xref: /dragonfly/sys/net/bridge/bridgestp.c (revision 0085a56d)
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/malloc.h>	/* for M_NOWAIT */
45 #include <sys/mbuf.h>
46 #include <sys/socket.h>
47 #include <sys/sockio.h>
48 #include <sys/kernel.h>
49 #include <sys/callout.h>
50 #include <sys/proc.h>
51 #include <sys/lock.h>
52 #include <sys/thread.h>
53 #include <sys/thread2.h>
54 #include <sys/msgport2.h>
55 
56 #include <net/if.h>
57 #include <net/if_dl.h>
58 #include <net/if_types.h>
59 #include <net/if_llc.h>
60 #include <net/if_media.h>
61 
62 #include <netinet/in.h>
63 #include <netinet/in_systm.h>
64 #include <netinet/in_var.h>
65 #include <netinet/if_ether.h>
66 #include <net/bridge/if_bridgevar.h>
67 
68 /* BPDU message types */
69 #define	BSTP_MSGTYPE_CFG	0x00		/* Configuration */
70 #define	BSTP_MSGTYPE_TCN	0x80		/* Topology chg notification */
71 
72 /* BPDU flags */
73 #define	BSTP_FLAG_TC		0x01		/* Topology change */
74 #define	BSTP_FLAG_TCA		0x80		/* Topology change ack */
75 
76 #define	BSTP_MESSAGE_AGE_INCR	(1 * 256)	/* in 256ths of a second */
77 #define	BSTP_TICK_VAL		(1 * 256)	/* in 256ths of a second */
78 
79 /*
80  * Because BPDU's do not make nicely aligned structures, two different
81  * declarations are used: bstp_?bpdu (wire representation, packed) and
82  * bstp_*_unit (internal, nicely aligned version).
83  */
84 
85 /* configuration bridge protocol data unit */
86 struct bstp_cbpdu {
87 	uint8_t		cbu_dsap;		/* LLC: destination sap */
88 	uint8_t		cbu_ssap;		/* LLC: source sap */
89 	uint8_t		cbu_ctl;		/* LLC: control */
90 	uint16_t	cbu_protoid;		/* protocol id */
91 	uint8_t		cbu_protover;		/* protocol version */
92 	uint8_t		cbu_bpdutype;		/* message type */
93 	uint8_t		cbu_flags;		/* flags (below) */
94 
95 	/* root id */
96 	uint16_t	cbu_rootpri;		/* root priority */
97 	uint8_t	cbu_rootaddr[6];	/* root address */
98 
99 	uint32_t	cbu_rootpathcost;	/* root path cost */
100 
101 	/* bridge id */
102 	uint16_t	cbu_bridgepri;		/* bridge priority */
103 	uint8_t		cbu_bridgeaddr[6];	/* bridge address */
104 
105 	uint16_t	cbu_portid;		/* port id */
106 	uint16_t	cbu_messageage;		/* current message age */
107 	uint16_t	cbu_maxage;		/* maximum age */
108 	uint16_t	cbu_hellotime;		/* hello time */
109 	uint16_t	cbu_forwarddelay;	/* forwarding delay */
110 } __attribute__((__packed__));
111 
112 /* topology change notification bridge protocol data unit */
113 struct bstp_tbpdu {
114 	uint8_t		tbu_dsap;		/* LLC: destination sap */
115 	uint8_t		tbu_ssap;		/* LLC: source sap */
116 	uint8_t		tbu_ctl;		/* LLC: control */
117 	uint16_t	tbu_protoid;		/* protocol id */
118 	uint8_t		tbu_protover;		/* protocol version */
119 	uint8_t		tbu_bpdutype;		/* message type */
120 } __attribute__((__packed__));
121 
122 const uint8_t bstp_etheraddr[] = { 0x01, 0x80, 0xc2, 0x00, 0x00, 0x00 };
123 
124 static void	bstp_initialize_port(struct bridge_softc *,
125 		    struct bridge_iflist *);
126 static void	bstp_ifupdstatus(struct bridge_softc *, struct bridge_iflist *);
127 static void	bstp_enable_port(struct bridge_softc *, struct bridge_iflist *);
128 static void	bstp_disable_port(struct bridge_softc *,
129 		    struct bridge_iflist *);
130 #ifdef notused
131 static void	bstp_enable_change_detection(struct bridge_iflist *);
132 static void	bstp_disable_change_detection(struct bridge_iflist *);
133 #endif /* notused */
134 static int	bstp_root_bridge(struct bridge_softc *sc);
135 static void	bstp_transmit_config(struct bridge_softc *,
136 		    struct bridge_iflist *);
137 static void	bstp_transmit_tcn(struct bridge_softc *);
138 static void	bstp_received_config_bpdu(struct bridge_softc *,
139 		    struct bridge_iflist *, struct bstp_config_unit *);
140 static void	bstp_received_tcn_bpdu(struct bridge_softc *,
141 		    struct bridge_iflist *, struct bstp_tcn_unit *);
142 static void	bstp_record_config_information(struct bridge_softc *,
143 		    struct bridge_iflist *, struct bstp_config_unit *);
144 static void	bstp_record_config_timeout_values(struct bridge_softc *,
145 		    struct bstp_config_unit *);
146 static void	bstp_config_bpdu_generation(struct bridge_softc *);
147 static void	bstp_send_config_bpdu(struct bridge_softc *,
148 		    struct bridge_iflist *, struct bstp_config_unit *);
149 static void	bstp_configuration_update(struct bridge_softc *);
150 static void	bstp_port_state_selection(struct bridge_softc *);
151 static void	bstp_clear_peer_info(struct bridge_softc *,
152 		    struct bridge_iflist *);
153 static void	bstp_make_forwarding(struct bridge_softc *,
154 		    struct bridge_iflist *);
155 static void	bstp_make_blocking(struct bridge_softc *,
156 		    struct bridge_iflist *);
157 static void	bstp_make_l1blocking(struct bridge_softc *sc,
158 		    struct bridge_iflist *bif);
159 static void	bstp_adjust_bonded_states(struct bridge_softc *sc,
160 		    struct bridge_iflist *obif);
161 static void	bstp_set_port_state(struct bridge_iflist *, uint8_t);
162 #ifdef notused
163 static void	bstp_set_bridge_priority(struct bridge_softc *, uint64_t);
164 static void	bstp_set_port_priority(struct bridge_softc *,
165 		    struct bridge_iflist *, uint16_t);
166 static void	bstp_set_path_cost(struct bridge_softc *,
167 		    struct bridge_iflist *, uint32_t);
168 #endif /* notused */
169 static void	bstp_topology_change_detection(struct bridge_softc *);
170 static void	bstp_topology_change_acknowledged(struct bridge_softc *);
171 static void	bstp_acknowledge_topology_change(struct bridge_softc *,
172 		    struct bridge_iflist *);
173 
174 static void	bstp_tick(void *);
175 static void	bstp_timer_start(struct bridge_timer *, uint16_t);
176 static void	bstp_timer_stop(struct bridge_timer *);
177 static int	bstp_timer_expired(struct bridge_timer *, uint16_t);
178 
179 static void	bstp_hold_timer_expiry(struct bridge_softc *,
180 		    struct bridge_iflist *);
181 static void	bstp_message_age_timer_expiry(struct bridge_softc *,
182 		    struct bridge_iflist *);
183 static void	bstp_forward_delay_timer_expiry(struct bridge_softc *,
184 		    struct bridge_iflist *);
185 static void	bstp_topology_change_timer_expiry(struct bridge_softc *);
186 static void	bstp_tcn_timer_expiry(struct bridge_softc *);
187 static void	bstp_hello_timer_expiry(struct bridge_softc *);
188 static int	bstp_addr_cmp(const uint8_t *, const uint8_t *);
189 
190 /*
191  * When transmitting a config we tack on our path cost to
192  * our aggregated path-to-root cost.
193  *
194  * Note that sc_designated_cost is the lowest root path cost from all
195  * incoming links.  When one is talking about two bridges with multiple
196  * links between them the root path cost the secondary bridge transmits
197  * will be the lowest root path cost it received across those three
198  * links and NOT necessarily the root path cost it received specifically
199  * on that particular link.  The receiving end must still add its local
200  * link cost back in for differentiation purposes.
201  */
202 static void
203 bstp_transmit_config(struct bridge_softc *sc, struct bridge_iflist *bif)
204 {
205 	if (bif->bif_hold_timer.active) {
206 		bif->bif_config_pending = 1;
207 		return;
208 	}
209 
210 	bif->bif_config_bpdu.cu_message_type = BSTP_MSGTYPE_CFG;
211 	bif->bif_config_bpdu.cu_rootid = sc->sc_designated_root;
212 	bif->bif_config_bpdu.cu_root_path_cost = sc->sc_designated_cost +
213 						 bif->bif_path_cost;
214 	bif->bif_config_bpdu.cu_bridge_id = sc->sc_bridge_id;
215 	bif->bif_config_bpdu.cu_port_id = bif->bif_port_id;
216 
217 	if (bstp_root_bridge(sc)) {
218 		bif->bif_config_bpdu.cu_message_age = 0;
219 	} else if (sc->sc_root_port) {
220 		bif->bif_config_bpdu.cu_message_age =
221 			sc->sc_root_port->bif_message_age_timer.value +
222 			BSTP_MESSAGE_AGE_INCR;
223 	} else {
224 		bif->bif_config_bpdu.cu_message_age = BSTP_MESSAGE_AGE_INCR;
225 	}
226 
227 	bif->bif_config_bpdu.cu_max_age = sc->sc_max_age;
228 	bif->bif_config_bpdu.cu_hello_time = sc->sc_hello_time;
229 	bif->bif_config_bpdu.cu_forward_delay = sc->sc_forward_delay;
230 	bif->bif_config_bpdu.cu_topology_change_acknowledgment
231 	    = bif->bif_topology_change_acknowledge;
232 	bif->bif_config_bpdu.cu_topology_change = sc->sc_topology_change;
233 
234 	if (bif->bif_config_bpdu.cu_message_age < sc->sc_max_age ||
235 	    (sc->sc_ifp->if_flags & IFF_LINK1)) {
236 		bif->bif_topology_change_acknowledge = 0;
237 		bif->bif_config_pending = 0;
238 		bstp_send_config_bpdu(sc, bif, &bif->bif_config_bpdu);
239 		bstp_timer_start(&bif->bif_hold_timer, 0);
240 	}
241 }
242 
243 static void
244 bstp_send_config_bpdu(struct bridge_softc *sc, struct bridge_iflist *bif,
245 		      struct bstp_config_unit *cu)
246 {
247 	struct ifnet *ifp;
248 	struct mbuf *m;
249 	struct ether_header *eh;
250 	struct bstp_cbpdu bpdu;
251 
252 	ifp = bif->bif_ifp;
253 
254 	if ((ifp->if_flags & IFF_RUNNING) == 0)
255 		return;
256 
257 	MGETHDR(m, M_NOWAIT, MT_DATA);
258 	if (m == NULL)
259 		return;
260 
261 	eh = mtod(m, struct ether_header *);
262 
263 	m->m_pkthdr.rcvif = ifp;
264 	m->m_pkthdr.len = sizeof(*eh) + sizeof(bpdu);
265 	m->m_len = m->m_pkthdr.len;
266 
267 	bpdu.cbu_ssap = bpdu.cbu_dsap = LLC_8021D_LSAP;
268 	bpdu.cbu_ctl = LLC_UI;
269 	bpdu.cbu_protoid = htons(0);
270 	bpdu.cbu_protover = 0;
271 	bpdu.cbu_bpdutype = cu->cu_message_type;
272 	bpdu.cbu_flags = (cu->cu_topology_change ? BSTP_FLAG_TC : 0) |
273 	    (cu->cu_topology_change_acknowledgment ? BSTP_FLAG_TCA : 0);
274 
275 	bpdu.cbu_rootpri = htons(cu->cu_rootid >> 48);
276 	bpdu.cbu_rootaddr[0] = cu->cu_rootid >> 40;
277 	bpdu.cbu_rootaddr[1] = cu->cu_rootid >> 32;
278 	bpdu.cbu_rootaddr[2] = cu->cu_rootid >> 24;
279 	bpdu.cbu_rootaddr[3] = cu->cu_rootid >> 16;
280 	bpdu.cbu_rootaddr[4] = cu->cu_rootid >> 8;
281 	bpdu.cbu_rootaddr[5] = cu->cu_rootid >> 0;
282 
283 	bpdu.cbu_rootpathcost = htonl(cu->cu_root_path_cost);
284 
285 	bpdu.cbu_bridgepri = htons(cu->cu_bridge_id >> 48);
286 	bpdu.cbu_bridgeaddr[0] = cu->cu_bridge_id >> 40;
287 	bpdu.cbu_bridgeaddr[1] = cu->cu_bridge_id >> 32;
288 	bpdu.cbu_bridgeaddr[2] = cu->cu_bridge_id >> 24;
289 	bpdu.cbu_bridgeaddr[3] = cu->cu_bridge_id >> 16;
290 	bpdu.cbu_bridgeaddr[4] = cu->cu_bridge_id >> 8;
291 	bpdu.cbu_bridgeaddr[5] = cu->cu_bridge_id >> 0;
292 
293 	bpdu.cbu_portid = htons(cu->cu_port_id);
294 	bpdu.cbu_messageage = htons(cu->cu_message_age);
295 	bpdu.cbu_maxage = htons(cu->cu_max_age);
296 	bpdu.cbu_hellotime = htons(cu->cu_hello_time);
297 	bpdu.cbu_forwarddelay = htons(cu->cu_forward_delay);
298 
299 	/*
300 	 * Packets sent from the bridge always use the bridge MAC
301 	 * as the source.
302 	 */
303 	memcpy(eh->ether_shost, IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
304 	memcpy(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN);
305 	eh->ether_type = htons(sizeof(bpdu));
306 
307 	memcpy(mtod(m, caddr_t) + sizeof(*eh), &bpdu, sizeof(bpdu));
308 
309 	bridge_enqueue(ifp, m);
310 }
311 
312 static int
313 bstp_root_bridge(struct bridge_softc *sc)
314 {
315 	return (sc->sc_designated_root == sc->sc_bridge_id);
316 }
317 
318 /*
319  * Returns TRUE if the recorded information from our peer has a shorter
320  * graph distance than our current best.
321  */
322 int
323 bstp_supersedes_port_info(struct bridge_softc *sc, struct bridge_iflist *bif)
324 {
325 	if (bif->bif_peer_root < sc->sc_designated_root)
326 		return (1);
327 	if (bif->bif_peer_root > sc->sc_designated_root)
328 		return (0);
329 
330 	/*
331 	 * bif_peer_cost and sc_designated_cost do not include the local
332 	 * bif_path_cost, otherwise we would not be able to make this
333 	 * comparison.
334 	 */
335 	if (bif->bif_peer_cost < sc->sc_designated_cost)
336 		return (1);
337 	if (bif->bif_peer_cost > sc->sc_designated_cost)
338 		return (0);
339 
340 	/*
341 	 * When the peer costs match we have to check the local path
342 	 * cost against the selected root port.  sc_designated_cost
343 	 * in an aggregation and cannot be used.
344 	 */
345 	if (sc->sc_root_port &&
346 	    bif->bif_path_cost < sc->sc_root_port->bif_path_cost)
347 		return (1);
348 	if (sc->sc_root_port &&
349 	    bif->bif_path_cost > sc->sc_root_port->bif_path_cost)
350 		return (0);
351 
352 	/*
353 	 * If the path costs are identical the bridge with the lowest
354 	 * bridge_id wins.
355 	 */
356 	if (bif->bif_peer_bridge < sc->sc_designated_bridge)
357 		return (1);
358 	if (bif->bif_peer_bridge > sc->sc_designated_bridge)
359 		return (0);
360 
361 	/*
362 	 * bridge_id or bridge+port collision w/peer returns TRUE.
363 	 *
364 	 * This case can also occur when two bridges are connected
365 	 * via multiple links whos path costs have not been set.
366 	 */
367 	if (bif->bif_peer_bridge != sc->sc_bridge_id)
368 		return (1);
369 	if (bif->bif_peer_port <= sc->sc_designated_port)
370 		return (1);
371 	return (0);
372 }
373 
374 /*
375  * The shorter graph distance represented by cu (most of which is also
376  * already stored in our bif_peer_* fields) becomes the designated info.
377  *
378  * NOTE: sc_designated_cost does not include bif_path_cost, it is added
379  *	 in later on a port-by-port basis as needed.
380  */
381 static void
382 bstp_record_config_information(struct bridge_softc *sc,
383 			       struct bridge_iflist *bif,
384 			       struct bstp_config_unit *cu)
385 {
386 	sc->sc_designated_root = bif->bif_peer_root;
387 	sc->sc_designated_cost = bif->bif_peer_cost;
388 	sc->sc_designated_bridge = bif->bif_peer_bridge;
389 	sc->sc_designated_port = bif->bif_peer_port;
390 	bstp_timer_start(&bif->bif_message_age_timer, cu->cu_message_age);
391 }
392 
393 static void
394 bstp_record_config_timeout_values(struct bridge_softc *sc,
395 				  struct bstp_config_unit *config)
396 {
397 	sc->sc_max_age = config->cu_max_age;
398 	sc->sc_hello_time = config->cu_hello_time;
399 	sc->sc_forward_delay = config->cu_forward_delay;
400 	sc->sc_topology_change = config->cu_topology_change;
401 }
402 
403 static void
404 bstp_config_bpdu_generation(struct bridge_softc *sc)
405 {
406 	struct bridge_iflist *bif, *nbif;
407 
408 	TAILQ_FOREACH_MUTABLE(bif, &sc->sc_iflists[mycpuid], bif_next, nbif) {
409 		if ((bif->bif_flags & IFBIF_STP) == 0)
410 			continue;
411 		if (bif->bif_state != BSTP_IFSTATE_DISABLED &&
412 		    ((sc->sc_ifp->if_flags & IFF_LINK1) ||
413 		     (bif->bif_flags & IFBIF_DESIGNATED))) {
414 			bstp_transmit_config(sc, bif);
415 		}
416 
417 		if (nbif != NULL && !nbif->bif_onlist) {
418 			KKASSERT(bif->bif_onlist);
419 			nbif = TAILQ_NEXT(bif, bif_next);
420 		}
421 	}
422 }
423 
424 static void
425 bstp_transmit_tcn(struct bridge_softc *sc)
426 {
427 	struct bstp_tbpdu bpdu;
428 	struct ifnet *ifp;
429 	struct ether_header *eh;
430 	struct mbuf *m;
431 
432 	if (sc->sc_root_port == NULL)	/* all iterfaces disabled */
433 		return;
434 
435 	ifp = sc->sc_root_port->bif_ifp;
436 	if ((ifp->if_flags & IFF_RUNNING) == 0)
437 		return;
438 
439 	MGETHDR(m, M_NOWAIT, MT_DATA);
440 	if (m == NULL)
441 		return;
442 
443 	m->m_pkthdr.rcvif = ifp;
444 	m->m_pkthdr.len = sizeof(*eh) + sizeof(bpdu);
445 	m->m_len = m->m_pkthdr.len;
446 
447 	eh = mtod(m, struct ether_header *);
448 
449 	/*
450 	 * Packets sent from the bridge always use the bridge MAC
451 	 * as the source.
452 	 */
453 	memcpy(eh->ether_shost, IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
454 	memcpy(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN);
455 	eh->ether_type = htons(sizeof(bpdu));
456 
457 	bpdu.tbu_ssap = bpdu.tbu_dsap = LLC_8021D_LSAP;
458 	bpdu.tbu_ctl = LLC_UI;
459 	bpdu.tbu_protoid = 0;
460 	bpdu.tbu_protover = 0;
461 	bpdu.tbu_bpdutype = BSTP_MSGTYPE_TCN;
462 
463 	memcpy(mtod(m, caddr_t) + sizeof(*eh), &bpdu, sizeof(bpdu));
464 
465 	bridge_enqueue(ifp, m);
466 }
467 
468 /*
469  * Recalculate sc->sc_designated* and sc->sc_root_port (if our bridge
470  * is calculated to be the root bridge).  We do this by initializing
471  * the designated variables to point at us and then scan our peers.
472  * Any uninitialized peers will have a max-value root.
473  *
474  * Clear IFBIF_DESIGNATED on any ports which no longer match the criteria
475  * required to be a designated port.  Only aged out ports and the root
476  * port can be designated.
477  *
478  * If we win we do a second scan to determine which port on our bridge
479  * is the best.
480  */
481 static void
482 bstp_configuration_update(struct bridge_softc *sc)
483 {
484 	uint64_t	designated_root = sc->sc_bridge_id;
485 	uint64_t	designated_bridge = sc->sc_bridge_id;
486 	uint32_t	designated_cost = 0xFFFFFFFFU;
487 	uint32_t	designated_root_cost = 0xFFFFFFFFU;
488 	uint16_t	designated_port = 65535;
489 	struct bridge_iflist *root_port = NULL;
490 	struct bridge_iflist *bif;
491 
492 	/*
493 	 * Resolve information from our peers.  Aged peers will have
494 	 * a maxed bif_peer_root and not come under consideration.
495 	 */
496 	TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
497 		if ((bif->bif_flags & IFBIF_STP) == 0)
498 			continue;
499 		if (bif->bif_state == BSTP_IFSTATE_DISABLED ||
500 		    bif->bif_state == BSTP_IFSTATE_L1BLOCKING) {
501 			continue;
502 		}
503 
504 		if (bif->bif_peer_root > designated_root)
505 			continue;
506 		if (bif->bif_peer_root < designated_root)
507 			goto set_port;
508 
509 		/*
510 		 * NOTE: The designated_cost temporary variable already added
511 		 *	 in the path code of the related root port.
512 		 */
513 		if (bif->bif_peer_cost + bif->bif_path_cost > designated_cost)
514 			continue;
515 		if (bif->bif_peer_cost + bif->bif_path_cost < designated_cost)
516 			goto set_port;
517 
518 		if (bif->bif_peer_bridge > designated_bridge)
519 			continue;
520 		if (bif->bif_peer_bridge < designated_bridge)
521 			goto set_port;
522 
523 		if (bif->bif_peer_port > designated_port)
524 			continue;
525 		if (bif->bif_peer_port < designated_port)
526 			goto set_port;
527 
528 		/*
529 		 * Same root, path cost, bridge, and port.  Set the root
530 		 * only if we do not already have it.
531 		 */
532 		if (root_port)
533 			continue;
534 
535 		/*
536 		 * New root port (from peers)
537 		 *
538 		 * NOTE: Temporarily add bif_path_cost into the designated
539 		 *	 cost to reduce complexity in the loop, it will be
540 		 *	 subtracted out when we are done.
541 		 */
542 set_port:
543 		designated_root = bif->bif_peer_root;
544 		designated_cost = bif->bif_peer_cost + bif->bif_path_cost;
545 		designated_root_cost = bif->bif_peer_cost;
546 		designated_bridge = bif->bif_peer_bridge;
547 		designated_port = bif->bif_peer_port;
548 		root_port = bif;
549 	}
550 
551 	/*
552 	 * root_port will be NULL at the start here if all of our
553 	 * peers are aged or are not as good a root as our bridge would
554 	 * be.  It can also be NULL due to all related links being
555 	 * disabled.
556 	 *
557 	 * If the root winds up being our bridge scan again against local
558 	 * information.  Unconditionally update IFBIF_DESIGNATED.
559 	 */
560 	TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
561 		bif->bif_flags &= ~(IFBIF_DESIGNATED | IFBIF_ROOT);
562 		if ((bif->bif_flags & IFBIF_STP) == 0)
563 			continue;
564 		if (bif->bif_state == BSTP_IFSTATE_DISABLED ||
565 		    bif->bif_state == BSTP_IFSTATE_L1BLOCKING) {
566 			continue;
567 		}
568 
569 		/*
570 		 * Set DESIGNATED for an aged or unknown peer.
571 		 */
572 		if (bif->bif_peer_bridge == 0xFFFFFFFFFFFFFFFFLLU)
573 			bif->bif_flags |= IFBIF_DESIGNATED;
574 		if (designated_root != sc->sc_bridge_id)
575 			continue;
576 
577 		/*
578 		 * This is only reached if our bridge is the root bridge,
579 		 * select the root port (IFBIF_DESIGNATED is set at the
580 		 * end).
581 		 *
582 		 * The peer cost will incorporate the peer bridge's best
583 		 * cost to the root bridge (us), which is NOT necessarily
584 		 * our path cost, plus the peer bridge's own path cost.
585 		 * Because of this we must still add in our path cost for
586 		 * further differentiation.  Because the peer's 'best' cost
587 		 * is applied to all of its outgoing messages we should
588 		 * still come to the same conclusion as the peer.
589 		 */
590 		if (bif->bif_peer_cost + bif->bif_path_cost > designated_cost)
591 			continue;
592 		if (bif->bif_peer_cost + bif->bif_path_cost < designated_cost)
593 			goto set_port2;
594 
595 		if (bif->bif_port_id > designated_port)
596 			continue;
597 		if (bif->bif_port_id < designated_port)
598 			goto set_port2;
599 		/* degenerate case (possible peer collision w/our key */
600 
601 		/*
602 		 * New port.  Since we are the root, the root cost is always
603 		 * 0.  The peer's cu_root_path_cost does not necessarily
604 		 * incorporate our peer path cost, but instead incorporates
605 		 * the lowest path cost to root for the target bridge
606 		 * when multiple links are present between the two bridges.
607 		 * Thus for selection purposes we must still add in our
608 		 * local path cost.
609 		 */
610 set_port2:
611 		designated_cost = bif->bif_peer_cost + bif->bif_path_cost;
612 		designated_root_cost = 0;
613 		designated_bridge = sc->sc_bridge_id;
614 		designated_port = bif->bif_port_id;
615 		root_port = bif;
616 	}
617 
618 	/*
619 	 * Update aggregate information.  The selected root port always
620 	 * becomes a designated port (along with aged ports).  This can
621 	 * either be the port whos peer is closest to the root or it
622 	 * can be one of our ports if our bridge is the root.
623 	 *
624 	 * The root cost we record in sc_designated_root does not include
625 	 * bif_path_cost of the root port, since we may transmit this
626 	 * out of any port we will add the cost back in later on on
627 	 * a per-port basis.
628 	 *
629 	 * root_port can be NULL here (if all links are disabled)
630 	 */
631 	if (root_port) {
632 		sc->sc_designated_root = designated_root;
633 		sc->sc_designated_cost = designated_root_cost;
634 		sc->sc_designated_bridge = designated_bridge;
635 		sc->sc_designated_port = designated_port;
636 		root_port->bif_flags |= IFBIF_DESIGNATED | IFBIF_ROOT;
637 	} else {
638 		sc->sc_designated_root = designated_root;
639 		sc->sc_designated_cost = designated_cost;
640 		sc->sc_designated_bridge = designated_bridge;
641 		sc->sc_designated_port = designated_port;
642 	}
643 	sc->sc_root_port = root_port;
644 }
645 
646 /*
647  * Calculate the desired state for each interface link on our bridge.
648  *
649  * The best port will match against sc->sc_root_port (whether we are root
650  * or whether that port is the closest to the root).  We push this port
651  * towards a FORWARDING state.
652  *
653  * Next come designated ports, either aged ports or ports with no peer info
654  * (yet), or the peer who is closest to the root. We push this port towards
655  * a FORWARDING state as well.
656  *
657  * Any remaining ports are pushed towards a BLOCKING state.  Both sides of
658  * the port (us and our peer) should wind up placing the two ends in this
659  * state or bad things happen.
660  */
661 static void
662 bstp_port_state_selection(struct bridge_softc *sc)
663 {
664 	struct bridge_iflist *bif, *nbif;
665 
666 	TAILQ_FOREACH_MUTABLE(bif, &sc->sc_iflists[mycpuid], bif_next, nbif) {
667 		if ((bif->bif_flags & IFBIF_STP) == 0)
668 			continue;
669 		if (sc->sc_root_port &&
670 		    bif->bif_info == sc->sc_root_port->bif_info) {
671 			bif->bif_config_pending = 0;
672 			bif->bif_topology_change_acknowledge = 0;
673 			bstp_make_forwarding(sc, bif);
674 		} else if (bif->bif_flags & IFBIF_DESIGNATED) {
675 			bstp_timer_stop(&bif->bif_message_age_timer);
676 			bstp_make_forwarding(sc, bif);
677 		} else {
678 			bif->bif_config_pending = 0;
679 			bif->bif_topology_change_acknowledge = 0;
680 			bstp_make_blocking(sc, bif);
681 		}
682 
683 		if (nbif != NULL && !nbif->bif_onlist) {
684 			KKASSERT(bif->bif_onlist);
685 			nbif = TAILQ_NEXT(bif, bif_next);
686 		}
687 	}
688 }
689 
690 /*
691  * Clear peer info, effectively makes the port looked aged out.
692  * It becomes a designated go-to port.
693  */
694 static void
695 bstp_clear_peer_info(struct bridge_softc *sc, struct bridge_iflist *bif)
696 {
697 	bif->bif_peer_root = 0xFFFFFFFFFFFFFFFFLLU;
698 	bif->bif_peer_cost = 0xFFFFFFFFU;
699 	bif->bif_peer_bridge = 0xFFFFFFFFFFFFFFFFLLU;
700 	bif->bif_peer_port = 0xFFFFU;
701 
702 	if (bif->bif_state != BSTP_IFSTATE_DISABLED &&
703 	    bif->bif_state != BSTP_IFSTATE_L1BLOCKING) {
704 		bif->bif_flags |= IFBIF_DESIGNATED;
705 	}
706 }
707 
708 static void
709 bstp_make_forwarding(struct bridge_softc *sc, struct bridge_iflist *bif)
710 {
711 	if (bif->bif_state == BSTP_IFSTATE_BLOCKING ||
712 	    bif->bif_state == BSTP_IFSTATE_BONDED) {
713 		bstp_set_port_state(bif, BSTP_IFSTATE_LISTENING);
714 		bstp_timer_start(&bif->bif_forward_delay_timer, 0);
715 	}
716 }
717 
718 static void
719 bstp_make_blocking(struct bridge_softc *sc, struct bridge_iflist *bif)
720 {
721 	if (bif->bif_state != BSTP_IFSTATE_DISABLED &&
722 	    bif->bif_state != BSTP_IFSTATE_BLOCKING &&
723 	    bif->bif_state != BSTP_IFSTATE_BONDED &&
724 	    bif->bif_state != BSTP_IFSTATE_L1BLOCKING) {
725 		if ((bif->bif_state == BSTP_IFSTATE_FORWARDING) ||
726 		    (bif->bif_state == BSTP_IFSTATE_LEARNING)) {
727 			if (bif->bif_change_detection_enabled) {
728 				bstp_topology_change_detection(sc);
729 			}
730 		}
731 		bstp_set_port_state(bif, BSTP_IFSTATE_BLOCKING);
732 		bridge_rtdelete(sc, bif->bif_ifp, IFBF_FLUSHDYN);
733 		bstp_timer_stop(&bif->bif_forward_delay_timer);
734 		if (sc->sc_ifp->if_flags & IFF_LINK2)
735 			bstp_adjust_bonded_states(sc, bif);
736 	}
737 }
738 
739 static void
740 bstp_make_l1blocking(struct bridge_softc *sc, struct bridge_iflist *bif)
741 {
742 	int was_forwarding = (bif->bif_state == BSTP_IFSTATE_FORWARDING);
743 
744 	switch(bif->bif_state) {
745 	case BSTP_IFSTATE_LISTENING:
746 	case BSTP_IFSTATE_LEARNING:
747 	case BSTP_IFSTATE_FORWARDING:
748 	case BSTP_IFSTATE_BLOCKING:
749 	case BSTP_IFSTATE_BONDED:
750 		bstp_set_port_state(bif, BSTP_IFSTATE_L1BLOCKING);
751 		bridge_rtdelete(sc, bif->bif_ifp, IFBF_FLUSHDYN);
752 		bstp_timer_stop(&bif->bif_forward_delay_timer);
753 		bstp_timer_stop(&bif->bif_link1_timer);
754 		if (bif->bif_flags & IFBIF_DESIGNATED) {
755 			bif->bif_flags &= ~IFBIF_DESIGNATED;
756 			bstp_configuration_update(sc);
757 			bstp_port_state_selection(sc);
758 		}
759 		if (was_forwarding && (sc->sc_ifp->if_flags & IFF_LINK2))
760 			bstp_adjust_bonded_states(sc, bif);
761 		break;
762 	default:
763 		break;
764 	}
765 }
766 
767 /*
768  * Member (bif) changes to or from a FORWARDING state.  All members in the
769  * same bonding group which are in a BLOCKING or BONDED state must be set
770  * to either BLOCKING or BONDED based on whether any members in the bonding
771  * group remain in the FORWARDING state.
772  *
773  * Going between the BLOCKING and BONDED states does not require a
774  * configuration update.
775  */
776 static void
777 bstp_adjust_bonded_states(struct bridge_softc *sc, struct bridge_iflist *obif)
778 {
779 	struct bridge_iflist *bif;
780 	int state = BSTP_IFSTATE_BLOCKING;
781 
782 	TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
783 		if ((bif->bif_flags & IFBIF_STP) == 0)
784 			continue;
785 		if (bif->bif_state != BSTP_IFSTATE_FORWARDING)
786 			continue;
787 		if (memcmp(IF_LLADDR(bif->bif_ifp), IF_LLADDR(obif->bif_ifp),
788 			   ETHER_ADDR_LEN) != 0) {
789 			continue;
790 		}
791 		state = BSTP_IFSTATE_BONDED;
792 		break;
793 	}
794 	TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
795 		if ((bif->bif_flags & IFBIF_STP) == 0)
796 			continue;
797 		if (bif->bif_state != BSTP_IFSTATE_BLOCKING &&
798 		    bif->bif_state != BSTP_IFSTATE_BONDED) {
799 			continue;
800 		}
801 		if (memcmp(IF_LLADDR(bif->bif_ifp), IF_LLADDR(obif->bif_ifp),
802 			   ETHER_ADDR_LEN) != 0) {
803 			continue;
804 		}
805 		if (bif->bif_bond_weight == 0)
806 			bif->bif_state = BSTP_IFSTATE_BLOCKING;
807 		else
808 			bif->bif_state = state;
809 	}
810 }
811 
812 static void
813 bstp_set_port_state(struct bridge_iflist *bif, uint8_t state)
814 {
815 	bif->bif_state = state;
816 }
817 
818 static void
819 bstp_topology_change_detection(struct bridge_softc *sc)
820 {
821 	if (bstp_root_bridge(sc)) {
822 		sc->sc_topology_change = 1;
823 		bstp_timer_start(&sc->sc_topology_change_timer, 0);
824 	} else if (!sc->sc_topology_change_detected) {
825 		bstp_transmit_tcn(sc);
826 		bstp_timer_start(&sc->sc_tcn_timer, 0);
827 	}
828 	sc->sc_topology_change_detected = 1;
829 }
830 
831 static void
832 bstp_topology_change_acknowledged(struct bridge_softc *sc)
833 {
834 	sc->sc_topology_change_detected = 0;
835 	bstp_timer_stop(&sc->sc_tcn_timer);
836 }
837 
838 static void
839 bstp_acknowledge_topology_change(struct bridge_softc *sc,
840 				 struct bridge_iflist *bif)
841 {
842 	bif->bif_topology_change_acknowledge = 1;
843 	bstp_transmit_config(sc, bif);
844 }
845 
846 void
847 bstp_input(struct bridge_softc *sc, struct bridge_iflist *bif, struct mbuf *m)
848 {
849 	struct ether_header *eh;
850 	struct bstp_tbpdu tpdu;
851 	struct bstp_cbpdu cpdu;
852 	struct bstp_config_unit cu;
853 	struct bstp_tcn_unit tu;
854 	uint16_t len;
855 
856 	if ((bif->bif_flags & IFBIF_STP) == 0)
857 		goto out;
858 
859 	/*
860 	 * The L1BLOCKING (ping pong failover) test needs to reset the
861 	 * timer if LINK1 is active.
862 	 */
863 	if (bif->bif_state == BSTP_IFSTATE_L1BLOCKING) {
864 		bstp_set_port_state(bif, BSTP_IFSTATE_BLOCKING);
865 		if (sc->sc_ifp->if_flags & IFF_LINK1)
866 			bstp_timer_start(&bif->bif_link1_timer, 0);
867 		bstp_make_forwarding(sc, bif);
868 	} else if (sc->sc_ifp->if_flags & IFF_LINK1) {
869 		bstp_timer_start(&bif->bif_link1_timer, 0);
870 	}
871 
872 	eh = mtod(m, struct ether_header *);
873 
874 	len = ntohs(eh->ether_type);
875 	if (len < sizeof(tpdu))
876 		goto out;
877 
878 	m_adj(m, ETHER_HDR_LEN);
879 
880 	if (m->m_pkthdr.len > len)
881 		m_adj(m, len - m->m_pkthdr.len);
882 	if (m->m_len < sizeof(tpdu) &&
883 	    (m = m_pullup(m, sizeof(tpdu))) == NULL)
884 		goto out;
885 
886 	memcpy(&tpdu, mtod(m, caddr_t), sizeof(tpdu));
887 
888 	if (tpdu.tbu_dsap != LLC_8021D_LSAP ||
889 	    tpdu.tbu_ssap != LLC_8021D_LSAP ||
890 	    tpdu.tbu_ctl != LLC_UI)
891 		goto out;
892 	if (tpdu.tbu_protoid != 0 || tpdu.tbu_protover != 0)
893 		goto out;
894 
895 	switch (tpdu.tbu_bpdutype) {
896 	case BSTP_MSGTYPE_TCN:
897 		tu.tu_message_type = tpdu.tbu_bpdutype;
898 		bstp_received_tcn_bpdu(sc, bif, &tu);
899 		break;
900 	case BSTP_MSGTYPE_CFG:
901 		if (m->m_len < sizeof(cpdu) &&
902 		    (m = m_pullup(m, sizeof(cpdu))) == NULL)
903 			goto out;
904 		memcpy(&cpdu, mtod(m, caddr_t), sizeof(cpdu));
905 
906 		cu.cu_rootid =
907 		    (((uint64_t)ntohs(cpdu.cbu_rootpri)) << 48) |
908 		    (((uint64_t)cpdu.cbu_rootaddr[0]) << 40) |
909 		    (((uint64_t)cpdu.cbu_rootaddr[1]) << 32) |
910 		    (((uint64_t)cpdu.cbu_rootaddr[2]) << 24) |
911 		    (((uint64_t)cpdu.cbu_rootaddr[3]) << 16) |
912 		    (((uint64_t)cpdu.cbu_rootaddr[4]) << 8) |
913 		    (((uint64_t)cpdu.cbu_rootaddr[5]) << 0);
914 
915 		cu.cu_bridge_id =
916 		    (((uint64_t)ntohs(cpdu.cbu_bridgepri)) << 48) |
917 		    (((uint64_t)cpdu.cbu_bridgeaddr[0]) << 40) |
918 		    (((uint64_t)cpdu.cbu_bridgeaddr[1]) << 32) |
919 		    (((uint64_t)cpdu.cbu_bridgeaddr[2]) << 24) |
920 		    (((uint64_t)cpdu.cbu_bridgeaddr[3]) << 16) |
921 		    (((uint64_t)cpdu.cbu_bridgeaddr[4]) << 8) |
922 		    (((uint64_t)cpdu.cbu_bridgeaddr[5]) << 0);
923 
924 		cu.cu_root_path_cost = ntohl(cpdu.cbu_rootpathcost);
925 		cu.cu_message_age = ntohs(cpdu.cbu_messageage);
926 		cu.cu_max_age = ntohs(cpdu.cbu_maxage);
927 		cu.cu_hello_time = ntohs(cpdu.cbu_hellotime);
928 		cu.cu_forward_delay = ntohs(cpdu.cbu_forwarddelay);
929 		cu.cu_port_id = ntohs(cpdu.cbu_portid);
930 		cu.cu_message_type = cpdu.cbu_bpdutype;
931 		cu.cu_topology_change_acknowledgment =
932 		    (cpdu.cbu_flags & BSTP_FLAG_TCA) ? 1 : 0;
933 		cu.cu_topology_change =
934 		    (cpdu.cbu_flags & BSTP_FLAG_TC) ? 1 : 0;
935 		bstp_received_config_bpdu(sc, bif, &cu);
936 		break;
937 	default:
938 		goto out;
939 	}
940 out:
941 	if (m)
942 		m_freem(m);
943 }
944 
945 static void
946 bstp_received_config_bpdu(struct bridge_softc *sc, struct bridge_iflist *bif,
947 			  struct bstp_config_unit *cu)
948 {
949 	int iamroot;
950 
951 	iamroot = bstp_root_bridge(sc);
952 
953 	if (bif->bif_state != BSTP_IFSTATE_DISABLED) {
954 		/*
955 		 * Record information from peer.  The peer_cost field
956 		 * does not include the local bif->bif_path_cost, it will
957 		 * be added in as needed (since it can be modified manually
958 		 * this way we don't have to worry about fixups).
959 		 */
960 		bif->bif_peer_root = cu->cu_rootid;
961 		bif->bif_peer_cost = cu->cu_root_path_cost;
962 		bif->bif_peer_bridge = cu->cu_bridge_id;
963 		bif->bif_peer_port = cu->cu_port_id;
964 
965 		if (bstp_supersedes_port_info(sc, bif)) {
966 			bstp_record_config_information(sc, bif, cu);
967 			bstp_configuration_update(sc);
968 			bstp_port_state_selection(sc);
969 
970 			/*
971 			 * If our bridge loses its root status (?)
972 			 *
973 			 * Hello's (unsolicited CFG packets) are generated
974 			 * every hello period of LINK1 is set, otherwise
975 			 * we are no longer the root bridge and must stop
976 			 * generating unsolicited CFG packets.
977 			 */
978 			if (iamroot && bstp_root_bridge(sc) == 0) {
979 				if ((sc->sc_ifp->if_flags & IFF_LINK1) == 0)
980 					bstp_timer_stop(&sc->sc_hello_timer);
981 
982 				if (sc->sc_topology_change_detected) {
983 					bstp_timer_stop(
984 					    &sc->sc_topology_change_timer);
985 					bstp_transmit_tcn(sc);
986 					bstp_timer_start(&sc->sc_tcn_timer, 0);
987 				}
988 			}
989 
990 			if (sc->sc_root_port &&
991 			    bif->bif_info == sc->sc_root_port->bif_info) {
992 				bstp_record_config_timeout_values(sc, cu);
993 				bstp_config_bpdu_generation(sc);
994 
995 				if (cu->cu_topology_change_acknowledgment)
996 					bstp_topology_change_acknowledged(sc);
997 			}
998 		} else if (bif->bif_flags & IFBIF_DESIGNATED) {
999 			/*
1000 			 * Update designated ports (aged out peers or
1001 			 * the port closest to the root) at a faster pace.
1002 			 *
1003 			 * Clear our designated flag if we aren't marked
1004 			 * as the root port.
1005 			 */
1006 			bstp_transmit_config(sc, bif);
1007 			if ((bif->bif_flags & IFBIF_ROOT) == 0) {
1008 				bif->bif_flags &= ~IFBIF_DESIGNATED;
1009 				bstp_configuration_update(sc);
1010 				bstp_port_state_selection(sc);
1011 			}
1012 		}
1013 	}
1014 }
1015 
1016 static void
1017 bstp_received_tcn_bpdu(struct bridge_softc *sc, struct bridge_iflist *bif,
1018 		       struct bstp_tcn_unit *tcn)
1019 {
1020 	if (bif->bif_state != BSTP_IFSTATE_DISABLED &&
1021 	    (bif->bif_flags & IFBIF_DESIGNATED)) {
1022 		bstp_topology_change_detection(sc);
1023 		bstp_acknowledge_topology_change(sc, bif);
1024 	}
1025 }
1026 
1027 /*
1028  * link1 forces continuous hello's (the bridge interface must be cycled
1029  * to start them up), so keep the timer hot if that is the case, otherwise
1030  * only send HELLO's if we are the root.
1031  */
1032 static void
1033 bstp_hello_timer_expiry(struct bridge_softc *sc)
1034 {
1035 	bstp_config_bpdu_generation(sc);
1036 
1037 	if ((sc->sc_ifp->if_flags & IFF_LINK1) || bstp_root_bridge(sc))
1038 		bstp_timer_start(&sc->sc_hello_timer, 0);
1039 }
1040 
1041 static void
1042 bstp_message_age_timer_expiry(struct bridge_softc *sc,
1043 			      struct bridge_iflist *bif)
1044 {
1045 	int iamroot;
1046 
1047 	iamroot = bstp_root_bridge(sc);
1048 	bstp_clear_peer_info(sc, bif);
1049 	bstp_configuration_update(sc);
1050 	bstp_port_state_selection(sc);
1051 
1052 	/*
1053 	 * If we've become the root and were not the root before
1054 	 * we have some cleanup to do.  This also occurs if we
1055 	 * wind up being completely isolated.
1056 	 */
1057 	if (iamroot == 0 && bstp_root_bridge(sc)) {
1058 		sc->sc_max_age = sc->sc_bridge_max_age;
1059 		sc->sc_hello_time = sc->sc_bridge_hello_time;
1060 		sc->sc_forward_delay = sc->sc_bridge_forward_delay;
1061 
1062 		bstp_topology_change_detection(sc);
1063 		bstp_timer_stop(&sc->sc_tcn_timer);
1064 		bstp_config_bpdu_generation(sc);
1065 		bstp_timer_start(&sc->sc_hello_timer, 0);
1066 	}
1067 }
1068 
1069 static void
1070 bstp_forward_delay_timer_expiry(struct bridge_softc *sc,
1071 				struct bridge_iflist *bif)
1072 {
1073 	if (bif->bif_state == BSTP_IFSTATE_LISTENING) {
1074 		bstp_set_port_state(bif, BSTP_IFSTATE_LEARNING);
1075 		bstp_timer_start(&bif->bif_forward_delay_timer, 0);
1076 	} else if (bif->bif_state == BSTP_IFSTATE_LEARNING) {
1077 		bstp_set_port_state(bif, BSTP_IFSTATE_FORWARDING);
1078 		if (sc->sc_designated_bridge == sc->sc_bridge_id &&
1079 		    bif->bif_change_detection_enabled) {
1080 			bstp_topology_change_detection(sc);
1081 		}
1082 		bstp_configuration_update(sc);
1083 		bstp_port_state_selection(sc);
1084 		if (sc->sc_ifp->if_flags & IFF_LINK2)
1085 			bstp_adjust_bonded_states(sc, bif);
1086 	}
1087 }
1088 
1089 static void
1090 bstp_tcn_timer_expiry(struct bridge_softc *sc)
1091 {
1092 	bstp_transmit_tcn(sc);
1093 	bstp_timer_start(&sc->sc_tcn_timer, 0);
1094 }
1095 
1096 static void
1097 bstp_topology_change_timer_expiry(struct bridge_softc *sc)
1098 {
1099 	sc->sc_topology_change_detected = 0;
1100 	sc->sc_topology_change = 0;
1101 }
1102 
1103 static void
1104 bstp_hold_timer_expiry(struct bridge_softc *sc, struct bridge_iflist *bif)
1105 {
1106 	if (bif->bif_config_pending)
1107 		bstp_transmit_config(sc, bif);
1108 }
1109 
1110 /*
1111  * If no traffic received directly on this port for the specified
1112  * period with link1 set we go into a special blocking mode to
1113  * fail-over traffic to another port.
1114  */
1115 static void
1116 bstp_link1_timer_expiry(struct bridge_softc *sc, struct bridge_iflist *bif)
1117 {
1118 	if (sc->sc_ifp->if_flags & IFF_LINK1)
1119 		bstp_make_l1blocking(sc, bif);
1120 }
1121 
1122 static int
1123 bstp_addr_cmp(const uint8_t *a, const uint8_t *b)
1124 {
1125 	int i, d;
1126 
1127 	for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
1128 		d = ((int)a[i]) - ((int)b[i]);
1129 	}
1130 
1131 	return (d);
1132 }
1133 
1134 void
1135 bstp_initialization(struct bridge_softc *sc)
1136 {
1137 	struct bridge_iflist *bif, *mif, *nbif;
1138 	u_char *e_addr;
1139 
1140 	KKASSERT(&curthread->td_msgport == BRIDGE_CFGPORT);
1141 
1142 	/*
1143 	 * Figure out our bridge ID, use the lowest-valued MAC.
1144 	 * Include the bridge's own random MAC in the calculation.
1145 	 */
1146 	mif = NULL;
1147 
1148 	TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
1149 		if ((bif->bif_flags & IFBIF_STP) == 0)
1150 			continue;
1151 		if (bif->bif_ifp->if_type != IFT_ETHER)
1152 			continue;
1153 		if (mif == NULL) {
1154 			mif = bif;
1155 			continue;
1156 		}
1157 
1158 		bif->bif_port_id = (bif->bif_priority << 8) |
1159 				   (bif->bif_ifp->if_index & 0xff);
1160 		if (bstp_addr_cmp(IF_LLADDR(bif->bif_ifp),
1161 				  IF_LLADDR(mif->bif_ifp)) < 0) {
1162 			mif = bif;
1163 			continue;
1164 		}
1165 	}
1166 	if (mif == NULL) {
1167 		bstp_stop(sc);
1168 		return;
1169 	}
1170 
1171 	if (bstp_addr_cmp(IF_LLADDR(sc->sc_ifp), IF_LLADDR(mif->bif_ifp)) < 0)
1172 		e_addr = IF_LLADDR(sc->sc_ifp);
1173 	else
1174 		e_addr = IF_LLADDR(mif->bif_ifp);
1175 
1176 	sc->sc_bridge_id =
1177 	    (((uint64_t)sc->sc_bridge_priority) << 48) |
1178 	    (((uint64_t)e_addr[0]) << 40) |
1179 	    (((uint64_t)e_addr[1]) << 32) |
1180 	    (((uint64_t)e_addr[2]) << 24) |
1181 	    (((uint64_t)e_addr[3]) << 16) |
1182 	    (((uint64_t)e_addr[4]) << 8) |
1183 	    (((uint64_t)e_addr[5]));
1184 
1185 	/*
1186 	 * Remainder of setup.
1187 	 */
1188 
1189 	sc->sc_designated_root = sc->sc_bridge_id;
1190 	sc->sc_designated_cost = 0;
1191 	sc->sc_root_port = NULL;
1192 
1193 	sc->sc_max_age = sc->sc_bridge_max_age;
1194 	sc->sc_hello_time = sc->sc_bridge_hello_time;
1195 	sc->sc_forward_delay = sc->sc_bridge_forward_delay;
1196 	sc->sc_topology_change_detected = 0;
1197 	sc->sc_topology_change = 0;
1198 	bstp_timer_stop(&sc->sc_tcn_timer);
1199 	bstp_timer_stop(&sc->sc_topology_change_timer);
1200 
1201 	if (callout_pending(&sc->sc_bstpcallout) == 0)
1202 		callout_reset(&sc->sc_bstpcallout, hz,
1203 		    bstp_tick, sc);
1204 
1205 	TAILQ_FOREACH_MUTABLE(bif, &sc->sc_iflists[mycpuid], bif_next, nbif) {
1206 		if (sc->sc_ifp->if_flags & IFF_LINK1)
1207 			bstp_timer_start(&bif->bif_link1_timer, 0);
1208 		if (bif->bif_flags & IFBIF_STP)
1209 			bstp_ifupdstatus(sc, bif);
1210 		else
1211 			bstp_disable_port(sc, bif);
1212 
1213 		if (nbif != NULL && !nbif->bif_onlist) {
1214 			KKASSERT(bif->bif_onlist);
1215 			nbif = TAILQ_NEXT(bif, bif_next);
1216 		}
1217 	}
1218 
1219 	bstp_port_state_selection(sc);
1220 	bstp_config_bpdu_generation(sc);
1221 	bstp_timer_start(&sc->sc_hello_timer, 0);
1222 }
1223 
1224 void
1225 bstp_stop(struct bridge_softc *sc)
1226 {
1227 	struct bridge_iflist *bif;
1228 
1229 	KKASSERT(&curthread->td_msgport == BRIDGE_CFGPORT);
1230 
1231 	TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
1232 		bstp_set_port_state(bif, BSTP_IFSTATE_DISABLED);
1233 		bstp_timer_stop(&bif->bif_hold_timer);
1234 		bstp_timer_stop(&bif->bif_message_age_timer);
1235 		bstp_timer_stop(&bif->bif_forward_delay_timer);
1236 		bstp_timer_stop(&bif->bif_link1_timer);
1237 	}
1238 
1239 	callout_stop(&sc->sc_bstpcallout);
1240 
1241 	bstp_timer_stop(&sc->sc_topology_change_timer);
1242 	bstp_timer_stop(&sc->sc_tcn_timer);
1243 	bstp_timer_stop(&sc->sc_hello_timer);
1244 
1245 	crit_enter();
1246 	lwkt_dropmsg(&sc->sc_bstptimemsg.lmsg);
1247 	crit_exit();
1248 }
1249 
1250 /*
1251  * [re]initialize a port.  The port is initialized to a L1BLOCKING state
1252  * or a BLOCKING state.  When link ping/pong is enabled (LINK1) we start
1253  * out in the L1BLOCKING state to prevent flapping from blowing up the
1254  * state of the other ports.
1255  *
1256  * Either way the first config packet received will knock the port into
1257  * the LISTENING state.
1258  */
1259 static void
1260 bstp_initialize_port(struct bridge_softc *sc, struct bridge_iflist *bif)
1261 {
1262 	int needs_adjust = (bif->bif_state == BSTP_IFSTATE_FORWARDING ||
1263 			    bif->bif_state == BSTP_IFSTATE_BLOCKING ||
1264 			    bif->bif_state == BSTP_IFSTATE_BONDED);
1265 
1266 	if (sc->sc_ifp->if_flags & IFF_LINK1)
1267 		bstp_set_port_state(bif, BSTP_IFSTATE_L1BLOCKING);
1268 	else
1269 		bstp_set_port_state(bif, BSTP_IFSTATE_BLOCKING);
1270 	bstp_clear_peer_info(sc, bif);
1271 	bif->bif_topology_change_acknowledge = 0;
1272 	bif->bif_config_pending = 0;
1273 	bif->bif_change_detection_enabled = 1;
1274 	bstp_timer_stop(&bif->bif_message_age_timer);
1275 	bstp_timer_stop(&bif->bif_forward_delay_timer);
1276 	bstp_timer_stop(&bif->bif_hold_timer);
1277 	bstp_timer_stop(&bif->bif_link1_timer);
1278 	if (needs_adjust && (sc->sc_ifp->if_flags & IFF_LINK2))
1279 		bstp_adjust_bonded_states(sc, bif);
1280 }
1281 
1282 /*
1283  * When enabling a port that was previously disabled no configuration
1284  * update occurs, otherwise down/up or running/not-running flapping
1285  * (e.g. by openvpn on a TAP interface) will blow up the other ports.
1286  */
1287 static void
1288 bstp_enable_port(struct bridge_softc *sc, struct bridge_iflist *bif)
1289 {
1290 	int was_disabled = (bif->bif_state == BSTP_IFSTATE_DISABLED ||
1291 			    bif->bif_state == BSTP_IFSTATE_L1BLOCKING);
1292 
1293 	bstp_initialize_port(sc, bif);
1294 	if (sc->sc_ifp->if_flags & IFF_LINK1)
1295 		bstp_timer_start(&bif->bif_link1_timer, 0);
1296 	if (was_disabled == 0) {
1297 		bstp_configuration_update(sc);
1298 		bstp_port_state_selection(sc);
1299 	}
1300 	bstp_adjust_bonded_states(sc, bif);
1301 }
1302 
1303 /*
1304  * When disabling a port that was previously in a non-forwarding or bonded
1305  * state no configuration update occurs, otherwise down/up or
1306  * running/not-running flapping (e.g. by openvpn on a TAP interface) will
1307  * blow up the other ports.
1308  */
1309 static void
1310 bstp_disable_port(struct bridge_softc *sc, struct bridge_iflist *bif)
1311 {
1312 	int was_forwarding = (bif->bif_state == BSTP_IFSTATE_FORWARDING);
1313 	int was_bonded = (bif->bif_state == BSTP_IFSTATE_BONDED);
1314 	int iamroot;
1315 
1316 	iamroot = bstp_root_bridge(sc);
1317 
1318 	bstp_clear_peer_info(sc, bif);
1319 	bstp_set_port_state(bif, BSTP_IFSTATE_DISABLED);
1320 	bif->bif_topology_change_acknowledge = 0;
1321 	bif->bif_config_pending = 0;
1322 	bstp_timer_stop(&bif->bif_message_age_timer);
1323 	bstp_timer_stop(&bif->bif_forward_delay_timer);
1324 	bstp_timer_stop(&bif->bif_link1_timer);
1325 	if (was_forwarding || was_bonded) {
1326 		bstp_configuration_update(sc);
1327 		bstp_port_state_selection(sc);
1328 	}
1329 	bridge_rtdelete(sc, bif->bif_ifp, IFBF_FLUSHDYN);
1330 	if (was_forwarding && (sc->sc_ifp->if_flags & IFF_LINK2))
1331 		bstp_adjust_bonded_states(sc, bif);
1332 
1333 	if (iamroot == 0 && bstp_root_bridge(sc)) {
1334 		sc->sc_max_age = sc->sc_bridge_max_age;
1335 		sc->sc_hello_time = sc->sc_bridge_hello_time;
1336 		sc->sc_forward_delay = sc->sc_bridge_forward_delay;
1337 
1338 		bstp_topology_change_detection(sc);
1339 		bstp_timer_stop(&sc->sc_tcn_timer);
1340 		bstp_config_bpdu_generation(sc);
1341 		bstp_timer_start(&sc->sc_hello_timer, 0);
1342 	}
1343 }
1344 
1345 void
1346 bstp_linkstate(struct ifnet *ifp, int state)
1347 {
1348 	struct bridge_softc *sc;
1349 	struct bridge_iflist *bif;
1350 
1351 	sc = ifp->if_bridge;
1352 	ifnet_serialize_all(sc->sc_ifp);
1353 
1354 	/*
1355 	 * bstp_ifupdstatus() may block, but it is the last
1356 	 * operation of the member iface iteration, so we
1357 	 * don't need to use LIST_FOREACH_MUTABLE()+bif_onlist
1358 	 * check here.
1359 	 */
1360 	TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
1361 		if ((bif->bif_flags & IFBIF_STP) == 0)
1362 			continue;
1363 
1364 		if (bif->bif_ifp == ifp) {
1365 			bstp_ifupdstatus(sc, bif);
1366 			break;
1367 		}
1368 	}
1369 	ifnet_deserialize_all(sc->sc_ifp);
1370 }
1371 
1372 static void
1373 bstp_ifupdstatus(struct bridge_softc *sc, struct bridge_iflist *bif)
1374 {
1375 	struct ifnet *ifp = bif->bif_ifp;
1376 	struct ifmediareq ifmr;
1377 	int error = 0;
1378 
1379 	bzero((char *)&ifmr, sizeof(ifmr));
1380 	ifnet_serialize_all(ifp);
1381 	error = (*ifp->if_ioctl)(ifp, SIOCGIFMEDIA, (caddr_t)&ifmr, NULL);
1382 	ifnet_deserialize_all(ifp);
1383 
1384 	if ((error == 0) && (ifp->if_flags & IFF_UP)) {
1385 	 	if (ifmr.ifm_status & IFM_ACTIVE) {
1386 			if (bif->bif_state == BSTP_IFSTATE_DISABLED)
1387 				bstp_enable_port(sc, bif);
1388 
1389 		} else {
1390 			if (bif->bif_state != BSTP_IFSTATE_DISABLED)
1391 				bstp_disable_port(sc, bif);
1392 		}
1393 		return;
1394 	}
1395 
1396 	if (bif->bif_state != BSTP_IFSTATE_DISABLED)
1397 		bstp_disable_port(sc, bif);
1398 }
1399 
1400 static void
1401 bstp_tick(void *arg)
1402 {
1403 	struct bridge_softc *sc = arg;
1404 	struct lwkt_msg *lmsg;
1405 
1406 	KKASSERT(mycpuid == BRIDGE_CFGCPU);
1407 
1408 	crit_enter();
1409 
1410 	if (callout_pending(&sc->sc_bstpcallout) ||
1411 	    !callout_active(&sc->sc_bstpcallout)) {
1412 		crit_exit();
1413 		return;
1414 	}
1415 	callout_deactivate(&sc->sc_bstpcallout);
1416 
1417 	lmsg = &sc->sc_bstptimemsg.lmsg;
1418 	KKASSERT(lmsg->ms_flags & MSGF_DONE);
1419 	lwkt_sendmsg_oncpu(BRIDGE_CFGPORT, lmsg);
1420 
1421 	crit_exit();
1422 }
1423 
1424 void
1425 bstp_tick_handler(netmsg_t msg)
1426 {
1427 	struct bridge_softc *sc = msg->lmsg.u.ms_resultp;
1428 	struct bridge_iflist *bif;
1429 
1430 	KKASSERT(&curthread->td_msgport == BRIDGE_CFGPORT);
1431 	crit_enter();
1432 	/* Reply ASAP */
1433 	lwkt_replymsg(&msg->lmsg, 0);
1434 	crit_exit();
1435 
1436 	ifnet_serialize_all(sc->sc_ifp);
1437 
1438 	/*
1439 	 * NOTE:
1440 	 * We don't need to worry that member iface is ripped
1441 	 * from the per-cpu list during the blocking operation
1442 	 * in the loop body, since deletion is serialized by
1443 	 * BRIDGE_CFGPORT
1444 	 */
1445 
1446 	TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
1447 		if ((bif->bif_flags & IFBIF_STP) == 0)
1448 			continue;
1449 		/*
1450 		 * XXX This can cause a lag in "link does away"
1451 		 * XXX and "spanning tree gets updated".  We need
1452 		 * XXX come sort of callback from the link state
1453 		 * XXX update code to kick spanning tree.
1454 		 * XXX --thorpej@NetBSD.org
1455 		 */
1456 		bstp_ifupdstatus(sc, bif);
1457 	}
1458 
1459 	if (bstp_timer_expired(&sc->sc_hello_timer, sc->sc_hello_time))
1460 		bstp_hello_timer_expiry(sc);
1461 
1462 	if (bstp_timer_expired(&sc->sc_tcn_timer, sc->sc_bridge_hello_time))
1463 		bstp_tcn_timer_expiry(sc);
1464 
1465 	if (bstp_timer_expired(&sc->sc_topology_change_timer,
1466 	    sc->sc_topology_change_time))
1467 		bstp_topology_change_timer_expiry(sc);
1468 
1469 	TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
1470 		if ((bif->bif_flags & IFBIF_STP) == 0)
1471 			continue;
1472 		if (bstp_timer_expired(&bif->bif_message_age_timer,
1473 		    sc->sc_max_age))
1474 			bstp_message_age_timer_expiry(sc, bif);
1475 	}
1476 
1477 	TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
1478 		if ((bif->bif_flags & IFBIF_STP) == 0)
1479 			continue;
1480 		if (bstp_timer_expired(&bif->bif_forward_delay_timer,
1481 		    sc->sc_forward_delay))
1482 			bstp_forward_delay_timer_expiry(sc, bif);
1483 
1484 		if (bstp_timer_expired(&bif->bif_hold_timer,
1485 		    sc->sc_hold_time))
1486 			bstp_hold_timer_expiry(sc, bif);
1487 
1488 		if (bstp_timer_expired(&bif->bif_link1_timer,
1489 		    sc->sc_hello_time * 10))
1490 			bstp_link1_timer_expiry(sc, bif);
1491 	}
1492 
1493 	if (sc->sc_ifp->if_flags & IFF_RUNNING)
1494 		callout_reset(&sc->sc_bstpcallout, hz, bstp_tick, sc);
1495 
1496 	ifnet_deserialize_all(sc->sc_ifp);
1497 }
1498 
1499 static void
1500 bstp_timer_start(struct bridge_timer *t, uint16_t v)
1501 {
1502 	t->value = v;
1503 	t->active = 1;
1504 }
1505 
1506 static void
1507 bstp_timer_stop(struct bridge_timer *t)
1508 {
1509 	t->value = 0;
1510 	t->active = 0;
1511 }
1512 
1513 static int
1514 bstp_timer_expired(struct bridge_timer *t, uint16_t v)
1515 {
1516 	if (t->active == 0)
1517 		return (0);
1518 	t->value += BSTP_TICK_VAL;
1519 	if (t->value >= v) {
1520 		bstp_timer_stop(t);
1521 		return (1);
1522 	}
1523 	return (0);
1524 
1525 }
1526