xref: /netbsd-src/sys/net/if_bridge.c (revision b7b7574d3bf8eeb51a1fa3977b59142ec6434a55)
1 /*	$NetBSD: if_bridge.c,v 1.86 2014/07/02 09:51:08 ozaki-r Exp $	*/
2 
3 /*
4  * Copyright 2001 Wasabi Systems, Inc.
5  * All rights reserved.
6  *
7  * Written by Jason R. Thorpe for Wasabi Systems, Inc.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. All advertising materials mentioning features or use of this software
18  *    must display the following acknowledgement:
19  *	This product includes software developed for the NetBSD Project by
20  *	Wasabi Systems, Inc.
21  * 4. The name of Wasabi Systems, Inc. may not be used to endorse
22  *    or promote products derived from this software without specific prior
23  *    written permission.
24  *
25  * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
27  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
28  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL WASABI SYSTEMS, INC
29  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
32  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
33  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
34  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
35  * POSSIBILITY OF SUCH DAMAGE.
36  */
37 
38 /*
39  * Copyright (c) 1999, 2000 Jason L. Wright (jason@thought.net)
40  * All rights reserved.
41  *
42  * Redistribution and use in source and binary forms, with or without
43  * modification, are permitted provided that the following conditions
44  * are met:
45  * 1. Redistributions of source code must retain the above copyright
46  *    notice, this list of conditions and the following disclaimer.
47  * 2. Redistributions in binary form must reproduce the above copyright
48  *    notice, this list of conditions and the following disclaimer in the
49  *    documentation and/or other materials provided with the distribution.
50  * 3. All advertising materials mentioning features or use of this software
51  *    must display the following acknowledgement:
52  *	This product includes software developed by Jason L. Wright
53  * 4. The name of the author may not be used to endorse or promote products
54  *    derived from this software without specific prior written permission.
55  *
56  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
57  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
58  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
59  * DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
60  * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
61  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
62  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
63  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
64  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
65  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
66  * POSSIBILITY OF SUCH DAMAGE.
67  *
68  * OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun Exp
69  */
70 
71 /*
72  * Network interface bridge support.
73  *
74  * TODO:
75  *
76  *	- Currently only supports Ethernet-like interfaces (Ethernet,
77  *	  802.11, VLANs on Ethernet, etc.)  Figure out a nice way
78  *	  to bridge other types of interfaces (FDDI-FDDI, and maybe
79  *	  consider heterogenous bridges).
80  */
81 
82 #include <sys/cdefs.h>
83 __KERNEL_RCSID(0, "$NetBSD: if_bridge.c,v 1.86 2014/07/02 09:51:08 ozaki-r Exp $");
84 
85 #ifdef _KERNEL_OPT
86 #include "opt_bridge_ipf.h"
87 #include "opt_inet.h"
88 #endif /* _KERNEL_OPT */
89 
90 #include <sys/param.h>
91 #include <sys/kernel.h>
92 #include <sys/mbuf.h>
93 #include <sys/queue.h>
94 #include <sys/socket.h>
95 #include <sys/socketvar.h> /* for softnet_lock */
96 #include <sys/sockio.h>
97 #include <sys/systm.h>
98 #include <sys/proc.h>
99 #include <sys/pool.h>
100 #include <sys/kauth.h>
101 #include <sys/cpu.h>
102 #include <sys/cprng.h>
103 #include <sys/mutex.h>
104 
105 #include <net/bpf.h>
106 #include <net/if.h>
107 #include <net/if_dl.h>
108 #include <net/if_types.h>
109 #include <net/if_llc.h>
110 #include <net/pktqueue.h>
111 
112 #include <net/if_ether.h>
113 #include <net/if_bridgevar.h>
114 
115 #if defined(BRIDGE_IPF)
116 /* Used for bridge_ip[6]_checkbasic */
117 #include <netinet/in.h>
118 #include <netinet/in_systm.h>
119 #include <netinet/ip.h>
120 #include <netinet/ip_var.h>
121 #include <netinet/ip_private.h>		/* XXX */
122 
123 #include <netinet/ip6.h>
124 #include <netinet6/in6_var.h>
125 #include <netinet6/ip6_var.h>
126 #include <netinet6/ip6_private.h>	/* XXX */
127 #endif /* BRIDGE_IPF */
128 
129 /*
130  * Size of the route hash table.  Must be a power of two.
131  */
132 #ifndef BRIDGE_RTHASH_SIZE
133 #define	BRIDGE_RTHASH_SIZE		1024
134 #endif
135 
136 #define	BRIDGE_RTHASH_MASK		(BRIDGE_RTHASH_SIZE - 1)
137 
138 #include "carp.h"
139 #if NCARP > 0
140 #include <netinet/in.h>
141 #include <netinet/in_var.h>
142 #include <netinet/ip_carp.h>
143 #endif
144 
145 /*
146  * Maximum number of addresses to cache.
147  */
148 #ifndef BRIDGE_RTABLE_MAX
149 #define	BRIDGE_RTABLE_MAX		100
150 #endif
151 
152 /*
153  * Spanning tree defaults.
154  */
155 #define	BSTP_DEFAULT_MAX_AGE		(20 * 256)
156 #define	BSTP_DEFAULT_HELLO_TIME		(2 * 256)
157 #define	BSTP_DEFAULT_FORWARD_DELAY	(15 * 256)
158 #define	BSTP_DEFAULT_HOLD_TIME		(1 * 256)
159 #define	BSTP_DEFAULT_BRIDGE_PRIORITY	0x8000
160 #define	BSTP_DEFAULT_PORT_PRIORITY	0x80
161 #define	BSTP_DEFAULT_PATH_COST		55
162 
163 /*
164  * Timeout (in seconds) for entries learned dynamically.
165  */
166 #ifndef BRIDGE_RTABLE_TIMEOUT
167 #define	BRIDGE_RTABLE_TIMEOUT		(20 * 60)	/* same as ARP */
168 #endif
169 
170 /*
171  * Number of seconds between walks of the route list.
172  */
173 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
174 #define	BRIDGE_RTABLE_PRUNE_PERIOD	(5 * 60)
175 #endif
176 
177 int	bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
178 
179 static struct pool bridge_rtnode_pool;
180 
181 void	bridgeattach(int);
182 
183 static int	bridge_clone_create(struct if_clone *, int);
184 static int	bridge_clone_destroy(struct ifnet *);
185 
186 static int	bridge_ioctl(struct ifnet *, u_long, void *);
187 static int	bridge_init(struct ifnet *);
188 static void	bridge_stop(struct ifnet *, int);
189 static void	bridge_start(struct ifnet *);
190 
191 static void	bridge_input(struct ifnet *, struct mbuf *);
192 static void	bridge_forward(void *);
193 
194 static void	bridge_timer(void *);
195 
196 static void	bridge_broadcast(struct bridge_softc *, struct ifnet *,
197 				 struct mbuf *);
198 
199 static int	bridge_rtupdate(struct bridge_softc *, const uint8_t *,
200 				struct ifnet *, int, uint8_t);
201 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *);
202 static void	bridge_rttrim(struct bridge_softc *);
203 static void	bridge_rtage(struct bridge_softc *);
204 static void	bridge_rtflush(struct bridge_softc *, int);
205 static int	bridge_rtdaddr(struct bridge_softc *, const uint8_t *);
206 static void	bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp);
207 
208 static int	bridge_rtable_init(struct bridge_softc *);
209 static void	bridge_rtable_fini(struct bridge_softc *);
210 
211 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
212 						  const uint8_t *);
213 static int	bridge_rtnode_insert(struct bridge_softc *,
214 				     struct bridge_rtnode *);
215 static void	bridge_rtnode_destroy(struct bridge_softc *,
216 				      struct bridge_rtnode *);
217 
218 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
219 						  const char *name);
220 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
221 						     struct ifnet *ifp);
222 static void	bridge_delete_member(struct bridge_softc *,
223 				     struct bridge_iflist *);
224 
225 static int	bridge_ioctl_add(struct bridge_softc *, void *);
226 static int	bridge_ioctl_del(struct bridge_softc *, void *);
227 static int	bridge_ioctl_gifflags(struct bridge_softc *, void *);
228 static int	bridge_ioctl_sifflags(struct bridge_softc *, void *);
229 static int	bridge_ioctl_scache(struct bridge_softc *, void *);
230 static int	bridge_ioctl_gcache(struct bridge_softc *, void *);
231 static int	bridge_ioctl_gifs(struct bridge_softc *, void *);
232 static int	bridge_ioctl_rts(struct bridge_softc *, void *);
233 static int	bridge_ioctl_saddr(struct bridge_softc *, void *);
234 static int	bridge_ioctl_sto(struct bridge_softc *, void *);
235 static int	bridge_ioctl_gto(struct bridge_softc *, void *);
236 static int	bridge_ioctl_daddr(struct bridge_softc *, void *);
237 static int	bridge_ioctl_flush(struct bridge_softc *, void *);
238 static int	bridge_ioctl_gpri(struct bridge_softc *, void *);
239 static int	bridge_ioctl_spri(struct bridge_softc *, void *);
240 static int	bridge_ioctl_ght(struct bridge_softc *, void *);
241 static int	bridge_ioctl_sht(struct bridge_softc *, void *);
242 static int	bridge_ioctl_gfd(struct bridge_softc *, void *);
243 static int	bridge_ioctl_sfd(struct bridge_softc *, void *);
244 static int	bridge_ioctl_gma(struct bridge_softc *, void *);
245 static int	bridge_ioctl_sma(struct bridge_softc *, void *);
246 static int	bridge_ioctl_sifprio(struct bridge_softc *, void *);
247 static int	bridge_ioctl_sifcost(struct bridge_softc *, void *);
248 #if defined(BRIDGE_IPF)
249 static int	bridge_ioctl_gfilt(struct bridge_softc *, void *);
250 static int	bridge_ioctl_sfilt(struct bridge_softc *, void *);
251 static int	bridge_ipf(void *, struct mbuf **, struct ifnet *, int);
252 static int	bridge_ip_checkbasic(struct mbuf **mp);
253 # ifdef INET6
254 static int	bridge_ip6_checkbasic(struct mbuf **mp);
255 # endif /* INET6 */
256 #endif /* BRIDGE_IPF */
257 
258 static void bridge_sysctl_fwdq_setup(struct sysctllog **clog,
259     struct bridge_softc *sc);
260 
261 struct bridge_control {
262 	int	(*bc_func)(struct bridge_softc *, void *);
263 	int	bc_argsize;
264 	int	bc_flags;
265 };
266 
267 #define	BC_F_COPYIN		0x01	/* copy arguments in */
268 #define	BC_F_COPYOUT		0x02	/* copy arguments out */
269 #define	BC_F_SUSER		0x04	/* do super-user check */
270 
271 static const struct bridge_control bridge_control_table[] = {
272 [BRDGADD] = {bridge_ioctl_add, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
273 [BRDGDEL] = {bridge_ioctl_del, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
274 
275 [BRDGGIFFLGS] = {bridge_ioctl_gifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_COPYOUT},
276 [BRDGSIFFLGS] = {bridge_ioctl_sifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
277 
278 [BRDGSCACHE] = {bridge_ioctl_scache, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
279 [BRDGGCACHE] = {bridge_ioctl_gcache, sizeof(struct ifbrparam), BC_F_COPYOUT},
280 
281 [BRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_COPYIN|BC_F_COPYOUT},
282 [BRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_COPYIN|BC_F_COPYOUT},
283 
284 [BRDGSADDR] = {bridge_ioctl_saddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
285 
286 [BRDGSTO] = {bridge_ioctl_sto, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
287 [BRDGGTO] = {bridge_ioctl_gto, sizeof(struct ifbrparam), BC_F_COPYOUT},
288 
289 [BRDGDADDR] = {bridge_ioctl_daddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
290 
291 [BRDGFLUSH] = {bridge_ioctl_flush, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
292 
293 [BRDGGPRI] = {bridge_ioctl_gpri, sizeof(struct ifbrparam), BC_F_COPYOUT},
294 [BRDGSPRI] = {bridge_ioctl_spri, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
295 
296 [BRDGGHT] = {bridge_ioctl_ght, sizeof(struct ifbrparam), BC_F_COPYOUT},
297 [BRDGSHT] = {bridge_ioctl_sht, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
298 
299 [BRDGGFD] = {bridge_ioctl_gfd, sizeof(struct ifbrparam), BC_F_COPYOUT},
300 [BRDGSFD] = {bridge_ioctl_sfd, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
301 
302 [BRDGGMA] = {bridge_ioctl_gma, sizeof(struct ifbrparam), BC_F_COPYOUT},
303 [BRDGSMA] = {bridge_ioctl_sma, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
304 
305 [BRDGSIFPRIO] = {bridge_ioctl_sifprio, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
306 
307 [BRDGSIFCOST] = {bridge_ioctl_sifcost, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
308 #if defined(BRIDGE_IPF)
309 [BRDGGFILT] = {bridge_ioctl_gfilt, sizeof(struct ifbrparam), BC_F_COPYOUT},
310 [BRDGSFILT] = {bridge_ioctl_sfilt, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
311 #endif /* BRIDGE_IPF */
312 };
313 static const int bridge_control_table_size = __arraycount(bridge_control_table);
314 
315 static LIST_HEAD(, bridge_softc) bridge_list;
316 static kmutex_t bridge_list_lock;
317 
318 static struct if_clone bridge_cloner =
319     IF_CLONE_INITIALIZER("bridge", bridge_clone_create, bridge_clone_destroy);
320 
321 /*
322  * bridgeattach:
323  *
324  *	Pseudo-device attach routine.
325  */
326 void
327 bridgeattach(int n)
328 {
329 
330 	pool_init(&bridge_rtnode_pool, sizeof(struct bridge_rtnode),
331 	    0, 0, 0, "brtpl", NULL, IPL_NET);
332 
333 	LIST_INIT(&bridge_list);
334 	mutex_init(&bridge_list_lock, MUTEX_DEFAULT, IPL_NET);
335 	if_clone_attach(&bridge_cloner);
336 }
337 
338 /*
339  * bridge_clone_create:
340  *
341  *	Create a new bridge instance.
342  */
343 static int
344 bridge_clone_create(struct if_clone *ifc, int unit)
345 {
346 	struct bridge_softc *sc;
347 	struct ifnet *ifp;
348 
349 	sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
350 	ifp = &sc->sc_if;
351 
352 	sc->sc_brtmax = BRIDGE_RTABLE_MAX;
353 	sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
354 	sc->sc_bridge_max_age = BSTP_DEFAULT_MAX_AGE;
355 	sc->sc_bridge_hello_time = BSTP_DEFAULT_HELLO_TIME;
356 	sc->sc_bridge_forward_delay = BSTP_DEFAULT_FORWARD_DELAY;
357 	sc->sc_bridge_priority = BSTP_DEFAULT_BRIDGE_PRIORITY;
358 	sc->sc_hold_time = BSTP_DEFAULT_HOLD_TIME;
359 	sc->sc_filter_flags = 0;
360 
361 	/* Initialize our routing table. */
362 	bridge_rtable_init(sc);
363 
364 	callout_init(&sc->sc_brcallout, 0);
365 	callout_init(&sc->sc_bstpcallout, 0);
366 
367 	LIST_INIT(&sc->sc_iflist);
368 
369 	if_initname(ifp, ifc->ifc_name, unit);
370 	ifp->if_softc = sc;
371 	ifp->if_mtu = ETHERMTU;
372 	ifp->if_ioctl = bridge_ioctl;
373 	ifp->if_output = bridge_output;
374 	ifp->if_start = bridge_start;
375 	ifp->if_stop = bridge_stop;
376 	ifp->if_init = bridge_init;
377 	ifp->if_type = IFT_BRIDGE;
378 	ifp->if_addrlen = 0;
379 	ifp->if_dlt = DLT_EN10MB;
380 	ifp->if_hdrlen = ETHER_HDR_LEN;
381 
382 	sc->sc_fwd_pktq = pktq_create(IFQ_MAXLEN, bridge_forward, sc);
383 	KASSERT(sc->sc_fwd_pktq != NULL);
384 
385 	bridge_sysctl_fwdq_setup(&ifp->if_sysctl_log, sc);
386 
387 	if_attach(ifp);
388 
389 	if_alloc_sadl(ifp);
390 
391 	mutex_enter(&bridge_list_lock);
392 	LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
393 	mutex_exit(&bridge_list_lock);
394 
395 	return (0);
396 }
397 
398 /*
399  * bridge_clone_destroy:
400  *
401  *	Destroy a bridge instance.
402  */
403 static int
404 bridge_clone_destroy(struct ifnet *ifp)
405 {
406 	struct bridge_softc *sc = ifp->if_softc;
407 	struct bridge_iflist *bif;
408 	int s;
409 
410 	/* Must be called during IFF_RUNNING, i.e., before bridge_stop */
411 	pktq_barrier(sc->sc_fwd_pktq);
412 
413 	s = splnet();
414 
415 	bridge_stop(ifp, 1);
416 
417 	while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
418 		bridge_delete_member(sc, bif);
419 
420 	mutex_enter(&bridge_list_lock);
421 	LIST_REMOVE(sc, sc_list);
422 	mutex_exit(&bridge_list_lock);
423 
424 	splx(s);
425 
426 	if_detach(ifp);
427 
428 	/* Should be called after if_detach for safe */
429 	pktq_flush(sc->sc_fwd_pktq);
430 	pktq_destroy(sc->sc_fwd_pktq);
431 
432 	/* Tear down the routing table. */
433 	bridge_rtable_fini(sc);
434 
435 	free(sc, M_DEVBUF);
436 
437 	return (0);
438 }
439 
440 static int
441 bridge_sysctl_fwdq_maxlen(SYSCTLFN_ARGS)
442 {
443 	struct sysctlnode node = *rnode;
444 	const struct bridge_softc *sc =	node.sysctl_data;
445 	return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), sc->sc_fwd_pktq);
446 }
447 
448 #define	SYSCTL_BRIDGE_PKTQ(cn, c)					\
449 	static int							\
450 	bridge_sysctl_fwdq_##cn(SYSCTLFN_ARGS)				\
451 	{								\
452 		struct sysctlnode node = *rnode;			\
453 		const struct bridge_softc *sc =	node.sysctl_data;	\
454 		return sysctl_pktq_count(SYSCTLFN_CALL(rnode),		\
455 					 sc->sc_fwd_pktq, c);		\
456 	}
457 
458 SYSCTL_BRIDGE_PKTQ(items, PKTQ_NITEMS)
459 SYSCTL_BRIDGE_PKTQ(drops, PKTQ_DROPS)
460 
461 static void
462 bridge_sysctl_fwdq_setup(struct sysctllog **clog, struct bridge_softc *sc)
463 {
464 	const struct sysctlnode *cnode, *rnode;
465 	sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL;
466 	const char *ifname = sc->sc_if.if_xname;
467 
468 	len_func = bridge_sysctl_fwdq_items;
469 	maxlen_func = bridge_sysctl_fwdq_maxlen;
470 	drops_func = bridge_sysctl_fwdq_drops;
471 
472 	if (sysctl_createv(clog, 0, NULL, &rnode,
473 			   CTLFLAG_PERMANENT,
474 			   CTLTYPE_NODE, "interfaces",
475 			   SYSCTL_DESCR("Per-interface controls"),
476 			   NULL, 0, NULL, 0,
477 			   CTL_NET, CTL_CREATE, CTL_EOL) != 0)
478 		goto bad;
479 
480 	if (sysctl_createv(clog, 0, &rnode, &rnode,
481 			   CTLFLAG_PERMANENT,
482 			   CTLTYPE_NODE, ifname,
483 			   SYSCTL_DESCR("Interface controls"),
484 			   NULL, 0, NULL, 0,
485 			   CTL_CREATE, CTL_EOL) != 0)
486 		goto bad;
487 
488 	if (sysctl_createv(clog, 0, &rnode, &rnode,
489 			   CTLFLAG_PERMANENT,
490 			   CTLTYPE_NODE, "fwdq",
491 			   SYSCTL_DESCR("Protocol input queue controls"),
492 			   NULL, 0, NULL, 0,
493 			   CTL_CREATE, CTL_EOL) != 0)
494 		goto bad;
495 
496 	if (sysctl_createv(clog, 0, &rnode, &cnode,
497 			   CTLFLAG_PERMANENT,
498 			   CTLTYPE_INT, "len",
499 			   SYSCTL_DESCR("Current forwarding queue length"),
500 			   len_func, 0, (void *)sc, 0,
501 			   CTL_CREATE, IFQCTL_LEN, CTL_EOL) != 0)
502 		goto bad;
503 
504 	if (sysctl_createv(clog, 0, &rnode, &cnode,
505 			   CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
506 			   CTLTYPE_INT, "maxlen",
507 			   SYSCTL_DESCR("Maximum allowed forwarding queue length"),
508 			   maxlen_func, 0, (void *)sc, 0,
509 			   CTL_CREATE, IFQCTL_MAXLEN, CTL_EOL) != 0)
510 		goto bad;
511 
512 	if (sysctl_createv(clog, 0, &rnode, &cnode,
513 			   CTLFLAG_PERMANENT,
514 			   CTLTYPE_INT, "drops",
515 			   SYSCTL_DESCR("Packets dropped due to full forwarding queue"),
516 			   drops_func, 0, (void *)sc, 0,
517 			   CTL_CREATE, IFQCTL_DROPS, CTL_EOL) != 0)
518 		goto bad;
519 
520 	return;
521 bad:
522 	aprint_error("%s: could not attach sysctl nodes\n", ifname);
523 	return;
524 }
525 
526 /*
527  * bridge_ioctl:
528  *
529  *	Handle a control request from the operator.
530  */
531 static int
532 bridge_ioctl(struct ifnet *ifp, u_long cmd, void *data)
533 {
534 	struct bridge_softc *sc = ifp->if_softc;
535 	struct lwp *l = curlwp;	/* XXX */
536 	union {
537 		struct ifbreq ifbreq;
538 		struct ifbifconf ifbifconf;
539 		struct ifbareq ifbareq;
540 		struct ifbaconf ifbaconf;
541 		struct ifbrparam ifbrparam;
542 	} args;
543 	struct ifdrv *ifd = (struct ifdrv *) data;
544 	const struct bridge_control *bc = NULL; /* XXXGCC */
545 	int s, error = 0;
546 
547 	/* Authorize command before calling splnet(). */
548 	switch (cmd) {
549 	case SIOCGDRVSPEC:
550 	case SIOCSDRVSPEC:
551 		if (ifd->ifd_cmd >= bridge_control_table_size) {
552 			error = EINVAL;
553 			return error;
554 		}
555 
556 		bc = &bridge_control_table[ifd->ifd_cmd];
557 
558 		/* We only care about BC_F_SUSER at this point. */
559 		if ((bc->bc_flags & BC_F_SUSER) == 0)
560 			break;
561 
562 		error = kauth_authorize_network(l->l_cred,
563 		    KAUTH_NETWORK_INTERFACE_BRIDGE,
564 		    cmd == SIOCGDRVSPEC ?
565 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_GETPRIV :
566 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_SETPRIV,
567 		     ifd, NULL, NULL);
568 		if (error)
569 			return (error);
570 
571 		break;
572 	}
573 
574 	s = splnet();
575 
576 	switch (cmd) {
577 	case SIOCGDRVSPEC:
578 	case SIOCSDRVSPEC:
579 		KASSERT(bc != NULL);
580 		if (cmd == SIOCGDRVSPEC &&
581 		    (bc->bc_flags & BC_F_COPYOUT) == 0) {
582 			error = EINVAL;
583 			break;
584 		}
585 		else if (cmd == SIOCSDRVSPEC &&
586 		    (bc->bc_flags & BC_F_COPYOUT) != 0) {
587 			error = EINVAL;
588 			break;
589 		}
590 
591 		/* BC_F_SUSER is checked above, before splnet(). */
592 
593 		if (ifd->ifd_len != bc->bc_argsize ||
594 		    ifd->ifd_len > sizeof(args)) {
595 			error = EINVAL;
596 			break;
597 		}
598 
599 		memset(&args, 0, sizeof(args));
600 		if (bc->bc_flags & BC_F_COPYIN) {
601 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
602 			if (error)
603 				break;
604 		}
605 
606 		error = (*bc->bc_func)(sc, &args);
607 		if (error)
608 			break;
609 
610 		if (bc->bc_flags & BC_F_COPYOUT)
611 			error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
612 
613 		break;
614 
615 	case SIOCSIFFLAGS:
616 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
617 			break;
618 		switch (ifp->if_flags & (IFF_UP|IFF_RUNNING)) {
619 		case IFF_RUNNING:
620 			/*
621 			 * If interface is marked down and it is running,
622 			 * then stop and disable it.
623 			 */
624 			(*ifp->if_stop)(ifp, 1);
625 			break;
626 		case IFF_UP:
627 			/*
628 			 * If interface is marked up and it is stopped, then
629 			 * start it.
630 			 */
631 			error = (*ifp->if_init)(ifp);
632 			break;
633 		default:
634 			break;
635 		}
636 		break;
637 
638 	default:
639 		error = ifioctl_common(ifp, cmd, data);
640 		break;
641 	}
642 
643 	splx(s);
644 
645 	return (error);
646 }
647 
648 /*
649  * bridge_lookup_member:
650  *
651  *	Lookup a bridge member interface.  Must be called at splnet().
652  */
653 static struct bridge_iflist *
654 bridge_lookup_member(struct bridge_softc *sc, const char *name)
655 {
656 	struct bridge_iflist *bif;
657 	struct ifnet *ifp;
658 
659 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
660 		ifp = bif->bif_ifp;
661 		if (strcmp(ifp->if_xname, name) == 0)
662 			return (bif);
663 	}
664 
665 	return (NULL);
666 }
667 
668 /*
669  * bridge_lookup_member_if:
670  *
671  *	Lookup a bridge member interface by ifnet*.  Must be called at splnet().
672  */
673 static struct bridge_iflist *
674 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
675 {
676 	struct bridge_iflist *bif;
677 
678 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
679 		if (bif->bif_ifp == member_ifp)
680 			return (bif);
681 	}
682 
683 	return (NULL);
684 }
685 
686 /*
687  * bridge_delete_member:
688  *
689  *	Delete the specified member interface.
690  */
691 static void
692 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif)
693 {
694 	struct ifnet *ifs = bif->bif_ifp;
695 
696 	switch (ifs->if_type) {
697 	case IFT_ETHER:
698 		/*
699 		 * Take the interface out of promiscuous mode.
700 		 */
701 		(void) ifpromisc(ifs, 0);
702 		break;
703 	default:
704 #ifdef DIAGNOSTIC
705 		panic("bridge_delete_member: impossible");
706 #endif
707 		break;
708 	}
709 
710 	ifs->if_input = ether_input;
711 	ifs->if_bridge = NULL;
712 	LIST_REMOVE(bif, bif_next);
713 
714 	bridge_rtdelete(sc, ifs);
715 
716 	free(bif, M_DEVBUF);
717 
718 	if (sc->sc_if.if_flags & IFF_RUNNING)
719 		bstp_initialization(sc);
720 }
721 
722 static int
723 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
724 {
725 	struct ifbreq *req = arg;
726 	struct bridge_iflist *bif = NULL;
727 	struct ifnet *ifs;
728 	int error = 0;
729 
730 	ifs = ifunit(req->ifbr_ifsname);
731 	if (ifs == NULL)
732 		return (ENOENT);
733 
734 	if (sc->sc_if.if_mtu != ifs->if_mtu)
735 		return (EINVAL);
736 
737 	if (ifs->if_bridge == sc)
738 		return (EEXIST);
739 
740 	if (ifs->if_bridge != NULL)
741 		return (EBUSY);
742 
743 	if (ifs->if_input != ether_input)
744 		return EINVAL;
745 
746 	bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT);
747 	if (bif == NULL)
748 		return (ENOMEM);
749 
750 	switch (ifs->if_type) {
751 	case IFT_ETHER:
752 		/*
753 		 * Place the interface into promiscuous mode.
754 		 */
755 		error = ifpromisc(ifs, 1);
756 		if (error)
757 			goto out;
758 		break;
759 	default:
760 		error = EINVAL;
761 		goto out;
762 	}
763 
764 	bif->bif_ifp = ifs;
765 	bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
766 	bif->bif_priority = BSTP_DEFAULT_PORT_PRIORITY;
767 	bif->bif_path_cost = BSTP_DEFAULT_PATH_COST;
768 
769 	ifs->if_bridge = sc;
770 	LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
771 	ifs->if_input = bridge_input;
772 
773 	if (sc->sc_if.if_flags & IFF_RUNNING)
774 		bstp_initialization(sc);
775 	else
776 		bstp_stop(sc);
777 
778  out:
779 	if (error) {
780 		if (bif != NULL)
781 			free(bif, M_DEVBUF);
782 	}
783 	return (error);
784 }
785 
786 static int
787 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
788 {
789 	struct ifbreq *req = arg;
790 	struct bridge_iflist *bif;
791 
792 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
793 	if (bif == NULL)
794 		return (ENOENT);
795 
796 	bridge_delete_member(sc, bif);
797 
798 	return (0);
799 }
800 
801 static int
802 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
803 {
804 	struct ifbreq *req = arg;
805 	struct bridge_iflist *bif;
806 
807 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
808 	if (bif == NULL)
809 		return (ENOENT);
810 
811 	req->ifbr_ifsflags = bif->bif_flags;
812 	req->ifbr_state = bif->bif_state;
813 	req->ifbr_priority = bif->bif_priority;
814 	req->ifbr_path_cost = bif->bif_path_cost;
815 	req->ifbr_portno = bif->bif_ifp->if_index & 0xff;
816 
817 	return (0);
818 }
819 
820 static int
821 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
822 {
823 	struct ifbreq *req = arg;
824 	struct bridge_iflist *bif;
825 
826 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
827 	if (bif == NULL)
828 		return (ENOENT);
829 
830 	if (req->ifbr_ifsflags & IFBIF_STP) {
831 		switch (bif->bif_ifp->if_type) {
832 		case IFT_ETHER:
833 			/* These can do spanning tree. */
834 			break;
835 
836 		default:
837 			/* Nothing else can. */
838 			return (EINVAL);
839 		}
840 	}
841 
842 	bif->bif_flags = req->ifbr_ifsflags;
843 
844 	if (sc->sc_if.if_flags & IFF_RUNNING)
845 		bstp_initialization(sc);
846 
847 	return (0);
848 }
849 
850 static int
851 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
852 {
853 	struct ifbrparam *param = arg;
854 
855 	sc->sc_brtmax = param->ifbrp_csize;
856 	bridge_rttrim(sc);
857 
858 	return (0);
859 }
860 
861 static int
862 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
863 {
864 	struct ifbrparam *param = arg;
865 
866 	param->ifbrp_csize = sc->sc_brtmax;
867 
868 	return (0);
869 }
870 
871 static int
872 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
873 {
874 	struct ifbifconf *bifc = arg;
875 	struct bridge_iflist *bif;
876 	struct ifbreq breq;
877 	int count, len, error = 0;
878 
879 	count = 0;
880 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
881 		count++;
882 
883 	if (bifc->ifbic_len == 0) {
884 		bifc->ifbic_len = sizeof(breq) * count;
885 		return (0);
886 	}
887 
888 	count = 0;
889 	len = bifc->ifbic_len;
890 	memset(&breq, 0, sizeof breq);
891 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
892 		if (len < sizeof(breq))
893 			break;
894 
895 		strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
896 		    sizeof(breq.ifbr_ifsname));
897 		breq.ifbr_ifsflags = bif->bif_flags;
898 		breq.ifbr_state = bif->bif_state;
899 		breq.ifbr_priority = bif->bif_priority;
900 		breq.ifbr_path_cost = bif->bif_path_cost;
901 		breq.ifbr_portno = bif->bif_ifp->if_index & 0xff;
902 		error = copyout(&breq, bifc->ifbic_req + count, sizeof(breq));
903 		if (error)
904 			break;
905 		count++;
906 		len -= sizeof(breq);
907 	}
908 
909 	bifc->ifbic_len = sizeof(breq) * count;
910 	return (error);
911 }
912 
913 static int
914 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
915 {
916 	struct ifbaconf *bac = arg;
917 	struct bridge_rtnode *brt;
918 	struct ifbareq bareq;
919 	int count = 0, error = 0, len;
920 
921 	if (bac->ifbac_len == 0)
922 		return (0);
923 
924 	len = bac->ifbac_len;
925 	LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
926 		if (len < sizeof(bareq))
927 			goto out;
928 		memset(&bareq, 0, sizeof(bareq));
929 		strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
930 		    sizeof(bareq.ifba_ifsname));
931 		memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
932 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
933 			bareq.ifba_expire = brt->brt_expire - time_uptime;
934 		} else
935 			bareq.ifba_expire = 0;
936 		bareq.ifba_flags = brt->brt_flags;
937 
938 		error = copyout(&bareq, bac->ifbac_req + count, sizeof(bareq));
939 		if (error)
940 			goto out;
941 		count++;
942 		len -= sizeof(bareq);
943 	}
944  out:
945 	bac->ifbac_len = sizeof(bareq) * count;
946 	return (error);
947 }
948 
949 static int
950 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
951 {
952 	struct ifbareq *req = arg;
953 	struct bridge_iflist *bif;
954 	int error;
955 
956 	bif = bridge_lookup_member(sc, req->ifba_ifsname);
957 	if (bif == NULL)
958 		return (ENOENT);
959 
960 	error = bridge_rtupdate(sc, req->ifba_dst, bif->bif_ifp, 1,
961 	    req->ifba_flags);
962 
963 	return (error);
964 }
965 
966 static int
967 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
968 {
969 	struct ifbrparam *param = arg;
970 
971 	sc->sc_brttimeout = param->ifbrp_ctime;
972 
973 	return (0);
974 }
975 
976 static int
977 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
978 {
979 	struct ifbrparam *param = arg;
980 
981 	param->ifbrp_ctime = sc->sc_brttimeout;
982 
983 	return (0);
984 }
985 
986 static int
987 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
988 {
989 	struct ifbareq *req = arg;
990 
991 	return (bridge_rtdaddr(sc, req->ifba_dst));
992 }
993 
994 static int
995 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
996 {
997 	struct ifbreq *req = arg;
998 
999 	bridge_rtflush(sc, req->ifbr_ifsflags);
1000 
1001 	return (0);
1002 }
1003 
1004 static int
1005 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1006 {
1007 	struct ifbrparam *param = arg;
1008 
1009 	param->ifbrp_prio = sc->sc_bridge_priority;
1010 
1011 	return (0);
1012 }
1013 
1014 static int
1015 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1016 {
1017 	struct ifbrparam *param = arg;
1018 
1019 	sc->sc_bridge_priority = param->ifbrp_prio;
1020 
1021 	if (sc->sc_if.if_flags & IFF_RUNNING)
1022 		bstp_initialization(sc);
1023 
1024 	return (0);
1025 }
1026 
1027 static int
1028 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1029 {
1030 	struct ifbrparam *param = arg;
1031 
1032 	param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8;
1033 
1034 	return (0);
1035 }
1036 
1037 static int
1038 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1039 {
1040 	struct ifbrparam *param = arg;
1041 
1042 	if (param->ifbrp_hellotime == 0)
1043 		return (EINVAL);
1044 	sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8;
1045 
1046 	if (sc->sc_if.if_flags & IFF_RUNNING)
1047 		bstp_initialization(sc);
1048 
1049 	return (0);
1050 }
1051 
1052 static int
1053 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1054 {
1055 	struct ifbrparam *param = arg;
1056 
1057 	param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8;
1058 
1059 	return (0);
1060 }
1061 
1062 static int
1063 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1064 {
1065 	struct ifbrparam *param = arg;
1066 
1067 	if (param->ifbrp_fwddelay == 0)
1068 		return (EINVAL);
1069 	sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8;
1070 
1071 	if (sc->sc_if.if_flags & IFF_RUNNING)
1072 		bstp_initialization(sc);
1073 
1074 	return (0);
1075 }
1076 
1077 static int
1078 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1079 {
1080 	struct ifbrparam *param = arg;
1081 
1082 	param->ifbrp_maxage = sc->sc_bridge_max_age >> 8;
1083 
1084 	return (0);
1085 }
1086 
1087 static int
1088 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1089 {
1090 	struct ifbrparam *param = arg;
1091 
1092 	if (param->ifbrp_maxage == 0)
1093 		return (EINVAL);
1094 	sc->sc_bridge_max_age = param->ifbrp_maxage << 8;
1095 
1096 	if (sc->sc_if.if_flags & IFF_RUNNING)
1097 		bstp_initialization(sc);
1098 
1099 	return (0);
1100 }
1101 
1102 static int
1103 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1104 {
1105 	struct ifbreq *req = arg;
1106 	struct bridge_iflist *bif;
1107 
1108 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1109 	if (bif == NULL)
1110 		return (ENOENT);
1111 
1112 	bif->bif_priority = req->ifbr_priority;
1113 
1114 	if (sc->sc_if.if_flags & IFF_RUNNING)
1115 		bstp_initialization(sc);
1116 
1117 	return (0);
1118 }
1119 
1120 #if defined(BRIDGE_IPF)
1121 static int
1122 bridge_ioctl_gfilt(struct bridge_softc *sc, void *arg)
1123 {
1124 	struct ifbrparam *param = arg;
1125 
1126 	param->ifbrp_filter = sc->sc_filter_flags;
1127 
1128 	return (0);
1129 }
1130 
1131 static int
1132 bridge_ioctl_sfilt(struct bridge_softc *sc, void *arg)
1133 {
1134 	struct ifbrparam *param = arg;
1135 	uint32_t nflags, oflags;
1136 
1137 	if (param->ifbrp_filter & ~IFBF_FILT_MASK)
1138 		return (EINVAL);
1139 
1140 	nflags = param->ifbrp_filter;
1141 	oflags = sc->sc_filter_flags;
1142 
1143 	if ((nflags & IFBF_FILT_USEIPF) && !(oflags & IFBF_FILT_USEIPF)) {
1144 		pfil_add_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
1145 			sc->sc_if.if_pfil);
1146 	}
1147 	if (!(nflags & IFBF_FILT_USEIPF) && (oflags & IFBF_FILT_USEIPF)) {
1148 		pfil_remove_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
1149 			sc->sc_if.if_pfil);
1150 	}
1151 
1152 	sc->sc_filter_flags = nflags;
1153 
1154 	return (0);
1155 }
1156 #endif /* BRIDGE_IPF */
1157 
1158 static int
1159 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1160 {
1161 	struct ifbreq *req = arg;
1162 	struct bridge_iflist *bif;
1163 
1164 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1165 	if (bif == NULL)
1166 		return (ENOENT);
1167 
1168 	bif->bif_path_cost = req->ifbr_path_cost;
1169 
1170 	if (sc->sc_if.if_flags & IFF_RUNNING)
1171 		bstp_initialization(sc);
1172 
1173 	return (0);
1174 }
1175 
1176 /*
1177  * bridge_ifdetach:
1178  *
1179  *	Detach an interface from a bridge.  Called when a member
1180  *	interface is detaching.
1181  */
1182 void
1183 bridge_ifdetach(struct ifnet *ifp)
1184 {
1185 	struct bridge_softc *sc = ifp->if_bridge;
1186 	struct ifbreq breq;
1187 
1188 	memset(&breq, 0, sizeof(breq));
1189 	strlcpy(breq.ifbr_ifsname, ifp->if_xname, sizeof(breq.ifbr_ifsname));
1190 
1191 	(void) bridge_ioctl_del(sc, &breq);
1192 }
1193 
1194 /*
1195  * bridge_init:
1196  *
1197  *	Initialize a bridge interface.
1198  */
1199 static int
1200 bridge_init(struct ifnet *ifp)
1201 {
1202 	struct bridge_softc *sc = ifp->if_softc;
1203 
1204 	if (ifp->if_flags & IFF_RUNNING)
1205 		return (0);
1206 
1207 	callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1208 	    bridge_timer, sc);
1209 
1210 	ifp->if_flags |= IFF_RUNNING;
1211 	bstp_initialization(sc);
1212 	return (0);
1213 }
1214 
1215 /*
1216  * bridge_stop:
1217  *
1218  *	Stop the bridge interface.
1219  */
1220 static void
1221 bridge_stop(struct ifnet *ifp, int disable)
1222 {
1223 	struct bridge_softc *sc = ifp->if_softc;
1224 
1225 	if ((ifp->if_flags & IFF_RUNNING) == 0)
1226 		return;
1227 
1228 	callout_stop(&sc->sc_brcallout);
1229 	bstp_stop(sc);
1230 
1231 	bridge_rtflush(sc, IFBF_FLUSHDYN);
1232 
1233 	ifp->if_flags &= ~IFF_RUNNING;
1234 }
1235 
1236 /*
1237  * bridge_enqueue:
1238  *
1239  *	Enqueue a packet on a bridge member interface.
1240  *
1241  *	NOTE: must be called at splnet().
1242  */
1243 void
1244 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m,
1245     int runfilt)
1246 {
1247 	ALTQ_DECL(struct altq_pktattr pktattr;)
1248 	int len, error;
1249 	short mflags;
1250 
1251 	/*
1252 	 * Clear any in-bound checksum flags for this packet.
1253 	 */
1254 	m->m_pkthdr.csum_flags = 0;
1255 
1256 	if (runfilt) {
1257 		if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
1258 		    dst_ifp, PFIL_OUT) != 0) {
1259 			if (m != NULL)
1260 				m_freem(m);
1261 			return;
1262 		}
1263 		if (m == NULL)
1264 			return;
1265 	}
1266 
1267 #ifdef ALTQ
1268 	/*
1269 	 * If ALTQ is enabled on the member interface, do
1270 	 * classification; the queueing discipline might
1271 	 * not require classification, but might require
1272 	 * the address family/header pointer in the pktattr.
1273 	 */
1274 	if (ALTQ_IS_ENABLED(&dst_ifp->if_snd)) {
1275 		/* XXX IFT_ETHER */
1276 		altq_etherclassify(&dst_ifp->if_snd, m, &pktattr);
1277 	}
1278 #endif /* ALTQ */
1279 
1280 	len = m->m_pkthdr.len;
1281 	m->m_flags |= M_PROTO1;
1282 	mflags = m->m_flags;
1283 	IFQ_ENQUEUE(&dst_ifp->if_snd, m, &pktattr, error);
1284 	if (error) {
1285 		/* mbuf is already freed */
1286 		sc->sc_if.if_oerrors++;
1287 		return;
1288 	}
1289 
1290 	sc->sc_if.if_opackets++;
1291 	sc->sc_if.if_obytes += len;
1292 
1293 	dst_ifp->if_obytes += len;
1294 
1295 	if (mflags & M_MCAST) {
1296 		sc->sc_if.if_omcasts++;
1297 		dst_ifp->if_omcasts++;
1298 	}
1299 
1300 	if ((dst_ifp->if_flags & IFF_OACTIVE) == 0)
1301 		(*dst_ifp->if_start)(dst_ifp);
1302 }
1303 
1304 /*
1305  * bridge_output:
1306  *
1307  *	Send output from a bridge member interface.  This
1308  *	performs the bridging function for locally originated
1309  *	packets.
1310  *
1311  *	The mbuf has the Ethernet header already attached.  We must
1312  *	enqueue or free the mbuf before returning.
1313  */
1314 int
1315 bridge_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa,
1316     struct rtentry *rt)
1317 {
1318 	struct ether_header *eh;
1319 	struct ifnet *dst_if;
1320 	struct bridge_softc *sc;
1321 	int s;
1322 
1323 	if (m->m_len < ETHER_HDR_LEN) {
1324 		m = m_pullup(m, ETHER_HDR_LEN);
1325 		if (m == NULL)
1326 			return (0);
1327 	}
1328 
1329 	eh = mtod(m, struct ether_header *);
1330 	sc = ifp->if_bridge;
1331 
1332 	s = splnet();
1333 
1334 	/*
1335 	 * If bridge is down, but the original output interface is up,
1336 	 * go ahead and send out that interface.  Otherwise, the packet
1337 	 * is dropped below.
1338 	 */
1339 	if ((sc->sc_if.if_flags & IFF_RUNNING) == 0) {
1340 		dst_if = ifp;
1341 		goto sendunicast;
1342 	}
1343 
1344 	/*
1345 	 * If the packet is a multicast, or we don't know a better way to
1346 	 * get there, send to all interfaces.
1347 	 */
1348 	if (ETHER_IS_MULTICAST(eh->ether_dhost))
1349 		dst_if = NULL;
1350 	else
1351 		dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1352 	if (dst_if == NULL) {
1353 		struct bridge_iflist *bif;
1354 		struct mbuf *mc;
1355 		int used = 0;
1356 
1357 		LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1358 			dst_if = bif->bif_ifp;
1359 			if ((dst_if->if_flags & IFF_RUNNING) == 0)
1360 				continue;
1361 
1362 			/*
1363 			 * If this is not the original output interface,
1364 			 * and the interface is participating in spanning
1365 			 * tree, make sure the port is in a state that
1366 			 * allows forwarding.
1367 			 */
1368 			if (dst_if != ifp &&
1369 			    (bif->bif_flags & IFBIF_STP) != 0) {
1370 				switch (bif->bif_state) {
1371 				case BSTP_IFSTATE_BLOCKING:
1372 				case BSTP_IFSTATE_LISTENING:
1373 				case BSTP_IFSTATE_DISABLED:
1374 					continue;
1375 				}
1376 			}
1377 
1378 			if (LIST_NEXT(bif, bif_next) == NULL) {
1379 				used = 1;
1380 				mc = m;
1381 			} else {
1382 				mc = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1383 				if (mc == NULL) {
1384 					sc->sc_if.if_oerrors++;
1385 					continue;
1386 				}
1387 			}
1388 
1389 			bridge_enqueue(sc, dst_if, mc, 0);
1390 		}
1391 		if (used == 0)
1392 			m_freem(m);
1393 		splx(s);
1394 		return (0);
1395 	}
1396 
1397  sendunicast:
1398 	/*
1399 	 * XXX Spanning tree consideration here?
1400 	 */
1401 
1402 	if ((dst_if->if_flags & IFF_RUNNING) == 0) {
1403 		m_freem(m);
1404 		splx(s);
1405 		return (0);
1406 	}
1407 
1408 	bridge_enqueue(sc, dst_if, m, 0);
1409 
1410 	splx(s);
1411 	return (0);
1412 }
1413 
1414 /*
1415  * bridge_start:
1416  *
1417  *	Start output on a bridge.
1418  *
1419  *	NOTE: This routine should never be called in this implementation.
1420  */
1421 static void
1422 bridge_start(struct ifnet *ifp)
1423 {
1424 
1425 	printf("%s: bridge_start() called\n", ifp->if_xname);
1426 }
1427 
1428 /*
1429  * bridge_forward:
1430  *
1431  *	The forwarding function of the bridge.
1432  */
1433 static void
1434 bridge_forward(void *v)
1435 {
1436 	struct bridge_softc *sc = v;
1437 	struct mbuf *m;
1438 	struct bridge_iflist *bif;
1439 	struct ifnet *src_if, *dst_if;
1440 	struct ether_header *eh;
1441 	int s;
1442 
1443 	KERNEL_LOCK(1, NULL);
1444 	mutex_enter(softnet_lock);
1445 	if ((sc->sc_if.if_flags & IFF_RUNNING) == 0) {
1446 		mutex_exit(softnet_lock);
1447 		KERNEL_UNLOCK_ONE(NULL);
1448 		return;
1449 	}
1450 
1451 	s = splnet();
1452 	while ((m = pktq_dequeue(sc->sc_fwd_pktq)) != NULL) {
1453 		src_if = m->m_pkthdr.rcvif;
1454 
1455 		sc->sc_if.if_ipackets++;
1456 		sc->sc_if.if_ibytes += m->m_pkthdr.len;
1457 
1458 		/*
1459 		 * Look up the bridge_iflist.
1460 		 */
1461 		bif = bridge_lookup_member_if(sc, src_if);
1462 		if (bif == NULL) {
1463 			/* Interface is not a bridge member (anymore?) */
1464 			m_freem(m);
1465 			continue;
1466 		}
1467 
1468 		if (bif->bif_flags & IFBIF_STP) {
1469 			switch (bif->bif_state) {
1470 			case BSTP_IFSTATE_BLOCKING:
1471 			case BSTP_IFSTATE_LISTENING:
1472 			case BSTP_IFSTATE_DISABLED:
1473 				m_freem(m);
1474 				continue;
1475 			}
1476 		}
1477 
1478 		eh = mtod(m, struct ether_header *);
1479 
1480 		/*
1481 		 * If the interface is learning, and the source
1482 		 * address is valid and not multicast, record
1483 		 * the address.
1484 		 */
1485 		if ((bif->bif_flags & IFBIF_LEARNING) != 0 &&
1486 		    ETHER_IS_MULTICAST(eh->ether_shost) == 0 &&
1487 		    (eh->ether_shost[0] == 0 &&
1488 		     eh->ether_shost[1] == 0 &&
1489 		     eh->ether_shost[2] == 0 &&
1490 		     eh->ether_shost[3] == 0 &&
1491 		     eh->ether_shost[4] == 0 &&
1492 		     eh->ether_shost[5] == 0) == 0) {
1493 			(void) bridge_rtupdate(sc, eh->ether_shost,
1494 			    src_if, 0, IFBAF_DYNAMIC);
1495 		}
1496 
1497 		if ((bif->bif_flags & IFBIF_STP) != 0 &&
1498 		    bif->bif_state == BSTP_IFSTATE_LEARNING) {
1499 			m_freem(m);
1500 			continue;
1501 		}
1502 
1503 		/*
1504 		 * At this point, the port either doesn't participate
1505 		 * in spanning tree or it is in the forwarding state.
1506 		 */
1507 
1508 		/*
1509 		 * If the packet is unicast, destined for someone on
1510 		 * "this" side of the bridge, drop it.
1511 		 */
1512 		if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
1513 			dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1514 			if (src_if == dst_if) {
1515 				m_freem(m);
1516 				continue;
1517 			}
1518 		} else {
1519 			/* ...forward it to all interfaces. */
1520 			sc->sc_if.if_imcasts++;
1521 			dst_if = NULL;
1522 		}
1523 
1524 		if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
1525 		    m->m_pkthdr.rcvif, PFIL_IN) != 0) {
1526 			if (m != NULL)
1527 				m_freem(m);
1528 			continue;
1529 		}
1530 		if (m == NULL)
1531 			continue;
1532 
1533 		if (dst_if == NULL) {
1534 			bridge_broadcast(sc, src_if, m);
1535 			continue;
1536 		}
1537 
1538 		/*
1539 		 * At this point, we're dealing with a unicast frame
1540 		 * going to a different interface.
1541 		 */
1542 		if ((dst_if->if_flags & IFF_RUNNING) == 0) {
1543 			m_freem(m);
1544 			continue;
1545 		}
1546 		bif = bridge_lookup_member_if(sc, dst_if);
1547 		if (bif == NULL) {
1548 			/* Not a member of the bridge (anymore?) */
1549 			m_freem(m);
1550 			continue;
1551 		}
1552 
1553 		if (bif->bif_flags & IFBIF_STP) {
1554 			switch (bif->bif_state) {
1555 			case BSTP_IFSTATE_DISABLED:
1556 			case BSTP_IFSTATE_BLOCKING:
1557 				m_freem(m);
1558 				continue;
1559 			}
1560 		}
1561 
1562 		bridge_enqueue(sc, dst_if, m, 1);
1563 	}
1564 	splx(s);
1565 	mutex_exit(softnet_lock);
1566 	KERNEL_UNLOCK_ONE(NULL);
1567 }
1568 
1569 static bool
1570 bstp_state_before_learning(struct bridge_iflist *bif)
1571 {
1572 	if (bif->bif_flags & IFBIF_STP) {
1573 		switch (bif->bif_state) {
1574 		case BSTP_IFSTATE_BLOCKING:
1575 		case BSTP_IFSTATE_LISTENING:
1576 		case BSTP_IFSTATE_DISABLED:
1577 			return true;
1578 		}
1579 	}
1580 	return false;
1581 }
1582 
1583 static bool
1584 bridge_ourether(struct bridge_iflist *bif, struct ether_header *eh, int src)
1585 {
1586 	uint8_t *ether = src ? eh->ether_shost : eh->ether_dhost;
1587 
1588 	if (memcmp(CLLADDR(bif->bif_ifp->if_sadl), ether, ETHER_ADDR_LEN) == 0
1589 #if NCARP > 0
1590 	    || (bif->bif_ifp->if_carp &&
1591 	        carp_ourether(bif->bif_ifp->if_carp, eh, IFT_ETHER, src) != NULL)
1592 #endif /* NCARP > 0 */
1593 	    )
1594 		return true;
1595 
1596 	return false;
1597 }
1598 
1599 /*
1600  * bridge_input:
1601  *
1602  *	Receive input from a member interface.  Queue the packet for
1603  *	bridging if it is not for us.
1604  *	should be called at splnet()
1605  */
1606 static void
1607 bridge_input(struct ifnet *ifp, struct mbuf *m)
1608 {
1609 	struct bridge_softc *sc = ifp->if_bridge;
1610 	struct bridge_iflist *bif;
1611 	struct ether_header *eh;
1612 
1613 	if ((sc->sc_if.if_flags & IFF_RUNNING) == 0) {
1614 		ether_input(ifp, m);
1615 		return;
1616 	}
1617 
1618 	bif = bridge_lookup_member_if(sc, ifp);
1619 	if (bif == NULL) {
1620 		ether_input(ifp, m);
1621 		return;
1622 	}
1623 
1624 	eh = mtod(m, struct ether_header *);
1625 
1626 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1627 		if (memcmp(etherbroadcastaddr,
1628 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
1629 			m->m_flags |= M_BCAST;
1630 		else
1631 			m->m_flags |= M_MCAST;
1632 	}
1633 
1634 	/*
1635 	 * A 'fast' path for packets addressed to interfaces that are
1636 	 * part of this bridge.
1637 	 */
1638 	if (!(m->m_flags & (M_BCAST|M_MCAST)) &&
1639 	    !bstp_state_before_learning(bif)) {
1640 		struct bridge_iflist *_bif;
1641 
1642 		LIST_FOREACH(_bif, &sc->sc_iflist, bif_next) {
1643 			/* It is destined for us. */
1644 			if (bridge_ourether(_bif, eh, 0)) {
1645 				if (_bif->bif_flags & IFBIF_LEARNING)
1646 					(void) bridge_rtupdate(sc,
1647 					    eh->ether_shost, ifp, 0, IFBAF_DYNAMIC);
1648 				m->m_pkthdr.rcvif = _bif->bif_ifp;
1649 				ether_input(_bif->bif_ifp, m);
1650 				return;
1651 			}
1652 
1653 			/* We just received a packet that we sent out. */
1654 			if (bridge_ourether(_bif, eh, 1)) {
1655 				m_freem(m);
1656 				return;
1657 			}
1658 		}
1659 	}
1660 
1661 	/* Tap off 802.1D packets; they do not get forwarded. */
1662 	if (bif->bif_flags & IFBIF_STP &&
1663 	    memcmp(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN) == 0) {
1664 		bstp_input(sc, bif, m);
1665 		return;
1666 	}
1667 
1668 	/*
1669 	 * A normal switch would discard the packet here, but that's not what
1670 	 * we've done historically. This also prevents some obnoxious behaviour.
1671 	 */
1672 	if (bstp_state_before_learning(bif)) {
1673 		ether_input(ifp, m);
1674 		return;
1675 	}
1676 
1677 	/* Queue the packet for bridge forwarding. */
1678 	if (__predict_false(!pktq_enqueue(sc->sc_fwd_pktq, m, 0)))
1679 		m_freem(m);
1680 }
1681 
1682 /*
1683  * bridge_broadcast:
1684  *
1685  *	Send a frame to all interfaces that are members of
1686  *	the bridge, except for the one on which the packet
1687  *	arrived.
1688  */
1689 static void
1690 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
1691     struct mbuf *m)
1692 {
1693 	struct bridge_iflist *bif;
1694 	struct mbuf *mc;
1695 	struct ifnet *dst_if;
1696 	bool used, bmcast;
1697 
1698 	used = bmcast = m->m_flags & (M_BCAST|M_MCAST);
1699 
1700 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1701 		dst_if = bif->bif_ifp;
1702 		if (dst_if == src_if)
1703 			continue;
1704 
1705 		if (bif->bif_flags & IFBIF_STP) {
1706 			switch (bif->bif_state) {
1707 			case BSTP_IFSTATE_BLOCKING:
1708 			case BSTP_IFSTATE_DISABLED:
1709 				continue;
1710 			}
1711 		}
1712 
1713 		if ((bif->bif_flags & IFBIF_DISCOVER) == 0 && !bmcast)
1714 			continue;
1715 
1716 		if ((dst_if->if_flags & IFF_RUNNING) == 0)
1717 			continue;
1718 
1719 		if (!used && LIST_NEXT(bif, bif_next) == NULL) {
1720 			mc = m;
1721 			used = true;
1722 		} else {
1723 			mc = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
1724 			if (mc == NULL) {
1725 				sc->sc_if.if_oerrors++;
1726 				continue;
1727 			}
1728 		}
1729 
1730 		bridge_enqueue(sc, dst_if, mc, 1);
1731 	}
1732 
1733 	if (bmcast)
1734 		ether_input(src_if, m);
1735 	else if (!used)
1736 		m_freem(m);
1737 }
1738 
1739 /*
1740  * bridge_rtupdate:
1741  *
1742  *	Add a bridge routing entry.
1743  */
1744 static int
1745 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
1746     struct ifnet *dst_if, int setflags, uint8_t flags)
1747 {
1748 	struct bridge_rtnode *brt;
1749 	int error;
1750 
1751 	/*
1752 	 * A route for this destination might already exist.  If so,
1753 	 * update it, otherwise create a new one.
1754 	 */
1755 	if ((brt = bridge_rtnode_lookup(sc, dst)) == NULL) {
1756 		if (sc->sc_brtcnt >= sc->sc_brtmax)
1757 			return (ENOSPC);
1758 
1759 		/*
1760 		 * Allocate a new bridge forwarding node, and
1761 		 * initialize the expiration time and Ethernet
1762 		 * address.
1763 		 */
1764 		brt = pool_get(&bridge_rtnode_pool, PR_NOWAIT);
1765 		if (brt == NULL)
1766 			return (ENOMEM);
1767 
1768 		memset(brt, 0, sizeof(*brt));
1769 		brt->brt_expire = time_uptime + sc->sc_brttimeout;
1770 		brt->brt_flags = IFBAF_DYNAMIC;
1771 		memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
1772 
1773 		if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
1774 			pool_put(&bridge_rtnode_pool, brt);
1775 			return (error);
1776 		}
1777 	}
1778 
1779 	brt->brt_ifp = dst_if;
1780 	if (setflags) {
1781 		brt->brt_flags = flags;
1782 		if (flags & IFBAF_STATIC)
1783 			brt->brt_expire = 0;
1784 		else
1785 			brt->brt_expire = time_uptime + sc->sc_brttimeout;
1786 	}
1787 
1788 	return (0);
1789 }
1790 
1791 /*
1792  * bridge_rtlookup:
1793  *
1794  *	Lookup the destination interface for an address.
1795  */
1796 static struct ifnet *
1797 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr)
1798 {
1799 	struct bridge_rtnode *brt;
1800 
1801 	if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL)
1802 		return (NULL);
1803 
1804 	return (brt->brt_ifp);
1805 }
1806 
1807 /*
1808  * bridge_rttrim:
1809  *
1810  *	Trim the routine table so that we have a number
1811  *	of routing entries less than or equal to the
1812  *	maximum number.
1813  */
1814 static void
1815 bridge_rttrim(struct bridge_softc *sc)
1816 {
1817 	struct bridge_rtnode *brt, *nbrt;
1818 
1819 	/* Make sure we actually need to do this. */
1820 	if (sc->sc_brtcnt <= sc->sc_brtmax)
1821 		return;
1822 
1823 	/* Force an aging cycle; this might trim enough addresses. */
1824 	bridge_rtage(sc);
1825 	if (sc->sc_brtcnt <= sc->sc_brtmax)
1826 		return;
1827 
1828 	for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
1829 		nbrt = LIST_NEXT(brt, brt_list);
1830 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
1831 			bridge_rtnode_destroy(sc, brt);
1832 			if (sc->sc_brtcnt <= sc->sc_brtmax)
1833 				return;
1834 		}
1835 	}
1836 }
1837 
1838 /*
1839  * bridge_timer:
1840  *
1841  *	Aging timer for the bridge.
1842  */
1843 static void
1844 bridge_timer(void *arg)
1845 {
1846 	struct bridge_softc *sc = arg;
1847 	int s;
1848 
1849 	s = splnet();
1850 	bridge_rtage(sc);
1851 	splx(s);
1852 
1853 	if (sc->sc_if.if_flags & IFF_RUNNING)
1854 		callout_reset(&sc->sc_brcallout,
1855 		    bridge_rtable_prune_period * hz, bridge_timer, sc);
1856 }
1857 
1858 /*
1859  * bridge_rtage:
1860  *
1861  *	Perform an aging cycle.
1862  */
1863 static void
1864 bridge_rtage(struct bridge_softc *sc)
1865 {
1866 	struct bridge_rtnode *brt, *nbrt;
1867 
1868 	for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
1869 		nbrt = LIST_NEXT(brt, brt_list);
1870 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
1871 			if (time_uptime >= brt->brt_expire)
1872 				bridge_rtnode_destroy(sc, brt);
1873 		}
1874 	}
1875 }
1876 
1877 /*
1878  * bridge_rtflush:
1879  *
1880  *	Remove all dynamic addresses from the bridge.
1881  */
1882 static void
1883 bridge_rtflush(struct bridge_softc *sc, int full)
1884 {
1885 	struct bridge_rtnode *brt, *nbrt;
1886 
1887 	for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
1888 		nbrt = LIST_NEXT(brt, brt_list);
1889 		if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
1890 			bridge_rtnode_destroy(sc, brt);
1891 	}
1892 }
1893 
1894 /*
1895  * bridge_rtdaddr:
1896  *
1897  *	Remove an address from the table.
1898  */
1899 static int
1900 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr)
1901 {
1902 	struct bridge_rtnode *brt;
1903 
1904 	if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL)
1905 		return (ENOENT);
1906 
1907 	bridge_rtnode_destroy(sc, brt);
1908 	return (0);
1909 }
1910 
1911 /*
1912  * bridge_rtdelete:
1913  *
1914  *	Delete routes to a speicifc member interface.
1915  */
1916 static void
1917 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp)
1918 {
1919 	struct bridge_rtnode *brt, *nbrt;
1920 
1921 	for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
1922 		nbrt = LIST_NEXT(brt, brt_list);
1923 		if (brt->brt_ifp == ifp)
1924 			bridge_rtnode_destroy(sc, brt);
1925 	}
1926 }
1927 
1928 /*
1929  * bridge_rtable_init:
1930  *
1931  *	Initialize the route table for this bridge.
1932  */
1933 static int
1934 bridge_rtable_init(struct bridge_softc *sc)
1935 {
1936 	int i;
1937 
1938 	sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
1939 	    M_DEVBUF, M_NOWAIT);
1940 	if (sc->sc_rthash == NULL)
1941 		return (ENOMEM);
1942 
1943 	for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
1944 		LIST_INIT(&sc->sc_rthash[i]);
1945 
1946 	sc->sc_rthash_key = cprng_fast32();
1947 
1948 	LIST_INIT(&sc->sc_rtlist);
1949 
1950 	return (0);
1951 }
1952 
1953 /*
1954  * bridge_rtable_fini:
1955  *
1956  *	Deconstruct the route table for this bridge.
1957  */
1958 static void
1959 bridge_rtable_fini(struct bridge_softc *sc)
1960 {
1961 
1962 	free(sc->sc_rthash, M_DEVBUF);
1963 }
1964 
1965 /*
1966  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
1967  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
1968  */
1969 #define	mix(a, b, c)							\
1970 do {									\
1971 	a -= b; a -= c; a ^= (c >> 13);					\
1972 	b -= c; b -= a; b ^= (a << 8);					\
1973 	c -= a; c -= b; c ^= (b >> 13);					\
1974 	a -= b; a -= c; a ^= (c >> 12);					\
1975 	b -= c; b -= a; b ^= (a << 16);					\
1976 	c -= a; c -= b; c ^= (b >> 5);					\
1977 	a -= b; a -= c; a ^= (c >> 3);					\
1978 	b -= c; b -= a; b ^= (a << 10);					\
1979 	c -= a; c -= b; c ^= (b >> 15);					\
1980 } while (/*CONSTCOND*/0)
1981 
1982 static inline uint32_t
1983 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
1984 {
1985 	uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
1986 
1987 	b += addr[5] << 8;
1988 	b += addr[4];
1989 	a += addr[3] << 24;
1990 	a += addr[2] << 16;
1991 	a += addr[1] << 8;
1992 	a += addr[0];
1993 
1994 	mix(a, b, c);
1995 
1996 	return (c & BRIDGE_RTHASH_MASK);
1997 }
1998 
1999 #undef mix
2000 
2001 /*
2002  * bridge_rtnode_lookup:
2003  *
2004  *	Look up a bridge route node for the specified destination.
2005  */
2006 static struct bridge_rtnode *
2007 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr)
2008 {
2009 	struct bridge_rtnode *brt;
2010 	uint32_t hash;
2011 	int dir;
2012 
2013 	hash = bridge_rthash(sc, addr);
2014 	LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
2015 		dir = memcmp(addr, brt->brt_addr, ETHER_ADDR_LEN);
2016 		if (dir == 0)
2017 			return (brt);
2018 		if (dir > 0)
2019 			return (NULL);
2020 	}
2021 
2022 	return (NULL);
2023 }
2024 
2025 /*
2026  * bridge_rtnode_insert:
2027  *
2028  *	Insert the specified bridge node into the route table.  We
2029  *	assume the entry is not already in the table.
2030  */
2031 static int
2032 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2033 {
2034 	struct bridge_rtnode *lbrt;
2035 	uint32_t hash;
2036 	int dir;
2037 
2038 	hash = bridge_rthash(sc, brt->brt_addr);
2039 
2040 	lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
2041 	if (lbrt == NULL) {
2042 		LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
2043 		goto out;
2044 	}
2045 
2046 	do {
2047 		dir = memcmp(brt->brt_addr, lbrt->brt_addr, ETHER_ADDR_LEN);
2048 		if (dir == 0)
2049 			return (EEXIST);
2050 		if (dir > 0) {
2051 			LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
2052 			goto out;
2053 		}
2054 		if (LIST_NEXT(lbrt, brt_hash) == NULL) {
2055 			LIST_INSERT_AFTER(lbrt, brt, brt_hash);
2056 			goto out;
2057 		}
2058 		lbrt = LIST_NEXT(lbrt, brt_hash);
2059 	} while (lbrt != NULL);
2060 
2061 #ifdef DIAGNOSTIC
2062 	panic("bridge_rtnode_insert: impossible");
2063 #endif
2064 
2065  out:
2066 	LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
2067 	sc->sc_brtcnt++;
2068 
2069 	return (0);
2070 }
2071 
2072 /*
2073  * bridge_rtnode_destroy:
2074  *
2075  *	Destroy a bridge rtnode.
2076  */
2077 static void
2078 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
2079 {
2080 	int s = splnet();
2081 
2082 	LIST_REMOVE(brt, brt_hash);
2083 
2084 	LIST_REMOVE(brt, brt_list);
2085 	sc->sc_brtcnt--;
2086 	pool_put(&bridge_rtnode_pool, brt);
2087 
2088 	splx(s);
2089 }
2090 
2091 #if defined(BRIDGE_IPF)
2092 extern pfil_head_t *inet_pfil_hook;                 /* XXX */
2093 extern pfil_head_t *inet6_pfil_hook;                /* XXX */
2094 
2095 /*
2096  * Send bridge packets through IPF if they are one of the types IPF can deal
2097  * with, or if they are ARP or REVARP.  (IPF will pass ARP and REVARP without
2098  * question.)
2099  */
2100 static int
2101 bridge_ipf(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
2102 {
2103 	int snap, error;
2104 	struct ether_header *eh1, eh2;
2105 	struct llc llc1;
2106 	uint16_t ether_type;
2107 
2108 	snap = 0;
2109 	error = -1;	/* Default error if not error == 0 */
2110 	eh1 = mtod(*mp, struct ether_header *);
2111 	ether_type = ntohs(eh1->ether_type);
2112 
2113 	/*
2114 	 * Check for SNAP/LLC.
2115 	 */
2116         if (ether_type < ETHERMTU) {
2117                 struct llc *llc2 = (struct llc *)(eh1 + 1);
2118 
2119                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
2120                     llc2->llc_dsap == LLC_SNAP_LSAP &&
2121                     llc2->llc_ssap == LLC_SNAP_LSAP &&
2122                     llc2->llc_control == LLC_UI) {
2123                 	ether_type = htons(llc2->llc_un.type_snap.ether_type);
2124 			snap = 1;
2125                 }
2126         }
2127 
2128 	/*
2129 	 * If we're trying to filter bridge traffic, don't look at anything
2130 	 * other than IP and ARP traffic.  If the filter doesn't understand
2131 	 * IPv6, don't allow IPv6 through the bridge either.  This is lame
2132 	 * since if we really wanted, say, an AppleTalk filter, we are hosed,
2133 	 * but of course we don't have an AppleTalk filter to begin with.
2134 	 * (Note that since IPF doesn't understand ARP it will pass *ALL*
2135 	 * ARP traffic.)
2136 	 */
2137 	switch (ether_type) {
2138 		case ETHERTYPE_ARP:
2139 		case ETHERTYPE_REVARP:
2140 			return 0; /* Automatically pass */
2141 		case ETHERTYPE_IP:
2142 # ifdef INET6
2143 		case ETHERTYPE_IPV6:
2144 # endif /* INET6 */
2145 			break;
2146 		default:
2147 			goto bad;
2148 	}
2149 
2150 	/* Strip off the Ethernet header and keep a copy. */
2151 	m_copydata(*mp, 0, ETHER_HDR_LEN, (void *) &eh2);
2152 	m_adj(*mp, ETHER_HDR_LEN);
2153 
2154 	/* Strip off snap header, if present */
2155 	if (snap) {
2156 		m_copydata(*mp, 0, sizeof(struct llc), (void *) &llc1);
2157 		m_adj(*mp, sizeof(struct llc));
2158 	}
2159 
2160 	/*
2161 	 * Check basic packet sanity and run IPF through pfil.
2162 	 */
2163 	KASSERT(!cpu_intr_p());
2164 	switch (ether_type)
2165 	{
2166 	case ETHERTYPE_IP :
2167 		error = (dir == PFIL_IN) ? bridge_ip_checkbasic(mp) : 0;
2168 		if (error == 0)
2169 			error = pfil_run_hooks(inet_pfil_hook, mp, ifp, dir);
2170 		break;
2171 # ifdef INET6
2172 	case ETHERTYPE_IPV6 :
2173 		error = (dir == PFIL_IN) ? bridge_ip6_checkbasic(mp) : 0;
2174 		if (error == 0)
2175 			error = pfil_run_hooks(inet6_pfil_hook, mp, ifp, dir);
2176 		break;
2177 # endif
2178 	default :
2179 		error = 0;
2180 		break;
2181 	}
2182 
2183 	if (*mp == NULL)
2184 		return error;
2185 	if (error != 0)
2186 		goto bad;
2187 
2188 	error = -1;
2189 
2190 	/*
2191 	 * Finally, put everything back the way it was and return
2192 	 */
2193 	if (snap) {
2194 		M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
2195 		if (*mp == NULL)
2196 			return error;
2197 		bcopy(&llc1, mtod(*mp, void *), sizeof(struct llc));
2198 	}
2199 
2200 	M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
2201 	if (*mp == NULL)
2202 		return error;
2203 	bcopy(&eh2, mtod(*mp, void *), ETHER_HDR_LEN);
2204 
2205 	return 0;
2206 
2207     bad:
2208 	m_freem(*mp);
2209 	*mp = NULL;
2210 	return error;
2211 }
2212 
2213 /*
2214  * Perform basic checks on header size since
2215  * IPF assumes ip_input has already processed
2216  * it for it.  Cut-and-pasted from ip_input.c.
2217  * Given how simple the IPv6 version is,
2218  * does the IPv4 version really need to be
2219  * this complicated?
2220  *
2221  * XXX Should we update ipstat here, or not?
2222  * XXX Right now we update ipstat but not
2223  * XXX csum_counter.
2224  */
2225 static int
2226 bridge_ip_checkbasic(struct mbuf **mp)
2227 {
2228 	struct mbuf *m = *mp;
2229 	struct ip *ip;
2230 	int len, hlen;
2231 
2232 	if (*mp == NULL)
2233 		return -1;
2234 
2235 	if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
2236 		if ((m = m_copyup(m, sizeof(struct ip),
2237 			(max_linkhdr + 3) & ~3)) == NULL) {
2238 			/* XXXJRT new stat, please */
2239 			ip_statinc(IP_STAT_TOOSMALL);
2240 			goto bad;
2241 		}
2242 	} else if (__predict_false(m->m_len < sizeof (struct ip))) {
2243 		if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
2244 			ip_statinc(IP_STAT_TOOSMALL);
2245 			goto bad;
2246 		}
2247 	}
2248 	ip = mtod(m, struct ip *);
2249 	if (ip == NULL) goto bad;
2250 
2251 	if (ip->ip_v != IPVERSION) {
2252 		ip_statinc(IP_STAT_BADVERS);
2253 		goto bad;
2254 	}
2255 	hlen = ip->ip_hl << 2;
2256 	if (hlen < sizeof(struct ip)) { /* minimum header length */
2257 		ip_statinc(IP_STAT_BADHLEN);
2258 		goto bad;
2259 	}
2260 	if (hlen > m->m_len) {
2261 		if ((m = m_pullup(m, hlen)) == 0) {
2262 			ip_statinc(IP_STAT_BADHLEN);
2263 			goto bad;
2264 		}
2265 		ip = mtod(m, struct ip *);
2266 		if (ip == NULL) goto bad;
2267 	}
2268 
2269         switch (m->m_pkthdr.csum_flags &
2270                 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) |
2271                  M_CSUM_IPv4_BAD)) {
2272         case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
2273                 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); */
2274                 goto bad;
2275 
2276         case M_CSUM_IPv4:
2277                 /* Checksum was okay. */
2278                 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); */
2279                 break;
2280 
2281         default:
2282                 /* Must compute it ourselves. */
2283                 /* INET_CSUM_COUNTER_INCR(&ip_swcsum); */
2284                 if (in_cksum(m, hlen) != 0)
2285                         goto bad;
2286                 break;
2287         }
2288 
2289         /* Retrieve the packet length. */
2290         len = ntohs(ip->ip_len);
2291 
2292         /*
2293          * Check for additional length bogosity
2294          */
2295         if (len < hlen) {
2296 		ip_statinc(IP_STAT_BADLEN);
2297                 goto bad;
2298         }
2299 
2300         /*
2301          * Check that the amount of data in the buffers
2302          * is as at least much as the IP header would have us expect.
2303          * Drop packet if shorter than we expect.
2304          */
2305         if (m->m_pkthdr.len < len) {
2306 		ip_statinc(IP_STAT_TOOSHORT);
2307                 goto bad;
2308         }
2309 
2310 	/* Checks out, proceed */
2311 	*mp = m;
2312 	return 0;
2313 
2314     bad:
2315 	*mp = m;
2316 	return -1;
2317 }
2318 
2319 # ifdef INET6
2320 /*
2321  * Same as above, but for IPv6.
2322  * Cut-and-pasted from ip6_input.c.
2323  * XXX Should we update ip6stat, or not?
2324  */
2325 static int
2326 bridge_ip6_checkbasic(struct mbuf **mp)
2327 {
2328 	struct mbuf *m = *mp;
2329 	struct ip6_hdr *ip6;
2330 
2331         /*
2332          * If the IPv6 header is not aligned, slurp it up into a new
2333          * mbuf with space for link headers, in the event we forward
2334          * it.  Otherwise, if it is aligned, make sure the entire base
2335          * IPv6 header is in the first mbuf of the chain.
2336          */
2337         if (IP6_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
2338                 struct ifnet *inifp = m->m_pkthdr.rcvif;
2339                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
2340                                   (max_linkhdr + 3) & ~3)) == NULL) {
2341                         /* XXXJRT new stat, please */
2342 			ip6_statinc(IP6_STAT_TOOSMALL);
2343                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
2344                         goto bad;
2345                 }
2346         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
2347                 struct ifnet *inifp = m->m_pkthdr.rcvif;
2348                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
2349 			ip6_statinc(IP6_STAT_TOOSMALL);
2350                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
2351                         goto bad;
2352                 }
2353         }
2354 
2355         ip6 = mtod(m, struct ip6_hdr *);
2356 
2357         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
2358 		ip6_statinc(IP6_STAT_BADVERS);
2359                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
2360                 goto bad;
2361         }
2362 
2363 	/* Checks out, proceed */
2364 	*mp = m;
2365 	return 0;
2366 
2367     bad:
2368 	*mp = m;
2369 	return -1;
2370 }
2371 # endif /* INET6 */
2372 #endif /* BRIDGE_IPF */
2373