Home | History | Annotate | Line # | Download | only in net
if_bridge.c revision 1.146
      1 /*	$NetBSD: if_bridge.c,v 1.146 2017/12/19 03:32:35 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 1.146 2017/12/19 03:32:35 ozaki-r Exp $");
     84 
     85 #ifdef _KERNEL_OPT
     86 #include "opt_bridge_ipf.h"
     87 #include "opt_inet.h"
     88 #include "opt_net_mpsafe.h"
     89 #endif /* _KERNEL_OPT */
     90 
     91 #include <sys/param.h>
     92 #include <sys/kernel.h>
     93 #include <sys/mbuf.h>
     94 #include <sys/queue.h>
     95 #include <sys/socket.h>
     96 #include <sys/socketvar.h> /* for softnet_lock */
     97 #include <sys/sockio.h>
     98 #include <sys/systm.h>
     99 #include <sys/proc.h>
    100 #include <sys/pool.h>
    101 #include <sys/kauth.h>
    102 #include <sys/cpu.h>
    103 #include <sys/cprng.h>
    104 #include <sys/mutex.h>
    105 #include <sys/kmem.h>
    106 
    107 #include <net/bpf.h>
    108 #include <net/if.h>
    109 #include <net/if_dl.h>
    110 #include <net/if_types.h>
    111 #include <net/if_llc.h>
    112 
    113 #include <net/if_ether.h>
    114 #include <net/if_bridgevar.h>
    115 
    116 #if defined(BRIDGE_IPF)
    117 /* Used for bridge_ip[6]_checkbasic */
    118 #include <netinet/in.h>
    119 #include <netinet/in_systm.h>
    120 #include <netinet/ip.h>
    121 #include <netinet/ip_var.h>
    122 #include <netinet/ip_private.h>		/* XXX */
    123 
    124 #include <netinet/ip6.h>
    125 #include <netinet6/in6_var.h>
    126 #include <netinet6/ip6_var.h>
    127 #include <netinet6/ip6_private.h>	/* XXX */
    128 #endif /* BRIDGE_IPF */
    129 
    130 /*
    131  * Size of the route hash table.  Must be a power of two.
    132  */
    133 #ifndef BRIDGE_RTHASH_SIZE
    134 #define	BRIDGE_RTHASH_SIZE		1024
    135 #endif
    136 
    137 #define	BRIDGE_RTHASH_MASK		(BRIDGE_RTHASH_SIZE - 1)
    138 
    139 #include "carp.h"
    140 #if NCARP > 0
    141 #include <netinet/in.h>
    142 #include <netinet/in_var.h>
    143 #include <netinet/ip_carp.h>
    144 #endif
    145 
    146 #include "ioconf.h"
    147 
    148 __CTASSERT(sizeof(struct ifbifconf) == sizeof(struct ifbaconf));
    149 __CTASSERT(offsetof(struct ifbifconf, ifbic_len) == offsetof(struct ifbaconf, ifbac_len));
    150 __CTASSERT(offsetof(struct ifbifconf, ifbic_buf) == offsetof(struct ifbaconf, ifbac_buf));
    151 
    152 /*
    153  * Maximum number of addresses to cache.
    154  */
    155 #ifndef BRIDGE_RTABLE_MAX
    156 #define	BRIDGE_RTABLE_MAX		100
    157 #endif
    158 
    159 /*
    160  * Spanning tree defaults.
    161  */
    162 #define	BSTP_DEFAULT_MAX_AGE		(20 * 256)
    163 #define	BSTP_DEFAULT_HELLO_TIME		(2 * 256)
    164 #define	BSTP_DEFAULT_FORWARD_DELAY	(15 * 256)
    165 #define	BSTP_DEFAULT_HOLD_TIME		(1 * 256)
    166 #define	BSTP_DEFAULT_BRIDGE_PRIORITY	0x8000
    167 #define	BSTP_DEFAULT_PORT_PRIORITY	0x80
    168 #define	BSTP_DEFAULT_PATH_COST		55
    169 
    170 /*
    171  * Timeout (in seconds) for entries learned dynamically.
    172  */
    173 #ifndef BRIDGE_RTABLE_TIMEOUT
    174 #define	BRIDGE_RTABLE_TIMEOUT		(20 * 60)	/* same as ARP */
    175 #endif
    176 
    177 /*
    178  * Number of seconds between walks of the route list.
    179  */
    180 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
    181 #define	BRIDGE_RTABLE_PRUNE_PERIOD	(5 * 60)
    182 #endif
    183 
    184 #define BRIDGE_RT_LOCK(_sc)	if ((_sc)->sc_rtlist_lock) \
    185 					mutex_enter((_sc)->sc_rtlist_lock)
    186 #define BRIDGE_RT_UNLOCK(_sc)	if ((_sc)->sc_rtlist_lock) \
    187 					mutex_exit((_sc)->sc_rtlist_lock)
    188 #define BRIDGE_RT_LOCKED(_sc)	(!(_sc)->sc_rtlist_lock || \
    189 				 mutex_owned((_sc)->sc_rtlist_lock))
    190 
    191 #define BRIDGE_RT_PSZ_PERFORM(_sc) \
    192 				if ((_sc)->sc_rtlist_psz != NULL) \
    193 					pserialize_perform((_sc)->sc_rtlist_psz);
    194 
    195 #define BRIDGE_RT_RENTER(__s)	do { __s = pserialize_read_enter(); } while (0)
    196 #define BRIDGE_RT_REXIT(__s)	do { pserialize_read_exit(__s); } while (0)
    197 
    198 
    199 #ifdef NET_MPSAFE
    200 #define DECLARE_LOCK_VARIABLE
    201 #define ACQUIRE_GLOBAL_LOCKS()	do { } while (0)
    202 #define RELEASE_GLOBAL_LOCKS()	do { } while (0)
    203 #else
    204 #define DECLARE_LOCK_VARIABLE	int __s
    205 #define ACQUIRE_GLOBAL_LOCKS()	do {					\
    206 					KERNEL_LOCK(1, NULL);		\
    207 					mutex_enter(softnet_lock);	\
    208 					__s = splsoftnet();		\
    209 				} while (0)
    210 #define RELEASE_GLOBAL_LOCKS()	do {					\
    211 					splx(__s);			\
    212 					mutex_exit(softnet_lock);	\
    213 					KERNEL_UNLOCK_ONE(NULL);	\
    214 				} while (0)
    215 #endif
    216 
    217 struct psref_class *bridge_psref_class __read_mostly;
    218 
    219 int	bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
    220 
    221 static struct pool bridge_rtnode_pool;
    222 
    223 static int	bridge_clone_create(struct if_clone *, int);
    224 static int	bridge_clone_destroy(struct ifnet *);
    225 
    226 static int	bridge_ioctl(struct ifnet *, u_long, void *);
    227 static int	bridge_init(struct ifnet *);
    228 static void	bridge_stop(struct ifnet *, int);
    229 static void	bridge_start(struct ifnet *);
    230 
    231 static void	bridge_input(struct ifnet *, struct mbuf *);
    232 static void	bridge_forward(struct bridge_softc *, struct mbuf *);
    233 
    234 static void	bridge_timer(void *);
    235 
    236 static void	bridge_broadcast(struct bridge_softc *, struct ifnet *,
    237 				 struct mbuf *);
    238 
    239 static int	bridge_rtupdate(struct bridge_softc *, const uint8_t *,
    240 				struct ifnet *, int, uint8_t);
    241 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *);
    242 static void	bridge_rttrim(struct bridge_softc *);
    243 static void	bridge_rtage(struct bridge_softc *);
    244 static void	bridge_rtage_work(struct work *, void *);
    245 static void	bridge_rtflush(struct bridge_softc *, int);
    246 static int	bridge_rtdaddr(struct bridge_softc *, const uint8_t *);
    247 static void	bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp);
    248 
    249 static void	bridge_rtable_init(struct bridge_softc *);
    250 static void	bridge_rtable_fini(struct bridge_softc *);
    251 
    252 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
    253 						  const uint8_t *);
    254 static int	bridge_rtnode_insert(struct bridge_softc *,
    255 				     struct bridge_rtnode *);
    256 static void	bridge_rtnode_remove(struct bridge_softc *,
    257 				     struct bridge_rtnode *);
    258 static void	bridge_rtnode_destroy(struct bridge_rtnode *);
    259 
    260 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
    261 						  const char *name,
    262 						  struct psref *);
    263 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
    264 						     struct ifnet *ifp,
    265 						     struct psref *);
    266 static void	bridge_release_member(struct bridge_softc *, struct bridge_iflist *,
    267                                       struct psref *);
    268 static void	bridge_delete_member(struct bridge_softc *,
    269 				     struct bridge_iflist *);
    270 static void	bridge_acquire_member(struct bridge_softc *sc,
    271                                       struct bridge_iflist *,
    272                                       struct psref *);
    273 
    274 static int	bridge_ioctl_add(struct bridge_softc *, void *);
    275 static int	bridge_ioctl_del(struct bridge_softc *, void *);
    276 static int	bridge_ioctl_gifflags(struct bridge_softc *, void *);
    277 static int	bridge_ioctl_sifflags(struct bridge_softc *, void *);
    278 static int	bridge_ioctl_scache(struct bridge_softc *, void *);
    279 static int	bridge_ioctl_gcache(struct bridge_softc *, void *);
    280 static int	bridge_ioctl_gifs(struct bridge_softc *, void *);
    281 static int	bridge_ioctl_rts(struct bridge_softc *, void *);
    282 static int	bridge_ioctl_saddr(struct bridge_softc *, void *);
    283 static int	bridge_ioctl_sto(struct bridge_softc *, void *);
    284 static int	bridge_ioctl_gto(struct bridge_softc *, void *);
    285 static int	bridge_ioctl_daddr(struct bridge_softc *, void *);
    286 static int	bridge_ioctl_flush(struct bridge_softc *, void *);
    287 static int	bridge_ioctl_gpri(struct bridge_softc *, void *);
    288 static int	bridge_ioctl_spri(struct bridge_softc *, void *);
    289 static int	bridge_ioctl_ght(struct bridge_softc *, void *);
    290 static int	bridge_ioctl_sht(struct bridge_softc *, void *);
    291 static int	bridge_ioctl_gfd(struct bridge_softc *, void *);
    292 static int	bridge_ioctl_sfd(struct bridge_softc *, void *);
    293 static int	bridge_ioctl_gma(struct bridge_softc *, void *);
    294 static int	bridge_ioctl_sma(struct bridge_softc *, void *);
    295 static int	bridge_ioctl_sifprio(struct bridge_softc *, void *);
    296 static int	bridge_ioctl_sifcost(struct bridge_softc *, void *);
    297 #if defined(BRIDGE_IPF)
    298 static int	bridge_ioctl_gfilt(struct bridge_softc *, void *);
    299 static int	bridge_ioctl_sfilt(struct bridge_softc *, void *);
    300 static int	bridge_ipf(void *, struct mbuf **, struct ifnet *, int);
    301 static int	bridge_ip_checkbasic(struct mbuf **mp);
    302 # ifdef INET6
    303 static int	bridge_ip6_checkbasic(struct mbuf **mp);
    304 # endif /* INET6 */
    305 #endif /* BRIDGE_IPF */
    306 
    307 struct bridge_control {
    308 	int	(*bc_func)(struct bridge_softc *, void *);
    309 	int	bc_argsize;
    310 	int	bc_flags;
    311 };
    312 
    313 #define	BC_F_COPYIN		0x01	/* copy arguments in */
    314 #define	BC_F_COPYOUT		0x02	/* copy arguments out */
    315 #define	BC_F_SUSER		0x04	/* do super-user check */
    316 #define BC_F_XLATEIN		0x08	/* xlate arguments in */
    317 #define BC_F_XLATEOUT		0x10	/* xlate arguments out */
    318 
    319 static const struct bridge_control bridge_control_table[] = {
    320 [BRDGADD] = {bridge_ioctl_add, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    321 [BRDGDEL] = {bridge_ioctl_del, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    322 
    323 [BRDGGIFFLGS] = {bridge_ioctl_gifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_COPYOUT},
    324 [BRDGSIFFLGS] = {bridge_ioctl_sifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    325 
    326 [BRDGSCACHE] = {bridge_ioctl_scache, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    327 [BRDGGCACHE] = {bridge_ioctl_gcache, sizeof(struct ifbrparam), BC_F_COPYOUT},
    328 
    329 [OBRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_COPYIN|BC_F_COPYOUT},
    330 [OBRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_COPYIN|BC_F_COPYOUT},
    331 
    332 [BRDGSADDR] = {bridge_ioctl_saddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
    333 
    334 [BRDGSTO] = {bridge_ioctl_sto, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    335 [BRDGGTO] = {bridge_ioctl_gto, sizeof(struct ifbrparam), BC_F_COPYOUT},
    336 
    337 [BRDGDADDR] = {bridge_ioctl_daddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
    338 
    339 [BRDGFLUSH] = {bridge_ioctl_flush, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    340 
    341 [BRDGGPRI] = {bridge_ioctl_gpri, sizeof(struct ifbrparam), BC_F_COPYOUT},
    342 [BRDGSPRI] = {bridge_ioctl_spri, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    343 
    344 [BRDGGHT] = {bridge_ioctl_ght, sizeof(struct ifbrparam), BC_F_COPYOUT},
    345 [BRDGSHT] = {bridge_ioctl_sht, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    346 
    347 [BRDGGFD] = {bridge_ioctl_gfd, sizeof(struct ifbrparam), BC_F_COPYOUT},
    348 [BRDGSFD] = {bridge_ioctl_sfd, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    349 
    350 [BRDGGMA] = {bridge_ioctl_gma, sizeof(struct ifbrparam), BC_F_COPYOUT},
    351 [BRDGSMA] = {bridge_ioctl_sma, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    352 
    353 [BRDGSIFPRIO] = {bridge_ioctl_sifprio, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    354 
    355 [BRDGSIFCOST] = {bridge_ioctl_sifcost, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    356 #if defined(BRIDGE_IPF)
    357 [BRDGGFILT] = {bridge_ioctl_gfilt, sizeof(struct ifbrparam), BC_F_COPYOUT},
    358 [BRDGSFILT] = {bridge_ioctl_sfilt, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    359 #endif /* BRIDGE_IPF */
    360 [BRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_XLATEIN|BC_F_XLATEOUT},
    361 [BRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_XLATEIN|BC_F_XLATEOUT},
    362 };
    363 
    364 static const int bridge_control_table_size = __arraycount(bridge_control_table);
    365 
    366 static struct if_clone bridge_cloner =
    367     IF_CLONE_INITIALIZER("bridge", bridge_clone_create, bridge_clone_destroy);
    368 
    369 /*
    370  * bridgeattach:
    371  *
    372  *	Pseudo-device attach routine.
    373  */
    374 void
    375 bridgeattach(int n)
    376 {
    377 
    378 	pool_init(&bridge_rtnode_pool, sizeof(struct bridge_rtnode),
    379 	    0, 0, 0, "brtpl", NULL, IPL_NET);
    380 
    381 	bridge_psref_class = psref_class_create("bridge", IPL_SOFTNET);
    382 
    383 	if_clone_attach(&bridge_cloner);
    384 }
    385 
    386 /*
    387  * bridge_clone_create:
    388  *
    389  *	Create a new bridge instance.
    390  */
    391 static int
    392 bridge_clone_create(struct if_clone *ifc, int unit)
    393 {
    394 	struct bridge_softc *sc;
    395 	struct ifnet *ifp;
    396 	int error;
    397 
    398 	sc = kmem_zalloc(sizeof(*sc),  KM_SLEEP);
    399 	ifp = &sc->sc_if;
    400 
    401 	sc->sc_brtmax = BRIDGE_RTABLE_MAX;
    402 	sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
    403 	sc->sc_bridge_max_age = BSTP_DEFAULT_MAX_AGE;
    404 	sc->sc_bridge_hello_time = BSTP_DEFAULT_HELLO_TIME;
    405 	sc->sc_bridge_forward_delay = BSTP_DEFAULT_FORWARD_DELAY;
    406 	sc->sc_bridge_priority = BSTP_DEFAULT_BRIDGE_PRIORITY;
    407 	sc->sc_hold_time = BSTP_DEFAULT_HOLD_TIME;
    408 	sc->sc_filter_flags = 0;
    409 
    410 	/* Initialize our routing table. */
    411 	bridge_rtable_init(sc);
    412 
    413 	error = workqueue_create(&sc->sc_rtage_wq, "bridge_rtage",
    414 	    bridge_rtage_work, sc, PRI_SOFTNET, IPL_SOFTNET, WQ_MPSAFE);
    415 	if (error)
    416 		panic("%s: workqueue_create %d\n", __func__, error);
    417 
    418 	callout_init(&sc->sc_brcallout, CALLOUT_MPSAFE);
    419 	callout_init(&sc->sc_bstpcallout, CALLOUT_MPSAFE);
    420 
    421 	mutex_init(&sc->sc_iflist_psref.bip_lock, MUTEX_DEFAULT, IPL_NONE);
    422 	PSLIST_INIT(&sc->sc_iflist_psref.bip_iflist);
    423 	sc->sc_iflist_psref.bip_psz = pserialize_create();
    424 
    425 	if_initname(ifp, ifc->ifc_name, unit);
    426 	ifp->if_softc = sc;
    427 	ifp->if_extflags = IFEF_NO_LINK_STATE_CHANGE;
    428 #ifdef NET_MPSAFE
    429 	ifp->if_extflags |= IFEF_MPSAFE;
    430 #endif
    431 	ifp->if_mtu = ETHERMTU;
    432 	ifp->if_ioctl = bridge_ioctl;
    433 	ifp->if_output = bridge_output;
    434 	ifp->if_start = bridge_start;
    435 	ifp->if_stop = bridge_stop;
    436 	ifp->if_init = bridge_init;
    437 	ifp->if_type = IFT_BRIDGE;
    438 	ifp->if_addrlen = 0;
    439 	ifp->if_dlt = DLT_EN10MB;
    440 	ifp->if_hdrlen = ETHER_HDR_LEN;
    441 
    442 	error = if_initialize(ifp);
    443 	if (error != 0) {
    444 		pserialize_destroy(sc->sc_iflist_psref.bip_psz);
    445 		mutex_destroy(&sc->sc_iflist_psref.bip_lock);
    446 		callout_destroy(&sc->sc_brcallout);
    447 		callout_destroy(&sc->sc_bstpcallout);
    448 		workqueue_destroy(sc->sc_rtage_wq);
    449 		bridge_rtable_fini(sc);
    450 		kmem_free(sc, sizeof(*sc));
    451 
    452 		return error;
    453 	}
    454 	if_register(ifp);
    455 
    456 	if_alloc_sadl(ifp);
    457 
    458 	return 0;
    459 }
    460 
    461 /*
    462  * bridge_clone_destroy:
    463  *
    464  *	Destroy a bridge instance.
    465  */
    466 static int
    467 bridge_clone_destroy(struct ifnet *ifp)
    468 {
    469 	struct bridge_softc *sc = ifp->if_softc;
    470 	struct bridge_iflist *bif;
    471 
    472 	bridge_stop(ifp, 1);
    473 
    474 	BRIDGE_LOCK(sc);
    475 	for (;;) {
    476 		bif = PSLIST_WRITER_FIRST(&sc->sc_iflist_psref.bip_iflist, struct bridge_iflist,
    477 		    bif_next);
    478 		if (bif == NULL)
    479 			break;
    480 		bridge_delete_member(sc, bif);
    481 	}
    482 	PSLIST_DESTROY(&sc->sc_iflist_psref.bip_iflist);
    483 	BRIDGE_UNLOCK(sc);
    484 
    485 	if_detach(ifp);
    486 
    487 	/* Tear down the routing table. */
    488 	bridge_rtable_fini(sc);
    489 
    490 	pserialize_destroy(sc->sc_iflist_psref.bip_psz);
    491 	mutex_destroy(&sc->sc_iflist_psref.bip_lock);
    492 	callout_destroy(&sc->sc_brcallout);
    493 	callout_destroy(&sc->sc_bstpcallout);
    494 	workqueue_destroy(sc->sc_rtage_wq);
    495 	kmem_free(sc, sizeof(*sc));
    496 
    497 	return 0;
    498 }
    499 
    500 /*
    501  * bridge_ioctl:
    502  *
    503  *	Handle a control request from the operator.
    504  */
    505 static int
    506 bridge_ioctl(struct ifnet *ifp, u_long cmd, void *data)
    507 {
    508 	struct bridge_softc *sc = ifp->if_softc;
    509 	struct lwp *l = curlwp;	/* XXX */
    510 	union {
    511 		struct ifbreq ifbreq;
    512 		struct ifbifconf ifbifconf;
    513 		struct ifbareq ifbareq;
    514 		struct ifbaconf ifbaconf;
    515 		struct ifbrparam ifbrparam;
    516 	} args;
    517 	struct ifdrv *ifd = (struct ifdrv *) data;
    518 	const struct bridge_control *bc = NULL; /* XXXGCC */
    519 	int s, error = 0;
    520 
    521 	/* Authorize command before calling splsoftnet(). */
    522 	switch (cmd) {
    523 	case SIOCGDRVSPEC:
    524 	case SIOCSDRVSPEC:
    525 		if (ifd->ifd_cmd >= bridge_control_table_size
    526 		    || (bc = &bridge_control_table[ifd->ifd_cmd]) == NULL) {
    527 			error = EINVAL;
    528 			return error;
    529 		}
    530 
    531 		/* We only care about BC_F_SUSER at this point. */
    532 		if ((bc->bc_flags & BC_F_SUSER) == 0)
    533 			break;
    534 
    535 		error = kauth_authorize_network(l->l_cred,
    536 		    KAUTH_NETWORK_INTERFACE_BRIDGE,
    537 		    cmd == SIOCGDRVSPEC ?
    538 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_GETPRIV :
    539 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_SETPRIV,
    540 		     ifd, NULL, NULL);
    541 		if (error)
    542 			return error;
    543 
    544 		break;
    545 	}
    546 
    547 	s = splsoftnet();
    548 
    549 	switch (cmd) {
    550 	case SIOCGDRVSPEC:
    551 	case SIOCSDRVSPEC:
    552 		KASSERT(bc != NULL);
    553 		if (cmd == SIOCGDRVSPEC &&
    554 		    (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) == 0) {
    555 			error = EINVAL;
    556 			break;
    557 		}
    558 		else if (cmd == SIOCSDRVSPEC &&
    559 		    (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) != 0) {
    560 			error = EINVAL;
    561 			break;
    562 		}
    563 
    564 		/* BC_F_SUSER is checked above, before splsoftnet(). */
    565 
    566 		if ((bc->bc_flags & (BC_F_XLATEIN|BC_F_XLATEOUT)) == 0
    567 		    && (ifd->ifd_len != bc->bc_argsize
    568 			|| ifd->ifd_len > sizeof(args))) {
    569 			error = EINVAL;
    570 			break;
    571 		}
    572 
    573 		memset(&args, 0, sizeof(args));
    574 		if (bc->bc_flags & BC_F_COPYIN) {
    575 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
    576 			if (error)
    577 				break;
    578 		} else if (bc->bc_flags & BC_F_XLATEIN) {
    579 			args.ifbifconf.ifbic_len = ifd->ifd_len;
    580 			args.ifbifconf.ifbic_buf = ifd->ifd_data;
    581 		}
    582 
    583 		error = (*bc->bc_func)(sc, &args);
    584 		if (error)
    585 			break;
    586 
    587 		if (bc->bc_flags & BC_F_COPYOUT) {
    588 			error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
    589 		} else if (bc->bc_flags & BC_F_XLATEOUT) {
    590 			ifd->ifd_len = args.ifbifconf.ifbic_len;
    591 			ifd->ifd_data = args.ifbifconf.ifbic_buf;
    592 		}
    593 		break;
    594 
    595 	case SIOCSIFFLAGS:
    596 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
    597 			break;
    598 		switch (ifp->if_flags & (IFF_UP|IFF_RUNNING)) {
    599 		case IFF_RUNNING:
    600 			/*
    601 			 * If interface is marked down and it is running,
    602 			 * then stop and disable it.
    603 			 */
    604 			(*ifp->if_stop)(ifp, 1);
    605 			break;
    606 		case IFF_UP:
    607 			/*
    608 			 * If interface is marked up and it is stopped, then
    609 			 * start it.
    610 			 */
    611 			error = (*ifp->if_init)(ifp);
    612 			break;
    613 		default:
    614 			break;
    615 		}
    616 		break;
    617 
    618 	case SIOCSIFMTU:
    619 		if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET)
    620 			error = 0;
    621 		break;
    622 
    623 	default:
    624 		error = ifioctl_common(ifp, cmd, data);
    625 		break;
    626 	}
    627 
    628 	splx(s);
    629 
    630 	return error;
    631 }
    632 
    633 /*
    634  * bridge_lookup_member:
    635  *
    636  *	Lookup a bridge member interface.
    637  */
    638 static struct bridge_iflist *
    639 bridge_lookup_member(struct bridge_softc *sc, const char *name, struct psref *psref)
    640 {
    641 	struct bridge_iflist *bif;
    642 	struct ifnet *ifp;
    643 	int s;
    644 
    645 	BRIDGE_PSZ_RENTER(s);
    646 
    647 	BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
    648 		ifp = bif->bif_ifp;
    649 		if (strcmp(ifp->if_xname, name) == 0)
    650 			break;
    651 	}
    652 	if (bif != NULL)
    653 		bridge_acquire_member(sc, bif, psref);
    654 
    655 	BRIDGE_PSZ_REXIT(s);
    656 
    657 	return bif;
    658 }
    659 
    660 /*
    661  * bridge_lookup_member_if:
    662  *
    663  *	Lookup a bridge member interface by ifnet*.
    664  */
    665 static struct bridge_iflist *
    666 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp,
    667     struct psref *psref)
    668 {
    669 	struct bridge_iflist *bif;
    670 	int s;
    671 
    672 	BRIDGE_PSZ_RENTER(s);
    673 
    674 	bif = member_ifp->if_bridgeif;
    675 	if (bif != NULL) {
    676 		psref_acquire(psref, &bif->bif_psref,
    677 		    bridge_psref_class);
    678 	}
    679 
    680 	BRIDGE_PSZ_REXIT(s);
    681 
    682 	return bif;
    683 }
    684 
    685 static void
    686 bridge_acquire_member(struct bridge_softc *sc, struct bridge_iflist *bif,
    687     struct psref *psref)
    688 {
    689 
    690 	psref_acquire(psref, &bif->bif_psref, bridge_psref_class);
    691 }
    692 
    693 /*
    694  * bridge_release_member:
    695  *
    696  *	Release the specified member interface.
    697  */
    698 static void
    699 bridge_release_member(struct bridge_softc *sc, struct bridge_iflist *bif,
    700     struct psref *psref)
    701 {
    702 
    703 	psref_release(psref, &bif->bif_psref, bridge_psref_class);
    704 }
    705 
    706 /*
    707  * bridge_delete_member:
    708  *
    709  *	Delete the specified member interface.
    710  */
    711 static void
    712 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif)
    713 {
    714 	struct ifnet *ifs = bif->bif_ifp;
    715 
    716 	KASSERT(BRIDGE_LOCKED(sc));
    717 
    718 	ifs->_if_input = ether_input;
    719 	ifs->if_bridge = NULL;
    720 	ifs->if_bridgeif = NULL;
    721 
    722 	PSLIST_WRITER_REMOVE(bif, bif_next);
    723 	BRIDGE_PSZ_PERFORM(sc);
    724 	BRIDGE_UNLOCK(sc);
    725 
    726 	psref_target_destroy(&bif->bif_psref, bridge_psref_class);
    727 
    728 	PSLIST_ENTRY_DESTROY(bif, bif_next);
    729 	kmem_free(bif, sizeof(*bif));
    730 
    731 	BRIDGE_LOCK(sc);
    732 }
    733 
    734 static int
    735 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
    736 {
    737 	struct ifbreq *req = arg;
    738 	struct bridge_iflist *bif = NULL;
    739 	struct ifnet *ifs;
    740 	int error = 0;
    741 	struct psref psref;
    742 
    743 	ifs = if_get(req->ifbr_ifsname, &psref);
    744 	if (ifs == NULL)
    745 		return ENOENT;
    746 
    747 	if (ifs->if_bridge == sc) {
    748 		error = EEXIST;
    749 		goto out;
    750 	}
    751 
    752 	if (ifs->if_bridge != NULL) {
    753 		error = EBUSY;
    754 		goto out;
    755 	}
    756 
    757 	if (ifs->_if_input != ether_input) {
    758 		error = EINVAL;
    759 		goto out;
    760 	}
    761 
    762 	/* FIXME: doesn't work with non-IFF_SIMPLEX interfaces */
    763 	if ((ifs->if_flags & IFF_SIMPLEX) == 0) {
    764 		error = EINVAL;
    765 		goto out;
    766 	}
    767 
    768 	bif = kmem_alloc(sizeof(*bif), KM_SLEEP);
    769 
    770 	switch (ifs->if_type) {
    771 	case IFT_ETHER:
    772 		if (sc->sc_if.if_mtu != ifs->if_mtu) {
    773 			error = EINVAL;
    774 			goto out;
    775 		}
    776 		/* FALLTHROUGH */
    777 	case IFT_L2TP:
    778 		IFNET_LOCK(ifs);
    779 		error = ether_enable_vlan_mtu(ifs);
    780 		IFNET_UNLOCK(ifs);
    781 		if (error > 0)
    782 			goto out;
    783 		/*
    784 		 * Place the interface into promiscuous mode.
    785 		 */
    786 		error = ifpromisc(ifs, 1);
    787 		if (error)
    788 			goto out;
    789 		break;
    790 	default:
    791 		error = EINVAL;
    792 		goto out;
    793 	}
    794 
    795 	bif->bif_ifp = ifs;
    796 	bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
    797 	bif->bif_priority = BSTP_DEFAULT_PORT_PRIORITY;
    798 	bif->bif_path_cost = BSTP_DEFAULT_PATH_COST;
    799 	PSLIST_ENTRY_INIT(bif, bif_next);
    800 	psref_target_init(&bif->bif_psref, bridge_psref_class);
    801 
    802 	BRIDGE_LOCK(sc);
    803 
    804 	ifs->if_bridge = sc;
    805 	ifs->if_bridgeif = bif;
    806 	PSLIST_WRITER_INSERT_HEAD(&sc->sc_iflist_psref.bip_iflist, bif, bif_next);
    807 	ifs->_if_input = bridge_input;
    808 
    809 	BRIDGE_UNLOCK(sc);
    810 
    811 	if (sc->sc_if.if_flags & IFF_RUNNING)
    812 		bstp_initialization(sc);
    813 	else
    814 		bstp_stop(sc);
    815 
    816  out:
    817 	if_put(ifs, &psref);
    818 	if (error) {
    819 		if (bif != NULL)
    820 			kmem_free(bif, sizeof(*bif));
    821 	}
    822 	return error;
    823 }
    824 
    825 static int
    826 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
    827 {
    828 	struct ifbreq *req = arg;
    829 	const char *name = req->ifbr_ifsname;
    830 	struct bridge_iflist *bif;
    831 	struct ifnet *ifs;
    832 
    833 	BRIDGE_LOCK(sc);
    834 
    835 	/*
    836 	 * Don't use bridge_lookup_member. We want to get a member
    837 	 * with bif_refs == 0.
    838 	 */
    839 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) {
    840 		ifs = bif->bif_ifp;
    841 		if (strcmp(ifs->if_xname, name) == 0)
    842 			break;
    843 	}
    844 
    845 	if (bif == NULL) {
    846 		BRIDGE_UNLOCK(sc);
    847 		return ENOENT;
    848 	}
    849 
    850 	bridge_delete_member(sc, bif);
    851 
    852 	BRIDGE_UNLOCK(sc);
    853 
    854 	switch (ifs->if_type) {
    855 	case IFT_ETHER:
    856 	case IFT_L2TP:
    857 		/*
    858 		 * Take the interface out of promiscuous mode.
    859 		 * Don't call it with holding a spin lock.
    860 		 */
    861 		(void) ifpromisc(ifs, 0);
    862 		IFNET_LOCK(ifs);
    863 		(void) ether_disable_vlan_mtu(ifs);
    864 		IFNET_UNLOCK(ifs);
    865 		break;
    866 	default:
    867 #ifdef DIAGNOSTIC
    868 		panic("bridge_delete_member: impossible");
    869 #endif
    870 		break;
    871 	}
    872 
    873 	bridge_rtdelete(sc, ifs);
    874 
    875 	if (sc->sc_if.if_flags & IFF_RUNNING)
    876 		bstp_initialization(sc);
    877 
    878 	return 0;
    879 }
    880 
    881 static int
    882 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
    883 {
    884 	struct ifbreq *req = arg;
    885 	struct bridge_iflist *bif;
    886 	struct psref psref;
    887 
    888 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
    889 	if (bif == NULL)
    890 		return ENOENT;
    891 
    892 	req->ifbr_ifsflags = bif->bif_flags;
    893 	req->ifbr_state = bif->bif_state;
    894 	req->ifbr_priority = bif->bif_priority;
    895 	req->ifbr_path_cost = bif->bif_path_cost;
    896 	req->ifbr_portno = bif->bif_ifp->if_index & 0xff;
    897 
    898 	bridge_release_member(sc, bif, &psref);
    899 
    900 	return 0;
    901 }
    902 
    903 static int
    904 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
    905 {
    906 	struct ifbreq *req = arg;
    907 	struct bridge_iflist *bif;
    908 	struct psref psref;
    909 
    910 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
    911 	if (bif == NULL)
    912 		return ENOENT;
    913 
    914 	if (req->ifbr_ifsflags & IFBIF_STP) {
    915 		switch (bif->bif_ifp->if_type) {
    916 		case IFT_ETHER:
    917 		case IFT_L2TP:
    918 			/* These can do spanning tree. */
    919 			break;
    920 
    921 		default:
    922 			/* Nothing else can. */
    923 			bridge_release_member(sc, bif, &psref);
    924 			return EINVAL;
    925 		}
    926 	}
    927 
    928 	bif->bif_flags = req->ifbr_ifsflags;
    929 
    930 	bridge_release_member(sc, bif, &psref);
    931 
    932 	if (sc->sc_if.if_flags & IFF_RUNNING)
    933 		bstp_initialization(sc);
    934 
    935 	return 0;
    936 }
    937 
    938 static int
    939 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
    940 {
    941 	struct ifbrparam *param = arg;
    942 
    943 	sc->sc_brtmax = param->ifbrp_csize;
    944 	bridge_rttrim(sc);
    945 
    946 	return 0;
    947 }
    948 
    949 static int
    950 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
    951 {
    952 	struct ifbrparam *param = arg;
    953 
    954 	param->ifbrp_csize = sc->sc_brtmax;
    955 
    956 	return 0;
    957 }
    958 
    959 static int
    960 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
    961 {
    962 	struct ifbifconf *bifc = arg;
    963 	struct bridge_iflist *bif;
    964 	struct ifbreq *breqs;
    965 	int i, count, error = 0;
    966 
    967 retry:
    968 	BRIDGE_LOCK(sc);
    969 	count = 0;
    970 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc)
    971 		count++;
    972 	BRIDGE_UNLOCK(sc);
    973 
    974 	if (count == 0) {
    975 		bifc->ifbic_len = 0;
    976 		return 0;
    977 	}
    978 
    979 	if (bifc->ifbic_len == 0 || bifc->ifbic_len < (sizeof(*breqs) * count)) {
    980 		/* Tell that a larger buffer is needed */
    981 		bifc->ifbic_len = sizeof(*breqs) * count;
    982 		return 0;
    983 	}
    984 
    985 	breqs = kmem_alloc(sizeof(*breqs) * count, KM_SLEEP);
    986 
    987 	BRIDGE_LOCK(sc);
    988 
    989 	i = 0;
    990 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc)
    991 		i++;
    992 	if (i > count) {
    993 		/*
    994 		 * The number of members has been increased.
    995 		 * We need more memory!
    996 		 */
    997 		BRIDGE_UNLOCK(sc);
    998 		kmem_free(breqs, sizeof(*breqs) * count);
    999 		goto retry;
   1000 	}
   1001 
   1002 	i = 0;
   1003 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) {
   1004 		struct ifbreq *breq = &breqs[i++];
   1005 		memset(breq, 0, sizeof(*breq));
   1006 
   1007 		strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname,
   1008 		    sizeof(breq->ifbr_ifsname));
   1009 		breq->ifbr_ifsflags = bif->bif_flags;
   1010 		breq->ifbr_state = bif->bif_state;
   1011 		breq->ifbr_priority = bif->bif_priority;
   1012 		breq->ifbr_path_cost = bif->bif_path_cost;
   1013 		breq->ifbr_portno = bif->bif_ifp->if_index & 0xff;
   1014 	}
   1015 
   1016 	/* Don't call copyout with holding the mutex */
   1017 	BRIDGE_UNLOCK(sc);
   1018 
   1019 	for (i = 0; i < count; i++) {
   1020 		error = copyout(&breqs[i], bifc->ifbic_req + i, sizeof(*breqs));
   1021 		if (error)
   1022 			break;
   1023 	}
   1024 	bifc->ifbic_len = sizeof(*breqs) * i;
   1025 
   1026 	kmem_free(breqs, sizeof(*breqs) * count);
   1027 
   1028 	return error;
   1029 }
   1030 
   1031 static int
   1032 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
   1033 {
   1034 	struct ifbaconf *bac = arg;
   1035 	struct bridge_rtnode *brt;
   1036 	struct ifbareq bareq;
   1037 	int count = 0, error = 0, len;
   1038 
   1039 	if (bac->ifbac_len == 0)
   1040 		return 0;
   1041 
   1042 	BRIDGE_RT_LOCK(sc);
   1043 
   1044 	len = bac->ifbac_len;
   1045 	LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
   1046 		if (len < sizeof(bareq))
   1047 			goto out;
   1048 		memset(&bareq, 0, sizeof(bareq));
   1049 		strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
   1050 		    sizeof(bareq.ifba_ifsname));
   1051 		memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
   1052 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
   1053 			bareq.ifba_expire = brt->brt_expire - time_uptime;
   1054 		} else
   1055 			bareq.ifba_expire = 0;
   1056 		bareq.ifba_flags = brt->brt_flags;
   1057 
   1058 		error = copyout(&bareq, bac->ifbac_req + count, sizeof(bareq));
   1059 		if (error)
   1060 			goto out;
   1061 		count++;
   1062 		len -= sizeof(bareq);
   1063 	}
   1064  out:
   1065 	BRIDGE_RT_UNLOCK(sc);
   1066 
   1067 	bac->ifbac_len = sizeof(bareq) * count;
   1068 	return error;
   1069 }
   1070 
   1071 static int
   1072 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
   1073 {
   1074 	struct ifbareq *req = arg;
   1075 	struct bridge_iflist *bif;
   1076 	int error;
   1077 	struct psref psref;
   1078 
   1079 	bif = bridge_lookup_member(sc, req->ifba_ifsname, &psref);
   1080 	if (bif == NULL)
   1081 		return ENOENT;
   1082 
   1083 	error = bridge_rtupdate(sc, req->ifba_dst, bif->bif_ifp, 1,
   1084 	    req->ifba_flags);
   1085 
   1086 	bridge_release_member(sc, bif, &psref);
   1087 
   1088 	return error;
   1089 }
   1090 
   1091 static int
   1092 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
   1093 {
   1094 	struct ifbrparam *param = arg;
   1095 
   1096 	sc->sc_brttimeout = param->ifbrp_ctime;
   1097 
   1098 	return 0;
   1099 }
   1100 
   1101 static int
   1102 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
   1103 {
   1104 	struct ifbrparam *param = arg;
   1105 
   1106 	param->ifbrp_ctime = sc->sc_brttimeout;
   1107 
   1108 	return 0;
   1109 }
   1110 
   1111 static int
   1112 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
   1113 {
   1114 	struct ifbareq *req = arg;
   1115 
   1116 	return (bridge_rtdaddr(sc, req->ifba_dst));
   1117 }
   1118 
   1119 static int
   1120 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
   1121 {
   1122 	struct ifbreq *req = arg;
   1123 
   1124 	bridge_rtflush(sc, req->ifbr_ifsflags);
   1125 
   1126 	return 0;
   1127 }
   1128 
   1129 static int
   1130 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
   1131 {
   1132 	struct ifbrparam *param = arg;
   1133 
   1134 	param->ifbrp_prio = sc->sc_bridge_priority;
   1135 
   1136 	return 0;
   1137 }
   1138 
   1139 static int
   1140 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
   1141 {
   1142 	struct ifbrparam *param = arg;
   1143 
   1144 	sc->sc_bridge_priority = param->ifbrp_prio;
   1145 
   1146 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1147 		bstp_initialization(sc);
   1148 
   1149 	return 0;
   1150 }
   1151 
   1152 static int
   1153 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
   1154 {
   1155 	struct ifbrparam *param = arg;
   1156 
   1157 	param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8;
   1158 
   1159 	return 0;
   1160 }
   1161 
   1162 static int
   1163 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
   1164 {
   1165 	struct ifbrparam *param = arg;
   1166 
   1167 	if (param->ifbrp_hellotime == 0)
   1168 		return EINVAL;
   1169 	sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8;
   1170 
   1171 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1172 		bstp_initialization(sc);
   1173 
   1174 	return 0;
   1175 }
   1176 
   1177 static int
   1178 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
   1179 {
   1180 	struct ifbrparam *param = arg;
   1181 
   1182 	param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8;
   1183 
   1184 	return 0;
   1185 }
   1186 
   1187 static int
   1188 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
   1189 {
   1190 	struct ifbrparam *param = arg;
   1191 
   1192 	if (param->ifbrp_fwddelay == 0)
   1193 		return EINVAL;
   1194 	sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8;
   1195 
   1196 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1197 		bstp_initialization(sc);
   1198 
   1199 	return 0;
   1200 }
   1201 
   1202 static int
   1203 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
   1204 {
   1205 	struct ifbrparam *param = arg;
   1206 
   1207 	param->ifbrp_maxage = sc->sc_bridge_max_age >> 8;
   1208 
   1209 	return 0;
   1210 }
   1211 
   1212 static int
   1213 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
   1214 {
   1215 	struct ifbrparam *param = arg;
   1216 
   1217 	if (param->ifbrp_maxage == 0)
   1218 		return EINVAL;
   1219 	sc->sc_bridge_max_age = param->ifbrp_maxage << 8;
   1220 
   1221 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1222 		bstp_initialization(sc);
   1223 
   1224 	return 0;
   1225 }
   1226 
   1227 static int
   1228 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
   1229 {
   1230 	struct ifbreq *req = arg;
   1231 	struct bridge_iflist *bif;
   1232 	struct psref psref;
   1233 
   1234 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
   1235 	if (bif == NULL)
   1236 		return ENOENT;
   1237 
   1238 	bif->bif_priority = req->ifbr_priority;
   1239 
   1240 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1241 		bstp_initialization(sc);
   1242 
   1243 	bridge_release_member(sc, bif, &psref);
   1244 
   1245 	return 0;
   1246 }
   1247 
   1248 #if defined(BRIDGE_IPF)
   1249 static int
   1250 bridge_ioctl_gfilt(struct bridge_softc *sc, void *arg)
   1251 {
   1252 	struct ifbrparam *param = arg;
   1253 
   1254 	param->ifbrp_filter = sc->sc_filter_flags;
   1255 
   1256 	return 0;
   1257 }
   1258 
   1259 static int
   1260 bridge_ioctl_sfilt(struct bridge_softc *sc, void *arg)
   1261 {
   1262 	struct ifbrparam *param = arg;
   1263 	uint32_t nflags, oflags;
   1264 
   1265 	if (param->ifbrp_filter & ~IFBF_FILT_MASK)
   1266 		return EINVAL;
   1267 
   1268 	nflags = param->ifbrp_filter;
   1269 	oflags = sc->sc_filter_flags;
   1270 
   1271 	if ((nflags & IFBF_FILT_USEIPF) && !(oflags & IFBF_FILT_USEIPF)) {
   1272 		pfil_add_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
   1273 			sc->sc_if.if_pfil);
   1274 	}
   1275 	if (!(nflags & IFBF_FILT_USEIPF) && (oflags & IFBF_FILT_USEIPF)) {
   1276 		pfil_remove_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
   1277 			sc->sc_if.if_pfil);
   1278 	}
   1279 
   1280 	sc->sc_filter_flags = nflags;
   1281 
   1282 	return 0;
   1283 }
   1284 #endif /* BRIDGE_IPF */
   1285 
   1286 static int
   1287 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
   1288 {
   1289 	struct ifbreq *req = arg;
   1290 	struct bridge_iflist *bif;
   1291 	struct psref psref;
   1292 
   1293 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
   1294 	if (bif == NULL)
   1295 		return ENOENT;
   1296 
   1297 	bif->bif_path_cost = req->ifbr_path_cost;
   1298 
   1299 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1300 		bstp_initialization(sc);
   1301 
   1302 	bridge_release_member(sc, bif, &psref);
   1303 
   1304 	return 0;
   1305 }
   1306 
   1307 /*
   1308  * bridge_ifdetach:
   1309  *
   1310  *	Detach an interface from a bridge.  Called when a member
   1311  *	interface is detaching.
   1312  */
   1313 void
   1314 bridge_ifdetach(struct ifnet *ifp)
   1315 {
   1316 	struct bridge_softc *sc = ifp->if_bridge;
   1317 	struct ifbreq breq;
   1318 
   1319 	/* ioctl_lock should prevent this from happening */
   1320 	KASSERT(sc != NULL);
   1321 
   1322 	memset(&breq, 0, sizeof(breq));
   1323 	strlcpy(breq.ifbr_ifsname, ifp->if_xname, sizeof(breq.ifbr_ifsname));
   1324 
   1325 	(void) bridge_ioctl_del(sc, &breq);
   1326 }
   1327 
   1328 /*
   1329  * bridge_init:
   1330  *
   1331  *	Initialize a bridge interface.
   1332  */
   1333 static int
   1334 bridge_init(struct ifnet *ifp)
   1335 {
   1336 	struct bridge_softc *sc = ifp->if_softc;
   1337 
   1338 	KASSERT((ifp->if_flags & IFF_RUNNING) == 0);
   1339 
   1340 	callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
   1341 	    bridge_timer, sc);
   1342 	bstp_initialization(sc);
   1343 
   1344 	ifp->if_flags |= IFF_RUNNING;
   1345 	return 0;
   1346 }
   1347 
   1348 /*
   1349  * bridge_stop:
   1350  *
   1351  *	Stop the bridge interface.
   1352  */
   1353 static void
   1354 bridge_stop(struct ifnet *ifp, int disable)
   1355 {
   1356 	struct bridge_softc *sc = ifp->if_softc;
   1357 
   1358 	KASSERT((ifp->if_flags & IFF_RUNNING) != 0);
   1359 	ifp->if_flags &= ~IFF_RUNNING;
   1360 
   1361 	callout_stop(&sc->sc_brcallout);
   1362 	bstp_stop(sc);
   1363 	bridge_rtflush(sc, IFBF_FLUSHDYN);
   1364 }
   1365 
   1366 /*
   1367  * bridge_enqueue:
   1368  *
   1369  *	Enqueue a packet on a bridge member interface.
   1370  */
   1371 void
   1372 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m,
   1373     int runfilt)
   1374 {
   1375 	int len, error;
   1376 	short mflags;
   1377 
   1378 	/*
   1379 	 * Clear any in-bound checksum flags for this packet.
   1380 	 */
   1381 	m->m_pkthdr.csum_flags = 0;
   1382 
   1383 	if (runfilt) {
   1384 		if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
   1385 		    dst_ifp, PFIL_OUT) != 0) {
   1386 			if (m != NULL)
   1387 				m_freem(m);
   1388 			return;
   1389 		}
   1390 		if (m == NULL)
   1391 			return;
   1392 	}
   1393 
   1394 #ifdef ALTQ
   1395 	KERNEL_LOCK(1, NULL);
   1396 	/*
   1397 	 * If ALTQ is enabled on the member interface, do
   1398 	 * classification; the queueing discipline might
   1399 	 * not require classification, but might require
   1400 	 * the address family/header pointer in the pktattr.
   1401 	 */
   1402 	if (ALTQ_IS_ENABLED(&dst_ifp->if_snd)) {
   1403 		/* XXX IFT_ETHER */
   1404 		altq_etherclassify(&dst_ifp->if_snd, m);
   1405 	}
   1406 	KERNEL_UNLOCK_ONE(NULL);
   1407 #endif /* ALTQ */
   1408 
   1409 	len = m->m_pkthdr.len;
   1410 	mflags = m->m_flags;
   1411 
   1412 	error = if_transmit_lock(dst_ifp, m);
   1413 	if (error) {
   1414 		/* mbuf is already freed */
   1415 		sc->sc_if.if_oerrors++;
   1416 		return;
   1417 	}
   1418 
   1419 	sc->sc_if.if_opackets++;
   1420 	sc->sc_if.if_obytes += len;
   1421 	if (mflags & M_MCAST)
   1422 		sc->sc_if.if_omcasts++;
   1423 }
   1424 
   1425 /*
   1426  * bridge_output:
   1427  *
   1428  *	Send output from a bridge member interface.  This
   1429  *	performs the bridging function for locally originated
   1430  *	packets.
   1431  *
   1432  *	The mbuf has the Ethernet header already attached.  We must
   1433  *	enqueue or free the mbuf before returning.
   1434  */
   1435 int
   1436 bridge_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa,
   1437     const struct rtentry *rt)
   1438 {
   1439 	struct ether_header *eh;
   1440 	struct ifnet *dst_if;
   1441 	struct bridge_softc *sc;
   1442 	int s;
   1443 
   1444 	/*
   1445 	 * bridge_output() is called from ether_output(), furthermore
   1446 	 * ifp argument doesn't point to bridge(4). So, don't assert
   1447 	 * IFEF_MPSAFE here.
   1448 	 */
   1449 
   1450 	if (m->m_len < ETHER_HDR_LEN) {
   1451 		m = m_pullup(m, ETHER_HDR_LEN);
   1452 		if (m == NULL)
   1453 			return 0;
   1454 	}
   1455 
   1456 	eh = mtod(m, struct ether_header *);
   1457 	sc = ifp->if_bridge;
   1458 
   1459 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
   1460 		if (memcmp(etherbroadcastaddr,
   1461 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
   1462 			m->m_flags |= M_BCAST;
   1463 		else
   1464 			m->m_flags |= M_MCAST;
   1465 	}
   1466 
   1467 	/*
   1468 	 * If bridge is down, but the original output interface is up,
   1469 	 * go ahead and send out that interface.  Otherwise, the packet
   1470 	 * is dropped below.
   1471 	 */
   1472 	if (__predict_false(sc == NULL) ||
   1473 	    (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
   1474 		dst_if = ifp;
   1475 		goto sendunicast;
   1476 	}
   1477 
   1478 	/*
   1479 	 * If the packet is a multicast, or we don't know a better way to
   1480 	 * get there, send to all interfaces.
   1481 	 */
   1482 	if ((m->m_flags & (M_MCAST | M_BCAST)) != 0)
   1483 		dst_if = NULL;
   1484 	else
   1485 		dst_if = bridge_rtlookup(sc, eh->ether_dhost);
   1486 	if (dst_if == NULL) {
   1487 		/* XXX Should call bridge_broadcast, but there are locking
   1488 		 * issues which need resolving first. */
   1489 		struct bridge_iflist *bif;
   1490 		struct mbuf *mc;
   1491 		bool used = false;
   1492 
   1493 		BRIDGE_PSZ_RENTER(s);
   1494 		BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
   1495 			struct psref psref;
   1496 
   1497 			bridge_acquire_member(sc, bif, &psref);
   1498 			BRIDGE_PSZ_REXIT(s);
   1499 
   1500 			dst_if = bif->bif_ifp;
   1501 			if ((dst_if->if_flags & IFF_RUNNING) == 0)
   1502 				goto next;
   1503 
   1504 			/*
   1505 			 * If this is not the original output interface,
   1506 			 * and the interface is participating in spanning
   1507 			 * tree, make sure the port is in a state that
   1508 			 * allows forwarding.
   1509 			 */
   1510 			if (dst_if != ifp &&
   1511 			    (bif->bif_flags & IFBIF_STP) != 0) {
   1512 				switch (bif->bif_state) {
   1513 				case BSTP_IFSTATE_BLOCKING:
   1514 				case BSTP_IFSTATE_LISTENING:
   1515 				case BSTP_IFSTATE_DISABLED:
   1516 					goto next;
   1517 				}
   1518 			}
   1519 
   1520 			if (PSLIST_READER_NEXT(bif, struct bridge_iflist,
   1521 			    bif_next) == NULL &&
   1522 			    ((m->m_flags & (M_MCAST | M_BCAST)) == 0 ||
   1523 			    dst_if == ifp))
   1524 			{
   1525 				used = true;
   1526 				mc = m;
   1527 			} else {
   1528 				mc = m_copym(m, 0, M_COPYALL, M_NOWAIT);
   1529 				if (mc == NULL) {
   1530 					sc->sc_if.if_oerrors++;
   1531 					goto next;
   1532 				}
   1533 			}
   1534 
   1535 			bridge_enqueue(sc, dst_if, mc, 0);
   1536 
   1537 			if ((m->m_flags & (M_MCAST | M_BCAST)) != 0 &&
   1538 			    dst_if != ifp)
   1539 			{
   1540 				if (PSLIST_READER_NEXT(bif,
   1541 				    struct bridge_iflist, bif_next) == NULL)
   1542 				{
   1543 					used = true;
   1544 					mc = m;
   1545 				} else {
   1546 					mc = m_copym(m, 0, M_COPYALL,
   1547 					    M_DONTWAIT);
   1548 					if (mc == NULL) {
   1549 						sc->sc_if.if_oerrors++;
   1550 						goto next;
   1551 					}
   1552 				}
   1553 
   1554 				m_set_rcvif(mc, dst_if);
   1555 				mc->m_flags &= ~M_PROMISC;
   1556 
   1557 #ifndef NET_MPSAFE
   1558 				s = splsoftnet();
   1559 #endif
   1560 				ether_input(dst_if, mc);
   1561 #ifndef NET_MPSAFE
   1562 				splx(s);
   1563 #endif
   1564 			}
   1565 
   1566 next:
   1567 			BRIDGE_PSZ_RENTER(s);
   1568 			bridge_release_member(sc, bif, &psref);
   1569 
   1570 			/* Guarantee we don't re-enter the loop as we already
   1571 			 * decided we're at the end. */
   1572 			if (used)
   1573 				break;
   1574 		}
   1575 		BRIDGE_PSZ_REXIT(s);
   1576 
   1577 		if (!used)
   1578 			m_freem(m);
   1579 		return 0;
   1580 	}
   1581 
   1582  sendunicast:
   1583 	/*
   1584 	 * XXX Spanning tree consideration here?
   1585 	 */
   1586 
   1587 	if ((dst_if->if_flags & IFF_RUNNING) == 0) {
   1588 		m_freem(m);
   1589 		return 0;
   1590 	}
   1591 
   1592 	bridge_enqueue(sc, dst_if, m, 0);
   1593 
   1594 	return 0;
   1595 }
   1596 
   1597 /*
   1598  * bridge_start:
   1599  *
   1600  *	Start output on a bridge.
   1601  *
   1602  *	NOTE: This routine should never be called in this implementation.
   1603  */
   1604 static void
   1605 bridge_start(struct ifnet *ifp)
   1606 {
   1607 
   1608 	printf("%s: bridge_start() called\n", ifp->if_xname);
   1609 }
   1610 
   1611 /*
   1612  * bridge_forward:
   1613  *
   1614  *	The forwarding function of the bridge.
   1615  */
   1616 static void
   1617 bridge_forward(struct bridge_softc *sc, struct mbuf *m)
   1618 {
   1619 	struct bridge_iflist *bif;
   1620 	struct ifnet *src_if, *dst_if;
   1621 	struct ether_header *eh;
   1622 	struct psref psref;
   1623 	struct psref psref_src;
   1624 	DECLARE_LOCK_VARIABLE;
   1625 
   1626 	if ((sc->sc_if.if_flags & IFF_RUNNING) == 0)
   1627 		return;
   1628 
   1629 	src_if = m_get_rcvif_psref(m, &psref_src);
   1630 	if (src_if == NULL) {
   1631 		/* Interface is being destroyed? */
   1632 		m_freem(m);
   1633 		goto out;
   1634 	}
   1635 
   1636 	sc->sc_if.if_ipackets++;
   1637 	sc->sc_if.if_ibytes += m->m_pkthdr.len;
   1638 
   1639 	/*
   1640 	 * Look up the bridge_iflist.
   1641 	 */
   1642 	bif = bridge_lookup_member_if(sc, src_if, &psref);
   1643 	if (bif == NULL) {
   1644 		/* Interface is not a bridge member (anymore?) */
   1645 		m_freem(m);
   1646 		goto out;
   1647 	}
   1648 
   1649 	if (bif->bif_flags & IFBIF_STP) {
   1650 		switch (bif->bif_state) {
   1651 		case BSTP_IFSTATE_BLOCKING:
   1652 		case BSTP_IFSTATE_LISTENING:
   1653 		case BSTP_IFSTATE_DISABLED:
   1654 			m_freem(m);
   1655 			bridge_release_member(sc, bif, &psref);
   1656 			goto out;
   1657 		}
   1658 	}
   1659 
   1660 	eh = mtod(m, struct ether_header *);
   1661 
   1662 	/*
   1663 	 * If the interface is learning, and the source
   1664 	 * address is valid and not multicast, record
   1665 	 * the address.
   1666 	 */
   1667 	if ((bif->bif_flags & IFBIF_LEARNING) != 0 &&
   1668 	    ETHER_IS_MULTICAST(eh->ether_shost) == 0 &&
   1669 	    (eh->ether_shost[0] == 0 &&
   1670 	     eh->ether_shost[1] == 0 &&
   1671 	     eh->ether_shost[2] == 0 &&
   1672 	     eh->ether_shost[3] == 0 &&
   1673 	     eh->ether_shost[4] == 0 &&
   1674 	     eh->ether_shost[5] == 0) == 0) {
   1675 		(void) bridge_rtupdate(sc, eh->ether_shost,
   1676 		    src_if, 0, IFBAF_DYNAMIC);
   1677 	}
   1678 
   1679 	if ((bif->bif_flags & IFBIF_STP) != 0 &&
   1680 	    bif->bif_state == BSTP_IFSTATE_LEARNING) {
   1681 		m_freem(m);
   1682 		bridge_release_member(sc, bif, &psref);
   1683 		goto out;
   1684 	}
   1685 
   1686 	bridge_release_member(sc, bif, &psref);
   1687 
   1688 	/*
   1689 	 * At this point, the port either doesn't participate
   1690 	 * in spanning tree or it is in the forwarding state.
   1691 	 */
   1692 
   1693 	/*
   1694 	 * If the packet is unicast, destined for someone on
   1695 	 * "this" side of the bridge, drop it.
   1696 	 */
   1697 	if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
   1698 		dst_if = bridge_rtlookup(sc, eh->ether_dhost);
   1699 		if (src_if == dst_if) {
   1700 			m_freem(m);
   1701 			goto out;
   1702 		}
   1703 	} else {
   1704 		/* ...forward it to all interfaces. */
   1705 		sc->sc_if.if_imcasts++;
   1706 		dst_if = NULL;
   1707 	}
   1708 
   1709 	if (pfil_run_hooks(sc->sc_if.if_pfil, &m, src_if, PFIL_IN) != 0) {
   1710 		if (m != NULL)
   1711 			m_freem(m);
   1712 		goto out;
   1713 	}
   1714 	if (m == NULL)
   1715 		goto out;
   1716 
   1717 	if (dst_if == NULL) {
   1718 		bridge_broadcast(sc, src_if, m);
   1719 		goto out;
   1720 	}
   1721 
   1722 	m_put_rcvif_psref(src_if, &psref_src);
   1723 	src_if = NULL;
   1724 
   1725 	/*
   1726 	 * At this point, we're dealing with a unicast frame
   1727 	 * going to a different interface.
   1728 	 */
   1729 	if ((dst_if->if_flags & IFF_RUNNING) == 0) {
   1730 		m_freem(m);
   1731 		goto out;
   1732 	}
   1733 
   1734 	bif = bridge_lookup_member_if(sc, dst_if, &psref);
   1735 	if (bif == NULL) {
   1736 		/* Not a member of the bridge (anymore?) */
   1737 		m_freem(m);
   1738 		goto out;
   1739 	}
   1740 
   1741 	if (bif->bif_flags & IFBIF_STP) {
   1742 		switch (bif->bif_state) {
   1743 		case BSTP_IFSTATE_DISABLED:
   1744 		case BSTP_IFSTATE_BLOCKING:
   1745 			m_freem(m);
   1746 			bridge_release_member(sc, bif, &psref);
   1747 			goto out;
   1748 		}
   1749 	}
   1750 
   1751 	bridge_release_member(sc, bif, &psref);
   1752 
   1753 	ACQUIRE_GLOBAL_LOCKS();
   1754 	bridge_enqueue(sc, dst_if, m, 1);
   1755 	RELEASE_GLOBAL_LOCKS();
   1756 out:
   1757 	if (src_if != NULL)
   1758 		m_put_rcvif_psref(src_if, &psref_src);
   1759 	return;
   1760 }
   1761 
   1762 static bool
   1763 bstp_state_before_learning(struct bridge_iflist *bif)
   1764 {
   1765 	if (bif->bif_flags & IFBIF_STP) {
   1766 		switch (bif->bif_state) {
   1767 		case BSTP_IFSTATE_BLOCKING:
   1768 		case BSTP_IFSTATE_LISTENING:
   1769 		case BSTP_IFSTATE_DISABLED:
   1770 			return true;
   1771 		}
   1772 	}
   1773 	return false;
   1774 }
   1775 
   1776 static bool
   1777 bridge_ourether(struct bridge_iflist *bif, struct ether_header *eh, int src)
   1778 {
   1779 	uint8_t *ether = src ? eh->ether_shost : eh->ether_dhost;
   1780 
   1781 	if (memcmp(CLLADDR(bif->bif_ifp->if_sadl), ether, ETHER_ADDR_LEN) == 0
   1782 #if NCARP > 0
   1783 	    || (bif->bif_ifp->if_carp &&
   1784 	        carp_ourether(bif->bif_ifp->if_carp, eh, IFT_ETHER, src) != NULL)
   1785 #endif /* NCARP > 0 */
   1786 	    )
   1787 		return true;
   1788 
   1789 	return false;
   1790 }
   1791 
   1792 /*
   1793  * bridge_input:
   1794  *
   1795  *	Receive input from a member interface.  Queue the packet for
   1796  *	bridging if it is not for us.
   1797  */
   1798 static void
   1799 bridge_input(struct ifnet *ifp, struct mbuf *m)
   1800 {
   1801 	struct bridge_softc *sc = ifp->if_bridge;
   1802 	struct bridge_iflist *bif;
   1803 	struct ether_header *eh;
   1804 	struct psref psref;
   1805 	int bound;
   1806 	DECLARE_LOCK_VARIABLE;
   1807 
   1808 	KASSERT(!cpu_intr_p());
   1809 
   1810 	if (__predict_false(sc == NULL) ||
   1811 	    (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
   1812 		ACQUIRE_GLOBAL_LOCKS();
   1813 		ether_input(ifp, m);
   1814 		RELEASE_GLOBAL_LOCKS();
   1815 		return;
   1816 	}
   1817 
   1818 	bound = curlwp_bind();
   1819 	bif = bridge_lookup_member_if(sc, ifp, &psref);
   1820 	if (bif == NULL) {
   1821 		curlwp_bindx(bound);
   1822 		ACQUIRE_GLOBAL_LOCKS();
   1823 		ether_input(ifp, m);
   1824 		RELEASE_GLOBAL_LOCKS();
   1825 		return;
   1826 	}
   1827 
   1828 	eh = mtod(m, struct ether_header *);
   1829 
   1830 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
   1831 		if (memcmp(etherbroadcastaddr,
   1832 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
   1833 			m->m_flags |= M_BCAST;
   1834 		else
   1835 			m->m_flags |= M_MCAST;
   1836 	}
   1837 
   1838 	/*
   1839 	 * A 'fast' path for packets addressed to interfaces that are
   1840 	 * part of this bridge.
   1841 	 */
   1842 	if (!(m->m_flags & (M_BCAST|M_MCAST)) &&
   1843 	    !bstp_state_before_learning(bif)) {
   1844 		struct bridge_iflist *_bif;
   1845 		struct ifnet *_ifp = NULL;
   1846 		int s;
   1847 		struct psref _psref;
   1848 
   1849 		BRIDGE_PSZ_RENTER(s);
   1850 		BRIDGE_IFLIST_READER_FOREACH(_bif, sc) {
   1851 			/* It is destined for us. */
   1852 			if (bridge_ourether(_bif, eh, 0)) {
   1853 				bridge_acquire_member(sc, _bif, &_psref);
   1854 				BRIDGE_PSZ_REXIT(s);
   1855 				if (_bif->bif_flags & IFBIF_LEARNING)
   1856 					(void) bridge_rtupdate(sc,
   1857 					    eh->ether_shost, ifp, 0, IFBAF_DYNAMIC);
   1858 				m_set_rcvif(m, _bif->bif_ifp);
   1859 				_ifp = _bif->bif_ifp;
   1860 				bridge_release_member(sc, _bif, &_psref);
   1861 				goto out;
   1862 			}
   1863 
   1864 			/* We just received a packet that we sent out. */
   1865 			if (bridge_ourether(_bif, eh, 1))
   1866 				break;
   1867 		}
   1868 		BRIDGE_PSZ_REXIT(s);
   1869 out:
   1870 
   1871 		if (_bif != NULL) {
   1872 			bridge_release_member(sc, bif, &psref);
   1873 			curlwp_bindx(bound);
   1874 			if (_ifp != NULL) {
   1875 				m->m_flags &= ~M_PROMISC;
   1876 				ACQUIRE_GLOBAL_LOCKS();
   1877 				ether_input(_ifp, m);
   1878 				RELEASE_GLOBAL_LOCKS();
   1879 			} else
   1880 				m_freem(m);
   1881 			return;
   1882 		}
   1883 	}
   1884 
   1885 	/* Tap off 802.1D packets; they do not get forwarded. */
   1886 	if (bif->bif_flags & IFBIF_STP &&
   1887 	    memcmp(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN) == 0) {
   1888 		bstp_input(sc, bif, m);
   1889 		bridge_release_member(sc, bif, &psref);
   1890 		curlwp_bindx(bound);
   1891 		return;
   1892 	}
   1893 
   1894 	/*
   1895 	 * A normal switch would discard the packet here, but that's not what
   1896 	 * we've done historically. This also prevents some obnoxious behaviour.
   1897 	 */
   1898 	if (bstp_state_before_learning(bif)) {
   1899 		bridge_release_member(sc, bif, &psref);
   1900 		curlwp_bindx(bound);
   1901 		ACQUIRE_GLOBAL_LOCKS();
   1902 		ether_input(ifp, m);
   1903 		RELEASE_GLOBAL_LOCKS();
   1904 		return;
   1905 	}
   1906 
   1907 	bridge_release_member(sc, bif, &psref);
   1908 
   1909 	bridge_forward(sc, m);
   1910 
   1911 	curlwp_bindx(bound);
   1912 }
   1913 
   1914 /*
   1915  * bridge_broadcast:
   1916  *
   1917  *	Send a frame to all interfaces that are members of
   1918  *	the bridge, except for the one on which the packet
   1919  *	arrived.
   1920  */
   1921 static void
   1922 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
   1923     struct mbuf *m)
   1924 {
   1925 	struct bridge_iflist *bif;
   1926 	struct mbuf *mc;
   1927 	struct ifnet *dst_if;
   1928 	bool bmcast;
   1929 	int s;
   1930 	DECLARE_LOCK_VARIABLE;
   1931 
   1932 	bmcast = m->m_flags & (M_BCAST|M_MCAST);
   1933 
   1934 	BRIDGE_PSZ_RENTER(s);
   1935 	BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
   1936 		struct psref psref;
   1937 
   1938 		bridge_acquire_member(sc, bif, &psref);
   1939 		BRIDGE_PSZ_REXIT(s);
   1940 
   1941 		dst_if = bif->bif_ifp;
   1942 
   1943 		if (bif->bif_flags & IFBIF_STP) {
   1944 			switch (bif->bif_state) {
   1945 			case BSTP_IFSTATE_BLOCKING:
   1946 			case BSTP_IFSTATE_DISABLED:
   1947 				goto next;
   1948 			}
   1949 		}
   1950 
   1951 		if ((bif->bif_flags & IFBIF_DISCOVER) == 0 && !bmcast)
   1952 			goto next;
   1953 
   1954 		if ((dst_if->if_flags & IFF_RUNNING) == 0)
   1955 			goto next;
   1956 
   1957 		if (dst_if != src_if) {
   1958 			mc = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
   1959 			if (mc == NULL) {
   1960 				sc->sc_if.if_oerrors++;
   1961 				goto next;
   1962 			}
   1963 			ACQUIRE_GLOBAL_LOCKS();
   1964 			bridge_enqueue(sc, dst_if, mc, 1);
   1965 			RELEASE_GLOBAL_LOCKS();
   1966 		}
   1967 
   1968 		if (bmcast) {
   1969 			mc = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
   1970 			if (mc == NULL) {
   1971 				sc->sc_if.if_oerrors++;
   1972 				goto next;
   1973 			}
   1974 
   1975 			m_set_rcvif(mc, dst_if);
   1976 			mc->m_flags &= ~M_PROMISC;
   1977 
   1978 			ACQUIRE_GLOBAL_LOCKS();
   1979 			ether_input(dst_if, mc);
   1980 			RELEASE_GLOBAL_LOCKS();
   1981 		}
   1982 next:
   1983 		BRIDGE_PSZ_RENTER(s);
   1984 		bridge_release_member(sc, bif, &psref);
   1985 	}
   1986 	BRIDGE_PSZ_REXIT(s);
   1987 
   1988 	m_freem(m);
   1989 }
   1990 
   1991 static int
   1992 bridge_rtalloc(struct bridge_softc *sc, const uint8_t *dst,
   1993     struct bridge_rtnode **brtp)
   1994 {
   1995 	struct bridge_rtnode *brt;
   1996 	int error;
   1997 
   1998 	if (sc->sc_brtcnt >= sc->sc_brtmax)
   1999 		return ENOSPC;
   2000 
   2001 	/*
   2002 	 * Allocate a new bridge forwarding node, and
   2003 	 * initialize the expiration time and Ethernet
   2004 	 * address.
   2005 	 */
   2006 	brt = pool_get(&bridge_rtnode_pool, PR_NOWAIT);
   2007 	if (brt == NULL)
   2008 		return ENOMEM;
   2009 
   2010 	memset(brt, 0, sizeof(*brt));
   2011 	brt->brt_expire = time_uptime + sc->sc_brttimeout;
   2012 	brt->brt_flags = IFBAF_DYNAMIC;
   2013 	memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
   2014 
   2015 	BRIDGE_RT_LOCK(sc);
   2016 	error = bridge_rtnode_insert(sc, brt);
   2017 	BRIDGE_RT_UNLOCK(sc);
   2018 
   2019 	if (error != 0) {
   2020 		pool_put(&bridge_rtnode_pool, brt);
   2021 		return error;
   2022 	}
   2023 
   2024 	*brtp = brt;
   2025 	return 0;
   2026 }
   2027 
   2028 /*
   2029  * bridge_rtupdate:
   2030  *
   2031  *	Add a bridge routing entry.
   2032  */
   2033 static int
   2034 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
   2035     struct ifnet *dst_if, int setflags, uint8_t flags)
   2036 {
   2037 	struct bridge_rtnode *brt;
   2038 	int s;
   2039 
   2040 again:
   2041 	/*
   2042 	 * A route for this destination might already exist.  If so,
   2043 	 * update it, otherwise create a new one.
   2044 	 */
   2045 	BRIDGE_RT_RENTER(s);
   2046 	brt = bridge_rtnode_lookup(sc, dst);
   2047 
   2048 	if (brt != NULL) {
   2049 		brt->brt_ifp = dst_if;
   2050 		if (setflags) {
   2051 			brt->brt_flags = flags;
   2052 			if (flags & IFBAF_STATIC)
   2053 				brt->brt_expire = 0;
   2054 			else
   2055 				brt->brt_expire = time_uptime + sc->sc_brttimeout;
   2056 		} else {
   2057 			if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
   2058 				brt->brt_expire = time_uptime + sc->sc_brttimeout;
   2059 		}
   2060 	}
   2061 	BRIDGE_RT_REXIT(s);
   2062 
   2063 	if (brt == NULL) {
   2064 		int r;
   2065 
   2066 		r = bridge_rtalloc(sc, dst, &brt);
   2067 		if (r != 0)
   2068 			return r;
   2069 		goto again;
   2070 	}
   2071 
   2072 	return 0;
   2073 }
   2074 
   2075 /*
   2076  * bridge_rtlookup:
   2077  *
   2078  *	Lookup the destination interface for an address.
   2079  */
   2080 static struct ifnet *
   2081 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr)
   2082 {
   2083 	struct bridge_rtnode *brt;
   2084 	struct ifnet *ifs = NULL;
   2085 	int s;
   2086 
   2087 	BRIDGE_RT_RENTER(s);
   2088 	brt = bridge_rtnode_lookup(sc, addr);
   2089 	if (brt != NULL)
   2090 		ifs = brt->brt_ifp;
   2091 	BRIDGE_RT_REXIT(s);
   2092 
   2093 	return ifs;
   2094 }
   2095 
   2096 typedef bool (*bridge_iterate_cb_t)
   2097     (struct bridge_softc *, struct bridge_rtnode *, bool *, void *);
   2098 
   2099 /*
   2100  * bridge_rtlist_iterate_remove:
   2101  *
   2102  *	It iterates on sc->sc_rtlist and removes rtnodes of it which func
   2103  *	callback judges to remove. Removals of rtnodes are done in a manner
   2104  *	of pserialize. To this end, all kmem_* operations are placed out of
   2105  *	mutexes.
   2106  */
   2107 static void
   2108 bridge_rtlist_iterate_remove(struct bridge_softc *sc, bridge_iterate_cb_t func, void *arg)
   2109 {
   2110 	struct bridge_rtnode *brt, *nbrt;
   2111 	struct bridge_rtnode **brt_list;
   2112 	int i, count;
   2113 
   2114 retry:
   2115 	count = sc->sc_brtcnt;
   2116 	if (count == 0)
   2117 		return;
   2118 	brt_list = kmem_alloc(sizeof(*brt_list) * count, KM_SLEEP);
   2119 
   2120 	BRIDGE_RT_LOCK(sc);
   2121 	if (__predict_false(sc->sc_brtcnt > count)) {
   2122 		/* The rtnodes increased, we need more memory */
   2123 		BRIDGE_RT_UNLOCK(sc);
   2124 		kmem_free(brt_list, sizeof(*brt_list) * count);
   2125 		goto retry;
   2126 	}
   2127 
   2128 	i = 0;
   2129 	LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
   2130 		bool need_break = false;
   2131 		if (func(sc, brt, &need_break, arg)) {
   2132 			bridge_rtnode_remove(sc, brt);
   2133 			brt_list[i++] = brt;
   2134 		}
   2135 		if (need_break)
   2136 			break;
   2137 	}
   2138 
   2139 	if (i > 0)
   2140 		BRIDGE_RT_PSZ_PERFORM(sc);
   2141 	BRIDGE_RT_UNLOCK(sc);
   2142 
   2143 	while (--i >= 0)
   2144 		bridge_rtnode_destroy(brt_list[i]);
   2145 
   2146 	kmem_free(brt_list, sizeof(*brt_list) * count);
   2147 }
   2148 
   2149 static bool
   2150 bridge_rttrim0_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
   2151     bool *need_break, void *arg)
   2152 {
   2153 	if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
   2154 		/* Take into account of the subsequent removal */
   2155 		if ((sc->sc_brtcnt - 1) <= sc->sc_brtmax)
   2156 			*need_break = true;
   2157 		return true;
   2158 	} else
   2159 		return false;
   2160 }
   2161 
   2162 static void
   2163 bridge_rttrim0(struct bridge_softc *sc)
   2164 {
   2165 	bridge_rtlist_iterate_remove(sc, bridge_rttrim0_cb, NULL);
   2166 }
   2167 
   2168 /*
   2169  * bridge_rttrim:
   2170  *
   2171  *	Trim the routine table so that we have a number
   2172  *	of routing entries less than or equal to the
   2173  *	maximum number.
   2174  */
   2175 static void
   2176 bridge_rttrim(struct bridge_softc *sc)
   2177 {
   2178 
   2179 	/* Make sure we actually need to do this. */
   2180 	if (sc->sc_brtcnt <= sc->sc_brtmax)
   2181 		return;
   2182 
   2183 	/* Force an aging cycle; this might trim enough addresses. */
   2184 	bridge_rtage(sc);
   2185 	if (sc->sc_brtcnt <= sc->sc_brtmax)
   2186 		return;
   2187 
   2188 	bridge_rttrim0(sc);
   2189 
   2190 	return;
   2191 }
   2192 
   2193 /*
   2194  * bridge_timer:
   2195  *
   2196  *	Aging timer for the bridge.
   2197  */
   2198 static void
   2199 bridge_timer(void *arg)
   2200 {
   2201 	struct bridge_softc *sc = arg;
   2202 
   2203 	workqueue_enqueue(sc->sc_rtage_wq, &sc->sc_rtage_wk, NULL);
   2204 }
   2205 
   2206 static void
   2207 bridge_rtage_work(struct work *wk, void *arg)
   2208 {
   2209 	struct bridge_softc *sc = arg;
   2210 
   2211 	KASSERT(wk == &sc->sc_rtage_wk);
   2212 
   2213 	bridge_rtage(sc);
   2214 
   2215 	if (sc->sc_if.if_flags & IFF_RUNNING)
   2216 		callout_reset(&sc->sc_brcallout,
   2217 		    bridge_rtable_prune_period * hz, bridge_timer, sc);
   2218 }
   2219 
   2220 static bool
   2221 bridge_rtage_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
   2222     bool *need_break, void *arg)
   2223 {
   2224 	if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
   2225 	    time_uptime >= brt->brt_expire)
   2226 		return true;
   2227 	else
   2228 		return false;
   2229 }
   2230 
   2231 /*
   2232  * bridge_rtage:
   2233  *
   2234  *	Perform an aging cycle.
   2235  */
   2236 static void
   2237 bridge_rtage(struct bridge_softc *sc)
   2238 {
   2239 	bridge_rtlist_iterate_remove(sc, bridge_rtage_cb, NULL);
   2240 }
   2241 
   2242 
   2243 static bool
   2244 bridge_rtflush_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
   2245     bool *need_break, void *arg)
   2246 {
   2247 	int full = *(int*)arg;
   2248 
   2249 	if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
   2250 		return true;
   2251 	else
   2252 		return false;
   2253 }
   2254 
   2255 /*
   2256  * bridge_rtflush:
   2257  *
   2258  *	Remove all dynamic addresses from the bridge.
   2259  */
   2260 static void
   2261 bridge_rtflush(struct bridge_softc *sc, int full)
   2262 {
   2263 	bridge_rtlist_iterate_remove(sc, bridge_rtflush_cb, &full);
   2264 }
   2265 
   2266 /*
   2267  * bridge_rtdaddr:
   2268  *
   2269  *	Remove an address from the table.
   2270  */
   2271 static int
   2272 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr)
   2273 {
   2274 	struct bridge_rtnode *brt;
   2275 
   2276 	BRIDGE_RT_LOCK(sc);
   2277 	if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL) {
   2278 		BRIDGE_RT_UNLOCK(sc);
   2279 		return ENOENT;
   2280 	}
   2281 	bridge_rtnode_remove(sc, brt);
   2282 	BRIDGE_RT_PSZ_PERFORM(sc);
   2283 	BRIDGE_RT_UNLOCK(sc);
   2284 
   2285 	bridge_rtnode_destroy(brt);
   2286 
   2287 	return 0;
   2288 }
   2289 
   2290 /*
   2291  * bridge_rtdelete:
   2292  *
   2293  *	Delete routes to a speicifc member interface.
   2294  */
   2295 static void
   2296 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp)
   2297 {
   2298 	struct bridge_rtnode *brt;
   2299 
   2300 	BRIDGE_RT_LOCK(sc);
   2301 	LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
   2302 		if (brt->brt_ifp == ifp)
   2303 			break;
   2304 	}
   2305 	if (brt == NULL) {
   2306 		BRIDGE_RT_UNLOCK(sc);
   2307 		return;
   2308 	}
   2309 	bridge_rtnode_remove(sc, brt);
   2310 	BRIDGE_RT_PSZ_PERFORM(sc);
   2311 	BRIDGE_RT_UNLOCK(sc);
   2312 
   2313 	bridge_rtnode_destroy(brt);
   2314 }
   2315 
   2316 /*
   2317  * bridge_rtable_init:
   2318  *
   2319  *	Initialize the route table for this bridge.
   2320  */
   2321 static void
   2322 bridge_rtable_init(struct bridge_softc *sc)
   2323 {
   2324 	int i;
   2325 
   2326 	sc->sc_rthash = kmem_alloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
   2327 	    KM_SLEEP);
   2328 
   2329 	for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
   2330 		LIST_INIT(&sc->sc_rthash[i]);
   2331 
   2332 	sc->sc_rthash_key = cprng_fast32();
   2333 
   2334 	LIST_INIT(&sc->sc_rtlist);
   2335 
   2336 	sc->sc_rtlist_psz = pserialize_create();
   2337 	sc->sc_rtlist_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_SOFTNET);
   2338 }
   2339 
   2340 /*
   2341  * bridge_rtable_fini:
   2342  *
   2343  *	Deconstruct the route table for this bridge.
   2344  */
   2345 static void
   2346 bridge_rtable_fini(struct bridge_softc *sc)
   2347 {
   2348 
   2349 	kmem_free(sc->sc_rthash, sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE);
   2350 	if (sc->sc_rtlist_lock)
   2351 		mutex_obj_free(sc->sc_rtlist_lock);
   2352 	if (sc->sc_rtlist_psz)
   2353 		pserialize_destroy(sc->sc_rtlist_psz);
   2354 }
   2355 
   2356 /*
   2357  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
   2358  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
   2359  */
   2360 #define	mix(a, b, c)							\
   2361 do {									\
   2362 	a -= b; a -= c; a ^= (c >> 13);					\
   2363 	b -= c; b -= a; b ^= (a << 8);					\
   2364 	c -= a; c -= b; c ^= (b >> 13);					\
   2365 	a -= b; a -= c; a ^= (c >> 12);					\
   2366 	b -= c; b -= a; b ^= (a << 16);					\
   2367 	c -= a; c -= b; c ^= (b >> 5);					\
   2368 	a -= b; a -= c; a ^= (c >> 3);					\
   2369 	b -= c; b -= a; b ^= (a << 10);					\
   2370 	c -= a; c -= b; c ^= (b >> 15);					\
   2371 } while (/*CONSTCOND*/0)
   2372 
   2373 static inline uint32_t
   2374 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
   2375 {
   2376 	uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
   2377 
   2378 	b += addr[5] << 8;
   2379 	b += addr[4];
   2380 	a += addr[3] << 24;
   2381 	a += addr[2] << 16;
   2382 	a += addr[1] << 8;
   2383 	a += addr[0];
   2384 
   2385 	mix(a, b, c);
   2386 
   2387 	return (c & BRIDGE_RTHASH_MASK);
   2388 }
   2389 
   2390 #undef mix
   2391 
   2392 /*
   2393  * bridge_rtnode_lookup:
   2394  *
   2395  *	Look up a bridge route node for the specified destination.
   2396  */
   2397 static struct bridge_rtnode *
   2398 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr)
   2399 {
   2400 	struct bridge_rtnode *brt;
   2401 	uint32_t hash;
   2402 	int dir;
   2403 
   2404 	hash = bridge_rthash(sc, addr);
   2405 	LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
   2406 		dir = memcmp(addr, brt->brt_addr, ETHER_ADDR_LEN);
   2407 		if (dir == 0)
   2408 			return brt;
   2409 		if (dir > 0)
   2410 			return NULL;
   2411 	}
   2412 
   2413 	return NULL;
   2414 }
   2415 
   2416 /*
   2417  * bridge_rtnode_insert:
   2418  *
   2419  *	Insert the specified bridge node into the route table.  We
   2420  *	assume the entry is not already in the table.
   2421  */
   2422 static int
   2423 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
   2424 {
   2425 	struct bridge_rtnode *lbrt;
   2426 	uint32_t hash;
   2427 	int dir;
   2428 
   2429 	KASSERT(BRIDGE_RT_LOCKED(sc));
   2430 
   2431 	hash = bridge_rthash(sc, brt->brt_addr);
   2432 
   2433 	lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
   2434 	if (lbrt == NULL) {
   2435 		LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
   2436 		goto out;
   2437 	}
   2438 
   2439 	do {
   2440 		dir = memcmp(brt->brt_addr, lbrt->brt_addr, ETHER_ADDR_LEN);
   2441 		if (dir == 0)
   2442 			return EEXIST;
   2443 		if (dir > 0) {
   2444 			LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
   2445 			goto out;
   2446 		}
   2447 		if (LIST_NEXT(lbrt, brt_hash) == NULL) {
   2448 			LIST_INSERT_AFTER(lbrt, brt, brt_hash);
   2449 			goto out;
   2450 		}
   2451 		lbrt = LIST_NEXT(lbrt, brt_hash);
   2452 	} while (lbrt != NULL);
   2453 
   2454 #ifdef DIAGNOSTIC
   2455 	panic("bridge_rtnode_insert: impossible");
   2456 #endif
   2457 
   2458  out:
   2459 	LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
   2460 	sc->sc_brtcnt++;
   2461 
   2462 	return 0;
   2463 }
   2464 
   2465 /*
   2466  * bridge_rtnode_remove:
   2467  *
   2468  *	Remove a bridge rtnode from the rthash and the rtlist of a bridge.
   2469  */
   2470 static void
   2471 bridge_rtnode_remove(struct bridge_softc *sc, struct bridge_rtnode *brt)
   2472 {
   2473 
   2474 	KASSERT(BRIDGE_RT_LOCKED(sc));
   2475 
   2476 	LIST_REMOVE(brt, brt_hash);
   2477 	LIST_REMOVE(brt, brt_list);
   2478 	sc->sc_brtcnt--;
   2479 }
   2480 
   2481 /*
   2482  * bridge_rtnode_destroy:
   2483  *
   2484  *	Destroy a bridge rtnode.
   2485  */
   2486 static void
   2487 bridge_rtnode_destroy(struct bridge_rtnode *brt)
   2488 {
   2489 
   2490 	pool_put(&bridge_rtnode_pool, brt);
   2491 }
   2492 
   2493 #if defined(BRIDGE_IPF)
   2494 extern pfil_head_t *inet_pfil_hook;                 /* XXX */
   2495 extern pfil_head_t *inet6_pfil_hook;                /* XXX */
   2496 
   2497 /*
   2498  * Send bridge packets through IPF if they are one of the types IPF can deal
   2499  * with, or if they are ARP or REVARP.  (IPF will pass ARP and REVARP without
   2500  * question.)
   2501  */
   2502 static int
   2503 bridge_ipf(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
   2504 {
   2505 	int snap, error;
   2506 	struct ether_header *eh1, eh2;
   2507 	struct llc llc1;
   2508 	uint16_t ether_type;
   2509 
   2510 	snap = 0;
   2511 	error = -1;	/* Default error if not error == 0 */
   2512 	eh1 = mtod(*mp, struct ether_header *);
   2513 	ether_type = ntohs(eh1->ether_type);
   2514 
   2515 	/*
   2516 	 * Check for SNAP/LLC.
   2517 	 */
   2518 	if (ether_type < ETHERMTU) {
   2519 		struct llc *llc2 = (struct llc *)(eh1 + 1);
   2520 
   2521 		if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
   2522 		    llc2->llc_dsap == LLC_SNAP_LSAP &&
   2523 		    llc2->llc_ssap == LLC_SNAP_LSAP &&
   2524 		    llc2->llc_control == LLC_UI) {
   2525 			ether_type = htons(llc2->llc_un.type_snap.ether_type);
   2526 			snap = 1;
   2527 		}
   2528 	}
   2529 
   2530 	/*
   2531 	 * If we're trying to filter bridge traffic, don't look at anything
   2532 	 * other than IP and ARP traffic.  If the filter doesn't understand
   2533 	 * IPv6, don't allow IPv6 through the bridge either.  This is lame
   2534 	 * since if we really wanted, say, an AppleTalk filter, we are hosed,
   2535 	 * but of course we don't have an AppleTalk filter to begin with.
   2536 	 * (Note that since IPF doesn't understand ARP it will pass *ALL*
   2537 	 * ARP traffic.)
   2538 	 */
   2539 	switch (ether_type) {
   2540 		case ETHERTYPE_ARP:
   2541 		case ETHERTYPE_REVARP:
   2542 			return 0; /* Automatically pass */
   2543 		case ETHERTYPE_IP:
   2544 # ifdef INET6
   2545 		case ETHERTYPE_IPV6:
   2546 # endif /* INET6 */
   2547 			break;
   2548 		default:
   2549 			goto bad;
   2550 	}
   2551 
   2552 	/* Strip off the Ethernet header and keep a copy. */
   2553 	m_copydata(*mp, 0, ETHER_HDR_LEN, (void *) &eh2);
   2554 	m_adj(*mp, ETHER_HDR_LEN);
   2555 
   2556 	/* Strip off snap header, if present */
   2557 	if (snap) {
   2558 		m_copydata(*mp, 0, sizeof(struct llc), (void *) &llc1);
   2559 		m_adj(*mp, sizeof(struct llc));
   2560 	}
   2561 
   2562 	/*
   2563 	 * Check basic packet sanity and run IPF through pfil.
   2564 	 */
   2565 	KASSERT(!cpu_intr_p());
   2566 	switch (ether_type)
   2567 	{
   2568 	case ETHERTYPE_IP :
   2569 		error = bridge_ip_checkbasic(mp);
   2570 		if (error == 0)
   2571 			error = pfil_run_hooks(inet_pfil_hook, mp, ifp, dir);
   2572 		break;
   2573 # ifdef INET6
   2574 	case ETHERTYPE_IPV6 :
   2575 		error = bridge_ip6_checkbasic(mp);
   2576 		if (error == 0)
   2577 			error = pfil_run_hooks(inet6_pfil_hook, mp, ifp, dir);
   2578 		break;
   2579 # endif
   2580 	default :
   2581 		error = 0;
   2582 		break;
   2583 	}
   2584 
   2585 	if (*mp == NULL)
   2586 		return error;
   2587 	if (error != 0)
   2588 		goto bad;
   2589 
   2590 	error = -1;
   2591 
   2592 	/*
   2593 	 * Finally, put everything back the way it was and return
   2594 	 */
   2595 	if (snap) {
   2596 		M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
   2597 		if (*mp == NULL)
   2598 			return error;
   2599 		bcopy(&llc1, mtod(*mp, void *), sizeof(struct llc));
   2600 	}
   2601 
   2602 	M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
   2603 	if (*mp == NULL)
   2604 		return error;
   2605 	bcopy(&eh2, mtod(*mp, void *), ETHER_HDR_LEN);
   2606 
   2607 	return 0;
   2608 
   2609     bad:
   2610 	m_freem(*mp);
   2611 	*mp = NULL;
   2612 	return error;
   2613 }
   2614 
   2615 /*
   2616  * Perform basic checks on header size since
   2617  * IPF assumes ip_input has already processed
   2618  * it for it.  Cut-and-pasted from ip_input.c.
   2619  * Given how simple the IPv6 version is,
   2620  * does the IPv4 version really need to be
   2621  * this complicated?
   2622  *
   2623  * XXX Should we update ipstat here, or not?
   2624  * XXX Right now we update ipstat but not
   2625  * XXX csum_counter.
   2626  */
   2627 static int
   2628 bridge_ip_checkbasic(struct mbuf **mp)
   2629 {
   2630 	struct mbuf *m = *mp;
   2631 	struct ip *ip;
   2632 	int len, hlen;
   2633 
   2634 	if (*mp == NULL)
   2635 		return -1;
   2636 
   2637 	if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
   2638 		if ((m = m_copyup(m, sizeof(struct ip),
   2639 			(max_linkhdr + 3) & ~3)) == NULL) {
   2640 			/* XXXJRT new stat, please */
   2641 			ip_statinc(IP_STAT_TOOSMALL);
   2642 			goto bad;
   2643 		}
   2644 	} else if (__predict_false(m->m_len < sizeof (struct ip))) {
   2645 		if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
   2646 			ip_statinc(IP_STAT_TOOSMALL);
   2647 			goto bad;
   2648 		}
   2649 	}
   2650 	ip = mtod(m, struct ip *);
   2651 	if (ip == NULL) goto bad;
   2652 
   2653 	if (ip->ip_v != IPVERSION) {
   2654 		ip_statinc(IP_STAT_BADVERS);
   2655 		goto bad;
   2656 	}
   2657 	hlen = ip->ip_hl << 2;
   2658 	if (hlen < sizeof(struct ip)) { /* minimum header length */
   2659 		ip_statinc(IP_STAT_BADHLEN);
   2660 		goto bad;
   2661 	}
   2662 	if (hlen > m->m_len) {
   2663 		if ((m = m_pullup(m, hlen)) == 0) {
   2664 			ip_statinc(IP_STAT_BADHLEN);
   2665 			goto bad;
   2666 		}
   2667 		ip = mtod(m, struct ip *);
   2668 		if (ip == NULL) goto bad;
   2669 	}
   2670 
   2671 	switch (m->m_pkthdr.csum_flags &
   2672 	        ((m_get_rcvif_NOMPSAFE(m)->if_csum_flags_rx & M_CSUM_IPv4) |
   2673 	         M_CSUM_IPv4_BAD)) {
   2674 	case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
   2675 		/* INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); */
   2676 		goto bad;
   2677 
   2678 	case M_CSUM_IPv4:
   2679 		/* Checksum was okay. */
   2680 		/* INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); */
   2681 		break;
   2682 
   2683 	default:
   2684 		/* Must compute it ourselves. */
   2685 		/* INET_CSUM_COUNTER_INCR(&ip_swcsum); */
   2686 		if (in_cksum(m, hlen) != 0)
   2687 			goto bad;
   2688 		break;
   2689 	}
   2690 
   2691 	/* Retrieve the packet length. */
   2692 	len = ntohs(ip->ip_len);
   2693 
   2694 	/*
   2695 	 * Check for additional length bogosity
   2696 	 */
   2697 	if (len < hlen) {
   2698 		ip_statinc(IP_STAT_BADLEN);
   2699 		goto bad;
   2700 	}
   2701 
   2702 	/*
   2703 	 * Check that the amount of data in the buffers
   2704 	 * is as at least much as the IP header would have us expect.
   2705 	 * Drop packet if shorter than we expect.
   2706 	 */
   2707 	if (m->m_pkthdr.len < len) {
   2708 		ip_statinc(IP_STAT_TOOSHORT);
   2709 		goto bad;
   2710 	}
   2711 
   2712 	/* Checks out, proceed */
   2713 	*mp = m;
   2714 	return 0;
   2715 
   2716     bad:
   2717 	*mp = m;
   2718 	return -1;
   2719 }
   2720 
   2721 # ifdef INET6
   2722 /*
   2723  * Same as above, but for IPv6.
   2724  * Cut-and-pasted from ip6_input.c.
   2725  * XXX Should we update ip6stat, or not?
   2726  */
   2727 static int
   2728 bridge_ip6_checkbasic(struct mbuf **mp)
   2729 {
   2730 	struct mbuf *m = *mp;
   2731 	struct ip6_hdr *ip6;
   2732 
   2733 	/*
   2734 	 * If the IPv6 header is not aligned, slurp it up into a new
   2735 	 * mbuf with space for link headers, in the event we forward
   2736 	 * it.  Otherwise, if it is aligned, make sure the entire base
   2737 	 * IPv6 header is in the first mbuf of the chain.
   2738 	 */
   2739 	if (IP6_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
   2740 		struct ifnet *inifp = m_get_rcvif_NOMPSAFE(m);
   2741 		if ((m = m_copyup(m, sizeof(struct ip6_hdr),
   2742 		                  (max_linkhdr + 3) & ~3)) == NULL) {
   2743 			/* XXXJRT new stat, please */
   2744 			ip6_statinc(IP6_STAT_TOOSMALL);
   2745 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
   2746 			goto bad;
   2747 		}
   2748 	} else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
   2749 		struct ifnet *inifp = m_get_rcvif_NOMPSAFE(m);
   2750 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
   2751 			ip6_statinc(IP6_STAT_TOOSMALL);
   2752 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
   2753 			goto bad;
   2754 		}
   2755 	}
   2756 
   2757 	ip6 = mtod(m, struct ip6_hdr *);
   2758 
   2759 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
   2760 		ip6_statinc(IP6_STAT_BADVERS);
   2761 		in6_ifstat_inc(m_get_rcvif_NOMPSAFE(m), ifs6_in_hdrerr);
   2762 		goto bad;
   2763 	}
   2764 
   2765 	/* Checks out, proceed */
   2766 	*mp = m;
   2767 	return 0;
   2768 
   2769     bad:
   2770 	*mp = m;
   2771 	return -1;
   2772 }
   2773 # endif /* INET6 */
   2774 #endif /* BRIDGE_IPF */
   2775