2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
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.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
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.
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
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
34 #include "opt_inet6.h"
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/eventhandler.h>
39 #include <sys/malloc.h>
41 #include <sys/syslog.h>
42 #include <sys/sysctl.h>
43 #include <sys/socket.h>
44 #include <sys/kernel.h>
46 #include <sys/mutex.h>
47 #include <sys/rwlock.h>
55 #include <netinet/in.h>
56 #include <net/if_llatbl.h>
58 #include <net/if_dl.h>
59 #include <net/if_var.h>
60 #include <net/route.h>
62 #include <netinet/if_ether.h>
63 #include <netinet6/in6_var.h>
64 #include <netinet6/nd6.h>
66 MALLOC_DEFINE(M_LLTABLE, "lltable", "link level address tables");
68 VNET_DEFINE_STATIC(SLIST_HEAD(, lltable), lltables) =
69 SLIST_HEAD_INITIALIZER(lltables);
70 #define V_lltables VNET(lltables)
72 static struct rwlock lltable_list_lock;
73 RW_SYSINIT(lltable_list_lock, &lltable_list_lock, "lltable_list_lock");
74 #define LLTABLE_LIST_RLOCK() rw_rlock(&lltable_list_lock)
75 #define LLTABLE_LIST_RUNLOCK() rw_runlock(&lltable_list_lock)
76 #define LLTABLE_LIST_WLOCK() rw_wlock(&lltable_list_lock)
77 #define LLTABLE_LIST_WUNLOCK() rw_wunlock(&lltable_list_lock)
78 #define LLTABLE_LIST_LOCK_ASSERT() rw_assert(&lltable_list_lock, RA_LOCKED)
80 static void lltable_unlink(struct lltable *llt);
81 static void llentries_unlink(struct lltable *llt, struct llentries *head);
84 * Dump lle state for a specific address family.
87 lltable_dump_af(struct lltable *llt, struct sysctl_req *wr)
89 struct epoch_tracker et;
92 LLTABLE_LIST_LOCK_ASSERT();
94 if (llt->llt_ifp->if_flags & IFF_LOOPBACK)
99 error = lltable_foreach_lle(llt,
100 (llt_foreach_cb_t *)llt->llt_dump_entry, wr);
107 * Dump arp state for a specific address family.
110 lltable_sysctl_dumparp(int af, struct sysctl_req *wr)
115 LLTABLE_LIST_RLOCK();
116 SLIST_FOREACH(llt, &V_lltables, llt_link) {
117 if (llt->llt_af == af) {
118 error = lltable_dump_af(llt, wr);
124 LLTABLE_LIST_RUNLOCK();
129 * Common function helpers for chained hash table.
133 * Runs specified callback for each entry in @llt.
134 * Caller does the locking.
138 htable_foreach_lle(struct lltable *llt, llt_foreach_cb_t *f, void *farg)
140 struct llentry *lle, *next;
145 for (i = 0; i < llt->llt_hsize; i++) {
146 CK_LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
147 error = f(llt, lle, farg);
157 * The htable_[un]link_entry() functions return:
158 * 0 if the entry was (un)linked already and nothing changed,
159 * 1 if the entry was added/removed to/from the table, and
160 * -1 on error (e.g., not being able to add the entry due to limits reached).
161 * While the "unlink" operation should never error, callers of
162 * lltable_link_entry() need to check for errors and handle them.
165 htable_link_entry(struct lltable *llt, struct llentry *lle)
167 struct llentries *lleh;
170 if ((lle->la_flags & LLE_LINKED) != 0)
173 IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
175 if (llt->llt_maxentries > 0 &&
176 llt->llt_entries >= llt->llt_maxentries)
179 hashidx = llt->llt_hash(lle, llt->llt_hsize);
180 lleh = &llt->lle_head[hashidx];
183 lle->lle_head = lleh;
184 lle->la_flags |= LLE_LINKED;
185 CK_LIST_INSERT_HEAD(lleh, lle, lle_next);
192 htable_unlink_entry(struct llentry *lle)
196 if ((lle->la_flags & LLE_LINKED) == 0)
200 IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
201 KASSERT(llt->llt_entries > 0, ("%s: lltable %p (%s) entries %d <= 0",
202 __func__, llt, if_name(llt->llt_ifp), llt->llt_entries));
204 CK_LIST_REMOVE(lle, lle_next);
205 lle->la_flags &= ~(LLE_VALID | LLE_LINKED);
208 lle->lle_head = NULL;
215 struct prefix_match_data {
216 const struct sockaddr *addr;
217 const struct sockaddr *mask;
218 struct llentries dchain;
223 htable_prefix_free_cb(struct lltable *llt, struct llentry *lle, void *farg)
225 struct prefix_match_data *pmd;
227 pmd = (struct prefix_match_data *)farg;
229 if (llt->llt_match_prefix(pmd->addr, pmd->mask, pmd->flags, lle)) {
231 CK_LIST_INSERT_HEAD(&pmd->dchain, lle, lle_chain);
238 htable_prefix_free(struct lltable *llt, const struct sockaddr *addr,
239 const struct sockaddr *mask, u_int flags)
241 struct llentry *lle, *next;
242 struct prefix_match_data pmd;
244 bzero(&pmd, sizeof(pmd));
248 CK_LIST_INIT(&pmd.dchain);
250 IF_AFDATA_WLOCK(llt->llt_ifp);
251 /* Push matching lles to chain */
252 lltable_foreach_lle(llt, htable_prefix_free_cb, &pmd);
254 llentries_unlink(llt, &pmd.dchain);
255 IF_AFDATA_WUNLOCK(llt->llt_ifp);
257 CK_LIST_FOREACH_SAFE(lle, &pmd.dchain, lle_chain, next)
258 lltable_free_entry(llt, lle);
262 htable_free_tbl(struct lltable *llt)
265 free(llt->lle_head, M_LLTABLE);
266 free(llt, M_LLTABLE);
270 llentries_unlink(struct lltable *llt, struct llentries *head)
272 struct llentry *lle, *next;
274 CK_LIST_FOREACH_SAFE(lle, head, lle_chain, next)
275 llt->llt_unlink_entry(lle);
279 * Helper function used to drop all mbufs in hold queue.
281 * Returns the number of held packets, if any, that were dropped.
284 lltable_drop_entry_queue(struct llentry *lle)
289 LLE_WLOCK_ASSERT(lle);
292 while ((lle->la_numheld > 0) && (lle->la_hold != NULL)) {
293 next = lle->la_hold->m_nextpkt;
294 m_freem(lle->la_hold);
300 KASSERT(lle->la_numheld == 0,
301 ("%s: la_numheld %d > 0, pkts_droped %zd", __func__,
302 lle->la_numheld, pkts_dropped));
304 return (pkts_dropped);
308 lltable_set_entry_addr(struct ifnet *ifp, struct llentry *lle,
309 const char *linkhdr, size_t linkhdrsize, int lladdr_off)
312 memcpy(lle->r_linkdata, linkhdr, linkhdrsize);
313 lle->r_hdrlen = linkhdrsize;
314 lle->ll_addr = &lle->r_linkdata[lladdr_off];
315 lle->la_flags |= LLE_VALID;
316 lle->r_flags |= RLLE_VALID;
320 * Tries to update @lle link-level address.
321 * Since update requires AFDATA WLOCK, function
322 * drops @lle lock, acquires AFDATA lock and then acquires
323 * @lle lock to maintain lock order.
325 * Returns 1 on success.
328 lltable_try_set_entry_addr(struct ifnet *ifp, struct llentry *lle,
329 const char *linkhdr, size_t linkhdrsize, int lladdr_off)
332 /* Perform real LLE update */
333 /* use afdata WLOCK to update fields */
334 LLE_WLOCK_ASSERT(lle);
337 IF_AFDATA_WLOCK(ifp);
341 * Since we droppped LLE lock, other thread might have deleted
342 * this lle. Check and return
344 if ((lle->la_flags & LLE_DELETED) != 0) {
345 IF_AFDATA_WUNLOCK(ifp);
346 LLE_FREE_LOCKED(lle);
351 lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize, lladdr_off);
353 IF_AFDATA_WUNLOCK(ifp);
361 * Helper function used to pre-compute full/partial link-layer
362 * header data suitable for feeding into if_output().
365 lltable_calc_llheader(struct ifnet *ifp, int family, char *lladdr,
366 char *buf, size_t *bufsize, int *lladdr_off)
368 struct if_encap_req ereq;
371 bzero(buf, *bufsize);
372 bzero(&ereq, sizeof(ereq));
374 ereq.bufsize = *bufsize;
375 ereq.rtype = IFENCAP_LL;
376 ereq.family = family;
377 ereq.lladdr = lladdr;
378 ereq.lladdr_len = ifp->if_addrlen;
379 error = ifp->if_requestencap(ifp, &ereq);
381 *bufsize = ereq.bufsize;
382 *lladdr_off = ereq.lladdr_off;
389 * Update link-layer header for given @lle after
390 * interface lladdr was changed.
393 llentry_update_ifaddr(struct lltable *llt, struct llentry *lle, void *farg)
396 u_char linkhdr[LLE_MAX_LINKHDR];
401 ifp = (struct ifnet *)farg;
403 lladdr = lle->ll_addr;
406 if ((lle->la_flags & LLE_VALID) == 0) {
411 if ((lle->la_flags & LLE_IFADDR) != 0)
412 lladdr = IF_LLADDR(ifp);
414 linkhdrsize = sizeof(linkhdr);
415 lltable_calc_llheader(ifp, llt->llt_af, lladdr, linkhdr, &linkhdrsize,
417 memcpy(lle->r_linkdata, linkhdr, linkhdrsize);
424 * Update all calculated headers for given @llt
427 lltable_update_ifaddr(struct lltable *llt)
430 if (llt->llt_ifp->if_flags & IFF_LOOPBACK)
433 IF_AFDATA_WLOCK(llt->llt_ifp);
434 lltable_foreach_lle(llt, llentry_update_ifaddr, llt->llt_ifp);
435 IF_AFDATA_WUNLOCK(llt->llt_ifp);
440 * Performs generic cleanup routines and frees lle.
442 * Called for non-linked entries, with callouts and
443 * other AF-specific cleanups performed.
445 * @lle must be passed WLOCK'ed
447 * Returns the number of held packets, if any, that were dropped.
450 llentry_free(struct llentry *lle)
454 LLE_WLOCK_ASSERT(lle);
456 KASSERT((lle->la_flags & LLE_LINKED) == 0, ("freeing linked lle"));
458 pkts_dropped = lltable_drop_entry_queue(lle);
461 if (callout_stop(&lle->lle_timer) > 0)
463 LLE_FREE_LOCKED(lle);
465 return (pkts_dropped);
469 * Free all entries from given table and free itself.
473 lltable_free_cb(struct lltable *llt, struct llentry *lle, void *farg)
475 struct llentries *dchain;
477 dchain = (struct llentries *)farg;
480 CK_LIST_INSERT_HEAD(dchain, lle, lle_chain);
486 * Free all entries from given table and free itself.
489 lltable_free(struct lltable *llt)
491 struct llentry *lle, *next;
492 struct llentries dchain;
494 KASSERT(llt != NULL, ("%s: llt is NULL", __func__));
498 CK_LIST_INIT(&dchain);
499 IF_AFDATA_WLOCK(llt->llt_ifp);
500 /* Push all lles to @dchain */
501 lltable_foreach_lle(llt, lltable_free_cb, &dchain);
502 llentries_unlink(llt, &dchain);
503 IF_AFDATA_WUNLOCK(llt->llt_ifp);
505 CK_LIST_FOREACH_SAFE(lle, &dchain, lle_chain, next) {
509 KASSERT(llt->llt_entries == 0, ("%s: lltable %p (%s) entires not 0: %d",
510 __func__, llt, llt->llt_ifp->if_xname, llt->llt_entries));
512 llt->llt_free_tbl(llt);
516 * Deletes an address from given lltable.
517 * Used for userland interaction to remove
518 * individual entries. Skips entries added by OS.
521 lltable_delete_addr(struct lltable *llt, u_int flags,
522 const struct sockaddr *l3addr)
528 IF_AFDATA_WLOCK(ifp);
529 lle = lla_lookup(llt, LLE_EXCLUSIVE, l3addr);
532 IF_AFDATA_WUNLOCK(ifp);
535 if ((lle->la_flags & LLE_IFADDR) != 0 && (flags & LLE_IFADDR) == 0) {
536 IF_AFDATA_WUNLOCK(ifp);
541 lltable_unlink_entry(llt, lle);
542 IF_AFDATA_WUNLOCK(ifp);
544 llt->llt_delete_entry(llt, lle);
550 lltable_prefix_free(int af, struct sockaddr *addr, struct sockaddr *mask,
555 LLTABLE_LIST_RLOCK();
556 SLIST_FOREACH(llt, &V_lltables, llt_link) {
557 if (llt->llt_af != af)
560 llt->llt_prefix_free(llt, addr, mask, flags);
562 LLTABLE_LIST_RUNLOCK();
566 lltable_allocate_htbl(uint32_t hsize)
571 llt = malloc(sizeof(struct lltable), M_LLTABLE, M_WAITOK | M_ZERO);
572 llt->llt_hsize = hsize;
573 llt->lle_head = malloc(sizeof(struct llentries) * hsize,
574 M_LLTABLE, M_WAITOK | M_ZERO);
576 for (i = 0; i < llt->llt_hsize; i++)
577 CK_LIST_INIT(&llt->lle_head[i]);
579 /* Set some default callbacks */
580 llt->llt_link_entry = htable_link_entry;
581 llt->llt_unlink_entry = htable_unlink_entry;
582 llt->llt_prefix_free = htable_prefix_free;
583 llt->llt_foreach_entry = htable_foreach_lle;
584 llt->llt_free_tbl = htable_free_tbl;
590 * Links lltable to global llt list.
593 lltable_link(struct lltable *llt)
596 LLTABLE_LIST_WLOCK();
597 SLIST_INSERT_HEAD(&V_lltables, llt, llt_link);
598 LLTABLE_LIST_WUNLOCK();
602 lltable_unlink(struct lltable *llt)
605 LLTABLE_LIST_WLOCK();
606 SLIST_REMOVE(&V_lltables, llt, lltable, llt_link);
607 LLTABLE_LIST_WUNLOCK();
612 * External methods used by lltable consumers
616 lltable_foreach_lle(struct lltable *llt, llt_foreach_cb_t *f, void *farg)
619 return (llt->llt_foreach_entry(llt, f, farg));
623 lltable_alloc_entry(struct lltable *llt, u_int flags,
624 const struct sockaddr *l3addr)
627 return (llt->llt_alloc_entry(llt, flags, l3addr));
631 lltable_free_entry(struct lltable *llt, struct llentry *lle)
634 llt->llt_free_entry(llt, lle);
638 lltable_link_entry(struct lltable *llt, struct llentry *lle)
641 return (llt->llt_link_entry(llt, lle));
645 lltable_unlink_entry(struct lltable *llt, struct llentry *lle)
648 return (llt->llt_unlink_entry(lle));
652 lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
657 llt->llt_fill_sa_entry(lle, sa);
661 lltable_get_ifp(const struct lltable *llt)
664 return (llt->llt_ifp);
668 lltable_get_af(const struct lltable *llt)
671 return (llt->llt_af);
675 * Called in route_output when rtm_flags contains RTF_LLDATA.
678 lla_rt_output(struct rt_msghdr *rtm, struct rt_addrinfo *info)
680 struct sockaddr_dl *dl =
681 (struct sockaddr_dl *)info->rti_info[RTAX_GATEWAY];
682 struct sockaddr *dst = (struct sockaddr *)info->rti_info[RTAX_DST];
685 struct llentry *lle, *lle_tmp;
686 uint8_t linkhdr[LLE_MAX_LINKHDR];
692 if (dl == NULL || dl->sdl_family != AF_LINK)
695 ifp = ifnet_byindex(dl->sdl_index);
697 log(LOG_INFO, "%s: invalid ifp (sdl_index %d)\n",
698 __func__, dl->sdl_index);
702 /* XXX linked list may be too expensive */
703 LLTABLE_LIST_RLOCK();
704 SLIST_FOREACH(llt, &V_lltables, llt_link) {
705 if (llt->llt_af == dst->sa_family &&
709 LLTABLE_LIST_RUNLOCK();
715 switch (rtm->rtm_type) {
719 if (rtm->rtm_rmx.rmx_expire == 0)
720 laflags = LLE_STATIC;
721 lle = lltable_alloc_entry(llt, laflags, dst);
725 linkhdrsize = sizeof(linkhdr);
726 if (lltable_calc_llheader(ifp, dst->sa_family, LLADDR(dl),
727 linkhdr, &linkhdrsize, &lladdr_off) != 0)
729 lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
731 if ((rtm->rtm_flags & RTF_ANNOUNCE))
732 lle->la_flags |= LLE_PUB;
733 lle->la_expire = rtm->rtm_rmx.rmx_expire;
735 laflags = lle->la_flags;
737 /* Try to link new entry */
739 IF_AFDATA_WLOCK(ifp);
741 lle_tmp = lla_lookup(llt, LLE_EXCLUSIVE, dst);
742 if (lle_tmp != NULL) {
743 /* Check if we are trying to replace immutable entry */
744 if ((lle_tmp->la_flags & LLE_IFADDR) != 0) {
745 IF_AFDATA_WUNLOCK(ifp);
746 LLE_WUNLOCK(lle_tmp);
747 lltable_free_entry(llt, lle);
750 /* Unlink existing entry from table */
751 lltable_unlink_entry(llt, lle_tmp);
753 lltable_link_entry(llt, lle);
754 IF_AFDATA_WUNLOCK(ifp);
756 if (lle_tmp != NULL) {
757 EVENTHANDLER_INVOKE(lle_event, lle_tmp,LLENTRY_EXPIRED);
758 lltable_free_entry(llt, lle_tmp);
762 * By invoking LLE handler here we might get
763 * two events on static LLE entry insertion
764 * in routing socket. However, since we might have
765 * other subscribers we need to generate this event.
767 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_RESOLVED);
771 if ((laflags & LLE_PUB) && dst->sa_family == AF_INET)
773 &((struct sockaddr_in *)dst)->sin_addr,
774 &((struct sockaddr_in *)dst)->sin_addr,
775 (u_char *)LLADDR(dl));
781 return (lltable_delete_addr(llt, 0, dst));
793 struct sockaddr l3_addr;
797 llatbl_lle_show(struct llentry_sa *la)
803 db_printf("lle=%p\n", lle);
804 db_printf(" lle_next=%p\n", lle->lle_next.cle_next);
805 db_printf(" lle_lock=%p\n", &lle->lle_lock);
806 db_printf(" lle_tbl=%p\n", lle->lle_tbl);
807 db_printf(" lle_head=%p\n", lle->lle_head);
808 db_printf(" la_hold=%p\n", lle->la_hold);
809 db_printf(" la_numheld=%d\n", lle->la_numheld);
810 db_printf(" la_expire=%ju\n", (uintmax_t)lle->la_expire);
811 db_printf(" la_flags=0x%04x\n", lle->la_flags);
812 db_printf(" la_asked=%u\n", lle->la_asked);
813 db_printf(" la_preempt=%u\n", lle->la_preempt);
814 db_printf(" ln_state=%d\n", lle->ln_state);
815 db_printf(" ln_router=%u\n", lle->ln_router);
816 db_printf(" ln_ntick=%ju\n", (uintmax_t)lle->ln_ntick);
817 db_printf(" lle_refcnt=%d\n", lle->lle_refcnt);
818 bcopy(lle->ll_addr, octet, sizeof(octet));
819 db_printf(" ll_addr=%02x:%02x:%02x:%02x:%02x:%02x\n",
820 octet[0], octet[1], octet[2], octet[3], octet[4], octet[5]);
821 db_printf(" lle_timer=%p\n", &lle->lle_timer);
823 switch (la->l3_addr.sa_family) {
827 struct sockaddr_in *sin;
828 char l3s[INET_ADDRSTRLEN];
830 sin = (struct sockaddr_in *)&la->l3_addr;
831 inet_ntoa_r(sin->sin_addr, l3s);
832 db_printf(" l3_addr=%s\n", l3s);
839 struct sockaddr_in6 *sin6;
840 char l3s[INET6_ADDRSTRLEN];
842 sin6 = (struct sockaddr_in6 *)&la->l3_addr;
843 ip6_sprintf(l3s, &sin6->sin6_addr);
844 db_printf(" l3_addr=%s\n", l3s);
849 db_printf(" l3_addr=N/A (af=%d)\n", la->l3_addr.sa_family);
854 DB_SHOW_COMMAND(llentry, db_show_llentry)
858 db_printf("usage: show llentry <struct llentry *>\n");
862 llatbl_lle_show((struct llentry_sa *)addr);
866 llatbl_llt_show(struct lltable *llt)
871 db_printf("llt=%p llt_af=%d llt_ifp=%p\n",
872 llt, llt->llt_af, llt->llt_ifp);
874 for (i = 0; i < llt->llt_hsize; i++) {
875 CK_LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
877 llatbl_lle_show((struct llentry_sa *)lle);
884 DB_SHOW_COMMAND(lltable, db_show_lltable)
888 db_printf("usage: show lltable <struct lltable *>\n");
892 llatbl_llt_show((struct lltable *)addr);
895 DB_SHOW_ALL_COMMAND(lltables, db_show_all_lltables)
897 VNET_ITERATOR_DECL(vnet_iter);
900 VNET_FOREACH(vnet_iter) {
901 CURVNET_SET_QUIET(vnet_iter);
903 db_printf("vnet=%p\n", curvnet);
905 SLIST_FOREACH(llt, &V_lltables, llt_link) {
906 db_printf("llt=%p llt_af=%d llt_ifp=%p(%s)\n",
907 llt, llt->llt_af, llt->llt_ifp,
908 (llt->llt_ifp != NULL) ?
909 llt->llt_ifp->if_xname : "?");
910 if (have_addr && addr != 0) /* verbose */
911 llatbl_llt_show(llt);