 { 142 struct llentry *lle, *next; 143 int i, error; 144 145 error = 0; 146 147 for (i = 0; i < llt->llt_hsize; i++) { 148 CK_LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) { 149 error = f(llt, lle, farg); 150 if (error != 0) 151 break; 152 } 153 } 154 155 return (error); 156 } 157 158 /* 159 * The htable_[un]link_entry() functions return: 160 * 0 if the entry was (un)linked already and nothing changed, 161 * 1 if the entry was added/removed to/from the table, and 162 * -1 on error (e.g., not being able to add the entry due to limits reached). 163 * While the "unlink" operation should never error, callers of 164 * lltable_link_entry() need to check for errors and handle them. 165 */ 166 static int 167 htable_link_entry(struct lltable *llt, struct llentry *lle) 168 { 169 struct llentries *lleh; 170 uint32_t hashidx; 171 172 if ((lle->la_flags & LLE_LINKED) != 0) 173 return (0); 174 175 IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp); 176 177 if (llt->llt_maxentries > 0 && 178 llt->llt_entries >= llt->llt_maxentries) 179 return (-1); 180 181 hashidx = llt->llt_hash(lle, llt->llt_hsize); 182 lleh = &llt->lle_head[hashidx]; 183 184 lle->lle_tbl = llt; 185 lle->lle_head = lleh; 186 lle->la_flags |= LLE_LINKED; 187 CK_LIST_INSERT_HEAD(lleh, lle, lle_next); 188 llt->llt_entries++; 189 190 return (1); 191 } 192 193 static int 194 htable_unlink_entry(struct llentry *lle) 195 { 196 struct lltable *llt; 197 198 if ((lle->la_flags & LLE_LINKED) == 0) 199 return (0); 200 201 llt = lle->lle_tbl; 202 IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp); 203 KASSERT(llt->llt_entries > 0, ("%s: lltable %p (%s) entries %d <= 0", 204 __func__, llt, if_name(llt->llt_ifp), llt->llt_entries)); 205 206 CK_LIST_REMOVE(lle, lle_next); 207 lle->la_flags &= ~(LLE_VALID | LLE_LINKED); 208 #if 0 209 lle->lle_tbl = NULL; 210 lle->lle_head = NULL; 211 #endif 212 llt->llt_entries--; 213 214 return (1); 215 } 216 217 struct prefix_match_data { 218 const struct sockaddr *addr; 219 const struct sockaddr *mask; 220 struct llentries dchain; 221 u_int flags; 222 }; 223 224 static int 225 htable_prefix_free_cb(struct lltable *llt, struct llentry *lle, void *farg) 226 { 227 struct prefix_match_data *pmd; 228 229 pmd = (struct prefix_match_data *)farg; 230 231 if (llt->llt_match_prefix(pmd->addr, pmd->mask, pmd->flags, lle)) { 232 LLE_WLOCK(lle); 233 CK_LIST_INSERT_HEAD(&pmd->dchain, lle, lle_chain); 234 } 235 236 return (0); 237 } 238 239 static void 240 htable_prefix_free(struct lltable *llt, const struct sockaddr *addr, 241 const struct sockaddr *mask, u_int flags) 242 { 243 struct llentry *lle, *next; 244 struct prefix_match_data pmd; 245 246 bzero(&pmd, sizeof(pmd)); 247 pmd.addr = addr; 248 pmd.mask = mask; 249 pmd.flags = flags; 250 CK_LIST_INIT(&pmd.dchain); 251 252 IF_AFDATA_WLOCK(llt->llt_ifp); 253 /* Push matching lles to chain */ 254 lltable_foreach_lle(llt, htable_prefix_free_cb, &pmd); 255 256 llentries_unlink(llt, &pmd.dchain); 257 IF_AFDATA_WUNLOCK(llt->llt_ifp); 258 259 CK_LIST_FOREACH_SAFE(lle, &pmd.dchain, lle_chain, next) 260 lltable_free_entry(llt, lle); 261 } 262 263 static void 264 htable_free_tbl(struct lltable *llt) 265 { 266 267 free(llt->lle_head, M_LLTABLE); 268 free(llt, M_LLTABLE); 269 } 270 271 static void 272 llentries_unlink(struct lltable *llt, struct llentries *head) 273 { 274 struct llentry *lle, *next; 275 276 CK_LIST_FOREACH_SAFE(lle, head, lle_chain, next) 277 llt->llt_unlink_entry(lle); 278 } 279 280 /* 281 * Helper function used to drop all mbufs in hold queue. 282 * 283 * Returns the number of held packets, if any, that were dropped. 284 */ 285 size_t 286 lltable_drop_entry_queue(struct llentry *lle) 287 { 288 size_t pkts_dropped; 289 struct mbuf *next; 290 291 LLE_WLOCK_ASSERT(lle); 292 293 pkts_dropped = 0; 294 while ((lle->la_numheld > 0) && (lle->la_hold != NULL)) { 295 next = lle->la_hold->m_nextpkt; 296 m_freem(lle->la_hold); 297 lle->la_hold = next; 298 lle->la_numheld--; 299 pkts_dropped++; 300 } 301 302 KASSERT(lle->la_numheld == 0, 303 ("%s: la_numheld %d > 0, pkts_droped %zd", __func__, 304 lle->la_numheld, pkts_dropped)); 305 306 return (pkts_dropped); 307 } 308 309 void 310 lltable_set_entry_addr(struct ifnet *ifp, struct llentry *lle, 311 const char *linkhdr, size_t linkhdrsize, int lladdr_off) 312 { 313 314 memcpy(lle->r_linkdata, linkhdr, linkhdrsize); 315 lle->r_hdrlen = linkhdrsize; 316 lle->ll_addr = &lle->r_linkdata[lladdr_off]; 317 lle->la_flags |= LLE_VALID; 318 lle->r_flags |= RLLE_VALID; 319 } 320 321 /* 322 * Acquires lltable write lock. 323 * 324 * Returns true on success, with both lltable and lle lock held. 325 * On failure, false is returned and lle wlock is still held. 326 */ 327 bool 328 lltable_acquire_wlock(struct ifnet *ifp, struct llentry *lle) 329 { 330 NET_EPOCH_ASSERT(); 331 332 /* Perform real LLE update */ 333 /* use afdata WLOCK to update fields */ 334 LLE_WUNLOCK(lle); 335 IF_AFDATA_WLOCK(ifp); 336 LLE_WLOCK(lle); 337 338 /* 339 * Since we droppped LLE lock, other thread might have deleted 340 * this lle. Check and return 341 */ 342 if ((lle->la_flags & LLE_DELETED) != 0) { 343 IF_AFDATA_WUNLOCK(ifp); 344 return (false); 345 } 346 347 return (true); 348 } 349 350 /* 351 * Tries to update @lle link-level address. 352 * Since update requires AFDATA WLOCK, function 353 * drops @lle lock, acquires AFDATA lock and then acquires 354 * @lle lock to maintain lock order. 355 * 356 * Returns 1 on success. 357 */ 358 int 359 lltable_try_set_entry_addr(struct ifnet *ifp, struct llentry *lle, 360 const char *linkhdr, size_t linkhdrsize, int lladdr_off) 361 { 362 363 if (!lltable_acquire_wlock(ifp, lle)) 364 return (0); 365 366 /* Update data */ 367 lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize, lladdr_off); 368 369 IF_AFDATA_WUNLOCK(ifp); 370 371 return (1); 372 } 373 374 /* 375 * Helper function used to pre-compute full/partial link-layer 376 * header data suitable for feeding into if_output(). 377 */ 378 int 379 lltable_calc_llheader(struct ifnet *ifp, int family, char *lladdr, 380 char *buf, size_t *bufsize, int *lladdr_off) 381 { 382 struct if_encap_req ereq; 383 int error; 384 385 bzero(buf, *bufsize); 386 bzero(&ereq, sizeof(ereq)); 387 ereq.buf = buf; 388 ereq.bufsize = *bufsize; 389 ereq.rtype = IFENCAP_LL; 390 ereq.family = family; 391 ereq.lladdr = lladdr; 392 ereq.lladdr_len = ifp->if_addrlen; 393 error = ifp->if_requestencap(ifp, &ereq); 394 if (error == 0) { 395 *bufsize = ereq.bufsize; 396 *lladdr_off = ereq.lladdr_off; 397 } 398 399 return (error); 400 } 401 402 /* 403 * Searches for the child entry matching @family inside @lle. 404 * Returns the entry or NULL. 405 */ 406 struct llentry * 407 llentry_lookup_family(struct llentry *lle, int family) 408 { 409 struct llentry *child_lle; 410 411 if (lle == NULL) 412 return (NULL); 413 414 CK_SLIST_FOREACH(child_lle, &lle->lle_children, lle_child_next) { 415 if (child_lle->r_family == family) 416 return (child_lle); 417 } 418 419 return (NULL); 420 } 421 422 /* 423 * Retrieves upper protocol family for the llentry. 424 * By default, all "normal" (e.g. upper_family == transport_family) 425 * llentries have r_family set to 0. 426 * Thus, use @default_family in that regard, otherwise use r_family. 427 * 428 * Returns upper protocol family 429 */ 430 int 431 llentry_get_upper_family(const struct llentry *lle, int default_family) 432 { 433 return (lle->r_family == 0 ? default_family : lle->r_family); 434 } 435 436 /* 437 * Prints llentry @lle data into provided buffer. 438 * Example: lle/inet/valid/em0/1.2.3.4 439 * 440 * Returns @buf. 441 */ 442 char * 443 llentry_print_buf(const struct llentry *lle, struct ifnet *ifp, int family, 444 char *buf, size_t bufsize) 445 { 446 char abuf[INET6_ADDRSTRLEN]; 447 448 const char *valid = (lle->r_flags & RLLE_VALID) ? "valid" : "no_l2"; 449 const char *upper_str = rib_print_family(llentry_get_upper_family(lle, family)); 450 451 switch (family) { 452 #ifdef INET 453 case AF_INET: 454 inet_ntop(AF_INET, &lle->r_l3addr.addr4, abuf, sizeof(abuf)); 455 snprintf(buf, bufsize, "lle/%s/%s/%s/%s", upper_str, 456 valid, if_name(ifp), abuf); 457 break; 458 #endif 459 #ifdef INET6 460 case AF_INET6: 461 inet_ntop(AF_INET6, &lle->r_l3addr.addr6, abuf, sizeof(abuf)); 462 snprintf(buf, bufsize, "lle/%s/%s/%s/%s", upper_str, 463 valid, if_name(ifp), abuf); 464 break; 465 #endif 466 default: 467 snprintf(buf, bufsize, "lle/%s/%s/%s/????", upper_str, 468 valid, if_name(ifp)); 469 break; 470 } 471 472 return (buf); 473 } 474 475 char * 476 llentry_print_buf_lltable(const struct llentry *lle, char *buf, size_t bufsize) 477 { 478 struct lltable *tbl = lle->lle_tbl; 479 480 return (llentry_print_buf(lle, lltable_get_ifp(tbl), lltable_get_af(tbl), buf, bufsize)); 481 } 482 483 /* 484 * Requests feedback from the datapath. 485 * First packet using @lle should result in 486 * setting r_skip_req back to 0 and updating 487 * lle_hittime to the current time_uptime. 488 */ 489 void 490 llentry_request_feedback(struct llentry *lle) 491 { 492 struct llentry *child_lle; 493 494 LLE_REQ_LOCK(lle); 495 lle->r_skip_req = 1; 496 LLE_REQ_UNLOCK(lle); 497 498 CK_SLIST_FOREACH(child_lle, &lle->lle_children, lle_child_next) { 499 LLE_REQ_LOCK(child_lle); 500 child_lle->r_skip_req = 1; 501 LLE_REQ_UNLOCK(child_lle); 502 } 503 } 504 505 /* 506 * Updates the lle state to mark it has been used 507 * and record the time. 508 * Used by the llentry_provide_feedback() wrapper. 509 */ 510 void 511 llentry_mark_used(struct llentry *lle) 512 { 513 LLE_REQ_LOCK(lle); 514 lle->r_skip_req = 0; 515 lle->lle_hittime = time_uptime; 516 LLE_REQ_UNLOCK(lle); 517 } 518 519 /* 520 * Fetches the time when lle was used. 521 * Return 0 if the entry was not used, relevant time_uptime 522 * otherwise. 523 */ 524 static time_t 525 llentry_get_hittime_raw(struct llentry *lle) 526 { 527 time_t lle_hittime = 0; 528 529 LLE_REQ_LOCK(lle); 530 if ((lle->r_skip_req == 0) && (lle_hittime < lle->lle_hittime)) 531 lle_hittime = lle->lle_hittime; 532 LLE_REQ_UNLOCK(lle); 533 534 return (lle_hittime); 535 } 536 537 time_t 538 llentry_get_hittime(struct llentry *lle) 539 { 540 time_t lle_hittime = 0; 541 struct llentry *child_lle; 542 543 lle_hittime = llentry_get_hittime_raw(lle); 544 545 CK_SLIST_FOREACH(child_lle, &lle->lle_children, lle_child_next) { 546 time_t hittime = llentry_get_hittime_raw(child_lle); 547 if (hittime > lle_hittime) 548 lle_hittime = hittime; 549 } 550 551 return (lle_hittime); 552 } 553 554 /* 555 * Update link-layer header for given @lle after 556 * interface lladdr was changed. 557 */ 558 static int 559 llentry_update_ifaddr(struct lltable *llt, struct llentry *lle, void *farg) 560 { 561 struct ifnet *ifp; 562 u_char linkhdr[LLE_MAX_LINKHDR]; 563 size_t linkhdrsize; 564 u_char *lladdr; 565 int lladdr_off; 566 567 ifp = (struct ifnet *)farg; 568 569 lladdr = lle->ll_addr; 570 571 LLE_WLOCK(lle); 572 if ((lle->la_flags & LLE_VALID) == 0) { 573 LLE_WUNLOCK(lle); 574 return (0); 575 } 576 577 if ((lle->la_flags & LLE_IFADDR) != 0) 578 lladdr = IF_LLADDR(ifp); 579 580 linkhdrsize = sizeof(linkhdr); 581 lltable_calc_llheader(ifp, llt->llt_af, lladdr, linkhdr, &linkhdrsize, 582 &lladdr_off); 583 memcpy(lle->r_linkdata, linkhdr, linkhdrsize); 584 LLE_WUNLOCK(lle); 585 586 return (0); 587 } 588 589 /* 590 * Update all calculated headers for given @llt 591 */ 592 void 593 lltable_update_ifaddr(struct lltable *llt) 594 { 595 596 if (llt->llt_ifp->if_flags & IFF_LOOPBACK) 597 return; 598 599 IF_AFDATA_WLOCK(llt->llt_ifp); 600 lltable_foreach_lle(llt, llentry_update_ifaddr, llt->llt_ifp); 601 IF_AFDATA_WUNLOCK(llt->llt_ifp); 602 } 603 604 /* 605 * 606 * Performs generic cleanup routines and frees lle. 607 * 608 * Called for non-linked entries, with callouts and 609 * other AF-specific cleanups performed. 610 * 611 * @lle must be passed WLOCK'ed 612 * 613 * Returns the number of held packets, if any, that were dropped. 614 */ 615 size_t 616 llentry_free(struct llentry *lle) 617 { 618 size_t pkts_dropped; 619 620 LLE_WLOCK_ASSERT(lle); 621 622 KASSERT((lle->la_flags & LLE_LINKED) == 0, ("freeing linked lle")); 623 624 pkts_dropped = lltable_drop_entry_queue(lle); 625 626 /* cancel timer */ 627 if (callout_stop(&lle->lle_timer) > 0) 628 LLE_REMREF(lle); 629 LLE_FREE_LOCKED(lle); 630 631 return (pkts_dropped); 632 } 633 634 /* 635 * Free all entries from given table and free itself. 636 */ 637 638 static int 639 lltable_free_cb(struct lltable *llt, struct llentry *lle, void *farg) 640 { 641 struct llentries *dchain; 642 643 dchain = (struct llentries *)farg; 644 645 LLE_WLOCK(lle); 646 CK_LIST_INSERT_HEAD(dchain, lle, lle_chain); 647 648 return (0); 649 } 650 651 /* 652 * Free all entries from given table and free itself. 653 */ 654 void 655 lltable_free(struct lltable *llt) 656 { 657 struct llentry *lle, *next; 658 struct llentries dchain; 659 660 KASSERT(llt != NULL, ("%s: llt is NULL", __func__)); 661 662 lltable_unlink(llt); 663 664 CK_LIST_INIT(&dchain); 665 IF_AFDATA_WLOCK(llt->llt_ifp); 666 /* Push all lles to @dchain */ 667 lltable_foreach_lle(llt, lltable_free_cb, &dchain); 668 llentries_unlink(llt, &dchain); 669 IF_AFDATA_WUNLOCK(llt->llt_ifp); 670 671 CK_LIST_FOREACH_SAFE(lle, &dchain, lle_chain, next) { 672 llentry_free(lle); 673 } 674 675 KASSERT(llt->llt_entries == 0, ("%s: lltable %p (%s) entires not 0: %d", 676 __func__, llt, llt->llt_ifp->if_xname, llt->llt_entries)); 677 678 llt->llt_free_tbl(llt); 679 } 680 681 /* 682 * Deletes an address from given lltable. 683 * Used for userland interaction to remove 684 * individual entries. Skips entries added by OS. 685 */ 686 int 687 lltable_delete_addr(struct lltable *llt, u_int flags, 688 const struct sockaddr *l3addr) 689 { 690 struct llentry *lle; 691 struct ifnet *ifp; 692 693 ifp = llt->llt_ifp; 694 IF_AFDATA_WLOCK(ifp); 695 lle = lla_lookup(llt, LLE_SF(l3addr->sa_family, LLE_EXCLUSIVE), l3addr); 696 697 if (lle == NULL) { 698 IF_AFDATA_WUNLOCK(ifp); 699 return (ENOENT); 700 } 701 if ((lle->la_flags & LLE_IFADDR) != 0 && (flags & LLE_IFADDR) == 0) { 702 IF_AFDATA_WUNLOCK(ifp); 703 LLE_WUNLOCK(lle); 704 return (EPERM); 705 } 706 707 lltable_unlink_entry(llt, lle); 708 IF_AFDATA_WUNLOCK(ifp); 709 710 llt->llt_delete_entry(llt, lle); 711 712 return (0); 713 } 714 715 void 716 lltable_prefix_free(int af, struct sockaddr *addr, struct sockaddr *mask, 717 u_int flags) 718 { 719 struct lltable *llt; 720 721 LLTABLE_LIST_RLOCK(); 722 SLIST_FOREACH(llt, &V_lltables, llt_link) { 723 if (llt->llt_af != af) 724 continue; 725 726 llt->llt_prefix_free(llt, addr, mask, flags); 727 } 728 LLTABLE_LIST_RUNLOCK(); 729 } 730 731 struct lltable * 732 lltable_allocate_htbl(uint32_t hsize) 733 { 734 struct lltable *llt; 735 int i; 736 737 llt = malloc(sizeof(struct lltable), M_LLTABLE, M_WAITOK | M_ZERO); 738 llt->llt_hsize = hsize; 739 llt->lle_head = malloc(sizeof(struct llentries) * hsize, 740 M_LLTABLE, M_WAITOK | M_ZERO); 741 742 for (i = 0; i < llt->llt_hsize; i++) 743 CK_LIST_INIT(&llt->lle_head[i]); 744 745 /* Set some default callbacks */ 746 llt->llt_link_entry = htable_link_entry; 747 llt->llt_unlink_entry = htable_unlink_entry; 748 llt->llt_prefix_free = htable_prefix_free; 749 llt->llt_foreach_entry = htable_foreach_lle; 750 llt->llt_free_tbl = htable_free_tbl; 751 752 return (llt); 753 } 754 755 /* 756 * Links lltable to global llt list. 757 */ 758 void 759 lltable_link(struct lltable *llt) 760 { 761 762 LLTABLE_LIST_WLOCK(); 763 SLIST_INSERT_HEAD(&V_lltables, llt, llt_link); 764 LLTABLE_LIST_WUNLOCK(); 765 } 766 767 static void 768 lltable_unlink(struct lltable *llt) 769 { 770 771 LLTABLE_LIST_WLOCK(); 772 SLIST_REMOVE(&V_lltables, llt, lltable, llt_link); 773 LLTABLE_LIST_WUNLOCK(); 774 775 } 776 777 /* 778 * Gets interface @ifp lltable for the specified @family 779 */ 780 struct lltable * 781 lltable_get(struct ifnet *ifp, int family) 782 { 783 switch (family) { 784 #ifdef INET 785 case AF_INET: 786 return (in_lltable_get(ifp)); 787 #endif 788 #ifdef INET6 789 case AF_INET6: 790 return (in6_lltable_get(ifp)); 791 #endif 792 } 793 794 return (NULL); 795 } 796 797 /* 798 * External methods used by lltable consumers 799 */ 800 801 int 802 lltable_foreach_lle(struct lltable *llt, llt_foreach_cb_t *f, void *farg) 803 { 804 805 return (llt->llt_foreach_entry(llt, f, farg)); 806 } 807 808 struct llentry * 809 lltable_alloc_entry(struct lltable *llt, u_int flags, 810 const struct sockaddr *l3addr) 811 { 812 813 return (llt->llt_alloc_entry(llt, flags, l3addr)); 814 } 815 816 void 817 lltable_free_entry(struct lltable *llt, struct llentry *lle) 818 { 819 820 llt->llt_free_entry(llt, lle); 821 } 822 823 int 824 lltable_link_entry(struct lltable *llt, struct llentry *lle) 825 { 826 827 return (llt->llt_link_entry(llt, lle)); 828 } 829 830 void 831 lltable_link_child_entry(struct llentry *lle, struct llentry *child_lle) 832 { 833 child_lle->lle_parent = lle; 834 child_lle->lle_tbl = lle->lle_tbl; 835 child_lle->la_flags |= LLE_LINKED; 836 CK_SLIST_INSERT_HEAD(&lle->lle_children, child_lle, lle_child_next); 837 } 838 839 void 840 lltable_unlink_child_entry(struct llentry *child_lle) 841 { 842 struct llentry *lle = child_lle->lle_parent; 843 844 child_lle->la_flags &= ~LLE_LINKED; 845 child_lle->lle_parent = NULL; 846 CK_SLIST_REMOVE(&lle->lle_children, child_lle, llentry, lle_child_next); 847 } 848 849 int 850 lltable_unlink_entry(struct lltable *llt, struct llentry *lle) 851 { 852 853 return (llt->llt_unlink_entry(lle)); 854 } 855 856 void 857 lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa) 858 { 859 struct lltable *llt; 860 861 llt = lle->lle_tbl; 862 llt->llt_fill_sa_entry(lle, sa); 863 } 864 865 struct ifnet * 866 lltable_get_ifp(const struct lltable *llt) 867 { 868 869 return (llt->llt_ifp); 870 } 871 872 int 873 lltable_get_af(const struct lltable *llt) 874 { 875 876 return (llt->llt_af); 877 } 878 879 /* 880 * Called in route_output when rtm_flags contains RTF_LLDATA. 881 */ 882 int 883 lla_rt_output(struct rt_msghdr *rtm, struct rt_addrinfo *info) 884 { 885 struct sockaddr_dl *dl = 886 (struct sockaddr_dl *)info->rti_info[RTAX_GATEWAY]; 887 struct sockaddr *dst = (struct sockaddr *)info->rti_info[RTAX_DST]; 888 struct ifnet *ifp; 889 struct lltable *llt; 890 struct llentry *lle, *lle_tmp; 891 uint8_t linkhdr[LLE_MAX_LINKHDR]; 892 size_t linkhdrsize; 893 int lladdr_off; 894 u_int laflags = 0; 895 int error; 896 897 if (dl == NULL || dl->sdl_family != AF_LINK) 898 return (EINVAL); 899 900 /* XXX: should be ntohs() */ 901 ifp = ifnet_byindex(dl->sdl_index); 902 if (ifp == NULL) { 903 log(LOG_INFO, "%s: invalid ifp (sdl_index %d)\n", 904 __func__, dl->sdl_index); 905 return EINVAL; 906 } 907 908 llt = lltable_get(ifp, dst->sa_family); 909 910 if (llt == NULL) 911 return (ESRCH); 912 913 error = 0; 914 915 switch (rtm->rtm_type) { 916 case RTM_ADD: 917 /* Add static LLE */ 918 laflags = 0; 919 if (rtm->rtm_rmx.rmx_expire == 0) 920 laflags = LLE_STATIC; 921 lle = lltable_alloc_entry(llt, laflags, dst); 922 if (lle == NULL) 923 return (ENOMEM); 924 925 linkhdrsize = sizeof(linkhdr); 926 if (lltable_calc_llheader(ifp, dst->sa_family, LLADDR(dl), 927 linkhdr, &linkhdrsize, &lladdr_off) != 0) 928 return (EINVAL); 929 lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize, 930 lladdr_off); 931 if ((rtm->rtm_flags & RTF_ANNOUNCE)) 932 lle->la_flags |= LLE_PUB; 933 lle->la_expire = rtm->rtm_rmx.rmx_expire; 934 935 laflags = lle->la_flags; 936 937 /* Try to link new entry */ 938 lle_tmp = NULL; 939 IF_AFDATA_WLOCK(ifp); 940 LLE_WLOCK(lle); 941 lle_tmp = lla_lookup(llt, LLE_EXCLUSIVE, dst); 942 if (lle_tmp != NULL) { 943 /* Check if we are trying to replace immutable entry */ 944 if ((lle_tmp->la_flags & LLE_IFADDR) != 0) { 945 IF_AFDATA_WUNLOCK(ifp); 946 LLE_WUNLOCK(lle_tmp); 947 lltable_free_entry(llt, lle); 948 return (EPERM); 949 } 950 /* Unlink existing entry from table */ 951 lltable_unlink_entry(llt, lle_tmp); 952 } 953 lltable_link_entry(llt, lle); 954 IF_AFDATA_WUNLOCK(ifp); 955 956 if (lle_tmp != NULL) { 957 EVENTHANDLER_INVOKE(lle_event, lle_tmp,LLENTRY_EXPIRED); 958 lltable_free_entry(llt, lle_tmp); 959 } 960 961 /* 962 * By invoking LLE handler here we might get 963 * two events on static LLE entry insertion 964 * in routing socket. However, since we might have 965 * other subscribers we need to generate this event. 966 */ 967 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_RESOLVED); 968 LLE_WUNLOCK(lle); 969 #ifdef INET 970 /* gratuitous ARP */ 971 if ((laflags & LLE_PUB) && dst->sa_family == AF_INET) 972 arprequest(ifp, 973 &((struct sockaddr_in *)dst)->sin_addr, 974 &((struct sockaddr_in *)dst)->sin_addr, 975 (u_char *)LLADDR(dl)); 976 #endif 977 978 break; 979 980 case RTM_DELETE: 981 return (lltable_delete_addr(llt, 0, dst)); 982 983 default: 984 error = EINVAL; 985 } 986 987 return (error); 988 } 989 990 #ifdef DDB 991 struct llentry_sa { 992 struct llentry base; 993 struct sockaddr l3_addr; 994 }; 995 996 static void 997 llatbl_lle_show(struct llentry_sa *la) 998 { 999 struct llentry *lle; 1000 uint8_t octet[6]; 1001 1002 lle = &la->base; 1003 db_printf("lle=%p\n", lle); 1004 db_printf(" lle_next=%p\n", lle->lle_next.cle_next); 1005 db_printf(" lle_lock=%p\n", &lle->lle_lock); 1006 db_printf(" lle_tbl=%p\n", lle->lle_tbl); 1007 db_printf(" lle_head=%p\n", lle->lle_head); 1008 db_printf(" la_hold=%p\n", lle->la_hold); 1009 db_printf(" la_numheld=%d\n", lle->la_numheld); 1010 db_printf(" la_expire=%ju\n", (uintmax_t)lle->la_expire); 1011 db_printf(" la_flags=0x%04x\n", lle->la_flags); 1012 db_printf(" la_asked=%u\n", lle->la_asked); 1013 db_printf(" la_preempt=%u\n", lle->la_preempt); 1014 db_printf(" ln_state=%d\n", lle->ln_state); 1015 db_printf(" ln_router=%u\n", lle->ln_router); 1016 db_printf(" ln_ntick=%ju\n", (uintmax_t)lle->ln_ntick); 1017 db_printf(" lle_refcnt=%d\n", lle->lle_refcnt); 1018 bcopy(lle->ll_addr, octet, sizeof(octet)); 1019 db_printf(" ll_addr=%02x:%02x:%02x:%02x:%02x:%02x\n", 1020 octet[0], octet[1], octet[2], octet[3], octet[4], octet[5]); 1021 db_printf(" lle_timer=%p\n", &lle->lle_timer); 1022 1023 switch (la->l3_addr.sa_family) { 1024 #ifdef INET 1025 case AF_INET: 1026 { 1027 struct sockaddr_in *sin; 1028 char l3s[INET_ADDRSTRLEN]; 1029 1030 sin = (struct sockaddr_in *)&la->l3_addr; 1031 inet_ntoa_r(sin->sin_addr, l3s); 1032 db_printf(" l3_addr=%s\n", l3s); 1033 break; 1034 } 1035 #endif 1036 #ifdef INET6 1037 case AF_INET6: 1038 { 1039 struct sockaddr_in6 *sin6; 1040 char l3s[INET6_ADDRSTRLEN]; 1041 1042 sin6 = (struct sockaddr_in6 *)&la->l3_addr; 1043 ip6_sprintf(l3s, &sin6->sin6_addr); 1044 db_printf(" l3_addr=%s\n", l3s); 1045 break; 1046 } 1047 #endif 1048 default: 1049 db_printf(" l3_addr=N/A (af=%d)\n", la->l3_addr.sa_family); 1050 break; 1051 } 1052 } 1053 1054 DB_SHOW_COMMAND(llentry, db_show_llentry) 1055 { 1056 1057 if (!have_addr) { 1058 db_printf("usage: show llentry <struct llentry *>\n"); 1059 return; 1060 } 1061 1062 llatbl_lle_show((struct llentry_sa *)addr); 1063 } 1064 1065 static void 1066 llatbl_llt_show(struct lltable *llt) 1067 { 1068 int i; 1069 struct llentry *lle; 1070 1071 db_printf("llt=%p llt_af=%d llt_ifp=%p\n", 1072 llt, llt->llt_af, llt->llt_ifp); 1073 1074 for (i = 0; i < llt->llt_hsize; i++) { 1075 CK_LIST_FOREACH(lle, &llt->lle_head[i], lle_next) { 1076 llatbl_lle_show((struct llentry_sa *)lle); 1077 if (db_pager_quit) 1078 return; 1079 } 1080 } 1081 } 1082 1083 DB_SHOW_COMMAND(lltable, db_show_lltable) 1084 { 1085 1086 if (!have_addr) { 1087 db_printf("usage: show lltable <struct lltable *>\n"); 1088 return; 1089 } 1090 1091 llatbl_llt_show((struct lltable *)addr); 1092 } 1093 1094 DB_SHOW_ALL_COMMAND(lltables, db_show_all_lltables) 1095 { 1096 VNET_ITERATOR_DECL(vnet_iter); 1097 struct lltable *llt; 1098 1099 VNET_FOREACH(vnet_iter) { 1100 CURVNET_SET_QUIET(vnet_iter); 1101 #ifdef VIMAGE 1102 db_printf("vnet=%p\n", curvnet); 1103 #endif 1104 SLIST_FOREACH(llt, &V_lltables, llt_link) { 1105 db_printf("llt=%p llt_af=%d llt_ifp=%p(%s)\n", 1106 llt, llt->llt_af, llt->llt_ifp, 1107 (llt->llt_ifp != NULL) ? 1108 llt->llt_ifp->if_xname : "?"); 1109 if (have_addr && addr != 0) /* verbose */ 1110 llatbl_llt_show(llt); 1111 if (db_pager_quit) { 1112 CURVNET_RESTORE(); 1113 return; 1114 } 1115 } 1116 CURVNET_RESTORE(); 1117 } 1118 } 1119 #endif 1120