xref: /dragonfly/sys/net/bridge/bridgestp.c (revision 0ca59c34)
1 /*
2  * Copyright (c) 2000 Jason L. Wright (jason@thought.net)
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *      This product includes software developed by Jason L. Wright
16  * 4. The name of the author may not be used to endorse or promote products
17  *    derived from this software without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
20  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
21  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
22  * DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
23  * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
24  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
25  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
27  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
28  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  *
31  * $OpenBSD: bridgestp.c,v 1.5 2001/03/22 03:48:29 jason Exp $
32  * $NetBSD: bridgestp.c,v 1.5 2003/11/28 08:56:48 keihan Exp $
33  * $FreeBSD: src/sys/net/bridgestp.c,v 1.7 2005/10/11 02:58:32 thompsa Exp $
34  */
35 
36 /*
37  * Implementation of the spanning tree protocol as defined in
38  * ISO/IEC Final DIS 15802-3 (IEEE P802.1D/D17), May 25, 1998.
39  * (In English: IEEE 802.1D, Draft 17, 1998)
40  */
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/mbuf.h>
45 #include <sys/socket.h>
46 #include <sys/sockio.h>
47 #include <sys/kernel.h>
48 #include <sys/callout.h>
49 #include <sys/proc.h>
50 #include <sys/lock.h>
51 #include <sys/thread.h>
52 #include <sys/thread2.h>
53 #include <sys/msgport2.h>
54 
55 #include <net/if.h>
56 #include <net/if_dl.h>
57 #include <net/if_types.h>
58 #include <net/if_llc.h>
59 #include <net/if_media.h>
60 
61 #include <netinet/in.h>
62 #include <netinet/in_systm.h>
63 #include <netinet/in_var.h>
64 #include <netinet/if_ether.h>
65 #include <net/bridge/if_bridgevar.h>
66 
67 /* BPDU message types */
68 #define	BSTP_MSGTYPE_CFG	0x00		/* Configuration */
69 #define	BSTP_MSGTYPE_TCN	0x80		/* Topology chg notification */
70 
71 /* BPDU flags */
72 #define	BSTP_FLAG_TC		0x01		/* Topology change */
73 #define	BSTP_FLAG_TCA		0x80		/* Topology change ack */
74 
75 #define	BSTP_MESSAGE_AGE_INCR	(1 * 256)	/* in 256ths of a second */
76 #define	BSTP_TICK_VAL		(1 * 256)	/* in 256ths of a second */
77 
78 /*
79  * Because BPDU's do not make nicely aligned structures, two different
80  * declarations are used: bstp_?bpdu (wire representation, packed) and
81  * bstp_*_unit (internal, nicely aligned version).
82  */
83 
84 /* configuration bridge protocol data unit */
85 struct bstp_cbpdu {
86 	uint8_t		cbu_dsap;		/* LLC: destination sap */
87 	uint8_t		cbu_ssap;		/* LLC: source sap */
88 	uint8_t		cbu_ctl;		/* LLC: control */
89 	uint16_t	cbu_protoid;		/* protocol id */
90 	uint8_t		cbu_protover;		/* protocol version */
91 	uint8_t		cbu_bpdutype;		/* message type */
92 	uint8_t		cbu_flags;		/* flags (below) */
93 
94 	/* root id */
95 	uint16_t	cbu_rootpri;		/* root priority */
96 	uint8_t	cbu_rootaddr[6];	/* root address */
97 
98 	uint32_t	cbu_rootpathcost;	/* root path cost */
99 
100 	/* bridge id */
101 	uint16_t	cbu_bridgepri;		/* bridge priority */
102 	uint8_t		cbu_bridgeaddr[6];	/* bridge address */
103 
104 	uint16_t	cbu_portid;		/* port id */
105 	uint16_t	cbu_messageage;		/* current message age */
106 	uint16_t	cbu_maxage;		/* maximum age */
107 	uint16_t	cbu_hellotime;		/* hello time */
108 	uint16_t	cbu_forwarddelay;	/* forwarding delay */
109 } __attribute__((__packed__));
110 
111 /* topology change notification bridge protocol data unit */
112 struct bstp_tbpdu {
113 	uint8_t		tbu_dsap;		/* LLC: destination sap */
114 	uint8_t		tbu_ssap;		/* LLC: source sap */
115 	uint8_t		tbu_ctl;		/* LLC: control */
116 	uint16_t	tbu_protoid;		/* protocol id */
117 	uint8_t		tbu_protover;		/* protocol version */
118 	uint8_t		tbu_bpdutype;		/* message type */
119 } __attribute__((__packed__));
120 
121 const uint8_t bstp_etheraddr[] = { 0x01, 0x80, 0xc2, 0x00, 0x00, 0x00 };
122 
123 static void	bstp_initialize_port(struct bridge_softc *,
124 		    struct bridge_iflist *);
125 static void	bstp_ifupdstatus(struct bridge_softc *, struct bridge_iflist *);
126 static void	bstp_enable_port(struct bridge_softc *, struct bridge_iflist *);
127 static void	bstp_disable_port(struct bridge_softc *,
128 		    struct bridge_iflist *);
129 #ifdef notused
130 static void	bstp_enable_change_detection(struct bridge_iflist *);
131 static void	bstp_disable_change_detection(struct bridge_iflist *);
132 #endif /* notused */
133 static int	bstp_root_bridge(struct bridge_softc *sc);
134 static void	bstp_transmit_config(struct bridge_softc *,
135 		    struct bridge_iflist *);
136 static void	bstp_transmit_tcn(struct bridge_softc *);
137 static void	bstp_received_config_bpdu(struct bridge_softc *,
138 		    struct bridge_iflist *, struct bstp_config_unit *);
139 static void	bstp_received_tcn_bpdu(struct bridge_softc *,
140 		    struct bridge_iflist *, struct bstp_tcn_unit *);
141 static void	bstp_record_config_information(struct bridge_softc *,
142 		    struct bridge_iflist *, struct bstp_config_unit *);
143 static void	bstp_record_config_timeout_values(struct bridge_softc *,
144 		    struct bstp_config_unit *);
145 static void	bstp_config_bpdu_generation(struct bridge_softc *);
146 static void	bstp_send_config_bpdu(struct bridge_softc *,
147 		    struct bridge_iflist *, struct bstp_config_unit *);
148 static void	bstp_configuration_update(struct bridge_softc *);
149 static void	bstp_port_state_selection(struct bridge_softc *);
150 static void	bstp_clear_peer_info(struct bridge_softc *,
151 		    struct bridge_iflist *);
152 static void	bstp_make_forwarding(struct bridge_softc *,
153 		    struct bridge_iflist *);
154 static void	bstp_make_blocking(struct bridge_softc *,
155 		    struct bridge_iflist *);
156 static void	bstp_make_l1blocking(struct bridge_softc *sc,
157 		    struct bridge_iflist *bif);
158 static void	bstp_adjust_bonded_states(struct bridge_softc *sc,
159 		    struct bridge_iflist *obif);
160 static void	bstp_set_port_state(struct bridge_iflist *, uint8_t);
161 #ifdef notused
162 static void	bstp_set_bridge_priority(struct bridge_softc *, uint64_t);
163 static void	bstp_set_port_priority(struct bridge_softc *,
164 		    struct bridge_iflist *, uint16_t);
165 static void	bstp_set_path_cost(struct bridge_softc *,
166 		    struct bridge_iflist *, uint32_t);
167 #endif /* notused */
168 static void	bstp_topology_change_detection(struct bridge_softc *);
169 static void	bstp_topology_change_acknowledged(struct bridge_softc *);
170 static void	bstp_acknowledge_topology_change(struct bridge_softc *,
171 		    struct bridge_iflist *);
172 
173 static void	bstp_tick(void *);
174 static void	bstp_timer_start(struct bridge_timer *, uint16_t);
175 static void	bstp_timer_stop(struct bridge_timer *);
176 static int	bstp_timer_expired(struct bridge_timer *, uint16_t);
177 
178 static void	bstp_hold_timer_expiry(struct bridge_softc *,
179 		    struct bridge_iflist *);
180 static void	bstp_message_age_timer_expiry(struct bridge_softc *,
181 		    struct bridge_iflist *);
182 static void	bstp_forward_delay_timer_expiry(struct bridge_softc *,
183 		    struct bridge_iflist *);
184 static void	bstp_topology_change_timer_expiry(struct bridge_softc *);
185 static void	bstp_tcn_timer_expiry(struct bridge_softc *);
186 static void	bstp_hello_timer_expiry(struct bridge_softc *);
187 static int	bstp_addr_cmp(const uint8_t *, const uint8_t *);
188 
189 /*
190  * When transmitting a config we tack on our path cost to
191  * our aggregated path-to-root cost.
192  *
193  * Note that sc_designated_cost is the lowest root path cost from all
194  * incoming links.  When one is talking about two bridges with multiple
195  * links between them the root path cost the secondary bridge transmits
196  * will be the lowest root path cost it received across those three
197  * links and NOT necessarily the root path cost it received specifically
198  * on that particular link.  The receiving end must still add its local
199  * link cost back in for differentiation purposes.
200  */
201 static void
202 bstp_transmit_config(struct bridge_softc *sc, struct bridge_iflist *bif)
203 {
204 	if (bif->bif_hold_timer.active) {
205 		bif->bif_config_pending = 1;
206 		return;
207 	}
208 
209 	bif->bif_config_bpdu.cu_message_type = BSTP_MSGTYPE_CFG;
210 	bif->bif_config_bpdu.cu_rootid = sc->sc_designated_root;
211 	bif->bif_config_bpdu.cu_root_path_cost = sc->sc_designated_cost +
212 						 bif->bif_path_cost;
213 	bif->bif_config_bpdu.cu_bridge_id = sc->sc_bridge_id;
214 	bif->bif_config_bpdu.cu_port_id = bif->bif_port_id;
215 
216 	if (bstp_root_bridge(sc)) {
217 		bif->bif_config_bpdu.cu_message_age = 0;
218 	} else if (sc->sc_root_port) {
219 		bif->bif_config_bpdu.cu_message_age =
220 			sc->sc_root_port->bif_message_age_timer.value +
221 			BSTP_MESSAGE_AGE_INCR;
222 	} else {
223 		bif->bif_config_bpdu.cu_message_age = BSTP_MESSAGE_AGE_INCR;
224 	}
225 
226 	bif->bif_config_bpdu.cu_max_age = sc->sc_max_age;
227 	bif->bif_config_bpdu.cu_hello_time = sc->sc_hello_time;
228 	bif->bif_config_bpdu.cu_forward_delay = sc->sc_forward_delay;
229 	bif->bif_config_bpdu.cu_topology_change_acknowledgment
230 	    = bif->bif_topology_change_acknowledge;
231 	bif->bif_config_bpdu.cu_topology_change = sc->sc_topology_change;
232 
233 	if (bif->bif_config_bpdu.cu_message_age < sc->sc_max_age ||
234 	    (sc->sc_ifp->if_flags & IFF_LINK1)) {
235 		bif->bif_topology_change_acknowledge = 0;
236 		bif->bif_config_pending = 0;
237 		bstp_send_config_bpdu(sc, bif, &bif->bif_config_bpdu);
238 		bstp_timer_start(&bif->bif_hold_timer, 0);
239 	}
240 }
241 
242 static void
243 bstp_send_config_bpdu(struct bridge_softc *sc, struct bridge_iflist *bif,
244 		      struct bstp_config_unit *cu)
245 {
246 	struct ifnet *ifp;
247 	struct mbuf *m;
248 	struct ether_header *eh;
249 	struct bstp_cbpdu bpdu;
250 
251 	ifp = bif->bif_ifp;
252 
253 	if ((ifp->if_flags & IFF_RUNNING) == 0)
254 		return;
255 
256 	MGETHDR(m, M_NOWAIT, MT_DATA);
257 	if (m == NULL)
258 		return;
259 
260 	eh = mtod(m, struct ether_header *);
261 
262 	m->m_pkthdr.rcvif = ifp;
263 	m->m_pkthdr.len = sizeof(*eh) + sizeof(bpdu);
264 	m->m_len = m->m_pkthdr.len;
265 
266 	bpdu.cbu_ssap = bpdu.cbu_dsap = LLC_8021D_LSAP;
267 	bpdu.cbu_ctl = LLC_UI;
268 	bpdu.cbu_protoid = htons(0);
269 	bpdu.cbu_protover = 0;
270 	bpdu.cbu_bpdutype = cu->cu_message_type;
271 	bpdu.cbu_flags = (cu->cu_topology_change ? BSTP_FLAG_TC : 0) |
272 	    (cu->cu_topology_change_acknowledgment ? BSTP_FLAG_TCA : 0);
273 
274 	bpdu.cbu_rootpri = htons(cu->cu_rootid >> 48);
275 	bpdu.cbu_rootaddr[0] = cu->cu_rootid >> 40;
276 	bpdu.cbu_rootaddr[1] = cu->cu_rootid >> 32;
277 	bpdu.cbu_rootaddr[2] = cu->cu_rootid >> 24;
278 	bpdu.cbu_rootaddr[3] = cu->cu_rootid >> 16;
279 	bpdu.cbu_rootaddr[4] = cu->cu_rootid >> 8;
280 	bpdu.cbu_rootaddr[5] = cu->cu_rootid >> 0;
281 
282 	bpdu.cbu_rootpathcost = htonl(cu->cu_root_path_cost);
283 
284 	bpdu.cbu_bridgepri = htons(cu->cu_bridge_id >> 48);
285 	bpdu.cbu_bridgeaddr[0] = cu->cu_bridge_id >> 40;
286 	bpdu.cbu_bridgeaddr[1] = cu->cu_bridge_id >> 32;
287 	bpdu.cbu_bridgeaddr[2] = cu->cu_bridge_id >> 24;
288 	bpdu.cbu_bridgeaddr[3] = cu->cu_bridge_id >> 16;
289 	bpdu.cbu_bridgeaddr[4] = cu->cu_bridge_id >> 8;
290 	bpdu.cbu_bridgeaddr[5] = cu->cu_bridge_id >> 0;
291 
292 	bpdu.cbu_portid = htons(cu->cu_port_id);
293 	bpdu.cbu_messageage = htons(cu->cu_message_age);
294 	bpdu.cbu_maxage = htons(cu->cu_max_age);
295 	bpdu.cbu_hellotime = htons(cu->cu_hello_time);
296 	bpdu.cbu_forwarddelay = htons(cu->cu_forward_delay);
297 
298 	/*
299 	 * Packets sent from the bridge always use the bridge MAC
300 	 * as the source.
301 	 */
302 	memcpy(eh->ether_shost, IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
303 	memcpy(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN);
304 	eh->ether_type = htons(sizeof(bpdu));
305 
306 	memcpy(mtod(m, caddr_t) + sizeof(*eh), &bpdu, sizeof(bpdu));
307 
308 	bridge_enqueue(ifp, m);
309 }
310 
311 static int
312 bstp_root_bridge(struct bridge_softc *sc)
313 {
314 	return (sc->sc_designated_root == sc->sc_bridge_id);
315 }
316 
317 /*
318  * Returns TRUE if the recorded information from our peer has a shorter
319  * graph distance than our current best.
320  */
321 int
322 bstp_supersedes_port_info(struct bridge_softc *sc, struct bridge_iflist *bif)
323 {
324 	if (bif->bif_peer_root < sc->sc_designated_root)
325 		return (1);
326 	if (bif->bif_peer_root > sc->sc_designated_root)
327 		return (0);
328 
329 	/*
330 	 * bif_peer_cost and sc_designated_cost do not include the local
331 	 * bif_path_cost, otherwise we would not be able to make this
332 	 * comparison.
333 	 */
334 	if (bif->bif_peer_cost < sc->sc_designated_cost)
335 		return (1);
336 	if (bif->bif_peer_cost > sc->sc_designated_cost)
337 		return (0);
338 
339 	/*
340 	 * When the peer costs match we have to check the local path
341 	 * cost against the selected root port.  sc_designated_cost
342 	 * in an aggregation and cannot be used.
343 	 */
344 	if (sc->sc_root_port &&
345 	    bif->bif_path_cost < sc->sc_root_port->bif_path_cost)
346 		return (1);
347 	if (sc->sc_root_port &&
348 	    bif->bif_path_cost > sc->sc_root_port->bif_path_cost)
349 		return (0);
350 
351 	/*
352 	 * If the path costs are identical the bridge with the lowest
353 	 * bridge_id wins.
354 	 */
355 	if (bif->bif_peer_bridge < sc->sc_designated_bridge)
356 		return (1);
357 	if (bif->bif_peer_bridge > sc->sc_designated_bridge)
358 		return (0);
359 
360 	/*
361 	 * bridge_id or bridge+port collision w/peer returns TRUE.
362 	 *
363 	 * This case can also occur when two bridges are connected
364 	 * via multiple links whos path costs have not been set.
365 	 */
366 	if (bif->bif_peer_bridge != sc->sc_bridge_id)
367 		return (1);
368 	if (bif->bif_peer_port <= sc->sc_designated_port)
369 		return (1);
370 	return (0);
371 }
372 
373 /*
374  * The shorter graph distance represented by cu (most of which is also
375  * already stored in our bif_peer_* fields) becomes the designated info.
376  *
377  * NOTE: sc_designated_cost does not include bif_path_cost, it is added
378  *	 in later on a port-by-port basis as needed.
379  */
380 static void
381 bstp_record_config_information(struct bridge_softc *sc,
382 			       struct bridge_iflist *bif,
383 			       struct bstp_config_unit *cu)
384 {
385 	sc->sc_designated_root = bif->bif_peer_root;
386 	sc->sc_designated_cost = bif->bif_peer_cost;
387 	sc->sc_designated_bridge = bif->bif_peer_bridge;
388 	sc->sc_designated_port = bif->bif_peer_port;
389 	bstp_timer_start(&bif->bif_message_age_timer, cu->cu_message_age);
390 }
391 
392 static void
393 bstp_record_config_timeout_values(struct bridge_softc *sc,
394 				  struct bstp_config_unit *config)
395 {
396 	sc->sc_max_age = config->cu_max_age;
397 	sc->sc_hello_time = config->cu_hello_time;
398 	sc->sc_forward_delay = config->cu_forward_delay;
399 	sc->sc_topology_change = config->cu_topology_change;
400 }
401 
402 static void
403 bstp_config_bpdu_generation(struct bridge_softc *sc)
404 {
405 	struct bridge_iflist *bif, *nbif;
406 
407 	TAILQ_FOREACH_MUTABLE(bif, &sc->sc_iflists[mycpuid], bif_next, nbif) {
408 		if ((bif->bif_flags & IFBIF_STP) == 0)
409 			continue;
410 		if (bif->bif_state != BSTP_IFSTATE_DISABLED &&
411 		    ((sc->sc_ifp->if_flags & IFF_LINK1) ||
412 		     (bif->bif_flags & IFBIF_DESIGNATED))) {
413 			bstp_transmit_config(sc, bif);
414 		}
415 
416 		if (nbif != NULL && !nbif->bif_onlist) {
417 			KKASSERT(bif->bif_onlist);
418 			nbif = TAILQ_NEXT(bif, bif_next);
419 		}
420 	}
421 }
422 
423 static void
424 bstp_transmit_tcn(struct bridge_softc *sc)
425 {
426 	struct bstp_tbpdu bpdu;
427 	struct ifnet *ifp;
428 	struct ether_header *eh;
429 	struct mbuf *m;
430 
431 	if (sc->sc_root_port == NULL)	/* all iterfaces disabled */
432 		return;
433 
434 	ifp = sc->sc_root_port->bif_ifp;
435 	if ((ifp->if_flags & IFF_RUNNING) == 0)
436 		return;
437 
438 	MGETHDR(m, M_NOWAIT, MT_DATA);
439 	if (m == NULL)
440 		return;
441 
442 	m->m_pkthdr.rcvif = ifp;
443 	m->m_pkthdr.len = sizeof(*eh) + sizeof(bpdu);
444 	m->m_len = m->m_pkthdr.len;
445 
446 	eh = mtod(m, struct ether_header *);
447 
448 	/*
449 	 * Packets sent from the bridge always use the bridge MAC
450 	 * as the source.
451 	 */
452 	memcpy(eh->ether_shost, IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
453 	memcpy(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN);
454 	eh->ether_type = htons(sizeof(bpdu));
455 
456 	bpdu.tbu_ssap = bpdu.tbu_dsap = LLC_8021D_LSAP;
457 	bpdu.tbu_ctl = LLC_UI;
458 	bpdu.tbu_protoid = 0;
459 	bpdu.tbu_protover = 0;
460 	bpdu.tbu_bpdutype = BSTP_MSGTYPE_TCN;
461 
462 	memcpy(mtod(m, caddr_t) + sizeof(*eh), &bpdu, sizeof(bpdu));
463 
464 	bridge_enqueue(ifp, m);
465 }
466 
467 /*
468  * Recalculate sc->sc_designated* and sc->sc_root_port (if our bridge
469  * is calculated to be the root bridge).  We do this by initializing
470  * the designated variables to point at us and then scan our peers.
471  * Any uninitialized peers will have a max-value root.
472  *
473  * Clear IFBIF_DESIGNATED on any ports which no longer match the criteria
474  * required to be a designated port.  Only aged out ports and the root
475  * port can be designated.
476  *
477  * If we win we do a second scan to determine which port on our bridge
478  * is the best.
479  */
480 static void
481 bstp_configuration_update(struct bridge_softc *sc)
482 {
483 	uint64_t	designated_root = sc->sc_bridge_id;
484 	uint64_t	designated_bridge = sc->sc_bridge_id;
485 	uint32_t	designated_cost = 0xFFFFFFFFU;
486 	uint32_t	designated_root_cost = 0xFFFFFFFFU;
487 	uint16_t	designated_port = 65535;
488 	struct bridge_iflist *root_port = NULL;
489 	struct bridge_iflist *bif;
490 
491 	/*
492 	 * Resolve information from our peers.  Aged peers will have
493 	 * a maxed bif_peer_root and not come under consideration.
494 	 */
495 	TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
496 		if ((bif->bif_flags & IFBIF_STP) == 0)
497 			continue;
498 		if (bif->bif_state == BSTP_IFSTATE_DISABLED ||
499 		    bif->bif_state == BSTP_IFSTATE_L1BLOCKING) {
500 			continue;
501 		}
502 
503 		if (bif->bif_peer_root > designated_root)
504 			continue;
505 		if (bif->bif_peer_root < designated_root)
506 			goto set_port;
507 
508 		/*
509 		 * NOTE: The designated_cost temporary variable already added
510 		 *	 in the path code of the related root port.
511 		 */
512 		if (bif->bif_peer_cost + bif->bif_path_cost > designated_cost)
513 			continue;
514 		if (bif->bif_peer_cost + bif->bif_path_cost < designated_cost)
515 			goto set_port;
516 
517 		if (bif->bif_peer_bridge > designated_bridge)
518 			continue;
519 		if (bif->bif_peer_bridge < designated_bridge)
520 			goto set_port;
521 
522 		if (bif->bif_peer_port > designated_port)
523 			continue;
524 		if (bif->bif_peer_port < designated_port)
525 			goto set_port;
526 
527 		/*
528 		 * Same root, path cost, bridge, and port.  Set the root
529 		 * only if we do not already have it.
530 		 */
531 		if (root_port)
532 			continue;
533 
534 		/*
535 		 * New root port (from peers)
536 		 *
537 		 * NOTE: Temporarily add bif_path_cost into the designated
538 		 *	 cost to reduce complexity in the loop, it will be
539 		 *	 subtracted out when we are done.
540 		 */
541 set_port:
542 		designated_root = bif->bif_peer_root;
543 		designated_cost = bif->bif_peer_cost + bif->bif_path_cost;
544 		designated_root_cost = bif->bif_peer_cost;
545 		designated_bridge = bif->bif_peer_bridge;
546 		designated_port = bif->bif_peer_port;
547 		root_port = bif;
548 	}
549 
550 	/*
551 	 * root_port will be NULL at the start here if all of our
552 	 * peers are aged or are not as good a root as our bridge would
553 	 * be.  It can also be NULL due to all related links being
554 	 * disabled.
555 	 *
556 	 * If the root winds up being our bridge scan again against local
557 	 * information.  Unconditionally update IFBIF_DESIGNATED.
558 	 */
559 	TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
560 		bif->bif_flags &= ~(IFBIF_DESIGNATED | IFBIF_ROOT);
561 		if ((bif->bif_flags & IFBIF_STP) == 0)
562 			continue;
563 		if (bif->bif_state == BSTP_IFSTATE_DISABLED ||
564 		    bif->bif_state == BSTP_IFSTATE_L1BLOCKING) {
565 			continue;
566 		}
567 
568 		/*
569 		 * Set DESIGNATED for an aged or unknown peer.
570 		 */
571 		if (bif->bif_peer_bridge == 0xFFFFFFFFFFFFFFFFLLU)
572 			bif->bif_flags |= IFBIF_DESIGNATED;
573 		if (designated_root != sc->sc_bridge_id)
574 			continue;
575 
576 		/*
577 		 * This is only reached if our bridge is the root bridge,
578 		 * select the root port (IFBIF_DESIGNATED is set at the
579 		 * end).
580 		 *
581 		 * The peer cost will incorporate the peer bridge's best
582 		 * cost to the root bridge (us), which is NOT necessarily
583 		 * our path cost, plus the peer bridge's own path cost.
584 		 * Because of this we must still add in our path cost for
585 		 * further differentiation.  Because the peer's 'best' cost
586 		 * is applied to all of its outgoing messages we should
587 		 * still come to the same conclusion as the peer.
588 		 */
589 		if (bif->bif_peer_cost + bif->bif_path_cost > designated_cost)
590 			continue;
591 		if (bif->bif_peer_cost + bif->bif_path_cost < designated_cost)
592 			goto set_port2;
593 
594 		if (bif->bif_port_id > designated_port)
595 			continue;
596 		if (bif->bif_port_id < designated_port)
597 			goto set_port2;
598 		/* degenerate case (possible peer collision w/our key */
599 
600 		/*
601 		 * New port.  Since we are the root, the root cost is always
602 		 * 0.  The peer's cu_root_path_cost does not necessarily
603 		 * incorporate our peer path cost, but instead incorporates
604 		 * the lowest path cost to root for the target bridge
605 		 * when multiple links are present between the two bridges.
606 		 * Thus for selection purposes we must still add in our
607 		 * local path cost.
608 		 */
609 set_port2:
610 		designated_cost = bif->bif_peer_cost + bif->bif_path_cost;
611 		designated_root_cost = 0;
612 		designated_bridge = sc->sc_bridge_id;
613 		designated_port = bif->bif_port_id;
614 		root_port = bif;
615 	}
616 
617 	/*
618 	 * Update aggregate information.  The selected root port always
619 	 * becomes a designated port (along with aged ports).  This can
620 	 * either be the port whos peer is closest to the root or it
621 	 * can be one of our ports if our bridge is the root.
622 	 *
623 	 * The root cost we record in sc_designated_root does not include
624 	 * bif_path_cost of the root port, since we may transmit this
625 	 * out of any port we will add the cost back in later on on
626 	 * a per-port basis.
627 	 *
628 	 * root_port can be NULL here (if all links are disabled)
629 	 */
630 	if (root_port) {
631 		sc->sc_designated_root = designated_root;
632 		sc->sc_designated_cost = designated_root_cost;
633 		sc->sc_designated_bridge = designated_bridge;
634 		sc->sc_designated_port = designated_port;
635 		root_port->bif_flags |= IFBIF_DESIGNATED | IFBIF_ROOT;
636 	} else {
637 		sc->sc_designated_root = designated_root;
638 		sc->sc_designated_cost = designated_cost;
639 		sc->sc_designated_bridge = designated_bridge;
640 		sc->sc_designated_port = designated_port;
641 	}
642 	sc->sc_root_port = root_port;
643 }
644 
645 /*
646  * Calculate the desired state for each interface link on our bridge.
647  *
648  * The best port will match against sc->sc_root_port (whether we are root
649  * or whether that port is the closest to the root).  We push this port
650  * towards a FORWARDING state.
651  *
652  * Next come designated ports, either aged ports or ports with no peer info
653  * (yet), or the peer who is closest to the root. We push this port towards
654  * a FORWARDING state as well.
655  *
656  * Any remaining ports are pushed towards a BLOCKING state.  Both sides of
657  * the port (us and our peer) should wind up placing the two ends in this
658  * state or bad things happen.
659  */
660 static void
661 bstp_port_state_selection(struct bridge_softc *sc)
662 {
663 	struct bridge_iflist *bif, *nbif;
664 
665 	TAILQ_FOREACH_MUTABLE(bif, &sc->sc_iflists[mycpuid], bif_next, nbif) {
666 		if ((bif->bif_flags & IFBIF_STP) == 0)
667 			continue;
668 		if (sc->sc_root_port &&
669 		    bif->bif_info == sc->sc_root_port->bif_info) {
670 			bif->bif_config_pending = 0;
671 			bif->bif_topology_change_acknowledge = 0;
672 			bstp_make_forwarding(sc, bif);
673 		} else if (bif->bif_flags & IFBIF_DESIGNATED) {
674 			bstp_timer_stop(&bif->bif_message_age_timer);
675 			bstp_make_forwarding(sc, bif);
676 		} else {
677 			bif->bif_config_pending = 0;
678 			bif->bif_topology_change_acknowledge = 0;
679 			bstp_make_blocking(sc, bif);
680 		}
681 
682 		if (nbif != NULL && !nbif->bif_onlist) {
683 			KKASSERT(bif->bif_onlist);
684 			nbif = TAILQ_NEXT(bif, bif_next);
685 		}
686 	}
687 }
688 
689 /*
690  * Clear peer info, effectively makes the port looked aged out.
691  * It becomes a designated go-to port.
692  */
693 static void
694 bstp_clear_peer_info(struct bridge_softc *sc, struct bridge_iflist *bif)
695 {
696 	bif->bif_peer_root = 0xFFFFFFFFFFFFFFFFLLU;
697 	bif->bif_peer_cost = 0xFFFFFFFFU;
698 	bif->bif_peer_bridge = 0xFFFFFFFFFFFFFFFFLLU;
699 	bif->bif_peer_port = 0xFFFFU;
700 
701 	if (bif->bif_state != BSTP_IFSTATE_DISABLED &&
702 	    bif->bif_state != BSTP_IFSTATE_L1BLOCKING) {
703 		bif->bif_flags |= IFBIF_DESIGNATED;
704 	}
705 }
706 
707 static void
708 bstp_make_forwarding(struct bridge_softc *sc, struct bridge_iflist *bif)
709 {
710 	if (bif->bif_state == BSTP_IFSTATE_BLOCKING ||
711 	    bif->bif_state == BSTP_IFSTATE_BONDED) {
712 		bstp_set_port_state(bif, BSTP_IFSTATE_LISTENING);
713 		bstp_timer_start(&bif->bif_forward_delay_timer, 0);
714 	}
715 }
716 
717 static void
718 bstp_make_blocking(struct bridge_softc *sc, struct bridge_iflist *bif)
719 {
720 	if (bif->bif_state != BSTP_IFSTATE_DISABLED &&
721 	    bif->bif_state != BSTP_IFSTATE_BLOCKING &&
722 	    bif->bif_state != BSTP_IFSTATE_BONDED &&
723 	    bif->bif_state != BSTP_IFSTATE_L1BLOCKING) {
724 		if ((bif->bif_state == BSTP_IFSTATE_FORWARDING) ||
725 		    (bif->bif_state == BSTP_IFSTATE_LEARNING)) {
726 			if (bif->bif_change_detection_enabled) {
727 				bstp_topology_change_detection(sc);
728 			}
729 		}
730 		bstp_set_port_state(bif, BSTP_IFSTATE_BLOCKING);
731 		bridge_rtdelete(sc, bif->bif_ifp, IFBF_FLUSHDYN);
732 		bstp_timer_stop(&bif->bif_forward_delay_timer);
733 		if (sc->sc_ifp->if_flags & IFF_LINK2)
734 			bstp_adjust_bonded_states(sc, bif);
735 	}
736 }
737 
738 static void
739 bstp_make_l1blocking(struct bridge_softc *sc, struct bridge_iflist *bif)
740 {
741 	int was_forwarding = (bif->bif_state == BSTP_IFSTATE_FORWARDING);
742 
743 	switch(bif->bif_state) {
744 	case BSTP_IFSTATE_LISTENING:
745 	case BSTP_IFSTATE_LEARNING:
746 	case BSTP_IFSTATE_FORWARDING:
747 	case BSTP_IFSTATE_BLOCKING:
748 	case BSTP_IFSTATE_BONDED:
749 		bstp_set_port_state(bif, BSTP_IFSTATE_L1BLOCKING);
750 		bridge_rtdelete(sc, bif->bif_ifp, IFBF_FLUSHDYN);
751 		bstp_timer_stop(&bif->bif_forward_delay_timer);
752 		bstp_timer_stop(&bif->bif_link1_timer);
753 		if (bif->bif_flags & IFBIF_DESIGNATED) {
754 			bif->bif_flags &= ~IFBIF_DESIGNATED;
755 			bstp_configuration_update(sc);
756 			bstp_port_state_selection(sc);
757 		}
758 		if (was_forwarding && (sc->sc_ifp->if_flags & IFF_LINK2))
759 			bstp_adjust_bonded_states(sc, bif);
760 		break;
761 	default:
762 		break;
763 	}
764 }
765 
766 /*
767  * Member (bif) changes to or from a FORWARDING state.  All members in the
768  * same bonding group which are in a BLOCKING or BONDED state must be set
769  * to either BLOCKING or BONDED based on whether any members in the bonding
770  * group remain in the FORWARDING state.
771  *
772  * Going between the BLOCKING and BONDED states does not require a
773  * configuration update.
774  */
775 static void
776 bstp_adjust_bonded_states(struct bridge_softc *sc, struct bridge_iflist *obif)
777 {
778 	struct bridge_iflist *bif;
779 	int state = BSTP_IFSTATE_BLOCKING;
780 
781 	TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
782 		if ((bif->bif_flags & IFBIF_STP) == 0)
783 			continue;
784 		if (bif->bif_state != BSTP_IFSTATE_FORWARDING)
785 			continue;
786 		if (memcmp(IF_LLADDR(bif->bif_ifp), IF_LLADDR(obif->bif_ifp),
787 			   ETHER_ADDR_LEN) != 0) {
788 			continue;
789 		}
790 		state = BSTP_IFSTATE_BONDED;
791 		break;
792 	}
793 	TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
794 		if ((bif->bif_flags & IFBIF_STP) == 0)
795 			continue;
796 		if (bif->bif_state != BSTP_IFSTATE_BLOCKING &&
797 		    bif->bif_state != BSTP_IFSTATE_BONDED) {
798 			continue;
799 		}
800 		if (memcmp(IF_LLADDR(bif->bif_ifp), IF_LLADDR(obif->bif_ifp),
801 			   ETHER_ADDR_LEN) != 0) {
802 			continue;
803 		}
804 		if (bif->bif_bond_weight == 0)
805 			bif->bif_state = BSTP_IFSTATE_BLOCKING;
806 		else
807 			bif->bif_state = state;
808 	}
809 }
810 
811 static void
812 bstp_set_port_state(struct bridge_iflist *bif, uint8_t state)
813 {
814 	bif->bif_state = state;
815 }
816 
817 static void
818 bstp_topology_change_detection(struct bridge_softc *sc)
819 {
820 	if (bstp_root_bridge(sc)) {
821 		sc->sc_topology_change = 1;
822 		bstp_timer_start(&sc->sc_topology_change_timer, 0);
823 	} else if (!sc->sc_topology_change_detected) {
824 		bstp_transmit_tcn(sc);
825 		bstp_timer_start(&sc->sc_tcn_timer, 0);
826 	}
827 	sc->sc_topology_change_detected = 1;
828 }
829 
830 static void
831 bstp_topology_change_acknowledged(struct bridge_softc *sc)
832 {
833 	sc->sc_topology_change_detected = 0;
834 	bstp_timer_stop(&sc->sc_tcn_timer);
835 }
836 
837 static void
838 bstp_acknowledge_topology_change(struct bridge_softc *sc,
839 				 struct bridge_iflist *bif)
840 {
841 	bif->bif_topology_change_acknowledge = 1;
842 	bstp_transmit_config(sc, bif);
843 }
844 
845 void
846 bstp_input(struct bridge_softc *sc, struct bridge_iflist *bif, struct mbuf *m)
847 {
848 	struct ether_header *eh;
849 	struct bstp_tbpdu tpdu;
850 	struct bstp_cbpdu cpdu;
851 	struct bstp_config_unit cu;
852 	struct bstp_tcn_unit tu;
853 	uint16_t len;
854 
855 	if ((bif->bif_flags & IFBIF_STP) == 0)
856 		goto out;
857 
858 	/*
859 	 * The L1BLOCKING (ping pong failover) test needs to reset the
860 	 * timer if LINK1 is active.
861 	 */
862 	if (bif->bif_state == BSTP_IFSTATE_L1BLOCKING) {
863 		bstp_set_port_state(bif, BSTP_IFSTATE_BLOCKING);
864 		if (sc->sc_ifp->if_flags & IFF_LINK1)
865 			bstp_timer_start(&bif->bif_link1_timer, 0);
866 		bstp_make_forwarding(sc, bif);
867 	} else if (sc->sc_ifp->if_flags & IFF_LINK1) {
868 		bstp_timer_start(&bif->bif_link1_timer, 0);
869 	}
870 
871 	eh = mtod(m, struct ether_header *);
872 
873 	len = ntohs(eh->ether_type);
874 	if (len < sizeof(tpdu))
875 		goto out;
876 
877 	m_adj(m, ETHER_HDR_LEN);
878 
879 	if (m->m_pkthdr.len > len)
880 		m_adj(m, len - m->m_pkthdr.len);
881 	if (m->m_len < sizeof(tpdu) &&
882 	    (m = m_pullup(m, sizeof(tpdu))) == NULL)
883 		goto out;
884 
885 	memcpy(&tpdu, mtod(m, caddr_t), sizeof(tpdu));
886 
887 	if (tpdu.tbu_dsap != LLC_8021D_LSAP ||
888 	    tpdu.tbu_ssap != LLC_8021D_LSAP ||
889 	    tpdu.tbu_ctl != LLC_UI)
890 		goto out;
891 	if (tpdu.tbu_protoid != 0 || tpdu.tbu_protover != 0)
892 		goto out;
893 
894 	switch (tpdu.tbu_bpdutype) {
895 	case BSTP_MSGTYPE_TCN:
896 		tu.tu_message_type = tpdu.tbu_bpdutype;
897 		bstp_received_tcn_bpdu(sc, bif, &tu);
898 		break;
899 	case BSTP_MSGTYPE_CFG:
900 		if (m->m_len < sizeof(cpdu) &&
901 		    (m = m_pullup(m, sizeof(cpdu))) == NULL)
902 			goto out;
903 		memcpy(&cpdu, mtod(m, caddr_t), sizeof(cpdu));
904 
905 		cu.cu_rootid =
906 		    (((uint64_t)ntohs(cpdu.cbu_rootpri)) << 48) |
907 		    (((uint64_t)cpdu.cbu_rootaddr[0]) << 40) |
908 		    (((uint64_t)cpdu.cbu_rootaddr[1]) << 32) |
909 		    (((uint64_t)cpdu.cbu_rootaddr[2]) << 24) |
910 		    (((uint64_t)cpdu.cbu_rootaddr[3]) << 16) |
911 		    (((uint64_t)cpdu.cbu_rootaddr[4]) << 8) |
912 		    (((uint64_t)cpdu.cbu_rootaddr[5]) << 0);
913 
914 		cu.cu_bridge_id =
915 		    (((uint64_t)ntohs(cpdu.cbu_bridgepri)) << 48) |
916 		    (((uint64_t)cpdu.cbu_bridgeaddr[0]) << 40) |
917 		    (((uint64_t)cpdu.cbu_bridgeaddr[1]) << 32) |
918 		    (((uint64_t)cpdu.cbu_bridgeaddr[2]) << 24) |
919 		    (((uint64_t)cpdu.cbu_bridgeaddr[3]) << 16) |
920 		    (((uint64_t)cpdu.cbu_bridgeaddr[4]) << 8) |
921 		    (((uint64_t)cpdu.cbu_bridgeaddr[5]) << 0);
922 
923 		cu.cu_root_path_cost = ntohl(cpdu.cbu_rootpathcost);
924 		cu.cu_message_age = ntohs(cpdu.cbu_messageage);
925 		cu.cu_max_age = ntohs(cpdu.cbu_maxage);
926 		cu.cu_hello_time = ntohs(cpdu.cbu_hellotime);
927 		cu.cu_forward_delay = ntohs(cpdu.cbu_forwarddelay);
928 		cu.cu_port_id = ntohs(cpdu.cbu_portid);
929 		cu.cu_message_type = cpdu.cbu_bpdutype;
930 		cu.cu_topology_change_acknowledgment =
931 		    (cpdu.cbu_flags & BSTP_FLAG_TCA) ? 1 : 0;
932 		cu.cu_topology_change =
933 		    (cpdu.cbu_flags & BSTP_FLAG_TC) ? 1 : 0;
934 		bstp_received_config_bpdu(sc, bif, &cu);
935 		break;
936 	default:
937 		goto out;
938 	}
939 out:
940 	if (m)
941 		m_freem(m);
942 }
943 
944 static void
945 bstp_received_config_bpdu(struct bridge_softc *sc, struct bridge_iflist *bif,
946 			  struct bstp_config_unit *cu)
947 {
948 	int iamroot;
949 
950 	iamroot = bstp_root_bridge(sc);
951 
952 	if (bif->bif_state != BSTP_IFSTATE_DISABLED) {
953 		/*
954 		 * Record information from peer.  The peer_cost field
955 		 * does not include the local bif->bif_path_cost, it will
956 		 * be added in as needed (since it can be modified manually
957 		 * this way we don't have to worry about fixups).
958 		 */
959 		bif->bif_peer_root = cu->cu_rootid;
960 		bif->bif_peer_cost = cu->cu_root_path_cost;
961 		bif->bif_peer_bridge = cu->cu_bridge_id;
962 		bif->bif_peer_port = cu->cu_port_id;
963 
964 		if (bstp_supersedes_port_info(sc, bif)) {
965 			bstp_record_config_information(sc, bif, cu);
966 			bstp_configuration_update(sc);
967 			bstp_port_state_selection(sc);
968 
969 			/*
970 			 * If our bridge loses its root status (?)
971 			 *
972 			 * Hello's (unsolicited CFG packets) are generated
973 			 * every hello period of LINK1 is set, otherwise
974 			 * we are no longer the root bridge and must stop
975 			 * generating unsolicited CFG packets.
976 			 */
977 			if (iamroot && bstp_root_bridge(sc) == 0) {
978 				if ((sc->sc_ifp->if_flags & IFF_LINK1) == 0)
979 					bstp_timer_stop(&sc->sc_hello_timer);
980 
981 				if (sc->sc_topology_change_detected) {
982 					bstp_timer_stop(
983 					    &sc->sc_topology_change_timer);
984 					bstp_transmit_tcn(sc);
985 					bstp_timer_start(&sc->sc_tcn_timer, 0);
986 				}
987 			}
988 
989 			if (sc->sc_root_port &&
990 			    bif->bif_info == sc->sc_root_port->bif_info) {
991 				bstp_record_config_timeout_values(sc, cu);
992 				bstp_config_bpdu_generation(sc);
993 
994 				if (cu->cu_topology_change_acknowledgment)
995 					bstp_topology_change_acknowledged(sc);
996 			}
997 		} else if (bif->bif_flags & IFBIF_DESIGNATED) {
998 			/*
999 			 * Update designated ports (aged out peers or
1000 			 * the port closest to the root) at a faster pace.
1001 			 *
1002 			 * Clear our designated flag if we aren't marked
1003 			 * as the root port.
1004 			 */
1005 			bstp_transmit_config(sc, bif);
1006 			if ((bif->bif_flags & IFBIF_ROOT) == 0) {
1007 				bif->bif_flags &= ~IFBIF_DESIGNATED;
1008 				bstp_configuration_update(sc);
1009 				bstp_port_state_selection(sc);
1010 			}
1011 		}
1012 	}
1013 }
1014 
1015 static void
1016 bstp_received_tcn_bpdu(struct bridge_softc *sc, struct bridge_iflist *bif,
1017 		       struct bstp_tcn_unit *tcn)
1018 {
1019 	if (bif->bif_state != BSTP_IFSTATE_DISABLED &&
1020 	    (bif->bif_flags & IFBIF_DESIGNATED)) {
1021 		bstp_topology_change_detection(sc);
1022 		bstp_acknowledge_topology_change(sc, bif);
1023 	}
1024 }
1025 
1026 /*
1027  * link1 forces continuous hello's (the bridge interface must be cycled
1028  * to start them up), so keep the timer hot if that is the case, otherwise
1029  * only send HELLO's if we are the root.
1030  */
1031 static void
1032 bstp_hello_timer_expiry(struct bridge_softc *sc)
1033 {
1034 	bstp_config_bpdu_generation(sc);
1035 
1036 	if ((sc->sc_ifp->if_flags & IFF_LINK1) || bstp_root_bridge(sc))
1037 		bstp_timer_start(&sc->sc_hello_timer, 0);
1038 }
1039 
1040 static void
1041 bstp_message_age_timer_expiry(struct bridge_softc *sc,
1042 			      struct bridge_iflist *bif)
1043 {
1044 	int iamroot;
1045 
1046 	iamroot = bstp_root_bridge(sc);
1047 	bstp_clear_peer_info(sc, bif);
1048 	bstp_configuration_update(sc);
1049 	bstp_port_state_selection(sc);
1050 
1051 	/*
1052 	 * If we've become the root and were not the root before
1053 	 * we have some cleanup to do.  This also occurs if we
1054 	 * wind up being completely isolated.
1055 	 */
1056 	if (iamroot == 0 && bstp_root_bridge(sc)) {
1057 		sc->sc_max_age = sc->sc_bridge_max_age;
1058 		sc->sc_hello_time = sc->sc_bridge_hello_time;
1059 		sc->sc_forward_delay = sc->sc_bridge_forward_delay;
1060 
1061 		bstp_topology_change_detection(sc);
1062 		bstp_timer_stop(&sc->sc_tcn_timer);
1063 		bstp_config_bpdu_generation(sc);
1064 		bstp_timer_start(&sc->sc_hello_timer, 0);
1065 	}
1066 }
1067 
1068 static void
1069 bstp_forward_delay_timer_expiry(struct bridge_softc *sc,
1070 				struct bridge_iflist *bif)
1071 {
1072 	if (bif->bif_state == BSTP_IFSTATE_LISTENING) {
1073 		bstp_set_port_state(bif, BSTP_IFSTATE_LEARNING);
1074 		bstp_timer_start(&bif->bif_forward_delay_timer, 0);
1075 	} else if (bif->bif_state == BSTP_IFSTATE_LEARNING) {
1076 		bstp_set_port_state(bif, BSTP_IFSTATE_FORWARDING);
1077 		if (sc->sc_designated_bridge == sc->sc_bridge_id &&
1078 		    bif->bif_change_detection_enabled) {
1079 			bstp_topology_change_detection(sc);
1080 		}
1081 		bstp_configuration_update(sc);
1082 		bstp_port_state_selection(sc);
1083 		if (sc->sc_ifp->if_flags & IFF_LINK2)
1084 			bstp_adjust_bonded_states(sc, bif);
1085 	}
1086 }
1087 
1088 static void
1089 bstp_tcn_timer_expiry(struct bridge_softc *sc)
1090 {
1091 	bstp_transmit_tcn(sc);
1092 	bstp_timer_start(&sc->sc_tcn_timer, 0);
1093 }
1094 
1095 static void
1096 bstp_topology_change_timer_expiry(struct bridge_softc *sc)
1097 {
1098 	sc->sc_topology_change_detected = 0;
1099 	sc->sc_topology_change = 0;
1100 }
1101 
1102 static void
1103 bstp_hold_timer_expiry(struct bridge_softc *sc, struct bridge_iflist *bif)
1104 {
1105 	if (bif->bif_config_pending)
1106 		bstp_transmit_config(sc, bif);
1107 }
1108 
1109 /*
1110  * If no traffic received directly on this port for the specified
1111  * period with link1 set we go into a special blocking mode to
1112  * fail-over traffic to another port.
1113  */
1114 static void
1115 bstp_link1_timer_expiry(struct bridge_softc *sc, struct bridge_iflist *bif)
1116 {
1117 	if (sc->sc_ifp->if_flags & IFF_LINK1)
1118 		bstp_make_l1blocking(sc, bif);
1119 }
1120 
1121 static int
1122 bstp_addr_cmp(const uint8_t *a, const uint8_t *b)
1123 {
1124 	int i, d;
1125 
1126 	for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
1127 		d = ((int)a[i]) - ((int)b[i]);
1128 	}
1129 
1130 	return (d);
1131 }
1132 
1133 void
1134 bstp_initialization(struct bridge_softc *sc)
1135 {
1136 	struct bridge_iflist *bif, *mif, *nbif;
1137 	u_char *e_addr;
1138 
1139 	KKASSERT(&curthread->td_msgport == BRIDGE_CFGPORT);
1140 
1141 	/*
1142 	 * Figure out our bridge ID, use the lowest-valued MAC.
1143 	 * Include the bridge's own random MAC in the calculation.
1144 	 */
1145 	mif = NULL;
1146 
1147 	TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
1148 		if ((bif->bif_flags & IFBIF_STP) == 0)
1149 			continue;
1150 		if (bif->bif_ifp->if_type != IFT_ETHER)
1151 			continue;
1152 		if (mif == NULL) {
1153 			mif = bif;
1154 			continue;
1155 		}
1156 
1157 		bif->bif_port_id = (bif->bif_priority << 8) |
1158 				   (bif->bif_ifp->if_index & 0xff);
1159 		if (bstp_addr_cmp(IF_LLADDR(bif->bif_ifp),
1160 				  IF_LLADDR(mif->bif_ifp)) < 0) {
1161 			mif = bif;
1162 			continue;
1163 		}
1164 	}
1165 	if (mif == NULL) {
1166 		bstp_stop(sc);
1167 		return;
1168 	}
1169 
1170 	if (bstp_addr_cmp(IF_LLADDR(sc->sc_ifp), IF_LLADDR(mif->bif_ifp)) < 0)
1171 		e_addr = IF_LLADDR(sc->sc_ifp);
1172 	else
1173 		e_addr = IF_LLADDR(mif->bif_ifp);
1174 
1175 	sc->sc_bridge_id =
1176 	    (((uint64_t)sc->sc_bridge_priority) << 48) |
1177 	    (((uint64_t)e_addr[0]) << 40) |
1178 	    (((uint64_t)e_addr[1]) << 32) |
1179 	    (((uint64_t)e_addr[2]) << 24) |
1180 	    (((uint64_t)e_addr[3]) << 16) |
1181 	    (((uint64_t)e_addr[4]) << 8) |
1182 	    (((uint64_t)e_addr[5]));
1183 
1184 	/*
1185 	 * Remainder of setup.
1186 	 */
1187 
1188 	sc->sc_designated_root = sc->sc_bridge_id;
1189 	sc->sc_designated_cost = 0;
1190 	sc->sc_root_port = NULL;
1191 
1192 	sc->sc_max_age = sc->sc_bridge_max_age;
1193 	sc->sc_hello_time = sc->sc_bridge_hello_time;
1194 	sc->sc_forward_delay = sc->sc_bridge_forward_delay;
1195 	sc->sc_topology_change_detected = 0;
1196 	sc->sc_topology_change = 0;
1197 	bstp_timer_stop(&sc->sc_tcn_timer);
1198 	bstp_timer_stop(&sc->sc_topology_change_timer);
1199 
1200 	if (callout_pending(&sc->sc_bstpcallout) == 0)
1201 		callout_reset(&sc->sc_bstpcallout, hz,
1202 		    bstp_tick, sc);
1203 
1204 	TAILQ_FOREACH_MUTABLE(bif, &sc->sc_iflists[mycpuid], bif_next, nbif) {
1205 		if (sc->sc_ifp->if_flags & IFF_LINK1)
1206 			bstp_timer_start(&bif->bif_link1_timer, 0);
1207 		if (bif->bif_flags & IFBIF_STP)
1208 			bstp_ifupdstatus(sc, bif);
1209 		else
1210 			bstp_disable_port(sc, bif);
1211 
1212 		if (nbif != NULL && !nbif->bif_onlist) {
1213 			KKASSERT(bif->bif_onlist);
1214 			nbif = TAILQ_NEXT(bif, bif_next);
1215 		}
1216 	}
1217 
1218 	bstp_port_state_selection(sc);
1219 	bstp_config_bpdu_generation(sc);
1220 	bstp_timer_start(&sc->sc_hello_timer, 0);
1221 }
1222 
1223 void
1224 bstp_stop(struct bridge_softc *sc)
1225 {
1226 	struct bridge_iflist *bif;
1227 
1228 	KKASSERT(&curthread->td_msgport == BRIDGE_CFGPORT);
1229 
1230 	TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
1231 		bstp_set_port_state(bif, BSTP_IFSTATE_DISABLED);
1232 		bstp_timer_stop(&bif->bif_hold_timer);
1233 		bstp_timer_stop(&bif->bif_message_age_timer);
1234 		bstp_timer_stop(&bif->bif_forward_delay_timer);
1235 		bstp_timer_stop(&bif->bif_link1_timer);
1236 	}
1237 
1238 	callout_stop(&sc->sc_bstpcallout);
1239 
1240 	bstp_timer_stop(&sc->sc_topology_change_timer);
1241 	bstp_timer_stop(&sc->sc_tcn_timer);
1242 	bstp_timer_stop(&sc->sc_hello_timer);
1243 
1244 	crit_enter();
1245 	lwkt_dropmsg(&sc->sc_bstptimemsg.lmsg);
1246 	crit_exit();
1247 }
1248 
1249 /*
1250  * [re]initialize a port.  The port is initialized to a L1BLOCKING state
1251  * or a BLOCKING state.  When link ping/pong is enabled (LINK1) we start
1252  * out in the L1BLOCKING state to prevent flapping from blowing up the
1253  * state of the other ports.
1254  *
1255  * Either way the first config packet received will knock the port into
1256  * the LISTENING state.
1257  */
1258 static void
1259 bstp_initialize_port(struct bridge_softc *sc, struct bridge_iflist *bif)
1260 {
1261 	int needs_adjust = (bif->bif_state == BSTP_IFSTATE_FORWARDING ||
1262 			    bif->bif_state == BSTP_IFSTATE_BLOCKING ||
1263 			    bif->bif_state == BSTP_IFSTATE_BONDED);
1264 
1265 	if (sc->sc_ifp->if_flags & IFF_LINK1)
1266 		bstp_set_port_state(bif, BSTP_IFSTATE_L1BLOCKING);
1267 	else
1268 		bstp_set_port_state(bif, BSTP_IFSTATE_BLOCKING);
1269 	bstp_clear_peer_info(sc, bif);
1270 	bif->bif_topology_change_acknowledge = 0;
1271 	bif->bif_config_pending = 0;
1272 	bif->bif_change_detection_enabled = 1;
1273 	bstp_timer_stop(&bif->bif_message_age_timer);
1274 	bstp_timer_stop(&bif->bif_forward_delay_timer);
1275 	bstp_timer_stop(&bif->bif_hold_timer);
1276 	bstp_timer_stop(&bif->bif_link1_timer);
1277 	if (needs_adjust && (sc->sc_ifp->if_flags & IFF_LINK2))
1278 		bstp_adjust_bonded_states(sc, bif);
1279 }
1280 
1281 /*
1282  * When enabling a port that was previously disabled no configuration
1283  * update occurs, otherwise down/up or running/not-running flapping
1284  * (e.g. by openvpn on a TAP interface) will blow up the other ports.
1285  */
1286 static void
1287 bstp_enable_port(struct bridge_softc *sc, struct bridge_iflist *bif)
1288 {
1289 	int was_disabled = (bif->bif_state == BSTP_IFSTATE_DISABLED ||
1290 			    bif->bif_state == BSTP_IFSTATE_L1BLOCKING);
1291 
1292 	bstp_initialize_port(sc, bif);
1293 	if (sc->sc_ifp->if_flags & IFF_LINK1)
1294 		bstp_timer_start(&bif->bif_link1_timer, 0);
1295 	if (was_disabled == 0) {
1296 		bstp_configuration_update(sc);
1297 		bstp_port_state_selection(sc);
1298 	}
1299 	bstp_adjust_bonded_states(sc, bif);
1300 }
1301 
1302 /*
1303  * When disabling a port that was previously in a non-forwarding or bonded
1304  * state no configuration update occurs, otherwise down/up or
1305  * running/not-running flapping (e.g. by openvpn on a TAP interface) will
1306  * blow up the other ports.
1307  */
1308 static void
1309 bstp_disable_port(struct bridge_softc *sc, struct bridge_iflist *bif)
1310 {
1311 	int was_forwarding = (bif->bif_state == BSTP_IFSTATE_FORWARDING);
1312 	int was_bonded = (bif->bif_state == BSTP_IFSTATE_BONDED);
1313 	int iamroot;
1314 
1315 	iamroot = bstp_root_bridge(sc);
1316 
1317 	bstp_clear_peer_info(sc, bif);
1318 	bstp_set_port_state(bif, BSTP_IFSTATE_DISABLED);
1319 	bif->bif_topology_change_acknowledge = 0;
1320 	bif->bif_config_pending = 0;
1321 	bstp_timer_stop(&bif->bif_message_age_timer);
1322 	bstp_timer_stop(&bif->bif_forward_delay_timer);
1323 	bstp_timer_stop(&bif->bif_link1_timer);
1324 	if (was_forwarding || was_bonded) {
1325 		bstp_configuration_update(sc);
1326 		bstp_port_state_selection(sc);
1327 	}
1328 	bridge_rtdelete(sc, bif->bif_ifp, IFBF_FLUSHDYN);
1329 	if (was_forwarding && (sc->sc_ifp->if_flags & IFF_LINK2))
1330 		bstp_adjust_bonded_states(sc, bif);
1331 
1332 	if (iamroot == 0 && bstp_root_bridge(sc)) {
1333 		sc->sc_max_age = sc->sc_bridge_max_age;
1334 		sc->sc_hello_time = sc->sc_bridge_hello_time;
1335 		sc->sc_forward_delay = sc->sc_bridge_forward_delay;
1336 
1337 		bstp_topology_change_detection(sc);
1338 		bstp_timer_stop(&sc->sc_tcn_timer);
1339 		bstp_config_bpdu_generation(sc);
1340 		bstp_timer_start(&sc->sc_hello_timer, 0);
1341 	}
1342 }
1343 
1344 void
1345 bstp_linkstate(struct ifnet *ifp, int state)
1346 {
1347 	struct bridge_softc *sc;
1348 	struct bridge_iflist *bif;
1349 
1350 	sc = ifp->if_bridge;
1351 	ifnet_serialize_all(sc->sc_ifp);
1352 
1353 	/*
1354 	 * bstp_ifupdstatus() may block, but it is the last
1355 	 * operation of the member iface iteration, so we
1356 	 * don't need to use LIST_FOREACH_MUTABLE()+bif_onlist
1357 	 * check here.
1358 	 */
1359 	TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
1360 		if ((bif->bif_flags & IFBIF_STP) == 0)
1361 			continue;
1362 
1363 		if (bif->bif_ifp == ifp) {
1364 			bstp_ifupdstatus(sc, bif);
1365 			break;
1366 		}
1367 	}
1368 	ifnet_deserialize_all(sc->sc_ifp);
1369 }
1370 
1371 static void
1372 bstp_ifupdstatus(struct bridge_softc *sc, struct bridge_iflist *bif)
1373 {
1374 	struct ifnet *ifp = bif->bif_ifp;
1375 	struct ifmediareq ifmr;
1376 	int error = 0;
1377 
1378 	bzero((char *)&ifmr, sizeof(ifmr));
1379 	ifnet_serialize_all(ifp);
1380 	error = (*ifp->if_ioctl)(ifp, SIOCGIFMEDIA, (caddr_t)&ifmr, NULL);
1381 	ifnet_deserialize_all(ifp);
1382 
1383 	if ((error == 0) && (ifp->if_flags & IFF_UP)) {
1384 	 	if (ifmr.ifm_status & IFM_ACTIVE) {
1385 			if (bif->bif_state == BSTP_IFSTATE_DISABLED)
1386 				bstp_enable_port(sc, bif);
1387 
1388 		} else {
1389 			if (bif->bif_state != BSTP_IFSTATE_DISABLED)
1390 				bstp_disable_port(sc, bif);
1391 		}
1392 		return;
1393 	}
1394 
1395 	if (bif->bif_state != BSTP_IFSTATE_DISABLED)
1396 		bstp_disable_port(sc, bif);
1397 }
1398 
1399 static void
1400 bstp_tick(void *arg)
1401 {
1402 	struct bridge_softc *sc = arg;
1403 	struct lwkt_msg *lmsg;
1404 
1405 	KKASSERT(mycpuid == BRIDGE_CFGCPU);
1406 
1407 	crit_enter();
1408 
1409 	if (callout_pending(&sc->sc_bstpcallout) ||
1410 	    !callout_active(&sc->sc_bstpcallout)) {
1411 		crit_exit();
1412 		return;
1413 	}
1414 	callout_deactivate(&sc->sc_bstpcallout);
1415 
1416 	lmsg = &sc->sc_bstptimemsg.lmsg;
1417 	KKASSERT(lmsg->ms_flags & MSGF_DONE);
1418 	lwkt_sendmsg_oncpu(BRIDGE_CFGPORT, lmsg);
1419 
1420 	crit_exit();
1421 }
1422 
1423 void
1424 bstp_tick_handler(netmsg_t msg)
1425 {
1426 	struct bridge_softc *sc = msg->lmsg.u.ms_resultp;
1427 	struct bridge_iflist *bif;
1428 
1429 	KKASSERT(&curthread->td_msgport == BRIDGE_CFGPORT);
1430 	crit_enter();
1431 	/* Reply ASAP */
1432 	lwkt_replymsg(&msg->lmsg, 0);
1433 	crit_exit();
1434 
1435 	ifnet_serialize_all(sc->sc_ifp);
1436 
1437 	/*
1438 	 * NOTE:
1439 	 * We don't need to worry that member iface is ripped
1440 	 * from the per-cpu list during the blocking operation
1441 	 * in the loop body, since deletion is serialized by
1442 	 * BRIDGE_CFGPORT
1443 	 */
1444 
1445 	TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
1446 		if ((bif->bif_flags & IFBIF_STP) == 0)
1447 			continue;
1448 		/*
1449 		 * XXX This can cause a lag in "link does away"
1450 		 * XXX and "spanning tree gets updated".  We need
1451 		 * XXX come sort of callback from the link state
1452 		 * XXX update code to kick spanning tree.
1453 		 * XXX --thorpej@NetBSD.org
1454 		 */
1455 		bstp_ifupdstatus(sc, bif);
1456 	}
1457 
1458 	if (bstp_timer_expired(&sc->sc_hello_timer, sc->sc_hello_time))
1459 		bstp_hello_timer_expiry(sc);
1460 
1461 	if (bstp_timer_expired(&sc->sc_tcn_timer, sc->sc_bridge_hello_time))
1462 		bstp_tcn_timer_expiry(sc);
1463 
1464 	if (bstp_timer_expired(&sc->sc_topology_change_timer,
1465 	    sc->sc_topology_change_time))
1466 		bstp_topology_change_timer_expiry(sc);
1467 
1468 	TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
1469 		if ((bif->bif_flags & IFBIF_STP) == 0)
1470 			continue;
1471 		if (bstp_timer_expired(&bif->bif_message_age_timer,
1472 		    sc->sc_max_age))
1473 			bstp_message_age_timer_expiry(sc, bif);
1474 	}
1475 
1476 	TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
1477 		if ((bif->bif_flags & IFBIF_STP) == 0)
1478 			continue;
1479 		if (bstp_timer_expired(&bif->bif_forward_delay_timer,
1480 		    sc->sc_forward_delay))
1481 			bstp_forward_delay_timer_expiry(sc, bif);
1482 
1483 		if (bstp_timer_expired(&bif->bif_hold_timer,
1484 		    sc->sc_hold_time))
1485 			bstp_hold_timer_expiry(sc, bif);
1486 
1487 		if (bstp_timer_expired(&bif->bif_link1_timer,
1488 		    sc->sc_hello_time * 10))
1489 			bstp_link1_timer_expiry(sc, bif);
1490 	}
1491 
1492 	if (sc->sc_ifp->if_flags & IFF_RUNNING)
1493 		callout_reset(&sc->sc_bstpcallout, hz, bstp_tick, sc);
1494 
1495 	ifnet_deserialize_all(sc->sc_ifp);
1496 }
1497 
1498 static void
1499 bstp_timer_start(struct bridge_timer *t, uint16_t v)
1500 {
1501 	t->value = v;
1502 	t->active = 1;
1503 }
1504 
1505 static void
1506 bstp_timer_stop(struct bridge_timer *t)
1507 {
1508 	t->value = 0;
1509 	t->active = 0;
1510 }
1511 
1512 static int
1513 bstp_timer_expired(struct bridge_timer *t, uint16_t v)
1514 {
1515 	if (t->active == 0)
1516 		return (0);
1517 	t->value += BSTP_TICK_VAL;
1518 	if (t->value >= v) {
1519 		bstp_timer_stop(t);
1520 		return (1);
1521 	}
1522 	return (0);
1523 
1524 }
1525