xref: /netbsd-src/sys/net/if_bridge.c (revision bdc22b2e01993381dcefeff2bc9b56ca75a4235c)
1 /*	$NetBSD: if_bridge.c,v 1.156 2018/05/25 04:40:27 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 1.156 2018/05/25 04:40:27 ozaki-r Exp $");
84 
85 #ifdef _KERNEL_OPT
86 #include "opt_bridge_ipf.h"
87 #include "opt_inet.h"
88 #include "opt_net_mpsafe.h"
89 #endif /* _KERNEL_OPT */
90 
91 #include <sys/param.h>
92 #include <sys/kernel.h>
93 #include <sys/mbuf.h>
94 #include <sys/queue.h>
95 #include <sys/socket.h>
96 #include <sys/socketvar.h> /* for softnet_lock */
97 #include <sys/sockio.h>
98 #include <sys/systm.h>
99 #include <sys/proc.h>
100 #include <sys/pool.h>
101 #include <sys/kauth.h>
102 #include <sys/cpu.h>
103 #include <sys/cprng.h>
104 #include <sys/mutex.h>
105 #include <sys/kmem.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 
116 #if defined(BRIDGE_IPF)
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 
124 #include <netinet/ip6.h>
125 #include <netinet6/in6_var.h>
126 #include <netinet6/ip6_var.h>
127 #include <netinet6/ip6_private.h>	/* XXX */
128 #endif /* BRIDGE_IPF */
129 
130 /*
131  * Size of the route hash table.  Must be a power of two.
132  */
133 #ifndef BRIDGE_RTHASH_SIZE
134 #define	BRIDGE_RTHASH_SIZE		1024
135 #endif
136 
137 #define	BRIDGE_RTHASH_MASK		(BRIDGE_RTHASH_SIZE - 1)
138 
139 #include "carp.h"
140 #if NCARP > 0
141 #include <netinet/in.h>
142 #include <netinet/in_var.h>
143 #include <netinet/ip_carp.h>
144 #endif
145 
146 #include "ioconf.h"
147 
148 __CTASSERT(sizeof(struct ifbifconf) == sizeof(struct ifbaconf));
149 __CTASSERT(offsetof(struct ifbifconf, ifbic_len) == offsetof(struct ifbaconf, ifbac_len));
150 __CTASSERT(offsetof(struct ifbifconf, ifbic_buf) == offsetof(struct ifbaconf, ifbac_buf));
151 
152 /*
153  * Maximum number of addresses to cache.
154  */
155 #ifndef BRIDGE_RTABLE_MAX
156 #define	BRIDGE_RTABLE_MAX		100
157 #endif
158 
159 /*
160  * Spanning tree defaults.
161  */
162 #define	BSTP_DEFAULT_MAX_AGE		(20 * 256)
163 #define	BSTP_DEFAULT_HELLO_TIME		(2 * 256)
164 #define	BSTP_DEFAULT_FORWARD_DELAY	(15 * 256)
165 #define	BSTP_DEFAULT_HOLD_TIME		(1 * 256)
166 #define	BSTP_DEFAULT_BRIDGE_PRIORITY	0x8000
167 #define	BSTP_DEFAULT_PORT_PRIORITY	0x80
168 #define	BSTP_DEFAULT_PATH_COST		55
169 
170 /*
171  * Timeout (in seconds) for entries learned dynamically.
172  */
173 #ifndef BRIDGE_RTABLE_TIMEOUT
174 #define	BRIDGE_RTABLE_TIMEOUT		(20 * 60)	/* same as ARP */
175 #endif
176 
177 /*
178  * Number of seconds between walks of the route list.
179  */
180 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
181 #define	BRIDGE_RTABLE_PRUNE_PERIOD	(5 * 60)
182 #endif
183 
184 #define BRIDGE_RT_LOCK(_sc)	mutex_enter((_sc)->sc_rtlist_lock)
185 #define BRIDGE_RT_UNLOCK(_sc)	mutex_exit((_sc)->sc_rtlist_lock)
186 #define BRIDGE_RT_LOCKED(_sc)	mutex_owned((_sc)->sc_rtlist_lock)
187 
188 #define BRIDGE_RT_PSZ_PERFORM(_sc) \
189 				pserialize_perform((_sc)->sc_rtlist_psz)
190 
191 #define BRIDGE_RT_RENTER(__s)	do { __s = pserialize_read_enter(); } while (0)
192 #define BRIDGE_RT_REXIT(__s)	do { pserialize_read_exit(__s); } while (0)
193 
194 #define BRIDGE_RTLIST_READER_FOREACH(_brt, _sc)			\
195 	PSLIST_READER_FOREACH((_brt), &((_sc)->sc_rtlist),		\
196 	    struct bridge_rtnode, brt_list)
197 #define BRIDGE_RTLIST_WRITER_FOREACH(_brt, _sc)			\
198 	PSLIST_WRITER_FOREACH((_brt), &((_sc)->sc_rtlist),		\
199 	    struct bridge_rtnode, brt_list)
200 #define BRIDGE_RTLIST_WRITER_INSERT_HEAD(_sc, _brt)			\
201 	PSLIST_WRITER_INSERT_HEAD(&(_sc)->sc_rtlist, brt, brt_list)
202 #define BRIDGE_RTLIST_WRITER_REMOVE(_brt)				\
203 	PSLIST_WRITER_REMOVE((_brt), brt_list)
204 
205 #define BRIDGE_RTHASH_READER_FOREACH(_brt, _sc, _hash)			\
206 	PSLIST_READER_FOREACH((_brt), &(_sc)->sc_rthash[(_hash)],	\
207 	    struct bridge_rtnode, brt_hash)
208 #define BRIDGE_RTHASH_WRITER_FOREACH(_brt, _sc, _hash)			\
209 	PSLIST_WRITER_FOREACH((_brt), &(_sc)->sc_rthash[(_hash)],	\
210 	    struct bridge_rtnode, brt_hash)
211 #define BRIDGE_RTHASH_WRITER_INSERT_HEAD(_sc, _hash, _brt)		\
212 	PSLIST_WRITER_INSERT_HEAD(&(_sc)->sc_rthash[(_hash)], brt, brt_hash)
213 #define BRIDGE_RTHASH_WRITER_INSERT_AFTER(_brt, _new)			\
214 	PSLIST_WRITER_INSERT_AFTER((_brt), (_new), brt_hash)
215 #define BRIDGE_RTHASH_WRITER_REMOVE(_brt)				\
216 	PSLIST_WRITER_REMOVE((_brt), brt_hash)
217 
218 #ifdef NET_MPSAFE
219 #define DECLARE_LOCK_VARIABLE
220 #define ACQUIRE_GLOBAL_LOCKS()	do { } while (0)
221 #define RELEASE_GLOBAL_LOCKS()	do { } while (0)
222 #else
223 #define DECLARE_LOCK_VARIABLE	int __s
224 #define ACQUIRE_GLOBAL_LOCKS()	do {					\
225 					KERNEL_LOCK(1, NULL);		\
226 					mutex_enter(softnet_lock);	\
227 					__s = splsoftnet();		\
228 				} while (0)
229 #define RELEASE_GLOBAL_LOCKS()	do {					\
230 					splx(__s);			\
231 					mutex_exit(softnet_lock);	\
232 					KERNEL_UNLOCK_ONE(NULL);	\
233 				} while (0)
234 #endif
235 
236 struct psref_class *bridge_psref_class __read_mostly;
237 
238 int	bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
239 
240 static struct pool bridge_rtnode_pool;
241 
242 static int	bridge_clone_create(struct if_clone *, int);
243 static int	bridge_clone_destroy(struct ifnet *);
244 
245 static int	bridge_ioctl(struct ifnet *, u_long, void *);
246 static int	bridge_init(struct ifnet *);
247 static void	bridge_stop(struct ifnet *, int);
248 static void	bridge_start(struct ifnet *);
249 
250 static void	bridge_input(struct ifnet *, struct mbuf *);
251 static void	bridge_forward(struct bridge_softc *, struct mbuf *);
252 
253 static void	bridge_timer(void *);
254 
255 static void	bridge_broadcast(struct bridge_softc *, struct ifnet *,
256 				 struct mbuf *);
257 
258 static int	bridge_rtupdate(struct bridge_softc *, const uint8_t *,
259 				struct ifnet *, int, uint8_t);
260 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *);
261 static void	bridge_rttrim(struct bridge_softc *);
262 static void	bridge_rtage(struct bridge_softc *);
263 static void	bridge_rtage_work(struct work *, void *);
264 static void	bridge_rtflush(struct bridge_softc *, int);
265 static int	bridge_rtdaddr(struct bridge_softc *, const uint8_t *);
266 static void	bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp);
267 
268 static void	bridge_rtable_init(struct bridge_softc *);
269 static void	bridge_rtable_fini(struct bridge_softc *);
270 
271 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
272 						  const uint8_t *);
273 static int	bridge_rtnode_insert(struct bridge_softc *,
274 				     struct bridge_rtnode *);
275 static void	bridge_rtnode_remove(struct bridge_softc *,
276 				     struct bridge_rtnode *);
277 static void	bridge_rtnode_destroy(struct bridge_rtnode *);
278 
279 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
280 						  const char *name,
281 						  struct psref *);
282 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
283 						     struct ifnet *ifp,
284 						     struct psref *);
285 static void	bridge_release_member(struct bridge_softc *, struct bridge_iflist *,
286                                       struct psref *);
287 static void	bridge_delete_member(struct bridge_softc *,
288 				     struct bridge_iflist *);
289 static void	bridge_acquire_member(struct bridge_softc *sc,
290                                       struct bridge_iflist *,
291                                       struct psref *);
292 
293 static int	bridge_ioctl_add(struct bridge_softc *, void *);
294 static int	bridge_ioctl_del(struct bridge_softc *, void *);
295 static int	bridge_ioctl_gifflags(struct bridge_softc *, void *);
296 static int	bridge_ioctl_sifflags(struct bridge_softc *, void *);
297 static int	bridge_ioctl_scache(struct bridge_softc *, void *);
298 static int	bridge_ioctl_gcache(struct bridge_softc *, void *);
299 static int	bridge_ioctl_gifs(struct bridge_softc *, void *);
300 static int	bridge_ioctl_rts(struct bridge_softc *, void *);
301 static int	bridge_ioctl_saddr(struct bridge_softc *, void *);
302 static int	bridge_ioctl_sto(struct bridge_softc *, void *);
303 static int	bridge_ioctl_gto(struct bridge_softc *, void *);
304 static int	bridge_ioctl_daddr(struct bridge_softc *, void *);
305 static int	bridge_ioctl_flush(struct bridge_softc *, void *);
306 static int	bridge_ioctl_gpri(struct bridge_softc *, void *);
307 static int	bridge_ioctl_spri(struct bridge_softc *, void *);
308 static int	bridge_ioctl_ght(struct bridge_softc *, void *);
309 static int	bridge_ioctl_sht(struct bridge_softc *, void *);
310 static int	bridge_ioctl_gfd(struct bridge_softc *, void *);
311 static int	bridge_ioctl_sfd(struct bridge_softc *, void *);
312 static int	bridge_ioctl_gma(struct bridge_softc *, void *);
313 static int	bridge_ioctl_sma(struct bridge_softc *, void *);
314 static int	bridge_ioctl_sifprio(struct bridge_softc *, void *);
315 static int	bridge_ioctl_sifcost(struct bridge_softc *, void *);
316 #if defined(BRIDGE_IPF)
317 static int	bridge_ioctl_gfilt(struct bridge_softc *, void *);
318 static int	bridge_ioctl_sfilt(struct bridge_softc *, void *);
319 static int	bridge_ipf(void *, struct mbuf **, struct ifnet *, int);
320 static int	bridge_ip_checkbasic(struct mbuf **mp);
321 # ifdef INET6
322 static int	bridge_ip6_checkbasic(struct mbuf **mp);
323 # endif /* INET6 */
324 #endif /* BRIDGE_IPF */
325 
326 struct bridge_control {
327 	int	(*bc_func)(struct bridge_softc *, void *);
328 	int	bc_argsize;
329 	int	bc_flags;
330 };
331 
332 #define	BC_F_COPYIN		0x01	/* copy arguments in */
333 #define	BC_F_COPYOUT		0x02	/* copy arguments out */
334 #define	BC_F_SUSER		0x04	/* do super-user check */
335 #define BC_F_XLATEIN		0x08	/* xlate arguments in */
336 #define BC_F_XLATEOUT		0x10	/* xlate arguments out */
337 
338 static const struct bridge_control bridge_control_table[] = {
339 [BRDGADD] = {bridge_ioctl_add, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
340 [BRDGDEL] = {bridge_ioctl_del, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
341 
342 [BRDGGIFFLGS] = {bridge_ioctl_gifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_COPYOUT},
343 [BRDGSIFFLGS] = {bridge_ioctl_sifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
344 
345 [BRDGSCACHE] = {bridge_ioctl_scache, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
346 [BRDGGCACHE] = {bridge_ioctl_gcache, sizeof(struct ifbrparam), BC_F_COPYOUT},
347 
348 [OBRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_COPYIN|BC_F_COPYOUT},
349 [OBRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_COPYIN|BC_F_COPYOUT},
350 
351 [BRDGSADDR] = {bridge_ioctl_saddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
352 
353 [BRDGSTO] = {bridge_ioctl_sto, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
354 [BRDGGTO] = {bridge_ioctl_gto, sizeof(struct ifbrparam), BC_F_COPYOUT},
355 
356 [BRDGDADDR] = {bridge_ioctl_daddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
357 
358 [BRDGFLUSH] = {bridge_ioctl_flush, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
359 
360 [BRDGGPRI] = {bridge_ioctl_gpri, sizeof(struct ifbrparam), BC_F_COPYOUT},
361 [BRDGSPRI] = {bridge_ioctl_spri, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
362 
363 [BRDGGHT] = {bridge_ioctl_ght, sizeof(struct ifbrparam), BC_F_COPYOUT},
364 [BRDGSHT] = {bridge_ioctl_sht, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
365 
366 [BRDGGFD] = {bridge_ioctl_gfd, sizeof(struct ifbrparam), BC_F_COPYOUT},
367 [BRDGSFD] = {bridge_ioctl_sfd, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
368 
369 [BRDGGMA] = {bridge_ioctl_gma, sizeof(struct ifbrparam), BC_F_COPYOUT},
370 [BRDGSMA] = {bridge_ioctl_sma, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
371 
372 [BRDGSIFPRIO] = {bridge_ioctl_sifprio, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
373 
374 [BRDGSIFCOST] = {bridge_ioctl_sifcost, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
375 #if defined(BRIDGE_IPF)
376 [BRDGGFILT] = {bridge_ioctl_gfilt, sizeof(struct ifbrparam), BC_F_COPYOUT},
377 [BRDGSFILT] = {bridge_ioctl_sfilt, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
378 #endif /* BRIDGE_IPF */
379 [BRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_XLATEIN|BC_F_XLATEOUT},
380 [BRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_XLATEIN|BC_F_XLATEOUT},
381 };
382 
383 static const int bridge_control_table_size = __arraycount(bridge_control_table);
384 
385 static struct if_clone bridge_cloner =
386     IF_CLONE_INITIALIZER("bridge", bridge_clone_create, bridge_clone_destroy);
387 
388 /*
389  * bridgeattach:
390  *
391  *	Pseudo-device attach routine.
392  */
393 void
394 bridgeattach(int n)
395 {
396 
397 	pool_init(&bridge_rtnode_pool, sizeof(struct bridge_rtnode),
398 	    0, 0, 0, "brtpl", NULL, IPL_NET);
399 
400 	bridge_psref_class = psref_class_create("bridge", IPL_SOFTNET);
401 
402 	if_clone_attach(&bridge_cloner);
403 }
404 
405 /*
406  * bridge_clone_create:
407  *
408  *	Create a new bridge instance.
409  */
410 static int
411 bridge_clone_create(struct if_clone *ifc, int unit)
412 {
413 	struct bridge_softc *sc;
414 	struct ifnet *ifp;
415 	int error;
416 
417 	sc = kmem_zalloc(sizeof(*sc),  KM_SLEEP);
418 	ifp = &sc->sc_if;
419 
420 	sc->sc_brtmax = BRIDGE_RTABLE_MAX;
421 	sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
422 	sc->sc_bridge_max_age = BSTP_DEFAULT_MAX_AGE;
423 	sc->sc_bridge_hello_time = BSTP_DEFAULT_HELLO_TIME;
424 	sc->sc_bridge_forward_delay = BSTP_DEFAULT_FORWARD_DELAY;
425 	sc->sc_bridge_priority = BSTP_DEFAULT_BRIDGE_PRIORITY;
426 	sc->sc_hold_time = BSTP_DEFAULT_HOLD_TIME;
427 	sc->sc_filter_flags = 0;
428 
429 	/* Initialize our routing table. */
430 	bridge_rtable_init(sc);
431 
432 	error = workqueue_create(&sc->sc_rtage_wq, "bridge_rtage",
433 	    bridge_rtage_work, sc, PRI_SOFTNET, IPL_SOFTNET, WQ_MPSAFE);
434 	if (error)
435 		panic("%s: workqueue_create %d\n", __func__, error);
436 
437 	callout_init(&sc->sc_brcallout, CALLOUT_MPSAFE);
438 	callout_init(&sc->sc_bstpcallout, CALLOUT_MPSAFE);
439 
440 	mutex_init(&sc->sc_iflist_psref.bip_lock, MUTEX_DEFAULT, IPL_NONE);
441 	PSLIST_INIT(&sc->sc_iflist_psref.bip_iflist);
442 	sc->sc_iflist_psref.bip_psz = pserialize_create();
443 
444 	if_initname(ifp, ifc->ifc_name, unit);
445 	ifp->if_softc = sc;
446 	ifp->if_extflags = IFEF_NO_LINK_STATE_CHANGE;
447 #ifdef NET_MPSAFE
448 	ifp->if_extflags |= IFEF_MPSAFE;
449 #endif
450 	ifp->if_mtu = ETHERMTU;
451 	ifp->if_ioctl = bridge_ioctl;
452 	ifp->if_output = bridge_output;
453 	ifp->if_start = bridge_start;
454 	ifp->if_stop = bridge_stop;
455 	ifp->if_init = bridge_init;
456 	ifp->if_type = IFT_BRIDGE;
457 	ifp->if_addrlen = 0;
458 	ifp->if_dlt = DLT_EN10MB;
459 	ifp->if_hdrlen = ETHER_HDR_LEN;
460 
461 	error = if_initialize(ifp);
462 	if (error != 0) {
463 		pserialize_destroy(sc->sc_iflist_psref.bip_psz);
464 		mutex_destroy(&sc->sc_iflist_psref.bip_lock);
465 		callout_destroy(&sc->sc_brcallout);
466 		callout_destroy(&sc->sc_bstpcallout);
467 		workqueue_destroy(sc->sc_rtage_wq);
468 		bridge_rtable_fini(sc);
469 		kmem_free(sc, sizeof(*sc));
470 
471 		return error;
472 	}
473 	if_alloc_sadl(ifp);
474 	if_register(ifp);
475 
476 	return 0;
477 }
478 
479 /*
480  * bridge_clone_destroy:
481  *
482  *	Destroy a bridge instance.
483  */
484 static int
485 bridge_clone_destroy(struct ifnet *ifp)
486 {
487 	struct bridge_softc *sc = ifp->if_softc;
488 	struct bridge_iflist *bif;
489 
490 	if ((ifp->if_flags & IFF_RUNNING) != 0)
491 		bridge_stop(ifp, 1);
492 
493 	BRIDGE_LOCK(sc);
494 	for (;;) {
495 		bif = PSLIST_WRITER_FIRST(&sc->sc_iflist_psref.bip_iflist, struct bridge_iflist,
496 		    bif_next);
497 		if (bif == NULL)
498 			break;
499 		bridge_delete_member(sc, bif);
500 	}
501 	PSLIST_DESTROY(&sc->sc_iflist_psref.bip_iflist);
502 	BRIDGE_UNLOCK(sc);
503 
504 	if_detach(ifp);
505 
506 	/* Tear down the routing table. */
507 	bridge_rtable_fini(sc);
508 
509 	pserialize_destroy(sc->sc_iflist_psref.bip_psz);
510 	mutex_destroy(&sc->sc_iflist_psref.bip_lock);
511 	callout_destroy(&sc->sc_brcallout);
512 	callout_destroy(&sc->sc_bstpcallout);
513 	workqueue_destroy(sc->sc_rtage_wq);
514 	kmem_free(sc, sizeof(*sc));
515 
516 	return 0;
517 }
518 
519 /*
520  * bridge_ioctl:
521  *
522  *	Handle a control request from the operator.
523  */
524 static int
525 bridge_ioctl(struct ifnet *ifp, u_long cmd, void *data)
526 {
527 	struct bridge_softc *sc = ifp->if_softc;
528 	struct lwp *l = curlwp;	/* XXX */
529 	union {
530 		struct ifbreq ifbreq;
531 		struct ifbifconf ifbifconf;
532 		struct ifbareq ifbareq;
533 		struct ifbaconf ifbaconf;
534 		struct ifbrparam ifbrparam;
535 	} args;
536 	struct ifdrv *ifd = (struct ifdrv *) data;
537 	const struct bridge_control *bc = NULL; /* XXXGCC */
538 	int s, error = 0;
539 
540 	/* Authorize command before calling splsoftnet(). */
541 	switch (cmd) {
542 	case SIOCGDRVSPEC:
543 	case SIOCSDRVSPEC:
544 		if (ifd->ifd_cmd >= bridge_control_table_size
545 		    || (bc = &bridge_control_table[ifd->ifd_cmd]) == NULL) {
546 			error = EINVAL;
547 			return error;
548 		}
549 
550 		/* We only care about BC_F_SUSER at this point. */
551 		if ((bc->bc_flags & BC_F_SUSER) == 0)
552 			break;
553 
554 		error = kauth_authorize_network(l->l_cred,
555 		    KAUTH_NETWORK_INTERFACE_BRIDGE,
556 		    cmd == SIOCGDRVSPEC ?
557 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_GETPRIV :
558 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_SETPRIV,
559 		     ifd, NULL, NULL);
560 		if (error)
561 			return error;
562 
563 		break;
564 	}
565 
566 	s = splsoftnet();
567 
568 	switch (cmd) {
569 	case SIOCGDRVSPEC:
570 	case SIOCSDRVSPEC:
571 		KASSERT(bc != NULL);
572 		if (cmd == SIOCGDRVSPEC &&
573 		    (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) == 0) {
574 			error = EINVAL;
575 			break;
576 		}
577 		else if (cmd == SIOCSDRVSPEC &&
578 		    (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) != 0) {
579 			error = EINVAL;
580 			break;
581 		}
582 
583 		/* BC_F_SUSER is checked above, before splsoftnet(). */
584 
585 		if ((bc->bc_flags & (BC_F_XLATEIN|BC_F_XLATEOUT)) == 0
586 		    && (ifd->ifd_len != bc->bc_argsize
587 			|| ifd->ifd_len > sizeof(args))) {
588 			error = EINVAL;
589 			break;
590 		}
591 
592 		memset(&args, 0, sizeof(args));
593 		if (bc->bc_flags & BC_F_COPYIN) {
594 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
595 			if (error)
596 				break;
597 		} else if (bc->bc_flags & BC_F_XLATEIN) {
598 			args.ifbifconf.ifbic_len = ifd->ifd_len;
599 			args.ifbifconf.ifbic_buf = ifd->ifd_data;
600 		}
601 
602 		error = (*bc->bc_func)(sc, &args);
603 		if (error)
604 			break;
605 
606 		if (bc->bc_flags & BC_F_COPYOUT) {
607 			error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
608 		} else if (bc->bc_flags & BC_F_XLATEOUT) {
609 			ifd->ifd_len = args.ifbifconf.ifbic_len;
610 			ifd->ifd_data = args.ifbifconf.ifbic_buf;
611 		}
612 		break;
613 
614 	case SIOCSIFFLAGS:
615 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
616 			break;
617 		switch (ifp->if_flags & (IFF_UP|IFF_RUNNING)) {
618 		case IFF_RUNNING:
619 			/*
620 			 * If interface is marked down and it is running,
621 			 * then stop and disable it.
622 			 */
623 			(*ifp->if_stop)(ifp, 1);
624 			break;
625 		case IFF_UP:
626 			/*
627 			 * If interface is marked up and it is stopped, then
628 			 * start it.
629 			 */
630 			error = (*ifp->if_init)(ifp);
631 			break;
632 		default:
633 			break;
634 		}
635 		break;
636 
637 	case SIOCSIFMTU:
638 		if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET)
639 			error = 0;
640 		break;
641 
642 	default:
643 		error = ifioctl_common(ifp, cmd, data);
644 		break;
645 	}
646 
647 	splx(s);
648 
649 	return error;
650 }
651 
652 /*
653  * bridge_lookup_member:
654  *
655  *	Lookup a bridge member interface.
656  */
657 static struct bridge_iflist *
658 bridge_lookup_member(struct bridge_softc *sc, const char *name, struct psref *psref)
659 {
660 	struct bridge_iflist *bif;
661 	struct ifnet *ifp;
662 	int s;
663 
664 	BRIDGE_PSZ_RENTER(s);
665 
666 	BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
667 		ifp = bif->bif_ifp;
668 		if (strcmp(ifp->if_xname, name) == 0)
669 			break;
670 	}
671 	if (bif != NULL)
672 		bridge_acquire_member(sc, bif, psref);
673 
674 	BRIDGE_PSZ_REXIT(s);
675 
676 	return bif;
677 }
678 
679 /*
680  * bridge_lookup_member_if:
681  *
682  *	Lookup a bridge member interface by ifnet*.
683  */
684 static struct bridge_iflist *
685 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp,
686     struct psref *psref)
687 {
688 	struct bridge_iflist *bif;
689 	int s;
690 
691 	BRIDGE_PSZ_RENTER(s);
692 
693 	bif = member_ifp->if_bridgeif;
694 	if (bif != NULL) {
695 		psref_acquire(psref, &bif->bif_psref,
696 		    bridge_psref_class);
697 	}
698 
699 	BRIDGE_PSZ_REXIT(s);
700 
701 	return bif;
702 }
703 
704 static void
705 bridge_acquire_member(struct bridge_softc *sc, struct bridge_iflist *bif,
706     struct psref *psref)
707 {
708 
709 	psref_acquire(psref, &bif->bif_psref, bridge_psref_class);
710 }
711 
712 /*
713  * bridge_release_member:
714  *
715  *	Release the specified member interface.
716  */
717 static void
718 bridge_release_member(struct bridge_softc *sc, struct bridge_iflist *bif,
719     struct psref *psref)
720 {
721 
722 	psref_release(psref, &bif->bif_psref, bridge_psref_class);
723 }
724 
725 /*
726  * bridge_delete_member:
727  *
728  *	Delete the specified member interface.
729  */
730 static void
731 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif)
732 {
733 	struct ifnet *ifs = bif->bif_ifp;
734 
735 	KASSERT(BRIDGE_LOCKED(sc));
736 
737 	ifs->_if_input = ether_input;
738 	ifs->if_bridge = NULL;
739 	ifs->if_bridgeif = NULL;
740 
741 	PSLIST_WRITER_REMOVE(bif, bif_next);
742 	BRIDGE_PSZ_PERFORM(sc);
743 	BRIDGE_UNLOCK(sc);
744 
745 	psref_target_destroy(&bif->bif_psref, bridge_psref_class);
746 
747 	PSLIST_ENTRY_DESTROY(bif, bif_next);
748 	kmem_free(bif, sizeof(*bif));
749 
750 	BRIDGE_LOCK(sc);
751 }
752 
753 static int
754 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
755 {
756 	struct ifbreq *req = arg;
757 	struct bridge_iflist *bif = NULL;
758 	struct ifnet *ifs;
759 	int error = 0;
760 	struct psref psref;
761 
762 	ifs = if_get(req->ifbr_ifsname, &psref);
763 	if (ifs == NULL)
764 		return ENOENT;
765 
766 	if (ifs->if_bridge == sc) {
767 		error = EEXIST;
768 		goto out;
769 	}
770 
771 	if (ifs->if_bridge != NULL) {
772 		error = EBUSY;
773 		goto out;
774 	}
775 
776 	if (ifs->_if_input != ether_input) {
777 		error = EINVAL;
778 		goto out;
779 	}
780 
781 	/* FIXME: doesn't work with non-IFF_SIMPLEX interfaces */
782 	if ((ifs->if_flags & IFF_SIMPLEX) == 0) {
783 		error = EINVAL;
784 		goto out;
785 	}
786 
787 	bif = kmem_alloc(sizeof(*bif), KM_SLEEP);
788 
789 	switch (ifs->if_type) {
790 	case IFT_ETHER:
791 		if (sc->sc_if.if_mtu != ifs->if_mtu) {
792 			error = EINVAL;
793 			goto out;
794 		}
795 		/* FALLTHROUGH */
796 	case IFT_L2TP:
797 		IFNET_LOCK(ifs);
798 		error = ether_enable_vlan_mtu(ifs);
799 		IFNET_UNLOCK(ifs);
800 		if (error > 0)
801 			goto out;
802 		/*
803 		 * Place the interface into promiscuous mode.
804 		 */
805 		error = ifpromisc(ifs, 1);
806 		if (error)
807 			goto out;
808 		break;
809 	default:
810 		error = EINVAL;
811 		goto out;
812 	}
813 
814 	bif->bif_ifp = ifs;
815 	bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
816 	bif->bif_priority = BSTP_DEFAULT_PORT_PRIORITY;
817 	bif->bif_path_cost = BSTP_DEFAULT_PATH_COST;
818 	PSLIST_ENTRY_INIT(bif, bif_next);
819 	psref_target_init(&bif->bif_psref, bridge_psref_class);
820 
821 	BRIDGE_LOCK(sc);
822 
823 	ifs->if_bridge = sc;
824 	ifs->if_bridgeif = bif;
825 	PSLIST_WRITER_INSERT_HEAD(&sc->sc_iflist_psref.bip_iflist, bif, bif_next);
826 	ifs->_if_input = bridge_input;
827 
828 	BRIDGE_UNLOCK(sc);
829 
830 	if (sc->sc_if.if_flags & IFF_RUNNING)
831 		bstp_initialization(sc);
832 	else
833 		bstp_stop(sc);
834 
835  out:
836 	if_put(ifs, &psref);
837 	if (error) {
838 		if (bif != NULL)
839 			kmem_free(bif, sizeof(*bif));
840 	}
841 	return error;
842 }
843 
844 static int
845 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
846 {
847 	struct ifbreq *req = arg;
848 	const char *name = req->ifbr_ifsname;
849 	struct bridge_iflist *bif;
850 	struct ifnet *ifs;
851 
852 	BRIDGE_LOCK(sc);
853 
854 	/*
855 	 * Don't use bridge_lookup_member. We want to get a member
856 	 * with bif_refs == 0.
857 	 */
858 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) {
859 		ifs = bif->bif_ifp;
860 		if (strcmp(ifs->if_xname, name) == 0)
861 			break;
862 	}
863 
864 	if (bif == NULL) {
865 		BRIDGE_UNLOCK(sc);
866 		return ENOENT;
867 	}
868 
869 	bridge_delete_member(sc, bif);
870 
871 	BRIDGE_UNLOCK(sc);
872 
873 	switch (ifs->if_type) {
874 	case IFT_ETHER:
875 	case IFT_L2TP:
876 		/*
877 		 * Take the interface out of promiscuous mode.
878 		 * Don't call it with holding a spin lock.
879 		 */
880 		(void) ifpromisc(ifs, 0);
881 		IFNET_LOCK(ifs);
882 		(void) ether_disable_vlan_mtu(ifs);
883 		IFNET_UNLOCK(ifs);
884 		break;
885 	default:
886 #ifdef DIAGNOSTIC
887 		panic("bridge_delete_member: impossible");
888 #endif
889 		break;
890 	}
891 
892 	bridge_rtdelete(sc, ifs);
893 
894 	if (sc->sc_if.if_flags & IFF_RUNNING)
895 		bstp_initialization(sc);
896 
897 	return 0;
898 }
899 
900 static int
901 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
902 {
903 	struct ifbreq *req = arg;
904 	struct bridge_iflist *bif;
905 	struct psref psref;
906 
907 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
908 	if (bif == NULL)
909 		return ENOENT;
910 
911 	req->ifbr_ifsflags = bif->bif_flags;
912 	req->ifbr_state = bif->bif_state;
913 	req->ifbr_priority = bif->bif_priority;
914 	req->ifbr_path_cost = bif->bif_path_cost;
915 	req->ifbr_portno = bif->bif_ifp->if_index & 0xff;
916 
917 	bridge_release_member(sc, bif, &psref);
918 
919 	return 0;
920 }
921 
922 static int
923 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
924 {
925 	struct ifbreq *req = arg;
926 	struct bridge_iflist *bif;
927 	struct psref psref;
928 
929 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
930 	if (bif == NULL)
931 		return ENOENT;
932 
933 	if (req->ifbr_ifsflags & IFBIF_STP) {
934 		switch (bif->bif_ifp->if_type) {
935 		case IFT_ETHER:
936 		case IFT_L2TP:
937 			/* These can do spanning tree. */
938 			break;
939 
940 		default:
941 			/* Nothing else can. */
942 			bridge_release_member(sc, bif, &psref);
943 			return EINVAL;
944 		}
945 	}
946 
947 	bif->bif_flags = req->ifbr_ifsflags;
948 
949 	bridge_release_member(sc, bif, &psref);
950 
951 	if (sc->sc_if.if_flags & IFF_RUNNING)
952 		bstp_initialization(sc);
953 
954 	return 0;
955 }
956 
957 static int
958 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
959 {
960 	struct ifbrparam *param = arg;
961 
962 	sc->sc_brtmax = param->ifbrp_csize;
963 	bridge_rttrim(sc);
964 
965 	return 0;
966 }
967 
968 static int
969 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
970 {
971 	struct ifbrparam *param = arg;
972 
973 	param->ifbrp_csize = sc->sc_brtmax;
974 
975 	return 0;
976 }
977 
978 static int
979 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
980 {
981 	struct ifbifconf *bifc = arg;
982 	struct bridge_iflist *bif;
983 	struct ifbreq *breqs;
984 	int i, count, error = 0;
985 
986 retry:
987 	BRIDGE_LOCK(sc);
988 	count = 0;
989 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc)
990 		count++;
991 	BRIDGE_UNLOCK(sc);
992 
993 	if (count == 0) {
994 		bifc->ifbic_len = 0;
995 		return 0;
996 	}
997 
998 	if (bifc->ifbic_len == 0 || bifc->ifbic_len < (sizeof(*breqs) * count)) {
999 		/* Tell that a larger buffer is needed */
1000 		bifc->ifbic_len = sizeof(*breqs) * count;
1001 		return 0;
1002 	}
1003 
1004 	breqs = kmem_alloc(sizeof(*breqs) * count, KM_SLEEP);
1005 
1006 	BRIDGE_LOCK(sc);
1007 
1008 	i = 0;
1009 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc)
1010 		i++;
1011 	if (i > count) {
1012 		/*
1013 		 * The number of members has been increased.
1014 		 * We need more memory!
1015 		 */
1016 		BRIDGE_UNLOCK(sc);
1017 		kmem_free(breqs, sizeof(*breqs) * count);
1018 		goto retry;
1019 	}
1020 
1021 	i = 0;
1022 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) {
1023 		struct ifbreq *breq = &breqs[i++];
1024 		memset(breq, 0, sizeof(*breq));
1025 
1026 		strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname,
1027 		    sizeof(breq->ifbr_ifsname));
1028 		breq->ifbr_ifsflags = bif->bif_flags;
1029 		breq->ifbr_state = bif->bif_state;
1030 		breq->ifbr_priority = bif->bif_priority;
1031 		breq->ifbr_path_cost = bif->bif_path_cost;
1032 		breq->ifbr_portno = bif->bif_ifp->if_index & 0xff;
1033 	}
1034 
1035 	/* Don't call copyout with holding the mutex */
1036 	BRIDGE_UNLOCK(sc);
1037 
1038 	for (i = 0; i < count; i++) {
1039 		error = copyout(&breqs[i], bifc->ifbic_req + i, sizeof(*breqs));
1040 		if (error)
1041 			break;
1042 	}
1043 	bifc->ifbic_len = sizeof(*breqs) * i;
1044 
1045 	kmem_free(breqs, sizeof(*breqs) * count);
1046 
1047 	return error;
1048 }
1049 
1050 static int
1051 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1052 {
1053 	struct ifbaconf *bac = arg;
1054 	struct bridge_rtnode *brt;
1055 	struct ifbareq bareq;
1056 	int count = 0, error = 0, len;
1057 
1058 	if (bac->ifbac_len == 0)
1059 		return 0;
1060 
1061 	BRIDGE_RT_LOCK(sc);
1062 
1063 	len = bac->ifbac_len;
1064 	BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) {
1065 		if (len < sizeof(bareq))
1066 			goto out;
1067 		memset(&bareq, 0, sizeof(bareq));
1068 		strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1069 		    sizeof(bareq.ifba_ifsname));
1070 		memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1071 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
1072 			bareq.ifba_expire = brt->brt_expire - time_uptime;
1073 		} else
1074 			bareq.ifba_expire = 0;
1075 		bareq.ifba_flags = brt->brt_flags;
1076 
1077 		error = copyout(&bareq, bac->ifbac_req + count, sizeof(bareq));
1078 		if (error)
1079 			goto out;
1080 		count++;
1081 		len -= sizeof(bareq);
1082 	}
1083  out:
1084 	BRIDGE_RT_UNLOCK(sc);
1085 
1086 	bac->ifbac_len = sizeof(bareq) * count;
1087 	return error;
1088 }
1089 
1090 static int
1091 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1092 {
1093 	struct ifbareq *req = arg;
1094 	struct bridge_iflist *bif;
1095 	int error;
1096 	struct psref psref;
1097 
1098 	bif = bridge_lookup_member(sc, req->ifba_ifsname, &psref);
1099 	if (bif == NULL)
1100 		return ENOENT;
1101 
1102 	error = bridge_rtupdate(sc, req->ifba_dst, bif->bif_ifp, 1,
1103 	    req->ifba_flags);
1104 
1105 	bridge_release_member(sc, bif, &psref);
1106 
1107 	return error;
1108 }
1109 
1110 static int
1111 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1112 {
1113 	struct ifbrparam *param = arg;
1114 
1115 	sc->sc_brttimeout = param->ifbrp_ctime;
1116 
1117 	return 0;
1118 }
1119 
1120 static int
1121 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1122 {
1123 	struct ifbrparam *param = arg;
1124 
1125 	param->ifbrp_ctime = sc->sc_brttimeout;
1126 
1127 	return 0;
1128 }
1129 
1130 static int
1131 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1132 {
1133 	struct ifbareq *req = arg;
1134 
1135 	return (bridge_rtdaddr(sc, req->ifba_dst));
1136 }
1137 
1138 static int
1139 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1140 {
1141 	struct ifbreq *req = arg;
1142 
1143 	bridge_rtflush(sc, req->ifbr_ifsflags);
1144 
1145 	return 0;
1146 }
1147 
1148 static int
1149 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1150 {
1151 	struct ifbrparam *param = arg;
1152 
1153 	param->ifbrp_prio = sc->sc_bridge_priority;
1154 
1155 	return 0;
1156 }
1157 
1158 static int
1159 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1160 {
1161 	struct ifbrparam *param = arg;
1162 
1163 	sc->sc_bridge_priority = param->ifbrp_prio;
1164 
1165 	if (sc->sc_if.if_flags & IFF_RUNNING)
1166 		bstp_initialization(sc);
1167 
1168 	return 0;
1169 }
1170 
1171 static int
1172 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1173 {
1174 	struct ifbrparam *param = arg;
1175 
1176 	param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8;
1177 
1178 	return 0;
1179 }
1180 
1181 static int
1182 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1183 {
1184 	struct ifbrparam *param = arg;
1185 
1186 	if (param->ifbrp_hellotime == 0)
1187 		return EINVAL;
1188 	sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8;
1189 
1190 	if (sc->sc_if.if_flags & IFF_RUNNING)
1191 		bstp_initialization(sc);
1192 
1193 	return 0;
1194 }
1195 
1196 static int
1197 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1198 {
1199 	struct ifbrparam *param = arg;
1200 
1201 	param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8;
1202 
1203 	return 0;
1204 }
1205 
1206 static int
1207 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1208 {
1209 	struct ifbrparam *param = arg;
1210 
1211 	if (param->ifbrp_fwddelay == 0)
1212 		return EINVAL;
1213 	sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8;
1214 
1215 	if (sc->sc_if.if_flags & IFF_RUNNING)
1216 		bstp_initialization(sc);
1217 
1218 	return 0;
1219 }
1220 
1221 static int
1222 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1223 {
1224 	struct ifbrparam *param = arg;
1225 
1226 	param->ifbrp_maxage = sc->sc_bridge_max_age >> 8;
1227 
1228 	return 0;
1229 }
1230 
1231 static int
1232 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1233 {
1234 	struct ifbrparam *param = arg;
1235 
1236 	if (param->ifbrp_maxage == 0)
1237 		return EINVAL;
1238 	sc->sc_bridge_max_age = param->ifbrp_maxage << 8;
1239 
1240 	if (sc->sc_if.if_flags & IFF_RUNNING)
1241 		bstp_initialization(sc);
1242 
1243 	return 0;
1244 }
1245 
1246 static int
1247 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1248 {
1249 	struct ifbreq *req = arg;
1250 	struct bridge_iflist *bif;
1251 	struct psref psref;
1252 
1253 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
1254 	if (bif == NULL)
1255 		return ENOENT;
1256 
1257 	bif->bif_priority = req->ifbr_priority;
1258 
1259 	if (sc->sc_if.if_flags & IFF_RUNNING)
1260 		bstp_initialization(sc);
1261 
1262 	bridge_release_member(sc, bif, &psref);
1263 
1264 	return 0;
1265 }
1266 
1267 #if defined(BRIDGE_IPF)
1268 static int
1269 bridge_ioctl_gfilt(struct bridge_softc *sc, void *arg)
1270 {
1271 	struct ifbrparam *param = arg;
1272 
1273 	param->ifbrp_filter = sc->sc_filter_flags;
1274 
1275 	return 0;
1276 }
1277 
1278 static int
1279 bridge_ioctl_sfilt(struct bridge_softc *sc, void *arg)
1280 {
1281 	struct ifbrparam *param = arg;
1282 	uint32_t nflags, oflags;
1283 
1284 	if (param->ifbrp_filter & ~IFBF_FILT_MASK)
1285 		return EINVAL;
1286 
1287 	nflags = param->ifbrp_filter;
1288 	oflags = sc->sc_filter_flags;
1289 
1290 	if ((nflags & IFBF_FILT_USEIPF) && !(oflags & IFBF_FILT_USEIPF)) {
1291 		pfil_add_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
1292 			sc->sc_if.if_pfil);
1293 	}
1294 	if (!(nflags & IFBF_FILT_USEIPF) && (oflags & IFBF_FILT_USEIPF)) {
1295 		pfil_remove_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
1296 			sc->sc_if.if_pfil);
1297 	}
1298 
1299 	sc->sc_filter_flags = nflags;
1300 
1301 	return 0;
1302 }
1303 #endif /* BRIDGE_IPF */
1304 
1305 static int
1306 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1307 {
1308 	struct ifbreq *req = arg;
1309 	struct bridge_iflist *bif;
1310 	struct psref psref;
1311 
1312 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
1313 	if (bif == NULL)
1314 		return ENOENT;
1315 
1316 	bif->bif_path_cost = req->ifbr_path_cost;
1317 
1318 	if (sc->sc_if.if_flags & IFF_RUNNING)
1319 		bstp_initialization(sc);
1320 
1321 	bridge_release_member(sc, bif, &psref);
1322 
1323 	return 0;
1324 }
1325 
1326 /*
1327  * bridge_ifdetach:
1328  *
1329  *	Detach an interface from a bridge.  Called when a member
1330  *	interface is detaching.
1331  */
1332 void
1333 bridge_ifdetach(struct ifnet *ifp)
1334 {
1335 	struct bridge_softc *sc = ifp->if_bridge;
1336 	struct ifbreq breq;
1337 
1338 	/* ioctl_lock should prevent this from happening */
1339 	KASSERT(sc != NULL);
1340 
1341 	memset(&breq, 0, sizeof(breq));
1342 	strlcpy(breq.ifbr_ifsname, ifp->if_xname, sizeof(breq.ifbr_ifsname));
1343 
1344 	(void) bridge_ioctl_del(sc, &breq);
1345 }
1346 
1347 /*
1348  * bridge_init:
1349  *
1350  *	Initialize a bridge interface.
1351  */
1352 static int
1353 bridge_init(struct ifnet *ifp)
1354 {
1355 	struct bridge_softc *sc = ifp->if_softc;
1356 
1357 	KASSERT((ifp->if_flags & IFF_RUNNING) == 0);
1358 
1359 	callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1360 	    bridge_timer, sc);
1361 	bstp_initialization(sc);
1362 
1363 	ifp->if_flags |= IFF_RUNNING;
1364 	return 0;
1365 }
1366 
1367 /*
1368  * bridge_stop:
1369  *
1370  *	Stop the bridge interface.
1371  */
1372 static void
1373 bridge_stop(struct ifnet *ifp, int disable)
1374 {
1375 	struct bridge_softc *sc = ifp->if_softc;
1376 
1377 	KASSERT((ifp->if_flags & IFF_RUNNING) != 0);
1378 	ifp->if_flags &= ~IFF_RUNNING;
1379 
1380 	callout_halt(&sc->sc_brcallout, NULL);
1381 	workqueue_wait(sc->sc_rtage_wq, &sc->sc_rtage_wk);
1382 	bstp_stop(sc);
1383 	bridge_rtflush(sc, IFBF_FLUSHDYN);
1384 }
1385 
1386 /*
1387  * bridge_enqueue:
1388  *
1389  *	Enqueue a packet on a bridge member interface.
1390  */
1391 void
1392 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m,
1393     int runfilt)
1394 {
1395 	int len, error;
1396 	short mflags;
1397 
1398 	/*
1399 	 * Clear any in-bound checksum flags for this packet.
1400 	 */
1401 	m->m_pkthdr.csum_flags = 0;
1402 
1403 	if (runfilt) {
1404 		if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
1405 		    dst_ifp, PFIL_OUT) != 0) {
1406 			if (m != NULL)
1407 				m_freem(m);
1408 			return;
1409 		}
1410 		if (m == NULL)
1411 			return;
1412 	}
1413 
1414 #ifdef ALTQ
1415 	KERNEL_LOCK(1, NULL);
1416 	/*
1417 	 * If ALTQ is enabled on the member interface, do
1418 	 * classification; the queueing discipline might
1419 	 * not require classification, but might require
1420 	 * the address family/header pointer in the pktattr.
1421 	 */
1422 	if (ALTQ_IS_ENABLED(&dst_ifp->if_snd)) {
1423 		/* XXX IFT_ETHER */
1424 		altq_etherclassify(&dst_ifp->if_snd, m);
1425 	}
1426 	KERNEL_UNLOCK_ONE(NULL);
1427 #endif /* ALTQ */
1428 
1429 	len = m->m_pkthdr.len;
1430 	mflags = m->m_flags;
1431 
1432 	error = if_transmit_lock(dst_ifp, m);
1433 	if (error) {
1434 		/* mbuf is already freed */
1435 		sc->sc_if.if_oerrors++;
1436 		return;
1437 	}
1438 
1439 	sc->sc_if.if_opackets++;
1440 	sc->sc_if.if_obytes += len;
1441 	if (mflags & M_MCAST)
1442 		sc->sc_if.if_omcasts++;
1443 }
1444 
1445 /*
1446  * bridge_output:
1447  *
1448  *	Send output from a bridge member interface.  This
1449  *	performs the bridging function for locally originated
1450  *	packets.
1451  *
1452  *	The mbuf has the Ethernet header already attached.  We must
1453  *	enqueue or free the mbuf before returning.
1454  */
1455 int
1456 bridge_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa,
1457     const struct rtentry *rt)
1458 {
1459 	struct ether_header *eh;
1460 	struct ifnet *dst_if;
1461 	struct bridge_softc *sc;
1462 	int s;
1463 
1464 	/*
1465 	 * bridge_output() is called from ether_output(), furthermore
1466 	 * ifp argument doesn't point to bridge(4). So, don't assert
1467 	 * IFEF_MPSAFE here.
1468 	 */
1469 
1470 	if (m->m_len < ETHER_HDR_LEN) {
1471 		m = m_pullup(m, ETHER_HDR_LEN);
1472 		if (m == NULL)
1473 			return 0;
1474 	}
1475 
1476 	eh = mtod(m, struct ether_header *);
1477 	sc = ifp->if_bridge;
1478 
1479 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1480 		if (memcmp(etherbroadcastaddr,
1481 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
1482 			m->m_flags |= M_BCAST;
1483 		else
1484 			m->m_flags |= M_MCAST;
1485 	}
1486 
1487 	/*
1488 	 * If bridge is down, but the original output interface is up,
1489 	 * go ahead and send out that interface.  Otherwise, the packet
1490 	 * is dropped below.
1491 	 */
1492 	if (__predict_false(sc == NULL) ||
1493 	    (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
1494 		dst_if = ifp;
1495 		goto sendunicast;
1496 	}
1497 
1498 	/*
1499 	 * If the packet is a multicast, or we don't know a better way to
1500 	 * get there, send to all interfaces.
1501 	 */
1502 	if ((m->m_flags & (M_MCAST | M_BCAST)) != 0)
1503 		dst_if = NULL;
1504 	else
1505 		dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1506 	if (dst_if == NULL) {
1507 		/* XXX Should call bridge_broadcast, but there are locking
1508 		 * issues which need resolving first. */
1509 		struct bridge_iflist *bif;
1510 		struct mbuf *mc;
1511 		bool used = false;
1512 
1513 		BRIDGE_PSZ_RENTER(s);
1514 		BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
1515 			struct psref psref;
1516 
1517 			bridge_acquire_member(sc, bif, &psref);
1518 			BRIDGE_PSZ_REXIT(s);
1519 
1520 			dst_if = bif->bif_ifp;
1521 			if ((dst_if->if_flags & IFF_RUNNING) == 0)
1522 				goto next;
1523 
1524 			/*
1525 			 * If this is not the original output interface,
1526 			 * and the interface is participating in spanning
1527 			 * tree, make sure the port is in a state that
1528 			 * allows forwarding.
1529 			 */
1530 			if (dst_if != ifp &&
1531 			    (bif->bif_flags & IFBIF_STP) != 0) {
1532 				switch (bif->bif_state) {
1533 				case BSTP_IFSTATE_BLOCKING:
1534 				case BSTP_IFSTATE_LISTENING:
1535 				case BSTP_IFSTATE_DISABLED:
1536 					goto next;
1537 				}
1538 			}
1539 
1540 			if (PSLIST_READER_NEXT(bif, struct bridge_iflist,
1541 			    bif_next) == NULL &&
1542 			    ((m->m_flags & (M_MCAST | M_BCAST)) == 0 ||
1543 			    dst_if == ifp))
1544 			{
1545 				used = true;
1546 				mc = m;
1547 			} else {
1548 				mc = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1549 				if (mc == NULL) {
1550 					sc->sc_if.if_oerrors++;
1551 					goto next;
1552 				}
1553 			}
1554 
1555 			bridge_enqueue(sc, dst_if, mc, 0);
1556 
1557 			if ((m->m_flags & (M_MCAST | M_BCAST)) != 0 &&
1558 			    dst_if != ifp)
1559 			{
1560 				if (PSLIST_READER_NEXT(bif,
1561 				    struct bridge_iflist, bif_next) == NULL)
1562 				{
1563 					used = true;
1564 					mc = m;
1565 				} else {
1566 					mc = m_copym(m, 0, M_COPYALL,
1567 					    M_DONTWAIT);
1568 					if (mc == NULL) {
1569 						sc->sc_if.if_oerrors++;
1570 						goto next;
1571 					}
1572 				}
1573 
1574 				m_set_rcvif(mc, dst_if);
1575 				mc->m_flags &= ~M_PROMISC;
1576 
1577 				s = splsoftnet();
1578 				KERNEL_LOCK_UNLESS_IFP_MPSAFE(dst_if);
1579 				ether_input(dst_if, mc);
1580 				KERNEL_UNLOCK_UNLESS_IFP_MPSAFE(dst_if);
1581 				splx(s);
1582 			}
1583 
1584 next:
1585 			BRIDGE_PSZ_RENTER(s);
1586 			bridge_release_member(sc, bif, &psref);
1587 
1588 			/* Guarantee we don't re-enter the loop as we already
1589 			 * decided we're at the end. */
1590 			if (used)
1591 				break;
1592 		}
1593 		BRIDGE_PSZ_REXIT(s);
1594 
1595 		if (!used)
1596 			m_freem(m);
1597 		return 0;
1598 	}
1599 
1600  sendunicast:
1601 	/*
1602 	 * XXX Spanning tree consideration here?
1603 	 */
1604 
1605 	if ((dst_if->if_flags & IFF_RUNNING) == 0) {
1606 		m_freem(m);
1607 		return 0;
1608 	}
1609 
1610 	bridge_enqueue(sc, dst_if, m, 0);
1611 
1612 	return 0;
1613 }
1614 
1615 /*
1616  * bridge_start:
1617  *
1618  *	Start output on a bridge.
1619  *
1620  *	NOTE: This routine should never be called in this implementation.
1621  */
1622 static void
1623 bridge_start(struct ifnet *ifp)
1624 {
1625 
1626 	printf("%s: bridge_start() called\n", ifp->if_xname);
1627 }
1628 
1629 /*
1630  * bridge_forward:
1631  *
1632  *	The forwarding function of the bridge.
1633  */
1634 static void
1635 bridge_forward(struct bridge_softc *sc, struct mbuf *m)
1636 {
1637 	struct bridge_iflist *bif;
1638 	struct ifnet *src_if, *dst_if;
1639 	struct ether_header *eh;
1640 	struct psref psref;
1641 	struct psref psref_src;
1642 	DECLARE_LOCK_VARIABLE;
1643 
1644 	if ((sc->sc_if.if_flags & IFF_RUNNING) == 0)
1645 		return;
1646 
1647 	src_if = m_get_rcvif_psref(m, &psref_src);
1648 	if (src_if == NULL) {
1649 		/* Interface is being destroyed? */
1650 		m_freem(m);
1651 		goto out;
1652 	}
1653 
1654 	sc->sc_if.if_ipackets++;
1655 	sc->sc_if.if_ibytes += m->m_pkthdr.len;
1656 
1657 	/*
1658 	 * Look up the bridge_iflist.
1659 	 */
1660 	bif = bridge_lookup_member_if(sc, src_if, &psref);
1661 	if (bif == NULL) {
1662 		/* Interface is not a bridge member (anymore?) */
1663 		m_freem(m);
1664 		goto out;
1665 	}
1666 
1667 	if (bif->bif_flags & IFBIF_STP) {
1668 		switch (bif->bif_state) {
1669 		case BSTP_IFSTATE_BLOCKING:
1670 		case BSTP_IFSTATE_LISTENING:
1671 		case BSTP_IFSTATE_DISABLED:
1672 			m_freem(m);
1673 			bridge_release_member(sc, bif, &psref);
1674 			goto out;
1675 		}
1676 	}
1677 
1678 	eh = mtod(m, struct ether_header *);
1679 
1680 	/*
1681 	 * If the interface is learning, and the source
1682 	 * address is valid and not multicast, record
1683 	 * the address.
1684 	 */
1685 	if ((bif->bif_flags & IFBIF_LEARNING) != 0 &&
1686 	    ETHER_IS_MULTICAST(eh->ether_shost) == 0 &&
1687 	    (eh->ether_shost[0] == 0 &&
1688 	     eh->ether_shost[1] == 0 &&
1689 	     eh->ether_shost[2] == 0 &&
1690 	     eh->ether_shost[3] == 0 &&
1691 	     eh->ether_shost[4] == 0 &&
1692 	     eh->ether_shost[5] == 0) == 0) {
1693 		(void) bridge_rtupdate(sc, eh->ether_shost,
1694 		    src_if, 0, IFBAF_DYNAMIC);
1695 	}
1696 
1697 	if ((bif->bif_flags & IFBIF_STP) != 0 &&
1698 	    bif->bif_state == BSTP_IFSTATE_LEARNING) {
1699 		m_freem(m);
1700 		bridge_release_member(sc, bif, &psref);
1701 		goto out;
1702 	}
1703 
1704 	bridge_release_member(sc, bif, &psref);
1705 
1706 	/*
1707 	 * At this point, the port either doesn't participate
1708 	 * in spanning tree or it is in the forwarding state.
1709 	 */
1710 
1711 	/*
1712 	 * If the packet is unicast, destined for someone on
1713 	 * "this" side of the bridge, drop it.
1714 	 */
1715 	if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
1716 		dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1717 		if (src_if == dst_if) {
1718 			m_freem(m);
1719 			goto out;
1720 		}
1721 	} else {
1722 		/* ...forward it to all interfaces. */
1723 		sc->sc_if.if_imcasts++;
1724 		dst_if = NULL;
1725 	}
1726 
1727 	if (pfil_run_hooks(sc->sc_if.if_pfil, &m, src_if, PFIL_IN) != 0) {
1728 		if (m != NULL)
1729 			m_freem(m);
1730 		goto out;
1731 	}
1732 	if (m == NULL)
1733 		goto out;
1734 
1735 	if (dst_if == NULL) {
1736 		bridge_broadcast(sc, src_if, m);
1737 		goto out;
1738 	}
1739 
1740 	m_put_rcvif_psref(src_if, &psref_src);
1741 	src_if = NULL;
1742 
1743 	/*
1744 	 * At this point, we're dealing with a unicast frame
1745 	 * going to a different interface.
1746 	 */
1747 	if ((dst_if->if_flags & IFF_RUNNING) == 0) {
1748 		m_freem(m);
1749 		goto out;
1750 	}
1751 
1752 	bif = bridge_lookup_member_if(sc, dst_if, &psref);
1753 	if (bif == NULL) {
1754 		/* Not a member of the bridge (anymore?) */
1755 		m_freem(m);
1756 		goto out;
1757 	}
1758 
1759 	if (bif->bif_flags & IFBIF_STP) {
1760 		switch (bif->bif_state) {
1761 		case BSTP_IFSTATE_DISABLED:
1762 		case BSTP_IFSTATE_BLOCKING:
1763 			m_freem(m);
1764 			bridge_release_member(sc, bif, &psref);
1765 			goto out;
1766 		}
1767 	}
1768 
1769 	bridge_release_member(sc, bif, &psref);
1770 
1771 	ACQUIRE_GLOBAL_LOCKS();
1772 	bridge_enqueue(sc, dst_if, m, 1);
1773 	RELEASE_GLOBAL_LOCKS();
1774 out:
1775 	if (src_if != NULL)
1776 		m_put_rcvif_psref(src_if, &psref_src);
1777 	return;
1778 }
1779 
1780 static bool
1781 bstp_state_before_learning(struct bridge_iflist *bif)
1782 {
1783 	if (bif->bif_flags & IFBIF_STP) {
1784 		switch (bif->bif_state) {
1785 		case BSTP_IFSTATE_BLOCKING:
1786 		case BSTP_IFSTATE_LISTENING:
1787 		case BSTP_IFSTATE_DISABLED:
1788 			return true;
1789 		}
1790 	}
1791 	return false;
1792 }
1793 
1794 static bool
1795 bridge_ourether(struct bridge_iflist *bif, struct ether_header *eh, int src)
1796 {
1797 	uint8_t *ether = src ? eh->ether_shost : eh->ether_dhost;
1798 
1799 	if (memcmp(CLLADDR(bif->bif_ifp->if_sadl), ether, ETHER_ADDR_LEN) == 0
1800 #if NCARP > 0
1801 	    || (bif->bif_ifp->if_carp &&
1802 	        carp_ourether(bif->bif_ifp->if_carp, eh, IFT_ETHER, src) != NULL)
1803 #endif /* NCARP > 0 */
1804 	    )
1805 		return true;
1806 
1807 	return false;
1808 }
1809 
1810 /*
1811  * bridge_input:
1812  *
1813  *	Receive input from a member interface.  Queue the packet for
1814  *	bridging if it is not for us.
1815  */
1816 static void
1817 bridge_input(struct ifnet *ifp, struct mbuf *m)
1818 {
1819 	struct bridge_softc *sc = ifp->if_bridge;
1820 	struct bridge_iflist *bif;
1821 	struct ether_header *eh;
1822 	struct psref psref;
1823 	int bound;
1824 	DECLARE_LOCK_VARIABLE;
1825 
1826 	KASSERT(!cpu_intr_p());
1827 
1828 	if (__predict_false(sc == NULL) ||
1829 	    (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
1830 		ACQUIRE_GLOBAL_LOCKS();
1831 		ether_input(ifp, m);
1832 		RELEASE_GLOBAL_LOCKS();
1833 		return;
1834 	}
1835 
1836 	bound = curlwp_bind();
1837 	bif = bridge_lookup_member_if(sc, ifp, &psref);
1838 	if (bif == NULL) {
1839 		curlwp_bindx(bound);
1840 		ACQUIRE_GLOBAL_LOCKS();
1841 		ether_input(ifp, m);
1842 		RELEASE_GLOBAL_LOCKS();
1843 		return;
1844 	}
1845 
1846 	eh = mtod(m, struct ether_header *);
1847 
1848 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1849 		if (memcmp(etherbroadcastaddr,
1850 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
1851 			m->m_flags |= M_BCAST;
1852 		else
1853 			m->m_flags |= M_MCAST;
1854 	}
1855 
1856 	/*
1857 	 * A 'fast' path for packets addressed to interfaces that are
1858 	 * part of this bridge.
1859 	 */
1860 	if (!(m->m_flags & (M_BCAST|M_MCAST)) &&
1861 	    !bstp_state_before_learning(bif)) {
1862 		struct bridge_iflist *_bif;
1863 		struct ifnet *_ifp = NULL;
1864 		int s;
1865 		struct psref _psref;
1866 
1867 		BRIDGE_PSZ_RENTER(s);
1868 		BRIDGE_IFLIST_READER_FOREACH(_bif, sc) {
1869 			/* It is destined for us. */
1870 			if (bridge_ourether(_bif, eh, 0)) {
1871 				bridge_acquire_member(sc, _bif, &_psref);
1872 				BRIDGE_PSZ_REXIT(s);
1873 				if (_bif->bif_flags & IFBIF_LEARNING)
1874 					(void) bridge_rtupdate(sc,
1875 					    eh->ether_shost, ifp, 0, IFBAF_DYNAMIC);
1876 				m_set_rcvif(m, _bif->bif_ifp);
1877 				_ifp = _bif->bif_ifp;
1878 				bridge_release_member(sc, _bif, &_psref);
1879 				goto out;
1880 			}
1881 
1882 			/* We just received a packet that we sent out. */
1883 			if (bridge_ourether(_bif, eh, 1))
1884 				break;
1885 		}
1886 		BRIDGE_PSZ_REXIT(s);
1887 out:
1888 
1889 		if (_bif != NULL) {
1890 			bridge_release_member(sc, bif, &psref);
1891 			curlwp_bindx(bound);
1892 			if (_ifp != NULL) {
1893 				m->m_flags &= ~M_PROMISC;
1894 				ACQUIRE_GLOBAL_LOCKS();
1895 				ether_input(_ifp, m);
1896 				RELEASE_GLOBAL_LOCKS();
1897 			} else
1898 				m_freem(m);
1899 			return;
1900 		}
1901 	}
1902 
1903 	/* Tap off 802.1D packets; they do not get forwarded. */
1904 	if (bif->bif_flags & IFBIF_STP &&
1905 	    memcmp(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN) == 0) {
1906 		bstp_input(sc, bif, m);
1907 		bridge_release_member(sc, bif, &psref);
1908 		curlwp_bindx(bound);
1909 		return;
1910 	}
1911 
1912 	/*
1913 	 * A normal switch would discard the packet here, but that's not what
1914 	 * we've done historically. This also prevents some obnoxious behaviour.
1915 	 */
1916 	if (bstp_state_before_learning(bif)) {
1917 		bridge_release_member(sc, bif, &psref);
1918 		curlwp_bindx(bound);
1919 		ACQUIRE_GLOBAL_LOCKS();
1920 		ether_input(ifp, m);
1921 		RELEASE_GLOBAL_LOCKS();
1922 		return;
1923 	}
1924 
1925 	bridge_release_member(sc, bif, &psref);
1926 
1927 	bridge_forward(sc, m);
1928 
1929 	curlwp_bindx(bound);
1930 }
1931 
1932 /*
1933  * bridge_broadcast:
1934  *
1935  *	Send a frame to all interfaces that are members of
1936  *	the bridge, except for the one on which the packet
1937  *	arrived.
1938  */
1939 static void
1940 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
1941     struct mbuf *m)
1942 {
1943 	struct bridge_iflist *bif;
1944 	struct mbuf *mc;
1945 	struct ifnet *dst_if;
1946 	bool bmcast;
1947 	int s;
1948 	DECLARE_LOCK_VARIABLE;
1949 
1950 	bmcast = m->m_flags & (M_BCAST|M_MCAST);
1951 
1952 	BRIDGE_PSZ_RENTER(s);
1953 	BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
1954 		struct psref psref;
1955 
1956 		bridge_acquire_member(sc, bif, &psref);
1957 		BRIDGE_PSZ_REXIT(s);
1958 
1959 		dst_if = bif->bif_ifp;
1960 
1961 		if (bif->bif_flags & IFBIF_STP) {
1962 			switch (bif->bif_state) {
1963 			case BSTP_IFSTATE_BLOCKING:
1964 			case BSTP_IFSTATE_DISABLED:
1965 				goto next;
1966 			}
1967 		}
1968 
1969 		if ((bif->bif_flags & IFBIF_DISCOVER) == 0 && !bmcast)
1970 			goto next;
1971 
1972 		if ((dst_if->if_flags & IFF_RUNNING) == 0)
1973 			goto next;
1974 
1975 		if (dst_if != src_if) {
1976 			mc = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
1977 			if (mc == NULL) {
1978 				sc->sc_if.if_oerrors++;
1979 				goto next;
1980 			}
1981 			ACQUIRE_GLOBAL_LOCKS();
1982 			bridge_enqueue(sc, dst_if, mc, 1);
1983 			RELEASE_GLOBAL_LOCKS();
1984 		}
1985 
1986 		if (bmcast) {
1987 			mc = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
1988 			if (mc == NULL) {
1989 				sc->sc_if.if_oerrors++;
1990 				goto next;
1991 			}
1992 
1993 			m_set_rcvif(mc, dst_if);
1994 			mc->m_flags &= ~M_PROMISC;
1995 
1996 			ACQUIRE_GLOBAL_LOCKS();
1997 			ether_input(dst_if, mc);
1998 			RELEASE_GLOBAL_LOCKS();
1999 		}
2000 next:
2001 		BRIDGE_PSZ_RENTER(s);
2002 		bridge_release_member(sc, bif, &psref);
2003 	}
2004 	BRIDGE_PSZ_REXIT(s);
2005 
2006 	m_freem(m);
2007 }
2008 
2009 static int
2010 bridge_rtalloc(struct bridge_softc *sc, const uint8_t *dst,
2011     struct bridge_rtnode **brtp)
2012 {
2013 	struct bridge_rtnode *brt;
2014 	int error;
2015 
2016 	if (sc->sc_brtcnt >= sc->sc_brtmax)
2017 		return ENOSPC;
2018 
2019 	/*
2020 	 * Allocate a new bridge forwarding node, and
2021 	 * initialize the expiration time and Ethernet
2022 	 * address.
2023 	 */
2024 	brt = pool_get(&bridge_rtnode_pool, PR_NOWAIT);
2025 	if (brt == NULL)
2026 		return ENOMEM;
2027 
2028 	memset(brt, 0, sizeof(*brt));
2029 	brt->brt_expire = time_uptime + sc->sc_brttimeout;
2030 	brt->brt_flags = IFBAF_DYNAMIC;
2031 	memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2032 	PSLIST_ENTRY_INIT(brt, brt_list);
2033 	PSLIST_ENTRY_INIT(brt, brt_hash);
2034 
2035 	BRIDGE_RT_LOCK(sc);
2036 	error = bridge_rtnode_insert(sc, brt);
2037 	BRIDGE_RT_UNLOCK(sc);
2038 
2039 	if (error != 0) {
2040 		pool_put(&bridge_rtnode_pool, brt);
2041 		return error;
2042 	}
2043 
2044 	*brtp = brt;
2045 	return 0;
2046 }
2047 
2048 /*
2049  * bridge_rtupdate:
2050  *
2051  *	Add a bridge routing entry.
2052  */
2053 static int
2054 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
2055     struct ifnet *dst_if, int setflags, uint8_t flags)
2056 {
2057 	struct bridge_rtnode *brt;
2058 	int s;
2059 
2060 again:
2061 	/*
2062 	 * A route for this destination might already exist.  If so,
2063 	 * update it, otherwise create a new one.
2064 	 */
2065 	BRIDGE_RT_RENTER(s);
2066 	brt = bridge_rtnode_lookup(sc, dst);
2067 
2068 	if (brt != NULL) {
2069 		brt->brt_ifp = dst_if;
2070 		if (setflags) {
2071 			brt->brt_flags = flags;
2072 			if (flags & IFBAF_STATIC)
2073 				brt->brt_expire = 0;
2074 			else
2075 				brt->brt_expire = time_uptime + sc->sc_brttimeout;
2076 		} else {
2077 			if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2078 				brt->brt_expire = time_uptime + sc->sc_brttimeout;
2079 		}
2080 	}
2081 	BRIDGE_RT_REXIT(s);
2082 
2083 	if (brt == NULL) {
2084 		int r;
2085 
2086 		r = bridge_rtalloc(sc, dst, &brt);
2087 		if (r != 0)
2088 			return r;
2089 		goto again;
2090 	}
2091 
2092 	return 0;
2093 }
2094 
2095 /*
2096  * bridge_rtlookup:
2097  *
2098  *	Lookup the destination interface for an address.
2099  */
2100 static struct ifnet *
2101 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr)
2102 {
2103 	struct bridge_rtnode *brt;
2104 	struct ifnet *ifs = NULL;
2105 	int s;
2106 
2107 	BRIDGE_RT_RENTER(s);
2108 	brt = bridge_rtnode_lookup(sc, addr);
2109 	if (brt != NULL)
2110 		ifs = brt->brt_ifp;
2111 	BRIDGE_RT_REXIT(s);
2112 
2113 	return ifs;
2114 }
2115 
2116 typedef bool (*bridge_iterate_cb_t)
2117     (struct bridge_softc *, struct bridge_rtnode *, bool *, void *);
2118 
2119 /*
2120  * bridge_rtlist_iterate_remove:
2121  *
2122  *	It iterates on sc->sc_rtlist and removes rtnodes of it which func
2123  *	callback judges to remove. Removals of rtnodes are done in a manner
2124  *	of pserialize. To this end, all kmem_* operations are placed out of
2125  *	mutexes.
2126  */
2127 static void
2128 bridge_rtlist_iterate_remove(struct bridge_softc *sc, bridge_iterate_cb_t func, void *arg)
2129 {
2130 	struct bridge_rtnode *brt;
2131 	struct bridge_rtnode **brt_list;
2132 	int i, count;
2133 
2134 retry:
2135 	count = sc->sc_brtcnt;
2136 	if (count == 0)
2137 		return;
2138 	brt_list = kmem_alloc(sizeof(*brt_list) * count, KM_SLEEP);
2139 
2140 	BRIDGE_RT_LOCK(sc);
2141 	if (__predict_false(sc->sc_brtcnt > count)) {
2142 		/* The rtnodes increased, we need more memory */
2143 		BRIDGE_RT_UNLOCK(sc);
2144 		kmem_free(brt_list, sizeof(*brt_list) * count);
2145 		goto retry;
2146 	}
2147 
2148 	i = 0;
2149 	/*
2150 	 * We don't need to use a _SAFE variant here because we know
2151 	 * that a removed item keeps its next pointer as-is thanks to
2152 	 * pslist(9) and isn't freed in the loop.
2153 	 */
2154 	BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) {
2155 		bool need_break = false;
2156 		if (func(sc, brt, &need_break, arg)) {
2157 			bridge_rtnode_remove(sc, brt);
2158 			brt_list[i++] = brt;
2159 		}
2160 		if (need_break)
2161 			break;
2162 	}
2163 
2164 	if (i > 0)
2165 		BRIDGE_RT_PSZ_PERFORM(sc);
2166 	BRIDGE_RT_UNLOCK(sc);
2167 
2168 	while (--i >= 0)
2169 		bridge_rtnode_destroy(brt_list[i]);
2170 
2171 	kmem_free(brt_list, sizeof(*brt_list) * count);
2172 }
2173 
2174 static bool
2175 bridge_rttrim0_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
2176     bool *need_break, void *arg)
2177 {
2178 	if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2179 		/* Take into account of the subsequent removal */
2180 		if ((sc->sc_brtcnt - 1) <= sc->sc_brtmax)
2181 			*need_break = true;
2182 		return true;
2183 	} else
2184 		return false;
2185 }
2186 
2187 static void
2188 bridge_rttrim0(struct bridge_softc *sc)
2189 {
2190 	bridge_rtlist_iterate_remove(sc, bridge_rttrim0_cb, NULL);
2191 }
2192 
2193 /*
2194  * bridge_rttrim:
2195  *
2196  *	Trim the routine table so that we have a number
2197  *	of routing entries less than or equal to the
2198  *	maximum number.
2199  */
2200 static void
2201 bridge_rttrim(struct bridge_softc *sc)
2202 {
2203 
2204 	/* Make sure we actually need to do this. */
2205 	if (sc->sc_brtcnt <= sc->sc_brtmax)
2206 		return;
2207 
2208 	/* Force an aging cycle; this might trim enough addresses. */
2209 	bridge_rtage(sc);
2210 	if (sc->sc_brtcnt <= sc->sc_brtmax)
2211 		return;
2212 
2213 	bridge_rttrim0(sc);
2214 
2215 	return;
2216 }
2217 
2218 /*
2219  * bridge_timer:
2220  *
2221  *	Aging timer for the bridge.
2222  */
2223 static void
2224 bridge_timer(void *arg)
2225 {
2226 	struct bridge_softc *sc = arg;
2227 
2228 	workqueue_enqueue(sc->sc_rtage_wq, &sc->sc_rtage_wk, NULL);
2229 }
2230 
2231 static void
2232 bridge_rtage_work(struct work *wk, void *arg)
2233 {
2234 	struct bridge_softc *sc = arg;
2235 
2236 	KASSERT(wk == &sc->sc_rtage_wk);
2237 
2238 	bridge_rtage(sc);
2239 
2240 	if (sc->sc_if.if_flags & IFF_RUNNING)
2241 		callout_reset(&sc->sc_brcallout,
2242 		    bridge_rtable_prune_period * hz, bridge_timer, sc);
2243 }
2244 
2245 static bool
2246 bridge_rtage_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
2247     bool *need_break, void *arg)
2248 {
2249 	if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2250 	    time_uptime >= brt->brt_expire)
2251 		return true;
2252 	else
2253 		return false;
2254 }
2255 
2256 /*
2257  * bridge_rtage:
2258  *
2259  *	Perform an aging cycle.
2260  */
2261 static void
2262 bridge_rtage(struct bridge_softc *sc)
2263 {
2264 	bridge_rtlist_iterate_remove(sc, bridge_rtage_cb, NULL);
2265 }
2266 
2267 
2268 static bool
2269 bridge_rtflush_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
2270     bool *need_break, void *arg)
2271 {
2272 	int full = *(int*)arg;
2273 
2274 	if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2275 		return true;
2276 	else
2277 		return false;
2278 }
2279 
2280 /*
2281  * bridge_rtflush:
2282  *
2283  *	Remove all dynamic addresses from the bridge.
2284  */
2285 static void
2286 bridge_rtflush(struct bridge_softc *sc, int full)
2287 {
2288 	bridge_rtlist_iterate_remove(sc, bridge_rtflush_cb, &full);
2289 }
2290 
2291 /*
2292  * bridge_rtdaddr:
2293  *
2294  *	Remove an address from the table.
2295  */
2296 static int
2297 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr)
2298 {
2299 	struct bridge_rtnode *brt;
2300 
2301 	BRIDGE_RT_LOCK(sc);
2302 	if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL) {
2303 		BRIDGE_RT_UNLOCK(sc);
2304 		return ENOENT;
2305 	}
2306 	bridge_rtnode_remove(sc, brt);
2307 	BRIDGE_RT_PSZ_PERFORM(sc);
2308 	BRIDGE_RT_UNLOCK(sc);
2309 
2310 	bridge_rtnode_destroy(brt);
2311 
2312 	return 0;
2313 }
2314 
2315 /*
2316  * bridge_rtdelete:
2317  *
2318  *	Delete routes to a speicifc member interface.
2319  */
2320 static void
2321 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp)
2322 {
2323 	struct bridge_rtnode *brt;
2324 
2325 	/* XXX pserialize_perform for each entry is slow */
2326 again:
2327 	BRIDGE_RT_LOCK(sc);
2328 	BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) {
2329 		if (brt->brt_ifp == ifp)
2330 			break;
2331 	}
2332 	if (brt == NULL) {
2333 		BRIDGE_RT_UNLOCK(sc);
2334 		return;
2335 	}
2336 	bridge_rtnode_remove(sc, brt);
2337 	BRIDGE_RT_PSZ_PERFORM(sc);
2338 	BRIDGE_RT_UNLOCK(sc);
2339 
2340 	bridge_rtnode_destroy(brt);
2341 
2342 	goto again;
2343 }
2344 
2345 /*
2346  * bridge_rtable_init:
2347  *
2348  *	Initialize the route table for this bridge.
2349  */
2350 static void
2351 bridge_rtable_init(struct bridge_softc *sc)
2352 {
2353 	int i;
2354 
2355 	sc->sc_rthash = kmem_alloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2356 	    KM_SLEEP);
2357 
2358 	for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2359 		PSLIST_INIT(&sc->sc_rthash[i]);
2360 
2361 	sc->sc_rthash_key = cprng_fast32();
2362 
2363 	PSLIST_INIT(&sc->sc_rtlist);
2364 
2365 	sc->sc_rtlist_psz = pserialize_create();
2366 	sc->sc_rtlist_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_SOFTNET);
2367 }
2368 
2369 /*
2370  * bridge_rtable_fini:
2371  *
2372  *	Deconstruct the route table for this bridge.
2373  */
2374 static void
2375 bridge_rtable_fini(struct bridge_softc *sc)
2376 {
2377 
2378 	kmem_free(sc->sc_rthash, sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE);
2379 	mutex_obj_free(sc->sc_rtlist_lock);
2380 	pserialize_destroy(sc->sc_rtlist_psz);
2381 }
2382 
2383 /*
2384  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2385  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2386  */
2387 #define	mix(a, b, c)							\
2388 do {									\
2389 	a -= b; a -= c; a ^= (c >> 13);					\
2390 	b -= c; b -= a; b ^= (a << 8);					\
2391 	c -= a; c -= b; c ^= (b >> 13);					\
2392 	a -= b; a -= c; a ^= (c >> 12);					\
2393 	b -= c; b -= a; b ^= (a << 16);					\
2394 	c -= a; c -= b; c ^= (b >> 5);					\
2395 	a -= b; a -= c; a ^= (c >> 3);					\
2396 	b -= c; b -= a; b ^= (a << 10);					\
2397 	c -= a; c -= b; c ^= (b >> 15);					\
2398 } while (/*CONSTCOND*/0)
2399 
2400 static inline uint32_t
2401 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2402 {
2403 	uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2404 
2405 	b += addr[5] << 8;
2406 	b += addr[4];
2407 	a += addr[3] << 24;
2408 	a += addr[2] << 16;
2409 	a += addr[1] << 8;
2410 	a += addr[0];
2411 
2412 	mix(a, b, c);
2413 
2414 	return (c & BRIDGE_RTHASH_MASK);
2415 }
2416 
2417 #undef mix
2418 
2419 /*
2420  * bridge_rtnode_lookup:
2421  *
2422  *	Look up a bridge route node for the specified destination.
2423  */
2424 static struct bridge_rtnode *
2425 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr)
2426 {
2427 	struct bridge_rtnode *brt;
2428 	uint32_t hash;
2429 	int dir;
2430 
2431 	hash = bridge_rthash(sc, addr);
2432 	BRIDGE_RTHASH_READER_FOREACH(brt, sc, hash) {
2433 		dir = memcmp(addr, brt->brt_addr, ETHER_ADDR_LEN);
2434 		if (dir == 0)
2435 			return brt;
2436 		if (dir > 0)
2437 			return NULL;
2438 	}
2439 
2440 	return NULL;
2441 }
2442 
2443 /*
2444  * bridge_rtnode_insert:
2445  *
2446  *	Insert the specified bridge node into the route table.  We
2447  *	assume the entry is not already in the table.
2448  */
2449 static int
2450 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2451 {
2452 	struct bridge_rtnode *lbrt, *prev = NULL;
2453 	uint32_t hash;
2454 
2455 	KASSERT(BRIDGE_RT_LOCKED(sc));
2456 
2457 	hash = bridge_rthash(sc, brt->brt_addr);
2458 	BRIDGE_RTHASH_WRITER_FOREACH(lbrt, sc, hash) {
2459 		int dir = memcmp(brt->brt_addr, lbrt->brt_addr, ETHER_ADDR_LEN);
2460 		if (dir == 0)
2461 			return EEXIST;
2462 		if (dir > 0)
2463 			break;
2464 		prev = lbrt;
2465 	}
2466 	if (prev == NULL)
2467 		BRIDGE_RTHASH_WRITER_INSERT_HEAD(sc, hash, brt);
2468 	else
2469 		BRIDGE_RTHASH_WRITER_INSERT_AFTER(prev, brt);
2470 
2471 	BRIDGE_RTLIST_WRITER_INSERT_HEAD(sc, brt);
2472 	sc->sc_brtcnt++;
2473 
2474 	return 0;
2475 }
2476 
2477 /*
2478  * bridge_rtnode_remove:
2479  *
2480  *	Remove a bridge rtnode from the rthash and the rtlist of a bridge.
2481  */
2482 static void
2483 bridge_rtnode_remove(struct bridge_softc *sc, struct bridge_rtnode *brt)
2484 {
2485 
2486 	KASSERT(BRIDGE_RT_LOCKED(sc));
2487 
2488 	BRIDGE_RTHASH_WRITER_REMOVE(brt);
2489 	BRIDGE_RTLIST_WRITER_REMOVE(brt);
2490 	sc->sc_brtcnt--;
2491 }
2492 
2493 /*
2494  * bridge_rtnode_destroy:
2495  *
2496  *	Destroy a bridge rtnode.
2497  */
2498 static void
2499 bridge_rtnode_destroy(struct bridge_rtnode *brt)
2500 {
2501 
2502 	PSLIST_ENTRY_DESTROY(brt, brt_list);
2503 	PSLIST_ENTRY_DESTROY(brt, brt_hash);
2504 	pool_put(&bridge_rtnode_pool, brt);
2505 }
2506 
2507 #if defined(BRIDGE_IPF)
2508 extern pfil_head_t *inet_pfil_hook;                 /* XXX */
2509 extern pfil_head_t *inet6_pfil_hook;                /* XXX */
2510 
2511 /*
2512  * Send bridge packets through IPF if they are one of the types IPF can deal
2513  * with, or if they are ARP or REVARP.  (IPF will pass ARP and REVARP without
2514  * question.)
2515  */
2516 static int
2517 bridge_ipf(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
2518 {
2519 	int snap, error;
2520 	struct ether_header *eh1, eh2;
2521 	struct llc llc1;
2522 	uint16_t ether_type;
2523 
2524 	snap = 0;
2525 	error = -1;	/* Default error if not error == 0 */
2526 	eh1 = mtod(*mp, struct ether_header *);
2527 	ether_type = ntohs(eh1->ether_type);
2528 
2529 	/*
2530 	 * Check for SNAP/LLC.
2531 	 */
2532 	if (ether_type < ETHERMTU) {
2533 		struct llc *llc2 = (struct llc *)(eh1 + 1);
2534 
2535 		if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
2536 		    llc2->llc_dsap == LLC_SNAP_LSAP &&
2537 		    llc2->llc_ssap == LLC_SNAP_LSAP &&
2538 		    llc2->llc_control == LLC_UI) {
2539 			ether_type = htons(llc2->llc_un.type_snap.ether_type);
2540 			snap = 1;
2541 		}
2542 	}
2543 
2544 	/*
2545 	 * If we're trying to filter bridge traffic, don't look at anything
2546 	 * other than IP and ARP traffic.  If the filter doesn't understand
2547 	 * IPv6, don't allow IPv6 through the bridge either.  This is lame
2548 	 * since if we really wanted, say, an AppleTalk filter, we are hosed,
2549 	 * but of course we don't have an AppleTalk filter to begin with.
2550 	 * (Note that since IPF doesn't understand ARP it will pass *ALL*
2551 	 * ARP traffic.)
2552 	 */
2553 	switch (ether_type) {
2554 		case ETHERTYPE_ARP:
2555 		case ETHERTYPE_REVARP:
2556 			return 0; /* Automatically pass */
2557 		case ETHERTYPE_IP:
2558 # ifdef INET6
2559 		case ETHERTYPE_IPV6:
2560 # endif /* INET6 */
2561 			break;
2562 		default:
2563 			goto bad;
2564 	}
2565 
2566 	/* Strip off the Ethernet header and keep a copy. */
2567 	m_copydata(*mp, 0, ETHER_HDR_LEN, (void *) &eh2);
2568 	m_adj(*mp, ETHER_HDR_LEN);
2569 
2570 	/* Strip off snap header, if present */
2571 	if (snap) {
2572 		m_copydata(*mp, 0, sizeof(struct llc), (void *) &llc1);
2573 		m_adj(*mp, sizeof(struct llc));
2574 	}
2575 
2576 	/*
2577 	 * Check basic packet sanity and run IPF through pfil.
2578 	 */
2579 	KASSERT(!cpu_intr_p());
2580 	switch (ether_type)
2581 	{
2582 	case ETHERTYPE_IP :
2583 		error = bridge_ip_checkbasic(mp);
2584 		if (error == 0)
2585 			error = pfil_run_hooks(inet_pfil_hook, mp, ifp, dir);
2586 		break;
2587 # ifdef INET6
2588 	case ETHERTYPE_IPV6 :
2589 		error = bridge_ip6_checkbasic(mp);
2590 		if (error == 0)
2591 			error = pfil_run_hooks(inet6_pfil_hook, mp, ifp, dir);
2592 		break;
2593 # endif
2594 	default :
2595 		error = 0;
2596 		break;
2597 	}
2598 
2599 	if (*mp == NULL)
2600 		return error;
2601 	if (error != 0)
2602 		goto bad;
2603 
2604 	error = -1;
2605 
2606 	/*
2607 	 * Finally, put everything back the way it was and return
2608 	 */
2609 	if (snap) {
2610 		M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
2611 		if (*mp == NULL)
2612 			return error;
2613 		bcopy(&llc1, mtod(*mp, void *), sizeof(struct llc));
2614 	}
2615 
2616 	M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
2617 	if (*mp == NULL)
2618 		return error;
2619 	bcopy(&eh2, mtod(*mp, void *), ETHER_HDR_LEN);
2620 
2621 	return 0;
2622 
2623     bad:
2624 	m_freem(*mp);
2625 	*mp = NULL;
2626 	return error;
2627 }
2628 
2629 /*
2630  * Perform basic checks on header size since
2631  * IPF assumes ip_input has already processed
2632  * it for it.  Cut-and-pasted from ip_input.c.
2633  * Given how simple the IPv6 version is,
2634  * does the IPv4 version really need to be
2635  * this complicated?
2636  *
2637  * XXX Should we update ipstat here, or not?
2638  * XXX Right now we update ipstat but not
2639  * XXX csum_counter.
2640  */
2641 static int
2642 bridge_ip_checkbasic(struct mbuf **mp)
2643 {
2644 	struct mbuf *m = *mp;
2645 	struct ip *ip;
2646 	int len, hlen;
2647 
2648 	if (*mp == NULL)
2649 		return -1;
2650 
2651 	if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
2652 		if ((m = m_copyup(m, sizeof(struct ip),
2653 			(max_linkhdr + 3) & ~3)) == NULL) {
2654 			/* XXXJRT new stat, please */
2655 			ip_statinc(IP_STAT_TOOSMALL);
2656 			goto bad;
2657 		}
2658 	} else if (__predict_false(m->m_len < sizeof (struct ip))) {
2659 		if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
2660 			ip_statinc(IP_STAT_TOOSMALL);
2661 			goto bad;
2662 		}
2663 	}
2664 	ip = mtod(m, struct ip *);
2665 	if (ip == NULL) goto bad;
2666 
2667 	if (ip->ip_v != IPVERSION) {
2668 		ip_statinc(IP_STAT_BADVERS);
2669 		goto bad;
2670 	}
2671 	hlen = ip->ip_hl << 2;
2672 	if (hlen < sizeof(struct ip)) { /* minimum header length */
2673 		ip_statinc(IP_STAT_BADHLEN);
2674 		goto bad;
2675 	}
2676 	if (hlen > m->m_len) {
2677 		if ((m = m_pullup(m, hlen)) == 0) {
2678 			ip_statinc(IP_STAT_BADHLEN);
2679 			goto bad;
2680 		}
2681 		ip = mtod(m, struct ip *);
2682 		if (ip == NULL) goto bad;
2683 	}
2684 
2685 	switch (m->m_pkthdr.csum_flags &
2686 	        ((m_get_rcvif_NOMPSAFE(m)->if_csum_flags_rx & M_CSUM_IPv4) |
2687 	         M_CSUM_IPv4_BAD)) {
2688 	case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
2689 		/* INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); */
2690 		goto bad;
2691 
2692 	case M_CSUM_IPv4:
2693 		/* Checksum was okay. */
2694 		/* INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); */
2695 		break;
2696 
2697 	default:
2698 		/* Must compute it ourselves. */
2699 		/* INET_CSUM_COUNTER_INCR(&ip_swcsum); */
2700 		if (in_cksum(m, hlen) != 0)
2701 			goto bad;
2702 		break;
2703 	}
2704 
2705 	/* Retrieve the packet length. */
2706 	len = ntohs(ip->ip_len);
2707 
2708 	/*
2709 	 * Check for additional length bogosity
2710 	 */
2711 	if (len < hlen) {
2712 		ip_statinc(IP_STAT_BADLEN);
2713 		goto bad;
2714 	}
2715 
2716 	/*
2717 	 * Check that the amount of data in the buffers
2718 	 * is as at least much as the IP header would have us expect.
2719 	 * Drop packet if shorter than we expect.
2720 	 */
2721 	if (m->m_pkthdr.len < len) {
2722 		ip_statinc(IP_STAT_TOOSHORT);
2723 		goto bad;
2724 	}
2725 
2726 	/* Checks out, proceed */
2727 	*mp = m;
2728 	return 0;
2729 
2730     bad:
2731 	*mp = m;
2732 	return -1;
2733 }
2734 
2735 # ifdef INET6
2736 /*
2737  * Same as above, but for IPv6.
2738  * Cut-and-pasted from ip6_input.c.
2739  * XXX Should we update ip6stat, or not?
2740  */
2741 static int
2742 bridge_ip6_checkbasic(struct mbuf **mp)
2743 {
2744 	struct mbuf *m = *mp;
2745 	struct ip6_hdr *ip6;
2746 
2747 	/*
2748 	 * If the IPv6 header is not aligned, slurp it up into a new
2749 	 * mbuf with space for link headers, in the event we forward
2750 	 * it.  Otherwise, if it is aligned, make sure the entire base
2751 	 * IPv6 header is in the first mbuf of the chain.
2752 	 */
2753 	if (IP6_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
2754 		struct ifnet *inifp = m_get_rcvif_NOMPSAFE(m);
2755 		if ((m = m_copyup(m, sizeof(struct ip6_hdr),
2756 		                  (max_linkhdr + 3) & ~3)) == NULL) {
2757 			/* XXXJRT new stat, please */
2758 			ip6_statinc(IP6_STAT_TOOSMALL);
2759 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
2760 			goto bad;
2761 		}
2762 	} else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
2763 		struct ifnet *inifp = m_get_rcvif_NOMPSAFE(m);
2764 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
2765 			ip6_statinc(IP6_STAT_TOOSMALL);
2766 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
2767 			goto bad;
2768 		}
2769 	}
2770 
2771 	ip6 = mtod(m, struct ip6_hdr *);
2772 
2773 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
2774 		ip6_statinc(IP6_STAT_BADVERS);
2775 		in6_ifstat_inc(m_get_rcvif_NOMPSAFE(m), ifs6_in_hdrerr);
2776 		goto bad;
2777 	}
2778 
2779 	/* Checks out, proceed */
2780 	*mp = m;
2781 	return 0;
2782 
2783     bad:
2784 	*mp = m;
2785 	return -1;
2786 }
2787 # endif /* INET6 */
2788 #endif /* BRIDGE_IPF */
2789