2 * Copyright (c) 2004 Luigi Rizzo, Alessandro Cerri. All rights reserved.
3 * Copyright (c) 2004-2008 Qing Li. All rights reserved.
4 * Copyright (c) 2008 Kip Macy. All rights reserved.
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
15 * THIS SOFTWARE IS PROVIDED BY AUTHOR AND CONTRIBUTORS ``AS IS'' AND
16 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18 * ARE DISCLAIMED. IN NO EVENT SHALL AUTHOR OR CONTRIBUTORS BE LIABLE
19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
21 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 #include <sys/cdefs.h>
28 __FBSDID("$FreeBSD$");
32 #include "opt_inet6.h"
34 #include <sys/param.h>
35 #include <sys/systm.h>
36 #include <sys/malloc.h>
38 #include <sys/syslog.h>
39 #include <sys/sysctl.h>
40 #include <sys/socket.h>
41 #include <sys/kernel.h>
43 #include <sys/mutex.h>
44 #include <sys/rwlock.h>
52 #include <netinet/in.h>
53 #include <net/if_llatbl.h>
55 #include <net/if_dl.h>
56 #include <net/if_var.h>
57 #include <net/route.h>
59 #include <netinet/if_ether.h>
60 #include <netinet6/in6_var.h>
61 #include <netinet6/nd6.h>
63 MALLOC_DEFINE(M_LLTABLE, "lltable", "link level address tables");
65 static VNET_DEFINE(SLIST_HEAD(, lltable), lltables) =
66 SLIST_HEAD_INITIALIZER(lltables);
67 #define V_lltables VNET(lltables)
69 struct rwlock lltable_rwlock;
70 RW_SYSINIT(lltable_rwlock, &lltable_rwlock, "lltable_rwlock");
72 static void lltable_unlink(struct lltable *llt);
73 static void llentries_unlink(struct lltable *llt, struct llentries *head);
75 static void htable_unlink_entry(struct llentry *lle);
76 static void htable_link_entry(struct lltable *llt, struct llentry *lle);
77 static int htable_foreach_lle(struct lltable *llt, llt_foreach_cb_t *f,
81 * Dump lle state for a specific address family.
84 lltable_dump_af(struct lltable *llt, struct sysctl_req *wr)
88 LLTABLE_LOCK_ASSERT();
90 if (llt->llt_ifp->if_flags & IFF_LOOPBACK)
94 IF_AFDATA_RLOCK(llt->llt_ifp);
95 error = lltable_foreach_lle(llt,
96 (llt_foreach_cb_t *)llt->llt_dump_entry, wr);
97 IF_AFDATA_RUNLOCK(llt->llt_ifp);
103 * Dump arp state for a specific address family.
106 lltable_sysctl_dumparp(int af, struct sysctl_req *wr)
112 SLIST_FOREACH(llt, &V_lltables, llt_link) {
113 if (llt->llt_af == af) {
114 error = lltable_dump_af(llt, wr);
125 * Common function helpers for chained hash table.
129 * Runs specified callback for each entry in @llt.
130 * Caller does the locking.
134 htable_foreach_lle(struct lltable *llt, llt_foreach_cb_t *f, void *farg)
136 struct llentry *lle, *next;
141 for (i = 0; i < llt->llt_hsize; i++) {
142 LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
143 error = f(llt, lle, farg);
153 htable_link_entry(struct lltable *llt, struct llentry *lle)
155 struct llentries *lleh;
158 if ((lle->la_flags & LLE_LINKED) != 0)
161 IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
163 hashidx = llt->llt_hash(lle, llt->llt_hsize);
164 lleh = &llt->lle_head[hashidx];
167 lle->lle_head = lleh;
168 lle->la_flags |= LLE_LINKED;
169 LIST_INSERT_HEAD(lleh, lle, lle_next);
173 htable_unlink_entry(struct llentry *lle)
176 if ((lle->la_flags & LLE_LINKED) != 0) {
177 IF_AFDATA_WLOCK_ASSERT(lle->lle_tbl->llt_ifp);
178 LIST_REMOVE(lle, lle_next);
179 lle->la_flags &= ~(LLE_VALID | LLE_LINKED);
182 lle->lle_head = NULL;
187 struct prefix_match_data {
188 const struct sockaddr *addr;
189 const struct sockaddr *mask;
190 struct llentries dchain;
195 htable_prefix_free_cb(struct lltable *llt, struct llentry *lle, void *farg)
197 struct prefix_match_data *pmd;
199 pmd = (struct prefix_match_data *)farg;
201 if (llt->llt_match_prefix(pmd->addr, pmd->mask, pmd->flags, lle)) {
203 LIST_INSERT_HEAD(&pmd->dchain, lle, lle_chain);
210 htable_prefix_free(struct lltable *llt, const struct sockaddr *addr,
211 const struct sockaddr *mask, u_int flags)
213 struct llentry *lle, *next;
214 struct prefix_match_data pmd;
216 bzero(&pmd, sizeof(pmd));
220 LIST_INIT(&pmd.dchain);
222 IF_AFDATA_WLOCK(llt->llt_ifp);
223 /* Push matching lles to chain */
224 lltable_foreach_lle(llt, htable_prefix_free_cb, &pmd);
226 llentries_unlink(llt, &pmd.dchain);
227 IF_AFDATA_WUNLOCK(llt->llt_ifp);
229 LIST_FOREACH_SAFE(lle, &pmd.dchain, lle_chain, next)
230 lltable_free_entry(llt, lle);
234 htable_free_tbl(struct lltable *llt)
237 free(llt->lle_head, M_LLTABLE);
238 free(llt, M_LLTABLE);
242 llentries_unlink(struct lltable *llt, struct llentries *head)
244 struct llentry *lle, *next;
246 LIST_FOREACH_SAFE(lle, head, lle_chain, next)
247 llt->llt_unlink_entry(lle);
251 * Helper function used to drop all mbufs in hold queue.
253 * Returns the number of held packets, if any, that were dropped.
256 lltable_drop_entry_queue(struct llentry *lle)
261 LLE_WLOCK_ASSERT(lle);
264 while ((lle->la_numheld > 0) && (lle->la_hold != NULL)) {
265 next = lle->la_hold->m_nextpkt;
266 m_freem(lle->la_hold);
272 KASSERT(lle->la_numheld == 0,
273 ("%s: la_numheld %d > 0, pkts_droped %zd", __func__,
274 lle->la_numheld, pkts_dropped));
276 return (pkts_dropped);
280 lltable_set_entry_addr(struct ifnet *ifp, struct llentry *lle,
284 bcopy(lladdr, &lle->ll_addr, ifp->if_addrlen);
285 lle->la_flags |= LLE_VALID;
286 lle->r_flags |= RLLE_VALID;
290 * Tries to update @lle link-level address.
291 * Since update requires AFDATA WLOCK, function
292 * drops @lle lock, acquires AFDATA lock and then acquires
293 * @lle lock to maintain lock order.
295 * Returns 1 on success.
298 lltable_try_set_entry_addr(struct ifnet *ifp, struct llentry *lle,
302 /* Perform real LLE update */
303 /* use afdata WLOCK to update fields */
304 LLE_WLOCK_ASSERT(lle);
307 IF_AFDATA_WLOCK(ifp);
311 * Since we droppped LLE lock, other thread might have deleted
312 * this lle. Check and return
314 if ((lle->la_flags & LLE_DELETED) != 0) {
315 IF_AFDATA_WUNLOCK(ifp);
316 LLE_FREE_LOCKED(lle);
321 lltable_set_entry_addr(ifp, lle, lladdr);
323 IF_AFDATA_WUNLOCK(ifp);
332 * Performes generic cleanup routines and frees lle.
334 * Called for non-linked entries, with callouts and
335 * other AF-specific cleanups performed.
337 * @lle must be passed WLOCK'ed
339 * Returns the number of held packets, if any, that were dropped.
342 llentry_free(struct llentry *lle)
346 LLE_WLOCK_ASSERT(lle);
348 KASSERT((lle->la_flags & LLE_LINKED) == 0, ("freeing linked lle"));
350 pkts_dropped = lltable_drop_entry_queue(lle);
352 LLE_FREE_LOCKED(lle);
354 return (pkts_dropped);
358 * (al)locate an llentry for address dst (equivalent to rtalloc for new-arp).
360 * If found the llentry * is returned referenced and unlocked.
363 llentry_alloc(struct ifnet *ifp, struct lltable *lt,
364 struct sockaddr_storage *dst)
366 struct llentry *la, *la_tmp;
368 IF_AFDATA_RLOCK(ifp);
369 la = lla_lookup(lt, LLE_EXCLUSIVE, (struct sockaddr *)dst);
370 IF_AFDATA_RUNLOCK(ifp);
378 if ((ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0) {
379 la = lltable_alloc_entry(lt, 0, (struct sockaddr *)dst);
382 IF_AFDATA_WLOCK(ifp);
384 /* Prefer any existing LLE over newly-created one */
385 la_tmp = lla_lookup(lt, LLE_EXCLUSIVE, (struct sockaddr *)dst);
387 lltable_link_entry(lt, la);
388 IF_AFDATA_WUNLOCK(ifp);
389 if (la_tmp != NULL) {
390 lltable_free_entry(lt, la);
401 * Free all entries from given table and free itself.
405 lltable_free_cb(struct lltable *llt, struct llentry *lle, void *farg)
407 struct llentries *dchain;
409 dchain = (struct llentries *)farg;
412 LIST_INSERT_HEAD(dchain, lle, lle_chain);
418 * Free all entries from given table and free itself.
421 lltable_free(struct lltable *llt)
423 struct llentry *lle, *next;
424 struct llentries dchain;
426 KASSERT(llt != NULL, ("%s: llt is NULL", __func__));
431 IF_AFDATA_WLOCK(llt->llt_ifp);
432 /* Push all lles to @dchain */
433 lltable_foreach_lle(llt, lltable_free_cb, &dchain);
434 llentries_unlink(llt, &dchain);
435 IF_AFDATA_WUNLOCK(llt->llt_ifp);
437 LIST_FOREACH_SAFE(lle, &dchain, lle_chain, next) {
438 if (callout_stop(&lle->lle_timer) > 0)
443 llt->llt_free_tbl(llt);
448 lltable_drain(int af)
455 SLIST_FOREACH(llt, &V_lltables, llt_link) {
456 if (llt->llt_af != af)
459 for (i=0; i < llt->llt_hsize; i++) {
460 LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
463 m_freem(lle->la_hold);
475 * Deletes an address from given lltable.
476 * Used for userland interaction to remove
477 * individual entries. Skips entries added by OS.
480 lltable_delete_addr(struct lltable *llt, u_int flags,
481 const struct sockaddr *l3addr)
487 IF_AFDATA_WLOCK(ifp);
488 lle = lla_lookup(llt, LLE_EXCLUSIVE, l3addr);
491 IF_AFDATA_WUNLOCK(ifp);
494 if ((lle->la_flags & LLE_IFADDR) != 0 && (flags & LLE_IFADDR) == 0) {
495 IF_AFDATA_WUNLOCK(ifp);
500 lltable_unlink_entry(llt, lle);
501 IF_AFDATA_WUNLOCK(ifp);
503 llt->llt_delete_entry(llt, lle);
509 lltable_prefix_free(int af, struct sockaddr *addr, struct sockaddr *mask,
515 SLIST_FOREACH(llt, &V_lltables, llt_link) {
516 if (llt->llt_af != af)
519 llt->llt_prefix_free(llt, addr, mask, flags);
525 lltable_allocate_htbl(uint32_t hsize)
530 llt = malloc(sizeof(struct lltable), M_LLTABLE, M_WAITOK | M_ZERO);
531 llt->llt_hsize = hsize;
532 llt->lle_head = malloc(sizeof(struct llentries) * hsize,
533 M_LLTABLE, M_WAITOK | M_ZERO);
535 for (i = 0; i < llt->llt_hsize; i++)
536 LIST_INIT(&llt->lle_head[i]);
538 /* Set some default callbacks */
539 llt->llt_link_entry = htable_link_entry;
540 llt->llt_unlink_entry = htable_unlink_entry;
541 llt->llt_prefix_free = htable_prefix_free;
542 llt->llt_foreach_entry = htable_foreach_lle;
543 llt->llt_free_tbl = htable_free_tbl;
549 * Links lltable to global llt list.
552 lltable_link(struct lltable *llt)
556 SLIST_INSERT_HEAD(&V_lltables, llt, llt_link);
561 lltable_unlink(struct lltable *llt)
565 SLIST_REMOVE(&V_lltables, llt, lltable, llt_link);
571 * External methods used by lltable consumers
575 lltable_foreach_lle(struct lltable *llt, llt_foreach_cb_t *f, void *farg)
578 return (llt->llt_foreach_entry(llt, f, farg));
582 lltable_alloc_entry(struct lltable *llt, u_int flags,
583 const struct sockaddr *l3addr)
586 return (llt->llt_alloc_entry(llt, flags, l3addr));
590 lltable_free_entry(struct lltable *llt, struct llentry *lle)
593 llt->llt_free_entry(llt, lle);
597 lltable_link_entry(struct lltable *llt, struct llentry *lle)
600 llt->llt_link_entry(llt, lle);
604 lltable_unlink_entry(struct lltable *llt, struct llentry *lle)
607 llt->llt_unlink_entry(lle);
611 lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
616 llt->llt_fill_sa_entry(lle, sa);
620 lltable_get_ifp(const struct lltable *llt)
623 return (llt->llt_ifp);
627 lltable_get_af(const struct lltable *llt)
630 return (llt->llt_af);
634 * Called in route_output when rtm_flags contains RTF_LLDATA.
637 lla_rt_output(struct rt_msghdr *rtm, struct rt_addrinfo *info)
639 struct sockaddr_dl *dl =
640 (struct sockaddr_dl *)info->rti_info[RTAX_GATEWAY];
641 struct sockaddr *dst = (struct sockaddr *)info->rti_info[RTAX_DST];
644 struct llentry *lle, *lle_tmp;
648 KASSERT(dl != NULL && dl->sdl_family == AF_LINK,
649 ("%s: invalid dl\n", __func__));
651 ifp = ifnet_byindex(dl->sdl_index);
653 log(LOG_INFO, "%s: invalid ifp (sdl_index %d)\n",
654 __func__, dl->sdl_index);
658 /* XXX linked list may be too expensive */
660 SLIST_FOREACH(llt, &V_lltables, llt_link) {
661 if (llt->llt_af == dst->sa_family &&
666 KASSERT(llt != NULL, ("Yep, ugly hacks are bad\n"));
670 switch (rtm->rtm_type) {
674 if (rtm->rtm_rmx.rmx_expire == 0)
675 laflags = LLE_STATIC;
676 lle = lltable_alloc_entry(llt, laflags, dst);
680 bcopy(LLADDR(dl), &lle->ll_addr, ifp->if_addrlen);
681 if ((rtm->rtm_flags & RTF_ANNOUNCE))
682 lle->la_flags |= LLE_PUB;
683 lle->la_flags |= LLE_VALID;
684 lle->r_flags |= RLLE_VALID;
685 lle->la_expire = rtm->rtm_rmx.rmx_expire;
687 laflags = lle->la_flags;
689 /* Try to link new entry */
691 IF_AFDATA_WLOCK(ifp);
693 lle_tmp = lla_lookup(llt, LLE_EXCLUSIVE, dst);
694 if (lle_tmp != NULL) {
695 /* Check if we are trying to replace immutable entry */
696 if ((lle_tmp->la_flags & LLE_IFADDR) != 0) {
697 IF_AFDATA_WUNLOCK(ifp);
698 LLE_WUNLOCK(lle_tmp);
699 lltable_free_entry(llt, lle);
702 /* Unlink existing entry from table */
703 lltable_unlink_entry(llt, lle_tmp);
705 lltable_link_entry(llt, lle);
706 IF_AFDATA_WUNLOCK(ifp);
708 if (lle_tmp != NULL) {
709 EVENTHANDLER_INVOKE(lle_event, lle_tmp,LLENTRY_EXPIRED);
710 lltable_free_entry(llt, lle_tmp);
714 * By invoking LLE handler here we might get
715 * two events on static LLE entry insertion
716 * in routing socket. However, since we might have
717 * other subscribers we need to generate this event.
719 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_RESOLVED);
723 if ((laflags & LLE_PUB) && dst->sa_family == AF_INET)
725 &((struct sockaddr_in *)dst)->sin_addr,
726 &((struct sockaddr_in *)dst)->sin_addr,
727 (u_char *)LLADDR(dl));
733 return (lltable_delete_addr(llt, 0, dst));
745 struct sockaddr l3_addr;
749 llatbl_lle_show(struct llentry_sa *la)
755 db_printf("lle=%p\n", lle);
756 db_printf(" lle_next=%p\n", lle->lle_next.le_next);
757 db_printf(" lle_lock=%p\n", &lle->lle_lock);
758 db_printf(" lle_tbl=%p\n", lle->lle_tbl);
759 db_printf(" lle_head=%p\n", lle->lle_head);
760 db_printf(" la_hold=%p\n", lle->la_hold);
761 db_printf(" la_numheld=%d\n", lle->la_numheld);
762 db_printf(" la_expire=%ju\n", (uintmax_t)lle->la_expire);
763 db_printf(" la_flags=0x%04x\n", lle->la_flags);
764 db_printf(" la_asked=%u\n", lle->la_asked);
765 db_printf(" la_preempt=%u\n", lle->la_preempt);
766 db_printf(" ln_state=%d\n", lle->ln_state);
767 db_printf(" ln_router=%u\n", lle->ln_router);
768 db_printf(" ln_ntick=%ju\n", (uintmax_t)lle->ln_ntick);
769 db_printf(" lle_refcnt=%d\n", lle->lle_refcnt);
770 bcopy(&lle->ll_addr.mac16, octet, sizeof(octet));
771 db_printf(" ll_addr=%02x:%02x:%02x:%02x:%02x:%02x\n",
772 octet[0], octet[1], octet[2], octet[3], octet[4], octet[5]);
773 db_printf(" lle_timer=%p\n", &lle->lle_timer);
775 switch (la->l3_addr.sa_family) {
779 struct sockaddr_in *sin;
780 char l3s[INET_ADDRSTRLEN];
782 sin = (struct sockaddr_in *)&la->l3_addr;
783 inet_ntoa_r(sin->sin_addr, l3s);
784 db_printf(" l3_addr=%s\n", l3s);
791 struct sockaddr_in6 *sin6;
792 char l3s[INET6_ADDRSTRLEN];
794 sin6 = (struct sockaddr_in6 *)&la->l3_addr;
795 ip6_sprintf(l3s, &sin6->sin6_addr);
796 db_printf(" l3_addr=%s\n", l3s);
801 db_printf(" l3_addr=N/A (af=%d)\n", la->l3_addr.sa_family);
806 DB_SHOW_COMMAND(llentry, db_show_llentry)
810 db_printf("usage: show llentry <struct llentry *>\n");
814 llatbl_lle_show((struct llentry_sa *)addr);
818 llatbl_llt_show(struct lltable *llt)
823 db_printf("llt=%p llt_af=%d llt_ifp=%p\n",
824 llt, llt->llt_af, llt->llt_ifp);
826 for (i = 0; i < llt->llt_hsize; i++) {
827 LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
829 llatbl_lle_show((struct llentry_sa *)lle);
836 DB_SHOW_COMMAND(lltable, db_show_lltable)
840 db_printf("usage: show lltable <struct lltable *>\n");
844 llatbl_llt_show((struct lltable *)addr);
847 DB_SHOW_ALL_COMMAND(lltables, db_show_all_lltables)
849 VNET_ITERATOR_DECL(vnet_iter);
852 VNET_FOREACH(vnet_iter) {
853 CURVNET_SET_QUIET(vnet_iter);
855 db_printf("vnet=%p\n", curvnet);
857 SLIST_FOREACH(llt, &V_lltables, llt_link) {
858 db_printf("llt=%p llt_af=%d llt_ifp=%p(%s)\n",
859 llt, llt->llt_af, llt->llt_ifp,
860 (llt->llt_ifp != NULL) ?
861 llt->llt_ifp->if_xname : "?");
862 if (have_addr && addr != 0) /* verbose */
863 llatbl_llt_show(llt);