xref: /netbsd-src/sys/net/if_bridge.c (revision c38e7cc395b1472a774ff828e46123de44c628e9)
1 /*	$NetBSD: if_bridge.c,v 1.154 2018/04/18 06:37:17 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 1.154 2018/04/18 06:37:17 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_register(ifp);
474 
475 	if_alloc_sadl(ifp);
476 
477 	return 0;
478 }
479 
480 /*
481  * bridge_clone_destroy:
482  *
483  *	Destroy a bridge instance.
484  */
485 static int
486 bridge_clone_destroy(struct ifnet *ifp)
487 {
488 	struct bridge_softc *sc = ifp->if_softc;
489 	struct bridge_iflist *bif;
490 
491 	if ((ifp->if_flags & IFF_RUNNING) != 0)
492 		bridge_stop(ifp, 1);
493 
494 	BRIDGE_LOCK(sc);
495 	for (;;) {
496 		bif = PSLIST_WRITER_FIRST(&sc->sc_iflist_psref.bip_iflist, struct bridge_iflist,
497 		    bif_next);
498 		if (bif == NULL)
499 			break;
500 		bridge_delete_member(sc, bif);
501 	}
502 	PSLIST_DESTROY(&sc->sc_iflist_psref.bip_iflist);
503 	BRIDGE_UNLOCK(sc);
504 
505 	if_detach(ifp);
506 
507 	/* Tear down the routing table. */
508 	bridge_rtable_fini(sc);
509 
510 	pserialize_destroy(sc->sc_iflist_psref.bip_psz);
511 	mutex_destroy(&sc->sc_iflist_psref.bip_lock);
512 	callout_destroy(&sc->sc_brcallout);
513 	callout_destroy(&sc->sc_bstpcallout);
514 	workqueue_destroy(sc->sc_rtage_wq);
515 	kmem_free(sc, sizeof(*sc));
516 
517 	return 0;
518 }
519 
520 /*
521  * bridge_ioctl:
522  *
523  *	Handle a control request from the operator.
524  */
525 static int
526 bridge_ioctl(struct ifnet *ifp, u_long cmd, void *data)
527 {
528 	struct bridge_softc *sc = ifp->if_softc;
529 	struct lwp *l = curlwp;	/* XXX */
530 	union {
531 		struct ifbreq ifbreq;
532 		struct ifbifconf ifbifconf;
533 		struct ifbareq ifbareq;
534 		struct ifbaconf ifbaconf;
535 		struct ifbrparam ifbrparam;
536 	} args;
537 	struct ifdrv *ifd = (struct ifdrv *) data;
538 	const struct bridge_control *bc = NULL; /* XXXGCC */
539 	int s, error = 0;
540 
541 	/* Authorize command before calling splsoftnet(). */
542 	switch (cmd) {
543 	case SIOCGDRVSPEC:
544 	case SIOCSDRVSPEC:
545 		if (ifd->ifd_cmd >= bridge_control_table_size
546 		    || (bc = &bridge_control_table[ifd->ifd_cmd]) == NULL) {
547 			error = EINVAL;
548 			return error;
549 		}
550 
551 		/* We only care about BC_F_SUSER at this point. */
552 		if ((bc->bc_flags & BC_F_SUSER) == 0)
553 			break;
554 
555 		error = kauth_authorize_network(l->l_cred,
556 		    KAUTH_NETWORK_INTERFACE_BRIDGE,
557 		    cmd == SIOCGDRVSPEC ?
558 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_GETPRIV :
559 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_SETPRIV,
560 		     ifd, NULL, NULL);
561 		if (error)
562 			return error;
563 
564 		break;
565 	}
566 
567 	s = splsoftnet();
568 
569 	switch (cmd) {
570 	case SIOCGDRVSPEC:
571 	case SIOCSDRVSPEC:
572 		KASSERT(bc != NULL);
573 		if (cmd == SIOCGDRVSPEC &&
574 		    (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) == 0) {
575 			error = EINVAL;
576 			break;
577 		}
578 		else if (cmd == SIOCSDRVSPEC &&
579 		    (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) != 0) {
580 			error = EINVAL;
581 			break;
582 		}
583 
584 		/* BC_F_SUSER is checked above, before splsoftnet(). */
585 
586 		if ((bc->bc_flags & (BC_F_XLATEIN|BC_F_XLATEOUT)) == 0
587 		    && (ifd->ifd_len != bc->bc_argsize
588 			|| ifd->ifd_len > sizeof(args))) {
589 			error = EINVAL;
590 			break;
591 		}
592 
593 		memset(&args, 0, sizeof(args));
594 		if (bc->bc_flags & BC_F_COPYIN) {
595 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
596 			if (error)
597 				break;
598 		} else if (bc->bc_flags & BC_F_XLATEIN) {
599 			args.ifbifconf.ifbic_len = ifd->ifd_len;
600 			args.ifbifconf.ifbic_buf = ifd->ifd_data;
601 		}
602 
603 		error = (*bc->bc_func)(sc, &args);
604 		if (error)
605 			break;
606 
607 		if (bc->bc_flags & BC_F_COPYOUT) {
608 			error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
609 		} else if (bc->bc_flags & BC_F_XLATEOUT) {
610 			ifd->ifd_len = args.ifbifconf.ifbic_len;
611 			ifd->ifd_data = args.ifbifconf.ifbic_buf;
612 		}
613 		break;
614 
615 	case SIOCSIFFLAGS:
616 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
617 			break;
618 		switch (ifp->if_flags & (IFF_UP|IFF_RUNNING)) {
619 		case IFF_RUNNING:
620 			/*
621 			 * If interface is marked down and it is running,
622 			 * then stop and disable it.
623 			 */
624 			(*ifp->if_stop)(ifp, 1);
625 			break;
626 		case IFF_UP:
627 			/*
628 			 * If interface is marked up and it is stopped, then
629 			 * start it.
630 			 */
631 			error = (*ifp->if_init)(ifp);
632 			break;
633 		default:
634 			break;
635 		}
636 		break;
637 
638 	case SIOCSIFMTU:
639 		if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET)
640 			error = 0;
641 		break;
642 
643 	default:
644 		error = ifioctl_common(ifp, cmd, data);
645 		break;
646 	}
647 
648 	splx(s);
649 
650 	return error;
651 }
652 
653 /*
654  * bridge_lookup_member:
655  *
656  *	Lookup a bridge member interface.
657  */
658 static struct bridge_iflist *
659 bridge_lookup_member(struct bridge_softc *sc, const char *name, struct psref *psref)
660 {
661 	struct bridge_iflist *bif;
662 	struct ifnet *ifp;
663 	int s;
664 
665 	BRIDGE_PSZ_RENTER(s);
666 
667 	BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
668 		ifp = bif->bif_ifp;
669 		if (strcmp(ifp->if_xname, name) == 0)
670 			break;
671 	}
672 	if (bif != NULL)
673 		bridge_acquire_member(sc, bif, psref);
674 
675 	BRIDGE_PSZ_REXIT(s);
676 
677 	return bif;
678 }
679 
680 /*
681  * bridge_lookup_member_if:
682  *
683  *	Lookup a bridge member interface by ifnet*.
684  */
685 static struct bridge_iflist *
686 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp,
687     struct psref *psref)
688 {
689 	struct bridge_iflist *bif;
690 	int s;
691 
692 	BRIDGE_PSZ_RENTER(s);
693 
694 	bif = member_ifp->if_bridgeif;
695 	if (bif != NULL) {
696 		psref_acquire(psref, &bif->bif_psref,
697 		    bridge_psref_class);
698 	}
699 
700 	BRIDGE_PSZ_REXIT(s);
701 
702 	return bif;
703 }
704 
705 static void
706 bridge_acquire_member(struct bridge_softc *sc, struct bridge_iflist *bif,
707     struct psref *psref)
708 {
709 
710 	psref_acquire(psref, &bif->bif_psref, bridge_psref_class);
711 }
712 
713 /*
714  * bridge_release_member:
715  *
716  *	Release the specified member interface.
717  */
718 static void
719 bridge_release_member(struct bridge_softc *sc, struct bridge_iflist *bif,
720     struct psref *psref)
721 {
722 
723 	psref_release(psref, &bif->bif_psref, bridge_psref_class);
724 }
725 
726 /*
727  * bridge_delete_member:
728  *
729  *	Delete the specified member interface.
730  */
731 static void
732 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif)
733 {
734 	struct ifnet *ifs = bif->bif_ifp;
735 
736 	KASSERT(BRIDGE_LOCKED(sc));
737 
738 	ifs->_if_input = ether_input;
739 	ifs->if_bridge = NULL;
740 	ifs->if_bridgeif = NULL;
741 
742 	PSLIST_WRITER_REMOVE(bif, bif_next);
743 	BRIDGE_PSZ_PERFORM(sc);
744 	BRIDGE_UNLOCK(sc);
745 
746 	psref_target_destroy(&bif->bif_psref, bridge_psref_class);
747 
748 	PSLIST_ENTRY_DESTROY(bif, bif_next);
749 	kmem_free(bif, sizeof(*bif));
750 
751 	BRIDGE_LOCK(sc);
752 }
753 
754 static int
755 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
756 {
757 	struct ifbreq *req = arg;
758 	struct bridge_iflist *bif = NULL;
759 	struct ifnet *ifs;
760 	int error = 0;
761 	struct psref psref;
762 
763 	ifs = if_get(req->ifbr_ifsname, &psref);
764 	if (ifs == NULL)
765 		return ENOENT;
766 
767 	if (ifs->if_bridge == sc) {
768 		error = EEXIST;
769 		goto out;
770 	}
771 
772 	if (ifs->if_bridge != NULL) {
773 		error = EBUSY;
774 		goto out;
775 	}
776 
777 	if (ifs->_if_input != ether_input) {
778 		error = EINVAL;
779 		goto out;
780 	}
781 
782 	/* FIXME: doesn't work with non-IFF_SIMPLEX interfaces */
783 	if ((ifs->if_flags & IFF_SIMPLEX) == 0) {
784 		error = EINVAL;
785 		goto out;
786 	}
787 
788 	bif = kmem_alloc(sizeof(*bif), KM_SLEEP);
789 
790 	switch (ifs->if_type) {
791 	case IFT_ETHER:
792 		if (sc->sc_if.if_mtu != ifs->if_mtu) {
793 			error = EINVAL;
794 			goto out;
795 		}
796 		/* FALLTHROUGH */
797 	case IFT_L2TP:
798 		IFNET_LOCK(ifs);
799 		error = ether_enable_vlan_mtu(ifs);
800 		IFNET_UNLOCK(ifs);
801 		if (error > 0)
802 			goto out;
803 		/*
804 		 * Place the interface into promiscuous mode.
805 		 */
806 		error = ifpromisc(ifs, 1);
807 		if (error)
808 			goto out;
809 		break;
810 	default:
811 		error = EINVAL;
812 		goto out;
813 	}
814 
815 	bif->bif_ifp = ifs;
816 	bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
817 	bif->bif_priority = BSTP_DEFAULT_PORT_PRIORITY;
818 	bif->bif_path_cost = BSTP_DEFAULT_PATH_COST;
819 	PSLIST_ENTRY_INIT(bif, bif_next);
820 	psref_target_init(&bif->bif_psref, bridge_psref_class);
821 
822 	BRIDGE_LOCK(sc);
823 
824 	ifs->if_bridge = sc;
825 	ifs->if_bridgeif = bif;
826 	PSLIST_WRITER_INSERT_HEAD(&sc->sc_iflist_psref.bip_iflist, bif, bif_next);
827 	ifs->_if_input = bridge_input;
828 
829 	BRIDGE_UNLOCK(sc);
830 
831 	if (sc->sc_if.if_flags & IFF_RUNNING)
832 		bstp_initialization(sc);
833 	else
834 		bstp_stop(sc);
835 
836  out:
837 	if_put(ifs, &psref);
838 	if (error) {
839 		if (bif != NULL)
840 			kmem_free(bif, sizeof(*bif));
841 	}
842 	return error;
843 }
844 
845 static int
846 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
847 {
848 	struct ifbreq *req = arg;
849 	const char *name = req->ifbr_ifsname;
850 	struct bridge_iflist *bif;
851 	struct ifnet *ifs;
852 
853 	BRIDGE_LOCK(sc);
854 
855 	/*
856 	 * Don't use bridge_lookup_member. We want to get a member
857 	 * with bif_refs == 0.
858 	 */
859 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) {
860 		ifs = bif->bif_ifp;
861 		if (strcmp(ifs->if_xname, name) == 0)
862 			break;
863 	}
864 
865 	if (bif == NULL) {
866 		BRIDGE_UNLOCK(sc);
867 		return ENOENT;
868 	}
869 
870 	bridge_delete_member(sc, bif);
871 
872 	BRIDGE_UNLOCK(sc);
873 
874 	switch (ifs->if_type) {
875 	case IFT_ETHER:
876 	case IFT_L2TP:
877 		/*
878 		 * Take the interface out of promiscuous mode.
879 		 * Don't call it with holding a spin lock.
880 		 */
881 		(void) ifpromisc(ifs, 0);
882 		IFNET_LOCK(ifs);
883 		(void) ether_disable_vlan_mtu(ifs);
884 		IFNET_UNLOCK(ifs);
885 		break;
886 	default:
887 #ifdef DIAGNOSTIC
888 		panic("bridge_delete_member: impossible");
889 #endif
890 		break;
891 	}
892 
893 	bridge_rtdelete(sc, ifs);
894 
895 	if (sc->sc_if.if_flags & IFF_RUNNING)
896 		bstp_initialization(sc);
897 
898 	return 0;
899 }
900 
901 static int
902 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
903 {
904 	struct ifbreq *req = arg;
905 	struct bridge_iflist *bif;
906 	struct psref psref;
907 
908 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
909 	if (bif == NULL)
910 		return ENOENT;
911 
912 	req->ifbr_ifsflags = bif->bif_flags;
913 	req->ifbr_state = bif->bif_state;
914 	req->ifbr_priority = bif->bif_priority;
915 	req->ifbr_path_cost = bif->bif_path_cost;
916 	req->ifbr_portno = bif->bif_ifp->if_index & 0xff;
917 
918 	bridge_release_member(sc, bif, &psref);
919 
920 	return 0;
921 }
922 
923 static int
924 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
925 {
926 	struct ifbreq *req = arg;
927 	struct bridge_iflist *bif;
928 	struct psref psref;
929 
930 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
931 	if (bif == NULL)
932 		return ENOENT;
933 
934 	if (req->ifbr_ifsflags & IFBIF_STP) {
935 		switch (bif->bif_ifp->if_type) {
936 		case IFT_ETHER:
937 		case IFT_L2TP:
938 			/* These can do spanning tree. */
939 			break;
940 
941 		default:
942 			/* Nothing else can. */
943 			bridge_release_member(sc, bif, &psref);
944 			return EINVAL;
945 		}
946 	}
947 
948 	bif->bif_flags = req->ifbr_ifsflags;
949 
950 	bridge_release_member(sc, bif, &psref);
951 
952 	if (sc->sc_if.if_flags & IFF_RUNNING)
953 		bstp_initialization(sc);
954 
955 	return 0;
956 }
957 
958 static int
959 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
960 {
961 	struct ifbrparam *param = arg;
962 
963 	sc->sc_brtmax = param->ifbrp_csize;
964 	bridge_rttrim(sc);
965 
966 	return 0;
967 }
968 
969 static int
970 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
971 {
972 	struct ifbrparam *param = arg;
973 
974 	param->ifbrp_csize = sc->sc_brtmax;
975 
976 	return 0;
977 }
978 
979 static int
980 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
981 {
982 	struct ifbifconf *bifc = arg;
983 	struct bridge_iflist *bif;
984 	struct ifbreq *breqs;
985 	int i, count, error = 0;
986 
987 retry:
988 	BRIDGE_LOCK(sc);
989 	count = 0;
990 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc)
991 		count++;
992 	BRIDGE_UNLOCK(sc);
993 
994 	if (count == 0) {
995 		bifc->ifbic_len = 0;
996 		return 0;
997 	}
998 
999 	if (bifc->ifbic_len == 0 || bifc->ifbic_len < (sizeof(*breqs) * count)) {
1000 		/* Tell that a larger buffer is needed */
1001 		bifc->ifbic_len = sizeof(*breqs) * count;
1002 		return 0;
1003 	}
1004 
1005 	breqs = kmem_alloc(sizeof(*breqs) * count, KM_SLEEP);
1006 
1007 	BRIDGE_LOCK(sc);
1008 
1009 	i = 0;
1010 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc)
1011 		i++;
1012 	if (i > count) {
1013 		/*
1014 		 * The number of members has been increased.
1015 		 * We need more memory!
1016 		 */
1017 		BRIDGE_UNLOCK(sc);
1018 		kmem_free(breqs, sizeof(*breqs) * count);
1019 		goto retry;
1020 	}
1021 
1022 	i = 0;
1023 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) {
1024 		struct ifbreq *breq = &breqs[i++];
1025 		memset(breq, 0, sizeof(*breq));
1026 
1027 		strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname,
1028 		    sizeof(breq->ifbr_ifsname));
1029 		breq->ifbr_ifsflags = bif->bif_flags;
1030 		breq->ifbr_state = bif->bif_state;
1031 		breq->ifbr_priority = bif->bif_priority;
1032 		breq->ifbr_path_cost = bif->bif_path_cost;
1033 		breq->ifbr_portno = bif->bif_ifp->if_index & 0xff;
1034 	}
1035 
1036 	/* Don't call copyout with holding the mutex */
1037 	BRIDGE_UNLOCK(sc);
1038 
1039 	for (i = 0; i < count; i++) {
1040 		error = copyout(&breqs[i], bifc->ifbic_req + i, sizeof(*breqs));
1041 		if (error)
1042 			break;
1043 	}
1044 	bifc->ifbic_len = sizeof(*breqs) * i;
1045 
1046 	kmem_free(breqs, sizeof(*breqs) * count);
1047 
1048 	return error;
1049 }
1050 
1051 static int
1052 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1053 {
1054 	struct ifbaconf *bac = arg;
1055 	struct bridge_rtnode *brt;
1056 	struct ifbareq bareq;
1057 	int count = 0, error = 0, len;
1058 
1059 	if (bac->ifbac_len == 0)
1060 		return 0;
1061 
1062 	BRIDGE_RT_LOCK(sc);
1063 
1064 	len = bac->ifbac_len;
1065 	BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) {
1066 		if (len < sizeof(bareq))
1067 			goto out;
1068 		memset(&bareq, 0, sizeof(bareq));
1069 		strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1070 		    sizeof(bareq.ifba_ifsname));
1071 		memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1072 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
1073 			bareq.ifba_expire = brt->brt_expire - time_uptime;
1074 		} else
1075 			bareq.ifba_expire = 0;
1076 		bareq.ifba_flags = brt->brt_flags;
1077 
1078 		error = copyout(&bareq, bac->ifbac_req + count, sizeof(bareq));
1079 		if (error)
1080 			goto out;
1081 		count++;
1082 		len -= sizeof(bareq);
1083 	}
1084  out:
1085 	BRIDGE_RT_UNLOCK(sc);
1086 
1087 	bac->ifbac_len = sizeof(bareq) * count;
1088 	return error;
1089 }
1090 
1091 static int
1092 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1093 {
1094 	struct ifbareq *req = arg;
1095 	struct bridge_iflist *bif;
1096 	int error;
1097 	struct psref psref;
1098 
1099 	bif = bridge_lookup_member(sc, req->ifba_ifsname, &psref);
1100 	if (bif == NULL)
1101 		return ENOENT;
1102 
1103 	error = bridge_rtupdate(sc, req->ifba_dst, bif->bif_ifp, 1,
1104 	    req->ifba_flags);
1105 
1106 	bridge_release_member(sc, bif, &psref);
1107 
1108 	return error;
1109 }
1110 
1111 static int
1112 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1113 {
1114 	struct ifbrparam *param = arg;
1115 
1116 	sc->sc_brttimeout = param->ifbrp_ctime;
1117 
1118 	return 0;
1119 }
1120 
1121 static int
1122 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1123 {
1124 	struct ifbrparam *param = arg;
1125 
1126 	param->ifbrp_ctime = sc->sc_brttimeout;
1127 
1128 	return 0;
1129 }
1130 
1131 static int
1132 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1133 {
1134 	struct ifbareq *req = arg;
1135 
1136 	return (bridge_rtdaddr(sc, req->ifba_dst));
1137 }
1138 
1139 static int
1140 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1141 {
1142 	struct ifbreq *req = arg;
1143 
1144 	bridge_rtflush(sc, req->ifbr_ifsflags);
1145 
1146 	return 0;
1147 }
1148 
1149 static int
1150 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1151 {
1152 	struct ifbrparam *param = arg;
1153 
1154 	param->ifbrp_prio = sc->sc_bridge_priority;
1155 
1156 	return 0;
1157 }
1158 
1159 static int
1160 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1161 {
1162 	struct ifbrparam *param = arg;
1163 
1164 	sc->sc_bridge_priority = param->ifbrp_prio;
1165 
1166 	if (sc->sc_if.if_flags & IFF_RUNNING)
1167 		bstp_initialization(sc);
1168 
1169 	return 0;
1170 }
1171 
1172 static int
1173 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1174 {
1175 	struct ifbrparam *param = arg;
1176 
1177 	param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8;
1178 
1179 	return 0;
1180 }
1181 
1182 static int
1183 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1184 {
1185 	struct ifbrparam *param = arg;
1186 
1187 	if (param->ifbrp_hellotime == 0)
1188 		return EINVAL;
1189 	sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8;
1190 
1191 	if (sc->sc_if.if_flags & IFF_RUNNING)
1192 		bstp_initialization(sc);
1193 
1194 	return 0;
1195 }
1196 
1197 static int
1198 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1199 {
1200 	struct ifbrparam *param = arg;
1201 
1202 	param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8;
1203 
1204 	return 0;
1205 }
1206 
1207 static int
1208 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1209 {
1210 	struct ifbrparam *param = arg;
1211 
1212 	if (param->ifbrp_fwddelay == 0)
1213 		return EINVAL;
1214 	sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8;
1215 
1216 	if (sc->sc_if.if_flags & IFF_RUNNING)
1217 		bstp_initialization(sc);
1218 
1219 	return 0;
1220 }
1221 
1222 static int
1223 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1224 {
1225 	struct ifbrparam *param = arg;
1226 
1227 	param->ifbrp_maxage = sc->sc_bridge_max_age >> 8;
1228 
1229 	return 0;
1230 }
1231 
1232 static int
1233 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1234 {
1235 	struct ifbrparam *param = arg;
1236 
1237 	if (param->ifbrp_maxage == 0)
1238 		return EINVAL;
1239 	sc->sc_bridge_max_age = param->ifbrp_maxage << 8;
1240 
1241 	if (sc->sc_if.if_flags & IFF_RUNNING)
1242 		bstp_initialization(sc);
1243 
1244 	return 0;
1245 }
1246 
1247 static int
1248 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1249 {
1250 	struct ifbreq *req = arg;
1251 	struct bridge_iflist *bif;
1252 	struct psref psref;
1253 
1254 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
1255 	if (bif == NULL)
1256 		return ENOENT;
1257 
1258 	bif->bif_priority = req->ifbr_priority;
1259 
1260 	if (sc->sc_if.if_flags & IFF_RUNNING)
1261 		bstp_initialization(sc);
1262 
1263 	bridge_release_member(sc, bif, &psref);
1264 
1265 	return 0;
1266 }
1267 
1268 #if defined(BRIDGE_IPF)
1269 static int
1270 bridge_ioctl_gfilt(struct bridge_softc *sc, void *arg)
1271 {
1272 	struct ifbrparam *param = arg;
1273 
1274 	param->ifbrp_filter = sc->sc_filter_flags;
1275 
1276 	return 0;
1277 }
1278 
1279 static int
1280 bridge_ioctl_sfilt(struct bridge_softc *sc, void *arg)
1281 {
1282 	struct ifbrparam *param = arg;
1283 	uint32_t nflags, oflags;
1284 
1285 	if (param->ifbrp_filter & ~IFBF_FILT_MASK)
1286 		return EINVAL;
1287 
1288 	nflags = param->ifbrp_filter;
1289 	oflags = sc->sc_filter_flags;
1290 
1291 	if ((nflags & IFBF_FILT_USEIPF) && !(oflags & IFBF_FILT_USEIPF)) {
1292 		pfil_add_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
1293 			sc->sc_if.if_pfil);
1294 	}
1295 	if (!(nflags & IFBF_FILT_USEIPF) && (oflags & IFBF_FILT_USEIPF)) {
1296 		pfil_remove_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
1297 			sc->sc_if.if_pfil);
1298 	}
1299 
1300 	sc->sc_filter_flags = nflags;
1301 
1302 	return 0;
1303 }
1304 #endif /* BRIDGE_IPF */
1305 
1306 static int
1307 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1308 {
1309 	struct ifbreq *req = arg;
1310 	struct bridge_iflist *bif;
1311 	struct psref psref;
1312 
1313 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
1314 	if (bif == NULL)
1315 		return ENOENT;
1316 
1317 	bif->bif_path_cost = req->ifbr_path_cost;
1318 
1319 	if (sc->sc_if.if_flags & IFF_RUNNING)
1320 		bstp_initialization(sc);
1321 
1322 	bridge_release_member(sc, bif, &psref);
1323 
1324 	return 0;
1325 }
1326 
1327 /*
1328  * bridge_ifdetach:
1329  *
1330  *	Detach an interface from a bridge.  Called when a member
1331  *	interface is detaching.
1332  */
1333 void
1334 bridge_ifdetach(struct ifnet *ifp)
1335 {
1336 	struct bridge_softc *sc = ifp->if_bridge;
1337 	struct ifbreq breq;
1338 
1339 	/* ioctl_lock should prevent this from happening */
1340 	KASSERT(sc != NULL);
1341 
1342 	memset(&breq, 0, sizeof(breq));
1343 	strlcpy(breq.ifbr_ifsname, ifp->if_xname, sizeof(breq.ifbr_ifsname));
1344 
1345 	(void) bridge_ioctl_del(sc, &breq);
1346 }
1347 
1348 /*
1349  * bridge_init:
1350  *
1351  *	Initialize a bridge interface.
1352  */
1353 static int
1354 bridge_init(struct ifnet *ifp)
1355 {
1356 	struct bridge_softc *sc = ifp->if_softc;
1357 
1358 	KASSERT((ifp->if_flags & IFF_RUNNING) == 0);
1359 
1360 	callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1361 	    bridge_timer, sc);
1362 	bstp_initialization(sc);
1363 
1364 	ifp->if_flags |= IFF_RUNNING;
1365 	return 0;
1366 }
1367 
1368 /*
1369  * bridge_stop:
1370  *
1371  *	Stop the bridge interface.
1372  */
1373 static void
1374 bridge_stop(struct ifnet *ifp, int disable)
1375 {
1376 	struct bridge_softc *sc = ifp->if_softc;
1377 
1378 	KASSERT((ifp->if_flags & IFF_RUNNING) != 0);
1379 	ifp->if_flags &= ~IFF_RUNNING;
1380 
1381 	callout_halt(&sc->sc_brcallout, NULL);
1382 	workqueue_wait(sc->sc_rtage_wq, &sc->sc_rtage_wk);
1383 	bstp_stop(sc);
1384 	bridge_rtflush(sc, IFBF_FLUSHDYN);
1385 }
1386 
1387 /*
1388  * bridge_enqueue:
1389  *
1390  *	Enqueue a packet on a bridge member interface.
1391  */
1392 void
1393 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m,
1394     int runfilt)
1395 {
1396 	int len, error;
1397 	short mflags;
1398 
1399 	/*
1400 	 * Clear any in-bound checksum flags for this packet.
1401 	 */
1402 	m->m_pkthdr.csum_flags = 0;
1403 
1404 	if (runfilt) {
1405 		if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
1406 		    dst_ifp, PFIL_OUT) != 0) {
1407 			if (m != NULL)
1408 				m_freem(m);
1409 			return;
1410 		}
1411 		if (m == NULL)
1412 			return;
1413 	}
1414 
1415 #ifdef ALTQ
1416 	KERNEL_LOCK(1, NULL);
1417 	/*
1418 	 * If ALTQ is enabled on the member interface, do
1419 	 * classification; the queueing discipline might
1420 	 * not require classification, but might require
1421 	 * the address family/header pointer in the pktattr.
1422 	 */
1423 	if (ALTQ_IS_ENABLED(&dst_ifp->if_snd)) {
1424 		/* XXX IFT_ETHER */
1425 		altq_etherclassify(&dst_ifp->if_snd, m);
1426 	}
1427 	KERNEL_UNLOCK_ONE(NULL);
1428 #endif /* ALTQ */
1429 
1430 	len = m->m_pkthdr.len;
1431 	mflags = m->m_flags;
1432 
1433 	error = if_transmit_lock(dst_ifp, m);
1434 	if (error) {
1435 		/* mbuf is already freed */
1436 		sc->sc_if.if_oerrors++;
1437 		return;
1438 	}
1439 
1440 	sc->sc_if.if_opackets++;
1441 	sc->sc_if.if_obytes += len;
1442 	if (mflags & M_MCAST)
1443 		sc->sc_if.if_omcasts++;
1444 }
1445 
1446 /*
1447  * bridge_output:
1448  *
1449  *	Send output from a bridge member interface.  This
1450  *	performs the bridging function for locally originated
1451  *	packets.
1452  *
1453  *	The mbuf has the Ethernet header already attached.  We must
1454  *	enqueue or free the mbuf before returning.
1455  */
1456 int
1457 bridge_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa,
1458     const struct rtentry *rt)
1459 {
1460 	struct ether_header *eh;
1461 	struct ifnet *dst_if;
1462 	struct bridge_softc *sc;
1463 	int s;
1464 
1465 	/*
1466 	 * bridge_output() is called from ether_output(), furthermore
1467 	 * ifp argument doesn't point to bridge(4). So, don't assert
1468 	 * IFEF_MPSAFE here.
1469 	 */
1470 
1471 	if (m->m_len < ETHER_HDR_LEN) {
1472 		m = m_pullup(m, ETHER_HDR_LEN);
1473 		if (m == NULL)
1474 			return 0;
1475 	}
1476 
1477 	eh = mtod(m, struct ether_header *);
1478 	sc = ifp->if_bridge;
1479 
1480 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1481 		if (memcmp(etherbroadcastaddr,
1482 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
1483 			m->m_flags |= M_BCAST;
1484 		else
1485 			m->m_flags |= M_MCAST;
1486 	}
1487 
1488 	/*
1489 	 * If bridge is down, but the original output interface is up,
1490 	 * go ahead and send out that interface.  Otherwise, the packet
1491 	 * is dropped below.
1492 	 */
1493 	if (__predict_false(sc == NULL) ||
1494 	    (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
1495 		dst_if = ifp;
1496 		goto sendunicast;
1497 	}
1498 
1499 	/*
1500 	 * If the packet is a multicast, or we don't know a better way to
1501 	 * get there, send to all interfaces.
1502 	 */
1503 	if ((m->m_flags & (M_MCAST | M_BCAST)) != 0)
1504 		dst_if = NULL;
1505 	else
1506 		dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1507 	if (dst_if == NULL) {
1508 		/* XXX Should call bridge_broadcast, but there are locking
1509 		 * issues which need resolving first. */
1510 		struct bridge_iflist *bif;
1511 		struct mbuf *mc;
1512 		bool used = false;
1513 
1514 		BRIDGE_PSZ_RENTER(s);
1515 		BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
1516 			struct psref psref;
1517 
1518 			bridge_acquire_member(sc, bif, &psref);
1519 			BRIDGE_PSZ_REXIT(s);
1520 
1521 			dst_if = bif->bif_ifp;
1522 			if ((dst_if->if_flags & IFF_RUNNING) == 0)
1523 				goto next;
1524 
1525 			/*
1526 			 * If this is not the original output interface,
1527 			 * and the interface is participating in spanning
1528 			 * tree, make sure the port is in a state that
1529 			 * allows forwarding.
1530 			 */
1531 			if (dst_if != ifp &&
1532 			    (bif->bif_flags & IFBIF_STP) != 0) {
1533 				switch (bif->bif_state) {
1534 				case BSTP_IFSTATE_BLOCKING:
1535 				case BSTP_IFSTATE_LISTENING:
1536 				case BSTP_IFSTATE_DISABLED:
1537 					goto next;
1538 				}
1539 			}
1540 
1541 			if (PSLIST_READER_NEXT(bif, struct bridge_iflist,
1542 			    bif_next) == NULL &&
1543 			    ((m->m_flags & (M_MCAST | M_BCAST)) == 0 ||
1544 			    dst_if == ifp))
1545 			{
1546 				used = true;
1547 				mc = m;
1548 			} else {
1549 				mc = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1550 				if (mc == NULL) {
1551 					sc->sc_if.if_oerrors++;
1552 					goto next;
1553 				}
1554 			}
1555 
1556 			bridge_enqueue(sc, dst_if, mc, 0);
1557 
1558 			if ((m->m_flags & (M_MCAST | M_BCAST)) != 0 &&
1559 			    dst_if != ifp)
1560 			{
1561 				if (PSLIST_READER_NEXT(bif,
1562 				    struct bridge_iflist, bif_next) == NULL)
1563 				{
1564 					used = true;
1565 					mc = m;
1566 				} else {
1567 					mc = m_copym(m, 0, M_COPYALL,
1568 					    M_DONTWAIT);
1569 					if (mc == NULL) {
1570 						sc->sc_if.if_oerrors++;
1571 						goto next;
1572 					}
1573 				}
1574 
1575 				m_set_rcvif(mc, dst_if);
1576 				mc->m_flags &= ~M_PROMISC;
1577 
1578 #ifndef NET_MPSAFE
1579 				s = splsoftnet();
1580 #endif
1581 				ether_input(dst_if, mc);
1582 #ifndef NET_MPSAFE
1583 				splx(s);
1584 #endif
1585 			}
1586 
1587 next:
1588 			BRIDGE_PSZ_RENTER(s);
1589 			bridge_release_member(sc, bif, &psref);
1590 
1591 			/* Guarantee we don't re-enter the loop as we already
1592 			 * decided we're at the end. */
1593 			if (used)
1594 				break;
1595 		}
1596 		BRIDGE_PSZ_REXIT(s);
1597 
1598 		if (!used)
1599 			m_freem(m);
1600 		return 0;
1601 	}
1602 
1603  sendunicast:
1604 	/*
1605 	 * XXX Spanning tree consideration here?
1606 	 */
1607 
1608 	if ((dst_if->if_flags & IFF_RUNNING) == 0) {
1609 		m_freem(m);
1610 		return 0;
1611 	}
1612 
1613 	bridge_enqueue(sc, dst_if, m, 0);
1614 
1615 	return 0;
1616 }
1617 
1618 /*
1619  * bridge_start:
1620  *
1621  *	Start output on a bridge.
1622  *
1623  *	NOTE: This routine should never be called in this implementation.
1624  */
1625 static void
1626 bridge_start(struct ifnet *ifp)
1627 {
1628 
1629 	printf("%s: bridge_start() called\n", ifp->if_xname);
1630 }
1631 
1632 /*
1633  * bridge_forward:
1634  *
1635  *	The forwarding function of the bridge.
1636  */
1637 static void
1638 bridge_forward(struct bridge_softc *sc, struct mbuf *m)
1639 {
1640 	struct bridge_iflist *bif;
1641 	struct ifnet *src_if, *dst_if;
1642 	struct ether_header *eh;
1643 	struct psref psref;
1644 	struct psref psref_src;
1645 	DECLARE_LOCK_VARIABLE;
1646 
1647 	if ((sc->sc_if.if_flags & IFF_RUNNING) == 0)
1648 		return;
1649 
1650 	src_if = m_get_rcvif_psref(m, &psref_src);
1651 	if (src_if == NULL) {
1652 		/* Interface is being destroyed? */
1653 		m_freem(m);
1654 		goto out;
1655 	}
1656 
1657 	sc->sc_if.if_ipackets++;
1658 	sc->sc_if.if_ibytes += m->m_pkthdr.len;
1659 
1660 	/*
1661 	 * Look up the bridge_iflist.
1662 	 */
1663 	bif = bridge_lookup_member_if(sc, src_if, &psref);
1664 	if (bif == NULL) {
1665 		/* Interface is not a bridge member (anymore?) */
1666 		m_freem(m);
1667 		goto out;
1668 	}
1669 
1670 	if (bif->bif_flags & IFBIF_STP) {
1671 		switch (bif->bif_state) {
1672 		case BSTP_IFSTATE_BLOCKING:
1673 		case BSTP_IFSTATE_LISTENING:
1674 		case BSTP_IFSTATE_DISABLED:
1675 			m_freem(m);
1676 			bridge_release_member(sc, bif, &psref);
1677 			goto out;
1678 		}
1679 	}
1680 
1681 	eh = mtod(m, struct ether_header *);
1682 
1683 	/*
1684 	 * If the interface is learning, and the source
1685 	 * address is valid and not multicast, record
1686 	 * the address.
1687 	 */
1688 	if ((bif->bif_flags & IFBIF_LEARNING) != 0 &&
1689 	    ETHER_IS_MULTICAST(eh->ether_shost) == 0 &&
1690 	    (eh->ether_shost[0] == 0 &&
1691 	     eh->ether_shost[1] == 0 &&
1692 	     eh->ether_shost[2] == 0 &&
1693 	     eh->ether_shost[3] == 0 &&
1694 	     eh->ether_shost[4] == 0 &&
1695 	     eh->ether_shost[5] == 0) == 0) {
1696 		(void) bridge_rtupdate(sc, eh->ether_shost,
1697 		    src_if, 0, IFBAF_DYNAMIC);
1698 	}
1699 
1700 	if ((bif->bif_flags & IFBIF_STP) != 0 &&
1701 	    bif->bif_state == BSTP_IFSTATE_LEARNING) {
1702 		m_freem(m);
1703 		bridge_release_member(sc, bif, &psref);
1704 		goto out;
1705 	}
1706 
1707 	bridge_release_member(sc, bif, &psref);
1708 
1709 	/*
1710 	 * At this point, the port either doesn't participate
1711 	 * in spanning tree or it is in the forwarding state.
1712 	 */
1713 
1714 	/*
1715 	 * If the packet is unicast, destined for someone on
1716 	 * "this" side of the bridge, drop it.
1717 	 */
1718 	if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
1719 		dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1720 		if (src_if == dst_if) {
1721 			m_freem(m);
1722 			goto out;
1723 		}
1724 	} else {
1725 		/* ...forward it to all interfaces. */
1726 		sc->sc_if.if_imcasts++;
1727 		dst_if = NULL;
1728 	}
1729 
1730 	if (pfil_run_hooks(sc->sc_if.if_pfil, &m, src_if, PFIL_IN) != 0) {
1731 		if (m != NULL)
1732 			m_freem(m);
1733 		goto out;
1734 	}
1735 	if (m == NULL)
1736 		goto out;
1737 
1738 	if (dst_if == NULL) {
1739 		bridge_broadcast(sc, src_if, m);
1740 		goto out;
1741 	}
1742 
1743 	m_put_rcvif_psref(src_if, &psref_src);
1744 	src_if = NULL;
1745 
1746 	/*
1747 	 * At this point, we're dealing with a unicast frame
1748 	 * going to a different interface.
1749 	 */
1750 	if ((dst_if->if_flags & IFF_RUNNING) == 0) {
1751 		m_freem(m);
1752 		goto out;
1753 	}
1754 
1755 	bif = bridge_lookup_member_if(sc, dst_if, &psref);
1756 	if (bif == NULL) {
1757 		/* Not a member of the bridge (anymore?) */
1758 		m_freem(m);
1759 		goto out;
1760 	}
1761 
1762 	if (bif->bif_flags & IFBIF_STP) {
1763 		switch (bif->bif_state) {
1764 		case BSTP_IFSTATE_DISABLED:
1765 		case BSTP_IFSTATE_BLOCKING:
1766 			m_freem(m);
1767 			bridge_release_member(sc, bif, &psref);
1768 			goto out;
1769 		}
1770 	}
1771 
1772 	bridge_release_member(sc, bif, &psref);
1773 
1774 	ACQUIRE_GLOBAL_LOCKS();
1775 	bridge_enqueue(sc, dst_if, m, 1);
1776 	RELEASE_GLOBAL_LOCKS();
1777 out:
1778 	if (src_if != NULL)
1779 		m_put_rcvif_psref(src_if, &psref_src);
1780 	return;
1781 }
1782 
1783 static bool
1784 bstp_state_before_learning(struct bridge_iflist *bif)
1785 {
1786 	if (bif->bif_flags & IFBIF_STP) {
1787 		switch (bif->bif_state) {
1788 		case BSTP_IFSTATE_BLOCKING:
1789 		case BSTP_IFSTATE_LISTENING:
1790 		case BSTP_IFSTATE_DISABLED:
1791 			return true;
1792 		}
1793 	}
1794 	return false;
1795 }
1796 
1797 static bool
1798 bridge_ourether(struct bridge_iflist *bif, struct ether_header *eh, int src)
1799 {
1800 	uint8_t *ether = src ? eh->ether_shost : eh->ether_dhost;
1801 
1802 	if (memcmp(CLLADDR(bif->bif_ifp->if_sadl), ether, ETHER_ADDR_LEN) == 0
1803 #if NCARP > 0
1804 	    || (bif->bif_ifp->if_carp &&
1805 	        carp_ourether(bif->bif_ifp->if_carp, eh, IFT_ETHER, src) != NULL)
1806 #endif /* NCARP > 0 */
1807 	    )
1808 		return true;
1809 
1810 	return false;
1811 }
1812 
1813 /*
1814  * bridge_input:
1815  *
1816  *	Receive input from a member interface.  Queue the packet for
1817  *	bridging if it is not for us.
1818  */
1819 static void
1820 bridge_input(struct ifnet *ifp, struct mbuf *m)
1821 {
1822 	struct bridge_softc *sc = ifp->if_bridge;
1823 	struct bridge_iflist *bif;
1824 	struct ether_header *eh;
1825 	struct psref psref;
1826 	int bound;
1827 	DECLARE_LOCK_VARIABLE;
1828 
1829 	KASSERT(!cpu_intr_p());
1830 
1831 	if (__predict_false(sc == NULL) ||
1832 	    (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
1833 		ACQUIRE_GLOBAL_LOCKS();
1834 		ether_input(ifp, m);
1835 		RELEASE_GLOBAL_LOCKS();
1836 		return;
1837 	}
1838 
1839 	bound = curlwp_bind();
1840 	bif = bridge_lookup_member_if(sc, ifp, &psref);
1841 	if (bif == NULL) {
1842 		curlwp_bindx(bound);
1843 		ACQUIRE_GLOBAL_LOCKS();
1844 		ether_input(ifp, m);
1845 		RELEASE_GLOBAL_LOCKS();
1846 		return;
1847 	}
1848 
1849 	eh = mtod(m, struct ether_header *);
1850 
1851 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1852 		if (memcmp(etherbroadcastaddr,
1853 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
1854 			m->m_flags |= M_BCAST;
1855 		else
1856 			m->m_flags |= M_MCAST;
1857 	}
1858 
1859 	/*
1860 	 * A 'fast' path for packets addressed to interfaces that are
1861 	 * part of this bridge.
1862 	 */
1863 	if (!(m->m_flags & (M_BCAST|M_MCAST)) &&
1864 	    !bstp_state_before_learning(bif)) {
1865 		struct bridge_iflist *_bif;
1866 		struct ifnet *_ifp = NULL;
1867 		int s;
1868 		struct psref _psref;
1869 
1870 		BRIDGE_PSZ_RENTER(s);
1871 		BRIDGE_IFLIST_READER_FOREACH(_bif, sc) {
1872 			/* It is destined for us. */
1873 			if (bridge_ourether(_bif, eh, 0)) {
1874 				bridge_acquire_member(sc, _bif, &_psref);
1875 				BRIDGE_PSZ_REXIT(s);
1876 				if (_bif->bif_flags & IFBIF_LEARNING)
1877 					(void) bridge_rtupdate(sc,
1878 					    eh->ether_shost, ifp, 0, IFBAF_DYNAMIC);
1879 				m_set_rcvif(m, _bif->bif_ifp);
1880 				_ifp = _bif->bif_ifp;
1881 				bridge_release_member(sc, _bif, &_psref);
1882 				goto out;
1883 			}
1884 
1885 			/* We just received a packet that we sent out. */
1886 			if (bridge_ourether(_bif, eh, 1))
1887 				break;
1888 		}
1889 		BRIDGE_PSZ_REXIT(s);
1890 out:
1891 
1892 		if (_bif != NULL) {
1893 			bridge_release_member(sc, bif, &psref);
1894 			curlwp_bindx(bound);
1895 			if (_ifp != NULL) {
1896 				m->m_flags &= ~M_PROMISC;
1897 				ACQUIRE_GLOBAL_LOCKS();
1898 				ether_input(_ifp, m);
1899 				RELEASE_GLOBAL_LOCKS();
1900 			} else
1901 				m_freem(m);
1902 			return;
1903 		}
1904 	}
1905 
1906 	/* Tap off 802.1D packets; they do not get forwarded. */
1907 	if (bif->bif_flags & IFBIF_STP &&
1908 	    memcmp(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN) == 0) {
1909 		bstp_input(sc, bif, m);
1910 		bridge_release_member(sc, bif, &psref);
1911 		curlwp_bindx(bound);
1912 		return;
1913 	}
1914 
1915 	/*
1916 	 * A normal switch would discard the packet here, but that's not what
1917 	 * we've done historically. This also prevents some obnoxious behaviour.
1918 	 */
1919 	if (bstp_state_before_learning(bif)) {
1920 		bridge_release_member(sc, bif, &psref);
1921 		curlwp_bindx(bound);
1922 		ACQUIRE_GLOBAL_LOCKS();
1923 		ether_input(ifp, m);
1924 		RELEASE_GLOBAL_LOCKS();
1925 		return;
1926 	}
1927 
1928 	bridge_release_member(sc, bif, &psref);
1929 
1930 	bridge_forward(sc, m);
1931 
1932 	curlwp_bindx(bound);
1933 }
1934 
1935 /*
1936  * bridge_broadcast:
1937  *
1938  *	Send a frame to all interfaces that are members of
1939  *	the bridge, except for the one on which the packet
1940  *	arrived.
1941  */
1942 static void
1943 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
1944     struct mbuf *m)
1945 {
1946 	struct bridge_iflist *bif;
1947 	struct mbuf *mc;
1948 	struct ifnet *dst_if;
1949 	bool bmcast;
1950 	int s;
1951 	DECLARE_LOCK_VARIABLE;
1952 
1953 	bmcast = m->m_flags & (M_BCAST|M_MCAST);
1954 
1955 	BRIDGE_PSZ_RENTER(s);
1956 	BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
1957 		struct psref psref;
1958 
1959 		bridge_acquire_member(sc, bif, &psref);
1960 		BRIDGE_PSZ_REXIT(s);
1961 
1962 		dst_if = bif->bif_ifp;
1963 
1964 		if (bif->bif_flags & IFBIF_STP) {
1965 			switch (bif->bif_state) {
1966 			case BSTP_IFSTATE_BLOCKING:
1967 			case BSTP_IFSTATE_DISABLED:
1968 				goto next;
1969 			}
1970 		}
1971 
1972 		if ((bif->bif_flags & IFBIF_DISCOVER) == 0 && !bmcast)
1973 			goto next;
1974 
1975 		if ((dst_if->if_flags & IFF_RUNNING) == 0)
1976 			goto next;
1977 
1978 		if (dst_if != src_if) {
1979 			mc = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
1980 			if (mc == NULL) {
1981 				sc->sc_if.if_oerrors++;
1982 				goto next;
1983 			}
1984 			ACQUIRE_GLOBAL_LOCKS();
1985 			bridge_enqueue(sc, dst_if, mc, 1);
1986 			RELEASE_GLOBAL_LOCKS();
1987 		}
1988 
1989 		if (bmcast) {
1990 			mc = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
1991 			if (mc == NULL) {
1992 				sc->sc_if.if_oerrors++;
1993 				goto next;
1994 			}
1995 
1996 			m_set_rcvif(mc, dst_if);
1997 			mc->m_flags &= ~M_PROMISC;
1998 
1999 			ACQUIRE_GLOBAL_LOCKS();
2000 			ether_input(dst_if, mc);
2001 			RELEASE_GLOBAL_LOCKS();
2002 		}
2003 next:
2004 		BRIDGE_PSZ_RENTER(s);
2005 		bridge_release_member(sc, bif, &psref);
2006 	}
2007 	BRIDGE_PSZ_REXIT(s);
2008 
2009 	m_freem(m);
2010 }
2011 
2012 static int
2013 bridge_rtalloc(struct bridge_softc *sc, const uint8_t *dst,
2014     struct bridge_rtnode **brtp)
2015 {
2016 	struct bridge_rtnode *brt;
2017 	int error;
2018 
2019 	if (sc->sc_brtcnt >= sc->sc_brtmax)
2020 		return ENOSPC;
2021 
2022 	/*
2023 	 * Allocate a new bridge forwarding node, and
2024 	 * initialize the expiration time and Ethernet
2025 	 * address.
2026 	 */
2027 	brt = pool_get(&bridge_rtnode_pool, PR_NOWAIT);
2028 	if (brt == NULL)
2029 		return ENOMEM;
2030 
2031 	memset(brt, 0, sizeof(*brt));
2032 	brt->brt_expire = time_uptime + sc->sc_brttimeout;
2033 	brt->brt_flags = IFBAF_DYNAMIC;
2034 	memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2035 	PSLIST_ENTRY_INIT(brt, brt_list);
2036 	PSLIST_ENTRY_INIT(brt, brt_hash);
2037 
2038 	BRIDGE_RT_LOCK(sc);
2039 	error = bridge_rtnode_insert(sc, brt);
2040 	BRIDGE_RT_UNLOCK(sc);
2041 
2042 	if (error != 0) {
2043 		pool_put(&bridge_rtnode_pool, brt);
2044 		return error;
2045 	}
2046 
2047 	*brtp = brt;
2048 	return 0;
2049 }
2050 
2051 /*
2052  * bridge_rtupdate:
2053  *
2054  *	Add a bridge routing entry.
2055  */
2056 static int
2057 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
2058     struct ifnet *dst_if, int setflags, uint8_t flags)
2059 {
2060 	struct bridge_rtnode *brt;
2061 	int s;
2062 
2063 again:
2064 	/*
2065 	 * A route for this destination might already exist.  If so,
2066 	 * update it, otherwise create a new one.
2067 	 */
2068 	BRIDGE_RT_RENTER(s);
2069 	brt = bridge_rtnode_lookup(sc, dst);
2070 
2071 	if (brt != NULL) {
2072 		brt->brt_ifp = dst_if;
2073 		if (setflags) {
2074 			brt->brt_flags = flags;
2075 			if (flags & IFBAF_STATIC)
2076 				brt->brt_expire = 0;
2077 			else
2078 				brt->brt_expire = time_uptime + sc->sc_brttimeout;
2079 		} else {
2080 			if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2081 				brt->brt_expire = time_uptime + sc->sc_brttimeout;
2082 		}
2083 	}
2084 	BRIDGE_RT_REXIT(s);
2085 
2086 	if (brt == NULL) {
2087 		int r;
2088 
2089 		r = bridge_rtalloc(sc, dst, &brt);
2090 		if (r != 0)
2091 			return r;
2092 		goto again;
2093 	}
2094 
2095 	return 0;
2096 }
2097 
2098 /*
2099  * bridge_rtlookup:
2100  *
2101  *	Lookup the destination interface for an address.
2102  */
2103 static struct ifnet *
2104 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr)
2105 {
2106 	struct bridge_rtnode *brt;
2107 	struct ifnet *ifs = NULL;
2108 	int s;
2109 
2110 	BRIDGE_RT_RENTER(s);
2111 	brt = bridge_rtnode_lookup(sc, addr);
2112 	if (brt != NULL)
2113 		ifs = brt->brt_ifp;
2114 	BRIDGE_RT_REXIT(s);
2115 
2116 	return ifs;
2117 }
2118 
2119 typedef bool (*bridge_iterate_cb_t)
2120     (struct bridge_softc *, struct bridge_rtnode *, bool *, void *);
2121 
2122 /*
2123  * bridge_rtlist_iterate_remove:
2124  *
2125  *	It iterates on sc->sc_rtlist and removes rtnodes of it which func
2126  *	callback judges to remove. Removals of rtnodes are done in a manner
2127  *	of pserialize. To this end, all kmem_* operations are placed out of
2128  *	mutexes.
2129  */
2130 static void
2131 bridge_rtlist_iterate_remove(struct bridge_softc *sc, bridge_iterate_cb_t func, void *arg)
2132 {
2133 	struct bridge_rtnode *brt;
2134 	struct bridge_rtnode **brt_list;
2135 	int i, count;
2136 
2137 retry:
2138 	count = sc->sc_brtcnt;
2139 	if (count == 0)
2140 		return;
2141 	brt_list = kmem_alloc(sizeof(*brt_list) * count, KM_SLEEP);
2142 
2143 	BRIDGE_RT_LOCK(sc);
2144 	if (__predict_false(sc->sc_brtcnt > count)) {
2145 		/* The rtnodes increased, we need more memory */
2146 		BRIDGE_RT_UNLOCK(sc);
2147 		kmem_free(brt_list, sizeof(*brt_list) * count);
2148 		goto retry;
2149 	}
2150 
2151 	i = 0;
2152 	/*
2153 	 * We don't need to use a _SAFE variant here because we know
2154 	 * that a removed item keeps its next pointer as-is thanks to
2155 	 * pslist(9) and isn't freed in the loop.
2156 	 */
2157 	BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) {
2158 		bool need_break = false;
2159 		if (func(sc, brt, &need_break, arg)) {
2160 			bridge_rtnode_remove(sc, brt);
2161 			brt_list[i++] = brt;
2162 		}
2163 		if (need_break)
2164 			break;
2165 	}
2166 
2167 	if (i > 0)
2168 		BRIDGE_RT_PSZ_PERFORM(sc);
2169 	BRIDGE_RT_UNLOCK(sc);
2170 
2171 	while (--i >= 0)
2172 		bridge_rtnode_destroy(brt_list[i]);
2173 
2174 	kmem_free(brt_list, sizeof(*brt_list) * count);
2175 }
2176 
2177 static bool
2178 bridge_rttrim0_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
2179     bool *need_break, void *arg)
2180 {
2181 	if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2182 		/* Take into account of the subsequent removal */
2183 		if ((sc->sc_brtcnt - 1) <= sc->sc_brtmax)
2184 			*need_break = true;
2185 		return true;
2186 	} else
2187 		return false;
2188 }
2189 
2190 static void
2191 bridge_rttrim0(struct bridge_softc *sc)
2192 {
2193 	bridge_rtlist_iterate_remove(sc, bridge_rttrim0_cb, NULL);
2194 }
2195 
2196 /*
2197  * bridge_rttrim:
2198  *
2199  *	Trim the routine table so that we have a number
2200  *	of routing entries less than or equal to the
2201  *	maximum number.
2202  */
2203 static void
2204 bridge_rttrim(struct bridge_softc *sc)
2205 {
2206 
2207 	/* Make sure we actually need to do this. */
2208 	if (sc->sc_brtcnt <= sc->sc_brtmax)
2209 		return;
2210 
2211 	/* Force an aging cycle; this might trim enough addresses. */
2212 	bridge_rtage(sc);
2213 	if (sc->sc_brtcnt <= sc->sc_brtmax)
2214 		return;
2215 
2216 	bridge_rttrim0(sc);
2217 
2218 	return;
2219 }
2220 
2221 /*
2222  * bridge_timer:
2223  *
2224  *	Aging timer for the bridge.
2225  */
2226 static void
2227 bridge_timer(void *arg)
2228 {
2229 	struct bridge_softc *sc = arg;
2230 
2231 	workqueue_enqueue(sc->sc_rtage_wq, &sc->sc_rtage_wk, NULL);
2232 }
2233 
2234 static void
2235 bridge_rtage_work(struct work *wk, void *arg)
2236 {
2237 	struct bridge_softc *sc = arg;
2238 
2239 	KASSERT(wk == &sc->sc_rtage_wk);
2240 
2241 	bridge_rtage(sc);
2242 
2243 	if (sc->sc_if.if_flags & IFF_RUNNING)
2244 		callout_reset(&sc->sc_brcallout,
2245 		    bridge_rtable_prune_period * hz, bridge_timer, sc);
2246 }
2247 
2248 static bool
2249 bridge_rtage_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
2250     bool *need_break, void *arg)
2251 {
2252 	if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2253 	    time_uptime >= brt->brt_expire)
2254 		return true;
2255 	else
2256 		return false;
2257 }
2258 
2259 /*
2260  * bridge_rtage:
2261  *
2262  *	Perform an aging cycle.
2263  */
2264 static void
2265 bridge_rtage(struct bridge_softc *sc)
2266 {
2267 	bridge_rtlist_iterate_remove(sc, bridge_rtage_cb, NULL);
2268 }
2269 
2270 
2271 static bool
2272 bridge_rtflush_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
2273     bool *need_break, void *arg)
2274 {
2275 	int full = *(int*)arg;
2276 
2277 	if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2278 		return true;
2279 	else
2280 		return false;
2281 }
2282 
2283 /*
2284  * bridge_rtflush:
2285  *
2286  *	Remove all dynamic addresses from the bridge.
2287  */
2288 static void
2289 bridge_rtflush(struct bridge_softc *sc, int full)
2290 {
2291 	bridge_rtlist_iterate_remove(sc, bridge_rtflush_cb, &full);
2292 }
2293 
2294 /*
2295  * bridge_rtdaddr:
2296  *
2297  *	Remove an address from the table.
2298  */
2299 static int
2300 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr)
2301 {
2302 	struct bridge_rtnode *brt;
2303 
2304 	BRIDGE_RT_LOCK(sc);
2305 	if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL) {
2306 		BRIDGE_RT_UNLOCK(sc);
2307 		return ENOENT;
2308 	}
2309 	bridge_rtnode_remove(sc, brt);
2310 	BRIDGE_RT_PSZ_PERFORM(sc);
2311 	BRIDGE_RT_UNLOCK(sc);
2312 
2313 	bridge_rtnode_destroy(brt);
2314 
2315 	return 0;
2316 }
2317 
2318 /*
2319  * bridge_rtdelete:
2320  *
2321  *	Delete routes to a speicifc member interface.
2322  */
2323 static void
2324 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp)
2325 {
2326 	struct bridge_rtnode *brt;
2327 
2328 	/* XXX pserialize_perform for each entry is slow */
2329 again:
2330 	BRIDGE_RT_LOCK(sc);
2331 	BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) {
2332 		if (brt->brt_ifp == ifp)
2333 			break;
2334 	}
2335 	if (brt == NULL) {
2336 		BRIDGE_RT_UNLOCK(sc);
2337 		return;
2338 	}
2339 	bridge_rtnode_remove(sc, brt);
2340 	BRIDGE_RT_PSZ_PERFORM(sc);
2341 	BRIDGE_RT_UNLOCK(sc);
2342 
2343 	bridge_rtnode_destroy(brt);
2344 
2345 	goto again;
2346 }
2347 
2348 /*
2349  * bridge_rtable_init:
2350  *
2351  *	Initialize the route table for this bridge.
2352  */
2353 static void
2354 bridge_rtable_init(struct bridge_softc *sc)
2355 {
2356 	int i;
2357 
2358 	sc->sc_rthash = kmem_alloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2359 	    KM_SLEEP);
2360 
2361 	for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2362 		PSLIST_INIT(&sc->sc_rthash[i]);
2363 
2364 	sc->sc_rthash_key = cprng_fast32();
2365 
2366 	PSLIST_INIT(&sc->sc_rtlist);
2367 
2368 	sc->sc_rtlist_psz = pserialize_create();
2369 	sc->sc_rtlist_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_SOFTNET);
2370 }
2371 
2372 /*
2373  * bridge_rtable_fini:
2374  *
2375  *	Deconstruct the route table for this bridge.
2376  */
2377 static void
2378 bridge_rtable_fini(struct bridge_softc *sc)
2379 {
2380 
2381 	kmem_free(sc->sc_rthash, sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE);
2382 	mutex_obj_free(sc->sc_rtlist_lock);
2383 	pserialize_destroy(sc->sc_rtlist_psz);
2384 }
2385 
2386 /*
2387  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2388  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2389  */
2390 #define	mix(a, b, c)							\
2391 do {									\
2392 	a -= b; a -= c; a ^= (c >> 13);					\
2393 	b -= c; b -= a; b ^= (a << 8);					\
2394 	c -= a; c -= b; c ^= (b >> 13);					\
2395 	a -= b; a -= c; a ^= (c >> 12);					\
2396 	b -= c; b -= a; b ^= (a << 16);					\
2397 	c -= a; c -= b; c ^= (b >> 5);					\
2398 	a -= b; a -= c; a ^= (c >> 3);					\
2399 	b -= c; b -= a; b ^= (a << 10);					\
2400 	c -= a; c -= b; c ^= (b >> 15);					\
2401 } while (/*CONSTCOND*/0)
2402 
2403 static inline uint32_t
2404 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2405 {
2406 	uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2407 
2408 	b += addr[5] << 8;
2409 	b += addr[4];
2410 	a += addr[3] << 24;
2411 	a += addr[2] << 16;
2412 	a += addr[1] << 8;
2413 	a += addr[0];
2414 
2415 	mix(a, b, c);
2416 
2417 	return (c & BRIDGE_RTHASH_MASK);
2418 }
2419 
2420 #undef mix
2421 
2422 /*
2423  * bridge_rtnode_lookup:
2424  *
2425  *	Look up a bridge route node for the specified destination.
2426  */
2427 static struct bridge_rtnode *
2428 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr)
2429 {
2430 	struct bridge_rtnode *brt;
2431 	uint32_t hash;
2432 	int dir;
2433 
2434 	hash = bridge_rthash(sc, addr);
2435 	BRIDGE_RTHASH_READER_FOREACH(brt, sc, hash) {
2436 		dir = memcmp(addr, brt->brt_addr, ETHER_ADDR_LEN);
2437 		if (dir == 0)
2438 			return brt;
2439 		if (dir > 0)
2440 			return NULL;
2441 	}
2442 
2443 	return NULL;
2444 }
2445 
2446 /*
2447  * bridge_rtnode_insert:
2448  *
2449  *	Insert the specified bridge node into the route table.  We
2450  *	assume the entry is not already in the table.
2451  */
2452 static int
2453 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2454 {
2455 	struct bridge_rtnode *lbrt, *prev = NULL;
2456 	uint32_t hash;
2457 
2458 	KASSERT(BRIDGE_RT_LOCKED(sc));
2459 
2460 	hash = bridge_rthash(sc, brt->brt_addr);
2461 	BRIDGE_RTHASH_WRITER_FOREACH(lbrt, sc, hash) {
2462 		int dir = memcmp(brt->brt_addr, lbrt->brt_addr, ETHER_ADDR_LEN);
2463 		if (dir == 0)
2464 			return EEXIST;
2465 		if (dir > 0)
2466 			break;
2467 		prev = lbrt;
2468 	}
2469 	if (prev == NULL)
2470 		BRIDGE_RTHASH_WRITER_INSERT_HEAD(sc, hash, brt);
2471 	else
2472 		BRIDGE_RTHASH_WRITER_INSERT_AFTER(prev, brt);
2473 
2474 	BRIDGE_RTLIST_WRITER_INSERT_HEAD(sc, brt);
2475 	sc->sc_brtcnt++;
2476 
2477 	return 0;
2478 }
2479 
2480 /*
2481  * bridge_rtnode_remove:
2482  *
2483  *	Remove a bridge rtnode from the rthash and the rtlist of a bridge.
2484  */
2485 static void
2486 bridge_rtnode_remove(struct bridge_softc *sc, struct bridge_rtnode *brt)
2487 {
2488 
2489 	KASSERT(BRIDGE_RT_LOCKED(sc));
2490 
2491 	BRIDGE_RTHASH_WRITER_REMOVE(brt);
2492 	BRIDGE_RTLIST_WRITER_REMOVE(brt);
2493 	sc->sc_brtcnt--;
2494 }
2495 
2496 /*
2497  * bridge_rtnode_destroy:
2498  *
2499  *	Destroy a bridge rtnode.
2500  */
2501 static void
2502 bridge_rtnode_destroy(struct bridge_rtnode *brt)
2503 {
2504 
2505 	PSLIST_ENTRY_DESTROY(brt, brt_list);
2506 	PSLIST_ENTRY_DESTROY(brt, brt_hash);
2507 	pool_put(&bridge_rtnode_pool, brt);
2508 }
2509 
2510 #if defined(BRIDGE_IPF)
2511 extern pfil_head_t *inet_pfil_hook;                 /* XXX */
2512 extern pfil_head_t *inet6_pfil_hook;                /* XXX */
2513 
2514 /*
2515  * Send bridge packets through IPF if they are one of the types IPF can deal
2516  * with, or if they are ARP or REVARP.  (IPF will pass ARP and REVARP without
2517  * question.)
2518  */
2519 static int
2520 bridge_ipf(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
2521 {
2522 	int snap, error;
2523 	struct ether_header *eh1, eh2;
2524 	struct llc llc1;
2525 	uint16_t ether_type;
2526 
2527 	snap = 0;
2528 	error = -1;	/* Default error if not error == 0 */
2529 	eh1 = mtod(*mp, struct ether_header *);
2530 	ether_type = ntohs(eh1->ether_type);
2531 
2532 	/*
2533 	 * Check for SNAP/LLC.
2534 	 */
2535 	if (ether_type < ETHERMTU) {
2536 		struct llc *llc2 = (struct llc *)(eh1 + 1);
2537 
2538 		if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
2539 		    llc2->llc_dsap == LLC_SNAP_LSAP &&
2540 		    llc2->llc_ssap == LLC_SNAP_LSAP &&
2541 		    llc2->llc_control == LLC_UI) {
2542 			ether_type = htons(llc2->llc_un.type_snap.ether_type);
2543 			snap = 1;
2544 		}
2545 	}
2546 
2547 	/*
2548 	 * If we're trying to filter bridge traffic, don't look at anything
2549 	 * other than IP and ARP traffic.  If the filter doesn't understand
2550 	 * IPv6, don't allow IPv6 through the bridge either.  This is lame
2551 	 * since if we really wanted, say, an AppleTalk filter, we are hosed,
2552 	 * but of course we don't have an AppleTalk filter to begin with.
2553 	 * (Note that since IPF doesn't understand ARP it will pass *ALL*
2554 	 * ARP traffic.)
2555 	 */
2556 	switch (ether_type) {
2557 		case ETHERTYPE_ARP:
2558 		case ETHERTYPE_REVARP:
2559 			return 0; /* Automatically pass */
2560 		case ETHERTYPE_IP:
2561 # ifdef INET6
2562 		case ETHERTYPE_IPV6:
2563 # endif /* INET6 */
2564 			break;
2565 		default:
2566 			goto bad;
2567 	}
2568 
2569 	/* Strip off the Ethernet header and keep a copy. */
2570 	m_copydata(*mp, 0, ETHER_HDR_LEN, (void *) &eh2);
2571 	m_adj(*mp, ETHER_HDR_LEN);
2572 
2573 	/* Strip off snap header, if present */
2574 	if (snap) {
2575 		m_copydata(*mp, 0, sizeof(struct llc), (void *) &llc1);
2576 		m_adj(*mp, sizeof(struct llc));
2577 	}
2578 
2579 	/*
2580 	 * Check basic packet sanity and run IPF through pfil.
2581 	 */
2582 	KASSERT(!cpu_intr_p());
2583 	switch (ether_type)
2584 	{
2585 	case ETHERTYPE_IP :
2586 		error = bridge_ip_checkbasic(mp);
2587 		if (error == 0)
2588 			error = pfil_run_hooks(inet_pfil_hook, mp, ifp, dir);
2589 		break;
2590 # ifdef INET6
2591 	case ETHERTYPE_IPV6 :
2592 		error = bridge_ip6_checkbasic(mp);
2593 		if (error == 0)
2594 			error = pfil_run_hooks(inet6_pfil_hook, mp, ifp, dir);
2595 		break;
2596 # endif
2597 	default :
2598 		error = 0;
2599 		break;
2600 	}
2601 
2602 	if (*mp == NULL)
2603 		return error;
2604 	if (error != 0)
2605 		goto bad;
2606 
2607 	error = -1;
2608 
2609 	/*
2610 	 * Finally, put everything back the way it was and return
2611 	 */
2612 	if (snap) {
2613 		M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
2614 		if (*mp == NULL)
2615 			return error;
2616 		bcopy(&llc1, mtod(*mp, void *), sizeof(struct llc));
2617 	}
2618 
2619 	M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
2620 	if (*mp == NULL)
2621 		return error;
2622 	bcopy(&eh2, mtod(*mp, void *), ETHER_HDR_LEN);
2623 
2624 	return 0;
2625 
2626     bad:
2627 	m_freem(*mp);
2628 	*mp = NULL;
2629 	return error;
2630 }
2631 
2632 /*
2633  * Perform basic checks on header size since
2634  * IPF assumes ip_input has already processed
2635  * it for it.  Cut-and-pasted from ip_input.c.
2636  * Given how simple the IPv6 version is,
2637  * does the IPv4 version really need to be
2638  * this complicated?
2639  *
2640  * XXX Should we update ipstat here, or not?
2641  * XXX Right now we update ipstat but not
2642  * XXX csum_counter.
2643  */
2644 static int
2645 bridge_ip_checkbasic(struct mbuf **mp)
2646 {
2647 	struct mbuf *m = *mp;
2648 	struct ip *ip;
2649 	int len, hlen;
2650 
2651 	if (*mp == NULL)
2652 		return -1;
2653 
2654 	if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
2655 		if ((m = m_copyup(m, sizeof(struct ip),
2656 			(max_linkhdr + 3) & ~3)) == NULL) {
2657 			/* XXXJRT new stat, please */
2658 			ip_statinc(IP_STAT_TOOSMALL);
2659 			goto bad;
2660 		}
2661 	} else if (__predict_false(m->m_len < sizeof (struct ip))) {
2662 		if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
2663 			ip_statinc(IP_STAT_TOOSMALL);
2664 			goto bad;
2665 		}
2666 	}
2667 	ip = mtod(m, struct ip *);
2668 	if (ip == NULL) goto bad;
2669 
2670 	if (ip->ip_v != IPVERSION) {
2671 		ip_statinc(IP_STAT_BADVERS);
2672 		goto bad;
2673 	}
2674 	hlen = ip->ip_hl << 2;
2675 	if (hlen < sizeof(struct ip)) { /* minimum header length */
2676 		ip_statinc(IP_STAT_BADHLEN);
2677 		goto bad;
2678 	}
2679 	if (hlen > m->m_len) {
2680 		if ((m = m_pullup(m, hlen)) == 0) {
2681 			ip_statinc(IP_STAT_BADHLEN);
2682 			goto bad;
2683 		}
2684 		ip = mtod(m, struct ip *);
2685 		if (ip == NULL) goto bad;
2686 	}
2687 
2688 	switch (m->m_pkthdr.csum_flags &
2689 	        ((m_get_rcvif_NOMPSAFE(m)->if_csum_flags_rx & M_CSUM_IPv4) |
2690 	         M_CSUM_IPv4_BAD)) {
2691 	case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
2692 		/* INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); */
2693 		goto bad;
2694 
2695 	case M_CSUM_IPv4:
2696 		/* Checksum was okay. */
2697 		/* INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); */
2698 		break;
2699 
2700 	default:
2701 		/* Must compute it ourselves. */
2702 		/* INET_CSUM_COUNTER_INCR(&ip_swcsum); */
2703 		if (in_cksum(m, hlen) != 0)
2704 			goto bad;
2705 		break;
2706 	}
2707 
2708 	/* Retrieve the packet length. */
2709 	len = ntohs(ip->ip_len);
2710 
2711 	/*
2712 	 * Check for additional length bogosity
2713 	 */
2714 	if (len < hlen) {
2715 		ip_statinc(IP_STAT_BADLEN);
2716 		goto bad;
2717 	}
2718 
2719 	/*
2720 	 * Check that the amount of data in the buffers
2721 	 * is as at least much as the IP header would have us expect.
2722 	 * Drop packet if shorter than we expect.
2723 	 */
2724 	if (m->m_pkthdr.len < len) {
2725 		ip_statinc(IP_STAT_TOOSHORT);
2726 		goto bad;
2727 	}
2728 
2729 	/* Checks out, proceed */
2730 	*mp = m;
2731 	return 0;
2732 
2733     bad:
2734 	*mp = m;
2735 	return -1;
2736 }
2737 
2738 # ifdef INET6
2739 /*
2740  * Same as above, but for IPv6.
2741  * Cut-and-pasted from ip6_input.c.
2742  * XXX Should we update ip6stat, or not?
2743  */
2744 static int
2745 bridge_ip6_checkbasic(struct mbuf **mp)
2746 {
2747 	struct mbuf *m = *mp;
2748 	struct ip6_hdr *ip6;
2749 
2750 	/*
2751 	 * If the IPv6 header is not aligned, slurp it up into a new
2752 	 * mbuf with space for link headers, in the event we forward
2753 	 * it.  Otherwise, if it is aligned, make sure the entire base
2754 	 * IPv6 header is in the first mbuf of the chain.
2755 	 */
2756 	if (IP6_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
2757 		struct ifnet *inifp = m_get_rcvif_NOMPSAFE(m);
2758 		if ((m = m_copyup(m, sizeof(struct ip6_hdr),
2759 		                  (max_linkhdr + 3) & ~3)) == NULL) {
2760 			/* XXXJRT new stat, please */
2761 			ip6_statinc(IP6_STAT_TOOSMALL);
2762 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
2763 			goto bad;
2764 		}
2765 	} else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
2766 		struct ifnet *inifp = m_get_rcvif_NOMPSAFE(m);
2767 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
2768 			ip6_statinc(IP6_STAT_TOOSMALL);
2769 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
2770 			goto bad;
2771 		}
2772 	}
2773 
2774 	ip6 = mtod(m, struct ip6_hdr *);
2775 
2776 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
2777 		ip6_statinc(IP6_STAT_BADVERS);
2778 		in6_ifstat_inc(m_get_rcvif_NOMPSAFE(m), ifs6_in_hdrerr);
2779 		goto bad;
2780 	}
2781 
2782 	/* Checks out, proceed */
2783 	*mp = m;
2784 	return 0;
2785 
2786     bad:
2787 	*mp = m;
2788 	return -1;
2789 }
2790 # endif /* INET6 */
2791 #endif /* BRIDGE_IPF */
2792