Home | History | Annotate | Line # | Download | only in net
if_bridge.c revision 1.84
      1 /*	$NetBSD: if_bridge.c,v 1.84 2014/06/23 06:27:58 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 1.84 2014/06/23 06:27:58 ozaki-r Exp $");
     84 
     85 #ifdef _KERNEL_OPT
     86 #include "opt_bridge_ipf.h"
     87 #include "opt_inet.h"
     88 #endif /* _KERNEL_OPT */
     89 
     90 #include <sys/param.h>
     91 #include <sys/kernel.h>
     92 #include <sys/mbuf.h>
     93 #include <sys/queue.h>
     94 #include <sys/socket.h>
     95 #include <sys/socketvar.h> /* for softnet_lock */
     96 #include <sys/sockio.h>
     97 #include <sys/systm.h>
     98 #include <sys/proc.h>
     99 #include <sys/pool.h>
    100 #include <sys/kauth.h>
    101 #include <sys/cpu.h>
    102 #include <sys/cprng.h>
    103 
    104 #include <net/bpf.h>
    105 #include <net/if.h>
    106 #include <net/if_dl.h>
    107 #include <net/if_types.h>
    108 #include <net/if_llc.h>
    109 #include <net/pktqueue.h>
    110 
    111 #include <net/if_ether.h>
    112 #include <net/if_bridgevar.h>
    113 
    114 #if defined(BRIDGE_IPF)
    115 /* Used for bridge_ip[6]_checkbasic */
    116 #include <netinet/in.h>
    117 #include <netinet/in_systm.h>
    118 #include <netinet/ip.h>
    119 #include <netinet/ip_var.h>
    120 #include <netinet/ip_private.h>		/* XXX */
    121 
    122 #include <netinet/ip6.h>
    123 #include <netinet6/in6_var.h>
    124 #include <netinet6/ip6_var.h>
    125 #include <netinet6/ip6_private.h>	/* XXX */
    126 #endif /* BRIDGE_IPF */
    127 
    128 /*
    129  * Size of the route hash table.  Must be a power of two.
    130  */
    131 #ifndef BRIDGE_RTHASH_SIZE
    132 #define	BRIDGE_RTHASH_SIZE		1024
    133 #endif
    134 
    135 #define	BRIDGE_RTHASH_MASK		(BRIDGE_RTHASH_SIZE - 1)
    136 
    137 #include "carp.h"
    138 #if NCARP > 0
    139 #include <netinet/in.h>
    140 #include <netinet/in_var.h>
    141 #include <netinet/ip_carp.h>
    142 #endif
    143 
    144 /*
    145  * Maximum number of addresses to cache.
    146  */
    147 #ifndef BRIDGE_RTABLE_MAX
    148 #define	BRIDGE_RTABLE_MAX		100
    149 #endif
    150 
    151 /*
    152  * Spanning tree defaults.
    153  */
    154 #define	BSTP_DEFAULT_MAX_AGE		(20 * 256)
    155 #define	BSTP_DEFAULT_HELLO_TIME		(2 * 256)
    156 #define	BSTP_DEFAULT_FORWARD_DELAY	(15 * 256)
    157 #define	BSTP_DEFAULT_HOLD_TIME		(1 * 256)
    158 #define	BSTP_DEFAULT_BRIDGE_PRIORITY	0x8000
    159 #define	BSTP_DEFAULT_PORT_PRIORITY	0x80
    160 #define	BSTP_DEFAULT_PATH_COST		55
    161 
    162 /*
    163  * Timeout (in seconds) for entries learned dynamically.
    164  */
    165 #ifndef BRIDGE_RTABLE_TIMEOUT
    166 #define	BRIDGE_RTABLE_TIMEOUT		(20 * 60)	/* same as ARP */
    167 #endif
    168 
    169 /*
    170  * Number of seconds between walks of the route list.
    171  */
    172 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
    173 #define	BRIDGE_RTABLE_PRUNE_PERIOD	(5 * 60)
    174 #endif
    175 
    176 int	bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
    177 
    178 static struct pool bridge_rtnode_pool;
    179 
    180 void	bridgeattach(int);
    181 
    182 static int	bridge_clone_create(struct if_clone *, int);
    183 static int	bridge_clone_destroy(struct ifnet *);
    184 
    185 static int	bridge_ioctl(struct ifnet *, u_long, void *);
    186 static int	bridge_init(struct ifnet *);
    187 static void	bridge_stop(struct ifnet *, int);
    188 static void	bridge_start(struct ifnet *);
    189 
    190 static void	bridge_input(struct ifnet *, struct mbuf *);
    191 static void	bridge_forward(void *);
    192 
    193 static void	bridge_timer(void *);
    194 
    195 static void	bridge_broadcast(struct bridge_softc *, struct ifnet *,
    196 				 struct mbuf *);
    197 
    198 static int	bridge_rtupdate(struct bridge_softc *, const uint8_t *,
    199 				struct ifnet *, int, uint8_t);
    200 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *);
    201 static void	bridge_rttrim(struct bridge_softc *);
    202 static void	bridge_rtage(struct bridge_softc *);
    203 static void	bridge_rtflush(struct bridge_softc *, int);
    204 static int	bridge_rtdaddr(struct bridge_softc *, const uint8_t *);
    205 static void	bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp);
    206 
    207 static int	bridge_rtable_init(struct bridge_softc *);
    208 static void	bridge_rtable_fini(struct bridge_softc *);
    209 
    210 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
    211 						  const uint8_t *);
    212 static int	bridge_rtnode_insert(struct bridge_softc *,
    213 				     struct bridge_rtnode *);
    214 static void	bridge_rtnode_destroy(struct bridge_softc *,
    215 				      struct bridge_rtnode *);
    216 
    217 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
    218 						  const char *name);
    219 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
    220 						     struct ifnet *ifp);
    221 static void	bridge_delete_member(struct bridge_softc *,
    222 				     struct bridge_iflist *);
    223 
    224 static int	bridge_ioctl_add(struct bridge_softc *, void *);
    225 static int	bridge_ioctl_del(struct bridge_softc *, void *);
    226 static int	bridge_ioctl_gifflags(struct bridge_softc *, void *);
    227 static int	bridge_ioctl_sifflags(struct bridge_softc *, void *);
    228 static int	bridge_ioctl_scache(struct bridge_softc *, void *);
    229 static int	bridge_ioctl_gcache(struct bridge_softc *, void *);
    230 static int	bridge_ioctl_gifs(struct bridge_softc *, void *);
    231 static int	bridge_ioctl_rts(struct bridge_softc *, void *);
    232 static int	bridge_ioctl_saddr(struct bridge_softc *, void *);
    233 static int	bridge_ioctl_sto(struct bridge_softc *, void *);
    234 static int	bridge_ioctl_gto(struct bridge_softc *, void *);
    235 static int	bridge_ioctl_daddr(struct bridge_softc *, void *);
    236 static int	bridge_ioctl_flush(struct bridge_softc *, void *);
    237 static int	bridge_ioctl_gpri(struct bridge_softc *, void *);
    238 static int	bridge_ioctl_spri(struct bridge_softc *, void *);
    239 static int	bridge_ioctl_ght(struct bridge_softc *, void *);
    240 static int	bridge_ioctl_sht(struct bridge_softc *, void *);
    241 static int	bridge_ioctl_gfd(struct bridge_softc *, void *);
    242 static int	bridge_ioctl_sfd(struct bridge_softc *, void *);
    243 static int	bridge_ioctl_gma(struct bridge_softc *, void *);
    244 static int	bridge_ioctl_sma(struct bridge_softc *, void *);
    245 static int	bridge_ioctl_sifprio(struct bridge_softc *, void *);
    246 static int	bridge_ioctl_sifcost(struct bridge_softc *, void *);
    247 #if defined(BRIDGE_IPF)
    248 static int	bridge_ioctl_gfilt(struct bridge_softc *, void *);
    249 static int	bridge_ioctl_sfilt(struct bridge_softc *, void *);
    250 static int	bridge_ipf(void *, struct mbuf **, struct ifnet *, int);
    251 static int	bridge_ip_checkbasic(struct mbuf **mp);
    252 # ifdef INET6
    253 static int	bridge_ip6_checkbasic(struct mbuf **mp);
    254 # endif /* INET6 */
    255 #endif /* BRIDGE_IPF */
    256 
    257 static void bridge_sysctl_fwdq_setup(struct sysctllog **clog,
    258     struct bridge_softc *sc);
    259 
    260 struct bridge_control {
    261 	int	(*bc_func)(struct bridge_softc *, void *);
    262 	int	bc_argsize;
    263 	int	bc_flags;
    264 };
    265 
    266 #define	BC_F_COPYIN		0x01	/* copy arguments in */
    267 #define	BC_F_COPYOUT		0x02	/* copy arguments out */
    268 #define	BC_F_SUSER		0x04	/* do super-user check */
    269 
    270 static const struct bridge_control bridge_control_table[] = {
    271 [BRDGADD] = {bridge_ioctl_add, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    272 [BRDGDEL] = {bridge_ioctl_del, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    273 
    274 [BRDGGIFFLGS] = {bridge_ioctl_gifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_COPYOUT},
    275 [BRDGSIFFLGS] = {bridge_ioctl_sifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    276 
    277 [BRDGSCACHE] = {bridge_ioctl_scache, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    278 [BRDGGCACHE] = {bridge_ioctl_gcache, sizeof(struct ifbrparam), BC_F_COPYOUT},
    279 
    280 [BRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_COPYIN|BC_F_COPYOUT},
    281 [BRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_COPYIN|BC_F_COPYOUT},
    282 
    283 [BRDGSADDR] = {bridge_ioctl_saddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
    284 
    285 [BRDGSTO] = {bridge_ioctl_sto, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    286 [BRDGGTO] = {bridge_ioctl_gto, sizeof(struct ifbrparam), BC_F_COPYOUT},
    287 
    288 [BRDGDADDR] = {bridge_ioctl_daddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
    289 
    290 [BRDGFLUSH] = {bridge_ioctl_flush, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    291 
    292 [BRDGGPRI] = {bridge_ioctl_gpri, sizeof(struct ifbrparam), BC_F_COPYOUT},
    293 [BRDGSPRI] = {bridge_ioctl_spri, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    294 
    295 [BRDGGHT] = {bridge_ioctl_ght, sizeof(struct ifbrparam), BC_F_COPYOUT},
    296 [BRDGSHT] = {bridge_ioctl_sht, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    297 
    298 [BRDGGFD] = {bridge_ioctl_gfd, sizeof(struct ifbrparam), BC_F_COPYOUT},
    299 [BRDGSFD] = {bridge_ioctl_sfd, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    300 
    301 [BRDGGMA] = {bridge_ioctl_gma, sizeof(struct ifbrparam), BC_F_COPYOUT},
    302 [BRDGSMA] = {bridge_ioctl_sma, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    303 
    304 [BRDGSIFPRIO] = {bridge_ioctl_sifprio, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    305 
    306 [BRDGSIFCOST] = {bridge_ioctl_sifcost, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    307 #if defined(BRIDGE_IPF)
    308 [BRDGGFILT] = {bridge_ioctl_gfilt, sizeof(struct ifbrparam), BC_F_COPYOUT},
    309 [BRDGSFILT] = {bridge_ioctl_sfilt, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    310 #endif /* BRIDGE_IPF */
    311 };
    312 static const int bridge_control_table_size = __arraycount(bridge_control_table);
    313 
    314 static LIST_HEAD(, bridge_softc) bridge_list;
    315 
    316 static struct if_clone bridge_cloner =
    317     IF_CLONE_INITIALIZER("bridge", bridge_clone_create, bridge_clone_destroy);
    318 
    319 /*
    320  * bridgeattach:
    321  *
    322  *	Pseudo-device attach routine.
    323  */
    324 void
    325 bridgeattach(int n)
    326 {
    327 
    328 	pool_init(&bridge_rtnode_pool, sizeof(struct bridge_rtnode),
    329 	    0, 0, 0, "brtpl", NULL, IPL_NET);
    330 
    331 	LIST_INIT(&bridge_list);
    332 	if_clone_attach(&bridge_cloner);
    333 }
    334 
    335 /*
    336  * bridge_clone_create:
    337  *
    338  *	Create a new bridge instance.
    339  */
    340 static int
    341 bridge_clone_create(struct if_clone *ifc, int unit)
    342 {
    343 	struct bridge_softc *sc;
    344 	struct ifnet *ifp;
    345 	int s;
    346 
    347 	sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
    348 	ifp = &sc->sc_if;
    349 
    350 	sc->sc_brtmax = BRIDGE_RTABLE_MAX;
    351 	sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
    352 	sc->sc_bridge_max_age = BSTP_DEFAULT_MAX_AGE;
    353 	sc->sc_bridge_hello_time = BSTP_DEFAULT_HELLO_TIME;
    354 	sc->sc_bridge_forward_delay = BSTP_DEFAULT_FORWARD_DELAY;
    355 	sc->sc_bridge_priority = BSTP_DEFAULT_BRIDGE_PRIORITY;
    356 	sc->sc_hold_time = BSTP_DEFAULT_HOLD_TIME;
    357 	sc->sc_filter_flags = 0;
    358 
    359 	/* Initialize our routing table. */
    360 	bridge_rtable_init(sc);
    361 
    362 	callout_init(&sc->sc_brcallout, 0);
    363 	callout_init(&sc->sc_bstpcallout, 0);
    364 
    365 	LIST_INIT(&sc->sc_iflist);
    366 
    367 	if_initname(ifp, ifc->ifc_name, unit);
    368 	ifp->if_softc = sc;
    369 	ifp->if_mtu = ETHERMTU;
    370 	ifp->if_ioctl = bridge_ioctl;
    371 	ifp->if_output = bridge_output;
    372 	ifp->if_start = bridge_start;
    373 	ifp->if_stop = bridge_stop;
    374 	ifp->if_init = bridge_init;
    375 	ifp->if_type = IFT_BRIDGE;
    376 	ifp->if_addrlen = 0;
    377 	ifp->if_dlt = DLT_EN10MB;
    378 	ifp->if_hdrlen = ETHER_HDR_LEN;
    379 	IFQ_SET_READY(&ifp->if_snd);
    380 
    381 	sc->sc_fwd_pktq = pktq_create(IFQ_MAXLEN, bridge_forward, sc);
    382 	KASSERT(sc->sc_fwd_pktq != NULL);
    383 
    384 	bridge_sysctl_fwdq_setup(&ifp->if_sysctl_log, sc);
    385 
    386 	if_attach(ifp);
    387 
    388 	if_alloc_sadl(ifp);
    389 
    390 	s = splnet();
    391 	LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
    392 	splx(s);
    393 
    394 	return (0);
    395 }
    396 
    397 /*
    398  * bridge_clone_destroy:
    399  *
    400  *	Destroy a bridge instance.
    401  */
    402 static int
    403 bridge_clone_destroy(struct ifnet *ifp)
    404 {
    405 	struct bridge_softc *sc = ifp->if_softc;
    406 	struct bridge_iflist *bif;
    407 	int s;
    408 
    409 	/* Must be called during IFF_RUNNING, i.e., before bridge_stop */
    410 	pktq_barrier(sc->sc_fwd_pktq);
    411 
    412 	s = splnet();
    413 
    414 	bridge_stop(ifp, 1);
    415 
    416 	while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
    417 		bridge_delete_member(sc, bif);
    418 
    419 	LIST_REMOVE(sc, sc_list);
    420 
    421 	splx(s);
    422 
    423 	if_detach(ifp);
    424 
    425 	/* Should be called after if_detach for safe */
    426 	pktq_flush(sc->sc_fwd_pktq);
    427 	pktq_destroy(sc->sc_fwd_pktq);
    428 
    429 	/* Tear down the routing table. */
    430 	bridge_rtable_fini(sc);
    431 
    432 	free(sc, M_DEVBUF);
    433 
    434 	return (0);
    435 }
    436 
    437 static int
    438 bridge_sysctl_fwdq_maxlen(SYSCTLFN_ARGS)
    439 {
    440 	struct sysctlnode node = *rnode;
    441 	const struct bridge_softc *sc =	node.sysctl_data;
    442 	return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), sc->sc_fwd_pktq);
    443 }
    444 
    445 #define	SYSCTL_BRIDGE_PKTQ(cn, c)					\
    446 	static int							\
    447 	bridge_sysctl_fwdq_##cn(SYSCTLFN_ARGS)				\
    448 	{								\
    449 		struct sysctlnode node = *rnode;			\
    450 		const struct bridge_softc *sc =	node.sysctl_data;	\
    451 		return sysctl_pktq_count(SYSCTLFN_CALL(rnode),		\
    452 					 sc->sc_fwd_pktq, c);		\
    453 	}
    454 
    455 SYSCTL_BRIDGE_PKTQ(items, PKTQ_NITEMS)
    456 SYSCTL_BRIDGE_PKTQ(drops, PKTQ_DROPS)
    457 
    458 static void
    459 bridge_sysctl_fwdq_setup(struct sysctllog **clog, struct bridge_softc *sc)
    460 {
    461 	const struct sysctlnode *cnode, *rnode;
    462 	sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL;
    463 	const char *ifname = sc->sc_if.if_xname;
    464 
    465 	len_func = bridge_sysctl_fwdq_items;
    466 	maxlen_func = bridge_sysctl_fwdq_maxlen;
    467 	drops_func = bridge_sysctl_fwdq_drops;
    468 
    469 	if (sysctl_createv(clog, 0, NULL, &rnode,
    470 			   CTLFLAG_PERMANENT,
    471 			   CTLTYPE_NODE, "interfaces",
    472 			   SYSCTL_DESCR("Per-interface controls"),
    473 			   NULL, 0, NULL, 0,
    474 			   CTL_NET, CTL_CREATE, CTL_EOL) != 0)
    475 		goto bad;
    476 
    477 	if (sysctl_createv(clog, 0, &rnode, &rnode,
    478 			   CTLFLAG_PERMANENT,
    479 			   CTLTYPE_NODE, ifname,
    480 			   SYSCTL_DESCR("Interface controls"),
    481 			   NULL, 0, NULL, 0,
    482 			   CTL_CREATE, CTL_EOL) != 0)
    483 		goto bad;
    484 
    485 	if (sysctl_createv(clog, 0, &rnode, &rnode,
    486 			   CTLFLAG_PERMANENT,
    487 			   CTLTYPE_NODE, "fwdq",
    488 			   SYSCTL_DESCR("Protocol input queue controls"),
    489 			   NULL, 0, NULL, 0,
    490 			   CTL_CREATE, CTL_EOL) != 0)
    491 		goto bad;
    492 
    493 	if (sysctl_createv(clog, 0, &rnode, &cnode,
    494 			   CTLFLAG_PERMANENT,
    495 			   CTLTYPE_INT, "len",
    496 			   SYSCTL_DESCR("Current forwarding queue length"),
    497 			   len_func, 0, (void *)sc, 0,
    498 			   CTL_CREATE, IFQCTL_LEN, CTL_EOL) != 0)
    499 		goto bad;
    500 
    501 	if (sysctl_createv(clog, 0, &rnode, &cnode,
    502 			   CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
    503 			   CTLTYPE_INT, "maxlen",
    504 			   SYSCTL_DESCR("Maximum allowed forwarding queue length"),
    505 			   maxlen_func, 0, (void *)sc, 0,
    506 			   CTL_CREATE, IFQCTL_MAXLEN, CTL_EOL) != 0)
    507 		goto bad;
    508 
    509 	if (sysctl_createv(clog, 0, &rnode, &cnode,
    510 			   CTLFLAG_PERMANENT,
    511 			   CTLTYPE_INT, "drops",
    512 			   SYSCTL_DESCR("Packets dropped due to full forwarding queue"),
    513 			   drops_func, 0, (void *)sc, 0,
    514 			   CTL_CREATE, IFQCTL_DROPS, CTL_EOL) != 0)
    515 		goto bad;
    516 
    517 	return;
    518 bad:
    519 	aprint_error("%s: could not attach sysctl nodes\n", ifname);
    520 	return;
    521 }
    522 
    523 /*
    524  * bridge_ioctl:
    525  *
    526  *	Handle a control request from the operator.
    527  */
    528 static int
    529 bridge_ioctl(struct ifnet *ifp, u_long cmd, void *data)
    530 {
    531 	struct bridge_softc *sc = ifp->if_softc;
    532 	struct lwp *l = curlwp;	/* XXX */
    533 	union {
    534 		struct ifbreq ifbreq;
    535 		struct ifbifconf ifbifconf;
    536 		struct ifbareq ifbareq;
    537 		struct ifbaconf ifbaconf;
    538 		struct ifbrparam ifbrparam;
    539 	} args;
    540 	struct ifdrv *ifd = (struct ifdrv *) data;
    541 	const struct bridge_control *bc = NULL; /* XXXGCC */
    542 	int s, error = 0;
    543 
    544 	/* Authorize command before calling splnet(). */
    545 	switch (cmd) {
    546 	case SIOCGDRVSPEC:
    547 	case SIOCSDRVSPEC:
    548 		if (ifd->ifd_cmd >= bridge_control_table_size) {
    549 			error = EINVAL;
    550 			return error;
    551 		}
    552 
    553 		bc = &bridge_control_table[ifd->ifd_cmd];
    554 
    555 		/* We only care about BC_F_SUSER at this point. */
    556 		if ((bc->bc_flags & BC_F_SUSER) == 0)
    557 			break;
    558 
    559 		error = kauth_authorize_network(l->l_cred,
    560 		    KAUTH_NETWORK_INTERFACE_BRIDGE,
    561 		    cmd == SIOCGDRVSPEC ?
    562 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_GETPRIV :
    563 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_SETPRIV,
    564 		     ifd, NULL, NULL);
    565 		if (error)
    566 			return (error);
    567 
    568 		break;
    569 	}
    570 
    571 	s = splnet();
    572 
    573 	switch (cmd) {
    574 	case SIOCGDRVSPEC:
    575 	case SIOCSDRVSPEC:
    576 		KASSERT(bc != NULL);
    577 		if (cmd == SIOCGDRVSPEC &&
    578 		    (bc->bc_flags & BC_F_COPYOUT) == 0) {
    579 			error = EINVAL;
    580 			break;
    581 		}
    582 		else if (cmd == SIOCSDRVSPEC &&
    583 		    (bc->bc_flags & BC_F_COPYOUT) != 0) {
    584 			error = EINVAL;
    585 			break;
    586 		}
    587 
    588 		/* BC_F_SUSER is checked above, before splnet(). */
    589 
    590 		if (ifd->ifd_len != bc->bc_argsize ||
    591 		    ifd->ifd_len > sizeof(args)) {
    592 			error = EINVAL;
    593 			break;
    594 		}
    595 
    596 		memset(&args, 0, sizeof(args));
    597 		if (bc->bc_flags & BC_F_COPYIN) {
    598 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
    599 			if (error)
    600 				break;
    601 		}
    602 
    603 		error = (*bc->bc_func)(sc, &args);
    604 		if (error)
    605 			break;
    606 
    607 		if (bc->bc_flags & BC_F_COPYOUT)
    608 			error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
    609 
    610 		break;
    611 
    612 	case SIOCSIFFLAGS:
    613 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
    614 			break;
    615 		switch (ifp->if_flags & (IFF_UP|IFF_RUNNING)) {
    616 		case IFF_RUNNING:
    617 			/*
    618 			 * If interface is marked down and it is running,
    619 			 * then stop and disable it.
    620 			 */
    621 			(*ifp->if_stop)(ifp, 1);
    622 			break;
    623 		case IFF_UP:
    624 			/*
    625 			 * If interface is marked up and it is stopped, then
    626 			 * start it.
    627 			 */
    628 			error = (*ifp->if_init)(ifp);
    629 			break;
    630 		default:
    631 			break;
    632 		}
    633 		break;
    634 
    635 	default:
    636 		error = ifioctl_common(ifp, cmd, data);
    637 		break;
    638 	}
    639 
    640 	splx(s);
    641 
    642 	return (error);
    643 }
    644 
    645 /*
    646  * bridge_lookup_member:
    647  *
    648  *	Lookup a bridge member interface.  Must be called at splnet().
    649  */
    650 static struct bridge_iflist *
    651 bridge_lookup_member(struct bridge_softc *sc, const char *name)
    652 {
    653 	struct bridge_iflist *bif;
    654 	struct ifnet *ifp;
    655 
    656 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
    657 		ifp = bif->bif_ifp;
    658 		if (strcmp(ifp->if_xname, name) == 0)
    659 			return (bif);
    660 	}
    661 
    662 	return (NULL);
    663 }
    664 
    665 /*
    666  * bridge_lookup_member_if:
    667  *
    668  *	Lookup a bridge member interface by ifnet*.  Must be called at splnet().
    669  */
    670 static struct bridge_iflist *
    671 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
    672 {
    673 	struct bridge_iflist *bif;
    674 
    675 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
    676 		if (bif->bif_ifp == member_ifp)
    677 			return (bif);
    678 	}
    679 
    680 	return (NULL);
    681 }
    682 
    683 /*
    684  * bridge_delete_member:
    685  *
    686  *	Delete the specified member interface.
    687  */
    688 static void
    689 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif)
    690 {
    691 	struct ifnet *ifs = bif->bif_ifp;
    692 
    693 	switch (ifs->if_type) {
    694 	case IFT_ETHER:
    695 		/*
    696 		 * Take the interface out of promiscuous mode.
    697 		 */
    698 		(void) ifpromisc(ifs, 0);
    699 		break;
    700 	default:
    701 #ifdef DIAGNOSTIC
    702 		panic("bridge_delete_member: impossible");
    703 #endif
    704 		break;
    705 	}
    706 
    707 	ifs->if_input = ether_input;
    708 	ifs->if_bridge = NULL;
    709 	LIST_REMOVE(bif, bif_next);
    710 
    711 	bridge_rtdelete(sc, ifs);
    712 
    713 	free(bif, M_DEVBUF);
    714 
    715 	if (sc->sc_if.if_flags & IFF_RUNNING)
    716 		bstp_initialization(sc);
    717 }
    718 
    719 static int
    720 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
    721 {
    722 	struct ifbreq *req = arg;
    723 	struct bridge_iflist *bif = NULL;
    724 	struct ifnet *ifs;
    725 	int error = 0;
    726 
    727 	ifs = ifunit(req->ifbr_ifsname);
    728 	if (ifs == NULL)
    729 		return (ENOENT);
    730 
    731 	if (sc->sc_if.if_mtu != ifs->if_mtu)
    732 		return (EINVAL);
    733 
    734 	if (ifs->if_bridge == sc)
    735 		return (EEXIST);
    736 
    737 	if (ifs->if_bridge != NULL)
    738 		return (EBUSY);
    739 
    740 	if (ifs->if_input != ether_input)
    741 		return EINVAL;
    742 
    743 	bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT);
    744 	if (bif == NULL)
    745 		return (ENOMEM);
    746 
    747 	switch (ifs->if_type) {
    748 	case IFT_ETHER:
    749 		/*
    750 		 * Place the interface into promiscuous mode.
    751 		 */
    752 		error = ifpromisc(ifs, 1);
    753 		if (error)
    754 			goto out;
    755 		break;
    756 	default:
    757 		error = EINVAL;
    758 		goto out;
    759 	}
    760 
    761 	bif->bif_ifp = ifs;
    762 	bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
    763 	bif->bif_priority = BSTP_DEFAULT_PORT_PRIORITY;
    764 	bif->bif_path_cost = BSTP_DEFAULT_PATH_COST;
    765 
    766 	ifs->if_bridge = sc;
    767 	LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
    768 	ifs->if_input = bridge_input;
    769 
    770 	if (sc->sc_if.if_flags & IFF_RUNNING)
    771 		bstp_initialization(sc);
    772 	else
    773 		bstp_stop(sc);
    774 
    775  out:
    776 	if (error) {
    777 		if (bif != NULL)
    778 			free(bif, M_DEVBUF);
    779 	}
    780 	return (error);
    781 }
    782 
    783 static int
    784 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
    785 {
    786 	struct ifbreq *req = arg;
    787 	struct bridge_iflist *bif;
    788 
    789 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
    790 	if (bif == NULL)
    791 		return (ENOENT);
    792 
    793 	bridge_delete_member(sc, bif);
    794 
    795 	return (0);
    796 }
    797 
    798 static int
    799 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
    800 {
    801 	struct ifbreq *req = arg;
    802 	struct bridge_iflist *bif;
    803 
    804 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
    805 	if (bif == NULL)
    806 		return (ENOENT);
    807 
    808 	req->ifbr_ifsflags = bif->bif_flags;
    809 	req->ifbr_state = bif->bif_state;
    810 	req->ifbr_priority = bif->bif_priority;
    811 	req->ifbr_path_cost = bif->bif_path_cost;
    812 	req->ifbr_portno = bif->bif_ifp->if_index & 0xff;
    813 
    814 	return (0);
    815 }
    816 
    817 static int
    818 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
    819 {
    820 	struct ifbreq *req = arg;
    821 	struct bridge_iflist *bif;
    822 
    823 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
    824 	if (bif == NULL)
    825 		return (ENOENT);
    826 
    827 	if (req->ifbr_ifsflags & IFBIF_STP) {
    828 		switch (bif->bif_ifp->if_type) {
    829 		case IFT_ETHER:
    830 			/* These can do spanning tree. */
    831 			break;
    832 
    833 		default:
    834 			/* Nothing else can. */
    835 			return (EINVAL);
    836 		}
    837 	}
    838 
    839 	bif->bif_flags = req->ifbr_ifsflags;
    840 
    841 	if (sc->sc_if.if_flags & IFF_RUNNING)
    842 		bstp_initialization(sc);
    843 
    844 	return (0);
    845 }
    846 
    847 static int
    848 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
    849 {
    850 	struct ifbrparam *param = arg;
    851 
    852 	sc->sc_brtmax = param->ifbrp_csize;
    853 	bridge_rttrim(sc);
    854 
    855 	return (0);
    856 }
    857 
    858 static int
    859 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
    860 {
    861 	struct ifbrparam *param = arg;
    862 
    863 	param->ifbrp_csize = sc->sc_brtmax;
    864 
    865 	return (0);
    866 }
    867 
    868 static int
    869 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
    870 {
    871 	struct ifbifconf *bifc = arg;
    872 	struct bridge_iflist *bif;
    873 	struct ifbreq breq;
    874 	int count, len, error = 0;
    875 
    876 	count = 0;
    877 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
    878 		count++;
    879 
    880 	if (bifc->ifbic_len == 0) {
    881 		bifc->ifbic_len = sizeof(breq) * count;
    882 		return (0);
    883 	}
    884 
    885 	count = 0;
    886 	len = bifc->ifbic_len;
    887 	memset(&breq, 0, sizeof breq);
    888 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
    889 		if (len < sizeof(breq))
    890 			break;
    891 
    892 		strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
    893 		    sizeof(breq.ifbr_ifsname));
    894 		breq.ifbr_ifsflags = bif->bif_flags;
    895 		breq.ifbr_state = bif->bif_state;
    896 		breq.ifbr_priority = bif->bif_priority;
    897 		breq.ifbr_path_cost = bif->bif_path_cost;
    898 		breq.ifbr_portno = bif->bif_ifp->if_index & 0xff;
    899 		error = copyout(&breq, bifc->ifbic_req + count, sizeof(breq));
    900 		if (error)
    901 			break;
    902 		count++;
    903 		len -= sizeof(breq);
    904 	}
    905 
    906 	bifc->ifbic_len = sizeof(breq) * count;
    907 	return (error);
    908 }
    909 
    910 static int
    911 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
    912 {
    913 	struct ifbaconf *bac = arg;
    914 	struct bridge_rtnode *brt;
    915 	struct ifbareq bareq;
    916 	int count = 0, error = 0, len;
    917 
    918 	if (bac->ifbac_len == 0)
    919 		return (0);
    920 
    921 	len = bac->ifbac_len;
    922 	LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
    923 		if (len < sizeof(bareq))
    924 			goto out;
    925 		memset(&bareq, 0, sizeof(bareq));
    926 		strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
    927 		    sizeof(bareq.ifba_ifsname));
    928 		memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
    929 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
    930 			bareq.ifba_expire = brt->brt_expire - time_uptime;
    931 		} else
    932 			bareq.ifba_expire = 0;
    933 		bareq.ifba_flags = brt->brt_flags;
    934 
    935 		error = copyout(&bareq, bac->ifbac_req + count, sizeof(bareq));
    936 		if (error)
    937 			goto out;
    938 		count++;
    939 		len -= sizeof(bareq);
    940 	}
    941  out:
    942 	bac->ifbac_len = sizeof(bareq) * count;
    943 	return (error);
    944 }
    945 
    946 static int
    947 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
    948 {
    949 	struct ifbareq *req = arg;
    950 	struct bridge_iflist *bif;
    951 	int error;
    952 
    953 	bif = bridge_lookup_member(sc, req->ifba_ifsname);
    954 	if (bif == NULL)
    955 		return (ENOENT);
    956 
    957 	error = bridge_rtupdate(sc, req->ifba_dst, bif->bif_ifp, 1,
    958 	    req->ifba_flags);
    959 
    960 	return (error);
    961 }
    962 
    963 static int
    964 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
    965 {
    966 	struct ifbrparam *param = arg;
    967 
    968 	sc->sc_brttimeout = param->ifbrp_ctime;
    969 
    970 	return (0);
    971 }
    972 
    973 static int
    974 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
    975 {
    976 	struct ifbrparam *param = arg;
    977 
    978 	param->ifbrp_ctime = sc->sc_brttimeout;
    979 
    980 	return (0);
    981 }
    982 
    983 static int
    984 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
    985 {
    986 	struct ifbareq *req = arg;
    987 
    988 	return (bridge_rtdaddr(sc, req->ifba_dst));
    989 }
    990 
    991 static int
    992 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
    993 {
    994 	struct ifbreq *req = arg;
    995 
    996 	bridge_rtflush(sc, req->ifbr_ifsflags);
    997 
    998 	return (0);
    999 }
   1000 
   1001 static int
   1002 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
   1003 {
   1004 	struct ifbrparam *param = arg;
   1005 
   1006 	param->ifbrp_prio = sc->sc_bridge_priority;
   1007 
   1008 	return (0);
   1009 }
   1010 
   1011 static int
   1012 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
   1013 {
   1014 	struct ifbrparam *param = arg;
   1015 
   1016 	sc->sc_bridge_priority = param->ifbrp_prio;
   1017 
   1018 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1019 		bstp_initialization(sc);
   1020 
   1021 	return (0);
   1022 }
   1023 
   1024 static int
   1025 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
   1026 {
   1027 	struct ifbrparam *param = arg;
   1028 
   1029 	param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8;
   1030 
   1031 	return (0);
   1032 }
   1033 
   1034 static int
   1035 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
   1036 {
   1037 	struct ifbrparam *param = arg;
   1038 
   1039 	if (param->ifbrp_hellotime == 0)
   1040 		return (EINVAL);
   1041 	sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8;
   1042 
   1043 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1044 		bstp_initialization(sc);
   1045 
   1046 	return (0);
   1047 }
   1048 
   1049 static int
   1050 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
   1051 {
   1052 	struct ifbrparam *param = arg;
   1053 
   1054 	param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8;
   1055 
   1056 	return (0);
   1057 }
   1058 
   1059 static int
   1060 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
   1061 {
   1062 	struct ifbrparam *param = arg;
   1063 
   1064 	if (param->ifbrp_fwddelay == 0)
   1065 		return (EINVAL);
   1066 	sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8;
   1067 
   1068 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1069 		bstp_initialization(sc);
   1070 
   1071 	return (0);
   1072 }
   1073 
   1074 static int
   1075 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
   1076 {
   1077 	struct ifbrparam *param = arg;
   1078 
   1079 	param->ifbrp_maxage = sc->sc_bridge_max_age >> 8;
   1080 
   1081 	return (0);
   1082 }
   1083 
   1084 static int
   1085 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
   1086 {
   1087 	struct ifbrparam *param = arg;
   1088 
   1089 	if (param->ifbrp_maxage == 0)
   1090 		return (EINVAL);
   1091 	sc->sc_bridge_max_age = param->ifbrp_maxage << 8;
   1092 
   1093 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1094 		bstp_initialization(sc);
   1095 
   1096 	return (0);
   1097 }
   1098 
   1099 static int
   1100 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
   1101 {
   1102 	struct ifbreq *req = arg;
   1103 	struct bridge_iflist *bif;
   1104 
   1105 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
   1106 	if (bif == NULL)
   1107 		return (ENOENT);
   1108 
   1109 	bif->bif_priority = req->ifbr_priority;
   1110 
   1111 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1112 		bstp_initialization(sc);
   1113 
   1114 	return (0);
   1115 }
   1116 
   1117 #if defined(BRIDGE_IPF)
   1118 static int
   1119 bridge_ioctl_gfilt(struct bridge_softc *sc, void *arg)
   1120 {
   1121 	struct ifbrparam *param = arg;
   1122 
   1123 	param->ifbrp_filter = sc->sc_filter_flags;
   1124 
   1125 	return (0);
   1126 }
   1127 
   1128 static int
   1129 bridge_ioctl_sfilt(struct bridge_softc *sc, void *arg)
   1130 {
   1131 	struct ifbrparam *param = arg;
   1132 	uint32_t nflags, oflags;
   1133 
   1134 	if (param->ifbrp_filter & ~IFBF_FILT_MASK)
   1135 		return (EINVAL);
   1136 
   1137 	nflags = param->ifbrp_filter;
   1138 	oflags = sc->sc_filter_flags;
   1139 
   1140 	if ((nflags & IFBF_FILT_USEIPF) && !(oflags & IFBF_FILT_USEIPF)) {
   1141 		pfil_add_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
   1142 			sc->sc_if.if_pfil);
   1143 	}
   1144 	if (!(nflags & IFBF_FILT_USEIPF) && (oflags & IFBF_FILT_USEIPF)) {
   1145 		pfil_remove_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
   1146 			sc->sc_if.if_pfil);
   1147 	}
   1148 
   1149 	sc->sc_filter_flags = nflags;
   1150 
   1151 	return (0);
   1152 }
   1153 #endif /* BRIDGE_IPF */
   1154 
   1155 static int
   1156 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
   1157 {
   1158 	struct ifbreq *req = arg;
   1159 	struct bridge_iflist *bif;
   1160 
   1161 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
   1162 	if (bif == NULL)
   1163 		return (ENOENT);
   1164 
   1165 	bif->bif_path_cost = req->ifbr_path_cost;
   1166 
   1167 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1168 		bstp_initialization(sc);
   1169 
   1170 	return (0);
   1171 }
   1172 
   1173 /*
   1174  * bridge_ifdetach:
   1175  *
   1176  *	Detach an interface from a bridge.  Called when a member
   1177  *	interface is detaching.
   1178  */
   1179 void
   1180 bridge_ifdetach(struct ifnet *ifp)
   1181 {
   1182 	struct bridge_softc *sc = ifp->if_bridge;
   1183 	struct ifbreq breq;
   1184 
   1185 	memset(&breq, 0, sizeof(breq));
   1186 	strlcpy(breq.ifbr_ifsname, ifp->if_xname, sizeof(breq.ifbr_ifsname));
   1187 
   1188 	(void) bridge_ioctl_del(sc, &breq);
   1189 }
   1190 
   1191 /*
   1192  * bridge_init:
   1193  *
   1194  *	Initialize a bridge interface.
   1195  */
   1196 static int
   1197 bridge_init(struct ifnet *ifp)
   1198 {
   1199 	struct bridge_softc *sc = ifp->if_softc;
   1200 
   1201 	if (ifp->if_flags & IFF_RUNNING)
   1202 		return (0);
   1203 
   1204 	callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
   1205 	    bridge_timer, sc);
   1206 
   1207 	ifp->if_flags |= IFF_RUNNING;
   1208 	bstp_initialization(sc);
   1209 	return (0);
   1210 }
   1211 
   1212 /*
   1213  * bridge_stop:
   1214  *
   1215  *	Stop the bridge interface.
   1216  */
   1217 static void
   1218 bridge_stop(struct ifnet *ifp, int disable)
   1219 {
   1220 	struct bridge_softc *sc = ifp->if_softc;
   1221 
   1222 	if ((ifp->if_flags & IFF_RUNNING) == 0)
   1223 		return;
   1224 
   1225 	callout_stop(&sc->sc_brcallout);
   1226 	bstp_stop(sc);
   1227 
   1228 	IF_PURGE(&ifp->if_snd);
   1229 
   1230 	bridge_rtflush(sc, IFBF_FLUSHDYN);
   1231 
   1232 	ifp->if_flags &= ~IFF_RUNNING;
   1233 }
   1234 
   1235 /*
   1236  * bridge_enqueue:
   1237  *
   1238  *	Enqueue a packet on a bridge member interface.
   1239  *
   1240  *	NOTE: must be called at splnet().
   1241  */
   1242 void
   1243 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m,
   1244     int runfilt)
   1245 {
   1246 	ALTQ_DECL(struct altq_pktattr pktattr;)
   1247 	int len, error;
   1248 	short mflags;
   1249 
   1250 	/*
   1251 	 * Clear any in-bound checksum flags for this packet.
   1252 	 */
   1253 	m->m_pkthdr.csum_flags = 0;
   1254 
   1255 	if (runfilt) {
   1256 		if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
   1257 		    dst_ifp, PFIL_OUT) != 0) {
   1258 			if (m != NULL)
   1259 				m_freem(m);
   1260 			return;
   1261 		}
   1262 		if (m == NULL)
   1263 			return;
   1264 	}
   1265 
   1266 #ifdef ALTQ
   1267 	/*
   1268 	 * If ALTQ is enabled on the member interface, do
   1269 	 * classification; the queueing discipline might
   1270 	 * not require classification, but might require
   1271 	 * the address family/header pointer in the pktattr.
   1272 	 */
   1273 	if (ALTQ_IS_ENABLED(&dst_ifp->if_snd)) {
   1274 		/* XXX IFT_ETHER */
   1275 		altq_etherclassify(&dst_ifp->if_snd, m, &pktattr);
   1276 	}
   1277 #endif /* ALTQ */
   1278 
   1279 	len = m->m_pkthdr.len;
   1280 	m->m_flags |= M_PROTO1;
   1281 	mflags = m->m_flags;
   1282 	IFQ_ENQUEUE(&dst_ifp->if_snd, m, &pktattr, error);
   1283 	if (error) {
   1284 		/* mbuf is already freed */
   1285 		sc->sc_if.if_oerrors++;
   1286 		return;
   1287 	}
   1288 
   1289 	sc->sc_if.if_opackets++;
   1290 	sc->sc_if.if_obytes += len;
   1291 
   1292 	dst_ifp->if_obytes += len;
   1293 
   1294 	if (mflags & M_MCAST) {
   1295 		sc->sc_if.if_omcasts++;
   1296 		dst_ifp->if_omcasts++;
   1297 	}
   1298 
   1299 	if ((dst_ifp->if_flags & IFF_OACTIVE) == 0)
   1300 		(*dst_ifp->if_start)(dst_ifp);
   1301 }
   1302 
   1303 /*
   1304  * bridge_output:
   1305  *
   1306  *	Send output from a bridge member interface.  This
   1307  *	performs the bridging function for locally originated
   1308  *	packets.
   1309  *
   1310  *	The mbuf has the Ethernet header already attached.  We must
   1311  *	enqueue or free the mbuf before returning.
   1312  */
   1313 int
   1314 bridge_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa,
   1315     struct rtentry *rt)
   1316 {
   1317 	struct ether_header *eh;
   1318 	struct ifnet *dst_if;
   1319 	struct bridge_softc *sc;
   1320 	int s;
   1321 
   1322 	if (m->m_len < ETHER_HDR_LEN) {
   1323 		m = m_pullup(m, ETHER_HDR_LEN);
   1324 		if (m == NULL)
   1325 			return (0);
   1326 	}
   1327 
   1328 	eh = mtod(m, struct ether_header *);
   1329 	sc = ifp->if_bridge;
   1330 
   1331 	s = splnet();
   1332 
   1333 	/*
   1334 	 * If bridge is down, but the original output interface is up,
   1335 	 * go ahead and send out that interface.  Otherwise, the packet
   1336 	 * is dropped below.
   1337 	 */
   1338 	if ((sc->sc_if.if_flags & IFF_RUNNING) == 0) {
   1339 		dst_if = ifp;
   1340 		goto sendunicast;
   1341 	}
   1342 
   1343 	/*
   1344 	 * If the packet is a multicast, or we don't know a better way to
   1345 	 * get there, send to all interfaces.
   1346 	 */
   1347 	if (ETHER_IS_MULTICAST(eh->ether_dhost))
   1348 		dst_if = NULL;
   1349 	else
   1350 		dst_if = bridge_rtlookup(sc, eh->ether_dhost);
   1351 	if (dst_if == NULL) {
   1352 		struct bridge_iflist *bif;
   1353 		struct mbuf *mc;
   1354 		int used = 0;
   1355 
   1356 		LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
   1357 			dst_if = bif->bif_ifp;
   1358 			if ((dst_if->if_flags & IFF_RUNNING) == 0)
   1359 				continue;
   1360 
   1361 			/*
   1362 			 * If this is not the original output interface,
   1363 			 * and the interface is participating in spanning
   1364 			 * tree, make sure the port is in a state that
   1365 			 * allows forwarding.
   1366 			 */
   1367 			if (dst_if != ifp &&
   1368 			    (bif->bif_flags & IFBIF_STP) != 0) {
   1369 				switch (bif->bif_state) {
   1370 				case BSTP_IFSTATE_BLOCKING:
   1371 				case BSTP_IFSTATE_LISTENING:
   1372 				case BSTP_IFSTATE_DISABLED:
   1373 					continue;
   1374 				}
   1375 			}
   1376 
   1377 			if (LIST_NEXT(bif, bif_next) == NULL) {
   1378 				used = 1;
   1379 				mc = m;
   1380 			} else {
   1381 				mc = m_copym(m, 0, M_COPYALL, M_NOWAIT);
   1382 				if (mc == NULL) {
   1383 					sc->sc_if.if_oerrors++;
   1384 					continue;
   1385 				}
   1386 			}
   1387 
   1388 			bridge_enqueue(sc, dst_if, mc, 0);
   1389 		}
   1390 		if (used == 0)
   1391 			m_freem(m);
   1392 		splx(s);
   1393 		return (0);
   1394 	}
   1395 
   1396  sendunicast:
   1397 	/*
   1398 	 * XXX Spanning tree consideration here?
   1399 	 */
   1400 
   1401 	if ((dst_if->if_flags & IFF_RUNNING) == 0) {
   1402 		m_freem(m);
   1403 		splx(s);
   1404 		return (0);
   1405 	}
   1406 
   1407 	bridge_enqueue(sc, dst_if, m, 0);
   1408 
   1409 	splx(s);
   1410 	return (0);
   1411 }
   1412 
   1413 /*
   1414  * bridge_start:
   1415  *
   1416  *	Start output on a bridge.
   1417  *
   1418  *	NOTE: This routine should never be called in this implementation.
   1419  */
   1420 static void
   1421 bridge_start(struct ifnet *ifp)
   1422 {
   1423 
   1424 	printf("%s: bridge_start() called\n", ifp->if_xname);
   1425 }
   1426 
   1427 /*
   1428  * bridge_forward:
   1429  *
   1430  *	The forwarding function of the bridge.
   1431  */
   1432 static void
   1433 bridge_forward(void *v)
   1434 {
   1435 	struct bridge_softc *sc = v;
   1436 	struct mbuf *m;
   1437 	struct bridge_iflist *bif;
   1438 	struct ifnet *src_if, *dst_if;
   1439 	struct ether_header *eh;
   1440 	int s;
   1441 
   1442 	KERNEL_LOCK(1, NULL);
   1443 	mutex_enter(softnet_lock);
   1444 	if ((sc->sc_if.if_flags & IFF_RUNNING) == 0) {
   1445 		mutex_exit(softnet_lock);
   1446 		KERNEL_UNLOCK_ONE(NULL);
   1447 		return;
   1448 	}
   1449 
   1450 	s = splnet();
   1451 	while ((m = pktq_dequeue(sc->sc_fwd_pktq)) != NULL) {
   1452 		src_if = m->m_pkthdr.rcvif;
   1453 
   1454 		sc->sc_if.if_ipackets++;
   1455 		sc->sc_if.if_ibytes += m->m_pkthdr.len;
   1456 
   1457 		/*
   1458 		 * Look up the bridge_iflist.
   1459 		 */
   1460 		bif = bridge_lookup_member_if(sc, src_if);
   1461 		if (bif == NULL) {
   1462 			/* Interface is not a bridge member (anymore?) */
   1463 			m_freem(m);
   1464 			continue;
   1465 		}
   1466 
   1467 		if (bif->bif_flags & IFBIF_STP) {
   1468 			switch (bif->bif_state) {
   1469 			case BSTP_IFSTATE_BLOCKING:
   1470 			case BSTP_IFSTATE_LISTENING:
   1471 			case BSTP_IFSTATE_DISABLED:
   1472 				m_freem(m);
   1473 				continue;
   1474 			}
   1475 		}
   1476 
   1477 		eh = mtod(m, struct ether_header *);
   1478 
   1479 		/*
   1480 		 * If the interface is learning, and the source
   1481 		 * address is valid and not multicast, record
   1482 		 * the address.
   1483 		 */
   1484 		if ((bif->bif_flags & IFBIF_LEARNING) != 0 &&
   1485 		    ETHER_IS_MULTICAST(eh->ether_shost) == 0 &&
   1486 		    (eh->ether_shost[0] == 0 &&
   1487 		     eh->ether_shost[1] == 0 &&
   1488 		     eh->ether_shost[2] == 0 &&
   1489 		     eh->ether_shost[3] == 0 &&
   1490 		     eh->ether_shost[4] == 0 &&
   1491 		     eh->ether_shost[5] == 0) == 0) {
   1492 			(void) bridge_rtupdate(sc, eh->ether_shost,
   1493 			    src_if, 0, IFBAF_DYNAMIC);
   1494 		}
   1495 
   1496 		if ((bif->bif_flags & IFBIF_STP) != 0 &&
   1497 		    bif->bif_state == BSTP_IFSTATE_LEARNING) {
   1498 			m_freem(m);
   1499 			continue;
   1500 		}
   1501 
   1502 		/*
   1503 		 * At this point, the port either doesn't participate
   1504 		 * in spanning tree or it is in the forwarding state.
   1505 		 */
   1506 
   1507 		/*
   1508 		 * If the packet is unicast, destined for someone on
   1509 		 * "this" side of the bridge, drop it.
   1510 		 */
   1511 		if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
   1512 			dst_if = bridge_rtlookup(sc, eh->ether_dhost);
   1513 			if (src_if == dst_if) {
   1514 				m_freem(m);
   1515 				continue;
   1516 			}
   1517 		} else {
   1518 			/* ...forward it to all interfaces. */
   1519 			sc->sc_if.if_imcasts++;
   1520 			dst_if = NULL;
   1521 		}
   1522 
   1523 		if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
   1524 		    m->m_pkthdr.rcvif, PFIL_IN) != 0) {
   1525 			if (m != NULL)
   1526 				m_freem(m);
   1527 			continue;
   1528 		}
   1529 		if (m == NULL)
   1530 			continue;
   1531 
   1532 		if (dst_if == NULL) {
   1533 			bridge_broadcast(sc, src_if, m);
   1534 			continue;
   1535 		}
   1536 
   1537 		/*
   1538 		 * At this point, we're dealing with a unicast frame
   1539 		 * going to a different interface.
   1540 		 */
   1541 		if ((dst_if->if_flags & IFF_RUNNING) == 0) {
   1542 			m_freem(m);
   1543 			continue;
   1544 		}
   1545 		bif = bridge_lookup_member_if(sc, dst_if);
   1546 		if (bif == NULL) {
   1547 			/* Not a member of the bridge (anymore?) */
   1548 			m_freem(m);
   1549 			continue;
   1550 		}
   1551 
   1552 		if (bif->bif_flags & IFBIF_STP) {
   1553 			switch (bif->bif_state) {
   1554 			case BSTP_IFSTATE_DISABLED:
   1555 			case BSTP_IFSTATE_BLOCKING:
   1556 				m_freem(m);
   1557 				continue;
   1558 			}
   1559 		}
   1560 
   1561 		bridge_enqueue(sc, dst_if, m, 1);
   1562 	}
   1563 	splx(s);
   1564 	mutex_exit(softnet_lock);
   1565 	KERNEL_UNLOCK_ONE(NULL);
   1566 }
   1567 
   1568 static bool
   1569 bstp_state_before_learning(struct bridge_iflist *bif)
   1570 {
   1571 	if (bif->bif_flags & IFBIF_STP) {
   1572 		switch (bif->bif_state) {
   1573 		case BSTP_IFSTATE_BLOCKING:
   1574 		case BSTP_IFSTATE_LISTENING:
   1575 		case BSTP_IFSTATE_DISABLED:
   1576 			return true;
   1577 		}
   1578 	}
   1579 	return false;
   1580 }
   1581 
   1582 static bool
   1583 bridge_ourether(struct bridge_iflist *bif, struct ether_header *eh, int src)
   1584 {
   1585 	uint8_t *ether = src ? eh->ether_shost : eh->ether_dhost;
   1586 
   1587 	if (memcmp(CLLADDR(bif->bif_ifp->if_sadl), ether, ETHER_ADDR_LEN) == 0
   1588 #if NCARP > 0
   1589 	    || (bif->bif_ifp->if_carp &&
   1590 	        carp_ourether(bif->bif_ifp->if_carp, eh, IFT_ETHER, src) != NULL)
   1591 #endif /* NCARP > 0 */
   1592 	    )
   1593 		return true;
   1594 
   1595 	return false;
   1596 }
   1597 
   1598 /*
   1599  * bridge_input:
   1600  *
   1601  *	Receive input from a member interface.  Queue the packet for
   1602  *	bridging if it is not for us.
   1603  *	should be called at splnet()
   1604  */
   1605 static void
   1606 bridge_input(struct ifnet *ifp, struct mbuf *m)
   1607 {
   1608 	struct bridge_softc *sc = ifp->if_bridge;
   1609 	struct bridge_iflist *bif;
   1610 	struct ether_header *eh;
   1611 
   1612 	if ((sc->sc_if.if_flags & IFF_RUNNING) == 0) {
   1613 		ether_input(ifp, m);
   1614 		return;
   1615 	}
   1616 
   1617 	bif = bridge_lookup_member_if(sc, ifp);
   1618 	if (bif == NULL) {
   1619 		ether_input(ifp, m);
   1620 		return;
   1621 	}
   1622 
   1623 	eh = mtod(m, struct ether_header *);
   1624 
   1625 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
   1626 		if (memcmp(etherbroadcastaddr,
   1627 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
   1628 			m->m_flags |= M_BCAST;
   1629 		else
   1630 			m->m_flags |= M_MCAST;
   1631 	}
   1632 
   1633 	/*
   1634 	 * A 'fast' path for packets addressed to interfaces that are
   1635 	 * part of this bridge.
   1636 	 */
   1637 	if (!(m->m_flags & (M_BCAST|M_MCAST)) &&
   1638 	    !bstp_state_before_learning(bif)) {
   1639 		struct bridge_iflist *_bif;
   1640 
   1641 		LIST_FOREACH(_bif, &sc->sc_iflist, bif_next) {
   1642 			/* It is destined for us. */
   1643 			if (bridge_ourether(_bif, eh, 0)) {
   1644 				if (_bif->bif_flags & IFBIF_LEARNING)
   1645 					(void) bridge_rtupdate(sc,
   1646 					    eh->ether_shost, ifp, 0, IFBAF_DYNAMIC);
   1647 				m->m_pkthdr.rcvif = _bif->bif_ifp;
   1648 				ether_input(_bif->bif_ifp, m);
   1649 				return;
   1650 			}
   1651 
   1652 			/* We just received a packet that we sent out. */
   1653 			if (bridge_ourether(_bif, eh, 1)) {
   1654 				m_freem(m);
   1655 				return;
   1656 			}
   1657 		}
   1658 	}
   1659 
   1660 	/* Tap off 802.1D packets; they do not get forwarded. */
   1661 	if (bif->bif_flags & IFBIF_STP &&
   1662 	    memcmp(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN) == 0) {
   1663 		bstp_input(sc, bif, m);
   1664 		return;
   1665 	}
   1666 
   1667 	/*
   1668 	 * A normal switch would discard the packet here, but that's not what
   1669 	 * we've done historically. This also prevents some obnoxious behaviour.
   1670 	 */
   1671 	if (bstp_state_before_learning(bif)) {
   1672 		ether_input(ifp, m);
   1673 		return;
   1674 	}
   1675 
   1676 	/* Queue the packet for bridge forwarding. */
   1677 	if (__predict_false(!pktq_enqueue(sc->sc_fwd_pktq, m, 0)))
   1678 		m_freem(m);
   1679 }
   1680 
   1681 /*
   1682  * bridge_broadcast:
   1683  *
   1684  *	Send a frame to all interfaces that are members of
   1685  *	the bridge, except for the one on which the packet
   1686  *	arrived.
   1687  */
   1688 static void
   1689 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
   1690     struct mbuf *m)
   1691 {
   1692 	struct bridge_iflist *bif;
   1693 	struct mbuf *mc;
   1694 	struct ifnet *dst_if;
   1695 	bool used, bmcast;
   1696 
   1697 	used = bmcast = m->m_flags & (M_BCAST|M_MCAST);
   1698 
   1699 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
   1700 		dst_if = bif->bif_ifp;
   1701 		if (dst_if == src_if)
   1702 			continue;
   1703 
   1704 		if (bif->bif_flags & IFBIF_STP) {
   1705 			switch (bif->bif_state) {
   1706 			case BSTP_IFSTATE_BLOCKING:
   1707 			case BSTP_IFSTATE_DISABLED:
   1708 				continue;
   1709 			}
   1710 		}
   1711 
   1712 		if ((bif->bif_flags & IFBIF_DISCOVER) == 0 && !bmcast)
   1713 			continue;
   1714 
   1715 		if ((dst_if->if_flags & IFF_RUNNING) == 0)
   1716 			continue;
   1717 
   1718 		if (!used && LIST_NEXT(bif, bif_next) == NULL) {
   1719 			mc = m;
   1720 			used = true;
   1721 		} else {
   1722 			mc = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
   1723 			if (mc == NULL) {
   1724 				sc->sc_if.if_oerrors++;
   1725 				continue;
   1726 			}
   1727 		}
   1728 
   1729 		bridge_enqueue(sc, dst_if, mc, 1);
   1730 	}
   1731 
   1732 	if (bmcast)
   1733 		ether_input(src_if, m);
   1734 	else if (!used)
   1735 		m_freem(m);
   1736 }
   1737 
   1738 /*
   1739  * bridge_rtupdate:
   1740  *
   1741  *	Add a bridge routing entry.
   1742  */
   1743 static int
   1744 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
   1745     struct ifnet *dst_if, int setflags, uint8_t flags)
   1746 {
   1747 	struct bridge_rtnode *brt;
   1748 	int error;
   1749 
   1750 	/*
   1751 	 * A route for this destination might already exist.  If so,
   1752 	 * update it, otherwise create a new one.
   1753 	 */
   1754 	if ((brt = bridge_rtnode_lookup(sc, dst)) == NULL) {
   1755 		if (sc->sc_brtcnt >= sc->sc_brtmax)
   1756 			return (ENOSPC);
   1757 
   1758 		/*
   1759 		 * Allocate a new bridge forwarding node, and
   1760 		 * initialize the expiration time and Ethernet
   1761 		 * address.
   1762 		 */
   1763 		brt = pool_get(&bridge_rtnode_pool, PR_NOWAIT);
   1764 		if (brt == NULL)
   1765 			return (ENOMEM);
   1766 
   1767 		memset(brt, 0, sizeof(*brt));
   1768 		brt->brt_expire = time_uptime + sc->sc_brttimeout;
   1769 		brt->brt_flags = IFBAF_DYNAMIC;
   1770 		memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
   1771 
   1772 		if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
   1773 			pool_put(&bridge_rtnode_pool, brt);
   1774 			return (error);
   1775 		}
   1776 	}
   1777 
   1778 	brt->brt_ifp = dst_if;
   1779 	if (setflags) {
   1780 		brt->brt_flags = flags;
   1781 		if (flags & IFBAF_STATIC)
   1782 			brt->brt_expire = 0;
   1783 		else
   1784 			brt->brt_expire = time_uptime + sc->sc_brttimeout;
   1785 	}
   1786 
   1787 	return (0);
   1788 }
   1789 
   1790 /*
   1791  * bridge_rtlookup:
   1792  *
   1793  *	Lookup the destination interface for an address.
   1794  */
   1795 static struct ifnet *
   1796 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr)
   1797 {
   1798 	struct bridge_rtnode *brt;
   1799 
   1800 	if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL)
   1801 		return (NULL);
   1802 
   1803 	return (brt->brt_ifp);
   1804 }
   1805 
   1806 /*
   1807  * bridge_rttrim:
   1808  *
   1809  *	Trim the routine table so that we have a number
   1810  *	of routing entries less than or equal to the
   1811  *	maximum number.
   1812  */
   1813 static void
   1814 bridge_rttrim(struct bridge_softc *sc)
   1815 {
   1816 	struct bridge_rtnode *brt, *nbrt;
   1817 
   1818 	/* Make sure we actually need to do this. */
   1819 	if (sc->sc_brtcnt <= sc->sc_brtmax)
   1820 		return;
   1821 
   1822 	/* Force an aging cycle; this might trim enough addresses. */
   1823 	bridge_rtage(sc);
   1824 	if (sc->sc_brtcnt <= sc->sc_brtmax)
   1825 		return;
   1826 
   1827 	for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
   1828 		nbrt = LIST_NEXT(brt, brt_list);
   1829 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
   1830 			bridge_rtnode_destroy(sc, brt);
   1831 			if (sc->sc_brtcnt <= sc->sc_brtmax)
   1832 				return;
   1833 		}
   1834 	}
   1835 }
   1836 
   1837 /*
   1838  * bridge_timer:
   1839  *
   1840  *	Aging timer for the bridge.
   1841  */
   1842 static void
   1843 bridge_timer(void *arg)
   1844 {
   1845 	struct bridge_softc *sc = arg;
   1846 	int s;
   1847 
   1848 	s = splnet();
   1849 	bridge_rtage(sc);
   1850 	splx(s);
   1851 
   1852 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1853 		callout_reset(&sc->sc_brcallout,
   1854 		    bridge_rtable_prune_period * hz, bridge_timer, sc);
   1855 }
   1856 
   1857 /*
   1858  * bridge_rtage:
   1859  *
   1860  *	Perform an aging cycle.
   1861  */
   1862 static void
   1863 bridge_rtage(struct bridge_softc *sc)
   1864 {
   1865 	struct bridge_rtnode *brt, *nbrt;
   1866 
   1867 	for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
   1868 		nbrt = LIST_NEXT(brt, brt_list);
   1869 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
   1870 			if (time_uptime >= brt->brt_expire)
   1871 				bridge_rtnode_destroy(sc, brt);
   1872 		}
   1873 	}
   1874 }
   1875 
   1876 /*
   1877  * bridge_rtflush:
   1878  *
   1879  *	Remove all dynamic addresses from the bridge.
   1880  */
   1881 static void
   1882 bridge_rtflush(struct bridge_softc *sc, int full)
   1883 {
   1884 	struct bridge_rtnode *brt, *nbrt;
   1885 
   1886 	for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
   1887 		nbrt = LIST_NEXT(brt, brt_list);
   1888 		if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
   1889 			bridge_rtnode_destroy(sc, brt);
   1890 	}
   1891 }
   1892 
   1893 /*
   1894  * bridge_rtdaddr:
   1895  *
   1896  *	Remove an address from the table.
   1897  */
   1898 static int
   1899 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr)
   1900 {
   1901 	struct bridge_rtnode *brt;
   1902 
   1903 	if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL)
   1904 		return (ENOENT);
   1905 
   1906 	bridge_rtnode_destroy(sc, brt);
   1907 	return (0);
   1908 }
   1909 
   1910 /*
   1911  * bridge_rtdelete:
   1912  *
   1913  *	Delete routes to a speicifc member interface.
   1914  */
   1915 static void
   1916 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp)
   1917 {
   1918 	struct bridge_rtnode *brt, *nbrt;
   1919 
   1920 	for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
   1921 		nbrt = LIST_NEXT(brt, brt_list);
   1922 		if (brt->brt_ifp == ifp)
   1923 			bridge_rtnode_destroy(sc, brt);
   1924 	}
   1925 }
   1926 
   1927 /*
   1928  * bridge_rtable_init:
   1929  *
   1930  *	Initialize the route table for this bridge.
   1931  */
   1932 static int
   1933 bridge_rtable_init(struct bridge_softc *sc)
   1934 {
   1935 	int i;
   1936 
   1937 	sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
   1938 	    M_DEVBUF, M_NOWAIT);
   1939 	if (sc->sc_rthash == NULL)
   1940 		return (ENOMEM);
   1941 
   1942 	for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
   1943 		LIST_INIT(&sc->sc_rthash[i]);
   1944 
   1945 	sc->sc_rthash_key = cprng_fast32();
   1946 
   1947 	LIST_INIT(&sc->sc_rtlist);
   1948 
   1949 	return (0);
   1950 }
   1951 
   1952 /*
   1953  * bridge_rtable_fini:
   1954  *
   1955  *	Deconstruct the route table for this bridge.
   1956  */
   1957 static void
   1958 bridge_rtable_fini(struct bridge_softc *sc)
   1959 {
   1960 
   1961 	free(sc->sc_rthash, M_DEVBUF);
   1962 }
   1963 
   1964 /*
   1965  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
   1966  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
   1967  */
   1968 #define	mix(a, b, c)							\
   1969 do {									\
   1970 	a -= b; a -= c; a ^= (c >> 13);					\
   1971 	b -= c; b -= a; b ^= (a << 8);					\
   1972 	c -= a; c -= b; c ^= (b >> 13);					\
   1973 	a -= b; a -= c; a ^= (c >> 12);					\
   1974 	b -= c; b -= a; b ^= (a << 16);					\
   1975 	c -= a; c -= b; c ^= (b >> 5);					\
   1976 	a -= b; a -= c; a ^= (c >> 3);					\
   1977 	b -= c; b -= a; b ^= (a << 10);					\
   1978 	c -= a; c -= b; c ^= (b >> 15);					\
   1979 } while (/*CONSTCOND*/0)
   1980 
   1981 static inline uint32_t
   1982 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
   1983 {
   1984 	uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
   1985 
   1986 	b += addr[5] << 8;
   1987 	b += addr[4];
   1988 	a += addr[3] << 24;
   1989 	a += addr[2] << 16;
   1990 	a += addr[1] << 8;
   1991 	a += addr[0];
   1992 
   1993 	mix(a, b, c);
   1994 
   1995 	return (c & BRIDGE_RTHASH_MASK);
   1996 }
   1997 
   1998 #undef mix
   1999 
   2000 /*
   2001  * bridge_rtnode_lookup:
   2002  *
   2003  *	Look up a bridge route node for the specified destination.
   2004  */
   2005 static struct bridge_rtnode *
   2006 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr)
   2007 {
   2008 	struct bridge_rtnode *brt;
   2009 	uint32_t hash;
   2010 	int dir;
   2011 
   2012 	hash = bridge_rthash(sc, addr);
   2013 	LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
   2014 		dir = memcmp(addr, brt->brt_addr, ETHER_ADDR_LEN);
   2015 		if (dir == 0)
   2016 			return (brt);
   2017 		if (dir > 0)
   2018 			return (NULL);
   2019 	}
   2020 
   2021 	return (NULL);
   2022 }
   2023 
   2024 /*
   2025  * bridge_rtnode_insert:
   2026  *
   2027  *	Insert the specified bridge node into the route table.  We
   2028  *	assume the entry is not already in the table.
   2029  */
   2030 static int
   2031 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
   2032 {
   2033 	struct bridge_rtnode *lbrt;
   2034 	uint32_t hash;
   2035 	int dir;
   2036 
   2037 	hash = bridge_rthash(sc, brt->brt_addr);
   2038 
   2039 	lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
   2040 	if (lbrt == NULL) {
   2041 		LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
   2042 		goto out;
   2043 	}
   2044 
   2045 	do {
   2046 		dir = memcmp(brt->brt_addr, lbrt->brt_addr, ETHER_ADDR_LEN);
   2047 		if (dir == 0)
   2048 			return (EEXIST);
   2049 		if (dir > 0) {
   2050 			LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
   2051 			goto out;
   2052 		}
   2053 		if (LIST_NEXT(lbrt, brt_hash) == NULL) {
   2054 			LIST_INSERT_AFTER(lbrt, brt, brt_hash);
   2055 			goto out;
   2056 		}
   2057 		lbrt = LIST_NEXT(lbrt, brt_hash);
   2058 	} while (lbrt != NULL);
   2059 
   2060 #ifdef DIAGNOSTIC
   2061 	panic("bridge_rtnode_insert: impossible");
   2062 #endif
   2063 
   2064  out:
   2065 	LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
   2066 	sc->sc_brtcnt++;
   2067 
   2068 	return (0);
   2069 }
   2070 
   2071 /*
   2072  * bridge_rtnode_destroy:
   2073  *
   2074  *	Destroy a bridge rtnode.
   2075  */
   2076 static void
   2077 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
   2078 {
   2079 	int s = splnet();
   2080 
   2081 	LIST_REMOVE(brt, brt_hash);
   2082 
   2083 	LIST_REMOVE(brt, brt_list);
   2084 	sc->sc_brtcnt--;
   2085 	pool_put(&bridge_rtnode_pool, brt);
   2086 
   2087 	splx(s);
   2088 }
   2089 
   2090 #if defined(BRIDGE_IPF)
   2091 extern pfil_head_t *inet_pfil_hook;                 /* XXX */
   2092 extern pfil_head_t *inet6_pfil_hook;                /* XXX */
   2093 
   2094 /*
   2095  * Send bridge packets through IPF if they are one of the types IPF can deal
   2096  * with, or if they are ARP or REVARP.  (IPF will pass ARP and REVARP without
   2097  * question.)
   2098  */
   2099 static int
   2100 bridge_ipf(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
   2101 {
   2102 	int snap, error;
   2103 	struct ether_header *eh1, eh2;
   2104 	struct llc llc1;
   2105 	uint16_t ether_type;
   2106 
   2107 	snap = 0;
   2108 	error = -1;	/* Default error if not error == 0 */
   2109 	eh1 = mtod(*mp, struct ether_header *);
   2110 	ether_type = ntohs(eh1->ether_type);
   2111 
   2112 	/*
   2113 	 * Check for SNAP/LLC.
   2114 	 */
   2115         if (ether_type < ETHERMTU) {
   2116                 struct llc *llc2 = (struct llc *)(eh1 + 1);
   2117 
   2118                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
   2119                     llc2->llc_dsap == LLC_SNAP_LSAP &&
   2120                     llc2->llc_ssap == LLC_SNAP_LSAP &&
   2121                     llc2->llc_control == LLC_UI) {
   2122                 	ether_type = htons(llc2->llc_un.type_snap.ether_type);
   2123 			snap = 1;
   2124                 }
   2125         }
   2126 
   2127 	/*
   2128 	 * If we're trying to filter bridge traffic, don't look at anything
   2129 	 * other than IP and ARP traffic.  If the filter doesn't understand
   2130 	 * IPv6, don't allow IPv6 through the bridge either.  This is lame
   2131 	 * since if we really wanted, say, an AppleTalk filter, we are hosed,
   2132 	 * but of course we don't have an AppleTalk filter to begin with.
   2133 	 * (Note that since IPF doesn't understand ARP it will pass *ALL*
   2134 	 * ARP traffic.)
   2135 	 */
   2136 	switch (ether_type) {
   2137 		case ETHERTYPE_ARP:
   2138 		case ETHERTYPE_REVARP:
   2139 			return 0; /* Automatically pass */
   2140 		case ETHERTYPE_IP:
   2141 # ifdef INET6
   2142 		case ETHERTYPE_IPV6:
   2143 # endif /* INET6 */
   2144 			break;
   2145 		default:
   2146 			goto bad;
   2147 	}
   2148 
   2149 	/* Strip off the Ethernet header and keep a copy. */
   2150 	m_copydata(*mp, 0, ETHER_HDR_LEN, (void *) &eh2);
   2151 	m_adj(*mp, ETHER_HDR_LEN);
   2152 
   2153 	/* Strip off snap header, if present */
   2154 	if (snap) {
   2155 		m_copydata(*mp, 0, sizeof(struct llc), (void *) &llc1);
   2156 		m_adj(*mp, sizeof(struct llc));
   2157 	}
   2158 
   2159 	/*
   2160 	 * Check basic packet sanity and run IPF through pfil.
   2161 	 */
   2162 	KASSERT(!cpu_intr_p());
   2163 	switch (ether_type)
   2164 	{
   2165 	case ETHERTYPE_IP :
   2166 		error = (dir == PFIL_IN) ? bridge_ip_checkbasic(mp) : 0;
   2167 		if (error == 0)
   2168 			error = pfil_run_hooks(inet_pfil_hook, mp, ifp, dir);
   2169 		break;
   2170 # ifdef INET6
   2171 	case ETHERTYPE_IPV6 :
   2172 		error = (dir == PFIL_IN) ? bridge_ip6_checkbasic(mp) : 0;
   2173 		if (error == 0)
   2174 			error = pfil_run_hooks(inet6_pfil_hook, mp, ifp, dir);
   2175 		break;
   2176 # endif
   2177 	default :
   2178 		error = 0;
   2179 		break;
   2180 	}
   2181 
   2182 	if (*mp == NULL)
   2183 		return error;
   2184 	if (error != 0)
   2185 		goto bad;
   2186 
   2187 	error = -1;
   2188 
   2189 	/*
   2190 	 * Finally, put everything back the way it was and return
   2191 	 */
   2192 	if (snap) {
   2193 		M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
   2194 		if (*mp == NULL)
   2195 			return error;
   2196 		bcopy(&llc1, mtod(*mp, void *), sizeof(struct llc));
   2197 	}
   2198 
   2199 	M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
   2200 	if (*mp == NULL)
   2201 		return error;
   2202 	bcopy(&eh2, mtod(*mp, void *), ETHER_HDR_LEN);
   2203 
   2204 	return 0;
   2205 
   2206     bad:
   2207 	m_freem(*mp);
   2208 	*mp = NULL;
   2209 	return error;
   2210 }
   2211 
   2212 /*
   2213  * Perform basic checks on header size since
   2214  * IPF assumes ip_input has already processed
   2215  * it for it.  Cut-and-pasted from ip_input.c.
   2216  * Given how simple the IPv6 version is,
   2217  * does the IPv4 version really need to be
   2218  * this complicated?
   2219  *
   2220  * XXX Should we update ipstat here, or not?
   2221  * XXX Right now we update ipstat but not
   2222  * XXX csum_counter.
   2223  */
   2224 static int
   2225 bridge_ip_checkbasic(struct mbuf **mp)
   2226 {
   2227 	struct mbuf *m = *mp;
   2228 	struct ip *ip;
   2229 	int len, hlen;
   2230 
   2231 	if (*mp == NULL)
   2232 		return -1;
   2233 
   2234 	if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
   2235 		if ((m = m_copyup(m, sizeof(struct ip),
   2236 			(max_linkhdr + 3) & ~3)) == NULL) {
   2237 			/* XXXJRT new stat, please */
   2238 			ip_statinc(IP_STAT_TOOSMALL);
   2239 			goto bad;
   2240 		}
   2241 	} else if (__predict_false(m->m_len < sizeof (struct ip))) {
   2242 		if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
   2243 			ip_statinc(IP_STAT_TOOSMALL);
   2244 			goto bad;
   2245 		}
   2246 	}
   2247 	ip = mtod(m, struct ip *);
   2248 	if (ip == NULL) goto bad;
   2249 
   2250 	if (ip->ip_v != IPVERSION) {
   2251 		ip_statinc(IP_STAT_BADVERS);
   2252 		goto bad;
   2253 	}
   2254 	hlen = ip->ip_hl << 2;
   2255 	if (hlen < sizeof(struct ip)) { /* minimum header length */
   2256 		ip_statinc(IP_STAT_BADHLEN);
   2257 		goto bad;
   2258 	}
   2259 	if (hlen > m->m_len) {
   2260 		if ((m = m_pullup(m, hlen)) == 0) {
   2261 			ip_statinc(IP_STAT_BADHLEN);
   2262 			goto bad;
   2263 		}
   2264 		ip = mtod(m, struct ip *);
   2265 		if (ip == NULL) goto bad;
   2266 	}
   2267 
   2268         switch (m->m_pkthdr.csum_flags &
   2269                 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) |
   2270                  M_CSUM_IPv4_BAD)) {
   2271         case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
   2272                 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); */
   2273                 goto bad;
   2274 
   2275         case M_CSUM_IPv4:
   2276                 /* Checksum was okay. */
   2277                 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); */
   2278                 break;
   2279 
   2280         default:
   2281                 /* Must compute it ourselves. */
   2282                 /* INET_CSUM_COUNTER_INCR(&ip_swcsum); */
   2283                 if (in_cksum(m, hlen) != 0)
   2284                         goto bad;
   2285                 break;
   2286         }
   2287 
   2288         /* Retrieve the packet length. */
   2289         len = ntohs(ip->ip_len);
   2290 
   2291         /*
   2292          * Check for additional length bogosity
   2293          */
   2294         if (len < hlen) {
   2295 		ip_statinc(IP_STAT_BADLEN);
   2296                 goto bad;
   2297         }
   2298 
   2299         /*
   2300          * Check that the amount of data in the buffers
   2301          * is as at least much as the IP header would have us expect.
   2302          * Drop packet if shorter than we expect.
   2303          */
   2304         if (m->m_pkthdr.len < len) {
   2305 		ip_statinc(IP_STAT_TOOSHORT);
   2306                 goto bad;
   2307         }
   2308 
   2309 	/* Checks out, proceed */
   2310 	*mp = m;
   2311 	return 0;
   2312 
   2313     bad:
   2314 	*mp = m;
   2315 	return -1;
   2316 }
   2317 
   2318 # ifdef INET6
   2319 /*
   2320  * Same as above, but for IPv6.
   2321  * Cut-and-pasted from ip6_input.c.
   2322  * XXX Should we update ip6stat, or not?
   2323  */
   2324 static int
   2325 bridge_ip6_checkbasic(struct mbuf **mp)
   2326 {
   2327 	struct mbuf *m = *mp;
   2328 	struct ip6_hdr *ip6;
   2329 
   2330         /*
   2331          * If the IPv6 header is not aligned, slurp it up into a new
   2332          * mbuf with space for link headers, in the event we forward
   2333          * it.  Otherwise, if it is aligned, make sure the entire base
   2334          * IPv6 header is in the first mbuf of the chain.
   2335          */
   2336         if (IP6_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
   2337                 struct ifnet *inifp = m->m_pkthdr.rcvif;
   2338                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
   2339                                   (max_linkhdr + 3) & ~3)) == NULL) {
   2340                         /* XXXJRT new stat, please */
   2341 			ip6_statinc(IP6_STAT_TOOSMALL);
   2342                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
   2343                         goto bad;
   2344                 }
   2345         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
   2346                 struct ifnet *inifp = m->m_pkthdr.rcvif;
   2347                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
   2348 			ip6_statinc(IP6_STAT_TOOSMALL);
   2349                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
   2350                         goto bad;
   2351                 }
   2352         }
   2353 
   2354         ip6 = mtod(m, struct ip6_hdr *);
   2355 
   2356         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
   2357 		ip6_statinc(IP6_STAT_BADVERS);
   2358                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
   2359                 goto bad;
   2360         }
   2361 
   2362 	/* Checks out, proceed */
   2363 	*mp = m;
   2364 	return 0;
   2365 
   2366     bad:
   2367 	*mp = m;
   2368 	return -1;
   2369 }
   2370 # endif /* INET6 */
   2371 #endif /* BRIDGE_IPF */
   2372