xref: /freebsd/sys/net/if_bridge.c (revision f621b087c00776de713944ae6561e44f83e7b541)
1 /*	$NetBSD: if_bridge.c,v 1.31 2005/06/01 19:45:34 jdc Exp $	*/
2 
3 /*-
4  * SPDX-License-Identifier: BSD-4-Clause
5  *
6  * Copyright 2001 Wasabi Systems, Inc.
7  * All rights reserved.
8  *
9  * Written by Jason R. Thorpe for Wasabi Systems, Inc.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed for the NetBSD Project by
22  *	Wasabi Systems, Inc.
23  * 4. The name of Wasabi Systems, Inc. may not be used to endorse
24  *    or promote products derived from this software without specific prior
25  *    written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
28  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL WASABI SYSTEMS, INC
31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37  * POSSIBILITY OF SUCH DAMAGE.
38  */
39 
40 /*
41  * Copyright (c) 1999, 2000 Jason L. Wright (jason@thought.net)
42  * All rights reserved.
43  *
44  * Redistribution and use in source and binary forms, with or without
45  * modification, are permitted provided that the following conditions
46  * are met:
47  * 1. Redistributions of source code must retain the above copyright
48  *    notice, this list of conditions and the following disclaimer.
49  * 2. Redistributions in binary form must reproduce the above copyright
50  *    notice, this list of conditions and the following disclaimer in the
51  *    documentation and/or other materials provided with the distribution.
52  *
53  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
54  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
55  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
56  * DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
57  * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
58  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
59  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
61  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
62  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
63  * POSSIBILITY OF SUCH DAMAGE.
64  *
65  * OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun Exp
66  */
67 
68 /*
69  * Network interface bridge support.
70  *
71  * TODO:
72  *
73  *	- Currently only supports Ethernet-like interfaces (Ethernet,
74  *	  802.11, VLANs on Ethernet, etc.)  Figure out a nice way
75  *	  to bridge other types of interfaces (maybe consider
76  *	  heterogeneous bridges).
77  */
78 
79 #include <sys/cdefs.h>
80 __FBSDID("$FreeBSD$");
81 
82 #include "opt_inet.h"
83 #include "opt_inet6.h"
84 
85 #include <sys/param.h>
86 #include <sys/eventhandler.h>
87 #include <sys/mbuf.h>
88 #include <sys/malloc.h>
89 #include <sys/protosw.h>
90 #include <sys/systm.h>
91 #include <sys/jail.h>
92 #include <sys/time.h>
93 #include <sys/socket.h> /* for net/if.h */
94 #include <sys/sockio.h>
95 #include <sys/ctype.h>  /* string functions */
96 #include <sys/kernel.h>
97 #include <sys/random.h>
98 #include <sys/syslog.h>
99 #include <sys/sysctl.h>
100 #include <vm/uma.h>
101 #include <sys/module.h>
102 #include <sys/priv.h>
103 #include <sys/proc.h>
104 #include <sys/lock.h>
105 #include <sys/mutex.h>
106 
107 #include <net/bpf.h>
108 #include <net/if.h>
109 #include <net/if_clone.h>
110 #include <net/if_dl.h>
111 #include <net/if_types.h>
112 #include <net/if_var.h>
113 #include <net/if_private.h>
114 #include <net/pfil.h>
115 #include <net/vnet.h>
116 
117 #include <netinet/in.h>
118 #include <netinet/in_systm.h>
119 #include <netinet/in_var.h>
120 #include <netinet/ip.h>
121 #include <netinet/ip_var.h>
122 #ifdef INET6
123 #include <netinet/ip6.h>
124 #include <netinet6/ip6_var.h>
125 #include <netinet6/in6_ifattach.h>
126 #endif
127 #if defined(INET) || defined(INET6)
128 #include <netinet/ip_carp.h>
129 #endif
130 #include <machine/in_cksum.h>
131 #include <netinet/if_ether.h>
132 #include <net/bridgestp.h>
133 #include <net/if_bridgevar.h>
134 #include <net/if_llc.h>
135 #include <net/if_vlan_var.h>
136 
137 #include <net/route.h>
138 
139 /*
140  * Size of the route hash table.  Must be a power of two.
141  */
142 #ifndef BRIDGE_RTHASH_SIZE
143 #define	BRIDGE_RTHASH_SIZE		1024
144 #endif
145 
146 #define	BRIDGE_RTHASH_MASK		(BRIDGE_RTHASH_SIZE - 1)
147 
148 /*
149  * Default maximum number of addresses to cache.
150  */
151 #ifndef BRIDGE_RTABLE_MAX
152 #define	BRIDGE_RTABLE_MAX		2000
153 #endif
154 
155 /*
156  * Timeout (in seconds) for entries learned dynamically.
157  */
158 #ifndef BRIDGE_RTABLE_TIMEOUT
159 #define	BRIDGE_RTABLE_TIMEOUT		(20 * 60)	/* same as ARP */
160 #endif
161 
162 /*
163  * Number of seconds between walks of the route list.
164  */
165 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
166 #define	BRIDGE_RTABLE_PRUNE_PERIOD	(5 * 60)
167 #endif
168 
169 /*
170  * List of capabilities to possibly mask on the member interface.
171  */
172 #define	BRIDGE_IFCAPS_MASK		(IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM|\
173 					 IFCAP_TXCSUM_IPV6)
174 
175 /*
176  * List of capabilities to strip
177  */
178 #define	BRIDGE_IFCAPS_STRIP		IFCAP_LRO
179 
180 /*
181  * Bridge locking
182  *
183  * The bridge relies heavily on the epoch(9) system to protect its data
184  * structures. This means we can safely use CK_LISTs while in NET_EPOCH, but we
185  * must ensure there is only one writer at a time.
186  *
187  * That is: for read accesses we only need to be in NET_EPOCH, but for write
188  * accesses we must hold:
189  *
190  *  - BRIDGE_RT_LOCK, for any change to bridge_rtnodes
191  *  - BRIDGE_LOCK, for any other change
192  *
193  * The BRIDGE_LOCK is a sleepable lock, because it is held across ioctl()
194  * calls to bridge member interfaces and these ioctl()s can sleep.
195  * The BRIDGE_RT_LOCK is a non-sleepable mutex, because it is sometimes
196  * required while we're in NET_EPOCH and then we're not allowed to sleep.
197  */
198 #define BRIDGE_LOCK_INIT(_sc)		do {			\
199 	sx_init(&(_sc)->sc_sx, "if_bridge");			\
200 	mtx_init(&(_sc)->sc_rt_mtx, "if_bridge rt", NULL, MTX_DEF);	\
201 } while (0)
202 #define BRIDGE_LOCK_DESTROY(_sc)	do {	\
203 	sx_destroy(&(_sc)->sc_sx);		\
204 	mtx_destroy(&(_sc)->sc_rt_mtx);		\
205 } while (0)
206 #define BRIDGE_LOCK(_sc)		sx_xlock(&(_sc)->sc_sx)
207 #define BRIDGE_UNLOCK(_sc)		sx_xunlock(&(_sc)->sc_sx)
208 #define BRIDGE_LOCK_ASSERT(_sc)		sx_assert(&(_sc)->sc_sx, SX_XLOCKED)
209 #define BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(_sc)	\
210 	    MPASS(in_epoch(net_epoch_preempt) || sx_xlocked(&(_sc)->sc_sx))
211 #define BRIDGE_UNLOCK_ASSERT(_sc)	sx_assert(&(_sc)->sc_sx, SX_UNLOCKED)
212 #define BRIDGE_RT_LOCK(_sc)		mtx_lock(&(_sc)->sc_rt_mtx)
213 #define BRIDGE_RT_UNLOCK(_sc)		mtx_unlock(&(_sc)->sc_rt_mtx)
214 #define BRIDGE_RT_LOCK_ASSERT(_sc)	mtx_assert(&(_sc)->sc_rt_mtx, MA_OWNED)
215 #define BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(_sc)	\
216 	    MPASS(in_epoch(net_epoch_preempt) || mtx_owned(&(_sc)->sc_rt_mtx))
217 
218 /*
219  * Bridge interface list entry.
220  */
221 struct bridge_iflist {
222 	CK_LIST_ENTRY(bridge_iflist) bif_next;
223 	struct ifnet		*bif_ifp;	/* member if */
224 	struct bstp_port	bif_stp;	/* STP state */
225 	uint32_t		bif_flags;	/* member if flags */
226 	int			bif_savedcaps;	/* saved capabilities */
227 	uint32_t		bif_addrmax;	/* max # of addresses */
228 	uint32_t		bif_addrcnt;	/* cur. # of addresses */
229 	uint32_t		bif_addrexceeded;/* # of address violations */
230 	struct epoch_context	bif_epoch_ctx;
231 };
232 
233 /*
234  * Bridge route node.
235  */
236 struct bridge_rtnode {
237 	CK_LIST_ENTRY(bridge_rtnode) brt_hash;	/* hash table linkage */
238 	CK_LIST_ENTRY(bridge_rtnode) brt_list;	/* list linkage */
239 	struct bridge_iflist	*brt_dst;	/* destination if */
240 	unsigned long		brt_expire;	/* expiration time */
241 	uint8_t			brt_flags;	/* address flags */
242 	uint8_t			brt_addr[ETHER_ADDR_LEN];
243 	uint16_t		brt_vlan;	/* vlan id */
244 	struct	vnet		*brt_vnet;
245 	struct	epoch_context	brt_epoch_ctx;
246 };
247 #define	brt_ifp			brt_dst->bif_ifp
248 
249 /*
250  * Software state for each bridge.
251  */
252 struct bridge_softc {
253 	struct ifnet		*sc_ifp;	/* make this an interface */
254 	LIST_ENTRY(bridge_softc) sc_list;
255 	struct sx		sc_sx;
256 	struct mtx		sc_rt_mtx;
257 	uint32_t		sc_brtmax;	/* max # of addresses */
258 	uint32_t		sc_brtcnt;	/* cur. # of addresses */
259 	uint32_t		sc_brttimeout;	/* rt timeout in seconds */
260 	struct callout		sc_brcallout;	/* bridge callout */
261 	CK_LIST_HEAD(, bridge_iflist) sc_iflist;	/* member interface list */
262 	CK_LIST_HEAD(, bridge_rtnode) *sc_rthash;	/* our forwarding table */
263 	CK_LIST_HEAD(, bridge_rtnode) sc_rtlist;	/* list version of above */
264 	uint32_t		sc_rthash_key;	/* key for hash */
265 	CK_LIST_HEAD(, bridge_iflist) sc_spanlist;	/* span ports list */
266 	struct bstp_state	sc_stp;		/* STP state */
267 	uint32_t		sc_brtexceeded;	/* # of cache drops */
268 	struct ifnet		*sc_ifaddr;	/* member mac copied from */
269 	struct ether_addr	sc_defaddr;	/* Default MAC address */
270 	if_input_fn_t		sc_if_input;	/* Saved copy of if_input */
271 	struct epoch_context	sc_epoch_ctx;
272 };
273 
274 VNET_DEFINE_STATIC(struct sx, bridge_list_sx);
275 #define	V_bridge_list_sx	VNET(bridge_list_sx)
276 static eventhandler_tag bridge_detach_cookie;
277 
278 int	bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
279 
280 VNET_DEFINE_STATIC(uma_zone_t, bridge_rtnode_zone);
281 #define	V_bridge_rtnode_zone	VNET(bridge_rtnode_zone)
282 
283 static int	bridge_clone_create(struct if_clone *, char *, size_t,
284 		    struct ifc_data *, struct ifnet **);
285 static int	bridge_clone_destroy(struct if_clone *, struct ifnet *, uint32_t);
286 
287 static int	bridge_ioctl(struct ifnet *, u_long, caddr_t);
288 static void	bridge_mutecaps(struct bridge_softc *);
289 static void	bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
290 		    int);
291 static void	bridge_ifdetach(void *arg __unused, struct ifnet *);
292 static void	bridge_init(void *);
293 static void	bridge_dummynet(struct mbuf *, struct ifnet *);
294 static void	bridge_stop(struct ifnet *, int);
295 static int	bridge_transmit(struct ifnet *, struct mbuf *);
296 #ifdef ALTQ
297 static void	bridge_altq_start(if_t);
298 static int	bridge_altq_transmit(if_t, struct mbuf *);
299 #endif
300 static void	bridge_qflush(struct ifnet *);
301 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
302 static void	bridge_inject(struct ifnet *, struct mbuf *);
303 static int	bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
304 		    struct rtentry *);
305 static int	bridge_enqueue(struct bridge_softc *, struct ifnet *,
306 		    struct mbuf *);
307 static void	bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
308 
309 static void	bridge_forward(struct bridge_softc *, struct bridge_iflist *,
310 		    struct mbuf *m);
311 
312 static void	bridge_timer(void *);
313 
314 static void	bridge_broadcast(struct bridge_softc *, struct ifnet *,
315 		    struct mbuf *, int);
316 static void	bridge_span(struct bridge_softc *, struct mbuf *);
317 
318 static int	bridge_rtupdate(struct bridge_softc *, const uint8_t *,
319 		    uint16_t, struct bridge_iflist *, int, uint8_t);
320 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
321 		    uint16_t);
322 static void	bridge_rttrim(struct bridge_softc *);
323 static void	bridge_rtage(struct bridge_softc *);
324 static void	bridge_rtflush(struct bridge_softc *, int);
325 static int	bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
326 		    uint16_t);
327 
328 static void	bridge_rtable_init(struct bridge_softc *);
329 static void	bridge_rtable_fini(struct bridge_softc *);
330 
331 static int	bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
332 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
333 		    const uint8_t *, uint16_t);
334 static int	bridge_rtnode_insert(struct bridge_softc *,
335 		    struct bridge_rtnode *);
336 static void	bridge_rtnode_destroy(struct bridge_softc *,
337 		    struct bridge_rtnode *);
338 static void	bridge_rtable_expire(struct ifnet *, int);
339 static void	bridge_state_change(struct ifnet *, int);
340 
341 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
342 		    const char *name);
343 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
344 		    struct ifnet *ifp);
345 static void	bridge_delete_member(struct bridge_softc *,
346 		    struct bridge_iflist *, int);
347 static void	bridge_delete_span(struct bridge_softc *,
348 		    struct bridge_iflist *);
349 
350 static int	bridge_ioctl_add(struct bridge_softc *, void *);
351 static int	bridge_ioctl_del(struct bridge_softc *, void *);
352 static int	bridge_ioctl_gifflags(struct bridge_softc *, void *);
353 static int	bridge_ioctl_sifflags(struct bridge_softc *, void *);
354 static int	bridge_ioctl_scache(struct bridge_softc *, void *);
355 static int	bridge_ioctl_gcache(struct bridge_softc *, void *);
356 static int	bridge_ioctl_gifs(struct bridge_softc *, void *);
357 static int	bridge_ioctl_rts(struct bridge_softc *, void *);
358 static int	bridge_ioctl_saddr(struct bridge_softc *, void *);
359 static int	bridge_ioctl_sto(struct bridge_softc *, void *);
360 static int	bridge_ioctl_gto(struct bridge_softc *, void *);
361 static int	bridge_ioctl_daddr(struct bridge_softc *, void *);
362 static int	bridge_ioctl_flush(struct bridge_softc *, void *);
363 static int	bridge_ioctl_gpri(struct bridge_softc *, void *);
364 static int	bridge_ioctl_spri(struct bridge_softc *, void *);
365 static int	bridge_ioctl_ght(struct bridge_softc *, void *);
366 static int	bridge_ioctl_sht(struct bridge_softc *, void *);
367 static int	bridge_ioctl_gfd(struct bridge_softc *, void *);
368 static int	bridge_ioctl_sfd(struct bridge_softc *, void *);
369 static int	bridge_ioctl_gma(struct bridge_softc *, void *);
370 static int	bridge_ioctl_sma(struct bridge_softc *, void *);
371 static int	bridge_ioctl_sifprio(struct bridge_softc *, void *);
372 static int	bridge_ioctl_sifcost(struct bridge_softc *, void *);
373 static int	bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
374 static int	bridge_ioctl_addspan(struct bridge_softc *, void *);
375 static int	bridge_ioctl_delspan(struct bridge_softc *, void *);
376 static int	bridge_ioctl_gbparam(struct bridge_softc *, void *);
377 static int	bridge_ioctl_grte(struct bridge_softc *, void *);
378 static int	bridge_ioctl_gifsstp(struct bridge_softc *, void *);
379 static int	bridge_ioctl_sproto(struct bridge_softc *, void *);
380 static int	bridge_ioctl_stxhc(struct bridge_softc *, void *);
381 static int	bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
382 		    int);
383 static int	bridge_ip_checkbasic(struct mbuf **mp);
384 #ifdef INET6
385 static int	bridge_ip6_checkbasic(struct mbuf **mp);
386 #endif /* INET6 */
387 static int	bridge_fragment(struct ifnet *, struct mbuf **mp,
388 		    struct ether_header *, int, struct llc *);
389 static void	bridge_linkstate(struct ifnet *ifp);
390 static void	bridge_linkcheck(struct bridge_softc *sc);
391 
392 /*
393  * Use the "null" value from IEEE 802.1Q-2014 Table 9-2
394  * to indicate untagged frames.
395  */
396 #define	VLANTAGOF(_m)	\
397     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : DOT1Q_VID_NULL
398 
399 static struct bstp_cb_ops bridge_ops = {
400 	.bcb_state = bridge_state_change,
401 	.bcb_rtage = bridge_rtable_expire
402 };
403 
404 SYSCTL_DECL(_net_link);
405 static SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
406     "Bridge");
407 
408 /* only pass IP[46] packets when pfil is enabled */
409 VNET_DEFINE_STATIC(int, pfil_onlyip) = 1;
410 #define	V_pfil_onlyip	VNET(pfil_onlyip)
411 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip,
412     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_onlyip), 0,
413     "Only pass IP packets when pfil is enabled");
414 
415 /* run pfil hooks on the bridge interface */
416 VNET_DEFINE_STATIC(int, pfil_bridge) = 0;
417 #define	V_pfil_bridge	VNET(pfil_bridge)
418 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge,
419     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_bridge), 0,
420     "Packet filter on the bridge interface");
421 
422 /* layer2 filter with ipfw */
423 VNET_DEFINE_STATIC(int, pfil_ipfw);
424 #define	V_pfil_ipfw	VNET(pfil_ipfw)
425 
426 /* layer2 ARP filter with ipfw */
427 VNET_DEFINE_STATIC(int, pfil_ipfw_arp);
428 #define	V_pfil_ipfw_arp	VNET(pfil_ipfw_arp)
429 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp,
430     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_ipfw_arp), 0,
431     "Filter ARP packets through IPFW layer2");
432 
433 /* run pfil hooks on the member interface */
434 VNET_DEFINE_STATIC(int, pfil_member) = 0;
435 #define	V_pfil_member	VNET(pfil_member)
436 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member,
437     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_member), 0,
438     "Packet filter on the member interface");
439 
440 /* run pfil hooks on the physical interface for locally destined packets */
441 VNET_DEFINE_STATIC(int, pfil_local_phys);
442 #define	V_pfil_local_phys	VNET(pfil_local_phys)
443 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys,
444     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_local_phys), 0,
445     "Packet filter on the physical interface for locally destined packets");
446 
447 /* log STP state changes */
448 VNET_DEFINE_STATIC(int, log_stp);
449 #define	V_log_stp	VNET(log_stp)
450 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp,
451     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(log_stp), 0,
452     "Log STP state changes");
453 
454 /* share MAC with first bridge member */
455 VNET_DEFINE_STATIC(int, bridge_inherit_mac);
456 #define	V_bridge_inherit_mac	VNET(bridge_inherit_mac)
457 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac,
458     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(bridge_inherit_mac), 0,
459     "Inherit MAC address from the first bridge member");
460 
461 VNET_DEFINE_STATIC(int, allow_llz_overlap) = 0;
462 #define	V_allow_llz_overlap	VNET(allow_llz_overlap)
463 SYSCTL_INT(_net_link_bridge, OID_AUTO, allow_llz_overlap,
464     CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(allow_llz_overlap), 0,
465     "Allow overlap of link-local scope "
466     "zones of a bridge interface and the member interfaces");
467 
468 /* log MAC address port flapping */
469 VNET_DEFINE_STATIC(bool, log_mac_flap) = true;
470 #define	V_log_mac_flap	VNET(log_mac_flap)
471 SYSCTL_BOOL(_net_link_bridge, OID_AUTO, log_mac_flap,
472     CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(log_mac_flap), true,
473     "Log MAC address port flapping");
474 
475 VNET_DEFINE_STATIC(int, log_interval) = 5;
476 VNET_DEFINE_STATIC(int, log_count) = 0;
477 VNET_DEFINE_STATIC(struct timeval, log_last) = { 0 };
478 
479 #define	V_log_interval	VNET(log_interval)
480 #define	V_log_count	VNET(log_count)
481 #define	V_log_last	VNET(log_last)
482 
483 struct bridge_control {
484 	int	(*bc_func)(struct bridge_softc *, void *);
485 	int	bc_argsize;
486 	int	bc_flags;
487 };
488 
489 #define	BC_F_COPYIN		0x01	/* copy arguments in */
490 #define	BC_F_COPYOUT		0x02	/* copy arguments out */
491 #define	BC_F_SUSER		0x04	/* do super-user check */
492 
493 static const struct bridge_control bridge_control_table[] = {
494 	{ bridge_ioctl_add,		sizeof(struct ifbreq),
495 	  BC_F_COPYIN|BC_F_SUSER },
496 	{ bridge_ioctl_del,		sizeof(struct ifbreq),
497 	  BC_F_COPYIN|BC_F_SUSER },
498 
499 	{ bridge_ioctl_gifflags,	sizeof(struct ifbreq),
500 	  BC_F_COPYIN|BC_F_COPYOUT },
501 	{ bridge_ioctl_sifflags,	sizeof(struct ifbreq),
502 	  BC_F_COPYIN|BC_F_SUSER },
503 
504 	{ bridge_ioctl_scache,		sizeof(struct ifbrparam),
505 	  BC_F_COPYIN|BC_F_SUSER },
506 	{ bridge_ioctl_gcache,		sizeof(struct ifbrparam),
507 	  BC_F_COPYOUT },
508 
509 	{ bridge_ioctl_gifs,		sizeof(struct ifbifconf),
510 	  BC_F_COPYIN|BC_F_COPYOUT },
511 	{ bridge_ioctl_rts,		sizeof(struct ifbaconf),
512 	  BC_F_COPYIN|BC_F_COPYOUT },
513 
514 	{ bridge_ioctl_saddr,		sizeof(struct ifbareq),
515 	  BC_F_COPYIN|BC_F_SUSER },
516 
517 	{ bridge_ioctl_sto,		sizeof(struct ifbrparam),
518 	  BC_F_COPYIN|BC_F_SUSER },
519 	{ bridge_ioctl_gto,		sizeof(struct ifbrparam),
520 	  BC_F_COPYOUT },
521 
522 	{ bridge_ioctl_daddr,		sizeof(struct ifbareq),
523 	  BC_F_COPYIN|BC_F_SUSER },
524 
525 	{ bridge_ioctl_flush,		sizeof(struct ifbreq),
526 	  BC_F_COPYIN|BC_F_SUSER },
527 
528 	{ bridge_ioctl_gpri,		sizeof(struct ifbrparam),
529 	  BC_F_COPYOUT },
530 	{ bridge_ioctl_spri,		sizeof(struct ifbrparam),
531 	  BC_F_COPYIN|BC_F_SUSER },
532 
533 	{ bridge_ioctl_ght,		sizeof(struct ifbrparam),
534 	  BC_F_COPYOUT },
535 	{ bridge_ioctl_sht,		sizeof(struct ifbrparam),
536 	  BC_F_COPYIN|BC_F_SUSER },
537 
538 	{ bridge_ioctl_gfd,		sizeof(struct ifbrparam),
539 	  BC_F_COPYOUT },
540 	{ bridge_ioctl_sfd,		sizeof(struct ifbrparam),
541 	  BC_F_COPYIN|BC_F_SUSER },
542 
543 	{ bridge_ioctl_gma,		sizeof(struct ifbrparam),
544 	  BC_F_COPYOUT },
545 	{ bridge_ioctl_sma,		sizeof(struct ifbrparam),
546 	  BC_F_COPYIN|BC_F_SUSER },
547 
548 	{ bridge_ioctl_sifprio,		sizeof(struct ifbreq),
549 	  BC_F_COPYIN|BC_F_SUSER },
550 
551 	{ bridge_ioctl_sifcost,		sizeof(struct ifbreq),
552 	  BC_F_COPYIN|BC_F_SUSER },
553 
554 	{ bridge_ioctl_addspan,		sizeof(struct ifbreq),
555 	  BC_F_COPYIN|BC_F_SUSER },
556 	{ bridge_ioctl_delspan,		sizeof(struct ifbreq),
557 	  BC_F_COPYIN|BC_F_SUSER },
558 
559 	{ bridge_ioctl_gbparam,		sizeof(struct ifbropreq),
560 	  BC_F_COPYOUT },
561 
562 	{ bridge_ioctl_grte,		sizeof(struct ifbrparam),
563 	  BC_F_COPYOUT },
564 
565 	{ bridge_ioctl_gifsstp,		sizeof(struct ifbpstpconf),
566 	  BC_F_COPYIN|BC_F_COPYOUT },
567 
568 	{ bridge_ioctl_sproto,		sizeof(struct ifbrparam),
569 	  BC_F_COPYIN|BC_F_SUSER },
570 
571 	{ bridge_ioctl_stxhc,		sizeof(struct ifbrparam),
572 	  BC_F_COPYIN|BC_F_SUSER },
573 
574 	{ bridge_ioctl_sifmaxaddr,	sizeof(struct ifbreq),
575 	  BC_F_COPYIN|BC_F_SUSER },
576 
577 };
578 static const int bridge_control_table_size = nitems(bridge_control_table);
579 
580 VNET_DEFINE_STATIC(LIST_HEAD(, bridge_softc), bridge_list);
581 #define	V_bridge_list	VNET(bridge_list)
582 #define	BRIDGE_LIST_LOCK_INIT(x)	sx_init(&V_bridge_list_sx,	\
583 					    "if_bridge list")
584 #define	BRIDGE_LIST_LOCK_DESTROY(x)	sx_destroy(&V_bridge_list_sx)
585 #define	BRIDGE_LIST_LOCK(x)		sx_xlock(&V_bridge_list_sx)
586 #define	BRIDGE_LIST_UNLOCK(x)		sx_xunlock(&V_bridge_list_sx)
587 
588 VNET_DEFINE_STATIC(struct if_clone *, bridge_cloner);
589 #define	V_bridge_cloner	VNET(bridge_cloner)
590 
591 static const char bridge_name[] = "bridge";
592 
593 static void
594 vnet_bridge_init(const void *unused __unused)
595 {
596 
597 	V_bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
598 	    sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
599 	    UMA_ALIGN_PTR, 0);
600 	BRIDGE_LIST_LOCK_INIT();
601 	LIST_INIT(&V_bridge_list);
602 
603 	struct if_clone_addreq req = {
604 		.create_f = bridge_clone_create,
605 		.destroy_f = bridge_clone_destroy,
606 		.flags = IFC_F_AUTOUNIT,
607 	};
608 	V_bridge_cloner = ifc_attach_cloner(bridge_name, &req);
609 }
610 VNET_SYSINIT(vnet_bridge_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
611     vnet_bridge_init, NULL);
612 
613 static void
614 vnet_bridge_uninit(const void *unused __unused)
615 {
616 
617 	ifc_detach_cloner(V_bridge_cloner);
618 	V_bridge_cloner = NULL;
619 	BRIDGE_LIST_LOCK_DESTROY();
620 
621 	/* Callbacks may use the UMA zone. */
622 	NET_EPOCH_DRAIN_CALLBACKS();
623 
624 	uma_zdestroy(V_bridge_rtnode_zone);
625 }
626 VNET_SYSUNINIT(vnet_bridge_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY,
627     vnet_bridge_uninit, NULL);
628 
629 static int
630 bridge_modevent(module_t mod, int type, void *data)
631 {
632 
633 	switch (type) {
634 	case MOD_LOAD:
635 		bridge_dn_p = bridge_dummynet;
636 		bridge_detach_cookie = EVENTHANDLER_REGISTER(
637 		    ifnet_departure_event, bridge_ifdetach, NULL,
638 		    EVENTHANDLER_PRI_ANY);
639 		break;
640 	case MOD_UNLOAD:
641 		EVENTHANDLER_DEREGISTER(ifnet_departure_event,
642 		    bridge_detach_cookie);
643 		bridge_dn_p = NULL;
644 		break;
645 	default:
646 		return (EOPNOTSUPP);
647 	}
648 	return (0);
649 }
650 
651 static moduledata_t bridge_mod = {
652 	"if_bridge",
653 	bridge_modevent,
654 	0
655 };
656 
657 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
658 MODULE_VERSION(if_bridge, 1);
659 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
660 
661 /*
662  * handler for net.link.bridge.ipfw
663  */
664 static int
665 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
666 {
667 	int enable = V_pfil_ipfw;
668 	int error;
669 
670 	error = sysctl_handle_int(oidp, &enable, 0, req);
671 	enable &= 1;
672 
673 	if (enable != V_pfil_ipfw) {
674 		V_pfil_ipfw = enable;
675 
676 		/*
677 		 * Disable pfil so that ipfw doesnt run twice, if the user
678 		 * really wants both then they can re-enable pfil_bridge and/or
679 		 * pfil_member. Also allow non-ip packets as ipfw can filter by
680 		 * layer2 type.
681 		 */
682 		if (V_pfil_ipfw) {
683 			V_pfil_onlyip = 0;
684 			V_pfil_bridge = 0;
685 			V_pfil_member = 0;
686 		}
687 	}
688 
689 	return (error);
690 }
691 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw,
692     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_VNET | CTLFLAG_NEEDGIANT,
693     &VNET_NAME(pfil_ipfw), 0, &sysctl_pfil_ipfw, "I",
694     "Layer2 filter with IPFW");
695 
696 #ifdef VIMAGE
697 static void
698 bridge_reassign(struct ifnet *ifp, struct vnet *newvnet, char *arg)
699 {
700 	struct bridge_softc *sc = ifp->if_softc;
701 	struct bridge_iflist *bif;
702 
703 	BRIDGE_LOCK(sc);
704 
705 	while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
706 		bridge_delete_member(sc, bif, 0);
707 
708 	while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
709 		bridge_delete_span(sc, bif);
710 	}
711 
712 	BRIDGE_UNLOCK(sc);
713 
714 	ether_reassign(ifp, newvnet, arg);
715 }
716 #endif
717 
718 /*
719  * bridge_clone_create:
720  *
721  *	Create a new bridge instance.
722  */
723 static int
724 bridge_clone_create(struct if_clone *ifc, char *name, size_t len,
725     struct ifc_data *ifd, struct ifnet **ifpp)
726 {
727 	struct bridge_softc *sc;
728 	struct ifnet *ifp;
729 
730 	sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
731 	ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
732 	if (ifp == NULL) {
733 		free(sc, M_DEVBUF);
734 		return (ENOSPC);
735 	}
736 
737 	BRIDGE_LOCK_INIT(sc);
738 	sc->sc_brtmax = BRIDGE_RTABLE_MAX;
739 	sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
740 
741 	/* Initialize our routing table. */
742 	bridge_rtable_init(sc);
743 
744 	callout_init_mtx(&sc->sc_brcallout, &sc->sc_rt_mtx, 0);
745 
746 	CK_LIST_INIT(&sc->sc_iflist);
747 	CK_LIST_INIT(&sc->sc_spanlist);
748 
749 	ifp->if_softc = sc;
750 	if_initname(ifp, bridge_name, ifd->unit);
751 	ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
752 	ifp->if_ioctl = bridge_ioctl;
753 #ifdef ALTQ
754 	ifp->if_start = bridge_altq_start;
755 	ifp->if_transmit = bridge_altq_transmit;
756 	IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
757 	ifp->if_snd.ifq_drv_maxlen = 0;
758 	IFQ_SET_READY(&ifp->if_snd);
759 #else
760 	ifp->if_transmit = bridge_transmit;
761 #endif
762 	ifp->if_qflush = bridge_qflush;
763 	ifp->if_init = bridge_init;
764 	ifp->if_type = IFT_BRIDGE;
765 
766 	ether_gen_addr(ifp, &sc->sc_defaddr);
767 
768 	bstp_attach(&sc->sc_stp, &bridge_ops);
769 	ether_ifattach(ifp, sc->sc_defaddr.octet);
770 	/* Now undo some of the damage... */
771 	ifp->if_baudrate = 0;
772 	ifp->if_type = IFT_BRIDGE;
773 #ifdef VIMAGE
774 	ifp->if_reassign = bridge_reassign;
775 #endif
776 	sc->sc_if_input = ifp->if_input;	/* ether_input */
777 	ifp->if_input = bridge_inject;
778 
779 	/*
780 	 * Allow BRIDGE_INPUT() to pass in packets originating from the bridge
781 	 * itself via bridge_inject().  This is required for netmap but
782 	 * otherwise has no effect.
783 	 */
784 	ifp->if_bridge_input = bridge_input;
785 
786 	BRIDGE_LIST_LOCK();
787 	LIST_INSERT_HEAD(&V_bridge_list, sc, sc_list);
788 	BRIDGE_LIST_UNLOCK();
789 	*ifpp = ifp;
790 
791 	return (0);
792 }
793 
794 static void
795 bridge_clone_destroy_cb(struct epoch_context *ctx)
796 {
797 	struct bridge_softc *sc;
798 
799 	sc = __containerof(ctx, struct bridge_softc, sc_epoch_ctx);
800 
801 	BRIDGE_LOCK_DESTROY(sc);
802 	free(sc, M_DEVBUF);
803 }
804 
805 /*
806  * bridge_clone_destroy:
807  *
808  *	Destroy a bridge instance.
809  */
810 static int
811 bridge_clone_destroy(struct if_clone *ifc, struct ifnet *ifp, uint32_t flags)
812 {
813 	struct bridge_softc *sc = ifp->if_softc;
814 	struct bridge_iflist *bif;
815 	struct epoch_tracker et;
816 
817 	BRIDGE_LOCK(sc);
818 
819 	bridge_stop(ifp, 1);
820 	ifp->if_flags &= ~IFF_UP;
821 
822 	while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
823 		bridge_delete_member(sc, bif, 0);
824 
825 	while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
826 		bridge_delete_span(sc, bif);
827 	}
828 
829 	/* Tear down the routing table. */
830 	bridge_rtable_fini(sc);
831 
832 	BRIDGE_UNLOCK(sc);
833 
834 	NET_EPOCH_ENTER(et);
835 
836 	callout_drain(&sc->sc_brcallout);
837 
838 	BRIDGE_LIST_LOCK();
839 	LIST_REMOVE(sc, sc_list);
840 	BRIDGE_LIST_UNLOCK();
841 
842 	bstp_detach(&sc->sc_stp);
843 #ifdef ALTQ
844 	IFQ_PURGE(&ifp->if_snd);
845 #endif
846 	NET_EPOCH_EXIT(et);
847 
848 	ether_ifdetach(ifp);
849 	if_free(ifp);
850 
851 	NET_EPOCH_CALL(bridge_clone_destroy_cb, &sc->sc_epoch_ctx);
852 
853 	return (0);
854 }
855 
856 /*
857  * bridge_ioctl:
858  *
859  *	Handle a control request from the operator.
860  */
861 static int
862 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
863 {
864 	struct bridge_softc *sc = ifp->if_softc;
865 	struct ifreq *ifr = (struct ifreq *)data;
866 	struct bridge_iflist *bif;
867 	struct thread *td = curthread;
868 	union {
869 		struct ifbreq ifbreq;
870 		struct ifbifconf ifbifconf;
871 		struct ifbareq ifbareq;
872 		struct ifbaconf ifbaconf;
873 		struct ifbrparam ifbrparam;
874 		struct ifbropreq ifbropreq;
875 	} args;
876 	struct ifdrv *ifd = (struct ifdrv *) data;
877 	const struct bridge_control *bc;
878 	int error = 0, oldmtu;
879 
880 	BRIDGE_LOCK(sc);
881 
882 	switch (cmd) {
883 	case SIOCADDMULTI:
884 	case SIOCDELMULTI:
885 		break;
886 
887 	case SIOCGDRVSPEC:
888 	case SIOCSDRVSPEC:
889 		if (ifd->ifd_cmd >= bridge_control_table_size) {
890 			error = EINVAL;
891 			break;
892 		}
893 		bc = &bridge_control_table[ifd->ifd_cmd];
894 
895 		if (cmd == SIOCGDRVSPEC &&
896 		    (bc->bc_flags & BC_F_COPYOUT) == 0) {
897 			error = EINVAL;
898 			break;
899 		}
900 		else if (cmd == SIOCSDRVSPEC &&
901 		    (bc->bc_flags & BC_F_COPYOUT) != 0) {
902 			error = EINVAL;
903 			break;
904 		}
905 
906 		if (bc->bc_flags & BC_F_SUSER) {
907 			error = priv_check(td, PRIV_NET_BRIDGE);
908 			if (error)
909 				break;
910 		}
911 
912 		if (ifd->ifd_len != bc->bc_argsize ||
913 		    ifd->ifd_len > sizeof(args)) {
914 			error = EINVAL;
915 			break;
916 		}
917 
918 		bzero(&args, sizeof(args));
919 		if (bc->bc_flags & BC_F_COPYIN) {
920 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
921 			if (error)
922 				break;
923 		}
924 
925 		oldmtu = ifp->if_mtu;
926 		error = (*bc->bc_func)(sc, &args);
927 		if (error)
928 			break;
929 
930 		/*
931 		 * Bridge MTU may change during addition of the first port.
932 		 * If it did, do network layer specific procedure.
933 		 */
934 		if (ifp->if_mtu != oldmtu)
935 			if_notifymtu(ifp);
936 
937 		if (bc->bc_flags & BC_F_COPYOUT)
938 			error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
939 
940 		break;
941 
942 	case SIOCSIFFLAGS:
943 		if (!(ifp->if_flags & IFF_UP) &&
944 		    (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
945 			/*
946 			 * If interface is marked down and it is running,
947 			 * then stop and disable it.
948 			 */
949 			bridge_stop(ifp, 1);
950 		} else if ((ifp->if_flags & IFF_UP) &&
951 		    !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
952 			/*
953 			 * If interface is marked up and it is stopped, then
954 			 * start it.
955 			 */
956 			BRIDGE_UNLOCK(sc);
957 			(*ifp->if_init)(sc);
958 			BRIDGE_LOCK(sc);
959 		}
960 		break;
961 
962 	case SIOCSIFMTU:
963 		oldmtu = sc->sc_ifp->if_mtu;
964 
965 		if (ifr->ifr_mtu < 576) {
966 			error = EINVAL;
967 			break;
968 		}
969 		if (CK_LIST_EMPTY(&sc->sc_iflist)) {
970 			sc->sc_ifp->if_mtu = ifr->ifr_mtu;
971 			break;
972 		}
973 		CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
974 			error = (*bif->bif_ifp->if_ioctl)(bif->bif_ifp,
975 			    SIOCSIFMTU, (caddr_t)ifr);
976 			if (error != 0) {
977 				log(LOG_NOTICE, "%s: invalid MTU: %u for"
978 				    " member %s\n", sc->sc_ifp->if_xname,
979 				    ifr->ifr_mtu,
980 				    bif->bif_ifp->if_xname);
981 				error = EINVAL;
982 				break;
983 			}
984 		}
985 		if (error) {
986 			/* Restore the previous MTU on all member interfaces. */
987 			ifr->ifr_mtu = oldmtu;
988 			CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
989 				(*bif->bif_ifp->if_ioctl)(bif->bif_ifp,
990 				    SIOCSIFMTU, (caddr_t)ifr);
991 			}
992 		} else {
993 			sc->sc_ifp->if_mtu = ifr->ifr_mtu;
994 		}
995 		break;
996 	default:
997 		/*
998 		 * drop the lock as ether_ioctl() will call bridge_start() and
999 		 * cause the lock to be recursed.
1000 		 */
1001 		BRIDGE_UNLOCK(sc);
1002 		error = ether_ioctl(ifp, cmd, data);
1003 		BRIDGE_LOCK(sc);
1004 		break;
1005 	}
1006 
1007 	BRIDGE_UNLOCK(sc);
1008 
1009 	return (error);
1010 }
1011 
1012 /*
1013  * bridge_mutecaps:
1014  *
1015  *	Clear or restore unwanted capabilities on the member interface
1016  */
1017 static void
1018 bridge_mutecaps(struct bridge_softc *sc)
1019 {
1020 	struct bridge_iflist *bif;
1021 	int enabled, mask;
1022 
1023 	BRIDGE_LOCK_ASSERT(sc);
1024 
1025 	/* Initial bitmask of capabilities to test */
1026 	mask = BRIDGE_IFCAPS_MASK;
1027 
1028 	CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1029 		/* Every member must support it or its disabled */
1030 		mask &= bif->bif_savedcaps;
1031 	}
1032 
1033 	CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1034 		enabled = bif->bif_ifp->if_capenable;
1035 		enabled &= ~BRIDGE_IFCAPS_STRIP;
1036 		/* strip off mask bits and enable them again if allowed */
1037 		enabled &= ~BRIDGE_IFCAPS_MASK;
1038 		enabled |= mask;
1039 		bridge_set_ifcap(sc, bif, enabled);
1040 	}
1041 }
1042 
1043 static void
1044 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
1045 {
1046 	struct ifnet *ifp = bif->bif_ifp;
1047 	struct ifreq ifr;
1048 	int error, mask, stuck;
1049 
1050 	bzero(&ifr, sizeof(ifr));
1051 	ifr.ifr_reqcap = set;
1052 
1053 	if (ifp->if_capenable != set) {
1054 		error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
1055 		if (error)
1056 			if_printf(sc->sc_ifp,
1057 			    "error setting capabilities on %s: %d\n",
1058 			    ifp->if_xname, error);
1059 		mask = BRIDGE_IFCAPS_MASK | BRIDGE_IFCAPS_STRIP;
1060 		stuck = ifp->if_capenable & mask & ~set;
1061 		if (stuck != 0)
1062 			if_printf(sc->sc_ifp,
1063 			    "can't disable some capabilities on %s: 0x%x\n",
1064 			    ifp->if_xname, stuck);
1065 	}
1066 }
1067 
1068 /*
1069  * bridge_lookup_member:
1070  *
1071  *	Lookup a bridge member interface.
1072  */
1073 static struct bridge_iflist *
1074 bridge_lookup_member(struct bridge_softc *sc, const char *name)
1075 {
1076 	struct bridge_iflist *bif;
1077 	struct ifnet *ifp;
1078 
1079 	BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1080 
1081 	CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1082 		ifp = bif->bif_ifp;
1083 		if (strcmp(ifp->if_xname, name) == 0)
1084 			return (bif);
1085 	}
1086 
1087 	return (NULL);
1088 }
1089 
1090 /*
1091  * bridge_lookup_member_if:
1092  *
1093  *	Lookup a bridge member interface by ifnet*.
1094  */
1095 static struct bridge_iflist *
1096 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
1097 {
1098 	struct bridge_iflist *bif;
1099 
1100 	BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1101 
1102 	CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1103 		if (bif->bif_ifp == member_ifp)
1104 			return (bif);
1105 	}
1106 
1107 	return (NULL);
1108 }
1109 
1110 static void
1111 bridge_delete_member_cb(struct epoch_context *ctx)
1112 {
1113 	struct bridge_iflist *bif;
1114 
1115 	bif = __containerof(ctx, struct bridge_iflist, bif_epoch_ctx);
1116 
1117 	free(bif, M_DEVBUF);
1118 }
1119 
1120 /*
1121  * bridge_delete_member:
1122  *
1123  *	Delete the specified member interface.
1124  */
1125 static void
1126 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
1127     int gone)
1128 {
1129 	struct ifnet *ifs = bif->bif_ifp;
1130 	struct ifnet *fif = NULL;
1131 	struct bridge_iflist *bifl;
1132 
1133 	BRIDGE_LOCK_ASSERT(sc);
1134 
1135 	if (bif->bif_flags & IFBIF_STP)
1136 		bstp_disable(&bif->bif_stp);
1137 
1138 	ifs->if_bridge = NULL;
1139 	CK_LIST_REMOVE(bif, bif_next);
1140 
1141 	/*
1142 	 * If removing the interface that gave the bridge its mac address, set
1143 	 * the mac address of the bridge to the address of the next member, or
1144 	 * to its default address if no members are left.
1145 	 */
1146 	if (V_bridge_inherit_mac && sc->sc_ifaddr == ifs) {
1147 		if (CK_LIST_EMPTY(&sc->sc_iflist)) {
1148 			bcopy(&sc->sc_defaddr,
1149 			    IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1150 			sc->sc_ifaddr = NULL;
1151 		} else {
1152 			bifl = CK_LIST_FIRST(&sc->sc_iflist);
1153 			fif = bifl->bif_ifp;
1154 			bcopy(IF_LLADDR(fif),
1155 			    IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1156 			sc->sc_ifaddr = fif;
1157 		}
1158 		EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1159 	}
1160 
1161 	bridge_linkcheck(sc);
1162 	bridge_mutecaps(sc);	/* recalcuate now this interface is removed */
1163 	BRIDGE_RT_LOCK(sc);
1164 	bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
1165 	BRIDGE_RT_UNLOCK(sc);
1166 	KASSERT(bif->bif_addrcnt == 0,
1167 	    ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
1168 
1169 	ifs->if_bridge_output = NULL;
1170 	ifs->if_bridge_input = NULL;
1171 	ifs->if_bridge_linkstate = NULL;
1172 	if (!gone) {
1173 		switch (ifs->if_type) {
1174 		case IFT_ETHER:
1175 		case IFT_L2VLAN:
1176 			/*
1177 			 * Take the interface out of promiscuous mode, but only
1178 			 * if it was promiscuous in the first place. It might
1179 			 * not be if we're in the bridge_ioctl_add() error path.
1180 			 */
1181 			if (ifs->if_flags & IFF_PROMISC)
1182 				(void) ifpromisc(ifs, 0);
1183 			break;
1184 
1185 		case IFT_GIF:
1186 			break;
1187 
1188 		default:
1189 #ifdef DIAGNOSTIC
1190 			panic("bridge_delete_member: impossible");
1191 #endif
1192 			break;
1193 		}
1194 		/* reneable any interface capabilities */
1195 		bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
1196 	}
1197 	bstp_destroy(&bif->bif_stp);	/* prepare to free */
1198 
1199 	NET_EPOCH_CALL(bridge_delete_member_cb, &bif->bif_epoch_ctx);
1200 }
1201 
1202 /*
1203  * bridge_delete_span:
1204  *
1205  *	Delete the specified span interface.
1206  */
1207 static void
1208 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1209 {
1210 	BRIDGE_LOCK_ASSERT(sc);
1211 
1212 	KASSERT(bif->bif_ifp->if_bridge == NULL,
1213 	    ("%s: not a span interface", __func__));
1214 
1215 	CK_LIST_REMOVE(bif, bif_next);
1216 
1217 	NET_EPOCH_CALL(bridge_delete_member_cb, &bif->bif_epoch_ctx);
1218 }
1219 
1220 static int
1221 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1222 {
1223 	struct ifbreq *req = arg;
1224 	struct bridge_iflist *bif = NULL;
1225 	struct ifnet *ifs;
1226 	int error = 0;
1227 
1228 	ifs = ifunit(req->ifbr_ifsname);
1229 	if (ifs == NULL)
1230 		return (ENOENT);
1231 	if (ifs->if_ioctl == NULL)	/* must be supported */
1232 		return (EINVAL);
1233 
1234 	/* If it's in the span list, it can't be a member. */
1235 	CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1236 		if (ifs == bif->bif_ifp)
1237 			return (EBUSY);
1238 
1239 	if (ifs->if_bridge == sc)
1240 		return (EEXIST);
1241 
1242 	if (ifs->if_bridge != NULL)
1243 		return (EBUSY);
1244 
1245 	switch (ifs->if_type) {
1246 	case IFT_ETHER:
1247 	case IFT_L2VLAN:
1248 	case IFT_GIF:
1249 		/* permitted interface types */
1250 		break;
1251 	default:
1252 		return (EINVAL);
1253 	}
1254 
1255 #ifdef INET6
1256 	/*
1257 	 * Two valid inet6 addresses with link-local scope must not be
1258 	 * on the parent interface and the member interfaces at the
1259 	 * same time.  This restriction is needed to prevent violation
1260 	 * of link-local scope zone.  Attempts to add a member
1261 	 * interface which has inet6 addresses when the parent has
1262 	 * inet6 triggers removal of all inet6 addresses on the member
1263 	 * interface.
1264 	 */
1265 
1266 	/* Check if the parent interface has a link-local scope addr. */
1267 	if (V_allow_llz_overlap == 0 &&
1268 	    in6ifa_llaonifp(sc->sc_ifp) != NULL) {
1269 		/*
1270 		 * If any, remove all inet6 addresses from the member
1271 		 * interfaces.
1272 		 */
1273 		CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1274  			if (in6ifa_llaonifp(bif->bif_ifp)) {
1275 				in6_ifdetach(bif->bif_ifp);
1276 				if_printf(sc->sc_ifp,
1277 				    "IPv6 addresses on %s have been removed "
1278 				    "before adding it as a member to prevent "
1279 				    "IPv6 address scope violation.\n",
1280 				    bif->bif_ifp->if_xname);
1281 			}
1282 		}
1283 		if (in6ifa_llaonifp(ifs)) {
1284 			in6_ifdetach(ifs);
1285 			if_printf(sc->sc_ifp,
1286 			    "IPv6 addresses on %s have been removed "
1287 			    "before adding it as a member to prevent "
1288 			    "IPv6 address scope violation.\n",
1289 			    ifs->if_xname);
1290 		}
1291 	}
1292 #endif
1293 	/* Allow the first Ethernet member to define the MTU */
1294 	if (CK_LIST_EMPTY(&sc->sc_iflist))
1295 		sc->sc_ifp->if_mtu = ifs->if_mtu;
1296 	else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1297 		struct ifreq ifr;
1298 
1299 		snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "%s",
1300 		    ifs->if_xname);
1301 		ifr.ifr_mtu = sc->sc_ifp->if_mtu;
1302 
1303 		error = (*ifs->if_ioctl)(ifs,
1304 		    SIOCSIFMTU, (caddr_t)&ifr);
1305 		if (error != 0) {
1306 			log(LOG_NOTICE, "%s: invalid MTU: %u for"
1307 			    " new member %s\n", sc->sc_ifp->if_xname,
1308 			    ifr.ifr_mtu,
1309 			    ifs->if_xname);
1310 			return (EINVAL);
1311 		}
1312 	}
1313 
1314 	bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1315 	if (bif == NULL)
1316 		return (ENOMEM);
1317 
1318 	bif->bif_ifp = ifs;
1319 	bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1320 	bif->bif_savedcaps = ifs->if_capenable;
1321 
1322 	/*
1323 	 * Assign the interface's MAC address to the bridge if it's the first
1324 	 * member and the MAC address of the bridge has not been changed from
1325 	 * the default randomly generated one.
1326 	 */
1327 	if (V_bridge_inherit_mac && CK_LIST_EMPTY(&sc->sc_iflist) &&
1328 	    !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr.octet, ETHER_ADDR_LEN)) {
1329 		bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1330 		sc->sc_ifaddr = ifs;
1331 		EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1332 	}
1333 
1334 	ifs->if_bridge = sc;
1335 	ifs->if_bridge_output = bridge_output;
1336 	ifs->if_bridge_input = bridge_input;
1337 	ifs->if_bridge_linkstate = bridge_linkstate;
1338 	bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1339 	/*
1340 	 * XXX: XLOCK HERE!?!
1341 	 *
1342 	 * NOTE: insert_***HEAD*** should be safe for the traversals.
1343 	 */
1344 	CK_LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1345 
1346 	/* Set interface capabilities to the intersection set of all members */
1347 	bridge_mutecaps(sc);
1348 	bridge_linkcheck(sc);
1349 
1350 	/* Place the interface into promiscuous mode */
1351 	switch (ifs->if_type) {
1352 		case IFT_ETHER:
1353 		case IFT_L2VLAN:
1354 			error = ifpromisc(ifs, 1);
1355 			break;
1356 	}
1357 
1358 	if (error)
1359 		bridge_delete_member(sc, bif, 0);
1360 	return (error);
1361 }
1362 
1363 static int
1364 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1365 {
1366 	struct ifbreq *req = arg;
1367 	struct bridge_iflist *bif;
1368 
1369 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1370 	if (bif == NULL)
1371 		return (ENOENT);
1372 
1373 	bridge_delete_member(sc, bif, 0);
1374 
1375 	return (0);
1376 }
1377 
1378 static int
1379 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1380 {
1381 	struct ifbreq *req = arg;
1382 	struct bridge_iflist *bif;
1383 	struct bstp_port *bp;
1384 
1385 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1386 	if (bif == NULL)
1387 		return (ENOENT);
1388 
1389 	bp = &bif->bif_stp;
1390 	req->ifbr_ifsflags = bif->bif_flags;
1391 	req->ifbr_state = bp->bp_state;
1392 	req->ifbr_priority = bp->bp_priority;
1393 	req->ifbr_path_cost = bp->bp_path_cost;
1394 	req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1395 	req->ifbr_proto = bp->bp_protover;
1396 	req->ifbr_role = bp->bp_role;
1397 	req->ifbr_stpflags = bp->bp_flags;
1398 	req->ifbr_addrcnt = bif->bif_addrcnt;
1399 	req->ifbr_addrmax = bif->bif_addrmax;
1400 	req->ifbr_addrexceeded = bif->bif_addrexceeded;
1401 
1402 	/* Copy STP state options as flags */
1403 	if (bp->bp_operedge)
1404 		req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1405 	if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1406 		req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1407 	if (bp->bp_ptp_link)
1408 		req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1409 	if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1410 		req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1411 	if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1412 		req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1413 	if (bp->bp_flags & BSTP_PORT_ADMCOST)
1414 		req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1415 	return (0);
1416 }
1417 
1418 static int
1419 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1420 {
1421 	struct epoch_tracker et;
1422 	struct ifbreq *req = arg;
1423 	struct bridge_iflist *bif;
1424 	struct bstp_port *bp;
1425 	int error;
1426 
1427 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1428 	if (bif == NULL)
1429 		return (ENOENT);
1430 	bp = &bif->bif_stp;
1431 
1432 	if (req->ifbr_ifsflags & IFBIF_SPAN)
1433 		/* SPAN is readonly */
1434 		return (EINVAL);
1435 
1436 	NET_EPOCH_ENTER(et);
1437 
1438 	if (req->ifbr_ifsflags & IFBIF_STP) {
1439 		if ((bif->bif_flags & IFBIF_STP) == 0) {
1440 			error = bstp_enable(&bif->bif_stp);
1441 			if (error) {
1442 				NET_EPOCH_EXIT(et);
1443 				return (error);
1444 			}
1445 		}
1446 	} else {
1447 		if ((bif->bif_flags & IFBIF_STP) != 0)
1448 			bstp_disable(&bif->bif_stp);
1449 	}
1450 
1451 	/* Pass on STP flags */
1452 	bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1453 	bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1454 	bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1455 	bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1456 
1457 	/* Save the bits relating to the bridge */
1458 	bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1459 
1460 	NET_EPOCH_EXIT(et);
1461 
1462 	return (0);
1463 }
1464 
1465 static int
1466 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1467 {
1468 	struct ifbrparam *param = arg;
1469 
1470 	sc->sc_brtmax = param->ifbrp_csize;
1471 	bridge_rttrim(sc);
1472 
1473 	return (0);
1474 }
1475 
1476 static int
1477 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1478 {
1479 	struct ifbrparam *param = arg;
1480 
1481 	param->ifbrp_csize = sc->sc_brtmax;
1482 
1483 	return (0);
1484 }
1485 
1486 static int
1487 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1488 {
1489 	struct ifbifconf *bifc = arg;
1490 	struct bridge_iflist *bif;
1491 	struct ifbreq breq;
1492 	char *buf, *outbuf;
1493 	int count, buflen, len, error = 0;
1494 
1495 	count = 0;
1496 	CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1497 		count++;
1498 	CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1499 		count++;
1500 
1501 	buflen = sizeof(breq) * count;
1502 	if (bifc->ifbic_len == 0) {
1503 		bifc->ifbic_len = buflen;
1504 		return (0);
1505 	}
1506 	outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1507 	if (outbuf == NULL)
1508 		return (ENOMEM);
1509 
1510 	count = 0;
1511 	buf = outbuf;
1512 	len = min(bifc->ifbic_len, buflen);
1513 	bzero(&breq, sizeof(breq));
1514 	CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1515 		if (len < sizeof(breq))
1516 			break;
1517 
1518 		strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1519 		    sizeof(breq.ifbr_ifsname));
1520 		/* Fill in the ifbreq structure */
1521 		error = bridge_ioctl_gifflags(sc, &breq);
1522 		if (error)
1523 			break;
1524 		memcpy(buf, &breq, sizeof(breq));
1525 		count++;
1526 		buf += sizeof(breq);
1527 		len -= sizeof(breq);
1528 	}
1529 	CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1530 		if (len < sizeof(breq))
1531 			break;
1532 
1533 		strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1534 		    sizeof(breq.ifbr_ifsname));
1535 		breq.ifbr_ifsflags = bif->bif_flags;
1536 		breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1537 		memcpy(buf, &breq, sizeof(breq));
1538 		count++;
1539 		buf += sizeof(breq);
1540 		len -= sizeof(breq);
1541 	}
1542 
1543 	bifc->ifbic_len = sizeof(breq) * count;
1544 	error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1545 	free(outbuf, M_TEMP);
1546 	return (error);
1547 }
1548 
1549 static int
1550 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1551 {
1552 	struct ifbaconf *bac = arg;
1553 	struct bridge_rtnode *brt;
1554 	struct ifbareq bareq;
1555 	char *buf, *outbuf;
1556 	int count, buflen, len, error = 0;
1557 
1558 	if (bac->ifbac_len == 0)
1559 		return (0);
1560 
1561 	count = 0;
1562 	CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1563 		count++;
1564 	buflen = sizeof(bareq) * count;
1565 
1566 	outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1567 	if (outbuf == NULL)
1568 		return (ENOMEM);
1569 
1570 	count = 0;
1571 	buf = outbuf;
1572 	len = min(bac->ifbac_len, buflen);
1573 	bzero(&bareq, sizeof(bareq));
1574 	CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1575 		if (len < sizeof(bareq))
1576 			goto out;
1577 		strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1578 		    sizeof(bareq.ifba_ifsname));
1579 		memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1580 		bareq.ifba_vlan = brt->brt_vlan;
1581 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1582 				time_uptime < brt->brt_expire)
1583 			bareq.ifba_expire = brt->brt_expire - time_uptime;
1584 		else
1585 			bareq.ifba_expire = 0;
1586 		bareq.ifba_flags = brt->brt_flags;
1587 
1588 		memcpy(buf, &bareq, sizeof(bareq));
1589 		count++;
1590 		buf += sizeof(bareq);
1591 		len -= sizeof(bareq);
1592 	}
1593 out:
1594 	bac->ifbac_len = sizeof(bareq) * count;
1595 	error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1596 	free(outbuf, M_TEMP);
1597 	return (error);
1598 }
1599 
1600 static int
1601 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1602 {
1603 	struct ifbareq *req = arg;
1604 	struct bridge_iflist *bif;
1605 	struct epoch_tracker et;
1606 	int error;
1607 
1608 	NET_EPOCH_ENTER(et);
1609 	bif = bridge_lookup_member(sc, req->ifba_ifsname);
1610 	if (bif == NULL) {
1611 		NET_EPOCH_EXIT(et);
1612 		return (ENOENT);
1613 	}
1614 
1615 	/* bridge_rtupdate() may acquire the lock. */
1616 	error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1617 	    req->ifba_flags);
1618 	NET_EPOCH_EXIT(et);
1619 
1620 	return (error);
1621 }
1622 
1623 static int
1624 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1625 {
1626 	struct ifbrparam *param = arg;
1627 
1628 	sc->sc_brttimeout = param->ifbrp_ctime;
1629 	return (0);
1630 }
1631 
1632 static int
1633 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1634 {
1635 	struct ifbrparam *param = arg;
1636 
1637 	param->ifbrp_ctime = sc->sc_brttimeout;
1638 	return (0);
1639 }
1640 
1641 static int
1642 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1643 {
1644 	struct ifbareq *req = arg;
1645 	int vlan = req->ifba_vlan;
1646 
1647 	/* Userspace uses '0' to mean 'any vlan' */
1648 	if (vlan == 0)
1649 		vlan = DOT1Q_VID_RSVD_IMPL;
1650 
1651 	return (bridge_rtdaddr(sc, req->ifba_dst, vlan));
1652 }
1653 
1654 static int
1655 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1656 {
1657 	struct ifbreq *req = arg;
1658 
1659 	BRIDGE_RT_LOCK(sc);
1660 	bridge_rtflush(sc, req->ifbr_ifsflags);
1661 	BRIDGE_RT_UNLOCK(sc);
1662 
1663 	return (0);
1664 }
1665 
1666 static int
1667 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1668 {
1669 	struct ifbrparam *param = arg;
1670 	struct bstp_state *bs = &sc->sc_stp;
1671 
1672 	param->ifbrp_prio = bs->bs_bridge_priority;
1673 	return (0);
1674 }
1675 
1676 static int
1677 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1678 {
1679 	struct ifbrparam *param = arg;
1680 
1681 	return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1682 }
1683 
1684 static int
1685 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1686 {
1687 	struct ifbrparam *param = arg;
1688 	struct bstp_state *bs = &sc->sc_stp;
1689 
1690 	param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1691 	return (0);
1692 }
1693 
1694 static int
1695 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1696 {
1697 	struct ifbrparam *param = arg;
1698 
1699 	return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1700 }
1701 
1702 static int
1703 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1704 {
1705 	struct ifbrparam *param = arg;
1706 	struct bstp_state *bs = &sc->sc_stp;
1707 
1708 	param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1709 	return (0);
1710 }
1711 
1712 static int
1713 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1714 {
1715 	struct ifbrparam *param = arg;
1716 
1717 	return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1718 }
1719 
1720 static int
1721 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1722 {
1723 	struct ifbrparam *param = arg;
1724 	struct bstp_state *bs = &sc->sc_stp;
1725 
1726 	param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1727 	return (0);
1728 }
1729 
1730 static int
1731 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1732 {
1733 	struct ifbrparam *param = arg;
1734 
1735 	return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1736 }
1737 
1738 static int
1739 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1740 {
1741 	struct ifbreq *req = arg;
1742 	struct bridge_iflist *bif;
1743 
1744 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1745 	if (bif == NULL)
1746 		return (ENOENT);
1747 
1748 	return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1749 }
1750 
1751 static int
1752 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1753 {
1754 	struct ifbreq *req = arg;
1755 	struct bridge_iflist *bif;
1756 
1757 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1758 	if (bif == NULL)
1759 		return (ENOENT);
1760 
1761 	return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1762 }
1763 
1764 static int
1765 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1766 {
1767 	struct ifbreq *req = arg;
1768 	struct bridge_iflist *bif;
1769 
1770 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1771 	if (bif == NULL)
1772 		return (ENOENT);
1773 
1774 	bif->bif_addrmax = req->ifbr_addrmax;
1775 	return (0);
1776 }
1777 
1778 static int
1779 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1780 {
1781 	struct ifbreq *req = arg;
1782 	struct bridge_iflist *bif = NULL;
1783 	struct ifnet *ifs;
1784 
1785 	ifs = ifunit(req->ifbr_ifsname);
1786 	if (ifs == NULL)
1787 		return (ENOENT);
1788 
1789 	CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1790 		if (ifs == bif->bif_ifp)
1791 			return (EBUSY);
1792 
1793 	if (ifs->if_bridge != NULL)
1794 		return (EBUSY);
1795 
1796 	switch (ifs->if_type) {
1797 		case IFT_ETHER:
1798 		case IFT_GIF:
1799 		case IFT_L2VLAN:
1800 			break;
1801 		default:
1802 			return (EINVAL);
1803 	}
1804 
1805 	bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1806 	if (bif == NULL)
1807 		return (ENOMEM);
1808 
1809 	bif->bif_ifp = ifs;
1810 	bif->bif_flags = IFBIF_SPAN;
1811 
1812 	CK_LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1813 
1814 	return (0);
1815 }
1816 
1817 static int
1818 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1819 {
1820 	struct ifbreq *req = arg;
1821 	struct bridge_iflist *bif;
1822 	struct ifnet *ifs;
1823 
1824 	ifs = ifunit(req->ifbr_ifsname);
1825 	if (ifs == NULL)
1826 		return (ENOENT);
1827 
1828 	CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1829 		if (ifs == bif->bif_ifp)
1830 			break;
1831 
1832 	if (bif == NULL)
1833 		return (ENOENT);
1834 
1835 	bridge_delete_span(sc, bif);
1836 
1837 	return (0);
1838 }
1839 
1840 static int
1841 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1842 {
1843 	struct ifbropreq *req = arg;
1844 	struct bstp_state *bs = &sc->sc_stp;
1845 	struct bstp_port *root_port;
1846 
1847 	req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1848 	req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1849 	req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1850 
1851 	root_port = bs->bs_root_port;
1852 	if (root_port == NULL)
1853 		req->ifbop_root_port = 0;
1854 	else
1855 		req->ifbop_root_port = root_port->bp_ifp->if_index;
1856 
1857 	req->ifbop_holdcount = bs->bs_txholdcount;
1858 	req->ifbop_priority = bs->bs_bridge_priority;
1859 	req->ifbop_protocol = bs->bs_protover;
1860 	req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1861 	req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1862 	req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1863 	req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1864 	req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1865 	req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1866 
1867 	return (0);
1868 }
1869 
1870 static int
1871 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1872 {
1873 	struct ifbrparam *param = arg;
1874 
1875 	param->ifbrp_cexceeded = sc->sc_brtexceeded;
1876 	return (0);
1877 }
1878 
1879 static int
1880 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1881 {
1882 	struct ifbpstpconf *bifstp = arg;
1883 	struct bridge_iflist *bif;
1884 	struct bstp_port *bp;
1885 	struct ifbpstpreq bpreq;
1886 	char *buf, *outbuf;
1887 	int count, buflen, len, error = 0;
1888 
1889 	count = 0;
1890 	CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1891 		if ((bif->bif_flags & IFBIF_STP) != 0)
1892 			count++;
1893 	}
1894 
1895 	buflen = sizeof(bpreq) * count;
1896 	if (bifstp->ifbpstp_len == 0) {
1897 		bifstp->ifbpstp_len = buflen;
1898 		return (0);
1899 	}
1900 
1901 	outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1902 	if (outbuf == NULL)
1903 		return (ENOMEM);
1904 
1905 	count = 0;
1906 	buf = outbuf;
1907 	len = min(bifstp->ifbpstp_len, buflen);
1908 	bzero(&bpreq, sizeof(bpreq));
1909 	CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1910 		if (len < sizeof(bpreq))
1911 			break;
1912 
1913 		if ((bif->bif_flags & IFBIF_STP) == 0)
1914 			continue;
1915 
1916 		bp = &bif->bif_stp;
1917 		bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1918 		bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1919 		bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1920 		bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1921 		bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1922 		bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1923 
1924 		memcpy(buf, &bpreq, sizeof(bpreq));
1925 		count++;
1926 		buf += sizeof(bpreq);
1927 		len -= sizeof(bpreq);
1928 	}
1929 
1930 	bifstp->ifbpstp_len = sizeof(bpreq) * count;
1931 	error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1932 	free(outbuf, M_TEMP);
1933 	return (error);
1934 }
1935 
1936 static int
1937 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1938 {
1939 	struct ifbrparam *param = arg;
1940 
1941 	return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1942 }
1943 
1944 static int
1945 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1946 {
1947 	struct ifbrparam *param = arg;
1948 
1949 	return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1950 }
1951 
1952 /*
1953  * bridge_ifdetach:
1954  *
1955  *	Detach an interface from a bridge.  Called when a member
1956  *	interface is detaching.
1957  */
1958 static void
1959 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1960 {
1961 	struct bridge_softc *sc = ifp->if_bridge;
1962 	struct bridge_iflist *bif;
1963 
1964 	if (ifp->if_flags & IFF_RENAMING)
1965 		return;
1966 	if (V_bridge_cloner == NULL) {
1967 		/*
1968 		 * This detach handler can be called after
1969 		 * vnet_bridge_uninit().  Just return in that case.
1970 		 */
1971 		return;
1972 	}
1973 	/* Check if the interface is a bridge member */
1974 	if (sc != NULL) {
1975 		BRIDGE_LOCK(sc);
1976 
1977 		bif = bridge_lookup_member_if(sc, ifp);
1978 		if (bif != NULL)
1979 			bridge_delete_member(sc, bif, 1);
1980 
1981 		BRIDGE_UNLOCK(sc);
1982 		return;
1983 	}
1984 
1985 	/* Check if the interface is a span port */
1986 	BRIDGE_LIST_LOCK();
1987 	LIST_FOREACH(sc, &V_bridge_list, sc_list) {
1988 		BRIDGE_LOCK(sc);
1989 		CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1990 			if (ifp == bif->bif_ifp) {
1991 				bridge_delete_span(sc, bif);
1992 				break;
1993 			}
1994 
1995 		BRIDGE_UNLOCK(sc);
1996 	}
1997 	BRIDGE_LIST_UNLOCK();
1998 }
1999 
2000 /*
2001  * bridge_init:
2002  *
2003  *	Initialize a bridge interface.
2004  */
2005 static void
2006 bridge_init(void *xsc)
2007 {
2008 	struct bridge_softc *sc = (struct bridge_softc *)xsc;
2009 	struct ifnet *ifp = sc->sc_ifp;
2010 
2011 	if (ifp->if_drv_flags & IFF_DRV_RUNNING)
2012 		return;
2013 
2014 	BRIDGE_LOCK(sc);
2015 	callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
2016 	    bridge_timer, sc);
2017 
2018 	ifp->if_drv_flags |= IFF_DRV_RUNNING;
2019 	bstp_init(&sc->sc_stp);		/* Initialize Spanning Tree */
2020 
2021 	BRIDGE_UNLOCK(sc);
2022 }
2023 
2024 /*
2025  * bridge_stop:
2026  *
2027  *	Stop the bridge interface.
2028  */
2029 static void
2030 bridge_stop(struct ifnet *ifp, int disable)
2031 {
2032 	struct bridge_softc *sc = ifp->if_softc;
2033 
2034 	BRIDGE_LOCK_ASSERT(sc);
2035 
2036 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2037 		return;
2038 
2039 	BRIDGE_RT_LOCK(sc);
2040 	callout_stop(&sc->sc_brcallout);
2041 
2042 	bstp_stop(&sc->sc_stp);
2043 
2044 	bridge_rtflush(sc, IFBF_FLUSHDYN);
2045 	BRIDGE_RT_UNLOCK(sc);
2046 
2047 	ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
2048 }
2049 
2050 /*
2051  * bridge_enqueue:
2052  *
2053  *	Enqueue a packet on a bridge member interface.
2054  *
2055  */
2056 static int
2057 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
2058 {
2059 	int len, err = 0;
2060 	short mflags;
2061 	struct mbuf *m0;
2062 
2063 	/* We may be sending a fragment so traverse the mbuf */
2064 	for (; m; m = m0) {
2065 		m0 = m->m_nextpkt;
2066 		m->m_nextpkt = NULL;
2067 		len = m->m_pkthdr.len;
2068 		mflags = m->m_flags;
2069 
2070 		/*
2071 		 * If underlying interface can not do VLAN tag insertion itself
2072 		 * then attach a packet tag that holds it.
2073 		 */
2074 		if ((m->m_flags & M_VLANTAG) &&
2075 		    (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
2076 			m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
2077 			if (m == NULL) {
2078 				if_printf(dst_ifp,
2079 				    "unable to prepend VLAN header\n");
2080 				if_inc_counter(dst_ifp, IFCOUNTER_OERRORS, 1);
2081 				continue;
2082 			}
2083 			m->m_flags &= ~M_VLANTAG;
2084 		}
2085 
2086 		M_ASSERTPKTHDR(m); /* We shouldn't transmit mbuf without pkthdr */
2087 		if ((err = dst_ifp->if_transmit(dst_ifp, m))) {
2088 			int n;
2089 
2090 			for (m = m0, n = 1; m != NULL; m = m0, n++) {
2091 				m0 = m->m_nextpkt;
2092 				m_freem(m);
2093 			}
2094 			if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, n);
2095 			break;
2096 		}
2097 
2098 		if_inc_counter(sc->sc_ifp, IFCOUNTER_OPACKETS, 1);
2099 		if_inc_counter(sc->sc_ifp, IFCOUNTER_OBYTES, len);
2100 		if (mflags & M_MCAST)
2101 			if_inc_counter(sc->sc_ifp, IFCOUNTER_OMCASTS, 1);
2102 	}
2103 
2104 	return (err);
2105 }
2106 
2107 /*
2108  * bridge_dummynet:
2109  *
2110  * 	Receive a queued packet from dummynet and pass it on to the output
2111  * 	interface.
2112  *
2113  *	The mbuf has the Ethernet header already attached.
2114  */
2115 static void
2116 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
2117 {
2118 	struct bridge_softc *sc;
2119 
2120 	sc = ifp->if_bridge;
2121 
2122 	/*
2123 	 * The packet didnt originate from a member interface. This should only
2124 	 * ever happen if a member interface is removed while packets are
2125 	 * queued for it.
2126 	 */
2127 	if (sc == NULL) {
2128 		m_freem(m);
2129 		return;
2130 	}
2131 
2132 	if (PFIL_HOOKED_OUT(V_inet_pfil_head)
2133 #ifdef INET6
2134 	    || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2135 #endif
2136 	    ) {
2137 		if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
2138 			return;
2139 		if (m == NULL)
2140 			return;
2141 	}
2142 
2143 	bridge_enqueue(sc, ifp, m);
2144 }
2145 
2146 /*
2147  * bridge_output:
2148  *
2149  *	Send output from a bridge member interface.  This
2150  *	performs the bridging function for locally originated
2151  *	packets.
2152  *
2153  *	The mbuf has the Ethernet header already attached.  We must
2154  *	enqueue or free the mbuf before returning.
2155  */
2156 static int
2157 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
2158     struct rtentry *rt)
2159 {
2160 	struct ether_header *eh;
2161 	struct ifnet *bifp, *dst_if;
2162 	struct bridge_softc *sc;
2163 	uint16_t vlan;
2164 
2165 	NET_EPOCH_ASSERT();
2166 
2167 	if (m->m_len < ETHER_HDR_LEN) {
2168 		m = m_pullup(m, ETHER_HDR_LEN);
2169 		if (m == NULL)
2170 			return (0);
2171 	}
2172 
2173 	eh = mtod(m, struct ether_header *);
2174 	sc = ifp->if_bridge;
2175 	vlan = VLANTAGOF(m);
2176 
2177 	bifp = sc->sc_ifp;
2178 
2179 	/*
2180 	 * If bridge is down, but the original output interface is up,
2181 	 * go ahead and send out that interface.  Otherwise, the packet
2182 	 * is dropped below.
2183 	 */
2184 	if ((bifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2185 		dst_if = ifp;
2186 		goto sendunicast;
2187 	}
2188 
2189 	/*
2190 	 * If the packet is a multicast, or we don't know a better way to
2191 	 * get there, send to all interfaces.
2192 	 */
2193 	if (ETHER_IS_MULTICAST(eh->ether_dhost))
2194 		dst_if = NULL;
2195 	else
2196 		dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
2197 	/* Tap any traffic not passing back out the originating interface */
2198 	if (dst_if != ifp)
2199 		ETHER_BPF_MTAP(bifp, m);
2200 	if (dst_if == NULL) {
2201 		struct bridge_iflist *bif;
2202 		struct mbuf *mc;
2203 		int used = 0;
2204 
2205 		bridge_span(sc, m);
2206 
2207 		CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
2208 			dst_if = bif->bif_ifp;
2209 
2210 			if (dst_if->if_type == IFT_GIF)
2211 				continue;
2212 			if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2213 				continue;
2214 
2215 			/*
2216 			 * If this is not the original output interface,
2217 			 * and the interface is participating in spanning
2218 			 * tree, make sure the port is in a state that
2219 			 * allows forwarding.
2220 			 */
2221 			if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
2222 			    bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2223 				continue;
2224 
2225 			if (CK_LIST_NEXT(bif, bif_next) == NULL) {
2226 				used = 1;
2227 				mc = m;
2228 			} else {
2229 				mc = m_dup(m, M_NOWAIT);
2230 				if (mc == NULL) {
2231 					if_inc_counter(bifp, IFCOUNTER_OERRORS, 1);
2232 					continue;
2233 				}
2234 			}
2235 
2236 			bridge_enqueue(sc, dst_if, mc);
2237 		}
2238 		if (used == 0)
2239 			m_freem(m);
2240 		return (0);
2241 	}
2242 
2243 sendunicast:
2244 	/*
2245 	 * XXX Spanning tree consideration here?
2246 	 */
2247 
2248 	bridge_span(sc, m);
2249 	if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2250 		m_freem(m);
2251 		return (0);
2252 	}
2253 
2254 	bridge_enqueue(sc, dst_if, m);
2255 	return (0);
2256 }
2257 
2258 /*
2259  * bridge_transmit:
2260  *
2261  *	Do output on a bridge.
2262  *
2263  */
2264 static int
2265 bridge_transmit(struct ifnet *ifp, struct mbuf *m)
2266 {
2267 	struct bridge_softc *sc;
2268 	struct ether_header *eh;
2269 	struct ifnet *dst_if;
2270 	int error = 0;
2271 
2272 	sc = ifp->if_softc;
2273 
2274 	ETHER_BPF_MTAP(ifp, m);
2275 
2276 	eh = mtod(m, struct ether_header *);
2277 
2278 	if (((m->m_flags & (M_BCAST|M_MCAST)) == 0) &&
2279 	    (dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1)) != NULL) {
2280 		error = bridge_enqueue(sc, dst_if, m);
2281 	} else
2282 		bridge_broadcast(sc, ifp, m, 0);
2283 
2284 	return (error);
2285 }
2286 
2287 #ifdef ALTQ
2288 static void
2289 bridge_altq_start(if_t ifp)
2290 {
2291 	struct ifaltq *ifq = &ifp->if_snd;
2292 	struct mbuf *m;
2293 
2294 	IFQ_LOCK(ifq);
2295 	IFQ_DEQUEUE_NOLOCK(ifq, m);
2296 	while (m != NULL) {
2297 		bridge_transmit(ifp, m);
2298 		IFQ_DEQUEUE_NOLOCK(ifq, m);
2299 	}
2300 	IFQ_UNLOCK(ifq);
2301 }
2302 
2303 static int
2304 bridge_altq_transmit(if_t ifp, struct mbuf *m)
2305 {
2306 	int err;
2307 
2308 	if (ALTQ_IS_ENABLED(&ifp->if_snd)) {
2309 		IFQ_ENQUEUE(&ifp->if_snd, m, err);
2310 		if (err == 0)
2311 			bridge_altq_start(ifp);
2312 	} else
2313 		err = bridge_transmit(ifp, m);
2314 
2315 	return (err);
2316 }
2317 #endif	/* ALTQ */
2318 
2319 /*
2320  * The ifp->if_qflush entry point for if_bridge(4) is no-op.
2321  */
2322 static void
2323 bridge_qflush(struct ifnet *ifp __unused)
2324 {
2325 }
2326 
2327 /*
2328  * bridge_forward:
2329  *
2330  *	The forwarding function of the bridge.
2331  *
2332  *	NOTE: Releases the lock on return.
2333  */
2334 static void
2335 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2336     struct mbuf *m)
2337 {
2338 	struct bridge_iflist *dbif;
2339 	struct ifnet *src_if, *dst_if, *ifp;
2340 	struct ether_header *eh;
2341 	uint16_t vlan;
2342 	uint8_t *dst;
2343 	int error;
2344 
2345 	NET_EPOCH_ASSERT();
2346 
2347 	src_if = m->m_pkthdr.rcvif;
2348 	ifp = sc->sc_ifp;
2349 
2350 	if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
2351 	if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2352 	vlan = VLANTAGOF(m);
2353 
2354 	if ((sbif->bif_flags & IFBIF_STP) &&
2355 	    sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2356 		goto drop;
2357 
2358 	eh = mtod(m, struct ether_header *);
2359 	dst = eh->ether_dhost;
2360 
2361 	/* If the interface is learning, record the address. */
2362 	if (sbif->bif_flags & IFBIF_LEARNING) {
2363 		error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2364 		    sbif, 0, IFBAF_DYNAMIC);
2365 		/*
2366 		 * If the interface has addresses limits then deny any source
2367 		 * that is not in the cache.
2368 		 */
2369 		if (error && sbif->bif_addrmax)
2370 			goto drop;
2371 	}
2372 
2373 	if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2374 	    sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2375 		goto drop;
2376 
2377 #ifdef DEV_NETMAP
2378 	/*
2379 	 * Hand the packet to netmap only if it wasn't injected by netmap
2380 	 * itself.
2381 	 */
2382 	if ((m->m_flags & M_BRIDGE_INJECT) == 0 &&
2383 	    (if_getcapenable(ifp) & IFCAP_NETMAP) != 0) {
2384 		ifp->if_input(ifp, m);
2385 		return;
2386 	}
2387 	m->m_flags &= ~M_BRIDGE_INJECT;
2388 #endif
2389 
2390 	/*
2391 	 * At this point, the port either doesn't participate
2392 	 * in spanning tree or it is in the forwarding state.
2393 	 */
2394 
2395 	/*
2396 	 * If the packet is unicast, destined for someone on
2397 	 * "this" side of the bridge, drop it.
2398 	 */
2399 	if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2400 		dst_if = bridge_rtlookup(sc, dst, vlan);
2401 		if (src_if == dst_if)
2402 			goto drop;
2403 	} else {
2404 		/*
2405 		 * Check if its a reserved multicast address, any address
2406 		 * listed in 802.1D section 7.12.6 may not be forwarded by the
2407 		 * bridge.
2408 		 * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2409 		 */
2410 		if (dst[0] == 0x01 && dst[1] == 0x80 &&
2411 		    dst[2] == 0xc2 && dst[3] == 0x00 &&
2412 		    dst[4] == 0x00 && dst[5] <= 0x0f)
2413 			goto drop;
2414 
2415 		/* ...forward it to all interfaces. */
2416 		if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
2417 		dst_if = NULL;
2418 	}
2419 
2420 	/*
2421 	 * If we have a destination interface which is a member of our bridge,
2422 	 * OR this is a unicast packet, push it through the bpf(4) machinery.
2423 	 * For broadcast or multicast packets, don't bother because it will
2424 	 * be reinjected into ether_input. We do this before we pass the packets
2425 	 * through the pfil(9) framework, as it is possible that pfil(9) will
2426 	 * drop the packet, or possibly modify it, making it difficult to debug
2427 	 * firewall issues on the bridge.
2428 	 */
2429 	if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2430 		ETHER_BPF_MTAP(ifp, m);
2431 
2432 	/* run the packet filter */
2433 	if (PFIL_HOOKED_IN(V_inet_pfil_head)
2434 #ifdef INET6
2435 	    || PFIL_HOOKED_IN(V_inet6_pfil_head)
2436 #endif
2437 	    ) {
2438 		if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2439 			return;
2440 		if (m == NULL)
2441 			return;
2442 	}
2443 
2444 	if (dst_if == NULL) {
2445 		bridge_broadcast(sc, src_if, m, 1);
2446 		return;
2447 	}
2448 
2449 	/*
2450 	 * At this point, we're dealing with a unicast frame
2451 	 * going to a different interface.
2452 	 */
2453 	if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2454 		goto drop;
2455 
2456 	dbif = bridge_lookup_member_if(sc, dst_if);
2457 	if (dbif == NULL)
2458 		/* Not a member of the bridge (anymore?) */
2459 		goto drop;
2460 
2461 	/* Private segments can not talk to each other */
2462 	if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2463 		goto drop;
2464 
2465 	if ((dbif->bif_flags & IFBIF_STP) &&
2466 	    dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2467 		goto drop;
2468 
2469 	if (PFIL_HOOKED_OUT(V_inet_pfil_head)
2470 #ifdef INET6
2471 	    || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2472 #endif
2473 	    ) {
2474 		if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2475 			return;
2476 		if (m == NULL)
2477 			return;
2478 	}
2479 
2480 	bridge_enqueue(sc, dst_if, m);
2481 	return;
2482 
2483 drop:
2484 	m_freem(m);
2485 }
2486 
2487 /*
2488  * bridge_input:
2489  *
2490  *	Receive input from a member interface.  Queue the packet for
2491  *	bridging if it is not for us.
2492  */
2493 static struct mbuf *
2494 bridge_input(struct ifnet *ifp, struct mbuf *m)
2495 {
2496 	struct bridge_softc *sc;
2497 	struct bridge_iflist *bif, *bif2;
2498 	struct ifnet *bifp;
2499 	struct ether_header *eh;
2500 	struct mbuf *mc, *mc2;
2501 	uint16_t vlan;
2502 	int error;
2503 
2504 	NET_EPOCH_ASSERT();
2505 
2506 	eh = mtod(m, struct ether_header *);
2507 	vlan = VLANTAGOF(m);
2508 
2509 	sc = ifp->if_bridge;
2510 	if (sc == NULL) {
2511 		/*
2512 		 * This packet originated from the bridge itself, so it must
2513 		 * have been transmitted by netmap.  Derive the "source"
2514 		 * interface from the source address and drop the packet if the
2515 		 * source address isn't known.
2516 		 */
2517 		KASSERT((m->m_flags & M_BRIDGE_INJECT) != 0,
2518 		    ("%s: ifnet %p missing a bridge softc", __func__, ifp));
2519 		sc = if_getsoftc(ifp);
2520 		ifp = bridge_rtlookup(sc, eh->ether_shost, vlan);
2521 		if (ifp == NULL) {
2522 			if_inc_counter(sc->sc_ifp, IFCOUNTER_IERRORS, 1);
2523 			m_freem(m);
2524 			return (NULL);
2525 		}
2526 		m->m_pkthdr.rcvif = ifp;
2527 	}
2528 	bifp = sc->sc_ifp;
2529 	if ((bifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2530 		return (m);
2531 
2532 	/*
2533 	 * Implement support for bridge monitoring. If this flag has been
2534 	 * set on this interface, discard the packet once we push it through
2535 	 * the bpf(4) machinery, but before we do, increment the byte and
2536 	 * packet counters associated with this interface.
2537 	 */
2538 	if ((bifp->if_flags & IFF_MONITOR) != 0) {
2539 		m->m_pkthdr.rcvif  = bifp;
2540 		ETHER_BPF_MTAP(bifp, m);
2541 		if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);
2542 		if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2543 		m_freem(m);
2544 		return (NULL);
2545 	}
2546 	bif = bridge_lookup_member_if(sc, ifp);
2547 	if (bif == NULL) {
2548 		return (m);
2549 	}
2550 
2551 	bridge_span(sc, m);
2552 
2553 	if (m->m_flags & (M_BCAST|M_MCAST)) {
2554 		/* Tap off 802.1D packets; they do not get forwarded. */
2555 		if (memcmp(eh->ether_dhost, bstp_etheraddr,
2556 		    ETHER_ADDR_LEN) == 0) {
2557 			bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2558 			return (NULL);
2559 		}
2560 
2561 		if ((bif->bif_flags & IFBIF_STP) &&
2562 		    bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2563 			return (m);
2564 		}
2565 
2566 		/*
2567 		 * Make a deep copy of the packet and enqueue the copy
2568 		 * for bridge processing; return the original packet for
2569 		 * local processing.
2570 		 */
2571 		mc = m_dup(m, M_NOWAIT);
2572 		if (mc == NULL) {
2573 			return (m);
2574 		}
2575 
2576 		/* Perform the bridge forwarding function with the copy. */
2577 		bridge_forward(sc, bif, mc);
2578 
2579 #ifdef DEV_NETMAP
2580 		/*
2581 		 * If netmap is enabled and has not already seen this packet,
2582 		 * then it will be consumed by bridge_forward().
2583 		 */
2584 		if ((if_getcapenable(bifp) & IFCAP_NETMAP) != 0 &&
2585 		    (m->m_flags & M_BRIDGE_INJECT) == 0) {
2586 			m_freem(m);
2587 			return (NULL);
2588 		}
2589 #endif
2590 
2591 		/*
2592 		 * Reinject the mbuf as arriving on the bridge so we have a
2593 		 * chance at claiming multicast packets. We can not loop back
2594 		 * here from ether_input as a bridge is never a member of a
2595 		 * bridge.
2596 		 */
2597 		KASSERT(bifp->if_bridge == NULL,
2598 		    ("loop created in bridge_input"));
2599 		mc2 = m_dup(m, M_NOWAIT);
2600 		if (mc2 != NULL) {
2601 			/* Keep the layer3 header aligned */
2602 			int i = min(mc2->m_pkthdr.len, max_protohdr);
2603 			mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2604 		}
2605 		if (mc2 != NULL) {
2606 			mc2->m_pkthdr.rcvif = bifp;
2607 			mc2->m_flags &= ~M_BRIDGE_INJECT;
2608 			sc->sc_if_input(bifp, mc2);
2609 		}
2610 
2611 		/* Return the original packet for local processing. */
2612 		return (m);
2613 	}
2614 
2615 	if ((bif->bif_flags & IFBIF_STP) &&
2616 	    bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2617 		return (m);
2618 	}
2619 
2620 #if defined(INET) || defined(INET6)
2621 #define	CARP_CHECK_WE_ARE_DST(iface) \
2622 	((iface)->if_carp && (*carp_forus_p)((iface), eh->ether_dhost))
2623 #define	CARP_CHECK_WE_ARE_SRC(iface) \
2624 	((iface)->if_carp && (*carp_forus_p)((iface), eh->ether_shost))
2625 #else
2626 #define	CARP_CHECK_WE_ARE_DST(iface)	false
2627 #define	CARP_CHECK_WE_ARE_SRC(iface)	false
2628 #endif
2629 
2630 #ifdef INET6
2631 #define	PFIL_HOOKED_INET6	PFIL_HOOKED_IN(V_inet6_pfil_head)
2632 #else
2633 #define	PFIL_HOOKED_INET6	false
2634 #endif
2635 
2636 #ifdef DEV_NETMAP
2637 #define	GRAB_FOR_NETMAP(ifp, m) do {					\
2638 	if ((if_getcapenable(ifp) & IFCAP_NETMAP) != 0 &&		\
2639 	    ((m)->m_flags & M_BRIDGE_INJECT) == 0) {			\
2640 		(ifp)->if_input(ifp, m);				\
2641 		return (NULL);						\
2642 	}								\
2643 } while (0)
2644 #else
2645 #define	GRAB_FOR_NETMAP(ifp, m)
2646 #endif
2647 
2648 #define GRAB_OUR_PACKETS(iface)						\
2649 	if ((iface)->if_type == IFT_GIF)				\
2650 		continue;						\
2651 	/* It is destined for us. */					\
2652 	if (memcmp(IF_LLADDR(iface), eh->ether_dhost, ETHER_ADDR_LEN) == 0 || \
2653 	    CARP_CHECK_WE_ARE_DST(iface)) {				\
2654 		if (bif->bif_flags & IFBIF_LEARNING) {			\
2655 			error = bridge_rtupdate(sc, eh->ether_shost,	\
2656 			    vlan, bif, 0, IFBAF_DYNAMIC);		\
2657 			if (error && bif->bif_addrmax) {		\
2658 				m_freem(m);				\
2659 				return (NULL);				\
2660 			}						\
2661 		}							\
2662 		m->m_pkthdr.rcvif = iface;				\
2663 		if ((iface) == ifp) {					\
2664 			/* Skip bridge processing... src == dest */	\
2665 			return (m);					\
2666 		}							\
2667 		/* It's passing over or to the bridge, locally. */	\
2668 		ETHER_BPF_MTAP(bifp, m);				\
2669 		if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);		\
2670 		if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);\
2671 		/* Hand the packet over to netmap if necessary. */	\
2672 		GRAB_FOR_NETMAP(bifp, m);				\
2673 		/* Filter on the physical interface. */			\
2674 		if (V_pfil_local_phys && (PFIL_HOOKED_IN(V_inet_pfil_head) || \
2675 		    PFIL_HOOKED_INET6)) {				\
2676 			if (bridge_pfil(&m, NULL, ifp,			\
2677 			    PFIL_IN) != 0 || m == NULL) {		\
2678 				return (NULL);				\
2679 			}						\
2680 		}							\
2681 		if ((iface) != bifp)					\
2682 			ETHER_BPF_MTAP(iface, m);			\
2683 		return (m);						\
2684 	}								\
2685 									\
2686 	/* We just received a packet that we sent out. */		\
2687 	if (memcmp(IF_LLADDR(iface), eh->ether_shost, ETHER_ADDR_LEN) == 0 || \
2688 	    CARP_CHECK_WE_ARE_SRC(iface)) {				\
2689 		m_freem(m);						\
2690 		return (NULL);						\
2691 	}
2692 
2693 	/*
2694 	 * Unicast.  Make sure it's not for the bridge.
2695 	 */
2696 	do { GRAB_OUR_PACKETS(bifp) } while (0);
2697 
2698 	/*
2699 	 * Give a chance for ifp at first priority. This will help when	the
2700 	 * packet comes through the interface like VLAN's with the same MACs
2701 	 * on several interfaces from the same bridge. This also will save
2702 	 * some CPU cycles in case the destination interface and the input
2703 	 * interface (eq ifp) are the same.
2704 	 */
2705 	do { GRAB_OUR_PACKETS(ifp) } while (0);
2706 
2707 	/* Now check the all bridge members. */
2708 	CK_LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2709 		GRAB_OUR_PACKETS(bif2->bif_ifp)
2710 	}
2711 
2712 #undef CARP_CHECK_WE_ARE_DST
2713 #undef CARP_CHECK_WE_ARE_SRC
2714 #undef PFIL_HOOKED_INET6
2715 #undef GRAB_FOR_NETMAP
2716 #undef GRAB_OUR_PACKETS
2717 
2718 	/* Perform the bridge forwarding function. */
2719 	bridge_forward(sc, bif, m);
2720 
2721 	return (NULL);
2722 }
2723 
2724 /*
2725  * Inject a packet back into the host ethernet stack.  This will generally only
2726  * be used by netmap when an application writes to the host TX ring.  The
2727  * M_BRIDGE_INJECT flag ensures that the packet is re-routed to the bridge
2728  * interface after ethernet processing.
2729  */
2730 static void
2731 bridge_inject(struct ifnet *ifp, struct mbuf *m)
2732 {
2733 	struct bridge_softc *sc;
2734 
2735 	KASSERT((if_getcapenable(ifp) & IFCAP_NETMAP) != 0,
2736 	    ("%s: iface %s is not running in netmap mode",
2737 	    __func__, if_name(ifp)));
2738 	KASSERT((m->m_flags & M_BRIDGE_INJECT) == 0,
2739 	    ("%s: mbuf %p has M_BRIDGE_INJECT set", __func__, m));
2740 
2741 	m->m_flags |= M_BRIDGE_INJECT;
2742 	sc = if_getsoftc(ifp);
2743 	sc->sc_if_input(ifp, m);
2744 }
2745 
2746 /*
2747  * bridge_broadcast:
2748  *
2749  *	Send a frame to all interfaces that are members of
2750  *	the bridge, except for the one on which the packet
2751  *	arrived.
2752  *
2753  *	NOTE: Releases the lock on return.
2754  */
2755 static void
2756 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2757     struct mbuf *m, int runfilt)
2758 {
2759 	struct bridge_iflist *dbif, *sbif;
2760 	struct mbuf *mc;
2761 	struct ifnet *dst_if;
2762 	int used = 0, i;
2763 
2764 	NET_EPOCH_ASSERT();
2765 
2766 	sbif = bridge_lookup_member_if(sc, src_if);
2767 
2768 	/* Filter on the bridge interface before broadcasting */
2769 	if (runfilt && (PFIL_HOOKED_OUT(V_inet_pfil_head)
2770 #ifdef INET6
2771 	    || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2772 #endif
2773 	    )) {
2774 		if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2775 			return;
2776 		if (m == NULL)
2777 			return;
2778 	}
2779 
2780 	CK_LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2781 		dst_if = dbif->bif_ifp;
2782 		if (dst_if == src_if)
2783 			continue;
2784 
2785 		/* Private segments can not talk to each other */
2786 		if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2787 			continue;
2788 
2789 		if ((dbif->bif_flags & IFBIF_STP) &&
2790 		    dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2791 			continue;
2792 
2793 		if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2794 		    (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2795 			continue;
2796 
2797 		if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2798 			continue;
2799 
2800 		if (CK_LIST_NEXT(dbif, bif_next) == NULL) {
2801 			mc = m;
2802 			used = 1;
2803 		} else {
2804 			mc = m_dup(m, M_NOWAIT);
2805 			if (mc == NULL) {
2806 				if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2807 				continue;
2808 			}
2809 		}
2810 
2811 		/*
2812 		 * Filter on the output interface. Pass a NULL bridge interface
2813 		 * pointer so we do not redundantly filter on the bridge for
2814 		 * each interface we broadcast on.
2815 		 */
2816 		if (runfilt && (PFIL_HOOKED_OUT(V_inet_pfil_head)
2817 #ifdef INET6
2818 		    || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2819 #endif
2820 		    )) {
2821 			if (used == 0) {
2822 				/* Keep the layer3 header aligned */
2823 				i = min(mc->m_pkthdr.len, max_protohdr);
2824 				mc = m_copyup(mc, i, ETHER_ALIGN);
2825 				if (mc == NULL) {
2826 					if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2827 					continue;
2828 				}
2829 			}
2830 			if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2831 				continue;
2832 			if (mc == NULL)
2833 				continue;
2834 		}
2835 
2836 		bridge_enqueue(sc, dst_if, mc);
2837 	}
2838 	if (used == 0)
2839 		m_freem(m);
2840 }
2841 
2842 /*
2843  * bridge_span:
2844  *
2845  *	Duplicate a packet out one or more interfaces that are in span mode,
2846  *	the original mbuf is unmodified.
2847  */
2848 static void
2849 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2850 {
2851 	struct bridge_iflist *bif;
2852 	struct ifnet *dst_if;
2853 	struct mbuf *mc;
2854 
2855 	NET_EPOCH_ASSERT();
2856 
2857 	if (CK_LIST_EMPTY(&sc->sc_spanlist))
2858 		return;
2859 
2860 	CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2861 		dst_if = bif->bif_ifp;
2862 
2863 		if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2864 			continue;
2865 
2866 		mc = m_dup(m, M_NOWAIT);
2867 		if (mc == NULL) {
2868 			if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2869 			continue;
2870 		}
2871 
2872 		bridge_enqueue(sc, dst_if, mc);
2873 	}
2874 }
2875 
2876 /*
2877  * bridge_rtupdate:
2878  *
2879  *	Add a bridge routing entry.
2880  */
2881 static int
2882 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2883     struct bridge_iflist *bif, int setflags, uint8_t flags)
2884 {
2885 	struct bridge_rtnode *brt;
2886 	struct bridge_iflist *obif;
2887 	int error;
2888 
2889 	BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
2890 
2891 	/* Check the source address is valid and not multicast. */
2892 	if (ETHER_IS_MULTICAST(dst) ||
2893 	    (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2894 	     dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2895 		return (EINVAL);
2896 
2897 	/*
2898 	 * A route for this destination might already exist.  If so,
2899 	 * update it, otherwise create a new one.
2900 	 */
2901 	if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2902 		BRIDGE_RT_LOCK(sc);
2903 
2904 		/* Check again, now that we have the lock. There could have
2905 		 * been a race and we only want to insert this once. */
2906 		if (bridge_rtnode_lookup(sc, dst, vlan) != NULL) {
2907 			BRIDGE_RT_UNLOCK(sc);
2908 			return (0);
2909 		}
2910 
2911 		if (sc->sc_brtcnt >= sc->sc_brtmax) {
2912 			sc->sc_brtexceeded++;
2913 			BRIDGE_RT_UNLOCK(sc);
2914 			return (ENOSPC);
2915 		}
2916 		/* Check per interface address limits (if enabled) */
2917 		if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2918 			bif->bif_addrexceeded++;
2919 			BRIDGE_RT_UNLOCK(sc);
2920 			return (ENOSPC);
2921 		}
2922 
2923 		/*
2924 		 * Allocate a new bridge forwarding node, and
2925 		 * initialize the expiration time and Ethernet
2926 		 * address.
2927 		 */
2928 		brt = uma_zalloc(V_bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2929 		if (brt == NULL) {
2930 			BRIDGE_RT_UNLOCK(sc);
2931 			return (ENOMEM);
2932 		}
2933 		brt->brt_vnet = curvnet;
2934 
2935 		if (bif->bif_flags & IFBIF_STICKY)
2936 			brt->brt_flags = IFBAF_STICKY;
2937 		else
2938 			brt->brt_flags = IFBAF_DYNAMIC;
2939 
2940 		memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2941 		brt->brt_vlan = vlan;
2942 
2943 		if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2944 			uma_zfree(V_bridge_rtnode_zone, brt);
2945 			BRIDGE_RT_UNLOCK(sc);
2946 			return (error);
2947 		}
2948 		brt->brt_dst = bif;
2949 		bif->bif_addrcnt++;
2950 
2951 		BRIDGE_RT_UNLOCK(sc);
2952 	}
2953 
2954 	if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2955 	    (obif = brt->brt_dst) != bif) {
2956 		BRIDGE_RT_LOCK(sc);
2957 		brt->brt_dst->bif_addrcnt--;
2958 		brt->brt_dst = bif;
2959 		brt->brt_dst->bif_addrcnt++;
2960 		BRIDGE_RT_UNLOCK(sc);
2961 
2962 		if (V_log_mac_flap &&
2963 		    ppsratecheck(&V_log_last, &V_log_count, V_log_interval)) {
2964 			log(LOG_NOTICE,
2965 			    "%s: mac address %6D vlan %d moved from %s to %s\n",
2966 			    sc->sc_ifp->if_xname,
2967 			    &brt->brt_addr[0], ":",
2968 			    brt->brt_vlan,
2969 			    obif->bif_ifp->if_xname,
2970 			    bif->bif_ifp->if_xname);
2971 		}
2972 	}
2973 
2974 	if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2975 		brt->brt_expire = time_uptime + sc->sc_brttimeout;
2976 	if (setflags)
2977 		brt->brt_flags = flags;
2978 
2979 	return (0);
2980 }
2981 
2982 /*
2983  * bridge_rtlookup:
2984  *
2985  *	Lookup the destination interface for an address.
2986  */
2987 static struct ifnet *
2988 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2989 {
2990 	struct bridge_rtnode *brt;
2991 
2992 	NET_EPOCH_ASSERT();
2993 
2994 	if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2995 		return (NULL);
2996 
2997 	return (brt->brt_ifp);
2998 }
2999 
3000 /*
3001  * bridge_rttrim:
3002  *
3003  *	Trim the routine table so that we have a number
3004  *	of routing entries less than or equal to the
3005  *	maximum number.
3006  */
3007 static void
3008 bridge_rttrim(struct bridge_softc *sc)
3009 {
3010 	struct bridge_rtnode *brt, *nbrt;
3011 
3012 	NET_EPOCH_ASSERT();
3013 	BRIDGE_RT_LOCK_ASSERT(sc);
3014 
3015 	/* Make sure we actually need to do this. */
3016 	if (sc->sc_brtcnt <= sc->sc_brtmax)
3017 		return;
3018 
3019 	/* Force an aging cycle; this might trim enough addresses. */
3020 	bridge_rtage(sc);
3021 	if (sc->sc_brtcnt <= sc->sc_brtmax)
3022 		return;
3023 
3024 	CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3025 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
3026 			bridge_rtnode_destroy(sc, brt);
3027 			if (sc->sc_brtcnt <= sc->sc_brtmax)
3028 				return;
3029 		}
3030 	}
3031 }
3032 
3033 /*
3034  * bridge_timer:
3035  *
3036  *	Aging timer for the bridge.
3037  */
3038 static void
3039 bridge_timer(void *arg)
3040 {
3041 	struct bridge_softc *sc = arg;
3042 
3043 	BRIDGE_RT_LOCK_ASSERT(sc);
3044 
3045 	/* Destruction of rtnodes requires a proper vnet context */
3046 	CURVNET_SET(sc->sc_ifp->if_vnet);
3047 	bridge_rtage(sc);
3048 
3049 	if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
3050 		callout_reset(&sc->sc_brcallout,
3051 		    bridge_rtable_prune_period * hz, bridge_timer, sc);
3052 	CURVNET_RESTORE();
3053 }
3054 
3055 /*
3056  * bridge_rtage:
3057  *
3058  *	Perform an aging cycle.
3059  */
3060 static void
3061 bridge_rtage(struct bridge_softc *sc)
3062 {
3063 	struct bridge_rtnode *brt, *nbrt;
3064 
3065 	BRIDGE_RT_LOCK_ASSERT(sc);
3066 
3067 	CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3068 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
3069 			if (time_uptime >= brt->brt_expire)
3070 				bridge_rtnode_destroy(sc, brt);
3071 		}
3072 	}
3073 }
3074 
3075 /*
3076  * bridge_rtflush:
3077  *
3078  *	Remove all dynamic addresses from the bridge.
3079  */
3080 static void
3081 bridge_rtflush(struct bridge_softc *sc, int full)
3082 {
3083 	struct bridge_rtnode *brt, *nbrt;
3084 
3085 	BRIDGE_RT_LOCK_ASSERT(sc);
3086 
3087 	CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3088 		if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3089 			bridge_rtnode_destroy(sc, brt);
3090 	}
3091 }
3092 
3093 /*
3094  * bridge_rtdaddr:
3095  *
3096  *	Remove an address from the table.
3097  */
3098 static int
3099 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
3100 {
3101 	struct bridge_rtnode *brt;
3102 	int found = 0;
3103 
3104 	BRIDGE_RT_LOCK(sc);
3105 
3106 	/*
3107 	 * If vlan is DOT1Q_VID_RSVD_IMPL then we want to delete for all vlans
3108 	 * so the lookup may return more than one.
3109 	 */
3110 	while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
3111 		bridge_rtnode_destroy(sc, brt);
3112 		found = 1;
3113 	}
3114 
3115 	BRIDGE_RT_UNLOCK(sc);
3116 
3117 	return (found ? 0 : ENOENT);
3118 }
3119 
3120 /*
3121  * bridge_rtdelete:
3122  *
3123  *	Delete routes to a speicifc member interface.
3124  */
3125 static void
3126 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
3127 {
3128 	struct bridge_rtnode *brt, *nbrt;
3129 
3130 	BRIDGE_RT_LOCK_ASSERT(sc);
3131 
3132 	CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3133 		if (brt->brt_ifp == ifp && (full ||
3134 			    (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
3135 			bridge_rtnode_destroy(sc, brt);
3136 	}
3137 }
3138 
3139 /*
3140  * bridge_rtable_init:
3141  *
3142  *	Initialize the route table for this bridge.
3143  */
3144 static void
3145 bridge_rtable_init(struct bridge_softc *sc)
3146 {
3147 	int i;
3148 
3149 	sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
3150 	    M_DEVBUF, M_WAITOK);
3151 
3152 	for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
3153 		CK_LIST_INIT(&sc->sc_rthash[i]);
3154 
3155 	sc->sc_rthash_key = arc4random();
3156 	CK_LIST_INIT(&sc->sc_rtlist);
3157 }
3158 
3159 /*
3160  * bridge_rtable_fini:
3161  *
3162  *	Deconstruct the route table for this bridge.
3163  */
3164 static void
3165 bridge_rtable_fini(struct bridge_softc *sc)
3166 {
3167 
3168 	KASSERT(sc->sc_brtcnt == 0,
3169 	    ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
3170 	free(sc->sc_rthash, M_DEVBUF);
3171 }
3172 
3173 /*
3174  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
3175  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
3176  */
3177 #define	mix(a, b, c)							\
3178 do {									\
3179 	a -= b; a -= c; a ^= (c >> 13);					\
3180 	b -= c; b -= a; b ^= (a << 8);					\
3181 	c -= a; c -= b; c ^= (b >> 13);					\
3182 	a -= b; a -= c; a ^= (c >> 12);					\
3183 	b -= c; b -= a; b ^= (a << 16);					\
3184 	c -= a; c -= b; c ^= (b >> 5);					\
3185 	a -= b; a -= c; a ^= (c >> 3);					\
3186 	b -= c; b -= a; b ^= (a << 10);					\
3187 	c -= a; c -= b; c ^= (b >> 15);					\
3188 } while (/*CONSTCOND*/0)
3189 
3190 static __inline uint32_t
3191 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
3192 {
3193 	uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
3194 
3195 	b += addr[5] << 8;
3196 	b += addr[4];
3197 	a += addr[3] << 24;
3198 	a += addr[2] << 16;
3199 	a += addr[1] << 8;
3200 	a += addr[0];
3201 
3202 	mix(a, b, c);
3203 
3204 	return (c & BRIDGE_RTHASH_MASK);
3205 }
3206 
3207 #undef mix
3208 
3209 static int
3210 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
3211 {
3212 	int i, d;
3213 
3214 	for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
3215 		d = ((int)a[i]) - ((int)b[i]);
3216 	}
3217 
3218 	return (d);
3219 }
3220 
3221 /*
3222  * bridge_rtnode_lookup:
3223  *
3224  *	Look up a bridge route node for the specified destination. Compare the
3225  *	vlan id or if zero then just return the first match.
3226  */
3227 static struct bridge_rtnode *
3228 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
3229 {
3230 	struct bridge_rtnode *brt;
3231 	uint32_t hash;
3232 	int dir;
3233 
3234 	BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(sc);
3235 
3236 	hash = bridge_rthash(sc, addr);
3237 	CK_LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
3238 		dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
3239 		if (dir == 0 && (brt->brt_vlan == vlan || vlan == DOT1Q_VID_RSVD_IMPL))
3240 			return (brt);
3241 		if (dir > 0)
3242 			return (NULL);
3243 	}
3244 
3245 	return (NULL);
3246 }
3247 
3248 /*
3249  * bridge_rtnode_insert:
3250  *
3251  *	Insert the specified bridge node into the route table.  We
3252  *	assume the entry is not already in the table.
3253  */
3254 static int
3255 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
3256 {
3257 	struct bridge_rtnode *lbrt;
3258 	uint32_t hash;
3259 	int dir;
3260 
3261 	BRIDGE_RT_LOCK_ASSERT(sc);
3262 
3263 	hash = bridge_rthash(sc, brt->brt_addr);
3264 
3265 	lbrt = CK_LIST_FIRST(&sc->sc_rthash[hash]);
3266 	if (lbrt == NULL) {
3267 		CK_LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
3268 		goto out;
3269 	}
3270 
3271 	do {
3272 		dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
3273 		if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
3274 			return (EEXIST);
3275 		if (dir > 0) {
3276 			CK_LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
3277 			goto out;
3278 		}
3279 		if (CK_LIST_NEXT(lbrt, brt_hash) == NULL) {
3280 			CK_LIST_INSERT_AFTER(lbrt, brt, brt_hash);
3281 			goto out;
3282 		}
3283 		lbrt = CK_LIST_NEXT(lbrt, brt_hash);
3284 	} while (lbrt != NULL);
3285 
3286 #ifdef DIAGNOSTIC
3287 	panic("bridge_rtnode_insert: impossible");
3288 #endif
3289 
3290 out:
3291 	CK_LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
3292 	sc->sc_brtcnt++;
3293 
3294 	return (0);
3295 }
3296 
3297 static void
3298 bridge_rtnode_destroy_cb(struct epoch_context *ctx)
3299 {
3300 	struct bridge_rtnode *brt;
3301 
3302 	brt = __containerof(ctx, struct bridge_rtnode, brt_epoch_ctx);
3303 
3304 	CURVNET_SET(brt->brt_vnet);
3305 	uma_zfree(V_bridge_rtnode_zone, brt);
3306 	CURVNET_RESTORE();
3307 }
3308 
3309 /*
3310  * bridge_rtnode_destroy:
3311  *
3312  *	Destroy a bridge rtnode.
3313  */
3314 static void
3315 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3316 {
3317 	BRIDGE_RT_LOCK_ASSERT(sc);
3318 
3319 	CK_LIST_REMOVE(brt, brt_hash);
3320 
3321 	CK_LIST_REMOVE(brt, brt_list);
3322 	sc->sc_brtcnt--;
3323 	brt->brt_dst->bif_addrcnt--;
3324 
3325 	NET_EPOCH_CALL(bridge_rtnode_destroy_cb, &brt->brt_epoch_ctx);
3326 }
3327 
3328 /*
3329  * bridge_rtable_expire:
3330  *
3331  *	Set the expiry time for all routes on an interface.
3332  */
3333 static void
3334 bridge_rtable_expire(struct ifnet *ifp, int age)
3335 {
3336 	struct bridge_softc *sc = ifp->if_bridge;
3337 	struct bridge_rtnode *brt;
3338 
3339 	CURVNET_SET(ifp->if_vnet);
3340 	BRIDGE_RT_LOCK(sc);
3341 
3342 	/*
3343 	 * If the age is zero then flush, otherwise set all the expiry times to
3344 	 * age for the interface
3345 	 */
3346 	if (age == 0)
3347 		bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
3348 	else {
3349 		CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
3350 			/* Cap the expiry time to 'age' */
3351 			if (brt->brt_ifp == ifp &&
3352 			    brt->brt_expire > time_uptime + age &&
3353 			    (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3354 				brt->brt_expire = time_uptime + age;
3355 		}
3356 	}
3357 	BRIDGE_RT_UNLOCK(sc);
3358 	CURVNET_RESTORE();
3359 }
3360 
3361 /*
3362  * bridge_state_change:
3363  *
3364  *	Callback from the bridgestp code when a port changes states.
3365  */
3366 static void
3367 bridge_state_change(struct ifnet *ifp, int state)
3368 {
3369 	struct bridge_softc *sc = ifp->if_bridge;
3370 	static const char *stpstates[] = {
3371 		"disabled",
3372 		"listening",
3373 		"learning",
3374 		"forwarding",
3375 		"blocking",
3376 		"discarding"
3377 	};
3378 
3379 	CURVNET_SET(ifp->if_vnet);
3380 	if (V_log_stp)
3381 		log(LOG_NOTICE, "%s: state changed to %s on %s\n",
3382 		    sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
3383 	CURVNET_RESTORE();
3384 }
3385 
3386 /*
3387  * Send bridge packets through pfil if they are one of the types pfil can deal
3388  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3389  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3390  * that interface.
3391  */
3392 static int
3393 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3394 {
3395 	int snap, error, i, hlen;
3396 	struct ether_header *eh1, eh2;
3397 	struct ip *ip;
3398 	struct llc llc1;
3399 	u_int16_t ether_type;
3400 	pfil_return_t rv;
3401 
3402 	snap = 0;
3403 	error = -1;	/* Default error if not error == 0 */
3404 
3405 #if 0
3406 	/* we may return with the IP fields swapped, ensure its not shared */
3407 	KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
3408 #endif
3409 
3410 	if (V_pfil_bridge == 0 && V_pfil_member == 0 && V_pfil_ipfw == 0)
3411 		return (0); /* filtering is disabled */
3412 
3413 	i = min((*mp)->m_pkthdr.len, max_protohdr);
3414 	if ((*mp)->m_len < i) {
3415 	    *mp = m_pullup(*mp, i);
3416 	    if (*mp == NULL) {
3417 		printf("%s: m_pullup failed\n", __func__);
3418 		return (-1);
3419 	    }
3420 	}
3421 
3422 	eh1 = mtod(*mp, struct ether_header *);
3423 	ether_type = ntohs(eh1->ether_type);
3424 
3425 	/*
3426 	 * Check for SNAP/LLC.
3427 	 */
3428 	if (ether_type < ETHERMTU) {
3429 		struct llc *llc2 = (struct llc *)(eh1 + 1);
3430 
3431 		if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3432 		    llc2->llc_dsap == LLC_SNAP_LSAP &&
3433 		    llc2->llc_ssap == LLC_SNAP_LSAP &&
3434 		    llc2->llc_control == LLC_UI) {
3435 			ether_type = htons(llc2->llc_un.type_snap.ether_type);
3436 			snap = 1;
3437 		}
3438 	}
3439 
3440 	/*
3441 	 * If we're trying to filter bridge traffic, don't look at anything
3442 	 * other than IP and ARP traffic.  If the filter doesn't understand
3443 	 * IPv6, don't allow IPv6 through the bridge either.  This is lame
3444 	 * since if we really wanted, say, an AppleTalk filter, we are hosed,
3445 	 * but of course we don't have an AppleTalk filter to begin with.
3446 	 * (Note that since pfil doesn't understand ARP it will pass *ALL*
3447 	 * ARP traffic.)
3448 	 */
3449 	switch (ether_type) {
3450 		case ETHERTYPE_ARP:
3451 		case ETHERTYPE_REVARP:
3452 			if (V_pfil_ipfw_arp == 0)
3453 				return (0); /* Automatically pass */
3454 			break;
3455 
3456 		case ETHERTYPE_IP:
3457 #ifdef INET6
3458 		case ETHERTYPE_IPV6:
3459 #endif /* INET6 */
3460 			break;
3461 		default:
3462 			/*
3463 			 * Check to see if the user wants to pass non-ip
3464 			 * packets, these will not be checked by pfil(9) and
3465 			 * passed unconditionally so the default is to drop.
3466 			 */
3467 			if (V_pfil_onlyip)
3468 				goto bad;
3469 	}
3470 
3471 	/* Run the packet through pfil before stripping link headers */
3472 	if (PFIL_HOOKED_OUT(V_link_pfil_head) && V_pfil_ipfw != 0 &&
3473 	    dir == PFIL_OUT && ifp != NULL) {
3474 		switch (pfil_mbuf_out(V_link_pfil_head, mp, ifp, NULL)) {
3475 		case PFIL_DROPPED:
3476 			return (EACCES);
3477 		case PFIL_CONSUMED:
3478 			return (0);
3479 		}
3480 	}
3481 
3482 	/* Strip off the Ethernet header and keep a copy. */
3483 	m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3484 	m_adj(*mp, ETHER_HDR_LEN);
3485 
3486 	/* Strip off snap header, if present */
3487 	if (snap) {
3488 		m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3489 		m_adj(*mp, sizeof(struct llc));
3490 	}
3491 
3492 	/*
3493 	 * Check the IP header for alignment and errors
3494 	 */
3495 	if (dir == PFIL_IN) {
3496 		switch (ether_type) {
3497 			case ETHERTYPE_IP:
3498 				error = bridge_ip_checkbasic(mp);
3499 				break;
3500 #ifdef INET6
3501 			case ETHERTYPE_IPV6:
3502 				error = bridge_ip6_checkbasic(mp);
3503 				break;
3504 #endif /* INET6 */
3505 			default:
3506 				error = 0;
3507 		}
3508 		if (error)
3509 			goto bad;
3510 	}
3511 
3512 	error = 0;
3513 
3514 	/*
3515 	 * Run the packet through pfil
3516 	 */
3517 	rv = PFIL_PASS;
3518 	switch (ether_type) {
3519 	case ETHERTYPE_IP:
3520 		/*
3521 		 * Run pfil on the member interface and the bridge, both can
3522 		 * be skipped by clearing pfil_member or pfil_bridge.
3523 		 *
3524 		 * Keep the order:
3525 		 *   in_if -> bridge_if -> out_if
3526 		 */
3527 		if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL && (rv =
3528 		    pfil_mbuf_out(V_inet_pfil_head, mp, bifp, NULL)) !=
3529 		    PFIL_PASS)
3530 			break;
3531 
3532 		if (V_pfil_member && ifp != NULL) {
3533 			rv = (dir == PFIL_OUT) ?
3534 			    pfil_mbuf_out(V_inet_pfil_head, mp, ifp, NULL) :
3535 			    pfil_mbuf_in(V_inet_pfil_head, mp, ifp, NULL);
3536 			if (rv != PFIL_PASS)
3537 				break;
3538 		}
3539 
3540 		if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv =
3541 		    pfil_mbuf_in(V_inet_pfil_head, mp, bifp, NULL)) !=
3542 		    PFIL_PASS)
3543 			break;
3544 
3545 		/* check if we need to fragment the packet */
3546 		/* bridge_fragment generates a mbuf chain of packets */
3547 		/* that already include eth headers */
3548 		if (V_pfil_member && ifp != NULL && dir == PFIL_OUT) {
3549 			i = (*mp)->m_pkthdr.len;
3550 			if (i > ifp->if_mtu) {
3551 				error = bridge_fragment(ifp, mp, &eh2, snap,
3552 					    &llc1);
3553 				return (error);
3554 			}
3555 		}
3556 
3557 		/* Recalculate the ip checksum. */
3558 		ip = mtod(*mp, struct ip *);
3559 		hlen = ip->ip_hl << 2;
3560 		if (hlen < sizeof(struct ip))
3561 			goto bad;
3562 		if (hlen > (*mp)->m_len) {
3563 			if ((*mp = m_pullup(*mp, hlen)) == NULL)
3564 				goto bad;
3565 			ip = mtod(*mp, struct ip *);
3566 			if (ip == NULL)
3567 				goto bad;
3568 		}
3569 		ip->ip_sum = 0;
3570 		if (hlen == sizeof(struct ip))
3571 			ip->ip_sum = in_cksum_hdr(ip);
3572 		else
3573 			ip->ip_sum = in_cksum(*mp, hlen);
3574 
3575 		break;
3576 #ifdef INET6
3577 	case ETHERTYPE_IPV6:
3578 		if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL && (rv =
3579 		    pfil_mbuf_out(V_inet6_pfil_head, mp, bifp, NULL)) !=
3580 		    PFIL_PASS)
3581 			break;
3582 
3583 		if (V_pfil_member && ifp != NULL) {
3584 			rv = (dir == PFIL_OUT) ?
3585 			    pfil_mbuf_out(V_inet6_pfil_head, mp, ifp, NULL) :
3586 			    pfil_mbuf_in(V_inet6_pfil_head, mp, ifp, NULL);
3587 			if (rv != PFIL_PASS)
3588 				break;
3589 		}
3590 
3591 		if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv =
3592 		    pfil_mbuf_in(V_inet6_pfil_head, mp, bifp, NULL)) !=
3593 		    PFIL_PASS)
3594 			break;
3595 		break;
3596 #endif
3597 	}
3598 
3599 	switch (rv) {
3600 	case PFIL_CONSUMED:
3601 		return (0);
3602 	case PFIL_DROPPED:
3603 		return (EACCES);
3604 	default:
3605 		break;
3606 	}
3607 
3608 	error = -1;
3609 
3610 	/*
3611 	 * Finally, put everything back the way it was and return
3612 	 */
3613 	if (snap) {
3614 		M_PREPEND(*mp, sizeof(struct llc), M_NOWAIT);
3615 		if (*mp == NULL)
3616 			return (error);
3617 		bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3618 	}
3619 
3620 	M_PREPEND(*mp, ETHER_HDR_LEN, M_NOWAIT);
3621 	if (*mp == NULL)
3622 		return (error);
3623 	bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3624 
3625 	return (0);
3626 
3627 bad:
3628 	m_freem(*mp);
3629 	*mp = NULL;
3630 	return (error);
3631 }
3632 
3633 /*
3634  * Perform basic checks on header size since
3635  * pfil assumes ip_input has already processed
3636  * it for it.  Cut-and-pasted from ip_input.c.
3637  * Given how simple the IPv6 version is,
3638  * does the IPv4 version really need to be
3639  * this complicated?
3640  *
3641  * XXX Should we update ipstat here, or not?
3642  * XXX Right now we update ipstat but not
3643  * XXX csum_counter.
3644  */
3645 static int
3646 bridge_ip_checkbasic(struct mbuf **mp)
3647 {
3648 	struct mbuf *m = *mp;
3649 	struct ip *ip;
3650 	int len, hlen;
3651 	u_short sum;
3652 
3653 	if (*mp == NULL)
3654 		return (-1);
3655 
3656 	if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3657 		if ((m = m_copyup(m, sizeof(struct ip),
3658 			(max_linkhdr + 3) & ~3)) == NULL) {
3659 			/* XXXJRT new stat, please */
3660 			KMOD_IPSTAT_INC(ips_toosmall);
3661 			goto bad;
3662 		}
3663 	} else if (__predict_false(m->m_len < sizeof (struct ip))) {
3664 		if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3665 			KMOD_IPSTAT_INC(ips_toosmall);
3666 			goto bad;
3667 		}
3668 	}
3669 	ip = mtod(m, struct ip *);
3670 	if (ip == NULL) goto bad;
3671 
3672 	if (ip->ip_v != IPVERSION) {
3673 		KMOD_IPSTAT_INC(ips_badvers);
3674 		goto bad;
3675 	}
3676 	hlen = ip->ip_hl << 2;
3677 	if (hlen < sizeof(struct ip)) { /* minimum header length */
3678 		KMOD_IPSTAT_INC(ips_badhlen);
3679 		goto bad;
3680 	}
3681 	if (hlen > m->m_len) {
3682 		if ((m = m_pullup(m, hlen)) == NULL) {
3683 			KMOD_IPSTAT_INC(ips_badhlen);
3684 			goto bad;
3685 		}
3686 		ip = mtod(m, struct ip *);
3687 		if (ip == NULL) goto bad;
3688 	}
3689 
3690 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3691 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3692 	} else {
3693 		if (hlen == sizeof(struct ip)) {
3694 			sum = in_cksum_hdr(ip);
3695 		} else {
3696 			sum = in_cksum(m, hlen);
3697 		}
3698 	}
3699 	if (sum) {
3700 		KMOD_IPSTAT_INC(ips_badsum);
3701 		goto bad;
3702 	}
3703 
3704 	/* Retrieve the packet length. */
3705 	len = ntohs(ip->ip_len);
3706 
3707 	/*
3708 	 * Check for additional length bogosity
3709 	 */
3710 	if (len < hlen) {
3711 		KMOD_IPSTAT_INC(ips_badlen);
3712 		goto bad;
3713 	}
3714 
3715 	/*
3716 	 * Check that the amount of data in the buffers
3717 	 * is as at least much as the IP header would have us expect.
3718 	 * Drop packet if shorter than we expect.
3719 	 */
3720 	if (m->m_pkthdr.len < len) {
3721 		KMOD_IPSTAT_INC(ips_tooshort);
3722 		goto bad;
3723 	}
3724 
3725 	/* Checks out, proceed */
3726 	*mp = m;
3727 	return (0);
3728 
3729 bad:
3730 	*mp = m;
3731 	return (-1);
3732 }
3733 
3734 #ifdef INET6
3735 /*
3736  * Same as above, but for IPv6.
3737  * Cut-and-pasted from ip6_input.c.
3738  * XXX Should we update ip6stat, or not?
3739  */
3740 static int
3741 bridge_ip6_checkbasic(struct mbuf **mp)
3742 {
3743 	struct mbuf *m = *mp;
3744 	struct ip6_hdr *ip6;
3745 
3746 	/*
3747 	 * If the IPv6 header is not aligned, slurp it up into a new
3748 	 * mbuf with space for link headers, in the event we forward
3749 	 * it.  Otherwise, if it is aligned, make sure the entire base
3750 	 * IPv6 header is in the first mbuf of the chain.
3751 	 */
3752 	if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3753 		struct ifnet *inifp = m->m_pkthdr.rcvif;
3754 		if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3755 			    (max_linkhdr + 3) & ~3)) == NULL) {
3756 			/* XXXJRT new stat, please */
3757 			IP6STAT_INC(ip6s_toosmall);
3758 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3759 			goto bad;
3760 		}
3761 	} else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3762 		struct ifnet *inifp = m->m_pkthdr.rcvif;
3763 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3764 			IP6STAT_INC(ip6s_toosmall);
3765 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3766 			goto bad;
3767 		}
3768 	}
3769 
3770 	ip6 = mtod(m, struct ip6_hdr *);
3771 
3772 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3773 		IP6STAT_INC(ip6s_badvers);
3774 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3775 		goto bad;
3776 	}
3777 
3778 	/* Checks out, proceed */
3779 	*mp = m;
3780 	return (0);
3781 
3782 bad:
3783 	*mp = m;
3784 	return (-1);
3785 }
3786 #endif /* INET6 */
3787 
3788 /*
3789  * bridge_fragment:
3790  *
3791  *	Fragment mbuf chain in multiple packets and prepend ethernet header.
3792  */
3793 static int
3794 bridge_fragment(struct ifnet *ifp, struct mbuf **mp, struct ether_header *eh,
3795     int snap, struct llc *llc)
3796 {
3797 	struct mbuf *m = *mp, *nextpkt = NULL, *mprev = NULL, *mcur = NULL;
3798 	struct ip *ip;
3799 	int error = -1;
3800 
3801 	if (m->m_len < sizeof(struct ip) &&
3802 	    (m = m_pullup(m, sizeof(struct ip))) == NULL)
3803 		goto dropit;
3804 	ip = mtod(m, struct ip *);
3805 
3806 	m->m_pkthdr.csum_flags |= CSUM_IP;
3807 	error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist);
3808 	if (error)
3809 		goto dropit;
3810 
3811 	/*
3812 	 * Walk the chain and re-add the Ethernet header for
3813 	 * each mbuf packet.
3814 	 */
3815 	for (mcur = m; mcur; mcur = mcur->m_nextpkt) {
3816 		nextpkt = mcur->m_nextpkt;
3817 		mcur->m_nextpkt = NULL;
3818 		if (snap) {
3819 			M_PREPEND(mcur, sizeof(struct llc), M_NOWAIT);
3820 			if (mcur == NULL) {
3821 				error = ENOBUFS;
3822 				if (mprev != NULL)
3823 					mprev->m_nextpkt = nextpkt;
3824 				goto dropit;
3825 			}
3826 			bcopy(llc, mtod(mcur, caddr_t),sizeof(struct llc));
3827 		}
3828 
3829 		M_PREPEND(mcur, ETHER_HDR_LEN, M_NOWAIT);
3830 		if (mcur == NULL) {
3831 			error = ENOBUFS;
3832 			if (mprev != NULL)
3833 				mprev->m_nextpkt = nextpkt;
3834 			goto dropit;
3835 		}
3836 		bcopy(eh, mtod(mcur, caddr_t), ETHER_HDR_LEN);
3837 
3838 		/*
3839 		 * The previous two M_PREPEND could have inserted one or two
3840 		 * mbufs in front so we have to update the previous packet's
3841 		 * m_nextpkt.
3842 		 */
3843 		mcur->m_nextpkt = nextpkt;
3844 		if (mprev != NULL)
3845 			mprev->m_nextpkt = mcur;
3846 		else {
3847 			/* The first mbuf in the original chain needs to be
3848 			 * updated. */
3849 			*mp = mcur;
3850 		}
3851 		mprev = mcur;
3852 	}
3853 
3854 	KMOD_IPSTAT_INC(ips_fragmented);
3855 	return (error);
3856 
3857 dropit:
3858 	for (mcur = *mp; mcur; mcur = m) { /* droping the full packet chain */
3859 		m = mcur->m_nextpkt;
3860 		m_freem(mcur);
3861 	}
3862 	return (error);
3863 }
3864 
3865 static void
3866 bridge_linkstate(struct ifnet *ifp)
3867 {
3868 	struct bridge_softc *sc = ifp->if_bridge;
3869 	struct bridge_iflist *bif;
3870 	struct epoch_tracker et;
3871 
3872 	NET_EPOCH_ENTER(et);
3873 
3874 	bif = bridge_lookup_member_if(sc, ifp);
3875 	if (bif == NULL) {
3876 		NET_EPOCH_EXIT(et);
3877 		return;
3878 	}
3879 	bridge_linkcheck(sc);
3880 
3881 	bstp_linkstate(&bif->bif_stp);
3882 
3883 	NET_EPOCH_EXIT(et);
3884 }
3885 
3886 static void
3887 bridge_linkcheck(struct bridge_softc *sc)
3888 {
3889 	struct bridge_iflist *bif;
3890 	int new_link, hasls;
3891 
3892 	BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
3893 
3894 	new_link = LINK_STATE_DOWN;
3895 	hasls = 0;
3896 	/* Our link is considered up if at least one of our ports is active */
3897 	CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
3898 		if (bif->bif_ifp->if_capabilities & IFCAP_LINKSTATE)
3899 			hasls++;
3900 		if (bif->bif_ifp->if_link_state == LINK_STATE_UP) {
3901 			new_link = LINK_STATE_UP;
3902 			break;
3903 		}
3904 	}
3905 	if (!CK_LIST_EMPTY(&sc->sc_iflist) && !hasls) {
3906 		/* If no interfaces support link-state then we default to up */
3907 		new_link = LINK_STATE_UP;
3908 	}
3909 	if_link_state_change(sc->sc_ifp, new_link);
3910 }
3911