xref: /netbsd-src/sys/net/if_bridge.c (revision 16dce51364ebe8aeafbae46bc5aa167b8115bc45)
1 /*	$NetBSD: if_bridge.c,v 1.149 2018/04/10 07:05:39 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 1.149 2018/04/10 07:05:39 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)	if ((_sc)->sc_rtlist_lock) \
185 					mutex_enter((_sc)->sc_rtlist_lock)
186 #define BRIDGE_RT_UNLOCK(_sc)	if ((_sc)->sc_rtlist_lock) \
187 					mutex_exit((_sc)->sc_rtlist_lock)
188 #define BRIDGE_RT_LOCKED(_sc)	(!(_sc)->sc_rtlist_lock || \
189 				 mutex_owned((_sc)->sc_rtlist_lock))
190 
191 #define BRIDGE_RT_PSZ_PERFORM(_sc) \
192 				if ((_sc)->sc_rtlist_psz != NULL) \
193 					pserialize_perform((_sc)->sc_rtlist_psz);
194 
195 #define BRIDGE_RT_RENTER(__s)	do { __s = pserialize_read_enter(); } while (0)
196 #define BRIDGE_RT_REXIT(__s)	do { pserialize_read_exit(__s); } while (0)
197 
198 
199 #ifdef NET_MPSAFE
200 #define DECLARE_LOCK_VARIABLE
201 #define ACQUIRE_GLOBAL_LOCKS()	do { } while (0)
202 #define RELEASE_GLOBAL_LOCKS()	do { } while (0)
203 #else
204 #define DECLARE_LOCK_VARIABLE	int __s
205 #define ACQUIRE_GLOBAL_LOCKS()	do {					\
206 					KERNEL_LOCK(1, NULL);		\
207 					mutex_enter(softnet_lock);	\
208 					__s = splsoftnet();		\
209 				} while (0)
210 #define RELEASE_GLOBAL_LOCKS()	do {					\
211 					splx(__s);			\
212 					mutex_exit(softnet_lock);	\
213 					KERNEL_UNLOCK_ONE(NULL);	\
214 				} while (0)
215 #endif
216 
217 struct psref_class *bridge_psref_class __read_mostly;
218 
219 int	bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
220 
221 static struct pool bridge_rtnode_pool;
222 
223 static int	bridge_clone_create(struct if_clone *, int);
224 static int	bridge_clone_destroy(struct ifnet *);
225 
226 static int	bridge_ioctl(struct ifnet *, u_long, void *);
227 static int	bridge_init(struct ifnet *);
228 static void	bridge_stop(struct ifnet *, int);
229 static void	bridge_start(struct ifnet *);
230 
231 static void	bridge_input(struct ifnet *, struct mbuf *);
232 static void	bridge_forward(struct bridge_softc *, struct mbuf *);
233 
234 static void	bridge_timer(void *);
235 
236 static void	bridge_broadcast(struct bridge_softc *, struct ifnet *,
237 				 struct mbuf *);
238 
239 static int	bridge_rtupdate(struct bridge_softc *, const uint8_t *,
240 				struct ifnet *, int, uint8_t);
241 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *);
242 static void	bridge_rttrim(struct bridge_softc *);
243 static void	bridge_rtage(struct bridge_softc *);
244 static void	bridge_rtage_work(struct work *, void *);
245 static void	bridge_rtflush(struct bridge_softc *, int);
246 static int	bridge_rtdaddr(struct bridge_softc *, const uint8_t *);
247 static void	bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp);
248 
249 static void	bridge_rtable_init(struct bridge_softc *);
250 static void	bridge_rtable_fini(struct bridge_softc *);
251 
252 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
253 						  const uint8_t *);
254 static int	bridge_rtnode_insert(struct bridge_softc *,
255 				     struct bridge_rtnode *);
256 static void	bridge_rtnode_remove(struct bridge_softc *,
257 				     struct bridge_rtnode *);
258 static void	bridge_rtnode_destroy(struct bridge_rtnode *);
259 
260 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
261 						  const char *name,
262 						  struct psref *);
263 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
264 						     struct ifnet *ifp,
265 						     struct psref *);
266 static void	bridge_release_member(struct bridge_softc *, struct bridge_iflist *,
267                                       struct psref *);
268 static void	bridge_delete_member(struct bridge_softc *,
269 				     struct bridge_iflist *);
270 static void	bridge_acquire_member(struct bridge_softc *sc,
271                                       struct bridge_iflist *,
272                                       struct psref *);
273 
274 static int	bridge_ioctl_add(struct bridge_softc *, void *);
275 static int	bridge_ioctl_del(struct bridge_softc *, void *);
276 static int	bridge_ioctl_gifflags(struct bridge_softc *, void *);
277 static int	bridge_ioctl_sifflags(struct bridge_softc *, void *);
278 static int	bridge_ioctl_scache(struct bridge_softc *, void *);
279 static int	bridge_ioctl_gcache(struct bridge_softc *, void *);
280 static int	bridge_ioctl_gifs(struct bridge_softc *, void *);
281 static int	bridge_ioctl_rts(struct bridge_softc *, void *);
282 static int	bridge_ioctl_saddr(struct bridge_softc *, void *);
283 static int	bridge_ioctl_sto(struct bridge_softc *, void *);
284 static int	bridge_ioctl_gto(struct bridge_softc *, void *);
285 static int	bridge_ioctl_daddr(struct bridge_softc *, void *);
286 static int	bridge_ioctl_flush(struct bridge_softc *, void *);
287 static int	bridge_ioctl_gpri(struct bridge_softc *, void *);
288 static int	bridge_ioctl_spri(struct bridge_softc *, void *);
289 static int	bridge_ioctl_ght(struct bridge_softc *, void *);
290 static int	bridge_ioctl_sht(struct bridge_softc *, void *);
291 static int	bridge_ioctl_gfd(struct bridge_softc *, void *);
292 static int	bridge_ioctl_sfd(struct bridge_softc *, void *);
293 static int	bridge_ioctl_gma(struct bridge_softc *, void *);
294 static int	bridge_ioctl_sma(struct bridge_softc *, void *);
295 static int	bridge_ioctl_sifprio(struct bridge_softc *, void *);
296 static int	bridge_ioctl_sifcost(struct bridge_softc *, void *);
297 #if defined(BRIDGE_IPF)
298 static int	bridge_ioctl_gfilt(struct bridge_softc *, void *);
299 static int	bridge_ioctl_sfilt(struct bridge_softc *, void *);
300 static int	bridge_ipf(void *, struct mbuf **, struct ifnet *, int);
301 static int	bridge_ip_checkbasic(struct mbuf **mp);
302 # ifdef INET6
303 static int	bridge_ip6_checkbasic(struct mbuf **mp);
304 # endif /* INET6 */
305 #endif /* BRIDGE_IPF */
306 
307 struct bridge_control {
308 	int	(*bc_func)(struct bridge_softc *, void *);
309 	int	bc_argsize;
310 	int	bc_flags;
311 };
312 
313 #define	BC_F_COPYIN		0x01	/* copy arguments in */
314 #define	BC_F_COPYOUT		0x02	/* copy arguments out */
315 #define	BC_F_SUSER		0x04	/* do super-user check */
316 #define BC_F_XLATEIN		0x08	/* xlate arguments in */
317 #define BC_F_XLATEOUT		0x10	/* xlate arguments out */
318 
319 static const struct bridge_control bridge_control_table[] = {
320 [BRDGADD] = {bridge_ioctl_add, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
321 [BRDGDEL] = {bridge_ioctl_del, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
322 
323 [BRDGGIFFLGS] = {bridge_ioctl_gifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_COPYOUT},
324 [BRDGSIFFLGS] = {bridge_ioctl_sifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
325 
326 [BRDGSCACHE] = {bridge_ioctl_scache, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
327 [BRDGGCACHE] = {bridge_ioctl_gcache, sizeof(struct ifbrparam), BC_F_COPYOUT},
328 
329 [OBRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_COPYIN|BC_F_COPYOUT},
330 [OBRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_COPYIN|BC_F_COPYOUT},
331 
332 [BRDGSADDR] = {bridge_ioctl_saddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
333 
334 [BRDGSTO] = {bridge_ioctl_sto, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
335 [BRDGGTO] = {bridge_ioctl_gto, sizeof(struct ifbrparam), BC_F_COPYOUT},
336 
337 [BRDGDADDR] = {bridge_ioctl_daddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
338 
339 [BRDGFLUSH] = {bridge_ioctl_flush, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
340 
341 [BRDGGPRI] = {bridge_ioctl_gpri, sizeof(struct ifbrparam), BC_F_COPYOUT},
342 [BRDGSPRI] = {bridge_ioctl_spri, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
343 
344 [BRDGGHT] = {bridge_ioctl_ght, sizeof(struct ifbrparam), BC_F_COPYOUT},
345 [BRDGSHT] = {bridge_ioctl_sht, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
346 
347 [BRDGGFD] = {bridge_ioctl_gfd, sizeof(struct ifbrparam), BC_F_COPYOUT},
348 [BRDGSFD] = {bridge_ioctl_sfd, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
349 
350 [BRDGGMA] = {bridge_ioctl_gma, sizeof(struct ifbrparam), BC_F_COPYOUT},
351 [BRDGSMA] = {bridge_ioctl_sma, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
352 
353 [BRDGSIFPRIO] = {bridge_ioctl_sifprio, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
354 
355 [BRDGSIFCOST] = {bridge_ioctl_sifcost, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
356 #if defined(BRIDGE_IPF)
357 [BRDGGFILT] = {bridge_ioctl_gfilt, sizeof(struct ifbrparam), BC_F_COPYOUT},
358 [BRDGSFILT] = {bridge_ioctl_sfilt, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
359 #endif /* BRIDGE_IPF */
360 [BRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_XLATEIN|BC_F_XLATEOUT},
361 [BRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_XLATEIN|BC_F_XLATEOUT},
362 };
363 
364 static const int bridge_control_table_size = __arraycount(bridge_control_table);
365 
366 static struct if_clone bridge_cloner =
367     IF_CLONE_INITIALIZER("bridge", bridge_clone_create, bridge_clone_destroy);
368 
369 /*
370  * bridgeattach:
371  *
372  *	Pseudo-device attach routine.
373  */
374 void
375 bridgeattach(int n)
376 {
377 
378 	pool_init(&bridge_rtnode_pool, sizeof(struct bridge_rtnode),
379 	    0, 0, 0, "brtpl", NULL, IPL_NET);
380 
381 	bridge_psref_class = psref_class_create("bridge", IPL_SOFTNET);
382 
383 	if_clone_attach(&bridge_cloner);
384 }
385 
386 /*
387  * bridge_clone_create:
388  *
389  *	Create a new bridge instance.
390  */
391 static int
392 bridge_clone_create(struct if_clone *ifc, int unit)
393 {
394 	struct bridge_softc *sc;
395 	struct ifnet *ifp;
396 	int error;
397 
398 	sc = kmem_zalloc(sizeof(*sc),  KM_SLEEP);
399 	ifp = &sc->sc_if;
400 
401 	sc->sc_brtmax = BRIDGE_RTABLE_MAX;
402 	sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
403 	sc->sc_bridge_max_age = BSTP_DEFAULT_MAX_AGE;
404 	sc->sc_bridge_hello_time = BSTP_DEFAULT_HELLO_TIME;
405 	sc->sc_bridge_forward_delay = BSTP_DEFAULT_FORWARD_DELAY;
406 	sc->sc_bridge_priority = BSTP_DEFAULT_BRIDGE_PRIORITY;
407 	sc->sc_hold_time = BSTP_DEFAULT_HOLD_TIME;
408 	sc->sc_filter_flags = 0;
409 
410 	/* Initialize our routing table. */
411 	bridge_rtable_init(sc);
412 
413 	error = workqueue_create(&sc->sc_rtage_wq, "bridge_rtage",
414 	    bridge_rtage_work, sc, PRI_SOFTNET, IPL_SOFTNET, WQ_MPSAFE);
415 	if (error)
416 		panic("%s: workqueue_create %d\n", __func__, error);
417 
418 	callout_init(&sc->sc_brcallout, CALLOUT_MPSAFE);
419 	callout_init(&sc->sc_bstpcallout, CALLOUT_MPSAFE);
420 
421 	mutex_init(&sc->sc_iflist_psref.bip_lock, MUTEX_DEFAULT, IPL_NONE);
422 	PSLIST_INIT(&sc->sc_iflist_psref.bip_iflist);
423 	sc->sc_iflist_psref.bip_psz = pserialize_create();
424 
425 	if_initname(ifp, ifc->ifc_name, unit);
426 	ifp->if_softc = sc;
427 	ifp->if_extflags = IFEF_NO_LINK_STATE_CHANGE;
428 #ifdef NET_MPSAFE
429 	ifp->if_extflags |= IFEF_MPSAFE;
430 #endif
431 	ifp->if_mtu = ETHERMTU;
432 	ifp->if_ioctl = bridge_ioctl;
433 	ifp->if_output = bridge_output;
434 	ifp->if_start = bridge_start;
435 	ifp->if_stop = bridge_stop;
436 	ifp->if_init = bridge_init;
437 	ifp->if_type = IFT_BRIDGE;
438 	ifp->if_addrlen = 0;
439 	ifp->if_dlt = DLT_EN10MB;
440 	ifp->if_hdrlen = ETHER_HDR_LEN;
441 
442 	error = if_initialize(ifp);
443 	if (error != 0) {
444 		pserialize_destroy(sc->sc_iflist_psref.bip_psz);
445 		mutex_destroy(&sc->sc_iflist_psref.bip_lock);
446 		callout_destroy(&sc->sc_brcallout);
447 		callout_destroy(&sc->sc_bstpcallout);
448 		workqueue_destroy(sc->sc_rtage_wq);
449 		bridge_rtable_fini(sc);
450 		kmem_free(sc, sizeof(*sc));
451 
452 		return error;
453 	}
454 	if_register(ifp);
455 
456 	if_alloc_sadl(ifp);
457 
458 	return 0;
459 }
460 
461 /*
462  * bridge_clone_destroy:
463  *
464  *	Destroy a bridge instance.
465  */
466 static int
467 bridge_clone_destroy(struct ifnet *ifp)
468 {
469 	struct bridge_softc *sc = ifp->if_softc;
470 	struct bridge_iflist *bif;
471 
472 	if ((ifp->if_flags & IFF_RUNNING) != 0)
473 		bridge_stop(ifp, 1);
474 
475 	BRIDGE_LOCK(sc);
476 	for (;;) {
477 		bif = PSLIST_WRITER_FIRST(&sc->sc_iflist_psref.bip_iflist, struct bridge_iflist,
478 		    bif_next);
479 		if (bif == NULL)
480 			break;
481 		bridge_delete_member(sc, bif);
482 	}
483 	PSLIST_DESTROY(&sc->sc_iflist_psref.bip_iflist);
484 	BRIDGE_UNLOCK(sc);
485 
486 	if_detach(ifp);
487 
488 	/* Tear down the routing table. */
489 	bridge_rtable_fini(sc);
490 
491 	pserialize_destroy(sc->sc_iflist_psref.bip_psz);
492 	mutex_destroy(&sc->sc_iflist_psref.bip_lock);
493 	callout_destroy(&sc->sc_brcallout);
494 	callout_destroy(&sc->sc_bstpcallout);
495 	workqueue_destroy(sc->sc_rtage_wq);
496 	kmem_free(sc, sizeof(*sc));
497 
498 	return 0;
499 }
500 
501 /*
502  * bridge_ioctl:
503  *
504  *	Handle a control request from the operator.
505  */
506 static int
507 bridge_ioctl(struct ifnet *ifp, u_long cmd, void *data)
508 {
509 	struct bridge_softc *sc = ifp->if_softc;
510 	struct lwp *l = curlwp;	/* XXX */
511 	union {
512 		struct ifbreq ifbreq;
513 		struct ifbifconf ifbifconf;
514 		struct ifbareq ifbareq;
515 		struct ifbaconf ifbaconf;
516 		struct ifbrparam ifbrparam;
517 	} args;
518 	struct ifdrv *ifd = (struct ifdrv *) data;
519 	const struct bridge_control *bc = NULL; /* XXXGCC */
520 	int s, error = 0;
521 
522 	/* Authorize command before calling splsoftnet(). */
523 	switch (cmd) {
524 	case SIOCGDRVSPEC:
525 	case SIOCSDRVSPEC:
526 		if (ifd->ifd_cmd >= bridge_control_table_size
527 		    || (bc = &bridge_control_table[ifd->ifd_cmd]) == NULL) {
528 			error = EINVAL;
529 			return error;
530 		}
531 
532 		/* We only care about BC_F_SUSER at this point. */
533 		if ((bc->bc_flags & BC_F_SUSER) == 0)
534 			break;
535 
536 		error = kauth_authorize_network(l->l_cred,
537 		    KAUTH_NETWORK_INTERFACE_BRIDGE,
538 		    cmd == SIOCGDRVSPEC ?
539 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_GETPRIV :
540 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_SETPRIV,
541 		     ifd, NULL, NULL);
542 		if (error)
543 			return error;
544 
545 		break;
546 	}
547 
548 	s = splsoftnet();
549 
550 	switch (cmd) {
551 	case SIOCGDRVSPEC:
552 	case SIOCSDRVSPEC:
553 		KASSERT(bc != NULL);
554 		if (cmd == SIOCGDRVSPEC &&
555 		    (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) == 0) {
556 			error = EINVAL;
557 			break;
558 		}
559 		else if (cmd == SIOCSDRVSPEC &&
560 		    (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) != 0) {
561 			error = EINVAL;
562 			break;
563 		}
564 
565 		/* BC_F_SUSER is checked above, before splsoftnet(). */
566 
567 		if ((bc->bc_flags & (BC_F_XLATEIN|BC_F_XLATEOUT)) == 0
568 		    && (ifd->ifd_len != bc->bc_argsize
569 			|| ifd->ifd_len > sizeof(args))) {
570 			error = EINVAL;
571 			break;
572 		}
573 
574 		memset(&args, 0, sizeof(args));
575 		if (bc->bc_flags & BC_F_COPYIN) {
576 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
577 			if (error)
578 				break;
579 		} else if (bc->bc_flags & BC_F_XLATEIN) {
580 			args.ifbifconf.ifbic_len = ifd->ifd_len;
581 			args.ifbifconf.ifbic_buf = ifd->ifd_data;
582 		}
583 
584 		error = (*bc->bc_func)(sc, &args);
585 		if (error)
586 			break;
587 
588 		if (bc->bc_flags & BC_F_COPYOUT) {
589 			error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
590 		} else if (bc->bc_flags & BC_F_XLATEOUT) {
591 			ifd->ifd_len = args.ifbifconf.ifbic_len;
592 			ifd->ifd_data = args.ifbifconf.ifbic_buf;
593 		}
594 		break;
595 
596 	case SIOCSIFFLAGS:
597 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
598 			break;
599 		switch (ifp->if_flags & (IFF_UP|IFF_RUNNING)) {
600 		case IFF_RUNNING:
601 			/*
602 			 * If interface is marked down and it is running,
603 			 * then stop and disable it.
604 			 */
605 			(*ifp->if_stop)(ifp, 1);
606 			break;
607 		case IFF_UP:
608 			/*
609 			 * If interface is marked up and it is stopped, then
610 			 * start it.
611 			 */
612 			error = (*ifp->if_init)(ifp);
613 			break;
614 		default:
615 			break;
616 		}
617 		break;
618 
619 	case SIOCSIFMTU:
620 		if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET)
621 			error = 0;
622 		break;
623 
624 	default:
625 		error = ifioctl_common(ifp, cmd, data);
626 		break;
627 	}
628 
629 	splx(s);
630 
631 	return error;
632 }
633 
634 /*
635  * bridge_lookup_member:
636  *
637  *	Lookup a bridge member interface.
638  */
639 static struct bridge_iflist *
640 bridge_lookup_member(struct bridge_softc *sc, const char *name, struct psref *psref)
641 {
642 	struct bridge_iflist *bif;
643 	struct ifnet *ifp;
644 	int s;
645 
646 	BRIDGE_PSZ_RENTER(s);
647 
648 	BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
649 		ifp = bif->bif_ifp;
650 		if (strcmp(ifp->if_xname, name) == 0)
651 			break;
652 	}
653 	if (bif != NULL)
654 		bridge_acquire_member(sc, bif, psref);
655 
656 	BRIDGE_PSZ_REXIT(s);
657 
658 	return bif;
659 }
660 
661 /*
662  * bridge_lookup_member_if:
663  *
664  *	Lookup a bridge member interface by ifnet*.
665  */
666 static struct bridge_iflist *
667 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp,
668     struct psref *psref)
669 {
670 	struct bridge_iflist *bif;
671 	int s;
672 
673 	BRIDGE_PSZ_RENTER(s);
674 
675 	bif = member_ifp->if_bridgeif;
676 	if (bif != NULL) {
677 		psref_acquire(psref, &bif->bif_psref,
678 		    bridge_psref_class);
679 	}
680 
681 	BRIDGE_PSZ_REXIT(s);
682 
683 	return bif;
684 }
685 
686 static void
687 bridge_acquire_member(struct bridge_softc *sc, struct bridge_iflist *bif,
688     struct psref *psref)
689 {
690 
691 	psref_acquire(psref, &bif->bif_psref, bridge_psref_class);
692 }
693 
694 /*
695  * bridge_release_member:
696  *
697  *	Release the specified member interface.
698  */
699 static void
700 bridge_release_member(struct bridge_softc *sc, struct bridge_iflist *bif,
701     struct psref *psref)
702 {
703 
704 	psref_release(psref, &bif->bif_psref, bridge_psref_class);
705 }
706 
707 /*
708  * bridge_delete_member:
709  *
710  *	Delete the specified member interface.
711  */
712 static void
713 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif)
714 {
715 	struct ifnet *ifs = bif->bif_ifp;
716 
717 	KASSERT(BRIDGE_LOCKED(sc));
718 
719 	ifs->_if_input = ether_input;
720 	ifs->if_bridge = NULL;
721 	ifs->if_bridgeif = NULL;
722 
723 	PSLIST_WRITER_REMOVE(bif, bif_next);
724 	BRIDGE_PSZ_PERFORM(sc);
725 	BRIDGE_UNLOCK(sc);
726 
727 	psref_target_destroy(&bif->bif_psref, bridge_psref_class);
728 
729 	PSLIST_ENTRY_DESTROY(bif, bif_next);
730 	kmem_free(bif, sizeof(*bif));
731 
732 	BRIDGE_LOCK(sc);
733 }
734 
735 static int
736 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
737 {
738 	struct ifbreq *req = arg;
739 	struct bridge_iflist *bif = NULL;
740 	struct ifnet *ifs;
741 	int error = 0;
742 	struct psref psref;
743 
744 	ifs = if_get(req->ifbr_ifsname, &psref);
745 	if (ifs == NULL)
746 		return ENOENT;
747 
748 	if (ifs->if_bridge == sc) {
749 		error = EEXIST;
750 		goto out;
751 	}
752 
753 	if (ifs->if_bridge != NULL) {
754 		error = EBUSY;
755 		goto out;
756 	}
757 
758 	if (ifs->_if_input != ether_input) {
759 		error = EINVAL;
760 		goto out;
761 	}
762 
763 	/* FIXME: doesn't work with non-IFF_SIMPLEX interfaces */
764 	if ((ifs->if_flags & IFF_SIMPLEX) == 0) {
765 		error = EINVAL;
766 		goto out;
767 	}
768 
769 	bif = kmem_alloc(sizeof(*bif), KM_SLEEP);
770 
771 	switch (ifs->if_type) {
772 	case IFT_ETHER:
773 		if (sc->sc_if.if_mtu != ifs->if_mtu) {
774 			error = EINVAL;
775 			goto out;
776 		}
777 		/* FALLTHROUGH */
778 	case IFT_L2TP:
779 		IFNET_LOCK(ifs);
780 		error = ether_enable_vlan_mtu(ifs);
781 		IFNET_UNLOCK(ifs);
782 		if (error > 0)
783 			goto out;
784 		/*
785 		 * Place the interface into promiscuous mode.
786 		 */
787 		error = ifpromisc(ifs, 1);
788 		if (error)
789 			goto out;
790 		break;
791 	default:
792 		error = EINVAL;
793 		goto out;
794 	}
795 
796 	bif->bif_ifp = ifs;
797 	bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
798 	bif->bif_priority = BSTP_DEFAULT_PORT_PRIORITY;
799 	bif->bif_path_cost = BSTP_DEFAULT_PATH_COST;
800 	PSLIST_ENTRY_INIT(bif, bif_next);
801 	psref_target_init(&bif->bif_psref, bridge_psref_class);
802 
803 	BRIDGE_LOCK(sc);
804 
805 	ifs->if_bridge = sc;
806 	ifs->if_bridgeif = bif;
807 	PSLIST_WRITER_INSERT_HEAD(&sc->sc_iflist_psref.bip_iflist, bif, bif_next);
808 	ifs->_if_input = bridge_input;
809 
810 	BRIDGE_UNLOCK(sc);
811 
812 	if (sc->sc_if.if_flags & IFF_RUNNING)
813 		bstp_initialization(sc);
814 	else
815 		bstp_stop(sc);
816 
817  out:
818 	if_put(ifs, &psref);
819 	if (error) {
820 		if (bif != NULL)
821 			kmem_free(bif, sizeof(*bif));
822 	}
823 	return error;
824 }
825 
826 static int
827 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
828 {
829 	struct ifbreq *req = arg;
830 	const char *name = req->ifbr_ifsname;
831 	struct bridge_iflist *bif;
832 	struct ifnet *ifs;
833 
834 	BRIDGE_LOCK(sc);
835 
836 	/*
837 	 * Don't use bridge_lookup_member. We want to get a member
838 	 * with bif_refs == 0.
839 	 */
840 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) {
841 		ifs = bif->bif_ifp;
842 		if (strcmp(ifs->if_xname, name) == 0)
843 			break;
844 	}
845 
846 	if (bif == NULL) {
847 		BRIDGE_UNLOCK(sc);
848 		return ENOENT;
849 	}
850 
851 	bridge_delete_member(sc, bif);
852 
853 	BRIDGE_UNLOCK(sc);
854 
855 	switch (ifs->if_type) {
856 	case IFT_ETHER:
857 	case IFT_L2TP:
858 		/*
859 		 * Take the interface out of promiscuous mode.
860 		 * Don't call it with holding a spin lock.
861 		 */
862 		(void) ifpromisc(ifs, 0);
863 		IFNET_LOCK(ifs);
864 		(void) ether_disable_vlan_mtu(ifs);
865 		IFNET_UNLOCK(ifs);
866 		break;
867 	default:
868 #ifdef DIAGNOSTIC
869 		panic("bridge_delete_member: impossible");
870 #endif
871 		break;
872 	}
873 
874 	bridge_rtdelete(sc, ifs);
875 
876 	if (sc->sc_if.if_flags & IFF_RUNNING)
877 		bstp_initialization(sc);
878 
879 	return 0;
880 }
881 
882 static int
883 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
884 {
885 	struct ifbreq *req = arg;
886 	struct bridge_iflist *bif;
887 	struct psref psref;
888 
889 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
890 	if (bif == NULL)
891 		return ENOENT;
892 
893 	req->ifbr_ifsflags = bif->bif_flags;
894 	req->ifbr_state = bif->bif_state;
895 	req->ifbr_priority = bif->bif_priority;
896 	req->ifbr_path_cost = bif->bif_path_cost;
897 	req->ifbr_portno = bif->bif_ifp->if_index & 0xff;
898 
899 	bridge_release_member(sc, bif, &psref);
900 
901 	return 0;
902 }
903 
904 static int
905 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
906 {
907 	struct ifbreq *req = arg;
908 	struct bridge_iflist *bif;
909 	struct psref psref;
910 
911 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
912 	if (bif == NULL)
913 		return ENOENT;
914 
915 	if (req->ifbr_ifsflags & IFBIF_STP) {
916 		switch (bif->bif_ifp->if_type) {
917 		case IFT_ETHER:
918 		case IFT_L2TP:
919 			/* These can do spanning tree. */
920 			break;
921 
922 		default:
923 			/* Nothing else can. */
924 			bridge_release_member(sc, bif, &psref);
925 			return EINVAL;
926 		}
927 	}
928 
929 	bif->bif_flags = req->ifbr_ifsflags;
930 
931 	bridge_release_member(sc, bif, &psref);
932 
933 	if (sc->sc_if.if_flags & IFF_RUNNING)
934 		bstp_initialization(sc);
935 
936 	return 0;
937 }
938 
939 static int
940 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
941 {
942 	struct ifbrparam *param = arg;
943 
944 	sc->sc_brtmax = param->ifbrp_csize;
945 	bridge_rttrim(sc);
946 
947 	return 0;
948 }
949 
950 static int
951 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
952 {
953 	struct ifbrparam *param = arg;
954 
955 	param->ifbrp_csize = sc->sc_brtmax;
956 
957 	return 0;
958 }
959 
960 static int
961 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
962 {
963 	struct ifbifconf *bifc = arg;
964 	struct bridge_iflist *bif;
965 	struct ifbreq *breqs;
966 	int i, count, error = 0;
967 
968 retry:
969 	BRIDGE_LOCK(sc);
970 	count = 0;
971 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc)
972 		count++;
973 	BRIDGE_UNLOCK(sc);
974 
975 	if (count == 0) {
976 		bifc->ifbic_len = 0;
977 		return 0;
978 	}
979 
980 	if (bifc->ifbic_len == 0 || bifc->ifbic_len < (sizeof(*breqs) * count)) {
981 		/* Tell that a larger buffer is needed */
982 		bifc->ifbic_len = sizeof(*breqs) * count;
983 		return 0;
984 	}
985 
986 	breqs = kmem_alloc(sizeof(*breqs) * count, KM_SLEEP);
987 
988 	BRIDGE_LOCK(sc);
989 
990 	i = 0;
991 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc)
992 		i++;
993 	if (i > count) {
994 		/*
995 		 * The number of members has been increased.
996 		 * We need more memory!
997 		 */
998 		BRIDGE_UNLOCK(sc);
999 		kmem_free(breqs, sizeof(*breqs) * count);
1000 		goto retry;
1001 	}
1002 
1003 	i = 0;
1004 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) {
1005 		struct ifbreq *breq = &breqs[i++];
1006 		memset(breq, 0, sizeof(*breq));
1007 
1008 		strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname,
1009 		    sizeof(breq->ifbr_ifsname));
1010 		breq->ifbr_ifsflags = bif->bif_flags;
1011 		breq->ifbr_state = bif->bif_state;
1012 		breq->ifbr_priority = bif->bif_priority;
1013 		breq->ifbr_path_cost = bif->bif_path_cost;
1014 		breq->ifbr_portno = bif->bif_ifp->if_index & 0xff;
1015 	}
1016 
1017 	/* Don't call copyout with holding the mutex */
1018 	BRIDGE_UNLOCK(sc);
1019 
1020 	for (i = 0; i < count; i++) {
1021 		error = copyout(&breqs[i], bifc->ifbic_req + i, sizeof(*breqs));
1022 		if (error)
1023 			break;
1024 	}
1025 	bifc->ifbic_len = sizeof(*breqs) * i;
1026 
1027 	kmem_free(breqs, sizeof(*breqs) * count);
1028 
1029 	return error;
1030 }
1031 
1032 static int
1033 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1034 {
1035 	struct ifbaconf *bac = arg;
1036 	struct bridge_rtnode *brt;
1037 	struct ifbareq bareq;
1038 	int count = 0, error = 0, len;
1039 
1040 	if (bac->ifbac_len == 0)
1041 		return 0;
1042 
1043 	BRIDGE_RT_LOCK(sc);
1044 
1045 	len = bac->ifbac_len;
1046 	LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1047 		if (len < sizeof(bareq))
1048 			goto out;
1049 		memset(&bareq, 0, sizeof(bareq));
1050 		strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1051 		    sizeof(bareq.ifba_ifsname));
1052 		memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1053 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
1054 			bareq.ifba_expire = brt->brt_expire - time_uptime;
1055 		} else
1056 			bareq.ifba_expire = 0;
1057 		bareq.ifba_flags = brt->brt_flags;
1058 
1059 		error = copyout(&bareq, bac->ifbac_req + count, sizeof(bareq));
1060 		if (error)
1061 			goto out;
1062 		count++;
1063 		len -= sizeof(bareq);
1064 	}
1065  out:
1066 	BRIDGE_RT_UNLOCK(sc);
1067 
1068 	bac->ifbac_len = sizeof(bareq) * count;
1069 	return error;
1070 }
1071 
1072 static int
1073 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1074 {
1075 	struct ifbareq *req = arg;
1076 	struct bridge_iflist *bif;
1077 	int error;
1078 	struct psref psref;
1079 
1080 	bif = bridge_lookup_member(sc, req->ifba_ifsname, &psref);
1081 	if (bif == NULL)
1082 		return ENOENT;
1083 
1084 	error = bridge_rtupdate(sc, req->ifba_dst, bif->bif_ifp, 1,
1085 	    req->ifba_flags);
1086 
1087 	bridge_release_member(sc, bif, &psref);
1088 
1089 	return error;
1090 }
1091 
1092 static int
1093 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1094 {
1095 	struct ifbrparam *param = arg;
1096 
1097 	sc->sc_brttimeout = param->ifbrp_ctime;
1098 
1099 	return 0;
1100 }
1101 
1102 static int
1103 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1104 {
1105 	struct ifbrparam *param = arg;
1106 
1107 	param->ifbrp_ctime = sc->sc_brttimeout;
1108 
1109 	return 0;
1110 }
1111 
1112 static int
1113 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1114 {
1115 	struct ifbareq *req = arg;
1116 
1117 	return (bridge_rtdaddr(sc, req->ifba_dst));
1118 }
1119 
1120 static int
1121 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1122 {
1123 	struct ifbreq *req = arg;
1124 
1125 	bridge_rtflush(sc, req->ifbr_ifsflags);
1126 
1127 	return 0;
1128 }
1129 
1130 static int
1131 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1132 {
1133 	struct ifbrparam *param = arg;
1134 
1135 	param->ifbrp_prio = sc->sc_bridge_priority;
1136 
1137 	return 0;
1138 }
1139 
1140 static int
1141 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1142 {
1143 	struct ifbrparam *param = arg;
1144 
1145 	sc->sc_bridge_priority = param->ifbrp_prio;
1146 
1147 	if (sc->sc_if.if_flags & IFF_RUNNING)
1148 		bstp_initialization(sc);
1149 
1150 	return 0;
1151 }
1152 
1153 static int
1154 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1155 {
1156 	struct ifbrparam *param = arg;
1157 
1158 	param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8;
1159 
1160 	return 0;
1161 }
1162 
1163 static int
1164 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1165 {
1166 	struct ifbrparam *param = arg;
1167 
1168 	if (param->ifbrp_hellotime == 0)
1169 		return EINVAL;
1170 	sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8;
1171 
1172 	if (sc->sc_if.if_flags & IFF_RUNNING)
1173 		bstp_initialization(sc);
1174 
1175 	return 0;
1176 }
1177 
1178 static int
1179 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1180 {
1181 	struct ifbrparam *param = arg;
1182 
1183 	param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8;
1184 
1185 	return 0;
1186 }
1187 
1188 static int
1189 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1190 {
1191 	struct ifbrparam *param = arg;
1192 
1193 	if (param->ifbrp_fwddelay == 0)
1194 		return EINVAL;
1195 	sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8;
1196 
1197 	if (sc->sc_if.if_flags & IFF_RUNNING)
1198 		bstp_initialization(sc);
1199 
1200 	return 0;
1201 }
1202 
1203 static int
1204 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1205 {
1206 	struct ifbrparam *param = arg;
1207 
1208 	param->ifbrp_maxage = sc->sc_bridge_max_age >> 8;
1209 
1210 	return 0;
1211 }
1212 
1213 static int
1214 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1215 {
1216 	struct ifbrparam *param = arg;
1217 
1218 	if (param->ifbrp_maxage == 0)
1219 		return EINVAL;
1220 	sc->sc_bridge_max_age = param->ifbrp_maxage << 8;
1221 
1222 	if (sc->sc_if.if_flags & IFF_RUNNING)
1223 		bstp_initialization(sc);
1224 
1225 	return 0;
1226 }
1227 
1228 static int
1229 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1230 {
1231 	struct ifbreq *req = arg;
1232 	struct bridge_iflist *bif;
1233 	struct psref psref;
1234 
1235 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
1236 	if (bif == NULL)
1237 		return ENOENT;
1238 
1239 	bif->bif_priority = req->ifbr_priority;
1240 
1241 	if (sc->sc_if.if_flags & IFF_RUNNING)
1242 		bstp_initialization(sc);
1243 
1244 	bridge_release_member(sc, bif, &psref);
1245 
1246 	return 0;
1247 }
1248 
1249 #if defined(BRIDGE_IPF)
1250 static int
1251 bridge_ioctl_gfilt(struct bridge_softc *sc, void *arg)
1252 {
1253 	struct ifbrparam *param = arg;
1254 
1255 	param->ifbrp_filter = sc->sc_filter_flags;
1256 
1257 	return 0;
1258 }
1259 
1260 static int
1261 bridge_ioctl_sfilt(struct bridge_softc *sc, void *arg)
1262 {
1263 	struct ifbrparam *param = arg;
1264 	uint32_t nflags, oflags;
1265 
1266 	if (param->ifbrp_filter & ~IFBF_FILT_MASK)
1267 		return EINVAL;
1268 
1269 	nflags = param->ifbrp_filter;
1270 	oflags = sc->sc_filter_flags;
1271 
1272 	if ((nflags & IFBF_FILT_USEIPF) && !(oflags & IFBF_FILT_USEIPF)) {
1273 		pfil_add_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
1274 			sc->sc_if.if_pfil);
1275 	}
1276 	if (!(nflags & IFBF_FILT_USEIPF) && (oflags & IFBF_FILT_USEIPF)) {
1277 		pfil_remove_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
1278 			sc->sc_if.if_pfil);
1279 	}
1280 
1281 	sc->sc_filter_flags = nflags;
1282 
1283 	return 0;
1284 }
1285 #endif /* BRIDGE_IPF */
1286 
1287 static int
1288 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1289 {
1290 	struct ifbreq *req = arg;
1291 	struct bridge_iflist *bif;
1292 	struct psref psref;
1293 
1294 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
1295 	if (bif == NULL)
1296 		return ENOENT;
1297 
1298 	bif->bif_path_cost = req->ifbr_path_cost;
1299 
1300 	if (sc->sc_if.if_flags & IFF_RUNNING)
1301 		bstp_initialization(sc);
1302 
1303 	bridge_release_member(sc, bif, &psref);
1304 
1305 	return 0;
1306 }
1307 
1308 /*
1309  * bridge_ifdetach:
1310  *
1311  *	Detach an interface from a bridge.  Called when a member
1312  *	interface is detaching.
1313  */
1314 void
1315 bridge_ifdetach(struct ifnet *ifp)
1316 {
1317 	struct bridge_softc *sc = ifp->if_bridge;
1318 	struct ifbreq breq;
1319 
1320 	/* ioctl_lock should prevent this from happening */
1321 	KASSERT(sc != NULL);
1322 
1323 	memset(&breq, 0, sizeof(breq));
1324 	strlcpy(breq.ifbr_ifsname, ifp->if_xname, sizeof(breq.ifbr_ifsname));
1325 
1326 	(void) bridge_ioctl_del(sc, &breq);
1327 }
1328 
1329 /*
1330  * bridge_init:
1331  *
1332  *	Initialize a bridge interface.
1333  */
1334 static int
1335 bridge_init(struct ifnet *ifp)
1336 {
1337 	struct bridge_softc *sc = ifp->if_softc;
1338 
1339 	KASSERT((ifp->if_flags & IFF_RUNNING) == 0);
1340 
1341 	callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1342 	    bridge_timer, sc);
1343 	bstp_initialization(sc);
1344 
1345 	ifp->if_flags |= IFF_RUNNING;
1346 	return 0;
1347 }
1348 
1349 /*
1350  * bridge_stop:
1351  *
1352  *	Stop the bridge interface.
1353  */
1354 static void
1355 bridge_stop(struct ifnet *ifp, int disable)
1356 {
1357 	struct bridge_softc *sc = ifp->if_softc;
1358 
1359 	KASSERT((ifp->if_flags & IFF_RUNNING) != 0);
1360 	ifp->if_flags &= ~IFF_RUNNING;
1361 
1362 	callout_halt(&sc->sc_brcallout, NULL);
1363 	workqueue_wait(sc->sc_rtage_wq, &sc->sc_rtage_wk);
1364 	bstp_stop(sc);
1365 	bridge_rtflush(sc, IFBF_FLUSHDYN);
1366 }
1367 
1368 /*
1369  * bridge_enqueue:
1370  *
1371  *	Enqueue a packet on a bridge member interface.
1372  */
1373 void
1374 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m,
1375     int runfilt)
1376 {
1377 	int len, error;
1378 	short mflags;
1379 
1380 	/*
1381 	 * Clear any in-bound checksum flags for this packet.
1382 	 */
1383 	m->m_pkthdr.csum_flags = 0;
1384 
1385 	if (runfilt) {
1386 		if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
1387 		    dst_ifp, PFIL_OUT) != 0) {
1388 			if (m != NULL)
1389 				m_freem(m);
1390 			return;
1391 		}
1392 		if (m == NULL)
1393 			return;
1394 	}
1395 
1396 #ifdef ALTQ
1397 	KERNEL_LOCK(1, NULL);
1398 	/*
1399 	 * If ALTQ is enabled on the member interface, do
1400 	 * classification; the queueing discipline might
1401 	 * not require classification, but might require
1402 	 * the address family/header pointer in the pktattr.
1403 	 */
1404 	if (ALTQ_IS_ENABLED(&dst_ifp->if_snd)) {
1405 		/* XXX IFT_ETHER */
1406 		altq_etherclassify(&dst_ifp->if_snd, m);
1407 	}
1408 	KERNEL_UNLOCK_ONE(NULL);
1409 #endif /* ALTQ */
1410 
1411 	len = m->m_pkthdr.len;
1412 	mflags = m->m_flags;
1413 
1414 	error = if_transmit_lock(dst_ifp, m);
1415 	if (error) {
1416 		/* mbuf is already freed */
1417 		sc->sc_if.if_oerrors++;
1418 		return;
1419 	}
1420 
1421 	sc->sc_if.if_opackets++;
1422 	sc->sc_if.if_obytes += len;
1423 	if (mflags & M_MCAST)
1424 		sc->sc_if.if_omcasts++;
1425 }
1426 
1427 /*
1428  * bridge_output:
1429  *
1430  *	Send output from a bridge member interface.  This
1431  *	performs the bridging function for locally originated
1432  *	packets.
1433  *
1434  *	The mbuf has the Ethernet header already attached.  We must
1435  *	enqueue or free the mbuf before returning.
1436  */
1437 int
1438 bridge_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa,
1439     const struct rtentry *rt)
1440 {
1441 	struct ether_header *eh;
1442 	struct ifnet *dst_if;
1443 	struct bridge_softc *sc;
1444 	int s;
1445 
1446 	/*
1447 	 * bridge_output() is called from ether_output(), furthermore
1448 	 * ifp argument doesn't point to bridge(4). So, don't assert
1449 	 * IFEF_MPSAFE here.
1450 	 */
1451 
1452 	if (m->m_len < ETHER_HDR_LEN) {
1453 		m = m_pullup(m, ETHER_HDR_LEN);
1454 		if (m == NULL)
1455 			return 0;
1456 	}
1457 
1458 	eh = mtod(m, struct ether_header *);
1459 	sc = ifp->if_bridge;
1460 
1461 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1462 		if (memcmp(etherbroadcastaddr,
1463 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
1464 			m->m_flags |= M_BCAST;
1465 		else
1466 			m->m_flags |= M_MCAST;
1467 	}
1468 
1469 	/*
1470 	 * If bridge is down, but the original output interface is up,
1471 	 * go ahead and send out that interface.  Otherwise, the packet
1472 	 * is dropped below.
1473 	 */
1474 	if (__predict_false(sc == NULL) ||
1475 	    (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
1476 		dst_if = ifp;
1477 		goto sendunicast;
1478 	}
1479 
1480 	/*
1481 	 * If the packet is a multicast, or we don't know a better way to
1482 	 * get there, send to all interfaces.
1483 	 */
1484 	if ((m->m_flags & (M_MCAST | M_BCAST)) != 0)
1485 		dst_if = NULL;
1486 	else
1487 		dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1488 	if (dst_if == NULL) {
1489 		/* XXX Should call bridge_broadcast, but there are locking
1490 		 * issues which need resolving first. */
1491 		struct bridge_iflist *bif;
1492 		struct mbuf *mc;
1493 		bool used = false;
1494 
1495 		BRIDGE_PSZ_RENTER(s);
1496 		BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
1497 			struct psref psref;
1498 
1499 			bridge_acquire_member(sc, bif, &psref);
1500 			BRIDGE_PSZ_REXIT(s);
1501 
1502 			dst_if = bif->bif_ifp;
1503 			if ((dst_if->if_flags & IFF_RUNNING) == 0)
1504 				goto next;
1505 
1506 			/*
1507 			 * If this is not the original output interface,
1508 			 * and the interface is participating in spanning
1509 			 * tree, make sure the port is in a state that
1510 			 * allows forwarding.
1511 			 */
1512 			if (dst_if != ifp &&
1513 			    (bif->bif_flags & IFBIF_STP) != 0) {
1514 				switch (bif->bif_state) {
1515 				case BSTP_IFSTATE_BLOCKING:
1516 				case BSTP_IFSTATE_LISTENING:
1517 				case BSTP_IFSTATE_DISABLED:
1518 					goto next;
1519 				}
1520 			}
1521 
1522 			if (PSLIST_READER_NEXT(bif, struct bridge_iflist,
1523 			    bif_next) == NULL &&
1524 			    ((m->m_flags & (M_MCAST | M_BCAST)) == 0 ||
1525 			    dst_if == ifp))
1526 			{
1527 				used = true;
1528 				mc = m;
1529 			} else {
1530 				mc = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1531 				if (mc == NULL) {
1532 					sc->sc_if.if_oerrors++;
1533 					goto next;
1534 				}
1535 			}
1536 
1537 			bridge_enqueue(sc, dst_if, mc, 0);
1538 
1539 			if ((m->m_flags & (M_MCAST | M_BCAST)) != 0 &&
1540 			    dst_if != ifp)
1541 			{
1542 				if (PSLIST_READER_NEXT(bif,
1543 				    struct bridge_iflist, bif_next) == NULL)
1544 				{
1545 					used = true;
1546 					mc = m;
1547 				} else {
1548 					mc = m_copym(m, 0, M_COPYALL,
1549 					    M_DONTWAIT);
1550 					if (mc == NULL) {
1551 						sc->sc_if.if_oerrors++;
1552 						goto next;
1553 					}
1554 				}
1555 
1556 				m_set_rcvif(mc, dst_if);
1557 				mc->m_flags &= ~M_PROMISC;
1558 
1559 #ifndef NET_MPSAFE
1560 				s = splsoftnet();
1561 #endif
1562 				ether_input(dst_if, mc);
1563 #ifndef NET_MPSAFE
1564 				splx(s);
1565 #endif
1566 			}
1567 
1568 next:
1569 			BRIDGE_PSZ_RENTER(s);
1570 			bridge_release_member(sc, bif, &psref);
1571 
1572 			/* Guarantee we don't re-enter the loop as we already
1573 			 * decided we're at the end. */
1574 			if (used)
1575 				break;
1576 		}
1577 		BRIDGE_PSZ_REXIT(s);
1578 
1579 		if (!used)
1580 			m_freem(m);
1581 		return 0;
1582 	}
1583 
1584  sendunicast:
1585 	/*
1586 	 * XXX Spanning tree consideration here?
1587 	 */
1588 
1589 	if ((dst_if->if_flags & IFF_RUNNING) == 0) {
1590 		m_freem(m);
1591 		return 0;
1592 	}
1593 
1594 	bridge_enqueue(sc, dst_if, m, 0);
1595 
1596 	return 0;
1597 }
1598 
1599 /*
1600  * bridge_start:
1601  *
1602  *	Start output on a bridge.
1603  *
1604  *	NOTE: This routine should never be called in this implementation.
1605  */
1606 static void
1607 bridge_start(struct ifnet *ifp)
1608 {
1609 
1610 	printf("%s: bridge_start() called\n", ifp->if_xname);
1611 }
1612 
1613 /*
1614  * bridge_forward:
1615  *
1616  *	The forwarding function of the bridge.
1617  */
1618 static void
1619 bridge_forward(struct bridge_softc *sc, struct mbuf *m)
1620 {
1621 	struct bridge_iflist *bif;
1622 	struct ifnet *src_if, *dst_if;
1623 	struct ether_header *eh;
1624 	struct psref psref;
1625 	struct psref psref_src;
1626 	DECLARE_LOCK_VARIABLE;
1627 
1628 	if ((sc->sc_if.if_flags & IFF_RUNNING) == 0)
1629 		return;
1630 
1631 	src_if = m_get_rcvif_psref(m, &psref_src);
1632 	if (src_if == NULL) {
1633 		/* Interface is being destroyed? */
1634 		m_freem(m);
1635 		goto out;
1636 	}
1637 
1638 	sc->sc_if.if_ipackets++;
1639 	sc->sc_if.if_ibytes += m->m_pkthdr.len;
1640 
1641 	/*
1642 	 * Look up the bridge_iflist.
1643 	 */
1644 	bif = bridge_lookup_member_if(sc, src_if, &psref);
1645 	if (bif == NULL) {
1646 		/* Interface is not a bridge member (anymore?) */
1647 		m_freem(m);
1648 		goto out;
1649 	}
1650 
1651 	if (bif->bif_flags & IFBIF_STP) {
1652 		switch (bif->bif_state) {
1653 		case BSTP_IFSTATE_BLOCKING:
1654 		case BSTP_IFSTATE_LISTENING:
1655 		case BSTP_IFSTATE_DISABLED:
1656 			m_freem(m);
1657 			bridge_release_member(sc, bif, &psref);
1658 			goto out;
1659 		}
1660 	}
1661 
1662 	eh = mtod(m, struct ether_header *);
1663 
1664 	/*
1665 	 * If the interface is learning, and the source
1666 	 * address is valid and not multicast, record
1667 	 * the address.
1668 	 */
1669 	if ((bif->bif_flags & IFBIF_LEARNING) != 0 &&
1670 	    ETHER_IS_MULTICAST(eh->ether_shost) == 0 &&
1671 	    (eh->ether_shost[0] == 0 &&
1672 	     eh->ether_shost[1] == 0 &&
1673 	     eh->ether_shost[2] == 0 &&
1674 	     eh->ether_shost[3] == 0 &&
1675 	     eh->ether_shost[4] == 0 &&
1676 	     eh->ether_shost[5] == 0) == 0) {
1677 		(void) bridge_rtupdate(sc, eh->ether_shost,
1678 		    src_if, 0, IFBAF_DYNAMIC);
1679 	}
1680 
1681 	if ((bif->bif_flags & IFBIF_STP) != 0 &&
1682 	    bif->bif_state == BSTP_IFSTATE_LEARNING) {
1683 		m_freem(m);
1684 		bridge_release_member(sc, bif, &psref);
1685 		goto out;
1686 	}
1687 
1688 	bridge_release_member(sc, bif, &psref);
1689 
1690 	/*
1691 	 * At this point, the port either doesn't participate
1692 	 * in spanning tree or it is in the forwarding state.
1693 	 */
1694 
1695 	/*
1696 	 * If the packet is unicast, destined for someone on
1697 	 * "this" side of the bridge, drop it.
1698 	 */
1699 	if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
1700 		dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1701 		if (src_if == dst_if) {
1702 			m_freem(m);
1703 			goto out;
1704 		}
1705 	} else {
1706 		/* ...forward it to all interfaces. */
1707 		sc->sc_if.if_imcasts++;
1708 		dst_if = NULL;
1709 	}
1710 
1711 	if (pfil_run_hooks(sc->sc_if.if_pfil, &m, src_if, PFIL_IN) != 0) {
1712 		if (m != NULL)
1713 			m_freem(m);
1714 		goto out;
1715 	}
1716 	if (m == NULL)
1717 		goto out;
1718 
1719 	if (dst_if == NULL) {
1720 		bridge_broadcast(sc, src_if, m);
1721 		goto out;
1722 	}
1723 
1724 	m_put_rcvif_psref(src_if, &psref_src);
1725 	src_if = NULL;
1726 
1727 	/*
1728 	 * At this point, we're dealing with a unicast frame
1729 	 * going to a different interface.
1730 	 */
1731 	if ((dst_if->if_flags & IFF_RUNNING) == 0) {
1732 		m_freem(m);
1733 		goto out;
1734 	}
1735 
1736 	bif = bridge_lookup_member_if(sc, dst_if, &psref);
1737 	if (bif == NULL) {
1738 		/* Not a member of the bridge (anymore?) */
1739 		m_freem(m);
1740 		goto out;
1741 	}
1742 
1743 	if (bif->bif_flags & IFBIF_STP) {
1744 		switch (bif->bif_state) {
1745 		case BSTP_IFSTATE_DISABLED:
1746 		case BSTP_IFSTATE_BLOCKING:
1747 			m_freem(m);
1748 			bridge_release_member(sc, bif, &psref);
1749 			goto out;
1750 		}
1751 	}
1752 
1753 	bridge_release_member(sc, bif, &psref);
1754 
1755 	ACQUIRE_GLOBAL_LOCKS();
1756 	bridge_enqueue(sc, dst_if, m, 1);
1757 	RELEASE_GLOBAL_LOCKS();
1758 out:
1759 	if (src_if != NULL)
1760 		m_put_rcvif_psref(src_if, &psref_src);
1761 	return;
1762 }
1763 
1764 static bool
1765 bstp_state_before_learning(struct bridge_iflist *bif)
1766 {
1767 	if (bif->bif_flags & IFBIF_STP) {
1768 		switch (bif->bif_state) {
1769 		case BSTP_IFSTATE_BLOCKING:
1770 		case BSTP_IFSTATE_LISTENING:
1771 		case BSTP_IFSTATE_DISABLED:
1772 			return true;
1773 		}
1774 	}
1775 	return false;
1776 }
1777 
1778 static bool
1779 bridge_ourether(struct bridge_iflist *bif, struct ether_header *eh, int src)
1780 {
1781 	uint8_t *ether = src ? eh->ether_shost : eh->ether_dhost;
1782 
1783 	if (memcmp(CLLADDR(bif->bif_ifp->if_sadl), ether, ETHER_ADDR_LEN) == 0
1784 #if NCARP > 0
1785 	    || (bif->bif_ifp->if_carp &&
1786 	        carp_ourether(bif->bif_ifp->if_carp, eh, IFT_ETHER, src) != NULL)
1787 #endif /* NCARP > 0 */
1788 	    )
1789 		return true;
1790 
1791 	return false;
1792 }
1793 
1794 /*
1795  * bridge_input:
1796  *
1797  *	Receive input from a member interface.  Queue the packet for
1798  *	bridging if it is not for us.
1799  */
1800 static void
1801 bridge_input(struct ifnet *ifp, struct mbuf *m)
1802 {
1803 	struct bridge_softc *sc = ifp->if_bridge;
1804 	struct bridge_iflist *bif;
1805 	struct ether_header *eh;
1806 	struct psref psref;
1807 	int bound;
1808 	DECLARE_LOCK_VARIABLE;
1809 
1810 	KASSERT(!cpu_intr_p());
1811 
1812 	if (__predict_false(sc == NULL) ||
1813 	    (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
1814 		ACQUIRE_GLOBAL_LOCKS();
1815 		ether_input(ifp, m);
1816 		RELEASE_GLOBAL_LOCKS();
1817 		return;
1818 	}
1819 
1820 	bound = curlwp_bind();
1821 	bif = bridge_lookup_member_if(sc, ifp, &psref);
1822 	if (bif == NULL) {
1823 		curlwp_bindx(bound);
1824 		ACQUIRE_GLOBAL_LOCKS();
1825 		ether_input(ifp, m);
1826 		RELEASE_GLOBAL_LOCKS();
1827 		return;
1828 	}
1829 
1830 	eh = mtod(m, struct ether_header *);
1831 
1832 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1833 		if (memcmp(etherbroadcastaddr,
1834 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
1835 			m->m_flags |= M_BCAST;
1836 		else
1837 			m->m_flags |= M_MCAST;
1838 	}
1839 
1840 	/*
1841 	 * A 'fast' path for packets addressed to interfaces that are
1842 	 * part of this bridge.
1843 	 */
1844 	if (!(m->m_flags & (M_BCAST|M_MCAST)) &&
1845 	    !bstp_state_before_learning(bif)) {
1846 		struct bridge_iflist *_bif;
1847 		struct ifnet *_ifp = NULL;
1848 		int s;
1849 		struct psref _psref;
1850 
1851 		BRIDGE_PSZ_RENTER(s);
1852 		BRIDGE_IFLIST_READER_FOREACH(_bif, sc) {
1853 			/* It is destined for us. */
1854 			if (bridge_ourether(_bif, eh, 0)) {
1855 				bridge_acquire_member(sc, _bif, &_psref);
1856 				BRIDGE_PSZ_REXIT(s);
1857 				if (_bif->bif_flags & IFBIF_LEARNING)
1858 					(void) bridge_rtupdate(sc,
1859 					    eh->ether_shost, ifp, 0, IFBAF_DYNAMIC);
1860 				m_set_rcvif(m, _bif->bif_ifp);
1861 				_ifp = _bif->bif_ifp;
1862 				bridge_release_member(sc, _bif, &_psref);
1863 				goto out;
1864 			}
1865 
1866 			/* We just received a packet that we sent out. */
1867 			if (bridge_ourether(_bif, eh, 1))
1868 				break;
1869 		}
1870 		BRIDGE_PSZ_REXIT(s);
1871 out:
1872 
1873 		if (_bif != NULL) {
1874 			bridge_release_member(sc, bif, &psref);
1875 			curlwp_bindx(bound);
1876 			if (_ifp != NULL) {
1877 				m->m_flags &= ~M_PROMISC;
1878 				ACQUIRE_GLOBAL_LOCKS();
1879 				ether_input(_ifp, m);
1880 				RELEASE_GLOBAL_LOCKS();
1881 			} else
1882 				m_freem(m);
1883 			return;
1884 		}
1885 	}
1886 
1887 	/* Tap off 802.1D packets; they do not get forwarded. */
1888 	if (bif->bif_flags & IFBIF_STP &&
1889 	    memcmp(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN) == 0) {
1890 		bstp_input(sc, bif, m);
1891 		bridge_release_member(sc, bif, &psref);
1892 		curlwp_bindx(bound);
1893 		return;
1894 	}
1895 
1896 	/*
1897 	 * A normal switch would discard the packet here, but that's not what
1898 	 * we've done historically. This also prevents some obnoxious behaviour.
1899 	 */
1900 	if (bstp_state_before_learning(bif)) {
1901 		bridge_release_member(sc, bif, &psref);
1902 		curlwp_bindx(bound);
1903 		ACQUIRE_GLOBAL_LOCKS();
1904 		ether_input(ifp, m);
1905 		RELEASE_GLOBAL_LOCKS();
1906 		return;
1907 	}
1908 
1909 	bridge_release_member(sc, bif, &psref);
1910 
1911 	bridge_forward(sc, m);
1912 
1913 	curlwp_bindx(bound);
1914 }
1915 
1916 /*
1917  * bridge_broadcast:
1918  *
1919  *	Send a frame to all interfaces that are members of
1920  *	the bridge, except for the one on which the packet
1921  *	arrived.
1922  */
1923 static void
1924 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
1925     struct mbuf *m)
1926 {
1927 	struct bridge_iflist *bif;
1928 	struct mbuf *mc;
1929 	struct ifnet *dst_if;
1930 	bool bmcast;
1931 	int s;
1932 	DECLARE_LOCK_VARIABLE;
1933 
1934 	bmcast = m->m_flags & (M_BCAST|M_MCAST);
1935 
1936 	BRIDGE_PSZ_RENTER(s);
1937 	BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
1938 		struct psref psref;
1939 
1940 		bridge_acquire_member(sc, bif, &psref);
1941 		BRIDGE_PSZ_REXIT(s);
1942 
1943 		dst_if = bif->bif_ifp;
1944 
1945 		if (bif->bif_flags & IFBIF_STP) {
1946 			switch (bif->bif_state) {
1947 			case BSTP_IFSTATE_BLOCKING:
1948 			case BSTP_IFSTATE_DISABLED:
1949 				goto next;
1950 			}
1951 		}
1952 
1953 		if ((bif->bif_flags & IFBIF_DISCOVER) == 0 && !bmcast)
1954 			goto next;
1955 
1956 		if ((dst_if->if_flags & IFF_RUNNING) == 0)
1957 			goto next;
1958 
1959 		if (dst_if != src_if) {
1960 			mc = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
1961 			if (mc == NULL) {
1962 				sc->sc_if.if_oerrors++;
1963 				goto next;
1964 			}
1965 			ACQUIRE_GLOBAL_LOCKS();
1966 			bridge_enqueue(sc, dst_if, mc, 1);
1967 			RELEASE_GLOBAL_LOCKS();
1968 		}
1969 
1970 		if (bmcast) {
1971 			mc = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
1972 			if (mc == NULL) {
1973 				sc->sc_if.if_oerrors++;
1974 				goto next;
1975 			}
1976 
1977 			m_set_rcvif(mc, dst_if);
1978 			mc->m_flags &= ~M_PROMISC;
1979 
1980 			ACQUIRE_GLOBAL_LOCKS();
1981 			ether_input(dst_if, mc);
1982 			RELEASE_GLOBAL_LOCKS();
1983 		}
1984 next:
1985 		BRIDGE_PSZ_RENTER(s);
1986 		bridge_release_member(sc, bif, &psref);
1987 	}
1988 	BRIDGE_PSZ_REXIT(s);
1989 
1990 	m_freem(m);
1991 }
1992 
1993 static int
1994 bridge_rtalloc(struct bridge_softc *sc, const uint8_t *dst,
1995     struct bridge_rtnode **brtp)
1996 {
1997 	struct bridge_rtnode *brt;
1998 	int error;
1999 
2000 	if (sc->sc_brtcnt >= sc->sc_brtmax)
2001 		return ENOSPC;
2002 
2003 	/*
2004 	 * Allocate a new bridge forwarding node, and
2005 	 * initialize the expiration time and Ethernet
2006 	 * address.
2007 	 */
2008 	brt = pool_get(&bridge_rtnode_pool, PR_NOWAIT);
2009 	if (brt == NULL)
2010 		return ENOMEM;
2011 
2012 	memset(brt, 0, sizeof(*brt));
2013 	brt->brt_expire = time_uptime + sc->sc_brttimeout;
2014 	brt->brt_flags = IFBAF_DYNAMIC;
2015 	memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2016 
2017 	BRIDGE_RT_LOCK(sc);
2018 	error = bridge_rtnode_insert(sc, brt);
2019 	BRIDGE_RT_UNLOCK(sc);
2020 
2021 	if (error != 0) {
2022 		pool_put(&bridge_rtnode_pool, brt);
2023 		return error;
2024 	}
2025 
2026 	*brtp = brt;
2027 	return 0;
2028 }
2029 
2030 /*
2031  * bridge_rtupdate:
2032  *
2033  *	Add a bridge routing entry.
2034  */
2035 static int
2036 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
2037     struct ifnet *dst_if, int setflags, uint8_t flags)
2038 {
2039 	struct bridge_rtnode *brt;
2040 	int s;
2041 
2042 again:
2043 	/*
2044 	 * A route for this destination might already exist.  If so,
2045 	 * update it, otherwise create a new one.
2046 	 */
2047 	BRIDGE_RT_RENTER(s);
2048 	brt = bridge_rtnode_lookup(sc, dst);
2049 
2050 	if (brt != NULL) {
2051 		brt->brt_ifp = dst_if;
2052 		if (setflags) {
2053 			brt->brt_flags = flags;
2054 			if (flags & IFBAF_STATIC)
2055 				brt->brt_expire = 0;
2056 			else
2057 				brt->brt_expire = time_uptime + sc->sc_brttimeout;
2058 		} else {
2059 			if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2060 				brt->brt_expire = time_uptime + sc->sc_brttimeout;
2061 		}
2062 	}
2063 	BRIDGE_RT_REXIT(s);
2064 
2065 	if (brt == NULL) {
2066 		int r;
2067 
2068 		r = bridge_rtalloc(sc, dst, &brt);
2069 		if (r != 0)
2070 			return r;
2071 		goto again;
2072 	}
2073 
2074 	return 0;
2075 }
2076 
2077 /*
2078  * bridge_rtlookup:
2079  *
2080  *	Lookup the destination interface for an address.
2081  */
2082 static struct ifnet *
2083 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr)
2084 {
2085 	struct bridge_rtnode *brt;
2086 	struct ifnet *ifs = NULL;
2087 	int s;
2088 
2089 	BRIDGE_RT_RENTER(s);
2090 	brt = bridge_rtnode_lookup(sc, addr);
2091 	if (brt != NULL)
2092 		ifs = brt->brt_ifp;
2093 	BRIDGE_RT_REXIT(s);
2094 
2095 	return ifs;
2096 }
2097 
2098 typedef bool (*bridge_iterate_cb_t)
2099     (struct bridge_softc *, struct bridge_rtnode *, bool *, void *);
2100 
2101 /*
2102  * bridge_rtlist_iterate_remove:
2103  *
2104  *	It iterates on sc->sc_rtlist and removes rtnodes of it which func
2105  *	callback judges to remove. Removals of rtnodes are done in a manner
2106  *	of pserialize. To this end, all kmem_* operations are placed out of
2107  *	mutexes.
2108  */
2109 static void
2110 bridge_rtlist_iterate_remove(struct bridge_softc *sc, bridge_iterate_cb_t func, void *arg)
2111 {
2112 	struct bridge_rtnode *brt, *nbrt;
2113 	struct bridge_rtnode **brt_list;
2114 	int i, count;
2115 
2116 retry:
2117 	count = sc->sc_brtcnt;
2118 	if (count == 0)
2119 		return;
2120 	brt_list = kmem_alloc(sizeof(*brt_list) * count, KM_SLEEP);
2121 
2122 	BRIDGE_RT_LOCK(sc);
2123 	if (__predict_false(sc->sc_brtcnt > count)) {
2124 		/* The rtnodes increased, we need more memory */
2125 		BRIDGE_RT_UNLOCK(sc);
2126 		kmem_free(brt_list, sizeof(*brt_list) * count);
2127 		goto retry;
2128 	}
2129 
2130 	i = 0;
2131 	LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2132 		bool need_break = false;
2133 		if (func(sc, brt, &need_break, arg)) {
2134 			bridge_rtnode_remove(sc, brt);
2135 			brt_list[i++] = brt;
2136 		}
2137 		if (need_break)
2138 			break;
2139 	}
2140 
2141 	if (i > 0)
2142 		BRIDGE_RT_PSZ_PERFORM(sc);
2143 	BRIDGE_RT_UNLOCK(sc);
2144 
2145 	while (--i >= 0)
2146 		bridge_rtnode_destroy(brt_list[i]);
2147 
2148 	kmem_free(brt_list, sizeof(*brt_list) * count);
2149 }
2150 
2151 static bool
2152 bridge_rttrim0_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
2153     bool *need_break, void *arg)
2154 {
2155 	if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2156 		/* Take into account of the subsequent removal */
2157 		if ((sc->sc_brtcnt - 1) <= sc->sc_brtmax)
2158 			*need_break = true;
2159 		return true;
2160 	} else
2161 		return false;
2162 }
2163 
2164 static void
2165 bridge_rttrim0(struct bridge_softc *sc)
2166 {
2167 	bridge_rtlist_iterate_remove(sc, bridge_rttrim0_cb, NULL);
2168 }
2169 
2170 /*
2171  * bridge_rttrim:
2172  *
2173  *	Trim the routine table so that we have a number
2174  *	of routing entries less than or equal to the
2175  *	maximum number.
2176  */
2177 static void
2178 bridge_rttrim(struct bridge_softc *sc)
2179 {
2180 
2181 	/* Make sure we actually need to do this. */
2182 	if (sc->sc_brtcnt <= sc->sc_brtmax)
2183 		return;
2184 
2185 	/* Force an aging cycle; this might trim enough addresses. */
2186 	bridge_rtage(sc);
2187 	if (sc->sc_brtcnt <= sc->sc_brtmax)
2188 		return;
2189 
2190 	bridge_rttrim0(sc);
2191 
2192 	return;
2193 }
2194 
2195 /*
2196  * bridge_timer:
2197  *
2198  *	Aging timer for the bridge.
2199  */
2200 static void
2201 bridge_timer(void *arg)
2202 {
2203 	struct bridge_softc *sc = arg;
2204 
2205 	workqueue_enqueue(sc->sc_rtage_wq, &sc->sc_rtage_wk, NULL);
2206 }
2207 
2208 static void
2209 bridge_rtage_work(struct work *wk, void *arg)
2210 {
2211 	struct bridge_softc *sc = arg;
2212 
2213 	KASSERT(wk == &sc->sc_rtage_wk);
2214 
2215 	bridge_rtage(sc);
2216 
2217 	if (sc->sc_if.if_flags & IFF_RUNNING)
2218 		callout_reset(&sc->sc_brcallout,
2219 		    bridge_rtable_prune_period * hz, bridge_timer, sc);
2220 }
2221 
2222 static bool
2223 bridge_rtage_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
2224     bool *need_break, void *arg)
2225 {
2226 	if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2227 	    time_uptime >= brt->brt_expire)
2228 		return true;
2229 	else
2230 		return false;
2231 }
2232 
2233 /*
2234  * bridge_rtage:
2235  *
2236  *	Perform an aging cycle.
2237  */
2238 static void
2239 bridge_rtage(struct bridge_softc *sc)
2240 {
2241 	bridge_rtlist_iterate_remove(sc, bridge_rtage_cb, NULL);
2242 }
2243 
2244 
2245 static bool
2246 bridge_rtflush_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
2247     bool *need_break, void *arg)
2248 {
2249 	int full = *(int*)arg;
2250 
2251 	if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2252 		return true;
2253 	else
2254 		return false;
2255 }
2256 
2257 /*
2258  * bridge_rtflush:
2259  *
2260  *	Remove all dynamic addresses from the bridge.
2261  */
2262 static void
2263 bridge_rtflush(struct bridge_softc *sc, int full)
2264 {
2265 	bridge_rtlist_iterate_remove(sc, bridge_rtflush_cb, &full);
2266 }
2267 
2268 /*
2269  * bridge_rtdaddr:
2270  *
2271  *	Remove an address from the table.
2272  */
2273 static int
2274 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr)
2275 {
2276 	struct bridge_rtnode *brt;
2277 
2278 	BRIDGE_RT_LOCK(sc);
2279 	if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL) {
2280 		BRIDGE_RT_UNLOCK(sc);
2281 		return ENOENT;
2282 	}
2283 	bridge_rtnode_remove(sc, brt);
2284 	BRIDGE_RT_PSZ_PERFORM(sc);
2285 	BRIDGE_RT_UNLOCK(sc);
2286 
2287 	bridge_rtnode_destroy(brt);
2288 
2289 	return 0;
2290 }
2291 
2292 /*
2293  * bridge_rtdelete:
2294  *
2295  *	Delete routes to a speicifc member interface.
2296  */
2297 static void
2298 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp)
2299 {
2300 	struct bridge_rtnode *brt;
2301 
2302 	/* XXX pserialize_perform for each entry is slow */
2303 again:
2304 	BRIDGE_RT_LOCK(sc);
2305 	LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
2306 		if (brt->brt_ifp == ifp)
2307 			break;
2308 	}
2309 	if (brt == NULL) {
2310 		BRIDGE_RT_UNLOCK(sc);
2311 		return;
2312 	}
2313 	bridge_rtnode_remove(sc, brt);
2314 	BRIDGE_RT_PSZ_PERFORM(sc);
2315 	BRIDGE_RT_UNLOCK(sc);
2316 
2317 	bridge_rtnode_destroy(brt);
2318 
2319 	goto again;
2320 }
2321 
2322 /*
2323  * bridge_rtable_init:
2324  *
2325  *	Initialize the route table for this bridge.
2326  */
2327 static void
2328 bridge_rtable_init(struct bridge_softc *sc)
2329 {
2330 	int i;
2331 
2332 	sc->sc_rthash = kmem_alloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2333 	    KM_SLEEP);
2334 
2335 	for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2336 		LIST_INIT(&sc->sc_rthash[i]);
2337 
2338 	sc->sc_rthash_key = cprng_fast32();
2339 
2340 	LIST_INIT(&sc->sc_rtlist);
2341 
2342 	sc->sc_rtlist_psz = pserialize_create();
2343 	sc->sc_rtlist_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_SOFTNET);
2344 }
2345 
2346 /*
2347  * bridge_rtable_fini:
2348  *
2349  *	Deconstruct the route table for this bridge.
2350  */
2351 static void
2352 bridge_rtable_fini(struct bridge_softc *sc)
2353 {
2354 
2355 	kmem_free(sc->sc_rthash, sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE);
2356 	if (sc->sc_rtlist_lock)
2357 		mutex_obj_free(sc->sc_rtlist_lock);
2358 	if (sc->sc_rtlist_psz)
2359 		pserialize_destroy(sc->sc_rtlist_psz);
2360 }
2361 
2362 /*
2363  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2364  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2365  */
2366 #define	mix(a, b, c)							\
2367 do {									\
2368 	a -= b; a -= c; a ^= (c >> 13);					\
2369 	b -= c; b -= a; b ^= (a << 8);					\
2370 	c -= a; c -= b; c ^= (b >> 13);					\
2371 	a -= b; a -= c; a ^= (c >> 12);					\
2372 	b -= c; b -= a; b ^= (a << 16);					\
2373 	c -= a; c -= b; c ^= (b >> 5);					\
2374 	a -= b; a -= c; a ^= (c >> 3);					\
2375 	b -= c; b -= a; b ^= (a << 10);					\
2376 	c -= a; c -= b; c ^= (b >> 15);					\
2377 } while (/*CONSTCOND*/0)
2378 
2379 static inline uint32_t
2380 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2381 {
2382 	uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2383 
2384 	b += addr[5] << 8;
2385 	b += addr[4];
2386 	a += addr[3] << 24;
2387 	a += addr[2] << 16;
2388 	a += addr[1] << 8;
2389 	a += addr[0];
2390 
2391 	mix(a, b, c);
2392 
2393 	return (c & BRIDGE_RTHASH_MASK);
2394 }
2395 
2396 #undef mix
2397 
2398 /*
2399  * bridge_rtnode_lookup:
2400  *
2401  *	Look up a bridge route node for the specified destination.
2402  */
2403 static struct bridge_rtnode *
2404 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr)
2405 {
2406 	struct bridge_rtnode *brt;
2407 	uint32_t hash;
2408 	int dir;
2409 
2410 	hash = bridge_rthash(sc, addr);
2411 	LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
2412 		dir = memcmp(addr, brt->brt_addr, ETHER_ADDR_LEN);
2413 		if (dir == 0)
2414 			return brt;
2415 		if (dir > 0)
2416 			return NULL;
2417 	}
2418 
2419 	return NULL;
2420 }
2421 
2422 /*
2423  * bridge_rtnode_insert:
2424  *
2425  *	Insert the specified bridge node into the route table.  We
2426  *	assume the entry is not already in the table.
2427  */
2428 static int
2429 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2430 {
2431 	struct bridge_rtnode *lbrt;
2432 	uint32_t hash;
2433 	int dir;
2434 
2435 	KASSERT(BRIDGE_RT_LOCKED(sc));
2436 
2437 	hash = bridge_rthash(sc, brt->brt_addr);
2438 
2439 	lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
2440 	if (lbrt == NULL) {
2441 		LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
2442 		goto out;
2443 	}
2444 
2445 	do {
2446 		dir = memcmp(brt->brt_addr, lbrt->brt_addr, ETHER_ADDR_LEN);
2447 		if (dir == 0)
2448 			return EEXIST;
2449 		if (dir > 0) {
2450 			LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
2451 			goto out;
2452 		}
2453 		if (LIST_NEXT(lbrt, brt_hash) == NULL) {
2454 			LIST_INSERT_AFTER(lbrt, brt, brt_hash);
2455 			goto out;
2456 		}
2457 		lbrt = LIST_NEXT(lbrt, brt_hash);
2458 	} while (lbrt != NULL);
2459 
2460 #ifdef DIAGNOSTIC
2461 	panic("bridge_rtnode_insert: impossible");
2462 #endif
2463 
2464  out:
2465 	LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
2466 	sc->sc_brtcnt++;
2467 
2468 	return 0;
2469 }
2470 
2471 /*
2472  * bridge_rtnode_remove:
2473  *
2474  *	Remove a bridge rtnode from the rthash and the rtlist of a bridge.
2475  */
2476 static void
2477 bridge_rtnode_remove(struct bridge_softc *sc, struct bridge_rtnode *brt)
2478 {
2479 
2480 	KASSERT(BRIDGE_RT_LOCKED(sc));
2481 
2482 	LIST_REMOVE(brt, brt_hash);
2483 	LIST_REMOVE(brt, brt_list);
2484 	sc->sc_brtcnt--;
2485 }
2486 
2487 /*
2488  * bridge_rtnode_destroy:
2489  *
2490  *	Destroy a bridge rtnode.
2491  */
2492 static void
2493 bridge_rtnode_destroy(struct bridge_rtnode *brt)
2494 {
2495 
2496 	pool_put(&bridge_rtnode_pool, brt);
2497 }
2498 
2499 #if defined(BRIDGE_IPF)
2500 extern pfil_head_t *inet_pfil_hook;                 /* XXX */
2501 extern pfil_head_t *inet6_pfil_hook;                /* XXX */
2502 
2503 /*
2504  * Send bridge packets through IPF if they are one of the types IPF can deal
2505  * with, or if they are ARP or REVARP.  (IPF will pass ARP and REVARP without
2506  * question.)
2507  */
2508 static int
2509 bridge_ipf(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
2510 {
2511 	int snap, error;
2512 	struct ether_header *eh1, eh2;
2513 	struct llc llc1;
2514 	uint16_t ether_type;
2515 
2516 	snap = 0;
2517 	error = -1;	/* Default error if not error == 0 */
2518 	eh1 = mtod(*mp, struct ether_header *);
2519 	ether_type = ntohs(eh1->ether_type);
2520 
2521 	/*
2522 	 * Check for SNAP/LLC.
2523 	 */
2524 	if (ether_type < ETHERMTU) {
2525 		struct llc *llc2 = (struct llc *)(eh1 + 1);
2526 
2527 		if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
2528 		    llc2->llc_dsap == LLC_SNAP_LSAP &&
2529 		    llc2->llc_ssap == LLC_SNAP_LSAP &&
2530 		    llc2->llc_control == LLC_UI) {
2531 			ether_type = htons(llc2->llc_un.type_snap.ether_type);
2532 			snap = 1;
2533 		}
2534 	}
2535 
2536 	/*
2537 	 * If we're trying to filter bridge traffic, don't look at anything
2538 	 * other than IP and ARP traffic.  If the filter doesn't understand
2539 	 * IPv6, don't allow IPv6 through the bridge either.  This is lame
2540 	 * since if we really wanted, say, an AppleTalk filter, we are hosed,
2541 	 * but of course we don't have an AppleTalk filter to begin with.
2542 	 * (Note that since IPF doesn't understand ARP it will pass *ALL*
2543 	 * ARP traffic.)
2544 	 */
2545 	switch (ether_type) {
2546 		case ETHERTYPE_ARP:
2547 		case ETHERTYPE_REVARP:
2548 			return 0; /* Automatically pass */
2549 		case ETHERTYPE_IP:
2550 # ifdef INET6
2551 		case ETHERTYPE_IPV6:
2552 # endif /* INET6 */
2553 			break;
2554 		default:
2555 			goto bad;
2556 	}
2557 
2558 	/* Strip off the Ethernet header and keep a copy. */
2559 	m_copydata(*mp, 0, ETHER_HDR_LEN, (void *) &eh2);
2560 	m_adj(*mp, ETHER_HDR_LEN);
2561 
2562 	/* Strip off snap header, if present */
2563 	if (snap) {
2564 		m_copydata(*mp, 0, sizeof(struct llc), (void *) &llc1);
2565 		m_adj(*mp, sizeof(struct llc));
2566 	}
2567 
2568 	/*
2569 	 * Check basic packet sanity and run IPF through pfil.
2570 	 */
2571 	KASSERT(!cpu_intr_p());
2572 	switch (ether_type)
2573 	{
2574 	case ETHERTYPE_IP :
2575 		error = bridge_ip_checkbasic(mp);
2576 		if (error == 0)
2577 			error = pfil_run_hooks(inet_pfil_hook, mp, ifp, dir);
2578 		break;
2579 # ifdef INET6
2580 	case ETHERTYPE_IPV6 :
2581 		error = bridge_ip6_checkbasic(mp);
2582 		if (error == 0)
2583 			error = pfil_run_hooks(inet6_pfil_hook, mp, ifp, dir);
2584 		break;
2585 # endif
2586 	default :
2587 		error = 0;
2588 		break;
2589 	}
2590 
2591 	if (*mp == NULL)
2592 		return error;
2593 	if (error != 0)
2594 		goto bad;
2595 
2596 	error = -1;
2597 
2598 	/*
2599 	 * Finally, put everything back the way it was and return
2600 	 */
2601 	if (snap) {
2602 		M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
2603 		if (*mp == NULL)
2604 			return error;
2605 		bcopy(&llc1, mtod(*mp, void *), sizeof(struct llc));
2606 	}
2607 
2608 	M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
2609 	if (*mp == NULL)
2610 		return error;
2611 	bcopy(&eh2, mtod(*mp, void *), ETHER_HDR_LEN);
2612 
2613 	return 0;
2614 
2615     bad:
2616 	m_freem(*mp);
2617 	*mp = NULL;
2618 	return error;
2619 }
2620 
2621 /*
2622  * Perform basic checks on header size since
2623  * IPF assumes ip_input has already processed
2624  * it for it.  Cut-and-pasted from ip_input.c.
2625  * Given how simple the IPv6 version is,
2626  * does the IPv4 version really need to be
2627  * this complicated?
2628  *
2629  * XXX Should we update ipstat here, or not?
2630  * XXX Right now we update ipstat but not
2631  * XXX csum_counter.
2632  */
2633 static int
2634 bridge_ip_checkbasic(struct mbuf **mp)
2635 {
2636 	struct mbuf *m = *mp;
2637 	struct ip *ip;
2638 	int len, hlen;
2639 
2640 	if (*mp == NULL)
2641 		return -1;
2642 
2643 	if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
2644 		if ((m = m_copyup(m, sizeof(struct ip),
2645 			(max_linkhdr + 3) & ~3)) == NULL) {
2646 			/* XXXJRT new stat, please */
2647 			ip_statinc(IP_STAT_TOOSMALL);
2648 			goto bad;
2649 		}
2650 	} else if (__predict_false(m->m_len < sizeof (struct ip))) {
2651 		if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
2652 			ip_statinc(IP_STAT_TOOSMALL);
2653 			goto bad;
2654 		}
2655 	}
2656 	ip = mtod(m, struct ip *);
2657 	if (ip == NULL) goto bad;
2658 
2659 	if (ip->ip_v != IPVERSION) {
2660 		ip_statinc(IP_STAT_BADVERS);
2661 		goto bad;
2662 	}
2663 	hlen = ip->ip_hl << 2;
2664 	if (hlen < sizeof(struct ip)) { /* minimum header length */
2665 		ip_statinc(IP_STAT_BADHLEN);
2666 		goto bad;
2667 	}
2668 	if (hlen > m->m_len) {
2669 		if ((m = m_pullup(m, hlen)) == 0) {
2670 			ip_statinc(IP_STAT_BADHLEN);
2671 			goto bad;
2672 		}
2673 		ip = mtod(m, struct ip *);
2674 		if (ip == NULL) goto bad;
2675 	}
2676 
2677 	switch (m->m_pkthdr.csum_flags &
2678 	        ((m_get_rcvif_NOMPSAFE(m)->if_csum_flags_rx & M_CSUM_IPv4) |
2679 	         M_CSUM_IPv4_BAD)) {
2680 	case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
2681 		/* INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); */
2682 		goto bad;
2683 
2684 	case M_CSUM_IPv4:
2685 		/* Checksum was okay. */
2686 		/* INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); */
2687 		break;
2688 
2689 	default:
2690 		/* Must compute it ourselves. */
2691 		/* INET_CSUM_COUNTER_INCR(&ip_swcsum); */
2692 		if (in_cksum(m, hlen) != 0)
2693 			goto bad;
2694 		break;
2695 	}
2696 
2697 	/* Retrieve the packet length. */
2698 	len = ntohs(ip->ip_len);
2699 
2700 	/*
2701 	 * Check for additional length bogosity
2702 	 */
2703 	if (len < hlen) {
2704 		ip_statinc(IP_STAT_BADLEN);
2705 		goto bad;
2706 	}
2707 
2708 	/*
2709 	 * Check that the amount of data in the buffers
2710 	 * is as at least much as the IP header would have us expect.
2711 	 * Drop packet if shorter than we expect.
2712 	 */
2713 	if (m->m_pkthdr.len < len) {
2714 		ip_statinc(IP_STAT_TOOSHORT);
2715 		goto bad;
2716 	}
2717 
2718 	/* Checks out, proceed */
2719 	*mp = m;
2720 	return 0;
2721 
2722     bad:
2723 	*mp = m;
2724 	return -1;
2725 }
2726 
2727 # ifdef INET6
2728 /*
2729  * Same as above, but for IPv6.
2730  * Cut-and-pasted from ip6_input.c.
2731  * XXX Should we update ip6stat, or not?
2732  */
2733 static int
2734 bridge_ip6_checkbasic(struct mbuf **mp)
2735 {
2736 	struct mbuf *m = *mp;
2737 	struct ip6_hdr *ip6;
2738 
2739 	/*
2740 	 * If the IPv6 header is not aligned, slurp it up into a new
2741 	 * mbuf with space for link headers, in the event we forward
2742 	 * it.  Otherwise, if it is aligned, make sure the entire base
2743 	 * IPv6 header is in the first mbuf of the chain.
2744 	 */
2745 	if (IP6_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
2746 		struct ifnet *inifp = m_get_rcvif_NOMPSAFE(m);
2747 		if ((m = m_copyup(m, sizeof(struct ip6_hdr),
2748 		                  (max_linkhdr + 3) & ~3)) == NULL) {
2749 			/* XXXJRT new stat, please */
2750 			ip6_statinc(IP6_STAT_TOOSMALL);
2751 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
2752 			goto bad;
2753 		}
2754 	} else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
2755 		struct ifnet *inifp = m_get_rcvif_NOMPSAFE(m);
2756 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
2757 			ip6_statinc(IP6_STAT_TOOSMALL);
2758 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
2759 			goto bad;
2760 		}
2761 	}
2762 
2763 	ip6 = mtod(m, struct ip6_hdr *);
2764 
2765 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
2766 		ip6_statinc(IP6_STAT_BADVERS);
2767 		in6_ifstat_inc(m_get_rcvif_NOMPSAFE(m), ifs6_in_hdrerr);
2768 		goto bad;
2769 	}
2770 
2771 	/* Checks out, proceed */
2772 	*mp = m;
2773 	return 0;
2774 
2775     bad:
2776 	*mp = m;
2777 	return -1;
2778 }
2779 # endif /* INET6 */
2780 #endif /* BRIDGE_IPF */
2781