1/*-
2 * SPDX-License-Identifier: BSD-2-Clause
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#include "opt_ddb.h"
31#include "opt_inet.h"
32#include "opt_inet6.h"
33
34#include <sys/param.h>
35#include <sys/systm.h>
36#include <sys/eventhandler.h>
37#include <sys/malloc.h>
38#include <sys/mbuf.h>
39#include <sys/syslog.h>
40#include <sys/sysctl.h>
41#include <sys/socket.h>
42#include <sys/kernel.h>
43#include <sys/lock.h>
44#include <sys/mutex.h>
45#include <sys/rwlock.h>
46
47#ifdef DDB
48#include <ddb/ddb.h>
49#endif
50
51#include <vm/uma.h>
52
53#include <netinet/in.h>
54#include <net/if_llatbl.h>
55#include <net/if.h>
56#include <net/if_dl.h>
57#include <net/if_var.h>
58#include <net/if_private.h>
59#include <net/route.h>
60#include <net/route/route_ctl.h>
61#include <net/route/route_debug.h>
62#include <net/vnet.h>
63#include <netinet/if_ether.h>
64#include <netinet6/in6_var.h>
65#include <netinet6/nd6.h>
66
67MALLOC_DEFINE(M_LLTABLE, "lltable", "link level address tables");
68
69VNET_DEFINE_STATIC(SLIST_HEAD(, lltable), lltables) =
70    SLIST_HEAD_INITIALIZER(lltables);
71#define	V_lltables	VNET(lltables)
72
73static struct rwlock lltable_list_lock;
74RW_SYSINIT(lltable_list_lock, &lltable_list_lock, "lltable_list_lock");
75#define	LLTABLE_LIST_RLOCK()		rw_rlock(&lltable_list_lock)
76#define	LLTABLE_LIST_RUNLOCK()		rw_runlock(&lltable_list_lock)
77#define	LLTABLE_LIST_WLOCK()		rw_wlock(&lltable_list_lock)
78#define	LLTABLE_LIST_WUNLOCK()		rw_wunlock(&lltable_list_lock)
79#define	LLTABLE_LIST_LOCK_ASSERT()	rw_assert(&lltable_list_lock, RA_LOCKED)
80
81static void lltable_unlink(struct lltable *llt);
82static void llentries_unlink(struct lltable *llt, struct llentries *head);
83
84/*
85 * Dump lle state for a specific address family.
86 */
87static int
88lltable_dump_af(struct lltable *llt, struct sysctl_req *wr)
89{
90	struct epoch_tracker et;
91	int error;
92
93	LLTABLE_LIST_LOCK_ASSERT();
94
95	if (llt->llt_ifp->if_flags & IFF_LOOPBACK)
96		return (0);
97	error = 0;
98
99	NET_EPOCH_ENTER(et);
100	error = lltable_foreach_lle(llt,
101	    (llt_foreach_cb_t *)llt->llt_dump_entry, wr);
102	NET_EPOCH_EXIT(et);
103
104	return (error);
105}
106
107/*
108 * Dump arp state for a specific address family.
109 */
110int
111lltable_sysctl_dumparp(int af, struct sysctl_req *wr)
112{
113	struct lltable *llt;
114	int error = 0;
115
116	LLTABLE_LIST_RLOCK();
117	SLIST_FOREACH(llt, &V_lltables, llt_link) {
118		if (llt->llt_af == af) {
119			error = lltable_dump_af(llt, wr);
120			if (error != 0)
121				goto done;
122		}
123	}
124done:
125	LLTABLE_LIST_RUNLOCK();
126	return (error);
127}
128
129/*
130 * Adds a mbuf to hold queue. Drops old packets if the queue is full.
131 *
132 * Returns the number of held packets that were dropped.
133 */
134size_t
135lltable_append_entry_queue(struct llentry *lle, struct mbuf *m,
136    size_t maxheld)
137{
138	size_t pkts_dropped = 0;
139
140	LLE_WLOCK_ASSERT(lle);
141
142	while (lle->la_numheld >= maxheld && lle->la_hold != NULL) {
143		struct mbuf *next = lle->la_hold->m_nextpkt;
144		m_freem(lle->la_hold);
145		lle->la_hold = next;
146		lle->la_numheld--;
147		pkts_dropped++;
148	}
149
150	if (lle->la_hold != NULL) {
151		struct mbuf *curr = lle->la_hold;
152		while (curr->m_nextpkt != NULL)
153			curr = curr->m_nextpkt;
154		curr->m_nextpkt = m;
155	} else
156		lle->la_hold = m;
157
158	lle->la_numheld++;
159
160	return pkts_dropped;
161}
162
163
164/*
165 * Common function helpers for chained hash table.
166 */
167
168/*
169 * Runs specified callback for each entry in @llt.
170 * Caller does the locking.
171 *
172 */
173static int
174htable_foreach_lle(struct lltable *llt, llt_foreach_cb_t *f, void *farg)
175{
176	struct llentry *lle, *next;
177	int i, error;
178
179	error = 0;
180
181	for (i = 0; i < llt->llt_hsize; i++) {
182		CK_LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
183			error = f(llt, lle, farg);
184			if (error != 0)
185				break;
186		}
187	}
188
189	return (error);
190}
191
192/*
193 * The htable_[un]link_entry() functions return:
194 * 0 if the entry was (un)linked already and nothing changed,
195 * 1 if the entry was added/removed to/from the table, and
196 * -1 on error (e.g., not being able to add the entry due to limits reached).
197 * While the "unlink" operation should never error, callers of
198 * lltable_link_entry() need to check for errors and handle them.
199 */
200static int
201htable_link_entry(struct lltable *llt, struct llentry *lle)
202{
203	struct llentries *lleh;
204	uint32_t hashidx;
205
206	if ((lle->la_flags & LLE_LINKED) != 0)
207		return (0);
208
209	IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
210
211	if (llt->llt_maxentries > 0 &&
212	    llt->llt_entries >= llt->llt_maxentries)
213		return (-1);
214
215	hashidx = llt->llt_hash(lle, llt->llt_hsize);
216	lleh = &llt->lle_head[hashidx];
217
218	lle->lle_tbl  = llt;
219	lle->lle_head = lleh;
220	lle->la_flags |= LLE_LINKED;
221	CK_LIST_INSERT_HEAD(lleh, lle, lle_next);
222	llt->llt_entries++;
223
224	return (1);
225}
226
227static int
228htable_unlink_entry(struct llentry *lle)
229{
230	struct lltable *llt;
231
232	if ((lle->la_flags & LLE_LINKED) == 0)
233		return (0);
234
235	llt = lle->lle_tbl;
236	IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
237	KASSERT(llt->llt_entries > 0, ("%s: lltable %p (%s) entries %d <= 0",
238	    __func__, llt, if_name(llt->llt_ifp), llt->llt_entries));
239
240	CK_LIST_REMOVE(lle, lle_next);
241	lle->la_flags &= ~(LLE_VALID | LLE_LINKED);
242#if 0
243	lle->lle_tbl = NULL;
244	lle->lle_head = NULL;
245#endif
246	llt->llt_entries--;
247
248	return (1);
249}
250
251struct prefix_match_data {
252	const struct sockaddr *addr;
253	const struct sockaddr *mask;
254	struct llentries dchain;
255	u_int flags;
256};
257
258static int
259htable_prefix_free_cb(struct lltable *llt, struct llentry *lle, void *farg)
260{
261	struct prefix_match_data *pmd;
262
263	pmd = (struct prefix_match_data *)farg;
264
265	if (llt->llt_match_prefix(pmd->addr, pmd->mask, pmd->flags, lle)) {
266		LLE_WLOCK(lle);
267		CK_LIST_INSERT_HEAD(&pmd->dchain, lle, lle_chain);
268	}
269
270	return (0);
271}
272
273static void
274htable_prefix_free(struct lltable *llt, const struct sockaddr *addr,
275    const struct sockaddr *mask, u_int flags)
276{
277	struct llentry *lle, *next;
278	struct prefix_match_data pmd;
279
280	bzero(&pmd, sizeof(pmd));
281	pmd.addr = addr;
282	pmd.mask = mask;
283	pmd.flags = flags;
284	CK_LIST_INIT(&pmd.dchain);
285
286	IF_AFDATA_WLOCK(llt->llt_ifp);
287	/* Push matching lles to chain */
288	lltable_foreach_lle(llt, htable_prefix_free_cb, &pmd);
289
290	llentries_unlink(llt, &pmd.dchain);
291	IF_AFDATA_WUNLOCK(llt->llt_ifp);
292
293	CK_LIST_FOREACH_SAFE(lle, &pmd.dchain, lle_chain, next)
294		lltable_free_entry(llt, lle);
295}
296
297static void
298htable_free_tbl(struct lltable *llt)
299{
300
301	free(llt->lle_head, M_LLTABLE);
302	free(llt, M_LLTABLE);
303}
304
305static void
306llentries_unlink(struct lltable *llt, struct llentries *head)
307{
308	struct llentry *lle, *next;
309
310	CK_LIST_FOREACH_SAFE(lle, head, lle_chain, next)
311		llt->llt_unlink_entry(lle);
312}
313
314/*
315 * Helper function used to drop all mbufs in hold queue.
316 *
317 * Returns the number of held packets, if any, that were dropped.
318 */
319size_t
320lltable_drop_entry_queue(struct llentry *lle)
321{
322	size_t pkts_dropped = 0;
323
324	LLE_WLOCK_ASSERT(lle);
325
326	while (lle->la_hold != NULL) {
327		struct mbuf *next = lle->la_hold->m_nextpkt;
328		m_freem(lle->la_hold);
329		lle->la_hold = next;
330		lle->la_numheld--;
331		pkts_dropped++;
332	}
333
334	KASSERT(lle->la_numheld == 0,
335		("%s: la_numheld %d > 0, pkts_dropped %zd", __func__,
336		 lle->la_numheld, pkts_dropped));
337
338	return (pkts_dropped);
339}
340
341void
342lltable_set_entry_addr(struct ifnet *ifp, struct llentry *lle,
343    const char *linkhdr, size_t linkhdrsize, int lladdr_off)
344{
345
346	memcpy(lle->r_linkdata, linkhdr, linkhdrsize);
347	lle->r_hdrlen = linkhdrsize;
348	lle->ll_addr = &lle->r_linkdata[lladdr_off];
349	lle->la_flags |= LLE_VALID;
350	lle->r_flags |= RLLE_VALID;
351}
352
353/*
354 * Acquires lltable write lock.
355 *
356 * Returns true on success, with both lltable and lle lock held.
357 * On failure, false is returned and lle wlock is still held.
358 */
359bool
360lltable_acquire_wlock(struct ifnet *ifp, struct llentry *lle)
361{
362	NET_EPOCH_ASSERT();
363
364	/* Perform real LLE update */
365	/* use afdata WLOCK to update fields */
366	LLE_WUNLOCK(lle);
367	IF_AFDATA_WLOCK(ifp);
368	LLE_WLOCK(lle);
369
370	/*
371	 * Since we droppped LLE lock, other thread might have deleted
372	 * this lle. Check and return
373	 */
374	if ((lle->la_flags & LLE_DELETED) != 0) {
375		IF_AFDATA_WUNLOCK(ifp);
376		return (false);
377	}
378
379	return (true);
380}
381
382/*
383 * Tries to update @lle link-level address.
384 * Since update requires AFDATA WLOCK, function
385 * drops @lle lock, acquires AFDATA lock and then acquires
386 * @lle lock to maintain lock order.
387 *
388 * Returns 1 on success.
389 */
390int
391lltable_try_set_entry_addr(struct ifnet *ifp, struct llentry *lle,
392    const char *linkhdr, size_t linkhdrsize, int lladdr_off)
393{
394
395	if (!lltable_acquire_wlock(ifp, lle))
396		return (0);
397
398	/* Update data */
399	lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize, lladdr_off);
400
401	IF_AFDATA_WUNLOCK(ifp);
402
403	return (1);
404}
405
406 /*
407 * Helper function used to pre-compute full/partial link-layer
408 * header data suitable for feeding into if_output().
409 */
410int
411lltable_calc_llheader(struct ifnet *ifp, int family, char *lladdr,
412    char *buf, size_t *bufsize, int *lladdr_off)
413{
414	struct if_encap_req ereq;
415	int error;
416
417	bzero(buf, *bufsize);
418	bzero(&ereq, sizeof(ereq));
419	ereq.buf = buf;
420	ereq.bufsize = *bufsize;
421	ereq.rtype = IFENCAP_LL;
422	ereq.family = family;
423	ereq.lladdr = lladdr;
424	ereq.lladdr_len = ifp->if_addrlen;
425	error = ifp->if_requestencap(ifp, &ereq);
426	if (error == 0) {
427		*bufsize = ereq.bufsize;
428		*lladdr_off = ereq.lladdr_off;
429	}
430
431	return (error);
432}
433
434/*
435 * Searches for the child entry matching @family inside @lle.
436 * Returns the entry or NULL.
437 */
438struct llentry *
439llentry_lookup_family(struct llentry *lle, int family)
440{
441	struct llentry *child_lle;
442
443	if (lle == NULL)
444		return (NULL);
445
446	CK_SLIST_FOREACH(child_lle, &lle->lle_children, lle_child_next) {
447		if (child_lle->r_family == family)
448			return (child_lle);
449	}
450
451	return (NULL);
452}
453
454/*
455 * Retrieves upper protocol family for the llentry.
456 * By default, all "normal" (e.g. upper_family == transport_family)
457 * llentries have r_family set to 0.
458 * Thus, use @default_family in that regard, otherwise use r_family.
459 *
460 * Returns upper protocol family
461 */
462int
463llentry_get_upper_family(const struct llentry *lle, int default_family)
464{
465	return (lle->r_family == 0 ? default_family : lle->r_family);
466}
467
468/*
469 * Prints llentry @lle data into provided buffer.
470 * Example: lle/inet/valid/em0/1.2.3.4
471 *
472 * Returns @buf.
473 */
474char *
475llentry_print_buf(const struct llentry *lle, struct ifnet *ifp, int family,
476    char *buf, size_t bufsize)
477{
478#if defined(INET) || defined(INET6)
479	char abuf[INET6_ADDRSTRLEN];
480#endif
481
482	const char *valid = (lle->r_flags & RLLE_VALID) ? "valid" : "no_l2";
483	const char *upper_str = rib_print_family(llentry_get_upper_family(lle, family));
484
485	switch (family) {
486#ifdef INET
487	case AF_INET:
488		inet_ntop(AF_INET, &lle->r_l3addr.addr4, abuf, sizeof(abuf));
489		snprintf(buf, bufsize, "lle/%s/%s/%s/%s", upper_str,
490		    valid, if_name(ifp), abuf);
491		break;
492#endif
493#ifdef INET6
494	case AF_INET6:
495		inet_ntop(AF_INET6, &lle->r_l3addr.addr6, abuf, sizeof(abuf));
496		snprintf(buf, bufsize, "lle/%s/%s/%s/%s", upper_str,
497		    valid, if_name(ifp), abuf);
498		break;
499#endif
500	default:
501		snprintf(buf, bufsize, "lle/%s/%s/%s/????", upper_str,
502		    valid, if_name(ifp));
503		break;
504	}
505
506	return (buf);
507}
508
509char *
510llentry_print_buf_lltable(const struct llentry *lle, char *buf, size_t bufsize)
511{
512	struct lltable *tbl = lle->lle_tbl;
513
514	return (llentry_print_buf(lle, lltable_get_ifp(tbl), lltable_get_af(tbl), buf, bufsize));
515}
516
517/*
518 * Requests feedback from the datapath.
519 * First packet using @lle should result in
520 * setting r_skip_req back to 0 and updating
521 * lle_hittime to the current time_uptime.
522 */
523void
524llentry_request_feedback(struct llentry *lle)
525{
526	struct llentry *child_lle;
527
528	LLE_REQ_LOCK(lle);
529	lle->r_skip_req = 1;
530	LLE_REQ_UNLOCK(lle);
531
532	CK_SLIST_FOREACH(child_lle, &lle->lle_children, lle_child_next) {
533		LLE_REQ_LOCK(child_lle);
534		child_lle->r_skip_req = 1;
535		LLE_REQ_UNLOCK(child_lle);
536	}
537}
538
539/*
540 * Updates the lle state to mark it has been used
541 * and record the time.
542 * Used by the llentry_provide_feedback() wrapper.
543 */
544void
545llentry_mark_used(struct llentry *lle)
546{
547	LLE_REQ_LOCK(lle);
548	lle->r_skip_req = 0;
549	lle->lle_hittime = time_uptime;
550	LLE_REQ_UNLOCK(lle);
551}
552
553/*
554 * Fetches the time when lle was used.
555 * Return 0 if the entry was not used, relevant time_uptime
556 *  otherwise.
557 */
558static time_t
559llentry_get_hittime_raw(struct llentry *lle)
560{
561	time_t lle_hittime = 0;
562
563	LLE_REQ_LOCK(lle);
564	if ((lle->r_skip_req == 0) && (lle_hittime < lle->lle_hittime))
565		lle_hittime = lle->lle_hittime;
566	LLE_REQ_UNLOCK(lle);
567
568	return (lle_hittime);
569}
570
571time_t
572llentry_get_hittime(struct llentry *lle)
573{
574	time_t lle_hittime = 0;
575	struct llentry *child_lle;
576
577	lle_hittime = llentry_get_hittime_raw(lle);
578
579	CK_SLIST_FOREACH(child_lle, &lle->lle_children, lle_child_next) {
580		time_t hittime = llentry_get_hittime_raw(child_lle);
581		if (hittime > lle_hittime)
582			lle_hittime = hittime;
583	}
584
585	return (lle_hittime);
586}
587
588/*
589 * Update link-layer header for given @lle after
590 * interface lladdr was changed.
591 */
592static int
593llentry_update_ifaddr(struct lltable *llt, struct llentry *lle, void *farg)
594{
595	struct ifnet *ifp;
596	u_char linkhdr[LLE_MAX_LINKHDR];
597	size_t linkhdrsize;
598	u_char *lladdr;
599	int lladdr_off;
600
601	ifp = (struct ifnet *)farg;
602
603	lladdr = lle->ll_addr;
604
605	LLE_WLOCK(lle);
606	if ((lle->la_flags & LLE_VALID) == 0) {
607		LLE_WUNLOCK(lle);
608		return (0);
609	}
610
611	if ((lle->la_flags & LLE_IFADDR) != 0)
612		lladdr = IF_LLADDR(ifp);
613
614	linkhdrsize = sizeof(linkhdr);
615	lltable_calc_llheader(ifp, llt->llt_af, lladdr, linkhdr, &linkhdrsize,
616	    &lladdr_off);
617	memcpy(lle->r_linkdata, linkhdr, linkhdrsize);
618	LLE_WUNLOCK(lle);
619
620	return (0);
621}
622
623/*
624 * Update all calculated headers for given @llt
625 */
626void
627lltable_update_ifaddr(struct lltable *llt)
628{
629
630	if (llt->llt_ifp->if_flags & IFF_LOOPBACK)
631		return;
632
633	IF_AFDATA_WLOCK(llt->llt_ifp);
634	lltable_foreach_lle(llt, llentry_update_ifaddr, llt->llt_ifp);
635	IF_AFDATA_WUNLOCK(llt->llt_ifp);
636}
637
638/*
639 *
640 * Performs generic cleanup routines and frees lle.
641 *
642 * Called for non-linked entries, with callouts and
643 * other AF-specific cleanups performed.
644 *
645 * @lle must be passed WLOCK'ed
646 *
647 * Returns the number of held packets, if any, that were dropped.
648 */
649size_t
650llentry_free(struct llentry *lle)
651{
652	size_t pkts_dropped;
653
654	LLE_WLOCK_ASSERT(lle);
655
656	KASSERT((lle->la_flags & LLE_LINKED) == 0, ("freeing linked lle"));
657
658	pkts_dropped = lltable_drop_entry_queue(lle);
659
660	/* cancel timer */
661	if (callout_stop(&lle->lle_timer) > 0)
662		LLE_REMREF(lle);
663	LLE_FREE_LOCKED(lle);
664
665	return (pkts_dropped);
666}
667
668/*
669 * Free all entries from given table and free itself.
670 */
671
672static int
673lltable_free_cb(struct lltable *llt, struct llentry *lle, void *farg)
674{
675	struct llentries *dchain;
676
677	dchain = (struct llentries *)farg;
678
679	LLE_WLOCK(lle);
680	CK_LIST_INSERT_HEAD(dchain, lle, lle_chain);
681
682	return (0);
683}
684
685/*
686 * Free all entries from given table and free itself.
687 */
688void
689lltable_free(struct lltable *llt)
690{
691	struct llentry *lle, *next;
692	struct llentries dchain;
693
694	KASSERT(llt != NULL, ("%s: llt is NULL", __func__));
695
696	lltable_unlink(llt);
697
698	CK_LIST_INIT(&dchain);
699	IF_AFDATA_WLOCK(llt->llt_ifp);
700	/* Push all lles to @dchain */
701	lltable_foreach_lle(llt, lltable_free_cb, &dchain);
702	llentries_unlink(llt, &dchain);
703	IF_AFDATA_WUNLOCK(llt->llt_ifp);
704
705	CK_LIST_FOREACH_SAFE(lle, &dchain, lle_chain, next) {
706		llentry_free(lle);
707	}
708
709	KASSERT(llt->llt_entries == 0, ("%s: lltable %p (%s) entries not 0: %d",
710	    __func__, llt, llt->llt_ifp->if_xname, llt->llt_entries));
711
712	llt->llt_free_tbl(llt);
713}
714
715/*
716 * Deletes an address from given lltable.
717 * Used for userland interaction to remove
718 * individual entries. Skips entries added by OS.
719 */
720int
721lltable_delete_addr(struct lltable *llt, u_int flags,
722    const struct sockaddr *l3addr)
723{
724	struct llentry *lle;
725	struct ifnet *ifp;
726
727	ifp = llt->llt_ifp;
728	IF_AFDATA_WLOCK(ifp);
729	lle = lla_lookup(llt, LLE_SF(l3addr->sa_family, LLE_EXCLUSIVE), l3addr);
730
731	if (lle == NULL) {
732		IF_AFDATA_WUNLOCK(ifp);
733		return (ENOENT);
734	}
735	if ((lle->la_flags & LLE_IFADDR) != 0 && (flags & LLE_IFADDR) == 0) {
736		IF_AFDATA_WUNLOCK(ifp);
737		LLE_WUNLOCK(lle);
738		return (EPERM);
739	}
740
741	lltable_unlink_entry(llt, lle);
742	IF_AFDATA_WUNLOCK(ifp);
743
744	llt->llt_delete_entry(llt, lle);
745
746	return (0);
747}
748
749void
750lltable_prefix_free(int af, struct sockaddr *addr, struct sockaddr *mask,
751    u_int flags)
752{
753	struct lltable *llt;
754
755	LLTABLE_LIST_RLOCK();
756	SLIST_FOREACH(llt, &V_lltables, llt_link) {
757		if (llt->llt_af != af)
758			continue;
759
760		llt->llt_prefix_free(llt, addr, mask, flags);
761	}
762	LLTABLE_LIST_RUNLOCK();
763}
764
765/*
766 * Delete llentries that func() returns true.
767 */
768struct lle_match_data {
769	struct llentries dchain;
770	llt_match_cb_t *func;
771	void *farg;
772};
773
774static int
775lltable_delete_conditional_cb(struct lltable *llt, struct llentry *lle,
776    void *farg)
777{
778	struct lle_match_data *lmd;
779
780	lmd = (struct lle_match_data *)farg;
781	if (lmd->func(llt, lle, lmd->farg)) {
782		LLE_WLOCK(lle);
783		CK_LIST_INSERT_HEAD(&lmd->dchain, lle, lle_chain);
784	}
785
786	return (0);
787}
788
789void
790lltable_delete_conditional(struct lltable *llt, llt_match_cb_t *func,
791    void *farg)
792{
793	struct llentry *lle, *next;
794	struct lle_match_data lmd;
795
796	bzero(&lmd, sizeof(lmd));
797	CK_LIST_INIT(&lmd.dchain);
798	lmd.func = func;
799	lmd.farg = farg;
800
801	IF_AFDATA_WLOCK(llt->llt_ifp);
802	lltable_foreach_lle(llt, lltable_delete_conditional_cb, &lmd);
803	llentries_unlink(llt, &lmd.dchain);
804	IF_AFDATA_WUNLOCK(llt->llt_ifp);
805
806	CK_LIST_FOREACH_SAFE(lle, &lmd.dchain, lle_chain, next)
807		llt->llt_delete_entry(llt, lle);
808}
809
810struct lltable *
811lltable_allocate_htbl(uint32_t hsize)
812{
813	struct lltable *llt;
814	int i;
815
816	llt = malloc(sizeof(struct lltable), M_LLTABLE, M_WAITOK | M_ZERO);
817	llt->llt_hsize = hsize;
818	llt->lle_head = malloc(sizeof(struct llentries) * hsize,
819	    M_LLTABLE, M_WAITOK | M_ZERO);
820
821	for (i = 0; i < llt->llt_hsize; i++)
822		CK_LIST_INIT(&llt->lle_head[i]);
823
824	/* Set some default callbacks */
825	llt->llt_link_entry = htable_link_entry;
826	llt->llt_unlink_entry = htable_unlink_entry;
827	llt->llt_prefix_free = htable_prefix_free;
828	llt->llt_foreach_entry = htable_foreach_lle;
829	llt->llt_free_tbl = htable_free_tbl;
830
831	return (llt);
832}
833
834/*
835 * Links lltable to global llt list.
836 */
837void
838lltable_link(struct lltable *llt)
839{
840
841	LLTABLE_LIST_WLOCK();
842	SLIST_INSERT_HEAD(&V_lltables, llt, llt_link);
843	LLTABLE_LIST_WUNLOCK();
844}
845
846static void
847lltable_unlink(struct lltable *llt)
848{
849
850	LLTABLE_LIST_WLOCK();
851	SLIST_REMOVE(&V_lltables, llt, lltable, llt_link);
852	LLTABLE_LIST_WUNLOCK();
853
854}
855
856/*
857 * Gets interface @ifp lltable for the specified @family
858 */
859struct lltable *
860lltable_get(struct ifnet *ifp, int family)
861{
862	switch (family) {
863#ifdef INET
864	case AF_INET:
865		return (in_lltable_get(ifp));
866#endif
867#ifdef INET6
868	case AF_INET6:
869		return (in6_lltable_get(ifp));
870#endif
871	}
872
873	return (NULL);
874}
875
876/*
877 * External methods used by lltable consumers
878 */
879
880int
881lltable_foreach_lle(struct lltable *llt, llt_foreach_cb_t *f, void *farg)
882{
883
884	return (llt->llt_foreach_entry(llt, f, farg));
885}
886
887struct llentry *
888lltable_alloc_entry(struct lltable *llt, u_int flags,
889    const struct sockaddr *l3addr)
890{
891
892	return (llt->llt_alloc_entry(llt, flags, l3addr));
893}
894
895void
896lltable_free_entry(struct lltable *llt, struct llentry *lle)
897{
898
899	llt->llt_free_entry(llt, lle);
900}
901
902int
903lltable_link_entry(struct lltable *llt, struct llentry *lle)
904{
905	int error = llt->llt_link_entry(llt, lle);
906
907	if (error == 0 && (lle->la_flags & LLE_PUB) != 0)
908		llt->llt_flags |= LLT_ADDEDPROXY;
909
910	return (error);
911}
912
913void
914lltable_link_child_entry(struct llentry *lle, struct llentry *child_lle)
915{
916	child_lle->lle_parent = lle;
917	child_lle->lle_tbl = lle->lle_tbl;
918	child_lle->la_flags |= LLE_LINKED;
919	CK_SLIST_INSERT_HEAD(&lle->lle_children, child_lle, lle_child_next);
920}
921
922void
923lltable_unlink_child_entry(struct llentry *child_lle)
924{
925	struct llentry *lle = child_lle->lle_parent;
926
927	child_lle->la_flags &= ~LLE_LINKED;
928	child_lle->lle_parent = NULL;
929	CK_SLIST_REMOVE(&lle->lle_children, child_lle, llentry, lle_child_next);
930}
931
932int
933lltable_unlink_entry(struct lltable *llt, struct llentry *lle)
934{
935
936	return (llt->llt_unlink_entry(lle));
937}
938
939void
940lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
941{
942	struct lltable *llt;
943
944	llt = lle->lle_tbl;
945	llt->llt_fill_sa_entry(lle, sa);
946}
947
948struct ifnet *
949lltable_get_ifp(const struct lltable *llt)
950{
951
952	return (llt->llt_ifp);
953}
954
955int
956lltable_get_af(const struct lltable *llt)
957{
958
959	return (llt->llt_af);
960}
961
962/*
963 * Called in route_output when rtm_flags contains RTF_LLDATA.
964 */
965int
966lla_rt_output(struct rt_msghdr *rtm, struct rt_addrinfo *info)
967{
968	struct sockaddr_dl *dl =
969	    (struct sockaddr_dl *)info->rti_info[RTAX_GATEWAY];
970	struct sockaddr *dst = (struct sockaddr *)info->rti_info[RTAX_DST];
971	struct ifnet *ifp;
972	struct lltable *llt;
973	struct llentry *lle, *lle_tmp;
974	uint8_t linkhdr[LLE_MAX_LINKHDR];
975	size_t linkhdrsize;
976	int lladdr_off;
977	u_int laflags = 0;
978	int error;
979
980	if (dl == NULL || dl->sdl_family != AF_LINK)
981		return (EINVAL);
982
983	/* XXX: should be ntohs() */
984	ifp = ifnet_byindex(dl->sdl_index);
985	if (ifp == NULL) {
986		log(LOG_INFO, "%s: invalid ifp (sdl_index %d)\n",
987		    __func__, dl->sdl_index);
988		return EINVAL;
989	}
990
991	llt = lltable_get(ifp, dst->sa_family);
992
993	if (llt == NULL)
994		return (ESRCH);
995
996	error = 0;
997
998	switch (rtm->rtm_type) {
999	case RTM_ADD:
1000		/* Add static LLE */
1001		laflags = 0;
1002		if (rtm->rtm_rmx.rmx_expire == 0)
1003			laflags = LLE_STATIC;
1004		lle = lltable_alloc_entry(llt, laflags, dst);
1005		if (lle == NULL)
1006			return (ENOMEM);
1007
1008		linkhdrsize = sizeof(linkhdr);
1009		if (lltable_calc_llheader(ifp, dst->sa_family, LLADDR(dl),
1010		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
1011			lltable_free_entry(llt, lle);
1012			return (EINVAL);
1013		}
1014		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
1015		    lladdr_off);
1016		if ((rtm->rtm_flags & RTF_ANNOUNCE))
1017			lle->la_flags |= LLE_PUB;
1018		lle->la_expire = rtm->rtm_rmx.rmx_expire;
1019
1020		laflags = lle->la_flags;
1021
1022		/* Try to link new entry */
1023		lle_tmp = NULL;
1024		IF_AFDATA_WLOCK(ifp);
1025		LLE_WLOCK(lle);
1026		lle_tmp = lla_lookup(llt, LLE_EXCLUSIVE, dst);
1027		if (lle_tmp != NULL) {
1028			/* Check if we are trying to replace immutable entry */
1029			if ((lle_tmp->la_flags & LLE_IFADDR) != 0) {
1030				IF_AFDATA_WUNLOCK(ifp);
1031				LLE_WUNLOCK(lle_tmp);
1032				lltable_free_entry(llt, lle);
1033				return (EPERM);
1034			}
1035			/* Unlink existing entry from table */
1036			lltable_unlink_entry(llt, lle_tmp);
1037		}
1038		lltable_link_entry(llt, lle);
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
1068static void
1069llatbl_lle_show(struct llentry *lle)
1070{
1071	uint8_t octet[6];
1072	sa_family_t af = AF_UNSPEC;
1073	char l3_addr_fmt[] = " l3_addr=%s (af=%d)\n";
1074
1075	db_printf("lle=%p\n", lle);
1076	db_printf(" lle_next=%p\n", lle->lle_next.cle_next);
1077	db_printf(" lle_lock=%p\n", &lle->lle_lock);
1078	db_printf(" lle_tbl=%p\n", lle->lle_tbl);
1079	db_printf(" lle_head=%p\n", lle->lle_head);
1080	db_printf(" la_hold=%p\n", lle->la_hold);
1081	db_printf(" la_numheld=%d\n", lle->la_numheld);
1082	db_printf(" la_expire=%ju\n", (uintmax_t)lle->la_expire);
1083	db_printf(" la_flags=0x%04x\n", lle->la_flags);
1084	db_printf(" la_asked=%u\n", lle->la_asked);
1085	db_printf(" la_preempt=%u\n", lle->la_preempt);
1086	db_printf(" ln_state=%d\n", lle->ln_state);
1087	db_printf(" ln_router=%u\n", lle->ln_router);
1088	db_printf(" ln_ntick=%ju\n", (uintmax_t)lle->ln_ntick);
1089	db_printf(" lle_refcnt=%d\n", lle->lle_refcnt);
1090	bcopy(lle->ll_addr, octet, sizeof(octet));
1091	db_printf(" ll_addr=%02x:%02x:%02x:%02x:%02x:%02x\n",
1092	    octet[0], octet[1], octet[2], octet[3], octet[4], octet[5]);
1093	db_printf(" lle_timer=%p\n", &lle->lle_timer);
1094
1095	if (lle->lle_tbl) {
1096		af = lle->lle_tbl->llt_af;
1097	}
1098
1099	switch (af) {
1100#ifdef INET
1101	case AF_INET:
1102	{
1103		struct sockaddr_in sin;
1104		char l3s[INET_ADDRSTRLEN];
1105
1106		lltable_fill_sa_entry(lle, (struct sockaddr *)&sin);
1107		(void) inet_ntop(af, &sin.sin_addr, l3s, sizeof(l3s));
1108		db_printf(l3_addr_fmt, l3s, af);
1109		break;
1110	}
1111#endif
1112#ifdef INET6
1113	case AF_INET6:
1114	{
1115		struct sockaddr_in6 sin6;
1116		char l3s[INET6_ADDRSTRLEN];
1117
1118		lltable_fill_sa_entry(lle, (struct sockaddr *)&sin6);
1119		(void) inet_ntop(af, &sin6.sin6_addr, l3s, sizeof(l3s));
1120		db_printf(l3_addr_fmt, l3s, af);
1121		break;
1122	}
1123#endif
1124	default:
1125		db_printf(l3_addr_fmt, "N/A", af);
1126		break;
1127	}
1128}
1129
1130DB_SHOW_COMMAND(llentry, db_show_llentry)
1131{
1132
1133	if (!have_addr) {
1134		db_printf("usage: show llentry <struct llentry *>\n");
1135		return;
1136	}
1137
1138	llatbl_lle_show((struct llentry *)addr);
1139}
1140
1141static void
1142llatbl_llt_show(struct lltable *llt)
1143{
1144	int i;
1145	struct llentry *lle;
1146
1147	db_printf("llt=%p llt_af=%d llt_ifp=%p\n",
1148	    llt, llt->llt_af, llt->llt_ifp);
1149
1150	for (i = 0; i < llt->llt_hsize; i++) {
1151		CK_LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
1152			llatbl_lle_show(lle);
1153			if (db_pager_quit)
1154				return;
1155		}
1156	}
1157}
1158
1159DB_SHOW_COMMAND(lltable, db_show_lltable)
1160{
1161
1162	if (!have_addr) {
1163		db_printf("usage: show lltable <struct lltable *>\n");
1164		return;
1165	}
1166
1167	llatbl_llt_show((struct lltable *)addr);
1168}
1169
1170DB_SHOW_ALL_COMMAND(lltables, db_show_all_lltables)
1171{
1172	VNET_ITERATOR_DECL(vnet_iter);
1173	struct lltable *llt;
1174
1175	VNET_FOREACH(vnet_iter) {
1176		CURVNET_SET_QUIET(vnet_iter);
1177#ifdef VIMAGE
1178		db_printf("vnet=%p\n", curvnet);
1179#endif
1180		SLIST_FOREACH(llt, &V_lltables, llt_link) {
1181			db_printf("llt=%p llt_af=%d llt_ifp=%p(%s)\n",
1182			    llt, llt->llt_af, llt->llt_ifp,
1183			    (llt->llt_ifp != NULL) ?
1184				llt->llt_ifp->if_xname : "?");
1185			if (have_addr && addr != 0) /* verbose */
1186				llatbl_llt_show(llt);
1187			if (db_pager_quit) {
1188				CURVNET_RESTORE();
1189				return;
1190			}
1191		}
1192		CURVNET_RESTORE();
1193	}
1194}
1195#endif
1196