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/if_private.h> 61 #include <net/route.h> 62 #include <net/route/route_ctl.h> 63 #include <net/route/route_debug.h> 64 #include <net/vnet.h> 65 #include <netinet/if_ether.h> 66 #include <netinet6/in6_var.h> 67 #include <netinet6/nd6.h> 68 69 MALLOC_DEFINE(M_LLTABLE, "lltable", "link level address tables"); 70 71 VNET_DEFINE_STATIC(SLIST_HEAD(, lltable), lltables) = 72 SLIST_HEAD_INITIALIZER(lltables); 73 #define V_lltables VNET(lltables) 74 75 static struct rwlock lltable_list_lock; 76 RW_SYSINIT(lltable_list_lock, &lltable_list_lock, "lltable_list_lock"); 77 #define LLTABLE_LIST_RLOCK() rw_rlock(&lltable_list_lock) 78 #define LLTABLE_LIST_RUNLOCK() rw_runlock(&lltable_list_lock) 79 #define LLTABLE_LIST_WLOCK() rw_wlock(&lltable_list_lock) 80 #define LLTABLE_LIST_WUNLOCK() rw_wunlock(&lltable_list_lock) 81 #define LLTABLE_LIST_LOCK_ASSERT() rw_assert(&lltable_list_lock, RA_LOCKED) 82 83 static void lltable_unlink(struct lltable *llt); 84 static void llentries_unlink(struct lltable *llt, struct llentries *head); 85 86 /* 87 * Dump lle state for a specific address family. 88 */ 89 static int 90 lltable_dump_af(struct lltable *llt, struct sysctl_req *wr) 91 { 92 struct epoch_tracker et; 93 int error; 94 95 LLTABLE_LIST_LOCK_ASSERT(); 96 97 if (llt->llt_ifp->if_flags & IFF_LOOPBACK) 98 return (0); 99 error = 0; 100 101 NET_EPOCH_ENTER(et); 102 error = lltable_foreach_lle(llt, 103 (llt_foreach_cb_t *)llt->llt_dump_entry, wr); 104 NET_EPOCH_EXIT(et); 105 106 return (error); 107 } 108 109 /* 110 * Dump arp state for a specific address family. 111 */ 112 int 113 lltable_sysctl_dumparp(int af, struct sysctl_req *wr) 114 { 115 struct lltable *llt; 116 int error = 0; 117 118 LLTABLE_LIST_RLOCK(); 119 SLIST_FOREACH(llt, &V_lltables, llt_link) { 120 if (llt->llt_af == af) { 121 error = lltable_dump_af(llt, wr); 122 if (error != 0) 123 goto done; 124 } 125 } 126 done: 127 LLTABLE_LIST_RUNLOCK(); 128 return (error); 129 } 130 131 /* 132 * Adds a mbuf to hold queue. Drops old packets if the queue is full. 133 * 134 * Returns the number of held packets that were dropped. 135 */ 136 size_t 137 lltable_append_entry_queue(struct llentry *lle, struct mbuf *m, 138 size_t maxheld) 139 { 140 size_t pkts_dropped = 0; 141 142 LLE_WLOCK_ASSERT(lle); 143 144 while (lle->la_numheld >= maxheld && lle->la_hold != NULL) { 145 struct mbuf *next = lle->la_hold->m_nextpkt; 146 m_freem(lle->la_hold); 147 lle->la_hold = next; 148 lle->la_numheld--; 149 pkts_dropped++; 150 } 151 152 if (lle->la_hold != NULL) { 153 struct mbuf *curr = lle->la_hold; 154 while (curr->m_nextpkt != NULL) 155 curr = curr->m_nextpkt; 156 curr->m_nextpkt = m; 157 } else 158 lle->la_hold = m; 159 160 lle->la_numheld++; 161 162 return pkts_dropped; 163 } 164 165 166 /* 167 * Common function helpers for chained hash table. 168 */ 169 170 /* 171 * Runs specified callback for each entry in @llt. 172 * Caller does the locking. 173 * 174 */ 175 static int 176 htable_foreach_lle(struct lltable *llt, llt_foreach_cb_t *f, void *farg) 177 { 178 struct llentry *lle, *next; 179 int i, error; 180 181 error = 0; 182 183 for (i = 0; i < llt->llt_hsize; i++) { 184 CK_LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) { 185 error = f(llt, lle, farg); 186 if (error != 0) 187 break; 188 } 189 } 190 191 return (error); 192 } 193 194 /* 195 * The htable_[un]link_entry() functions return: 196 * 0 if the entry was (un)linked already and nothing changed, 197 * 1 if the entry was added/removed to/from the table, and 198 * -1 on error (e.g., not being able to add the entry due to limits reached). 199 * While the "unlink" operation should never error, callers of 200 * lltable_link_entry() need to check for errors and handle them. 201 */ 202 static int 203 htable_link_entry(struct lltable *llt, struct llentry *lle) 204 { 205 struct llentries *lleh; 206 uint32_t hashidx; 207 208 if ((lle->la_flags & LLE_LINKED) != 0) 209 return (0); 210 211 IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp); 212 213 if (llt->llt_maxentries > 0 && 214 llt->llt_entries >= llt->llt_maxentries) 215 return (-1); 216 217 hashidx = llt->llt_hash(lle, llt->llt_hsize); 218 lleh = &llt->lle_head[hashidx]; 219 220 lle->lle_tbl = llt; 221 lle->lle_head = lleh; 222 lle->la_flags |= LLE_LINKED; 223 CK_LIST_INSERT_HEAD(lleh, lle, lle_next); 224 llt->llt_entries++; 225 226 return (1); 227 } 228 229 static int 230 htable_unlink_entry(struct llentry *lle) 231 { 232 struct lltable *llt; 233 234 if ((lle->la_flags & LLE_LINKED) == 0) 235 return (0); 236 237 llt = lle->lle_tbl; 238 IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp); 239 KASSERT(llt->llt_entries > 0, ("%s: lltable %p (%s) entries %d <= 0", 240 __func__, llt, if_name(llt->llt_ifp), llt->llt_entries)); 241 242 CK_LIST_REMOVE(lle, lle_next); 243 lle->la_flags &= ~(LLE_VALID | LLE_LINKED); 244 #if 0 245 lle->lle_tbl = NULL; 246 lle->lle_head = NULL; 247 #endif 248 llt->llt_entries--; 249 250 return (1); 251 } 252 253 struct prefix_match_data { 254 const struct sockaddr *addr; 255 const struct sockaddr *mask; 256 struct llentries dchain; 257 u_int flags; 258 }; 259 260 static int 261 htable_prefix_free_cb(struct lltable *llt, struct llentry *lle, void *farg) 262 { 263 struct prefix_match_data *pmd; 264 265 pmd = (struct prefix_match_data *)farg; 266 267 if (llt->llt_match_prefix(pmd->addr, pmd->mask, pmd->flags, lle)) { 268 LLE_WLOCK(lle); 269 CK_LIST_INSERT_HEAD(&pmd->dchain, lle, lle_chain); 270 } 271 272 return (0); 273 } 274 275 static void 276 htable_prefix_free(struct lltable *llt, const struct sockaddr *addr, 277 const struct sockaddr *mask, u_int flags) 278 { 279 struct llentry *lle, *next; 280 struct prefix_match_data pmd; 281 282 bzero(&pmd, sizeof(pmd)); 283 pmd.addr = addr; 284 pmd.mask = mask; 285 pmd.flags = flags; 286 CK_LIST_INIT(&pmd.dchain); 287 288 IF_AFDATA_WLOCK(llt->llt_ifp); 289 /* Push matching lles to chain */ 290 lltable_foreach_lle(llt, htable_prefix_free_cb, &pmd); 291 292 llentries_unlink(llt, &pmd.dchain); 293 IF_AFDATA_WUNLOCK(llt->llt_ifp); 294 295 CK_LIST_FOREACH_SAFE(lle, &pmd.dchain, lle_chain, next) 296 lltable_free_entry(llt, lle); 297 } 298 299 static void 300 htable_free_tbl(struct lltable *llt) 301 { 302 303 free(llt->lle_head, M_LLTABLE); 304 free(llt, M_LLTABLE); 305 } 306 307 static void 308 llentries_unlink(struct lltable *llt, struct llentries *head) 309 { 310 struct llentry *lle, *next; 311 312 CK_LIST_FOREACH_SAFE(lle, head, lle_chain, next) 313 llt->llt_unlink_entry(lle); 314 } 315 316 /* 317 * Helper function used to drop all mbufs in hold queue. 318 * 319 * Returns the number of held packets, if any, that were dropped. 320 */ 321 size_t 322 lltable_drop_entry_queue(struct llentry *lle) 323 { 324 size_t pkts_dropped = 0; 325 326 LLE_WLOCK_ASSERT(lle); 327 328 while (lle->la_hold != NULL) { 329 struct mbuf *next = lle->la_hold->m_nextpkt; 330 m_freem(lle->la_hold); 331 lle->la_hold = next; 332 lle->la_numheld--; 333 pkts_dropped++; 334 } 335 336 KASSERT(lle->la_numheld == 0, 337 ("%s: la_numheld %d > 0, pkts_dropped %zd", __func__, 338 lle->la_numheld, pkts_dropped)); 339 340 return (pkts_dropped); 341 } 342 343 void 344 lltable_set_entry_addr(struct ifnet *ifp, struct llentry *lle, 345 const char *linkhdr, size_t linkhdrsize, int lladdr_off) 346 { 347 348 memcpy(lle->r_linkdata, linkhdr, linkhdrsize); 349 lle->r_hdrlen = linkhdrsize; 350 lle->ll_addr = &lle->r_linkdata[lladdr_off]; 351 lle->la_flags |= LLE_VALID; 352 lle->r_flags |= RLLE_VALID; 353 } 354 355 /* 356 * Acquires lltable write lock. 357 * 358 * Returns true on success, with both lltable and lle lock held. 359 * On failure, false is returned and lle wlock is still held. 360 */ 361 bool 362 lltable_acquire_wlock(struct ifnet *ifp, struct llentry *lle) 363 { 364 NET_EPOCH_ASSERT(); 365 366 /* Perform real LLE update */ 367 /* use afdata WLOCK to update fields */ 368 LLE_WUNLOCK(lle); 369 IF_AFDATA_WLOCK(ifp); 370 LLE_WLOCK(lle); 371 372 /* 373 * Since we droppped LLE lock, other thread might have deleted 374 * this lle. Check and return 375 */ 376 if ((lle->la_flags & LLE_DELETED) != 0) { 377 IF_AFDATA_WUNLOCK(ifp); 378 return (false); 379 } 380 381 return (true); 382 } 383 384 /* 385 * Tries to update @lle link-level address. 386 * Since update requires AFDATA WLOCK, function 387 * drops @lle lock, acquires AFDATA lock and then acquires 388 * @lle lock to maintain lock order. 389 * 390 * Returns 1 on success. 391 */ 392 int 393 lltable_try_set_entry_addr(struct ifnet *ifp, struct llentry *lle, 394 const char *linkhdr, size_t linkhdrsize, int lladdr_off) 395 { 396 397 if (!lltable_acquire_wlock(ifp, lle)) 398 return (0); 399 400 /* Update data */ 401 lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize, lladdr_off); 402 403 IF_AFDATA_WUNLOCK(ifp); 404 405 return (1); 406 } 407 408 /* 409 * Helper function used to pre-compute full/partial link-layer 410 * header data suitable for feeding into if_output(). 411 */ 412 int 413 lltable_calc_llheader(struct ifnet *ifp, int family, char *lladdr, 414 char *buf, size_t *bufsize, int *lladdr_off) 415 { 416 struct if_encap_req ereq; 417 int error; 418 419 bzero(buf, *bufsize); 420 bzero(&ereq, sizeof(ereq)); 421 ereq.buf = buf; 422 ereq.bufsize = *bufsize; 423 ereq.rtype = IFENCAP_LL; 424 ereq.family = family; 425 ereq.lladdr = lladdr; 426 ereq.lladdr_len = ifp->if_addrlen; 427 error = ifp->if_requestencap(ifp, &ereq); 428 if (error == 0) { 429 *bufsize = ereq.bufsize; 430 *lladdr_off = ereq.lladdr_off; 431 } 432 433 return (error); 434 } 435 436 /* 437 * Searches for the child entry matching @family inside @lle. 438 * Returns the entry or NULL. 439 */ 440 struct llentry * 441 llentry_lookup_family(struct llentry *lle, int family) 442 { 443 struct llentry *child_lle; 444 445 if (lle == NULL) 446 return (NULL); 447 448 CK_SLIST_FOREACH(child_lle, &lle->lle_children, lle_child_next) { 449 if (child_lle->r_family == family) 450 return (child_lle); 451 } 452 453 return (NULL); 454 } 455 456 /* 457 * Retrieves upper protocol family for the llentry. 458 * By default, all "normal" (e.g. upper_family == transport_family) 459 * llentries have r_family set to 0. 460 * Thus, use @default_family in that regard, otherwise use r_family. 461 * 462 * Returns upper protocol family 463 */ 464 int 465 llentry_get_upper_family(const struct llentry *lle, int default_family) 466 { 467 return (lle->r_family == 0 ? default_family : lle->r_family); 468 } 469 470 /* 471 * Prints llentry @lle data into provided buffer. 472 * Example: lle/inet/valid/em0/1.2.3.4 473 * 474 * Returns @buf. 475 */ 476 char * 477 llentry_print_buf(const struct llentry *lle, struct ifnet *ifp, int family, 478 char *buf, size_t bufsize) 479 { 480 #if defined(INET) || defined(INET6) 481 char abuf[INET6_ADDRSTRLEN]; 482 #endif 483 484 const char *valid = (lle->r_flags & RLLE_VALID) ? "valid" : "no_l2"; 485 const char *upper_str = rib_print_family(llentry_get_upper_family(lle, family)); 486 487 switch (family) { 488 #ifdef INET 489 case AF_INET: 490 inet_ntop(AF_INET, &lle->r_l3addr.addr4, abuf, sizeof(abuf)); 491 snprintf(buf, bufsize, "lle/%s/%s/%s/%s", upper_str, 492 valid, if_name(ifp), abuf); 493 break; 494 #endif 495 #ifdef INET6 496 case AF_INET6: 497 inet_ntop(AF_INET6, &lle->r_l3addr.addr6, abuf, sizeof(abuf)); 498 snprintf(buf, bufsize, "lle/%s/%s/%s/%s", upper_str, 499 valid, if_name(ifp), abuf); 500 break; 501 #endif 502 default: 503 snprintf(buf, bufsize, "lle/%s/%s/%s/????", upper_str, 504 valid, if_name(ifp)); 505 break; 506 } 507 508 return (buf); 509 } 510 511 char * 512 llentry_print_buf_lltable(const struct llentry *lle, char *buf, size_t bufsize) 513 { 514 struct lltable *tbl = lle->lle_tbl; 515 516 return (llentry_print_buf(lle, lltable_get_ifp(tbl), lltable_get_af(tbl), buf, bufsize)); 517 } 518 519 /* 520 * Requests feedback from the datapath. 521 * First packet using @lle should result in 522 * setting r_skip_req back to 0 and updating 523 * lle_hittime to the current time_uptime. 524 */ 525 void 526 llentry_request_feedback(struct llentry *lle) 527 { 528 struct llentry *child_lle; 529 530 LLE_REQ_LOCK(lle); 531 lle->r_skip_req = 1; 532 LLE_REQ_UNLOCK(lle); 533 534 CK_SLIST_FOREACH(child_lle, &lle->lle_children, lle_child_next) { 535 LLE_REQ_LOCK(child_lle); 536 child_lle->r_skip_req = 1; 537 LLE_REQ_UNLOCK(child_lle); 538 } 539 } 540 541 /* 542 * Updates the lle state to mark it has been used 543 * and record the time. 544 * Used by the llentry_provide_feedback() wrapper. 545 */ 546 void 547 llentry_mark_used(struct llentry *lle) 548 { 549 LLE_REQ_LOCK(lle); 550 lle->r_skip_req = 0; 551 lle->lle_hittime = time_uptime; 552 LLE_REQ_UNLOCK(lle); 553 } 554 555 /* 556 * Fetches the time when lle was used. 557 * Return 0 if the entry was not used, relevant time_uptime 558 * otherwise. 559 */ 560 static time_t 561 llentry_get_hittime_raw(struct llentry *lle) 562 { 563 time_t lle_hittime = 0; 564 565 LLE_REQ_LOCK(lle); 566 if ((lle->r_skip_req == 0) && (lle_hittime < lle->lle_hittime)) 567 lle_hittime = lle->lle_hittime; 568 LLE_REQ_UNLOCK(lle); 569 570 return (lle_hittime); 571 } 572 573 time_t 574 llentry_get_hittime(struct llentry *lle) 575 { 576 time_t lle_hittime = 0; 577 struct llentry *child_lle; 578 579 lle_hittime = llentry_get_hittime_raw(lle); 580 581 CK_SLIST_FOREACH(child_lle, &lle->lle_children, lle_child_next) { 582 time_t hittime = llentry_get_hittime_raw(child_lle); 583 if (hittime > lle_hittime) 584 lle_hittime = hittime; 585 } 586 587 return (lle_hittime); 588 } 589 590 /* 591 * Update link-layer header for given @lle after 592 * interface lladdr was changed. 593 */ 594 static int 595 llentry_update_ifaddr(struct lltable *llt, struct llentry *lle, void *farg) 596 { 597 struct ifnet *ifp; 598 u_char linkhdr[LLE_MAX_LINKHDR]; 599 size_t linkhdrsize; 600 u_char *lladdr; 601 int lladdr_off; 602 603 ifp = (struct ifnet *)farg; 604 605 lladdr = lle->ll_addr; 606 607 LLE_WLOCK(lle); 608 if ((lle->la_flags & LLE_VALID) == 0) { 609 LLE_WUNLOCK(lle); 610 return (0); 611 } 612 613 if ((lle->la_flags & LLE_IFADDR) != 0) 614 lladdr = IF_LLADDR(ifp); 615 616 linkhdrsize = sizeof(linkhdr); 617 lltable_calc_llheader(ifp, llt->llt_af, lladdr, linkhdr, &linkhdrsize, 618 &lladdr_off); 619 memcpy(lle->r_linkdata, linkhdr, linkhdrsize); 620 LLE_WUNLOCK(lle); 621 622 return (0); 623 } 624 625 /* 626 * Update all calculated headers for given @llt 627 */ 628 void 629 lltable_update_ifaddr(struct lltable *llt) 630 { 631 632 if (llt->llt_ifp->if_flags & IFF_LOOPBACK) 633 return; 634 635 IF_AFDATA_WLOCK(llt->llt_ifp); 636 lltable_foreach_lle(llt, llentry_update_ifaddr, llt->llt_ifp); 637 IF_AFDATA_WUNLOCK(llt->llt_ifp); 638 } 639 640 /* 641 * 642 * Performs generic cleanup routines and frees lle. 643 * 644 * Called for non-linked entries, with callouts and 645 * other AF-specific cleanups performed. 646 * 647 * @lle must be passed WLOCK'ed 648 * 649 * Returns the number of held packets, if any, that were dropped. 650 */ 651 size_t 652 llentry_free(struct llentry *lle) 653 { 654 size_t pkts_dropped; 655 656 LLE_WLOCK_ASSERT(lle); 657 658 KASSERT((lle->la_flags & LLE_LINKED) == 0, ("freeing linked lle")); 659 660 pkts_dropped = lltable_drop_entry_queue(lle); 661 662 /* cancel timer */ 663 if (callout_stop(&lle->lle_timer) > 0) 664 LLE_REMREF(lle); 665 LLE_FREE_LOCKED(lle); 666 667 return (pkts_dropped); 668 } 669 670 /* 671 * Free all entries from given table and free itself. 672 */ 673 674 static int 675 lltable_free_cb(struct lltable *llt, struct llentry *lle, void *farg) 676 { 677 struct llentries *dchain; 678 679 dchain = (struct llentries *)farg; 680 681 LLE_WLOCK(lle); 682 CK_LIST_INSERT_HEAD(dchain, lle, lle_chain); 683 684 return (0); 685 } 686 687 /* 688 * Free all entries from given table and free itself. 689 */ 690 void 691 lltable_free(struct lltable *llt) 692 { 693 struct llentry *lle, *next; 694 struct llentries dchain; 695 696 KASSERT(llt != NULL, ("%s: llt is NULL", __func__)); 697 698 lltable_unlink(llt); 699 700 CK_LIST_INIT(&dchain); 701 IF_AFDATA_WLOCK(llt->llt_ifp); 702 /* Push all lles to @dchain */ 703 lltable_foreach_lle(llt, lltable_free_cb, &dchain); 704 llentries_unlink(llt, &dchain); 705 IF_AFDATA_WUNLOCK(llt->llt_ifp); 706 707 CK_LIST_FOREACH_SAFE(lle, &dchain, lle_chain, next) { 708 llentry_free(lle); 709 } 710 711 KASSERT(llt->llt_entries == 0, ("%s: lltable %p (%s) entires not 0: %d", 712 __func__, llt, llt->llt_ifp->if_xname, llt->llt_entries)); 713 714 llt->llt_free_tbl(llt); 715 } 716 717 /* 718 * Deletes an address from given lltable. 719 * Used for userland interaction to remove 720 * individual entries. Skips entries added by OS. 721 */ 722 int 723 lltable_delete_addr(struct lltable *llt, u_int flags, 724 const struct sockaddr *l3addr) 725 { 726 struct llentry *lle; 727 struct ifnet *ifp; 728 729 ifp = llt->llt_ifp; 730 IF_AFDATA_WLOCK(ifp); 731 lle = lla_lookup(llt, LLE_SF(l3addr->sa_family, LLE_EXCLUSIVE), l3addr); 732 733 if (lle == NULL) { 734 IF_AFDATA_WUNLOCK(ifp); 735 return (ENOENT); 736 } 737 if ((lle->la_flags & LLE_IFADDR) != 0 && (flags & LLE_IFADDR) == 0) { 738 IF_AFDATA_WUNLOCK(ifp); 739 LLE_WUNLOCK(lle); 740 return (EPERM); 741 } 742 743 lltable_unlink_entry(llt, lle); 744 IF_AFDATA_WUNLOCK(ifp); 745 746 llt->llt_delete_entry(llt, lle); 747 748 return (0); 749 } 750 751 void 752 lltable_prefix_free(int af, struct sockaddr *addr, struct sockaddr *mask, 753 u_int flags) 754 { 755 struct lltable *llt; 756 757 LLTABLE_LIST_RLOCK(); 758 SLIST_FOREACH(llt, &V_lltables, llt_link) { 759 if (llt->llt_af != af) 760 continue; 761 762 llt->llt_prefix_free(llt, addr, mask, flags); 763 } 764 LLTABLE_LIST_RUNLOCK(); 765 } 766 767 /* 768 * Delete llentries that func() returns true. 769 */ 770 struct lle_match_data { 771 struct llentries dchain; 772 llt_match_cb_t *func; 773 void *farg; 774 }; 775 776 static int 777 lltable_delete_conditional_cb(struct lltable *llt, struct llentry *lle, 778 void *farg) 779 { 780 struct lle_match_data *lmd; 781 782 lmd = (struct lle_match_data *)farg; 783 if (lmd->func(llt, lle, lmd->farg)) { 784 LLE_WLOCK(lle); 785 CK_LIST_INSERT_HEAD(&lmd->dchain, lle, lle_chain); 786 } 787 788 return (0); 789 } 790 791 void 792 lltable_delete_conditional(struct lltable *llt, llt_match_cb_t *func, 793 void *farg) 794 { 795 struct llentry *lle, *next; 796 struct lle_match_data lmd; 797 798 bzero(&lmd, sizeof(lmd)); 799 CK_LIST_INIT(&lmd.dchain); 800 lmd.func = func; 801 lmd.farg = farg; 802 803 IF_AFDATA_WLOCK(llt->llt_ifp); 804 lltable_foreach_lle(llt, lltable_delete_conditional_cb, &lmd); 805 llentries_unlink(llt, &lmd.dchain); 806 IF_AFDATA_WUNLOCK(llt->llt_ifp); 807 808 CK_LIST_FOREACH_SAFE(lle, &lmd.dchain, lle_chain, next) 809 llt->llt_delete_entry(llt, lle); 810 } 811 812 struct lltable * 813 lltable_allocate_htbl(uint32_t hsize) 814 { 815 struct lltable *llt; 816 int i; 817 818 llt = malloc(sizeof(struct lltable), M_LLTABLE, M_WAITOK | M_ZERO); 819 llt->llt_hsize = hsize; 820 llt->lle_head = malloc(sizeof(struct llentries) * hsize, 821 M_LLTABLE, M_WAITOK | M_ZERO); 822 823 for (i = 0; i < llt->llt_hsize; i++) 824 CK_LIST_INIT(&llt->lle_head[i]); 825 826 /* Set some default callbacks */ 827 llt->llt_link_entry = htable_link_entry; 828 llt->llt_unlink_entry = htable_unlink_entry; 829 llt->llt_prefix_free = htable_prefix_free; 830 llt->llt_foreach_entry = htable_foreach_lle; 831 llt->llt_free_tbl = htable_free_tbl; 832 833 return (llt); 834 } 835 836 /* 837 * Links lltable to global llt list. 838 */ 839 void 840 lltable_link(struct lltable *llt) 841 { 842 843 LLTABLE_LIST_WLOCK(); 844 SLIST_INSERT_HEAD(&V_lltables, llt, llt_link); 845 LLTABLE_LIST_WUNLOCK(); 846 } 847 848 static void 849 lltable_unlink(struct lltable *llt) 850 { 851 852 LLTABLE_LIST_WLOCK(); 853 SLIST_REMOVE(&V_lltables, llt, lltable, llt_link); 854 LLTABLE_LIST_WUNLOCK(); 855 856 } 857 858 /* 859 * Gets interface @ifp lltable for the specified @family 860 */ 861 struct lltable * 862 lltable_get(struct ifnet *ifp, int family) 863 { 864 switch (family) { 865 #ifdef INET 866 case AF_INET: 867 return (in_lltable_get(ifp)); 868 #endif 869 #ifdef INET6 870 case AF_INET6: 871 return (in6_lltable_get(ifp)); 872 #endif 873 } 874 875 return (NULL); 876 } 877 878 /* 879 * External methods used by lltable consumers 880 */ 881 882 int 883 lltable_foreach_lle(struct lltable *llt, llt_foreach_cb_t *f, void *farg) 884 { 885 886 return (llt->llt_foreach_entry(llt, f, farg)); 887 } 888 889 struct llentry * 890 lltable_alloc_entry(struct lltable *llt, u_int flags, 891 const struct sockaddr *l3addr) 892 { 893 894 return (llt->llt_alloc_entry(llt, flags, l3addr)); 895 } 896 897 void 898 lltable_free_entry(struct lltable *llt, struct llentry *lle) 899 { 900 901 llt->llt_free_entry(llt, lle); 902 } 903 904 int 905 lltable_link_entry(struct lltable *llt, struct llentry *lle) 906 { 907 908 return (llt->llt_link_entry(llt, lle)); 909 } 910 911 void 912 lltable_link_child_entry(struct llentry *lle, struct llentry *child_lle) 913 { 914 child_lle->lle_parent = lle; 915 child_lle->lle_tbl = lle->lle_tbl; 916 child_lle->la_flags |= LLE_LINKED; 917 CK_SLIST_INSERT_HEAD(&lle->lle_children, child_lle, lle_child_next); 918 } 919 920 void 921 lltable_unlink_child_entry(struct llentry *child_lle) 922 { 923 struct llentry *lle = child_lle->lle_parent; 924 925 child_lle->la_flags &= ~LLE_LINKED; 926 child_lle->lle_parent = NULL; 927 CK_SLIST_REMOVE(&lle->lle_children, child_lle, llentry, lle_child_next); 928 } 929 930 int 931 lltable_unlink_entry(struct lltable *llt, struct llentry *lle) 932 { 933 934 return (llt->llt_unlink_entry(lle)); 935 } 936 937 void 938 lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa) 939 { 940 struct lltable *llt; 941 942 llt = lle->lle_tbl; 943 llt->llt_fill_sa_entry(lle, sa); 944 } 945 946 struct ifnet * 947 lltable_get_ifp(const struct lltable *llt) 948 { 949 950 return (llt->llt_ifp); 951 } 952 953 int 954 lltable_get_af(const struct lltable *llt) 955 { 956 957 return (llt->llt_af); 958 } 959 960 /* 961 * Called in route_output when rtm_flags contains RTF_LLDATA. 962 */ 963 int 964 lla_rt_output(struct rt_msghdr *rtm, struct rt_addrinfo *info) 965 { 966 struct sockaddr_dl *dl = 967 (struct sockaddr_dl *)info->rti_info[RTAX_GATEWAY]; 968 struct sockaddr *dst = (struct sockaddr *)info->rti_info[RTAX_DST]; 969 struct ifnet *ifp; 970 struct lltable *llt; 971 struct llentry *lle, *lle_tmp; 972 uint8_t linkhdr[LLE_MAX_LINKHDR]; 973 size_t linkhdrsize; 974 int lladdr_off; 975 u_int laflags = 0; 976 int error; 977 978 if (dl == NULL || dl->sdl_family != AF_LINK) 979 return (EINVAL); 980 981 /* XXX: should be ntohs() */ 982 ifp = ifnet_byindex(dl->sdl_index); 983 if (ifp == NULL) { 984 log(LOG_INFO, "%s: invalid ifp (sdl_index %d)\n", 985 __func__, dl->sdl_index); 986 return EINVAL; 987 } 988 989 llt = lltable_get(ifp, dst->sa_family); 990 991 if (llt == NULL) 992 return (ESRCH); 993 994 error = 0; 995 996 switch (rtm->rtm_type) { 997 case RTM_ADD: 998 /* Add static LLE */ 999 laflags = 0; 1000 if (rtm->rtm_rmx.rmx_expire == 0) 1001 laflags = LLE_STATIC; 1002 lle = lltable_alloc_entry(llt, laflags, dst); 1003 if (lle == NULL) 1004 return (ENOMEM); 1005 1006 linkhdrsize = sizeof(linkhdr); 1007 if (lltable_calc_llheader(ifp, dst->sa_family, LLADDR(dl), 1008 linkhdr, &linkhdrsize, &lladdr_off) != 0) { 1009 lltable_free_entry(llt, lle); 1010 return (EINVAL); 1011 } 1012 lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize, 1013 lladdr_off); 1014 if ((rtm->rtm_flags & RTF_ANNOUNCE)) 1015 lle->la_flags |= LLE_PUB; 1016 lle->la_expire = rtm->rtm_rmx.rmx_expire; 1017 1018 laflags = lle->la_flags; 1019 1020 /* Try to link new entry */ 1021 lle_tmp = NULL; 1022 IF_AFDATA_WLOCK(ifp); 1023 LLE_WLOCK(lle); 1024 lle_tmp = lla_lookup(llt, LLE_EXCLUSIVE, dst); 1025 if (lle_tmp != NULL) { 1026 /* Check if we are trying to replace immutable entry */ 1027 if ((lle_tmp->la_flags & LLE_IFADDR) != 0) { 1028 IF_AFDATA_WUNLOCK(ifp); 1029 LLE_WUNLOCK(lle_tmp); 1030 lltable_free_entry(llt, lle); 1031 return (EPERM); 1032 } 1033 /* Unlink existing entry from table */ 1034 lltable_unlink_entry(llt, lle_tmp); 1035 } 1036 lltable_link_entry(llt, lle); 1037 if ((lle->la_flags & LLE_PUB) != 0 && 1038 (llt->llt_flags & LLT_ADDEDPROXY) == 0) 1039 llt->llt_flags |= LLT_ADDEDPROXY; 1040 IF_AFDATA_WUNLOCK(ifp); 1041 1042 if (lle_tmp != NULL) { 1043 EVENTHANDLER_INVOKE(lle_event, lle_tmp,LLENTRY_EXPIRED); 1044 lltable_free_entry(llt, lle_tmp); 1045 } 1046 1047 /* 1048 * By invoking LLE handler here we might get 1049 * two events on static LLE entry insertion 1050 * in routing socket. However, since we might have 1051 * other subscribers we need to generate this event. 1052 */ 1053 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_RESOLVED); 1054 LLE_WUNLOCK(lle); 1055 llt->llt_post_resolved(llt, lle); 1056 break; 1057 1058 case RTM_DELETE: 1059 return (lltable_delete_addr(llt, 0, dst)); 1060 1061 default: 1062 error = EINVAL; 1063 } 1064 1065 return (error); 1066 } 1067 1068 #ifdef DDB 1069 struct llentry_sa { 1070 struct llentry base; 1071 struct sockaddr l3_addr; 1072 }; 1073 1074 static void 1075 llatbl_lle_show(struct llentry_sa *la) 1076 { 1077 struct llentry *lle; 1078 uint8_t octet[6]; 1079 1080 lle = &la->base; 1081 db_printf("lle=%p\n", lle); 1082 db_printf(" lle_next=%p\n", lle->lle_next.cle_next); 1083 db_printf(" lle_lock=%p\n", &lle->lle_lock); 1084 db_printf(" lle_tbl=%p\n", lle->lle_tbl); 1085 db_printf(" lle_head=%p\n", lle->lle_head); 1086 db_printf(" la_hold=%p\n", lle->la_hold); 1087 db_printf(" la_numheld=%d\n", lle->la_numheld); 1088 db_printf(" la_expire=%ju\n", (uintmax_t)lle->la_expire); 1089 db_printf(" la_flags=0x%04x\n", lle->la_flags); 1090 db_printf(" la_asked=%u\n", lle->la_asked); 1091 db_printf(" la_preempt=%u\n", lle->la_preempt); 1092 db_printf(" ln_state=%d\n", lle->ln_state); 1093 db_printf(" ln_router=%u\n", lle->ln_router); 1094 db_printf(" ln_ntick=%ju\n", (uintmax_t)lle->ln_ntick); 1095 db_printf(" lle_refcnt=%d\n", lle->lle_refcnt); 1096 bcopy(lle->ll_addr, octet, sizeof(octet)); 1097 db_printf(" ll_addr=%02x:%02x:%02x:%02x:%02x:%02x\n", 1098 octet[0], octet[1], octet[2], octet[3], octet[4], octet[5]); 1099 db_printf(" lle_timer=%p\n", &lle->lle_timer); 1100 1101 switch (la->l3_addr.sa_family) { 1102 #ifdef INET 1103 case AF_INET: 1104 { 1105 struct sockaddr_in *sin; 1106 char l3s[INET_ADDRSTRLEN]; 1107 1108 sin = (struct sockaddr_in *)&la->l3_addr; 1109 inet_ntoa_r(sin->sin_addr, l3s); 1110 db_printf(" l3_addr=%s\n", l3s); 1111 break; 1112 } 1113 #endif 1114 #ifdef INET6 1115 case AF_INET6: 1116 { 1117 struct sockaddr_in6 *sin6; 1118 char l3s[INET6_ADDRSTRLEN]; 1119 1120 sin6 = (struct sockaddr_in6 *)&la->l3_addr; 1121 ip6_sprintf(l3s, &sin6->sin6_addr); 1122 db_printf(" l3_addr=%s\n", l3s); 1123 break; 1124 } 1125 #endif 1126 default: 1127 db_printf(" l3_addr=N/A (af=%d)\n", la->l3_addr.sa_family); 1128 break; 1129 } 1130 } 1131 1132 DB_SHOW_COMMAND(llentry, db_show_llentry) 1133 { 1134 1135 if (!have_addr) { 1136 db_printf("usage: show llentry <struct llentry *>\n"); 1137 return; 1138 } 1139 1140 llatbl_lle_show((struct llentry_sa *)addr); 1141 } 1142 1143 static void 1144 llatbl_llt_show(struct lltable *llt) 1145 { 1146 int i; 1147 struct llentry *lle; 1148 1149 db_printf("llt=%p llt_af=%d llt_ifp=%p\n", 1150 llt, llt->llt_af, llt->llt_ifp); 1151 1152 for (i = 0; i < llt->llt_hsize; i++) { 1153 CK_LIST_FOREACH(lle, &llt->lle_head[i], lle_next) { 1154 llatbl_lle_show((struct llentry_sa *)lle); 1155 if (db_pager_quit) 1156 return; 1157 } 1158 } 1159 } 1160 1161 DB_SHOW_COMMAND(lltable, db_show_lltable) 1162 { 1163 1164 if (!have_addr) { 1165 db_printf("usage: show lltable <struct lltable *>\n"); 1166 return; 1167 } 1168 1169 llatbl_llt_show((struct lltable *)addr); 1170 } 1171 1172 DB_SHOW_ALL_COMMAND(lltables, db_show_all_lltables) 1173 { 1174 VNET_ITERATOR_DECL(vnet_iter); 1175 struct lltable *llt; 1176 1177 VNET_FOREACH(vnet_iter) { 1178 CURVNET_SET_QUIET(vnet_iter); 1179 #ifdef VIMAGE 1180 db_printf("vnet=%p\n", curvnet); 1181 #endif 1182 SLIST_FOREACH(llt, &V_lltables, llt_link) { 1183 db_printf("llt=%p llt_af=%d llt_ifp=%p(%s)\n", 1184 llt, llt->llt_af, llt->llt_ifp, 1185 (llt->llt_ifp != NULL) ? 1186 llt->llt_ifp->if_xname : "?"); 1187 if (have_addr && addr != 0) /* verbose */ 1188 llatbl_llt_show(llt); 1189 if (db_pager_quit) { 1190 CURVNET_RESTORE(); 1191 return; 1192 } 1193 } 1194 CURVNET_RESTORE(); 1195 } 1196 } 1197 #endif 1198