xref: /freebsd/sys/net/ieee8023ad_lacp.c (revision b28624fde638caadd4a89f50c9b7e7da0f98c4d2)
1 /*	$NetBSD: ieee8023ad_lacp.c,v 1.3 2005/12/11 12:24:54 christos Exp $	*/
2 
3 /*-
4  * Copyright (c)2005 YAMAMOTO Takashi,
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  */
28 
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
31 
32 #include <sys/param.h>
33 #include <sys/callout.h>
34 #include <sys/mbuf.h>
35 #include <sys/systm.h>
36 #include <sys/malloc.h>
37 #include <sys/kernel.h> /* hz */
38 #include <sys/socket.h> /* for net/if.h */
39 #include <sys/sockio.h>
40 #include <machine/stdarg.h>
41 #include <sys/lock.h>
42 #include <sys/rwlock.h>
43 #include <sys/taskqueue.h>
44 
45 #include <net/if.h>
46 #include <net/if_dl.h>
47 #include <net/ethernet.h>
48 #include <net/if_media.h>
49 #include <net/if_types.h>
50 
51 #include <net/if_lagg.h>
52 #include <net/ieee8023ad_lacp.h>
53 
54 /*
55  * actor system priority and port priority.
56  * XXX should be configurable.
57  */
58 
59 #define	LACP_SYSTEM_PRIO	0x8000
60 #define	LACP_PORT_PRIO		0x8000
61 
62 const uint8_t ethermulticastaddr_slowprotocols[ETHER_ADDR_LEN] =
63     { 0x01, 0x80, 0xc2, 0x00, 0x00, 0x02 };
64 
65 static const struct tlv_template lacp_info_tlv_template[] = {
66 	{ LACP_TYPE_ACTORINFO,
67 	    sizeof(struct tlvhdr) + sizeof(struct lacp_peerinfo) },
68 	{ LACP_TYPE_PARTNERINFO,
69 	    sizeof(struct tlvhdr) + sizeof(struct lacp_peerinfo) },
70 	{ LACP_TYPE_COLLECTORINFO,
71 	    sizeof(struct tlvhdr) + sizeof(struct lacp_collectorinfo) },
72 	{ 0, 0 },
73 };
74 
75 typedef void (*lacp_timer_func_t)(struct lacp_port *);
76 
77 static const struct tlv_template marker_info_tlv_template[] = {
78 	{ MARKER_TYPE_INFO,
79 	    sizeof(struct tlvhdr) + sizeof(struct lacp_markerinfo) },
80 	{ 0, 0 },
81 };
82 
83 static const struct tlv_template marker_response_tlv_template[] = {
84 	{ MARKER_TYPE_RESPONSE,
85 	    sizeof(struct tlvhdr) + sizeof(struct lacp_markerinfo) },
86 	{ 0, 0 },
87 };
88 
89 static void	lacp_fill_actorinfo(struct lacp_port *, struct lacp_peerinfo *);
90 static void	lacp_fill_markerinfo(struct lacp_port *,
91 		    struct lacp_markerinfo *);
92 
93 static uint64_t	lacp_aggregator_bandwidth(struct lacp_aggregator *);
94 static void	lacp_suppress_distributing(struct lacp_softc *,
95 		    struct lacp_aggregator *);
96 static void	lacp_transit_expire(void *);
97 static void	lacp_select_active_aggregator(struct lacp_softc *);
98 static uint16_t	lacp_compose_key(struct lacp_port *);
99 static int	tlv_check(const void *, size_t, const struct tlvhdr *,
100 		    const struct tlv_template *, boolean_t);
101 static void	lacp_tick(void *);
102 
103 static void	lacp_fill_aggregator_id(struct lacp_aggregator *,
104 		    const struct lacp_port *);
105 static void	lacp_fill_aggregator_id_peer(struct lacp_peerinfo *,
106 		    const struct lacp_peerinfo *);
107 static int	lacp_aggregator_is_compatible(const struct lacp_aggregator *,
108 		    const struct lacp_port *);
109 static int	lacp_peerinfo_is_compatible(const struct lacp_peerinfo *,
110 		    const struct lacp_peerinfo *);
111 
112 static struct lacp_aggregator *lacp_aggregator_get(struct lacp_softc *,
113 		    struct lacp_port *);
114 static void	lacp_aggregator_addref(struct lacp_softc *,
115 		    struct lacp_aggregator *);
116 static void	lacp_aggregator_delref(struct lacp_softc *,
117 		    struct lacp_aggregator *);
118 
119 /* receive machine */
120 
121 static void	lacp_dequeue(void *, int);
122 static int	lacp_pdu_input(struct lagg_port *, struct mbuf *);
123 static int	lacp_marker_input(struct lagg_port *, struct mbuf *);
124 static void	lacp_sm_rx(struct lacp_port *, const struct lacpdu *);
125 static void	lacp_sm_rx_timer(struct lacp_port *);
126 static void	lacp_sm_rx_set_expired(struct lacp_port *);
127 static void	lacp_sm_rx_update_ntt(struct lacp_port *,
128 		    const struct lacpdu *);
129 static void	lacp_sm_rx_record_pdu(struct lacp_port *,
130 		    const struct lacpdu *);
131 static void	lacp_sm_rx_update_selected(struct lacp_port *,
132 		    const struct lacpdu *);
133 static void	lacp_sm_rx_record_default(struct lacp_port *);
134 static void	lacp_sm_rx_update_default_selected(struct lacp_port *);
135 static void	lacp_sm_rx_update_selected_from_peerinfo(struct lacp_port *,
136 		    const struct lacp_peerinfo *);
137 
138 /* mux machine */
139 
140 static void	lacp_sm_mux(struct lacp_port *);
141 static void	lacp_set_mux(struct lacp_port *, enum lacp_mux_state);
142 static void	lacp_sm_mux_timer(struct lacp_port *);
143 
144 /* periodic transmit machine */
145 
146 static void	lacp_sm_ptx_update_timeout(struct lacp_port *, uint8_t);
147 static void	lacp_sm_ptx_tx_schedule(struct lacp_port *);
148 static void	lacp_sm_ptx_timer(struct lacp_port *);
149 
150 /* transmit machine */
151 
152 static void	lacp_sm_tx(struct lacp_port *);
153 static void	lacp_sm_assert_ntt(struct lacp_port *);
154 
155 static void	lacp_run_timers(struct lacp_port *);
156 static int	lacp_compare_peerinfo(const struct lacp_peerinfo *,
157 		    const struct lacp_peerinfo *);
158 static int	lacp_compare_systemid(const struct lacp_systemid *,
159 		    const struct lacp_systemid *);
160 static void	lacp_port_enable(struct lacp_port *);
161 static void	lacp_port_disable(struct lacp_port *);
162 static void	lacp_select(struct lacp_port *);
163 static void	lacp_unselect(struct lacp_port *);
164 static void	lacp_disable_collecting(struct lacp_port *);
165 static void	lacp_enable_collecting(struct lacp_port *);
166 static void	lacp_disable_distributing(struct lacp_port *);
167 static void	lacp_enable_distributing(struct lacp_port *);
168 static int	lacp_xmit_lacpdu(struct lacp_port *);
169 static int	lacp_xmit_marker(struct lacp_port *);
170 
171 #if defined(LACP_DEBUG)
172 static void	lacp_dump_lacpdu(const struct lacpdu *);
173 static const char *lacp_format_partner(const struct lacp_peerinfo *, char *,
174 		    size_t);
175 static const char *lacp_format_lagid(const struct lacp_peerinfo *,
176 		    const struct lacp_peerinfo *, char *, size_t);
177 static const char *lacp_format_lagid_aggregator(const struct lacp_aggregator *,
178 		    char *, size_t);
179 static const char *lacp_format_state(uint8_t, char *, size_t);
180 static const char *lacp_format_mac(const uint8_t *, char *, size_t);
181 static const char *lacp_format_systemid(const struct lacp_systemid *, char *,
182 		    size_t);
183 static const char *lacp_format_portid(const struct lacp_portid *, char *,
184 		    size_t);
185 static void	lacp_dprintf(const struct lacp_port *, const char *, ...)
186 		    __attribute__((__format__(__printf__, 2, 3)));
187 #define	LACP_DPRINTF(a)	lacp_dprintf a
188 #else
189 #define LACP_DPRINTF(a) /* nothing */
190 #endif
191 
192 /*
193  * partner administration variables.
194  * XXX should be configurable.
195  */
196 
197 static const struct lacp_peerinfo lacp_partner_admin = {
198 	.lip_systemid = { .lsi_prio = 0xffff },
199 	.lip_portid = { .lpi_prio = 0xffff },
200 #if 1
201 	/* optimistic */
202 	.lip_state = LACP_STATE_SYNC | LACP_STATE_AGGREGATION |
203 	    LACP_STATE_COLLECTING | LACP_STATE_DISTRIBUTING,
204 #else
205 	/* pessimistic */
206 	.lip_state = 0,
207 #endif
208 };
209 
210 static const lacp_timer_func_t lacp_timer_funcs[LACP_NTIMER] = {
211 	[LACP_TIMER_CURRENT_WHILE] = lacp_sm_rx_timer,
212 	[LACP_TIMER_PERIODIC] = lacp_sm_ptx_timer,
213 	[LACP_TIMER_WAIT_WHILE] = lacp_sm_mux_timer,
214 };
215 
216 void
217 lacp_input(struct lagg_port *lgp, struct mbuf *m)
218 {
219 	struct lagg_softc *sc = lgp->lp_softc;
220 	struct lacp_softc *lsc = LACP_SOFTC(sc);
221 	uint8_t subtype;
222 
223 	if (m->m_pkthdr.len < sizeof(struct ether_header) + sizeof(subtype)) {
224 		m_freem(m);
225 		return;
226 	}
227 
228 	m_copydata(m, sizeof(struct ether_header), sizeof(subtype), &subtype);
229 	switch (subtype) {
230 		case SLOWPROTOCOLS_SUBTYPE_LACP:
231 			IF_HANDOFF(&lsc->lsc_queue, m, NULL);
232 			taskqueue_enqueue(taskqueue_swi, &lsc->lsc_qtask);
233 			break;
234 
235 		case SLOWPROTOCOLS_SUBTYPE_MARKER:
236 			lacp_marker_input(lgp, m);
237 			break;
238 
239 		default:
240 			/* Unknown LACP packet type */
241 			m_freem(m);
242 			break;
243 	}
244 }
245 
246 static void
247 lacp_dequeue(void *arg, int pending)
248 {
249 	struct lacp_softc *lsc = (struct lacp_softc *)arg;
250 	struct lagg_softc *sc = lsc->lsc_softc;
251 	struct lagg_port *lgp;
252 	struct mbuf *m;
253 
254 	LAGG_WLOCK(sc);
255 	for (;;) {
256 		IF_DEQUEUE(&lsc->lsc_queue, m);
257 		if (m == NULL)
258 			break;
259 		lgp = m->m_pkthdr.rcvif->if_lagg;
260 		lacp_pdu_input(lgp, m);
261 	}
262 	LAGG_WUNLOCK(sc);
263 }
264 
265 /*
266  * lacp_pdu_input: process lacpdu
267  */
268 static int
269 lacp_pdu_input(struct lagg_port *lgp, struct mbuf *m)
270 {
271 	struct lacp_port *lp = LACP_PORT(lgp);
272 	struct lacpdu *du;
273 	int error = 0;
274 
275 	LAGG_WLOCK_ASSERT(lgp->lp_softc);
276 
277 	if (m->m_pkthdr.len != sizeof(*du)) {
278 		goto bad;
279 	}
280 
281 	if ((m->m_flags & M_MCAST) == 0) {
282 		goto bad;
283 	}
284 
285 	if (m->m_len < sizeof(*du)) {
286 		m = m_pullup(m, sizeof(*du));
287 		if (m == NULL) {
288 			return (ENOMEM);
289 		}
290 	}
291 
292 	du = mtod(m, struct lacpdu *);
293 
294 	if (memcmp(&du->ldu_eh.ether_dhost,
295 	    &ethermulticastaddr_slowprotocols, ETHER_ADDR_LEN)) {
296 		goto bad;
297 	}
298 
299 	/*
300 	 * ignore the version for compatibility with
301 	 * the future protocol revisions.
302 	 */
303 #if 0
304 	if (du->ldu_sph.sph_version != 1) {
305 		goto bad;
306 	}
307 #endif
308 
309 	/*
310 	 * ignore tlv types for compatibility with
311 	 * the future protocol revisions.
312 	 */
313 	if (tlv_check(du, sizeof(*du), &du->ldu_tlv_actor,
314 	    lacp_info_tlv_template, FALSE)) {
315 		goto bad;
316 	}
317 
318 #if defined(LACP_DEBUG)
319 	LACP_DPRINTF((lp, "lacpdu receive\n"));
320 	lacp_dump_lacpdu(du);
321 #endif /* defined(LACP_DEBUG) */
322 	lacp_sm_rx(lp, du);
323 
324 	m_freem(m);
325 
326 	return (error);
327 
328 bad:
329 	m_freem(m);
330 	return (EINVAL);
331 }
332 
333 static void
334 lacp_fill_actorinfo(struct lacp_port *lp, struct lacp_peerinfo *info)
335 {
336 	struct lagg_port *lgp = lp->lp_lagg;
337 	struct lagg_softc *sc = lgp->lp_softc;
338 
339 	info->lip_systemid.lsi_prio = htons(LACP_SYSTEM_PRIO);
340 	memcpy(&info->lip_systemid.lsi_mac,
341 	    IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
342 	info->lip_portid.lpi_prio = htons(LACP_PORT_PRIO);
343 	info->lip_portid.lpi_portno = htons(lp->lp_ifp->if_index);
344 	info->lip_state = lp->lp_state;
345 }
346 
347 static void
348 lacp_fill_markerinfo(struct lacp_port *lp, struct lacp_markerinfo *info)
349 {
350 	struct ifnet *ifp = lp->lp_ifp;
351 
352 	/* Fill in the port index and system id (encoded as the MAC) */
353 	info->mi_rq_port = htons(ifp->if_index);
354 	memcpy(&info->mi_rq_system, lp->lp_systemid.lsi_mac, ETHER_ADDR_LEN);
355 	info->mi_rq_xid = htonl(0);
356 }
357 
358 static int
359 lacp_xmit_lacpdu(struct lacp_port *lp)
360 {
361 	struct lagg_port *lgp = lp->lp_lagg;
362 	struct mbuf *m;
363 	struct lacpdu *du;
364 	int error;
365 
366 	LAGG_WLOCK_ASSERT(lgp->lp_softc);
367 
368 	m = m_gethdr(M_DONTWAIT, MT_DATA);
369 	if (m == NULL) {
370 		return (ENOMEM);
371 	}
372 	m->m_len = m->m_pkthdr.len = sizeof(*du);
373 
374 	du = mtod(m, struct lacpdu *);
375 	memset(du, 0, sizeof(*du));
376 
377 	memcpy(&du->ldu_eh.ether_dhost, ethermulticastaddr_slowprotocols,
378 	    ETHER_ADDR_LEN);
379 	memcpy(&du->ldu_eh.ether_shost, lgp->lp_lladdr, ETHER_ADDR_LEN);
380 	du->ldu_eh.ether_type = htons(ETHERTYPE_SLOW);
381 
382 	du->ldu_sph.sph_subtype = SLOWPROTOCOLS_SUBTYPE_LACP;
383 	du->ldu_sph.sph_version = 1;
384 
385 	TLV_SET(&du->ldu_tlv_actor, LACP_TYPE_ACTORINFO, sizeof(du->ldu_actor));
386 	du->ldu_actor = lp->lp_actor;
387 
388 	TLV_SET(&du->ldu_tlv_partner, LACP_TYPE_PARTNERINFO,
389 	    sizeof(du->ldu_partner));
390 	du->ldu_partner = lp->lp_partner;
391 
392 	TLV_SET(&du->ldu_tlv_collector, LACP_TYPE_COLLECTORINFO,
393 	    sizeof(du->ldu_collector));
394 	du->ldu_collector.lci_maxdelay = 0;
395 
396 #if defined(LACP_DEBUG)
397 	LACP_DPRINTF((lp, "lacpdu transmit\n"));
398 	lacp_dump_lacpdu(du);
399 #endif /* defined(LACP_DEBUG) */
400 
401 	m->m_flags |= M_MCAST;
402 
403 	/*
404 	 * XXX should use higher priority queue.
405 	 * otherwise network congestion can break aggregation.
406 	 */
407 
408 	error = lagg_enqueue(lp->lp_ifp, m);
409 	return (error);
410 }
411 
412 static int
413 lacp_xmit_marker(struct lacp_port *lp)
414 {
415 	struct lagg_port *lgp = lp->lp_lagg;
416 	struct mbuf *m;
417 	struct markerdu *mdu;
418 	int error;
419 
420 	LAGG_WLOCK_ASSERT(lgp->lp_softc);
421 
422 	m = m_gethdr(M_DONTWAIT, MT_DATA);
423 	if (m == NULL) {
424 		return (ENOMEM);
425 	}
426 	m->m_len = m->m_pkthdr.len = sizeof(*mdu);
427 
428 	mdu = mtod(m, struct markerdu *);
429 	memset(mdu, 0, sizeof(*mdu));
430 
431 	memcpy(&mdu->mdu_eh.ether_dhost, ethermulticastaddr_slowprotocols,
432 	    ETHER_ADDR_LEN);
433 	memcpy(&mdu->mdu_eh.ether_shost, lgp->lp_lladdr, ETHER_ADDR_LEN);
434 	mdu->mdu_eh.ether_type = htons(ETHERTYPE_SLOW);
435 
436 	mdu->mdu_sph.sph_subtype = SLOWPROTOCOLS_SUBTYPE_MARKER;
437 	mdu->mdu_sph.sph_version = 1;
438 
439 	/* Bump the transaction id and copy over the marker info */
440 	lp->lp_marker.mi_rq_xid = htonl(ntohl(lp->lp_marker.mi_rq_xid) + 1);
441 	TLV_SET(&mdu->mdu_tlv, MARKER_TYPE_INFO, sizeof(mdu->mdu_info));
442 	mdu->mdu_info = lp->lp_marker;
443 
444 	LACP_DPRINTF((lp, "marker transmit, port=%u, sys=%6D, id=%u\n",
445 	    ntohs(mdu->mdu_info.mi_rq_port), mdu->mdu_info.mi_rq_system, ":",
446 	    ntohl(mdu->mdu_info.mi_rq_xid)));
447 
448 	m->m_flags |= M_MCAST;
449 	error = lagg_enqueue(lp->lp_ifp, m);
450 	return (error);
451 }
452 void
453 lacp_linkstate(struct lagg_port *lgp)
454 {
455 	struct lacp_port *lp = LACP_PORT(lgp);
456 	struct ifnet *ifp = lgp->lp_ifp;
457 	struct ifmediareq ifmr;
458 	int error = 0;
459 	u_int media;
460 	uint8_t old_state;
461 	uint16_t old_key;
462 
463 	LAGG_WLOCK_ASSERT(lgp->lp_softc);
464 
465 	bzero((char *)&ifmr, sizeof(ifmr));
466 	error = (*ifp->if_ioctl)(ifp, SIOCGIFMEDIA, (caddr_t)&ifmr);
467 	if (error != 0)
468 		return;
469 
470 	media = ifmr.ifm_active;
471 	LACP_DPRINTF((lp, "media changed 0x%x -> 0x%x, ether = %d, fdx = %d, "
472 	    "link = %d\n", lp->lp_media, media, IFM_TYPE(media) == IFM_ETHER,
473 	    (media & IFM_FDX) != 0, ifp->if_link_state == LINK_STATE_UP));
474 	old_state = lp->lp_state;
475 	old_key = lp->lp_key;
476 
477 	lp->lp_media = media;
478 	/*
479 	 * If the port is not an active full duplex Ethernet link then it can
480 	 * not be aggregated.
481 	 */
482 	if (IFM_TYPE(media) != IFM_ETHER || (media & IFM_FDX) == 0 ||
483 	    ifp->if_link_state != LINK_STATE_UP) {
484 		lacp_port_disable(lp);
485 	} else {
486 		lacp_port_enable(lp);
487 	}
488 	lp->lp_key = lacp_compose_key(lp);
489 
490 	if (old_state != lp->lp_state || old_key != lp->lp_key) {
491 		LACP_DPRINTF((lp, "-> UNSELECTED\n"));
492 		lp->lp_selected = LACP_UNSELECTED;
493 	}
494 }
495 
496 static void
497 lacp_tick(void *arg)
498 {
499 	struct lacp_softc *lsc = arg;
500 	struct lagg_softc *sc = lsc->lsc_softc;
501 	struct lacp_port *lp;
502 
503 	LAGG_WLOCK(sc);
504 	LIST_FOREACH(lp, &lsc->lsc_ports, lp_next) {
505 		if ((lp->lp_state & LACP_STATE_AGGREGATION) == 0)
506 			continue;
507 
508 		lacp_run_timers(lp);
509 
510 		lacp_select(lp);
511 		lacp_sm_mux(lp);
512 		lacp_sm_tx(lp);
513 		lacp_sm_ptx_tx_schedule(lp);
514 	}
515 	LAGG_WUNLOCK(sc);
516 	callout_reset(&lsc->lsc_callout, hz, lacp_tick, lsc);
517 }
518 
519 int
520 lacp_port_create(struct lagg_port *lgp)
521 {
522 	struct lagg_softc *sc = lgp->lp_softc;
523 	struct lacp_softc *lsc = LACP_SOFTC(sc);
524 	struct lacp_port *lp;
525 	struct ifnet *ifp = lgp->lp_ifp;
526 	struct sockaddr_dl sdl;
527 	struct ifmultiaddr *rifma = NULL;
528 	int error;
529 
530 	boolean_t active = TRUE; /* XXX should be configurable */
531 	boolean_t fast = FALSE; /* XXX should be configurable */
532 
533 	LAGG_WLOCK_ASSERT(sc);
534 
535 	bzero((char *)&sdl, sizeof(sdl));
536 	sdl.sdl_len = sizeof(sdl);
537 	sdl.sdl_family = AF_LINK;
538 	sdl.sdl_index = ifp->if_index;
539 	sdl.sdl_type = IFT_ETHER;
540 	sdl.sdl_alen = ETHER_ADDR_LEN;
541 
542 	bcopy(&ethermulticastaddr_slowprotocols,
543 	    LLADDR(&sdl), ETHER_ADDR_LEN);
544 	error = if_addmulti(ifp, (struct sockaddr *)&sdl, &rifma);
545 	if (error) {
546 		printf("%s: ADDMULTI failed on %s\n", __func__, lgp->lp_ifname);
547 		return (error);
548 	}
549 
550 	lp = malloc(sizeof(struct lacp_port),
551 	    M_DEVBUF, M_NOWAIT|M_ZERO);
552 	if (lp == NULL)
553 		return (ENOMEM);
554 
555 	lgp->lp_psc = (caddr_t)lp;
556 	lp->lp_ifp = ifp;
557 	lp->lp_lagg = lgp;
558 	lp->lp_lsc = lsc;
559 	lp->lp_ifma = rifma;
560 
561 	LIST_INSERT_HEAD(&lsc->lsc_ports, lp, lp_next);
562 
563 	lacp_fill_actorinfo(lp, &lp->lp_actor);
564 	lacp_fill_markerinfo(lp, &lp->lp_marker);
565 	lp->lp_state =
566 	    (active ? LACP_STATE_ACTIVITY : 0) |
567 	    (fast ? LACP_STATE_TIMEOUT : 0);
568 	lp->lp_aggregator = NULL;
569 	lacp_linkstate(lgp);
570 	lacp_sm_rx_set_expired(lp);
571 
572 	return (0);
573 }
574 
575 void
576 lacp_port_destroy(struct lagg_port *lgp)
577 {
578 	struct lacp_port *lp = LACP_PORT(lgp);
579 	int i;
580 
581 	LAGG_WLOCK_ASSERT(lgp->lp_softc);
582 
583 	for (i = 0; i < LACP_NTIMER; i++) {
584 		LACP_TIMER_DISARM(lp, i);
585 	}
586 
587 	lacp_disable_collecting(lp);
588 	lacp_disable_distributing(lp);
589 	lacp_unselect(lp);
590 	lgp->lp_flags &= ~LAGG_PORT_DISABLED;
591 
592 	/* The address may have already been removed by if_purgemaddrs() */
593 	if (!lgp->lp_detaching)
594 		if_delmulti_ifma(lp->lp_ifma);
595 
596 	LIST_REMOVE(lp, lp_next);
597 	free(lp, M_DEVBUF);
598 }
599 
600 int
601 lacp_port_isactive(struct lagg_port *lgp)
602 {
603 	struct lacp_port *lp = LACP_PORT(lgp);
604 	struct lacp_softc *lsc = lp->lp_lsc;
605 	struct lacp_aggregator *la = lp->lp_aggregator;
606 
607 	/* This port is joined to the active aggregator */
608 	if (la != NULL && la == lsc->lsc_active_aggregator)
609 		return (1);
610 
611 	return (0);
612 }
613 
614 void
615 lacp_req(struct lagg_softc *sc, caddr_t data)
616 {
617 	struct lacp_opreq *req = (struct lacp_opreq *)data;
618 	struct lacp_softc *lsc = LACP_SOFTC(sc);
619 	struct lacp_aggregator *la = lsc->lsc_active_aggregator;
620 
621 	bzero(req, sizeof(struct lacp_opreq));
622 	if (la != NULL) {
623 		req->actor_prio = ntohs(la->la_actor.lip_systemid.lsi_prio);
624 		memcpy(&req->actor_mac, &la->la_actor.lip_systemid.lsi_mac,
625 		    ETHER_ADDR_LEN);
626 		req->actor_key = ntohs(la->la_actor.lip_key);
627 		req->actor_portprio = ntohs(la->la_actor.lip_portid.lpi_prio);
628 		req->actor_portno = ntohs(la->la_actor.lip_portid.lpi_portno);
629 		req->actor_state = la->la_actor.lip_state;
630 
631 		req->partner_prio = ntohs(la->la_partner.lip_systemid.lsi_prio);
632 		memcpy(&req->partner_mac, &la->la_partner.lip_systemid.lsi_mac,
633 		    ETHER_ADDR_LEN);
634 		req->partner_key = ntohs(la->la_partner.lip_key);
635 		req->partner_portprio = ntohs(la->la_partner.lip_portid.lpi_prio);
636 		req->partner_portno = ntohs(la->la_partner.lip_portid.lpi_portno);
637 		req->partner_state = la->la_partner.lip_state;
638 	}
639 }
640 
641 void
642 lacp_portreq(struct lagg_port *lgp, caddr_t data)
643 {
644 	struct lacp_opreq *req = (struct lacp_opreq *)data;
645 	struct lacp_port *lp = LACP_PORT(lgp);
646 
647 	req->actor_prio = ntohs(lp->lp_actor.lip_systemid.lsi_prio);
648 	memcpy(&req->actor_mac, &lp->lp_actor.lip_systemid.lsi_mac,
649 	    ETHER_ADDR_LEN);
650 	req->actor_key = ntohs(lp->lp_actor.lip_key);
651 	req->actor_portprio = ntohs(lp->lp_actor.lip_portid.lpi_prio);
652 	req->actor_portno = ntohs(lp->lp_actor.lip_portid.lpi_portno);
653 	req->actor_state = lp->lp_actor.lip_state;
654 
655 	req->partner_prio = ntohs(lp->lp_partner.lip_systemid.lsi_prio);
656 	memcpy(&req->partner_mac, &lp->lp_partner.lip_systemid.lsi_mac,
657 	    ETHER_ADDR_LEN);
658 	req->partner_key = ntohs(lp->lp_partner.lip_key);
659 	req->partner_portprio = ntohs(lp->lp_partner.lip_portid.lpi_prio);
660 	req->partner_portno = ntohs(lp->lp_partner.lip_portid.lpi_portno);
661 	req->partner_state = lp->lp_partner.lip_state;
662 }
663 
664 static void
665 lacp_disable_collecting(struct lacp_port *lp)
666 {
667 	struct lagg_port *lgp = lp->lp_lagg;
668 
669 	LACP_DPRINTF((lp, "collecting disabled\n"));
670 
671 	lp->lp_state &= ~LACP_STATE_COLLECTING;
672 	lgp->lp_flags &= ~LAGG_PORT_COLLECTING;
673 }
674 
675 static void
676 lacp_enable_collecting(struct lacp_port *lp)
677 {
678 	struct lagg_port *lgp = lp->lp_lagg;
679 
680 	LACP_DPRINTF((lp, "collecting enabled\n"));
681 
682 	lp->lp_state |= LACP_STATE_COLLECTING;
683 	lgp->lp_flags |= LAGG_PORT_COLLECTING;
684 }
685 
686 static void
687 lacp_disable_distributing(struct lacp_port *lp)
688 {
689 	struct lacp_aggregator *la = lp->lp_aggregator;
690 	struct lacp_softc *lsc = lp->lp_lsc;
691 	struct lagg_port *lgp = lp->lp_lagg;
692 #if defined(LACP_DEBUG)
693 	char buf[LACP_LAGIDSTR_MAX+1];
694 #endif /* defined(LACP_DEBUG) */
695 
696 	LAGG_WLOCK_ASSERT(lgp->lp_softc);
697 
698 	if (la == NULL || (lp->lp_state & LACP_STATE_DISTRIBUTING) == 0) {
699 		return;
700 	}
701 
702 	KASSERT(!TAILQ_EMPTY(&la->la_ports), ("no aggregator ports"));
703 	KASSERT(la->la_nports > 0, ("nports invalid (%d)", la->la_nports));
704 	KASSERT(la->la_refcnt >= la->la_nports, ("aggregator refcnt invalid"));
705 
706 	LACP_DPRINTF((lp, "disable distributing on aggregator %s, "
707 	    "nports %d -> %d\n",
708 	    lacp_format_lagid_aggregator(la, buf, sizeof(buf)),
709 	    la->la_nports, la->la_nports - 1));
710 
711 	TAILQ_REMOVE(&la->la_ports, lp, lp_dist_q);
712 	la->la_nports--;
713 
714 	lacp_suppress_distributing(lsc, la);
715 
716 	lp->lp_state &= ~LACP_STATE_DISTRIBUTING;
717 	lgp->lp_flags &= ~LAGG_PORT_DISTRIBUTING;
718 
719 	if (lsc->lsc_active_aggregator == la) {
720 		lacp_select_active_aggregator(lsc);
721 	}
722 }
723 
724 static void
725 lacp_enable_distributing(struct lacp_port *lp)
726 {
727 	struct lacp_aggregator *la = lp->lp_aggregator;
728 	struct lacp_softc *lsc = lp->lp_lsc;
729 	struct lagg_port *lgp = lp->lp_lagg;
730 #if defined(LACP_DEBUG)
731 	char buf[LACP_LAGIDSTR_MAX+1];
732 #endif /* defined(LACP_DEBUG) */
733 
734 	LAGG_WLOCK_ASSERT(lgp->lp_softc);
735 
736 	if ((lp->lp_state & LACP_STATE_DISTRIBUTING) != 0) {
737 		return;
738 	}
739 
740 	LACP_DPRINTF((lp, "enable distributing on aggregator %s, "
741 	    "nports %d -> %d\n",
742 	    lacp_format_lagid_aggregator(la, buf, sizeof(buf)),
743 	    la->la_nports, la->la_nports + 1));
744 
745 	KASSERT(la->la_refcnt > la->la_nports, ("aggregator refcnt invalid"));
746 	TAILQ_INSERT_HEAD(&la->la_ports, lp, lp_dist_q);
747 	la->la_nports++;
748 
749 	lacp_suppress_distributing(lsc, la);
750 
751 	lp->lp_state |= LACP_STATE_DISTRIBUTING;
752 	lgp->lp_flags |= LAGG_PORT_DISTRIBUTING;
753 
754 	if (lsc->lsc_active_aggregator != la) {
755 		lacp_select_active_aggregator(lsc);
756 	}
757 }
758 
759 static void
760 lacp_transit_expire(void *vp)
761 {
762 	struct lacp_softc *lsc = vp;
763 
764 	LACP_DPRINTF((NULL, "%s\n", __func__));
765 	lsc->lsc_suppress_distributing = FALSE;
766 }
767 
768 int
769 lacp_attach(struct lagg_softc *sc)
770 {
771 	struct lacp_softc *lsc;
772 
773 	LAGG_WLOCK_ASSERT(sc);
774 
775 	lsc = malloc(sizeof(struct lacp_softc),
776 	    M_DEVBUF, M_NOWAIT|M_ZERO);
777 	if (lsc == NULL)
778 		return (ENOMEM);
779 
780 	sc->sc_psc = (caddr_t)lsc;
781 	lsc->lsc_softc = sc;
782 
783 	lsc->lsc_hashkey = arc4random();
784 	lsc->lsc_active_aggregator = NULL;
785 	TAILQ_INIT(&lsc->lsc_aggregators);
786 	LIST_INIT(&lsc->lsc_ports);
787 
788 	TASK_INIT(&lsc->lsc_qtask, 0, lacp_dequeue, lsc);
789 	mtx_init(&lsc->lsc_queue.ifq_mtx, "lacp queue", NULL, MTX_DEF);
790 	lsc->lsc_queue.ifq_maxlen = ifqmaxlen;
791 
792 	callout_init(&lsc->lsc_transit_callout, CALLOUT_MPSAFE);
793 	callout_init(&lsc->lsc_callout, CALLOUT_MPSAFE);
794 
795 	/* if the lagg is already up then do the same */
796 	if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
797 		lacp_init(sc);
798 
799 	return (0);
800 }
801 
802 int
803 lacp_detach(struct lagg_softc *sc)
804 {
805 	struct lacp_softc *lsc = LACP_SOFTC(sc);
806 
807 	KASSERT(TAILQ_EMPTY(&lsc->lsc_aggregators),
808 	    ("aggregators still active"));
809 	KASSERT(lsc->lsc_active_aggregator == NULL,
810 	    ("aggregator still attached"));
811 
812 	sc->sc_psc = NULL;
813 	callout_drain(&lsc->lsc_transit_callout);
814 	callout_drain(&lsc->lsc_callout);
815 	taskqueue_drain(taskqueue_swi, &lsc->lsc_qtask);
816 	IF_DRAIN(&lsc->lsc_queue);
817 	mtx_destroy(&lsc->lsc_queue.ifq_mtx);
818 
819 	free(lsc, M_DEVBUF);
820 	return (0);
821 }
822 
823 void
824 lacp_init(struct lagg_softc *sc)
825 {
826 	struct lacp_softc *lsc = LACP_SOFTC(sc);
827 
828 	callout_reset(&lsc->lsc_callout, hz, lacp_tick, lsc);
829 }
830 
831 void
832 lacp_stop(struct lagg_softc *sc)
833 {
834 	struct lacp_softc *lsc = LACP_SOFTC(sc);
835 
836 	callout_stop(&lsc->lsc_transit_callout);
837 	callout_stop(&lsc->lsc_callout);
838 }
839 
840 struct lagg_port *
841 lacp_select_tx_port(struct lagg_softc *sc, struct mbuf *m)
842 {
843 	struct lacp_softc *lsc = LACP_SOFTC(sc);
844 	struct lacp_aggregator *la;
845 	struct lacp_port *lp;
846 	uint32_t hash;
847 	int nports;
848 
849 	LAGG_RLOCK_ASSERT(sc);
850 
851 	if (__predict_false(lsc->lsc_suppress_distributing)) {
852 		LACP_DPRINTF((NULL, "%s: waiting transit\n", __func__));
853 		return (NULL);
854 	}
855 
856 	la = lsc->lsc_active_aggregator;
857 	if (__predict_false(la == NULL)) {
858 		LACP_DPRINTF((NULL, "%s: no active aggregator\n", __func__));
859 		return (NULL);
860 	}
861 
862 	nports = la->la_nports;
863 	KASSERT(nports > 0, ("no ports available"));
864 
865 	hash = lagg_hashmbuf(m, lsc->lsc_hashkey);
866 	hash %= nports;
867 	lp = TAILQ_FIRST(&la->la_ports);
868 	while (hash--) {
869 		lp = TAILQ_NEXT(lp, lp_dist_q);
870 	}
871 
872 	KASSERT((lp->lp_state & LACP_STATE_DISTRIBUTING) != 0,
873 	    ("aggregated port is not distributing"));
874 
875 	return (lp->lp_lagg);
876 }
877 /*
878  * lacp_suppress_distributing: drop transmit packets for a while
879  * to preserve packet ordering.
880  */
881 
882 static void
883 lacp_suppress_distributing(struct lacp_softc *lsc, struct lacp_aggregator *la)
884 {
885 	struct lacp_port *lp;
886 
887 	if (lsc->lsc_active_aggregator != la) {
888 		return;
889 	}
890 
891 	LACP_DPRINTF((NULL, "%s\n", __func__));
892 	lsc->lsc_suppress_distributing = TRUE;
893 
894 	/* send a marker frame down each port to verify the queues are empty */
895 	LIST_FOREACH(lp, &lsc->lsc_ports, lp_next) {
896 		lp->lp_flags |= LACP_PORT_MARK;
897 		lacp_xmit_marker(lp);
898 	}
899 
900 	/* set a timeout for the marker frames */
901 	callout_reset(&lsc->lsc_transit_callout,
902 	    LACP_TRANSIT_DELAY * hz / 1000, lacp_transit_expire, lsc);
903 }
904 
905 static int
906 lacp_compare_peerinfo(const struct lacp_peerinfo *a,
907     const struct lacp_peerinfo *b)
908 {
909 	return (memcmp(a, b, offsetof(struct lacp_peerinfo, lip_state)));
910 }
911 
912 static int
913 lacp_compare_systemid(const struct lacp_systemid *a,
914     const struct lacp_systemid *b)
915 {
916 	return (memcmp(a, b, sizeof(*a)));
917 }
918 
919 #if 0	/* unused */
920 static int
921 lacp_compare_portid(const struct lacp_portid *a,
922     const struct lacp_portid *b)
923 {
924 	return (memcmp(a, b, sizeof(*a)));
925 }
926 #endif
927 
928 static uint64_t
929 lacp_aggregator_bandwidth(struct lacp_aggregator *la)
930 {
931 	struct lacp_port *lp;
932 	uint64_t speed;
933 
934 	lp = TAILQ_FIRST(&la->la_ports);
935 	if (lp == NULL) {
936 		return (0);
937 	}
938 
939 	speed = ifmedia_baudrate(lp->lp_media);
940 	speed *= la->la_nports;
941 	if (speed == 0) {
942 		LACP_DPRINTF((lp, "speed 0? media=0x%x nports=%d\n",
943 		    lp->lp_media, la->la_nports));
944 	}
945 
946 	return (speed);
947 }
948 
949 /*
950  * lacp_select_active_aggregator: select an aggregator to be used to transmit
951  * packets from lagg(4) interface.
952  */
953 
954 static void
955 lacp_select_active_aggregator(struct lacp_softc *lsc)
956 {
957 	struct lacp_aggregator *la;
958 	struct lacp_aggregator *best_la = NULL;
959 	uint64_t best_speed = 0;
960 #if defined(LACP_DEBUG)
961 	char buf[LACP_LAGIDSTR_MAX+1];
962 #endif /* defined(LACP_DEBUG) */
963 
964 	LACP_DPRINTF((NULL, "%s:\n", __func__));
965 
966 	TAILQ_FOREACH(la, &lsc->lsc_aggregators, la_q) {
967 		uint64_t speed;
968 
969 		if (la->la_nports == 0) {
970 			continue;
971 		}
972 
973 		speed = lacp_aggregator_bandwidth(la);
974 		LACP_DPRINTF((NULL, "%s, speed=%jd, nports=%d\n",
975 		    lacp_format_lagid_aggregator(la, buf, sizeof(buf)),
976 		    speed, la->la_nports));
977 
978 		/* This aggregator is chosen if
979 		 *      the partner has a better system priority
980 		 *  or, the total aggregated speed is higher
981 		 *  or, it is already the chosen aggregator
982 		 */
983 		if ((best_la != NULL && LACP_SYS_PRI(la->la_partner) <
984 		     LACP_SYS_PRI(best_la->la_partner)) ||
985 		    speed > best_speed ||
986 		    (speed == best_speed &&
987 		    la == lsc->lsc_active_aggregator)) {
988 			best_la = la;
989 			best_speed = speed;
990 		}
991 	}
992 
993 	KASSERT(best_la == NULL || best_la->la_nports > 0,
994 	    ("invalid aggregator refcnt"));
995 	KASSERT(best_la == NULL || !TAILQ_EMPTY(&best_la->la_ports),
996 	    ("invalid aggregator list"));
997 
998 #if defined(LACP_DEBUG)
999 	if (lsc->lsc_active_aggregator != best_la) {
1000 		LACP_DPRINTF((NULL, "active aggregator changed\n"));
1001 		LACP_DPRINTF((NULL, "old %s\n",
1002 		    lacp_format_lagid_aggregator(lsc->lsc_active_aggregator,
1003 		    buf, sizeof(buf))));
1004 	} else {
1005 		LACP_DPRINTF((NULL, "active aggregator not changed\n"));
1006 	}
1007 	LACP_DPRINTF((NULL, "new %s\n",
1008 	    lacp_format_lagid_aggregator(best_la, buf, sizeof(buf))));
1009 #endif /* defined(LACP_DEBUG) */
1010 
1011 	if (lsc->lsc_active_aggregator != best_la) {
1012 		lsc->lsc_active_aggregator = best_la;
1013 		if (best_la) {
1014 			lacp_suppress_distributing(lsc, best_la);
1015 		}
1016 	}
1017 }
1018 
1019 static uint16_t
1020 lacp_compose_key(struct lacp_port *lp)
1021 {
1022 	struct lagg_port *lgp = lp->lp_lagg;
1023 	struct lagg_softc *sc = lgp->lp_softc;
1024 	u_int media = lp->lp_media;
1025 	uint16_t key;
1026 
1027 	if ((lp->lp_state & LACP_STATE_AGGREGATION) == 0) {
1028 
1029 		/*
1030 		 * non-aggregatable links should have unique keys.
1031 		 *
1032 		 * XXX this isn't really unique as if_index is 16 bit.
1033 		 */
1034 
1035 		/* bit 0..14:	(some bits of) if_index of this port */
1036 		key = lp->lp_ifp->if_index;
1037 		/* bit 15:	1 */
1038 		key |= 0x8000;
1039 	} else {
1040 		u_int subtype = IFM_SUBTYPE(media);
1041 
1042 		KASSERT(IFM_TYPE(media) == IFM_ETHER, ("invalid media type"));
1043 		KASSERT((media & IFM_FDX) != 0, ("aggregating HDX interface"));
1044 
1045 		/* bit 0..4:	IFM_SUBTYPE */
1046 		key = subtype;
1047 		/* bit 5..14:	(some bits of) if_index of lagg device */
1048 		key |= 0x7fe0 & ((sc->sc_ifp->if_index) << 5);
1049 		/* bit 15:	0 */
1050 	}
1051 	return (htons(key));
1052 }
1053 
1054 static void
1055 lacp_aggregator_addref(struct lacp_softc *lsc, struct lacp_aggregator *la)
1056 {
1057 #if defined(LACP_DEBUG)
1058 	char buf[LACP_LAGIDSTR_MAX+1];
1059 #endif
1060 
1061 	LACP_DPRINTF((NULL, "%s: lagid=%s, refcnt %d -> %d\n",
1062 	    __func__,
1063 	    lacp_format_lagid(&la->la_actor, &la->la_partner,
1064 	    buf, sizeof(buf)),
1065 	    la->la_refcnt, la->la_refcnt + 1));
1066 
1067 	KASSERT(la->la_refcnt > 0, ("refcount <= 0"));
1068 	la->la_refcnt++;
1069 	KASSERT(la->la_refcnt > la->la_nports, ("invalid refcount"));
1070 }
1071 
1072 static void
1073 lacp_aggregator_delref(struct lacp_softc *lsc, struct lacp_aggregator *la)
1074 {
1075 #if defined(LACP_DEBUG)
1076 	char buf[LACP_LAGIDSTR_MAX+1];
1077 #endif
1078 
1079 	LACP_DPRINTF((NULL, "%s: lagid=%s, refcnt %d -> %d\n",
1080 	    __func__,
1081 	    lacp_format_lagid(&la->la_actor, &la->la_partner,
1082 	    buf, sizeof(buf)),
1083 	    la->la_refcnt, la->la_refcnt - 1));
1084 
1085 	KASSERT(la->la_refcnt > la->la_nports, ("invalid refcnt"));
1086 	la->la_refcnt--;
1087 	if (la->la_refcnt > 0) {
1088 		return;
1089 	}
1090 
1091 	KASSERT(la->la_refcnt == 0, ("refcount not zero"));
1092 	KASSERT(lsc->lsc_active_aggregator != la, ("aggregator active"));
1093 
1094 	TAILQ_REMOVE(&lsc->lsc_aggregators, la, la_q);
1095 
1096 	free(la, M_DEVBUF);
1097 }
1098 
1099 /*
1100  * lacp_aggregator_get: allocate an aggregator.
1101  */
1102 
1103 static struct lacp_aggregator *
1104 lacp_aggregator_get(struct lacp_softc *lsc, struct lacp_port *lp)
1105 {
1106 	struct lacp_aggregator *la;
1107 
1108 	la = malloc(sizeof(*la), M_DEVBUF, M_NOWAIT);
1109 	if (la) {
1110 		la->la_refcnt = 1;
1111 		la->la_nports = 0;
1112 		TAILQ_INIT(&la->la_ports);
1113 		la->la_pending = 0;
1114 		TAILQ_INSERT_TAIL(&lsc->lsc_aggregators, la, la_q);
1115 	}
1116 
1117 	return (la);
1118 }
1119 
1120 /*
1121  * lacp_fill_aggregator_id: setup a newly allocated aggregator from a port.
1122  */
1123 
1124 static void
1125 lacp_fill_aggregator_id(struct lacp_aggregator *la, const struct lacp_port *lp)
1126 {
1127 	lacp_fill_aggregator_id_peer(&la->la_partner, &lp->lp_partner);
1128 	lacp_fill_aggregator_id_peer(&la->la_actor, &lp->lp_actor);
1129 
1130 	la->la_actor.lip_state = lp->lp_state & LACP_STATE_AGGREGATION;
1131 }
1132 
1133 static void
1134 lacp_fill_aggregator_id_peer(struct lacp_peerinfo *lpi_aggr,
1135     const struct lacp_peerinfo *lpi_port)
1136 {
1137 	memset(lpi_aggr, 0, sizeof(*lpi_aggr));
1138 	lpi_aggr->lip_systemid = lpi_port->lip_systemid;
1139 	lpi_aggr->lip_key = lpi_port->lip_key;
1140 }
1141 
1142 /*
1143  * lacp_aggregator_is_compatible: check if a port can join to an aggregator.
1144  */
1145 
1146 static int
1147 lacp_aggregator_is_compatible(const struct lacp_aggregator *la,
1148     const struct lacp_port *lp)
1149 {
1150 	if (!(lp->lp_state & LACP_STATE_AGGREGATION) ||
1151 	    !(lp->lp_partner.lip_state & LACP_STATE_AGGREGATION)) {
1152 		return (0);
1153 	}
1154 
1155 	if (!(la->la_actor.lip_state & LACP_STATE_AGGREGATION)) {
1156 		return (0);
1157 	}
1158 
1159 	if (!lacp_peerinfo_is_compatible(&la->la_partner, &lp->lp_partner)) {
1160 		return (0);
1161 	}
1162 
1163 	if (!lacp_peerinfo_is_compatible(&la->la_actor, &lp->lp_actor)) {
1164 		return (0);
1165 	}
1166 
1167 	return (1);
1168 }
1169 
1170 static int
1171 lacp_peerinfo_is_compatible(const struct lacp_peerinfo *a,
1172     const struct lacp_peerinfo *b)
1173 {
1174 	if (memcmp(&a->lip_systemid, &b->lip_systemid,
1175 	    sizeof(a->lip_systemid))) {
1176 		return (0);
1177 	}
1178 
1179 	if (memcmp(&a->lip_key, &b->lip_key, sizeof(a->lip_key))) {
1180 		return (0);
1181 	}
1182 
1183 	return (1);
1184 }
1185 
1186 static void
1187 lacp_port_enable(struct lacp_port *lp)
1188 {
1189 	struct lagg_port *lgp = lp->lp_lagg;
1190 
1191 	lp->lp_state |= LACP_STATE_AGGREGATION;
1192 	lgp->lp_flags &= ~LAGG_PORT_DISABLED;
1193 }
1194 
1195 static void
1196 lacp_port_disable(struct lacp_port *lp)
1197 {
1198 	struct lagg_port *lgp = lp->lp_lagg;
1199 
1200 	lacp_set_mux(lp, LACP_MUX_DETACHED);
1201 
1202 	lp->lp_state &= ~LACP_STATE_AGGREGATION;
1203 	lp->lp_selected = LACP_UNSELECTED;
1204 	lacp_sm_rx_record_default(lp);
1205 	lp->lp_partner.lip_state &= ~LACP_STATE_AGGREGATION;
1206 	lp->lp_state &= ~LACP_STATE_EXPIRED;
1207 	lgp->lp_flags |= LAGG_PORT_DISABLED;
1208 }
1209 
1210 /*
1211  * lacp_select: select an aggregator.  create one if necessary.
1212  */
1213 static void
1214 lacp_select(struct lacp_port *lp)
1215 {
1216 	struct lacp_softc *lsc = lp->lp_lsc;
1217 	struct lacp_aggregator *la;
1218 #if defined(LACP_DEBUG)
1219 	char buf[LACP_LAGIDSTR_MAX+1];
1220 #endif
1221 
1222 	if (lp->lp_aggregator) {
1223 		return;
1224 	}
1225 
1226 	KASSERT(!LACP_TIMER_ISARMED(lp, LACP_TIMER_WAIT_WHILE),
1227 	    ("timer_wait_while still active"));
1228 
1229 	LACP_DPRINTF((lp, "port lagid=%s\n",
1230 	    lacp_format_lagid(&lp->lp_actor, &lp->lp_partner,
1231 	    buf, sizeof(buf))));
1232 
1233 	TAILQ_FOREACH(la, &lsc->lsc_aggregators, la_q) {
1234 		if (lacp_aggregator_is_compatible(la, lp)) {
1235 			break;
1236 		}
1237 	}
1238 
1239 	if (la == NULL) {
1240 		la = lacp_aggregator_get(lsc, lp);
1241 		if (la == NULL) {
1242 			LACP_DPRINTF((lp, "aggregator creation failed\n"));
1243 
1244 			/*
1245 			 * will retry on the next tick.
1246 			 */
1247 
1248 			return;
1249 		}
1250 		lacp_fill_aggregator_id(la, lp);
1251 		LACP_DPRINTF((lp, "aggregator created\n"));
1252 	} else {
1253 		LACP_DPRINTF((lp, "compatible aggregator found\n"));
1254 		lacp_aggregator_addref(lsc, la);
1255 	}
1256 
1257 	LACP_DPRINTF((lp, "aggregator lagid=%s\n",
1258 	    lacp_format_lagid(&la->la_actor, &la->la_partner,
1259 	    buf, sizeof(buf))));
1260 
1261 	lp->lp_aggregator = la;
1262 	lp->lp_selected = LACP_SELECTED;
1263 }
1264 
1265 /*
1266  * lacp_unselect: finish unselect/detach process.
1267  */
1268 
1269 static void
1270 lacp_unselect(struct lacp_port *lp)
1271 {
1272 	struct lacp_softc *lsc = lp->lp_lsc;
1273 	struct lacp_aggregator *la = lp->lp_aggregator;
1274 
1275 	KASSERT(!LACP_TIMER_ISARMED(lp, LACP_TIMER_WAIT_WHILE),
1276 	    ("timer_wait_while still active"));
1277 
1278 	if (la == NULL) {
1279 		return;
1280 	}
1281 
1282 	lp->lp_aggregator = NULL;
1283 	lacp_aggregator_delref(lsc, la);
1284 }
1285 
1286 /* mux machine */
1287 
1288 static void
1289 lacp_sm_mux(struct lacp_port *lp)
1290 {
1291 	enum lacp_mux_state new_state;
1292 	boolean_t p_sync =
1293 		    (lp->lp_partner.lip_state & LACP_STATE_SYNC) != 0;
1294 	boolean_t p_collecting =
1295 	    (lp->lp_partner.lip_state & LACP_STATE_COLLECTING) != 0;
1296 	enum lacp_selected selected = lp->lp_selected;
1297 	struct lacp_aggregator *la;
1298 
1299 	/* LACP_DPRINTF((lp, "%s: state %d\n", __func__, lp->lp_mux_state)); */
1300 
1301 re_eval:
1302 	la = lp->lp_aggregator;
1303 	KASSERT(lp->lp_mux_state == LACP_MUX_DETACHED || la != NULL,
1304 	    ("MUX not detached"));
1305 	new_state = lp->lp_mux_state;
1306 	switch (lp->lp_mux_state) {
1307 	case LACP_MUX_DETACHED:
1308 		if (selected != LACP_UNSELECTED) {
1309 			new_state = LACP_MUX_WAITING;
1310 		}
1311 		break;
1312 	case LACP_MUX_WAITING:
1313 		KASSERT(la->la_pending > 0 ||
1314 		    !LACP_TIMER_ISARMED(lp, LACP_TIMER_WAIT_WHILE),
1315 		    ("timer_wait_while still active"));
1316 		if (selected == LACP_SELECTED && la->la_pending == 0) {
1317 			new_state = LACP_MUX_ATTACHED;
1318 		} else if (selected == LACP_UNSELECTED) {
1319 			new_state = LACP_MUX_DETACHED;
1320 		}
1321 		break;
1322 	case LACP_MUX_ATTACHED:
1323 		if (selected == LACP_SELECTED && p_sync) {
1324 			new_state = LACP_MUX_COLLECTING;
1325 		} else if (selected != LACP_SELECTED) {
1326 			new_state = LACP_MUX_DETACHED;
1327 		}
1328 		break;
1329 	case LACP_MUX_COLLECTING:
1330 		if (selected == LACP_SELECTED && p_sync && p_collecting) {
1331 			new_state = LACP_MUX_DISTRIBUTING;
1332 		} else if (selected != LACP_SELECTED || !p_sync) {
1333 			new_state = LACP_MUX_ATTACHED;
1334 		}
1335 		break;
1336 	case LACP_MUX_DISTRIBUTING:
1337 		if (selected != LACP_SELECTED || !p_sync || !p_collecting) {
1338 			new_state = LACP_MUX_COLLECTING;
1339 		}
1340 		break;
1341 	default:
1342 		panic("%s: unknown state", __func__);
1343 	}
1344 
1345 	if (lp->lp_mux_state == new_state) {
1346 		return;
1347 	}
1348 
1349 	lacp_set_mux(lp, new_state);
1350 	goto re_eval;
1351 }
1352 
1353 static void
1354 lacp_set_mux(struct lacp_port *lp, enum lacp_mux_state new_state)
1355 {
1356 	struct lacp_aggregator *la = lp->lp_aggregator;
1357 
1358 	if (lp->lp_mux_state == new_state) {
1359 		return;
1360 	}
1361 
1362 	switch (new_state) {
1363 	case LACP_MUX_DETACHED:
1364 		lp->lp_state &= ~LACP_STATE_SYNC;
1365 		lacp_disable_distributing(lp);
1366 		lacp_disable_collecting(lp);
1367 		lacp_sm_assert_ntt(lp);
1368 		/* cancel timer */
1369 		if (LACP_TIMER_ISARMED(lp, LACP_TIMER_WAIT_WHILE)) {
1370 			KASSERT(la->la_pending > 0,
1371 			    ("timer_wait_while not active"));
1372 			la->la_pending--;
1373 		}
1374 		LACP_TIMER_DISARM(lp, LACP_TIMER_WAIT_WHILE);
1375 		lacp_unselect(lp);
1376 		break;
1377 	case LACP_MUX_WAITING:
1378 		LACP_TIMER_ARM(lp, LACP_TIMER_WAIT_WHILE,
1379 		    LACP_AGGREGATE_WAIT_TIME);
1380 		la->la_pending++;
1381 		break;
1382 	case LACP_MUX_ATTACHED:
1383 		lp->lp_state |= LACP_STATE_SYNC;
1384 		lacp_disable_collecting(lp);
1385 		lacp_sm_assert_ntt(lp);
1386 		break;
1387 	case LACP_MUX_COLLECTING:
1388 		lacp_enable_collecting(lp);
1389 		lacp_disable_distributing(lp);
1390 		lacp_sm_assert_ntt(lp);
1391 		break;
1392 	case LACP_MUX_DISTRIBUTING:
1393 		lacp_enable_distributing(lp);
1394 		break;
1395 	default:
1396 		panic("%s: unknown state", __func__);
1397 	}
1398 
1399 	LACP_DPRINTF((lp, "mux_state %d -> %d\n", lp->lp_mux_state, new_state));
1400 
1401 	lp->lp_mux_state = new_state;
1402 }
1403 
1404 static void
1405 lacp_sm_mux_timer(struct lacp_port *lp)
1406 {
1407 	struct lacp_aggregator *la = lp->lp_aggregator;
1408 #if defined(LACP_DEBUG)
1409 	char buf[LACP_LAGIDSTR_MAX+1];
1410 #endif
1411 
1412 	KASSERT(la->la_pending > 0, ("no pending event"));
1413 
1414 	LACP_DPRINTF((lp, "%s: aggregator %s, pending %d -> %d\n", __func__,
1415 	    lacp_format_lagid(&la->la_actor, &la->la_partner,
1416 	    buf, sizeof(buf)),
1417 	    la->la_pending, la->la_pending - 1));
1418 
1419 	la->la_pending--;
1420 }
1421 
1422 /* periodic transmit machine */
1423 
1424 static void
1425 lacp_sm_ptx_update_timeout(struct lacp_port *lp, uint8_t oldpstate)
1426 {
1427 	if (LACP_STATE_EQ(oldpstate, lp->lp_partner.lip_state,
1428 	    LACP_STATE_TIMEOUT)) {
1429 		return;
1430 	}
1431 
1432 	LACP_DPRINTF((lp, "partner timeout changed\n"));
1433 
1434 	/*
1435 	 * FAST_PERIODIC -> SLOW_PERIODIC
1436 	 * or
1437 	 * SLOW_PERIODIC (-> PERIODIC_TX) -> FAST_PERIODIC
1438 	 *
1439 	 * let lacp_sm_ptx_tx_schedule to update timeout.
1440 	 */
1441 
1442 	LACP_TIMER_DISARM(lp, LACP_TIMER_PERIODIC);
1443 
1444 	/*
1445 	 * if timeout has been shortened, assert NTT.
1446 	 */
1447 
1448 	if ((lp->lp_partner.lip_state & LACP_STATE_TIMEOUT)) {
1449 		lacp_sm_assert_ntt(lp);
1450 	}
1451 }
1452 
1453 static void
1454 lacp_sm_ptx_tx_schedule(struct lacp_port *lp)
1455 {
1456 	int timeout;
1457 
1458 	if (!(lp->lp_state & LACP_STATE_ACTIVITY) &&
1459 	    !(lp->lp_partner.lip_state & LACP_STATE_ACTIVITY)) {
1460 
1461 		/*
1462 		 * NO_PERIODIC
1463 		 */
1464 
1465 		LACP_TIMER_DISARM(lp, LACP_TIMER_PERIODIC);
1466 		return;
1467 	}
1468 
1469 	if (LACP_TIMER_ISARMED(lp, LACP_TIMER_PERIODIC)) {
1470 		return;
1471 	}
1472 
1473 	timeout = (lp->lp_partner.lip_state & LACP_STATE_TIMEOUT) ?
1474 	    LACP_FAST_PERIODIC_TIME : LACP_SLOW_PERIODIC_TIME;
1475 
1476 	LACP_TIMER_ARM(lp, LACP_TIMER_PERIODIC, timeout);
1477 }
1478 
1479 static void
1480 lacp_sm_ptx_timer(struct lacp_port *lp)
1481 {
1482 	lacp_sm_assert_ntt(lp);
1483 }
1484 
1485 static void
1486 lacp_sm_rx(struct lacp_port *lp, const struct lacpdu *du)
1487 {
1488 	int timeout;
1489 
1490 	/*
1491 	 * check LACP_DISABLED first
1492 	 */
1493 
1494 	if (!(lp->lp_state & LACP_STATE_AGGREGATION)) {
1495 		return;
1496 	}
1497 
1498 	/*
1499 	 * check loopback condition.
1500 	 */
1501 
1502 	if (!lacp_compare_systemid(&du->ldu_actor.lip_systemid,
1503 	    &lp->lp_actor.lip_systemid)) {
1504 		return;
1505 	}
1506 
1507 	/*
1508 	 * EXPIRED, DEFAULTED, CURRENT -> CURRENT
1509 	 */
1510 
1511 	lacp_sm_rx_update_selected(lp, du);
1512 	lacp_sm_rx_update_ntt(lp, du);
1513 	lacp_sm_rx_record_pdu(lp, du);
1514 
1515 	timeout = (lp->lp_state & LACP_STATE_TIMEOUT) ?
1516 	    LACP_SHORT_TIMEOUT_TIME : LACP_LONG_TIMEOUT_TIME;
1517 	LACP_TIMER_ARM(lp, LACP_TIMER_CURRENT_WHILE, timeout);
1518 
1519 	lp->lp_state &= ~LACP_STATE_EXPIRED;
1520 
1521 	/*
1522 	 * kick transmit machine without waiting the next tick.
1523 	 */
1524 
1525 	lacp_sm_tx(lp);
1526 }
1527 
1528 static void
1529 lacp_sm_rx_set_expired(struct lacp_port *lp)
1530 {
1531 	lp->lp_partner.lip_state &= ~LACP_STATE_SYNC;
1532 	lp->lp_partner.lip_state |= LACP_STATE_TIMEOUT;
1533 	LACP_TIMER_ARM(lp, LACP_TIMER_CURRENT_WHILE, LACP_SHORT_TIMEOUT_TIME);
1534 	lp->lp_state |= LACP_STATE_EXPIRED;
1535 }
1536 
1537 static void
1538 lacp_sm_rx_timer(struct lacp_port *lp)
1539 {
1540 	if ((lp->lp_state & LACP_STATE_EXPIRED) == 0) {
1541 		/* CURRENT -> EXPIRED */
1542 		LACP_DPRINTF((lp, "%s: CURRENT -> EXPIRED\n", __func__));
1543 		lacp_sm_rx_set_expired(lp);
1544 	} else {
1545 		/* EXPIRED -> DEFAULTED */
1546 		LACP_DPRINTF((lp, "%s: EXPIRED -> DEFAULTED\n", __func__));
1547 		lacp_sm_rx_update_default_selected(lp);
1548 		lacp_sm_rx_record_default(lp);
1549 		lp->lp_state &= ~LACP_STATE_EXPIRED;
1550 	}
1551 }
1552 
1553 static void
1554 lacp_sm_rx_record_pdu(struct lacp_port *lp, const struct lacpdu *du)
1555 {
1556 	boolean_t active;
1557 	uint8_t oldpstate;
1558 #if defined(LACP_DEBUG)
1559 	char buf[LACP_STATESTR_MAX+1];
1560 #endif
1561 
1562 	/* LACP_DPRINTF((lp, "%s\n", __func__)); */
1563 
1564 	oldpstate = lp->lp_partner.lip_state;
1565 
1566 	active = (du->ldu_actor.lip_state & LACP_STATE_ACTIVITY)
1567 	    || ((lp->lp_state & LACP_STATE_ACTIVITY) &&
1568 	    (du->ldu_partner.lip_state & LACP_STATE_ACTIVITY));
1569 
1570 	lp->lp_partner = du->ldu_actor;
1571 	if (active &&
1572 	    ((LACP_STATE_EQ(lp->lp_state, du->ldu_partner.lip_state,
1573 	    LACP_STATE_AGGREGATION) &&
1574 	    !lacp_compare_peerinfo(&lp->lp_actor, &du->ldu_partner))
1575 	    || (du->ldu_partner.lip_state & LACP_STATE_AGGREGATION) == 0)) {
1576 		/* XXX nothing? */
1577 	} else {
1578 		lp->lp_partner.lip_state &= ~LACP_STATE_SYNC;
1579 	}
1580 
1581 	lp->lp_state &= ~LACP_STATE_DEFAULTED;
1582 
1583 	if (oldpstate != lp->lp_partner.lip_state) {
1584 		LACP_DPRINTF((lp, "old pstate %s\n",
1585 		    lacp_format_state(oldpstate, buf, sizeof(buf))));
1586 		LACP_DPRINTF((lp, "new pstate %s\n",
1587 		    lacp_format_state(lp->lp_partner.lip_state, buf,
1588 		    sizeof(buf))));
1589 	}
1590 
1591 	lacp_sm_ptx_update_timeout(lp, oldpstate);
1592 }
1593 
1594 static void
1595 lacp_sm_rx_update_ntt(struct lacp_port *lp, const struct lacpdu *du)
1596 {
1597 	/* LACP_DPRINTF((lp, "%s\n", __func__)); */
1598 
1599 	if (lacp_compare_peerinfo(&lp->lp_actor, &du->ldu_partner) ||
1600 	    !LACP_STATE_EQ(lp->lp_state, du->ldu_partner.lip_state,
1601 	    LACP_STATE_ACTIVITY | LACP_STATE_SYNC | LACP_STATE_AGGREGATION)) {
1602 		LACP_DPRINTF((lp, "%s: assert ntt\n", __func__));
1603 		lacp_sm_assert_ntt(lp);
1604 	}
1605 }
1606 
1607 static void
1608 lacp_sm_rx_record_default(struct lacp_port *lp)
1609 {
1610 	uint8_t oldpstate;
1611 
1612 	/* LACP_DPRINTF((lp, "%s\n", __func__)); */
1613 
1614 	oldpstate = lp->lp_partner.lip_state;
1615 	lp->lp_partner = lacp_partner_admin;
1616 	lp->lp_state |= LACP_STATE_DEFAULTED;
1617 	lacp_sm_ptx_update_timeout(lp, oldpstate);
1618 }
1619 
1620 static void
1621 lacp_sm_rx_update_selected_from_peerinfo(struct lacp_port *lp,
1622     const struct lacp_peerinfo *info)
1623 {
1624 	/* LACP_DPRINTF((lp, "%s\n", __func__)); */
1625 
1626 	if (lacp_compare_peerinfo(&lp->lp_partner, info) ||
1627 	    !LACP_STATE_EQ(lp->lp_partner.lip_state, info->lip_state,
1628 	    LACP_STATE_AGGREGATION)) {
1629 		lp->lp_selected = LACP_UNSELECTED;
1630 		/* mux machine will clean up lp->lp_aggregator */
1631 	}
1632 }
1633 
1634 static void
1635 lacp_sm_rx_update_selected(struct lacp_port *lp, const struct lacpdu *du)
1636 {
1637 	/* LACP_DPRINTF((lp, "%s\n", __func__)); */
1638 
1639 	lacp_sm_rx_update_selected_from_peerinfo(lp, &du->ldu_actor);
1640 }
1641 
1642 static void
1643 lacp_sm_rx_update_default_selected(struct lacp_port *lp)
1644 {
1645 	/* LACP_DPRINTF((lp, "%s\n", __func__)); */
1646 
1647 	lacp_sm_rx_update_selected_from_peerinfo(lp, &lacp_partner_admin);
1648 }
1649 
1650 /* transmit machine */
1651 
1652 static void
1653 lacp_sm_tx(struct lacp_port *lp)
1654 {
1655 	int error;
1656 
1657 	if (!(lp->lp_state & LACP_STATE_AGGREGATION)
1658 #if 1
1659 	    || (!(lp->lp_state & LACP_STATE_ACTIVITY)
1660 	    && !(lp->lp_partner.lip_state & LACP_STATE_ACTIVITY))
1661 #endif
1662 	    ) {
1663 		lp->lp_flags &= ~LACP_PORT_NTT;
1664 	}
1665 
1666 	if (!(lp->lp_flags & LACP_PORT_NTT)) {
1667 		return;
1668 	}
1669 
1670 	/* Rate limit to 3 PDUs per LACP_FAST_PERIODIC_TIME */
1671 	if (ppsratecheck(&lp->lp_last_lacpdu, &lp->lp_lacpdu_sent,
1672 		    (3 / LACP_FAST_PERIODIC_TIME)) == 0) {
1673 		LACP_DPRINTF((lp, "rate limited pdu\n"));
1674 		return;
1675 	}
1676 
1677 	error = lacp_xmit_lacpdu(lp);
1678 
1679 	if (error == 0) {
1680 		lp->lp_flags &= ~LACP_PORT_NTT;
1681 	} else {
1682 		LACP_DPRINTF((lp, "lacpdu transmit failure, error %d\n",
1683 		    error));
1684 	}
1685 }
1686 
1687 static void
1688 lacp_sm_assert_ntt(struct lacp_port *lp)
1689 {
1690 
1691 	lp->lp_flags |= LACP_PORT_NTT;
1692 }
1693 
1694 static void
1695 lacp_run_timers(struct lacp_port *lp)
1696 {
1697 	int i;
1698 
1699 	for (i = 0; i < LACP_NTIMER; i++) {
1700 		KASSERT(lp->lp_timer[i] >= 0,
1701 		    ("invalid timer value %d", lp->lp_timer[i]));
1702 		if (lp->lp_timer[i] == 0) {
1703 			continue;
1704 		} else if (--lp->lp_timer[i] <= 0) {
1705 			if (lacp_timer_funcs[i]) {
1706 				(*lacp_timer_funcs[i])(lp);
1707 			}
1708 		}
1709 	}
1710 }
1711 
1712 int
1713 lacp_marker_input(struct lagg_port *lgp, struct mbuf *m)
1714 {
1715 	struct lacp_port *lp = LACP_PORT(lgp);
1716 	struct lacp_port *lp2;
1717 	struct lacp_softc *lsc = lp->lp_lsc;
1718 	struct markerdu *mdu;
1719 	int error = 0;
1720 	int pending = 0;
1721 
1722 	LAGG_RLOCK_ASSERT(lgp->lp_softc);
1723 
1724 	if (m->m_pkthdr.len != sizeof(*mdu)) {
1725 		goto bad;
1726 	}
1727 
1728 	if ((m->m_flags & M_MCAST) == 0) {
1729 		goto bad;
1730 	}
1731 
1732 	if (m->m_len < sizeof(*mdu)) {
1733 		m = m_pullup(m, sizeof(*mdu));
1734 		if (m == NULL) {
1735 			return (ENOMEM);
1736 		}
1737 	}
1738 
1739 	mdu = mtod(m, struct markerdu *);
1740 
1741 	if (memcmp(&mdu->mdu_eh.ether_dhost,
1742 	    &ethermulticastaddr_slowprotocols, ETHER_ADDR_LEN)) {
1743 		goto bad;
1744 	}
1745 
1746 	if (mdu->mdu_sph.sph_version != 1) {
1747 		goto bad;
1748 	}
1749 
1750 	switch (mdu->mdu_tlv.tlv_type) {
1751 	case MARKER_TYPE_INFO:
1752 		if (tlv_check(mdu, sizeof(*mdu), &mdu->mdu_tlv,
1753 		    marker_info_tlv_template, TRUE)) {
1754 			goto bad;
1755 		}
1756 		mdu->mdu_tlv.tlv_type = MARKER_TYPE_RESPONSE;
1757 		memcpy(&mdu->mdu_eh.ether_dhost,
1758 		    &ethermulticastaddr_slowprotocols, ETHER_ADDR_LEN);
1759 		memcpy(&mdu->mdu_eh.ether_shost,
1760 		    lgp->lp_lladdr, ETHER_ADDR_LEN);
1761 		error = lagg_enqueue(lp->lp_ifp, m);
1762 		break;
1763 
1764 	case MARKER_TYPE_RESPONSE:
1765 		if (tlv_check(mdu, sizeof(*mdu), &mdu->mdu_tlv,
1766 		    marker_response_tlv_template, TRUE)) {
1767 			goto bad;
1768 		}
1769 		LACP_DPRINTF((lp, "marker response, port=%u, sys=%6D, id=%u\n",
1770 		    ntohs(mdu->mdu_info.mi_rq_port), mdu->mdu_info.mi_rq_system,
1771 		    ":", ntohl(mdu->mdu_info.mi_rq_xid)));
1772 
1773 		/* Verify that it is the last marker we sent out */
1774 		if (memcmp(&mdu->mdu_info, &lp->lp_marker,
1775 		    sizeof(struct lacp_markerinfo)))
1776 			goto bad;
1777 
1778 		lp->lp_flags &= ~LACP_PORT_MARK;
1779 
1780 		if (lsc->lsc_suppress_distributing) {
1781 			/* Check if any ports are waiting for a response */
1782 			LIST_FOREACH(lp2, &lsc->lsc_ports, lp_next) {
1783 				if (lp2->lp_flags & LACP_PORT_MARK) {
1784 					pending = 1;
1785 					break;
1786 				}
1787 			}
1788 
1789 			if (pending == 0) {
1790 				/* All interface queues are clear */
1791 				LACP_DPRINTF((NULL, "queue flush complete\n"));
1792 				lsc->lsc_suppress_distributing = FALSE;
1793 			}
1794 		}
1795 
1796 		m_freem(m);
1797 		break;
1798 
1799 	default:
1800 		goto bad;
1801 	}
1802 
1803 	return (error);
1804 
1805 bad:
1806 	LACP_DPRINTF((lp, "bad marker frame\n"));
1807 	m_freem(m);
1808 	return (EINVAL);
1809 }
1810 
1811 static int
1812 tlv_check(const void *p, size_t size, const struct tlvhdr *tlv,
1813     const struct tlv_template *tmpl, boolean_t check_type)
1814 {
1815 	while (/* CONSTCOND */ 1) {
1816 		if ((const char *)tlv - (const char *)p + sizeof(*tlv) > size) {
1817 			return (EINVAL);
1818 		}
1819 		if ((check_type && tlv->tlv_type != tmpl->tmpl_type) ||
1820 		    tlv->tlv_length != tmpl->tmpl_length) {
1821 			return (EINVAL);
1822 		}
1823 		if (tmpl->tmpl_type == 0) {
1824 			break;
1825 		}
1826 		tlv = (const struct tlvhdr *)
1827 		    ((const char *)tlv + tlv->tlv_length);
1828 		tmpl++;
1829 	}
1830 
1831 	return (0);
1832 }
1833 
1834 #if defined(LACP_DEBUG)
1835 const char *
1836 lacp_format_mac(const uint8_t *mac, char *buf, size_t buflen)
1837 {
1838 	snprintf(buf, buflen, "%02X-%02X-%02X-%02X-%02X-%02X",
1839 	    (int)mac[0],
1840 	    (int)mac[1],
1841 	    (int)mac[2],
1842 	    (int)mac[3],
1843 	    (int)mac[4],
1844 	    (int)mac[5]);
1845 
1846 	return (buf);
1847 }
1848 
1849 const char *
1850 lacp_format_systemid(const struct lacp_systemid *sysid,
1851     char *buf, size_t buflen)
1852 {
1853 	char macbuf[LACP_MACSTR_MAX+1];
1854 
1855 	snprintf(buf, buflen, "%04X,%s",
1856 	    ntohs(sysid->lsi_prio),
1857 	    lacp_format_mac(sysid->lsi_mac, macbuf, sizeof(macbuf)));
1858 
1859 	return (buf);
1860 }
1861 
1862 const char *
1863 lacp_format_portid(const struct lacp_portid *portid, char *buf, size_t buflen)
1864 {
1865 	snprintf(buf, buflen, "%04X,%04X",
1866 	    ntohs(portid->lpi_prio),
1867 	    ntohs(portid->lpi_portno));
1868 
1869 	return (buf);
1870 }
1871 
1872 const char *
1873 lacp_format_partner(const struct lacp_peerinfo *peer, char *buf, size_t buflen)
1874 {
1875 	char sysid[LACP_SYSTEMIDSTR_MAX+1];
1876 	char portid[LACP_PORTIDSTR_MAX+1];
1877 
1878 	snprintf(buf, buflen, "(%s,%04X,%s)",
1879 	    lacp_format_systemid(&peer->lip_systemid, sysid, sizeof(sysid)),
1880 	    ntohs(peer->lip_key),
1881 	    lacp_format_portid(&peer->lip_portid, portid, sizeof(portid)));
1882 
1883 	return (buf);
1884 }
1885 
1886 const char *
1887 lacp_format_lagid(const struct lacp_peerinfo *a,
1888     const struct lacp_peerinfo *b, char *buf, size_t buflen)
1889 {
1890 	char astr[LACP_PARTNERSTR_MAX+1];
1891 	char bstr[LACP_PARTNERSTR_MAX+1];
1892 
1893 #if 0
1894 	/*
1895 	 * there's a convention to display small numbered peer
1896 	 * in the left.
1897 	 */
1898 
1899 	if (lacp_compare_peerinfo(a, b) > 0) {
1900 		const struct lacp_peerinfo *t;
1901 
1902 		t = a;
1903 		a = b;
1904 		b = t;
1905 	}
1906 #endif
1907 
1908 	snprintf(buf, buflen, "[%s,%s]",
1909 	    lacp_format_partner(a, astr, sizeof(astr)),
1910 	    lacp_format_partner(b, bstr, sizeof(bstr)));
1911 
1912 	return (buf);
1913 }
1914 
1915 const char *
1916 lacp_format_lagid_aggregator(const struct lacp_aggregator *la,
1917     char *buf, size_t buflen)
1918 {
1919 	if (la == NULL) {
1920 		return ("(none)");
1921 	}
1922 
1923 	return (lacp_format_lagid(&la->la_actor, &la->la_partner, buf, buflen));
1924 }
1925 
1926 const char *
1927 lacp_format_state(uint8_t state, char *buf, size_t buflen)
1928 {
1929 	snprintf(buf, buflen, "%b", state, LACP_STATE_BITS);
1930 	return (buf);
1931 }
1932 
1933 static void
1934 lacp_dump_lacpdu(const struct lacpdu *du)
1935 {
1936 	char buf[LACP_PARTNERSTR_MAX+1];
1937 	char buf2[LACP_STATESTR_MAX+1];
1938 
1939 	printf("actor=%s\n",
1940 	    lacp_format_partner(&du->ldu_actor, buf, sizeof(buf)));
1941 	printf("actor.state=%s\n",
1942 	    lacp_format_state(du->ldu_actor.lip_state, buf2, sizeof(buf2)));
1943 	printf("partner=%s\n",
1944 	    lacp_format_partner(&du->ldu_partner, buf, sizeof(buf)));
1945 	printf("partner.state=%s\n",
1946 	    lacp_format_state(du->ldu_partner.lip_state, buf2, sizeof(buf2)));
1947 
1948 	printf("maxdelay=%d\n", ntohs(du->ldu_collector.lci_maxdelay));
1949 }
1950 
1951 static void
1952 lacp_dprintf(const struct lacp_port *lp, const char *fmt, ...)
1953 {
1954 	va_list va;
1955 
1956 	if (lp) {
1957 		printf("%s: ", lp->lp_ifp->if_xname);
1958 	}
1959 
1960 	va_start(va, fmt);
1961 	vprintf(fmt, va);
1962 	va_end(va);
1963 }
1964 #endif
1965