Home | History | Annotate | Line # | Download | only in net
if_bridge.c revision 1.169
      1 /*	$NetBSD: if_bridge.c,v 1.169 2020/03/24 13:30:54 jdolecek 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 1.169 2020/03/24 13:30:54 jdolecek 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 #include <net/ether_sw_offload.h>
    116 
    117 #if defined(BRIDGE_IPF)
    118 /* Used for bridge_ip[6]_checkbasic */
    119 #include <netinet/in.h>
    120 #include <netinet/in_systm.h>
    121 #include <netinet/ip.h>
    122 #include <netinet/ip_var.h>
    123 #include <netinet/ip_private.h>		/* XXX */
    124 
    125 #include <netinet/ip6.h>
    126 #include <netinet6/in6_var.h>
    127 #include <netinet6/ip6_var.h>
    128 #include <netinet6/ip6_private.h>	/* XXX */
    129 #endif /* BRIDGE_IPF */
    130 
    131 /*
    132  * Size of the route hash table.  Must be a power of two.
    133  */
    134 #ifndef BRIDGE_RTHASH_SIZE
    135 #define	BRIDGE_RTHASH_SIZE		1024
    136 #endif
    137 
    138 #define	BRIDGE_RTHASH_MASK		(BRIDGE_RTHASH_SIZE - 1)
    139 
    140 #include "carp.h"
    141 #if NCARP > 0
    142 #include <netinet/in.h>
    143 #include <netinet/in_var.h>
    144 #include <netinet/ip_carp.h>
    145 #endif
    146 
    147 #include "ioconf.h"
    148 
    149 __CTASSERT(sizeof(struct ifbifconf) == sizeof(struct ifbaconf));
    150 __CTASSERT(offsetof(struct ifbifconf, ifbic_len) == offsetof(struct ifbaconf, ifbac_len));
    151 __CTASSERT(offsetof(struct ifbifconf, ifbic_buf) == offsetof(struct ifbaconf, ifbac_buf));
    152 
    153 /*
    154  * Maximum number of addresses to cache.
    155  */
    156 #ifndef BRIDGE_RTABLE_MAX
    157 #define	BRIDGE_RTABLE_MAX		100
    158 #endif
    159 
    160 /*
    161  * Spanning tree defaults.
    162  */
    163 #define	BSTP_DEFAULT_MAX_AGE		(20 * 256)
    164 #define	BSTP_DEFAULT_HELLO_TIME		(2 * 256)
    165 #define	BSTP_DEFAULT_FORWARD_DELAY	(15 * 256)
    166 #define	BSTP_DEFAULT_HOLD_TIME		(1 * 256)
    167 #define	BSTP_DEFAULT_BRIDGE_PRIORITY	0x8000
    168 #define	BSTP_DEFAULT_PORT_PRIORITY	0x80
    169 #define	BSTP_DEFAULT_PATH_COST		55
    170 
    171 /*
    172  * Timeout (in seconds) for entries learned dynamically.
    173  */
    174 #ifndef BRIDGE_RTABLE_TIMEOUT
    175 #define	BRIDGE_RTABLE_TIMEOUT		(20 * 60)	/* same as ARP */
    176 #endif
    177 
    178 /*
    179  * Number of seconds between walks of the route list.
    180  */
    181 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
    182 #define	BRIDGE_RTABLE_PRUNE_PERIOD	(5 * 60)
    183 #endif
    184 
    185 #define BRIDGE_RT_LOCK(_sc)	mutex_enter((_sc)->sc_rtlist_lock)
    186 #define BRIDGE_RT_UNLOCK(_sc)	mutex_exit((_sc)->sc_rtlist_lock)
    187 #define BRIDGE_RT_LOCKED(_sc)	mutex_owned((_sc)->sc_rtlist_lock)
    188 
    189 #define BRIDGE_RT_PSZ_PERFORM(_sc) \
    190 				pserialize_perform((_sc)->sc_rtlist_psz)
    191 
    192 #define BRIDGE_RT_RENTER(__s)	do { __s = pserialize_read_enter(); } while (0)
    193 #define BRIDGE_RT_REXIT(__s)	do { pserialize_read_exit(__s); } while (0)
    194 
    195 #define BRIDGE_RTLIST_READER_FOREACH(_brt, _sc)			\
    196 	PSLIST_READER_FOREACH((_brt), &((_sc)->sc_rtlist),		\
    197 	    struct bridge_rtnode, brt_list)
    198 #define BRIDGE_RTLIST_WRITER_FOREACH(_brt, _sc)			\
    199 	PSLIST_WRITER_FOREACH((_brt), &((_sc)->sc_rtlist),		\
    200 	    struct bridge_rtnode, brt_list)
    201 #define BRIDGE_RTLIST_WRITER_INSERT_HEAD(_sc, _brt)			\
    202 	PSLIST_WRITER_INSERT_HEAD(&(_sc)->sc_rtlist, brt, brt_list)
    203 #define BRIDGE_RTLIST_WRITER_REMOVE(_brt)				\
    204 	PSLIST_WRITER_REMOVE((_brt), brt_list)
    205 
    206 #define BRIDGE_RTHASH_READER_FOREACH(_brt, _sc, _hash)			\
    207 	PSLIST_READER_FOREACH((_brt), &(_sc)->sc_rthash[(_hash)],	\
    208 	    struct bridge_rtnode, brt_hash)
    209 #define BRIDGE_RTHASH_WRITER_FOREACH(_brt, _sc, _hash)			\
    210 	PSLIST_WRITER_FOREACH((_brt), &(_sc)->sc_rthash[(_hash)],	\
    211 	    struct bridge_rtnode, brt_hash)
    212 #define BRIDGE_RTHASH_WRITER_INSERT_HEAD(_sc, _hash, _brt)		\
    213 	PSLIST_WRITER_INSERT_HEAD(&(_sc)->sc_rthash[(_hash)], brt, brt_hash)
    214 #define BRIDGE_RTHASH_WRITER_INSERT_AFTER(_brt, _new)			\
    215 	PSLIST_WRITER_INSERT_AFTER((_brt), (_new), brt_hash)
    216 #define BRIDGE_RTHASH_WRITER_REMOVE(_brt)				\
    217 	PSLIST_WRITER_REMOVE((_brt), brt_hash)
    218 
    219 #ifdef NET_MPSAFE
    220 #define DECLARE_LOCK_VARIABLE
    221 #define ACQUIRE_GLOBAL_LOCKS()	do { } while (0)
    222 #define RELEASE_GLOBAL_LOCKS()	do { } while (0)
    223 #else
    224 #define DECLARE_LOCK_VARIABLE	int __s
    225 #define ACQUIRE_GLOBAL_LOCKS()	do {					\
    226 					KERNEL_LOCK(1, NULL);		\
    227 					mutex_enter(softnet_lock);	\
    228 					__s = splsoftnet();		\
    229 				} while (0)
    230 #define RELEASE_GLOBAL_LOCKS()	do {					\
    231 					splx(__s);			\
    232 					mutex_exit(softnet_lock);	\
    233 					KERNEL_UNLOCK_ONE(NULL);	\
    234 				} while (0)
    235 #endif
    236 
    237 struct psref_class *bridge_psref_class __read_mostly;
    238 
    239 int	bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
    240 
    241 static struct pool bridge_rtnode_pool;
    242 
    243 static int	bridge_clone_create(struct if_clone *, int);
    244 static int	bridge_clone_destroy(struct ifnet *);
    245 
    246 static int	bridge_ioctl(struct ifnet *, u_long, void *);
    247 static int	bridge_init(struct ifnet *);
    248 static void	bridge_stop(struct ifnet *, int);
    249 static void	bridge_start(struct ifnet *);
    250 
    251 static void	bridge_input(struct ifnet *, struct mbuf *);
    252 static void	bridge_forward(struct bridge_softc *, struct mbuf *);
    253 
    254 static void	bridge_timer(void *);
    255 
    256 static void	bridge_broadcast(struct bridge_softc *, struct ifnet *,
    257 				 struct mbuf *);
    258 
    259 static int	bridge_rtupdate(struct bridge_softc *, const uint8_t *,
    260 				struct ifnet *, int, uint8_t);
    261 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *);
    262 static void	bridge_rttrim(struct bridge_softc *);
    263 static void	bridge_rtage(struct bridge_softc *);
    264 static void	bridge_rtage_work(struct work *, void *);
    265 static void	bridge_rtflush(struct bridge_softc *, int);
    266 static int	bridge_rtdaddr(struct bridge_softc *, const uint8_t *);
    267 static void	bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp);
    268 
    269 static void	bridge_rtable_init(struct bridge_softc *);
    270 static void	bridge_rtable_fini(struct bridge_softc *);
    271 
    272 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
    273 						  const uint8_t *);
    274 static int	bridge_rtnode_insert(struct bridge_softc *,
    275 				     struct bridge_rtnode *);
    276 static void	bridge_rtnode_remove(struct bridge_softc *,
    277 				     struct bridge_rtnode *);
    278 static void	bridge_rtnode_destroy(struct bridge_rtnode *);
    279 
    280 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
    281 						  const char *name,
    282 						  struct psref *);
    283 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
    284 						     struct ifnet *ifp,
    285 						     struct psref *);
    286 static void	bridge_release_member(struct bridge_softc *, struct bridge_iflist *,
    287                                       struct psref *);
    288 static void	bridge_delete_member(struct bridge_softc *,
    289 				     struct bridge_iflist *);
    290 static void	bridge_acquire_member(struct bridge_softc *sc,
    291                                       struct bridge_iflist *,
    292                                       struct psref *);
    293 
    294 static int	bridge_ioctl_add(struct bridge_softc *, void *);
    295 static int	bridge_ioctl_del(struct bridge_softc *, void *);
    296 static int	bridge_ioctl_gifflags(struct bridge_softc *, void *);
    297 static int	bridge_ioctl_sifflags(struct bridge_softc *, void *);
    298 static int	bridge_ioctl_scache(struct bridge_softc *, void *);
    299 static int	bridge_ioctl_gcache(struct bridge_softc *, void *);
    300 static int	bridge_ioctl_gifs(struct bridge_softc *, void *);
    301 static int	bridge_ioctl_rts(struct bridge_softc *, void *);
    302 static int	bridge_ioctl_saddr(struct bridge_softc *, void *);
    303 static int	bridge_ioctl_sto(struct bridge_softc *, void *);
    304 static int	bridge_ioctl_gto(struct bridge_softc *, void *);
    305 static int	bridge_ioctl_daddr(struct bridge_softc *, void *);
    306 static int	bridge_ioctl_flush(struct bridge_softc *, void *);
    307 static int	bridge_ioctl_gpri(struct bridge_softc *, void *);
    308 static int	bridge_ioctl_spri(struct bridge_softc *, void *);
    309 static int	bridge_ioctl_ght(struct bridge_softc *, void *);
    310 static int	bridge_ioctl_sht(struct bridge_softc *, void *);
    311 static int	bridge_ioctl_gfd(struct bridge_softc *, void *);
    312 static int	bridge_ioctl_sfd(struct bridge_softc *, void *);
    313 static int	bridge_ioctl_gma(struct bridge_softc *, void *);
    314 static int	bridge_ioctl_sma(struct bridge_softc *, void *);
    315 static int	bridge_ioctl_sifprio(struct bridge_softc *, void *);
    316 static int	bridge_ioctl_sifcost(struct bridge_softc *, void *);
    317 #if defined(BRIDGE_IPF)
    318 static int	bridge_ioctl_gfilt(struct bridge_softc *, void *);
    319 static int	bridge_ioctl_sfilt(struct bridge_softc *, void *);
    320 static int	bridge_ipf(void *, struct mbuf **, struct ifnet *, int);
    321 static int	bridge_ip_checkbasic(struct mbuf **mp);
    322 # ifdef INET6
    323 static int	bridge_ip6_checkbasic(struct mbuf **mp);
    324 # endif /* INET6 */
    325 #endif /* BRIDGE_IPF */
    326 
    327 struct bridge_control {
    328 	int	(*bc_func)(struct bridge_softc *, void *);
    329 	int	bc_argsize;
    330 	int	bc_flags;
    331 };
    332 
    333 #define	BC_F_COPYIN		0x01	/* copy arguments in */
    334 #define	BC_F_COPYOUT		0x02	/* copy arguments out */
    335 #define	BC_F_SUSER		0x04	/* do super-user check */
    336 #define BC_F_XLATEIN		0x08	/* xlate arguments in */
    337 #define BC_F_XLATEOUT		0x10	/* xlate arguments out */
    338 
    339 static const struct bridge_control bridge_control_table[] = {
    340 [BRDGADD] = {bridge_ioctl_add, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    341 [BRDGDEL] = {bridge_ioctl_del, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    342 
    343 [BRDGGIFFLGS] = {bridge_ioctl_gifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_COPYOUT},
    344 [BRDGSIFFLGS] = {bridge_ioctl_sifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    345 
    346 [BRDGSCACHE] = {bridge_ioctl_scache, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    347 [BRDGGCACHE] = {bridge_ioctl_gcache, sizeof(struct ifbrparam), BC_F_COPYOUT},
    348 
    349 [OBRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_COPYIN|BC_F_COPYOUT},
    350 [OBRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_COPYIN|BC_F_COPYOUT},
    351 
    352 [BRDGSADDR] = {bridge_ioctl_saddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
    353 
    354 [BRDGSTO] = {bridge_ioctl_sto, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    355 [BRDGGTO] = {bridge_ioctl_gto, sizeof(struct ifbrparam), BC_F_COPYOUT},
    356 
    357 [BRDGDADDR] = {bridge_ioctl_daddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
    358 
    359 [BRDGFLUSH] = {bridge_ioctl_flush, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    360 
    361 [BRDGGPRI] = {bridge_ioctl_gpri, sizeof(struct ifbrparam), BC_F_COPYOUT},
    362 [BRDGSPRI] = {bridge_ioctl_spri, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    363 
    364 [BRDGGHT] = {bridge_ioctl_ght, sizeof(struct ifbrparam), BC_F_COPYOUT},
    365 [BRDGSHT] = {bridge_ioctl_sht, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    366 
    367 [BRDGGFD] = {bridge_ioctl_gfd, sizeof(struct ifbrparam), BC_F_COPYOUT},
    368 [BRDGSFD] = {bridge_ioctl_sfd, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    369 
    370 [BRDGGMA] = {bridge_ioctl_gma, sizeof(struct ifbrparam), BC_F_COPYOUT},
    371 [BRDGSMA] = {bridge_ioctl_sma, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    372 
    373 [BRDGSIFPRIO] = {bridge_ioctl_sifprio, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    374 
    375 [BRDGSIFCOST] = {bridge_ioctl_sifcost, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    376 #if defined(BRIDGE_IPF)
    377 [BRDGGFILT] = {bridge_ioctl_gfilt, sizeof(struct ifbrparam), BC_F_COPYOUT},
    378 [BRDGSFILT] = {bridge_ioctl_sfilt, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    379 #endif /* BRIDGE_IPF */
    380 [BRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_XLATEIN|BC_F_XLATEOUT},
    381 [BRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_XLATEIN|BC_F_XLATEOUT},
    382 };
    383 
    384 static const int bridge_control_table_size = __arraycount(bridge_control_table);
    385 
    386 static struct if_clone bridge_cloner =
    387     IF_CLONE_INITIALIZER("bridge", bridge_clone_create, bridge_clone_destroy);
    388 
    389 /*
    390  * bridgeattach:
    391  *
    392  *	Pseudo-device attach routine.
    393  */
    394 void
    395 bridgeattach(int n)
    396 {
    397 
    398 	pool_init(&bridge_rtnode_pool, sizeof(struct bridge_rtnode),
    399 	    0, 0, 0, "brtpl", NULL, IPL_NET);
    400 
    401 	bridge_psref_class = psref_class_create("bridge", IPL_SOFTNET);
    402 
    403 	if_clone_attach(&bridge_cloner);
    404 }
    405 
    406 /*
    407  * bridge_clone_create:
    408  *
    409  *	Create a new bridge instance.
    410  */
    411 static int
    412 bridge_clone_create(struct if_clone *ifc, int unit)
    413 {
    414 	struct bridge_softc *sc;
    415 	struct ifnet *ifp;
    416 	int error;
    417 
    418 	sc = kmem_zalloc(sizeof(*sc),  KM_SLEEP);
    419 	ifp = &sc->sc_if;
    420 
    421 	sc->sc_brtmax = BRIDGE_RTABLE_MAX;
    422 	sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
    423 	sc->sc_bridge_max_age = BSTP_DEFAULT_MAX_AGE;
    424 	sc->sc_bridge_hello_time = BSTP_DEFAULT_HELLO_TIME;
    425 	sc->sc_bridge_forward_delay = BSTP_DEFAULT_FORWARD_DELAY;
    426 	sc->sc_bridge_priority = BSTP_DEFAULT_BRIDGE_PRIORITY;
    427 	sc->sc_hold_time = BSTP_DEFAULT_HOLD_TIME;
    428 	sc->sc_filter_flags = 0;
    429 
    430 	/* Initialize our routing table. */
    431 	bridge_rtable_init(sc);
    432 
    433 	error = workqueue_create(&sc->sc_rtage_wq, "bridge_rtage",
    434 	    bridge_rtage_work, sc, PRI_SOFTNET, IPL_SOFTNET, WQ_MPSAFE);
    435 	if (error)
    436 		panic("%s: workqueue_create %d\n", __func__, error);
    437 
    438 	callout_init(&sc->sc_brcallout, CALLOUT_MPSAFE);
    439 	callout_init(&sc->sc_bstpcallout, CALLOUT_MPSAFE);
    440 
    441 	mutex_init(&sc->sc_iflist_psref.bip_lock, MUTEX_DEFAULT, IPL_NONE);
    442 	PSLIST_INIT(&sc->sc_iflist_psref.bip_iflist);
    443 	sc->sc_iflist_psref.bip_psz = pserialize_create();
    444 
    445 	if_initname(ifp, ifc->ifc_name, unit);
    446 	ifp->if_softc = sc;
    447 	ifp->if_extflags = IFEF_NO_LINK_STATE_CHANGE;
    448 #ifdef NET_MPSAFE
    449 	ifp->if_extflags |= IFEF_MPSAFE;
    450 #endif
    451 	ifp->if_mtu = ETHERMTU;
    452 	ifp->if_ioctl = bridge_ioctl;
    453 	ifp->if_output = bridge_output;
    454 	ifp->if_start = bridge_start;
    455 	ifp->if_stop = bridge_stop;
    456 	ifp->if_init = bridge_init;
    457 	ifp->if_type = IFT_BRIDGE;
    458 	ifp->if_addrlen = 0;
    459 	ifp->if_dlt = DLT_EN10MB;
    460 	ifp->if_hdrlen = ETHER_HDR_LEN;
    461 
    462 	error = if_initialize(ifp);
    463 	if (error != 0) {
    464 		pserialize_destroy(sc->sc_iflist_psref.bip_psz);
    465 		mutex_destroy(&sc->sc_iflist_psref.bip_lock);
    466 		callout_destroy(&sc->sc_brcallout);
    467 		callout_destroy(&sc->sc_bstpcallout);
    468 		workqueue_destroy(sc->sc_rtage_wq);
    469 		bridge_rtable_fini(sc);
    470 		kmem_free(sc, sizeof(*sc));
    471 
    472 		return error;
    473 	}
    474 	if_alloc_sadl(ifp);
    475 	if_register(ifp);
    476 
    477 	return 0;
    478 }
    479 
    480 /*
    481  * bridge_clone_destroy:
    482  *
    483  *	Destroy a bridge instance.
    484  */
    485 static int
    486 bridge_clone_destroy(struct ifnet *ifp)
    487 {
    488 	struct bridge_softc *sc = ifp->if_softc;
    489 	struct bridge_iflist *bif;
    490 
    491 	if ((ifp->if_flags & IFF_RUNNING) != 0)
    492 		bridge_stop(ifp, 1);
    493 
    494 	BRIDGE_LOCK(sc);
    495 	for (;;) {
    496 		bif = PSLIST_WRITER_FIRST(&sc->sc_iflist_psref.bip_iflist, struct bridge_iflist,
    497 		    bif_next);
    498 		if (bif == NULL)
    499 			break;
    500 		bridge_delete_member(sc, bif);
    501 	}
    502 	PSLIST_DESTROY(&sc->sc_iflist_psref.bip_iflist);
    503 	BRIDGE_UNLOCK(sc);
    504 
    505 	if_detach(ifp);
    506 
    507 	/* Tear down the routing table. */
    508 	bridge_rtable_fini(sc);
    509 
    510 	pserialize_destroy(sc->sc_iflist_psref.bip_psz);
    511 	mutex_destroy(&sc->sc_iflist_psref.bip_lock);
    512 	callout_destroy(&sc->sc_brcallout);
    513 	callout_destroy(&sc->sc_bstpcallout);
    514 	workqueue_destroy(sc->sc_rtage_wq);
    515 	kmem_free(sc, sizeof(*sc));
    516 
    517 	return 0;
    518 }
    519 
    520 /*
    521  * bridge_ioctl:
    522  *
    523  *	Handle a control request from the operator.
    524  */
    525 static int
    526 bridge_ioctl(struct ifnet *ifp, u_long cmd, void *data)
    527 {
    528 	struct bridge_softc *sc = ifp->if_softc;
    529 	struct lwp *l = curlwp;	/* XXX */
    530 	union {
    531 		struct ifbreq ifbreq;
    532 		struct ifbifconf ifbifconf;
    533 		struct ifbareq ifbareq;
    534 		struct ifbaconf ifbaconf;
    535 		struct ifbrparam ifbrparam;
    536 	} args;
    537 	struct ifdrv *ifd = (struct ifdrv *) data;
    538 	const struct bridge_control *bc = NULL; /* XXXGCC */
    539 	int s, error = 0;
    540 
    541 	/* Authorize command before calling splsoftnet(). */
    542 	switch (cmd) {
    543 	case SIOCGDRVSPEC:
    544 	case SIOCSDRVSPEC:
    545 		if (ifd->ifd_cmd >= bridge_control_table_size
    546 		    || (bc = &bridge_control_table[ifd->ifd_cmd]) == NULL) {
    547 			error = EINVAL;
    548 			return error;
    549 		}
    550 
    551 		/* We only care about BC_F_SUSER at this point. */
    552 		if ((bc->bc_flags & BC_F_SUSER) == 0)
    553 			break;
    554 
    555 		error = kauth_authorize_network(l->l_cred,
    556 		    KAUTH_NETWORK_INTERFACE_BRIDGE,
    557 		    cmd == SIOCGDRVSPEC ?
    558 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_GETPRIV :
    559 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_SETPRIV,
    560 		     ifd, NULL, NULL);
    561 		if (error)
    562 			return error;
    563 
    564 		break;
    565 	}
    566 
    567 	s = splsoftnet();
    568 
    569 	switch (cmd) {
    570 	case SIOCGDRVSPEC:
    571 	case SIOCSDRVSPEC:
    572 		KASSERT(bc != NULL);
    573 		if (cmd == SIOCGDRVSPEC &&
    574 		    (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) == 0) {
    575 			error = EINVAL;
    576 			break;
    577 		}
    578 		else if (cmd == SIOCSDRVSPEC &&
    579 		    (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) != 0) {
    580 			error = EINVAL;
    581 			break;
    582 		}
    583 
    584 		/* BC_F_SUSER is checked above, before splsoftnet(). */
    585 
    586 		if ((bc->bc_flags & (BC_F_XLATEIN|BC_F_XLATEOUT)) == 0
    587 		    && (ifd->ifd_len != bc->bc_argsize
    588 			|| ifd->ifd_len > sizeof(args))) {
    589 			error = EINVAL;
    590 			break;
    591 		}
    592 
    593 		memset(&args, 0, sizeof(args));
    594 		if (bc->bc_flags & BC_F_COPYIN) {
    595 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
    596 			if (error)
    597 				break;
    598 		} else if (bc->bc_flags & BC_F_XLATEIN) {
    599 			args.ifbifconf.ifbic_len = ifd->ifd_len;
    600 			args.ifbifconf.ifbic_buf = ifd->ifd_data;
    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 		} else if (bc->bc_flags & BC_F_XLATEOUT) {
    610 			ifd->ifd_len = args.ifbifconf.ifbic_len;
    611 			ifd->ifd_data = args.ifbifconf.ifbic_buf;
    612 		}
    613 		break;
    614 
    615 	case SIOCSIFFLAGS:
    616 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
    617 			break;
    618 		switch (ifp->if_flags & (IFF_UP|IFF_RUNNING)) {
    619 		case IFF_RUNNING:
    620 			/*
    621 			 * If interface is marked down and it is running,
    622 			 * then stop and disable it.
    623 			 */
    624 			(*ifp->if_stop)(ifp, 1);
    625 			break;
    626 		case IFF_UP:
    627 			/*
    628 			 * If interface is marked up and it is stopped, then
    629 			 * start it.
    630 			 */
    631 			error = (*ifp->if_init)(ifp);
    632 			break;
    633 		default:
    634 			break;
    635 		}
    636 		break;
    637 
    638 	case SIOCSIFMTU:
    639 		if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET)
    640 			error = 0;
    641 		break;
    642 
    643 	default:
    644 		error = ifioctl_common(ifp, cmd, data);
    645 		break;
    646 	}
    647 
    648 	splx(s);
    649 
    650 	return error;
    651 }
    652 
    653 /*
    654  * bridge_lookup_member:
    655  *
    656  *	Lookup a bridge member interface.
    657  */
    658 static struct bridge_iflist *
    659 bridge_lookup_member(struct bridge_softc *sc, const char *name, struct psref *psref)
    660 {
    661 	struct bridge_iflist *bif;
    662 	struct ifnet *ifp;
    663 	int s;
    664 
    665 	BRIDGE_PSZ_RENTER(s);
    666 
    667 	BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
    668 		ifp = bif->bif_ifp;
    669 		if (strcmp(ifp->if_xname, name) == 0)
    670 			break;
    671 	}
    672 	if (bif != NULL)
    673 		bridge_acquire_member(sc, bif, psref);
    674 
    675 	BRIDGE_PSZ_REXIT(s);
    676 
    677 	return bif;
    678 }
    679 
    680 /*
    681  * bridge_lookup_member_if:
    682  *
    683  *	Lookup a bridge member interface by ifnet*.
    684  */
    685 static struct bridge_iflist *
    686 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp,
    687     struct psref *psref)
    688 {
    689 	struct bridge_iflist *bif;
    690 	int s;
    691 
    692 	BRIDGE_PSZ_RENTER(s);
    693 
    694 	bif = member_ifp->if_bridgeif;
    695 	if (bif != NULL) {
    696 		psref_acquire(psref, &bif->bif_psref,
    697 		    bridge_psref_class);
    698 	}
    699 
    700 	BRIDGE_PSZ_REXIT(s);
    701 
    702 	return bif;
    703 }
    704 
    705 static void
    706 bridge_acquire_member(struct bridge_softc *sc, struct bridge_iflist *bif,
    707     struct psref *psref)
    708 {
    709 
    710 	psref_acquire(psref, &bif->bif_psref, bridge_psref_class);
    711 }
    712 
    713 /*
    714  * bridge_release_member:
    715  *
    716  *	Release the specified member interface.
    717  */
    718 static void
    719 bridge_release_member(struct bridge_softc *sc, struct bridge_iflist *bif,
    720     struct psref *psref)
    721 {
    722 
    723 	psref_release(psref, &bif->bif_psref, bridge_psref_class);
    724 }
    725 
    726 /*
    727  * bridge_delete_member:
    728  *
    729  *	Delete the specified member interface.
    730  */
    731 static void
    732 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif)
    733 {
    734 	struct ifnet *ifs = bif->bif_ifp;
    735 
    736 	KASSERT(BRIDGE_LOCKED(sc));
    737 
    738 	ifs->_if_input = ether_input;
    739 	ifs->if_bridge = NULL;
    740 	ifs->if_bridgeif = NULL;
    741 
    742 	PSLIST_WRITER_REMOVE(bif, bif_next);
    743 	BRIDGE_PSZ_PERFORM(sc);
    744 	BRIDGE_UNLOCK(sc);
    745 
    746 	switch (ifs->if_type) {
    747 	case IFT_ETHER:
    748 	case IFT_L2TP:
    749 		/*
    750 		 * Take the interface out of promiscuous mode.
    751 		 * Don't call it with holding a spin lock.
    752 		 */
    753 		(void) ifpromisc(ifs, 0);
    754 		IFNET_LOCK(ifs);
    755 		(void) ether_disable_vlan_mtu(ifs);
    756 		IFNET_UNLOCK(ifs);
    757 		break;
    758 	default:
    759 #ifdef DIAGNOSTIC
    760 		panic("%s: impossible", __func__);
    761 #endif
    762 		break;
    763 	}
    764 
    765 	psref_target_destroy(&bif->bif_psref, bridge_psref_class);
    766 
    767 	PSLIST_ENTRY_DESTROY(bif, bif_next);
    768 	kmem_free(bif, sizeof(*bif));
    769 
    770 	BRIDGE_LOCK(sc);
    771 }
    772 
    773 /*
    774  * bridge_calc_csum_flags:
    775  *
    776  *	Calculate logical and b/w csum flags each member interface supports.
    777  */
    778 void
    779 bridge_calc_csum_flags(struct bridge_softc *sc)
    780 {
    781 	struct bridge_iflist *bif;
    782 	struct ifnet *ifs;
    783 	int flags = ~0;
    784 
    785 	BRIDGE_LOCK(sc);
    786 	BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
    787 		ifs = bif->bif_ifp;
    788 		flags &= ifs->if_csum_flags_tx;
    789 	}
    790 	sc->sc_csum_flags_tx = flags;
    791 	BRIDGE_UNLOCK(sc);
    792 }
    793 
    794 static int
    795 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
    796 {
    797 	struct ifbreq *req = arg;
    798 	struct bridge_iflist *bif = NULL;
    799 	struct ifnet *ifs;
    800 	int error = 0;
    801 	struct psref psref;
    802 
    803 	ifs = if_get(req->ifbr_ifsname, &psref);
    804 	if (ifs == NULL)
    805 		return ENOENT;
    806 
    807 	if (ifs->if_bridge == sc) {
    808 		error = EEXIST;
    809 		goto out;
    810 	}
    811 
    812 	if (ifs->if_bridge != NULL) {
    813 		error = EBUSY;
    814 		goto out;
    815 	}
    816 
    817 	if (ifs->_if_input != ether_input) {
    818 		error = EINVAL;
    819 		goto out;
    820 	}
    821 
    822 	/* FIXME: doesn't work with non-IFF_SIMPLEX interfaces */
    823 	if ((ifs->if_flags & IFF_SIMPLEX) == 0) {
    824 		error = EINVAL;
    825 		goto out;
    826 	}
    827 
    828 	bif = kmem_alloc(sizeof(*bif), KM_SLEEP);
    829 
    830 	switch (ifs->if_type) {
    831 	case IFT_ETHER:
    832 		if (sc->sc_if.if_mtu != ifs->if_mtu) {
    833 			error = EINVAL;
    834 			goto out;
    835 		}
    836 		/* FALLTHROUGH */
    837 	case IFT_L2TP:
    838 		IFNET_LOCK(ifs);
    839 		error = ether_enable_vlan_mtu(ifs);
    840 		IFNET_UNLOCK(ifs);
    841 		if (error > 0)
    842 			goto out;
    843 		/*
    844 		 * Place the interface into promiscuous mode.
    845 		 */
    846 		error = ifpromisc(ifs, 1);
    847 		if (error)
    848 			goto out;
    849 		break;
    850 	default:
    851 		error = EINVAL;
    852 		goto out;
    853 	}
    854 
    855 	bif->bif_ifp = ifs;
    856 	bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
    857 	bif->bif_priority = BSTP_DEFAULT_PORT_PRIORITY;
    858 	bif->bif_path_cost = BSTP_DEFAULT_PATH_COST;
    859 	PSLIST_ENTRY_INIT(bif, bif_next);
    860 	psref_target_init(&bif->bif_psref, bridge_psref_class);
    861 
    862 	BRIDGE_LOCK(sc);
    863 
    864 	ifs->if_bridge = sc;
    865 	ifs->if_bridgeif = bif;
    866 	PSLIST_WRITER_INSERT_HEAD(&sc->sc_iflist_psref.bip_iflist, bif, bif_next);
    867 	ifs->_if_input = bridge_input;
    868 
    869 	BRIDGE_UNLOCK(sc);
    870 
    871 	bridge_calc_csum_flags(sc);
    872 
    873 	if (sc->sc_if.if_flags & IFF_RUNNING)
    874 		bstp_initialization(sc);
    875 	else
    876 		bstp_stop(sc);
    877 
    878 out:
    879 	if_put(ifs, &psref);
    880 	if (error) {
    881 		if (bif != NULL)
    882 			kmem_free(bif, sizeof(*bif));
    883 	}
    884 	return error;
    885 }
    886 
    887 static int
    888 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
    889 {
    890 	struct ifbreq *req = arg;
    891 	const char *name = req->ifbr_ifsname;
    892 	struct bridge_iflist *bif;
    893 	struct ifnet *ifs;
    894 
    895 	BRIDGE_LOCK(sc);
    896 
    897 	/*
    898 	 * Don't use bridge_lookup_member. We want to get a member
    899 	 * with bif_refs == 0.
    900 	 */
    901 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) {
    902 		ifs = bif->bif_ifp;
    903 		if (strcmp(ifs->if_xname, name) == 0)
    904 			break;
    905 	}
    906 
    907 	if (bif == NULL) {
    908 		BRIDGE_UNLOCK(sc);
    909 		return ENOENT;
    910 	}
    911 
    912 	bridge_delete_member(sc, bif);
    913 
    914 	BRIDGE_UNLOCK(sc);
    915 
    916 	bridge_rtdelete(sc, ifs);
    917 	bridge_calc_csum_flags(sc);
    918 
    919 	if (sc->sc_if.if_flags & IFF_RUNNING)
    920 		bstp_initialization(sc);
    921 
    922 	return 0;
    923 }
    924 
    925 static int
    926 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
    927 {
    928 	struct ifbreq *req = arg;
    929 	struct bridge_iflist *bif;
    930 	struct psref psref;
    931 
    932 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
    933 	if (bif == NULL)
    934 		return ENOENT;
    935 
    936 	req->ifbr_ifsflags = bif->bif_flags;
    937 	req->ifbr_state = bif->bif_state;
    938 	req->ifbr_priority = bif->bif_priority;
    939 	req->ifbr_path_cost = bif->bif_path_cost;
    940 	req->ifbr_portno = bif->bif_ifp->if_index & 0xff;
    941 
    942 	bridge_release_member(sc, bif, &psref);
    943 
    944 	return 0;
    945 }
    946 
    947 static int
    948 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
    949 {
    950 	struct ifbreq *req = arg;
    951 	struct bridge_iflist *bif;
    952 	struct psref psref;
    953 
    954 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
    955 	if (bif == NULL)
    956 		return ENOENT;
    957 
    958 	if (req->ifbr_ifsflags & IFBIF_STP) {
    959 		switch (bif->bif_ifp->if_type) {
    960 		case IFT_ETHER:
    961 		case IFT_L2TP:
    962 			/* These can do spanning tree. */
    963 			break;
    964 
    965 		default:
    966 			/* Nothing else can. */
    967 			bridge_release_member(sc, bif, &psref);
    968 			return EINVAL;
    969 		}
    970 	}
    971 
    972 	bif->bif_flags = req->ifbr_ifsflags;
    973 
    974 	bridge_release_member(sc, bif, &psref);
    975 
    976 	if (sc->sc_if.if_flags & IFF_RUNNING)
    977 		bstp_initialization(sc);
    978 
    979 	return 0;
    980 }
    981 
    982 static int
    983 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
    984 {
    985 	struct ifbrparam *param = arg;
    986 
    987 	sc->sc_brtmax = param->ifbrp_csize;
    988 	bridge_rttrim(sc);
    989 
    990 	return 0;
    991 }
    992 
    993 static int
    994 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
    995 {
    996 	struct ifbrparam *param = arg;
    997 
    998 	param->ifbrp_csize = sc->sc_brtmax;
    999 
   1000 	return 0;
   1001 }
   1002 
   1003 static int
   1004 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
   1005 {
   1006 	struct ifbifconf *bifc = arg;
   1007 	struct bridge_iflist *bif;
   1008 	struct ifbreq *breqs;
   1009 	int i, count, error = 0;
   1010 
   1011 retry:
   1012 	BRIDGE_LOCK(sc);
   1013 	count = 0;
   1014 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc)
   1015 		count++;
   1016 	BRIDGE_UNLOCK(sc);
   1017 
   1018 	if (count == 0) {
   1019 		bifc->ifbic_len = 0;
   1020 		return 0;
   1021 	}
   1022 
   1023 	if (bifc->ifbic_len == 0 || bifc->ifbic_len < (sizeof(*breqs) * count)) {
   1024 		/* Tell that a larger buffer is needed */
   1025 		bifc->ifbic_len = sizeof(*breqs) * count;
   1026 		return 0;
   1027 	}
   1028 
   1029 	breqs = kmem_alloc(sizeof(*breqs) * count, KM_SLEEP);
   1030 
   1031 	BRIDGE_LOCK(sc);
   1032 
   1033 	i = 0;
   1034 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc)
   1035 		i++;
   1036 	if (i > count) {
   1037 		/*
   1038 		 * The number of members has been increased.
   1039 		 * We need more memory!
   1040 		 */
   1041 		BRIDGE_UNLOCK(sc);
   1042 		kmem_free(breqs, sizeof(*breqs) * count);
   1043 		goto retry;
   1044 	}
   1045 
   1046 	i = 0;
   1047 	BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) {
   1048 		struct ifbreq *breq = &breqs[i++];
   1049 		memset(breq, 0, sizeof(*breq));
   1050 
   1051 		strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname,
   1052 		    sizeof(breq->ifbr_ifsname));
   1053 		breq->ifbr_ifsflags = bif->bif_flags;
   1054 		breq->ifbr_state = bif->bif_state;
   1055 		breq->ifbr_priority = bif->bif_priority;
   1056 		breq->ifbr_path_cost = bif->bif_path_cost;
   1057 		breq->ifbr_portno = bif->bif_ifp->if_index & 0xff;
   1058 	}
   1059 
   1060 	/* Don't call copyout with holding the mutex */
   1061 	BRIDGE_UNLOCK(sc);
   1062 
   1063 	for (i = 0; i < count; i++) {
   1064 		error = copyout(&breqs[i], bifc->ifbic_req + i, sizeof(*breqs));
   1065 		if (error)
   1066 			break;
   1067 	}
   1068 	bifc->ifbic_len = sizeof(*breqs) * i;
   1069 
   1070 	kmem_free(breqs, sizeof(*breqs) * count);
   1071 
   1072 	return error;
   1073 }
   1074 
   1075 static int
   1076 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
   1077 {
   1078 	struct ifbaconf *bac = arg;
   1079 	struct bridge_rtnode *brt;
   1080 	struct ifbareq bareq;
   1081 	int count = 0, error = 0, len;
   1082 
   1083 	if (bac->ifbac_len == 0)
   1084 		return 0;
   1085 
   1086 	BRIDGE_RT_LOCK(sc);
   1087 
   1088 	/* The passed buffer is not enough, tell a required size. */
   1089 	if (bac->ifbac_len < (sizeof(bareq) * sc->sc_brtcnt)) {
   1090 		count = sc->sc_brtcnt;
   1091 		goto out;
   1092 	}
   1093 
   1094 	len = bac->ifbac_len;
   1095 	BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) {
   1096 		if (len < sizeof(bareq))
   1097 			goto out;
   1098 		memset(&bareq, 0, sizeof(bareq));
   1099 		strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
   1100 		    sizeof(bareq.ifba_ifsname));
   1101 		memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
   1102 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
   1103 			bareq.ifba_expire = brt->brt_expire - time_uptime;
   1104 		} else
   1105 			bareq.ifba_expire = 0;
   1106 		bareq.ifba_flags = brt->brt_flags;
   1107 
   1108 		error = copyout(&bareq, bac->ifbac_req + count, sizeof(bareq));
   1109 		if (error)
   1110 			goto out;
   1111 		count++;
   1112 		len -= sizeof(bareq);
   1113 	}
   1114 out:
   1115 	BRIDGE_RT_UNLOCK(sc);
   1116 
   1117 	bac->ifbac_len = sizeof(bareq) * count;
   1118 	return error;
   1119 }
   1120 
   1121 static int
   1122 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
   1123 {
   1124 	struct ifbareq *req = arg;
   1125 	struct bridge_iflist *bif;
   1126 	int error;
   1127 	struct psref psref;
   1128 
   1129 	bif = bridge_lookup_member(sc, req->ifba_ifsname, &psref);
   1130 	if (bif == NULL)
   1131 		return ENOENT;
   1132 
   1133 	error = bridge_rtupdate(sc, req->ifba_dst, bif->bif_ifp, 1,
   1134 	    req->ifba_flags);
   1135 
   1136 	bridge_release_member(sc, bif, &psref);
   1137 
   1138 	return error;
   1139 }
   1140 
   1141 static int
   1142 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
   1143 {
   1144 	struct ifbrparam *param = arg;
   1145 
   1146 	sc->sc_brttimeout = param->ifbrp_ctime;
   1147 
   1148 	return 0;
   1149 }
   1150 
   1151 static int
   1152 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
   1153 {
   1154 	struct ifbrparam *param = arg;
   1155 
   1156 	param->ifbrp_ctime = sc->sc_brttimeout;
   1157 
   1158 	return 0;
   1159 }
   1160 
   1161 static int
   1162 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
   1163 {
   1164 	struct ifbareq *req = arg;
   1165 
   1166 	return (bridge_rtdaddr(sc, req->ifba_dst));
   1167 }
   1168 
   1169 static int
   1170 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
   1171 {
   1172 	struct ifbreq *req = arg;
   1173 
   1174 	bridge_rtflush(sc, req->ifbr_ifsflags);
   1175 
   1176 	return 0;
   1177 }
   1178 
   1179 static int
   1180 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
   1181 {
   1182 	struct ifbrparam *param = arg;
   1183 
   1184 	param->ifbrp_prio = sc->sc_bridge_priority;
   1185 
   1186 	return 0;
   1187 }
   1188 
   1189 static int
   1190 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
   1191 {
   1192 	struct ifbrparam *param = arg;
   1193 
   1194 	sc->sc_bridge_priority = param->ifbrp_prio;
   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_ght(struct bridge_softc *sc, void *arg)
   1204 {
   1205 	struct ifbrparam *param = arg;
   1206 
   1207 	param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8;
   1208 
   1209 	return 0;
   1210 }
   1211 
   1212 static int
   1213 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
   1214 {
   1215 	struct ifbrparam *param = arg;
   1216 
   1217 	if (param->ifbrp_hellotime == 0)
   1218 		return EINVAL;
   1219 	sc->sc_bridge_hello_time = param->ifbrp_hellotime << 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_gfd(struct bridge_softc *sc, void *arg)
   1229 {
   1230 	struct ifbrparam *param = arg;
   1231 
   1232 	param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8;
   1233 
   1234 	return 0;
   1235 }
   1236 
   1237 static int
   1238 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
   1239 {
   1240 	struct ifbrparam *param = arg;
   1241 
   1242 	if (param->ifbrp_fwddelay == 0)
   1243 		return EINVAL;
   1244 	sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8;
   1245 
   1246 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1247 		bstp_initialization(sc);
   1248 
   1249 	return 0;
   1250 }
   1251 
   1252 static int
   1253 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
   1254 {
   1255 	struct ifbrparam *param = arg;
   1256 
   1257 	param->ifbrp_maxage = sc->sc_bridge_max_age >> 8;
   1258 
   1259 	return 0;
   1260 }
   1261 
   1262 static int
   1263 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
   1264 {
   1265 	struct ifbrparam *param = arg;
   1266 
   1267 	if (param->ifbrp_maxage == 0)
   1268 		return EINVAL;
   1269 	sc->sc_bridge_max_age = param->ifbrp_maxage << 8;
   1270 
   1271 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1272 		bstp_initialization(sc);
   1273 
   1274 	return 0;
   1275 }
   1276 
   1277 static int
   1278 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
   1279 {
   1280 	struct ifbreq *req = arg;
   1281 	struct bridge_iflist *bif;
   1282 	struct psref psref;
   1283 
   1284 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
   1285 	if (bif == NULL)
   1286 		return ENOENT;
   1287 
   1288 	bif->bif_priority = req->ifbr_priority;
   1289 
   1290 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1291 		bstp_initialization(sc);
   1292 
   1293 	bridge_release_member(sc, bif, &psref);
   1294 
   1295 	return 0;
   1296 }
   1297 
   1298 #if defined(BRIDGE_IPF)
   1299 static int
   1300 bridge_ioctl_gfilt(struct bridge_softc *sc, void *arg)
   1301 {
   1302 	struct ifbrparam *param = arg;
   1303 
   1304 	param->ifbrp_filter = sc->sc_filter_flags;
   1305 
   1306 	return 0;
   1307 }
   1308 
   1309 static int
   1310 bridge_ioctl_sfilt(struct bridge_softc *sc, void *arg)
   1311 {
   1312 	struct ifbrparam *param = arg;
   1313 	uint32_t nflags, oflags;
   1314 
   1315 	if (param->ifbrp_filter & ~IFBF_FILT_MASK)
   1316 		return EINVAL;
   1317 
   1318 	nflags = param->ifbrp_filter;
   1319 	oflags = sc->sc_filter_flags;
   1320 
   1321 	if ((nflags & IFBF_FILT_USEIPF) && !(oflags & IFBF_FILT_USEIPF)) {
   1322 		pfil_add_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
   1323 			sc->sc_if.if_pfil);
   1324 	}
   1325 	if (!(nflags & IFBF_FILT_USEIPF) && (oflags & IFBF_FILT_USEIPF)) {
   1326 		pfil_remove_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
   1327 			sc->sc_if.if_pfil);
   1328 	}
   1329 
   1330 	sc->sc_filter_flags = nflags;
   1331 
   1332 	return 0;
   1333 }
   1334 #endif /* BRIDGE_IPF */
   1335 
   1336 static int
   1337 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
   1338 {
   1339 	struct ifbreq *req = arg;
   1340 	struct bridge_iflist *bif;
   1341 	struct psref psref;
   1342 
   1343 	bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
   1344 	if (bif == NULL)
   1345 		return ENOENT;
   1346 
   1347 	bif->bif_path_cost = req->ifbr_path_cost;
   1348 
   1349 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1350 		bstp_initialization(sc);
   1351 
   1352 	bridge_release_member(sc, bif, &psref);
   1353 
   1354 	return 0;
   1355 }
   1356 
   1357 /*
   1358  * bridge_ifdetach:
   1359  *
   1360  *	Detach an interface from a bridge.  Called when a member
   1361  *	interface is detaching.
   1362  */
   1363 void
   1364 bridge_ifdetach(struct ifnet *ifp)
   1365 {
   1366 	struct bridge_softc *sc = ifp->if_bridge;
   1367 	struct ifbreq breq;
   1368 
   1369 	/* ioctl_lock should prevent this from happening */
   1370 	KASSERT(sc != NULL);
   1371 
   1372 	memset(&breq, 0, sizeof(breq));
   1373 	strlcpy(breq.ifbr_ifsname, ifp->if_xname, sizeof(breq.ifbr_ifsname));
   1374 
   1375 	(void) bridge_ioctl_del(sc, &breq);
   1376 }
   1377 
   1378 /*
   1379  * bridge_init:
   1380  *
   1381  *	Initialize a bridge interface.
   1382  */
   1383 static int
   1384 bridge_init(struct ifnet *ifp)
   1385 {
   1386 	struct bridge_softc *sc = ifp->if_softc;
   1387 
   1388 	KASSERT((ifp->if_flags & IFF_RUNNING) == 0);
   1389 
   1390 	callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
   1391 	    bridge_timer, sc);
   1392 	bstp_initialization(sc);
   1393 
   1394 	ifp->if_flags |= IFF_RUNNING;
   1395 	return 0;
   1396 }
   1397 
   1398 /*
   1399  * bridge_stop:
   1400  *
   1401  *	Stop the bridge interface.
   1402  */
   1403 static void
   1404 bridge_stop(struct ifnet *ifp, int disable)
   1405 {
   1406 	struct bridge_softc *sc = ifp->if_softc;
   1407 
   1408 	KASSERT((ifp->if_flags & IFF_RUNNING) != 0);
   1409 	ifp->if_flags &= ~IFF_RUNNING;
   1410 
   1411 	callout_halt(&sc->sc_brcallout, NULL);
   1412 	workqueue_wait(sc->sc_rtage_wq, &sc->sc_rtage_wk);
   1413 	bstp_stop(sc);
   1414 	bridge_rtflush(sc, IFBF_FLUSHDYN);
   1415 }
   1416 
   1417 /*
   1418  * bridge_enqueue:
   1419  *
   1420  *	Enqueue a packet on a bridge member interface.
   1421  */
   1422 void
   1423 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m,
   1424     int runfilt)
   1425 {
   1426 	int len, error;
   1427 	short mflags;
   1428 
   1429 	if (runfilt) {
   1430 		if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
   1431 		    dst_ifp, PFIL_OUT) != 0) {
   1432 			if (m != NULL)
   1433 				m_freem(m);
   1434 			return;
   1435 		}
   1436 		if (m == NULL)
   1437 			return;
   1438 	}
   1439 
   1440 #ifdef ALTQ
   1441 	KERNEL_LOCK(1, NULL);
   1442 	/*
   1443 	 * If ALTQ is enabled on the member interface, do
   1444 	 * classification; the queueing discipline might
   1445 	 * not require classification, but might require
   1446 	 * the address family/header pointer in the pktattr.
   1447 	 */
   1448 	if (ALTQ_IS_ENABLED(&dst_ifp->if_snd)) {
   1449 		/* XXX IFT_ETHER */
   1450 		altq_etherclassify(&dst_ifp->if_snd, m);
   1451 	}
   1452 	KERNEL_UNLOCK_ONE(NULL);
   1453 #endif /* ALTQ */
   1454 
   1455 	len = m->m_pkthdr.len;
   1456 	mflags = m->m_flags;
   1457 
   1458 	error = if_transmit_lock(dst_ifp, m);
   1459 	if (error) {
   1460 		/* mbuf is already freed */
   1461 		if_statinc(&sc->sc_if, if_oerrors);
   1462 		return;
   1463 	}
   1464 
   1465 	net_stat_ref_t nsr = IF_STAT_GETREF(&sc->sc_if);
   1466 	if_statinc_ref(nsr, if_opackets);
   1467 	if_statadd_ref(nsr, if_obytes, len);
   1468 	if (mflags & M_MCAST)
   1469 		if_statinc_ref(nsr, if_omcasts);
   1470 	IF_STAT_PUTREF(&sc->sc_if);
   1471 }
   1472 
   1473 /*
   1474  * bridge_output:
   1475  *
   1476  *	Send output from a bridge member interface.  This
   1477  *	performs the bridging function for locally originated
   1478  *	packets.
   1479  *
   1480  *	The mbuf has the Ethernet header already attached.  We must
   1481  *	enqueue or free the mbuf before returning.
   1482  */
   1483 int
   1484 bridge_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa,
   1485     const struct rtentry *rt)
   1486 {
   1487 	struct ether_header *eh;
   1488 	struct ifnet *dst_if;
   1489 	struct bridge_softc *sc;
   1490 	struct mbuf *n;
   1491 	int s;
   1492 
   1493 	/*
   1494 	 * bridge_output() is called from ether_output(), furthermore
   1495 	 * ifp argument doesn't point to bridge(4). So, don't assert
   1496 	 * IFEF_MPSAFE here.
   1497 	 */
   1498 
   1499 	if (m->m_len < ETHER_HDR_LEN) {
   1500 		m = m_pullup(m, ETHER_HDR_LEN);
   1501 		if (m == NULL)
   1502 			return 0;
   1503 	}
   1504 
   1505 	eh = mtod(m, struct ether_header *);
   1506 	sc = ifp->if_bridge;
   1507 
   1508 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
   1509 		if (memcmp(etherbroadcastaddr,
   1510 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
   1511 			m->m_flags |= M_BCAST;
   1512 		else
   1513 			m->m_flags |= M_MCAST;
   1514 	}
   1515 
   1516 	/*
   1517 	 * If bridge is down, but the original output interface is up,
   1518 	 * go ahead and send out that interface.  Otherwise, the packet
   1519 	 * is dropped below.
   1520 	 */
   1521 	if (__predict_false(sc == NULL) ||
   1522 	    (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
   1523 		dst_if = ifp;
   1524 		goto unicast_asis;
   1525 	}
   1526 
   1527 	/*
   1528 	 * If the packet is a multicast, or we don't know a better way to
   1529 	 * get there, send to all interfaces.
   1530 	 */
   1531 	if ((m->m_flags & (M_MCAST | M_BCAST)) != 0)
   1532 		dst_if = NULL;
   1533 	else
   1534 		dst_if = bridge_rtlookup(sc, eh->ether_dhost);
   1535 
   1536 	/*
   1537 	 * In general, we need to handle TX offload in software before
   1538 	 * enqueueing a packet. However, we can send it as is in the
   1539 	 * cases of unicast via (1) the source interface, or (2) an
   1540 	 * interface which supports the specified offload options.
   1541 	 * For multicast or broadcast, send it as is only if (3) all
   1542 	 * the member interfaces support the specified options.
   1543 	 */
   1544 
   1545 	/*
   1546 	 * Unicast via the source interface.
   1547 	 */
   1548 	if (dst_if == ifp)
   1549 		goto unicast_asis;
   1550 
   1551 	/*
   1552 	 * Unicast via other interface.
   1553 	 */
   1554 	if (dst_if != NULL) {
   1555 		KASSERT(m->m_flags & M_PKTHDR);
   1556 		if (TX_OFFLOAD_SUPPORTED(dst_if->if_csum_flags_tx,
   1557 		    m->m_pkthdr.csum_flags)) {
   1558 			/*
   1559 			 * Unicast via an interface which supports the
   1560 			 * specified offload options.
   1561 			 */
   1562 			goto unicast_asis;
   1563 		}
   1564 
   1565 		/*
   1566 		 * Handle TX offload in software. For TSO, a packet is
   1567 		 * split into multiple chunks. Thus, the return value of
   1568 		 * ether_sw_offload_tx() is mbuf queue consists of them.
   1569 		 */
   1570 		m = ether_sw_offload_tx(ifp, m);
   1571 		if (m == NULL)
   1572 			return 0;
   1573 
   1574 		do {
   1575 			n = m->m_nextpkt;
   1576 			if ((dst_if->if_flags & IFF_RUNNING) == 0)
   1577 				m_freem(m);
   1578 			else
   1579 				bridge_enqueue(sc, dst_if, m, 0);
   1580 			m = n;
   1581 		} while (m != NULL);
   1582 
   1583 		return 0;
   1584 	}
   1585 
   1586 	/*
   1587 	 * Multicast or broadcast.
   1588 	 */
   1589 	if (TX_OFFLOAD_SUPPORTED(sc->sc_csum_flags_tx,
   1590 	    m->m_pkthdr.csum_flags)) {
   1591 		/*
   1592 		 * Specified TX offload options are supported by all
   1593 		 * the member interfaces of this bridge.
   1594 		 */
   1595 		m->m_nextpkt = NULL;	/* XXX */
   1596 	} else {
   1597 		/*
   1598 		 * Otherwise, handle TX offload in software.
   1599 		 */
   1600 		m = ether_sw_offload_tx(ifp, m);
   1601 		if (m == NULL)
   1602 			return 0;
   1603 	}
   1604 
   1605 	do {
   1606 		/* XXX Should call bridge_broadcast, but there are locking
   1607 		 * issues which need resolving first. */
   1608 		struct bridge_iflist *bif;
   1609 		struct mbuf *mc;
   1610 		bool used = false;
   1611 
   1612 		n = m->m_nextpkt;
   1613 
   1614 		BRIDGE_PSZ_RENTER(s);
   1615 		BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
   1616 			struct psref psref;
   1617 
   1618 			bridge_acquire_member(sc, bif, &psref);
   1619 			BRIDGE_PSZ_REXIT(s);
   1620 
   1621 			dst_if = bif->bif_ifp;
   1622 			if ((dst_if->if_flags & IFF_RUNNING) == 0)
   1623 				goto next;
   1624 
   1625 			/*
   1626 			 * If this is not the original output interface,
   1627 			 * and the interface is participating in spanning
   1628 			 * tree, make sure the port is in a state that
   1629 			 * allows forwarding.
   1630 			 */
   1631 			if (dst_if != ifp &&
   1632 			    (bif->bif_flags & IFBIF_STP) != 0) {
   1633 				switch (bif->bif_state) {
   1634 				case BSTP_IFSTATE_BLOCKING:
   1635 				case BSTP_IFSTATE_LISTENING:
   1636 				case BSTP_IFSTATE_DISABLED:
   1637 					goto next;
   1638 				}
   1639 			}
   1640 
   1641 			if (PSLIST_READER_NEXT(bif, struct bridge_iflist,
   1642 			    bif_next) == NULL &&
   1643 			    ((m->m_flags & (M_MCAST | M_BCAST)) == 0 ||
   1644 			    dst_if == ifp))
   1645 			{
   1646 				used = true;
   1647 				mc = m;
   1648 			} else {
   1649 				mc = m_copypacket(m, M_DONTWAIT);
   1650 				if (mc == NULL) {
   1651 					if_statinc(&sc->sc_if, if_oerrors);
   1652 					goto next;
   1653 				}
   1654 			}
   1655 
   1656 			bridge_enqueue(sc, dst_if, mc, 0);
   1657 
   1658 			if ((m->m_flags & (M_MCAST | M_BCAST)) != 0 &&
   1659 			    dst_if != ifp)
   1660 			{
   1661 				if (PSLIST_READER_NEXT(bif,
   1662 				    struct bridge_iflist, bif_next) == NULL)
   1663 				{
   1664 					used = true;
   1665 					mc = m;
   1666 				} else {
   1667 					mc = m_copypacket(m, M_DONTWAIT);
   1668 					if (mc == NULL) {
   1669 						if_statinc(&sc->sc_if,
   1670 						    if_oerrors);
   1671 						goto next;
   1672 					}
   1673 				}
   1674 
   1675 				m_set_rcvif(mc, dst_if);
   1676 				mc->m_flags &= ~M_PROMISC;
   1677 
   1678 				s = splsoftnet();
   1679 				KERNEL_LOCK_UNLESS_IFP_MPSAFE(dst_if);
   1680 				ether_input(dst_if, mc);
   1681 				KERNEL_UNLOCK_UNLESS_IFP_MPSAFE(dst_if);
   1682 				splx(s);
   1683 			}
   1684 
   1685 next:
   1686 			BRIDGE_PSZ_RENTER(s);
   1687 			bridge_release_member(sc, bif, &psref);
   1688 
   1689 			/* Guarantee we don't re-enter the loop as we already
   1690 			 * decided we're at the end. */
   1691 			if (used)
   1692 				break;
   1693 		}
   1694 		BRIDGE_PSZ_REXIT(s);
   1695 
   1696 		if (!used)
   1697 			m_freem(m);
   1698 
   1699 		m = n;
   1700 	} while (m != NULL);
   1701 	return 0;
   1702 
   1703 unicast_asis:
   1704 	/*
   1705 	 * XXX Spanning tree consideration here?
   1706 	 */
   1707 	if ((dst_if->if_flags & IFF_RUNNING) == 0)
   1708 		m_freem(m);
   1709 	else
   1710 		bridge_enqueue(sc, dst_if, m, 0);
   1711 	return 0;
   1712 }
   1713 
   1714 /*
   1715  * bridge_start:
   1716  *
   1717  *	Start output on a bridge.
   1718  *
   1719  *	NOTE: This routine should never be called in this implementation.
   1720  */
   1721 static void
   1722 bridge_start(struct ifnet *ifp)
   1723 {
   1724 
   1725 	printf("%s: bridge_start() called\n", ifp->if_xname);
   1726 }
   1727 
   1728 /*
   1729  * bridge_forward:
   1730  *
   1731  *	The forwarding function of the bridge.
   1732  */
   1733 static void
   1734 bridge_forward(struct bridge_softc *sc, struct mbuf *m)
   1735 {
   1736 	struct bridge_iflist *bif;
   1737 	struct ifnet *src_if, *dst_if;
   1738 	struct ether_header *eh;
   1739 	struct psref psref;
   1740 	struct psref psref_src;
   1741 	DECLARE_LOCK_VARIABLE;
   1742 
   1743 	if ((sc->sc_if.if_flags & IFF_RUNNING) == 0)
   1744 		return;
   1745 
   1746 	src_if = m_get_rcvif_psref(m, &psref_src);
   1747 	if (src_if == NULL) {
   1748 		/* Interface is being destroyed? */
   1749 		m_freem(m);
   1750 		goto out;
   1751 	}
   1752 
   1753 	if_statadd2(&sc->sc_if, if_ipackets, 1, if_ibytes, m->m_pkthdr.len);
   1754 
   1755 	/*
   1756 	 * Look up the bridge_iflist.
   1757 	 */
   1758 	bif = bridge_lookup_member_if(sc, src_if, &psref);
   1759 	if (bif == NULL) {
   1760 		/* Interface is not a bridge member (anymore?) */
   1761 		m_freem(m);
   1762 		goto out;
   1763 	}
   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 			m_freem(m);
   1771 			bridge_release_member(sc, bif, &psref);
   1772 			goto out;
   1773 		}
   1774 	}
   1775 
   1776 	eh = mtod(m, struct ether_header *);
   1777 
   1778 	/*
   1779 	 * If the interface is learning, and the source
   1780 	 * address is valid and not multicast, record
   1781 	 * the address.
   1782 	 */
   1783 	if ((bif->bif_flags & IFBIF_LEARNING) != 0 &&
   1784 	    ETHER_IS_MULTICAST(eh->ether_shost) == 0 &&
   1785 	    (eh->ether_shost[0] == 0 &&
   1786 	     eh->ether_shost[1] == 0 &&
   1787 	     eh->ether_shost[2] == 0 &&
   1788 	     eh->ether_shost[3] == 0 &&
   1789 	     eh->ether_shost[4] == 0 &&
   1790 	     eh->ether_shost[5] == 0) == 0) {
   1791 		(void) bridge_rtupdate(sc, eh->ether_shost,
   1792 		    src_if, 0, IFBAF_DYNAMIC);
   1793 	}
   1794 
   1795 	if ((bif->bif_flags & IFBIF_STP) != 0 &&
   1796 	    bif->bif_state == BSTP_IFSTATE_LEARNING) {
   1797 		m_freem(m);
   1798 		bridge_release_member(sc, bif, &psref);
   1799 		goto out;
   1800 	}
   1801 
   1802 	bridge_release_member(sc, bif, &psref);
   1803 
   1804 	/*
   1805 	 * At this point, the port either doesn't participate
   1806 	 * in spanning tree or it is in the forwarding state.
   1807 	 */
   1808 
   1809 	/*
   1810 	 * If the packet is unicast, destined for someone on
   1811 	 * "this" side of the bridge, drop it.
   1812 	 */
   1813 	if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
   1814 		dst_if = bridge_rtlookup(sc, eh->ether_dhost);
   1815 		if (src_if == dst_if) {
   1816 			m_freem(m);
   1817 			goto out;
   1818 		}
   1819 	} else {
   1820 		/* ...forward it to all interfaces. */
   1821 		if_statinc(&sc->sc_if, if_imcasts);
   1822 		dst_if = NULL;
   1823 	}
   1824 
   1825 	if (pfil_run_hooks(sc->sc_if.if_pfil, &m, src_if, PFIL_IN) != 0) {
   1826 		if (m != NULL)
   1827 			m_freem(m);
   1828 		goto out;
   1829 	}
   1830 	if (m == NULL)
   1831 		goto out;
   1832 
   1833 	if (dst_if == NULL) {
   1834 		bridge_broadcast(sc, src_if, m);
   1835 		goto out;
   1836 	}
   1837 
   1838 	m_put_rcvif_psref(src_if, &psref_src);
   1839 	src_if = NULL;
   1840 
   1841 	/*
   1842 	 * At this point, we're dealing with a unicast frame
   1843 	 * going to a different interface.
   1844 	 */
   1845 	if ((dst_if->if_flags & IFF_RUNNING) == 0) {
   1846 		m_freem(m);
   1847 		goto out;
   1848 	}
   1849 
   1850 	bif = bridge_lookup_member_if(sc, dst_if, &psref);
   1851 	if (bif == NULL) {
   1852 		/* Not a member of the bridge (anymore?) */
   1853 		m_freem(m);
   1854 		goto out;
   1855 	}
   1856 
   1857 	if (bif->bif_flags & IFBIF_STP) {
   1858 		switch (bif->bif_state) {
   1859 		case BSTP_IFSTATE_DISABLED:
   1860 		case BSTP_IFSTATE_BLOCKING:
   1861 			m_freem(m);
   1862 			bridge_release_member(sc, bif, &psref);
   1863 			goto out;
   1864 		}
   1865 	}
   1866 
   1867 	bridge_release_member(sc, bif, &psref);
   1868 
   1869 	/*
   1870 	 * Before enqueueing this packet to the destination interface,
   1871 	 * clear any in-bound checksum flags to prevent them from being
   1872 	 * misused as out-bound flags.
   1873 	 */
   1874 	m->m_pkthdr.csum_flags = 0;
   1875 
   1876 	ACQUIRE_GLOBAL_LOCKS();
   1877 	bridge_enqueue(sc, dst_if, m, 1);
   1878 	RELEASE_GLOBAL_LOCKS();
   1879 out:
   1880 	if (src_if != NULL)
   1881 		m_put_rcvif_psref(src_if, &psref_src);
   1882 	return;
   1883 }
   1884 
   1885 static bool
   1886 bstp_state_before_learning(struct bridge_iflist *bif)
   1887 {
   1888 	if (bif->bif_flags & IFBIF_STP) {
   1889 		switch (bif->bif_state) {
   1890 		case BSTP_IFSTATE_BLOCKING:
   1891 		case BSTP_IFSTATE_LISTENING:
   1892 		case BSTP_IFSTATE_DISABLED:
   1893 			return true;
   1894 		}
   1895 	}
   1896 	return false;
   1897 }
   1898 
   1899 static bool
   1900 bridge_ourether(struct bridge_iflist *bif, struct ether_header *eh, int src)
   1901 {
   1902 	uint8_t *ether = src ? eh->ether_shost : eh->ether_dhost;
   1903 
   1904 	if (memcmp(CLLADDR(bif->bif_ifp->if_sadl), ether, ETHER_ADDR_LEN) == 0
   1905 #if NCARP > 0
   1906 	    || (bif->bif_ifp->if_carp &&
   1907 	        carp_ourether(bif->bif_ifp->if_carp, eh, IFT_ETHER, src) != NULL)
   1908 #endif /* NCARP > 0 */
   1909 	    )
   1910 		return true;
   1911 
   1912 	return false;
   1913 }
   1914 
   1915 /*
   1916  * bridge_input:
   1917  *
   1918  *	Receive input from a member interface.  Queue the packet for
   1919  *	bridging if it is not for us.
   1920  */
   1921 static void
   1922 bridge_input(struct ifnet *ifp, struct mbuf *m)
   1923 {
   1924 	struct bridge_softc *sc = ifp->if_bridge;
   1925 	struct bridge_iflist *bif;
   1926 	struct ether_header *eh;
   1927 	struct psref psref;
   1928 	int bound;
   1929 	DECLARE_LOCK_VARIABLE;
   1930 
   1931 	KASSERT(!cpu_intr_p());
   1932 
   1933 	if (__predict_false(sc == NULL) ||
   1934 	    (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
   1935 		ACQUIRE_GLOBAL_LOCKS();
   1936 		ether_input(ifp, m);
   1937 		RELEASE_GLOBAL_LOCKS();
   1938 		return;
   1939 	}
   1940 
   1941 	bound = curlwp_bind();
   1942 	bif = bridge_lookup_member_if(sc, ifp, &psref);
   1943 	if (bif == NULL) {
   1944 		curlwp_bindx(bound);
   1945 		ACQUIRE_GLOBAL_LOCKS();
   1946 		ether_input(ifp, m);
   1947 		RELEASE_GLOBAL_LOCKS();
   1948 		return;
   1949 	}
   1950 
   1951 	eh = mtod(m, struct ether_header *);
   1952 
   1953 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
   1954 		if (memcmp(etherbroadcastaddr,
   1955 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
   1956 			m->m_flags |= M_BCAST;
   1957 		else
   1958 			m->m_flags |= M_MCAST;
   1959 	}
   1960 
   1961 	/*
   1962 	 * A 'fast' path for packets addressed to interfaces that are
   1963 	 * part of this bridge.
   1964 	 */
   1965 	if (!(m->m_flags & (M_BCAST|M_MCAST)) &&
   1966 	    !bstp_state_before_learning(bif)) {
   1967 		struct bridge_iflist *_bif;
   1968 		struct ifnet *_ifp = NULL;
   1969 		int s;
   1970 		struct psref _psref;
   1971 
   1972 		BRIDGE_PSZ_RENTER(s);
   1973 		BRIDGE_IFLIST_READER_FOREACH(_bif, sc) {
   1974 			/* It is destined for us. */
   1975 			if (bridge_ourether(_bif, eh, 0)) {
   1976 				bridge_acquire_member(sc, _bif, &_psref);
   1977 				BRIDGE_PSZ_REXIT(s);
   1978 				if (_bif->bif_flags & IFBIF_LEARNING)
   1979 					(void) bridge_rtupdate(sc,
   1980 					    eh->ether_shost, ifp, 0, IFBAF_DYNAMIC);
   1981 				m_set_rcvif(m, _bif->bif_ifp);
   1982 				_ifp = _bif->bif_ifp;
   1983 				bridge_release_member(sc, _bif, &_psref);
   1984 				goto out;
   1985 			}
   1986 
   1987 			/* We just received a packet that we sent out. */
   1988 			if (bridge_ourether(_bif, eh, 1))
   1989 				break;
   1990 		}
   1991 		BRIDGE_PSZ_REXIT(s);
   1992 out:
   1993 
   1994 		if (_bif != NULL) {
   1995 			bridge_release_member(sc, bif, &psref);
   1996 			curlwp_bindx(bound);
   1997 			if (_ifp != NULL) {
   1998 				m->m_flags &= ~M_PROMISC;
   1999 				ACQUIRE_GLOBAL_LOCKS();
   2000 				ether_input(_ifp, m);
   2001 				RELEASE_GLOBAL_LOCKS();
   2002 			} else
   2003 				m_freem(m);
   2004 			return;
   2005 		}
   2006 	}
   2007 
   2008 	/* Tap off 802.1D packets; they do not get forwarded. */
   2009 	if (bif->bif_flags & IFBIF_STP &&
   2010 	    memcmp(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN) == 0) {
   2011 		bstp_input(sc, bif, m);
   2012 		bridge_release_member(sc, bif, &psref);
   2013 		curlwp_bindx(bound);
   2014 		return;
   2015 	}
   2016 
   2017 	/*
   2018 	 * A normal switch would discard the packet here, but that's not what
   2019 	 * we've done historically. This also prevents some obnoxious behaviour.
   2020 	 */
   2021 	if (bstp_state_before_learning(bif)) {
   2022 		bridge_release_member(sc, bif, &psref);
   2023 		curlwp_bindx(bound);
   2024 		ACQUIRE_GLOBAL_LOCKS();
   2025 		ether_input(ifp, m);
   2026 		RELEASE_GLOBAL_LOCKS();
   2027 		return;
   2028 	}
   2029 
   2030 	bridge_release_member(sc, bif, &psref);
   2031 
   2032 	bridge_forward(sc, m);
   2033 
   2034 	curlwp_bindx(bound);
   2035 }
   2036 
   2037 /*
   2038  * bridge_broadcast:
   2039  *
   2040  *	Send a frame to all interfaces that are members of
   2041  *	the bridge, except for the one on which the packet
   2042  *	arrived.
   2043  */
   2044 static void
   2045 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
   2046     struct mbuf *m)
   2047 {
   2048 	struct bridge_iflist *bif;
   2049 	struct mbuf *mc;
   2050 	struct ifnet *dst_if;
   2051 	bool bmcast;
   2052 	int s;
   2053 	DECLARE_LOCK_VARIABLE;
   2054 
   2055 	bmcast = m->m_flags & (M_BCAST|M_MCAST);
   2056 
   2057 	BRIDGE_PSZ_RENTER(s);
   2058 	BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
   2059 		struct psref psref;
   2060 
   2061 		bridge_acquire_member(sc, bif, &psref);
   2062 		BRIDGE_PSZ_REXIT(s);
   2063 
   2064 		dst_if = bif->bif_ifp;
   2065 
   2066 		if (bif->bif_flags & IFBIF_STP) {
   2067 			switch (bif->bif_state) {
   2068 			case BSTP_IFSTATE_BLOCKING:
   2069 			case BSTP_IFSTATE_DISABLED:
   2070 				goto next;
   2071 			}
   2072 		}
   2073 
   2074 		if ((bif->bif_flags & IFBIF_DISCOVER) == 0 && !bmcast)
   2075 			goto next;
   2076 
   2077 		if ((dst_if->if_flags & IFF_RUNNING) == 0)
   2078 			goto next;
   2079 
   2080 		if (dst_if != src_if) {
   2081 			mc = m_copypacket(m, M_DONTWAIT);
   2082 			if (mc == NULL) {
   2083 				if_statinc(&sc->sc_if, if_oerrors);
   2084 				goto next;
   2085 			}
   2086 			/*
   2087 			 * Before enqueueing this packet to the destination
   2088 			 * interface, clear any in-bound checksum flags to
   2089 			 * prevent them from being misused as out-bound flags.
   2090 			 */
   2091 			mc->m_pkthdr.csum_flags = 0;
   2092 
   2093 			ACQUIRE_GLOBAL_LOCKS();
   2094 			bridge_enqueue(sc, dst_if, mc, 1);
   2095 			RELEASE_GLOBAL_LOCKS();
   2096 		}
   2097 
   2098 		if (bmcast) {
   2099 			mc = m_copypacket(m, M_DONTWAIT);
   2100 			if (mc == NULL) {
   2101 				if_statinc(&sc->sc_if, if_oerrors);
   2102 				goto next;
   2103 			}
   2104 			/*
   2105 			 * Before enqueueing this packet to the destination
   2106 			 * interface, clear any in-bound checksum flags to
   2107 			 * prevent them from being misused as out-bound flags.
   2108 			 */
   2109 			mc->m_pkthdr.csum_flags = 0;
   2110 
   2111 			m_set_rcvif(mc, dst_if);
   2112 			mc->m_flags &= ~M_PROMISC;
   2113 
   2114 			ACQUIRE_GLOBAL_LOCKS();
   2115 			ether_input(dst_if, mc);
   2116 			RELEASE_GLOBAL_LOCKS();
   2117 		}
   2118 next:
   2119 		BRIDGE_PSZ_RENTER(s);
   2120 		bridge_release_member(sc, bif, &psref);
   2121 	}
   2122 	BRIDGE_PSZ_REXIT(s);
   2123 
   2124 	m_freem(m);
   2125 }
   2126 
   2127 static int
   2128 bridge_rtalloc(struct bridge_softc *sc, const uint8_t *dst,
   2129     struct bridge_rtnode **brtp)
   2130 {
   2131 	struct bridge_rtnode *brt;
   2132 	int error;
   2133 
   2134 	if (sc->sc_brtcnt >= sc->sc_brtmax)
   2135 		return ENOSPC;
   2136 
   2137 	/*
   2138 	 * Allocate a new bridge forwarding node, and
   2139 	 * initialize the expiration time and Ethernet
   2140 	 * address.
   2141 	 */
   2142 	brt = pool_get(&bridge_rtnode_pool, PR_NOWAIT);
   2143 	if (brt == NULL)
   2144 		return ENOMEM;
   2145 
   2146 	memset(brt, 0, sizeof(*brt));
   2147 	brt->brt_expire = time_uptime + sc->sc_brttimeout;
   2148 	brt->brt_flags = IFBAF_DYNAMIC;
   2149 	memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
   2150 	PSLIST_ENTRY_INIT(brt, brt_list);
   2151 	PSLIST_ENTRY_INIT(brt, brt_hash);
   2152 
   2153 	BRIDGE_RT_LOCK(sc);
   2154 	error = bridge_rtnode_insert(sc, brt);
   2155 	BRIDGE_RT_UNLOCK(sc);
   2156 
   2157 	if (error != 0) {
   2158 		pool_put(&bridge_rtnode_pool, brt);
   2159 		return error;
   2160 	}
   2161 
   2162 	*brtp = brt;
   2163 	return 0;
   2164 }
   2165 
   2166 /*
   2167  * bridge_rtupdate:
   2168  *
   2169  *	Add a bridge routing entry.
   2170  */
   2171 static int
   2172 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
   2173     struct ifnet *dst_if, int setflags, uint8_t flags)
   2174 {
   2175 	struct bridge_rtnode *brt;
   2176 	int s;
   2177 
   2178 again:
   2179 	/*
   2180 	 * A route for this destination might already exist.  If so,
   2181 	 * update it, otherwise create a new one.
   2182 	 */
   2183 	BRIDGE_RT_RENTER(s);
   2184 	brt = bridge_rtnode_lookup(sc, dst);
   2185 
   2186 	if (brt != NULL) {
   2187 		brt->brt_ifp = dst_if;
   2188 		if (setflags) {
   2189 			brt->brt_flags = flags;
   2190 			if (flags & IFBAF_STATIC)
   2191 				brt->brt_expire = 0;
   2192 			else
   2193 				brt->brt_expire = time_uptime + sc->sc_brttimeout;
   2194 		} else {
   2195 			if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
   2196 				brt->brt_expire = time_uptime + sc->sc_brttimeout;
   2197 		}
   2198 	}
   2199 	BRIDGE_RT_REXIT(s);
   2200 
   2201 	if (brt == NULL) {
   2202 		int r;
   2203 
   2204 		r = bridge_rtalloc(sc, dst, &brt);
   2205 		if (r != 0)
   2206 			return r;
   2207 		goto again;
   2208 	}
   2209 
   2210 	return 0;
   2211 }
   2212 
   2213 /*
   2214  * bridge_rtlookup:
   2215  *
   2216  *	Lookup the destination interface for an address.
   2217  */
   2218 static struct ifnet *
   2219 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr)
   2220 {
   2221 	struct bridge_rtnode *brt;
   2222 	struct ifnet *ifs = NULL;
   2223 	int s;
   2224 
   2225 	BRIDGE_RT_RENTER(s);
   2226 	brt = bridge_rtnode_lookup(sc, addr);
   2227 	if (brt != NULL)
   2228 		ifs = brt->brt_ifp;
   2229 	BRIDGE_RT_REXIT(s);
   2230 
   2231 	return ifs;
   2232 }
   2233 
   2234 typedef bool (*bridge_iterate_cb_t)
   2235     (struct bridge_softc *, struct bridge_rtnode *, bool *, void *);
   2236 
   2237 /*
   2238  * bridge_rtlist_iterate_remove:
   2239  *
   2240  *	It iterates on sc->sc_rtlist and removes rtnodes of it which func
   2241  *	callback judges to remove. Removals of rtnodes are done in a manner
   2242  *	of pserialize. To this end, all kmem_* operations are placed out of
   2243  *	mutexes.
   2244  */
   2245 static void
   2246 bridge_rtlist_iterate_remove(struct bridge_softc *sc, bridge_iterate_cb_t func, void *arg)
   2247 {
   2248 	struct bridge_rtnode *brt;
   2249 	struct bridge_rtnode **brt_list;
   2250 	int i, count;
   2251 
   2252 retry:
   2253 	count = sc->sc_brtcnt;
   2254 	if (count == 0)
   2255 		return;
   2256 	brt_list = kmem_alloc(sizeof(*brt_list) * count, KM_SLEEP);
   2257 
   2258 	BRIDGE_RT_LOCK(sc);
   2259 	if (__predict_false(sc->sc_brtcnt > count)) {
   2260 		/* The rtnodes increased, we need more memory */
   2261 		BRIDGE_RT_UNLOCK(sc);
   2262 		kmem_free(brt_list, sizeof(*brt_list) * count);
   2263 		goto retry;
   2264 	}
   2265 
   2266 	i = 0;
   2267 	/*
   2268 	 * We don't need to use a _SAFE variant here because we know
   2269 	 * that a removed item keeps its next pointer as-is thanks to
   2270 	 * pslist(9) and isn't freed in the loop.
   2271 	 */
   2272 	BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) {
   2273 		bool need_break = false;
   2274 		if (func(sc, brt, &need_break, arg)) {
   2275 			bridge_rtnode_remove(sc, brt);
   2276 			brt_list[i++] = brt;
   2277 		}
   2278 		if (need_break)
   2279 			break;
   2280 	}
   2281 
   2282 	if (i > 0)
   2283 		BRIDGE_RT_PSZ_PERFORM(sc);
   2284 	BRIDGE_RT_UNLOCK(sc);
   2285 
   2286 	while (--i >= 0)
   2287 		bridge_rtnode_destroy(brt_list[i]);
   2288 
   2289 	kmem_free(brt_list, sizeof(*brt_list) * count);
   2290 }
   2291 
   2292 static bool
   2293 bridge_rttrim0_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
   2294     bool *need_break, void *arg)
   2295 {
   2296 	if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
   2297 		/* Take into account of the subsequent removal */
   2298 		if ((sc->sc_brtcnt - 1) <= sc->sc_brtmax)
   2299 			*need_break = true;
   2300 		return true;
   2301 	} else
   2302 		return false;
   2303 }
   2304 
   2305 static void
   2306 bridge_rttrim0(struct bridge_softc *sc)
   2307 {
   2308 	bridge_rtlist_iterate_remove(sc, bridge_rttrim0_cb, NULL);
   2309 }
   2310 
   2311 /*
   2312  * bridge_rttrim:
   2313  *
   2314  *	Trim the routine table so that we have a number
   2315  *	of routing entries less than or equal to the
   2316  *	maximum number.
   2317  */
   2318 static void
   2319 bridge_rttrim(struct bridge_softc *sc)
   2320 {
   2321 
   2322 	/* Make sure we actually need to do this. */
   2323 	if (sc->sc_brtcnt <= sc->sc_brtmax)
   2324 		return;
   2325 
   2326 	/* Force an aging cycle; this might trim enough addresses. */
   2327 	bridge_rtage(sc);
   2328 	if (sc->sc_brtcnt <= sc->sc_brtmax)
   2329 		return;
   2330 
   2331 	bridge_rttrim0(sc);
   2332 
   2333 	return;
   2334 }
   2335 
   2336 /*
   2337  * bridge_timer:
   2338  *
   2339  *	Aging timer for the bridge.
   2340  */
   2341 static void
   2342 bridge_timer(void *arg)
   2343 {
   2344 	struct bridge_softc *sc = arg;
   2345 
   2346 	workqueue_enqueue(sc->sc_rtage_wq, &sc->sc_rtage_wk, NULL);
   2347 }
   2348 
   2349 static void
   2350 bridge_rtage_work(struct work *wk, void *arg)
   2351 {
   2352 	struct bridge_softc *sc = arg;
   2353 
   2354 	KASSERT(wk == &sc->sc_rtage_wk);
   2355 
   2356 	bridge_rtage(sc);
   2357 
   2358 	if (sc->sc_if.if_flags & IFF_RUNNING)
   2359 		callout_reset(&sc->sc_brcallout,
   2360 		    bridge_rtable_prune_period * hz, bridge_timer, sc);
   2361 }
   2362 
   2363 static bool
   2364 bridge_rtage_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
   2365     bool *need_break, void *arg)
   2366 {
   2367 	if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
   2368 	    time_uptime >= brt->brt_expire)
   2369 		return true;
   2370 	else
   2371 		return false;
   2372 }
   2373 
   2374 /*
   2375  * bridge_rtage:
   2376  *
   2377  *	Perform an aging cycle.
   2378  */
   2379 static void
   2380 bridge_rtage(struct bridge_softc *sc)
   2381 {
   2382 	bridge_rtlist_iterate_remove(sc, bridge_rtage_cb, NULL);
   2383 }
   2384 
   2385 
   2386 static bool
   2387 bridge_rtflush_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
   2388     bool *need_break, void *arg)
   2389 {
   2390 	int full = *(int*)arg;
   2391 
   2392 	if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
   2393 		return true;
   2394 	else
   2395 		return false;
   2396 }
   2397 
   2398 /*
   2399  * bridge_rtflush:
   2400  *
   2401  *	Remove all dynamic addresses from the bridge.
   2402  */
   2403 static void
   2404 bridge_rtflush(struct bridge_softc *sc, int full)
   2405 {
   2406 	bridge_rtlist_iterate_remove(sc, bridge_rtflush_cb, &full);
   2407 }
   2408 
   2409 /*
   2410  * bridge_rtdaddr:
   2411  *
   2412  *	Remove an address from the table.
   2413  */
   2414 static int
   2415 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr)
   2416 {
   2417 	struct bridge_rtnode *brt;
   2418 
   2419 	BRIDGE_RT_LOCK(sc);
   2420 	if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL) {
   2421 		BRIDGE_RT_UNLOCK(sc);
   2422 		return ENOENT;
   2423 	}
   2424 	bridge_rtnode_remove(sc, brt);
   2425 	BRIDGE_RT_PSZ_PERFORM(sc);
   2426 	BRIDGE_RT_UNLOCK(sc);
   2427 
   2428 	bridge_rtnode_destroy(brt);
   2429 
   2430 	return 0;
   2431 }
   2432 
   2433 /*
   2434  * bridge_rtdelete:
   2435  *
   2436  *	Delete routes to a speicifc member interface.
   2437  */
   2438 static void
   2439 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp)
   2440 {
   2441 	struct bridge_rtnode *brt;
   2442 
   2443 	/* XXX pserialize_perform for each entry is slow */
   2444 again:
   2445 	BRIDGE_RT_LOCK(sc);
   2446 	BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) {
   2447 		if (brt->brt_ifp == ifp)
   2448 			break;
   2449 	}
   2450 	if (brt == NULL) {
   2451 		BRIDGE_RT_UNLOCK(sc);
   2452 		return;
   2453 	}
   2454 	bridge_rtnode_remove(sc, brt);
   2455 	BRIDGE_RT_PSZ_PERFORM(sc);
   2456 	BRIDGE_RT_UNLOCK(sc);
   2457 
   2458 	bridge_rtnode_destroy(brt);
   2459 
   2460 	goto again;
   2461 }
   2462 
   2463 /*
   2464  * bridge_rtable_init:
   2465  *
   2466  *	Initialize the route table for this bridge.
   2467  */
   2468 static void
   2469 bridge_rtable_init(struct bridge_softc *sc)
   2470 {
   2471 	int i;
   2472 
   2473 	sc->sc_rthash = kmem_alloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
   2474 	    KM_SLEEP);
   2475 
   2476 	for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
   2477 		PSLIST_INIT(&sc->sc_rthash[i]);
   2478 
   2479 	sc->sc_rthash_key = cprng_fast32();
   2480 
   2481 	PSLIST_INIT(&sc->sc_rtlist);
   2482 
   2483 	sc->sc_rtlist_psz = pserialize_create();
   2484 	sc->sc_rtlist_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_SOFTNET);
   2485 }
   2486 
   2487 /*
   2488  * bridge_rtable_fini:
   2489  *
   2490  *	Deconstruct the route table for this bridge.
   2491  */
   2492 static void
   2493 bridge_rtable_fini(struct bridge_softc *sc)
   2494 {
   2495 
   2496 	kmem_free(sc->sc_rthash, sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE);
   2497 	mutex_obj_free(sc->sc_rtlist_lock);
   2498 	pserialize_destroy(sc->sc_rtlist_psz);
   2499 }
   2500 
   2501 /*
   2502  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
   2503  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
   2504  */
   2505 #define	mix(a, b, c)							\
   2506 do {									\
   2507 	a -= b; a -= c; a ^= (c >> 13);					\
   2508 	b -= c; b -= a; b ^= (a << 8);					\
   2509 	c -= a; c -= b; c ^= (b >> 13);					\
   2510 	a -= b; a -= c; a ^= (c >> 12);					\
   2511 	b -= c; b -= a; b ^= (a << 16);					\
   2512 	c -= a; c -= b; c ^= (b >> 5);					\
   2513 	a -= b; a -= c; a ^= (c >> 3);					\
   2514 	b -= c; b -= a; b ^= (a << 10);					\
   2515 	c -= a; c -= b; c ^= (b >> 15);					\
   2516 } while (/*CONSTCOND*/0)
   2517 
   2518 static inline uint32_t
   2519 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
   2520 {
   2521 	uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
   2522 
   2523 	b += addr[5] << 8;
   2524 	b += addr[4];
   2525 	a += (uint32_t)addr[3] << 24;
   2526 	a += addr[2] << 16;
   2527 	a += addr[1] << 8;
   2528 	a += addr[0];
   2529 
   2530 	mix(a, b, c);
   2531 
   2532 	return (c & BRIDGE_RTHASH_MASK);
   2533 }
   2534 
   2535 #undef mix
   2536 
   2537 /*
   2538  * bridge_rtnode_lookup:
   2539  *
   2540  *	Look up a bridge route node for the specified destination.
   2541  */
   2542 static struct bridge_rtnode *
   2543 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr)
   2544 {
   2545 	struct bridge_rtnode *brt;
   2546 	uint32_t hash;
   2547 	int dir;
   2548 
   2549 	hash = bridge_rthash(sc, addr);
   2550 	BRIDGE_RTHASH_READER_FOREACH(brt, sc, hash) {
   2551 		dir = memcmp(addr, brt->brt_addr, ETHER_ADDR_LEN);
   2552 		if (dir == 0)
   2553 			return brt;
   2554 		if (dir > 0)
   2555 			return NULL;
   2556 	}
   2557 
   2558 	return NULL;
   2559 }
   2560 
   2561 /*
   2562  * bridge_rtnode_insert:
   2563  *
   2564  *	Insert the specified bridge node into the route table.  We
   2565  *	assume the entry is not already in the table.
   2566  */
   2567 static int
   2568 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
   2569 {
   2570 	struct bridge_rtnode *lbrt, *prev = NULL;
   2571 	uint32_t hash;
   2572 
   2573 	KASSERT(BRIDGE_RT_LOCKED(sc));
   2574 
   2575 	hash = bridge_rthash(sc, brt->brt_addr);
   2576 	BRIDGE_RTHASH_WRITER_FOREACH(lbrt, sc, hash) {
   2577 		int dir = memcmp(brt->brt_addr, lbrt->brt_addr, ETHER_ADDR_LEN);
   2578 		if (dir == 0)
   2579 			return EEXIST;
   2580 		if (dir > 0)
   2581 			break;
   2582 		prev = lbrt;
   2583 	}
   2584 	if (prev == NULL)
   2585 		BRIDGE_RTHASH_WRITER_INSERT_HEAD(sc, hash, brt);
   2586 	else
   2587 		BRIDGE_RTHASH_WRITER_INSERT_AFTER(prev, brt);
   2588 
   2589 	BRIDGE_RTLIST_WRITER_INSERT_HEAD(sc, brt);
   2590 	sc->sc_brtcnt++;
   2591 
   2592 	return 0;
   2593 }
   2594 
   2595 /*
   2596  * bridge_rtnode_remove:
   2597  *
   2598  *	Remove a bridge rtnode from the rthash and the rtlist of a bridge.
   2599  */
   2600 static void
   2601 bridge_rtnode_remove(struct bridge_softc *sc, struct bridge_rtnode *brt)
   2602 {
   2603 
   2604 	KASSERT(BRIDGE_RT_LOCKED(sc));
   2605 
   2606 	BRIDGE_RTHASH_WRITER_REMOVE(brt);
   2607 	BRIDGE_RTLIST_WRITER_REMOVE(brt);
   2608 	sc->sc_brtcnt--;
   2609 }
   2610 
   2611 /*
   2612  * bridge_rtnode_destroy:
   2613  *
   2614  *	Destroy a bridge rtnode.
   2615  */
   2616 static void
   2617 bridge_rtnode_destroy(struct bridge_rtnode *brt)
   2618 {
   2619 
   2620 	PSLIST_ENTRY_DESTROY(brt, brt_list);
   2621 	PSLIST_ENTRY_DESTROY(brt, brt_hash);
   2622 	pool_put(&bridge_rtnode_pool, brt);
   2623 }
   2624 
   2625 #if defined(BRIDGE_IPF)
   2626 extern pfil_head_t *inet_pfil_hook;                 /* XXX */
   2627 extern pfil_head_t *inet6_pfil_hook;                /* XXX */
   2628 
   2629 /*
   2630  * Send bridge packets through IPF if they are one of the types IPF can deal
   2631  * with, or if they are ARP or REVARP.  (IPF will pass ARP and REVARP without
   2632  * question.)
   2633  */
   2634 static int
   2635 bridge_ipf(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
   2636 {
   2637 	int snap, error;
   2638 	struct ether_header *eh1, eh2;
   2639 	struct llc llc1;
   2640 	uint16_t ether_type;
   2641 
   2642 	snap = 0;
   2643 	error = -1;	/* Default error if not error == 0 */
   2644 	eh1 = mtod(*mp, struct ether_header *);
   2645 	ether_type = ntohs(eh1->ether_type);
   2646 
   2647 	/*
   2648 	 * Check for SNAP/LLC.
   2649 	 */
   2650 	if (ether_type < ETHERMTU) {
   2651 		struct llc *llc2 = (struct llc *)(eh1 + 1);
   2652 
   2653 		if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
   2654 		    llc2->llc_dsap == LLC_SNAP_LSAP &&
   2655 		    llc2->llc_ssap == LLC_SNAP_LSAP &&
   2656 		    llc2->llc_control == LLC_UI) {
   2657 			ether_type = htons(llc2->llc_un.type_snap.ether_type);
   2658 			snap = 1;
   2659 		}
   2660 	}
   2661 
   2662 	/*
   2663 	 * If we're trying to filter bridge traffic, don't look at anything
   2664 	 * other than IP and ARP traffic.  If the filter doesn't understand
   2665 	 * IPv6, don't allow IPv6 through the bridge either.  This is lame
   2666 	 * since if we really wanted, say, an AppleTalk filter, we are hosed,
   2667 	 * but of course we don't have an AppleTalk filter to begin with.
   2668 	 * (Note that since IPF doesn't understand ARP it will pass *ALL*
   2669 	 * ARP traffic.)
   2670 	 */
   2671 	switch (ether_type) {
   2672 		case ETHERTYPE_ARP:
   2673 		case ETHERTYPE_REVARP:
   2674 			return 0; /* Automatically pass */
   2675 		case ETHERTYPE_IP:
   2676 # ifdef INET6
   2677 		case ETHERTYPE_IPV6:
   2678 # endif /* INET6 */
   2679 			break;
   2680 		default:
   2681 			goto bad;
   2682 	}
   2683 
   2684 	/* Strip off the Ethernet header and keep a copy. */
   2685 	m_copydata(*mp, 0, ETHER_HDR_LEN, (void *) &eh2);
   2686 	m_adj(*mp, ETHER_HDR_LEN);
   2687 
   2688 	/* Strip off snap header, if present */
   2689 	if (snap) {
   2690 		m_copydata(*mp, 0, sizeof(struct llc), (void *) &llc1);
   2691 		m_adj(*mp, sizeof(struct llc));
   2692 	}
   2693 
   2694 	/*
   2695 	 * Check basic packet sanity and run IPF through pfil.
   2696 	 */
   2697 	KASSERT(!cpu_intr_p());
   2698 	switch (ether_type)
   2699 	{
   2700 	case ETHERTYPE_IP :
   2701 		error = bridge_ip_checkbasic(mp);
   2702 		if (error == 0)
   2703 			error = pfil_run_hooks(inet_pfil_hook, mp, ifp, dir);
   2704 		break;
   2705 # ifdef INET6
   2706 	case ETHERTYPE_IPV6 :
   2707 		error = bridge_ip6_checkbasic(mp);
   2708 		if (error == 0)
   2709 			error = pfil_run_hooks(inet6_pfil_hook, mp, ifp, dir);
   2710 		break;
   2711 # endif
   2712 	default :
   2713 		error = 0;
   2714 		break;
   2715 	}
   2716 
   2717 	if (*mp == NULL)
   2718 		return error;
   2719 	if (error != 0)
   2720 		goto bad;
   2721 
   2722 	error = -1;
   2723 
   2724 	/*
   2725 	 * Finally, put everything back the way it was and return
   2726 	 */
   2727 	if (snap) {
   2728 		M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
   2729 		if (*mp == NULL)
   2730 			return error;
   2731 		bcopy(&llc1, mtod(*mp, void *), sizeof(struct llc));
   2732 	}
   2733 
   2734 	M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
   2735 	if (*mp == NULL)
   2736 		return error;
   2737 	bcopy(&eh2, mtod(*mp, void *), ETHER_HDR_LEN);
   2738 
   2739 	return 0;
   2740 
   2741     bad:
   2742 	m_freem(*mp);
   2743 	*mp = NULL;
   2744 	return error;
   2745 }
   2746 
   2747 /*
   2748  * Perform basic checks on header size since
   2749  * IPF assumes ip_input has already processed
   2750  * it for it.  Cut-and-pasted from ip_input.c.
   2751  * Given how simple the IPv6 version is,
   2752  * does the IPv4 version really need to be
   2753  * this complicated?
   2754  *
   2755  * XXX Should we update ipstat here, or not?
   2756  * XXX Right now we update ipstat but not
   2757  * XXX csum_counter.
   2758  */
   2759 static int
   2760 bridge_ip_checkbasic(struct mbuf **mp)
   2761 {
   2762 	struct mbuf *m = *mp;
   2763 	struct ip *ip;
   2764 	int len, hlen;
   2765 
   2766 	if (*mp == NULL)
   2767 		return -1;
   2768 
   2769 	if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
   2770 		if ((m = m_copyup(m, sizeof(struct ip),
   2771 			(max_linkhdr + 3) & ~3)) == NULL) {
   2772 			/* XXXJRT new stat, please */
   2773 			ip_statinc(IP_STAT_TOOSMALL);
   2774 			goto bad;
   2775 		}
   2776 	} else if (__predict_false(m->m_len < sizeof (struct ip))) {
   2777 		if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
   2778 			ip_statinc(IP_STAT_TOOSMALL);
   2779 			goto bad;
   2780 		}
   2781 	}
   2782 	ip = mtod(m, struct ip *);
   2783 	if (ip == NULL) goto bad;
   2784 
   2785 	if (ip->ip_v != IPVERSION) {
   2786 		ip_statinc(IP_STAT_BADVERS);
   2787 		goto bad;
   2788 	}
   2789 	hlen = ip->ip_hl << 2;
   2790 	if (hlen < sizeof(struct ip)) { /* minimum header length */
   2791 		ip_statinc(IP_STAT_BADHLEN);
   2792 		goto bad;
   2793 	}
   2794 	if (hlen > m->m_len) {
   2795 		if ((m = m_pullup(m, hlen)) == 0) {
   2796 			ip_statinc(IP_STAT_BADHLEN);
   2797 			goto bad;
   2798 		}
   2799 		ip = mtod(m, struct ip *);
   2800 		if (ip == NULL) goto bad;
   2801 	}
   2802 
   2803 	switch (m->m_pkthdr.csum_flags &
   2804 	        ((m_get_rcvif_NOMPSAFE(m)->if_csum_flags_rx & M_CSUM_IPv4) |
   2805 	         M_CSUM_IPv4_BAD)) {
   2806 	case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
   2807 		/* INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); */
   2808 		goto bad;
   2809 
   2810 	case M_CSUM_IPv4:
   2811 		/* Checksum was okay. */
   2812 		/* INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); */
   2813 		break;
   2814 
   2815 	default:
   2816 		/* Must compute it ourselves. */
   2817 		/* INET_CSUM_COUNTER_INCR(&ip_swcsum); */
   2818 		if (in_cksum(m, hlen) != 0)
   2819 			goto bad;
   2820 		break;
   2821 	}
   2822 
   2823 	/* Retrieve the packet length. */
   2824 	len = ntohs(ip->ip_len);
   2825 
   2826 	/*
   2827 	 * Check for additional length bogosity
   2828 	 */
   2829 	if (len < hlen) {
   2830 		ip_statinc(IP_STAT_BADLEN);
   2831 		goto bad;
   2832 	}
   2833 
   2834 	/*
   2835 	 * Check that the amount of data in the buffers
   2836 	 * is as at least much as the IP header would have us expect.
   2837 	 * Drop packet if shorter than we expect.
   2838 	 */
   2839 	if (m->m_pkthdr.len < len) {
   2840 		ip_statinc(IP_STAT_TOOSHORT);
   2841 		goto bad;
   2842 	}
   2843 
   2844 	/* Checks out, proceed */
   2845 	*mp = m;
   2846 	return 0;
   2847 
   2848     bad:
   2849 	*mp = m;
   2850 	return -1;
   2851 }
   2852 
   2853 # ifdef INET6
   2854 /*
   2855  * Same as above, but for IPv6.
   2856  * Cut-and-pasted from ip6_input.c.
   2857  * XXX Should we update ip6stat, or not?
   2858  */
   2859 static int
   2860 bridge_ip6_checkbasic(struct mbuf **mp)
   2861 {
   2862 	struct mbuf *m = *mp;
   2863 	struct ip6_hdr *ip6;
   2864 
   2865 	/*
   2866 	 * If the IPv6 header is not aligned, slurp it up into a new
   2867 	 * mbuf with space for link headers, in the event we forward
   2868 	 * it.  Otherwise, if it is aligned, make sure the entire base
   2869 	 * IPv6 header is in the first mbuf of the chain.
   2870 	 */
   2871 	if (IP6_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
   2872 		struct ifnet *inifp = m_get_rcvif_NOMPSAFE(m);
   2873 		if ((m = m_copyup(m, sizeof(struct ip6_hdr),
   2874 		                  (max_linkhdr + 3) & ~3)) == NULL) {
   2875 			/* XXXJRT new stat, please */
   2876 			ip6_statinc(IP6_STAT_TOOSMALL);
   2877 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
   2878 			goto bad;
   2879 		}
   2880 	} else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
   2881 		struct ifnet *inifp = m_get_rcvif_NOMPSAFE(m);
   2882 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
   2883 			ip6_statinc(IP6_STAT_TOOSMALL);
   2884 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
   2885 			goto bad;
   2886 		}
   2887 	}
   2888 
   2889 	ip6 = mtod(m, struct ip6_hdr *);
   2890 
   2891 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
   2892 		ip6_statinc(IP6_STAT_BADVERS);
   2893 		in6_ifstat_inc(m_get_rcvif_NOMPSAFE(m), ifs6_in_hdrerr);
   2894 		goto bad;
   2895 	}
   2896 
   2897 	/* Checks out, proceed */
   2898 	*mp = m;
   2899 	return 0;
   2900 
   2901     bad:
   2902 	*mp = m;
   2903 	return -1;
   2904 }
   2905 # endif /* INET6 */
   2906 #endif /* BRIDGE_IPF */
   2907