1 /* $NetBSD: if_bridge.c,v 1.31 2005/06/01 19:45:34 jdc Exp $ */
4 * SPDX-License-Identifier: BSD-4-Clause
6 * Copyright 2001 Wasabi Systems, Inc.
9 * Written by Jason R. Thorpe for Wasabi Systems, Inc.
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
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
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.
41 * Copyright (c) 1999, 2000 Jason L. Wright (jason@thought.net)
42 * All rights reserved.
44 * Redistribution and use in source and binary forms, with or without
45 * modification, are permitted provided that the following conditions
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.
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.
65 * OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun Exp
69 * Network interface bridge support.
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).
79 #include <sys/cdefs.h>
80 __FBSDID("$FreeBSD$");
83 #include "opt_inet6.h"
85 #include <sys/param.h>
86 #include <sys/eventhandler.h>
88 #include <sys/malloc.h>
89 #include <sys/protosw.h>
90 #include <sys/systm.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>
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>
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>
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>
123 #include <netinet/ip6.h>
124 #include <netinet6/ip6_var.h>
125 #include <netinet6/in6_ifattach.h>
127 #if defined(INET) || defined(INET6)
128 #include <netinet/ip_carp.h>
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>
137 #include <net/route.h>
140 * Size of the route hash table. Must be a power of two.
142 #ifndef BRIDGE_RTHASH_SIZE
143 #define BRIDGE_RTHASH_SIZE 1024
146 #define BRIDGE_RTHASH_MASK (BRIDGE_RTHASH_SIZE - 1)
149 * Default maximum number of addresses to cache.
151 #ifndef BRIDGE_RTABLE_MAX
152 #define BRIDGE_RTABLE_MAX 2000
156 * Timeout (in seconds) for entries learned dynamically.
158 #ifndef BRIDGE_RTABLE_TIMEOUT
159 #define BRIDGE_RTABLE_TIMEOUT (20 * 60) /* same as ARP */
163 * Number of seconds between walks of the route list.
165 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
166 #define BRIDGE_RTABLE_PRUNE_PERIOD (5 * 60)
170 * List of capabilities to possibly mask on the member interface.
172 #define BRIDGE_IFCAPS_MASK (IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM|\
176 * List of capabilities to strip
178 #define BRIDGE_IFCAPS_STRIP IFCAP_LRO
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.
187 * That is: for read accesses we only need to be in NET_EPOCH, but for write
188 * accesses we must hold:
190 * - BRIDGE_RT_LOCK, for any change to bridge_rtnodes
191 * - BRIDGE_LOCK, for any other change
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.
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); \
202 #define BRIDGE_LOCK_DESTROY(_sc) do { \
203 sx_destroy(&(_sc)->sc_sx); \
204 mtx_destroy(&(_sc)->sc_rt_mtx); \
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))
219 * Bridge interface list entry.
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;
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;
247 #define brt_ifp brt_dst->bif_ifp
250 * Software state for each bridge.
252 struct bridge_softc {
253 struct ifnet *sc_ifp; /* make this an interface */
254 LIST_ENTRY(bridge_softc) sc_list;
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;
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;
278 int bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
280 VNET_DEFINE_STATIC(uma_zone_t, bridge_rtnode_zone);
281 #define V_bridge_rtnode_zone VNET(bridge_rtnode_zone)
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);
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 *,
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 *);
297 static void bridge_altq_start(if_t);
298 static int bridge_altq_transmit(if_t, struct mbuf *);
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 *,
305 static int bridge_enqueue(struct bridge_softc *, struct ifnet *,
307 static void bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
309 static void bridge_forward(struct bridge_softc *, struct bridge_iflist *,
312 static void bridge_timer(void *);
314 static void bridge_broadcast(struct bridge_softc *, struct ifnet *,
316 static void bridge_span(struct bridge_softc *, struct mbuf *);
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 *,
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 *,
328 static void bridge_rtable_init(struct bridge_softc *);
329 static void bridge_rtable_fini(struct bridge_softc *);
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);
341 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
343 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
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 *);
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 *,
383 static int bridge_ip_checkbasic(struct mbuf **mp);
385 static int bridge_ip6_checkbasic(struct mbuf **mp);
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);
393 * Use the "null" value from IEEE 802.1Q-2014 Table 9-2
394 * to indicate untagged frames.
396 #define VLANTAGOF(_m) \
397 (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : DOT1Q_VID_NULL
399 static struct bstp_cb_ops bridge_ops = {
400 .bcb_state = bridge_state_change,
401 .bcb_rtage = bridge_rtable_expire
404 SYSCTL_DECL(_net_link);
405 static SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
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");
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");
422 /* layer2 filter with ipfw */
423 VNET_DEFINE_STATIC(int, pfil_ipfw);
424 #define V_pfil_ipfw VNET(pfil_ipfw)
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");
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");
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");
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");
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");
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");
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");
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 };
479 #define V_log_interval VNET(log_interval)
480 #define V_log_count VNET(log_count)
481 #define V_log_last VNET(log_last)
483 struct bridge_control {
484 int (*bc_func)(struct bridge_softc *, void *);
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 */
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 },
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 },
504 { bridge_ioctl_scache, sizeof(struct ifbrparam),
505 BC_F_COPYIN|BC_F_SUSER },
506 { bridge_ioctl_gcache, sizeof(struct ifbrparam),
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 },
514 { bridge_ioctl_saddr, sizeof(struct ifbareq),
515 BC_F_COPYIN|BC_F_SUSER },
517 { bridge_ioctl_sto, sizeof(struct ifbrparam),
518 BC_F_COPYIN|BC_F_SUSER },
519 { bridge_ioctl_gto, sizeof(struct ifbrparam),
522 { bridge_ioctl_daddr, sizeof(struct ifbareq),
523 BC_F_COPYIN|BC_F_SUSER },
525 { bridge_ioctl_flush, sizeof(struct ifbreq),
526 BC_F_COPYIN|BC_F_SUSER },
528 { bridge_ioctl_gpri, sizeof(struct ifbrparam),
530 { bridge_ioctl_spri, sizeof(struct ifbrparam),
531 BC_F_COPYIN|BC_F_SUSER },
533 { bridge_ioctl_ght, sizeof(struct ifbrparam),
535 { bridge_ioctl_sht, sizeof(struct ifbrparam),
536 BC_F_COPYIN|BC_F_SUSER },
538 { bridge_ioctl_gfd, sizeof(struct ifbrparam),
540 { bridge_ioctl_sfd, sizeof(struct ifbrparam),
541 BC_F_COPYIN|BC_F_SUSER },
543 { bridge_ioctl_gma, sizeof(struct ifbrparam),
545 { bridge_ioctl_sma, sizeof(struct ifbrparam),
546 BC_F_COPYIN|BC_F_SUSER },
548 { bridge_ioctl_sifprio, sizeof(struct ifbreq),
549 BC_F_COPYIN|BC_F_SUSER },
551 { bridge_ioctl_sifcost, sizeof(struct ifbreq),
552 BC_F_COPYIN|BC_F_SUSER },
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 },
559 { bridge_ioctl_gbparam, sizeof(struct ifbropreq),
562 { bridge_ioctl_grte, sizeof(struct ifbrparam),
565 { bridge_ioctl_gifsstp, sizeof(struct ifbpstpconf),
566 BC_F_COPYIN|BC_F_COPYOUT },
568 { bridge_ioctl_sproto, sizeof(struct ifbrparam),
569 BC_F_COPYIN|BC_F_SUSER },
571 { bridge_ioctl_stxhc, sizeof(struct ifbrparam),
572 BC_F_COPYIN|BC_F_SUSER },
574 { bridge_ioctl_sifmaxaddr, sizeof(struct ifbreq),
575 BC_F_COPYIN|BC_F_SUSER },
578 static const int bridge_control_table_size = nitems(bridge_control_table);
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, \
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)
588 VNET_DEFINE_STATIC(struct if_clone *, bridge_cloner);
589 #define V_bridge_cloner VNET(bridge_cloner)
591 static const char bridge_name[] = "bridge";
594 vnet_bridge_init(const void *unused __unused)
597 V_bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
598 sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
600 BRIDGE_LIST_LOCK_INIT();
601 LIST_INIT(&V_bridge_list);
603 struct if_clone_addreq req = {
604 .create_f = bridge_clone_create,
605 .destroy_f = bridge_clone_destroy,
606 .flags = IFC_F_AUTOUNIT,
608 V_bridge_cloner = ifc_attach_cloner(bridge_name, &req);
610 VNET_SYSINIT(vnet_bridge_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
611 vnet_bridge_init, NULL);
614 vnet_bridge_uninit(const void *unused __unused)
617 ifc_detach_cloner(V_bridge_cloner);
618 V_bridge_cloner = NULL;
619 BRIDGE_LIST_LOCK_DESTROY();
621 /* Callbacks may use the UMA zone. */
622 NET_EPOCH_DRAIN_CALLBACKS();
624 uma_zdestroy(V_bridge_rtnode_zone);
626 VNET_SYSUNINIT(vnet_bridge_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY,
627 vnet_bridge_uninit, NULL);
630 bridge_modevent(module_t mod, int type, void *data)
635 bridge_dn_p = bridge_dummynet;
636 bridge_detach_cookie = EVENTHANDLER_REGISTER(
637 ifnet_departure_event, bridge_ifdetach, NULL,
638 EVENTHANDLER_PRI_ANY);
641 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
642 bridge_detach_cookie);
651 static moduledata_t bridge_mod = {
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);
662 * handler for net.link.bridge.ipfw
665 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
667 int enable = V_pfil_ipfw;
670 error = sysctl_handle_int(oidp, &enable, 0, req);
673 if (enable != V_pfil_ipfw) {
674 V_pfil_ipfw = enable;
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
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");
698 bridge_reassign(struct ifnet *ifp, struct vnet *newvnet, char *arg)
700 struct bridge_softc *sc = ifp->if_softc;
701 struct bridge_iflist *bif;
705 while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
706 bridge_delete_member(sc, bif, 0);
708 while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
709 bridge_delete_span(sc, bif);
714 ether_reassign(ifp, newvnet, arg);
719 * bridge_clone_create:
721 * Create a new bridge instance.
724 bridge_clone_create(struct if_clone *ifc, char *name, size_t len,
725 struct ifc_data *ifd, struct ifnet **ifpp)
727 struct bridge_softc *sc;
730 sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
731 ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
737 BRIDGE_LOCK_INIT(sc);
738 sc->sc_brtmax = BRIDGE_RTABLE_MAX;
739 sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
741 /* Initialize our routing table. */
742 bridge_rtable_init(sc);
744 callout_init_mtx(&sc->sc_brcallout, &sc->sc_rt_mtx, 0);
746 CK_LIST_INIT(&sc->sc_iflist);
747 CK_LIST_INIT(&sc->sc_spanlist);
750 if_initname(ifp, bridge_name, ifd->unit);
751 ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
752 ifp->if_ioctl = bridge_ioctl;
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);
760 ifp->if_transmit = bridge_transmit;
762 ifp->if_qflush = bridge_qflush;
763 ifp->if_init = bridge_init;
764 ifp->if_type = IFT_BRIDGE;
766 ether_gen_addr(ifp, &sc->sc_defaddr);
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;
774 ifp->if_reassign = bridge_reassign;
776 sc->sc_if_input = ifp->if_input; /* ether_input */
777 ifp->if_input = bridge_inject;
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.
784 ifp->if_bridge_input = bridge_input;
787 LIST_INSERT_HEAD(&V_bridge_list, sc, sc_list);
788 BRIDGE_LIST_UNLOCK();
795 bridge_clone_destroy_cb(struct epoch_context *ctx)
797 struct bridge_softc *sc;
799 sc = __containerof(ctx, struct bridge_softc, sc_epoch_ctx);
801 BRIDGE_LOCK_DESTROY(sc);
806 * bridge_clone_destroy:
808 * Destroy a bridge instance.
811 bridge_clone_destroy(struct if_clone *ifc, struct ifnet *ifp, uint32_t flags)
813 struct bridge_softc *sc = ifp->if_softc;
814 struct bridge_iflist *bif;
815 struct epoch_tracker et;
820 ifp->if_flags &= ~IFF_UP;
822 while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
823 bridge_delete_member(sc, bif, 0);
825 while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
826 bridge_delete_span(sc, bif);
829 /* Tear down the routing table. */
830 bridge_rtable_fini(sc);
836 callout_drain(&sc->sc_brcallout);
839 LIST_REMOVE(sc, sc_list);
840 BRIDGE_LIST_UNLOCK();
842 bstp_detach(&sc->sc_stp);
844 IFQ_PURGE(&ifp->if_snd);
851 NET_EPOCH_CALL(bridge_clone_destroy_cb, &sc->sc_epoch_ctx);
859 * Handle a control request from the operator.
862 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
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;
869 struct ifbreq ifbreq;
870 struct ifbifconf ifbifconf;
871 struct ifbareq ifbareq;
872 struct ifbaconf ifbaconf;
873 struct ifbrparam ifbrparam;
874 struct ifbropreq ifbropreq;
876 struct ifdrv *ifd = (struct ifdrv *) data;
877 const struct bridge_control *bc;
878 int error = 0, oldmtu;
889 if (ifd->ifd_cmd >= bridge_control_table_size) {
893 bc = &bridge_control_table[ifd->ifd_cmd];
895 if (cmd == SIOCGDRVSPEC &&
896 (bc->bc_flags & BC_F_COPYOUT) == 0) {
900 else if (cmd == SIOCSDRVSPEC &&
901 (bc->bc_flags & BC_F_COPYOUT) != 0) {
906 if (bc->bc_flags & BC_F_SUSER) {
907 error = priv_check(td, PRIV_NET_BRIDGE);
912 if (ifd->ifd_len != bc->bc_argsize ||
913 ifd->ifd_len > sizeof(args)) {
918 bzero(&args, sizeof(args));
919 if (bc->bc_flags & BC_F_COPYIN) {
920 error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
925 oldmtu = ifp->if_mtu;
926 error = (*bc->bc_func)(sc, &args);
931 * Bridge MTU may change during addition of the first port.
932 * If it did, do network layer specific procedure.
934 if (ifp->if_mtu != oldmtu)
937 if (bc->bc_flags & BC_F_COPYOUT)
938 error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
943 if (!(ifp->if_flags & IFF_UP) &&
944 (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
946 * If interface is marked down and it is running,
947 * then stop and disable it.
950 } else if ((ifp->if_flags & IFF_UP) &&
951 !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
953 * If interface is marked up and it is stopped, then
963 oldmtu = sc->sc_ifp->if_mtu;
965 if (ifr->ifr_mtu < 576) {
969 if (CK_LIST_EMPTY(&sc->sc_iflist)) {
970 sc->sc_ifp->if_mtu = ifr->ifr_mtu;
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);
977 log(LOG_NOTICE, "%s: invalid MTU: %u for"
978 " member %s\n", sc->sc_ifp->if_xname,
980 bif->bif_ifp->if_xname);
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);
993 sc->sc_ifp->if_mtu = ifr->ifr_mtu;
998 * drop the lock as ether_ioctl() will call bridge_start() and
999 * cause the lock to be recursed.
1002 error = ether_ioctl(ifp, cmd, data);
1015 * Clear or restore unwanted capabilities on the member interface
1018 bridge_mutecaps(struct bridge_softc *sc)
1020 struct bridge_iflist *bif;
1023 BRIDGE_LOCK_ASSERT(sc);
1025 /* Initial bitmask of capabilities to test */
1026 mask = BRIDGE_IFCAPS_MASK;
1028 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1029 /* Every member must support it or its disabled */
1030 mask &= bif->bif_savedcaps;
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;
1039 bridge_set_ifcap(sc, bif, enabled);
1044 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
1046 struct ifnet *ifp = bif->bif_ifp;
1048 int error, mask, stuck;
1050 bzero(&ifr, sizeof(ifr));
1051 ifr.ifr_reqcap = set;
1053 if (ifp->if_capenable != set) {
1054 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
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;
1062 if_printf(sc->sc_ifp,
1063 "can't disable some capabilities on %s: 0x%x\n",
1064 ifp->if_xname, stuck);
1069 * bridge_lookup_member:
1071 * Lookup a bridge member interface.
1073 static struct bridge_iflist *
1074 bridge_lookup_member(struct bridge_softc *sc, const char *name)
1076 struct bridge_iflist *bif;
1079 BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1081 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1083 if (strcmp(ifp->if_xname, name) == 0)
1091 * bridge_lookup_member_if:
1093 * Lookup a bridge member interface by ifnet*.
1095 static struct bridge_iflist *
1096 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
1098 struct bridge_iflist *bif;
1100 BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1102 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1103 if (bif->bif_ifp == member_ifp)
1111 bridge_delete_member_cb(struct epoch_context *ctx)
1113 struct bridge_iflist *bif;
1115 bif = __containerof(ctx, struct bridge_iflist, bif_epoch_ctx);
1117 free(bif, M_DEVBUF);
1121 * bridge_delete_member:
1123 * Delete the specified member interface.
1126 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
1129 struct ifnet *ifs = bif->bif_ifp;
1130 struct ifnet *fif = NULL;
1131 struct bridge_iflist *bifl;
1133 BRIDGE_LOCK_ASSERT(sc);
1135 if (bif->bif_flags & IFBIF_STP)
1136 bstp_disable(&bif->bif_stp);
1138 ifs->if_bridge = NULL;
1139 CK_LIST_REMOVE(bif, bif_next);
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.
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;
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;
1158 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1161 bridge_linkcheck(sc);
1162 bridge_mutecaps(sc); /* recalcuate now this interface is removed */
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));
1169 ifs->if_bridge_output = NULL;
1170 ifs->if_bridge_input = NULL;
1171 ifs->if_bridge_linkstate = NULL;
1173 switch (ifs->if_type) {
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.
1181 if (ifs->if_flags & IFF_PROMISC)
1182 (void) ifpromisc(ifs, 0);
1190 panic("bridge_delete_member: impossible");
1194 /* reneable any interface capabilities */
1195 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
1197 bstp_destroy(&bif->bif_stp); /* prepare to free */
1199 NET_EPOCH_CALL(bridge_delete_member_cb, &bif->bif_epoch_ctx);
1203 * bridge_delete_span:
1205 * Delete the specified span interface.
1208 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1210 BRIDGE_LOCK_ASSERT(sc);
1212 KASSERT(bif->bif_ifp->if_bridge == NULL,
1213 ("%s: not a span interface", __func__));
1215 CK_LIST_REMOVE(bif, bif_next);
1217 NET_EPOCH_CALL(bridge_delete_member_cb, &bif->bif_epoch_ctx);
1221 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1223 struct ifbreq *req = arg;
1224 struct bridge_iflist *bif = NULL;
1228 ifs = ifunit(req->ifbr_ifsname);
1231 if (ifs->if_ioctl == NULL) /* must be supported */
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)
1239 if (ifs->if_bridge == sc)
1242 if (ifs->if_bridge != NULL)
1245 switch (ifs->if_type) {
1249 /* permitted interface types */
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
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) {
1270 * If any, remove all inet6 addresses from the member
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);
1283 if (in6ifa_llaonifp(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",
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) {
1299 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "%s",
1301 ifr.ifr_mtu = sc->sc_ifp->if_mtu;
1303 error = (*ifs->if_ioctl)(ifs,
1304 SIOCSIFMTU, (caddr_t)&ifr);
1306 log(LOG_NOTICE, "%s: invalid MTU: %u for"
1307 " new member %s\n", sc->sc_ifp->if_xname,
1314 bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1319 bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1320 bif->bif_savedcaps = ifs->if_capenable;
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.
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);
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);
1340 * XXX: XLOCK HERE!?!
1342 * NOTE: insert_***HEAD*** should be safe for the traversals.
1344 CK_LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1346 /* Set interface capabilities to the intersection set of all members */
1347 bridge_mutecaps(sc);
1348 bridge_linkcheck(sc);
1350 /* Place the interface into promiscuous mode */
1351 switch (ifs->if_type) {
1354 error = ifpromisc(ifs, 1);
1359 bridge_delete_member(sc, bif, 0);
1364 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1366 struct ifbreq *req = arg;
1367 struct bridge_iflist *bif;
1369 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1373 bridge_delete_member(sc, bif, 0);
1379 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1381 struct ifbreq *req = arg;
1382 struct bridge_iflist *bif;
1383 struct bstp_port *bp;
1385 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
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;
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;
1419 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1421 struct epoch_tracker et;
1422 struct ifbreq *req = arg;
1423 struct bridge_iflist *bif;
1424 struct bstp_port *bp;
1427 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1432 if (req->ifbr_ifsflags & IFBIF_SPAN)
1433 /* SPAN is readonly */
1436 NET_EPOCH_ENTER(et);
1438 if (req->ifbr_ifsflags & IFBIF_STP) {
1439 if ((bif->bif_flags & IFBIF_STP) == 0) {
1440 error = bstp_enable(&bif->bif_stp);
1447 if ((bif->bif_flags & IFBIF_STP) != 0)
1448 bstp_disable(&bif->bif_stp);
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);
1457 /* Save the bits relating to the bridge */
1458 bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1466 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1468 struct ifbrparam *param = arg;
1470 sc->sc_brtmax = param->ifbrp_csize;
1477 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1479 struct ifbrparam *param = arg;
1481 param->ifbrp_csize = sc->sc_brtmax;
1487 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1489 struct ifbifconf *bifc = arg;
1490 struct bridge_iflist *bif;
1493 int count, buflen, len, error = 0;
1496 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1498 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1501 buflen = sizeof(breq) * count;
1502 if (bifc->ifbic_len == 0) {
1503 bifc->ifbic_len = buflen;
1506 outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
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))
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);
1524 memcpy(buf, &breq, sizeof(breq));
1526 buf += sizeof(breq);
1527 len -= sizeof(breq);
1529 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1530 if (len < sizeof(breq))
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));
1539 buf += sizeof(breq);
1540 len -= sizeof(breq);
1543 bifc->ifbic_len = sizeof(breq) * count;
1544 error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1545 free(outbuf, M_TEMP);
1550 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1552 struct ifbaconf *bac = arg;
1553 struct bridge_rtnode *brt;
1554 struct ifbareq bareq;
1556 int count, buflen, len, error = 0;
1558 if (bac->ifbac_len == 0)
1562 CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1564 buflen = sizeof(bareq) * count;
1566 outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
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))
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;
1585 bareq.ifba_expire = 0;
1586 bareq.ifba_flags = brt->brt_flags;
1588 memcpy(buf, &bareq, sizeof(bareq));
1590 buf += sizeof(bareq);
1591 len -= sizeof(bareq);
1594 bac->ifbac_len = sizeof(bareq) * count;
1595 error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1596 free(outbuf, M_TEMP);
1601 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1603 struct ifbareq *req = arg;
1604 struct bridge_iflist *bif;
1605 struct epoch_tracker et;
1608 NET_EPOCH_ENTER(et);
1609 bif = bridge_lookup_member(sc, req->ifba_ifsname);
1615 /* bridge_rtupdate() may acquire the lock. */
1616 error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1624 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1626 struct ifbrparam *param = arg;
1628 sc->sc_brttimeout = param->ifbrp_ctime;
1633 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1635 struct ifbrparam *param = arg;
1637 param->ifbrp_ctime = sc->sc_brttimeout;
1642 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1644 struct ifbareq *req = arg;
1645 int vlan = req->ifba_vlan;
1647 /* Userspace uses '0' to mean 'any vlan' */
1649 vlan = DOT1Q_VID_RSVD_IMPL;
1651 return (bridge_rtdaddr(sc, req->ifba_dst, vlan));
1655 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1657 struct ifbreq *req = arg;
1660 bridge_rtflush(sc, req->ifbr_ifsflags);
1661 BRIDGE_RT_UNLOCK(sc);
1667 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1669 struct ifbrparam *param = arg;
1670 struct bstp_state *bs = &sc->sc_stp;
1672 param->ifbrp_prio = bs->bs_bridge_priority;
1677 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1679 struct ifbrparam *param = arg;
1681 return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1685 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1687 struct ifbrparam *param = arg;
1688 struct bstp_state *bs = &sc->sc_stp;
1690 param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1695 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1697 struct ifbrparam *param = arg;
1699 return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1703 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1705 struct ifbrparam *param = arg;
1706 struct bstp_state *bs = &sc->sc_stp;
1708 param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1713 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1715 struct ifbrparam *param = arg;
1717 return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1721 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1723 struct ifbrparam *param = arg;
1724 struct bstp_state *bs = &sc->sc_stp;
1726 param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1731 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1733 struct ifbrparam *param = arg;
1735 return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1739 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1741 struct ifbreq *req = arg;
1742 struct bridge_iflist *bif;
1744 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1748 return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1752 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1754 struct ifbreq *req = arg;
1755 struct bridge_iflist *bif;
1757 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1761 return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1765 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1767 struct ifbreq *req = arg;
1768 struct bridge_iflist *bif;
1770 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1774 bif->bif_addrmax = req->ifbr_addrmax;
1779 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1781 struct ifbreq *req = arg;
1782 struct bridge_iflist *bif = NULL;
1785 ifs = ifunit(req->ifbr_ifsname);
1789 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1790 if (ifs == bif->bif_ifp)
1793 if (ifs->if_bridge != NULL)
1796 switch (ifs->if_type) {
1805 bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1810 bif->bif_flags = IFBIF_SPAN;
1812 CK_LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1818 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1820 struct ifbreq *req = arg;
1821 struct bridge_iflist *bif;
1824 ifs = ifunit(req->ifbr_ifsname);
1828 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1829 if (ifs == bif->bif_ifp)
1835 bridge_delete_span(sc, bif);
1841 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1843 struct ifbropreq *req = arg;
1844 struct bstp_state *bs = &sc->sc_stp;
1845 struct bstp_port *root_port;
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;
1851 root_port = bs->bs_root_port;
1852 if (root_port == NULL)
1853 req->ifbop_root_port = 0;
1855 req->ifbop_root_port = root_port->bp_ifp->if_index;
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;
1871 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1873 struct ifbrparam *param = arg;
1875 param->ifbrp_cexceeded = sc->sc_brtexceeded;
1880 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1882 struct ifbpstpconf *bifstp = arg;
1883 struct bridge_iflist *bif;
1884 struct bstp_port *bp;
1885 struct ifbpstpreq bpreq;
1887 int count, buflen, len, error = 0;
1890 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1891 if ((bif->bif_flags & IFBIF_STP) != 0)
1895 buflen = sizeof(bpreq) * count;
1896 if (bifstp->ifbpstp_len == 0) {
1897 bifstp->ifbpstp_len = buflen;
1901 outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
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))
1913 if ((bif->bif_flags & IFBIF_STP) == 0)
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;
1924 memcpy(buf, &bpreq, sizeof(bpreq));
1926 buf += sizeof(bpreq);
1927 len -= sizeof(bpreq);
1930 bifstp->ifbpstp_len = sizeof(bpreq) * count;
1931 error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1932 free(outbuf, M_TEMP);
1937 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1939 struct ifbrparam *param = arg;
1941 return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1945 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1947 struct ifbrparam *param = arg;
1949 return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1955 * Detach an interface from a bridge. Called when a member
1956 * interface is detaching.
1959 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1961 struct bridge_softc *sc = ifp->if_bridge;
1962 struct bridge_iflist *bif;
1964 if (ifp->if_flags & IFF_RENAMING)
1966 if (V_bridge_cloner == NULL) {
1968 * This detach handler can be called after
1969 * vnet_bridge_uninit(). Just return in that case.
1973 /* Check if the interface is a bridge member */
1977 bif = bridge_lookup_member_if(sc, ifp);
1979 bridge_delete_member(sc, bif, 1);
1985 /* Check if the interface is a span port */
1987 LIST_FOREACH(sc, &V_bridge_list, sc_list) {
1989 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1990 if (ifp == bif->bif_ifp) {
1991 bridge_delete_span(sc, bif);
1997 BRIDGE_LIST_UNLOCK();
2003 * Initialize a bridge interface.
2006 bridge_init(void *xsc)
2008 struct bridge_softc *sc = (struct bridge_softc *)xsc;
2009 struct ifnet *ifp = sc->sc_ifp;
2011 if (ifp->if_drv_flags & IFF_DRV_RUNNING)
2015 callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
2018 ifp->if_drv_flags |= IFF_DRV_RUNNING;
2019 bstp_init(&sc->sc_stp); /* Initialize Spanning Tree */
2027 * Stop the bridge interface.
2030 bridge_stop(struct ifnet *ifp, int disable)
2032 struct bridge_softc *sc = ifp->if_softc;
2034 BRIDGE_LOCK_ASSERT(sc);
2036 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2040 callout_stop(&sc->sc_brcallout);
2042 bstp_stop(&sc->sc_stp);
2044 bridge_rtflush(sc, IFBF_FLUSHDYN);
2045 BRIDGE_RT_UNLOCK(sc);
2047 ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
2053 * Enqueue a packet on a bridge member interface.
2057 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
2063 /* We may be sending a fragment so traverse the mbuf */
2066 m->m_nextpkt = NULL;
2067 len = m->m_pkthdr.len;
2068 mflags = m->m_flags;
2071 * If underlying interface can not do VLAN tag insertion itself
2072 * then attach a packet tag that holds it.
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);
2079 "unable to prepend VLAN header\n");
2080 if_inc_counter(dst_ifp, IFCOUNTER_OERRORS, 1);
2083 m->m_flags &= ~M_VLANTAG;
2086 M_ASSERTPKTHDR(m); /* We shouldn't transmit mbuf without pkthdr */
2087 if ((err = dst_ifp->if_transmit(dst_ifp, m))) {
2090 for (m = m0, n = 1; m != NULL; m = m0, n++) {
2094 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, n);
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);
2110 * Receive a queued packet from dummynet and pass it on to the output
2113 * The mbuf has the Ethernet header already attached.
2116 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
2118 struct bridge_softc *sc;
2120 sc = ifp->if_bridge;
2123 * The packet didnt originate from a member interface. This should only
2124 * ever happen if a member interface is removed while packets are
2132 if (PFIL_HOOKED_OUT(V_inet_pfil_head)
2134 || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2137 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
2143 bridge_enqueue(sc, ifp, m);
2149 * Send output from a bridge member interface. This
2150 * performs the bridging function for locally originated
2153 * The mbuf has the Ethernet header already attached. We must
2154 * enqueue or free the mbuf before returning.
2157 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
2160 struct ether_header *eh;
2161 struct ifnet *bifp, *dst_if;
2162 struct bridge_softc *sc;
2167 if (m->m_len < ETHER_HDR_LEN) {
2168 m = m_pullup(m, ETHER_HDR_LEN);
2173 eh = mtod(m, struct ether_header *);
2174 sc = ifp->if_bridge;
2175 vlan = VLANTAGOF(m);
2180 * If bridge is down, but the original output interface is up,
2181 * go ahead and send out that interface. Otherwise, the packet
2184 if ((bifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2190 * If the packet is a multicast, or we don't know a better way to
2191 * get there, send to all interfaces.
2193 if (ETHER_IS_MULTICAST(eh->ether_dhost))
2196 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
2197 /* Tap any traffic not passing back out the originating interface */
2199 ETHER_BPF_MTAP(bifp, m);
2200 if (dst_if == NULL) {
2201 struct bridge_iflist *bif;
2207 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
2208 dst_if = bif->bif_ifp;
2210 if (dst_if->if_type == IFT_GIF)
2212 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
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.
2221 if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
2222 bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2225 if (CK_LIST_NEXT(bif, bif_next) == NULL) {
2229 mc = m_dup(m, M_NOWAIT);
2231 if_inc_counter(bifp, IFCOUNTER_OERRORS, 1);
2236 bridge_enqueue(sc, dst_if, mc);
2245 * XXX Spanning tree consideration here?
2249 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2254 bridge_enqueue(sc, dst_if, m);
2261 * Do output on a bridge.
2265 bridge_transmit(struct ifnet *ifp, struct mbuf *m)
2267 struct bridge_softc *sc;
2268 struct ether_header *eh;
2269 struct ifnet *dst_if;
2274 ETHER_BPF_MTAP(ifp, m);
2276 eh = mtod(m, struct ether_header *);
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);
2282 bridge_broadcast(sc, ifp, m, 0);
2289 bridge_altq_start(if_t ifp)
2291 struct ifaltq *ifq = &ifp->if_snd;
2295 IFQ_DEQUEUE_NOLOCK(ifq, m);
2297 bridge_transmit(ifp, m);
2298 IFQ_DEQUEUE_NOLOCK(ifq, m);
2304 bridge_altq_transmit(if_t ifp, struct mbuf *m)
2308 if (ALTQ_IS_ENABLED(&ifp->if_snd)) {
2309 IFQ_ENQUEUE(&ifp->if_snd, m, err);
2311 bridge_altq_start(ifp);
2313 err = bridge_transmit(ifp, m);
2320 * The ifp->if_qflush entry point for if_bridge(4) is no-op.
2323 bridge_qflush(struct ifnet *ifp __unused)
2330 * The forwarding function of the bridge.
2332 * NOTE: Releases the lock on return.
2335 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2338 struct bridge_iflist *dbif;
2339 struct ifnet *src_if, *dst_if, *ifp;
2340 struct ether_header *eh;
2347 src_if = m->m_pkthdr.rcvif;
2350 if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
2351 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2352 vlan = VLANTAGOF(m);
2354 if ((sbif->bif_flags & IFBIF_STP) &&
2355 sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2358 eh = mtod(m, struct ether_header *);
2359 dst = eh->ether_dhost;
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);
2366 * If the interface has addresses limits then deny any source
2367 * that is not in the cache.
2369 if (error && sbif->bif_addrmax)
2373 if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2374 sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2379 * Hand the packet to netmap only if it wasn't injected by netmap
2382 if ((m->m_flags & M_BRIDGE_INJECT) == 0 &&
2383 (if_getcapenable(ifp) & IFCAP_NETMAP) != 0) {
2384 ifp->if_input(ifp, m);
2387 m->m_flags &= ~M_BRIDGE_INJECT;
2391 * At this point, the port either doesn't participate
2392 * in spanning tree or it is in the forwarding state.
2396 * If the packet is unicast, destined for someone on
2397 * "this" side of the bridge, drop it.
2399 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2400 dst_if = bridge_rtlookup(sc, dst, vlan);
2401 if (src_if == dst_if)
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
2408 * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2410 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2411 dst[2] == 0xc2 && dst[3] == 0x00 &&
2412 dst[4] == 0x00 && dst[5] <= 0x0f)
2415 /* ...forward it to all interfaces. */
2416 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
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.
2429 if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2430 ETHER_BPF_MTAP(ifp, m);
2432 /* run the packet filter */
2433 if (PFIL_HOOKED_IN(V_inet_pfil_head)
2435 || PFIL_HOOKED_IN(V_inet6_pfil_head)
2438 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2444 if (dst_if == NULL) {
2445 bridge_broadcast(sc, src_if, m, 1);
2450 * At this point, we're dealing with a unicast frame
2451 * going to a different interface.
2453 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2456 dbif = bridge_lookup_member_if(sc, dst_if);
2458 /* Not a member of the bridge (anymore?) */
2461 /* Private segments can not talk to each other */
2462 if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2465 if ((dbif->bif_flags & IFBIF_STP) &&
2466 dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2469 if (PFIL_HOOKED_OUT(V_inet_pfil_head)
2471 || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2474 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2480 bridge_enqueue(sc, dst_if, m);
2490 * Receive input from a member interface. Queue the packet for
2491 * bridging if it is not for us.
2493 static struct mbuf *
2494 bridge_input(struct ifnet *ifp, struct mbuf *m)
2496 struct bridge_softc *sc;
2497 struct bridge_iflist *bif, *bif2;
2499 struct ether_header *eh;
2500 struct mbuf *mc, *mc2;
2506 eh = mtod(m, struct ether_header *);
2507 vlan = VLANTAGOF(m);
2509 sc = ifp->if_bridge;
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.
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);
2522 if_inc_counter(sc->sc_ifp, IFCOUNTER_IERRORS, 1);
2526 m->m_pkthdr.rcvif = ifp;
2529 if ((bifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
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.
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);
2546 bif = bridge_lookup_member_if(sc, ifp);
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 */
2561 if ((bif->bif_flags & IFBIF_STP) &&
2562 bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2567 * Make a deep copy of the packet and enqueue the copy
2568 * for bridge processing; return the original packet for
2571 mc = m_dup(m, M_NOWAIT);
2576 /* Perform the bridge forwarding function with the copy. */
2577 bridge_forward(sc, bif, mc);
2581 * If netmap is enabled and has not already seen this packet,
2582 * then it will be consumed by bridge_forward().
2584 if ((if_getcapenable(bifp) & IFCAP_NETMAP) != 0 &&
2585 (m->m_flags & M_BRIDGE_INJECT) == 0) {
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
2597 KASSERT(bifp->if_bridge == NULL,
2598 ("loop created in bridge_input"));
2599 mc2 = m_dup(m, M_NOWAIT);
2601 /* Keep the layer3 header aligned */
2602 int i = min(mc2->m_pkthdr.len, max_protohdr);
2603 mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2606 mc2->m_pkthdr.rcvif = bifp;
2607 mc2->m_flags &= ~M_BRIDGE_INJECT;
2608 sc->sc_if_input(bifp, mc2);
2611 /* Return the original packet for local processing. */
2615 if ((bif->bif_flags & IFBIF_STP) &&
2616 bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
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))
2626 #define CARP_CHECK_WE_ARE_DST(iface) false
2627 #define CARP_CHECK_WE_ARE_SRC(iface) false
2631 #define PFIL_HOOKED_INET6 PFIL_HOOKED_IN(V_inet6_pfil_head)
2633 #define PFIL_HOOKED_INET6 false
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); \
2645 #define GRAB_FOR_NETMAP(ifp, m)
2648 #define GRAB_OUR_PACKETS(iface) \
2649 if ((iface)->if_type == IFT_GIF) \
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) { \
2662 m->m_pkthdr.rcvif = iface; \
2663 if ((iface) == ifp) { \
2664 /* Skip bridge processing... src == dest */ \
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) { \
2681 if ((iface) != bifp) \
2682 ETHER_BPF_MTAP(iface, m); \
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)) { \
2694 * Unicast. Make sure it's not for the bridge.
2696 do { GRAB_OUR_PACKETS(bifp) } while (0);
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.
2705 do { GRAB_OUR_PACKETS(ifp) } while (0);
2707 /* Now check the all bridge members. */
2708 CK_LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2709 GRAB_OUR_PACKETS(bif2->bif_ifp)
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
2718 /* Perform the bridge forwarding function. */
2719 bridge_forward(sc, bif, m);
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.
2731 bridge_inject(struct ifnet *ifp, struct mbuf *m)
2733 struct bridge_softc *sc;
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));
2741 m->m_flags |= M_BRIDGE_INJECT;
2742 sc = if_getsoftc(ifp);
2743 sc->sc_if_input(ifp, m);
2749 * Send a frame to all interfaces that are members of
2750 * the bridge, except for the one on which the packet
2753 * NOTE: Releases the lock on return.
2756 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2757 struct mbuf *m, int runfilt)
2759 struct bridge_iflist *dbif, *sbif;
2761 struct ifnet *dst_if;
2766 sbif = bridge_lookup_member_if(sc, src_if);
2768 /* Filter on the bridge interface before broadcasting */
2769 if (runfilt && (PFIL_HOOKED_OUT(V_inet_pfil_head)
2771 || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2774 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2780 CK_LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2781 dst_if = dbif->bif_ifp;
2782 if (dst_if == src_if)
2785 /* Private segments can not talk to each other */
2786 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2789 if ((dbif->bif_flags & IFBIF_STP) &&
2790 dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2793 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2794 (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2797 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2800 if (CK_LIST_NEXT(dbif, bif_next) == NULL) {
2804 mc = m_dup(m, M_NOWAIT);
2806 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
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.
2816 if (runfilt && (PFIL_HOOKED_OUT(V_inet_pfil_head)
2818 || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2822 /* Keep the layer3 header aligned */
2823 i = min(mc->m_pkthdr.len, max_protohdr);
2824 mc = m_copyup(mc, i, ETHER_ALIGN);
2826 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2830 if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2836 bridge_enqueue(sc, dst_if, mc);
2845 * Duplicate a packet out one or more interfaces that are in span mode,
2846 * the original mbuf is unmodified.
2849 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2851 struct bridge_iflist *bif;
2852 struct ifnet *dst_if;
2857 if (CK_LIST_EMPTY(&sc->sc_spanlist))
2860 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2861 dst_if = bif->bif_ifp;
2863 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2866 mc = m_dup(m, M_NOWAIT);
2868 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2872 bridge_enqueue(sc, dst_if, mc);
2879 * Add a bridge routing entry.
2882 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2883 struct bridge_iflist *bif, int setflags, uint8_t flags)
2885 struct bridge_rtnode *brt;
2886 struct bridge_iflist *obif;
2889 BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
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)
2898 * A route for this destination might already exist. If so,
2899 * update it, otherwise create a new one.
2901 if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
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);
2911 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2912 sc->sc_brtexceeded++;
2913 BRIDGE_RT_UNLOCK(sc);
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);
2924 * Allocate a new bridge forwarding node, and
2925 * initialize the expiration time and Ethernet
2928 brt = uma_zalloc(V_bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2930 BRIDGE_RT_UNLOCK(sc);
2933 brt->brt_vnet = curvnet;
2935 if (bif->bif_flags & IFBIF_STICKY)
2936 brt->brt_flags = IFBAF_STICKY;
2938 brt->brt_flags = IFBAF_DYNAMIC;
2940 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2941 brt->brt_vlan = vlan;
2943 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2944 uma_zfree(V_bridge_rtnode_zone, brt);
2945 BRIDGE_RT_UNLOCK(sc);
2951 BRIDGE_RT_UNLOCK(sc);
2954 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2955 (obif = brt->brt_dst) != bif) {
2957 brt->brt_dst->bif_addrcnt--;
2959 brt->brt_dst->bif_addrcnt++;
2960 BRIDGE_RT_UNLOCK(sc);
2962 if (V_log_mac_flap &&
2963 ppsratecheck(&V_log_last, &V_log_count, V_log_interval)) {
2965 "%s: mac address %6D vlan %d moved from %s to %s\n",
2966 sc->sc_ifp->if_xname,
2967 &brt->brt_addr[0], ":",
2969 obif->bif_ifp->if_xname,
2970 bif->bif_ifp->if_xname);
2974 if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2975 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2977 brt->brt_flags = flags;
2985 * Lookup the destination interface for an address.
2987 static struct ifnet *
2988 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2990 struct bridge_rtnode *brt;
2994 if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2997 return (brt->brt_ifp);
3003 * Trim the routine table so that we have a number
3004 * of routing entries less than or equal to the
3008 bridge_rttrim(struct bridge_softc *sc)
3010 struct bridge_rtnode *brt, *nbrt;
3013 BRIDGE_RT_LOCK_ASSERT(sc);
3015 /* Make sure we actually need to do this. */
3016 if (sc->sc_brtcnt <= sc->sc_brtmax)
3019 /* Force an aging cycle; this might trim enough addresses. */
3021 if (sc->sc_brtcnt <= sc->sc_brtmax)
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)
3036 * Aging timer for the bridge.
3039 bridge_timer(void *arg)
3041 struct bridge_softc *sc = arg;
3043 BRIDGE_RT_LOCK_ASSERT(sc);
3045 /* Destruction of rtnodes requires a proper vnet context */
3046 CURVNET_SET(sc->sc_ifp->if_vnet);
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);
3058 * Perform an aging cycle.
3061 bridge_rtage(struct bridge_softc *sc)
3063 struct bridge_rtnode *brt, *nbrt;
3065 BRIDGE_RT_LOCK_ASSERT(sc);
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);
3078 * Remove all dynamic addresses from the bridge.
3081 bridge_rtflush(struct bridge_softc *sc, int full)
3083 struct bridge_rtnode *brt, *nbrt;
3085 BRIDGE_RT_LOCK_ASSERT(sc);
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);
3096 * Remove an address from the table.
3099 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
3101 struct bridge_rtnode *brt;
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.
3110 while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
3111 bridge_rtnode_destroy(sc, brt);
3115 BRIDGE_RT_UNLOCK(sc);
3117 return (found ? 0 : ENOENT);
3123 * Delete routes to a speicifc member interface.
3126 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
3128 struct bridge_rtnode *brt, *nbrt;
3130 BRIDGE_RT_LOCK_ASSERT(sc);
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);
3140 * bridge_rtable_init:
3142 * Initialize the route table for this bridge.
3145 bridge_rtable_init(struct bridge_softc *sc)
3149 sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
3150 M_DEVBUF, M_WAITOK);
3152 for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
3153 CK_LIST_INIT(&sc->sc_rthash[i]);
3155 sc->sc_rthash_key = arc4random();
3156 CK_LIST_INIT(&sc->sc_rtlist);
3160 * bridge_rtable_fini:
3162 * Deconstruct the route table for this bridge.
3165 bridge_rtable_fini(struct bridge_softc *sc)
3168 KASSERT(sc->sc_brtcnt == 0,
3169 ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
3170 free(sc->sc_rthash, M_DEVBUF);
3174 * The following hash function is adapted from "Hash Functions" by Bob Jenkins
3175 * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
3177 #define mix(a, b, c) \
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)
3190 static __inline uint32_t
3191 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
3193 uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
3204 return (c & BRIDGE_RTHASH_MASK);
3210 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
3214 for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
3215 d = ((int)a[i]) - ((int)b[i]);
3222 * bridge_rtnode_lookup:
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.
3227 static struct bridge_rtnode *
3228 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
3230 struct bridge_rtnode *brt;
3234 BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(sc);
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))
3249 * bridge_rtnode_insert:
3251 * Insert the specified bridge node into the route table. We
3252 * assume the entry is not already in the table.
3255 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
3257 struct bridge_rtnode *lbrt;
3261 BRIDGE_RT_LOCK_ASSERT(sc);
3263 hash = bridge_rthash(sc, brt->brt_addr);
3265 lbrt = CK_LIST_FIRST(&sc->sc_rthash[hash]);
3267 CK_LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
3272 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
3273 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
3276 CK_LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
3279 if (CK_LIST_NEXT(lbrt, brt_hash) == NULL) {
3280 CK_LIST_INSERT_AFTER(lbrt, brt, brt_hash);
3283 lbrt = CK_LIST_NEXT(lbrt, brt_hash);
3284 } while (lbrt != NULL);
3287 panic("bridge_rtnode_insert: impossible");
3291 CK_LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
3298 bridge_rtnode_destroy_cb(struct epoch_context *ctx)
3300 struct bridge_rtnode *brt;
3302 brt = __containerof(ctx, struct bridge_rtnode, brt_epoch_ctx);
3304 CURVNET_SET(brt->brt_vnet);
3305 uma_zfree(V_bridge_rtnode_zone, brt);
3310 * bridge_rtnode_destroy:
3312 * Destroy a bridge rtnode.
3315 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3317 BRIDGE_RT_LOCK_ASSERT(sc);
3319 CK_LIST_REMOVE(brt, brt_hash);
3321 CK_LIST_REMOVE(brt, brt_list);
3323 brt->brt_dst->bif_addrcnt--;
3325 NET_EPOCH_CALL(bridge_rtnode_destroy_cb, &brt->brt_epoch_ctx);
3329 * bridge_rtable_expire:
3331 * Set the expiry time for all routes on an interface.
3334 bridge_rtable_expire(struct ifnet *ifp, int age)
3336 struct bridge_softc *sc = ifp->if_bridge;
3337 struct bridge_rtnode *brt;
3339 CURVNET_SET(ifp->if_vnet);
3343 * If the age is zero then flush, otherwise set all the expiry times to
3344 * age for the interface
3347 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
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;
3357 BRIDGE_RT_UNLOCK(sc);
3362 * bridge_state_change:
3364 * Callback from the bridgestp code when a port changes states.
3367 bridge_state_change(struct ifnet *ifp, int state)
3369 struct bridge_softc *sc = ifp->if_bridge;
3370 static const char *stpstates[] = {
3379 CURVNET_SET(ifp->if_vnet);
3381 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
3382 sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
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
3393 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3395 int snap, error, i, hlen;
3396 struct ether_header *eh1, eh2;
3399 u_int16_t ether_type;
3403 error = -1; /* Default error if not error == 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__));
3410 if (V_pfil_bridge == 0 && V_pfil_member == 0 && V_pfil_ipfw == 0)
3411 return (0); /* filtering is disabled */
3413 i = min((*mp)->m_pkthdr.len, max_protohdr);
3414 if ((*mp)->m_len < i) {
3415 *mp = m_pullup(*mp, i);
3417 printf("%s: m_pullup failed\n", __func__);
3422 eh1 = mtod(*mp, struct ether_header *);
3423 ether_type = ntohs(eh1->ether_type);
3426 * Check for SNAP/LLC.
3428 if (ether_type < ETHERMTU) {
3429 struct llc *llc2 = (struct llc *)(eh1 + 1);
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);
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*
3449 switch (ether_type) {
3451 case ETHERTYPE_REVARP:
3452 if (V_pfil_ipfw_arp == 0)
3453 return (0); /* Automatically pass */
3458 case ETHERTYPE_IPV6:
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.
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)) {
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);
3486 /* Strip off snap header, if present */
3488 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3489 m_adj(*mp, sizeof(struct llc));
3493 * Check the IP header for alignment and errors
3495 if (dir == PFIL_IN) {
3496 switch (ether_type) {
3498 error = bridge_ip_checkbasic(mp);
3501 case ETHERTYPE_IPV6:
3502 error = bridge_ip6_checkbasic(mp);
3515 * Run the packet through pfil
3518 switch (ether_type) {
3521 * Run pfil on the member interface and the bridge, both can
3522 * be skipped by clearing pfil_member or pfil_bridge.
3525 * in_if -> bridge_if -> out_if
3527 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL && (rv =
3528 pfil_mbuf_out(V_inet_pfil_head, mp, bifp, NULL)) !=
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)
3540 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv =
3541 pfil_mbuf_in(V_inet_pfil_head, mp, bifp, NULL)) !=
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,
3557 /* Recalculate the ip checksum. */
3558 ip = mtod(*mp, struct ip *);
3559 hlen = ip->ip_hl << 2;
3560 if (hlen < sizeof(struct ip))
3562 if (hlen > (*mp)->m_len) {
3563 if ((*mp = m_pullup(*mp, hlen)) == NULL)
3565 ip = mtod(*mp, struct ip *);
3570 if (hlen == sizeof(struct ip))
3571 ip->ip_sum = in_cksum_hdr(ip);
3573 ip->ip_sum = in_cksum(*mp, hlen);
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)) !=
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)
3591 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv =
3592 pfil_mbuf_in(V_inet6_pfil_head, mp, bifp, NULL)) !=
3611 * Finally, put everything back the way it was and return
3614 M_PREPEND(*mp, sizeof(struct llc), M_NOWAIT);
3617 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3620 M_PREPEND(*mp, ETHER_HDR_LEN, M_NOWAIT);
3623 bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
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
3641 * XXX Should we update ipstat here, or not?
3642 * XXX Right now we update ipstat but not
3646 bridge_ip_checkbasic(struct mbuf **mp)
3648 struct mbuf *m = *mp;
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);
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);
3669 ip = mtod(m, struct ip *);
3670 if (ip == NULL) goto bad;
3672 if (ip->ip_v != IPVERSION) {
3673 KMOD_IPSTAT_INC(ips_badvers);
3676 hlen = ip->ip_hl << 2;
3677 if (hlen < sizeof(struct ip)) { /* minimum header length */
3678 KMOD_IPSTAT_INC(ips_badhlen);
3681 if (hlen > m->m_len) {
3682 if ((m = m_pullup(m, hlen)) == NULL) {
3683 KMOD_IPSTAT_INC(ips_badhlen);
3686 ip = mtod(m, struct ip *);
3687 if (ip == NULL) goto bad;
3690 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3691 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3693 if (hlen == sizeof(struct ip)) {
3694 sum = in_cksum_hdr(ip);
3696 sum = in_cksum(m, hlen);
3700 KMOD_IPSTAT_INC(ips_badsum);
3704 /* Retrieve the packet length. */
3705 len = ntohs(ip->ip_len);
3708 * Check for additional length bogosity
3711 KMOD_IPSTAT_INC(ips_badlen);
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.
3720 if (m->m_pkthdr.len < len) {
3721 KMOD_IPSTAT_INC(ips_tooshort);
3725 /* Checks out, proceed */
3736 * Same as above, but for IPv6.
3737 * Cut-and-pasted from ip6_input.c.
3738 * XXX Should we update ip6stat, or not?
3741 bridge_ip6_checkbasic(struct mbuf **mp)
3743 struct mbuf *m = *mp;
3744 struct ip6_hdr *ip6;
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.
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);
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);
3770 ip6 = mtod(m, struct ip6_hdr *);
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);
3778 /* Checks out, proceed */
3791 * Fragment mbuf chain in multiple packets and prepend ethernet header.
3794 bridge_fragment(struct ifnet *ifp, struct mbuf **mp, struct ether_header *eh,
3795 int snap, struct llc *llc)
3797 struct mbuf *m = *mp, *nextpkt = NULL, *mprev = NULL, *mcur = NULL;
3801 if (m->m_len < sizeof(struct ip) &&
3802 (m = m_pullup(m, sizeof(struct ip))) == NULL)
3804 ip = mtod(m, struct ip *);
3806 m->m_pkthdr.csum_flags |= CSUM_IP;
3807 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist);
3812 * Walk the chain and re-add the Ethernet header for
3815 for (mcur = m; mcur; mcur = mcur->m_nextpkt) {
3816 nextpkt = mcur->m_nextpkt;
3817 mcur->m_nextpkt = NULL;
3819 M_PREPEND(mcur, sizeof(struct llc), M_NOWAIT);
3823 mprev->m_nextpkt = nextpkt;
3826 bcopy(llc, mtod(mcur, caddr_t),sizeof(struct llc));
3829 M_PREPEND(mcur, ETHER_HDR_LEN, M_NOWAIT);
3833 mprev->m_nextpkt = nextpkt;
3836 bcopy(eh, mtod(mcur, caddr_t), ETHER_HDR_LEN);
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
3843 mcur->m_nextpkt = nextpkt;
3845 mprev->m_nextpkt = mcur;
3847 /* The first mbuf in the original chain needs to be
3854 KMOD_IPSTAT_INC(ips_fragmented);
3858 for (mcur = *mp; mcur; mcur = m) { /* droping the full packet chain */
3859 m = mcur->m_nextpkt;
3866 bridge_linkstate(struct ifnet *ifp)
3868 struct bridge_softc *sc = ifp->if_bridge;
3869 struct bridge_iflist *bif;
3870 struct epoch_tracker et;
3872 NET_EPOCH_ENTER(et);
3874 bif = bridge_lookup_member_if(sc, ifp);
3879 bridge_linkcheck(sc);
3881 bstp_linkstate(&bif->bif_stp);
3887 bridge_linkcheck(struct bridge_softc *sc)
3889 struct bridge_iflist *bif;
3890 int new_link, hasls;
3892 BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
3894 new_link = LINK_STATE_DOWN;
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)
3900 if (bif->bif_ifp->if_link_state == LINK_STATE_UP) {
3901 new_link = LINK_STATE_UP;
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;
3909 if_link_state_change(sc->sc_ifp, new_link);