]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/blob - sys/net/if_bridge.c
Mark more nodes as CTLFLAG_MPSAFE or CTLFLAG_NEEDGIANT (17 of many)
[FreeBSD/FreeBSD.git] / sys / net / if_bridge.c
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/pfil.h>
114 #include <net/vnet.h>
115
116 #include <netinet/in.h>
117 #include <netinet/in_systm.h>
118 #include <netinet/in_var.h>
119 #include <netinet/ip.h>
120 #include <netinet/ip_var.h>
121 #ifdef INET6
122 #include <netinet/ip6.h>
123 #include <netinet6/ip6_var.h>
124 #include <netinet6/in6_ifattach.h>
125 #endif
126 #if defined(INET) || defined(INET6)
127 #include <netinet/ip_carp.h>
128 #endif
129 #include <machine/in_cksum.h>
130 #include <netinet/if_ether.h>
131 #include <net/bridgestp.h>
132 #include <net/if_bridgevar.h>
133 #include <net/if_llc.h>
134 #include <net/if_vlan_var.h>
135
136 #include <net/route.h>
137
138 #ifdef INET6
139 /*
140  * XXX: declare here to avoid to include many inet6 related files..
141  * should be more generalized?
142  */
143 extern void     nd6_setmtu(struct ifnet *);
144 #endif
145
146 /*
147  * Size of the route hash table.  Must be a power of two.
148  */
149 #ifndef BRIDGE_RTHASH_SIZE
150 #define BRIDGE_RTHASH_SIZE              1024
151 #endif
152
153 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
154
155 /*
156  * Default maximum number of addresses to cache.
157  */
158 #ifndef BRIDGE_RTABLE_MAX
159 #define BRIDGE_RTABLE_MAX               2000
160 #endif
161
162 /*
163  * Timeout (in seconds) for entries learned dynamically.
164  */
165 #ifndef BRIDGE_RTABLE_TIMEOUT
166 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
167 #endif
168
169 /*
170  * Number of seconds between walks of the route list.
171  */
172 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
173 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
174 #endif
175
176 /*
177  * List of capabilities to possibly mask on the member interface.
178  */
179 #define BRIDGE_IFCAPS_MASK              (IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM|\
180                                          IFCAP_TXCSUM_IPV6)
181
182 /*
183  * List of capabilities to strip
184  */
185 #define BRIDGE_IFCAPS_STRIP             IFCAP_LRO
186
187 /*
188  * Bridge locking
189  */
190 #define BRIDGE_LOCK_INIT(_sc)           do {                    \
191         mtx_init(&(_sc)->sc_mtx, "if_bridge", NULL, MTX_DEF);   \
192         cv_init(&(_sc)->sc_cv, "if_bridge_cv");                 \
193 } while (0)
194 #define BRIDGE_LOCK_DESTROY(_sc)        do {    \
195         mtx_destroy(&(_sc)->sc_mtx);            \
196         cv_destroy(&(_sc)->sc_cv);              \
197 } while (0)
198 #define BRIDGE_LOCK(_sc)                mtx_lock(&(_sc)->sc_mtx)
199 #define BRIDGE_UNLOCK(_sc)              mtx_unlock(&(_sc)->sc_mtx)
200 #define BRIDGE_LOCK_ASSERT(_sc)         mtx_assert(&(_sc)->sc_mtx, MA_OWNED)
201 #define BRIDGE_UNLOCK_ASSERT(_sc)       mtx_assert(&(_sc)->sc_mtx, MA_NOTOWNED)
202 #define BRIDGE_LOCK2REF(_sc, _err)      do {    \
203         mtx_assert(&(_sc)->sc_mtx, MA_OWNED);   \
204         if ((_sc)->sc_iflist_xcnt > 0)          \
205                 (_err) = EBUSY;                 \
206         else                                    \
207                 (_sc)->sc_iflist_ref++;         \
208         mtx_unlock(&(_sc)->sc_mtx);             \
209 } while (0)
210 #define BRIDGE_UNREF(_sc)               do {                            \
211         mtx_lock(&(_sc)->sc_mtx);                                       \
212         (_sc)->sc_iflist_ref--;                                         \
213         if (((_sc)->sc_iflist_xcnt > 0) && ((_sc)->sc_iflist_ref == 0)) \
214                 cv_broadcast(&(_sc)->sc_cv);                            \
215         mtx_unlock(&(_sc)->sc_mtx);                                     \
216 } while (0)
217 #define BRIDGE_XLOCK(_sc)               do {            \
218         mtx_assert(&(_sc)->sc_mtx, MA_OWNED);           \
219         (_sc)->sc_iflist_xcnt++;                        \
220         while ((_sc)->sc_iflist_ref > 0)                \
221                 cv_wait(&(_sc)->sc_cv, &(_sc)->sc_mtx); \
222 } while (0)
223 #define BRIDGE_XDROP(_sc)               do {    \
224         mtx_assert(&(_sc)->sc_mtx, MA_OWNED);   \
225         (_sc)->sc_iflist_xcnt--;                \
226 } while (0)
227
228 /*
229  * Bridge interface list entry.
230  */
231 struct bridge_iflist {
232         LIST_ENTRY(bridge_iflist) bif_next;
233         struct ifnet            *bif_ifp;       /* member if */
234         struct bstp_port        bif_stp;        /* STP state */
235         uint32_t                bif_flags;      /* member if flags */
236         int                     bif_savedcaps;  /* saved capabilities */
237         uint32_t                bif_addrmax;    /* max # of addresses */
238         uint32_t                bif_addrcnt;    /* cur. # of addresses */
239         uint32_t                bif_addrexceeded;/* # of address violations */
240 };
241
242 /*
243  * Bridge route node.
244  */
245 struct bridge_rtnode {
246         LIST_ENTRY(bridge_rtnode) brt_hash;     /* hash table linkage */
247         LIST_ENTRY(bridge_rtnode) brt_list;     /* list linkage */
248         struct bridge_iflist    *brt_dst;       /* destination if */
249         unsigned long           brt_expire;     /* expiration time */
250         uint8_t                 brt_flags;      /* address flags */
251         uint8_t                 brt_addr[ETHER_ADDR_LEN];
252         uint16_t                brt_vlan;       /* vlan id */
253 };
254 #define brt_ifp                 brt_dst->bif_ifp
255
256 /*
257  * Software state for each bridge.
258  */
259 struct bridge_softc {
260         struct ifnet            *sc_ifp;        /* make this an interface */
261         LIST_ENTRY(bridge_softc) sc_list;
262         struct mtx              sc_mtx;
263         struct cv               sc_cv;
264         uint32_t                sc_brtmax;      /* max # of addresses */
265         uint32_t                sc_brtcnt;      /* cur. # of addresses */
266         uint32_t                sc_brttimeout;  /* rt timeout in seconds */
267         struct callout          sc_brcallout;   /* bridge callout */
268         uint32_t                sc_iflist_ref;  /* refcount for sc_iflist */
269         uint32_t                sc_iflist_xcnt; /* refcount for sc_iflist */
270         LIST_HEAD(, bridge_iflist) sc_iflist;   /* member interface list */
271         LIST_HEAD(, bridge_rtnode) *sc_rthash;  /* our forwarding table */
272         LIST_HEAD(, bridge_rtnode) sc_rtlist;   /* list version of above */
273         uint32_t                sc_rthash_key;  /* key for hash */
274         LIST_HEAD(, bridge_iflist) sc_spanlist; /* span ports list */
275         struct bstp_state       sc_stp;         /* STP state */
276         uint32_t                sc_brtexceeded; /* # of cache drops */
277         struct ifnet            *sc_ifaddr;     /* member mac copied from */
278         struct ether_addr       sc_defaddr;     /* Default MAC address */
279 };
280
281 VNET_DEFINE_STATIC(struct mtx, bridge_list_mtx);
282 #define V_bridge_list_mtx       VNET(bridge_list_mtx)
283 static eventhandler_tag bridge_detach_cookie;
284
285 int     bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
286
287 VNET_DEFINE_STATIC(uma_zone_t, bridge_rtnode_zone);
288 #define V_bridge_rtnode_zone    VNET(bridge_rtnode_zone)
289
290 static int      bridge_clone_create(struct if_clone *, int, caddr_t);
291 static void     bridge_clone_destroy(struct ifnet *);
292
293 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t);
294 static void     bridge_mutecaps(struct bridge_softc *);
295 static void     bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
296                     int);
297 static void     bridge_ifdetach(void *arg __unused, struct ifnet *);
298 static void     bridge_init(void *);
299 static void     bridge_dummynet(struct mbuf *, struct ifnet *);
300 static void     bridge_stop(struct ifnet *, int);
301 static int      bridge_transmit(struct ifnet *, struct mbuf *);
302 static void     bridge_qflush(struct ifnet *);
303 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
304 static int      bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
305                     struct rtentry *);
306 static int      bridge_enqueue(struct bridge_softc *, struct ifnet *,
307                     struct mbuf *);
308 static void     bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
309
310 static void     bridge_forward(struct bridge_softc *, struct bridge_iflist *,
311                     struct mbuf *m);
312
313 static void     bridge_timer(void *);
314
315 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
316                     struct mbuf *, int);
317 static void     bridge_span(struct bridge_softc *, struct mbuf *);
318
319 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
320                     uint16_t, struct bridge_iflist *, int, uint8_t);
321 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
322                     uint16_t);
323 static void     bridge_rttrim(struct bridge_softc *);
324 static void     bridge_rtage(struct bridge_softc *);
325 static void     bridge_rtflush(struct bridge_softc *, int);
326 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
327                     uint16_t);
328
329 static void     bridge_rtable_init(struct bridge_softc *);
330 static void     bridge_rtable_fini(struct bridge_softc *);
331
332 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
333 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
334                     const uint8_t *, uint16_t);
335 static int      bridge_rtnode_insert(struct bridge_softc *,
336                     struct bridge_rtnode *);
337 static void     bridge_rtnode_destroy(struct bridge_softc *,
338                     struct bridge_rtnode *);
339 static void     bridge_rtable_expire(struct ifnet *, int);
340 static void     bridge_state_change(struct ifnet *, int);
341
342 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
343                     const char *name);
344 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
345                     struct ifnet *ifp);
346 static void     bridge_delete_member(struct bridge_softc *,
347                     struct bridge_iflist *, int);
348 static void     bridge_delete_span(struct bridge_softc *,
349                     struct bridge_iflist *);
350
351 static int      bridge_ioctl_add(struct bridge_softc *, void *);
352 static int      bridge_ioctl_del(struct bridge_softc *, void *);
353 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
354 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
355 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
356 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
357 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
358 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
359 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
360 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
361 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
362 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
363 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
364 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
365 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
366 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
367 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
368 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
369 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
370 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
371 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
372 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
373 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
374 static int      bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
375 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
376 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
377 static int      bridge_ioctl_gbparam(struct bridge_softc *, void *);
378 static int      bridge_ioctl_grte(struct bridge_softc *, void *);
379 static int      bridge_ioctl_gifsstp(struct bridge_softc *, void *);
380 static int      bridge_ioctl_sproto(struct bridge_softc *, void *);
381 static int      bridge_ioctl_stxhc(struct bridge_softc *, void *);
382 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
383                     int);
384 static int      bridge_ip_checkbasic(struct mbuf **mp);
385 #ifdef INET6
386 static int      bridge_ip6_checkbasic(struct mbuf **mp);
387 #endif /* INET6 */
388 static int      bridge_fragment(struct ifnet *, struct mbuf **mp,
389                     struct ether_header *, int, struct llc *);
390 static void     bridge_linkstate(struct ifnet *ifp);
391 static void     bridge_linkcheck(struct bridge_softc *sc);
392
393
394 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
395 #define VLANTAGOF(_m)   \
396     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : 1
397
398 static struct bstp_cb_ops bridge_ops = {
399         .bcb_state = bridge_state_change,
400         .bcb_rtage = bridge_rtable_expire
401 };
402
403 SYSCTL_DECL(_net_link);
404 static SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
405     "Bridge");
406
407 /* only pass IP[46] packets when pfil is enabled */
408 VNET_DEFINE_STATIC(int, pfil_onlyip) = 1;
409 #define V_pfil_onlyip   VNET(pfil_onlyip)
410 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip,
411     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_onlyip), 0,
412     "Only pass IP packets when pfil is enabled");
413
414 /* run pfil hooks on the bridge interface */
415 VNET_DEFINE_STATIC(int, pfil_bridge) = 1;
416 #define V_pfil_bridge   VNET(pfil_bridge)
417 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge,
418     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_bridge), 0,
419     "Packet filter on the bridge interface");
420
421 /* layer2 filter with ipfw */
422 VNET_DEFINE_STATIC(int, pfil_ipfw);
423 #define V_pfil_ipfw     VNET(pfil_ipfw)
424
425 /* layer2 ARP filter with ipfw */
426 VNET_DEFINE_STATIC(int, pfil_ipfw_arp);
427 #define V_pfil_ipfw_arp VNET(pfil_ipfw_arp)
428 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp,
429     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_ipfw_arp), 0,
430     "Filter ARP packets through IPFW layer2");
431
432 /* run pfil hooks on the member interface */
433 VNET_DEFINE_STATIC(int, pfil_member) = 1;
434 #define V_pfil_member   VNET(pfil_member)
435 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member,
436     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_member), 0,
437     "Packet filter on the member interface");
438
439 /* run pfil hooks on the physical interface for locally destined packets */
440 VNET_DEFINE_STATIC(int, pfil_local_phys);
441 #define V_pfil_local_phys       VNET(pfil_local_phys)
442 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys,
443     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_local_phys), 0,
444     "Packet filter on the physical interface for locally destined packets");
445
446 /* log STP state changes */
447 VNET_DEFINE_STATIC(int, log_stp);
448 #define V_log_stp       VNET(log_stp)
449 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp,
450     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(log_stp), 0,
451     "Log STP state changes");
452
453 /* share MAC with first bridge member */
454 VNET_DEFINE_STATIC(int, bridge_inherit_mac);
455 #define V_bridge_inherit_mac    VNET(bridge_inherit_mac)
456 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac,
457     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(bridge_inherit_mac), 0,
458     "Inherit MAC address from the first bridge member");
459
460 VNET_DEFINE_STATIC(int, allow_llz_overlap) = 0;
461 #define V_allow_llz_overlap     VNET(allow_llz_overlap)
462 SYSCTL_INT(_net_link_bridge, OID_AUTO, allow_llz_overlap,
463     CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(allow_llz_overlap), 0,
464     "Allow overlap of link-local scope "
465     "zones of a bridge interface and the member interfaces");
466
467 struct bridge_control {
468         int     (*bc_func)(struct bridge_softc *, void *);
469         int     bc_argsize;
470         int     bc_flags;
471 };
472
473 #define BC_F_COPYIN             0x01    /* copy arguments in */
474 #define BC_F_COPYOUT            0x02    /* copy arguments out */
475 #define BC_F_SUSER              0x04    /* do super-user check */
476
477 const struct bridge_control bridge_control_table[] = {
478         { bridge_ioctl_add,             sizeof(struct ifbreq),
479           BC_F_COPYIN|BC_F_SUSER },
480         { bridge_ioctl_del,             sizeof(struct ifbreq),
481           BC_F_COPYIN|BC_F_SUSER },
482
483         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
484           BC_F_COPYIN|BC_F_COPYOUT },
485         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
486           BC_F_COPYIN|BC_F_SUSER },
487
488         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
489           BC_F_COPYIN|BC_F_SUSER },
490         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
491           BC_F_COPYOUT },
492
493         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
494           BC_F_COPYIN|BC_F_COPYOUT },
495         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
496           BC_F_COPYIN|BC_F_COPYOUT },
497
498         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
499           BC_F_COPYIN|BC_F_SUSER },
500
501         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
502           BC_F_COPYIN|BC_F_SUSER },
503         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
504           BC_F_COPYOUT },
505
506         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
507           BC_F_COPYIN|BC_F_SUSER },
508
509         { bridge_ioctl_flush,           sizeof(struct ifbreq),
510           BC_F_COPYIN|BC_F_SUSER },
511
512         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
513           BC_F_COPYOUT },
514         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
515           BC_F_COPYIN|BC_F_SUSER },
516
517         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
518           BC_F_COPYOUT },
519         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
520           BC_F_COPYIN|BC_F_SUSER },
521
522         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
523           BC_F_COPYOUT },
524         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
525           BC_F_COPYIN|BC_F_SUSER },
526
527         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
528           BC_F_COPYOUT },
529         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
530           BC_F_COPYIN|BC_F_SUSER },
531
532         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
533           BC_F_COPYIN|BC_F_SUSER },
534
535         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
536           BC_F_COPYIN|BC_F_SUSER },
537
538         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
539           BC_F_COPYIN|BC_F_SUSER },
540         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
541           BC_F_COPYIN|BC_F_SUSER },
542
543         { bridge_ioctl_gbparam,         sizeof(struct ifbropreq),
544           BC_F_COPYOUT },
545
546         { bridge_ioctl_grte,            sizeof(struct ifbrparam),
547           BC_F_COPYOUT },
548
549         { bridge_ioctl_gifsstp,         sizeof(struct ifbpstpconf),
550           BC_F_COPYIN|BC_F_COPYOUT },
551
552         { bridge_ioctl_sproto,          sizeof(struct ifbrparam),
553           BC_F_COPYIN|BC_F_SUSER },
554
555         { bridge_ioctl_stxhc,           sizeof(struct ifbrparam),
556           BC_F_COPYIN|BC_F_SUSER },
557
558         { bridge_ioctl_sifmaxaddr,      sizeof(struct ifbreq),
559           BC_F_COPYIN|BC_F_SUSER },
560
561 };
562 const int bridge_control_table_size = nitems(bridge_control_table);
563
564 VNET_DEFINE_STATIC(LIST_HEAD(, bridge_softc), bridge_list);
565 #define V_bridge_list   VNET(bridge_list)
566 #define BRIDGE_LIST_LOCK_INIT(x)        mtx_init(&V_bridge_list_mtx,    \
567                                             "if_bridge list", NULL, MTX_DEF)
568 #define BRIDGE_LIST_LOCK_DESTROY(x)     mtx_destroy(&V_bridge_list_mtx)
569 #define BRIDGE_LIST_LOCK(x)             mtx_lock(&V_bridge_list_mtx)
570 #define BRIDGE_LIST_UNLOCK(x)           mtx_unlock(&V_bridge_list_mtx)
571
572 VNET_DEFINE_STATIC(struct if_clone *, bridge_cloner);
573 #define V_bridge_cloner VNET(bridge_cloner)
574
575 static const char bridge_name[] = "bridge";
576
577 static void
578 vnet_bridge_init(const void *unused __unused)
579 {
580
581         V_bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
582             sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
583             UMA_ALIGN_PTR, 0);
584         BRIDGE_LIST_LOCK_INIT();
585         LIST_INIT(&V_bridge_list);
586         V_bridge_cloner = if_clone_simple(bridge_name,
587             bridge_clone_create, bridge_clone_destroy, 0);
588 }
589 VNET_SYSINIT(vnet_bridge_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
590     vnet_bridge_init, NULL);
591
592 static void
593 vnet_bridge_uninit(const void *unused __unused)
594 {
595
596         if_clone_detach(V_bridge_cloner);
597         V_bridge_cloner = NULL;
598         BRIDGE_LIST_LOCK_DESTROY();
599         uma_zdestroy(V_bridge_rtnode_zone);
600 }
601 VNET_SYSUNINIT(vnet_bridge_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY,
602     vnet_bridge_uninit, NULL);
603
604 static int
605 bridge_modevent(module_t mod, int type, void *data)
606 {
607
608         switch (type) {
609         case MOD_LOAD:
610                 bridge_dn_p = bridge_dummynet;
611                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
612                     ifnet_departure_event, bridge_ifdetach, NULL,
613                     EVENTHANDLER_PRI_ANY);
614                 break;
615         case MOD_UNLOAD:
616                 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
617                     bridge_detach_cookie);
618                 bridge_dn_p = NULL;
619                 break;
620         default:
621                 return (EOPNOTSUPP);
622         }
623         return (0);
624 }
625
626 static moduledata_t bridge_mod = {
627         "if_bridge",
628         bridge_modevent,
629         0
630 };
631
632 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
633 MODULE_VERSION(if_bridge, 1);
634 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
635
636 /*
637  * handler for net.link.bridge.ipfw
638  */
639 static int
640 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
641 {
642         int enable = V_pfil_ipfw;
643         int error;
644
645         error = sysctl_handle_int(oidp, &enable, 0, req);
646         enable &= 1;
647
648         if (enable != V_pfil_ipfw) {
649                 V_pfil_ipfw = enable;
650
651                 /*
652                  * Disable pfil so that ipfw doesnt run twice, if the user
653                  * really wants both then they can re-enable pfil_bridge and/or
654                  * pfil_member. Also allow non-ip packets as ipfw can filter by
655                  * layer2 type.
656                  */
657                 if (V_pfil_ipfw) {
658                         V_pfil_onlyip = 0;
659                         V_pfil_bridge = 0;
660                         V_pfil_member = 0;
661                 }
662         }
663
664         return (error);
665 }
666 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw,
667     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_VNET | CTLFLAG_NEEDGIANT,
668     &VNET_NAME(pfil_ipfw), 0, &sysctl_pfil_ipfw, "I",
669     "Layer2 filter with IPFW");
670
671 /*
672  * bridge_clone_create:
673  *
674  *      Create a new bridge instance.
675  */
676 static int
677 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t params)
678 {
679         struct bridge_softc *sc, *sc2;
680         struct ifnet *bifp, *ifp;
681         int fb, retry;
682         unsigned long hostid;
683
684         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
685         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
686         if (ifp == NULL) {
687                 free(sc, M_DEVBUF);
688                 return (ENOSPC);
689         }
690
691         BRIDGE_LOCK_INIT(sc);
692         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
693         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
694
695         /* Initialize our routing table. */
696         bridge_rtable_init(sc);
697
698         callout_init_mtx(&sc->sc_brcallout, &sc->sc_mtx, 0);
699
700         LIST_INIT(&sc->sc_iflist);
701         LIST_INIT(&sc->sc_spanlist);
702
703         ifp->if_softc = sc;
704         if_initname(ifp, bridge_name, unit);
705         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
706         ifp->if_ioctl = bridge_ioctl;
707         ifp->if_transmit = bridge_transmit;
708         ifp->if_qflush = bridge_qflush;
709         ifp->if_init = bridge_init;
710         ifp->if_type = IFT_BRIDGE;
711
712         /*
713          * Generate an ethernet address with a locally administered address.
714          *
715          * Since we are using random ethernet addresses for the bridge, it is
716          * possible that we might have address collisions, so make sure that
717          * this hardware address isn't already in use on another bridge.
718          * The first try uses the hostid and falls back to arc4rand().
719          */
720         fb = 0;
721         getcredhostid(curthread->td_ucred, &hostid);
722         do {
723                 if (fb || hostid == 0) {
724                         ether_gen_addr(ifp, &sc->sc_defaddr);
725                 } else {
726                         sc->sc_defaddr.octet[0] = 0x2;
727                         sc->sc_defaddr.octet[1] = (hostid >> 24) & 0xff;
728                         sc->sc_defaddr.octet[2] = (hostid >> 16) & 0xff;
729                         sc->sc_defaddr.octet[3] = (hostid >> 8 ) & 0xff;
730                         sc->sc_defaddr.octet[4] =  hostid        & 0xff;
731                         sc->sc_defaddr.octet[5] = ifp->if_dunit & 0xff;
732                 }
733
734                 fb = 1;
735                 retry = 0;
736                 BRIDGE_LIST_LOCK();
737                 LIST_FOREACH(sc2, &V_bridge_list, sc_list) {
738                         bifp = sc2->sc_ifp;
739                         if (memcmp(sc->sc_defaddr.octet,
740                             IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0) {
741                                 retry = 1;
742                                 break;
743                         }
744                 }
745                 BRIDGE_LIST_UNLOCK();
746         } while (retry == 1);
747
748         bstp_attach(&sc->sc_stp, &bridge_ops);
749         ether_ifattach(ifp, sc->sc_defaddr.octet);
750         /* Now undo some of the damage... */
751         ifp->if_baudrate = 0;
752         ifp->if_type = IFT_BRIDGE;
753
754         BRIDGE_LIST_LOCK();
755         LIST_INSERT_HEAD(&V_bridge_list, sc, sc_list);
756         BRIDGE_LIST_UNLOCK();
757
758         return (0);
759 }
760
761 /*
762  * bridge_clone_destroy:
763  *
764  *      Destroy a bridge instance.
765  */
766 static void
767 bridge_clone_destroy(struct ifnet *ifp)
768 {
769         struct bridge_softc *sc = ifp->if_softc;
770         struct bridge_iflist *bif;
771
772         BRIDGE_LOCK(sc);
773
774         bridge_stop(ifp, 1);
775         ifp->if_flags &= ~IFF_UP;
776
777         while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
778                 bridge_delete_member(sc, bif, 0);
779
780         while ((bif = LIST_FIRST(&sc->sc_spanlist)) != NULL) {
781                 bridge_delete_span(sc, bif);
782         }
783
784         /* Tear down the routing table. */
785         bridge_rtable_fini(sc);
786
787         BRIDGE_UNLOCK(sc);
788
789         callout_drain(&sc->sc_brcallout);
790
791         BRIDGE_LIST_LOCK();
792         LIST_REMOVE(sc, sc_list);
793         BRIDGE_LIST_UNLOCK();
794
795         bstp_detach(&sc->sc_stp);
796         ether_ifdetach(ifp);
797         if_free(ifp);
798
799         BRIDGE_LOCK_DESTROY(sc);
800         free(sc, M_DEVBUF);
801 }
802
803 /*
804  * bridge_ioctl:
805  *
806  *      Handle a control request from the operator.
807  */
808 static int
809 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
810 {
811         struct bridge_softc *sc = ifp->if_softc;
812         struct ifreq *ifr = (struct ifreq *)data;
813         struct bridge_iflist *bif;
814         struct thread *td = curthread;
815         union {
816                 struct ifbreq ifbreq;
817                 struct ifbifconf ifbifconf;
818                 struct ifbareq ifbareq;
819                 struct ifbaconf ifbaconf;
820                 struct ifbrparam ifbrparam;
821                 struct ifbropreq ifbropreq;
822         } args;
823         struct ifdrv *ifd = (struct ifdrv *) data;
824         const struct bridge_control *bc;
825         int error = 0, oldmtu;
826
827         switch (cmd) {
828
829         case SIOCADDMULTI:
830         case SIOCDELMULTI:
831                 break;
832
833         case SIOCGDRVSPEC:
834         case SIOCSDRVSPEC:
835                 if (ifd->ifd_cmd >= bridge_control_table_size) {
836                         error = EINVAL;
837                         break;
838                 }
839                 bc = &bridge_control_table[ifd->ifd_cmd];
840
841                 if (cmd == SIOCGDRVSPEC &&
842                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
843                         error = EINVAL;
844                         break;
845                 }
846                 else if (cmd == SIOCSDRVSPEC &&
847                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
848                         error = EINVAL;
849                         break;
850                 }
851
852                 if (bc->bc_flags & BC_F_SUSER) {
853                         error = priv_check(td, PRIV_NET_BRIDGE);
854                         if (error)
855                                 break;
856                 }
857
858                 if (ifd->ifd_len != bc->bc_argsize ||
859                     ifd->ifd_len > sizeof(args)) {
860                         error = EINVAL;
861                         break;
862                 }
863
864                 bzero(&args, sizeof(args));
865                 if (bc->bc_flags & BC_F_COPYIN) {
866                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
867                         if (error)
868                                 break;
869                 }
870
871                 oldmtu = ifp->if_mtu;
872                 BRIDGE_LOCK(sc);
873                 error = (*bc->bc_func)(sc, &args);
874                 BRIDGE_UNLOCK(sc);
875                 if (error)
876                         break;
877
878                 /*
879                  * Bridge MTU may change during addition of the first port.
880                  * If it did, do network layer specific procedure.
881                  */
882                 if (ifp->if_mtu != oldmtu) {
883 #ifdef INET6
884                         nd6_setmtu(ifp);
885 #endif
886                         rt_updatemtu(ifp);
887                 }
888
889                 if (bc->bc_flags & BC_F_COPYOUT)
890                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
891
892                 break;
893
894         case SIOCSIFFLAGS:
895                 if (!(ifp->if_flags & IFF_UP) &&
896                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
897                         /*
898                          * If interface is marked down and it is running,
899                          * then stop and disable it.
900                          */
901                         BRIDGE_LOCK(sc);
902                         bridge_stop(ifp, 1);
903                         BRIDGE_UNLOCK(sc);
904                 } else if ((ifp->if_flags & IFF_UP) &&
905                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
906                         /*
907                          * If interface is marked up and it is stopped, then
908                          * start it.
909                          */
910                         (*ifp->if_init)(sc);
911                 }
912                 break;
913
914         case SIOCSIFMTU:
915                 if (ifr->ifr_mtu < 576) {
916                         error = EINVAL;
917                         break;
918                 }
919                 if (LIST_EMPTY(&sc->sc_iflist)) {
920                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
921                         break;
922                 }
923                 BRIDGE_LOCK(sc);
924                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
925                         if (bif->bif_ifp->if_mtu != ifr->ifr_mtu) {
926                                 log(LOG_NOTICE, "%s: invalid MTU: %u(%s)"
927                                     " != %d\n", sc->sc_ifp->if_xname,
928                                     bif->bif_ifp->if_mtu,
929                                     bif->bif_ifp->if_xname, ifr->ifr_mtu);
930                                 error = EINVAL;
931                                 break;
932                         }
933                 }
934                 if (!error)
935                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
936                 BRIDGE_UNLOCK(sc);
937                 break;
938         default:
939                 /*
940                  * drop the lock as ether_ioctl() will call bridge_start() and
941                  * cause the lock to be recursed.
942                  */
943                 error = ether_ioctl(ifp, cmd, data);
944                 break;
945         }
946
947         return (error);
948 }
949
950 /*
951  * bridge_mutecaps:
952  *
953  *      Clear or restore unwanted capabilities on the member interface
954  */
955 static void
956 bridge_mutecaps(struct bridge_softc *sc)
957 {
958         struct bridge_iflist *bif;
959         int enabled, mask;
960
961         /* Initial bitmask of capabilities to test */
962         mask = BRIDGE_IFCAPS_MASK;
963
964         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
965                 /* Every member must support it or its disabled */
966                 mask &= bif->bif_savedcaps;
967         }
968
969         BRIDGE_XLOCK(sc);
970         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
971                 enabled = bif->bif_ifp->if_capenable;
972                 enabled &= ~BRIDGE_IFCAPS_STRIP;
973                 /* strip off mask bits and enable them again if allowed */
974                 enabled &= ~BRIDGE_IFCAPS_MASK;
975                 enabled |= mask;
976                 BRIDGE_UNLOCK(sc);
977                 bridge_set_ifcap(sc, bif, enabled);
978                 BRIDGE_LOCK(sc);
979         }
980         BRIDGE_XDROP(sc);
981
982 }
983
984 static void
985 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
986 {
987         struct ifnet *ifp = bif->bif_ifp;
988         struct ifreq ifr;
989         int error, mask, stuck;
990
991         BRIDGE_UNLOCK_ASSERT(sc);
992
993         bzero(&ifr, sizeof(ifr));
994         ifr.ifr_reqcap = set;
995
996         if (ifp->if_capenable != set) {
997                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
998                 if (error)
999                         if_printf(sc->sc_ifp,
1000                             "error setting capabilities on %s: %d\n",
1001                             ifp->if_xname, error);
1002                 mask = BRIDGE_IFCAPS_MASK | BRIDGE_IFCAPS_STRIP;
1003                 stuck = ifp->if_capenable & mask & ~set;
1004                 if (stuck != 0)
1005                         if_printf(sc->sc_ifp,
1006                             "can't disable some capabilities on %s: 0x%x\n",
1007                             ifp->if_xname, stuck);
1008         }
1009 }
1010
1011 /*
1012  * bridge_lookup_member:
1013  *
1014  *      Lookup a bridge member interface.
1015  */
1016 static struct bridge_iflist *
1017 bridge_lookup_member(struct bridge_softc *sc, const char *name)
1018 {
1019         struct bridge_iflist *bif;
1020         struct ifnet *ifp;
1021
1022         BRIDGE_LOCK_ASSERT(sc);
1023
1024         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1025                 ifp = bif->bif_ifp;
1026                 if (strcmp(ifp->if_xname, name) == 0)
1027                         return (bif);
1028         }
1029
1030         return (NULL);
1031 }
1032
1033 /*
1034  * bridge_lookup_member_if:
1035  *
1036  *      Lookup a bridge member interface by ifnet*.
1037  */
1038 static struct bridge_iflist *
1039 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
1040 {
1041         struct bridge_iflist *bif;
1042
1043         BRIDGE_LOCK_ASSERT(sc);
1044
1045         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1046                 if (bif->bif_ifp == member_ifp)
1047                         return (bif);
1048         }
1049
1050         return (NULL);
1051 }
1052
1053 /*
1054  * bridge_delete_member:
1055  *
1056  *      Delete the specified member interface.
1057  */
1058 static void
1059 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
1060     int gone)
1061 {
1062         struct ifnet *ifs = bif->bif_ifp;
1063         struct ifnet *fif = NULL;
1064
1065         BRIDGE_LOCK_ASSERT(sc);
1066
1067         if (bif->bif_flags & IFBIF_STP)
1068                 bstp_disable(&bif->bif_stp);
1069
1070         ifs->if_bridge = NULL;
1071         BRIDGE_XLOCK(sc);
1072         LIST_REMOVE(bif, bif_next);
1073         BRIDGE_XDROP(sc);
1074
1075         /*
1076          * If removing the interface that gave the bridge its mac address, set
1077          * the mac address of the bridge to the address of the next member, or
1078          * to its default address if no members are left.
1079          */
1080         if (V_bridge_inherit_mac && sc->sc_ifaddr == ifs) {
1081                 if (LIST_EMPTY(&sc->sc_iflist)) {
1082                         bcopy(&sc->sc_defaddr,
1083                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1084                         sc->sc_ifaddr = NULL;
1085                 } else {
1086                         fif = LIST_FIRST(&sc->sc_iflist)->bif_ifp;
1087                         bcopy(IF_LLADDR(fif),
1088                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1089                         sc->sc_ifaddr = fif;
1090                 }
1091                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1092         }
1093
1094         bridge_linkcheck(sc);
1095         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
1096         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
1097         KASSERT(bif->bif_addrcnt == 0,
1098             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
1099
1100         ifs->if_bridge_output = NULL;
1101         ifs->if_bridge_input = NULL;
1102         ifs->if_bridge_linkstate = NULL;
1103         BRIDGE_UNLOCK(sc);
1104         if (!gone) {
1105                 switch (ifs->if_type) {
1106                 case IFT_ETHER:
1107                 case IFT_L2VLAN:
1108                         /*
1109                          * Take the interface out of promiscuous mode, but only
1110                          * if it was promiscuous in the first place. It might
1111                          * not be if we're in the bridge_ioctl_add() error path.
1112                          */
1113                         if (ifs->if_flags & IFF_PROMISC)
1114                                 (void) ifpromisc(ifs, 0);
1115                         break;
1116
1117                 case IFT_GIF:
1118                         break;
1119
1120                 default:
1121 #ifdef DIAGNOSTIC
1122                         panic("bridge_delete_member: impossible");
1123 #endif
1124                         break;
1125                 }
1126                 /* reneable any interface capabilities */
1127                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
1128         }
1129         bstp_destroy(&bif->bif_stp);    /* prepare to free */
1130         BRIDGE_LOCK(sc);
1131         free(bif, M_DEVBUF);
1132 }
1133
1134 /*
1135  * bridge_delete_span:
1136  *
1137  *      Delete the specified span interface.
1138  */
1139 static void
1140 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1141 {
1142         BRIDGE_LOCK_ASSERT(sc);
1143
1144         KASSERT(bif->bif_ifp->if_bridge == NULL,
1145             ("%s: not a span interface", __func__));
1146
1147         LIST_REMOVE(bif, bif_next);
1148         free(bif, M_DEVBUF);
1149 }
1150
1151 static int
1152 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1153 {
1154         struct ifbreq *req = arg;
1155         struct bridge_iflist *bif = NULL;
1156         struct ifnet *ifs;
1157         int error = 0;
1158
1159         ifs = ifunit(req->ifbr_ifsname);
1160         if (ifs == NULL)
1161                 return (ENOENT);
1162         if (ifs->if_ioctl == NULL)      /* must be supported */
1163                 return (EINVAL);
1164
1165         /* If it's in the span list, it can't be a member. */
1166         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1167                 if (ifs == bif->bif_ifp)
1168                         return (EBUSY);
1169
1170         if (ifs->if_bridge == sc)
1171                 return (EEXIST);
1172
1173         if (ifs->if_bridge != NULL)
1174                 return (EBUSY);
1175
1176         switch (ifs->if_type) {
1177         case IFT_ETHER:
1178         case IFT_L2VLAN:
1179         case IFT_GIF:
1180                 /* permitted interface types */
1181                 break;
1182         default:
1183                 return (EINVAL);
1184         }
1185
1186 #ifdef INET6
1187         /*
1188          * Two valid inet6 addresses with link-local scope must not be
1189          * on the parent interface and the member interfaces at the
1190          * same time.  This restriction is needed to prevent violation
1191          * of link-local scope zone.  Attempts to add a member
1192          * interface which has inet6 addresses when the parent has
1193          * inet6 triggers removal of all inet6 addresses on the member
1194          * interface.
1195          */
1196
1197         /* Check if the parent interface has a link-local scope addr. */
1198         if (V_allow_llz_overlap == 0 &&
1199             in6ifa_llaonifp(sc->sc_ifp) != NULL) {
1200                 /*
1201                  * If any, remove all inet6 addresses from the member
1202                  * interfaces.
1203                  */
1204                 BRIDGE_XLOCK(sc);
1205                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1206                         if (in6ifa_llaonifp(bif->bif_ifp)) {
1207                                 BRIDGE_UNLOCK(sc);
1208                                 in6_ifdetach(bif->bif_ifp);
1209                                 BRIDGE_LOCK(sc);
1210                                 if_printf(sc->sc_ifp,
1211                                     "IPv6 addresses on %s have been removed "
1212                                     "before adding it as a member to prevent "
1213                                     "IPv6 address scope violation.\n",
1214                                     bif->bif_ifp->if_xname);
1215                         }
1216                 }
1217                 BRIDGE_XDROP(sc);
1218                 if (in6ifa_llaonifp(ifs)) {
1219                         BRIDGE_UNLOCK(sc);
1220                         in6_ifdetach(ifs);
1221                         BRIDGE_LOCK(sc);
1222                         if_printf(sc->sc_ifp,
1223                             "IPv6 addresses on %s have been removed "
1224                             "before adding it as a member to prevent "
1225                             "IPv6 address scope violation.\n",
1226                             ifs->if_xname);
1227                 }
1228         }
1229 #endif
1230         /* Allow the first Ethernet member to define the MTU */
1231         if (LIST_EMPTY(&sc->sc_iflist))
1232                 sc->sc_ifp->if_mtu = ifs->if_mtu;
1233         else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1234                 if_printf(sc->sc_ifp, "invalid MTU: %u(%s) != %u\n",
1235                     ifs->if_mtu, ifs->if_xname, sc->sc_ifp->if_mtu);
1236                 return (EINVAL);
1237         }
1238
1239         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1240         if (bif == NULL)
1241                 return (ENOMEM);
1242
1243         bif->bif_ifp = ifs;
1244         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1245         bif->bif_savedcaps = ifs->if_capenable;
1246
1247         /*
1248          * Assign the interface's MAC address to the bridge if it's the first
1249          * member and the MAC address of the bridge has not been changed from
1250          * the default randomly generated one.
1251          */
1252         if (V_bridge_inherit_mac && LIST_EMPTY(&sc->sc_iflist) &&
1253             !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr.octet, ETHER_ADDR_LEN)) {
1254                 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1255                 sc->sc_ifaddr = ifs;
1256                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1257         }
1258
1259         ifs->if_bridge = sc;
1260         ifs->if_bridge_output = bridge_output;
1261         ifs->if_bridge_input = bridge_input;
1262         ifs->if_bridge_linkstate = bridge_linkstate;
1263         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1264         /*
1265          * XXX: XLOCK HERE!?!
1266          *
1267          * NOTE: insert_***HEAD*** should be safe for the traversals.
1268          */
1269         LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1270
1271         /* Set interface capabilities to the intersection set of all members */
1272         bridge_mutecaps(sc);
1273         bridge_linkcheck(sc);
1274
1275         /* Place the interface into promiscuous mode */
1276         switch (ifs->if_type) {
1277                 case IFT_ETHER:
1278                 case IFT_L2VLAN:
1279                         BRIDGE_UNLOCK(sc);
1280                         error = ifpromisc(ifs, 1);
1281                         BRIDGE_LOCK(sc);
1282                         break;
1283         }
1284
1285         if (error)
1286                 bridge_delete_member(sc, bif, 0);
1287         return (error);
1288 }
1289
1290 static int
1291 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1292 {
1293         struct ifbreq *req = arg;
1294         struct bridge_iflist *bif;
1295
1296         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1297         if (bif == NULL)
1298                 return (ENOENT);
1299
1300         bridge_delete_member(sc, bif, 0);
1301
1302         return (0);
1303 }
1304
1305 static int
1306 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1307 {
1308         struct ifbreq *req = arg;
1309         struct bridge_iflist *bif;
1310         struct bstp_port *bp;
1311
1312         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1313         if (bif == NULL)
1314                 return (ENOENT);
1315
1316         bp = &bif->bif_stp;
1317         req->ifbr_ifsflags = bif->bif_flags;
1318         req->ifbr_state = bp->bp_state;
1319         req->ifbr_priority = bp->bp_priority;
1320         req->ifbr_path_cost = bp->bp_path_cost;
1321         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1322         req->ifbr_proto = bp->bp_protover;
1323         req->ifbr_role = bp->bp_role;
1324         req->ifbr_stpflags = bp->bp_flags;
1325         req->ifbr_addrcnt = bif->bif_addrcnt;
1326         req->ifbr_addrmax = bif->bif_addrmax;
1327         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1328
1329         /* Copy STP state options as flags */
1330         if (bp->bp_operedge)
1331                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1332         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1333                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1334         if (bp->bp_ptp_link)
1335                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1336         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1337                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1338         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1339                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1340         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1341                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1342         return (0);
1343 }
1344
1345 static int
1346 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1347 {
1348         struct ifbreq *req = arg;
1349         struct bridge_iflist *bif;
1350         struct bstp_port *bp;
1351         int error;
1352
1353         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1354         if (bif == NULL)
1355                 return (ENOENT);
1356         bp = &bif->bif_stp;
1357
1358         if (req->ifbr_ifsflags & IFBIF_SPAN)
1359                 /* SPAN is readonly */
1360                 return (EINVAL);
1361
1362         if (req->ifbr_ifsflags & IFBIF_STP) {
1363                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1364                         error = bstp_enable(&bif->bif_stp);
1365                         if (error)
1366                                 return (error);
1367                 }
1368         } else {
1369                 if ((bif->bif_flags & IFBIF_STP) != 0)
1370                         bstp_disable(&bif->bif_stp);
1371         }
1372
1373         /* Pass on STP flags */
1374         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1375         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1376         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1377         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1378
1379         /* Save the bits relating to the bridge */
1380         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1381
1382         return (0);
1383 }
1384
1385 static int
1386 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1387 {
1388         struct ifbrparam *param = arg;
1389
1390         sc->sc_brtmax = param->ifbrp_csize;
1391         bridge_rttrim(sc);
1392
1393         return (0);
1394 }
1395
1396 static int
1397 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1398 {
1399         struct ifbrparam *param = arg;
1400
1401         param->ifbrp_csize = sc->sc_brtmax;
1402
1403         return (0);
1404 }
1405
1406 static int
1407 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1408 {
1409         struct ifbifconf *bifc = arg;
1410         struct bridge_iflist *bif;
1411         struct ifbreq breq;
1412         char *buf, *outbuf;
1413         int count, buflen, len, error = 0;
1414
1415         count = 0;
1416         LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1417                 count++;
1418         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1419                 count++;
1420
1421         buflen = sizeof(breq) * count;
1422         if (bifc->ifbic_len == 0) {
1423                 bifc->ifbic_len = buflen;
1424                 return (0);
1425         }
1426         BRIDGE_UNLOCK(sc);
1427         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1428         BRIDGE_LOCK(sc);
1429
1430         count = 0;
1431         buf = outbuf;
1432         len = min(bifc->ifbic_len, buflen);
1433         bzero(&breq, sizeof(breq));
1434         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1435                 if (len < sizeof(breq))
1436                         break;
1437
1438                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1439                     sizeof(breq.ifbr_ifsname));
1440                 /* Fill in the ifbreq structure */
1441                 error = bridge_ioctl_gifflags(sc, &breq);
1442                 if (error)
1443                         break;
1444                 memcpy(buf, &breq, sizeof(breq));
1445                 count++;
1446                 buf += sizeof(breq);
1447                 len -= sizeof(breq);
1448         }
1449         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1450                 if (len < sizeof(breq))
1451                         break;
1452
1453                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1454                     sizeof(breq.ifbr_ifsname));
1455                 breq.ifbr_ifsflags = bif->bif_flags;
1456                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1457                 memcpy(buf, &breq, sizeof(breq));
1458                 count++;
1459                 buf += sizeof(breq);
1460                 len -= sizeof(breq);
1461         }
1462
1463         BRIDGE_UNLOCK(sc);
1464         bifc->ifbic_len = sizeof(breq) * count;
1465         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1466         BRIDGE_LOCK(sc);
1467         free(outbuf, M_TEMP);
1468         return (error);
1469 }
1470
1471 static int
1472 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1473 {
1474         struct ifbaconf *bac = arg;
1475         struct bridge_rtnode *brt;
1476         struct ifbareq bareq;
1477         char *buf, *outbuf;
1478         int count, buflen, len, error = 0;
1479
1480         if (bac->ifbac_len == 0)
1481                 return (0);
1482
1483         count = 0;
1484         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1485                 count++;
1486         buflen = sizeof(bareq) * count;
1487
1488         BRIDGE_UNLOCK(sc);
1489         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1490         BRIDGE_LOCK(sc);
1491
1492         count = 0;
1493         buf = outbuf;
1494         len = min(bac->ifbac_len, buflen);
1495         bzero(&bareq, sizeof(bareq));
1496         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1497                 if (len < sizeof(bareq))
1498                         goto out;
1499                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1500                     sizeof(bareq.ifba_ifsname));
1501                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1502                 bareq.ifba_vlan = brt->brt_vlan;
1503                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1504                                 time_uptime < brt->brt_expire)
1505                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1506                 else
1507                         bareq.ifba_expire = 0;
1508                 bareq.ifba_flags = brt->brt_flags;
1509
1510                 memcpy(buf, &bareq, sizeof(bareq));
1511                 count++;
1512                 buf += sizeof(bareq);
1513                 len -= sizeof(bareq);
1514         }
1515 out:
1516         BRIDGE_UNLOCK(sc);
1517         bac->ifbac_len = sizeof(bareq) * count;
1518         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1519         BRIDGE_LOCK(sc);
1520         free(outbuf, M_TEMP);
1521         return (error);
1522 }
1523
1524 static int
1525 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1526 {
1527         struct ifbareq *req = arg;
1528         struct bridge_iflist *bif;
1529         int error;
1530
1531         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1532         if (bif == NULL)
1533                 return (ENOENT);
1534
1535         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1536             req->ifba_flags);
1537
1538         return (error);
1539 }
1540
1541 static int
1542 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1543 {
1544         struct ifbrparam *param = arg;
1545
1546         sc->sc_brttimeout = param->ifbrp_ctime;
1547         return (0);
1548 }
1549
1550 static int
1551 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1552 {
1553         struct ifbrparam *param = arg;
1554
1555         param->ifbrp_ctime = sc->sc_brttimeout;
1556         return (0);
1557 }
1558
1559 static int
1560 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1561 {
1562         struct ifbareq *req = arg;
1563
1564         return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1565 }
1566
1567 static int
1568 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1569 {
1570         struct ifbreq *req = arg;
1571
1572         bridge_rtflush(sc, req->ifbr_ifsflags);
1573         return (0);
1574 }
1575
1576 static int
1577 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1578 {
1579         struct ifbrparam *param = arg;
1580         struct bstp_state *bs = &sc->sc_stp;
1581
1582         param->ifbrp_prio = bs->bs_bridge_priority;
1583         return (0);
1584 }
1585
1586 static int
1587 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1588 {
1589         struct ifbrparam *param = arg;
1590
1591         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1592 }
1593
1594 static int
1595 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1596 {
1597         struct ifbrparam *param = arg;
1598         struct bstp_state *bs = &sc->sc_stp;
1599
1600         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1601         return (0);
1602 }
1603
1604 static int
1605 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1606 {
1607         struct ifbrparam *param = arg;
1608
1609         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1610 }
1611
1612 static int
1613 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1614 {
1615         struct ifbrparam *param = arg;
1616         struct bstp_state *bs = &sc->sc_stp;
1617
1618         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1619         return (0);
1620 }
1621
1622 static int
1623 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1624 {
1625         struct ifbrparam *param = arg;
1626
1627         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1628 }
1629
1630 static int
1631 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1632 {
1633         struct ifbrparam *param = arg;
1634         struct bstp_state *bs = &sc->sc_stp;
1635
1636         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1637         return (0);
1638 }
1639
1640 static int
1641 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1642 {
1643         struct ifbrparam *param = arg;
1644
1645         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1646 }
1647
1648 static int
1649 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1650 {
1651         struct ifbreq *req = arg;
1652         struct bridge_iflist *bif;
1653
1654         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1655         if (bif == NULL)
1656                 return (ENOENT);
1657
1658         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1659 }
1660
1661 static int
1662 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1663 {
1664         struct ifbreq *req = arg;
1665         struct bridge_iflist *bif;
1666
1667         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1668         if (bif == NULL)
1669                 return (ENOENT);
1670
1671         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1672 }
1673
1674 static int
1675 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1676 {
1677         struct ifbreq *req = arg;
1678         struct bridge_iflist *bif;
1679
1680         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1681         if (bif == NULL)
1682                 return (ENOENT);
1683
1684         bif->bif_addrmax = req->ifbr_addrmax;
1685         return (0);
1686 }
1687
1688 static int
1689 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1690 {
1691         struct ifbreq *req = arg;
1692         struct bridge_iflist *bif = NULL;
1693         struct ifnet *ifs;
1694
1695         ifs = ifunit(req->ifbr_ifsname);
1696         if (ifs == NULL)
1697                 return (ENOENT);
1698
1699         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1700                 if (ifs == bif->bif_ifp)
1701                         return (EBUSY);
1702
1703         if (ifs->if_bridge != NULL)
1704                 return (EBUSY);
1705
1706         switch (ifs->if_type) {
1707                 case IFT_ETHER:
1708                 case IFT_GIF:
1709                 case IFT_L2VLAN:
1710                         break;
1711                 default:
1712                         return (EINVAL);
1713         }
1714
1715         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1716         if (bif == NULL)
1717                 return (ENOMEM);
1718
1719         bif->bif_ifp = ifs;
1720         bif->bif_flags = IFBIF_SPAN;
1721
1722         LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1723
1724         return (0);
1725 }
1726
1727 static int
1728 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1729 {
1730         struct ifbreq *req = arg;
1731         struct bridge_iflist *bif;
1732         struct ifnet *ifs;
1733
1734         ifs = ifunit(req->ifbr_ifsname);
1735         if (ifs == NULL)
1736                 return (ENOENT);
1737
1738         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1739                 if (ifs == bif->bif_ifp)
1740                         break;
1741
1742         if (bif == NULL)
1743                 return (ENOENT);
1744
1745         bridge_delete_span(sc, bif);
1746
1747         return (0);
1748 }
1749
1750 static int
1751 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1752 {
1753         struct ifbropreq *req = arg;
1754         struct bstp_state *bs = &sc->sc_stp;
1755         struct bstp_port *root_port;
1756
1757         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1758         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1759         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1760
1761         root_port = bs->bs_root_port;
1762         if (root_port == NULL)
1763                 req->ifbop_root_port = 0;
1764         else
1765                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1766
1767         req->ifbop_holdcount = bs->bs_txholdcount;
1768         req->ifbop_priority = bs->bs_bridge_priority;
1769         req->ifbop_protocol = bs->bs_protover;
1770         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1771         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1772         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1773         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1774         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1775         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1776
1777         return (0);
1778 }
1779
1780 static int
1781 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1782 {
1783         struct ifbrparam *param = arg;
1784
1785         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1786         return (0);
1787 }
1788
1789 static int
1790 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1791 {
1792         struct ifbpstpconf *bifstp = arg;
1793         struct bridge_iflist *bif;
1794         struct bstp_port *bp;
1795         struct ifbpstpreq bpreq;
1796         char *buf, *outbuf;
1797         int count, buflen, len, error = 0;
1798
1799         count = 0;
1800         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1801                 if ((bif->bif_flags & IFBIF_STP) != 0)
1802                         count++;
1803         }
1804
1805         buflen = sizeof(bpreq) * count;
1806         if (bifstp->ifbpstp_len == 0) {
1807                 bifstp->ifbpstp_len = buflen;
1808                 return (0);
1809         }
1810
1811         BRIDGE_UNLOCK(sc);
1812         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1813         BRIDGE_LOCK(sc);
1814
1815         count = 0;
1816         buf = outbuf;
1817         len = min(bifstp->ifbpstp_len, buflen);
1818         bzero(&bpreq, sizeof(bpreq));
1819         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1820                 if (len < sizeof(bpreq))
1821                         break;
1822
1823                 if ((bif->bif_flags & IFBIF_STP) == 0)
1824                         continue;
1825
1826                 bp = &bif->bif_stp;
1827                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1828                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1829                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1830                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1831                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1832                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1833
1834                 memcpy(buf, &bpreq, sizeof(bpreq));
1835                 count++;
1836                 buf += sizeof(bpreq);
1837                 len -= sizeof(bpreq);
1838         }
1839
1840         BRIDGE_UNLOCK(sc);
1841         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1842         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1843         BRIDGE_LOCK(sc);
1844         free(outbuf, M_TEMP);
1845         return (error);
1846 }
1847
1848 static int
1849 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1850 {
1851         struct ifbrparam *param = arg;
1852
1853         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1854 }
1855
1856 static int
1857 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1858 {
1859         struct ifbrparam *param = arg;
1860
1861         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1862 }
1863
1864 /*
1865  * bridge_ifdetach:
1866  *
1867  *      Detach an interface from a bridge.  Called when a member
1868  *      interface is detaching.
1869  */
1870 static void
1871 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1872 {
1873         struct bridge_softc *sc = ifp->if_bridge;
1874         struct bridge_iflist *bif;
1875
1876         if (ifp->if_flags & IFF_RENAMING)
1877                 return;
1878         if (V_bridge_cloner == NULL) {
1879                 /*
1880                  * This detach handler can be called after
1881                  * vnet_bridge_uninit().  Just return in that case.
1882                  */
1883                 return;
1884         }
1885         /* Check if the interface is a bridge member */
1886         if (sc != NULL) {
1887                 BRIDGE_LOCK(sc);
1888
1889                 bif = bridge_lookup_member_if(sc, ifp);
1890                 if (bif != NULL)
1891                         bridge_delete_member(sc, bif, 1);
1892
1893                 BRIDGE_UNLOCK(sc);
1894                 return;
1895         }
1896
1897         /* Check if the interface is a span port */
1898         BRIDGE_LIST_LOCK();
1899         LIST_FOREACH(sc, &V_bridge_list, sc_list) {
1900                 BRIDGE_LOCK(sc);
1901                 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1902                         if (ifp == bif->bif_ifp) {
1903                                 bridge_delete_span(sc, bif);
1904                                 break;
1905                         }
1906
1907                 BRIDGE_UNLOCK(sc);
1908         }
1909         BRIDGE_LIST_UNLOCK();
1910 }
1911
1912 /*
1913  * bridge_init:
1914  *
1915  *      Initialize a bridge interface.
1916  */
1917 static void
1918 bridge_init(void *xsc)
1919 {
1920         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1921         struct ifnet *ifp = sc->sc_ifp;
1922
1923         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1924                 return;
1925
1926         BRIDGE_LOCK(sc);
1927         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1928             bridge_timer, sc);
1929
1930         ifp->if_drv_flags |= IFF_DRV_RUNNING;
1931         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
1932
1933         BRIDGE_UNLOCK(sc);
1934 }
1935
1936 /*
1937  * bridge_stop:
1938  *
1939  *      Stop the bridge interface.
1940  */
1941 static void
1942 bridge_stop(struct ifnet *ifp, int disable)
1943 {
1944         struct bridge_softc *sc = ifp->if_softc;
1945
1946         BRIDGE_LOCK_ASSERT(sc);
1947
1948         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1949                 return;
1950
1951         callout_stop(&sc->sc_brcallout);
1952         bstp_stop(&sc->sc_stp);
1953
1954         bridge_rtflush(sc, IFBF_FLUSHDYN);
1955
1956         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1957 }
1958
1959 /*
1960  * bridge_enqueue:
1961  *
1962  *      Enqueue a packet on a bridge member interface.
1963  *
1964  */
1965 static int
1966 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
1967 {
1968         int len, err = 0;
1969         short mflags;
1970         struct mbuf *m0;
1971
1972         /* We may be sending a fragment so traverse the mbuf */
1973         for (; m; m = m0) {
1974                 m0 = m->m_nextpkt;
1975                 m->m_nextpkt = NULL;
1976                 len = m->m_pkthdr.len;
1977                 mflags = m->m_flags;
1978
1979                 /*
1980                  * If underlying interface can not do VLAN tag insertion itself
1981                  * then attach a packet tag that holds it.
1982                  */
1983                 if ((m->m_flags & M_VLANTAG) &&
1984                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
1985                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
1986                         if (m == NULL) {
1987                                 if_printf(dst_ifp,
1988                                     "unable to prepend VLAN header\n");
1989                                 if_inc_counter(dst_ifp, IFCOUNTER_OERRORS, 1);
1990                                 continue;
1991                         }
1992                         m->m_flags &= ~M_VLANTAG;
1993                 }
1994
1995                 M_ASSERTPKTHDR(m); /* We shouldn't transmit mbuf without pkthdr */
1996                 if ((err = dst_ifp->if_transmit(dst_ifp, m))) {
1997                         m_freem(m0);
1998                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
1999                         break;
2000                 }
2001
2002                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OPACKETS, 1);
2003                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OBYTES, len);
2004                 if (mflags & M_MCAST)
2005                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OMCASTS, 1);
2006         }
2007
2008         return (err);
2009 }
2010
2011 /*
2012  * bridge_dummynet:
2013  *
2014  *      Receive a queued packet from dummynet and pass it on to the output
2015  *      interface.
2016  *
2017  *      The mbuf has the Ethernet header already attached.
2018  */
2019 static void
2020 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
2021 {
2022         struct bridge_softc *sc;
2023
2024         sc = ifp->if_bridge;
2025
2026         /*
2027          * The packet didnt originate from a member interface. This should only
2028          * ever happen if a member interface is removed while packets are
2029          * queued for it.
2030          */
2031         if (sc == NULL) {
2032                 m_freem(m);
2033                 return;
2034         }
2035
2036         if (PFIL_HOOKED_OUT(V_inet_pfil_head)
2037 #ifdef INET6
2038             || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2039 #endif
2040             ) {
2041                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
2042                         return;
2043                 if (m == NULL)
2044                         return;
2045         }
2046
2047         bridge_enqueue(sc, ifp, m);
2048 }
2049
2050 /*
2051  * bridge_output:
2052  *
2053  *      Send output from a bridge member interface.  This
2054  *      performs the bridging function for locally originated
2055  *      packets.
2056  *
2057  *      The mbuf has the Ethernet header already attached.  We must
2058  *      enqueue or free the mbuf before returning.
2059  */
2060 static int
2061 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
2062     struct rtentry *rt)
2063 {
2064         struct ether_header *eh;
2065         struct ifnet *bifp, *dst_if;
2066         struct bridge_softc *sc;
2067         uint16_t vlan;
2068
2069         if (m->m_len < ETHER_HDR_LEN) {
2070                 m = m_pullup(m, ETHER_HDR_LEN);
2071                 if (m == NULL)
2072                         return (0);
2073         }
2074
2075         eh = mtod(m, struct ether_header *);
2076         sc = ifp->if_bridge;
2077         vlan = VLANTAGOF(m);
2078
2079         BRIDGE_LOCK(sc);
2080         bifp = sc->sc_ifp;
2081
2082         /*
2083          * If bridge is down, but the original output interface is up,
2084          * go ahead and send out that interface.  Otherwise, the packet
2085          * is dropped below.
2086          */
2087         if ((bifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2088                 dst_if = ifp;
2089                 goto sendunicast;
2090         }
2091
2092         /*
2093          * If the packet is a multicast, or we don't know a better way to
2094          * get there, send to all interfaces.
2095          */
2096         if (ETHER_IS_MULTICAST(eh->ether_dhost))
2097                 dst_if = NULL;
2098         else
2099                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
2100         /* Tap any traffic not passing back out the originating interface */
2101         if (dst_if != ifp)
2102                 ETHER_BPF_MTAP(bifp, m);
2103         if (dst_if == NULL) {
2104                 struct bridge_iflist *bif;
2105                 struct mbuf *mc;
2106                 int error = 0, used = 0;
2107
2108                 bridge_span(sc, m);
2109
2110                 BRIDGE_LOCK2REF(sc, error);
2111                 if (error) {
2112                         m_freem(m);
2113                         return (0);
2114                 }
2115
2116                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
2117                         dst_if = bif->bif_ifp;
2118
2119                         if (dst_if->if_type == IFT_GIF)
2120                                 continue;
2121                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2122                                 continue;
2123
2124                         /*
2125                          * If this is not the original output interface,
2126                          * and the interface is participating in spanning
2127                          * tree, make sure the port is in a state that
2128                          * allows forwarding.
2129                          */
2130                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
2131                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2132                                 continue;
2133
2134                         if (LIST_NEXT(bif, bif_next) == NULL) {
2135                                 used = 1;
2136                                 mc = m;
2137                         } else {
2138                                 mc = m_copypacket(m, M_NOWAIT);
2139                                 if (mc == NULL) {
2140                                         if_inc_counter(bifp, IFCOUNTER_OERRORS, 1);
2141                                         continue;
2142                                 }
2143                         }
2144
2145                         bridge_enqueue(sc, dst_if, mc);
2146                 }
2147                 if (used == 0)
2148                         m_freem(m);
2149                 BRIDGE_UNREF(sc);
2150                 return (0);
2151         }
2152
2153 sendunicast:
2154         /*
2155          * XXX Spanning tree consideration here?
2156          */
2157
2158         bridge_span(sc, m);
2159         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2160                 m_freem(m);
2161                 BRIDGE_UNLOCK(sc);
2162                 return (0);
2163         }
2164
2165         BRIDGE_UNLOCK(sc);
2166         bridge_enqueue(sc, dst_if, m);
2167         return (0);
2168 }
2169
2170 /*
2171  * bridge_transmit:
2172  *
2173  *      Do output on a bridge.
2174  *
2175  */
2176 static int
2177 bridge_transmit(struct ifnet *ifp, struct mbuf *m)
2178 {
2179         struct bridge_softc *sc;
2180         struct ether_header *eh;
2181         struct ifnet *dst_if;
2182         int error = 0;
2183
2184         sc = ifp->if_softc;
2185
2186         ETHER_BPF_MTAP(ifp, m);
2187
2188         eh = mtod(m, struct ether_header *);
2189
2190         BRIDGE_LOCK(sc);
2191         if (((m->m_flags & (M_BCAST|M_MCAST)) == 0) &&
2192             (dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1)) != NULL) {
2193                 BRIDGE_UNLOCK(sc);
2194                 error = bridge_enqueue(sc, dst_if, m);
2195         } else
2196                 bridge_broadcast(sc, ifp, m, 0);
2197
2198         return (error);
2199 }
2200
2201 /*
2202  * The ifp->if_qflush entry point for if_bridge(4) is no-op.
2203  */
2204 static void
2205 bridge_qflush(struct ifnet *ifp __unused)
2206 {
2207 }
2208
2209 /*
2210  * bridge_forward:
2211  *
2212  *      The forwarding function of the bridge.
2213  *
2214  *      NOTE: Releases the lock on return.
2215  */
2216 static void
2217 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2218     struct mbuf *m)
2219 {
2220         struct bridge_iflist *dbif;
2221         struct ifnet *src_if, *dst_if, *ifp;
2222         struct ether_header *eh;
2223         uint16_t vlan;
2224         uint8_t *dst;
2225         int error;
2226
2227         src_if = m->m_pkthdr.rcvif;
2228         ifp = sc->sc_ifp;
2229
2230         if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
2231         if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2232         vlan = VLANTAGOF(m);
2233
2234         if ((sbif->bif_flags & IFBIF_STP) &&
2235             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2236                 goto drop;
2237
2238         eh = mtod(m, struct ether_header *);
2239         dst = eh->ether_dhost;
2240
2241         /* If the interface is learning, record the address. */
2242         if (sbif->bif_flags & IFBIF_LEARNING) {
2243                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2244                     sbif, 0, IFBAF_DYNAMIC);
2245                 /*
2246                  * If the interface has addresses limits then deny any source
2247                  * that is not in the cache.
2248                  */
2249                 if (error && sbif->bif_addrmax)
2250                         goto drop;
2251         }
2252
2253         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2254             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2255                 goto drop;
2256
2257         /*
2258          * At this point, the port either doesn't participate
2259          * in spanning tree or it is in the forwarding state.
2260          */
2261
2262         /*
2263          * If the packet is unicast, destined for someone on
2264          * "this" side of the bridge, drop it.
2265          */
2266         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2267                 dst_if = bridge_rtlookup(sc, dst, vlan);
2268                 if (src_if == dst_if)
2269                         goto drop;
2270         } else {
2271                 /*
2272                  * Check if its a reserved multicast address, any address
2273                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2274                  * bridge.
2275                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2276                  */
2277                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2278                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2279                     dst[4] == 0x00 && dst[5] <= 0x0f)
2280                         goto drop;
2281
2282                 /* ...forward it to all interfaces. */
2283                 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
2284                 dst_if = NULL;
2285         }
2286
2287         /*
2288          * If we have a destination interface which is a member of our bridge,
2289          * OR this is a unicast packet, push it through the bpf(4) machinery.
2290          * For broadcast or multicast packets, don't bother because it will
2291          * be reinjected into ether_input. We do this before we pass the packets
2292          * through the pfil(9) framework, as it is possible that pfil(9) will
2293          * drop the packet, or possibly modify it, making it difficult to debug
2294          * firewall issues on the bridge.
2295          */
2296         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2297                 ETHER_BPF_MTAP(ifp, m);
2298
2299         /* run the packet filter */
2300         if (PFIL_HOOKED_IN(V_inet_pfil_head)
2301 #ifdef INET6
2302             || PFIL_HOOKED_IN(V_inet6_pfil_head)
2303 #endif
2304             ) {
2305                 BRIDGE_UNLOCK(sc);
2306                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2307                         return;
2308                 if (m == NULL)
2309                         return;
2310                 BRIDGE_LOCK(sc);
2311         }
2312
2313         if (dst_if == NULL) {
2314                 bridge_broadcast(sc, src_if, m, 1);
2315                 return;
2316         }
2317
2318         /*
2319          * At this point, we're dealing with a unicast frame
2320          * going to a different interface.
2321          */
2322         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2323                 goto drop;
2324
2325         dbif = bridge_lookup_member_if(sc, dst_if);
2326         if (dbif == NULL)
2327                 /* Not a member of the bridge (anymore?) */
2328                 goto drop;
2329
2330         /* Private segments can not talk to each other */
2331         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2332                 goto drop;
2333
2334         if ((dbif->bif_flags & IFBIF_STP) &&
2335             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2336                 goto drop;
2337
2338         BRIDGE_UNLOCK(sc);
2339
2340         if (PFIL_HOOKED_OUT(V_inet_pfil_head)
2341 #ifdef INET6
2342             || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2343 #endif
2344             ) {
2345                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2346                         return;
2347                 if (m == NULL)
2348                         return;
2349         }
2350
2351         bridge_enqueue(sc, dst_if, m);
2352         return;
2353
2354 drop:
2355         BRIDGE_UNLOCK(sc);
2356         m_freem(m);
2357 }
2358
2359 /*
2360  * bridge_input:
2361  *
2362  *      Receive input from a member interface.  Queue the packet for
2363  *      bridging if it is not for us.
2364  */
2365 static struct mbuf *
2366 bridge_input(struct ifnet *ifp, struct mbuf *m)
2367 {
2368         struct bridge_softc *sc = ifp->if_bridge;
2369         struct bridge_iflist *bif, *bif2;
2370         struct ifnet *bifp;
2371         struct ether_header *eh;
2372         struct mbuf *mc, *mc2;
2373         uint16_t vlan;
2374         int error;
2375
2376         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2377                 return (m);
2378
2379         bifp = sc->sc_ifp;
2380         vlan = VLANTAGOF(m);
2381
2382         /*
2383          * Implement support for bridge monitoring. If this flag has been
2384          * set on this interface, discard the packet once we push it through
2385          * the bpf(4) machinery, but before we do, increment the byte and
2386          * packet counters associated with this interface.
2387          */
2388         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2389                 m->m_pkthdr.rcvif  = bifp;
2390                 ETHER_BPF_MTAP(bifp, m);
2391                 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);
2392                 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2393                 m_freem(m);
2394                 return (NULL);
2395         }
2396         BRIDGE_LOCK(sc);
2397         bif = bridge_lookup_member_if(sc, ifp);
2398         if (bif == NULL) {
2399                 BRIDGE_UNLOCK(sc);
2400                 return (m);
2401         }
2402
2403         eh = mtod(m, struct ether_header *);
2404
2405         bridge_span(sc, m);
2406
2407         if (m->m_flags & (M_BCAST|M_MCAST)) {
2408                 /* Tap off 802.1D packets; they do not get forwarded. */
2409                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2410                     ETHER_ADDR_LEN) == 0) {
2411                         bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2412                         BRIDGE_UNLOCK(sc);
2413                         return (NULL);
2414                 }
2415
2416                 if ((bif->bif_flags & IFBIF_STP) &&
2417                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2418                         BRIDGE_UNLOCK(sc);
2419                         return (m);
2420                 }
2421
2422                 /*
2423                  * Make a deep copy of the packet and enqueue the copy
2424                  * for bridge processing; return the original packet for
2425                  * local processing.
2426                  */
2427                 mc = m_dup(m, M_NOWAIT);
2428                 if (mc == NULL) {
2429                         BRIDGE_UNLOCK(sc);
2430                         return (m);
2431                 }
2432
2433                 /* Perform the bridge forwarding function with the copy. */
2434                 bridge_forward(sc, bif, mc);
2435
2436                 /*
2437                  * Reinject the mbuf as arriving on the bridge so we have a
2438                  * chance at claiming multicast packets. We can not loop back
2439                  * here from ether_input as a bridge is never a member of a
2440                  * bridge.
2441                  */
2442                 KASSERT(bifp->if_bridge == NULL,
2443                     ("loop created in bridge_input"));
2444                 mc2 = m_dup(m, M_NOWAIT);
2445                 if (mc2 != NULL) {
2446                         /* Keep the layer3 header aligned */
2447                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2448                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2449                 }
2450                 if (mc2 != NULL) {
2451                         mc2->m_pkthdr.rcvif = bifp;
2452                         (*bifp->if_input)(bifp, mc2);
2453                 }
2454
2455                 /* Return the original packet for local processing. */
2456                 return (m);
2457         }
2458
2459         if ((bif->bif_flags & IFBIF_STP) &&
2460             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2461                 BRIDGE_UNLOCK(sc);
2462                 return (m);
2463         }
2464
2465 #if (defined(INET) || defined(INET6))
2466 #   define OR_CARP_CHECK_WE_ARE_DST(iface) \
2467         || ((iface)->if_carp \
2468             && (*carp_forus_p)((iface), eh->ether_dhost))
2469 #   define OR_CARP_CHECK_WE_ARE_SRC(iface) \
2470         || ((iface)->if_carp \
2471             && (*carp_forus_p)((iface), eh->ether_shost))
2472 #else
2473 #   define OR_CARP_CHECK_WE_ARE_DST(iface)
2474 #   define OR_CARP_CHECK_WE_ARE_SRC(iface)
2475 #endif
2476
2477 #ifdef INET6
2478 #   define OR_PFIL_HOOKED_INET6 \
2479         || PFIL_HOOKED_IN(V_inet6_pfil_head)
2480 #else
2481 #   define OR_PFIL_HOOKED_INET6
2482 #endif
2483
2484 #define GRAB_OUR_PACKETS(iface) \
2485         if ((iface)->if_type == IFT_GIF) \
2486                 continue; \
2487         /* It is destined for us. */ \
2488         if (memcmp(IF_LLADDR((iface)), eh->ether_dhost,  ETHER_ADDR_LEN) == 0 \
2489             OR_CARP_CHECK_WE_ARE_DST((iface))                           \
2490             ) {                                                         \
2491                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2492                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2493                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2494                         if (error && bif->bif_addrmax) {                \
2495                                 BRIDGE_UNLOCK(sc);                      \
2496                                 m_freem(m);                             \
2497                                 return (NULL);                          \
2498                         }                                               \
2499                 }                                                       \
2500                 m->m_pkthdr.rcvif = iface;                              \
2501                 if ((iface) == ifp) {                                   \
2502                         /* Skip bridge processing... src == dest */     \
2503                         BRIDGE_UNLOCK(sc);                              \
2504                         return (m);                                     \
2505                 }                                                       \
2506                 /* It's passing over or to the bridge, locally. */      \
2507                 ETHER_BPF_MTAP(bifp, m);                                \
2508                 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);            \
2509                 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len); \
2510                 /* Filter on the physical interface. */                 \
2511                 if (V_pfil_local_phys && (PFIL_HOOKED_IN(V_inet_pfil_head) \
2512                      OR_PFIL_HOOKED_INET6)) {                           \
2513                         if (bridge_pfil(&m, NULL, ifp,                  \
2514                             PFIL_IN) != 0 || m == NULL) {               \
2515                                 BRIDGE_UNLOCK(sc);                      \
2516                                 return (NULL);                          \
2517                         }                                               \
2518                 }                                                       \
2519                 if ((iface) != bifp)                                    \
2520                         ETHER_BPF_MTAP(iface, m);                       \
2521                 BRIDGE_UNLOCK(sc);                                      \
2522                 return (m);                                             \
2523         }                                                               \
2524                                                                         \
2525         /* We just received a packet that we sent out. */               \
2526         if (memcmp(IF_LLADDR((iface)), eh->ether_shost, ETHER_ADDR_LEN) == 0 \
2527             OR_CARP_CHECK_WE_ARE_SRC((iface))                   \
2528             ) {                                                         \
2529                 BRIDGE_UNLOCK(sc);                                      \
2530                 m_freem(m);                                             \
2531                 return (NULL);                                          \
2532         }
2533
2534         /*
2535          * Unicast.  Make sure it's not for the bridge.
2536          */
2537         do { GRAB_OUR_PACKETS(bifp) } while (0);
2538
2539         /*
2540          * Give a chance for ifp at first priority. This will help when the
2541          * packet comes through the interface like VLAN's with the same MACs
2542          * on several interfaces from the same bridge. This also will save
2543          * some CPU cycles in case the destination interface and the input
2544          * interface (eq ifp) are the same.
2545          */
2546         do { GRAB_OUR_PACKETS(ifp) } while (0);
2547
2548         /* Now check the all bridge members. */
2549         LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2550                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2551         }
2552
2553 #undef OR_CARP_CHECK_WE_ARE_DST
2554 #undef OR_CARP_CHECK_WE_ARE_SRC
2555 #undef OR_PFIL_HOOKED_INET6
2556 #undef GRAB_OUR_PACKETS
2557
2558         /* Perform the bridge forwarding function. */
2559         bridge_forward(sc, bif, m);
2560
2561         return (NULL);
2562 }
2563
2564 /*
2565  * bridge_broadcast:
2566  *
2567  *      Send a frame to all interfaces that are members of
2568  *      the bridge, except for the one on which the packet
2569  *      arrived.
2570  *
2571  *      NOTE: Releases the lock on return.
2572  */
2573 static void
2574 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2575     struct mbuf *m, int runfilt)
2576 {
2577         struct bridge_iflist *dbif, *sbif;
2578         struct mbuf *mc;
2579         struct ifnet *dst_if;
2580         int error = 0, used = 0, i;
2581
2582         sbif = bridge_lookup_member_if(sc, src_if);
2583
2584         BRIDGE_LOCK2REF(sc, error);
2585         if (error) {
2586                 m_freem(m);
2587                 return;
2588         }
2589
2590         /* Filter on the bridge interface before broadcasting */
2591         if (runfilt && (PFIL_HOOKED_OUT(V_inet_pfil_head)
2592 #ifdef INET6
2593             || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2594 #endif
2595             )) {
2596                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2597                         goto out;
2598                 if (m == NULL)
2599                         goto out;
2600         }
2601
2602         LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2603                 dst_if = dbif->bif_ifp;
2604                 if (dst_if == src_if)
2605                         continue;
2606
2607                 /* Private segments can not talk to each other */
2608                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2609                         continue;
2610
2611                 if ((dbif->bif_flags & IFBIF_STP) &&
2612                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2613                         continue;
2614
2615                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2616                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2617                         continue;
2618
2619                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2620                         continue;
2621
2622                 if (LIST_NEXT(dbif, bif_next) == NULL) {
2623                         mc = m;
2624                         used = 1;
2625                 } else {
2626                         mc = m_dup(m, M_NOWAIT);
2627                         if (mc == NULL) {
2628                                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2629                                 continue;
2630                         }
2631                 }
2632
2633                 /*
2634                  * Filter on the output interface. Pass a NULL bridge interface
2635                  * pointer so we do not redundantly filter on the bridge for
2636                  * each interface we broadcast on.
2637                  */
2638                 if (runfilt && (PFIL_HOOKED_OUT(V_inet_pfil_head)
2639 #ifdef INET6
2640                     || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2641 #endif
2642                     )) {
2643                         if (used == 0) {
2644                                 /* Keep the layer3 header aligned */
2645                                 i = min(mc->m_pkthdr.len, max_protohdr);
2646                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2647                                 if (mc == NULL) {
2648                                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2649                                         continue;
2650                                 }
2651                         }
2652                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2653                                 continue;
2654                         if (mc == NULL)
2655                                 continue;
2656                 }
2657
2658                 bridge_enqueue(sc, dst_if, mc);
2659         }
2660         if (used == 0)
2661                 m_freem(m);
2662
2663 out:
2664         BRIDGE_UNREF(sc);
2665 }
2666
2667 /*
2668  * bridge_span:
2669  *
2670  *      Duplicate a packet out one or more interfaces that are in span mode,
2671  *      the original mbuf is unmodified.
2672  */
2673 static void
2674 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2675 {
2676         struct bridge_iflist *bif;
2677         struct ifnet *dst_if;
2678         struct mbuf *mc;
2679
2680         if (LIST_EMPTY(&sc->sc_spanlist))
2681                 return;
2682
2683         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2684                 dst_if = bif->bif_ifp;
2685
2686                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2687                         continue;
2688
2689                 mc = m_copypacket(m, M_NOWAIT);
2690                 if (mc == NULL) {
2691                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2692                         continue;
2693                 }
2694
2695                 bridge_enqueue(sc, dst_if, mc);
2696         }
2697 }
2698
2699 /*
2700  * bridge_rtupdate:
2701  *
2702  *      Add a bridge routing entry.
2703  */
2704 static int
2705 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2706     struct bridge_iflist *bif, int setflags, uint8_t flags)
2707 {
2708         struct bridge_rtnode *brt;
2709         int error;
2710
2711         BRIDGE_LOCK_ASSERT(sc);
2712
2713         /* Check the source address is valid and not multicast. */
2714         if (ETHER_IS_MULTICAST(dst) ||
2715             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2716              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2717                 return (EINVAL);
2718
2719         /* 802.1p frames map to vlan 1 */
2720         if (vlan == 0)
2721                 vlan = 1;
2722
2723         /*
2724          * A route for this destination might already exist.  If so,
2725          * update it, otherwise create a new one.
2726          */
2727         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2728                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2729                         sc->sc_brtexceeded++;
2730                         return (ENOSPC);
2731                 }
2732                 /* Check per interface address limits (if enabled) */
2733                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2734                         bif->bif_addrexceeded++;
2735                         return (ENOSPC);
2736                 }
2737
2738                 /*
2739                  * Allocate a new bridge forwarding node, and
2740                  * initialize the expiration time and Ethernet
2741                  * address.
2742                  */
2743                 brt = uma_zalloc(V_bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2744                 if (brt == NULL)
2745                         return (ENOMEM);
2746
2747                 if (bif->bif_flags & IFBIF_STICKY)
2748                         brt->brt_flags = IFBAF_STICKY;
2749                 else
2750                         brt->brt_flags = IFBAF_DYNAMIC;
2751
2752                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2753                 brt->brt_vlan = vlan;
2754
2755                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2756                         uma_zfree(V_bridge_rtnode_zone, brt);
2757                         return (error);
2758                 }
2759                 brt->brt_dst = bif;
2760                 bif->bif_addrcnt++;
2761         }
2762
2763         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2764             brt->brt_dst != bif) {
2765                 brt->brt_dst->bif_addrcnt--;
2766                 brt->brt_dst = bif;
2767                 brt->brt_dst->bif_addrcnt++;
2768         }
2769
2770         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2771                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2772         if (setflags)
2773                 brt->brt_flags = flags;
2774
2775         return (0);
2776 }
2777
2778 /*
2779  * bridge_rtlookup:
2780  *
2781  *      Lookup the destination interface for an address.
2782  */
2783 static struct ifnet *
2784 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2785 {
2786         struct bridge_rtnode *brt;
2787
2788         BRIDGE_LOCK_ASSERT(sc);
2789
2790         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2791                 return (NULL);
2792
2793         return (brt->brt_ifp);
2794 }
2795
2796 /*
2797  * bridge_rttrim:
2798  *
2799  *      Trim the routine table so that we have a number
2800  *      of routing entries less than or equal to the
2801  *      maximum number.
2802  */
2803 static void
2804 bridge_rttrim(struct bridge_softc *sc)
2805 {
2806         struct bridge_rtnode *brt, *nbrt;
2807
2808         BRIDGE_LOCK_ASSERT(sc);
2809
2810         /* Make sure we actually need to do this. */
2811         if (sc->sc_brtcnt <= sc->sc_brtmax)
2812                 return;
2813
2814         /* Force an aging cycle; this might trim enough addresses. */
2815         bridge_rtage(sc);
2816         if (sc->sc_brtcnt <= sc->sc_brtmax)
2817                 return;
2818
2819         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2820                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2821                         bridge_rtnode_destroy(sc, brt);
2822                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2823                                 return;
2824                 }
2825         }
2826 }
2827
2828 /*
2829  * bridge_timer:
2830  *
2831  *      Aging timer for the bridge.
2832  */
2833 static void
2834 bridge_timer(void *arg)
2835 {
2836         struct bridge_softc *sc = arg;
2837
2838         BRIDGE_LOCK_ASSERT(sc);
2839
2840         /* Destruction of rtnodes requires a proper vnet context */
2841         CURVNET_SET(sc->sc_ifp->if_vnet);
2842         bridge_rtage(sc);
2843
2844         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2845                 callout_reset(&sc->sc_brcallout,
2846                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2847         CURVNET_RESTORE();
2848 }
2849
2850 /*
2851  * bridge_rtage:
2852  *
2853  *      Perform an aging cycle.
2854  */
2855 static void
2856 bridge_rtage(struct bridge_softc *sc)
2857 {
2858         struct bridge_rtnode *brt, *nbrt;
2859
2860         BRIDGE_LOCK_ASSERT(sc);
2861
2862         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2863                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2864                         if (time_uptime >= brt->brt_expire)
2865                                 bridge_rtnode_destroy(sc, brt);
2866                 }
2867         }
2868 }
2869
2870 /*
2871  * bridge_rtflush:
2872  *
2873  *      Remove all dynamic addresses from the bridge.
2874  */
2875 static void
2876 bridge_rtflush(struct bridge_softc *sc, int full)
2877 {
2878         struct bridge_rtnode *brt, *nbrt;
2879
2880         BRIDGE_LOCK_ASSERT(sc);
2881
2882         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2883                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2884                         bridge_rtnode_destroy(sc, brt);
2885         }
2886 }
2887
2888 /*
2889  * bridge_rtdaddr:
2890  *
2891  *      Remove an address from the table.
2892  */
2893 static int
2894 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2895 {
2896         struct bridge_rtnode *brt;
2897         int found = 0;
2898
2899         BRIDGE_LOCK_ASSERT(sc);
2900
2901         /*
2902          * If vlan is zero then we want to delete for all vlans so the lookup
2903          * may return more than one.
2904          */
2905         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
2906                 bridge_rtnode_destroy(sc, brt);
2907                 found = 1;
2908         }
2909
2910         return (found ? 0 : ENOENT);
2911 }
2912
2913 /*
2914  * bridge_rtdelete:
2915  *
2916  *      Delete routes to a speicifc member interface.
2917  */
2918 static void
2919 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2920 {
2921         struct bridge_rtnode *brt, *nbrt;
2922
2923         BRIDGE_LOCK_ASSERT(sc);
2924
2925         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2926                 if (brt->brt_ifp == ifp && (full ||
2927                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
2928                         bridge_rtnode_destroy(sc, brt);
2929         }
2930 }
2931
2932 /*
2933  * bridge_rtable_init:
2934  *
2935  *      Initialize the route table for this bridge.
2936  */
2937 static void
2938 bridge_rtable_init(struct bridge_softc *sc)
2939 {
2940         int i;
2941
2942         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2943             M_DEVBUF, M_WAITOK);
2944
2945         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2946                 LIST_INIT(&sc->sc_rthash[i]);
2947
2948         sc->sc_rthash_key = arc4random();
2949         LIST_INIT(&sc->sc_rtlist);
2950 }
2951
2952 /*
2953  * bridge_rtable_fini:
2954  *
2955  *      Deconstruct the route table for this bridge.
2956  */
2957 static void
2958 bridge_rtable_fini(struct bridge_softc *sc)
2959 {
2960
2961         KASSERT(sc->sc_brtcnt == 0,
2962             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
2963         free(sc->sc_rthash, M_DEVBUF);
2964 }
2965
2966 /*
2967  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2968  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2969  */
2970 #define mix(a, b, c)                                                    \
2971 do {                                                                    \
2972         a -= b; a -= c; a ^= (c >> 13);                                 \
2973         b -= c; b -= a; b ^= (a << 8);                                  \
2974         c -= a; c -= b; c ^= (b >> 13);                                 \
2975         a -= b; a -= c; a ^= (c >> 12);                                 \
2976         b -= c; b -= a; b ^= (a << 16);                                 \
2977         c -= a; c -= b; c ^= (b >> 5);                                  \
2978         a -= b; a -= c; a ^= (c >> 3);                                  \
2979         b -= c; b -= a; b ^= (a << 10);                                 \
2980         c -= a; c -= b; c ^= (b >> 15);                                 \
2981 } while (/*CONSTCOND*/0)
2982
2983 static __inline uint32_t
2984 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2985 {
2986         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2987
2988         b += addr[5] << 8;
2989         b += addr[4];
2990         a += addr[3] << 24;
2991         a += addr[2] << 16;
2992         a += addr[1] << 8;
2993         a += addr[0];
2994
2995         mix(a, b, c);
2996
2997         return (c & BRIDGE_RTHASH_MASK);
2998 }
2999
3000 #undef mix
3001
3002 static int
3003 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
3004 {
3005         int i, d;
3006
3007         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
3008                 d = ((int)a[i]) - ((int)b[i]);
3009         }
3010
3011         return (d);
3012 }
3013
3014 /*
3015  * bridge_rtnode_lookup:
3016  *
3017  *      Look up a bridge route node for the specified destination. Compare the
3018  *      vlan id or if zero then just return the first match.
3019  */
3020 static struct bridge_rtnode *
3021 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
3022 {
3023         struct bridge_rtnode *brt;
3024         uint32_t hash;
3025         int dir;
3026
3027         BRIDGE_LOCK_ASSERT(sc);
3028
3029         hash = bridge_rthash(sc, addr);
3030         LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
3031                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
3032                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
3033                         return (brt);
3034                 if (dir > 0)
3035                         return (NULL);
3036         }
3037
3038         return (NULL);
3039 }
3040
3041 /*
3042  * bridge_rtnode_insert:
3043  *
3044  *      Insert the specified bridge node into the route table.  We
3045  *      assume the entry is not already in the table.
3046  */
3047 static int
3048 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
3049 {
3050         struct bridge_rtnode *lbrt;
3051         uint32_t hash;
3052         int dir;
3053
3054         BRIDGE_LOCK_ASSERT(sc);
3055
3056         hash = bridge_rthash(sc, brt->brt_addr);
3057
3058         lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
3059         if (lbrt == NULL) {
3060                 LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
3061                 goto out;
3062         }
3063
3064         do {
3065                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
3066                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
3067                         return (EEXIST);
3068                 if (dir > 0) {
3069                         LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
3070                         goto out;
3071                 }
3072                 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
3073                         LIST_INSERT_AFTER(lbrt, brt, brt_hash);
3074                         goto out;
3075                 }
3076                 lbrt = LIST_NEXT(lbrt, brt_hash);
3077         } while (lbrt != NULL);
3078
3079 #ifdef DIAGNOSTIC
3080         panic("bridge_rtnode_insert: impossible");
3081 #endif
3082
3083 out:
3084         LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
3085         sc->sc_brtcnt++;
3086
3087         return (0);
3088 }
3089
3090 /*
3091  * bridge_rtnode_destroy:
3092  *
3093  *      Destroy a bridge rtnode.
3094  */
3095 static void
3096 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3097 {
3098         BRIDGE_LOCK_ASSERT(sc);
3099
3100         LIST_REMOVE(brt, brt_hash);
3101
3102         LIST_REMOVE(brt, brt_list);
3103         sc->sc_brtcnt--;
3104         brt->brt_dst->bif_addrcnt--;
3105         uma_zfree(V_bridge_rtnode_zone, brt);
3106 }
3107
3108 /*
3109  * bridge_rtable_expire:
3110  *
3111  *      Set the expiry time for all routes on an interface.
3112  */
3113 static void
3114 bridge_rtable_expire(struct ifnet *ifp, int age)
3115 {
3116         struct bridge_softc *sc = ifp->if_bridge;
3117         struct bridge_rtnode *brt;
3118
3119         CURVNET_SET(ifp->if_vnet);
3120         BRIDGE_LOCK(sc);
3121
3122         /*
3123          * If the age is zero then flush, otherwise set all the expiry times to
3124          * age for the interface
3125          */
3126         if (age == 0)
3127                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
3128         else {
3129                 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
3130                         /* Cap the expiry time to 'age' */
3131                         if (brt->brt_ifp == ifp &&
3132                             brt->brt_expire > time_uptime + age &&
3133                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3134                                 brt->brt_expire = time_uptime + age;
3135                 }
3136         }
3137         BRIDGE_UNLOCK(sc);
3138         CURVNET_RESTORE();
3139 }
3140
3141 /*
3142  * bridge_state_change:
3143  *
3144  *      Callback from the bridgestp code when a port changes states.
3145  */
3146 static void
3147 bridge_state_change(struct ifnet *ifp, int state)
3148 {
3149         struct bridge_softc *sc = ifp->if_bridge;
3150         static const char *stpstates[] = {
3151                 "disabled",
3152                 "listening",
3153                 "learning",
3154                 "forwarding",
3155                 "blocking",
3156                 "discarding"
3157         };
3158
3159         CURVNET_SET(ifp->if_vnet);
3160         if (V_log_stp)
3161                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
3162                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
3163         CURVNET_RESTORE();
3164 }
3165
3166 /*
3167  * Send bridge packets through pfil if they are one of the types pfil can deal
3168  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3169  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3170  * that interface.
3171  */
3172 static int
3173 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3174 {
3175         int snap, error, i, hlen;
3176         struct ether_header *eh1, eh2;
3177         struct ip *ip;
3178         struct llc llc1;
3179         u_int16_t ether_type;
3180         pfil_return_t rv;
3181
3182         snap = 0;
3183         error = -1;     /* Default error if not error == 0 */
3184
3185 #if 0
3186         /* we may return with the IP fields swapped, ensure its not shared */
3187         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
3188 #endif
3189
3190         if (V_pfil_bridge == 0 && V_pfil_member == 0 && V_pfil_ipfw == 0)
3191                 return (0); /* filtering is disabled */
3192
3193         i = min((*mp)->m_pkthdr.len, max_protohdr);
3194         if ((*mp)->m_len < i) {
3195             *mp = m_pullup(*mp, i);
3196             if (*mp == NULL) {
3197                 printf("%s: m_pullup failed\n", __func__);
3198                 return (-1);
3199             }
3200         }
3201
3202         eh1 = mtod(*mp, struct ether_header *);
3203         ether_type = ntohs(eh1->ether_type);
3204
3205         /*
3206          * Check for SNAP/LLC.
3207          */
3208         if (ether_type < ETHERMTU) {
3209                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3210
3211                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3212                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3213                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3214                     llc2->llc_control == LLC_UI) {
3215                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3216                         snap = 1;
3217                 }
3218         }
3219
3220         /*
3221          * If we're trying to filter bridge traffic, don't look at anything
3222          * other than IP and ARP traffic.  If the filter doesn't understand
3223          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3224          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3225          * but of course we don't have an AppleTalk filter to begin with.
3226          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3227          * ARP traffic.)
3228          */
3229         switch (ether_type) {
3230                 case ETHERTYPE_ARP:
3231                 case ETHERTYPE_REVARP:
3232                         if (V_pfil_ipfw_arp == 0)
3233                                 return (0); /* Automatically pass */
3234                         break;
3235
3236                 case ETHERTYPE_IP:
3237 #ifdef INET6
3238                 case ETHERTYPE_IPV6:
3239 #endif /* INET6 */
3240                         break;
3241                 default:
3242                         /*
3243                          * Check to see if the user wants to pass non-ip
3244                          * packets, these will not be checked by pfil(9) and
3245                          * passed unconditionally so the default is to drop.
3246                          */
3247                         if (V_pfil_onlyip)
3248                                 goto bad;
3249         }
3250
3251         /* Run the packet through pfil before stripping link headers */
3252         if (PFIL_HOOKED_OUT(V_link_pfil_head) && V_pfil_ipfw != 0 &&
3253             dir == PFIL_OUT && ifp != NULL) {
3254                 switch (pfil_run_hooks(V_link_pfil_head, mp, ifp, dir, NULL)) {
3255                 case PFIL_DROPPED:
3256                         return (EACCES);
3257                 case PFIL_CONSUMED:
3258                         return (0);
3259                 }
3260         }
3261
3262         /* Strip off the Ethernet header and keep a copy. */
3263         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3264         m_adj(*mp, ETHER_HDR_LEN);
3265
3266         /* Strip off snap header, if present */
3267         if (snap) {
3268                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3269                 m_adj(*mp, sizeof(struct llc));
3270         }
3271
3272         /*
3273          * Check the IP header for alignment and errors
3274          */
3275         if (dir == PFIL_IN) {
3276                 switch (ether_type) {
3277                         case ETHERTYPE_IP:
3278                                 error = bridge_ip_checkbasic(mp);
3279                                 break;
3280 #ifdef INET6
3281                         case ETHERTYPE_IPV6:
3282                                 error = bridge_ip6_checkbasic(mp);
3283                                 break;
3284 #endif /* INET6 */
3285                         default:
3286                                 error = 0;
3287                 }
3288                 if (error)
3289                         goto bad;
3290         }
3291
3292         error = 0;
3293
3294         /*
3295          * Run the packet through pfil
3296          */
3297         rv = PFIL_PASS;
3298         switch (ether_type) {
3299         case ETHERTYPE_IP:
3300                 /*
3301                  * Run pfil on the member interface and the bridge, both can
3302                  * be skipped by clearing pfil_member or pfil_bridge.
3303                  *
3304                  * Keep the order:
3305                  *   in_if -> bridge_if -> out_if
3306                  */
3307                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL && (rv =
3308                     pfil_run_hooks(V_inet_pfil_head, mp, bifp, dir, NULL)) !=
3309                     PFIL_PASS)
3310                         break;
3311
3312                 if (V_pfil_member && ifp != NULL && (rv =
3313                     pfil_run_hooks(V_inet_pfil_head, mp, ifp, dir, NULL)) !=
3314                     PFIL_PASS)
3315                         break;
3316
3317                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv =
3318                     pfil_run_hooks(V_inet_pfil_head, mp, bifp, dir, NULL)) !=
3319                     PFIL_PASS)
3320                         break;
3321
3322                 /* check if we need to fragment the packet */
3323                 /* bridge_fragment generates a mbuf chain of packets */
3324                 /* that already include eth headers */
3325                 if (V_pfil_member && ifp != NULL && dir == PFIL_OUT) {
3326                         i = (*mp)->m_pkthdr.len;
3327                         if (i > ifp->if_mtu) {
3328                                 error = bridge_fragment(ifp, mp, &eh2, snap,
3329                                             &llc1);
3330                                 return (error);
3331                         }
3332                 }
3333
3334                 /* Recalculate the ip checksum. */
3335                 ip = mtod(*mp, struct ip *);
3336                 hlen = ip->ip_hl << 2;
3337                 if (hlen < sizeof(struct ip))
3338                         goto bad;
3339                 if (hlen > (*mp)->m_len) {
3340                         if ((*mp = m_pullup(*mp, hlen)) == NULL)
3341                                 goto bad;
3342                         ip = mtod(*mp, struct ip *);
3343                         if (ip == NULL)
3344                                 goto bad;
3345                 }
3346                 ip->ip_sum = 0;
3347                 if (hlen == sizeof(struct ip))
3348                         ip->ip_sum = in_cksum_hdr(ip);
3349                 else
3350                         ip->ip_sum = in_cksum(*mp, hlen);
3351
3352                 break;
3353 #ifdef INET6
3354         case ETHERTYPE_IPV6:
3355                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL && (rv =
3356                     pfil_run_hooks(V_inet6_pfil_head, mp, bifp, dir, NULL)) !=
3357                     PFIL_PASS)
3358                         break;
3359
3360                 if (V_pfil_member && ifp != NULL && (rv =
3361                     pfil_run_hooks(V_inet6_pfil_head, mp, ifp, dir, NULL)) !=
3362                     PFIL_PASS)
3363                         break;
3364
3365                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv =
3366                     pfil_run_hooks(V_inet6_pfil_head, mp, bifp, dir, NULL)) !=
3367                     PFIL_PASS)
3368                         break;
3369                 break;
3370 #endif
3371         }
3372
3373         switch (rv) {
3374         case PFIL_CONSUMED:
3375                 return (0);
3376         case PFIL_DROPPED:
3377                 return (EACCES);
3378         default:
3379                 break;
3380         }
3381
3382         error = -1;
3383
3384         /*
3385          * Finally, put everything back the way it was and return
3386          */
3387         if (snap) {
3388                 M_PREPEND(*mp, sizeof(struct llc), M_NOWAIT);
3389                 if (*mp == NULL)
3390                         return (error);
3391                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3392         }
3393
3394         M_PREPEND(*mp, ETHER_HDR_LEN, M_NOWAIT);
3395         if (*mp == NULL)
3396                 return (error);
3397         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3398
3399         return (0);
3400
3401 bad:
3402         m_freem(*mp);
3403         *mp = NULL;
3404         return (error);
3405 }
3406
3407 /*
3408  * Perform basic checks on header size since
3409  * pfil assumes ip_input has already processed
3410  * it for it.  Cut-and-pasted from ip_input.c.
3411  * Given how simple the IPv6 version is,
3412  * does the IPv4 version really need to be
3413  * this complicated?
3414  *
3415  * XXX Should we update ipstat here, or not?
3416  * XXX Right now we update ipstat but not
3417  * XXX csum_counter.
3418  */
3419 static int
3420 bridge_ip_checkbasic(struct mbuf **mp)
3421 {
3422         struct mbuf *m = *mp;
3423         struct ip *ip;
3424         int len, hlen;
3425         u_short sum;
3426
3427         if (*mp == NULL)
3428                 return (-1);
3429
3430         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3431                 if ((m = m_copyup(m, sizeof(struct ip),
3432                         (max_linkhdr + 3) & ~3)) == NULL) {
3433                         /* XXXJRT new stat, please */
3434                         KMOD_IPSTAT_INC(ips_toosmall);
3435                         goto bad;
3436                 }
3437         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3438                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3439                         KMOD_IPSTAT_INC(ips_toosmall);
3440                         goto bad;
3441                 }
3442         }
3443         ip = mtod(m, struct ip *);
3444         if (ip == NULL) goto bad;
3445
3446         if (ip->ip_v != IPVERSION) {
3447                 KMOD_IPSTAT_INC(ips_badvers);
3448                 goto bad;
3449         }
3450         hlen = ip->ip_hl << 2;
3451         if (hlen < sizeof(struct ip)) { /* minimum header length */
3452                 KMOD_IPSTAT_INC(ips_badhlen);
3453                 goto bad;
3454         }
3455         if (hlen > m->m_len) {
3456                 if ((m = m_pullup(m, hlen)) == NULL) {
3457                         KMOD_IPSTAT_INC(ips_badhlen);
3458                         goto bad;
3459                 }
3460                 ip = mtod(m, struct ip *);
3461                 if (ip == NULL) goto bad;
3462         }
3463
3464         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3465                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3466         } else {
3467                 if (hlen == sizeof(struct ip)) {
3468                         sum = in_cksum_hdr(ip);
3469                 } else {
3470                         sum = in_cksum(m, hlen);
3471                 }
3472         }
3473         if (sum) {
3474                 KMOD_IPSTAT_INC(ips_badsum);
3475                 goto bad;
3476         }
3477
3478         /* Retrieve the packet length. */
3479         len = ntohs(ip->ip_len);
3480
3481         /*
3482          * Check for additional length bogosity
3483          */
3484         if (len < hlen) {
3485                 KMOD_IPSTAT_INC(ips_badlen);
3486                 goto bad;
3487         }
3488
3489         /*
3490          * Check that the amount of data in the buffers
3491          * is as at least much as the IP header would have us expect.
3492          * Drop packet if shorter than we expect.
3493          */
3494         if (m->m_pkthdr.len < len) {
3495                 KMOD_IPSTAT_INC(ips_tooshort);
3496                 goto bad;
3497         }
3498
3499         /* Checks out, proceed */
3500         *mp = m;
3501         return (0);
3502
3503 bad:
3504         *mp = m;
3505         return (-1);
3506 }
3507
3508 #ifdef INET6
3509 /*
3510  * Same as above, but for IPv6.
3511  * Cut-and-pasted from ip6_input.c.
3512  * XXX Should we update ip6stat, or not?
3513  */
3514 static int
3515 bridge_ip6_checkbasic(struct mbuf **mp)
3516 {
3517         struct mbuf *m = *mp;
3518         struct ip6_hdr *ip6;
3519
3520         /*
3521          * If the IPv6 header is not aligned, slurp it up into a new
3522          * mbuf with space for link headers, in the event we forward
3523          * it.  Otherwise, if it is aligned, make sure the entire base
3524          * IPv6 header is in the first mbuf of the chain.
3525          */
3526         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3527                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3528                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3529                             (max_linkhdr + 3) & ~3)) == NULL) {
3530                         /* XXXJRT new stat, please */
3531                         IP6STAT_INC(ip6s_toosmall);
3532                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3533                         goto bad;
3534                 }
3535         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3536                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3537                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3538                         IP6STAT_INC(ip6s_toosmall);
3539                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3540                         goto bad;
3541                 }
3542         }
3543
3544         ip6 = mtod(m, struct ip6_hdr *);
3545
3546         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3547                 IP6STAT_INC(ip6s_badvers);
3548                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3549                 goto bad;
3550         }
3551
3552         /* Checks out, proceed */
3553         *mp = m;
3554         return (0);
3555
3556 bad:
3557         *mp = m;
3558         return (-1);
3559 }
3560 #endif /* INET6 */
3561
3562 /*
3563  * bridge_fragment:
3564  *
3565  *      Fragment mbuf chain in multiple packets and prepend ethernet header.
3566  */
3567 static int
3568 bridge_fragment(struct ifnet *ifp, struct mbuf **mp, struct ether_header *eh,
3569     int snap, struct llc *llc)
3570 {
3571         struct mbuf *m = *mp, *nextpkt = NULL, *mprev = NULL, *mcur = NULL;
3572         struct ip *ip;
3573         int error = -1;
3574
3575         if (m->m_len < sizeof(struct ip) &&
3576             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3577                 goto dropit;
3578         ip = mtod(m, struct ip *);
3579
3580         m->m_pkthdr.csum_flags |= CSUM_IP;
3581         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist);
3582         if (error)
3583                 goto dropit;
3584
3585         /*
3586          * Walk the chain and re-add the Ethernet header for
3587          * each mbuf packet.
3588          */
3589         for (mcur = m; mcur; mcur = mcur->m_nextpkt) {
3590                 nextpkt = mcur->m_nextpkt;
3591                 mcur->m_nextpkt = NULL;
3592                 if (snap) {
3593                         M_PREPEND(mcur, sizeof(struct llc), M_NOWAIT);
3594                         if (mcur == NULL) {
3595                                 error = ENOBUFS;
3596                                 if (mprev != NULL)
3597                                         mprev->m_nextpkt = nextpkt;
3598                                 goto dropit;
3599                         }
3600                         bcopy(llc, mtod(mcur, caddr_t),sizeof(struct llc));
3601                 }
3602
3603                 M_PREPEND(mcur, ETHER_HDR_LEN, M_NOWAIT);
3604                 if (mcur == NULL) {
3605                         error = ENOBUFS;
3606                         if (mprev != NULL)
3607                                 mprev->m_nextpkt = nextpkt;
3608                         goto dropit;
3609                 }
3610                 bcopy(eh, mtod(mcur, caddr_t), ETHER_HDR_LEN);
3611
3612                 /*
3613                  * The previous two M_PREPEND could have inserted one or two
3614                  * mbufs in front so we have to update the previous packet's
3615                  * m_nextpkt.
3616                  */
3617                 mcur->m_nextpkt = nextpkt;
3618                 if (mprev != NULL)
3619                         mprev->m_nextpkt = mcur;
3620                 else {
3621                         /* The first mbuf in the original chain needs to be
3622                          * updated. */
3623                         *mp = mcur;
3624                 }
3625                 mprev = mcur;
3626         }
3627
3628         KMOD_IPSTAT_INC(ips_fragmented);
3629         return (error);
3630
3631 dropit:
3632         for (mcur = *mp; mcur; mcur = m) { /* droping the full packet chain */
3633                 m = mcur->m_nextpkt;
3634                 m_freem(mcur);
3635         }
3636         return (error);
3637 }
3638
3639 static void
3640 bridge_linkstate(struct ifnet *ifp)
3641 {
3642         struct bridge_softc *sc = ifp->if_bridge;
3643         struct bridge_iflist *bif;
3644
3645         BRIDGE_LOCK(sc);
3646         bif = bridge_lookup_member_if(sc, ifp);
3647         if (bif == NULL) {
3648                 BRIDGE_UNLOCK(sc);
3649                 return;
3650         }
3651         bridge_linkcheck(sc);
3652         BRIDGE_UNLOCK(sc);
3653
3654         bstp_linkstate(&bif->bif_stp);
3655 }
3656
3657 static void
3658 bridge_linkcheck(struct bridge_softc *sc)
3659 {
3660         struct bridge_iflist *bif;
3661         int new_link, hasls;
3662
3663         BRIDGE_LOCK_ASSERT(sc);
3664         new_link = LINK_STATE_DOWN;
3665         hasls = 0;
3666         /* Our link is considered up if at least one of our ports is active */
3667         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
3668                 if (bif->bif_ifp->if_capabilities & IFCAP_LINKSTATE)
3669                         hasls++;
3670                 if (bif->bif_ifp->if_link_state == LINK_STATE_UP) {
3671                         new_link = LINK_STATE_UP;
3672                         break;
3673                 }
3674         }
3675         if (!LIST_EMPTY(&sc->sc_iflist) && !hasls) {
3676                 /* If no interfaces support link-state then we default to up */
3677                 new_link = LINK_STATE_UP;
3678         }
3679         if_link_state_change(sc->sc_ifp, new_link);
3680 }