Home | History | Annotate | Line # | Download | only in net
if_bridge.c revision 1.91.2.5
      1 /*	$NetBSD: if_bridge.c,v 1.91.2.5 2016/03/19 11:30:32 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 1.91.2.5 2016/03/19 11:30:32 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 
    112 #include <net/if_ether.h>
    113 #include <net/if_bridgevar.h>
    114 
    115 #if defined(BRIDGE_IPF)
    116 /* Used for bridge_ip[6]_checkbasic */
    117 #include <netinet/in.h>
    118 #include <netinet/in_systm.h>
    119 #include <netinet/ip.h>
    120 #include <netinet/ip_var.h>
    121 #include <netinet/ip_private.h>		/* XXX */
    122 
    123 #include <netinet/ip6.h>
    124 #include <netinet6/in6_var.h>
    125 #include <netinet6/ip6_var.h>
    126 #include <netinet6/ip6_private.h>	/* XXX */
    127 #endif /* BRIDGE_IPF */
    128 
    129 /*
    130  * Size of the route hash table.  Must be a power of two.
    131  */
    132 #ifndef BRIDGE_RTHASH_SIZE
    133 #define	BRIDGE_RTHASH_SIZE		1024
    134 #endif
    135 
    136 #define	BRIDGE_RTHASH_MASK		(BRIDGE_RTHASH_SIZE - 1)
    137 
    138 #include "carp.h"
    139 #if NCARP > 0
    140 #include <netinet/in.h>
    141 #include <netinet/in_var.h>
    142 #include <netinet/ip_carp.h>
    143 #endif
    144 
    145 #include "ioconf.h"
    146 
    147 __CTASSERT(sizeof(struct ifbifconf) == sizeof(struct ifbaconf));
    148 __CTASSERT(offsetof(struct ifbifconf, ifbic_len) == offsetof(struct ifbaconf, ifbac_len));
    149 __CTASSERT(offsetof(struct ifbifconf, ifbic_buf) == offsetof(struct ifbaconf, ifbac_buf));
    150 
    151 /*
    152  * Maximum number of addresses to cache.
    153  */
    154 #ifndef BRIDGE_RTABLE_MAX
    155 #define	BRIDGE_RTABLE_MAX		100
    156 #endif
    157 
    158 /*
    159  * Spanning tree defaults.
    160  */
    161 #define	BSTP_DEFAULT_MAX_AGE		(20 * 256)
    162 #define	BSTP_DEFAULT_HELLO_TIME		(2 * 256)
    163 #define	BSTP_DEFAULT_FORWARD_DELAY	(15 * 256)
    164 #define	BSTP_DEFAULT_HOLD_TIME		(1 * 256)
    165 #define	BSTP_DEFAULT_BRIDGE_PRIORITY	0x8000
    166 #define	BSTP_DEFAULT_PORT_PRIORITY	0x80
    167 #define	BSTP_DEFAULT_PATH_COST		55
    168 
    169 /*
    170  * Timeout (in seconds) for entries learned dynamically.
    171  */
    172 #ifndef BRIDGE_RTABLE_TIMEOUT
    173 #define	BRIDGE_RTABLE_TIMEOUT		(20 * 60)	/* same as ARP */
    174 #endif
    175 
    176 /*
    177  * Number of seconds between walks of the route list.
    178  */
    179 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
    180 #define	BRIDGE_RTABLE_PRUNE_PERIOD	(5 * 60)
    181 #endif
    182 
    183 #define BRIDGE_RT_LOCK(_sc)	if ((_sc)->sc_rtlist_lock) \
    184 					mutex_enter((_sc)->sc_rtlist_lock)
    185 #define BRIDGE_RT_UNLOCK(_sc)	if ((_sc)->sc_rtlist_lock) \
    186 					mutex_exit((_sc)->sc_rtlist_lock)
    187 #define BRIDGE_RT_LOCKED(_sc)	(!(_sc)->sc_rtlist_lock || \
    188 				 mutex_owned((_sc)->sc_rtlist_lock))
    189 
    190 #define BRIDGE_RT_PSZ_PERFORM(_sc) \
    191 				if ((_sc)->sc_rtlist_psz != NULL) \
    192 					pserialize_perform((_sc)->sc_rtlist_psz);
    193 
    194 #ifdef BRIDGE_MPSAFE
    195 #define BRIDGE_RT_RENTER(__s)	do { __s = pserialize_read_enter(); } while (0)
    196 #define BRIDGE_RT_REXIT(__s)	do { pserialize_read_exit(__s); } while (0)
    197 #else /* BRIDGE_MPSAFE */
    198 #define BRIDGE_RT_RENTER(__s)	do { __s = 0; } while (0)
    199 #define BRIDGE_RT_REXIT(__s)	do { (void)__s; } while (0)
    200 #endif /* BRIDGE_MPSAFE */
    201 
    202 int	bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
    203 
    204 static struct pool bridge_rtnode_pool;
    205 
    206 static int	bridge_clone_create(struct if_clone *, int);
    207 static int	bridge_clone_destroy(struct ifnet *);
    208 
    209 static int	bridge_ioctl(struct ifnet *, u_long, void *);
    210 static int	bridge_init(struct ifnet *);
    211 static void	bridge_stop(struct ifnet *, int);
    212 static void	bridge_start(struct ifnet *);
    213 
    214 static void	bridge_input(struct ifnet *, struct mbuf *);
    215 static void	bridge_forward(struct bridge_softc *, struct mbuf *);
    216 
    217 static void	bridge_timer(void *);
    218 
    219 static void	bridge_broadcast(struct bridge_softc *, struct ifnet *,
    220 				 struct mbuf *);
    221 
    222 static int	bridge_rtupdate(struct bridge_softc *, const uint8_t *,
    223 				struct ifnet *, int, uint8_t);
    224 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *);
    225 static void	bridge_rttrim(struct bridge_softc *);
    226 static void	bridge_rtage(struct bridge_softc *);
    227 static void	bridge_rtage_work(struct work *, void *);
    228 static void	bridge_rtflush(struct bridge_softc *, int);
    229 static int	bridge_rtdaddr(struct bridge_softc *, const uint8_t *);
    230 static void	bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp);
    231 
    232 static void	bridge_rtable_init(struct bridge_softc *);
    233 static void	bridge_rtable_fini(struct bridge_softc *);
    234 
    235 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
    236 						  const uint8_t *);
    237 static int	bridge_rtnode_insert(struct bridge_softc *,
    238 				     struct bridge_rtnode *);
    239 static void	bridge_rtnode_remove(struct bridge_softc *,
    240 				     struct bridge_rtnode *);
    241 static void	bridge_rtnode_destroy(struct bridge_rtnode *);
    242 
    243 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
    244 						  const char *name);
    245 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
    246 						     struct ifnet *ifp);
    247 static void	bridge_release_member(struct bridge_softc *, struct bridge_iflist *);
    248 static void	bridge_delete_member(struct bridge_softc *,
    249 				     struct bridge_iflist *);
    250 static struct bridge_iflist *bridge_try_hold_bif(struct bridge_iflist *);
    251 
    252 static int	bridge_ioctl_add(struct bridge_softc *, void *);
    253 static int	bridge_ioctl_del(struct bridge_softc *, void *);
    254 static int	bridge_ioctl_gifflags(struct bridge_softc *, void *);
    255 static int	bridge_ioctl_sifflags(struct bridge_softc *, void *);
    256 static int	bridge_ioctl_scache(struct bridge_softc *, void *);
    257 static int	bridge_ioctl_gcache(struct bridge_softc *, void *);
    258 static int	bridge_ioctl_gifs(struct bridge_softc *, void *);
    259 static int	bridge_ioctl_rts(struct bridge_softc *, void *);
    260 static int	bridge_ioctl_saddr(struct bridge_softc *, void *);
    261 static int	bridge_ioctl_sto(struct bridge_softc *, void *);
    262 static int	bridge_ioctl_gto(struct bridge_softc *, void *);
    263 static int	bridge_ioctl_daddr(struct bridge_softc *, void *);
    264 static int	bridge_ioctl_flush(struct bridge_softc *, void *);
    265 static int	bridge_ioctl_gpri(struct bridge_softc *, void *);
    266 static int	bridge_ioctl_spri(struct bridge_softc *, void *);
    267 static int	bridge_ioctl_ght(struct bridge_softc *, void *);
    268 static int	bridge_ioctl_sht(struct bridge_softc *, void *);
    269 static int	bridge_ioctl_gfd(struct bridge_softc *, void *);
    270 static int	bridge_ioctl_sfd(struct bridge_softc *, void *);
    271 static int	bridge_ioctl_gma(struct bridge_softc *, void *);
    272 static int	bridge_ioctl_sma(struct bridge_softc *, void *);
    273 static int	bridge_ioctl_sifprio(struct bridge_softc *, void *);
    274 static int	bridge_ioctl_sifcost(struct bridge_softc *, void *);
    275 #if defined(BRIDGE_IPF)
    276 static int	bridge_ioctl_gfilt(struct bridge_softc *, void *);
    277 static int	bridge_ioctl_sfilt(struct bridge_softc *, void *);
    278 static int	bridge_ipf(void *, struct mbuf **, struct ifnet *, int);
    279 static int	bridge_ip_checkbasic(struct mbuf **mp);
    280 # ifdef INET6
    281 static int	bridge_ip6_checkbasic(struct mbuf **mp);
    282 # endif /* INET6 */
    283 #endif /* BRIDGE_IPF */
    284 
    285 struct bridge_control {
    286 	int	(*bc_func)(struct bridge_softc *, void *);
    287 	int	bc_argsize;
    288 	int	bc_flags;
    289 };
    290 
    291 #define	BC_F_COPYIN		0x01	/* copy arguments in */
    292 #define	BC_F_COPYOUT		0x02	/* copy arguments out */
    293 #define	BC_F_SUSER		0x04	/* do super-user check */
    294 #define BC_F_XLATEIN		0x08	/* xlate arguments in */
    295 #define BC_F_XLATEOUT		0x10	/* xlate arguments out */
    296 
    297 static const struct bridge_control bridge_control_table[] = {
    298 [BRDGADD] = {bridge_ioctl_add, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    299 [BRDGDEL] = {bridge_ioctl_del, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    300 
    301 [BRDGGIFFLGS] = {bridge_ioctl_gifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_COPYOUT},
    302 [BRDGSIFFLGS] = {bridge_ioctl_sifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    303 
    304 [BRDGSCACHE] = {bridge_ioctl_scache, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    305 [BRDGGCACHE] = {bridge_ioctl_gcache, sizeof(struct ifbrparam), BC_F_COPYOUT},
    306 
    307 [OBRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_COPYIN|BC_F_COPYOUT},
    308 [OBRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_COPYIN|BC_F_COPYOUT},
    309 
    310 [BRDGSADDR] = {bridge_ioctl_saddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
    311 
    312 [BRDGSTO] = {bridge_ioctl_sto, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    313 [BRDGGTO] = {bridge_ioctl_gto, sizeof(struct ifbrparam), BC_F_COPYOUT},
    314 
    315 [BRDGDADDR] = {bridge_ioctl_daddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
    316 
    317 [BRDGFLUSH] = {bridge_ioctl_flush, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    318 
    319 [BRDGGPRI] = {bridge_ioctl_gpri, sizeof(struct ifbrparam), BC_F_COPYOUT},
    320 [BRDGSPRI] = {bridge_ioctl_spri, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    321 
    322 [BRDGGHT] = {bridge_ioctl_ght, sizeof(struct ifbrparam), BC_F_COPYOUT},
    323 [BRDGSHT] = {bridge_ioctl_sht, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    324 
    325 [BRDGGFD] = {bridge_ioctl_gfd, sizeof(struct ifbrparam), BC_F_COPYOUT},
    326 [BRDGSFD] = {bridge_ioctl_sfd, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    327 
    328 [BRDGGMA] = {bridge_ioctl_gma, sizeof(struct ifbrparam), BC_F_COPYOUT},
    329 [BRDGSMA] = {bridge_ioctl_sma, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    330 
    331 [BRDGSIFPRIO] = {bridge_ioctl_sifprio, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    332 
    333 [BRDGSIFCOST] = {bridge_ioctl_sifcost, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    334 #if defined(BRIDGE_IPF)
    335 [BRDGGFILT] = {bridge_ioctl_gfilt, sizeof(struct ifbrparam), BC_F_COPYOUT},
    336 [BRDGSFILT] = {bridge_ioctl_sfilt, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    337 #endif /* BRIDGE_IPF */
    338 [BRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_XLATEIN|BC_F_XLATEOUT},
    339 [BRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_XLATEIN|BC_F_XLATEOUT},
    340 };
    341 
    342 static const int bridge_control_table_size = __arraycount(bridge_control_table);
    343 
    344 static LIST_HEAD(, bridge_softc) bridge_list;
    345 static kmutex_t bridge_list_lock;
    346 
    347 static struct if_clone bridge_cloner =
    348     IF_CLONE_INITIALIZER("bridge", bridge_clone_create, bridge_clone_destroy);
    349 
    350 /*
    351  * bridgeattach:
    352  *
    353  *	Pseudo-device attach routine.
    354  */
    355 void
    356 bridgeattach(int n)
    357 {
    358 
    359 	pool_init(&bridge_rtnode_pool, sizeof(struct bridge_rtnode),
    360 	    0, 0, 0, "brtpl", NULL, IPL_NET);
    361 
    362 	LIST_INIT(&bridge_list);
    363 	mutex_init(&bridge_list_lock, MUTEX_DEFAULT, IPL_NET);
    364 	if_clone_attach(&bridge_cloner);
    365 }
    366 
    367 /*
    368  * bridge_clone_create:
    369  *
    370  *	Create a new bridge instance.
    371  */
    372 static int
    373 bridge_clone_create(struct if_clone *ifc, int unit)
    374 {
    375 	struct bridge_softc *sc;
    376 	struct ifnet *ifp;
    377 	int error, flags;
    378 
    379 	sc = kmem_zalloc(sizeof(*sc),  KM_SLEEP);
    380 	ifp = &sc->sc_if;
    381 
    382 	sc->sc_brtmax = BRIDGE_RTABLE_MAX;
    383 	sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
    384 	sc->sc_bridge_max_age = BSTP_DEFAULT_MAX_AGE;
    385 	sc->sc_bridge_hello_time = BSTP_DEFAULT_HELLO_TIME;
    386 	sc->sc_bridge_forward_delay = BSTP_DEFAULT_FORWARD_DELAY;
    387 	sc->sc_bridge_priority = BSTP_DEFAULT_BRIDGE_PRIORITY;
    388 	sc->sc_hold_time = BSTP_DEFAULT_HOLD_TIME;
    389 	sc->sc_filter_flags = 0;
    390 
    391 	/* Initialize our routing table. */
    392 	bridge_rtable_init(sc);
    393 
    394 #ifdef BRIDGE_MPSAFE
    395 	flags = WQ_MPSAFE;
    396 #else
    397 	flags = 0;
    398 #endif
    399 	error = workqueue_create(&sc->sc_rtage_wq, "bridge_rtage",
    400 	    bridge_rtage_work, sc, PRI_SOFTNET, IPL_SOFTNET, flags);
    401 	if (error)
    402 		panic("%s: workqueue_create %d\n", __func__, error);
    403 
    404 	callout_init(&sc->sc_brcallout, 0);
    405 	callout_init(&sc->sc_bstpcallout, 0);
    406 
    407 	LIST_INIT(&sc->sc_iflist);
    408 #ifdef BRIDGE_MPSAFE
    409 	sc->sc_iflist_psz = pserialize_create();
    410 	sc->sc_iflist_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_SOFTNET);
    411 #else
    412 	sc->sc_iflist_psz = NULL;
    413 	sc->sc_iflist_lock = NULL;
    414 #endif
    415 	cv_init(&sc->sc_iflist_cv, "if_bridge_cv");
    416 
    417 	if_initname(ifp, ifc->ifc_name, unit);
    418 	ifp->if_softc = sc;
    419 	ifp->if_mtu = ETHERMTU;
    420 	ifp->if_ioctl = bridge_ioctl;
    421 	ifp->if_output = bridge_output;
    422 	ifp->if_start = bridge_start;
    423 	ifp->if_stop = bridge_stop;
    424 	ifp->if_init = bridge_init;
    425 	ifp->if_type = IFT_BRIDGE;
    426 	ifp->if_addrlen = 0;
    427 	ifp->if_dlt = DLT_EN10MB;
    428 	ifp->if_hdrlen = ETHER_HDR_LEN;
    429 
    430 	if_initialize(ifp);
    431 	if_register(ifp);
    432 
    433 	if_alloc_sadl(ifp);
    434 
    435 	mutex_enter(&bridge_list_lock);
    436 	LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
    437 	mutex_exit(&bridge_list_lock);
    438 
    439 	return (0);
    440 }
    441 
    442 /*
    443  * bridge_clone_destroy:
    444  *
    445  *	Destroy a bridge instance.
    446  */
    447 static int
    448 bridge_clone_destroy(struct ifnet *ifp)
    449 {
    450 	struct bridge_softc *sc = ifp->if_softc;
    451 	struct bridge_iflist *bif;
    452 	int s;
    453 
    454 	s = splnet();
    455 
    456 	bridge_stop(ifp, 1);
    457 
    458 	BRIDGE_LOCK(sc);
    459 	while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
    460 		bridge_delete_member(sc, bif);
    461 	BRIDGE_UNLOCK(sc);
    462 
    463 	mutex_enter(&bridge_list_lock);
    464 	LIST_REMOVE(sc, sc_list);
    465 	mutex_exit(&bridge_list_lock);
    466 
    467 	splx(s);
    468 
    469 	if_detach(ifp);
    470 
    471 	/* Tear down the routing table. */
    472 	bridge_rtable_fini(sc);
    473 
    474 	cv_destroy(&sc->sc_iflist_cv);
    475 
    476 	if (sc->sc_iflist_psz)
    477 		pserialize_destroy(sc->sc_iflist_psz);
    478 	if (sc->sc_iflist_lock)
    479 		mutex_obj_free(sc->sc_iflist_lock);
    480 
    481 	workqueue_destroy(sc->sc_rtage_wq);
    482 
    483 	kmem_free(sc, sizeof(*sc));
    484 
    485 	return (0);
    486 }
    487 
    488 /*
    489  * bridge_ioctl:
    490  *
    491  *	Handle a control request from the operator.
    492  */
    493 static int
    494 bridge_ioctl(struct ifnet *ifp, u_long cmd, void *data)
    495 {
    496 	struct bridge_softc *sc = ifp->if_softc;
    497 	struct lwp *l = curlwp;	/* XXX */
    498 	union {
    499 		struct ifbreq ifbreq;
    500 		struct ifbifconf ifbifconf;
    501 		struct ifbareq ifbareq;
    502 		struct ifbaconf ifbaconf;
    503 		struct ifbrparam ifbrparam;
    504 	} args;
    505 	struct ifdrv *ifd = (struct ifdrv *) data;
    506 	const struct bridge_control *bc = NULL; /* XXXGCC */
    507 	int s, error = 0;
    508 
    509 	/* Authorize command before calling splnet(). */
    510 	switch (cmd) {
    511 	case SIOCGDRVSPEC:
    512 	case SIOCSDRVSPEC:
    513 		if (ifd->ifd_cmd >= bridge_control_table_size
    514 		    || (bc = &bridge_control_table[ifd->ifd_cmd]) == NULL) {
    515 			error = EINVAL;
    516 			return error;
    517 		}
    518 
    519 		/* We only care about BC_F_SUSER at this point. */
    520 		if ((bc->bc_flags & BC_F_SUSER) == 0)
    521 			break;
    522 
    523 		error = kauth_authorize_network(l->l_cred,
    524 		    KAUTH_NETWORK_INTERFACE_BRIDGE,
    525 		    cmd == SIOCGDRVSPEC ?
    526 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_GETPRIV :
    527 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_SETPRIV,
    528 		     ifd, NULL, NULL);
    529 		if (error)
    530 			return (error);
    531 
    532 		break;
    533 	}
    534 
    535 	s = splnet();
    536 
    537 	switch (cmd) {
    538 	case SIOCGDRVSPEC:
    539 	case SIOCSDRVSPEC:
    540 		KASSERT(bc != NULL);
    541 		if (cmd == SIOCGDRVSPEC &&
    542 		    (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) == 0) {
    543 			error = EINVAL;
    544 			break;
    545 		}
    546 		else if (cmd == SIOCSDRVSPEC &&
    547 		    (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) != 0) {
    548 			error = EINVAL;
    549 			break;
    550 		}
    551 
    552 		/* BC_F_SUSER is checked above, before splnet(). */
    553 
    554 		if ((bc->bc_flags & (BC_F_XLATEIN|BC_F_XLATEOUT)) == 0
    555 		    && (ifd->ifd_len != bc->bc_argsize
    556 			|| ifd->ifd_len > sizeof(args))) {
    557 			error = EINVAL;
    558 			break;
    559 		}
    560 
    561 		memset(&args, 0, sizeof(args));
    562 		if (bc->bc_flags & BC_F_COPYIN) {
    563 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
    564 			if (error)
    565 				break;
    566 		} else if (bc->bc_flags & BC_F_XLATEIN) {
    567 			args.ifbifconf.ifbic_len = ifd->ifd_len;
    568 			args.ifbifconf.ifbic_buf = ifd->ifd_data;
    569 		}
    570 
    571 		error = (*bc->bc_func)(sc, &args);
    572 		if (error)
    573 			break;
    574 
    575 		if (bc->bc_flags & BC_F_COPYOUT) {
    576 			error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
    577 		} else if (bc->bc_flags & BC_F_XLATEOUT) {
    578 			ifd->ifd_len = args.ifbifconf.ifbic_len;
    579 			ifd->ifd_data = args.ifbifconf.ifbic_buf;
    580 		}
    581 		break;
    582 
    583 	case SIOCSIFFLAGS:
    584 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
    585 			break;
    586 		switch (ifp->if_flags & (IFF_UP|IFF_RUNNING)) {
    587 		case IFF_RUNNING:
    588 			/*
    589 			 * If interface is marked down and it is running,
    590 			 * then stop and disable it.
    591 			 */
    592 			(*ifp->if_stop)(ifp, 1);
    593 			break;
    594 		case IFF_UP:
    595 			/*
    596 			 * If interface is marked up and it is stopped, then
    597 			 * start it.
    598 			 */
    599 			error = (*ifp->if_init)(ifp);
    600 			break;
    601 		default:
    602 			break;
    603 		}
    604 		break;
    605 
    606 	case SIOCSIFMTU:
    607 		if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET)
    608 			error = 0;
    609 		break;
    610 
    611 	default:
    612 		error = ifioctl_common(ifp, cmd, data);
    613 		break;
    614 	}
    615 
    616 	splx(s);
    617 
    618 	return (error);
    619 }
    620 
    621 /*
    622  * bridge_lookup_member:
    623  *
    624  *	Lookup a bridge member interface.
    625  */
    626 static struct bridge_iflist *
    627 bridge_lookup_member(struct bridge_softc *sc, const char *name)
    628 {
    629 	struct bridge_iflist *bif;
    630 	struct ifnet *ifp;
    631 	int s;
    632 
    633 	BRIDGE_PSZ_RENTER(s);
    634 
    635 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
    636 		ifp = bif->bif_ifp;
    637 		if (strcmp(ifp->if_xname, name) == 0)
    638 			break;
    639 	}
    640 	bif = bridge_try_hold_bif(bif);
    641 
    642 	BRIDGE_PSZ_REXIT(s);
    643 
    644 	return bif;
    645 }
    646 
    647 /*
    648  * bridge_lookup_member_if:
    649  *
    650  *	Lookup a bridge member interface by ifnet*.
    651  */
    652 static struct bridge_iflist *
    653 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
    654 {
    655 	struct bridge_iflist *bif;
    656 	int s;
    657 
    658 	BRIDGE_PSZ_RENTER(s);
    659 
    660 	bif = member_ifp->if_bridgeif;
    661 	bif = bridge_try_hold_bif(bif);
    662 
    663 	BRIDGE_PSZ_REXIT(s);
    664 
    665 	return bif;
    666 }
    667 
    668 static struct bridge_iflist *
    669 bridge_try_hold_bif(struct bridge_iflist *bif)
    670 {
    671 #ifdef BRIDGE_MPSAFE
    672 	if (bif != NULL) {
    673 		if (bif->bif_waiting)
    674 			bif = NULL;
    675 		else
    676 			atomic_inc_32(&bif->bif_refs);
    677 	}
    678 #endif
    679 	return bif;
    680 }
    681 
    682 /*
    683  * bridge_release_member:
    684  *
    685  *	Release the specified member interface.
    686  */
    687 static void
    688 bridge_release_member(struct bridge_softc *sc, struct bridge_iflist *bif)
    689 {
    690 #ifdef BRIDGE_MPSAFE
    691 	uint32_t refs;
    692 
    693 	refs = atomic_dec_uint_nv(&bif->bif_refs);
    694 	if (__predict_false(refs == 0 && bif->bif_waiting)) {
    695 		BRIDGE_LOCK(sc);
    696 		cv_broadcast(&sc->sc_iflist_cv);
    697 		BRIDGE_UNLOCK(sc);
    698 	}
    699 #else
    700 	(void)sc;
    701 	(void)bif;
    702 #endif
    703 }
    704 
    705 /*
    706  * bridge_delete_member:
    707  *
    708  *	Delete the specified member interface.
    709  */
    710 static void
    711 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif)
    712 {
    713 	struct ifnet *ifs = bif->bif_ifp;
    714 
    715 	KASSERT(BRIDGE_LOCKED(sc));
    716 
    717 	ifs->_if_input = ether_input;
    718 	ifs->if_bridge = NULL;
    719 	ifs->if_bridgeif = NULL;
    720 
    721 	LIST_REMOVE(bif, bif_next);
    722 
    723 	BRIDGE_PSZ_PERFORM(sc);
    724 
    725 #ifdef BRIDGE_MPSAFE
    726 	bif->bif_waiting = true;
    727 	membar_sync();
    728 	while (bif->bif_refs > 0) {
    729 		aprint_debug("%s: cv_wait on iflist\n", __func__);
    730 		cv_wait(&sc->sc_iflist_cv, sc->sc_iflist_lock);
    731 	}
    732 	bif->bif_waiting = false;
    733 #endif
    734 	BRIDGE_UNLOCK(sc);
    735 
    736 	kmem_free(bif, sizeof(*bif));
    737 
    738 	BRIDGE_LOCK(sc);
    739 }
    740 
    741 static int
    742 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
    743 {
    744 	struct ifbreq *req = arg;
    745 	struct bridge_iflist *bif = NULL;
    746 	struct ifnet *ifs;
    747 	int error = 0;
    748 
    749 	ifs = ifunit(req->ifbr_ifsname);
    750 	if (ifs == NULL)
    751 		return (ENOENT);
    752 
    753 	if (sc->sc_if.if_mtu != ifs->if_mtu)
    754 		return (EINVAL);
    755 
    756 	if (ifs->if_bridge == sc)
    757 		return (EEXIST);
    758 
    759 	if (ifs->if_bridge != NULL)
    760 		return (EBUSY);
    761 
    762 	if (ifs->_if_input != ether_input)
    763 		return EINVAL;
    764 
    765 	/* FIXME: doesn't work with non-IFF_SIMPLEX interfaces */
    766 	if ((ifs->if_flags & IFF_SIMPLEX) == 0)
    767 		return EINVAL;
    768 
    769 	bif = kmem_alloc(sizeof(*bif), KM_SLEEP);
    770 
    771 	switch (ifs->if_type) {
    772 	case IFT_ETHER:
    773 		if ((error = ether_enable_vlan_mtu(ifs)) > 0)
    774 			goto out;
    775 		/*
    776 		 * Place the interface into promiscuous mode.
    777 		 */
    778 		error = ifpromisc(ifs, 1);
    779 		if (error)
    780 			goto out;
    781 		break;
    782 	default:
    783 		error = EINVAL;
    784 		goto out;
    785 	}
    786 
    787 	bif->bif_ifp = ifs;
    788 	bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
    789 	bif->bif_priority = BSTP_DEFAULT_PORT_PRIORITY;
    790 	bif->bif_path_cost = BSTP_DEFAULT_PATH_COST;
    791 	bif->bif_refs = 0;
    792 	bif->bif_waiting = false;
    793 
    794 	BRIDGE_LOCK(sc);
    795 
    796 	ifs->if_bridge = sc;
    797 	ifs->if_bridgeif = bif;
    798 	LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
    799 	ifs->_if_input = bridge_input;
    800 
    801 	BRIDGE_UNLOCK(sc);
    802 
    803 	if (sc->sc_if.if_flags & IFF_RUNNING)
    804 		bstp_initialization(sc);
    805 	else
    806 		bstp_stop(sc);
    807 
    808  out:
    809 	if (error) {
    810 		if (bif != NULL)
    811 			kmem_free(bif, sizeof(*bif));
    812 	}
    813 	return (error);
    814 }
    815 
    816 static int
    817 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
    818 {
    819 	struct ifbreq *req = arg;
    820 	const char *name = req->ifbr_ifsname;
    821 	struct bridge_iflist *bif;
    822 	struct ifnet *ifs;
    823 
    824 	BRIDGE_LOCK(sc);
    825 
    826 	/*
    827 	 * Don't use bridge_lookup_member. We want to get a member
    828 	 * with bif_refs == 0.
    829 	 */
    830 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
    831 		ifs = bif->bif_ifp;
    832 		if (strcmp(ifs->if_xname, name) == 0)
    833 			break;
    834 	}
    835 
    836 	if (bif == NULL) {
    837 		BRIDGE_UNLOCK(sc);
    838 		return ENOENT;
    839 	}
    840 
    841 	bridge_delete_member(sc, bif);
    842 
    843 	BRIDGE_UNLOCK(sc);
    844 
    845 	switch (ifs->if_type) {
    846 	case IFT_ETHER:
    847 		/*
    848 		 * Take the interface out of promiscuous mode.
    849 		 * Don't call it with holding a spin lock.
    850 		 */
    851 		(void) ifpromisc(ifs, 0);
    852 		(void) ether_disable_vlan_mtu(ifs);
    853 		break;
    854 	default:
    855 #ifdef DIAGNOSTIC
    856 		panic("bridge_delete_member: impossible");
    857 #endif
    858 		break;
    859 	}
    860 
    861 	bridge_rtdelete(sc, ifs);
    862 
    863 	if (sc->sc_if.if_flags & IFF_RUNNING)
    864 		bstp_initialization(sc);
    865 
    866 	return 0;
    867 }
    868 
    869 static int
    870 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
    871 {
    872 	struct ifbreq *req = arg;
    873 	struct bridge_iflist *bif;
    874 
    875 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
    876 	if (bif == NULL)
    877 		return (ENOENT);
    878 
    879 	req->ifbr_ifsflags = bif->bif_flags;
    880 	req->ifbr_state = bif->bif_state;
    881 	req->ifbr_priority = bif->bif_priority;
    882 	req->ifbr_path_cost = bif->bif_path_cost;
    883 	req->ifbr_portno = bif->bif_ifp->if_index & 0xff;
    884 
    885 	bridge_release_member(sc, bif);
    886 
    887 	return (0);
    888 }
    889 
    890 static int
    891 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
    892 {
    893 	struct ifbreq *req = arg;
    894 	struct bridge_iflist *bif;
    895 
    896 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
    897 	if (bif == NULL)
    898 		return (ENOENT);
    899 
    900 	if (req->ifbr_ifsflags & IFBIF_STP) {
    901 		switch (bif->bif_ifp->if_type) {
    902 		case IFT_ETHER:
    903 			/* These can do spanning tree. */
    904 			break;
    905 
    906 		default:
    907 			/* Nothing else can. */
    908 			bridge_release_member(sc, bif);
    909 			return (EINVAL);
    910 		}
    911 	}
    912 
    913 	bif->bif_flags = req->ifbr_ifsflags;
    914 
    915 	bridge_release_member(sc, bif);
    916 
    917 	if (sc->sc_if.if_flags & IFF_RUNNING)
    918 		bstp_initialization(sc);
    919 
    920 	return (0);
    921 }
    922 
    923 static int
    924 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
    925 {
    926 	struct ifbrparam *param = arg;
    927 
    928 	sc->sc_brtmax = param->ifbrp_csize;
    929 	bridge_rttrim(sc);
    930 
    931 	return (0);
    932 }
    933 
    934 static int
    935 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
    936 {
    937 	struct ifbrparam *param = arg;
    938 
    939 	param->ifbrp_csize = sc->sc_brtmax;
    940 
    941 	return (0);
    942 }
    943 
    944 static int
    945 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
    946 {
    947 	struct ifbifconf *bifc = arg;
    948 	struct bridge_iflist *bif;
    949 	struct ifbreq *breqs;
    950 	int i, count, error = 0;
    951 
    952 retry:
    953 	BRIDGE_LOCK(sc);
    954 	count = 0;
    955 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
    956 		count++;
    957 	BRIDGE_UNLOCK(sc);
    958 
    959 	if (count == 0) {
    960 		bifc->ifbic_len = 0;
    961 		return 0;
    962 	}
    963 
    964 	if (bifc->ifbic_len == 0 || bifc->ifbic_len < (sizeof(*breqs) * count)) {
    965 		/* Tell that a larger buffer is needed */
    966 		bifc->ifbic_len = sizeof(*breqs) * count;
    967 		return 0;
    968 	}
    969 
    970 	breqs = kmem_alloc(sizeof(*breqs) * count, KM_SLEEP);
    971 
    972 	BRIDGE_LOCK(sc);
    973 
    974 	i = 0;
    975 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
    976 		i++;
    977 	if (i > count) {
    978 		/*
    979 		 * The number of members has been increased.
    980 		 * We need more memory!
    981 		 */
    982 		BRIDGE_UNLOCK(sc);
    983 		kmem_free(breqs, sizeof(*breqs) * count);
    984 		goto retry;
    985 	}
    986 
    987 	i = 0;
    988 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
    989 		struct ifbreq *breq = &breqs[i++];
    990 		memset(breq, 0, sizeof(*breq));
    991 
    992 		strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname,
    993 		    sizeof(breq->ifbr_ifsname));
    994 		breq->ifbr_ifsflags = bif->bif_flags;
    995 		breq->ifbr_state = bif->bif_state;
    996 		breq->ifbr_priority = bif->bif_priority;
    997 		breq->ifbr_path_cost = bif->bif_path_cost;
    998 		breq->ifbr_portno = bif->bif_ifp->if_index & 0xff;
    999 	}
   1000 
   1001 	/* Don't call copyout with holding the mutex */
   1002 	BRIDGE_UNLOCK(sc);
   1003 
   1004 	for (i = 0; i < count; i++) {
   1005 		error = copyout(&breqs[i], bifc->ifbic_req + i, sizeof(*breqs));
   1006 		if (error)
   1007 			break;
   1008 	}
   1009 	bifc->ifbic_len = sizeof(*breqs) * i;
   1010 
   1011 	kmem_free(breqs, sizeof(*breqs) * count);
   1012 
   1013 	return error;
   1014 }
   1015 
   1016 static int
   1017 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
   1018 {
   1019 	struct ifbaconf *bac = arg;
   1020 	struct bridge_rtnode *brt;
   1021 	struct ifbareq bareq;
   1022 	int count = 0, error = 0, len;
   1023 
   1024 	if (bac->ifbac_len == 0)
   1025 		return (0);
   1026 
   1027 	BRIDGE_RT_LOCK(sc);
   1028 
   1029 	len = bac->ifbac_len;
   1030 	LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
   1031 		if (len < sizeof(bareq))
   1032 			goto out;
   1033 		memset(&bareq, 0, sizeof(bareq));
   1034 		strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
   1035 		    sizeof(bareq.ifba_ifsname));
   1036 		memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
   1037 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
   1038 			bareq.ifba_expire = brt->brt_expire - time_uptime;
   1039 		} else
   1040 			bareq.ifba_expire = 0;
   1041 		bareq.ifba_flags = brt->brt_flags;
   1042 
   1043 		error = copyout(&bareq, bac->ifbac_req + count, sizeof(bareq));
   1044 		if (error)
   1045 			goto out;
   1046 		count++;
   1047 		len -= sizeof(bareq);
   1048 	}
   1049  out:
   1050 	BRIDGE_RT_UNLOCK(sc);
   1051 
   1052 	bac->ifbac_len = sizeof(bareq) * count;
   1053 	return (error);
   1054 }
   1055 
   1056 static int
   1057 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
   1058 {
   1059 	struct ifbareq *req = arg;
   1060 	struct bridge_iflist *bif;
   1061 	int error;
   1062 
   1063 	bif = bridge_lookup_member(sc, req->ifba_ifsname);
   1064 	if (bif == NULL)
   1065 		return (ENOENT);
   1066 
   1067 	error = bridge_rtupdate(sc, req->ifba_dst, bif->bif_ifp, 1,
   1068 	    req->ifba_flags);
   1069 
   1070 	bridge_release_member(sc, bif);
   1071 
   1072 	return (error);
   1073 }
   1074 
   1075 static int
   1076 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
   1077 {
   1078 	struct ifbrparam *param = arg;
   1079 
   1080 	sc->sc_brttimeout = param->ifbrp_ctime;
   1081 
   1082 	return (0);
   1083 }
   1084 
   1085 static int
   1086 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
   1087 {
   1088 	struct ifbrparam *param = arg;
   1089 
   1090 	param->ifbrp_ctime = sc->sc_brttimeout;
   1091 
   1092 	return (0);
   1093 }
   1094 
   1095 static int
   1096 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
   1097 {
   1098 	struct ifbareq *req = arg;
   1099 
   1100 	return (bridge_rtdaddr(sc, req->ifba_dst));
   1101 }
   1102 
   1103 static int
   1104 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
   1105 {
   1106 	struct ifbreq *req = arg;
   1107 
   1108 	bridge_rtflush(sc, req->ifbr_ifsflags);
   1109 
   1110 	return (0);
   1111 }
   1112 
   1113 static int
   1114 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
   1115 {
   1116 	struct ifbrparam *param = arg;
   1117 
   1118 	param->ifbrp_prio = sc->sc_bridge_priority;
   1119 
   1120 	return (0);
   1121 }
   1122 
   1123 static int
   1124 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
   1125 {
   1126 	struct ifbrparam *param = arg;
   1127 
   1128 	sc->sc_bridge_priority = param->ifbrp_prio;
   1129 
   1130 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1131 		bstp_initialization(sc);
   1132 
   1133 	return (0);
   1134 }
   1135 
   1136 static int
   1137 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
   1138 {
   1139 	struct ifbrparam *param = arg;
   1140 
   1141 	param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8;
   1142 
   1143 	return (0);
   1144 }
   1145 
   1146 static int
   1147 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
   1148 {
   1149 	struct ifbrparam *param = arg;
   1150 
   1151 	if (param->ifbrp_hellotime == 0)
   1152 		return (EINVAL);
   1153 	sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8;
   1154 
   1155 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1156 		bstp_initialization(sc);
   1157 
   1158 	return (0);
   1159 }
   1160 
   1161 static int
   1162 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
   1163 {
   1164 	struct ifbrparam *param = arg;
   1165 
   1166 	param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8;
   1167 
   1168 	return (0);
   1169 }
   1170 
   1171 static int
   1172 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
   1173 {
   1174 	struct ifbrparam *param = arg;
   1175 
   1176 	if (param->ifbrp_fwddelay == 0)
   1177 		return (EINVAL);
   1178 	sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8;
   1179 
   1180 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1181 		bstp_initialization(sc);
   1182 
   1183 	return (0);
   1184 }
   1185 
   1186 static int
   1187 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
   1188 {
   1189 	struct ifbrparam *param = arg;
   1190 
   1191 	param->ifbrp_maxage = sc->sc_bridge_max_age >> 8;
   1192 
   1193 	return (0);
   1194 }
   1195 
   1196 static int
   1197 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
   1198 {
   1199 	struct ifbrparam *param = arg;
   1200 
   1201 	if (param->ifbrp_maxage == 0)
   1202 		return (EINVAL);
   1203 	sc->sc_bridge_max_age = param->ifbrp_maxage << 8;
   1204 
   1205 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1206 		bstp_initialization(sc);
   1207 
   1208 	return (0);
   1209 }
   1210 
   1211 static int
   1212 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
   1213 {
   1214 	struct ifbreq *req = arg;
   1215 	struct bridge_iflist *bif;
   1216 
   1217 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
   1218 	if (bif == NULL)
   1219 		return (ENOENT);
   1220 
   1221 	bif->bif_priority = req->ifbr_priority;
   1222 
   1223 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1224 		bstp_initialization(sc);
   1225 
   1226 	bridge_release_member(sc, bif);
   1227 
   1228 	return (0);
   1229 }
   1230 
   1231 #if defined(BRIDGE_IPF)
   1232 static int
   1233 bridge_ioctl_gfilt(struct bridge_softc *sc, void *arg)
   1234 {
   1235 	struct ifbrparam *param = arg;
   1236 
   1237 	param->ifbrp_filter = sc->sc_filter_flags;
   1238 
   1239 	return (0);
   1240 }
   1241 
   1242 static int
   1243 bridge_ioctl_sfilt(struct bridge_softc *sc, void *arg)
   1244 {
   1245 	struct ifbrparam *param = arg;
   1246 	uint32_t nflags, oflags;
   1247 
   1248 	if (param->ifbrp_filter & ~IFBF_FILT_MASK)
   1249 		return (EINVAL);
   1250 
   1251 	nflags = param->ifbrp_filter;
   1252 	oflags = sc->sc_filter_flags;
   1253 
   1254 	if ((nflags & IFBF_FILT_USEIPF) && !(oflags & IFBF_FILT_USEIPF)) {
   1255 		pfil_add_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
   1256 			sc->sc_if.if_pfil);
   1257 	}
   1258 	if (!(nflags & IFBF_FILT_USEIPF) && (oflags & IFBF_FILT_USEIPF)) {
   1259 		pfil_remove_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
   1260 			sc->sc_if.if_pfil);
   1261 	}
   1262 
   1263 	sc->sc_filter_flags = nflags;
   1264 
   1265 	return (0);
   1266 }
   1267 #endif /* BRIDGE_IPF */
   1268 
   1269 static int
   1270 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
   1271 {
   1272 	struct ifbreq *req = arg;
   1273 	struct bridge_iflist *bif;
   1274 
   1275 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
   1276 	if (bif == NULL)
   1277 		return (ENOENT);
   1278 
   1279 	bif->bif_path_cost = req->ifbr_path_cost;
   1280 
   1281 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1282 		bstp_initialization(sc);
   1283 
   1284 	bridge_release_member(sc, bif);
   1285 
   1286 	return (0);
   1287 }
   1288 
   1289 /*
   1290  * bridge_ifdetach:
   1291  *
   1292  *	Detach an interface from a bridge.  Called when a member
   1293  *	interface is detaching.
   1294  */
   1295 void
   1296 bridge_ifdetach(struct ifnet *ifp)
   1297 {
   1298 	struct bridge_softc *sc = ifp->if_bridge;
   1299 	struct ifbreq breq;
   1300 
   1301 	/* ioctl_lock should prevent this from happening */
   1302 	KASSERT(sc != NULL);
   1303 
   1304 	memset(&breq, 0, sizeof(breq));
   1305 	strlcpy(breq.ifbr_ifsname, ifp->if_xname, sizeof(breq.ifbr_ifsname));
   1306 
   1307 	(void) bridge_ioctl_del(sc, &breq);
   1308 }
   1309 
   1310 /*
   1311  * bridge_init:
   1312  *
   1313  *	Initialize a bridge interface.
   1314  */
   1315 static int
   1316 bridge_init(struct ifnet *ifp)
   1317 {
   1318 	struct bridge_softc *sc = ifp->if_softc;
   1319 
   1320 	if (ifp->if_flags & IFF_RUNNING)
   1321 		return (0);
   1322 
   1323 	callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
   1324 	    bridge_timer, sc);
   1325 
   1326 	ifp->if_flags |= IFF_RUNNING;
   1327 	bstp_initialization(sc);
   1328 	return (0);
   1329 }
   1330 
   1331 /*
   1332  * bridge_stop:
   1333  *
   1334  *	Stop the bridge interface.
   1335  */
   1336 static void
   1337 bridge_stop(struct ifnet *ifp, int disable)
   1338 {
   1339 	struct bridge_softc *sc = ifp->if_softc;
   1340 
   1341 	if ((ifp->if_flags & IFF_RUNNING) == 0)
   1342 		return;
   1343 
   1344 	callout_stop(&sc->sc_brcallout);
   1345 	bstp_stop(sc);
   1346 
   1347 	bridge_rtflush(sc, IFBF_FLUSHDYN);
   1348 
   1349 	ifp->if_flags &= ~IFF_RUNNING;
   1350 }
   1351 
   1352 /*
   1353  * bridge_enqueue:
   1354  *
   1355  *	Enqueue a packet on a bridge member interface.
   1356  */
   1357 void
   1358 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m,
   1359     int runfilt)
   1360 {
   1361 	ALTQ_DECL(struct altq_pktattr pktattr;)
   1362 	int len, error;
   1363 	short mflags;
   1364 
   1365 	/*
   1366 	 * Clear any in-bound checksum flags for this packet.
   1367 	 */
   1368 	m->m_pkthdr.csum_flags = 0;
   1369 
   1370 	if (runfilt) {
   1371 		if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
   1372 		    dst_ifp, PFIL_OUT) != 0) {
   1373 			if (m != NULL)
   1374 				m_freem(m);
   1375 			return;
   1376 		}
   1377 		if (m == NULL)
   1378 			return;
   1379 	}
   1380 
   1381 #ifdef ALTQ
   1382 	/*
   1383 	 * If ALTQ is enabled on the member interface, do
   1384 	 * classification; the queueing discipline might
   1385 	 * not require classification, but might require
   1386 	 * the address family/header pointer in the pktattr.
   1387 	 */
   1388 	if (ALTQ_IS_ENABLED(&dst_ifp->if_snd)) {
   1389 		/* XXX IFT_ETHER */
   1390 		altq_etherclassify(&dst_ifp->if_snd, m, &pktattr);
   1391 	}
   1392 #endif /* ALTQ */
   1393 
   1394 	len = m->m_pkthdr.len;
   1395 	mflags = m->m_flags;
   1396 
   1397 	IFQ_ENQUEUE(&dst_ifp->if_snd, m, &pktattr, error);
   1398 
   1399 	if (error) {
   1400 		/* mbuf is already freed */
   1401 		sc->sc_if.if_oerrors++;
   1402 		return;
   1403 	}
   1404 
   1405 	sc->sc_if.if_opackets++;
   1406 	sc->sc_if.if_obytes += len;
   1407 
   1408 	dst_ifp->if_obytes += len;
   1409 
   1410 	if (mflags & M_MCAST) {
   1411 		sc->sc_if.if_omcasts++;
   1412 		dst_ifp->if_omcasts++;
   1413 	}
   1414 
   1415 	if ((dst_ifp->if_flags & IFF_OACTIVE) == 0)
   1416 		(*dst_ifp->if_start)(dst_ifp);
   1417 }
   1418 
   1419 /*
   1420  * bridge_output:
   1421  *
   1422  *	Send output from a bridge member interface.  This
   1423  *	performs the bridging function for locally originated
   1424  *	packets.
   1425  *
   1426  *	The mbuf has the Ethernet header already attached.  We must
   1427  *	enqueue or free the mbuf before returning.
   1428  */
   1429 int
   1430 bridge_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa,
   1431     struct rtentry *rt)
   1432 {
   1433 	struct ether_header *eh;
   1434 	struct ifnet *dst_if;
   1435 	struct bridge_softc *sc;
   1436 #ifndef BRIDGE_MPSAFE
   1437 	int s;
   1438 #endif
   1439 
   1440 	if (m->m_len < ETHER_HDR_LEN) {
   1441 		m = m_pullup(m, ETHER_HDR_LEN);
   1442 		if (m == NULL)
   1443 			return (0);
   1444 	}
   1445 
   1446 	eh = mtod(m, struct ether_header *);
   1447 	sc = ifp->if_bridge;
   1448 
   1449 #ifndef BRIDGE_MPSAFE
   1450 	s = splnet();
   1451 #endif
   1452 
   1453 	/*
   1454 	 * If bridge is down, but the original output interface is up,
   1455 	 * go ahead and send out that interface.  Otherwise, the packet
   1456 	 * is dropped below.
   1457 	 */
   1458 	if (__predict_false(sc == NULL) ||
   1459 	    (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
   1460 		dst_if = ifp;
   1461 		goto sendunicast;
   1462 	}
   1463 
   1464 	/*
   1465 	 * If the packet is a multicast, or we don't know a better way to
   1466 	 * get there, send to all interfaces.
   1467 	 */
   1468 	if (ETHER_IS_MULTICAST(eh->ether_dhost))
   1469 		dst_if = NULL;
   1470 	else
   1471 		dst_if = bridge_rtlookup(sc, eh->ether_dhost);
   1472 	if (dst_if == NULL) {
   1473 		struct bridge_iflist *bif;
   1474 		struct mbuf *mc;
   1475 		int used = 0;
   1476 		int ss;
   1477 
   1478 		BRIDGE_PSZ_RENTER(ss);
   1479 		LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
   1480 			bif = bridge_try_hold_bif(bif);
   1481 			if (bif == NULL)
   1482 				continue;
   1483 			BRIDGE_PSZ_REXIT(ss);
   1484 
   1485 			dst_if = bif->bif_ifp;
   1486 			if ((dst_if->if_flags & IFF_RUNNING) == 0)
   1487 				goto next;
   1488 
   1489 			/*
   1490 			 * If this is not the original output interface,
   1491 			 * and the interface is participating in spanning
   1492 			 * tree, make sure the port is in a state that
   1493 			 * allows forwarding.
   1494 			 */
   1495 			if (dst_if != ifp &&
   1496 			    (bif->bif_flags & IFBIF_STP) != 0) {
   1497 				switch (bif->bif_state) {
   1498 				case BSTP_IFSTATE_BLOCKING:
   1499 				case BSTP_IFSTATE_LISTENING:
   1500 				case BSTP_IFSTATE_DISABLED:
   1501 					goto next;
   1502 				}
   1503 			}
   1504 
   1505 			if (LIST_NEXT(bif, bif_next) == NULL) {
   1506 				used = 1;
   1507 				mc = m;
   1508 			} else {
   1509 				mc = m_copym(m, 0, M_COPYALL, M_NOWAIT);
   1510 				if (mc == NULL) {
   1511 					sc->sc_if.if_oerrors++;
   1512 					goto next;
   1513 				}
   1514 			}
   1515 
   1516 			bridge_enqueue(sc, dst_if, mc, 0);
   1517 next:
   1518 			bridge_release_member(sc, bif);
   1519 			BRIDGE_PSZ_RENTER(ss);
   1520 		}
   1521 		BRIDGE_PSZ_REXIT(ss);
   1522 
   1523 		if (used == 0)
   1524 			m_freem(m);
   1525 #ifndef BRIDGE_MPSAFE
   1526 		splx(s);
   1527 #endif
   1528 		return (0);
   1529 	}
   1530 
   1531  sendunicast:
   1532 	/*
   1533 	 * XXX Spanning tree consideration here?
   1534 	 */
   1535 
   1536 	if ((dst_if->if_flags & IFF_RUNNING) == 0) {
   1537 		m_freem(m);
   1538 #ifndef BRIDGE_MPSAFE
   1539 		splx(s);
   1540 #endif
   1541 		return (0);
   1542 	}
   1543 
   1544 	bridge_enqueue(sc, dst_if, m, 0);
   1545 
   1546 #ifndef BRIDGE_MPSAFE
   1547 	splx(s);
   1548 #endif
   1549 	return (0);
   1550 }
   1551 
   1552 /*
   1553  * bridge_start:
   1554  *
   1555  *	Start output on a bridge.
   1556  *
   1557  *	NOTE: This routine should never be called in this implementation.
   1558  */
   1559 static void
   1560 bridge_start(struct ifnet *ifp)
   1561 {
   1562 
   1563 	printf("%s: bridge_start() called\n", ifp->if_xname);
   1564 }
   1565 
   1566 /*
   1567  * bridge_forward:
   1568  *
   1569  *	The forwarding function of the bridge.
   1570  */
   1571 static void
   1572 bridge_forward(struct bridge_softc *sc, struct mbuf *m)
   1573 {
   1574 	struct bridge_iflist *bif;
   1575 	struct ifnet *src_if, *dst_if;
   1576 	struct ether_header *eh;
   1577 #ifndef BRIDGE_MPSAFE
   1578 	int s;
   1579 
   1580 	KERNEL_LOCK(1, NULL);
   1581 	mutex_enter(softnet_lock);
   1582 #endif
   1583 
   1584 	if ((sc->sc_if.if_flags & IFF_RUNNING) == 0) {
   1585 #ifndef BRIDGE_MPSAFE
   1586 		mutex_exit(softnet_lock);
   1587 		KERNEL_UNLOCK_ONE(NULL);
   1588 #endif
   1589 		return;
   1590 	}
   1591 
   1592 #ifndef BRIDGE_MPSAFE
   1593 	s = splnet();
   1594 #endif
   1595 
   1596 	src_if = m->m_pkthdr.rcvif;
   1597 
   1598 	sc->sc_if.if_ipackets++;
   1599 	sc->sc_if.if_ibytes += m->m_pkthdr.len;
   1600 
   1601 	/*
   1602 	 * Look up the bridge_iflist.
   1603 	 */
   1604 	bif = bridge_lookup_member_if(sc, src_if);
   1605 	if (bif == NULL) {
   1606 		/* Interface is not a bridge member (anymore?) */
   1607 		m_freem(m);
   1608 		goto out;
   1609 	}
   1610 
   1611 	if (bif->bif_flags & IFBIF_STP) {
   1612 		switch (bif->bif_state) {
   1613 		case BSTP_IFSTATE_BLOCKING:
   1614 		case BSTP_IFSTATE_LISTENING:
   1615 		case BSTP_IFSTATE_DISABLED:
   1616 			m_freem(m);
   1617 			bridge_release_member(sc, bif);
   1618 			goto out;
   1619 		}
   1620 	}
   1621 
   1622 	eh = mtod(m, struct ether_header *);
   1623 
   1624 	/*
   1625 	 * If the interface is learning, and the source
   1626 	 * address is valid and not multicast, record
   1627 	 * the address.
   1628 	 */
   1629 	if ((bif->bif_flags & IFBIF_LEARNING) != 0 &&
   1630 	    ETHER_IS_MULTICAST(eh->ether_shost) == 0 &&
   1631 	    (eh->ether_shost[0] == 0 &&
   1632 	     eh->ether_shost[1] == 0 &&
   1633 	     eh->ether_shost[2] == 0 &&
   1634 	     eh->ether_shost[3] == 0 &&
   1635 	     eh->ether_shost[4] == 0 &&
   1636 	     eh->ether_shost[5] == 0) == 0) {
   1637 		(void) bridge_rtupdate(sc, eh->ether_shost,
   1638 		    src_if, 0, IFBAF_DYNAMIC);
   1639 	}
   1640 
   1641 	if ((bif->bif_flags & IFBIF_STP) != 0 &&
   1642 	    bif->bif_state == BSTP_IFSTATE_LEARNING) {
   1643 		m_freem(m);
   1644 		bridge_release_member(sc, bif);
   1645 		goto out;
   1646 	}
   1647 
   1648 	bridge_release_member(sc, bif);
   1649 
   1650 	/*
   1651 	 * At this point, the port either doesn't participate
   1652 	 * in spanning tree or it is in the forwarding state.
   1653 	 */
   1654 
   1655 	/*
   1656 	 * If the packet is unicast, destined for someone on
   1657 	 * "this" side of the bridge, drop it.
   1658 	 */
   1659 	if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
   1660 		dst_if = bridge_rtlookup(sc, eh->ether_dhost);
   1661 		if (src_if == dst_if) {
   1662 			m_freem(m);
   1663 			goto out;
   1664 		}
   1665 	} else {
   1666 		/* ...forward it to all interfaces. */
   1667 		sc->sc_if.if_imcasts++;
   1668 		dst_if = NULL;
   1669 	}
   1670 
   1671 	if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
   1672 	    m->m_pkthdr.rcvif, PFIL_IN) != 0) {
   1673 		if (m != NULL)
   1674 			m_freem(m);
   1675 		goto out;
   1676 	}
   1677 	if (m == NULL)
   1678 		goto out;
   1679 
   1680 	if (dst_if == NULL) {
   1681 		bridge_broadcast(sc, src_if, m);
   1682 		goto out;
   1683 	}
   1684 
   1685 	/*
   1686 	 * At this point, we're dealing with a unicast frame
   1687 	 * going to a different interface.
   1688 	 */
   1689 	if ((dst_if->if_flags & IFF_RUNNING) == 0) {
   1690 		m_freem(m);
   1691 		goto out;
   1692 	}
   1693 
   1694 	bif = bridge_lookup_member_if(sc, dst_if);
   1695 	if (bif == NULL) {
   1696 		/* Not a member of the bridge (anymore?) */
   1697 		m_freem(m);
   1698 		goto out;
   1699 	}
   1700 
   1701 	if (bif->bif_flags & IFBIF_STP) {
   1702 		switch (bif->bif_state) {
   1703 		case BSTP_IFSTATE_DISABLED:
   1704 		case BSTP_IFSTATE_BLOCKING:
   1705 			m_freem(m);
   1706 			bridge_release_member(sc, bif);
   1707 			goto out;
   1708 		}
   1709 	}
   1710 
   1711 	bridge_release_member(sc, bif);
   1712 
   1713 	bridge_enqueue(sc, dst_if, m, 1);
   1714 out:
   1715 #ifndef BRIDGE_MPSAFE
   1716 	splx(s);
   1717 	mutex_exit(softnet_lock);
   1718 	KERNEL_UNLOCK_ONE(NULL);
   1719 #else
   1720 	/* XXX gcc */
   1721 	return;
   1722 #endif
   1723 }
   1724 
   1725 static bool
   1726 bstp_state_before_learning(struct bridge_iflist *bif)
   1727 {
   1728 	if (bif->bif_flags & IFBIF_STP) {
   1729 		switch (bif->bif_state) {
   1730 		case BSTP_IFSTATE_BLOCKING:
   1731 		case BSTP_IFSTATE_LISTENING:
   1732 		case BSTP_IFSTATE_DISABLED:
   1733 			return true;
   1734 		}
   1735 	}
   1736 	return false;
   1737 }
   1738 
   1739 static bool
   1740 bridge_ourether(struct bridge_iflist *bif, struct ether_header *eh, int src)
   1741 {
   1742 	uint8_t *ether = src ? eh->ether_shost : eh->ether_dhost;
   1743 
   1744 	if (memcmp(CLLADDR(bif->bif_ifp->if_sadl), ether, ETHER_ADDR_LEN) == 0
   1745 #if NCARP > 0
   1746 	    || (bif->bif_ifp->if_carp &&
   1747 	        carp_ourether(bif->bif_ifp->if_carp, eh, IFT_ETHER, src) != NULL)
   1748 #endif /* NCARP > 0 */
   1749 	    )
   1750 		return true;
   1751 
   1752 	return false;
   1753 }
   1754 
   1755 /*
   1756  * bridge_input:
   1757  *
   1758  *	Receive input from a member interface.  Queue the packet for
   1759  *	bridging if it is not for us.
   1760  */
   1761 static void
   1762 bridge_input(struct ifnet *ifp, struct mbuf *m)
   1763 {
   1764 	struct bridge_softc *sc = ifp->if_bridge;
   1765 	struct bridge_iflist *bif;
   1766 	struct ether_header *eh;
   1767 
   1768 	KASSERT(!cpu_intr_p());
   1769 
   1770 	if (__predict_false(sc == NULL) ||
   1771 	    (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
   1772 		ether_input(ifp, m);
   1773 		return;
   1774 	}
   1775 
   1776 	bif = bridge_lookup_member_if(sc, ifp);
   1777 	if (bif == NULL) {
   1778 		ether_input(ifp, m);
   1779 		return;
   1780 	}
   1781 
   1782 	eh = mtod(m, struct ether_header *);
   1783 
   1784 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
   1785 		if (memcmp(etherbroadcastaddr,
   1786 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
   1787 			m->m_flags |= M_BCAST;
   1788 		else
   1789 			m->m_flags |= M_MCAST;
   1790 	}
   1791 
   1792 	/*
   1793 	 * A 'fast' path for packets addressed to interfaces that are
   1794 	 * part of this bridge.
   1795 	 */
   1796 	if (!(m->m_flags & (M_BCAST|M_MCAST)) &&
   1797 	    !bstp_state_before_learning(bif)) {
   1798 		struct bridge_iflist *_bif;
   1799 		struct ifnet *_ifp = NULL;
   1800 		int s;
   1801 
   1802 		BRIDGE_PSZ_RENTER(s);
   1803 		LIST_FOREACH(_bif, &sc->sc_iflist, bif_next) {
   1804 			/* It is destined for us. */
   1805 			if (bridge_ourether(_bif, eh, 0)) {
   1806 				_bif = bridge_try_hold_bif(_bif);
   1807 				BRIDGE_PSZ_REXIT(s);
   1808 				if (_bif == NULL)
   1809 					goto out;
   1810 				if (_bif->bif_flags & IFBIF_LEARNING)
   1811 					(void) bridge_rtupdate(sc,
   1812 					    eh->ether_shost, ifp, 0, IFBAF_DYNAMIC);
   1813 				_ifp = m->m_pkthdr.rcvif = _bif->bif_ifp;
   1814 				bridge_release_member(sc, _bif);
   1815 				goto out;
   1816 			}
   1817 
   1818 			/* We just received a packet that we sent out. */
   1819 			if (bridge_ourether(_bif, eh, 1))
   1820 				break;
   1821 		}
   1822 		BRIDGE_PSZ_REXIT(s);
   1823 out:
   1824 
   1825 		if (_bif != NULL) {
   1826 			bridge_release_member(sc, bif);
   1827 			if (_ifp != NULL) {
   1828 				m->m_flags &= ~M_PROMISC;
   1829 				ether_input(_ifp, m);
   1830 			} else
   1831 				m_freem(m);
   1832 			return;
   1833 		}
   1834 	}
   1835 
   1836 	/* Tap off 802.1D packets; they do not get forwarded. */
   1837 	if (bif->bif_flags & IFBIF_STP &&
   1838 	    memcmp(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN) == 0) {
   1839 		bstp_input(sc, bif, m);
   1840 		bridge_release_member(sc, bif);
   1841 		return;
   1842 	}
   1843 
   1844 	/*
   1845 	 * A normal switch would discard the packet here, but that's not what
   1846 	 * we've done historically. This also prevents some obnoxious behaviour.
   1847 	 */
   1848 	if (bstp_state_before_learning(bif)) {
   1849 		bridge_release_member(sc, bif);
   1850 		ether_input(ifp, m);
   1851 		return;
   1852 	}
   1853 
   1854 	bridge_release_member(sc, bif);
   1855 
   1856 	bridge_forward(sc, m);
   1857 }
   1858 
   1859 /*
   1860  * bridge_broadcast:
   1861  *
   1862  *	Send a frame to all interfaces that are members of
   1863  *	the bridge, except for the one on which the packet
   1864  *	arrived.
   1865  */
   1866 static void
   1867 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
   1868     struct mbuf *m)
   1869 {
   1870 	struct bridge_iflist *bif;
   1871 	struct mbuf *mc;
   1872 	struct ifnet *dst_if;
   1873 	bool bmcast;
   1874 	int s;
   1875 
   1876 	bmcast = m->m_flags & (M_BCAST|M_MCAST);
   1877 
   1878 	BRIDGE_PSZ_RENTER(s);
   1879 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
   1880 		bif = bridge_try_hold_bif(bif);
   1881 		if (bif == NULL)
   1882 			continue;
   1883 		BRIDGE_PSZ_REXIT(s);
   1884 
   1885 		dst_if = bif->bif_ifp;
   1886 
   1887 		if (bif->bif_flags & IFBIF_STP) {
   1888 			switch (bif->bif_state) {
   1889 			case BSTP_IFSTATE_BLOCKING:
   1890 			case BSTP_IFSTATE_DISABLED:
   1891 				goto next;
   1892 			}
   1893 		}
   1894 
   1895 		if ((bif->bif_flags & IFBIF_DISCOVER) == 0 && !bmcast)
   1896 			goto next;
   1897 
   1898 		if ((dst_if->if_flags & IFF_RUNNING) == 0)
   1899 			goto next;
   1900 
   1901 		if (dst_if != src_if) {
   1902 			mc = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
   1903 			if (mc == NULL) {
   1904 				sc->sc_if.if_oerrors++;
   1905 				goto next;
   1906 			}
   1907 			bridge_enqueue(sc, dst_if, mc, 1);
   1908 		}
   1909 
   1910 		if (bmcast) {
   1911 			mc = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
   1912 			if (mc == NULL) {
   1913 				sc->sc_if.if_oerrors++;
   1914 				goto next;
   1915 			}
   1916 
   1917 			mc->m_pkthdr.rcvif = dst_if;
   1918 			mc->m_flags &= ~M_PROMISC;
   1919 			ether_input(dst_if, mc);
   1920 		}
   1921 next:
   1922 		bridge_release_member(sc, bif);
   1923 		BRIDGE_PSZ_RENTER(s);
   1924 	}
   1925 	BRIDGE_PSZ_REXIT(s);
   1926 
   1927 	m_freem(m);
   1928 }
   1929 
   1930 static int
   1931 bridge_rtalloc(struct bridge_softc *sc, const uint8_t *dst,
   1932     struct bridge_rtnode **brtp)
   1933 {
   1934 	struct bridge_rtnode *brt;
   1935 	int error;
   1936 
   1937 	if (sc->sc_brtcnt >= sc->sc_brtmax)
   1938 		return ENOSPC;
   1939 
   1940 	/*
   1941 	 * Allocate a new bridge forwarding node, and
   1942 	 * initialize the expiration time and Ethernet
   1943 	 * address.
   1944 	 */
   1945 	brt = pool_get(&bridge_rtnode_pool, PR_NOWAIT);
   1946 	if (brt == NULL)
   1947 		return ENOMEM;
   1948 
   1949 	memset(brt, 0, sizeof(*brt));
   1950 	brt->brt_expire = time_uptime + sc->sc_brttimeout;
   1951 	brt->brt_flags = IFBAF_DYNAMIC;
   1952 	memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
   1953 
   1954 	BRIDGE_RT_LOCK(sc);
   1955 	error = bridge_rtnode_insert(sc, brt);
   1956 	BRIDGE_RT_UNLOCK(sc);
   1957 
   1958 	if (error != 0) {
   1959 		pool_put(&bridge_rtnode_pool, brt);
   1960 		return error;
   1961 	}
   1962 
   1963 	*brtp = brt;
   1964 	return 0;
   1965 }
   1966 
   1967 /*
   1968  * bridge_rtupdate:
   1969  *
   1970  *	Add a bridge routing entry.
   1971  */
   1972 static int
   1973 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
   1974     struct ifnet *dst_if, int setflags, uint8_t flags)
   1975 {
   1976 	struct bridge_rtnode *brt;
   1977 	int s;
   1978 
   1979 again:
   1980 	/*
   1981 	 * A route for this destination might already exist.  If so,
   1982 	 * update it, otherwise create a new one.
   1983 	 */
   1984 	BRIDGE_RT_RENTER(s);
   1985 	brt = bridge_rtnode_lookup(sc, dst);
   1986 
   1987 	if (brt != NULL) {
   1988 		brt->brt_ifp = dst_if;
   1989 		if (setflags) {
   1990 			brt->brt_flags = flags;
   1991 			if (flags & IFBAF_STATIC)
   1992 				brt->brt_expire = 0;
   1993 			else
   1994 				brt->brt_expire = time_uptime + sc->sc_brttimeout;
   1995 		} else {
   1996 			if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
   1997 				brt->brt_expire = time_uptime + sc->sc_brttimeout;
   1998 		}
   1999 	}
   2000 	BRIDGE_RT_REXIT(s);
   2001 
   2002 	if (brt == NULL) {
   2003 		int r;
   2004 
   2005 		r = bridge_rtalloc(sc, dst, &brt);
   2006 		if (r != 0)
   2007 			return r;
   2008 		goto again;
   2009 	}
   2010 
   2011 	return 0;
   2012 }
   2013 
   2014 /*
   2015  * bridge_rtlookup:
   2016  *
   2017  *	Lookup the destination interface for an address.
   2018  */
   2019 static struct ifnet *
   2020 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr)
   2021 {
   2022 	struct bridge_rtnode *brt;
   2023 	struct ifnet *ifs = NULL;
   2024 	int s;
   2025 
   2026 	BRIDGE_RT_RENTER(s);
   2027 	brt = bridge_rtnode_lookup(sc, addr);
   2028 	if (brt != NULL)
   2029 		ifs = brt->brt_ifp;
   2030 	BRIDGE_RT_REXIT(s);
   2031 
   2032 	return ifs;
   2033 }
   2034 
   2035 typedef bool (*bridge_iterate_cb_t)
   2036     (struct bridge_softc *, struct bridge_rtnode *, bool *, void *);
   2037 
   2038 /*
   2039  * bridge_rtlist_iterate_remove:
   2040  *
   2041  *	It iterates on sc->sc_rtlist and removes rtnodes of it which func
   2042  *	callback judges to remove. Removals of rtnodes are done in a manner
   2043  *	of pserialize. To this end, all kmem_* operations are placed out of
   2044  *	mutexes.
   2045  */
   2046 static void
   2047 bridge_rtlist_iterate_remove(struct bridge_softc *sc, bridge_iterate_cb_t func, void *arg)
   2048 {
   2049 	struct bridge_rtnode *brt, *nbrt;
   2050 	struct bridge_rtnode **brt_list;
   2051 	int i, count;
   2052 
   2053 retry:
   2054 	count = sc->sc_brtcnt;
   2055 	if (count == 0)
   2056 		return;
   2057 	brt_list = kmem_alloc(sizeof(*brt_list) * count, KM_SLEEP);
   2058 
   2059 	BRIDGE_RT_LOCK(sc);
   2060 	if (__predict_false(sc->sc_brtcnt > count)) {
   2061 		/* The rtnodes increased, we need more memory */
   2062 		BRIDGE_RT_UNLOCK(sc);
   2063 		kmem_free(brt_list, sizeof(*brt_list) * count);
   2064 		goto retry;
   2065 	}
   2066 
   2067 	i = 0;
   2068 	LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
   2069 		bool need_break = false;
   2070 		if (func(sc, brt, &need_break, arg)) {
   2071 			bridge_rtnode_remove(sc, brt);
   2072 			brt_list[i++] = brt;
   2073 		}
   2074 		if (need_break)
   2075 			break;
   2076 	}
   2077 
   2078 	if (i > 0)
   2079 		BRIDGE_RT_PSZ_PERFORM(sc);
   2080 	BRIDGE_RT_UNLOCK(sc);
   2081 
   2082 	while (--i >= 0)
   2083 		bridge_rtnode_destroy(brt_list[i]);
   2084 
   2085 	kmem_free(brt_list, sizeof(*brt_list) * count);
   2086 }
   2087 
   2088 static bool
   2089 bridge_rttrim0_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
   2090     bool *need_break, void *arg)
   2091 {
   2092 	if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
   2093 		/* Take into account of the subsequent removal */
   2094 		if ((sc->sc_brtcnt - 1) <= sc->sc_brtmax)
   2095 			*need_break = true;
   2096 		return true;
   2097 	} else
   2098 		return false;
   2099 }
   2100 
   2101 static void
   2102 bridge_rttrim0(struct bridge_softc *sc)
   2103 {
   2104 	bridge_rtlist_iterate_remove(sc, bridge_rttrim0_cb, NULL);
   2105 }
   2106 
   2107 /*
   2108  * bridge_rttrim:
   2109  *
   2110  *	Trim the routine table so that we have a number
   2111  *	of routing entries less than or equal to the
   2112  *	maximum number.
   2113  */
   2114 static void
   2115 bridge_rttrim(struct bridge_softc *sc)
   2116 {
   2117 
   2118 	/* Make sure we actually need to do this. */
   2119 	if (sc->sc_brtcnt <= sc->sc_brtmax)
   2120 		return;
   2121 
   2122 	/* Force an aging cycle; this might trim enough addresses. */
   2123 	bridge_rtage(sc);
   2124 	if (sc->sc_brtcnt <= sc->sc_brtmax)
   2125 		return;
   2126 
   2127 	bridge_rttrim0(sc);
   2128 
   2129 	return;
   2130 }
   2131 
   2132 /*
   2133  * bridge_timer:
   2134  *
   2135  *	Aging timer for the bridge.
   2136  */
   2137 static void
   2138 bridge_timer(void *arg)
   2139 {
   2140 	struct bridge_softc *sc = arg;
   2141 
   2142 	workqueue_enqueue(sc->sc_rtage_wq, &sc->sc_rtage_wk, NULL);
   2143 }
   2144 
   2145 static void
   2146 bridge_rtage_work(struct work *wk, void *arg)
   2147 {
   2148 	struct bridge_softc *sc = arg;
   2149 
   2150 	KASSERT(wk == &sc->sc_rtage_wk);
   2151 
   2152 	bridge_rtage(sc);
   2153 
   2154 	if (sc->sc_if.if_flags & IFF_RUNNING)
   2155 		callout_reset(&sc->sc_brcallout,
   2156 		    bridge_rtable_prune_period * hz, bridge_timer, sc);
   2157 }
   2158 
   2159 static bool
   2160 bridge_rtage_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
   2161     bool *need_break, void *arg)
   2162 {
   2163 	if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
   2164 	    time_uptime >= brt->brt_expire)
   2165 		return true;
   2166 	else
   2167 		return false;
   2168 }
   2169 
   2170 /*
   2171  * bridge_rtage:
   2172  *
   2173  *	Perform an aging cycle.
   2174  */
   2175 static void
   2176 bridge_rtage(struct bridge_softc *sc)
   2177 {
   2178 	bridge_rtlist_iterate_remove(sc, bridge_rtage_cb, NULL);
   2179 }
   2180 
   2181 
   2182 static bool
   2183 bridge_rtflush_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
   2184     bool *need_break, void *arg)
   2185 {
   2186 	int full = *(int*)arg;
   2187 
   2188 	if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
   2189 		return true;
   2190 	else
   2191 		return false;
   2192 }
   2193 
   2194 /*
   2195  * bridge_rtflush:
   2196  *
   2197  *	Remove all dynamic addresses from the bridge.
   2198  */
   2199 static void
   2200 bridge_rtflush(struct bridge_softc *sc, int full)
   2201 {
   2202 	bridge_rtlist_iterate_remove(sc, bridge_rtflush_cb, &full);
   2203 }
   2204 
   2205 /*
   2206  * bridge_rtdaddr:
   2207  *
   2208  *	Remove an address from the table.
   2209  */
   2210 static int
   2211 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr)
   2212 {
   2213 	struct bridge_rtnode *brt;
   2214 
   2215 	BRIDGE_RT_LOCK(sc);
   2216 	if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL) {
   2217 		BRIDGE_RT_UNLOCK(sc);
   2218 		return ENOENT;
   2219 	}
   2220 	bridge_rtnode_remove(sc, brt);
   2221 	BRIDGE_RT_PSZ_PERFORM(sc);
   2222 	BRIDGE_RT_UNLOCK(sc);
   2223 
   2224 	bridge_rtnode_destroy(brt);
   2225 
   2226 	return 0;
   2227 }
   2228 
   2229 /*
   2230  * bridge_rtdelete:
   2231  *
   2232  *	Delete routes to a speicifc member interface.
   2233  */
   2234 static void
   2235 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp)
   2236 {
   2237 	struct bridge_rtnode *brt, *nbrt;
   2238 
   2239 	BRIDGE_RT_LOCK(sc);
   2240 	LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
   2241 		if (brt->brt_ifp == ifp)
   2242 			break;
   2243 	}
   2244 	if (brt == NULL) {
   2245 		BRIDGE_RT_UNLOCK(sc);
   2246 		return;
   2247 	}
   2248 	bridge_rtnode_remove(sc, brt);
   2249 	BRIDGE_RT_PSZ_PERFORM(sc);
   2250 	BRIDGE_RT_UNLOCK(sc);
   2251 
   2252 	bridge_rtnode_destroy(brt);
   2253 }
   2254 
   2255 /*
   2256  * bridge_rtable_init:
   2257  *
   2258  *	Initialize the route table for this bridge.
   2259  */
   2260 static void
   2261 bridge_rtable_init(struct bridge_softc *sc)
   2262 {
   2263 	int i;
   2264 
   2265 	sc->sc_rthash = kmem_alloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
   2266 	    KM_SLEEP);
   2267 
   2268 	for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
   2269 		LIST_INIT(&sc->sc_rthash[i]);
   2270 
   2271 	sc->sc_rthash_key = cprng_fast32();
   2272 
   2273 	LIST_INIT(&sc->sc_rtlist);
   2274 
   2275 #ifdef BRIDGE_MPSAFE
   2276 	sc->sc_rtlist_psz = pserialize_create();
   2277 	sc->sc_rtlist_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_SOFTNET);
   2278 #else
   2279 	sc->sc_rtlist_psz = NULL;
   2280 	sc->sc_rtlist_lock = NULL;
   2281 #endif
   2282 }
   2283 
   2284 /*
   2285  * bridge_rtable_fini:
   2286  *
   2287  *	Deconstruct the route table for this bridge.
   2288  */
   2289 static void
   2290 bridge_rtable_fini(struct bridge_softc *sc)
   2291 {
   2292 
   2293 	kmem_free(sc->sc_rthash, sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE);
   2294 	if (sc->sc_rtlist_lock)
   2295 		mutex_obj_free(sc->sc_rtlist_lock);
   2296 	if (sc->sc_rtlist_psz)
   2297 		pserialize_destroy(sc->sc_rtlist_psz);
   2298 }
   2299 
   2300 /*
   2301  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
   2302  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
   2303  */
   2304 #define	mix(a, b, c)							\
   2305 do {									\
   2306 	a -= b; a -= c; a ^= (c >> 13);					\
   2307 	b -= c; b -= a; b ^= (a << 8);					\
   2308 	c -= a; c -= b; c ^= (b >> 13);					\
   2309 	a -= b; a -= c; a ^= (c >> 12);					\
   2310 	b -= c; b -= a; b ^= (a << 16);					\
   2311 	c -= a; c -= b; c ^= (b >> 5);					\
   2312 	a -= b; a -= c; a ^= (c >> 3);					\
   2313 	b -= c; b -= a; b ^= (a << 10);					\
   2314 	c -= a; c -= b; c ^= (b >> 15);					\
   2315 } while (/*CONSTCOND*/0)
   2316 
   2317 static inline uint32_t
   2318 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
   2319 {
   2320 	uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
   2321 
   2322 	b += addr[5] << 8;
   2323 	b += addr[4];
   2324 	a += addr[3] << 24;
   2325 	a += addr[2] << 16;
   2326 	a += addr[1] << 8;
   2327 	a += addr[0];
   2328 
   2329 	mix(a, b, c);
   2330 
   2331 	return (c & BRIDGE_RTHASH_MASK);
   2332 }
   2333 
   2334 #undef mix
   2335 
   2336 /*
   2337  * bridge_rtnode_lookup:
   2338  *
   2339  *	Look up a bridge route node for the specified destination.
   2340  */
   2341 static struct bridge_rtnode *
   2342 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr)
   2343 {
   2344 	struct bridge_rtnode *brt;
   2345 	uint32_t hash;
   2346 	int dir;
   2347 
   2348 	hash = bridge_rthash(sc, addr);
   2349 	LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
   2350 		dir = memcmp(addr, brt->brt_addr, ETHER_ADDR_LEN);
   2351 		if (dir == 0)
   2352 			return (brt);
   2353 		if (dir > 0)
   2354 			return (NULL);
   2355 	}
   2356 
   2357 	return (NULL);
   2358 }
   2359 
   2360 /*
   2361  * bridge_rtnode_insert:
   2362  *
   2363  *	Insert the specified bridge node into the route table.  We
   2364  *	assume the entry is not already in the table.
   2365  */
   2366 static int
   2367 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
   2368 {
   2369 	struct bridge_rtnode *lbrt;
   2370 	uint32_t hash;
   2371 	int dir;
   2372 
   2373 	KASSERT(BRIDGE_RT_LOCKED(sc));
   2374 
   2375 	hash = bridge_rthash(sc, brt->brt_addr);
   2376 
   2377 	lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
   2378 	if (lbrt == NULL) {
   2379 		LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
   2380 		goto out;
   2381 	}
   2382 
   2383 	do {
   2384 		dir = memcmp(brt->brt_addr, lbrt->brt_addr, ETHER_ADDR_LEN);
   2385 		if (dir == 0)
   2386 			return (EEXIST);
   2387 		if (dir > 0) {
   2388 			LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
   2389 			goto out;
   2390 		}
   2391 		if (LIST_NEXT(lbrt, brt_hash) == NULL) {
   2392 			LIST_INSERT_AFTER(lbrt, brt, brt_hash);
   2393 			goto out;
   2394 		}
   2395 		lbrt = LIST_NEXT(lbrt, brt_hash);
   2396 	} while (lbrt != NULL);
   2397 
   2398 #ifdef DIAGNOSTIC
   2399 	panic("bridge_rtnode_insert: impossible");
   2400 #endif
   2401 
   2402  out:
   2403 	LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
   2404 	sc->sc_brtcnt++;
   2405 
   2406 	return (0);
   2407 }
   2408 
   2409 /*
   2410  * bridge_rtnode_remove:
   2411  *
   2412  *	Remove a bridge rtnode from the rthash and the rtlist of a bridge.
   2413  */
   2414 static void
   2415 bridge_rtnode_remove(struct bridge_softc *sc, struct bridge_rtnode *brt)
   2416 {
   2417 
   2418 	KASSERT(BRIDGE_RT_LOCKED(sc));
   2419 
   2420 	LIST_REMOVE(brt, brt_hash);
   2421 	LIST_REMOVE(brt, brt_list);
   2422 	sc->sc_brtcnt--;
   2423 }
   2424 
   2425 /*
   2426  * bridge_rtnode_destroy:
   2427  *
   2428  *	Destroy a bridge rtnode.
   2429  */
   2430 static void
   2431 bridge_rtnode_destroy(struct bridge_rtnode *brt)
   2432 {
   2433 
   2434 	pool_put(&bridge_rtnode_pool, brt);
   2435 }
   2436 
   2437 #if defined(BRIDGE_IPF)
   2438 extern pfil_head_t *inet_pfil_hook;                 /* XXX */
   2439 extern pfil_head_t *inet6_pfil_hook;                /* XXX */
   2440 
   2441 /*
   2442  * Send bridge packets through IPF if they are one of the types IPF can deal
   2443  * with, or if they are ARP or REVARP.  (IPF will pass ARP and REVARP without
   2444  * question.)
   2445  */
   2446 static int
   2447 bridge_ipf(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
   2448 {
   2449 	int snap, error;
   2450 	struct ether_header *eh1, eh2;
   2451 	struct llc llc1;
   2452 	uint16_t ether_type;
   2453 
   2454 	snap = 0;
   2455 	error = -1;	/* Default error if not error == 0 */
   2456 	eh1 = mtod(*mp, struct ether_header *);
   2457 	ether_type = ntohs(eh1->ether_type);
   2458 
   2459 	/*
   2460 	 * Check for SNAP/LLC.
   2461 	 */
   2462         if (ether_type < ETHERMTU) {
   2463                 struct llc *llc2 = (struct llc *)(eh1 + 1);
   2464 
   2465                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
   2466                     llc2->llc_dsap == LLC_SNAP_LSAP &&
   2467                     llc2->llc_ssap == LLC_SNAP_LSAP &&
   2468                     llc2->llc_control == LLC_UI) {
   2469                 	ether_type = htons(llc2->llc_un.type_snap.ether_type);
   2470 			snap = 1;
   2471                 }
   2472         }
   2473 
   2474 	/*
   2475 	 * If we're trying to filter bridge traffic, don't look at anything
   2476 	 * other than IP and ARP traffic.  If the filter doesn't understand
   2477 	 * IPv6, don't allow IPv6 through the bridge either.  This is lame
   2478 	 * since if we really wanted, say, an AppleTalk filter, we are hosed,
   2479 	 * but of course we don't have an AppleTalk filter to begin with.
   2480 	 * (Note that since IPF doesn't understand ARP it will pass *ALL*
   2481 	 * ARP traffic.)
   2482 	 */
   2483 	switch (ether_type) {
   2484 		case ETHERTYPE_ARP:
   2485 		case ETHERTYPE_REVARP:
   2486 			return 0; /* Automatically pass */
   2487 		case ETHERTYPE_IP:
   2488 # ifdef INET6
   2489 		case ETHERTYPE_IPV6:
   2490 # endif /* INET6 */
   2491 			break;
   2492 		default:
   2493 			goto bad;
   2494 	}
   2495 
   2496 	/* Strip off the Ethernet header and keep a copy. */
   2497 	m_copydata(*mp, 0, ETHER_HDR_LEN, (void *) &eh2);
   2498 	m_adj(*mp, ETHER_HDR_LEN);
   2499 
   2500 	/* Strip off snap header, if present */
   2501 	if (snap) {
   2502 		m_copydata(*mp, 0, sizeof(struct llc), (void *) &llc1);
   2503 		m_adj(*mp, sizeof(struct llc));
   2504 	}
   2505 
   2506 	/*
   2507 	 * Check basic packet sanity and run IPF through pfil.
   2508 	 */
   2509 	KASSERT(!cpu_intr_p());
   2510 	switch (ether_type)
   2511 	{
   2512 	case ETHERTYPE_IP :
   2513 		error = (dir == PFIL_IN) ? bridge_ip_checkbasic(mp) : 0;
   2514 		if (error == 0)
   2515 			error = pfil_run_hooks(inet_pfil_hook, mp, ifp, dir);
   2516 		break;
   2517 # ifdef INET6
   2518 	case ETHERTYPE_IPV6 :
   2519 		error = (dir == PFIL_IN) ? bridge_ip6_checkbasic(mp) : 0;
   2520 		if (error == 0)
   2521 			error = pfil_run_hooks(inet6_pfil_hook, mp, ifp, dir);
   2522 		break;
   2523 # endif
   2524 	default :
   2525 		error = 0;
   2526 		break;
   2527 	}
   2528 
   2529 	if (*mp == NULL)
   2530 		return error;
   2531 	if (error != 0)
   2532 		goto bad;
   2533 
   2534 	error = -1;
   2535 
   2536 	/*
   2537 	 * Finally, put everything back the way it was and return
   2538 	 */
   2539 	if (snap) {
   2540 		M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
   2541 		if (*mp == NULL)
   2542 			return error;
   2543 		bcopy(&llc1, mtod(*mp, void *), sizeof(struct llc));
   2544 	}
   2545 
   2546 	M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
   2547 	if (*mp == NULL)
   2548 		return error;
   2549 	bcopy(&eh2, mtod(*mp, void *), ETHER_HDR_LEN);
   2550 
   2551 	return 0;
   2552 
   2553     bad:
   2554 	m_freem(*mp);
   2555 	*mp = NULL;
   2556 	return error;
   2557 }
   2558 
   2559 /*
   2560  * Perform basic checks on header size since
   2561  * IPF assumes ip_input has already processed
   2562  * it for it.  Cut-and-pasted from ip_input.c.
   2563  * Given how simple the IPv6 version is,
   2564  * does the IPv4 version really need to be
   2565  * this complicated?
   2566  *
   2567  * XXX Should we update ipstat here, or not?
   2568  * XXX Right now we update ipstat but not
   2569  * XXX csum_counter.
   2570  */
   2571 static int
   2572 bridge_ip_checkbasic(struct mbuf **mp)
   2573 {
   2574 	struct mbuf *m = *mp;
   2575 	struct ip *ip;
   2576 	int len, hlen;
   2577 
   2578 	if (*mp == NULL)
   2579 		return -1;
   2580 
   2581 	if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
   2582 		if ((m = m_copyup(m, sizeof(struct ip),
   2583 			(max_linkhdr + 3) & ~3)) == NULL) {
   2584 			/* XXXJRT new stat, please */
   2585 			ip_statinc(IP_STAT_TOOSMALL);
   2586 			goto bad;
   2587 		}
   2588 	} else if (__predict_false(m->m_len < sizeof (struct ip))) {
   2589 		if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
   2590 			ip_statinc(IP_STAT_TOOSMALL);
   2591 			goto bad;
   2592 		}
   2593 	}
   2594 	ip = mtod(m, struct ip *);
   2595 	if (ip == NULL) goto bad;
   2596 
   2597 	if (ip->ip_v != IPVERSION) {
   2598 		ip_statinc(IP_STAT_BADVERS);
   2599 		goto bad;
   2600 	}
   2601 	hlen = ip->ip_hl << 2;
   2602 	if (hlen < sizeof(struct ip)) { /* minimum header length */
   2603 		ip_statinc(IP_STAT_BADHLEN);
   2604 		goto bad;
   2605 	}
   2606 	if (hlen > m->m_len) {
   2607 		if ((m = m_pullup(m, hlen)) == 0) {
   2608 			ip_statinc(IP_STAT_BADHLEN);
   2609 			goto bad;
   2610 		}
   2611 		ip = mtod(m, struct ip *);
   2612 		if (ip == NULL) goto bad;
   2613 	}
   2614 
   2615         switch (m->m_pkthdr.csum_flags &
   2616                 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) |
   2617                  M_CSUM_IPv4_BAD)) {
   2618         case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
   2619                 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); */
   2620                 goto bad;
   2621 
   2622         case M_CSUM_IPv4:
   2623                 /* Checksum was okay. */
   2624                 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); */
   2625                 break;
   2626 
   2627         default:
   2628                 /* Must compute it ourselves. */
   2629                 /* INET_CSUM_COUNTER_INCR(&ip_swcsum); */
   2630                 if (in_cksum(m, hlen) != 0)
   2631                         goto bad;
   2632                 break;
   2633         }
   2634 
   2635         /* Retrieve the packet length. */
   2636         len = ntohs(ip->ip_len);
   2637 
   2638         /*
   2639          * Check for additional length bogosity
   2640          */
   2641         if (len < hlen) {
   2642 		ip_statinc(IP_STAT_BADLEN);
   2643                 goto bad;
   2644         }
   2645 
   2646         /*
   2647          * Check that the amount of data in the buffers
   2648          * is as at least much as the IP header would have us expect.
   2649          * Drop packet if shorter than we expect.
   2650          */
   2651         if (m->m_pkthdr.len < len) {
   2652 		ip_statinc(IP_STAT_TOOSHORT);
   2653                 goto bad;
   2654         }
   2655 
   2656 	/* Checks out, proceed */
   2657 	*mp = m;
   2658 	return 0;
   2659 
   2660     bad:
   2661 	*mp = m;
   2662 	return -1;
   2663 }
   2664 
   2665 # ifdef INET6
   2666 /*
   2667  * Same as above, but for IPv6.
   2668  * Cut-and-pasted from ip6_input.c.
   2669  * XXX Should we update ip6stat, or not?
   2670  */
   2671 static int
   2672 bridge_ip6_checkbasic(struct mbuf **mp)
   2673 {
   2674 	struct mbuf *m = *mp;
   2675 	struct ip6_hdr *ip6;
   2676 
   2677         /*
   2678          * If the IPv6 header is not aligned, slurp it up into a new
   2679          * mbuf with space for link headers, in the event we forward
   2680          * it.  Otherwise, if it is aligned, make sure the entire base
   2681          * IPv6 header is in the first mbuf of the chain.
   2682          */
   2683         if (IP6_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
   2684                 struct ifnet *inifp = m->m_pkthdr.rcvif;
   2685                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
   2686                                   (max_linkhdr + 3) & ~3)) == NULL) {
   2687                         /* XXXJRT new stat, please */
   2688 			ip6_statinc(IP6_STAT_TOOSMALL);
   2689                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
   2690                         goto bad;
   2691                 }
   2692         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
   2693                 struct ifnet *inifp = m->m_pkthdr.rcvif;
   2694                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
   2695 			ip6_statinc(IP6_STAT_TOOSMALL);
   2696                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
   2697                         goto bad;
   2698                 }
   2699         }
   2700 
   2701         ip6 = mtod(m, struct ip6_hdr *);
   2702 
   2703         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
   2704 		ip6_statinc(IP6_STAT_BADVERS);
   2705                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
   2706                 goto bad;
   2707         }
   2708 
   2709 	/* Checks out, proceed */
   2710 	*mp = m;
   2711 	return 0;
   2712 
   2713     bad:
   2714 	*mp = m;
   2715 	return -1;
   2716 }
   2717 # endif /* INET6 */
   2718 #endif /* BRIDGE_IPF */
   2719