xref: /netbsd-src/sys/net/if_bridge.c (revision c9e0a13b6e3d74db73a38ceca211df00b41debde)
1 /*	$NetBSD: if_bridge.c,v 1.196 2024/12/16 05:21:24 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 1.196 2024/12/16 05:21:24 ozaki-r Exp $");
84 
85 #ifdef _KERNEL_OPT
86 #include "opt_inet.h"
87 #include "opt_net_mpsafe.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 #include <sys/kmem.h>
105 #include <sys/syslog.h>
106 
107 #include <net/bpf.h>
108 #include <net/if.h>
109 #include <net/if_dl.h>
110 #include <net/if_types.h>
111 #include <net/if_llc.h>
112 
113 #include <net/if_ether.h>
114 #include <net/if_bridgevar.h>
115 #include <net/ether_sw_offload.h>
116 
117 /* Used for bridge_ip[6]_checkbasic */
118 #include <netinet/in.h>
119 #include <netinet/in_systm.h>
120 #include <netinet/ip.h>
121 #include <netinet/ip_var.h>
122 #include <netinet/ip_private.h>		/* XXX */
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 
128 /*
129  * Size of the route hash table.  Must be a power of two.
130  */
131 #ifndef BRIDGE_RTHASH_SIZE
132 #define	BRIDGE_RTHASH_SIZE		1024
133 #endif
134 
135 #define	BRIDGE_RTHASH_MASK		(BRIDGE_RTHASH_SIZE - 1)
136 
137 #include "carp.h"
138 #if NCARP > 0
139 #include <netinet/in.h>
140 #include <netinet/in_var.h>
141 #include <netinet/ip_carp.h>
142 #endif
143 
144 #include "ioconf.h"
145 
146 __CTASSERT(sizeof(struct ifbifconf) == sizeof(struct ifbaconf));
147 __CTASSERT(offsetof(struct ifbifconf, ifbic_len) == offsetof(struct ifbaconf, ifbac_len));
148 __CTASSERT(offsetof(struct ifbifconf, ifbic_buf) == offsetof(struct ifbaconf, ifbac_buf));
149 
150 /*
151  * Maximum number of addresses to cache.
152  */
153 #ifndef BRIDGE_RTABLE_MAX
154 #define	BRIDGE_RTABLE_MAX		100
155 #endif
156 
157 /*
158  * Spanning tree defaults.
159  */
160 #define	BSTP_DEFAULT_MAX_AGE		(20 * 256)
161 #define	BSTP_DEFAULT_HELLO_TIME		(2 * 256)
162 #define	BSTP_DEFAULT_FORWARD_DELAY	(15 * 256)
163 #define	BSTP_DEFAULT_HOLD_TIME		(1 * 256)
164 #define	BSTP_DEFAULT_BRIDGE_PRIORITY	0x8000
165 #define	BSTP_DEFAULT_PORT_PRIORITY	0x80
166 #define	BSTP_DEFAULT_PATH_COST		55
167 
168 /*
169  * Timeout (in seconds) for entries learned dynamically.
170  */
171 #ifndef BRIDGE_RTABLE_TIMEOUT
172 #define	BRIDGE_RTABLE_TIMEOUT		(20 * 60)	/* same as ARP */
173 #endif
174 
175 /*
176  * Number of seconds between walks of the route list.
177  */
178 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
179 #define	BRIDGE_RTABLE_PRUNE_PERIOD	(5 * 60)
180 #endif
181 
182 #define BRIDGE_RT_LOCK(_sc)	mutex_enter((_sc)->sc_rtlist_lock)
183 #define BRIDGE_RT_UNLOCK(_sc)	mutex_exit((_sc)->sc_rtlist_lock)
184 #define BRIDGE_RT_LOCKED(_sc)	mutex_owned((_sc)->sc_rtlist_lock)
185 
186 #define BRIDGE_RT_PSZ_PERFORM(_sc) \
187 				pserialize_perform((_sc)->sc_rtlist_psz)
188 
189 #define BRIDGE_RTLIST_READER_FOREACH(_brt, _sc)			\
190 	PSLIST_READER_FOREACH((_brt), &((_sc)->sc_rtlist),		\
191 	    struct bridge_rtnode, brt_list)
192 #define BRIDGE_RTLIST_WRITER_FOREACH(_brt, _sc)			\
193 	PSLIST_WRITER_FOREACH((_brt), &((_sc)->sc_rtlist),		\
194 	    struct bridge_rtnode, brt_list)
195 #define BRIDGE_RTLIST_WRITER_INSERT_HEAD(_sc, _brt)			\
196 	PSLIST_WRITER_INSERT_HEAD(&(_sc)->sc_rtlist, brt, brt_list)
197 #define BRIDGE_RTLIST_WRITER_REMOVE(_brt)				\
198 	PSLIST_WRITER_REMOVE((_brt), brt_list)
199 
200 #define BRIDGE_RTHASH_READER_FOREACH(_brt, _sc, _hash)			\
201 	PSLIST_READER_FOREACH((_brt), &(_sc)->sc_rthash[(_hash)],	\
202 	    struct bridge_rtnode, brt_hash)
203 #define BRIDGE_RTHASH_WRITER_FOREACH(_brt, _sc, _hash)			\
204 	PSLIST_WRITER_FOREACH((_brt), &(_sc)->sc_rthash[(_hash)],	\
205 	    struct bridge_rtnode, brt_hash)
206 #define BRIDGE_RTHASH_WRITER_INSERT_HEAD(_sc, _hash, _brt)		\
207 	PSLIST_WRITER_INSERT_HEAD(&(_sc)->sc_rthash[(_hash)], brt, brt_hash)
208 #define BRIDGE_RTHASH_WRITER_INSERT_AFTER(_brt, _new)			\
209 	PSLIST_WRITER_INSERT_AFTER((_brt), (_new), brt_hash)
210 #define BRIDGE_RTHASH_WRITER_REMOVE(_brt)				\
211 	PSLIST_WRITER_REMOVE((_brt), brt_hash)
212 
213 #ifdef NET_MPSAFE
214 #define DECLARE_LOCK_VARIABLE
215 #define ACQUIRE_GLOBAL_LOCKS()	do { } while (0)
216 #define RELEASE_GLOBAL_LOCKS()	do { } while (0)
217 #else
218 #define DECLARE_LOCK_VARIABLE	int __s
219 #define ACQUIRE_GLOBAL_LOCKS()	do {					\
220 					KERNEL_LOCK(1, NULL);		\
221 					mutex_enter(softnet_lock);	\
222 					__s = splsoftnet();		\
223 				} while (0)
224 #define RELEASE_GLOBAL_LOCKS()	do {					\
225 					splx(__s);			\
226 					mutex_exit(softnet_lock);	\
227 					KERNEL_UNLOCK_ONE(NULL);	\
228 				} while (0)
229 #endif
230 
231 struct psref_class *bridge_psref_class __read_mostly;
232 
233 int	bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
234 
235 static struct pool bridge_rtnode_pool;
236 
237 static int	bridge_clone_create(struct if_clone *, int);
238 static int	bridge_clone_destroy(struct ifnet *);
239 
240 static int	bridge_ioctl(struct ifnet *, u_long, void *);
241 static int	bridge_init(struct ifnet *);
242 static void	bridge_stop(struct ifnet *, int);
243 static void	bridge_start(struct ifnet *);
244 static void	bridge_ifdetach(void *);
245 
246 static void	bridge_input(struct ifnet *, struct mbuf *);
247 static void	bridge_forward(struct bridge_softc *, struct mbuf *);
248 
249 static void	bridge_timer(void *);
250 
251 static void	bridge_broadcast(struct bridge_softc *, struct ifnet *, bool,
252 				 struct mbuf *);
253 
254 static int	bridge_rtupdate(struct bridge_softc *, const uint8_t *,
255 				struct ifnet *, int, uint8_t);
256 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *);
257 static void	bridge_rttrim(struct bridge_softc *);
258 static void	bridge_rtage(struct bridge_softc *);
259 static void	bridge_rtage_work(struct work *, void *);
260 static void	bridge_rtflush(struct bridge_softc *, int);
261 static int	bridge_rtdaddr(struct bridge_softc *, const uint8_t *);
262 static void	bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp);
263 
264 static void	bridge_rtable_init(struct bridge_softc *);
265 static void	bridge_rtable_fini(struct bridge_softc *);
266 
267 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
268 						  const uint8_t *);
269 static int	bridge_rtnode_insert(struct bridge_softc *,
270 				     struct bridge_rtnode *);
271 static void	bridge_rtnode_remove(struct bridge_softc *,
272 				     struct bridge_rtnode *);
273 static void	bridge_rtnode_destroy(struct bridge_rtnode *);
274 
275 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
276 						  const char *name,
277 						  struct psref *);
278 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
279 						     struct ifnet *ifp,
280 						     struct psref *);
281 static void	bridge_release_member(struct bridge_softc *, struct bridge_iflist *,
282                                       struct psref *);
283 static void	bridge_delete_member(struct bridge_softc *,
284 				     struct bridge_iflist *);
285 static void	bridge_acquire_member(struct bridge_softc *sc,
286                                       struct bridge_iflist *,
287                                       struct psref *);
288 
289 static int	bridge_ioctl_add(struct bridge_softc *, void *);
290 static int	bridge_ioctl_del(struct bridge_softc *, void *);
291 static int	bridge_ioctl_gifflags(struct bridge_softc *, void *);
292 static int	bridge_ioctl_sifflags(struct bridge_softc *, void *);
293 static int	bridge_ioctl_scache(struct bridge_softc *, void *);
294 static int	bridge_ioctl_gcache(struct bridge_softc *, void *);
295 static int	bridge_ioctl_gifs(struct bridge_softc *, void *);
296 static int	bridge_ioctl_rts(struct bridge_softc *, void *);
297 static int	bridge_ioctl_saddr(struct bridge_softc *, void *);
298 static int	bridge_ioctl_sto(struct bridge_softc *, void *);
299 static int	bridge_ioctl_gto(struct bridge_softc *, void *);
300 static int	bridge_ioctl_daddr(struct bridge_softc *, void *);
301 static int	bridge_ioctl_flush(struct bridge_softc *, void *);
302 static int	bridge_ioctl_gpri(struct bridge_softc *, void *);
303 static int	bridge_ioctl_spri(struct bridge_softc *, void *);
304 static int	bridge_ioctl_ght(struct bridge_softc *, void *);
305 static int	bridge_ioctl_sht(struct bridge_softc *, void *);
306 static int	bridge_ioctl_gfd(struct bridge_softc *, void *);
307 static int	bridge_ioctl_sfd(struct bridge_softc *, void *);
308 static int	bridge_ioctl_gma(struct bridge_softc *, void *);
309 static int	bridge_ioctl_sma(struct bridge_softc *, void *);
310 static int	bridge_ioctl_sifprio(struct bridge_softc *, void *);
311 static int	bridge_ioctl_sifcost(struct bridge_softc *, void *);
312 static int	bridge_ioctl_gfilt(struct bridge_softc *, void *);
313 static int	bridge_ioctl_sfilt(struct bridge_softc *, void *);
314 static int	bridge_ipf(void *, struct mbuf **, struct ifnet *, int);
315 static int	bridge_ip_checkbasic(struct mbuf **mp);
316 # ifdef INET6
317 static int	bridge_ip6_checkbasic(struct mbuf **mp);
318 # endif /* INET6 */
319 
320 struct bridge_control {
321 	int	(*bc_func)(struct bridge_softc *, void *);
322 	int	bc_argsize;
323 	int	bc_flags;
324 };
325 
326 #define	BC_F_COPYIN		0x01	/* copy arguments in */
327 #define	BC_F_COPYOUT		0x02	/* copy arguments out */
328 #define	BC_F_SUSER		0x04	/* do super-user check */
329 #define BC_F_XLATEIN		0x08	/* xlate arguments in */
330 #define BC_F_XLATEOUT		0x10	/* xlate arguments out */
331 
332 static const struct bridge_control bridge_control_table[] = {
333 [BRDGADD] = {bridge_ioctl_add, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
334 [BRDGDEL] = {bridge_ioctl_del, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
335 
336 [BRDGGIFFLGS] = {bridge_ioctl_gifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_COPYOUT},
337 [BRDGSIFFLGS] = {bridge_ioctl_sifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
338 
339 [BRDGSCACHE] = {bridge_ioctl_scache, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
340 [BRDGGCACHE] = {bridge_ioctl_gcache, sizeof(struct ifbrparam), BC_F_COPYOUT},
341 
342 [OBRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_COPYIN|BC_F_COPYOUT},
343 [OBRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_COPYIN|BC_F_COPYOUT},
344 
345 [BRDGSADDR] = {bridge_ioctl_saddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
346 
347 [BRDGSTO] = {bridge_ioctl_sto, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
348 [BRDGGTO] = {bridge_ioctl_gto, sizeof(struct ifbrparam), BC_F_COPYOUT},
349 
350 [BRDGDADDR] = {bridge_ioctl_daddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
351 
352 [BRDGFLUSH] = {bridge_ioctl_flush, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
353 
354 [BRDGGPRI] = {bridge_ioctl_gpri, sizeof(struct ifbrparam), BC_F_COPYOUT},
355 [BRDGSPRI] = {bridge_ioctl_spri, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
356 
357 [BRDGGHT] = {bridge_ioctl_ght, sizeof(struct ifbrparam), BC_F_COPYOUT},
358 [BRDGSHT] = {bridge_ioctl_sht, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
359 
360 [BRDGGFD] = {bridge_ioctl_gfd, sizeof(struct ifbrparam), BC_F_COPYOUT},
361 [BRDGSFD] = {bridge_ioctl_sfd, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
362 
363 [BRDGGMA] = {bridge_ioctl_gma, sizeof(struct ifbrparam), BC_F_COPYOUT},
364 [BRDGSMA] = {bridge_ioctl_sma, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
365 
366 [BRDGSIFPRIO] = {bridge_ioctl_sifprio, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
367 
368 [BRDGSIFCOST] = {bridge_ioctl_sifcost, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
369 
370 [BRDGGFILT] = {bridge_ioctl_gfilt, sizeof(struct ifbrparam), BC_F_COPYOUT},
371 [BRDGSFILT] = {bridge_ioctl_sfilt, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
372 
373 [BRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_XLATEIN|BC_F_XLATEOUT},
374 [BRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_XLATEIN|BC_F_XLATEOUT},
375 };
376 
377 static const int bridge_control_table_size = __arraycount(bridge_control_table);
378 
379 static struct if_clone bridge_cloner =
380     IF_CLONE_INITIALIZER("bridge", bridge_clone_create, bridge_clone_destroy);
381 
382 /*
383  * bridgeattach:
384  *
385  *	Pseudo-device attach routine.
386  */
387 void
388 bridgeattach(int n)
389 {
390 
391 	pool_init(&bridge_rtnode_pool, sizeof(struct bridge_rtnode),
392 	    0, 0, 0, "brtpl", NULL, IPL_NET);
393 
394 	bridge_psref_class = psref_class_create("bridge", IPL_SOFTNET);
395 
396 	if_clone_attach(&bridge_cloner);
397 }
398 
399 /*
400  * bridge_clone_create:
401  *
402  *	Create a new bridge instance.
403  */
404 static int
405 bridge_clone_create(struct if_clone *ifc, int unit)
406 {
407 	struct bridge_softc *sc;
408 	struct ifnet *ifp;
409 	int error;
410 
411 	sc = kmem_zalloc(sizeof(*sc),  KM_SLEEP);
412 	ifp = &sc->sc_if;
413 
414 	sc->sc_brtmax = BRIDGE_RTABLE_MAX;
415 	sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
416 	sc->sc_bridge_max_age = BSTP_DEFAULT_MAX_AGE;
417 	sc->sc_bridge_hello_time = BSTP_DEFAULT_HELLO_TIME;
418 	sc->sc_bridge_forward_delay = BSTP_DEFAULT_FORWARD_DELAY;
419 	sc->sc_bridge_priority = BSTP_DEFAULT_BRIDGE_PRIORITY;
420 	sc->sc_hold_time = BSTP_DEFAULT_HOLD_TIME;
421 	sc->sc_filter_flags = 0;
422 
423 	/* Initialize our routing table. */
424 	bridge_rtable_init(sc);
425 
426 	error = workqueue_create(&sc->sc_rtage_wq, "bridge_rtage",
427 	    bridge_rtage_work, sc, PRI_SOFTNET, IPL_SOFTNET, WQ_MPSAFE);
428 	if (error)
429 		panic("%s: workqueue_create %d\n", __func__, error);
430 
431 	callout_init(&sc->sc_brcallout, CALLOUT_MPSAFE);
432 	callout_init(&sc->sc_bstpcallout, CALLOUT_MPSAFE);
433 
434 	mutex_init(&sc->sc_iflist_psref.bip_lock, MUTEX_DEFAULT, IPL_NONE);
435 	PSLIST_INIT(&sc->sc_iflist_psref.bip_iflist);
436 	sc->sc_iflist_psref.bip_psz = pserialize_create();
437 
438 	if_initname(ifp, ifc->ifc_name, unit);
439 	ifp->if_softc = sc;
440 #ifdef NET_MPSAFE
441 	ifp->if_extflags = IFEF_MPSAFE;
442 #endif
443 	ifp->if_mtu = ETHERMTU;
444 	ifp->if_ioctl = bridge_ioctl;
445 	ifp->if_output = bridge_output;
446 	ifp->if_start = bridge_start;
447 	ifp->if_stop = bridge_stop;
448 	ifp->if_init = bridge_init;
449 	ifp->if_type = IFT_BRIDGE;
450 	ifp->if_addrlen = 0;
451 	ifp->if_dlt = DLT_EN10MB;
452 	ifp->if_hdrlen = ETHER_HDR_LEN;
453 	if_initialize(ifp);
454 
455 	/*
456 	 * Set the link state to down.
457 	 * When interfaces are added the link state will reflect
458 	 * the best link state of the combined interfaces.
459 	 */
460 	ifp->if_link_state = LINK_STATE_DOWN;
461 
462 	if_alloc_sadl(ifp);
463 	if_register(ifp);
464 
465 	return 0;
466 }
467 
468 /*
469  * bridge_clone_destroy:
470  *
471  *	Destroy a bridge instance.
472  */
473 static int
474 bridge_clone_destroy(struct ifnet *ifp)
475 {
476 	struct bridge_softc *sc = ifp->if_softc;
477 	struct bridge_iflist *bif;
478 
479 	if ((ifp->if_flags & IFF_RUNNING) != 0)
480 		bridge_stop(ifp, 1);
481 
482 	BRIDGE_LOCK(sc);
483 	for (;;) {
484 		bif = PSLIST_WRITER_FIRST(&sc->sc_iflist_psref.bip_iflist, struct bridge_iflist,
485 		    bif_next);
486 		if (bif == NULL)
487 			break;
488 		bridge_delete_member(sc, bif);
489 	}
490 	PSLIST_DESTROY(&sc->sc_iflist_psref.bip_iflist);
491 	BRIDGE_UNLOCK(sc);
492 
493 	if_detach(ifp);
494 
495 	/* Tear down the routing table. */
496 	bridge_rtable_fini(sc);
497 
498 	pserialize_destroy(sc->sc_iflist_psref.bip_psz);
499 	mutex_destroy(&sc->sc_iflist_psref.bip_lock);
500 	callout_destroy(&sc->sc_brcallout);
501 	callout_destroy(&sc->sc_bstpcallout);
502 	workqueue_destroy(sc->sc_rtage_wq);
503 	kmem_free(sc, sizeof(*sc));
504 
505 	return 0;
506 }
507 
508 /*
509  * bridge_ioctl:
510  *
511  *	Handle a control request from the operator.
512  */
513 static int
514 bridge_ioctl(struct ifnet *ifp, u_long cmd, void *data)
515 {
516 	struct bridge_softc *sc = ifp->if_softc;
517 	struct lwp *l = curlwp;	/* XXX */
518 	union {
519 		struct ifbreq ifbreq;
520 		struct ifbifconf ifbifconf;
521 		struct ifbareq ifbareq;
522 		struct ifbaconf ifbaconf;
523 		struct ifbrparam ifbrparam;
524 	} args;
525 	struct ifdrv *ifd = (struct ifdrv *) data;
526 	const struct bridge_control *bc = NULL; /* XXXGCC */
527 	int error = 0;
528 
529 	/* Authorize command before calling splsoftnet(). */
530 	switch (cmd) {
531 	case SIOCGDRVSPEC:
532 	case SIOCSDRVSPEC:
533 		if (ifd->ifd_cmd >= bridge_control_table_size
534 		    || (bc = &bridge_control_table[ifd->ifd_cmd]) == NULL) {
535 			error = EINVAL;
536 			return error;
537 		}
538 
539 		/* We only care about BC_F_SUSER at this point. */
540 		if ((bc->bc_flags & BC_F_SUSER) == 0)
541 			break;
542 
543 		error = kauth_authorize_network(l->l_cred,
544 		    KAUTH_NETWORK_INTERFACE_BRIDGE,
545 		    cmd == SIOCGDRVSPEC ?
546 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_GETPRIV :
547 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_SETPRIV,
548 		     ifd, NULL, NULL);
549 		if (error)
550 			return error;
551 
552 		break;
553 	}
554 
555 	const int s = splsoftnet();
556 
557 	switch (cmd) {
558 	case SIOCGDRVSPEC:
559 	case SIOCSDRVSPEC:
560 		KASSERT(bc != NULL);
561 		if (cmd == SIOCGDRVSPEC &&
562 		    (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) == 0) {
563 			error = EINVAL;
564 			break;
565 		}
566 		else if (cmd == SIOCSDRVSPEC &&
567 		    (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) != 0) {
568 			error = EINVAL;
569 			break;
570 		}
571 
572 		/* BC_F_SUSER is checked above, before splsoftnet(). */
573 
574 		if ((bc->bc_flags & (BC_F_XLATEIN|BC_F_XLATEOUT)) == 0
575 		    && (ifd->ifd_len != bc->bc_argsize
576 			|| ifd->ifd_len > sizeof(args))) {
577 			error = EINVAL;
578 			break;
579 		}
580 
581 		memset(&args, 0, sizeof(args));
582 		if (bc->bc_flags & BC_F_COPYIN) {
583 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
584 			if (error)
585 				break;
586 		} else if (bc->bc_flags & BC_F_XLATEIN) {
587 			args.ifbifconf.ifbic_len = ifd->ifd_len;
588 			args.ifbifconf.ifbic_buf = ifd->ifd_data;
589 		}
590 
591 		error = (*bc->bc_func)(sc, &args);
592 		if (error)
593 			break;
594 
595 		if (bc->bc_flags & BC_F_COPYOUT) {
596 			error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
597 		} else if (bc->bc_flags & BC_F_XLATEOUT) {
598 			ifd->ifd_len = args.ifbifconf.ifbic_len;
599 			ifd->ifd_data = args.ifbifconf.ifbic_buf;
600 		}
601 		break;
602 
603 	case SIOCSIFFLAGS:
604 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
605 			break;
606 		switch (ifp->if_flags & (IFF_UP|IFF_RUNNING)) {
607 		case IFF_RUNNING:
608 			/*
609 			 * If interface is marked down and it is running,
610 			 * then stop and disable it.
611 			 */
612 			if_stop(ifp, 1);
613 			break;
614 		case IFF_UP:
615 			/*
616 			 * If interface is marked up and it is stopped, then
617 			 * start it.
618 			 */
619 			error = if_init(ifp);
620 			break;
621 		default:
622 			break;
623 		}
624 		break;
625 
626 	case SIOCSIFMTU:
627 		if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET)
628 			error = 0;
629 		break;
630 
631         case SIOCGIFCAP:
632 	    {
633 		struct ifcapreq *ifcr = (struct ifcapreq *)data;
634                 ifcr->ifcr_capabilities = sc->sc_capenable;
635                 ifcr->ifcr_capenable = sc->sc_capenable;
636 		break;
637 	    }
638 
639 	default:
640 		error = ifioctl_common(ifp, cmd, data);
641 		break;
642 	}
643 
644 	splx(s);
645 
646 	return error;
647 }
648 
649 /*
650  * bridge_lookup_member:
651  *
652  *	Lookup a bridge member interface.
653  */
654 static struct bridge_iflist *
655 bridge_lookup_member(struct bridge_softc *sc, const char *name, struct psref *psref)
656 {
657 	struct bridge_iflist *bif;
658 	struct ifnet *ifp;
659 	int s;
660 
661 	s = pserialize_read_enter();
662 
663 	BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
664 		ifp = bif->bif_ifp;
665 		if (strcmp(ifp->if_xname, name) == 0)
666 			break;
667 	}
668 	if (bif != NULL)
669 		bridge_acquire_member(sc, bif, psref);
670 
671 	pserialize_read_exit(s);
672 
673 	return bif;
674 }
675 
676 /*
677  * bridge_lookup_member_if:
678  *
679  *	Lookup a bridge member interface by ifnet*.
680  */
681 static struct bridge_iflist *
682 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp,
683     struct psref *psref)
684 {
685 	struct bridge_iflist *bif;
686 	int s;
687 
688 	s = pserialize_read_enter();
689 
690 	bif = member_ifp->if_bridgeif;
691 	if (bif != NULL) {
692 		psref_acquire(psref, &bif->bif_psref,
693 		    bridge_psref_class);
694 	}
695 
696 	pserialize_read_exit(s);
697 
698 	return bif;
699 }
700 
701 static void
702 bridge_acquire_member(struct bridge_softc *sc, struct bridge_iflist *bif,
703     struct psref *psref)
704 {
705 
706 	psref_acquire(psref, &bif->bif_psref, bridge_psref_class);
707 }
708 
709 /*
710  * bridge_release_member:
711  *
712  *	Release the specified member interface.
713  */
714 static void
715 bridge_release_member(struct bridge_softc *sc, struct bridge_iflist *bif,
716     struct psref *psref)
717 {
718 
719 	psref_release(psref, &bif->bif_psref, bridge_psref_class);
720 }
721 
722 /*
723  * bridge_delete_member:
724  *
725  *	Delete the specified member interface.
726  */
727 static void
728 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif)
729 {
730 	struct ifnet *ifs = bif->bif_ifp;
731 
732 	KASSERT(BRIDGE_LOCKED(sc));
733 
734 	ifs->_if_input = ether_input;
735 	ifs->if_bridge = NULL;
736 	ifs->if_bridgeif = NULL;
737 
738 	PSLIST_WRITER_REMOVE(bif, bif_next);
739 	BRIDGE_PSZ_PERFORM(sc);
740 
741 	if_linkstate_change_disestablish(ifs,
742 	    bif->bif_linkstate_hook, BRIDGE_LOCK_OBJ(sc));
743 	ether_ifdetachhook_disestablish(ifs,
744 	    bif->bif_ifdetach_hook, BRIDGE_LOCK_OBJ(sc));
745 
746 	BRIDGE_UNLOCK(sc);
747 
748 	switch (ifs->if_type) {
749 	case IFT_ETHER:
750 	case IFT_L2TP:
751 		/*
752 		 * Take the interface out of promiscuous mode.
753 		 * Don't call it with holding a spin lock.
754 		 */
755 		(void) ifpromisc(ifs, 0);
756 		IFNET_LOCK(ifs);
757 		(void) ether_disable_vlan_mtu(ifs);
758 		IFNET_UNLOCK(ifs);
759 		break;
760 	default:
761 #ifdef DIAGNOSTIC
762 		panic("%s: impossible", __func__);
763 #endif
764 		break;
765 	}
766 
767 	psref_target_destroy(&bif->bif_psref, bridge_psref_class);
768 
769 	PSLIST_ENTRY_DESTROY(bif, bif_next);
770 	kmem_free(bif, sizeof(*bif));
771 
772 	BRIDGE_LOCK(sc);
773 }
774 
775 /*
776  * bridge_calc_csum_flags:
777  *
778  *	Calculate logical and b/w csum flags each member interface supports.
779  */
780 void
781 bridge_calc_csum_flags(struct bridge_softc *sc)
782 {
783 	struct bridge_iflist *bif;
784 	struct ifnet *ifs = NULL;
785 	int flags = ~0;
786 	int capenable = ~0;
787 
788 	BRIDGE_LOCK(sc);
789 	BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
790 		ifs = bif->bif_ifp;
791 		flags &= ifs->if_csum_flags_tx;
792 		capenable &= ifs->if_capenable;
793 	}
794 	sc->sc_csum_flags_tx = flags;
795 	sc->sc_capenable = (ifs != NULL) ? capenable : 0;
796 	BRIDGE_UNLOCK(sc);
797 }
798 
799 /*
800  * bridge_calc_link_state:
801  *
802  *	Calculate the link state based on each member interface.
803  */
804 static void
805 bridge_calc_link_state(void *xsc)
806 {
807 	struct bridge_softc *sc = xsc;
808 	struct bridge_iflist *bif;
809 	struct ifnet *ifs;
810 	int link_state = LINK_STATE_DOWN;
811 
812 	BRIDGE_LOCK(sc);
813 	BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
814 		ifs = bif->bif_ifp;
815 		if (ifs->if_link_state == LINK_STATE_UP) {
816 			link_state = LINK_STATE_UP;
817 			break;
818 		}
819 		if (ifs->if_link_state == LINK_STATE_UNKNOWN)
820 			link_state = LINK_STATE_UNKNOWN;
821 	}
822 	if_link_state_change(&sc->sc_if, link_state);
823 	BRIDGE_UNLOCK(sc);
824 }
825 
826 static int
827 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
828 {
829 	struct ifbreq *req = arg;
830 	struct bridge_iflist *bif = NULL;
831 	struct ifnet *ifs;
832 	int error = 0;
833 	struct psref psref;
834 
835 	ifs = if_get(req->ifbr_ifsname, &psref);
836 	if (ifs == NULL)
837 		return ENOENT;
838 
839 	if (ifs->if_bridge == sc) {
840 		error = EEXIST;
841 		goto out;
842 	}
843 
844 	if (ifs->if_bridge != NULL) {
845 		error = EBUSY;
846 		goto out;
847 	}
848 
849 	if (ifs->_if_input != ether_input) {
850 		error = EINVAL;
851 		goto out;
852 	}
853 
854 	/* FIXME: doesn't work with non-IFF_SIMPLEX interfaces */
855 	if ((ifs->if_flags & IFF_SIMPLEX) == 0) {
856 		error = EINVAL;
857 		goto out;
858 	}
859 
860 	bif = kmem_alloc(sizeof(*bif), KM_SLEEP);
861 
862 	switch (ifs->if_type) {
863 	case IFT_ETHER:
864 		if (sc->sc_if.if_mtu != ifs->if_mtu) {
865 			/* Change MTU of added interface to bridge MTU */
866 			struct ifreq ifr;
867 			memset(&ifr, 0, sizeof(ifr));
868 			ifr.ifr_mtu = sc->sc_if.if_mtu;
869 			IFNET_LOCK(ifs);
870 			error = if_ioctl(ifs, SIOCSIFMTU, &ifr);
871 			IFNET_UNLOCK(ifs);
872 			if (error != 0)
873 				goto out;
874 		}
875 		/* FALLTHROUGH */
876 	case IFT_L2TP:
877 		IFNET_LOCK(ifs);
878 		error = ether_enable_vlan_mtu(ifs);
879 		IFNET_UNLOCK(ifs);
880 		if (error > 0)
881 			goto out;
882 		/*
883 		 * Place the interface into promiscuous mode.
884 		 */
885 		error = ifpromisc(ifs, 1);
886 		if (error)
887 			goto out;
888 		break;
889 	default:
890 		error = EINVAL;
891 		goto out;
892 	}
893 
894 	bif->bif_ifp = ifs;
895 	bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
896 	bif->bif_priority = BSTP_DEFAULT_PORT_PRIORITY;
897 	bif->bif_path_cost = BSTP_DEFAULT_PATH_COST;
898 	bif->bif_linkstate_hook = if_linkstate_change_establish(ifs,
899 	    bridge_calc_link_state, sc);
900 	PSLIST_ENTRY_INIT(bif, bif_next);
901 	psref_target_init(&bif->bif_psref, bridge_psref_class);
902 
903 	BRIDGE_LOCK(sc);
904 
905 	ifs->if_bridge = sc;
906 	ifs->if_bridgeif = bif;
907 	PSLIST_WRITER_INSERT_HEAD(&sc->sc_iflist_psref.bip_iflist, bif, bif_next);
908 	ifs->_if_input = bridge_input;
909 
910 	BRIDGE_UNLOCK(sc);
911 
912 	bif->bif_ifdetach_hook = ether_ifdetachhook_establish(ifs,
913 	    bridge_ifdetach, (void *)ifs);
914 
915 	bridge_calc_csum_flags(sc);
916 	bridge_calc_link_state(sc);
917 
918 	if (sc->sc_if.if_flags & IFF_RUNNING)
919 		bstp_initialization(sc);
920 	else
921 		bstp_stop(sc);
922 
923 out:
924 	if_put(ifs, &psref);
925 	if (error) {
926 		if (bif != NULL)
927 			kmem_free(bif, sizeof(*bif));
928 	}
929 	return error;
930 }
931 
932 static int
933 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
934 {
935 	struct ifbreq *req = arg;
936 	const char *name = req->ifbr_ifsname;
937 	struct bridge_iflist *bif;
938 	struct ifnet *ifs;
939 
940 	BRIDGE_LOCK(sc);
941 
942 	/*
943 	 * Don't use bridge_lookup_member. We want to get a member
944 	 * with bif_refs == 0.
945 	 */
946 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) {
947 		ifs = bif->bif_ifp;
948 		if (strcmp(ifs->if_xname, name) == 0)
949 			break;
950 	}
951 
952 	if (bif == NULL) {
953 		BRIDGE_UNLOCK(sc);
954 		return ENOENT;
955 	}
956 
957 	bridge_delete_member(sc, bif);
958 
959 	BRIDGE_UNLOCK(sc);
960 
961 	bridge_rtdelete(sc, ifs);
962 	bridge_calc_csum_flags(sc);
963 	bridge_calc_link_state(sc);
964 
965 	if (sc->sc_if.if_flags & IFF_RUNNING)
966 		bstp_initialization(sc);
967 
968 	return 0;
969 }
970 
971 static int
972 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
973 {
974 	struct ifbreq *req = arg;
975 	struct bridge_iflist *bif;
976 	struct psref psref;
977 
978 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
979 	if (bif == NULL)
980 		return ENOENT;
981 
982 	req->ifbr_ifsflags = bif->bif_flags;
983 	req->ifbr_state = bif->bif_state;
984 	req->ifbr_priority = bif->bif_priority;
985 	req->ifbr_path_cost = bif->bif_path_cost;
986 	req->ifbr_portno = bif->bif_ifp->if_index & 0xff;
987 
988 	bridge_release_member(sc, bif, &psref);
989 
990 	return 0;
991 }
992 
993 static int
994 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
995 {
996 	struct ifbreq *req = arg;
997 	struct bridge_iflist *bif;
998 	struct psref psref;
999 
1000 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
1001 	if (bif == NULL)
1002 		return ENOENT;
1003 
1004 	if (req->ifbr_ifsflags & IFBIF_STP) {
1005 		switch (bif->bif_ifp->if_type) {
1006 		case IFT_ETHER:
1007 		case IFT_L2TP:
1008 			/* These can do spanning tree. */
1009 			break;
1010 
1011 		default:
1012 			/* Nothing else can. */
1013 			bridge_release_member(sc, bif, &psref);
1014 			return EINVAL;
1015 		}
1016 	}
1017 
1018 	if (bif->bif_flags & IFBIF_PROTECTED) {
1019 		if ((req->ifbr_ifsflags & IFBIF_PROTECTED) == 0) {
1020 			log(LOG_INFO, "%s: disabling protection on %s\n",
1021 			    sc->sc_if.if_xname, bif->bif_ifp->if_xname);
1022 		}
1023 	} else {
1024 		if (req->ifbr_ifsflags & IFBIF_PROTECTED) {
1025 			log(LOG_INFO, "%s: enabling protection on %s\n",
1026 			    sc->sc_if.if_xname, bif->bif_ifp->if_xname);
1027 		}
1028 	}
1029 
1030 	bif->bif_flags = req->ifbr_ifsflags;
1031 
1032 	bridge_release_member(sc, bif, &psref);
1033 
1034 	if (sc->sc_if.if_flags & IFF_RUNNING)
1035 		bstp_initialization(sc);
1036 
1037 	return 0;
1038 }
1039 
1040 static int
1041 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1042 {
1043 	struct ifbrparam *param = arg;
1044 
1045 	sc->sc_brtmax = param->ifbrp_csize;
1046 	bridge_rttrim(sc);
1047 
1048 	return 0;
1049 }
1050 
1051 static int
1052 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1053 {
1054 	struct ifbrparam *param = arg;
1055 
1056 	param->ifbrp_csize = sc->sc_brtmax;
1057 
1058 	return 0;
1059 }
1060 
1061 static int
1062 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1063 {
1064 	struct ifbifconf *bifc = arg;
1065 	struct bridge_iflist *bif;
1066 	struct ifbreq *breqs;
1067 	int i, count, error = 0;
1068 
1069 retry:
1070 	BRIDGE_LOCK(sc);
1071 	count = 0;
1072 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc)
1073 		count++;
1074 	BRIDGE_UNLOCK(sc);
1075 
1076 	if (count == 0) {
1077 		bifc->ifbic_len = 0;
1078 		return 0;
1079 	}
1080 
1081 	if (bifc->ifbic_len == 0 || bifc->ifbic_len < (sizeof(*breqs) * count)) {
1082 		/* Tell that a larger buffer is needed */
1083 		bifc->ifbic_len = sizeof(*breqs) * count;
1084 		return 0;
1085 	}
1086 
1087 	breqs = kmem_alloc(sizeof(*breqs) * count, KM_SLEEP);
1088 
1089 	BRIDGE_LOCK(sc);
1090 
1091 	i = 0;
1092 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc)
1093 		i++;
1094 	if (i > count) {
1095 		/*
1096 		 * The number of members has been increased.
1097 		 * We need more memory!
1098 		 */
1099 		BRIDGE_UNLOCK(sc);
1100 		kmem_free(breqs, sizeof(*breqs) * count);
1101 		goto retry;
1102 	}
1103 
1104 	i = 0;
1105 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) {
1106 		struct ifbreq *breq = &breqs[i++];
1107 		memset(breq, 0, sizeof(*breq));
1108 
1109 		strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname,
1110 		    sizeof(breq->ifbr_ifsname));
1111 		breq->ifbr_ifsflags = bif->bif_flags;
1112 		breq->ifbr_state = bif->bif_state;
1113 		breq->ifbr_priority = bif->bif_priority;
1114 		breq->ifbr_path_cost = bif->bif_path_cost;
1115 		breq->ifbr_portno = bif->bif_ifp->if_index & 0xff;
1116 	}
1117 
1118 	/* Don't call copyout with holding the mutex */
1119 	BRIDGE_UNLOCK(sc);
1120 
1121 	for (i = 0; i < count; i++) {
1122 		error = copyout(&breqs[i], bifc->ifbic_req + i, sizeof(*breqs));
1123 		if (error)
1124 			break;
1125 	}
1126 	bifc->ifbic_len = sizeof(*breqs) * i;
1127 
1128 	kmem_free(breqs, sizeof(*breqs) * count);
1129 
1130 	return error;
1131 }
1132 
1133 static int
1134 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1135 {
1136 	struct ifbaconf *bac = arg;
1137 	struct bridge_rtnode *brt;
1138 	struct ifbareq bareq;
1139 	int count = 0, error = 0, len;
1140 
1141 	if (bac->ifbac_len == 0)
1142 		return 0;
1143 
1144 	BRIDGE_RT_LOCK(sc);
1145 
1146 	/* The passed buffer is not enough, tell a required size. */
1147 	if (bac->ifbac_len < (sizeof(bareq) * sc->sc_brtcnt)) {
1148 		count = sc->sc_brtcnt;
1149 		goto out;
1150 	}
1151 
1152 	len = bac->ifbac_len;
1153 	BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) {
1154 		if (len < sizeof(bareq))
1155 			goto out;
1156 		memset(&bareq, 0, sizeof(bareq));
1157 		strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1158 		    sizeof(bareq.ifba_ifsname));
1159 		memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1160 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
1161 			bareq.ifba_expire = brt->brt_expire - time_uptime;
1162 		} else
1163 			bareq.ifba_expire = 0;
1164 		bareq.ifba_flags = brt->brt_flags;
1165 
1166 		error = copyout(&bareq, bac->ifbac_req + count, sizeof(bareq));
1167 		if (error)
1168 			goto out;
1169 		count++;
1170 		len -= sizeof(bareq);
1171 	}
1172 out:
1173 	BRIDGE_RT_UNLOCK(sc);
1174 
1175 	bac->ifbac_len = sizeof(bareq) * count;
1176 	return error;
1177 }
1178 
1179 static int
1180 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1181 {
1182 	struct ifbareq *req = arg;
1183 	struct bridge_iflist *bif;
1184 	int error;
1185 	struct psref psref;
1186 
1187 	bif = bridge_lookup_member(sc, req->ifba_ifsname, &psref);
1188 	if (bif == NULL)
1189 		return ENOENT;
1190 
1191 	error = bridge_rtupdate(sc, req->ifba_dst, bif->bif_ifp, 1,
1192 	    req->ifba_flags);
1193 
1194 	bridge_release_member(sc, bif, &psref);
1195 
1196 	return error;
1197 }
1198 
1199 static int
1200 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1201 {
1202 	struct ifbrparam *param = arg;
1203 
1204 	sc->sc_brttimeout = param->ifbrp_ctime;
1205 
1206 	return 0;
1207 }
1208 
1209 static int
1210 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1211 {
1212 	struct ifbrparam *param = arg;
1213 
1214 	param->ifbrp_ctime = sc->sc_brttimeout;
1215 
1216 	return 0;
1217 }
1218 
1219 static int
1220 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1221 {
1222 	struct ifbareq *req = arg;
1223 
1224 	return (bridge_rtdaddr(sc, req->ifba_dst));
1225 }
1226 
1227 static int
1228 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1229 {
1230 	struct ifbreq *req = arg;
1231 
1232 	bridge_rtflush(sc, req->ifbr_ifsflags);
1233 
1234 	return 0;
1235 }
1236 
1237 static int
1238 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1239 {
1240 	struct ifbrparam *param = arg;
1241 
1242 	param->ifbrp_prio = sc->sc_bridge_priority;
1243 
1244 	return 0;
1245 }
1246 
1247 static int
1248 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1249 {
1250 	struct ifbrparam *param = arg;
1251 
1252 	sc->sc_bridge_priority = param->ifbrp_prio;
1253 
1254 	if (sc->sc_if.if_flags & IFF_RUNNING)
1255 		bstp_initialization(sc);
1256 
1257 	return 0;
1258 }
1259 
1260 static int
1261 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1262 {
1263 	struct ifbrparam *param = arg;
1264 
1265 	param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8;
1266 
1267 	return 0;
1268 }
1269 
1270 static int
1271 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1272 {
1273 	struct ifbrparam *param = arg;
1274 
1275 	if (param->ifbrp_hellotime == 0)
1276 		return EINVAL;
1277 	sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8;
1278 
1279 	if (sc->sc_if.if_flags & IFF_RUNNING)
1280 		bstp_initialization(sc);
1281 
1282 	return 0;
1283 }
1284 
1285 static int
1286 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1287 {
1288 	struct ifbrparam *param = arg;
1289 
1290 	param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8;
1291 
1292 	return 0;
1293 }
1294 
1295 static int
1296 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1297 {
1298 	struct ifbrparam *param = arg;
1299 
1300 	if (param->ifbrp_fwddelay == 0)
1301 		return EINVAL;
1302 	sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8;
1303 
1304 	if (sc->sc_if.if_flags & IFF_RUNNING)
1305 		bstp_initialization(sc);
1306 
1307 	return 0;
1308 }
1309 
1310 static int
1311 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1312 {
1313 	struct ifbrparam *param = arg;
1314 
1315 	param->ifbrp_maxage = sc->sc_bridge_max_age >> 8;
1316 
1317 	return 0;
1318 }
1319 
1320 static int
1321 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1322 {
1323 	struct ifbrparam *param = arg;
1324 
1325 	if (param->ifbrp_maxage == 0)
1326 		return EINVAL;
1327 	sc->sc_bridge_max_age = param->ifbrp_maxage << 8;
1328 
1329 	if (sc->sc_if.if_flags & IFF_RUNNING)
1330 		bstp_initialization(sc);
1331 
1332 	return 0;
1333 }
1334 
1335 static int
1336 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1337 {
1338 	struct ifbreq *req = arg;
1339 	struct bridge_iflist *bif;
1340 	struct psref psref;
1341 
1342 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
1343 	if (bif == NULL)
1344 		return ENOENT;
1345 
1346 	bif->bif_priority = req->ifbr_priority;
1347 
1348 	if (sc->sc_if.if_flags & IFF_RUNNING)
1349 		bstp_initialization(sc);
1350 
1351 	bridge_release_member(sc, bif, &psref);
1352 
1353 	return 0;
1354 }
1355 
1356 static int
1357 bridge_ioctl_gfilt(struct bridge_softc *sc, void *arg)
1358 {
1359 	struct ifbrparam *param = arg;
1360 
1361 	param->ifbrp_filter = sc->sc_filter_flags;
1362 
1363 	return 0;
1364 }
1365 
1366 static int
1367 bridge_ioctl_sfilt(struct bridge_softc *sc, void *arg)
1368 {
1369 	struct ifbrparam *param = arg;
1370 	uint32_t nflags, oflags;
1371 
1372 	if (param->ifbrp_filter & ~IFBF_FILT_MASK)
1373 		return EINVAL;
1374 
1375 	nflags = param->ifbrp_filter;
1376 	oflags = sc->sc_filter_flags;
1377 
1378 	if ((nflags & IFBF_FILT_USEIPF) && !(oflags & IFBF_FILT_USEIPF)) {
1379 		pfil_add_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
1380 			sc->sc_if.if_pfil);
1381 	}
1382 	if (!(nflags & IFBF_FILT_USEIPF) && (oflags & IFBF_FILT_USEIPF)) {
1383 		pfil_remove_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
1384 			sc->sc_if.if_pfil);
1385 	}
1386 
1387 	sc->sc_filter_flags = nflags;
1388 
1389 	return 0;
1390 }
1391 
1392 static int
1393 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1394 {
1395 	struct ifbreq *req = arg;
1396 	struct bridge_iflist *bif;
1397 	struct psref psref;
1398 
1399 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
1400 	if (bif == NULL)
1401 		return ENOENT;
1402 
1403 	bif->bif_path_cost = req->ifbr_path_cost;
1404 
1405 	if (sc->sc_if.if_flags & IFF_RUNNING)
1406 		bstp_initialization(sc);
1407 
1408 	bridge_release_member(sc, bif, &psref);
1409 
1410 	return 0;
1411 }
1412 
1413 /*
1414  * bridge_ifdetach:
1415  *
1416  *	Detach an interface from a bridge.  Called when a member
1417  *	interface is detaching.
1418  */
1419 static void
1420 bridge_ifdetach(void *xifs)
1421 {
1422 	struct ifnet *ifs;
1423 	struct bridge_softc *sc;
1424 	struct ifbreq breq;
1425 
1426 	ifs = (struct ifnet *)xifs;
1427 	sc = ifs->if_bridge;
1428 
1429 	/* ioctl_lock should prevent this from happening */
1430 	KASSERT(sc != NULL);
1431 
1432 	memset(&breq, 0, sizeof(breq));
1433 	strlcpy(breq.ifbr_ifsname, ifs->if_xname, sizeof(breq.ifbr_ifsname));
1434 
1435 	(void) bridge_ioctl_del(sc, &breq);
1436 }
1437 
1438 /*
1439  * bridge_init:
1440  *
1441  *	Initialize a bridge interface.
1442  */
1443 static int
1444 bridge_init(struct ifnet *ifp)
1445 {
1446 	struct bridge_softc *sc = ifp->if_softc;
1447 
1448 	KASSERT((ifp->if_flags & IFF_RUNNING) == 0);
1449 
1450 	callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1451 	    bridge_timer, sc);
1452 	bstp_initialization(sc);
1453 
1454 	ifp->if_flags |= IFF_RUNNING;
1455 	return 0;
1456 }
1457 
1458 /*
1459  * bridge_stop:
1460  *
1461  *	Stop the bridge interface.
1462  */
1463 static void
1464 bridge_stop(struct ifnet *ifp, int disable)
1465 {
1466 	struct bridge_softc *sc = ifp->if_softc;
1467 
1468 	KASSERT((ifp->if_flags & IFF_RUNNING) != 0);
1469 	ifp->if_flags &= ~IFF_RUNNING;
1470 
1471 	callout_halt(&sc->sc_brcallout, NULL);
1472 	workqueue_wait(sc->sc_rtage_wq, &sc->sc_rtage_wk);
1473 	bstp_stop(sc);
1474 	bridge_rtflush(sc, IFBF_FLUSHDYN);
1475 }
1476 
1477 /*
1478  * bridge_enqueue:
1479  *
1480  *	Enqueue a packet on a bridge member interface.
1481  */
1482 void
1483 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m,
1484     int runfilt)
1485 {
1486 	int len, error;
1487 	short mflags;
1488 
1489 	if (runfilt) {
1490 		if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
1491 		    dst_ifp, PFIL_OUT) != 0) {
1492 			m_freem(m);
1493 			return;
1494 		}
1495 		if (m == NULL)
1496 			return;
1497 	}
1498 
1499 #ifdef ALTQ
1500 	KERNEL_LOCK(1, NULL);
1501 	/*
1502 	 * If ALTQ is enabled on the member interface, do
1503 	 * classification; the queueing discipline might
1504 	 * not require classification, but might require
1505 	 * the address family/header pointer in the pktattr.
1506 	 */
1507 	if (ALTQ_IS_ENABLED(&dst_ifp->if_snd)) {
1508 		/* XXX IFT_ETHER */
1509 		altq_etherclassify(&dst_ifp->if_snd, m);
1510 	}
1511 	KERNEL_UNLOCK_ONE(NULL);
1512 #endif /* ALTQ */
1513 
1514 	if (vlan_has_tag(m) &&
1515 	    !vlan_is_hwtag_enabled(dst_ifp)) {
1516 		(void)ether_inject_vlantag(&m, ETHERTYPE_VLAN,
1517 		    vlan_get_tag(m));
1518 		if (m == NULL) {
1519 			if_statinc(&sc->sc_if, if_oerrors);
1520 			return;
1521 		}
1522 	}
1523 
1524 	len = m->m_pkthdr.len;
1525 	mflags = m->m_flags;
1526 
1527 	error = if_transmit_lock(dst_ifp, m);
1528 	if (error) {
1529 		/* mbuf is already freed */
1530 		if_statinc(&sc->sc_if, if_oerrors);
1531 		return;
1532 	}
1533 
1534 	net_stat_ref_t nsr = IF_STAT_GETREF(&sc->sc_if);
1535 	if_statinc_ref(&sc->sc_if, nsr, if_opackets);
1536 	if_statadd_ref(&sc->sc_if, nsr, if_obytes, len);
1537 	if (mflags & M_MCAST)
1538 		if_statinc_ref(&sc->sc_if, nsr, if_omcasts);
1539 	IF_STAT_PUTREF(&sc->sc_if);
1540 }
1541 
1542 /*
1543  * bridge_output:
1544  *
1545  *	Send output from a bridge member interface.  This
1546  *	performs the bridging function for locally originated
1547  *	packets.
1548  *
1549  *	The mbuf has the Ethernet header already attached.  We must
1550  *	enqueue or free the mbuf before returning.
1551  */
1552 int
1553 bridge_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa,
1554     const struct rtentry *rt)
1555 {
1556 	struct ether_header *eh;
1557 	struct ifnet *dst_if;
1558 	struct bridge_softc *sc;
1559 	struct mbuf *n;
1560 	int s, bound;
1561 
1562 	/*
1563 	 * bridge_output() is called from ether_output(), furthermore
1564 	 * ifp argument doesn't point to bridge(4). So, don't assert
1565 	 * IFEF_MPSAFE here.
1566 	 */
1567 
1568 	KASSERT(m->m_len >= ETHER_HDR_LEN);
1569 
1570 	eh = mtod(m, struct ether_header *);
1571 	sc = ifp->if_bridge;
1572 
1573 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1574 		if (memcmp(etherbroadcastaddr,
1575 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
1576 			m->m_flags |= M_BCAST;
1577 		else
1578 			m->m_flags |= M_MCAST;
1579 	}
1580 
1581 	/*
1582 	 * If bridge is down, but the original output interface is up,
1583 	 * go ahead and send out that interface.  Otherwise, the packet
1584 	 * is dropped below.
1585 	 */
1586 	if (__predict_false(sc == NULL) ||
1587 	    (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
1588 		dst_if = ifp;
1589 		goto unicast_asis;
1590 	}
1591 
1592 	/*
1593 	 * If the packet is a multicast, or we don't know a better way to
1594 	 * get there, send to all interfaces.
1595 	 */
1596 	if ((m->m_flags & (M_MCAST | M_BCAST)) != 0)
1597 		dst_if = NULL;
1598 	else
1599 		dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1600 
1601 	/*
1602 	 * In general, we need to handle TX offload in software before
1603 	 * enqueueing a packet. However, we can send it as is in the
1604 	 * cases of unicast via (1) the source interface, or (2) an
1605 	 * interface which supports the specified offload options.
1606 	 * For multicast or broadcast, send it as is only if (3) all
1607 	 * the member interfaces support the specified options.
1608 	 */
1609 
1610 	/*
1611 	 * Unicast via the source interface.
1612 	 */
1613 	if (dst_if == ifp)
1614 		goto unicast_asis;
1615 
1616 	/*
1617 	 * Unicast via other interface.
1618 	 */
1619 	if (dst_if != NULL) {
1620 		KASSERT(m->m_flags & M_PKTHDR);
1621 		if (TX_OFFLOAD_SUPPORTED(dst_if->if_csum_flags_tx,
1622 		    m->m_pkthdr.csum_flags)) {
1623 			/*
1624 			 * Unicast via an interface which supports the
1625 			 * specified offload options.
1626 			 */
1627 			goto unicast_asis;
1628 		}
1629 
1630 		/*
1631 		 * Handle TX offload in software. For TSO, a packet is
1632 		 * split into multiple chunks. Thus, the return value of
1633 		 * ether_sw_offload_tx() is mbuf queue consists of them.
1634 		 */
1635 		m = ether_sw_offload_tx(ifp, m);
1636 		if (m == NULL)
1637 			return 0;
1638 
1639 		do {
1640 			n = m->m_nextpkt;
1641 			if ((dst_if->if_flags & IFF_RUNNING) == 0)
1642 				m_freem(m);
1643 			else
1644 				bridge_enqueue(sc, dst_if, m, 0);
1645 			m = n;
1646 		} while (m != NULL);
1647 
1648 		return 0;
1649 	}
1650 
1651 	/*
1652 	 * Multicast or broadcast.
1653 	 */
1654 	if (TX_OFFLOAD_SUPPORTED(sc->sc_csum_flags_tx,
1655 	    m->m_pkthdr.csum_flags)) {
1656 		/*
1657 		 * Specified TX offload options are supported by all
1658 		 * the member interfaces of this bridge.
1659 		 */
1660 		m->m_nextpkt = NULL;	/* XXX */
1661 	} else {
1662 		/*
1663 		 * Otherwise, handle TX offload in software.
1664 		 */
1665 		m = ether_sw_offload_tx(ifp, m);
1666 		if (m == NULL)
1667 			return 0;
1668 	}
1669 
1670 	/*
1671 	 * When we use pppoe over bridge, bridge_output() can be called
1672 	 * in a lwp context by pppoe_timeout_wk().
1673 	 */
1674 	bound = curlwp_bind();
1675 	do {
1676 		/* XXX Should call bridge_broadcast, but there are locking
1677 		 * issues which need resolving first. */
1678 		struct bridge_iflist *bif;
1679 		struct mbuf *mc;
1680 		bool used = false;
1681 
1682 		n = m->m_nextpkt;
1683 
1684 		s = pserialize_read_enter();
1685 		BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
1686 			struct psref psref;
1687 
1688 			bridge_acquire_member(sc, bif, &psref);
1689 			pserialize_read_exit(s);
1690 
1691 			dst_if = bif->bif_ifp;
1692 			if ((dst_if->if_flags & IFF_RUNNING) == 0)
1693 				goto next;
1694 
1695 			/*
1696 			 * If this is not the original output interface,
1697 			 * and the interface is participating in spanning
1698 			 * tree, make sure the port is in a state that
1699 			 * allows forwarding.
1700 			 */
1701 			if (dst_if != ifp &&
1702 			    (bif->bif_flags & IFBIF_STP) != 0) {
1703 				switch (bif->bif_state) {
1704 				case BSTP_IFSTATE_BLOCKING:
1705 				case BSTP_IFSTATE_LISTENING:
1706 				case BSTP_IFSTATE_DISABLED:
1707 					goto next;
1708 				}
1709 			}
1710 
1711 			if (PSLIST_READER_NEXT(bif, struct bridge_iflist,
1712 			    bif_next) == NULL &&
1713 			    ((m->m_flags & (M_MCAST | M_BCAST)) == 0 ||
1714 			    dst_if == ifp))
1715 			{
1716 				used = true;
1717 				mc = m;
1718 			} else {
1719 				mc = m_copypacket(m, M_DONTWAIT);
1720 				if (mc == NULL) {
1721 					if_statinc(&sc->sc_if, if_oerrors);
1722 					goto next;
1723 				}
1724 			}
1725 
1726 			bridge_enqueue(sc, dst_if, mc, 0);
1727 
1728 			if ((m->m_flags & (M_MCAST | M_BCAST)) != 0 &&
1729 			    dst_if != ifp)
1730 			{
1731 				if (PSLIST_READER_NEXT(bif,
1732 				    struct bridge_iflist, bif_next) == NULL)
1733 				{
1734 					used = true;
1735 					mc = m;
1736 				} else {
1737 					mc = m_copypacket(m, M_DONTWAIT);
1738 					if (mc == NULL) {
1739 						if_statinc(&sc->sc_if,
1740 						    if_oerrors);
1741 						goto next;
1742 					}
1743 				}
1744 
1745 				m_set_rcvif(mc, dst_if);
1746 				mc->m_flags &= ~M_PROMISC;
1747 
1748 				const int _s = splsoftnet();
1749 				KERNEL_LOCK_UNLESS_IFP_MPSAFE(dst_if);
1750 				ether_input(dst_if, mc);
1751 				KERNEL_UNLOCK_UNLESS_IFP_MPSAFE(dst_if);
1752 				splx(_s);
1753 			}
1754 
1755 next:
1756 			s = pserialize_read_enter();
1757 			bridge_release_member(sc, bif, &psref);
1758 
1759 			/* Guarantee we don't re-enter the loop as we already
1760 			 * decided we're at the end. */
1761 			if (used)
1762 				break;
1763 		}
1764 		pserialize_read_exit(s);
1765 
1766 		if (!used)
1767 			m_freem(m);
1768 
1769 		m = n;
1770 	} while (m != NULL);
1771 	curlwp_bindx(bound);
1772 
1773 	return 0;
1774 
1775 unicast_asis:
1776 	/*
1777 	 * XXX Spanning tree consideration here?
1778 	 */
1779 	if ((dst_if->if_flags & IFF_RUNNING) == 0)
1780 		m_freem(m);
1781 	else
1782 		bridge_enqueue(sc, dst_if, m, 0);
1783 	return 0;
1784 }
1785 
1786 /*
1787  * bridge_start:
1788  *
1789  *	Start output on a bridge.
1790  *
1791  *	NOTE: This routine should never be called in this implementation.
1792  */
1793 static void
1794 bridge_start(struct ifnet *ifp)
1795 {
1796 
1797 	printf("%s: bridge_start() called\n", ifp->if_xname);
1798 }
1799 
1800 /*
1801  * bridge_forward:
1802  *
1803  *	The forwarding function of the bridge.
1804  */
1805 static void
1806 bridge_forward(struct bridge_softc *sc, struct mbuf *m)
1807 {
1808 	struct bridge_iflist *bif;
1809 	struct ifnet *src_if, *dst_if;
1810 	struct ether_header *eh;
1811 	struct psref psref;
1812 	struct psref psref_src;
1813 	DECLARE_LOCK_VARIABLE;
1814 	bool src_if_protected;
1815 
1816 	src_if = m_get_rcvif_psref(m, &psref_src);
1817 	if (src_if == NULL) {
1818 		/* Interface is being destroyed? */
1819 		goto discard;
1820 	}
1821 
1822 	if_statadd2(&sc->sc_if, if_ipackets, 1, if_ibytes, m->m_pkthdr.len);
1823 
1824 	/*
1825 	 * Look up the bridge_iflist.
1826 	 */
1827 	bif = bridge_lookup_member_if(sc, src_if, &psref);
1828 	if (bif == NULL) {
1829 		/* Interface is not a bridge member (anymore?) */
1830 		goto discard;
1831 	}
1832 
1833 	if (bif->bif_flags & IFBIF_STP) {
1834 		switch (bif->bif_state) {
1835 		case BSTP_IFSTATE_BLOCKING:
1836 		case BSTP_IFSTATE_LISTENING:
1837 		case BSTP_IFSTATE_DISABLED:
1838 			bridge_release_member(sc, bif, &psref);
1839 			goto discard;
1840 		}
1841 	}
1842 
1843 	eh = mtod(m, struct ether_header *);
1844 
1845 	/*
1846 	 * If the interface is learning, and the source
1847 	 * address is valid and not multicast, record
1848 	 * the address.
1849 	 */
1850 	if ((bif->bif_flags & IFBIF_LEARNING) != 0 &&
1851 	    ETHER_IS_MULTICAST(eh->ether_shost) == 0 &&
1852 	    (eh->ether_shost[0] == 0 &&
1853 	     eh->ether_shost[1] == 0 &&
1854 	     eh->ether_shost[2] == 0 &&
1855 	     eh->ether_shost[3] == 0 &&
1856 	     eh->ether_shost[4] == 0 &&
1857 	     eh->ether_shost[5] == 0) == 0) {
1858 		(void) bridge_rtupdate(sc, eh->ether_shost,
1859 		    src_if, 0, IFBAF_DYNAMIC);
1860 	}
1861 
1862 	if ((bif->bif_flags & IFBIF_STP) != 0 &&
1863 	    bif->bif_state == BSTP_IFSTATE_LEARNING) {
1864 		bridge_release_member(sc, bif, &psref);
1865 		goto discard;
1866 	}
1867 
1868 	src_if_protected = ((bif->bif_flags & IFBIF_PROTECTED) != 0);
1869 
1870 	bridge_release_member(sc, bif, &psref);
1871 
1872 	/*
1873 	 * At this point, the port either doesn't participate
1874 	 * in spanning tree or it is in the forwarding state.
1875 	 */
1876 
1877 	/*
1878 	 * If the packet is unicast, destined for someone on
1879 	 * "this" side of the bridge, drop it.
1880 	 */
1881 	if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
1882 		dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1883 		if (src_if == dst_if)
1884 			goto discard;
1885 	} else {
1886 		/* ...forward it to all interfaces. */
1887 		if_statinc(&sc->sc_if, if_imcasts);
1888 		dst_if = NULL;
1889 	}
1890 
1891 	if (pfil_run_hooks(sc->sc_if.if_pfil, &m, src_if, PFIL_IN) != 0 ||
1892 	    m == NULL) {
1893 		goto discard;
1894 	}
1895 
1896 	if (dst_if == NULL) {
1897 		bridge_broadcast(sc, src_if, src_if_protected, m);
1898 		goto out;
1899 	}
1900 
1901 	m_put_rcvif_psref(src_if, &psref_src);
1902 	src_if = NULL;
1903 
1904 	/*
1905 	 * At this point, we're dealing with a unicast frame
1906 	 * going to a different interface.
1907 	 */
1908 	if ((dst_if->if_flags & IFF_RUNNING) == 0)
1909 		goto discard;
1910 
1911 	bif = bridge_lookup_member_if(sc, dst_if, &psref);
1912 	if (bif == NULL) {
1913 		/* Not a member of the bridge (anymore?) */
1914 		goto discard;
1915 	}
1916 
1917 	if (bif->bif_flags & IFBIF_STP) {
1918 		switch (bif->bif_state) {
1919 		case BSTP_IFSTATE_DISABLED:
1920 		case BSTP_IFSTATE_BLOCKING:
1921 			bridge_release_member(sc, bif, &psref);
1922 			goto discard;
1923 		}
1924 	}
1925 
1926 	if ((bif->bif_flags & IFBIF_PROTECTED) && src_if_protected) {
1927 		bridge_release_member(sc, bif, &psref);
1928 		goto discard;
1929 	}
1930 
1931 	bridge_release_member(sc, bif, &psref);
1932 
1933 	/*
1934 	 * Before enqueueing this packet to the destination interface,
1935 	 * clear any in-bound checksum flags to prevent them from being
1936 	 * misused as out-bound flags.
1937 	 */
1938 	m->m_pkthdr.csum_flags = 0;
1939 
1940 	ACQUIRE_GLOBAL_LOCKS();
1941 	bridge_enqueue(sc, dst_if, m, 1);
1942 	RELEASE_GLOBAL_LOCKS();
1943 out:
1944 	if (src_if != NULL)
1945 		m_put_rcvif_psref(src_if, &psref_src);
1946 	return;
1947 
1948 discard:
1949 	m_freem(m);
1950 	goto out;
1951 }
1952 
1953 static bool
1954 bstp_state_before_learning(struct bridge_iflist *bif)
1955 {
1956 	if (bif->bif_flags & IFBIF_STP) {
1957 		switch (bif->bif_state) {
1958 		case BSTP_IFSTATE_BLOCKING:
1959 		case BSTP_IFSTATE_LISTENING:
1960 		case BSTP_IFSTATE_DISABLED:
1961 			return true;
1962 		}
1963 	}
1964 	return false;
1965 }
1966 
1967 static bool
1968 bridge_ourether(struct bridge_iflist *bif, struct ether_header *eh, int src)
1969 {
1970 	uint8_t *ether = src ? eh->ether_shost : eh->ether_dhost;
1971 
1972 	if (memcmp(CLLADDR(bif->bif_ifp->if_sadl), ether, ETHER_ADDR_LEN) == 0
1973 #if NCARP > 0
1974 	    || (bif->bif_ifp->if_carp &&
1975 	        carp_ourether(bif->bif_ifp->if_carp, eh, IFT_ETHER, src) != NULL)
1976 #endif /* NCARP > 0 */
1977 	    )
1978 		return true;
1979 
1980 	return false;
1981 }
1982 
1983 /*
1984  * bridge_input:
1985  *
1986  *	Receive input from a member interface.  Queue the packet for
1987  *	bridging if it is not for us.
1988  */
1989 static void
1990 bridge_input(struct ifnet *ifp, struct mbuf *m)
1991 {
1992 	struct bridge_softc *sc = ifp->if_bridge;
1993 	struct bridge_iflist *bif;
1994 	struct ether_header *eh;
1995 	struct psref psref;
1996 	int bound;
1997 	DECLARE_LOCK_VARIABLE;
1998 
1999 	KASSERT(!cpu_intr_p());
2000 
2001 	if (__predict_false(sc == NULL) ||
2002 	    (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
2003 		ACQUIRE_GLOBAL_LOCKS();
2004 		ether_input(ifp, m);
2005 		RELEASE_GLOBAL_LOCKS();
2006 		return;
2007 	}
2008 
2009 	bound = curlwp_bind();
2010 	bif = bridge_lookup_member_if(sc, ifp, &psref);
2011 	if (bif == NULL) {
2012 		curlwp_bindx(bound);
2013 		ACQUIRE_GLOBAL_LOCKS();
2014 		ether_input(ifp, m);
2015 		RELEASE_GLOBAL_LOCKS();
2016 		return;
2017 	}
2018 
2019 	eh = mtod(m, struct ether_header *);
2020 
2021 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
2022 		if (memcmp(etherbroadcastaddr,
2023 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
2024 			m->m_flags |= M_BCAST;
2025 		else
2026 			m->m_flags |= M_MCAST;
2027 	}
2028 
2029 	/*
2030 	 * A 'fast' path for packets addressed to interfaces that are
2031 	 * part of this bridge.
2032 	 */
2033 	if (!(m->m_flags & (M_BCAST|M_MCAST)) &&
2034 	    !bstp_state_before_learning(bif)) {
2035 		struct bridge_iflist *_bif;
2036 		struct ifnet *_ifp = NULL;
2037 		int s;
2038 		struct psref _psref;
2039 
2040 		s = pserialize_read_enter();
2041 		BRIDGE_IFLIST_READER_FOREACH(_bif, sc) {
2042 			/* It is destined for us. */
2043 			if (bridge_ourether(_bif, eh, 0)) {
2044 				bridge_acquire_member(sc, _bif, &_psref);
2045 				pserialize_read_exit(s);
2046 				if (_bif->bif_flags & IFBIF_LEARNING)
2047 					(void) bridge_rtupdate(sc,
2048 					    eh->ether_shost, ifp, 0, IFBAF_DYNAMIC);
2049 				m_set_rcvif(m, _bif->bif_ifp);
2050 				_ifp = _bif->bif_ifp;
2051 				bridge_release_member(sc, _bif, &_psref);
2052 				goto out;
2053 			}
2054 
2055 			/* We just received a packet that we sent out. */
2056 			if (bridge_ourether(_bif, eh, 1))
2057 				break;
2058 		}
2059 		pserialize_read_exit(s);
2060 out:
2061 
2062 		if (_bif != NULL) {
2063 			bridge_release_member(sc, bif, &psref);
2064 			curlwp_bindx(bound);
2065 			if (_ifp != NULL) {
2066 				m->m_flags &= ~M_PROMISC;
2067 				ACQUIRE_GLOBAL_LOCKS();
2068 				ether_input(_ifp, m);
2069 				RELEASE_GLOBAL_LOCKS();
2070 			} else
2071 				m_freem(m);
2072 			return;
2073 		}
2074 	}
2075 
2076 	/* Tap off 802.1D packets; they do not get forwarded. */
2077 	if (bif->bif_flags & IFBIF_STP &&
2078 	    memcmp(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN) == 0) {
2079 		bstp_input(sc, bif, m);
2080 		bridge_release_member(sc, bif, &psref);
2081 		curlwp_bindx(bound);
2082 		return;
2083 	}
2084 
2085 	/*
2086 	 * A normal switch would discard the packet here, but that's not what
2087 	 * we've done historically. This also prevents some obnoxious behaviour.
2088 	 */
2089 	if (bstp_state_before_learning(bif)) {
2090 		bridge_release_member(sc, bif, &psref);
2091 		curlwp_bindx(bound);
2092 		ACQUIRE_GLOBAL_LOCKS();
2093 		ether_input(ifp, m);
2094 		RELEASE_GLOBAL_LOCKS();
2095 		return;
2096 	}
2097 
2098 	bridge_release_member(sc, bif, &psref);
2099 
2100 	bridge_forward(sc, m);
2101 
2102 	curlwp_bindx(bound);
2103 }
2104 
2105 /*
2106  * bridge_broadcast:
2107  *
2108  *	Send a frame to all interfaces that are members of
2109  *	the bridge, except for the one on which the packet
2110  *	arrived.
2111  */
2112 static void
2113 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2114     bool src_if_protected, struct mbuf *m)
2115 {
2116 	struct bridge_iflist *bif;
2117 	struct mbuf *mc;
2118 	struct ifnet *dst_if;
2119 	bool bmcast;
2120 	int s;
2121 	DECLARE_LOCK_VARIABLE;
2122 
2123 	bmcast = m->m_flags & (M_BCAST|M_MCAST);
2124 
2125 	s = pserialize_read_enter();
2126 	BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
2127 		struct psref psref;
2128 
2129 		bridge_acquire_member(sc, bif, &psref);
2130 		pserialize_read_exit(s);
2131 
2132 		dst_if = bif->bif_ifp;
2133 
2134 		if (bif->bif_flags & IFBIF_STP) {
2135 			switch (bif->bif_state) {
2136 			case BSTP_IFSTATE_BLOCKING:
2137 			case BSTP_IFSTATE_DISABLED:
2138 				goto next;
2139 			}
2140 		}
2141 
2142 		if ((bif->bif_flags & IFBIF_DISCOVER) == 0 && !bmcast)
2143 			goto next;
2144 
2145 		if ((dst_if->if_flags & IFF_RUNNING) == 0)
2146 			goto next;
2147 
2148 		if (dst_if != src_if) {
2149 			if ((bif->bif_flags & IFBIF_PROTECTED) &&
2150 			    src_if_protected) {
2151 				goto next;
2152 			}
2153 
2154 			mc = m_copypacket(m, M_DONTWAIT);
2155 			if (mc == NULL) {
2156 				if_statinc(&sc->sc_if, if_oerrors);
2157 				goto next;
2158 			}
2159 			/*
2160 			 * Before enqueueing this packet to the destination
2161 			 * interface, clear any in-bound checksum flags to
2162 			 * prevent them from being misused as out-bound flags.
2163 			 */
2164 			mc->m_pkthdr.csum_flags = 0;
2165 
2166 			ACQUIRE_GLOBAL_LOCKS();
2167 			bridge_enqueue(sc, dst_if, mc, 1);
2168 			RELEASE_GLOBAL_LOCKS();
2169 		}
2170 
2171 		if (bmcast) {
2172 			mc = m_copypacket(m, M_DONTWAIT);
2173 			if (mc == NULL) {
2174 				if_statinc(&sc->sc_if, if_oerrors);
2175 				goto next;
2176 			}
2177 			/*
2178 			 * Before enqueueing this packet to the destination
2179 			 * interface, clear any in-bound checksum flags to
2180 			 * prevent them from being misused as out-bound flags.
2181 			 */
2182 			mc->m_pkthdr.csum_flags = 0;
2183 
2184 			m_set_rcvif(mc, dst_if);
2185 			mc->m_flags &= ~M_PROMISC;
2186 
2187 			ACQUIRE_GLOBAL_LOCKS();
2188 			ether_input(dst_if, mc);
2189 			RELEASE_GLOBAL_LOCKS();
2190 		}
2191 next:
2192 		s = pserialize_read_enter();
2193 		bridge_release_member(sc, bif, &psref);
2194 	}
2195 	pserialize_read_exit(s);
2196 
2197 	m_freem(m);
2198 }
2199 
2200 static int
2201 bridge_rtalloc(struct bridge_softc *sc, const uint8_t *dst,
2202     struct bridge_rtnode **brtp)
2203 {
2204 	struct bridge_rtnode *brt;
2205 	int error;
2206 
2207 	if (sc->sc_brtcnt >= sc->sc_brtmax)
2208 		return ENOSPC;
2209 
2210 	/*
2211 	 * Allocate a new bridge forwarding node, and
2212 	 * initialize the expiration time and Ethernet
2213 	 * address.
2214 	 */
2215 	brt = pool_get(&bridge_rtnode_pool, PR_NOWAIT);
2216 	if (brt == NULL)
2217 		return ENOMEM;
2218 
2219 	memset(brt, 0, sizeof(*brt));
2220 	brt->brt_expire = time_uptime + sc->sc_brttimeout;
2221 	brt->brt_flags = IFBAF_DYNAMIC;
2222 	memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2223 	PSLIST_ENTRY_INIT(brt, brt_list);
2224 	PSLIST_ENTRY_INIT(brt, brt_hash);
2225 
2226 	BRIDGE_RT_LOCK(sc);
2227 	error = bridge_rtnode_insert(sc, brt);
2228 	BRIDGE_RT_UNLOCK(sc);
2229 
2230 	if (error != 0) {
2231 		pool_put(&bridge_rtnode_pool, brt);
2232 		return error;
2233 	}
2234 
2235 	*brtp = brt;
2236 	return 0;
2237 }
2238 
2239 /*
2240  * bridge_rtupdate:
2241  *
2242  *	Add a bridge routing entry.
2243  */
2244 static int
2245 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
2246     struct ifnet *dst_if, int setflags, uint8_t flags)
2247 {
2248 	struct bridge_rtnode *brt;
2249 	int s;
2250 
2251 again:
2252 	/*
2253 	 * A route for this destination might already exist.  If so,
2254 	 * update it, otherwise create a new one.
2255 	 */
2256 	s = pserialize_read_enter();
2257 	brt = bridge_rtnode_lookup(sc, dst);
2258 
2259 	if (brt != NULL) {
2260 		brt->brt_ifp = dst_if;
2261 		if (setflags) {
2262 			brt->brt_flags = flags;
2263 			if (flags & IFBAF_STATIC)
2264 				brt->brt_expire = 0;
2265 			else
2266 				brt->brt_expire = time_uptime + sc->sc_brttimeout;
2267 		} else {
2268 			if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2269 				brt->brt_expire = time_uptime + sc->sc_brttimeout;
2270 		}
2271 	}
2272 	pserialize_read_exit(s);
2273 
2274 	if (brt == NULL) {
2275 		int r;
2276 
2277 		r = bridge_rtalloc(sc, dst, &brt);
2278 		if (r != 0)
2279 			return r;
2280 		goto again;
2281 	}
2282 
2283 	return 0;
2284 }
2285 
2286 /*
2287  * bridge_rtlookup:
2288  *
2289  *	Lookup the destination interface for an address.
2290  */
2291 static struct ifnet *
2292 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr)
2293 {
2294 	struct bridge_rtnode *brt;
2295 	struct ifnet *ifs = NULL;
2296 	int s;
2297 
2298 	s = pserialize_read_enter();
2299 	brt = bridge_rtnode_lookup(sc, addr);
2300 	if (brt != NULL)
2301 		ifs = brt->brt_ifp;
2302 	pserialize_read_exit(s);
2303 
2304 	return ifs;
2305 }
2306 
2307 typedef bool (*bridge_iterate_cb_t)
2308     (struct bridge_softc *, struct bridge_rtnode *, bool *, void *);
2309 
2310 /*
2311  * bridge_rtlist_iterate_remove:
2312  *
2313  *	It iterates on sc->sc_rtlist and removes rtnodes of it which func
2314  *	callback judges to remove. Removals of rtnodes are done in a manner
2315  *	of pserialize. To this end, all kmem_* operations are placed out of
2316  *	mutexes.
2317  */
2318 static void
2319 bridge_rtlist_iterate_remove(struct bridge_softc *sc, bridge_iterate_cb_t func, void *arg)
2320 {
2321 	struct bridge_rtnode *brt;
2322 	struct bridge_rtnode **brt_list;
2323 	int i, count;
2324 
2325 retry:
2326 	count = sc->sc_brtcnt;
2327 	if (count == 0)
2328 		return;
2329 	brt_list = kmem_alloc(sizeof(*brt_list) * count, KM_SLEEP);
2330 
2331 	BRIDGE_RT_LOCK(sc);
2332 	if (__predict_false(sc->sc_brtcnt > count)) {
2333 		/* The rtnodes increased, we need more memory */
2334 		BRIDGE_RT_UNLOCK(sc);
2335 		kmem_free(brt_list, sizeof(*brt_list) * count);
2336 		goto retry;
2337 	}
2338 
2339 	i = 0;
2340 	/*
2341 	 * We don't need to use a _SAFE variant here because we know
2342 	 * that a removed item keeps its next pointer as-is thanks to
2343 	 * pslist(9) and isn't freed in the loop.
2344 	 */
2345 	BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) {
2346 		bool need_break = false;
2347 		if (func(sc, brt, &need_break, arg)) {
2348 			bridge_rtnode_remove(sc, brt);
2349 			brt_list[i++] = brt;
2350 		}
2351 		if (need_break)
2352 			break;
2353 	}
2354 
2355 	if (i > 0)
2356 		BRIDGE_RT_PSZ_PERFORM(sc);
2357 	BRIDGE_RT_UNLOCK(sc);
2358 
2359 	while (--i >= 0)
2360 		bridge_rtnode_destroy(brt_list[i]);
2361 
2362 	kmem_free(brt_list, sizeof(*brt_list) * count);
2363 }
2364 
2365 static bool
2366 bridge_rttrim0_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
2367     bool *need_break, void *arg)
2368 {
2369 	if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2370 		/* Take into account of the subsequent removal */
2371 		if ((sc->sc_brtcnt - 1) <= sc->sc_brtmax)
2372 			*need_break = true;
2373 		return true;
2374 	} else
2375 		return false;
2376 }
2377 
2378 static void
2379 bridge_rttrim0(struct bridge_softc *sc)
2380 {
2381 	bridge_rtlist_iterate_remove(sc, bridge_rttrim0_cb, NULL);
2382 }
2383 
2384 /*
2385  * bridge_rttrim:
2386  *
2387  *	Trim the routine table so that we have a number
2388  *	of routing entries less than or equal to the
2389  *	maximum number.
2390  */
2391 static void
2392 bridge_rttrim(struct bridge_softc *sc)
2393 {
2394 
2395 	/* Make sure we actually need to do this. */
2396 	if (sc->sc_brtcnt <= sc->sc_brtmax)
2397 		return;
2398 
2399 	/* Force an aging cycle; this might trim enough addresses. */
2400 	bridge_rtage(sc);
2401 	if (sc->sc_brtcnt <= sc->sc_brtmax)
2402 		return;
2403 
2404 	bridge_rttrim0(sc);
2405 
2406 	return;
2407 }
2408 
2409 /*
2410  * bridge_timer:
2411  *
2412  *	Aging timer for the bridge.
2413  */
2414 static void
2415 bridge_timer(void *arg)
2416 {
2417 	struct bridge_softc *sc = arg;
2418 
2419 	workqueue_enqueue(sc->sc_rtage_wq, &sc->sc_rtage_wk, NULL);
2420 }
2421 
2422 static void
2423 bridge_rtage_work(struct work *wk, void *arg)
2424 {
2425 	struct bridge_softc *sc = arg;
2426 
2427 	KASSERT(wk == &sc->sc_rtage_wk);
2428 
2429 	bridge_rtage(sc);
2430 
2431 	if (sc->sc_if.if_flags & IFF_RUNNING)
2432 		callout_reset(&sc->sc_brcallout,
2433 		    bridge_rtable_prune_period * hz, bridge_timer, sc);
2434 }
2435 
2436 static bool
2437 bridge_rtage_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
2438     bool *need_break, void *arg)
2439 {
2440 	if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2441 	    time_uptime >= brt->brt_expire)
2442 		return true;
2443 	else
2444 		return false;
2445 }
2446 
2447 /*
2448  * bridge_rtage:
2449  *
2450  *	Perform an aging cycle.
2451  */
2452 static void
2453 bridge_rtage(struct bridge_softc *sc)
2454 {
2455 	bridge_rtlist_iterate_remove(sc, bridge_rtage_cb, NULL);
2456 }
2457 
2458 
2459 static bool
2460 bridge_rtflush_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
2461     bool *need_break, void *arg)
2462 {
2463 	int full = *(int*)arg;
2464 
2465 	if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2466 		return true;
2467 	else
2468 		return false;
2469 }
2470 
2471 /*
2472  * bridge_rtflush:
2473  *
2474  *	Remove all dynamic addresses from the bridge.
2475  */
2476 static void
2477 bridge_rtflush(struct bridge_softc *sc, int full)
2478 {
2479 	bridge_rtlist_iterate_remove(sc, bridge_rtflush_cb, &full);
2480 }
2481 
2482 /*
2483  * bridge_rtdaddr:
2484  *
2485  *	Remove an address from the table.
2486  */
2487 static int
2488 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr)
2489 {
2490 	struct bridge_rtnode *brt;
2491 
2492 	BRIDGE_RT_LOCK(sc);
2493 	if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL) {
2494 		BRIDGE_RT_UNLOCK(sc);
2495 		return ENOENT;
2496 	}
2497 	bridge_rtnode_remove(sc, brt);
2498 	BRIDGE_RT_PSZ_PERFORM(sc);
2499 	BRIDGE_RT_UNLOCK(sc);
2500 
2501 	bridge_rtnode_destroy(brt);
2502 
2503 	return 0;
2504 }
2505 
2506 /*
2507  * bridge_rtdelete:
2508  *
2509  *	Delete routes to a speicifc member interface.
2510  */
2511 static void
2512 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp)
2513 {
2514 	struct bridge_rtnode *brt;
2515 
2516 	/* XXX pserialize_perform for each entry is slow */
2517 again:
2518 	BRIDGE_RT_LOCK(sc);
2519 	BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) {
2520 		if (brt->brt_ifp == ifp)
2521 			break;
2522 	}
2523 	if (brt == NULL) {
2524 		BRIDGE_RT_UNLOCK(sc);
2525 		return;
2526 	}
2527 	bridge_rtnode_remove(sc, brt);
2528 	BRIDGE_RT_PSZ_PERFORM(sc);
2529 	BRIDGE_RT_UNLOCK(sc);
2530 
2531 	bridge_rtnode_destroy(brt);
2532 
2533 	goto again;
2534 }
2535 
2536 /*
2537  * bridge_rtable_init:
2538  *
2539  *	Initialize the route table for this bridge.
2540  */
2541 static void
2542 bridge_rtable_init(struct bridge_softc *sc)
2543 {
2544 	int i;
2545 
2546 	sc->sc_rthash = kmem_alloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2547 	    KM_SLEEP);
2548 
2549 	for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2550 		PSLIST_INIT(&sc->sc_rthash[i]);
2551 
2552 	sc->sc_rthash_key = cprng_fast32();
2553 
2554 	PSLIST_INIT(&sc->sc_rtlist);
2555 
2556 	sc->sc_rtlist_psz = pserialize_create();
2557 	sc->sc_rtlist_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_SOFTNET);
2558 }
2559 
2560 /*
2561  * bridge_rtable_fini:
2562  *
2563  *	Deconstruct the route table for this bridge.
2564  */
2565 static void
2566 bridge_rtable_fini(struct bridge_softc *sc)
2567 {
2568 
2569 	kmem_free(sc->sc_rthash, sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE);
2570 	mutex_obj_free(sc->sc_rtlist_lock);
2571 	pserialize_destroy(sc->sc_rtlist_psz);
2572 }
2573 
2574 /*
2575  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2576  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2577  */
2578 #define	mix(a, b, c)							\
2579 do {									\
2580 	a -= b; a -= c; a ^= (c >> 13);					\
2581 	b -= c; b -= a; b ^= (a << 8);					\
2582 	c -= a; c -= b; c ^= (b >> 13);					\
2583 	a -= b; a -= c; a ^= (c >> 12);					\
2584 	b -= c; b -= a; b ^= (a << 16);					\
2585 	c -= a; c -= b; c ^= (b >> 5);					\
2586 	a -= b; a -= c; a ^= (c >> 3);					\
2587 	b -= c; b -= a; b ^= (a << 10);					\
2588 	c -= a; c -= b; c ^= (b >> 15);					\
2589 } while (/*CONSTCOND*/0)
2590 
2591 static inline uint32_t
2592 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2593 {
2594 	uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2595 
2596 	b += addr[5] << 8;
2597 	b += addr[4];
2598 	a += (uint32_t)addr[3] << 24;
2599 	a += addr[2] << 16;
2600 	a += addr[1] << 8;
2601 	a += addr[0];
2602 
2603 	mix(a, b, c);
2604 
2605 	return (c & BRIDGE_RTHASH_MASK);
2606 }
2607 
2608 #undef mix
2609 
2610 /*
2611  * bridge_rtnode_lookup:
2612  *
2613  *	Look up a bridge route node for the specified destination.
2614  */
2615 static struct bridge_rtnode *
2616 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr)
2617 {
2618 	struct bridge_rtnode *brt;
2619 	uint32_t hash;
2620 	int dir;
2621 
2622 	hash = bridge_rthash(sc, addr);
2623 	BRIDGE_RTHASH_READER_FOREACH(brt, sc, hash) {
2624 		dir = memcmp(addr, brt->brt_addr, ETHER_ADDR_LEN);
2625 		if (dir == 0)
2626 			return brt;
2627 		if (dir > 0)
2628 			return NULL;
2629 	}
2630 
2631 	return NULL;
2632 }
2633 
2634 /*
2635  * bridge_rtnode_insert:
2636  *
2637  *	Insert the specified bridge node into the route table.  We
2638  *	assume the entry is not already in the table.
2639  */
2640 static int
2641 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2642 {
2643 	struct bridge_rtnode *lbrt, *prev = NULL;
2644 	uint32_t hash;
2645 
2646 	KASSERT(BRIDGE_RT_LOCKED(sc));
2647 
2648 	hash = bridge_rthash(sc, brt->brt_addr);
2649 	BRIDGE_RTHASH_WRITER_FOREACH(lbrt, sc, hash) {
2650 		int dir = memcmp(brt->brt_addr, lbrt->brt_addr, ETHER_ADDR_LEN);
2651 		if (dir == 0)
2652 			return EEXIST;
2653 		if (dir > 0)
2654 			break;
2655 		prev = lbrt;
2656 	}
2657 	if (prev == NULL)
2658 		BRIDGE_RTHASH_WRITER_INSERT_HEAD(sc, hash, brt);
2659 	else
2660 		BRIDGE_RTHASH_WRITER_INSERT_AFTER(prev, brt);
2661 
2662 	BRIDGE_RTLIST_WRITER_INSERT_HEAD(sc, brt);
2663 	sc->sc_brtcnt++;
2664 
2665 	return 0;
2666 }
2667 
2668 /*
2669  * bridge_rtnode_remove:
2670  *
2671  *	Remove a bridge rtnode from the rthash and the rtlist of a bridge.
2672  */
2673 static void
2674 bridge_rtnode_remove(struct bridge_softc *sc, struct bridge_rtnode *brt)
2675 {
2676 
2677 	KASSERT(BRIDGE_RT_LOCKED(sc));
2678 
2679 	BRIDGE_RTHASH_WRITER_REMOVE(brt);
2680 	BRIDGE_RTLIST_WRITER_REMOVE(brt);
2681 	sc->sc_brtcnt--;
2682 }
2683 
2684 /*
2685  * bridge_rtnode_destroy:
2686  *
2687  *	Destroy a bridge rtnode.
2688  */
2689 static void
2690 bridge_rtnode_destroy(struct bridge_rtnode *brt)
2691 {
2692 
2693 	PSLIST_ENTRY_DESTROY(brt, brt_list);
2694 	PSLIST_ENTRY_DESTROY(brt, brt_hash);
2695 	pool_put(&bridge_rtnode_pool, brt);
2696 }
2697 
2698 extern pfil_head_t *inet_pfil_hook;                 /* XXX */
2699 extern pfil_head_t *inet6_pfil_hook;                /* XXX */
2700 
2701 /*
2702  * Send bridge packets through IPF if they are one of the types IPF can deal
2703  * with, or if they are ARP or REVARP.  (IPF will pass ARP and REVARP without
2704  * question.)
2705  */
2706 static int
2707 bridge_ipf(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
2708 {
2709 	int snap, error;
2710 	struct ether_header *eh1, eh2;
2711 	struct llc llc1;
2712 	uint16_t ether_type;
2713 
2714 	snap = 0;
2715 	error = -1;	/* Default error if not error == 0 */
2716 	eh1 = mtod(*mp, struct ether_header *);
2717 	ether_type = ntohs(eh1->ether_type);
2718 
2719 	/*
2720 	 * Check for SNAP/LLC.
2721 	 */
2722 	if (ether_type < ETHERMTU) {
2723 		struct llc *llc2 = (struct llc *)(eh1 + 1);
2724 
2725 		if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
2726 		    llc2->llc_dsap == LLC_SNAP_LSAP &&
2727 		    llc2->llc_ssap == LLC_SNAP_LSAP &&
2728 		    llc2->llc_control == LLC_UI) {
2729 			ether_type = htons(llc2->llc_un.type_snap.ether_type);
2730 			snap = 1;
2731 		}
2732 	}
2733 
2734 	/* drop VLAN traffic untagged by hardware offloading */
2735 	if (vlan_has_tag(*mp))
2736 		goto bad;
2737 
2738 	/*
2739 	 * If we're trying to filter bridge traffic, don't look at anything
2740 	 * other than IP and ARP traffic.  If the filter doesn't understand
2741 	 * IPv6, don't allow IPv6 through the bridge either.  This is lame
2742 	 * since if we really wanted, say, an AppleTalk filter, we are hosed,
2743 	 * but of course we don't have an AppleTalk filter to begin with.
2744 	 * (Note that since IPF doesn't understand ARP it will pass *ALL*
2745 	 * ARP traffic.)
2746 	 */
2747 	switch (ether_type) {
2748 		case ETHERTYPE_ARP:
2749 		case ETHERTYPE_REVARP:
2750 			return 0; /* Automatically pass */
2751 		case ETHERTYPE_IP:
2752 # ifdef INET6
2753 		case ETHERTYPE_IPV6:
2754 # endif /* INET6 */
2755 			break;
2756 		default:
2757 			goto bad;
2758 	}
2759 
2760 	/* Strip off the Ethernet header and keep a copy. */
2761 	m_copydata(*mp, 0, ETHER_HDR_LEN, (void *) &eh2);
2762 	m_adj(*mp, ETHER_HDR_LEN);
2763 
2764 	/* Strip off snap header, if present */
2765 	if (snap) {
2766 		m_copydata(*mp, 0, sizeof(struct llc), (void *) &llc1);
2767 		m_adj(*mp, sizeof(struct llc));
2768 	}
2769 
2770 	/*
2771 	 * Check basic packet sanity and run IPF through pfil.
2772 	 */
2773 	KASSERT(!cpu_intr_p());
2774 	switch (ether_type)
2775 	{
2776 	case ETHERTYPE_IP :
2777 		error = bridge_ip_checkbasic(mp);
2778 		if (error == 0)
2779 			error = pfil_run_hooks(inet_pfil_hook, mp, ifp, dir);
2780 		break;
2781 # ifdef INET6
2782 	case ETHERTYPE_IPV6 :
2783 		error = bridge_ip6_checkbasic(mp);
2784 		if (error == 0)
2785 			error = pfil_run_hooks(inet6_pfil_hook, mp, ifp, dir);
2786 		break;
2787 # endif
2788 	default :
2789 		error = 0;
2790 		break;
2791 	}
2792 
2793 	if (*mp == NULL)
2794 		return error;
2795 	if (error != 0)
2796 		goto bad;
2797 
2798 	error = -1;
2799 
2800 	/*
2801 	 * Finally, put everything back the way it was and return
2802 	 */
2803 	if (snap) {
2804 		M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
2805 		if (*mp == NULL)
2806 			return error;
2807 		bcopy(&llc1, mtod(*mp, void *), sizeof(struct llc));
2808 	}
2809 
2810 	M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
2811 	if (*mp == NULL)
2812 		return error;
2813 	bcopy(&eh2, mtod(*mp, void *), ETHER_HDR_LEN);
2814 
2815 	return 0;
2816 
2817     bad:
2818 	m_freem(*mp);
2819 	*mp = NULL;
2820 	return error;
2821 }
2822 
2823 /*
2824  * Perform basic checks on header size since
2825  * IPF assumes ip_input has already processed
2826  * it for it.  Cut-and-pasted from ip_input.c.
2827  * Given how simple the IPv6 version is,
2828  * does the IPv4 version really need to be
2829  * this complicated?
2830  *
2831  * XXX Should we update ipstat here, or not?
2832  * XXX Right now we update ipstat but not
2833  * XXX csum_counter.
2834  */
2835 static int
2836 bridge_ip_checkbasic(struct mbuf **mp)
2837 {
2838 	struct mbuf *m = *mp;
2839 	struct ip *ip;
2840 	int len, hlen;
2841 
2842 	if (*mp == NULL)
2843 		return -1;
2844 
2845 	if (M_GET_ALIGNED_HDR(&m, struct ip, true) != 0) {
2846 		/* XXXJRT new stat, please */
2847 		ip_statinc(IP_STAT_TOOSMALL);
2848 		goto bad;
2849 	}
2850 	ip = mtod(m, struct ip *);
2851 	if (ip == NULL) goto bad;
2852 
2853 	if (ip->ip_v != IPVERSION) {
2854 		ip_statinc(IP_STAT_BADVERS);
2855 		goto bad;
2856 	}
2857 	hlen = ip->ip_hl << 2;
2858 	if (hlen < sizeof(struct ip)) { /* minimum header length */
2859 		ip_statinc(IP_STAT_BADHLEN);
2860 		goto bad;
2861 	}
2862 	if (hlen > m->m_len) {
2863 		if ((m = m_pullup(m, hlen)) == 0) {
2864 			ip_statinc(IP_STAT_BADHLEN);
2865 			goto bad;
2866 		}
2867 		ip = mtod(m, struct ip *);
2868 		if (ip == NULL) goto bad;
2869 	}
2870 
2871 	switch (m->m_pkthdr.csum_flags &
2872 	        ((m_get_rcvif_NOMPSAFE(m)->if_csum_flags_rx & M_CSUM_IPv4) |
2873 	         M_CSUM_IPv4_BAD)) {
2874 	case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
2875 		/* INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); */
2876 		goto bad;
2877 
2878 	case M_CSUM_IPv4:
2879 		/* Checksum was okay. */
2880 		/* INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); */
2881 		break;
2882 
2883 	default:
2884 		/* Must compute it ourselves. */
2885 		/* INET_CSUM_COUNTER_INCR(&ip_swcsum); */
2886 		if (in_cksum(m, hlen) != 0)
2887 			goto bad;
2888 		break;
2889 	}
2890 
2891 	/* Retrieve the packet length. */
2892 	len = ntohs(ip->ip_len);
2893 
2894 	/*
2895 	 * Check for additional length bogosity
2896 	 */
2897 	if (len < hlen) {
2898 		ip_statinc(IP_STAT_BADLEN);
2899 		goto bad;
2900 	}
2901 
2902 	/*
2903 	 * Check that the amount of data in the buffers
2904 	 * is as at least much as the IP header would have us expect.
2905 	 * Drop packet if shorter than we expect.
2906 	 */
2907 	if (m->m_pkthdr.len < len) {
2908 		ip_statinc(IP_STAT_TOOSHORT);
2909 		goto bad;
2910 	}
2911 
2912 	/* Checks out, proceed */
2913 	*mp = m;
2914 	return 0;
2915 
2916     bad:
2917 	*mp = m;
2918 	return -1;
2919 }
2920 
2921 # ifdef INET6
2922 /*
2923  * Same as above, but for IPv6.
2924  * Cut-and-pasted from ip6_input.c.
2925  * XXX Should we update ip6stat, or not?
2926  */
2927 static int
2928 bridge_ip6_checkbasic(struct mbuf **mp)
2929 {
2930 	struct mbuf *m = *mp;
2931 	struct ip6_hdr *ip6;
2932 
2933 	/*
2934 	 * If the IPv6 header is not aligned, slurp it up into a new
2935 	 * mbuf with space for link headers, in the event we forward
2936 	 * it.  Otherwise, if it is aligned, make sure the entire base
2937 	 * IPv6 header is in the first mbuf of the chain.
2938 	 */
2939 	if (M_GET_ALIGNED_HDR(&m, struct ip6_hdr, true) != 0) {
2940 		struct ifnet *inifp = m_get_rcvif_NOMPSAFE(m);
2941 		/* XXXJRT new stat, please */
2942 		ip6_statinc(IP6_STAT_TOOSMALL);
2943 		in6_ifstat_inc(inifp, ifs6_in_hdrerr);
2944 		goto bad;
2945 	}
2946 
2947 	ip6 = mtod(m, struct ip6_hdr *);
2948 
2949 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
2950 		ip6_statinc(IP6_STAT_BADVERS);
2951 		in6_ifstat_inc(m_get_rcvif_NOMPSAFE(m), ifs6_in_hdrerr);
2952 		goto bad;
2953 	}
2954 
2955 	/* Checks out, proceed */
2956 	*mp = m;
2957 	return 0;
2958 
2959     bad:
2960 	*mp = m;
2961 	return -1;
2962 }
2963 # endif /* INET6 */
2964