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