Home | History | Annotate | Line # | Download | only in net
if_bridge.c revision 1.91.2.1
      1 /*	$NetBSD: if_bridge.c,v 1.91.2.1 2015/04/06 15:18:22 skrll 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 (at) 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 1.91.2.1 2015/04/06 15:18:22 skrll Exp $");
     84 
     85 #ifdef _KERNEL_OPT
     86 #include "opt_bridge_ipf.h"
     87 #include "opt_inet.h"
     88 #endif /* _KERNEL_OPT */
     89 
     90 #include <sys/param.h>
     91 #include <sys/kernel.h>
     92 #include <sys/mbuf.h>
     93 #include <sys/queue.h>
     94 #include <sys/socket.h>
     95 #include <sys/socketvar.h> /* for softnet_lock */
     96 #include <sys/sockio.h>
     97 #include <sys/systm.h>
     98 #include <sys/proc.h>
     99 #include <sys/pool.h>
    100 #include <sys/kauth.h>
    101 #include <sys/cpu.h>
    102 #include <sys/cprng.h>
    103 #include <sys/mutex.h>
    104 #include <sys/kmem.h>
    105 
    106 #include <net/bpf.h>
    107 #include <net/if.h>
    108 #include <net/if_dl.h>
    109 #include <net/if_types.h>
    110 #include <net/if_llc.h>
    111 #include <net/pktqueue.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 /*
    147  * Maximum number of addresses to cache.
    148  */
    149 #ifndef BRIDGE_RTABLE_MAX
    150 #define	BRIDGE_RTABLE_MAX		100
    151 #endif
    152 
    153 /*
    154  * Spanning tree defaults.
    155  */
    156 #define	BSTP_DEFAULT_MAX_AGE		(20 * 256)
    157 #define	BSTP_DEFAULT_HELLO_TIME		(2 * 256)
    158 #define	BSTP_DEFAULT_FORWARD_DELAY	(15 * 256)
    159 #define	BSTP_DEFAULT_HOLD_TIME		(1 * 256)
    160 #define	BSTP_DEFAULT_BRIDGE_PRIORITY	0x8000
    161 #define	BSTP_DEFAULT_PORT_PRIORITY	0x80
    162 #define	BSTP_DEFAULT_PATH_COST		55
    163 
    164 /*
    165  * Timeout (in seconds) for entries learned dynamically.
    166  */
    167 #ifndef BRIDGE_RTABLE_TIMEOUT
    168 #define	BRIDGE_RTABLE_TIMEOUT		(20 * 60)	/* same as ARP */
    169 #endif
    170 
    171 /*
    172  * Number of seconds between walks of the route list.
    173  */
    174 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
    175 #define	BRIDGE_RTABLE_PRUNE_PERIOD	(5 * 60)
    176 #endif
    177 
    178 #define BRIDGE_RT_INTR_LOCK(_sc)	mutex_enter((_sc)->sc_rtlist_intr_lock)
    179 #define BRIDGE_RT_INTR_UNLOCK(_sc)	mutex_exit((_sc)->sc_rtlist_intr_lock)
    180 #define BRIDGE_RT_INTR_LOCKED(_sc)	mutex_owned((_sc)->sc_rtlist_intr_lock)
    181 
    182 #define BRIDGE_RT_LOCK(_sc)	if ((_sc)->sc_rtlist_lock) \
    183 					mutex_enter((_sc)->sc_rtlist_lock)
    184 #define BRIDGE_RT_UNLOCK(_sc)	if ((_sc)->sc_rtlist_lock) \
    185 					mutex_exit((_sc)->sc_rtlist_lock)
    186 #define BRIDGE_RT_LOCKED(_sc)	(!(_sc)->sc_rtlist_lock || \
    187 				 mutex_owned((_sc)->sc_rtlist_lock))
    188 
    189 #define BRIDGE_RT_PSZ_PERFORM(_sc) \
    190 				if ((_sc)->sc_rtlist_psz != NULL) \
    191 					pserialize_perform((_sc)->sc_rtlist_psz);
    192 
    193 #ifdef BRIDGE_MPSAFE
    194 #define BRIDGE_RT_RENTER(__s)	do { \
    195 					if (!cpu_intr_p()) \
    196 						__s = pserialize_read_enter(); \
    197 					else \
    198 						__s = splhigh(); \
    199 				} while (0)
    200 #define BRIDGE_RT_REXIT(__s)	do { \
    201 					if (!cpu_intr_p()) \
    202 						pserialize_read_exit(__s); \
    203 					else \
    204 						splx(__s); \
    205 				} while (0)
    206 #else /* BRIDGE_MPSAFE */
    207 #define BRIDGE_RT_RENTER(__s)	do { __s = 0; } while (0)
    208 #define BRIDGE_RT_REXIT(__s)	do { (void)__s; } while (0)
    209 #endif /* BRIDGE_MPSAFE */
    210 
    211 int	bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
    212 
    213 static struct pool bridge_rtnode_pool;
    214 static struct work bridge_rtage_wk;
    215 
    216 void	bridgeattach(int);
    217 
    218 static int	bridge_clone_create(struct if_clone *, int);
    219 static int	bridge_clone_destroy(struct ifnet *);
    220 
    221 static int	bridge_ioctl(struct ifnet *, u_long, void *);
    222 static int	bridge_init(struct ifnet *);
    223 static void	bridge_stop(struct ifnet *, int);
    224 static void	bridge_start(struct ifnet *);
    225 
    226 static void	bridge_input(struct ifnet *, struct mbuf *);
    227 static void	bridge_forward(void *);
    228 
    229 static void	bridge_timer(void *);
    230 
    231 static void	bridge_broadcast(struct bridge_softc *, struct ifnet *,
    232 				 struct mbuf *);
    233 
    234 static int	bridge_rtupdate(struct bridge_softc *, const uint8_t *,
    235 				struct ifnet *, int, uint8_t);
    236 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *);
    237 static void	bridge_rttrim(struct bridge_softc *);
    238 static void	bridge_rtage(struct bridge_softc *);
    239 static void	bridge_rtage_work(struct work *, void *);
    240 static void	bridge_rtflush(struct bridge_softc *, int);
    241 static int	bridge_rtdaddr(struct bridge_softc *, const uint8_t *);
    242 static void	bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp);
    243 
    244 static void	bridge_rtable_init(struct bridge_softc *);
    245 static void	bridge_rtable_fini(struct bridge_softc *);
    246 
    247 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
    248 						  const uint8_t *);
    249 static int	bridge_rtnode_insert(struct bridge_softc *,
    250 				     struct bridge_rtnode *);
    251 static void	bridge_rtnode_remove(struct bridge_softc *,
    252 				     struct bridge_rtnode *);
    253 static void	bridge_rtnode_destroy(struct bridge_rtnode *);
    254 
    255 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
    256 						  const char *name);
    257 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
    258 						     struct ifnet *ifp);
    259 static void	bridge_release_member(struct bridge_softc *, struct bridge_iflist *);
    260 static void	bridge_delete_member(struct bridge_softc *,
    261 				     struct bridge_iflist *);
    262 static struct bridge_iflist *bridge_try_hold_bif(struct bridge_iflist *);
    263 
    264 static int	bridge_ioctl_add(struct bridge_softc *, void *);
    265 static int	bridge_ioctl_del(struct bridge_softc *, void *);
    266 static int	bridge_ioctl_gifflags(struct bridge_softc *, void *);
    267 static int	bridge_ioctl_sifflags(struct bridge_softc *, void *);
    268 static int	bridge_ioctl_scache(struct bridge_softc *, void *);
    269 static int	bridge_ioctl_gcache(struct bridge_softc *, void *);
    270 static int	bridge_ioctl_gifs(struct bridge_softc *, void *);
    271 static int	bridge_ioctl_rts(struct bridge_softc *, void *);
    272 static int	bridge_ioctl_saddr(struct bridge_softc *, void *);
    273 static int	bridge_ioctl_sto(struct bridge_softc *, void *);
    274 static int	bridge_ioctl_gto(struct bridge_softc *, void *);
    275 static int	bridge_ioctl_daddr(struct bridge_softc *, void *);
    276 static int	bridge_ioctl_flush(struct bridge_softc *, void *);
    277 static int	bridge_ioctl_gpri(struct bridge_softc *, void *);
    278 static int	bridge_ioctl_spri(struct bridge_softc *, void *);
    279 static int	bridge_ioctl_ght(struct bridge_softc *, void *);
    280 static int	bridge_ioctl_sht(struct bridge_softc *, void *);
    281 static int	bridge_ioctl_gfd(struct bridge_softc *, void *);
    282 static int	bridge_ioctl_sfd(struct bridge_softc *, void *);
    283 static int	bridge_ioctl_gma(struct bridge_softc *, void *);
    284 static int	bridge_ioctl_sma(struct bridge_softc *, void *);
    285 static int	bridge_ioctl_sifprio(struct bridge_softc *, void *);
    286 static int	bridge_ioctl_sifcost(struct bridge_softc *, void *);
    287 #if defined(BRIDGE_IPF)
    288 static int	bridge_ioctl_gfilt(struct bridge_softc *, void *);
    289 static int	bridge_ioctl_sfilt(struct bridge_softc *, void *);
    290 static int	bridge_ipf(void *, struct mbuf **, struct ifnet *, int);
    291 static int	bridge_ip_checkbasic(struct mbuf **mp);
    292 # ifdef INET6
    293 static int	bridge_ip6_checkbasic(struct mbuf **mp);
    294 # endif /* INET6 */
    295 #endif /* BRIDGE_IPF */
    296 
    297 static void bridge_sysctl_fwdq_setup(struct sysctllog **clog,
    298     struct bridge_softc *sc);
    299 
    300 struct bridge_control {
    301 	int	(*bc_func)(struct bridge_softc *, void *);
    302 	int	bc_argsize;
    303 	int	bc_flags;
    304 };
    305 
    306 #define	BC_F_COPYIN		0x01	/* copy arguments in */
    307 #define	BC_F_COPYOUT		0x02	/* copy arguments out */
    308 #define	BC_F_SUSER		0x04	/* do super-user check */
    309 
    310 static const struct bridge_control bridge_control_table[] = {
    311 [BRDGADD] = {bridge_ioctl_add, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    312 [BRDGDEL] = {bridge_ioctl_del, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    313 
    314 [BRDGGIFFLGS] = {bridge_ioctl_gifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_COPYOUT},
    315 [BRDGSIFFLGS] = {bridge_ioctl_sifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    316 
    317 [BRDGSCACHE] = {bridge_ioctl_scache, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    318 [BRDGGCACHE] = {bridge_ioctl_gcache, sizeof(struct ifbrparam), BC_F_COPYOUT},
    319 
    320 [BRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_COPYIN|BC_F_COPYOUT},
    321 [BRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_COPYIN|BC_F_COPYOUT},
    322 
    323 [BRDGSADDR] = {bridge_ioctl_saddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
    324 
    325 [BRDGSTO] = {bridge_ioctl_sto, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    326 [BRDGGTO] = {bridge_ioctl_gto, sizeof(struct ifbrparam), BC_F_COPYOUT},
    327 
    328 [BRDGDADDR] = {bridge_ioctl_daddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
    329 
    330 [BRDGFLUSH] = {bridge_ioctl_flush, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    331 
    332 [BRDGGPRI] = {bridge_ioctl_gpri, sizeof(struct ifbrparam), BC_F_COPYOUT},
    333 [BRDGSPRI] = {bridge_ioctl_spri, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    334 
    335 [BRDGGHT] = {bridge_ioctl_ght, sizeof(struct ifbrparam), BC_F_COPYOUT},
    336 [BRDGSHT] = {bridge_ioctl_sht, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    337 
    338 [BRDGGFD] = {bridge_ioctl_gfd, sizeof(struct ifbrparam), BC_F_COPYOUT},
    339 [BRDGSFD] = {bridge_ioctl_sfd, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    340 
    341 [BRDGGMA] = {bridge_ioctl_gma, sizeof(struct ifbrparam), BC_F_COPYOUT},
    342 [BRDGSMA] = {bridge_ioctl_sma, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    343 
    344 [BRDGSIFPRIO] = {bridge_ioctl_sifprio, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    345 
    346 [BRDGSIFCOST] = {bridge_ioctl_sifcost, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    347 #if defined(BRIDGE_IPF)
    348 [BRDGGFILT] = {bridge_ioctl_gfilt, sizeof(struct ifbrparam), BC_F_COPYOUT},
    349 [BRDGSFILT] = {bridge_ioctl_sfilt, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    350 #endif /* BRIDGE_IPF */
    351 };
    352 static const int bridge_control_table_size = __arraycount(bridge_control_table);
    353 
    354 static LIST_HEAD(, bridge_softc) bridge_list;
    355 static kmutex_t bridge_list_lock;
    356 
    357 static struct if_clone bridge_cloner =
    358     IF_CLONE_INITIALIZER("bridge", bridge_clone_create, bridge_clone_destroy);
    359 
    360 /*
    361  * bridgeattach:
    362  *
    363  *	Pseudo-device attach routine.
    364  */
    365 void
    366 bridgeattach(int n)
    367 {
    368 
    369 	pool_init(&bridge_rtnode_pool, sizeof(struct bridge_rtnode),
    370 	    0, 0, 0, "brtpl", NULL, IPL_NET);
    371 
    372 	LIST_INIT(&bridge_list);
    373 	mutex_init(&bridge_list_lock, MUTEX_DEFAULT, IPL_NET);
    374 	if_clone_attach(&bridge_cloner);
    375 }
    376 
    377 /*
    378  * bridge_clone_create:
    379  *
    380  *	Create a new bridge instance.
    381  */
    382 static int
    383 bridge_clone_create(struct if_clone *ifc, int unit)
    384 {
    385 	struct bridge_softc *sc;
    386 	struct ifnet *ifp;
    387 	int error, flags;
    388 
    389 	sc = kmem_zalloc(sizeof(*sc),  KM_SLEEP);
    390 	ifp = &sc->sc_if;
    391 
    392 	sc->sc_brtmax = BRIDGE_RTABLE_MAX;
    393 	sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
    394 	sc->sc_bridge_max_age = BSTP_DEFAULT_MAX_AGE;
    395 	sc->sc_bridge_hello_time = BSTP_DEFAULT_HELLO_TIME;
    396 	sc->sc_bridge_forward_delay = BSTP_DEFAULT_FORWARD_DELAY;
    397 	sc->sc_bridge_priority = BSTP_DEFAULT_BRIDGE_PRIORITY;
    398 	sc->sc_hold_time = BSTP_DEFAULT_HOLD_TIME;
    399 	sc->sc_filter_flags = 0;
    400 
    401 	/* Initialize our routing table. */
    402 	bridge_rtable_init(sc);
    403 
    404 #ifdef BRIDGE_MPSAFE
    405 	flags = WQ_MPSAFE;
    406 #else
    407 	flags = 0;
    408 #endif
    409 	error = workqueue_create(&sc->sc_rtage_wq, "bridge_rtage",
    410 	    bridge_rtage_work, sc, PRI_SOFTNET, IPL_SOFTNET, flags);
    411 	if (error)
    412 		panic("%s: workqueue_create %d\n", __func__, error);
    413 
    414 	callout_init(&sc->sc_brcallout, 0);
    415 	callout_init(&sc->sc_bstpcallout, 0);
    416 
    417 	LIST_INIT(&sc->sc_iflist);
    418 #ifdef BRIDGE_MPSAFE
    419 	sc->sc_iflist_intr_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
    420 	sc->sc_iflist_psz = pserialize_create();
    421 	sc->sc_iflist_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_SOFTNET);
    422 #else
    423 	sc->sc_iflist_intr_lock = NULL;
    424 	sc->sc_iflist_psz = NULL;
    425 	sc->sc_iflist_lock = NULL;
    426 #endif
    427 	cv_init(&sc->sc_iflist_cv, "if_bridge_cv");
    428 
    429 	if_initname(ifp, ifc->ifc_name, unit);
    430 	ifp->if_softc = sc;
    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 	sc->sc_fwd_pktq = pktq_create(IFQ_MAXLEN, bridge_forward, sc);
    443 	KASSERT(sc->sc_fwd_pktq != NULL);
    444 
    445 	bridge_sysctl_fwdq_setup(&ifp->if_sysctl_log, sc);
    446 
    447 	if_attach(ifp);
    448 
    449 	if_alloc_sadl(ifp);
    450 
    451 	mutex_enter(&bridge_list_lock);
    452 	LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
    453 	mutex_exit(&bridge_list_lock);
    454 
    455 	return (0);
    456 }
    457 
    458 /*
    459  * bridge_clone_destroy:
    460  *
    461  *	Destroy a bridge instance.
    462  */
    463 static int
    464 bridge_clone_destroy(struct ifnet *ifp)
    465 {
    466 	struct bridge_softc *sc = ifp->if_softc;
    467 	struct bridge_iflist *bif;
    468 	int s;
    469 
    470 	/* Must be called during IFF_RUNNING, i.e., before bridge_stop */
    471 	pktq_barrier(sc->sc_fwd_pktq);
    472 
    473 	s = splnet();
    474 
    475 	bridge_stop(ifp, 1);
    476 
    477 	while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
    478 		bridge_delete_member(sc, bif);
    479 
    480 	mutex_enter(&bridge_list_lock);
    481 	LIST_REMOVE(sc, sc_list);
    482 	mutex_exit(&bridge_list_lock);
    483 
    484 	splx(s);
    485 
    486 	if_detach(ifp);
    487 
    488 	/* Should be called after if_detach for safe */
    489 	pktq_flush(sc->sc_fwd_pktq);
    490 	pktq_destroy(sc->sc_fwd_pktq);
    491 
    492 	/* Tear down the routing table. */
    493 	bridge_rtable_fini(sc);
    494 
    495 	cv_destroy(&sc->sc_iflist_cv);
    496 	if (sc->sc_iflist_intr_lock)
    497 		mutex_obj_free(sc->sc_iflist_intr_lock);
    498 
    499 	if (sc->sc_iflist_psz)
    500 		pserialize_destroy(sc->sc_iflist_psz);
    501 	if (sc->sc_iflist_lock)
    502 		mutex_obj_free(sc->sc_iflist_lock);
    503 
    504 	workqueue_destroy(sc->sc_rtage_wq);
    505 
    506 	kmem_free(sc, sizeof(*sc));
    507 
    508 	return (0);
    509 }
    510 
    511 static int
    512 bridge_sysctl_fwdq_maxlen(SYSCTLFN_ARGS)
    513 {
    514 	struct sysctlnode node = *rnode;
    515 	const struct bridge_softc *sc =	node.sysctl_data;
    516 	return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), sc->sc_fwd_pktq);
    517 }
    518 
    519 #define	SYSCTL_BRIDGE_PKTQ(cn, c)					\
    520 	static int							\
    521 	bridge_sysctl_fwdq_##cn(SYSCTLFN_ARGS)				\
    522 	{								\
    523 		struct sysctlnode node = *rnode;			\
    524 		const struct bridge_softc *sc =	node.sysctl_data;	\
    525 		return sysctl_pktq_count(SYSCTLFN_CALL(rnode),		\
    526 					 sc->sc_fwd_pktq, c);		\
    527 	}
    528 
    529 SYSCTL_BRIDGE_PKTQ(items, PKTQ_NITEMS)
    530 SYSCTL_BRIDGE_PKTQ(drops, PKTQ_DROPS)
    531 
    532 static void
    533 bridge_sysctl_fwdq_setup(struct sysctllog **clog, struct bridge_softc *sc)
    534 {
    535 	const struct sysctlnode *cnode, *rnode;
    536 	sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL;
    537 	const char *ifname = sc->sc_if.if_xname;
    538 
    539 	len_func = bridge_sysctl_fwdq_items;
    540 	maxlen_func = bridge_sysctl_fwdq_maxlen;
    541 	drops_func = bridge_sysctl_fwdq_drops;
    542 
    543 	if (sysctl_createv(clog, 0, NULL, &rnode,
    544 			   CTLFLAG_PERMANENT,
    545 			   CTLTYPE_NODE, "interfaces",
    546 			   SYSCTL_DESCR("Per-interface controls"),
    547 			   NULL, 0, NULL, 0,
    548 			   CTL_NET, CTL_CREATE, CTL_EOL) != 0)
    549 		goto bad;
    550 
    551 	if (sysctl_createv(clog, 0, &rnode, &rnode,
    552 			   CTLFLAG_PERMANENT,
    553 			   CTLTYPE_NODE, ifname,
    554 			   SYSCTL_DESCR("Interface controls"),
    555 			   NULL, 0, NULL, 0,
    556 			   CTL_CREATE, CTL_EOL) != 0)
    557 		goto bad;
    558 
    559 	if (sysctl_createv(clog, 0, &rnode, &rnode,
    560 			   CTLFLAG_PERMANENT,
    561 			   CTLTYPE_NODE, "fwdq",
    562 			   SYSCTL_DESCR("Protocol input queue controls"),
    563 			   NULL, 0, NULL, 0,
    564 			   CTL_CREATE, CTL_EOL) != 0)
    565 		goto bad;
    566 
    567 	if (sysctl_createv(clog, 0, &rnode, &cnode,
    568 			   CTLFLAG_PERMANENT,
    569 			   CTLTYPE_INT, "len",
    570 			   SYSCTL_DESCR("Current forwarding queue length"),
    571 			   len_func, 0, (void *)sc, 0,
    572 			   CTL_CREATE, IFQCTL_LEN, CTL_EOL) != 0)
    573 		goto bad;
    574 
    575 	if (sysctl_createv(clog, 0, &rnode, &cnode,
    576 			   CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
    577 			   CTLTYPE_INT, "maxlen",
    578 			   SYSCTL_DESCR("Maximum allowed forwarding queue length"),
    579 			   maxlen_func, 0, (void *)sc, 0,
    580 			   CTL_CREATE, IFQCTL_MAXLEN, CTL_EOL) != 0)
    581 		goto bad;
    582 
    583 	if (sysctl_createv(clog, 0, &rnode, &cnode,
    584 			   CTLFLAG_PERMANENT,
    585 			   CTLTYPE_INT, "drops",
    586 			   SYSCTL_DESCR("Packets dropped due to full forwarding queue"),
    587 			   drops_func, 0, (void *)sc, 0,
    588 			   CTL_CREATE, IFQCTL_DROPS, CTL_EOL) != 0)
    589 		goto bad;
    590 
    591 	return;
    592 bad:
    593 	aprint_error("%s: could not attach sysctl nodes\n", ifname);
    594 	return;
    595 }
    596 
    597 /*
    598  * bridge_ioctl:
    599  *
    600  *	Handle a control request from the operator.
    601  */
    602 static int
    603 bridge_ioctl(struct ifnet *ifp, u_long cmd, void *data)
    604 {
    605 	struct bridge_softc *sc = ifp->if_softc;
    606 	struct lwp *l = curlwp;	/* XXX */
    607 	union {
    608 		struct ifbreq ifbreq;
    609 		struct ifbifconf ifbifconf;
    610 		struct ifbareq ifbareq;
    611 		struct ifbaconf ifbaconf;
    612 		struct ifbrparam ifbrparam;
    613 	} args;
    614 	struct ifdrv *ifd = (struct ifdrv *) data;
    615 	const struct bridge_control *bc = NULL; /* XXXGCC */
    616 	int s, error = 0;
    617 
    618 	/* Authorize command before calling splnet(). */
    619 	switch (cmd) {
    620 	case SIOCGDRVSPEC:
    621 	case SIOCSDRVSPEC:
    622 		if (ifd->ifd_cmd >= bridge_control_table_size) {
    623 			error = EINVAL;
    624 			return error;
    625 		}
    626 
    627 		bc = &bridge_control_table[ifd->ifd_cmd];
    628 
    629 		/* We only care about BC_F_SUSER at this point. */
    630 		if ((bc->bc_flags & BC_F_SUSER) == 0)
    631 			break;
    632 
    633 		error = kauth_authorize_network(l->l_cred,
    634 		    KAUTH_NETWORK_INTERFACE_BRIDGE,
    635 		    cmd == SIOCGDRVSPEC ?
    636 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_GETPRIV :
    637 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_SETPRIV,
    638 		     ifd, NULL, NULL);
    639 		if (error)
    640 			return (error);
    641 
    642 		break;
    643 	}
    644 
    645 	s = splnet();
    646 
    647 	switch (cmd) {
    648 	case SIOCGDRVSPEC:
    649 	case SIOCSDRVSPEC:
    650 		KASSERT(bc != NULL);
    651 		if (cmd == SIOCGDRVSPEC &&
    652 		    (bc->bc_flags & BC_F_COPYOUT) == 0) {
    653 			error = EINVAL;
    654 			break;
    655 		}
    656 		else if (cmd == SIOCSDRVSPEC &&
    657 		    (bc->bc_flags & BC_F_COPYOUT) != 0) {
    658 			error = EINVAL;
    659 			break;
    660 		}
    661 
    662 		/* BC_F_SUSER is checked above, before splnet(). */
    663 
    664 		if (ifd->ifd_len != bc->bc_argsize ||
    665 		    ifd->ifd_len > sizeof(args)) {
    666 			error = EINVAL;
    667 			break;
    668 		}
    669 
    670 		memset(&args, 0, sizeof(args));
    671 		if (bc->bc_flags & BC_F_COPYIN) {
    672 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
    673 			if (error)
    674 				break;
    675 		}
    676 
    677 		error = (*bc->bc_func)(sc, &args);
    678 		if (error)
    679 			break;
    680 
    681 		if (bc->bc_flags & BC_F_COPYOUT)
    682 			error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
    683 
    684 		break;
    685 
    686 	case SIOCSIFFLAGS:
    687 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
    688 			break;
    689 		switch (ifp->if_flags & (IFF_UP|IFF_RUNNING)) {
    690 		case IFF_RUNNING:
    691 			/*
    692 			 * If interface is marked down and it is running,
    693 			 * then stop and disable it.
    694 			 */
    695 			(*ifp->if_stop)(ifp, 1);
    696 			break;
    697 		case IFF_UP:
    698 			/*
    699 			 * If interface is marked up and it is stopped, then
    700 			 * start it.
    701 			 */
    702 			error = (*ifp->if_init)(ifp);
    703 			break;
    704 		default:
    705 			break;
    706 		}
    707 		break;
    708 
    709 	case SIOCSIFMTU:
    710 		if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET)
    711 			error = 0;
    712 		break;
    713 
    714 	default:
    715 		error = ifioctl_common(ifp, cmd, data);
    716 		break;
    717 	}
    718 
    719 	splx(s);
    720 
    721 	return (error);
    722 }
    723 
    724 /*
    725  * bridge_lookup_member:
    726  *
    727  *	Lookup a bridge member interface.
    728  */
    729 static struct bridge_iflist *
    730 bridge_lookup_member(struct bridge_softc *sc, const char *name)
    731 {
    732 	struct bridge_iflist *bif;
    733 	struct ifnet *ifp;
    734 	int s;
    735 
    736 	BRIDGE_PSZ_RENTER(s);
    737 
    738 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
    739 		ifp = bif->bif_ifp;
    740 		if (strcmp(ifp->if_xname, name) == 0)
    741 			break;
    742 	}
    743 	bif = bridge_try_hold_bif(bif);
    744 
    745 	BRIDGE_PSZ_REXIT(s);
    746 
    747 	return bif;
    748 }
    749 
    750 /*
    751  * bridge_lookup_member_if:
    752  *
    753  *	Lookup a bridge member interface by ifnet*.
    754  */
    755 static struct bridge_iflist *
    756 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
    757 {
    758 	struct bridge_iflist *bif;
    759 	int s;
    760 
    761 	BRIDGE_PSZ_RENTER(s);
    762 
    763 	bif = member_ifp->if_bridgeif;
    764 	bif = bridge_try_hold_bif(bif);
    765 
    766 	BRIDGE_PSZ_REXIT(s);
    767 
    768 	return bif;
    769 }
    770 
    771 static struct bridge_iflist *
    772 bridge_try_hold_bif(struct bridge_iflist *bif)
    773 {
    774 #ifdef BRIDGE_MPSAFE
    775 	if (bif != NULL) {
    776 		if (bif->bif_waiting)
    777 			bif = NULL;
    778 		else
    779 			atomic_inc_32(&bif->bif_refs);
    780 	}
    781 #endif
    782 	return bif;
    783 }
    784 
    785 /*
    786  * bridge_release_member:
    787  *
    788  *	Release the specified member interface.
    789  */
    790 static void
    791 bridge_release_member(struct bridge_softc *sc, struct bridge_iflist *bif)
    792 {
    793 #ifdef BRIDGE_MPSAFE
    794 	uint32_t refs;
    795 
    796 	refs = atomic_dec_uint_nv(&bif->bif_refs);
    797 	if (__predict_false(refs == 0 && bif->bif_waiting)) {
    798 		BRIDGE_INTR_LOCK(sc);
    799 		cv_broadcast(&sc->sc_iflist_cv);
    800 		BRIDGE_INTR_UNLOCK(sc);
    801 	}
    802 #else
    803 	(void)sc;
    804 	(void)bif;
    805 #endif
    806 }
    807 
    808 /*
    809  * bridge_delete_member:
    810  *
    811  *	Delete the specified member interface.
    812  */
    813 static void
    814 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif)
    815 {
    816 	struct ifnet *ifs = bif->bif_ifp;
    817 
    818 	BRIDGE_LOCK(sc);
    819 
    820 	ifs->if_input = ether_input;
    821 	ifs->if_bridge = NULL;
    822 	ifs->if_bridgeif = NULL;
    823 
    824 	LIST_REMOVE(bif, bif_next);
    825 
    826 	BRIDGE_PSZ_PERFORM(sc);
    827 
    828 	BRIDGE_UNLOCK(sc);
    829 
    830 #ifdef BRIDGE_MPSAFE
    831 	BRIDGE_INTR_LOCK(sc);
    832 	bif->bif_waiting = true;
    833 	membar_sync();
    834 	while (bif->bif_refs > 0) {
    835 		aprint_debug("%s: cv_wait on iflist\n", __func__);
    836 		cv_wait(&sc->sc_iflist_cv, sc->sc_iflist_intr_lock);
    837 	}
    838 	bif->bif_waiting = false;
    839 	BRIDGE_INTR_UNLOCK(sc);
    840 #endif
    841 
    842 	kmem_free(bif, sizeof(*bif));
    843 }
    844 
    845 static int
    846 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
    847 {
    848 	struct ifbreq *req = arg;
    849 	struct bridge_iflist *bif = NULL;
    850 	struct ifnet *ifs;
    851 	int error = 0;
    852 
    853 	ifs = ifunit(req->ifbr_ifsname);
    854 	if (ifs == NULL)
    855 		return (ENOENT);
    856 
    857 	if (sc->sc_if.if_mtu != ifs->if_mtu)
    858 		return (EINVAL);
    859 
    860 	if (ifs->if_bridge == sc)
    861 		return (EEXIST);
    862 
    863 	if (ifs->if_bridge != NULL)
    864 		return (EBUSY);
    865 
    866 	if (ifs->if_input != ether_input)
    867 		return EINVAL;
    868 
    869 	/* FIXME: doesn't work with non-IFF_SIMPLEX interfaces */
    870 	if ((ifs->if_flags & IFF_SIMPLEX) == 0)
    871 		return EINVAL;
    872 
    873 	bif = kmem_alloc(sizeof(*bif), KM_SLEEP);
    874 
    875 	switch (ifs->if_type) {
    876 	case IFT_ETHER:
    877 		/*
    878 		 * Place the interface into promiscuous mode.
    879 		 */
    880 		error = ifpromisc(ifs, 1);
    881 		if (error)
    882 			goto out;
    883 		break;
    884 	default:
    885 		error = EINVAL;
    886 		goto out;
    887 	}
    888 
    889 	bif->bif_ifp = ifs;
    890 	bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
    891 	bif->bif_priority = BSTP_DEFAULT_PORT_PRIORITY;
    892 	bif->bif_path_cost = BSTP_DEFAULT_PATH_COST;
    893 	bif->bif_refs = 0;
    894 	bif->bif_waiting = false;
    895 
    896 	BRIDGE_LOCK(sc);
    897 
    898 	ifs->if_bridge = sc;
    899 	ifs->if_bridgeif = bif;
    900 	LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
    901 	ifs->if_input = bridge_input;
    902 
    903 	BRIDGE_UNLOCK(sc);
    904 
    905 	if (sc->sc_if.if_flags & IFF_RUNNING)
    906 		bstp_initialization(sc);
    907 	else
    908 		bstp_stop(sc);
    909 
    910  out:
    911 	if (error) {
    912 		if (bif != NULL)
    913 			kmem_free(bif, sizeof(*bif));
    914 	}
    915 	return (error);
    916 }
    917 
    918 static int
    919 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
    920 {
    921 	struct ifbreq *req = arg;
    922 	const char *name = req->ifbr_ifsname;
    923 	struct bridge_iflist *bif;
    924 	struct ifnet *ifs;
    925 
    926 	BRIDGE_LOCK(sc);
    927 
    928 	/*
    929 	 * Don't use bridge_lookup_member. We want to get a member
    930 	 * with bif_refs == 0.
    931 	 */
    932 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
    933 		ifs = bif->bif_ifp;
    934 		if (strcmp(ifs->if_xname, name) == 0)
    935 			break;
    936 	}
    937 
    938 	if (bif == NULL) {
    939 		BRIDGE_UNLOCK(sc);
    940 		return ENOENT;
    941 	}
    942 
    943 	BRIDGE_UNLOCK(sc);
    944 
    945 	bridge_delete_member(sc, bif);
    946 
    947 
    948 	switch (ifs->if_type) {
    949 	case IFT_ETHER:
    950 		/*
    951 		 * Take the interface out of promiscuous mode.
    952 		 * Don't call it with holding a spin lock.
    953 		 */
    954 		(void) ifpromisc(ifs, 0);
    955 		break;
    956 	default:
    957 #ifdef DIAGNOSTIC
    958 		panic("bridge_delete_member: impossible");
    959 #endif
    960 		break;
    961 	}
    962 
    963 	bridge_rtdelete(sc, ifs);
    964 
    965 	if (sc->sc_if.if_flags & IFF_RUNNING)
    966 		bstp_initialization(sc);
    967 
    968 	return 0;
    969 }
    970 
    971 static int
    972 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
    973 {
    974 	struct ifbreq *req = arg;
    975 	struct bridge_iflist *bif;
    976 
    977 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
    978 	if (bif == NULL)
    979 		return (ENOENT);
    980 
    981 	req->ifbr_ifsflags = bif->bif_flags;
    982 	req->ifbr_state = bif->bif_state;
    983 	req->ifbr_priority = bif->bif_priority;
    984 	req->ifbr_path_cost = bif->bif_path_cost;
    985 	req->ifbr_portno = bif->bif_ifp->if_index & 0xff;
    986 
    987 	bridge_release_member(sc, bif);
    988 
    989 	return (0);
    990 }
    991 
    992 static int
    993 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
    994 {
    995 	struct ifbreq *req = arg;
    996 	struct bridge_iflist *bif;
    997 
    998 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
    999 	if (bif == NULL)
   1000 		return (ENOENT);
   1001 
   1002 	if (req->ifbr_ifsflags & IFBIF_STP) {
   1003 		switch (bif->bif_ifp->if_type) {
   1004 		case IFT_ETHER:
   1005 			/* These can do spanning tree. */
   1006 			break;
   1007 
   1008 		default:
   1009 			/* Nothing else can. */
   1010 			bridge_release_member(sc, bif);
   1011 			return (EINVAL);
   1012 		}
   1013 	}
   1014 
   1015 	bif->bif_flags = req->ifbr_ifsflags;
   1016 
   1017 	bridge_release_member(sc, bif);
   1018 
   1019 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1020 		bstp_initialization(sc);
   1021 
   1022 	return (0);
   1023 }
   1024 
   1025 static int
   1026 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
   1027 {
   1028 	struct ifbrparam *param = arg;
   1029 
   1030 	sc->sc_brtmax = param->ifbrp_csize;
   1031 	bridge_rttrim(sc);
   1032 
   1033 	return (0);
   1034 }
   1035 
   1036 static int
   1037 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
   1038 {
   1039 	struct ifbrparam *param = arg;
   1040 
   1041 	param->ifbrp_csize = sc->sc_brtmax;
   1042 
   1043 	return (0);
   1044 }
   1045 
   1046 static int
   1047 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
   1048 {
   1049 	struct ifbifconf *bifc = arg;
   1050 	struct bridge_iflist *bif;
   1051 	struct ifbreq *breqs;
   1052 	int i, count, error = 0;
   1053 
   1054 retry:
   1055 	BRIDGE_LOCK(sc);
   1056 	count = 0;
   1057 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
   1058 		count++;
   1059 	BRIDGE_UNLOCK(sc);
   1060 
   1061 	if (count == 0) {
   1062 		bifc->ifbic_len = 0;
   1063 		return 0;
   1064 	}
   1065 
   1066 	if (bifc->ifbic_len == 0 || bifc->ifbic_len < (sizeof(*breqs) * count)) {
   1067 		/* Tell that a larger buffer is needed */
   1068 		bifc->ifbic_len = sizeof(*breqs) * count;
   1069 		return 0;
   1070 	}
   1071 
   1072 	breqs = kmem_alloc(sizeof(*breqs) * count, KM_SLEEP);
   1073 
   1074 	BRIDGE_LOCK(sc);
   1075 
   1076 	i = 0;
   1077 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
   1078 		i++;
   1079 	if (i > count) {
   1080 		/*
   1081 		 * The number of members has been increased.
   1082 		 * We need more memory!
   1083 		 */
   1084 		BRIDGE_UNLOCK(sc);
   1085 		kmem_free(breqs, sizeof(*breqs) * count);
   1086 		goto retry;
   1087 	}
   1088 
   1089 	i = 0;
   1090 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
   1091 		struct ifbreq *breq = &breqs[i++];
   1092 		memset(breq, 0, sizeof(*breq));
   1093 
   1094 		strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname,
   1095 		    sizeof(breq->ifbr_ifsname));
   1096 		breq->ifbr_ifsflags = bif->bif_flags;
   1097 		breq->ifbr_state = bif->bif_state;
   1098 		breq->ifbr_priority = bif->bif_priority;
   1099 		breq->ifbr_path_cost = bif->bif_path_cost;
   1100 		breq->ifbr_portno = bif->bif_ifp->if_index & 0xff;
   1101 	}
   1102 
   1103 	/* Don't call copyout with holding the mutex */
   1104 	BRIDGE_UNLOCK(sc);
   1105 
   1106 	for (i = 0; i < count; i++) {
   1107 		error = copyout(&breqs[i], bifc->ifbic_req + i, sizeof(*breqs));
   1108 		if (error)
   1109 			break;
   1110 	}
   1111 	bifc->ifbic_len = sizeof(*breqs) * i;
   1112 
   1113 	kmem_free(breqs, sizeof(*breqs) * count);
   1114 
   1115 	return error;
   1116 }
   1117 
   1118 static int
   1119 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
   1120 {
   1121 	struct ifbaconf *bac = arg;
   1122 	struct bridge_rtnode *brt;
   1123 	struct ifbareq bareq;
   1124 	int count = 0, error = 0, len;
   1125 
   1126 	if (bac->ifbac_len == 0)
   1127 		return (0);
   1128 
   1129 	BRIDGE_RT_INTR_LOCK(sc);
   1130 
   1131 	len = bac->ifbac_len;
   1132 	LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
   1133 		if (len < sizeof(bareq))
   1134 			goto out;
   1135 		memset(&bareq, 0, sizeof(bareq));
   1136 		strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
   1137 		    sizeof(bareq.ifba_ifsname));
   1138 		memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
   1139 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
   1140 			bareq.ifba_expire = brt->brt_expire - time_uptime;
   1141 		} else
   1142 			bareq.ifba_expire = 0;
   1143 		bareq.ifba_flags = brt->brt_flags;
   1144 
   1145 		error = copyout(&bareq, bac->ifbac_req + count, sizeof(bareq));
   1146 		if (error)
   1147 			goto out;
   1148 		count++;
   1149 		len -= sizeof(bareq);
   1150 	}
   1151  out:
   1152 	BRIDGE_RT_INTR_UNLOCK(sc);
   1153 
   1154 	bac->ifbac_len = sizeof(bareq) * count;
   1155 	return (error);
   1156 }
   1157 
   1158 static int
   1159 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
   1160 {
   1161 	struct ifbareq *req = arg;
   1162 	struct bridge_iflist *bif;
   1163 	int error;
   1164 
   1165 	bif = bridge_lookup_member(sc, req->ifba_ifsname);
   1166 	if (bif == NULL)
   1167 		return (ENOENT);
   1168 
   1169 	error = bridge_rtupdate(sc, req->ifba_dst, bif->bif_ifp, 1,
   1170 	    req->ifba_flags);
   1171 
   1172 	bridge_release_member(sc, bif);
   1173 
   1174 	return (error);
   1175 }
   1176 
   1177 static int
   1178 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
   1179 {
   1180 	struct ifbrparam *param = arg;
   1181 
   1182 	sc->sc_brttimeout = param->ifbrp_ctime;
   1183 
   1184 	return (0);
   1185 }
   1186 
   1187 static int
   1188 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
   1189 {
   1190 	struct ifbrparam *param = arg;
   1191 
   1192 	param->ifbrp_ctime = sc->sc_brttimeout;
   1193 
   1194 	return (0);
   1195 }
   1196 
   1197 static int
   1198 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
   1199 {
   1200 	struct ifbareq *req = arg;
   1201 
   1202 	return (bridge_rtdaddr(sc, req->ifba_dst));
   1203 }
   1204 
   1205 static int
   1206 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
   1207 {
   1208 	struct ifbreq *req = arg;
   1209 
   1210 	bridge_rtflush(sc, req->ifbr_ifsflags);
   1211 
   1212 	return (0);
   1213 }
   1214 
   1215 static int
   1216 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
   1217 {
   1218 	struct ifbrparam *param = arg;
   1219 
   1220 	param->ifbrp_prio = sc->sc_bridge_priority;
   1221 
   1222 	return (0);
   1223 }
   1224 
   1225 static int
   1226 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
   1227 {
   1228 	struct ifbrparam *param = arg;
   1229 
   1230 	sc->sc_bridge_priority = param->ifbrp_prio;
   1231 
   1232 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1233 		bstp_initialization(sc);
   1234 
   1235 	return (0);
   1236 }
   1237 
   1238 static int
   1239 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
   1240 {
   1241 	struct ifbrparam *param = arg;
   1242 
   1243 	param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8;
   1244 
   1245 	return (0);
   1246 }
   1247 
   1248 static int
   1249 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
   1250 {
   1251 	struct ifbrparam *param = arg;
   1252 
   1253 	if (param->ifbrp_hellotime == 0)
   1254 		return (EINVAL);
   1255 	sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8;
   1256 
   1257 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1258 		bstp_initialization(sc);
   1259 
   1260 	return (0);
   1261 }
   1262 
   1263 static int
   1264 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
   1265 {
   1266 	struct ifbrparam *param = arg;
   1267 
   1268 	param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8;
   1269 
   1270 	return (0);
   1271 }
   1272 
   1273 static int
   1274 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
   1275 {
   1276 	struct ifbrparam *param = arg;
   1277 
   1278 	if (param->ifbrp_fwddelay == 0)
   1279 		return (EINVAL);
   1280 	sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8;
   1281 
   1282 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1283 		bstp_initialization(sc);
   1284 
   1285 	return (0);
   1286 }
   1287 
   1288 static int
   1289 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
   1290 {
   1291 	struct ifbrparam *param = arg;
   1292 
   1293 	param->ifbrp_maxage = sc->sc_bridge_max_age >> 8;
   1294 
   1295 	return (0);
   1296 }
   1297 
   1298 static int
   1299 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
   1300 {
   1301 	struct ifbrparam *param = arg;
   1302 
   1303 	if (param->ifbrp_maxage == 0)
   1304 		return (EINVAL);
   1305 	sc->sc_bridge_max_age = param->ifbrp_maxage << 8;
   1306 
   1307 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1308 		bstp_initialization(sc);
   1309 
   1310 	return (0);
   1311 }
   1312 
   1313 static int
   1314 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
   1315 {
   1316 	struct ifbreq *req = arg;
   1317 	struct bridge_iflist *bif;
   1318 
   1319 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
   1320 	if (bif == NULL)
   1321 		return (ENOENT);
   1322 
   1323 	bif->bif_priority = req->ifbr_priority;
   1324 
   1325 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1326 		bstp_initialization(sc);
   1327 
   1328 	bridge_release_member(sc, bif);
   1329 
   1330 	return (0);
   1331 }
   1332 
   1333 #if defined(BRIDGE_IPF)
   1334 static int
   1335 bridge_ioctl_gfilt(struct bridge_softc *sc, void *arg)
   1336 {
   1337 	struct ifbrparam *param = arg;
   1338 
   1339 	param->ifbrp_filter = sc->sc_filter_flags;
   1340 
   1341 	return (0);
   1342 }
   1343 
   1344 static int
   1345 bridge_ioctl_sfilt(struct bridge_softc *sc, void *arg)
   1346 {
   1347 	struct ifbrparam *param = arg;
   1348 	uint32_t nflags, oflags;
   1349 
   1350 	if (param->ifbrp_filter & ~IFBF_FILT_MASK)
   1351 		return (EINVAL);
   1352 
   1353 	nflags = param->ifbrp_filter;
   1354 	oflags = sc->sc_filter_flags;
   1355 
   1356 	if ((nflags & IFBF_FILT_USEIPF) && !(oflags & IFBF_FILT_USEIPF)) {
   1357 		pfil_add_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
   1358 			sc->sc_if.if_pfil);
   1359 	}
   1360 	if (!(nflags & IFBF_FILT_USEIPF) && (oflags & IFBF_FILT_USEIPF)) {
   1361 		pfil_remove_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
   1362 			sc->sc_if.if_pfil);
   1363 	}
   1364 
   1365 	sc->sc_filter_flags = nflags;
   1366 
   1367 	return (0);
   1368 }
   1369 #endif /* BRIDGE_IPF */
   1370 
   1371 static int
   1372 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
   1373 {
   1374 	struct ifbreq *req = arg;
   1375 	struct bridge_iflist *bif;
   1376 
   1377 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
   1378 	if (bif == NULL)
   1379 		return (ENOENT);
   1380 
   1381 	bif->bif_path_cost = req->ifbr_path_cost;
   1382 
   1383 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1384 		bstp_initialization(sc);
   1385 
   1386 	bridge_release_member(sc, bif);
   1387 
   1388 	return (0);
   1389 }
   1390 
   1391 /*
   1392  * bridge_ifdetach:
   1393  *
   1394  *	Detach an interface from a bridge.  Called when a member
   1395  *	interface is detaching.
   1396  */
   1397 void
   1398 bridge_ifdetach(struct ifnet *ifp)
   1399 {
   1400 	struct bridge_softc *sc = ifp->if_bridge;
   1401 	struct ifbreq breq;
   1402 
   1403 	/* ioctl_lock should prevent this from happening */
   1404 	KASSERT(sc != NULL);
   1405 
   1406 	memset(&breq, 0, sizeof(breq));
   1407 	strlcpy(breq.ifbr_ifsname, ifp->if_xname, sizeof(breq.ifbr_ifsname));
   1408 
   1409 	(void) bridge_ioctl_del(sc, &breq);
   1410 }
   1411 
   1412 /*
   1413  * bridge_init:
   1414  *
   1415  *	Initialize a bridge interface.
   1416  */
   1417 static int
   1418 bridge_init(struct ifnet *ifp)
   1419 {
   1420 	struct bridge_softc *sc = ifp->if_softc;
   1421 
   1422 	if (ifp->if_flags & IFF_RUNNING)
   1423 		return (0);
   1424 
   1425 	callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
   1426 	    bridge_timer, sc);
   1427 
   1428 	ifp->if_flags |= IFF_RUNNING;
   1429 	bstp_initialization(sc);
   1430 	return (0);
   1431 }
   1432 
   1433 /*
   1434  * bridge_stop:
   1435  *
   1436  *	Stop the bridge interface.
   1437  */
   1438 static void
   1439 bridge_stop(struct ifnet *ifp, int disable)
   1440 {
   1441 	struct bridge_softc *sc = ifp->if_softc;
   1442 
   1443 	if ((ifp->if_flags & IFF_RUNNING) == 0)
   1444 		return;
   1445 
   1446 	callout_stop(&sc->sc_brcallout);
   1447 	bstp_stop(sc);
   1448 
   1449 	bridge_rtflush(sc, IFBF_FLUSHDYN);
   1450 
   1451 	ifp->if_flags &= ~IFF_RUNNING;
   1452 }
   1453 
   1454 /*
   1455  * bridge_enqueue:
   1456  *
   1457  *	Enqueue a packet on a bridge member interface.
   1458  */
   1459 void
   1460 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m,
   1461     int runfilt)
   1462 {
   1463 	ALTQ_DECL(struct altq_pktattr pktattr;)
   1464 	int len, error;
   1465 	short mflags;
   1466 
   1467 	/*
   1468 	 * Clear any in-bound checksum flags for this packet.
   1469 	 */
   1470 	m->m_pkthdr.csum_flags = 0;
   1471 
   1472 	if (runfilt) {
   1473 		if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
   1474 		    dst_ifp, PFIL_OUT) != 0) {
   1475 			if (m != NULL)
   1476 				m_freem(m);
   1477 			return;
   1478 		}
   1479 		if (m == NULL)
   1480 			return;
   1481 	}
   1482 
   1483 #ifdef ALTQ
   1484 	/*
   1485 	 * If ALTQ is enabled on the member interface, do
   1486 	 * classification; the queueing discipline might
   1487 	 * not require classification, but might require
   1488 	 * the address family/header pointer in the pktattr.
   1489 	 */
   1490 	if (ALTQ_IS_ENABLED(&dst_ifp->if_snd)) {
   1491 		/* XXX IFT_ETHER */
   1492 		altq_etherclassify(&dst_ifp->if_snd, m, &pktattr);
   1493 	}
   1494 #endif /* ALTQ */
   1495 
   1496 	len = m->m_pkthdr.len;
   1497 	m->m_flags |= M_PROTO1;
   1498 	mflags = m->m_flags;
   1499 
   1500 	IFQ_ENQUEUE(&dst_ifp->if_snd, m, &pktattr, error);
   1501 
   1502 	if (error) {
   1503 		/* mbuf is already freed */
   1504 		sc->sc_if.if_oerrors++;
   1505 		return;
   1506 	}
   1507 
   1508 	sc->sc_if.if_opackets++;
   1509 	sc->sc_if.if_obytes += len;
   1510 
   1511 	dst_ifp->if_obytes += len;
   1512 
   1513 	if (mflags & M_MCAST) {
   1514 		sc->sc_if.if_omcasts++;
   1515 		dst_ifp->if_omcasts++;
   1516 	}
   1517 
   1518 	if ((dst_ifp->if_flags & IFF_OACTIVE) == 0)
   1519 		(*dst_ifp->if_start)(dst_ifp);
   1520 }
   1521 
   1522 /*
   1523  * bridge_output:
   1524  *
   1525  *	Send output from a bridge member interface.  This
   1526  *	performs the bridging function for locally originated
   1527  *	packets.
   1528  *
   1529  *	The mbuf has the Ethernet header already attached.  We must
   1530  *	enqueue or free the mbuf before returning.
   1531  */
   1532 int
   1533 bridge_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa,
   1534     struct rtentry *rt)
   1535 {
   1536 	struct ether_header *eh;
   1537 	struct ifnet *dst_if;
   1538 	struct bridge_softc *sc;
   1539 #ifndef BRIDGE_MPSAFE
   1540 	int s;
   1541 #endif
   1542 
   1543 	if (m->m_len < ETHER_HDR_LEN) {
   1544 		m = m_pullup(m, ETHER_HDR_LEN);
   1545 		if (m == NULL)
   1546 			return (0);
   1547 	}
   1548 
   1549 	eh = mtod(m, struct ether_header *);
   1550 	sc = ifp->if_bridge;
   1551 
   1552 #ifndef BRIDGE_MPSAFE
   1553 	s = splnet();
   1554 #endif
   1555 
   1556 	/*
   1557 	 * If bridge is down, but the original output interface is up,
   1558 	 * go ahead and send out that interface.  Otherwise, the packet
   1559 	 * is dropped below.
   1560 	 */
   1561 	if (__predict_false(sc == NULL) ||
   1562 	    (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
   1563 		dst_if = ifp;
   1564 		goto sendunicast;
   1565 	}
   1566 
   1567 	/*
   1568 	 * If the packet is a multicast, or we don't know a better way to
   1569 	 * get there, send to all interfaces.
   1570 	 */
   1571 	if (ETHER_IS_MULTICAST(eh->ether_dhost))
   1572 		dst_if = NULL;
   1573 	else
   1574 		dst_if = bridge_rtlookup(sc, eh->ether_dhost);
   1575 	if (dst_if == NULL) {
   1576 		struct bridge_iflist *bif;
   1577 		struct mbuf *mc;
   1578 		int used = 0;
   1579 		int ss;
   1580 
   1581 		BRIDGE_PSZ_RENTER(ss);
   1582 		LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
   1583 			bif = bridge_try_hold_bif(bif);
   1584 			if (bif == NULL)
   1585 				continue;
   1586 			BRIDGE_PSZ_REXIT(ss);
   1587 
   1588 			dst_if = bif->bif_ifp;
   1589 			if ((dst_if->if_flags & IFF_RUNNING) == 0)
   1590 				goto next;
   1591 
   1592 			/*
   1593 			 * If this is not the original output interface,
   1594 			 * and the interface is participating in spanning
   1595 			 * tree, make sure the port is in a state that
   1596 			 * allows forwarding.
   1597 			 */
   1598 			if (dst_if != ifp &&
   1599 			    (bif->bif_flags & IFBIF_STP) != 0) {
   1600 				switch (bif->bif_state) {
   1601 				case BSTP_IFSTATE_BLOCKING:
   1602 				case BSTP_IFSTATE_LISTENING:
   1603 				case BSTP_IFSTATE_DISABLED:
   1604 					goto next;
   1605 				}
   1606 			}
   1607 
   1608 			if (LIST_NEXT(bif, bif_next) == NULL) {
   1609 				used = 1;
   1610 				mc = m;
   1611 			} else {
   1612 				mc = m_copym(m, 0, M_COPYALL, M_NOWAIT);
   1613 				if (mc == NULL) {
   1614 					sc->sc_if.if_oerrors++;
   1615 					goto next;
   1616 				}
   1617 			}
   1618 
   1619 			bridge_enqueue(sc, dst_if, mc, 0);
   1620 next:
   1621 			bridge_release_member(sc, bif);
   1622 			BRIDGE_PSZ_RENTER(ss);
   1623 		}
   1624 		BRIDGE_PSZ_REXIT(ss);
   1625 
   1626 		if (used == 0)
   1627 			m_freem(m);
   1628 #ifndef BRIDGE_MPSAFE
   1629 		splx(s);
   1630 #endif
   1631 		return (0);
   1632 	}
   1633 
   1634  sendunicast:
   1635 	/*
   1636 	 * XXX Spanning tree consideration here?
   1637 	 */
   1638 
   1639 	if ((dst_if->if_flags & IFF_RUNNING) == 0) {
   1640 		m_freem(m);
   1641 #ifndef BRIDGE_MPSAFE
   1642 		splx(s);
   1643 #endif
   1644 		return (0);
   1645 	}
   1646 
   1647 	bridge_enqueue(sc, dst_if, m, 0);
   1648 
   1649 #ifndef BRIDGE_MPSAFE
   1650 	splx(s);
   1651 #endif
   1652 	return (0);
   1653 }
   1654 
   1655 /*
   1656  * bridge_start:
   1657  *
   1658  *	Start output on a bridge.
   1659  *
   1660  *	NOTE: This routine should never be called in this implementation.
   1661  */
   1662 static void
   1663 bridge_start(struct ifnet *ifp)
   1664 {
   1665 
   1666 	printf("%s: bridge_start() called\n", ifp->if_xname);
   1667 }
   1668 
   1669 /*
   1670  * bridge_forward:
   1671  *
   1672  *	The forwarding function of the bridge.
   1673  */
   1674 static void
   1675 bridge_forward(void *v)
   1676 {
   1677 	struct bridge_softc *sc = v;
   1678 	struct mbuf *m;
   1679 	struct bridge_iflist *bif;
   1680 	struct ifnet *src_if, *dst_if;
   1681 	struct ether_header *eh;
   1682 #ifndef BRIDGE_MPSAFE
   1683 	int s;
   1684 
   1685 	KERNEL_LOCK(1, NULL);
   1686 	mutex_enter(softnet_lock);
   1687 #endif
   1688 
   1689 	if ((sc->sc_if.if_flags & IFF_RUNNING) == 0) {
   1690 #ifndef BRIDGE_MPSAFE
   1691 		mutex_exit(softnet_lock);
   1692 		KERNEL_UNLOCK_ONE(NULL);
   1693 #endif
   1694 		return;
   1695 	}
   1696 
   1697 #ifndef BRIDGE_MPSAFE
   1698 	s = splnet();
   1699 #endif
   1700 	while ((m = pktq_dequeue(sc->sc_fwd_pktq)) != NULL) {
   1701 		src_if = m->m_pkthdr.rcvif;
   1702 
   1703 		sc->sc_if.if_ipackets++;
   1704 		sc->sc_if.if_ibytes += m->m_pkthdr.len;
   1705 
   1706 		/*
   1707 		 * Look up the bridge_iflist.
   1708 		 */
   1709 		bif = bridge_lookup_member_if(sc, src_if);
   1710 		if (bif == NULL) {
   1711 			/* Interface is not a bridge member (anymore?) */
   1712 			m_freem(m);
   1713 			continue;
   1714 		}
   1715 
   1716 		if (bif->bif_flags & IFBIF_STP) {
   1717 			switch (bif->bif_state) {
   1718 			case BSTP_IFSTATE_BLOCKING:
   1719 			case BSTP_IFSTATE_LISTENING:
   1720 			case BSTP_IFSTATE_DISABLED:
   1721 				m_freem(m);
   1722 				bridge_release_member(sc, bif);
   1723 				continue;
   1724 			}
   1725 		}
   1726 
   1727 		eh = mtod(m, struct ether_header *);
   1728 
   1729 		/*
   1730 		 * If the interface is learning, and the source
   1731 		 * address is valid and not multicast, record
   1732 		 * the address.
   1733 		 */
   1734 		if ((bif->bif_flags & IFBIF_LEARNING) != 0 &&
   1735 		    ETHER_IS_MULTICAST(eh->ether_shost) == 0 &&
   1736 		    (eh->ether_shost[0] == 0 &&
   1737 		     eh->ether_shost[1] == 0 &&
   1738 		     eh->ether_shost[2] == 0 &&
   1739 		     eh->ether_shost[3] == 0 &&
   1740 		     eh->ether_shost[4] == 0 &&
   1741 		     eh->ether_shost[5] == 0) == 0) {
   1742 			(void) bridge_rtupdate(sc, eh->ether_shost,
   1743 			    src_if, 0, IFBAF_DYNAMIC);
   1744 		}
   1745 
   1746 		if ((bif->bif_flags & IFBIF_STP) != 0 &&
   1747 		    bif->bif_state == BSTP_IFSTATE_LEARNING) {
   1748 			m_freem(m);
   1749 			bridge_release_member(sc, bif);
   1750 			continue;
   1751 		}
   1752 
   1753 		bridge_release_member(sc, bif);
   1754 
   1755 		/*
   1756 		 * At this point, the port either doesn't participate
   1757 		 * in spanning tree or it is in the forwarding state.
   1758 		 */
   1759 
   1760 		/*
   1761 		 * If the packet is unicast, destined for someone on
   1762 		 * "this" side of the bridge, drop it.
   1763 		 */
   1764 		if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
   1765 			dst_if = bridge_rtlookup(sc, eh->ether_dhost);
   1766 			if (src_if == dst_if) {
   1767 				m_freem(m);
   1768 				continue;
   1769 			}
   1770 		} else {
   1771 			/* ...forward it to all interfaces. */
   1772 			sc->sc_if.if_imcasts++;
   1773 			dst_if = NULL;
   1774 		}
   1775 
   1776 		if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
   1777 		    m->m_pkthdr.rcvif, PFIL_IN) != 0) {
   1778 			if (m != NULL)
   1779 				m_freem(m);
   1780 			continue;
   1781 		}
   1782 		if (m == NULL)
   1783 			continue;
   1784 
   1785 		if (dst_if == NULL) {
   1786 			bridge_broadcast(sc, src_if, m);
   1787 			continue;
   1788 		}
   1789 
   1790 		/*
   1791 		 * At this point, we're dealing with a unicast frame
   1792 		 * going to a different interface.
   1793 		 */
   1794 		if ((dst_if->if_flags & IFF_RUNNING) == 0) {
   1795 			m_freem(m);
   1796 			continue;
   1797 		}
   1798 
   1799 		bif = bridge_lookup_member_if(sc, dst_if);
   1800 		if (bif == NULL) {
   1801 			/* Not a member of the bridge (anymore?) */
   1802 			m_freem(m);
   1803 			continue;
   1804 		}
   1805 
   1806 		if (bif->bif_flags & IFBIF_STP) {
   1807 			switch (bif->bif_state) {
   1808 			case BSTP_IFSTATE_DISABLED:
   1809 			case BSTP_IFSTATE_BLOCKING:
   1810 				m_freem(m);
   1811 				bridge_release_member(sc, bif);
   1812 				continue;
   1813 			}
   1814 		}
   1815 
   1816 		bridge_release_member(sc, bif);
   1817 
   1818 		bridge_enqueue(sc, dst_if, m, 1);
   1819 	}
   1820 #ifndef BRIDGE_MPSAFE
   1821 	splx(s);
   1822 	mutex_exit(softnet_lock);
   1823 	KERNEL_UNLOCK_ONE(NULL);
   1824 #endif
   1825 }
   1826 
   1827 static bool
   1828 bstp_state_before_learning(struct bridge_iflist *bif)
   1829 {
   1830 	if (bif->bif_flags & IFBIF_STP) {
   1831 		switch (bif->bif_state) {
   1832 		case BSTP_IFSTATE_BLOCKING:
   1833 		case BSTP_IFSTATE_LISTENING:
   1834 		case BSTP_IFSTATE_DISABLED:
   1835 			return true;
   1836 		}
   1837 	}
   1838 	return false;
   1839 }
   1840 
   1841 static bool
   1842 bridge_ourether(struct bridge_iflist *bif, struct ether_header *eh, int src)
   1843 {
   1844 	uint8_t *ether = src ? eh->ether_shost : eh->ether_dhost;
   1845 
   1846 	if (memcmp(CLLADDR(bif->bif_ifp->if_sadl), ether, ETHER_ADDR_LEN) == 0
   1847 #if NCARP > 0
   1848 	    || (bif->bif_ifp->if_carp &&
   1849 	        carp_ourether(bif->bif_ifp->if_carp, eh, IFT_ETHER, src) != NULL)
   1850 #endif /* NCARP > 0 */
   1851 	    )
   1852 		return true;
   1853 
   1854 	return false;
   1855 }
   1856 
   1857 /*
   1858  * bridge_input:
   1859  *
   1860  *	Receive input from a member interface.  Queue the packet for
   1861  *	bridging if it is not for us.
   1862  */
   1863 static void
   1864 bridge_input(struct ifnet *ifp, struct mbuf *m)
   1865 {
   1866 	struct bridge_softc *sc = ifp->if_bridge;
   1867 	struct bridge_iflist *bif;
   1868 	struct ether_header *eh;
   1869 
   1870 	if (__predict_false(sc == NULL) ||
   1871 	    (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
   1872 		ether_input(ifp, m);
   1873 		return;
   1874 	}
   1875 
   1876 	bif = bridge_lookup_member_if(sc, ifp);
   1877 	if (bif == NULL) {
   1878 		ether_input(ifp, m);
   1879 		return;
   1880 	}
   1881 
   1882 	eh = mtod(m, struct ether_header *);
   1883 
   1884 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
   1885 		if (memcmp(etherbroadcastaddr,
   1886 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
   1887 			m->m_flags |= M_BCAST;
   1888 		else
   1889 			m->m_flags |= M_MCAST;
   1890 	}
   1891 
   1892 	/*
   1893 	 * A 'fast' path for packets addressed to interfaces that are
   1894 	 * part of this bridge.
   1895 	 */
   1896 	if (!(m->m_flags & (M_BCAST|M_MCAST)) &&
   1897 	    !bstp_state_before_learning(bif)) {
   1898 		struct bridge_iflist *_bif;
   1899 		struct ifnet *_ifp = NULL;
   1900 		int s;
   1901 
   1902 		BRIDGE_PSZ_RENTER(s);
   1903 		LIST_FOREACH(_bif, &sc->sc_iflist, bif_next) {
   1904 			/* It is destined for us. */
   1905 			if (bridge_ourether(_bif, eh, 0)) {
   1906 				_bif = bridge_try_hold_bif(_bif);
   1907 				BRIDGE_PSZ_REXIT(s);
   1908 				if (_bif == NULL)
   1909 					goto out;
   1910 				if (_bif->bif_flags & IFBIF_LEARNING)
   1911 					(void) bridge_rtupdate(sc,
   1912 					    eh->ether_shost, ifp, 0, IFBAF_DYNAMIC);
   1913 				_ifp = m->m_pkthdr.rcvif = _bif->bif_ifp;
   1914 				bridge_release_member(sc, _bif);
   1915 				goto out;
   1916 			}
   1917 
   1918 			/* We just received a packet that we sent out. */
   1919 			if (bridge_ourether(_bif, eh, 1))
   1920 				break;
   1921 		}
   1922 		BRIDGE_PSZ_REXIT(s);
   1923 out:
   1924 
   1925 		if (_bif != NULL) {
   1926 			bridge_release_member(sc, bif);
   1927 			if (_ifp != NULL)
   1928 				ether_input(_ifp, m);
   1929 			else
   1930 				m_freem(m);
   1931 			return;
   1932 		}
   1933 	}
   1934 
   1935 	/* Tap off 802.1D packets; they do not get forwarded. */
   1936 	if (bif->bif_flags & IFBIF_STP &&
   1937 	    memcmp(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN) == 0) {
   1938 		bstp_input(sc, bif, m);
   1939 		bridge_release_member(sc, bif);
   1940 		return;
   1941 	}
   1942 
   1943 	/*
   1944 	 * A normal switch would discard the packet here, but that's not what
   1945 	 * we've done historically. This also prevents some obnoxious behaviour.
   1946 	 */
   1947 	if (bstp_state_before_learning(bif)) {
   1948 		bridge_release_member(sc, bif);
   1949 		ether_input(ifp, m);
   1950 		return;
   1951 	}
   1952 
   1953 	bridge_release_member(sc, bif);
   1954 
   1955 	/* Queue the packet for bridge forwarding. */
   1956 	if (__predict_false(!pktq_enqueue(sc->sc_fwd_pktq, m, 0)))
   1957 		m_freem(m);
   1958 }
   1959 
   1960 /*
   1961  * bridge_broadcast:
   1962  *
   1963  *	Send a frame to all interfaces that are members of
   1964  *	the bridge, except for the one on which the packet
   1965  *	arrived.
   1966  */
   1967 static void
   1968 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
   1969     struct mbuf *m)
   1970 {
   1971 	struct bridge_iflist *bif;
   1972 	struct mbuf *mc;
   1973 	struct ifnet *dst_if;
   1974 	bool used, bmcast;
   1975 	int s;
   1976 
   1977 	used = bmcast = m->m_flags & (M_BCAST|M_MCAST);
   1978 
   1979 	BRIDGE_PSZ_RENTER(s);
   1980 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
   1981 		bif = bridge_try_hold_bif(bif);
   1982 		if (bif == NULL)
   1983 			continue;
   1984 		BRIDGE_PSZ_REXIT(s);
   1985 
   1986 		dst_if = bif->bif_ifp;
   1987 		if (dst_if == src_if)
   1988 			goto next;
   1989 
   1990 		if (bif->bif_flags & IFBIF_STP) {
   1991 			switch (bif->bif_state) {
   1992 			case BSTP_IFSTATE_BLOCKING:
   1993 			case BSTP_IFSTATE_DISABLED:
   1994 				goto next;
   1995 			}
   1996 		}
   1997 
   1998 		if ((bif->bif_flags & IFBIF_DISCOVER) == 0 && !bmcast)
   1999 			goto next;
   2000 
   2001 		if ((dst_if->if_flags & IFF_RUNNING) == 0)
   2002 			goto next;
   2003 
   2004 		if (!used && LIST_NEXT(bif, bif_next) == NULL) {
   2005 			mc = m;
   2006 			used = true;
   2007 		} else {
   2008 			mc = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
   2009 			if (mc == NULL) {
   2010 				sc->sc_if.if_oerrors++;
   2011 				goto next;
   2012 			}
   2013 		}
   2014 
   2015 		bridge_enqueue(sc, dst_if, mc, 1);
   2016 next:
   2017 		bridge_release_member(sc, bif);
   2018 		BRIDGE_PSZ_RENTER(s);
   2019 	}
   2020 	BRIDGE_PSZ_REXIT(s);
   2021 
   2022 	if (bmcast)
   2023 		ether_input(src_if, m);
   2024 	else if (!used)
   2025 		m_freem(m);
   2026 }
   2027 
   2028 static int
   2029 bridge_rtalloc(struct bridge_softc *sc, const uint8_t *dst,
   2030     struct bridge_rtnode **brtp)
   2031 {
   2032 	struct bridge_rtnode *brt;
   2033 	int error;
   2034 
   2035 	if (sc->sc_brtcnt >= sc->sc_brtmax)
   2036 		return ENOSPC;
   2037 
   2038 	/*
   2039 	 * Allocate a new bridge forwarding node, and
   2040 	 * initialize the expiration time and Ethernet
   2041 	 * address.
   2042 	 */
   2043 	brt = pool_get(&bridge_rtnode_pool, PR_NOWAIT);
   2044 	if (brt == NULL)
   2045 		return ENOMEM;
   2046 
   2047 	memset(brt, 0, sizeof(*brt));
   2048 	brt->brt_expire = time_uptime + sc->sc_brttimeout;
   2049 	brt->brt_flags = IFBAF_DYNAMIC;
   2050 	memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
   2051 
   2052 	BRIDGE_RT_INTR_LOCK(sc);
   2053 	error = bridge_rtnode_insert(sc, brt);
   2054 	BRIDGE_RT_INTR_UNLOCK(sc);
   2055 
   2056 	if (error != 0) {
   2057 		pool_put(&bridge_rtnode_pool, brt);
   2058 		return error;
   2059 	}
   2060 
   2061 	*brtp = brt;
   2062 	return 0;
   2063 }
   2064 
   2065 /*
   2066  * bridge_rtupdate:
   2067  *
   2068  *	Add a bridge routing entry.
   2069  */
   2070 static int
   2071 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
   2072     struct ifnet *dst_if, int setflags, uint8_t flags)
   2073 {
   2074 	struct bridge_rtnode *brt;
   2075 	int s;
   2076 
   2077 again:
   2078 	/*
   2079 	 * A route for this destination might already exist.  If so,
   2080 	 * update it, otherwise create a new one.
   2081 	 */
   2082 	BRIDGE_RT_RENTER(s);
   2083 	brt = bridge_rtnode_lookup(sc, dst);
   2084 
   2085 	if (brt != NULL) {
   2086 		brt->brt_ifp = dst_if;
   2087 		if (setflags) {
   2088 			brt->brt_flags = flags;
   2089 			if (flags & IFBAF_STATIC)
   2090 				brt->brt_expire = 0;
   2091 			else
   2092 				brt->brt_expire = time_uptime + sc->sc_brttimeout;
   2093 		} else {
   2094 			if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
   2095 				brt->brt_expire = time_uptime + sc->sc_brttimeout;
   2096 		}
   2097 	}
   2098 	BRIDGE_RT_REXIT(s);
   2099 
   2100 	if (brt == NULL) {
   2101 		int r;
   2102 
   2103 		r = bridge_rtalloc(sc, dst, &brt);
   2104 		if (r != 0)
   2105 			return r;
   2106 		goto again;
   2107 	}
   2108 
   2109 	return 0;
   2110 }
   2111 
   2112 /*
   2113  * bridge_rtlookup:
   2114  *
   2115  *	Lookup the destination interface for an address.
   2116  */
   2117 static struct ifnet *
   2118 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr)
   2119 {
   2120 	struct bridge_rtnode *brt;
   2121 	struct ifnet *ifs = NULL;
   2122 	int s;
   2123 
   2124 	BRIDGE_RT_RENTER(s);
   2125 	brt = bridge_rtnode_lookup(sc, addr);
   2126 	if (brt != NULL)
   2127 		ifs = brt->brt_ifp;
   2128 	BRIDGE_RT_REXIT(s);
   2129 
   2130 	return ifs;
   2131 }
   2132 
   2133 typedef bool (*bridge_iterate_cb_t)
   2134     (struct bridge_softc *, struct bridge_rtnode *, bool *, void *);
   2135 
   2136 /*
   2137  * bridge_rtlist_iterate_remove:
   2138  *
   2139  *	It iterates on sc->sc_rtlist and removes rtnodes of it which func
   2140  *	callback judges to remove. Removals of rtnodes are done in a manner
   2141  *	of pserialize. To this end, all kmem_* operations are placed out of
   2142  *	mutexes.
   2143  */
   2144 static void
   2145 bridge_rtlist_iterate_remove(struct bridge_softc *sc, bridge_iterate_cb_t func, void *arg)
   2146 {
   2147 	struct bridge_rtnode *brt, *nbrt;
   2148 	struct bridge_rtnode **brt_list;
   2149 	int i, count;
   2150 
   2151 retry:
   2152 	count = sc->sc_brtcnt;
   2153 	if (count == 0)
   2154 		return;
   2155 	brt_list = kmem_alloc(sizeof(struct bridge_rtnode *) * count, KM_SLEEP);
   2156 
   2157 	BRIDGE_RT_LOCK(sc);
   2158 	BRIDGE_RT_INTR_LOCK(sc);
   2159 	if (__predict_false(sc->sc_brtcnt > count)) {
   2160 		/* The rtnodes increased, we need more memory */
   2161 		BRIDGE_RT_INTR_UNLOCK(sc);
   2162 		BRIDGE_RT_UNLOCK(sc);
   2163 		kmem_free(brt_list, sizeof(*brt_list) * count);
   2164 		goto retry;
   2165 	}
   2166 
   2167 	i = 0;
   2168 	LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
   2169 		bool need_break = false;
   2170 		if (func(sc, brt, &need_break, arg)) {
   2171 			bridge_rtnode_remove(sc, brt);
   2172 			brt_list[i++] = brt;
   2173 		}
   2174 		if (need_break)
   2175 			break;
   2176 	}
   2177 	BRIDGE_RT_INTR_UNLOCK(sc);
   2178 
   2179 	if (i > 0)
   2180 		BRIDGE_RT_PSZ_PERFORM(sc);
   2181 	BRIDGE_RT_UNLOCK(sc);
   2182 
   2183 	while (--i >= 0)
   2184 		bridge_rtnode_destroy(brt_list[i]);
   2185 
   2186 	kmem_free(brt_list, sizeof(*brt_list) * count);
   2187 }
   2188 
   2189 static bool
   2190 bridge_rttrim0_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
   2191     bool *need_break, void *arg)
   2192 {
   2193 	if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
   2194 		/* Take into account of the subsequent removal */
   2195 		if ((sc->sc_brtcnt - 1) <= sc->sc_brtmax)
   2196 			*need_break = true;
   2197 		return true;
   2198 	} else
   2199 		return false;
   2200 }
   2201 
   2202 static void
   2203 bridge_rttrim0(struct bridge_softc *sc)
   2204 {
   2205 	bridge_rtlist_iterate_remove(sc, bridge_rttrim0_cb, NULL);
   2206 }
   2207 
   2208 /*
   2209  * bridge_rttrim:
   2210  *
   2211  *	Trim the routine table so that we have a number
   2212  *	of routing entries less than or equal to the
   2213  *	maximum number.
   2214  */
   2215 static void
   2216 bridge_rttrim(struct bridge_softc *sc)
   2217 {
   2218 
   2219 	/* Make sure we actually need to do this. */
   2220 	if (sc->sc_brtcnt <= sc->sc_brtmax)
   2221 		return;
   2222 
   2223 	/* Force an aging cycle; this might trim enough addresses. */
   2224 	bridge_rtage(sc);
   2225 	if (sc->sc_brtcnt <= sc->sc_brtmax)
   2226 		return;
   2227 
   2228 	bridge_rttrim0(sc);
   2229 
   2230 	return;
   2231 }
   2232 
   2233 /*
   2234  * bridge_timer:
   2235  *
   2236  *	Aging timer for the bridge.
   2237  */
   2238 static void
   2239 bridge_timer(void *arg)
   2240 {
   2241 	struct bridge_softc *sc = arg;
   2242 
   2243 	workqueue_enqueue(sc->sc_rtage_wq, &bridge_rtage_wk, NULL);
   2244 }
   2245 
   2246 static void
   2247 bridge_rtage_work(struct work *wk, void *arg)
   2248 {
   2249 	struct bridge_softc *sc = arg;
   2250 
   2251 	KASSERT(wk == &bridge_rtage_wk);
   2252 
   2253 	bridge_rtage(sc);
   2254 
   2255 	if (sc->sc_if.if_flags & IFF_RUNNING)
   2256 		callout_reset(&sc->sc_brcallout,
   2257 		    bridge_rtable_prune_period * hz, bridge_timer, sc);
   2258 }
   2259 
   2260 static bool
   2261 bridge_rtage_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
   2262     bool *need_break, void *arg)
   2263 {
   2264 	if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
   2265 	    time_uptime >= brt->brt_expire)
   2266 		return true;
   2267 	else
   2268 		return false;
   2269 }
   2270 
   2271 /*
   2272  * bridge_rtage:
   2273  *
   2274  *	Perform an aging cycle.
   2275  */
   2276 static void
   2277 bridge_rtage(struct bridge_softc *sc)
   2278 {
   2279 	bridge_rtlist_iterate_remove(sc, bridge_rtage_cb, NULL);
   2280 }
   2281 
   2282 
   2283 static bool
   2284 bridge_rtflush_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
   2285     bool *need_break, void *arg)
   2286 {
   2287 	int full = *(int*)arg;
   2288 
   2289 	if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
   2290 		return true;
   2291 	else
   2292 		return false;
   2293 }
   2294 
   2295 /*
   2296  * bridge_rtflush:
   2297  *
   2298  *	Remove all dynamic addresses from the bridge.
   2299  */
   2300 static void
   2301 bridge_rtflush(struct bridge_softc *sc, int full)
   2302 {
   2303 	bridge_rtlist_iterate_remove(sc, bridge_rtflush_cb, &full);
   2304 }
   2305 
   2306 /*
   2307  * bridge_rtdaddr:
   2308  *
   2309  *	Remove an address from the table.
   2310  */
   2311 static int
   2312 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr)
   2313 {
   2314 	struct bridge_rtnode *brt;
   2315 
   2316 	BRIDGE_RT_LOCK(sc);
   2317 	BRIDGE_RT_INTR_LOCK(sc);
   2318 	if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL) {
   2319 		BRIDGE_RT_INTR_UNLOCK(sc);
   2320 		BRIDGE_RT_UNLOCK(sc);
   2321 		return ENOENT;
   2322 	}
   2323 	bridge_rtnode_remove(sc, brt);
   2324 	BRIDGE_RT_INTR_UNLOCK(sc);
   2325 	BRIDGE_RT_PSZ_PERFORM(sc);
   2326 	BRIDGE_RT_UNLOCK(sc);
   2327 
   2328 	bridge_rtnode_destroy(brt);
   2329 
   2330 	return 0;
   2331 }
   2332 
   2333 /*
   2334  * bridge_rtdelete:
   2335  *
   2336  *	Delete routes to a speicifc member interface.
   2337  */
   2338 static void
   2339 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp)
   2340 {
   2341 	struct bridge_rtnode *brt, *nbrt;
   2342 
   2343 	BRIDGE_RT_LOCK(sc);
   2344 	BRIDGE_RT_INTR_LOCK(sc);
   2345 	LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
   2346 		if (brt->brt_ifp == ifp)
   2347 			break;
   2348 	}
   2349 	if (brt == NULL) {
   2350 		BRIDGE_RT_INTR_UNLOCK(sc);
   2351 		BRIDGE_RT_UNLOCK(sc);
   2352 		return;
   2353 	}
   2354 	bridge_rtnode_remove(sc, brt);
   2355 	BRIDGE_RT_INTR_UNLOCK(sc);
   2356 	BRIDGE_RT_PSZ_PERFORM(sc);
   2357 	BRIDGE_RT_UNLOCK(sc);
   2358 
   2359 	bridge_rtnode_destroy(brt);
   2360 }
   2361 
   2362 /*
   2363  * bridge_rtable_init:
   2364  *
   2365  *	Initialize the route table for this bridge.
   2366  */
   2367 static void
   2368 bridge_rtable_init(struct bridge_softc *sc)
   2369 {
   2370 	int i;
   2371 
   2372 	sc->sc_rthash = kmem_alloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
   2373 	    KM_SLEEP);
   2374 
   2375 	for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
   2376 		LIST_INIT(&sc->sc_rthash[i]);
   2377 
   2378 	sc->sc_rthash_key = cprng_fast32();
   2379 
   2380 	LIST_INIT(&sc->sc_rtlist);
   2381 
   2382 	sc->sc_rtlist_intr_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
   2383 #ifdef BRIDGE_MPSAFE
   2384 	sc->sc_rtlist_psz = pserialize_create();
   2385 	sc->sc_rtlist_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_SOFTNET);
   2386 #else
   2387 	sc->sc_rtlist_psz = NULL;
   2388 	sc->sc_rtlist_lock = NULL;
   2389 #endif
   2390 }
   2391 
   2392 /*
   2393  * bridge_rtable_fini:
   2394  *
   2395  *	Deconstruct the route table for this bridge.
   2396  */
   2397 static void
   2398 bridge_rtable_fini(struct bridge_softc *sc)
   2399 {
   2400 
   2401 	kmem_free(sc->sc_rthash, sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE);
   2402 	if (sc->sc_rtlist_intr_lock)
   2403 		mutex_obj_free(sc->sc_rtlist_intr_lock);
   2404 	if (sc->sc_rtlist_lock)
   2405 		mutex_obj_free(sc->sc_rtlist_lock);
   2406 	if (sc->sc_rtlist_psz)
   2407 		pserialize_destroy(sc->sc_rtlist_psz);
   2408 }
   2409 
   2410 /*
   2411  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
   2412  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
   2413  */
   2414 #define	mix(a, b, c)							\
   2415 do {									\
   2416 	a -= b; a -= c; a ^= (c >> 13);					\
   2417 	b -= c; b -= a; b ^= (a << 8);					\
   2418 	c -= a; c -= b; c ^= (b >> 13);					\
   2419 	a -= b; a -= c; a ^= (c >> 12);					\
   2420 	b -= c; b -= a; b ^= (a << 16);					\
   2421 	c -= a; c -= b; c ^= (b >> 5);					\
   2422 	a -= b; a -= c; a ^= (c >> 3);					\
   2423 	b -= c; b -= a; b ^= (a << 10);					\
   2424 	c -= a; c -= b; c ^= (b >> 15);					\
   2425 } while (/*CONSTCOND*/0)
   2426 
   2427 static inline uint32_t
   2428 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
   2429 {
   2430 	uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
   2431 
   2432 	b += addr[5] << 8;
   2433 	b += addr[4];
   2434 	a += addr[3] << 24;
   2435 	a += addr[2] << 16;
   2436 	a += addr[1] << 8;
   2437 	a += addr[0];
   2438 
   2439 	mix(a, b, c);
   2440 
   2441 	return (c & BRIDGE_RTHASH_MASK);
   2442 }
   2443 
   2444 #undef mix
   2445 
   2446 /*
   2447  * bridge_rtnode_lookup:
   2448  *
   2449  *	Look up a bridge route node for the specified destination.
   2450  */
   2451 static struct bridge_rtnode *
   2452 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr)
   2453 {
   2454 	struct bridge_rtnode *brt;
   2455 	uint32_t hash;
   2456 	int dir;
   2457 
   2458 	hash = bridge_rthash(sc, addr);
   2459 	LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
   2460 		dir = memcmp(addr, brt->brt_addr, ETHER_ADDR_LEN);
   2461 		if (dir == 0)
   2462 			return (brt);
   2463 		if (dir > 0)
   2464 			return (NULL);
   2465 	}
   2466 
   2467 	return (NULL);
   2468 }
   2469 
   2470 /*
   2471  * bridge_rtnode_insert:
   2472  *
   2473  *	Insert the specified bridge node into the route table.  We
   2474  *	assume the entry is not already in the table.
   2475  */
   2476 static int
   2477 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
   2478 {
   2479 	struct bridge_rtnode *lbrt;
   2480 	uint32_t hash;
   2481 	int dir;
   2482 
   2483 	KASSERT(BRIDGE_RT_INTR_LOCKED(sc));
   2484 
   2485 	hash = bridge_rthash(sc, brt->brt_addr);
   2486 
   2487 	lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
   2488 	if (lbrt == NULL) {
   2489 		LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
   2490 		goto out;
   2491 	}
   2492 
   2493 	do {
   2494 		dir = memcmp(brt->brt_addr, lbrt->brt_addr, ETHER_ADDR_LEN);
   2495 		if (dir == 0)
   2496 			return (EEXIST);
   2497 		if (dir > 0) {
   2498 			LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
   2499 			goto out;
   2500 		}
   2501 		if (LIST_NEXT(lbrt, brt_hash) == NULL) {
   2502 			LIST_INSERT_AFTER(lbrt, brt, brt_hash);
   2503 			goto out;
   2504 		}
   2505 		lbrt = LIST_NEXT(lbrt, brt_hash);
   2506 	} while (lbrt != NULL);
   2507 
   2508 #ifdef DIAGNOSTIC
   2509 	panic("bridge_rtnode_insert: impossible");
   2510 #endif
   2511 
   2512  out:
   2513 	LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
   2514 	sc->sc_brtcnt++;
   2515 
   2516 	return (0);
   2517 }
   2518 
   2519 /*
   2520  * bridge_rtnode_remove:
   2521  *
   2522  *	Remove a bridge rtnode from the rthash and the rtlist of a bridge.
   2523  */
   2524 static void
   2525 bridge_rtnode_remove(struct bridge_softc *sc, struct bridge_rtnode *brt)
   2526 {
   2527 
   2528 	KASSERT(BRIDGE_RT_INTR_LOCKED(sc));
   2529 
   2530 	LIST_REMOVE(brt, brt_hash);
   2531 	LIST_REMOVE(brt, brt_list);
   2532 	sc->sc_brtcnt--;
   2533 }
   2534 
   2535 /*
   2536  * bridge_rtnode_destroy:
   2537  *
   2538  *	Destroy a bridge rtnode.
   2539  */
   2540 static void
   2541 bridge_rtnode_destroy(struct bridge_rtnode *brt)
   2542 {
   2543 
   2544 	pool_put(&bridge_rtnode_pool, brt);
   2545 }
   2546 
   2547 #if defined(BRIDGE_IPF)
   2548 extern pfil_head_t *inet_pfil_hook;                 /* XXX */
   2549 extern pfil_head_t *inet6_pfil_hook;                /* XXX */
   2550 
   2551 /*
   2552  * Send bridge packets through IPF if they are one of the types IPF can deal
   2553  * with, or if they are ARP or REVARP.  (IPF will pass ARP and REVARP without
   2554  * question.)
   2555  */
   2556 static int
   2557 bridge_ipf(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
   2558 {
   2559 	int snap, error;
   2560 	struct ether_header *eh1, eh2;
   2561 	struct llc llc1;
   2562 	uint16_t ether_type;
   2563 
   2564 	snap = 0;
   2565 	error = -1;	/* Default error if not error == 0 */
   2566 	eh1 = mtod(*mp, struct ether_header *);
   2567 	ether_type = ntohs(eh1->ether_type);
   2568 
   2569 	/*
   2570 	 * Check for SNAP/LLC.
   2571 	 */
   2572         if (ether_type < ETHERMTU) {
   2573                 struct llc *llc2 = (struct llc *)(eh1 + 1);
   2574 
   2575                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
   2576                     llc2->llc_dsap == LLC_SNAP_LSAP &&
   2577                     llc2->llc_ssap == LLC_SNAP_LSAP &&
   2578                     llc2->llc_control == LLC_UI) {
   2579                 	ether_type = htons(llc2->llc_un.type_snap.ether_type);
   2580 			snap = 1;
   2581                 }
   2582         }
   2583 
   2584 	/*
   2585 	 * If we're trying to filter bridge traffic, don't look at anything
   2586 	 * other than IP and ARP traffic.  If the filter doesn't understand
   2587 	 * IPv6, don't allow IPv6 through the bridge either.  This is lame
   2588 	 * since if we really wanted, say, an AppleTalk filter, we are hosed,
   2589 	 * but of course we don't have an AppleTalk filter to begin with.
   2590 	 * (Note that since IPF doesn't understand ARP it will pass *ALL*
   2591 	 * ARP traffic.)
   2592 	 */
   2593 	switch (ether_type) {
   2594 		case ETHERTYPE_ARP:
   2595 		case ETHERTYPE_REVARP:
   2596 			return 0; /* Automatically pass */
   2597 		case ETHERTYPE_IP:
   2598 # ifdef INET6
   2599 		case ETHERTYPE_IPV6:
   2600 # endif /* INET6 */
   2601 			break;
   2602 		default:
   2603 			goto bad;
   2604 	}
   2605 
   2606 	/* Strip off the Ethernet header and keep a copy. */
   2607 	m_copydata(*mp, 0, ETHER_HDR_LEN, (void *) &eh2);
   2608 	m_adj(*mp, ETHER_HDR_LEN);
   2609 
   2610 	/* Strip off snap header, if present */
   2611 	if (snap) {
   2612 		m_copydata(*mp, 0, sizeof(struct llc), (void *) &llc1);
   2613 		m_adj(*mp, sizeof(struct llc));
   2614 	}
   2615 
   2616 	/*
   2617 	 * Check basic packet sanity and run IPF through pfil.
   2618 	 */
   2619 	KASSERT(!cpu_intr_p());
   2620 	switch (ether_type)
   2621 	{
   2622 	case ETHERTYPE_IP :
   2623 		error = (dir == PFIL_IN) ? bridge_ip_checkbasic(mp) : 0;
   2624 		if (error == 0)
   2625 			error = pfil_run_hooks(inet_pfil_hook, mp, ifp, dir);
   2626 		break;
   2627 # ifdef INET6
   2628 	case ETHERTYPE_IPV6 :
   2629 		error = (dir == PFIL_IN) ? bridge_ip6_checkbasic(mp) : 0;
   2630 		if (error == 0)
   2631 			error = pfil_run_hooks(inet6_pfil_hook, mp, ifp, dir);
   2632 		break;
   2633 # endif
   2634 	default :
   2635 		error = 0;
   2636 		break;
   2637 	}
   2638 
   2639 	if (*mp == NULL)
   2640 		return error;
   2641 	if (error != 0)
   2642 		goto bad;
   2643 
   2644 	error = -1;
   2645 
   2646 	/*
   2647 	 * Finally, put everything back the way it was and return
   2648 	 */
   2649 	if (snap) {
   2650 		M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
   2651 		if (*mp == NULL)
   2652 			return error;
   2653 		bcopy(&llc1, mtod(*mp, void *), sizeof(struct llc));
   2654 	}
   2655 
   2656 	M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
   2657 	if (*mp == NULL)
   2658 		return error;
   2659 	bcopy(&eh2, mtod(*mp, void *), ETHER_HDR_LEN);
   2660 
   2661 	return 0;
   2662 
   2663     bad:
   2664 	m_freem(*mp);
   2665 	*mp = NULL;
   2666 	return error;
   2667 }
   2668 
   2669 /*
   2670  * Perform basic checks on header size since
   2671  * IPF assumes ip_input has already processed
   2672  * it for it.  Cut-and-pasted from ip_input.c.
   2673  * Given how simple the IPv6 version is,
   2674  * does the IPv4 version really need to be
   2675  * this complicated?
   2676  *
   2677  * XXX Should we update ipstat here, or not?
   2678  * XXX Right now we update ipstat but not
   2679  * XXX csum_counter.
   2680  */
   2681 static int
   2682 bridge_ip_checkbasic(struct mbuf **mp)
   2683 {
   2684 	struct mbuf *m = *mp;
   2685 	struct ip *ip;
   2686 	int len, hlen;
   2687 
   2688 	if (*mp == NULL)
   2689 		return -1;
   2690 
   2691 	if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
   2692 		if ((m = m_copyup(m, sizeof(struct ip),
   2693 			(max_linkhdr + 3) & ~3)) == NULL) {
   2694 			/* XXXJRT new stat, please */
   2695 			ip_statinc(IP_STAT_TOOSMALL);
   2696 			goto bad;
   2697 		}
   2698 	} else if (__predict_false(m->m_len < sizeof (struct ip))) {
   2699 		if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
   2700 			ip_statinc(IP_STAT_TOOSMALL);
   2701 			goto bad;
   2702 		}
   2703 	}
   2704 	ip = mtod(m, struct ip *);
   2705 	if (ip == NULL) goto bad;
   2706 
   2707 	if (ip->ip_v != IPVERSION) {
   2708 		ip_statinc(IP_STAT_BADVERS);
   2709 		goto bad;
   2710 	}
   2711 	hlen = ip->ip_hl << 2;
   2712 	if (hlen < sizeof(struct ip)) { /* minimum header length */
   2713 		ip_statinc(IP_STAT_BADHLEN);
   2714 		goto bad;
   2715 	}
   2716 	if (hlen > m->m_len) {
   2717 		if ((m = m_pullup(m, hlen)) == 0) {
   2718 			ip_statinc(IP_STAT_BADHLEN);
   2719 			goto bad;
   2720 		}
   2721 		ip = mtod(m, struct ip *);
   2722 		if (ip == NULL) goto bad;
   2723 	}
   2724 
   2725         switch (m->m_pkthdr.csum_flags &
   2726                 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) |
   2727                  M_CSUM_IPv4_BAD)) {
   2728         case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
   2729                 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); */
   2730                 goto bad;
   2731 
   2732         case M_CSUM_IPv4:
   2733                 /* Checksum was okay. */
   2734                 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); */
   2735                 break;
   2736 
   2737         default:
   2738                 /* Must compute it ourselves. */
   2739                 /* INET_CSUM_COUNTER_INCR(&ip_swcsum); */
   2740                 if (in_cksum(m, hlen) != 0)
   2741                         goto bad;
   2742                 break;
   2743         }
   2744 
   2745         /* Retrieve the packet length. */
   2746         len = ntohs(ip->ip_len);
   2747 
   2748         /*
   2749          * Check for additional length bogosity
   2750          */
   2751         if (len < hlen) {
   2752 		ip_statinc(IP_STAT_BADLEN);
   2753                 goto bad;
   2754         }
   2755 
   2756         /*
   2757          * Check that the amount of data in the buffers
   2758          * is as at least much as the IP header would have us expect.
   2759          * Drop packet if shorter than we expect.
   2760          */
   2761         if (m->m_pkthdr.len < len) {
   2762 		ip_statinc(IP_STAT_TOOSHORT);
   2763                 goto bad;
   2764         }
   2765 
   2766 	/* Checks out, proceed */
   2767 	*mp = m;
   2768 	return 0;
   2769 
   2770     bad:
   2771 	*mp = m;
   2772 	return -1;
   2773 }
   2774 
   2775 # ifdef INET6
   2776 /*
   2777  * Same as above, but for IPv6.
   2778  * Cut-and-pasted from ip6_input.c.
   2779  * XXX Should we update ip6stat, or not?
   2780  */
   2781 static int
   2782 bridge_ip6_checkbasic(struct mbuf **mp)
   2783 {
   2784 	struct mbuf *m = *mp;
   2785 	struct ip6_hdr *ip6;
   2786 
   2787         /*
   2788          * If the IPv6 header is not aligned, slurp it up into a new
   2789          * mbuf with space for link headers, in the event we forward
   2790          * it.  Otherwise, if it is aligned, make sure the entire base
   2791          * IPv6 header is in the first mbuf of the chain.
   2792          */
   2793         if (IP6_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
   2794                 struct ifnet *inifp = m->m_pkthdr.rcvif;
   2795                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
   2796                                   (max_linkhdr + 3) & ~3)) == NULL) {
   2797                         /* XXXJRT new stat, please */
   2798 			ip6_statinc(IP6_STAT_TOOSMALL);
   2799                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
   2800                         goto bad;
   2801                 }
   2802         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
   2803                 struct ifnet *inifp = m->m_pkthdr.rcvif;
   2804                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
   2805 			ip6_statinc(IP6_STAT_TOOSMALL);
   2806                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
   2807                         goto bad;
   2808                 }
   2809         }
   2810 
   2811         ip6 = mtod(m, struct ip6_hdr *);
   2812 
   2813         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
   2814 		ip6_statinc(IP6_STAT_BADVERS);
   2815                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
   2816                 goto bad;
   2817         }
   2818 
   2819 	/* Checks out, proceed */
   2820 	*mp = m;
   2821 	return 0;
   2822 
   2823     bad:
   2824 	*mp = m;
   2825 	return -1;
   2826 }
   2827 # endif /* INET6 */
   2828 #endif /* BRIDGE_IPF */
   2829