Home | History | Annotate | Line # | Download | only in net
if_bridge.c revision 1.106
      1 /*	$NetBSD: if_bridge.c,v 1.106 2016/02/09 08:32:12 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 (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 1.106 2016/02/09 08:32:12 ozaki-r 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 #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_INTR_LOCK(_sc)	mutex_enter((_sc)->sc_rtlist_intr_lock)
    185 #define BRIDGE_RT_INTR_UNLOCK(_sc)	mutex_exit((_sc)->sc_rtlist_intr_lock)
    186 #define BRIDGE_RT_INTR_LOCKED(_sc)	mutex_owned((_sc)->sc_rtlist_intr_lock)
    187 
    188 #define BRIDGE_RT_LOCK(_sc)	if ((_sc)->sc_rtlist_lock) \
    189 					mutex_enter((_sc)->sc_rtlist_lock)
    190 #define BRIDGE_RT_UNLOCK(_sc)	if ((_sc)->sc_rtlist_lock) \
    191 					mutex_exit((_sc)->sc_rtlist_lock)
    192 #define BRIDGE_RT_LOCKED(_sc)	(!(_sc)->sc_rtlist_lock || \
    193 				 mutex_owned((_sc)->sc_rtlist_lock))
    194 
    195 #define BRIDGE_RT_PSZ_PERFORM(_sc) \
    196 				if ((_sc)->sc_rtlist_psz != NULL) \
    197 					pserialize_perform((_sc)->sc_rtlist_psz);
    198 
    199 #ifdef BRIDGE_MPSAFE
    200 #define BRIDGE_RT_RENTER(__s)	do { \
    201 					if (!cpu_intr_p()) \
    202 						__s = pserialize_read_enter(); \
    203 					else \
    204 						__s = splhigh(); \
    205 				} while (0)
    206 #define BRIDGE_RT_REXIT(__s)	do { \
    207 					if (!cpu_intr_p()) \
    208 						pserialize_read_exit(__s); \
    209 					else \
    210 						splx(__s); \
    211 				} while (0)
    212 #else /* BRIDGE_MPSAFE */
    213 #define BRIDGE_RT_RENTER(__s)	do { __s = 0; } while (0)
    214 #define BRIDGE_RT_REXIT(__s)	do { (void)__s; } while (0)
    215 #endif /* BRIDGE_MPSAFE */
    216 
    217 int	bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
    218 
    219 static struct pool bridge_rtnode_pool;
    220 static struct work bridge_rtage_wk;
    221 
    222 static int	bridge_clone_create(struct if_clone *, int);
    223 static int	bridge_clone_destroy(struct ifnet *);
    224 
    225 static int	bridge_ioctl(struct ifnet *, u_long, void *);
    226 static int	bridge_init(struct ifnet *);
    227 static void	bridge_stop(struct ifnet *, int);
    228 static void	bridge_start(struct ifnet *);
    229 
    230 static void	bridge_input(struct ifnet *, struct mbuf *);
    231 static void	bridge_forward(void *);
    232 
    233 static void	bridge_timer(void *);
    234 
    235 static void	bridge_broadcast(struct bridge_softc *, struct ifnet *,
    236 				 struct mbuf *);
    237 
    238 static int	bridge_rtupdate(struct bridge_softc *, const uint8_t *,
    239 				struct ifnet *, int, uint8_t);
    240 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *);
    241 static void	bridge_rttrim(struct bridge_softc *);
    242 static void	bridge_rtage(struct bridge_softc *);
    243 static void	bridge_rtage_work(struct work *, void *);
    244 static void	bridge_rtflush(struct bridge_softc *, int);
    245 static int	bridge_rtdaddr(struct bridge_softc *, const uint8_t *);
    246 static void	bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp);
    247 
    248 static void	bridge_rtable_init(struct bridge_softc *);
    249 static void	bridge_rtable_fini(struct bridge_softc *);
    250 
    251 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
    252 						  const uint8_t *);
    253 static int	bridge_rtnode_insert(struct bridge_softc *,
    254 				     struct bridge_rtnode *);
    255 static void	bridge_rtnode_remove(struct bridge_softc *,
    256 				     struct bridge_rtnode *);
    257 static void	bridge_rtnode_destroy(struct bridge_rtnode *);
    258 
    259 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
    260 						  const char *name);
    261 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
    262 						     struct ifnet *ifp);
    263 static void	bridge_release_member(struct bridge_softc *, struct bridge_iflist *);
    264 static void	bridge_delete_member(struct bridge_softc *,
    265 				     struct bridge_iflist *);
    266 static struct bridge_iflist *bridge_try_hold_bif(struct bridge_iflist *);
    267 
    268 static int	bridge_ioctl_add(struct bridge_softc *, void *);
    269 static int	bridge_ioctl_del(struct bridge_softc *, void *);
    270 static int	bridge_ioctl_gifflags(struct bridge_softc *, void *);
    271 static int	bridge_ioctl_sifflags(struct bridge_softc *, void *);
    272 static int	bridge_ioctl_scache(struct bridge_softc *, void *);
    273 static int	bridge_ioctl_gcache(struct bridge_softc *, void *);
    274 static int	bridge_ioctl_gifs(struct bridge_softc *, void *);
    275 static int	bridge_ioctl_rts(struct bridge_softc *, void *);
    276 static int	bridge_ioctl_saddr(struct bridge_softc *, void *);
    277 static int	bridge_ioctl_sto(struct bridge_softc *, void *);
    278 static int	bridge_ioctl_gto(struct bridge_softc *, void *);
    279 static int	bridge_ioctl_daddr(struct bridge_softc *, void *);
    280 static int	bridge_ioctl_flush(struct bridge_softc *, void *);
    281 static int	bridge_ioctl_gpri(struct bridge_softc *, void *);
    282 static int	bridge_ioctl_spri(struct bridge_softc *, void *);
    283 static int	bridge_ioctl_ght(struct bridge_softc *, void *);
    284 static int	bridge_ioctl_sht(struct bridge_softc *, void *);
    285 static int	bridge_ioctl_gfd(struct bridge_softc *, void *);
    286 static int	bridge_ioctl_sfd(struct bridge_softc *, void *);
    287 static int	bridge_ioctl_gma(struct bridge_softc *, void *);
    288 static int	bridge_ioctl_sma(struct bridge_softc *, void *);
    289 static int	bridge_ioctl_sifprio(struct bridge_softc *, void *);
    290 static int	bridge_ioctl_sifcost(struct bridge_softc *, void *);
    291 #if defined(BRIDGE_IPF)
    292 static int	bridge_ioctl_gfilt(struct bridge_softc *, void *);
    293 static int	bridge_ioctl_sfilt(struct bridge_softc *, void *);
    294 static int	bridge_ipf(void *, struct mbuf **, struct ifnet *, int);
    295 static int	bridge_ip_checkbasic(struct mbuf **mp);
    296 # ifdef INET6
    297 static int	bridge_ip6_checkbasic(struct mbuf **mp);
    298 # endif /* INET6 */
    299 #endif /* BRIDGE_IPF */
    300 
    301 static void bridge_sysctl_fwdq_setup(struct sysctllog **clog,
    302     struct bridge_softc *sc);
    303 
    304 struct bridge_control {
    305 	int	(*bc_func)(struct bridge_softc *, void *);
    306 	int	bc_argsize;
    307 	int	bc_flags;
    308 };
    309 
    310 #define	BC_F_COPYIN		0x01	/* copy arguments in */
    311 #define	BC_F_COPYOUT		0x02	/* copy arguments out */
    312 #define	BC_F_SUSER		0x04	/* do super-user check */
    313 #define BC_F_XLATEIN		0x08	/* xlate arguments in */
    314 #define BC_F_XLATEOUT		0x10	/* xlate arguments out */
    315 
    316 static const struct bridge_control bridge_control_table[] = {
    317 [BRDGADD] = {bridge_ioctl_add, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    318 [BRDGDEL] = {bridge_ioctl_del, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    319 
    320 [BRDGGIFFLGS] = {bridge_ioctl_gifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_COPYOUT},
    321 [BRDGSIFFLGS] = {bridge_ioctl_sifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    322 
    323 [BRDGSCACHE] = {bridge_ioctl_scache, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    324 [BRDGGCACHE] = {bridge_ioctl_gcache, sizeof(struct ifbrparam), BC_F_COPYOUT},
    325 
    326 [OBRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_COPYIN|BC_F_COPYOUT},
    327 [OBRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_COPYIN|BC_F_COPYOUT},
    328 
    329 [BRDGSADDR] = {bridge_ioctl_saddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
    330 
    331 [BRDGSTO] = {bridge_ioctl_sto, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    332 [BRDGGTO] = {bridge_ioctl_gto, sizeof(struct ifbrparam), BC_F_COPYOUT},
    333 
    334 [BRDGDADDR] = {bridge_ioctl_daddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
    335 
    336 [BRDGFLUSH] = {bridge_ioctl_flush, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    337 
    338 [BRDGGPRI] = {bridge_ioctl_gpri, sizeof(struct ifbrparam), BC_F_COPYOUT},
    339 [BRDGSPRI] = {bridge_ioctl_spri, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    340 
    341 [BRDGGHT] = {bridge_ioctl_ght, sizeof(struct ifbrparam), BC_F_COPYOUT},
    342 [BRDGSHT] = {bridge_ioctl_sht, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    343 
    344 [BRDGGFD] = {bridge_ioctl_gfd, sizeof(struct ifbrparam), BC_F_COPYOUT},
    345 [BRDGSFD] = {bridge_ioctl_sfd, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    346 
    347 [BRDGGMA] = {bridge_ioctl_gma, sizeof(struct ifbrparam), BC_F_COPYOUT},
    348 [BRDGSMA] = {bridge_ioctl_sma, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    349 
    350 [BRDGSIFPRIO] = {bridge_ioctl_sifprio, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    351 
    352 [BRDGSIFCOST] = {bridge_ioctl_sifcost, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    353 #if defined(BRIDGE_IPF)
    354 [BRDGGFILT] = {bridge_ioctl_gfilt, sizeof(struct ifbrparam), BC_F_COPYOUT},
    355 [BRDGSFILT] = {bridge_ioctl_sfilt, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    356 #endif /* BRIDGE_IPF */
    357 [BRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_XLATEIN|BC_F_XLATEOUT},
    358 [BRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_XLATEIN|BC_F_XLATEOUT},
    359 };
    360 
    361 static const int bridge_control_table_size = __arraycount(bridge_control_table);
    362 
    363 static LIST_HEAD(, bridge_softc) bridge_list;
    364 static kmutex_t bridge_list_lock;
    365 
    366 static struct if_clone bridge_cloner =
    367     IF_CLONE_INITIALIZER("bridge", bridge_clone_create, bridge_clone_destroy);
    368 
    369 /*
    370  * bridgeattach:
    371  *
    372  *	Pseudo-device attach routine.
    373  */
    374 void
    375 bridgeattach(int n)
    376 {
    377 
    378 	pool_init(&bridge_rtnode_pool, sizeof(struct bridge_rtnode),
    379 	    0, 0, 0, "brtpl", NULL, IPL_NET);
    380 
    381 	LIST_INIT(&bridge_list);
    382 	mutex_init(&bridge_list_lock, MUTEX_DEFAULT, IPL_NET);
    383 	if_clone_attach(&bridge_cloner);
    384 }
    385 
    386 /*
    387  * bridge_clone_create:
    388  *
    389  *	Create a new bridge instance.
    390  */
    391 static int
    392 bridge_clone_create(struct if_clone *ifc, int unit)
    393 {
    394 	struct bridge_softc *sc;
    395 	struct ifnet *ifp;
    396 	int error, flags;
    397 
    398 	sc = kmem_zalloc(sizeof(*sc),  KM_SLEEP);
    399 	ifp = &sc->sc_if;
    400 
    401 	sc->sc_brtmax = BRIDGE_RTABLE_MAX;
    402 	sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
    403 	sc->sc_bridge_max_age = BSTP_DEFAULT_MAX_AGE;
    404 	sc->sc_bridge_hello_time = BSTP_DEFAULT_HELLO_TIME;
    405 	sc->sc_bridge_forward_delay = BSTP_DEFAULT_FORWARD_DELAY;
    406 	sc->sc_bridge_priority = BSTP_DEFAULT_BRIDGE_PRIORITY;
    407 	sc->sc_hold_time = BSTP_DEFAULT_HOLD_TIME;
    408 	sc->sc_filter_flags = 0;
    409 
    410 	/* Initialize our routing table. */
    411 	bridge_rtable_init(sc);
    412 
    413 #ifdef BRIDGE_MPSAFE
    414 	flags = WQ_MPSAFE;
    415 #else
    416 	flags = 0;
    417 #endif
    418 	error = workqueue_create(&sc->sc_rtage_wq, "bridge_rtage",
    419 	    bridge_rtage_work, sc, PRI_SOFTNET, IPL_SOFTNET, flags);
    420 	if (error)
    421 		panic("%s: workqueue_create %d\n", __func__, error);
    422 
    423 	callout_init(&sc->sc_brcallout, 0);
    424 	callout_init(&sc->sc_bstpcallout, 0);
    425 
    426 	LIST_INIT(&sc->sc_iflist);
    427 #ifdef BRIDGE_MPSAFE
    428 	sc->sc_iflist_intr_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
    429 	sc->sc_iflist_psz = pserialize_create();
    430 	sc->sc_iflist_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_SOFTNET);
    431 #else
    432 	sc->sc_iflist_intr_lock = NULL;
    433 	sc->sc_iflist_psz = NULL;
    434 	sc->sc_iflist_lock = NULL;
    435 #endif
    436 	cv_init(&sc->sc_iflist_cv, "if_bridge_cv");
    437 
    438 	if_initname(ifp, ifc->ifc_name, unit);
    439 	ifp->if_softc = sc;
    440 	ifp->if_mtu = ETHERMTU;
    441 	ifp->if_ioctl = bridge_ioctl;
    442 	ifp->if_output = bridge_output;
    443 	ifp->if_start = bridge_start;
    444 	ifp->if_stop = bridge_stop;
    445 	ifp->if_init = bridge_init;
    446 	ifp->if_type = IFT_BRIDGE;
    447 	ifp->if_addrlen = 0;
    448 	ifp->if_dlt = DLT_EN10MB;
    449 	ifp->if_hdrlen = ETHER_HDR_LEN;
    450 
    451 	sc->sc_fwd_pktq = pktq_create(IFQ_MAXLEN, bridge_forward, sc);
    452 	KASSERT(sc->sc_fwd_pktq != NULL);
    453 
    454 	bridge_sysctl_fwdq_setup(&ifp->if_sysctl_log, sc);
    455 
    456 	if_initialize(ifp);
    457 	if_register(ifp);
    458 
    459 	if_alloc_sadl(ifp);
    460 
    461 	mutex_enter(&bridge_list_lock);
    462 	LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
    463 	mutex_exit(&bridge_list_lock);
    464 
    465 	return (0);
    466 }
    467 
    468 /*
    469  * bridge_clone_destroy:
    470  *
    471  *	Destroy a bridge instance.
    472  */
    473 static int
    474 bridge_clone_destroy(struct ifnet *ifp)
    475 {
    476 	struct bridge_softc *sc = ifp->if_softc;
    477 	struct bridge_iflist *bif;
    478 	int s;
    479 
    480 	/* Must be called during IFF_RUNNING, i.e., before bridge_stop */
    481 	pktq_barrier(sc->sc_fwd_pktq);
    482 
    483 	s = splnet();
    484 
    485 	bridge_stop(ifp, 1);
    486 
    487 	BRIDGE_LOCK(sc);
    488 	while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
    489 		bridge_delete_member(sc, bif);
    490 	BRIDGE_UNLOCK(sc);
    491 
    492 	mutex_enter(&bridge_list_lock);
    493 	LIST_REMOVE(sc, sc_list);
    494 	mutex_exit(&bridge_list_lock);
    495 
    496 	splx(s);
    497 
    498 	if_detach(ifp);
    499 
    500 	/* Should be called after if_detach for safe */
    501 	pktq_flush(sc->sc_fwd_pktq);
    502 	pktq_destroy(sc->sc_fwd_pktq);
    503 
    504 	/* Tear down the routing table. */
    505 	bridge_rtable_fini(sc);
    506 
    507 	cv_destroy(&sc->sc_iflist_cv);
    508 	if (sc->sc_iflist_intr_lock)
    509 		mutex_obj_free(sc->sc_iflist_intr_lock);
    510 
    511 	if (sc->sc_iflist_psz)
    512 		pserialize_destroy(sc->sc_iflist_psz);
    513 	if (sc->sc_iflist_lock)
    514 		mutex_obj_free(sc->sc_iflist_lock);
    515 
    516 	workqueue_destroy(sc->sc_rtage_wq);
    517 
    518 	kmem_free(sc, sizeof(*sc));
    519 
    520 	return (0);
    521 }
    522 
    523 static int
    524 bridge_sysctl_fwdq_maxlen(SYSCTLFN_ARGS)
    525 {
    526 	struct sysctlnode node = *rnode;
    527 	const struct bridge_softc *sc =	node.sysctl_data;
    528 	return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), sc->sc_fwd_pktq);
    529 }
    530 
    531 #define	SYSCTL_BRIDGE_PKTQ(cn, c)					\
    532 	static int							\
    533 	bridge_sysctl_fwdq_##cn(SYSCTLFN_ARGS)				\
    534 	{								\
    535 		struct sysctlnode node = *rnode;			\
    536 		const struct bridge_softc *sc =	node.sysctl_data;	\
    537 		return sysctl_pktq_count(SYSCTLFN_CALL(rnode),		\
    538 					 sc->sc_fwd_pktq, c);		\
    539 	}
    540 
    541 SYSCTL_BRIDGE_PKTQ(items, PKTQ_NITEMS)
    542 SYSCTL_BRIDGE_PKTQ(drops, PKTQ_DROPS)
    543 
    544 static void
    545 bridge_sysctl_fwdq_setup(struct sysctllog **clog, struct bridge_softc *sc)
    546 {
    547 	const struct sysctlnode *cnode, *rnode;
    548 	sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL;
    549 	const char *ifname = sc->sc_if.if_xname;
    550 
    551 	len_func = bridge_sysctl_fwdq_items;
    552 	maxlen_func = bridge_sysctl_fwdq_maxlen;
    553 	drops_func = bridge_sysctl_fwdq_drops;
    554 
    555 	if (sysctl_createv(clog, 0, NULL, &rnode,
    556 			   CTLFLAG_PERMANENT,
    557 			   CTLTYPE_NODE, "interfaces",
    558 			   SYSCTL_DESCR("Per-interface controls"),
    559 			   NULL, 0, NULL, 0,
    560 			   CTL_NET, CTL_CREATE, CTL_EOL) != 0)
    561 		goto bad;
    562 
    563 	if (sysctl_createv(clog, 0, &rnode, &rnode,
    564 			   CTLFLAG_PERMANENT,
    565 			   CTLTYPE_NODE, ifname,
    566 			   SYSCTL_DESCR("Interface controls"),
    567 			   NULL, 0, NULL, 0,
    568 			   CTL_CREATE, CTL_EOL) != 0)
    569 		goto bad;
    570 
    571 	if (sysctl_createv(clog, 0, &rnode, &rnode,
    572 			   CTLFLAG_PERMANENT,
    573 			   CTLTYPE_NODE, "fwdq",
    574 			   SYSCTL_DESCR("Protocol input queue controls"),
    575 			   NULL, 0, NULL, 0,
    576 			   CTL_CREATE, CTL_EOL) != 0)
    577 		goto bad;
    578 
    579 	if (sysctl_createv(clog, 0, &rnode, &cnode,
    580 			   CTLFLAG_PERMANENT,
    581 			   CTLTYPE_INT, "len",
    582 			   SYSCTL_DESCR("Current forwarding queue length"),
    583 			   len_func, 0, (void *)sc, 0,
    584 			   CTL_CREATE, IFQCTL_LEN, CTL_EOL) != 0)
    585 		goto bad;
    586 
    587 	if (sysctl_createv(clog, 0, &rnode, &cnode,
    588 			   CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
    589 			   CTLTYPE_INT, "maxlen",
    590 			   SYSCTL_DESCR("Maximum allowed forwarding queue length"),
    591 			   maxlen_func, 0, (void *)sc, 0,
    592 			   CTL_CREATE, IFQCTL_MAXLEN, CTL_EOL) != 0)
    593 		goto bad;
    594 
    595 	if (sysctl_createv(clog, 0, &rnode, &cnode,
    596 			   CTLFLAG_PERMANENT,
    597 			   CTLTYPE_INT, "drops",
    598 			   SYSCTL_DESCR("Packets dropped due to full forwarding queue"),
    599 			   drops_func, 0, (void *)sc, 0,
    600 			   CTL_CREATE, IFQCTL_DROPS, CTL_EOL) != 0)
    601 		goto bad;
    602 
    603 	return;
    604 bad:
    605 	aprint_error("%s: could not attach sysctl nodes\n", ifname);
    606 	return;
    607 }
    608 
    609 /*
    610  * bridge_ioctl:
    611  *
    612  *	Handle a control request from the operator.
    613  */
    614 static int
    615 bridge_ioctl(struct ifnet *ifp, u_long cmd, void *data)
    616 {
    617 	struct bridge_softc *sc = ifp->if_softc;
    618 	struct lwp *l = curlwp;	/* XXX */
    619 	union {
    620 		struct ifbreq ifbreq;
    621 		struct ifbifconf ifbifconf;
    622 		struct ifbareq ifbareq;
    623 		struct ifbaconf ifbaconf;
    624 		struct ifbrparam ifbrparam;
    625 	} args;
    626 	struct ifdrv *ifd = (struct ifdrv *) data;
    627 	const struct bridge_control *bc = NULL; /* XXXGCC */
    628 	int s, error = 0;
    629 
    630 	/* Authorize command before calling splnet(). */
    631 	switch (cmd) {
    632 	case SIOCGDRVSPEC:
    633 	case SIOCSDRVSPEC:
    634 		if (ifd->ifd_cmd >= bridge_control_table_size
    635 		    || (bc = &bridge_control_table[ifd->ifd_cmd]) == NULL) {
    636 			error = EINVAL;
    637 			return error;
    638 		}
    639 
    640 		/* We only care about BC_F_SUSER at this point. */
    641 		if ((bc->bc_flags & BC_F_SUSER) == 0)
    642 			break;
    643 
    644 		error = kauth_authorize_network(l->l_cred,
    645 		    KAUTH_NETWORK_INTERFACE_BRIDGE,
    646 		    cmd == SIOCGDRVSPEC ?
    647 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_GETPRIV :
    648 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_SETPRIV,
    649 		     ifd, NULL, NULL);
    650 		if (error)
    651 			return (error);
    652 
    653 		break;
    654 	}
    655 
    656 	s = splnet();
    657 
    658 	switch (cmd) {
    659 	case SIOCGDRVSPEC:
    660 	case SIOCSDRVSPEC:
    661 		KASSERT(bc != NULL);
    662 		if (cmd == SIOCGDRVSPEC &&
    663 		    (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) == 0) {
    664 			error = EINVAL;
    665 			break;
    666 		}
    667 		else if (cmd == SIOCSDRVSPEC &&
    668 		    (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) != 0) {
    669 			error = EINVAL;
    670 			break;
    671 		}
    672 
    673 		/* BC_F_SUSER is checked above, before splnet(). */
    674 
    675 		if ((bc->bc_flags & (BC_F_XLATEIN|BC_F_XLATEOUT)) == 0
    676 		    && (ifd->ifd_len != bc->bc_argsize
    677 			|| ifd->ifd_len > sizeof(args))) {
    678 			error = EINVAL;
    679 			break;
    680 		}
    681 
    682 		memset(&args, 0, sizeof(args));
    683 		if (bc->bc_flags & BC_F_COPYIN) {
    684 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
    685 			if (error)
    686 				break;
    687 		} else if (bc->bc_flags & BC_F_XLATEIN) {
    688 			args.ifbifconf.ifbic_len = ifd->ifd_len;
    689 			args.ifbifconf.ifbic_buf = ifd->ifd_data;
    690 		}
    691 
    692 		error = (*bc->bc_func)(sc, &args);
    693 		if (error)
    694 			break;
    695 
    696 		if (bc->bc_flags & BC_F_COPYOUT) {
    697 			error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
    698 		} else if (bc->bc_flags & BC_F_XLATEOUT) {
    699 			ifd->ifd_len = args.ifbifconf.ifbic_len;
    700 			ifd->ifd_data = args.ifbifconf.ifbic_buf;
    701 		}
    702 		break;
    703 
    704 	case SIOCSIFFLAGS:
    705 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
    706 			break;
    707 		switch (ifp->if_flags & (IFF_UP|IFF_RUNNING)) {
    708 		case IFF_RUNNING:
    709 			/*
    710 			 * If interface is marked down and it is running,
    711 			 * then stop and disable it.
    712 			 */
    713 			(*ifp->if_stop)(ifp, 1);
    714 			break;
    715 		case IFF_UP:
    716 			/*
    717 			 * If interface is marked up and it is stopped, then
    718 			 * start it.
    719 			 */
    720 			error = (*ifp->if_init)(ifp);
    721 			break;
    722 		default:
    723 			break;
    724 		}
    725 		break;
    726 
    727 	case SIOCSIFMTU:
    728 		if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET)
    729 			error = 0;
    730 		break;
    731 
    732 	default:
    733 		error = ifioctl_common(ifp, cmd, data);
    734 		break;
    735 	}
    736 
    737 	splx(s);
    738 
    739 	return (error);
    740 }
    741 
    742 /*
    743  * bridge_lookup_member:
    744  *
    745  *	Lookup a bridge member interface.
    746  */
    747 static struct bridge_iflist *
    748 bridge_lookup_member(struct bridge_softc *sc, const char *name)
    749 {
    750 	struct bridge_iflist *bif;
    751 	struct ifnet *ifp;
    752 	int s;
    753 
    754 	BRIDGE_PSZ_RENTER(s);
    755 
    756 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
    757 		ifp = bif->bif_ifp;
    758 		if (strcmp(ifp->if_xname, name) == 0)
    759 			break;
    760 	}
    761 	bif = bridge_try_hold_bif(bif);
    762 
    763 	BRIDGE_PSZ_REXIT(s);
    764 
    765 	return bif;
    766 }
    767 
    768 /*
    769  * bridge_lookup_member_if:
    770  *
    771  *	Lookup a bridge member interface by ifnet*.
    772  */
    773 static struct bridge_iflist *
    774 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
    775 {
    776 	struct bridge_iflist *bif;
    777 	int s;
    778 
    779 	BRIDGE_PSZ_RENTER(s);
    780 
    781 	bif = member_ifp->if_bridgeif;
    782 	bif = bridge_try_hold_bif(bif);
    783 
    784 	BRIDGE_PSZ_REXIT(s);
    785 
    786 	return bif;
    787 }
    788 
    789 static struct bridge_iflist *
    790 bridge_try_hold_bif(struct bridge_iflist *bif)
    791 {
    792 #ifdef BRIDGE_MPSAFE
    793 	if (bif != NULL) {
    794 		if (bif->bif_waiting)
    795 			bif = NULL;
    796 		else
    797 			atomic_inc_32(&bif->bif_refs);
    798 	}
    799 #endif
    800 	return bif;
    801 }
    802 
    803 /*
    804  * bridge_release_member:
    805  *
    806  *	Release the specified member interface.
    807  */
    808 static void
    809 bridge_release_member(struct bridge_softc *sc, struct bridge_iflist *bif)
    810 {
    811 #ifdef BRIDGE_MPSAFE
    812 	uint32_t refs;
    813 
    814 	refs = atomic_dec_uint_nv(&bif->bif_refs);
    815 	if (__predict_false(refs == 0 && bif->bif_waiting)) {
    816 		BRIDGE_INTR_LOCK(sc);
    817 		cv_broadcast(&sc->sc_iflist_cv);
    818 		BRIDGE_INTR_UNLOCK(sc);
    819 	}
    820 #else
    821 	(void)sc;
    822 	(void)bif;
    823 #endif
    824 }
    825 
    826 /*
    827  * bridge_delete_member:
    828  *
    829  *	Delete the specified member interface.
    830  */
    831 static void
    832 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif)
    833 {
    834 	struct ifnet *ifs = bif->bif_ifp;
    835 
    836 	KASSERT(BRIDGE_LOCKED(sc));
    837 
    838 	ifs->_if_input = ether_input;
    839 	ifs->if_bridge = NULL;
    840 	ifs->if_bridgeif = NULL;
    841 
    842 	LIST_REMOVE(bif, bif_next);
    843 
    844 	BRIDGE_PSZ_PERFORM(sc);
    845 
    846 	BRIDGE_UNLOCK(sc);
    847 
    848 #ifdef BRIDGE_MPSAFE
    849 	BRIDGE_INTR_LOCK(sc);
    850 	bif->bif_waiting = true;
    851 	membar_sync();
    852 	while (bif->bif_refs > 0) {
    853 		aprint_debug("%s: cv_wait on iflist\n", __func__);
    854 		cv_wait(&sc->sc_iflist_cv, sc->sc_iflist_intr_lock);
    855 	}
    856 	bif->bif_waiting = false;
    857 	BRIDGE_INTR_UNLOCK(sc);
    858 #endif
    859 
    860 	kmem_free(bif, sizeof(*bif));
    861 
    862 	BRIDGE_LOCK(sc);
    863 }
    864 
    865 static int
    866 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
    867 {
    868 	struct ifbreq *req = arg;
    869 	struct bridge_iflist *bif = NULL;
    870 	struct ifnet *ifs;
    871 	int error = 0;
    872 
    873 	ifs = ifunit(req->ifbr_ifsname);
    874 	if (ifs == NULL)
    875 		return (ENOENT);
    876 
    877 	if (sc->sc_if.if_mtu != ifs->if_mtu)
    878 		return (EINVAL);
    879 
    880 	if (ifs->if_bridge == sc)
    881 		return (EEXIST);
    882 
    883 	if (ifs->if_bridge != NULL)
    884 		return (EBUSY);
    885 
    886 	if (ifs->_if_input != ether_input)
    887 		return EINVAL;
    888 
    889 	/* FIXME: doesn't work with non-IFF_SIMPLEX interfaces */
    890 	if ((ifs->if_flags & IFF_SIMPLEX) == 0)
    891 		return EINVAL;
    892 
    893 	bif = kmem_alloc(sizeof(*bif), KM_SLEEP);
    894 
    895 	switch (ifs->if_type) {
    896 	case IFT_ETHER:
    897 		if ((error = ether_enable_vlan_mtu(ifs)) > 0)
    898 			goto out;
    899 		/*
    900 		 * Place the interface into promiscuous mode.
    901 		 */
    902 		error = ifpromisc(ifs, 1);
    903 		if (error)
    904 			goto out;
    905 		break;
    906 	default:
    907 		error = EINVAL;
    908 		goto out;
    909 	}
    910 
    911 	bif->bif_ifp = ifs;
    912 	bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
    913 	bif->bif_priority = BSTP_DEFAULT_PORT_PRIORITY;
    914 	bif->bif_path_cost = BSTP_DEFAULT_PATH_COST;
    915 	bif->bif_refs = 0;
    916 	bif->bif_waiting = false;
    917 
    918 	BRIDGE_LOCK(sc);
    919 
    920 	ifs->if_bridge = sc;
    921 	ifs->if_bridgeif = bif;
    922 	LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
    923 	ifs->_if_input = bridge_input;
    924 
    925 	BRIDGE_UNLOCK(sc);
    926 
    927 	if (sc->sc_if.if_flags & IFF_RUNNING)
    928 		bstp_initialization(sc);
    929 	else
    930 		bstp_stop(sc);
    931 
    932  out:
    933 	if (error) {
    934 		if (bif != NULL)
    935 			kmem_free(bif, sizeof(*bif));
    936 	}
    937 	return (error);
    938 }
    939 
    940 static int
    941 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
    942 {
    943 	struct ifbreq *req = arg;
    944 	const char *name = req->ifbr_ifsname;
    945 	struct bridge_iflist *bif;
    946 	struct ifnet *ifs;
    947 
    948 	BRIDGE_LOCK(sc);
    949 
    950 	/*
    951 	 * Don't use bridge_lookup_member. We want to get a member
    952 	 * with bif_refs == 0.
    953 	 */
    954 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
    955 		ifs = bif->bif_ifp;
    956 		if (strcmp(ifs->if_xname, name) == 0)
    957 			break;
    958 	}
    959 
    960 	if (bif == NULL) {
    961 		BRIDGE_UNLOCK(sc);
    962 		return ENOENT;
    963 	}
    964 
    965 	bridge_delete_member(sc, bif);
    966 
    967 	BRIDGE_UNLOCK(sc);
    968 
    969 	switch (ifs->if_type) {
    970 	case IFT_ETHER:
    971 		/*
    972 		 * Take the interface out of promiscuous mode.
    973 		 * Don't call it with holding a spin lock.
    974 		 */
    975 		(void) ifpromisc(ifs, 0);
    976 		(void) ether_disable_vlan_mtu(ifs);
    977 		break;
    978 	default:
    979 #ifdef DIAGNOSTIC
    980 		panic("bridge_delete_member: impossible");
    981 #endif
    982 		break;
    983 	}
    984 
    985 	bridge_rtdelete(sc, ifs);
    986 
    987 	if (sc->sc_if.if_flags & IFF_RUNNING)
    988 		bstp_initialization(sc);
    989 
    990 	return 0;
    991 }
    992 
    993 static int
    994 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
    995 {
    996 	struct ifbreq *req = arg;
    997 	struct bridge_iflist *bif;
    998 
    999 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
   1000 	if (bif == NULL)
   1001 		return (ENOENT);
   1002 
   1003 	req->ifbr_ifsflags = bif->bif_flags;
   1004 	req->ifbr_state = bif->bif_state;
   1005 	req->ifbr_priority = bif->bif_priority;
   1006 	req->ifbr_path_cost = bif->bif_path_cost;
   1007 	req->ifbr_portno = bif->bif_ifp->if_index & 0xff;
   1008 
   1009 	bridge_release_member(sc, bif);
   1010 
   1011 	return (0);
   1012 }
   1013 
   1014 static int
   1015 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
   1016 {
   1017 	struct ifbreq *req = arg;
   1018 	struct bridge_iflist *bif;
   1019 
   1020 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
   1021 	if (bif == NULL)
   1022 		return (ENOENT);
   1023 
   1024 	if (req->ifbr_ifsflags & IFBIF_STP) {
   1025 		switch (bif->bif_ifp->if_type) {
   1026 		case IFT_ETHER:
   1027 			/* These can do spanning tree. */
   1028 			break;
   1029 
   1030 		default:
   1031 			/* Nothing else can. */
   1032 			bridge_release_member(sc, bif);
   1033 			return (EINVAL);
   1034 		}
   1035 	}
   1036 
   1037 	bif->bif_flags = req->ifbr_ifsflags;
   1038 
   1039 	bridge_release_member(sc, bif);
   1040 
   1041 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1042 		bstp_initialization(sc);
   1043 
   1044 	return (0);
   1045 }
   1046 
   1047 static int
   1048 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
   1049 {
   1050 	struct ifbrparam *param = arg;
   1051 
   1052 	sc->sc_brtmax = param->ifbrp_csize;
   1053 	bridge_rttrim(sc);
   1054 
   1055 	return (0);
   1056 }
   1057 
   1058 static int
   1059 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
   1060 {
   1061 	struct ifbrparam *param = arg;
   1062 
   1063 	param->ifbrp_csize = sc->sc_brtmax;
   1064 
   1065 	return (0);
   1066 }
   1067 
   1068 static int
   1069 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
   1070 {
   1071 	struct ifbifconf *bifc = arg;
   1072 	struct bridge_iflist *bif;
   1073 	struct ifbreq *breqs;
   1074 	int i, count, error = 0;
   1075 
   1076 retry:
   1077 	BRIDGE_LOCK(sc);
   1078 	count = 0;
   1079 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
   1080 		count++;
   1081 	BRIDGE_UNLOCK(sc);
   1082 
   1083 	if (count == 0) {
   1084 		bifc->ifbic_len = 0;
   1085 		return 0;
   1086 	}
   1087 
   1088 	if (bifc->ifbic_len == 0 || bifc->ifbic_len < (sizeof(*breqs) * count)) {
   1089 		/* Tell that a larger buffer is needed */
   1090 		bifc->ifbic_len = sizeof(*breqs) * count;
   1091 		return 0;
   1092 	}
   1093 
   1094 	breqs = kmem_alloc(sizeof(*breqs) * count, KM_SLEEP);
   1095 
   1096 	BRIDGE_LOCK(sc);
   1097 
   1098 	i = 0;
   1099 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
   1100 		i++;
   1101 	if (i > count) {
   1102 		/*
   1103 		 * The number of members has been increased.
   1104 		 * We need more memory!
   1105 		 */
   1106 		BRIDGE_UNLOCK(sc);
   1107 		kmem_free(breqs, sizeof(*breqs) * count);
   1108 		goto retry;
   1109 	}
   1110 
   1111 	i = 0;
   1112 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
   1113 		struct ifbreq *breq = &breqs[i++];
   1114 		memset(breq, 0, sizeof(*breq));
   1115 
   1116 		strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname,
   1117 		    sizeof(breq->ifbr_ifsname));
   1118 		breq->ifbr_ifsflags = bif->bif_flags;
   1119 		breq->ifbr_state = bif->bif_state;
   1120 		breq->ifbr_priority = bif->bif_priority;
   1121 		breq->ifbr_path_cost = bif->bif_path_cost;
   1122 		breq->ifbr_portno = bif->bif_ifp->if_index & 0xff;
   1123 	}
   1124 
   1125 	/* Don't call copyout with holding the mutex */
   1126 	BRIDGE_UNLOCK(sc);
   1127 
   1128 	for (i = 0; i < count; i++) {
   1129 		error = copyout(&breqs[i], bifc->ifbic_req + i, sizeof(*breqs));
   1130 		if (error)
   1131 			break;
   1132 	}
   1133 	bifc->ifbic_len = sizeof(*breqs) * i;
   1134 
   1135 	kmem_free(breqs, sizeof(*breqs) * count);
   1136 
   1137 	return error;
   1138 }
   1139 
   1140 static int
   1141 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
   1142 {
   1143 	struct ifbaconf *bac = arg;
   1144 	struct bridge_rtnode *brt;
   1145 	struct ifbareq bareq;
   1146 	int count = 0, error = 0, len;
   1147 
   1148 	if (bac->ifbac_len == 0)
   1149 		return (0);
   1150 
   1151 	BRIDGE_RT_INTR_LOCK(sc);
   1152 
   1153 	len = bac->ifbac_len;
   1154 	LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
   1155 		if (len < sizeof(bareq))
   1156 			goto out;
   1157 		memset(&bareq, 0, sizeof(bareq));
   1158 		strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
   1159 		    sizeof(bareq.ifba_ifsname));
   1160 		memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
   1161 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
   1162 			bareq.ifba_expire = brt->brt_expire - time_uptime;
   1163 		} else
   1164 			bareq.ifba_expire = 0;
   1165 		bareq.ifba_flags = brt->brt_flags;
   1166 
   1167 		error = copyout(&bareq, bac->ifbac_req + count, sizeof(bareq));
   1168 		if (error)
   1169 			goto out;
   1170 		count++;
   1171 		len -= sizeof(bareq);
   1172 	}
   1173  out:
   1174 	BRIDGE_RT_INTR_UNLOCK(sc);
   1175 
   1176 	bac->ifbac_len = sizeof(bareq) * count;
   1177 	return (error);
   1178 }
   1179 
   1180 static int
   1181 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
   1182 {
   1183 	struct ifbareq *req = arg;
   1184 	struct bridge_iflist *bif;
   1185 	int error;
   1186 
   1187 	bif = bridge_lookup_member(sc, req->ifba_ifsname);
   1188 	if (bif == NULL)
   1189 		return (ENOENT);
   1190 
   1191 	error = bridge_rtupdate(sc, req->ifba_dst, bif->bif_ifp, 1,
   1192 	    req->ifba_flags);
   1193 
   1194 	bridge_release_member(sc, bif);
   1195 
   1196 	return (error);
   1197 }
   1198 
   1199 static int
   1200 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
   1201 {
   1202 	struct ifbrparam *param = arg;
   1203 
   1204 	sc->sc_brttimeout = param->ifbrp_ctime;
   1205 
   1206 	return (0);
   1207 }
   1208 
   1209 static int
   1210 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
   1211 {
   1212 	struct ifbrparam *param = arg;
   1213 
   1214 	param->ifbrp_ctime = sc->sc_brttimeout;
   1215 
   1216 	return (0);
   1217 }
   1218 
   1219 static int
   1220 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
   1221 {
   1222 	struct ifbareq *req = arg;
   1223 
   1224 	return (bridge_rtdaddr(sc, req->ifba_dst));
   1225 }
   1226 
   1227 static int
   1228 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
   1229 {
   1230 	struct ifbreq *req = arg;
   1231 
   1232 	bridge_rtflush(sc, req->ifbr_ifsflags);
   1233 
   1234 	return (0);
   1235 }
   1236 
   1237 static int
   1238 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
   1239 {
   1240 	struct ifbrparam *param = arg;
   1241 
   1242 	param->ifbrp_prio = sc->sc_bridge_priority;
   1243 
   1244 	return (0);
   1245 }
   1246 
   1247 static int
   1248 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
   1249 {
   1250 	struct ifbrparam *param = arg;
   1251 
   1252 	sc->sc_bridge_priority = param->ifbrp_prio;
   1253 
   1254 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1255 		bstp_initialization(sc);
   1256 
   1257 	return (0);
   1258 }
   1259 
   1260 static int
   1261 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
   1262 {
   1263 	struct ifbrparam *param = arg;
   1264 
   1265 	param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8;
   1266 
   1267 	return (0);
   1268 }
   1269 
   1270 static int
   1271 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
   1272 {
   1273 	struct ifbrparam *param = arg;
   1274 
   1275 	if (param->ifbrp_hellotime == 0)
   1276 		return (EINVAL);
   1277 	sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8;
   1278 
   1279 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1280 		bstp_initialization(sc);
   1281 
   1282 	return (0);
   1283 }
   1284 
   1285 static int
   1286 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
   1287 {
   1288 	struct ifbrparam *param = arg;
   1289 
   1290 	param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8;
   1291 
   1292 	return (0);
   1293 }
   1294 
   1295 static int
   1296 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
   1297 {
   1298 	struct ifbrparam *param = arg;
   1299 
   1300 	if (param->ifbrp_fwddelay == 0)
   1301 		return (EINVAL);
   1302 	sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8;
   1303 
   1304 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1305 		bstp_initialization(sc);
   1306 
   1307 	return (0);
   1308 }
   1309 
   1310 static int
   1311 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
   1312 {
   1313 	struct ifbrparam *param = arg;
   1314 
   1315 	param->ifbrp_maxage = sc->sc_bridge_max_age >> 8;
   1316 
   1317 	return (0);
   1318 }
   1319 
   1320 static int
   1321 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
   1322 {
   1323 	struct ifbrparam *param = arg;
   1324 
   1325 	if (param->ifbrp_maxage == 0)
   1326 		return (EINVAL);
   1327 	sc->sc_bridge_max_age = param->ifbrp_maxage << 8;
   1328 
   1329 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1330 		bstp_initialization(sc);
   1331 
   1332 	return (0);
   1333 }
   1334 
   1335 static int
   1336 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
   1337 {
   1338 	struct ifbreq *req = arg;
   1339 	struct bridge_iflist *bif;
   1340 
   1341 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
   1342 	if (bif == NULL)
   1343 		return (ENOENT);
   1344 
   1345 	bif->bif_priority = req->ifbr_priority;
   1346 
   1347 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1348 		bstp_initialization(sc);
   1349 
   1350 	bridge_release_member(sc, bif);
   1351 
   1352 	return (0);
   1353 }
   1354 
   1355 #if defined(BRIDGE_IPF)
   1356 static int
   1357 bridge_ioctl_gfilt(struct bridge_softc *sc, void *arg)
   1358 {
   1359 	struct ifbrparam *param = arg;
   1360 
   1361 	param->ifbrp_filter = sc->sc_filter_flags;
   1362 
   1363 	return (0);
   1364 }
   1365 
   1366 static int
   1367 bridge_ioctl_sfilt(struct bridge_softc *sc, void *arg)
   1368 {
   1369 	struct ifbrparam *param = arg;
   1370 	uint32_t nflags, oflags;
   1371 
   1372 	if (param->ifbrp_filter & ~IFBF_FILT_MASK)
   1373 		return (EINVAL);
   1374 
   1375 	nflags = param->ifbrp_filter;
   1376 	oflags = sc->sc_filter_flags;
   1377 
   1378 	if ((nflags & IFBF_FILT_USEIPF) && !(oflags & IFBF_FILT_USEIPF)) {
   1379 		pfil_add_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
   1380 			sc->sc_if.if_pfil);
   1381 	}
   1382 	if (!(nflags & IFBF_FILT_USEIPF) && (oflags & IFBF_FILT_USEIPF)) {
   1383 		pfil_remove_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
   1384 			sc->sc_if.if_pfil);
   1385 	}
   1386 
   1387 	sc->sc_filter_flags = nflags;
   1388 
   1389 	return (0);
   1390 }
   1391 #endif /* BRIDGE_IPF */
   1392 
   1393 static int
   1394 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
   1395 {
   1396 	struct ifbreq *req = arg;
   1397 	struct bridge_iflist *bif;
   1398 
   1399 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
   1400 	if (bif == NULL)
   1401 		return (ENOENT);
   1402 
   1403 	bif->bif_path_cost = req->ifbr_path_cost;
   1404 
   1405 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1406 		bstp_initialization(sc);
   1407 
   1408 	bridge_release_member(sc, bif);
   1409 
   1410 	return (0);
   1411 }
   1412 
   1413 /*
   1414  * bridge_ifdetach:
   1415  *
   1416  *	Detach an interface from a bridge.  Called when a member
   1417  *	interface is detaching.
   1418  */
   1419 void
   1420 bridge_ifdetach(struct ifnet *ifp)
   1421 {
   1422 	struct bridge_softc *sc = ifp->if_bridge;
   1423 	struct ifbreq breq;
   1424 
   1425 	/* ioctl_lock should prevent this from happening */
   1426 	KASSERT(sc != NULL);
   1427 
   1428 	memset(&breq, 0, sizeof(breq));
   1429 	strlcpy(breq.ifbr_ifsname, ifp->if_xname, sizeof(breq.ifbr_ifsname));
   1430 
   1431 	(void) bridge_ioctl_del(sc, &breq);
   1432 }
   1433 
   1434 /*
   1435  * bridge_init:
   1436  *
   1437  *	Initialize a bridge interface.
   1438  */
   1439 static int
   1440 bridge_init(struct ifnet *ifp)
   1441 {
   1442 	struct bridge_softc *sc = ifp->if_softc;
   1443 
   1444 	if (ifp->if_flags & IFF_RUNNING)
   1445 		return (0);
   1446 
   1447 	callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
   1448 	    bridge_timer, sc);
   1449 
   1450 	ifp->if_flags |= IFF_RUNNING;
   1451 	bstp_initialization(sc);
   1452 	return (0);
   1453 }
   1454 
   1455 /*
   1456  * bridge_stop:
   1457  *
   1458  *	Stop the bridge interface.
   1459  */
   1460 static void
   1461 bridge_stop(struct ifnet *ifp, int disable)
   1462 {
   1463 	struct bridge_softc *sc = ifp->if_softc;
   1464 
   1465 	if ((ifp->if_flags & IFF_RUNNING) == 0)
   1466 		return;
   1467 
   1468 	callout_stop(&sc->sc_brcallout);
   1469 	bstp_stop(sc);
   1470 
   1471 	bridge_rtflush(sc, IFBF_FLUSHDYN);
   1472 
   1473 	ifp->if_flags &= ~IFF_RUNNING;
   1474 }
   1475 
   1476 /*
   1477  * bridge_enqueue:
   1478  *
   1479  *	Enqueue a packet on a bridge member interface.
   1480  */
   1481 void
   1482 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m,
   1483     int runfilt)
   1484 {
   1485 	ALTQ_DECL(struct altq_pktattr pktattr;)
   1486 	int len, error;
   1487 	short mflags;
   1488 
   1489 	/*
   1490 	 * Clear any in-bound checksum flags for this packet.
   1491 	 */
   1492 	m->m_pkthdr.csum_flags = 0;
   1493 
   1494 	if (runfilt) {
   1495 		if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
   1496 		    dst_ifp, PFIL_OUT) != 0) {
   1497 			if (m != NULL)
   1498 				m_freem(m);
   1499 			return;
   1500 		}
   1501 		if (m == NULL)
   1502 			return;
   1503 	}
   1504 
   1505 #ifdef ALTQ
   1506 	/*
   1507 	 * If ALTQ is enabled on the member interface, do
   1508 	 * classification; the queueing discipline might
   1509 	 * not require classification, but might require
   1510 	 * the address family/header pointer in the pktattr.
   1511 	 */
   1512 	if (ALTQ_IS_ENABLED(&dst_ifp->if_snd)) {
   1513 		/* XXX IFT_ETHER */
   1514 		altq_etherclassify(&dst_ifp->if_snd, m, &pktattr);
   1515 	}
   1516 #endif /* ALTQ */
   1517 
   1518 	len = m->m_pkthdr.len;
   1519 	mflags = m->m_flags;
   1520 
   1521 	IFQ_ENQUEUE(&dst_ifp->if_snd, m, &pktattr, error);
   1522 
   1523 	if (error) {
   1524 		/* mbuf is already freed */
   1525 		sc->sc_if.if_oerrors++;
   1526 		return;
   1527 	}
   1528 
   1529 	sc->sc_if.if_opackets++;
   1530 	sc->sc_if.if_obytes += len;
   1531 
   1532 	dst_ifp->if_obytes += len;
   1533 
   1534 	if (mflags & M_MCAST) {
   1535 		sc->sc_if.if_omcasts++;
   1536 		dst_ifp->if_omcasts++;
   1537 	}
   1538 
   1539 	if ((dst_ifp->if_flags & IFF_OACTIVE) == 0)
   1540 		(*dst_ifp->if_start)(dst_ifp);
   1541 }
   1542 
   1543 /*
   1544  * bridge_output:
   1545  *
   1546  *	Send output from a bridge member interface.  This
   1547  *	performs the bridging function for locally originated
   1548  *	packets.
   1549  *
   1550  *	The mbuf has the Ethernet header already attached.  We must
   1551  *	enqueue or free the mbuf before returning.
   1552  */
   1553 int
   1554 bridge_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa,
   1555     struct rtentry *rt)
   1556 {
   1557 	struct ether_header *eh;
   1558 	struct ifnet *dst_if;
   1559 	struct bridge_softc *sc;
   1560 #ifndef BRIDGE_MPSAFE
   1561 	int s;
   1562 #endif
   1563 
   1564 	if (m->m_len < ETHER_HDR_LEN) {
   1565 		m = m_pullup(m, ETHER_HDR_LEN);
   1566 		if (m == NULL)
   1567 			return (0);
   1568 	}
   1569 
   1570 	eh = mtod(m, struct ether_header *);
   1571 	sc = ifp->if_bridge;
   1572 
   1573 #ifndef BRIDGE_MPSAFE
   1574 	s = splnet();
   1575 #endif
   1576 
   1577 	/*
   1578 	 * If bridge is down, but the original output interface is up,
   1579 	 * go ahead and send out that interface.  Otherwise, the packet
   1580 	 * is dropped below.
   1581 	 */
   1582 	if (__predict_false(sc == NULL) ||
   1583 	    (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
   1584 		dst_if = ifp;
   1585 		goto sendunicast;
   1586 	}
   1587 
   1588 	/*
   1589 	 * If the packet is a multicast, or we don't know a better way to
   1590 	 * get there, send to all interfaces.
   1591 	 */
   1592 	if (ETHER_IS_MULTICAST(eh->ether_dhost))
   1593 		dst_if = NULL;
   1594 	else
   1595 		dst_if = bridge_rtlookup(sc, eh->ether_dhost);
   1596 	if (dst_if == NULL) {
   1597 		struct bridge_iflist *bif;
   1598 		struct mbuf *mc;
   1599 		int used = 0;
   1600 		int ss;
   1601 
   1602 		BRIDGE_PSZ_RENTER(ss);
   1603 		LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
   1604 			bif = bridge_try_hold_bif(bif);
   1605 			if (bif == NULL)
   1606 				continue;
   1607 			BRIDGE_PSZ_REXIT(ss);
   1608 
   1609 			dst_if = bif->bif_ifp;
   1610 			if ((dst_if->if_flags & IFF_RUNNING) == 0)
   1611 				goto next;
   1612 
   1613 			/*
   1614 			 * If this is not the original output interface,
   1615 			 * and the interface is participating in spanning
   1616 			 * tree, make sure the port is in a state that
   1617 			 * allows forwarding.
   1618 			 */
   1619 			if (dst_if != ifp &&
   1620 			    (bif->bif_flags & IFBIF_STP) != 0) {
   1621 				switch (bif->bif_state) {
   1622 				case BSTP_IFSTATE_BLOCKING:
   1623 				case BSTP_IFSTATE_LISTENING:
   1624 				case BSTP_IFSTATE_DISABLED:
   1625 					goto next;
   1626 				}
   1627 			}
   1628 
   1629 			if (LIST_NEXT(bif, bif_next) == NULL) {
   1630 				used = 1;
   1631 				mc = m;
   1632 			} else {
   1633 				mc = m_copym(m, 0, M_COPYALL, M_NOWAIT);
   1634 				if (mc == NULL) {
   1635 					sc->sc_if.if_oerrors++;
   1636 					goto next;
   1637 				}
   1638 			}
   1639 
   1640 			bridge_enqueue(sc, dst_if, mc, 0);
   1641 next:
   1642 			bridge_release_member(sc, bif);
   1643 			BRIDGE_PSZ_RENTER(ss);
   1644 		}
   1645 		BRIDGE_PSZ_REXIT(ss);
   1646 
   1647 		if (used == 0)
   1648 			m_freem(m);
   1649 #ifndef BRIDGE_MPSAFE
   1650 		splx(s);
   1651 #endif
   1652 		return (0);
   1653 	}
   1654 
   1655  sendunicast:
   1656 	/*
   1657 	 * XXX Spanning tree consideration here?
   1658 	 */
   1659 
   1660 	if ((dst_if->if_flags & IFF_RUNNING) == 0) {
   1661 		m_freem(m);
   1662 #ifndef BRIDGE_MPSAFE
   1663 		splx(s);
   1664 #endif
   1665 		return (0);
   1666 	}
   1667 
   1668 	bridge_enqueue(sc, dst_if, m, 0);
   1669 
   1670 #ifndef BRIDGE_MPSAFE
   1671 	splx(s);
   1672 #endif
   1673 	return (0);
   1674 }
   1675 
   1676 /*
   1677  * bridge_start:
   1678  *
   1679  *	Start output on a bridge.
   1680  *
   1681  *	NOTE: This routine should never be called in this implementation.
   1682  */
   1683 static void
   1684 bridge_start(struct ifnet *ifp)
   1685 {
   1686 
   1687 	printf("%s: bridge_start() called\n", ifp->if_xname);
   1688 }
   1689 
   1690 /*
   1691  * bridge_forward:
   1692  *
   1693  *	The forwarding function of the bridge.
   1694  */
   1695 static void
   1696 bridge_forward(void *v)
   1697 {
   1698 	struct bridge_softc *sc = v;
   1699 	struct mbuf *m;
   1700 	struct bridge_iflist *bif;
   1701 	struct ifnet *src_if, *dst_if;
   1702 	struct ether_header *eh;
   1703 #ifndef BRIDGE_MPSAFE
   1704 	int s;
   1705 
   1706 	KERNEL_LOCK(1, NULL);
   1707 	mutex_enter(softnet_lock);
   1708 #endif
   1709 
   1710 	if ((sc->sc_if.if_flags & IFF_RUNNING) == 0) {
   1711 #ifndef BRIDGE_MPSAFE
   1712 		mutex_exit(softnet_lock);
   1713 		KERNEL_UNLOCK_ONE(NULL);
   1714 #endif
   1715 		return;
   1716 	}
   1717 
   1718 #ifndef BRIDGE_MPSAFE
   1719 	s = splnet();
   1720 #endif
   1721 	while ((m = pktq_dequeue(sc->sc_fwd_pktq)) != NULL) {
   1722 		src_if = m->m_pkthdr.rcvif;
   1723 
   1724 		sc->sc_if.if_ipackets++;
   1725 		sc->sc_if.if_ibytes += m->m_pkthdr.len;
   1726 
   1727 		/*
   1728 		 * Look up the bridge_iflist.
   1729 		 */
   1730 		bif = bridge_lookup_member_if(sc, src_if);
   1731 		if (bif == NULL) {
   1732 			/* Interface is not a bridge member (anymore?) */
   1733 			m_freem(m);
   1734 			continue;
   1735 		}
   1736 
   1737 		if (bif->bif_flags & IFBIF_STP) {
   1738 			switch (bif->bif_state) {
   1739 			case BSTP_IFSTATE_BLOCKING:
   1740 			case BSTP_IFSTATE_LISTENING:
   1741 			case BSTP_IFSTATE_DISABLED:
   1742 				m_freem(m);
   1743 				bridge_release_member(sc, bif);
   1744 				continue;
   1745 			}
   1746 		}
   1747 
   1748 		eh = mtod(m, struct ether_header *);
   1749 
   1750 		/*
   1751 		 * If the interface is learning, and the source
   1752 		 * address is valid and not multicast, record
   1753 		 * the address.
   1754 		 */
   1755 		if ((bif->bif_flags & IFBIF_LEARNING) != 0 &&
   1756 		    ETHER_IS_MULTICAST(eh->ether_shost) == 0 &&
   1757 		    (eh->ether_shost[0] == 0 &&
   1758 		     eh->ether_shost[1] == 0 &&
   1759 		     eh->ether_shost[2] == 0 &&
   1760 		     eh->ether_shost[3] == 0 &&
   1761 		     eh->ether_shost[4] == 0 &&
   1762 		     eh->ether_shost[5] == 0) == 0) {
   1763 			(void) bridge_rtupdate(sc, eh->ether_shost,
   1764 			    src_if, 0, IFBAF_DYNAMIC);
   1765 		}
   1766 
   1767 		if ((bif->bif_flags & IFBIF_STP) != 0 &&
   1768 		    bif->bif_state == BSTP_IFSTATE_LEARNING) {
   1769 			m_freem(m);
   1770 			bridge_release_member(sc, bif);
   1771 			continue;
   1772 		}
   1773 
   1774 		bridge_release_member(sc, bif);
   1775 
   1776 		/*
   1777 		 * At this point, the port either doesn't participate
   1778 		 * in spanning tree or it is in the forwarding state.
   1779 		 */
   1780 
   1781 		/*
   1782 		 * If the packet is unicast, destined for someone on
   1783 		 * "this" side of the bridge, drop it.
   1784 		 */
   1785 		if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
   1786 			dst_if = bridge_rtlookup(sc, eh->ether_dhost);
   1787 			if (src_if == dst_if) {
   1788 				m_freem(m);
   1789 				continue;
   1790 			}
   1791 		} else {
   1792 			/* ...forward it to all interfaces. */
   1793 			sc->sc_if.if_imcasts++;
   1794 			dst_if = NULL;
   1795 		}
   1796 
   1797 		if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
   1798 		    m->m_pkthdr.rcvif, PFIL_IN) != 0) {
   1799 			if (m != NULL)
   1800 				m_freem(m);
   1801 			continue;
   1802 		}
   1803 		if (m == NULL)
   1804 			continue;
   1805 
   1806 		if (dst_if == NULL) {
   1807 			bridge_broadcast(sc, src_if, m);
   1808 			continue;
   1809 		}
   1810 
   1811 		/*
   1812 		 * At this point, we're dealing with a unicast frame
   1813 		 * going to a different interface.
   1814 		 */
   1815 		if ((dst_if->if_flags & IFF_RUNNING) == 0) {
   1816 			m_freem(m);
   1817 			continue;
   1818 		}
   1819 
   1820 		bif = bridge_lookup_member_if(sc, dst_if);
   1821 		if (bif == NULL) {
   1822 			/* Not a member of the bridge (anymore?) */
   1823 			m_freem(m);
   1824 			continue;
   1825 		}
   1826 
   1827 		if (bif->bif_flags & IFBIF_STP) {
   1828 			switch (bif->bif_state) {
   1829 			case BSTP_IFSTATE_DISABLED:
   1830 			case BSTP_IFSTATE_BLOCKING:
   1831 				m_freem(m);
   1832 				bridge_release_member(sc, bif);
   1833 				continue;
   1834 			}
   1835 		}
   1836 
   1837 		bridge_release_member(sc, bif);
   1838 
   1839 		bridge_enqueue(sc, dst_if, m, 1);
   1840 	}
   1841 #ifndef BRIDGE_MPSAFE
   1842 	splx(s);
   1843 	mutex_exit(softnet_lock);
   1844 	KERNEL_UNLOCK_ONE(NULL);
   1845 #endif
   1846 }
   1847 
   1848 static bool
   1849 bstp_state_before_learning(struct bridge_iflist *bif)
   1850 {
   1851 	if (bif->bif_flags & IFBIF_STP) {
   1852 		switch (bif->bif_state) {
   1853 		case BSTP_IFSTATE_BLOCKING:
   1854 		case BSTP_IFSTATE_LISTENING:
   1855 		case BSTP_IFSTATE_DISABLED:
   1856 			return true;
   1857 		}
   1858 	}
   1859 	return false;
   1860 }
   1861 
   1862 static bool
   1863 bridge_ourether(struct bridge_iflist *bif, struct ether_header *eh, int src)
   1864 {
   1865 	uint8_t *ether = src ? eh->ether_shost : eh->ether_dhost;
   1866 
   1867 	if (memcmp(CLLADDR(bif->bif_ifp->if_sadl), ether, ETHER_ADDR_LEN) == 0
   1868 #if NCARP > 0
   1869 	    || (bif->bif_ifp->if_carp &&
   1870 	        carp_ourether(bif->bif_ifp->if_carp, eh, IFT_ETHER, src) != NULL)
   1871 #endif /* NCARP > 0 */
   1872 	    )
   1873 		return true;
   1874 
   1875 	return false;
   1876 }
   1877 
   1878 /*
   1879  * bridge_input:
   1880  *
   1881  *	Receive input from a member interface.  Queue the packet for
   1882  *	bridging if it is not for us.
   1883  */
   1884 static void
   1885 bridge_input(struct ifnet *ifp, struct mbuf *m)
   1886 {
   1887 	struct bridge_softc *sc = ifp->if_bridge;
   1888 	struct bridge_iflist *bif;
   1889 	struct ether_header *eh;
   1890 
   1891 	KASSERT(!cpu_intr_p());
   1892 
   1893 	if (__predict_false(sc == NULL) ||
   1894 	    (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
   1895 		ether_input(ifp, m);
   1896 		return;
   1897 	}
   1898 
   1899 	bif = bridge_lookup_member_if(sc, ifp);
   1900 	if (bif == NULL) {
   1901 		ether_input(ifp, m);
   1902 		return;
   1903 	}
   1904 
   1905 	eh = mtod(m, struct ether_header *);
   1906 
   1907 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
   1908 		if (memcmp(etherbroadcastaddr,
   1909 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
   1910 			m->m_flags |= M_BCAST;
   1911 		else
   1912 			m->m_flags |= M_MCAST;
   1913 	}
   1914 
   1915 	/*
   1916 	 * A 'fast' path for packets addressed to interfaces that are
   1917 	 * part of this bridge.
   1918 	 */
   1919 	if (!(m->m_flags & (M_BCAST|M_MCAST)) &&
   1920 	    !bstp_state_before_learning(bif)) {
   1921 		struct bridge_iflist *_bif;
   1922 		struct ifnet *_ifp = NULL;
   1923 		int s;
   1924 
   1925 		BRIDGE_PSZ_RENTER(s);
   1926 		LIST_FOREACH(_bif, &sc->sc_iflist, bif_next) {
   1927 			/* It is destined for us. */
   1928 			if (bridge_ourether(_bif, eh, 0)) {
   1929 				_bif = bridge_try_hold_bif(_bif);
   1930 				BRIDGE_PSZ_REXIT(s);
   1931 				if (_bif == NULL)
   1932 					goto out;
   1933 				if (_bif->bif_flags & IFBIF_LEARNING)
   1934 					(void) bridge_rtupdate(sc,
   1935 					    eh->ether_shost, ifp, 0, IFBAF_DYNAMIC);
   1936 				_ifp = m->m_pkthdr.rcvif = _bif->bif_ifp;
   1937 				bridge_release_member(sc, _bif);
   1938 				goto out;
   1939 			}
   1940 
   1941 			/* We just received a packet that we sent out. */
   1942 			if (bridge_ourether(_bif, eh, 1))
   1943 				break;
   1944 		}
   1945 		BRIDGE_PSZ_REXIT(s);
   1946 out:
   1947 
   1948 		if (_bif != NULL) {
   1949 			bridge_release_member(sc, bif);
   1950 			if (_ifp != NULL) {
   1951 				m->m_flags &= ~M_PROMISC;
   1952 				ether_input(_ifp, m);
   1953 			} else
   1954 				m_freem(m);
   1955 			return;
   1956 		}
   1957 	}
   1958 
   1959 	/* Tap off 802.1D packets; they do not get forwarded. */
   1960 	if (bif->bif_flags & IFBIF_STP &&
   1961 	    memcmp(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN) == 0) {
   1962 		bstp_input(sc, bif, m);
   1963 		bridge_release_member(sc, bif);
   1964 		return;
   1965 	}
   1966 
   1967 	/*
   1968 	 * A normal switch would discard the packet here, but that's not what
   1969 	 * we've done historically. This also prevents some obnoxious behaviour.
   1970 	 */
   1971 	if (bstp_state_before_learning(bif)) {
   1972 		bridge_release_member(sc, bif);
   1973 		ether_input(ifp, m);
   1974 		return;
   1975 	}
   1976 
   1977 	bridge_release_member(sc, bif);
   1978 
   1979 	/* Queue the packet for bridge forwarding. */
   1980 	{
   1981 		/*
   1982 		 * Force to enqueue to curcpu's pktq (RX can run on a CPU
   1983 		 * other than CPU#0). XXX need fundamental solution.
   1984 		 */
   1985 		const unsigned hash = curcpu()->ci_index;
   1986 
   1987 		if (__predict_false(!pktq_enqueue(sc->sc_fwd_pktq, m, hash)))
   1988 			m_freem(m);
   1989 	}
   1990 }
   1991 
   1992 /*
   1993  * bridge_broadcast:
   1994  *
   1995  *	Send a frame to all interfaces that are members of
   1996  *	the bridge, except for the one on which the packet
   1997  *	arrived.
   1998  */
   1999 static void
   2000 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
   2001     struct mbuf *m)
   2002 {
   2003 	struct bridge_iflist *bif;
   2004 	struct mbuf *mc;
   2005 	struct ifnet *dst_if;
   2006 	bool bmcast;
   2007 	int s;
   2008 
   2009 	bmcast = m->m_flags & (M_BCAST|M_MCAST);
   2010 
   2011 	BRIDGE_PSZ_RENTER(s);
   2012 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
   2013 		bif = bridge_try_hold_bif(bif);
   2014 		if (bif == NULL)
   2015 			continue;
   2016 		BRIDGE_PSZ_REXIT(s);
   2017 
   2018 		dst_if = bif->bif_ifp;
   2019 
   2020 		if (bif->bif_flags & IFBIF_STP) {
   2021 			switch (bif->bif_state) {
   2022 			case BSTP_IFSTATE_BLOCKING:
   2023 			case BSTP_IFSTATE_DISABLED:
   2024 				goto next;
   2025 			}
   2026 		}
   2027 
   2028 		if ((bif->bif_flags & IFBIF_DISCOVER) == 0 && !bmcast)
   2029 			goto next;
   2030 
   2031 		if ((dst_if->if_flags & IFF_RUNNING) == 0)
   2032 			goto next;
   2033 
   2034 		if (dst_if != src_if) {
   2035 			mc = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
   2036 			if (mc == NULL) {
   2037 				sc->sc_if.if_oerrors++;
   2038 				goto next;
   2039 			}
   2040 			bridge_enqueue(sc, dst_if, mc, 1);
   2041 		}
   2042 
   2043 		if (bmcast) {
   2044 			mc = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
   2045 			if (mc == NULL) {
   2046 				sc->sc_if.if_oerrors++;
   2047 				goto next;
   2048 			}
   2049 
   2050 			mc->m_pkthdr.rcvif = dst_if;
   2051 			mc->m_flags &= ~M_PROMISC;
   2052 			ether_input(dst_if, mc);
   2053 		}
   2054 next:
   2055 		bridge_release_member(sc, bif);
   2056 		BRIDGE_PSZ_RENTER(s);
   2057 	}
   2058 	BRIDGE_PSZ_REXIT(s);
   2059 
   2060 	m_freem(m);
   2061 }
   2062 
   2063 static int
   2064 bridge_rtalloc(struct bridge_softc *sc, const uint8_t *dst,
   2065     struct bridge_rtnode **brtp)
   2066 {
   2067 	struct bridge_rtnode *brt;
   2068 	int error;
   2069 
   2070 	if (sc->sc_brtcnt >= sc->sc_brtmax)
   2071 		return ENOSPC;
   2072 
   2073 	/*
   2074 	 * Allocate a new bridge forwarding node, and
   2075 	 * initialize the expiration time and Ethernet
   2076 	 * address.
   2077 	 */
   2078 	brt = pool_get(&bridge_rtnode_pool, PR_NOWAIT);
   2079 	if (brt == NULL)
   2080 		return ENOMEM;
   2081 
   2082 	memset(brt, 0, sizeof(*brt));
   2083 	brt->brt_expire = time_uptime + sc->sc_brttimeout;
   2084 	brt->brt_flags = IFBAF_DYNAMIC;
   2085 	memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
   2086 
   2087 	BRIDGE_RT_INTR_LOCK(sc);
   2088 	error = bridge_rtnode_insert(sc, brt);
   2089 	BRIDGE_RT_INTR_UNLOCK(sc);
   2090 
   2091 	if (error != 0) {
   2092 		pool_put(&bridge_rtnode_pool, brt);
   2093 		return error;
   2094 	}
   2095 
   2096 	*brtp = brt;
   2097 	return 0;
   2098 }
   2099 
   2100 /*
   2101  * bridge_rtupdate:
   2102  *
   2103  *	Add a bridge routing entry.
   2104  */
   2105 static int
   2106 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
   2107     struct ifnet *dst_if, int setflags, uint8_t flags)
   2108 {
   2109 	struct bridge_rtnode *brt;
   2110 	int s;
   2111 
   2112 again:
   2113 	/*
   2114 	 * A route for this destination might already exist.  If so,
   2115 	 * update it, otherwise create a new one.
   2116 	 */
   2117 	BRIDGE_RT_RENTER(s);
   2118 	brt = bridge_rtnode_lookup(sc, dst);
   2119 
   2120 	if (brt != NULL) {
   2121 		brt->brt_ifp = dst_if;
   2122 		if (setflags) {
   2123 			brt->brt_flags = flags;
   2124 			if (flags & IFBAF_STATIC)
   2125 				brt->brt_expire = 0;
   2126 			else
   2127 				brt->brt_expire = time_uptime + sc->sc_brttimeout;
   2128 		} else {
   2129 			if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
   2130 				brt->brt_expire = time_uptime + sc->sc_brttimeout;
   2131 		}
   2132 	}
   2133 	BRIDGE_RT_REXIT(s);
   2134 
   2135 	if (brt == NULL) {
   2136 		int r;
   2137 
   2138 		r = bridge_rtalloc(sc, dst, &brt);
   2139 		if (r != 0)
   2140 			return r;
   2141 		goto again;
   2142 	}
   2143 
   2144 	return 0;
   2145 }
   2146 
   2147 /*
   2148  * bridge_rtlookup:
   2149  *
   2150  *	Lookup the destination interface for an address.
   2151  */
   2152 static struct ifnet *
   2153 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr)
   2154 {
   2155 	struct bridge_rtnode *brt;
   2156 	struct ifnet *ifs = NULL;
   2157 	int s;
   2158 
   2159 	BRIDGE_RT_RENTER(s);
   2160 	brt = bridge_rtnode_lookup(sc, addr);
   2161 	if (brt != NULL)
   2162 		ifs = brt->brt_ifp;
   2163 	BRIDGE_RT_REXIT(s);
   2164 
   2165 	return ifs;
   2166 }
   2167 
   2168 typedef bool (*bridge_iterate_cb_t)
   2169     (struct bridge_softc *, struct bridge_rtnode *, bool *, void *);
   2170 
   2171 /*
   2172  * bridge_rtlist_iterate_remove:
   2173  *
   2174  *	It iterates on sc->sc_rtlist and removes rtnodes of it which func
   2175  *	callback judges to remove. Removals of rtnodes are done in a manner
   2176  *	of pserialize. To this end, all kmem_* operations are placed out of
   2177  *	mutexes.
   2178  */
   2179 static void
   2180 bridge_rtlist_iterate_remove(struct bridge_softc *sc, bridge_iterate_cb_t func, void *arg)
   2181 {
   2182 	struct bridge_rtnode *brt, *nbrt;
   2183 	struct bridge_rtnode **brt_list;
   2184 	int i, count;
   2185 
   2186 retry:
   2187 	count = sc->sc_brtcnt;
   2188 	if (count == 0)
   2189 		return;
   2190 	brt_list = kmem_alloc(sizeof(*brt_list) * count, KM_SLEEP);
   2191 
   2192 	BRIDGE_RT_LOCK(sc);
   2193 	BRIDGE_RT_INTR_LOCK(sc);
   2194 	if (__predict_false(sc->sc_brtcnt > count)) {
   2195 		/* The rtnodes increased, we need more memory */
   2196 		BRIDGE_RT_INTR_UNLOCK(sc);
   2197 		BRIDGE_RT_UNLOCK(sc);
   2198 		kmem_free(brt_list, sizeof(*brt_list) * count);
   2199 		goto retry;
   2200 	}
   2201 
   2202 	i = 0;
   2203 	LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
   2204 		bool need_break = false;
   2205 		if (func(sc, brt, &need_break, arg)) {
   2206 			bridge_rtnode_remove(sc, brt);
   2207 			brt_list[i++] = brt;
   2208 		}
   2209 		if (need_break)
   2210 			break;
   2211 	}
   2212 	BRIDGE_RT_INTR_UNLOCK(sc);
   2213 
   2214 	if (i > 0)
   2215 		BRIDGE_RT_PSZ_PERFORM(sc);
   2216 	BRIDGE_RT_UNLOCK(sc);
   2217 
   2218 	while (--i >= 0)
   2219 		bridge_rtnode_destroy(brt_list[i]);
   2220 
   2221 	kmem_free(brt_list, sizeof(*brt_list) * count);
   2222 }
   2223 
   2224 static bool
   2225 bridge_rttrim0_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
   2226     bool *need_break, void *arg)
   2227 {
   2228 	if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
   2229 		/* Take into account of the subsequent removal */
   2230 		if ((sc->sc_brtcnt - 1) <= sc->sc_brtmax)
   2231 			*need_break = true;
   2232 		return true;
   2233 	} else
   2234 		return false;
   2235 }
   2236 
   2237 static void
   2238 bridge_rttrim0(struct bridge_softc *sc)
   2239 {
   2240 	bridge_rtlist_iterate_remove(sc, bridge_rttrim0_cb, NULL);
   2241 }
   2242 
   2243 /*
   2244  * bridge_rttrim:
   2245  *
   2246  *	Trim the routine table so that we have a number
   2247  *	of routing entries less than or equal to the
   2248  *	maximum number.
   2249  */
   2250 static void
   2251 bridge_rttrim(struct bridge_softc *sc)
   2252 {
   2253 
   2254 	/* Make sure we actually need to do this. */
   2255 	if (sc->sc_brtcnt <= sc->sc_brtmax)
   2256 		return;
   2257 
   2258 	/* Force an aging cycle; this might trim enough addresses. */
   2259 	bridge_rtage(sc);
   2260 	if (sc->sc_brtcnt <= sc->sc_brtmax)
   2261 		return;
   2262 
   2263 	bridge_rttrim0(sc);
   2264 
   2265 	return;
   2266 }
   2267 
   2268 /*
   2269  * bridge_timer:
   2270  *
   2271  *	Aging timer for the bridge.
   2272  */
   2273 static void
   2274 bridge_timer(void *arg)
   2275 {
   2276 	struct bridge_softc *sc = arg;
   2277 
   2278 	workqueue_enqueue(sc->sc_rtage_wq, &bridge_rtage_wk, NULL);
   2279 }
   2280 
   2281 static void
   2282 bridge_rtage_work(struct work *wk, void *arg)
   2283 {
   2284 	struct bridge_softc *sc = arg;
   2285 
   2286 	KASSERT(wk == &bridge_rtage_wk);
   2287 
   2288 	bridge_rtage(sc);
   2289 
   2290 	if (sc->sc_if.if_flags & IFF_RUNNING)
   2291 		callout_reset(&sc->sc_brcallout,
   2292 		    bridge_rtable_prune_period * hz, bridge_timer, sc);
   2293 }
   2294 
   2295 static bool
   2296 bridge_rtage_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
   2297     bool *need_break, void *arg)
   2298 {
   2299 	if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
   2300 	    time_uptime >= brt->brt_expire)
   2301 		return true;
   2302 	else
   2303 		return false;
   2304 }
   2305 
   2306 /*
   2307  * bridge_rtage:
   2308  *
   2309  *	Perform an aging cycle.
   2310  */
   2311 static void
   2312 bridge_rtage(struct bridge_softc *sc)
   2313 {
   2314 	bridge_rtlist_iterate_remove(sc, bridge_rtage_cb, NULL);
   2315 }
   2316 
   2317 
   2318 static bool
   2319 bridge_rtflush_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
   2320     bool *need_break, void *arg)
   2321 {
   2322 	int full = *(int*)arg;
   2323 
   2324 	if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
   2325 		return true;
   2326 	else
   2327 		return false;
   2328 }
   2329 
   2330 /*
   2331  * bridge_rtflush:
   2332  *
   2333  *	Remove all dynamic addresses from the bridge.
   2334  */
   2335 static void
   2336 bridge_rtflush(struct bridge_softc *sc, int full)
   2337 {
   2338 	bridge_rtlist_iterate_remove(sc, bridge_rtflush_cb, &full);
   2339 }
   2340 
   2341 /*
   2342  * bridge_rtdaddr:
   2343  *
   2344  *	Remove an address from the table.
   2345  */
   2346 static int
   2347 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr)
   2348 {
   2349 	struct bridge_rtnode *brt;
   2350 
   2351 	BRIDGE_RT_LOCK(sc);
   2352 	BRIDGE_RT_INTR_LOCK(sc);
   2353 	if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL) {
   2354 		BRIDGE_RT_INTR_UNLOCK(sc);
   2355 		BRIDGE_RT_UNLOCK(sc);
   2356 		return ENOENT;
   2357 	}
   2358 	bridge_rtnode_remove(sc, brt);
   2359 	BRIDGE_RT_INTR_UNLOCK(sc);
   2360 	BRIDGE_RT_PSZ_PERFORM(sc);
   2361 	BRIDGE_RT_UNLOCK(sc);
   2362 
   2363 	bridge_rtnode_destroy(brt);
   2364 
   2365 	return 0;
   2366 }
   2367 
   2368 /*
   2369  * bridge_rtdelete:
   2370  *
   2371  *	Delete routes to a speicifc member interface.
   2372  */
   2373 static void
   2374 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp)
   2375 {
   2376 	struct bridge_rtnode *brt, *nbrt;
   2377 
   2378 	BRIDGE_RT_LOCK(sc);
   2379 	BRIDGE_RT_INTR_LOCK(sc);
   2380 	LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
   2381 		if (brt->brt_ifp == ifp)
   2382 			break;
   2383 	}
   2384 	if (brt == NULL) {
   2385 		BRIDGE_RT_INTR_UNLOCK(sc);
   2386 		BRIDGE_RT_UNLOCK(sc);
   2387 		return;
   2388 	}
   2389 	bridge_rtnode_remove(sc, brt);
   2390 	BRIDGE_RT_INTR_UNLOCK(sc);
   2391 	BRIDGE_RT_PSZ_PERFORM(sc);
   2392 	BRIDGE_RT_UNLOCK(sc);
   2393 
   2394 	bridge_rtnode_destroy(brt);
   2395 }
   2396 
   2397 /*
   2398  * bridge_rtable_init:
   2399  *
   2400  *	Initialize the route table for this bridge.
   2401  */
   2402 static void
   2403 bridge_rtable_init(struct bridge_softc *sc)
   2404 {
   2405 	int i;
   2406 
   2407 	sc->sc_rthash = kmem_alloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
   2408 	    KM_SLEEP);
   2409 
   2410 	for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
   2411 		LIST_INIT(&sc->sc_rthash[i]);
   2412 
   2413 	sc->sc_rthash_key = cprng_fast32();
   2414 
   2415 	LIST_INIT(&sc->sc_rtlist);
   2416 
   2417 	sc->sc_rtlist_intr_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
   2418 #ifdef BRIDGE_MPSAFE
   2419 	sc->sc_rtlist_psz = pserialize_create();
   2420 	sc->sc_rtlist_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_SOFTNET);
   2421 #else
   2422 	sc->sc_rtlist_psz = NULL;
   2423 	sc->sc_rtlist_lock = NULL;
   2424 #endif
   2425 }
   2426 
   2427 /*
   2428  * bridge_rtable_fini:
   2429  *
   2430  *	Deconstruct the route table for this bridge.
   2431  */
   2432 static void
   2433 bridge_rtable_fini(struct bridge_softc *sc)
   2434 {
   2435 
   2436 	kmem_free(sc->sc_rthash, sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE);
   2437 	if (sc->sc_rtlist_intr_lock)
   2438 		mutex_obj_free(sc->sc_rtlist_intr_lock);
   2439 	if (sc->sc_rtlist_lock)
   2440 		mutex_obj_free(sc->sc_rtlist_lock);
   2441 	if (sc->sc_rtlist_psz)
   2442 		pserialize_destroy(sc->sc_rtlist_psz);
   2443 }
   2444 
   2445 /*
   2446  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
   2447  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
   2448  */
   2449 #define	mix(a, b, c)							\
   2450 do {									\
   2451 	a -= b; a -= c; a ^= (c >> 13);					\
   2452 	b -= c; b -= a; b ^= (a << 8);					\
   2453 	c -= a; c -= b; c ^= (b >> 13);					\
   2454 	a -= b; a -= c; a ^= (c >> 12);					\
   2455 	b -= c; b -= a; b ^= (a << 16);					\
   2456 	c -= a; c -= b; c ^= (b >> 5);					\
   2457 	a -= b; a -= c; a ^= (c >> 3);					\
   2458 	b -= c; b -= a; b ^= (a << 10);					\
   2459 	c -= a; c -= b; c ^= (b >> 15);					\
   2460 } while (/*CONSTCOND*/0)
   2461 
   2462 static inline uint32_t
   2463 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
   2464 {
   2465 	uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
   2466 
   2467 	b += addr[5] << 8;
   2468 	b += addr[4];
   2469 	a += addr[3] << 24;
   2470 	a += addr[2] << 16;
   2471 	a += addr[1] << 8;
   2472 	a += addr[0];
   2473 
   2474 	mix(a, b, c);
   2475 
   2476 	return (c & BRIDGE_RTHASH_MASK);
   2477 }
   2478 
   2479 #undef mix
   2480 
   2481 /*
   2482  * bridge_rtnode_lookup:
   2483  *
   2484  *	Look up a bridge route node for the specified destination.
   2485  */
   2486 static struct bridge_rtnode *
   2487 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr)
   2488 {
   2489 	struct bridge_rtnode *brt;
   2490 	uint32_t hash;
   2491 	int dir;
   2492 
   2493 	hash = bridge_rthash(sc, addr);
   2494 	LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
   2495 		dir = memcmp(addr, brt->brt_addr, ETHER_ADDR_LEN);
   2496 		if (dir == 0)
   2497 			return (brt);
   2498 		if (dir > 0)
   2499 			return (NULL);
   2500 	}
   2501 
   2502 	return (NULL);
   2503 }
   2504 
   2505 /*
   2506  * bridge_rtnode_insert:
   2507  *
   2508  *	Insert the specified bridge node into the route table.  We
   2509  *	assume the entry is not already in the table.
   2510  */
   2511 static int
   2512 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
   2513 {
   2514 	struct bridge_rtnode *lbrt;
   2515 	uint32_t hash;
   2516 	int dir;
   2517 
   2518 	KASSERT(BRIDGE_RT_INTR_LOCKED(sc));
   2519 
   2520 	hash = bridge_rthash(sc, brt->brt_addr);
   2521 
   2522 	lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
   2523 	if (lbrt == NULL) {
   2524 		LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
   2525 		goto out;
   2526 	}
   2527 
   2528 	do {
   2529 		dir = memcmp(brt->brt_addr, lbrt->brt_addr, ETHER_ADDR_LEN);
   2530 		if (dir == 0)
   2531 			return (EEXIST);
   2532 		if (dir > 0) {
   2533 			LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
   2534 			goto out;
   2535 		}
   2536 		if (LIST_NEXT(lbrt, brt_hash) == NULL) {
   2537 			LIST_INSERT_AFTER(lbrt, brt, brt_hash);
   2538 			goto out;
   2539 		}
   2540 		lbrt = LIST_NEXT(lbrt, brt_hash);
   2541 	} while (lbrt != NULL);
   2542 
   2543 #ifdef DIAGNOSTIC
   2544 	panic("bridge_rtnode_insert: impossible");
   2545 #endif
   2546 
   2547  out:
   2548 	LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
   2549 	sc->sc_brtcnt++;
   2550 
   2551 	return (0);
   2552 }
   2553 
   2554 /*
   2555  * bridge_rtnode_remove:
   2556  *
   2557  *	Remove a bridge rtnode from the rthash and the rtlist of a bridge.
   2558  */
   2559 static void
   2560 bridge_rtnode_remove(struct bridge_softc *sc, struct bridge_rtnode *brt)
   2561 {
   2562 
   2563 	KASSERT(BRIDGE_RT_INTR_LOCKED(sc));
   2564 
   2565 	LIST_REMOVE(brt, brt_hash);
   2566 	LIST_REMOVE(brt, brt_list);
   2567 	sc->sc_brtcnt--;
   2568 }
   2569 
   2570 /*
   2571  * bridge_rtnode_destroy:
   2572  *
   2573  *	Destroy a bridge rtnode.
   2574  */
   2575 static void
   2576 bridge_rtnode_destroy(struct bridge_rtnode *brt)
   2577 {
   2578 
   2579 	pool_put(&bridge_rtnode_pool, brt);
   2580 }
   2581 
   2582 #if defined(BRIDGE_IPF)
   2583 extern pfil_head_t *inet_pfil_hook;                 /* XXX */
   2584 extern pfil_head_t *inet6_pfil_hook;                /* XXX */
   2585 
   2586 /*
   2587  * Send bridge packets through IPF if they are one of the types IPF can deal
   2588  * with, or if they are ARP or REVARP.  (IPF will pass ARP and REVARP without
   2589  * question.)
   2590  */
   2591 static int
   2592 bridge_ipf(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
   2593 {
   2594 	int snap, error;
   2595 	struct ether_header *eh1, eh2;
   2596 	struct llc llc1;
   2597 	uint16_t ether_type;
   2598 
   2599 	snap = 0;
   2600 	error = -1;	/* Default error if not error == 0 */
   2601 	eh1 = mtod(*mp, struct ether_header *);
   2602 	ether_type = ntohs(eh1->ether_type);
   2603 
   2604 	/*
   2605 	 * Check for SNAP/LLC.
   2606 	 */
   2607         if (ether_type < ETHERMTU) {
   2608                 struct llc *llc2 = (struct llc *)(eh1 + 1);
   2609 
   2610                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
   2611                     llc2->llc_dsap == LLC_SNAP_LSAP &&
   2612                     llc2->llc_ssap == LLC_SNAP_LSAP &&
   2613                     llc2->llc_control == LLC_UI) {
   2614                 	ether_type = htons(llc2->llc_un.type_snap.ether_type);
   2615 			snap = 1;
   2616                 }
   2617         }
   2618 
   2619 	/*
   2620 	 * If we're trying to filter bridge traffic, don't look at anything
   2621 	 * other than IP and ARP traffic.  If the filter doesn't understand
   2622 	 * IPv6, don't allow IPv6 through the bridge either.  This is lame
   2623 	 * since if we really wanted, say, an AppleTalk filter, we are hosed,
   2624 	 * but of course we don't have an AppleTalk filter to begin with.
   2625 	 * (Note that since IPF doesn't understand ARP it will pass *ALL*
   2626 	 * ARP traffic.)
   2627 	 */
   2628 	switch (ether_type) {
   2629 		case ETHERTYPE_ARP:
   2630 		case ETHERTYPE_REVARP:
   2631 			return 0; /* Automatically pass */
   2632 		case ETHERTYPE_IP:
   2633 # ifdef INET6
   2634 		case ETHERTYPE_IPV6:
   2635 # endif /* INET6 */
   2636 			break;
   2637 		default:
   2638 			goto bad;
   2639 	}
   2640 
   2641 	/* Strip off the Ethernet header and keep a copy. */
   2642 	m_copydata(*mp, 0, ETHER_HDR_LEN, (void *) &eh2);
   2643 	m_adj(*mp, ETHER_HDR_LEN);
   2644 
   2645 	/* Strip off snap header, if present */
   2646 	if (snap) {
   2647 		m_copydata(*mp, 0, sizeof(struct llc), (void *) &llc1);
   2648 		m_adj(*mp, sizeof(struct llc));
   2649 	}
   2650 
   2651 	/*
   2652 	 * Check basic packet sanity and run IPF through pfil.
   2653 	 */
   2654 	KASSERT(!cpu_intr_p());
   2655 	switch (ether_type)
   2656 	{
   2657 	case ETHERTYPE_IP :
   2658 		error = (dir == PFIL_IN) ? bridge_ip_checkbasic(mp) : 0;
   2659 		if (error == 0)
   2660 			error = pfil_run_hooks(inet_pfil_hook, mp, ifp, dir);
   2661 		break;
   2662 # ifdef INET6
   2663 	case ETHERTYPE_IPV6 :
   2664 		error = (dir == PFIL_IN) ? bridge_ip6_checkbasic(mp) : 0;
   2665 		if (error == 0)
   2666 			error = pfil_run_hooks(inet6_pfil_hook, mp, ifp, dir);
   2667 		break;
   2668 # endif
   2669 	default :
   2670 		error = 0;
   2671 		break;
   2672 	}
   2673 
   2674 	if (*mp == NULL)
   2675 		return error;
   2676 	if (error != 0)
   2677 		goto bad;
   2678 
   2679 	error = -1;
   2680 
   2681 	/*
   2682 	 * Finally, put everything back the way it was and return
   2683 	 */
   2684 	if (snap) {
   2685 		M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
   2686 		if (*mp == NULL)
   2687 			return error;
   2688 		bcopy(&llc1, mtod(*mp, void *), sizeof(struct llc));
   2689 	}
   2690 
   2691 	M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
   2692 	if (*mp == NULL)
   2693 		return error;
   2694 	bcopy(&eh2, mtod(*mp, void *), ETHER_HDR_LEN);
   2695 
   2696 	return 0;
   2697 
   2698     bad:
   2699 	m_freem(*mp);
   2700 	*mp = NULL;
   2701 	return error;
   2702 }
   2703 
   2704 /*
   2705  * Perform basic checks on header size since
   2706  * IPF assumes ip_input has already processed
   2707  * it for it.  Cut-and-pasted from ip_input.c.
   2708  * Given how simple the IPv6 version is,
   2709  * does the IPv4 version really need to be
   2710  * this complicated?
   2711  *
   2712  * XXX Should we update ipstat here, or not?
   2713  * XXX Right now we update ipstat but not
   2714  * XXX csum_counter.
   2715  */
   2716 static int
   2717 bridge_ip_checkbasic(struct mbuf **mp)
   2718 {
   2719 	struct mbuf *m = *mp;
   2720 	struct ip *ip;
   2721 	int len, hlen;
   2722 
   2723 	if (*mp == NULL)
   2724 		return -1;
   2725 
   2726 	if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
   2727 		if ((m = m_copyup(m, sizeof(struct ip),
   2728 			(max_linkhdr + 3) & ~3)) == NULL) {
   2729 			/* XXXJRT new stat, please */
   2730 			ip_statinc(IP_STAT_TOOSMALL);
   2731 			goto bad;
   2732 		}
   2733 	} else if (__predict_false(m->m_len < sizeof (struct ip))) {
   2734 		if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
   2735 			ip_statinc(IP_STAT_TOOSMALL);
   2736 			goto bad;
   2737 		}
   2738 	}
   2739 	ip = mtod(m, struct ip *);
   2740 	if (ip == NULL) goto bad;
   2741 
   2742 	if (ip->ip_v != IPVERSION) {
   2743 		ip_statinc(IP_STAT_BADVERS);
   2744 		goto bad;
   2745 	}
   2746 	hlen = ip->ip_hl << 2;
   2747 	if (hlen < sizeof(struct ip)) { /* minimum header length */
   2748 		ip_statinc(IP_STAT_BADHLEN);
   2749 		goto bad;
   2750 	}
   2751 	if (hlen > m->m_len) {
   2752 		if ((m = m_pullup(m, hlen)) == 0) {
   2753 			ip_statinc(IP_STAT_BADHLEN);
   2754 			goto bad;
   2755 		}
   2756 		ip = mtod(m, struct ip *);
   2757 		if (ip == NULL) goto bad;
   2758 	}
   2759 
   2760         switch (m->m_pkthdr.csum_flags &
   2761                 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) |
   2762                  M_CSUM_IPv4_BAD)) {
   2763         case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
   2764                 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); */
   2765                 goto bad;
   2766 
   2767         case M_CSUM_IPv4:
   2768                 /* Checksum was okay. */
   2769                 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); */
   2770                 break;
   2771 
   2772         default:
   2773                 /* Must compute it ourselves. */
   2774                 /* INET_CSUM_COUNTER_INCR(&ip_swcsum); */
   2775                 if (in_cksum(m, hlen) != 0)
   2776                         goto bad;
   2777                 break;
   2778         }
   2779 
   2780         /* Retrieve the packet length. */
   2781         len = ntohs(ip->ip_len);
   2782 
   2783         /*
   2784          * Check for additional length bogosity
   2785          */
   2786         if (len < hlen) {
   2787 		ip_statinc(IP_STAT_BADLEN);
   2788                 goto bad;
   2789         }
   2790 
   2791         /*
   2792          * Check that the amount of data in the buffers
   2793          * is as at least much as the IP header would have us expect.
   2794          * Drop packet if shorter than we expect.
   2795          */
   2796         if (m->m_pkthdr.len < len) {
   2797 		ip_statinc(IP_STAT_TOOSHORT);
   2798                 goto bad;
   2799         }
   2800 
   2801 	/* Checks out, proceed */
   2802 	*mp = m;
   2803 	return 0;
   2804 
   2805     bad:
   2806 	*mp = m;
   2807 	return -1;
   2808 }
   2809 
   2810 # ifdef INET6
   2811 /*
   2812  * Same as above, but for IPv6.
   2813  * Cut-and-pasted from ip6_input.c.
   2814  * XXX Should we update ip6stat, or not?
   2815  */
   2816 static int
   2817 bridge_ip6_checkbasic(struct mbuf **mp)
   2818 {
   2819 	struct mbuf *m = *mp;
   2820 	struct ip6_hdr *ip6;
   2821 
   2822         /*
   2823          * If the IPv6 header is not aligned, slurp it up into a new
   2824          * mbuf with space for link headers, in the event we forward
   2825          * it.  Otherwise, if it is aligned, make sure the entire base
   2826          * IPv6 header is in the first mbuf of the chain.
   2827          */
   2828         if (IP6_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
   2829                 struct ifnet *inifp = m->m_pkthdr.rcvif;
   2830                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
   2831                                   (max_linkhdr + 3) & ~3)) == NULL) {
   2832                         /* XXXJRT new stat, please */
   2833 			ip6_statinc(IP6_STAT_TOOSMALL);
   2834                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
   2835                         goto bad;
   2836                 }
   2837         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
   2838                 struct ifnet *inifp = m->m_pkthdr.rcvif;
   2839                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
   2840 			ip6_statinc(IP6_STAT_TOOSMALL);
   2841                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
   2842                         goto bad;
   2843                 }
   2844         }
   2845 
   2846         ip6 = mtod(m, struct ip6_hdr *);
   2847 
   2848         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
   2849 		ip6_statinc(IP6_STAT_BADVERS);
   2850                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
   2851                 goto bad;
   2852         }
   2853 
   2854 	/* Checks out, proceed */
   2855 	*mp = m;
   2856 	return 0;
   2857 
   2858     bad:
   2859 	*mp = m;
   2860 	return -1;
   2861 }
   2862 # endif /* INET6 */
   2863 #endif /* BRIDGE_IPF */
   2864