]> CyberLeo.Net >> Repos - FreeBSD/stable/9.git/blob - sys/net/if_bridge.c
MFC r234487
[FreeBSD/stable/9.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  * Copyright 2001 Wasabi Systems, Inc.
5  * All rights reserved.
6  *
7  * Written by Jason R. Thorpe for Wasabi Systems, Inc.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. All advertising materials mentioning features or use of this software
18  *    must display the following acknowledgement:
19  *      This product includes software developed for the NetBSD Project by
20  *      Wasabi Systems, Inc.
21  * 4. The name of Wasabi Systems, Inc. may not be used to endorse
22  *    or promote products derived from this software without specific prior
23  *    written permission.
24  *
25  * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
27  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
28  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL WASABI SYSTEMS, INC
29  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
32  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
33  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
34  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
35  * POSSIBILITY OF SUCH DAMAGE.
36  */
37
38 /*
39  * Copyright (c) 1999, 2000 Jason L. Wright (jason@thought.net)
40  * All rights reserved.
41  *
42  * Redistribution and use in source and binary forms, with or without
43  * modification, are permitted provided that the following conditions
44  * are met:
45  * 1. Redistributions of source code must retain the above copyright
46  *    notice, this list of conditions and the following disclaimer.
47  * 2. Redistributions in binary form must reproduce the above copyright
48  *    notice, this list of conditions and the following disclaimer in the
49  *    documentation and/or other materials provided with the distribution.
50  *
51  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
52  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
53  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
54  * DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
55  * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
56  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
57  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
58  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
59  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
60  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
61  * POSSIBILITY OF SUCH DAMAGE.
62  *
63  * OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun Exp
64  */
65
66 /*
67  * Network interface bridge support.
68  *
69  * TODO:
70  *
71  *      - Currently only supports Ethernet-like interfaces (Ethernet,
72  *        802.11, VLANs on Ethernet, etc.)  Figure out a nice way
73  *        to bridge other types of interfaces (FDDI-FDDI, and maybe
74  *        consider heterogenous bridges).
75  */
76
77 #include <sys/cdefs.h>
78 __FBSDID("$FreeBSD$");
79
80 #include "opt_inet.h"
81 #include "opt_inet6.h"
82
83 #include <sys/param.h>
84 #include <sys/mbuf.h>
85 #include <sys/malloc.h>
86 #include <sys/protosw.h>
87 #include <sys/systm.h>
88 #include <sys/jail.h>
89 #include <sys/time.h>
90 #include <sys/socket.h> /* for net/if.h */
91 #include <sys/sockio.h>
92 #include <sys/ctype.h>  /* string functions */
93 #include <sys/kernel.h>
94 #include <sys/random.h>
95 #include <sys/syslog.h>
96 #include <sys/sysctl.h>
97 #include <vm/uma.h>
98 #include <sys/module.h>
99 #include <sys/priv.h>
100 #include <sys/proc.h>
101 #include <sys/lock.h>
102 #include <sys/mutex.h>
103 #include <sys/rwlock.h>
104
105 #include <net/bpf.h>
106 #include <net/if.h>
107 #include <net/if_clone.h>
108 #include <net/if_dl.h>
109 #include <net/if_types.h>
110 #include <net/if_var.h>
111 #include <net/pfil.h>
112 #include <net/vnet.h>
113
114 #include <netinet/in.h> /* for struct arpcom */
115 #include <netinet/in_systm.h>
116 #include <netinet/in_var.h>
117 #include <netinet/ip.h>
118 #include <netinet/ip_var.h>
119 #ifdef INET6
120 #include <netinet/ip6.h>
121 #include <netinet6/ip6_var.h>
122 #endif
123 #if defined(INET) || defined(INET6)
124 #include <netinet/ip_carp.h>
125 #endif
126 #include <machine/in_cksum.h>
127 #include <netinet/if_ether.h> /* for struct arpcom */
128 #include <net/bridgestp.h>
129 #include <net/if_bridgevar.h>
130 #include <net/if_llc.h>
131 #include <net/if_vlan_var.h>
132
133 #include <net/route.h>
134 #include <netinet/ip_fw.h>
135 #include <netinet/ipfw/ip_fw_private.h>
136
137 /*
138  * Size of the route hash table.  Must be a power of two.
139  */
140 #ifndef BRIDGE_RTHASH_SIZE
141 #define BRIDGE_RTHASH_SIZE              1024
142 #endif
143
144 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
145
146 /*
147  * Default maximum number of addresses to cache.
148  */
149 #ifndef BRIDGE_RTABLE_MAX
150 #define BRIDGE_RTABLE_MAX               2000
151 #endif
152
153 /*
154  * Timeout (in seconds) for entries learned dynamically.
155  */
156 #ifndef BRIDGE_RTABLE_TIMEOUT
157 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
158 #endif
159
160 /*
161  * Number of seconds between walks of the route list.
162  */
163 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
164 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
165 #endif
166
167 /*
168  * List of capabilities to possibly mask on the member interface.
169  */
170 #define BRIDGE_IFCAPS_MASK              (IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM)
171
172 /*
173  * List of capabilities to strip
174  */
175 #define BRIDGE_IFCAPS_STRIP             IFCAP_LRO
176
177 /*
178  * Bridge interface list entry.
179  */
180 struct bridge_iflist {
181         LIST_ENTRY(bridge_iflist) bif_next;
182         struct ifnet            *bif_ifp;       /* member if */
183         struct bstp_port        bif_stp;        /* STP state */
184         uint32_t                bif_flags;      /* member if flags */
185         int                     bif_savedcaps;  /* saved capabilities */
186         uint32_t                bif_addrmax;    /* max # of addresses */
187         uint32_t                bif_addrcnt;    /* cur. # of addresses */
188         uint32_t                bif_addrexceeded;/* # of address violations */
189 };
190
191 /*
192  * Bridge route node.
193  */
194 struct bridge_rtnode {
195         LIST_ENTRY(bridge_rtnode) brt_hash;     /* hash table linkage */
196         LIST_ENTRY(bridge_rtnode) brt_list;     /* list linkage */
197         struct bridge_iflist    *brt_dst;       /* destination if */
198         unsigned long           brt_expire;     /* expiration time */
199         uint8_t                 brt_flags;      /* address flags */
200         uint8_t                 brt_addr[ETHER_ADDR_LEN];
201         uint16_t                brt_vlan;       /* vlan id */
202 };
203 #define brt_ifp                 brt_dst->bif_ifp
204
205 /*
206  * Software state for each bridge.
207  */
208 struct bridge_softc {
209         struct ifnet            *sc_ifp;        /* make this an interface */
210         LIST_ENTRY(bridge_softc) sc_list;
211         struct mtx              sc_mtx;
212         struct cv               sc_cv;
213         uint32_t                sc_brtmax;      /* max # of addresses */
214         uint32_t                sc_brtcnt;      /* cur. # of addresses */
215         uint32_t                sc_brttimeout;  /* rt timeout in seconds */
216         struct callout          sc_brcallout;   /* bridge callout */
217         uint32_t                sc_iflist_ref;  /* refcount for sc_iflist */
218         uint32_t                sc_iflist_xcnt; /* refcount for sc_iflist */
219         LIST_HEAD(, bridge_iflist) sc_iflist;   /* member interface list */
220         LIST_HEAD(, bridge_rtnode) *sc_rthash;  /* our forwarding table */
221         LIST_HEAD(, bridge_rtnode) sc_rtlist;   /* list version of above */
222         uint32_t                sc_rthash_key;  /* key for hash */
223         LIST_HEAD(, bridge_iflist) sc_spanlist; /* span ports list */
224         struct bstp_state       sc_stp;         /* STP state */
225         uint32_t                sc_brtexceeded; /* # of cache drops */
226         struct ifnet            *sc_ifaddr;     /* member mac copied from */
227         u_char                  sc_defaddr[6];  /* Default MAC address */
228 };
229
230 static struct mtx       bridge_list_mtx;
231 eventhandler_tag        bridge_detach_cookie = NULL;
232
233 int     bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
234
235 uma_zone_t bridge_rtnode_zone;
236
237 static int      bridge_clone_create(struct if_clone *, int, caddr_t);
238 static void     bridge_clone_destroy(struct ifnet *);
239
240 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t);
241 static void     bridge_mutecaps(struct bridge_softc *);
242 static void     bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
243                     int);
244 static void     bridge_ifdetach(void *arg __unused, struct ifnet *);
245 static void     bridge_init(void *);
246 static void     bridge_dummynet(struct mbuf *, struct ifnet *);
247 static void     bridge_stop(struct ifnet *, int);
248 static void     bridge_start(struct ifnet *);
249 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
250 static int      bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
251                     struct rtentry *);
252 static void     bridge_enqueue(struct bridge_softc *, struct ifnet *,
253                     struct mbuf *);
254 static void     bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
255
256 static void     bridge_forward(struct bridge_softc *, struct bridge_iflist *,
257                     struct mbuf *m);
258
259 static void     bridge_timer(void *);
260
261 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
262                     struct mbuf *, int);
263 static void     bridge_span(struct bridge_softc *, struct mbuf *);
264
265 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
266                     uint16_t, struct bridge_iflist *, int, uint8_t);
267 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
268                     uint16_t);
269 static void     bridge_rttrim(struct bridge_softc *);
270 static void     bridge_rtage(struct bridge_softc *);
271 static void     bridge_rtflush(struct bridge_softc *, int);
272 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
273                     uint16_t);
274
275 static int      bridge_rtable_init(struct bridge_softc *);
276 static void     bridge_rtable_fini(struct bridge_softc *);
277
278 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
279 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
280                     const uint8_t *, uint16_t);
281 static int      bridge_rtnode_insert(struct bridge_softc *,
282                     struct bridge_rtnode *);
283 static void     bridge_rtnode_destroy(struct bridge_softc *,
284                     struct bridge_rtnode *);
285 static void     bridge_rtable_expire(struct ifnet *, int);
286 static void     bridge_state_change(struct ifnet *, int);
287
288 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
289                     const char *name);
290 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
291                     struct ifnet *ifp);
292 static void     bridge_delete_member(struct bridge_softc *,
293                     struct bridge_iflist *, int);
294 static void     bridge_delete_span(struct bridge_softc *,
295                     struct bridge_iflist *);
296
297 static int      bridge_ioctl_add(struct bridge_softc *, void *);
298 static int      bridge_ioctl_del(struct bridge_softc *, void *);
299 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
300 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
301 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
302 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
303 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
304 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
305 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
306 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
307 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
308 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
309 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
310 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
311 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
312 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
313 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
314 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
315 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
316 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
317 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
318 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
319 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
320 static int      bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
321 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
322 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
323 static int      bridge_ioctl_gbparam(struct bridge_softc *, void *);
324 static int      bridge_ioctl_grte(struct bridge_softc *, void *);
325 static int      bridge_ioctl_gifsstp(struct bridge_softc *, void *);
326 static int      bridge_ioctl_sproto(struct bridge_softc *, void *);
327 static int      bridge_ioctl_stxhc(struct bridge_softc *, void *);
328 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
329                     int);
330 static int      bridge_ip_checkbasic(struct mbuf **mp);
331 #ifdef INET6
332 static int      bridge_ip6_checkbasic(struct mbuf **mp);
333 #endif /* INET6 */
334 static int      bridge_fragment(struct ifnet *, struct mbuf *,
335                     struct ether_header *, int, struct llc *);
336 static void     bridge_linkstate(struct ifnet *ifp);
337
338 extern void (*bridge_linkstate_p)(struct ifnet *ifp);
339
340 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
341 #define VLANTAGOF(_m)   \
342     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : 1
343
344 static struct bstp_cb_ops bridge_ops = {
345         .bcb_state = bridge_state_change,
346         .bcb_rtage = bridge_rtable_expire
347 };
348
349 SYSCTL_DECL(_net_link);
350 SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge");
351
352 static int pfil_onlyip = 1; /* only pass IP[46] packets when pfil is enabled */
353 static int pfil_bridge = 1; /* run pfil hooks on the bridge interface */
354 static int pfil_member = 1; /* run pfil hooks on the member interface */
355 static int pfil_ipfw = 0;   /* layer2 filter with ipfw */
356 static int pfil_ipfw_arp = 0;   /* layer2 filter with ipfw */
357 static int pfil_local_phys = 0; /* run pfil hooks on the physical interface for
358                                    locally destined packets */
359 static int log_stp   = 0;   /* log STP state changes */
360 static int bridge_inherit_mac = 0;   /* share MAC with first bridge member */
361 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip, CTLFLAG_RW,
362     &pfil_onlyip, 0, "Only pass IP packets when pfil is enabled");
363 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp, CTLFLAG_RW,
364     &pfil_ipfw_arp, 0, "Filter ARP packets through IPFW layer2");
365 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge, CTLFLAG_RW,
366     &pfil_bridge, 0, "Packet filter on the bridge interface");
367 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member, CTLFLAG_RW,
368     &pfil_member, 0, "Packet filter on the member interface");
369 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys, CTLFLAG_RW,
370     &pfil_local_phys, 0,
371     "Packet filter on the physical interface for locally destined packets");
372 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp, CTLFLAG_RW,
373     &log_stp, 0, "Log STP state changes");
374 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac, CTLFLAG_RW,
375     &bridge_inherit_mac, 0,
376     "Inherit MAC address from the first bridge member");
377
378 struct bridge_control {
379         int     (*bc_func)(struct bridge_softc *, void *);
380         int     bc_argsize;
381         int     bc_flags;
382 };
383
384 #define BC_F_COPYIN             0x01    /* copy arguments in */
385 #define BC_F_COPYOUT            0x02    /* copy arguments out */
386 #define BC_F_SUSER              0x04    /* do super-user check */
387
388 const struct bridge_control bridge_control_table[] = {
389         { bridge_ioctl_add,             sizeof(struct ifbreq),
390           BC_F_COPYIN|BC_F_SUSER },
391         { bridge_ioctl_del,             sizeof(struct ifbreq),
392           BC_F_COPYIN|BC_F_SUSER },
393
394         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
395           BC_F_COPYIN|BC_F_COPYOUT },
396         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
397           BC_F_COPYIN|BC_F_SUSER },
398
399         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
400           BC_F_COPYIN|BC_F_SUSER },
401         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
402           BC_F_COPYOUT },
403
404         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
405           BC_F_COPYIN|BC_F_COPYOUT },
406         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
407           BC_F_COPYIN|BC_F_COPYOUT },
408
409         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
410           BC_F_COPYIN|BC_F_SUSER },
411
412         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
413           BC_F_COPYIN|BC_F_SUSER },
414         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
415           BC_F_COPYOUT },
416
417         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
418           BC_F_COPYIN|BC_F_SUSER },
419
420         { bridge_ioctl_flush,           sizeof(struct ifbreq),
421           BC_F_COPYIN|BC_F_SUSER },
422
423         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
424           BC_F_COPYOUT },
425         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
426           BC_F_COPYIN|BC_F_SUSER },
427
428         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
429           BC_F_COPYOUT },
430         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
431           BC_F_COPYIN|BC_F_SUSER },
432
433         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
434           BC_F_COPYOUT },
435         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
436           BC_F_COPYIN|BC_F_SUSER },
437
438         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
439           BC_F_COPYOUT },
440         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
441           BC_F_COPYIN|BC_F_SUSER },
442
443         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
444           BC_F_COPYIN|BC_F_SUSER },
445
446         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
447           BC_F_COPYIN|BC_F_SUSER },
448
449         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
450           BC_F_COPYIN|BC_F_SUSER },
451         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
452           BC_F_COPYIN|BC_F_SUSER },
453
454         { bridge_ioctl_gbparam,         sizeof(struct ifbropreq),
455           BC_F_COPYOUT },
456
457         { bridge_ioctl_grte,            sizeof(struct ifbrparam),
458           BC_F_COPYOUT },
459
460         { bridge_ioctl_gifsstp,         sizeof(struct ifbpstpconf),
461           BC_F_COPYIN|BC_F_COPYOUT },
462
463         { bridge_ioctl_sproto,          sizeof(struct ifbrparam),
464           BC_F_COPYIN|BC_F_SUSER },
465
466         { bridge_ioctl_stxhc,           sizeof(struct ifbrparam),
467           BC_F_COPYIN|BC_F_SUSER },
468
469         { bridge_ioctl_sifmaxaddr,      sizeof(struct ifbreq),
470           BC_F_COPYIN|BC_F_SUSER },
471
472 };
473 const int bridge_control_table_size =
474     sizeof(bridge_control_table) / sizeof(bridge_control_table[0]);
475
476 LIST_HEAD(, bridge_softc) bridge_list;
477
478 IFC_SIMPLE_DECLARE(bridge, 0);
479
480 static int
481 bridge_modevent(module_t mod, int type, void *data)
482 {
483
484         switch (type) {
485         case MOD_LOAD:
486                 mtx_init(&bridge_list_mtx, "if_bridge list", NULL, MTX_DEF);
487                 if_clone_attach(&bridge_cloner);
488                 bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
489                     sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
490                     UMA_ALIGN_PTR, 0);
491                 LIST_INIT(&bridge_list);
492                 bridge_input_p = bridge_input;
493                 bridge_output_p = bridge_output;
494                 bridge_dn_p = bridge_dummynet;
495                 bridge_linkstate_p = bridge_linkstate;
496                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
497                     ifnet_departure_event, bridge_ifdetach, NULL,
498                     EVENTHANDLER_PRI_ANY);
499                 break;
500         case MOD_UNLOAD:
501                 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
502                     bridge_detach_cookie);
503                 if_clone_detach(&bridge_cloner);
504                 uma_zdestroy(bridge_rtnode_zone);
505                 bridge_input_p = NULL;
506                 bridge_output_p = NULL;
507                 bridge_dn_p = NULL;
508                 bridge_linkstate_p = NULL;
509                 mtx_destroy(&bridge_list_mtx);
510                 break;
511         default:
512                 return (EOPNOTSUPP);
513         }
514         return (0);
515 }
516
517 static moduledata_t bridge_mod = {
518         "if_bridge",
519         bridge_modevent,
520         0
521 };
522
523 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
524 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
525
526 /*
527  * handler for net.link.bridge.pfil_ipfw
528  */
529 static int
530 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
531 {
532         int enable = pfil_ipfw;
533         int error;
534
535         error = sysctl_handle_int(oidp, &enable, 0, req);
536         enable = (enable) ? 1 : 0;
537
538         if (enable != pfil_ipfw) {
539                 pfil_ipfw = enable;
540
541                 /*
542                  * Disable pfil so that ipfw doesnt run twice, if the user
543                  * really wants both then they can re-enable pfil_bridge and/or
544                  * pfil_member. Also allow non-ip packets as ipfw can filter by
545                  * layer2 type.
546                  */
547                 if (pfil_ipfw) {
548                         pfil_onlyip = 0;
549                         pfil_bridge = 0;
550                         pfil_member = 0;
551                 }
552         }
553
554         return (error);
555 }
556 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw, CTLTYPE_INT|CTLFLAG_RW,
557             &pfil_ipfw, 0, &sysctl_pfil_ipfw, "I", "Layer2 filter with IPFW");
558
559 /*
560  * bridge_clone_create:
561  *
562  *      Create a new bridge instance.
563  */
564 static int
565 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t params)
566 {
567         struct bridge_softc *sc, *sc2;
568         struct ifnet *bifp, *ifp;
569         int fb, retry;
570         unsigned long hostid;
571
572         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
573         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
574         if (ifp == NULL) {
575                 free(sc, M_DEVBUF);
576                 return (ENOSPC);
577         }
578
579         BRIDGE_LOCK_INIT(sc);
580         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
581         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
582
583         /* Initialize our routing table. */
584         bridge_rtable_init(sc);
585
586         callout_init_mtx(&sc->sc_brcallout, &sc->sc_mtx, 0);
587
588         LIST_INIT(&sc->sc_iflist);
589         LIST_INIT(&sc->sc_spanlist);
590
591         ifp->if_softc = sc;
592         if_initname(ifp, ifc->ifc_name, unit);
593         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
594         ifp->if_ioctl = bridge_ioctl;
595         ifp->if_start = bridge_start;
596         ifp->if_init = bridge_init;
597         ifp->if_type = IFT_BRIDGE;
598         IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
599         ifp->if_snd.ifq_drv_maxlen = ifqmaxlen;
600         IFQ_SET_READY(&ifp->if_snd);
601
602         /*
603          * Generate an ethernet address with a locally administered address.
604          *
605          * Since we are using random ethernet addresses for the bridge, it is
606          * possible that we might have address collisions, so make sure that
607          * this hardware address isn't already in use on another bridge.
608          * The first try uses the hostid and falls back to arc4rand().
609          */
610         fb = 0;
611         getcredhostid(curthread->td_ucred, &hostid);
612         for (retry = 1; retry != 0;) {
613                 if (fb || hostid == 0) {
614                         arc4rand(sc->sc_defaddr, ETHER_ADDR_LEN, 1);
615                         sc->sc_defaddr[0] &= ~1;/* clear multicast bit */
616                         sc->sc_defaddr[0] |= 2; /* set the LAA bit */
617                 } else {
618                         sc->sc_defaddr[0] = 0x2;
619                         sc->sc_defaddr[1] = (hostid >> 24) & 0xff;
620                         sc->sc_defaddr[2] = (hostid >> 16) & 0xff;
621                         sc->sc_defaddr[3] = (hostid >> 8 ) & 0xff;
622                         sc->sc_defaddr[4] =  hostid        & 0xff;
623                         sc->sc_defaddr[5] = ifp->if_dunit & 0xff;
624                 }
625
626                 fb = 1;
627                 retry = 0;
628                 mtx_lock(&bridge_list_mtx);
629                 LIST_FOREACH(sc2, &bridge_list, sc_list) {
630                         bifp = sc2->sc_ifp;
631                         if (memcmp(sc->sc_defaddr,
632                             IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0)
633                                 retry = 1;
634                 }
635                 mtx_unlock(&bridge_list_mtx);
636         }
637
638         bstp_attach(&sc->sc_stp, &bridge_ops);
639         ether_ifattach(ifp, sc->sc_defaddr);
640         /* Now undo some of the damage... */
641         ifp->if_baudrate = 0;
642         ifp->if_type = IFT_BRIDGE;
643
644         mtx_lock(&bridge_list_mtx);
645         LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
646         mtx_unlock(&bridge_list_mtx);
647
648         return (0);
649 }
650
651 /*
652  * bridge_clone_destroy:
653  *
654  *      Destroy a bridge instance.
655  */
656 static void
657 bridge_clone_destroy(struct ifnet *ifp)
658 {
659         struct bridge_softc *sc = ifp->if_softc;
660         struct bridge_iflist *bif;
661
662         BRIDGE_LOCK(sc);
663
664         bridge_stop(ifp, 1);
665         ifp->if_flags &= ~IFF_UP;
666
667         while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
668                 bridge_delete_member(sc, bif, 0);
669
670         while ((bif = LIST_FIRST(&sc->sc_spanlist)) != NULL) {
671                 bridge_delete_span(sc, bif);
672         }
673
674         BRIDGE_UNLOCK(sc);
675
676         callout_drain(&sc->sc_brcallout);
677
678         mtx_lock(&bridge_list_mtx);
679         LIST_REMOVE(sc, sc_list);
680         mtx_unlock(&bridge_list_mtx);
681
682         bstp_detach(&sc->sc_stp);
683         ether_ifdetach(ifp);
684         if_free_type(ifp, IFT_ETHER);
685
686         /* Tear down the routing table. */
687         bridge_rtable_fini(sc);
688
689         BRIDGE_LOCK_DESTROY(sc);
690         free(sc, M_DEVBUF);
691 }
692
693 /*
694  * bridge_ioctl:
695  *
696  *      Handle a control request from the operator.
697  */
698 static int
699 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
700 {
701         struct bridge_softc *sc = ifp->if_softc;
702         struct ifreq *ifr = (struct ifreq *)data;
703         struct bridge_iflist *bif;
704         struct thread *td = curthread;
705         union {
706                 struct ifbreq ifbreq;
707                 struct ifbifconf ifbifconf;
708                 struct ifbareq ifbareq;
709                 struct ifbaconf ifbaconf;
710                 struct ifbrparam ifbrparam;
711                 struct ifbropreq ifbropreq;
712         } args;
713         struct ifdrv *ifd = (struct ifdrv *) data;
714         const struct bridge_control *bc;
715         int error = 0;
716
717         switch (cmd) {
718
719         case SIOCADDMULTI:
720         case SIOCDELMULTI:
721                 break;
722
723         case SIOCGDRVSPEC:
724         case SIOCSDRVSPEC:
725                 if (ifd->ifd_cmd >= bridge_control_table_size) {
726                         error = EINVAL;
727                         break;
728                 }
729                 bc = &bridge_control_table[ifd->ifd_cmd];
730
731                 if (cmd == SIOCGDRVSPEC &&
732                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
733                         error = EINVAL;
734                         break;
735                 }
736                 else if (cmd == SIOCSDRVSPEC &&
737                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
738                         error = EINVAL;
739                         break;
740                 }
741
742                 if (bc->bc_flags & BC_F_SUSER) {
743                         error = priv_check(td, PRIV_NET_BRIDGE);
744                         if (error)
745                                 break;
746                 }
747
748                 if (ifd->ifd_len != bc->bc_argsize ||
749                     ifd->ifd_len > sizeof(args)) {
750                         error = EINVAL;
751                         break;
752                 }
753
754                 bzero(&args, sizeof(args));
755                 if (bc->bc_flags & BC_F_COPYIN) {
756                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
757                         if (error)
758                                 break;
759                 }
760
761                 BRIDGE_LOCK(sc);
762                 error = (*bc->bc_func)(sc, &args);
763                 BRIDGE_UNLOCK(sc);
764                 if (error)
765                         break;
766
767                 if (bc->bc_flags & BC_F_COPYOUT)
768                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
769
770                 break;
771
772         case SIOCSIFFLAGS:
773                 if (!(ifp->if_flags & IFF_UP) &&
774                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
775                         /*
776                          * If interface is marked down and it is running,
777                          * then stop and disable it.
778                          */
779                         BRIDGE_LOCK(sc);
780                         bridge_stop(ifp, 1);
781                         BRIDGE_UNLOCK(sc);
782                 } else if ((ifp->if_flags & IFF_UP) &&
783                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
784                         /*
785                          * If interface is marked up and it is stopped, then
786                          * start it.
787                          */
788                         (*ifp->if_init)(sc);
789                 }
790                 break;
791
792         case SIOCSIFMTU:
793                 if (ifr->ifr_mtu < 576) {
794                         error = EINVAL;
795                         break;
796                 }
797                 if (LIST_EMPTY(&sc->sc_iflist)) {
798                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
799                         break;
800                 }
801                 BRIDGE_LOCK(sc);
802                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
803                         if (bif->bif_ifp->if_mtu != ifr->ifr_mtu) {
804                                 log(LOG_NOTICE, "%s: invalid MTU: %lu(%s)"
805                                     " != %d\n", sc->sc_ifp->if_xname,
806                                     bif->bif_ifp->if_mtu,
807                                     bif->bif_ifp->if_xname, ifr->ifr_mtu);
808                                 error = EINVAL;
809                                 break;
810                         }
811                 }
812                 if (!error)
813                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
814                 BRIDGE_UNLOCK(sc);
815                 break;
816         default:
817                 /*
818                  * drop the lock as ether_ioctl() will call bridge_start() and
819                  * cause the lock to be recursed.
820                  */
821                 error = ether_ioctl(ifp, cmd, data);
822                 break;
823         }
824
825         return (error);
826 }
827
828 /*
829  * bridge_mutecaps:
830  *
831  *      Clear or restore unwanted capabilities on the member interface
832  */
833 static void
834 bridge_mutecaps(struct bridge_softc *sc)
835 {
836         struct bridge_iflist *bif;
837         int enabled, mask;
838
839         /* Initial bitmask of capabilities to test */
840         mask = BRIDGE_IFCAPS_MASK;
841
842         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
843                 /* Every member must support it or its disabled */
844                 mask &= bif->bif_savedcaps;
845         }
846
847         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
848                 enabled = bif->bif_ifp->if_capenable;
849                 enabled &= ~BRIDGE_IFCAPS_STRIP;
850                 /* strip off mask bits and enable them again if allowed */
851                 enabled &= ~BRIDGE_IFCAPS_MASK;
852                 enabled |= mask;
853                 bridge_set_ifcap(sc, bif, enabled);
854         }
855
856 }
857
858 static void
859 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
860 {
861         struct ifnet *ifp = bif->bif_ifp;
862         struct ifreq ifr;
863         int error;
864
865         bzero(&ifr, sizeof(ifr));
866         ifr.ifr_reqcap = set;
867
868         if (ifp->if_capenable != set) {
869                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
870                 if (error)
871                         if_printf(sc->sc_ifp,
872                             "error setting interface capabilities on %s\n",
873                             ifp->if_xname);
874         }
875 }
876
877 /*
878  * bridge_lookup_member:
879  *
880  *      Lookup a bridge member interface.
881  */
882 static struct bridge_iflist *
883 bridge_lookup_member(struct bridge_softc *sc, const char *name)
884 {
885         struct bridge_iflist *bif;
886         struct ifnet *ifp;
887
888         BRIDGE_LOCK_ASSERT(sc);
889
890         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
891                 ifp = bif->bif_ifp;
892                 if (strcmp(ifp->if_xname, name) == 0)
893                         return (bif);
894         }
895
896         return (NULL);
897 }
898
899 /*
900  * bridge_lookup_member_if:
901  *
902  *      Lookup a bridge member interface by ifnet*.
903  */
904 static struct bridge_iflist *
905 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
906 {
907         struct bridge_iflist *bif;
908
909         BRIDGE_LOCK_ASSERT(sc);
910
911         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
912                 if (bif->bif_ifp == member_ifp)
913                         return (bif);
914         }
915
916         return (NULL);
917 }
918
919 /*
920  * bridge_delete_member:
921  *
922  *      Delete the specified member interface.
923  */
924 static void
925 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
926     int gone)
927 {
928         struct ifnet *ifs = bif->bif_ifp;
929         struct ifnet *fif = NULL;
930
931         BRIDGE_LOCK_ASSERT(sc);
932
933         if (bif->bif_flags & IFBIF_STP)
934                 bstp_disable(&bif->bif_stp);
935
936         ifs->if_bridge = NULL;
937         BRIDGE_XLOCK(sc);
938         LIST_REMOVE(bif, bif_next);
939         BRIDGE_XDROP(sc);
940
941         /*
942          * If removing the interface that gave the bridge its mac address, set
943          * the mac address of the bridge to the address of the next member, or
944          * to its default address if no members are left.
945          */
946         if (bridge_inherit_mac && sc->sc_ifaddr == ifs) {
947                 if (LIST_EMPTY(&sc->sc_iflist)) {
948                         bcopy(sc->sc_defaddr,
949                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
950                         sc->sc_ifaddr = NULL;
951                 } else {
952                         fif = LIST_FIRST(&sc->sc_iflist)->bif_ifp;
953                         bcopy(IF_LLADDR(fif),
954                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
955                         sc->sc_ifaddr = fif;
956                 }
957                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
958         }
959
960         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
961         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
962         KASSERT(bif->bif_addrcnt == 0,
963             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
964
965         BRIDGE_UNLOCK(sc);
966         if (!gone) {
967                 switch (ifs->if_type) {
968                 case IFT_ETHER:
969                 case IFT_L2VLAN:
970                         /*
971                          * Take the interface out of promiscuous mode.
972                          */
973                         (void) ifpromisc(ifs, 0);
974                         break;
975
976                 case IFT_GIF:
977                         break;
978
979                 default:
980 #ifdef DIAGNOSTIC
981                         panic("bridge_delete_member: impossible");
982 #endif
983                         break;
984                 }
985                 /* reneable any interface capabilities */
986                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
987         }
988         bstp_destroy(&bif->bif_stp);    /* prepare to free */
989         bridge_linkstate(ifs);
990         BRIDGE_LOCK(sc);
991         free(bif, M_DEVBUF);
992 }
993
994 /*
995  * bridge_delete_span:
996  *
997  *      Delete the specified span interface.
998  */
999 static void
1000 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1001 {
1002         BRIDGE_LOCK_ASSERT(sc);
1003
1004         KASSERT(bif->bif_ifp->if_bridge == NULL,
1005             ("%s: not a span interface", __func__));
1006
1007         LIST_REMOVE(bif, bif_next);
1008         free(bif, M_DEVBUF);
1009 }
1010
1011 static int
1012 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1013 {
1014         struct ifbreq *req = arg;
1015         struct bridge_iflist *bif = NULL;
1016         struct ifnet *ifs;
1017         int error = 0;
1018
1019         ifs = ifunit(req->ifbr_ifsname);
1020         if (ifs == NULL)
1021                 return (ENOENT);
1022         if (ifs->if_ioctl == NULL)      /* must be supported */
1023                 return (EINVAL);
1024
1025         /* If it's in the span list, it can't be a member. */
1026         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1027                 if (ifs == bif->bif_ifp)
1028                         return (EBUSY);
1029
1030         if (ifs->if_bridge == sc)
1031                 return (EEXIST);
1032
1033         if (ifs->if_bridge != NULL)
1034                 return (EBUSY);
1035
1036         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1037         if (bif == NULL)
1038                 return (ENOMEM);
1039
1040         bif->bif_ifp = ifs;
1041         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1042         bif->bif_savedcaps = ifs->if_capenable;
1043
1044         switch (ifs->if_type) {
1045         case IFT_ETHER:
1046         case IFT_L2VLAN:
1047         case IFT_GIF:
1048                 /* permitted interface types */
1049                 break;
1050         default:
1051                 error = EINVAL;
1052                 goto out;
1053         }
1054
1055         /* Allow the first Ethernet member to define the MTU */
1056         if (LIST_EMPTY(&sc->sc_iflist))
1057                 sc->sc_ifp->if_mtu = ifs->if_mtu;
1058         else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1059                 if_printf(sc->sc_ifp, "invalid MTU: %lu(%s) != %lu\n",
1060                     ifs->if_mtu, ifs->if_xname, sc->sc_ifp->if_mtu);
1061                 error = EINVAL;
1062                 goto out;
1063         }
1064
1065         /*
1066          * Assign the interface's MAC address to the bridge if it's the first
1067          * member and the MAC address of the bridge has not been changed from
1068          * the default randomly generated one.
1069          */
1070         if (bridge_inherit_mac && LIST_EMPTY(&sc->sc_iflist) &&
1071             !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr, ETHER_ADDR_LEN)) {
1072                 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1073                 sc->sc_ifaddr = ifs;
1074                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1075         }
1076
1077         ifs->if_bridge = sc;
1078         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1079         /*
1080          * XXX: XLOCK HERE!?!
1081          *
1082          * NOTE: insert_***HEAD*** should be safe for the traversals.
1083          */
1084         LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1085
1086         /* Set interface capabilities to the intersection set of all members */
1087         bridge_mutecaps(sc);
1088
1089         BRIDGE_UNLOCK(sc);
1090         /* Update the linkstate for the bridge */
1091         bridge_linkstate(ifs);
1092         /* Place the interface into promiscuous mode */
1093         switch (ifs->if_type) {
1094                 case IFT_ETHER:
1095                 case IFT_L2VLAN:
1096                         error = ifpromisc(ifs, 1);
1097                         break;
1098         }
1099         BRIDGE_LOCK(sc);
1100         if (error)
1101                 bridge_delete_member(sc, bif, 0);
1102 out:
1103         if (error) {
1104                 if (bif != NULL)
1105                         free(bif, M_DEVBUF);
1106         }
1107         return (error);
1108 }
1109
1110 static int
1111 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1112 {
1113         struct ifbreq *req = arg;
1114         struct bridge_iflist *bif;
1115
1116         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1117         if (bif == NULL)
1118                 return (ENOENT);
1119
1120         bridge_delete_member(sc, bif, 0);
1121
1122         return (0);
1123 }
1124
1125 static int
1126 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1127 {
1128         struct ifbreq *req = arg;
1129         struct bridge_iflist *bif;
1130         struct bstp_port *bp;
1131
1132         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1133         if (bif == NULL)
1134                 return (ENOENT);
1135
1136         bp = &bif->bif_stp;
1137         req->ifbr_ifsflags = bif->bif_flags;
1138         req->ifbr_state = bp->bp_state;
1139         req->ifbr_priority = bp->bp_priority;
1140         req->ifbr_path_cost = bp->bp_path_cost;
1141         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1142         req->ifbr_proto = bp->bp_protover;
1143         req->ifbr_role = bp->bp_role;
1144         req->ifbr_stpflags = bp->bp_flags;
1145         req->ifbr_addrcnt = bif->bif_addrcnt;
1146         req->ifbr_addrmax = bif->bif_addrmax;
1147         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1148
1149         /* Copy STP state options as flags */
1150         if (bp->bp_operedge)
1151                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1152         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1153                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1154         if (bp->bp_ptp_link)
1155                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1156         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1157                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1158         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1159                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1160         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1161                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1162         return (0);
1163 }
1164
1165 static int
1166 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1167 {
1168         struct ifbreq *req = arg;
1169         struct bridge_iflist *bif;
1170         struct bstp_port *bp;
1171         int error;
1172
1173         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1174         if (bif == NULL)
1175                 return (ENOENT);
1176         bp = &bif->bif_stp;
1177
1178         if (req->ifbr_ifsflags & IFBIF_SPAN)
1179                 /* SPAN is readonly */
1180                 return (EINVAL);
1181
1182         if (req->ifbr_ifsflags & IFBIF_STP) {
1183                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1184                         error = bstp_enable(&bif->bif_stp);
1185                         if (error)
1186                                 return (error);
1187                 }
1188         } else {
1189                 if ((bif->bif_flags & IFBIF_STP) != 0)
1190                         bstp_disable(&bif->bif_stp);
1191         }
1192
1193         /* Pass on STP flags */
1194         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1195         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1196         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1197         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1198
1199         /* Save the bits relating to the bridge */
1200         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1201
1202         return (0);
1203 }
1204
1205 static int
1206 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1207 {
1208         struct ifbrparam *param = arg;
1209
1210         sc->sc_brtmax = param->ifbrp_csize;
1211         bridge_rttrim(sc);
1212
1213         return (0);
1214 }
1215
1216 static int
1217 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1218 {
1219         struct ifbrparam *param = arg;
1220
1221         param->ifbrp_csize = sc->sc_brtmax;
1222
1223         return (0);
1224 }
1225
1226 static int
1227 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1228 {
1229         struct ifbifconf *bifc = arg;
1230         struct bridge_iflist *bif;
1231         struct ifbreq breq;
1232         char *buf, *outbuf;
1233         int count, buflen, len, error = 0;
1234
1235         count = 0;
1236         LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1237                 count++;
1238         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1239                 count++;
1240
1241         buflen = sizeof(breq) * count;
1242         if (bifc->ifbic_len == 0) {
1243                 bifc->ifbic_len = buflen;
1244                 return (0);
1245         }
1246         BRIDGE_UNLOCK(sc);
1247         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1248         BRIDGE_LOCK(sc);
1249
1250         count = 0;
1251         buf = outbuf;
1252         len = min(bifc->ifbic_len, buflen);
1253         bzero(&breq, sizeof(breq));
1254         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1255                 if (len < sizeof(breq))
1256                         break;
1257
1258                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1259                     sizeof(breq.ifbr_ifsname));
1260                 /* Fill in the ifbreq structure */
1261                 error = bridge_ioctl_gifflags(sc, &breq);
1262                 if (error)
1263                         break;
1264                 memcpy(buf, &breq, sizeof(breq));
1265                 count++;
1266                 buf += sizeof(breq);
1267                 len -= sizeof(breq);
1268         }
1269         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1270                 if (len < sizeof(breq))
1271                         break;
1272
1273                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1274                     sizeof(breq.ifbr_ifsname));
1275                 breq.ifbr_ifsflags = bif->bif_flags;
1276                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1277                 memcpy(buf, &breq, sizeof(breq));
1278                 count++;
1279                 buf += sizeof(breq);
1280                 len -= sizeof(breq);
1281         }
1282
1283         BRIDGE_UNLOCK(sc);
1284         bifc->ifbic_len = sizeof(breq) * count;
1285         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1286         BRIDGE_LOCK(sc);
1287         free(outbuf, M_TEMP);
1288         return (error);
1289 }
1290
1291 static int
1292 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1293 {
1294         struct ifbaconf *bac = arg;
1295         struct bridge_rtnode *brt;
1296         struct ifbareq bareq;
1297         char *buf, *outbuf;
1298         int count, buflen, len, error = 0;
1299
1300         if (bac->ifbac_len == 0)
1301                 return (0);
1302
1303         count = 0;
1304         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1305                 count++;
1306         buflen = sizeof(bareq) * count;
1307
1308         BRIDGE_UNLOCK(sc);
1309         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1310         BRIDGE_LOCK(sc);
1311
1312         count = 0;
1313         buf = outbuf;
1314         len = min(bac->ifbac_len, buflen);
1315         bzero(&bareq, sizeof(bareq));
1316         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1317                 if (len < sizeof(bareq))
1318                         goto out;
1319                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1320                     sizeof(bareq.ifba_ifsname));
1321                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1322                 bareq.ifba_vlan = brt->brt_vlan;
1323                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1324                                 time_uptime < brt->brt_expire)
1325                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1326                 else
1327                         bareq.ifba_expire = 0;
1328                 bareq.ifba_flags = brt->brt_flags;
1329
1330                 memcpy(buf, &bareq, sizeof(bareq));
1331                 count++;
1332                 buf += sizeof(bareq);
1333                 len -= sizeof(bareq);
1334         }
1335 out:
1336         BRIDGE_UNLOCK(sc);
1337         bac->ifbac_len = sizeof(bareq) * count;
1338         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1339         BRIDGE_LOCK(sc);
1340         free(outbuf, M_TEMP);
1341         return (error);
1342 }
1343
1344 static int
1345 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1346 {
1347         struct ifbareq *req = arg;
1348         struct bridge_iflist *bif;
1349         int error;
1350
1351         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1352         if (bif == NULL)
1353                 return (ENOENT);
1354
1355         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1356             req->ifba_flags);
1357
1358         return (error);
1359 }
1360
1361 static int
1362 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1363 {
1364         struct ifbrparam *param = arg;
1365
1366         sc->sc_brttimeout = param->ifbrp_ctime;
1367         return (0);
1368 }
1369
1370 static int
1371 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1372 {
1373         struct ifbrparam *param = arg;
1374
1375         param->ifbrp_ctime = sc->sc_brttimeout;
1376         return (0);
1377 }
1378
1379 static int
1380 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1381 {
1382         struct ifbareq *req = arg;
1383
1384         return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1385 }
1386
1387 static int
1388 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1389 {
1390         struct ifbreq *req = arg;
1391
1392         bridge_rtflush(sc, req->ifbr_ifsflags);
1393         return (0);
1394 }
1395
1396 static int
1397 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1398 {
1399         struct ifbrparam *param = arg;
1400         struct bstp_state *bs = &sc->sc_stp;
1401
1402         param->ifbrp_prio = bs->bs_bridge_priority;
1403         return (0);
1404 }
1405
1406 static int
1407 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1408 {
1409         struct ifbrparam *param = arg;
1410
1411         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1412 }
1413
1414 static int
1415 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1416 {
1417         struct ifbrparam *param = arg;
1418         struct bstp_state *bs = &sc->sc_stp;
1419
1420         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1421         return (0);
1422 }
1423
1424 static int
1425 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1426 {
1427         struct ifbrparam *param = arg;
1428
1429         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1430 }
1431
1432 static int
1433 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1434 {
1435         struct ifbrparam *param = arg;
1436         struct bstp_state *bs = &sc->sc_stp;
1437
1438         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1439         return (0);
1440 }
1441
1442 static int
1443 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1444 {
1445         struct ifbrparam *param = arg;
1446
1447         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1448 }
1449
1450 static int
1451 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1452 {
1453         struct ifbrparam *param = arg;
1454         struct bstp_state *bs = &sc->sc_stp;
1455
1456         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1457         return (0);
1458 }
1459
1460 static int
1461 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1462 {
1463         struct ifbrparam *param = arg;
1464
1465         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1466 }
1467
1468 static int
1469 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1470 {
1471         struct ifbreq *req = arg;
1472         struct bridge_iflist *bif;
1473
1474         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1475         if (bif == NULL)
1476                 return (ENOENT);
1477
1478         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1479 }
1480
1481 static int
1482 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1483 {
1484         struct ifbreq *req = arg;
1485         struct bridge_iflist *bif;
1486
1487         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1488         if (bif == NULL)
1489                 return (ENOENT);
1490
1491         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1492 }
1493
1494 static int
1495 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1496 {
1497         struct ifbreq *req = arg;
1498         struct bridge_iflist *bif;
1499
1500         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1501         if (bif == NULL)
1502                 return (ENOENT);
1503
1504         bif->bif_addrmax = req->ifbr_addrmax;
1505         return (0);
1506 }
1507
1508 static int
1509 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1510 {
1511         struct ifbreq *req = arg;
1512         struct bridge_iflist *bif = NULL;
1513         struct ifnet *ifs;
1514
1515         ifs = ifunit(req->ifbr_ifsname);
1516         if (ifs == NULL)
1517                 return (ENOENT);
1518
1519         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1520                 if (ifs == bif->bif_ifp)
1521                         return (EBUSY);
1522
1523         if (ifs->if_bridge != NULL)
1524                 return (EBUSY);
1525
1526         switch (ifs->if_type) {
1527                 case IFT_ETHER:
1528                 case IFT_GIF:
1529                 case IFT_L2VLAN:
1530                         break;
1531                 default:
1532                         return (EINVAL);
1533         }
1534
1535         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1536         if (bif == NULL)
1537                 return (ENOMEM);
1538
1539         bif->bif_ifp = ifs;
1540         bif->bif_flags = IFBIF_SPAN;
1541
1542         LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1543
1544         return (0);
1545 }
1546
1547 static int
1548 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1549 {
1550         struct ifbreq *req = arg;
1551         struct bridge_iflist *bif;
1552         struct ifnet *ifs;
1553
1554         ifs = ifunit(req->ifbr_ifsname);
1555         if (ifs == NULL)
1556                 return (ENOENT);
1557
1558         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1559                 if (ifs == bif->bif_ifp)
1560                         break;
1561
1562         if (bif == NULL)
1563                 return (ENOENT);
1564
1565         bridge_delete_span(sc, bif);
1566
1567         return (0);
1568 }
1569
1570 static int
1571 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1572 {
1573         struct ifbropreq *req = arg;
1574         struct bstp_state *bs = &sc->sc_stp;
1575         struct bstp_port *root_port;
1576
1577         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1578         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1579         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1580
1581         root_port = bs->bs_root_port;
1582         if (root_port == NULL)
1583                 req->ifbop_root_port = 0;
1584         else
1585                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1586
1587         req->ifbop_holdcount = bs->bs_txholdcount;
1588         req->ifbop_priority = bs->bs_bridge_priority;
1589         req->ifbop_protocol = bs->bs_protover;
1590         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1591         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1592         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1593         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1594         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1595         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1596
1597         return (0);
1598 }
1599
1600 static int
1601 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1602 {
1603         struct ifbrparam *param = arg;
1604
1605         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1606         return (0);
1607 }
1608
1609 static int
1610 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1611 {
1612         struct ifbpstpconf *bifstp = arg;
1613         struct bridge_iflist *bif;
1614         struct bstp_port *bp;
1615         struct ifbpstpreq bpreq;
1616         char *buf, *outbuf;
1617         int count, buflen, len, error = 0;
1618
1619         count = 0;
1620         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1621                 if ((bif->bif_flags & IFBIF_STP) != 0)
1622                         count++;
1623         }
1624
1625         buflen = sizeof(bpreq) * count;
1626         if (bifstp->ifbpstp_len == 0) {
1627                 bifstp->ifbpstp_len = buflen;
1628                 return (0);
1629         }
1630
1631         BRIDGE_UNLOCK(sc);
1632         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1633         BRIDGE_LOCK(sc);
1634
1635         count = 0;
1636         buf = outbuf;
1637         len = min(bifstp->ifbpstp_len, buflen);
1638         bzero(&bpreq, sizeof(bpreq));
1639         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1640                 if (len < sizeof(bpreq))
1641                         break;
1642
1643                 if ((bif->bif_flags & IFBIF_STP) == 0)
1644                         continue;
1645
1646                 bp = &bif->bif_stp;
1647                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1648                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1649                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1650                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1651                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1652                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1653
1654                 memcpy(buf, &bpreq, sizeof(bpreq));
1655                 count++;
1656                 buf += sizeof(bpreq);
1657                 len -= sizeof(bpreq);
1658         }
1659
1660         BRIDGE_UNLOCK(sc);
1661         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1662         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1663         BRIDGE_LOCK(sc);
1664         free(outbuf, M_TEMP);
1665         return (error);
1666 }
1667
1668 static int
1669 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1670 {
1671         struct ifbrparam *param = arg;
1672
1673         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1674 }
1675
1676 static int
1677 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1678 {
1679         struct ifbrparam *param = arg;
1680
1681         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1682 }
1683
1684 /*
1685  * bridge_ifdetach:
1686  *
1687  *      Detach an interface from a bridge.  Called when a member
1688  *      interface is detaching.
1689  */
1690 static void
1691 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1692 {
1693         struct bridge_softc *sc = ifp->if_bridge;
1694         struct bridge_iflist *bif;
1695
1696         /* Check if the interface is a bridge member */
1697         if (sc != NULL) {
1698                 BRIDGE_LOCK(sc);
1699
1700                 bif = bridge_lookup_member_if(sc, ifp);
1701                 if (bif != NULL)
1702                         bridge_delete_member(sc, bif, 1);
1703
1704                 BRIDGE_UNLOCK(sc);
1705                 return;
1706         }
1707
1708         /* Check if the interface is a span port */
1709         mtx_lock(&bridge_list_mtx);
1710         LIST_FOREACH(sc, &bridge_list, sc_list) {
1711                 BRIDGE_LOCK(sc);
1712                 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1713                         if (ifp == bif->bif_ifp) {
1714                                 bridge_delete_span(sc, bif);
1715                                 break;
1716                         }
1717
1718                 BRIDGE_UNLOCK(sc);
1719         }
1720         mtx_unlock(&bridge_list_mtx);
1721 }
1722
1723 /*
1724  * bridge_init:
1725  *
1726  *      Initialize a bridge interface.
1727  */
1728 static void
1729 bridge_init(void *xsc)
1730 {
1731         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1732         struct ifnet *ifp = sc->sc_ifp;
1733
1734         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1735                 return;
1736
1737         BRIDGE_LOCK(sc);
1738         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1739             bridge_timer, sc);
1740
1741         ifp->if_drv_flags |= IFF_DRV_RUNNING;
1742         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
1743
1744         BRIDGE_UNLOCK(sc);
1745 }
1746
1747 /*
1748  * bridge_stop:
1749  *
1750  *      Stop the bridge interface.
1751  */
1752 static void
1753 bridge_stop(struct ifnet *ifp, int disable)
1754 {
1755         struct bridge_softc *sc = ifp->if_softc;
1756
1757         BRIDGE_LOCK_ASSERT(sc);
1758
1759         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1760                 return;
1761
1762         callout_stop(&sc->sc_brcallout);
1763         bstp_stop(&sc->sc_stp);
1764
1765         bridge_rtflush(sc, IFBF_FLUSHDYN);
1766
1767         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1768 }
1769
1770 /*
1771  * bridge_enqueue:
1772  *
1773  *      Enqueue a packet on a bridge member interface.
1774  *
1775  */
1776 static void
1777 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
1778 {
1779         int len, err = 0;
1780         short mflags;
1781         struct mbuf *m0;
1782
1783         len = m->m_pkthdr.len;
1784         mflags = m->m_flags;
1785
1786         /* We may be sending a fragment so traverse the mbuf */
1787         for (; m; m = m0) {
1788                 m0 = m->m_nextpkt;
1789                 m->m_nextpkt = NULL;
1790
1791                 /*
1792                  * If underlying interface can not do VLAN tag insertion itself
1793                  * then attach a packet tag that holds it.
1794                  */
1795                 if ((m->m_flags & M_VLANTAG) &&
1796                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
1797                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
1798                         if (m == NULL) {
1799                                 if_printf(dst_ifp,
1800                                     "unable to prepend VLAN header\n");
1801                                 dst_ifp->if_oerrors++;
1802                                 continue;
1803                         }
1804                         m->m_flags &= ~M_VLANTAG;
1805                 }
1806
1807                 if (err == 0)
1808                         dst_ifp->if_transmit(dst_ifp, m);
1809         }
1810
1811         if (err == 0) {
1812                 sc->sc_ifp->if_opackets++;
1813                 sc->sc_ifp->if_obytes += len;
1814                 if (mflags & M_MCAST)
1815                         sc->sc_ifp->if_omcasts++;
1816         }
1817 }
1818
1819 /*
1820  * bridge_dummynet:
1821  *
1822  *      Receive a queued packet from dummynet and pass it on to the output
1823  *      interface.
1824  *
1825  *      The mbuf has the Ethernet header already attached.
1826  */
1827 static void
1828 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
1829 {
1830         struct bridge_softc *sc;
1831
1832         sc = ifp->if_bridge;
1833
1834         /*
1835          * The packet didnt originate from a member interface. This should only
1836          * ever happen if a member interface is removed while packets are
1837          * queued for it.
1838          */
1839         if (sc == NULL) {
1840                 m_freem(m);
1841                 return;
1842         }
1843
1844         if (PFIL_HOOKED(&V_inet_pfil_hook)
1845 #ifdef INET6
1846             || PFIL_HOOKED(&V_inet6_pfil_hook)
1847 #endif
1848             ) {
1849                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
1850                         return;
1851                 if (m == NULL)
1852                         return;
1853         }
1854
1855         bridge_enqueue(sc, ifp, m);
1856 }
1857
1858 /*
1859  * bridge_output:
1860  *
1861  *      Send output from a bridge member interface.  This
1862  *      performs the bridging function for locally originated
1863  *      packets.
1864  *
1865  *      The mbuf has the Ethernet header already attached.  We must
1866  *      enqueue or free the mbuf before returning.
1867  */
1868 static int
1869 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
1870     struct rtentry *rt)
1871 {
1872         struct ether_header *eh;
1873         struct ifnet *dst_if;
1874         struct bridge_softc *sc;
1875         uint16_t vlan;
1876
1877         if (m->m_len < ETHER_HDR_LEN) {
1878                 m = m_pullup(m, ETHER_HDR_LEN);
1879                 if (m == NULL)
1880                         return (0);
1881         }
1882
1883         eh = mtod(m, struct ether_header *);
1884         sc = ifp->if_bridge;
1885         vlan = VLANTAGOF(m);
1886
1887         BRIDGE_LOCK(sc);
1888
1889         /*
1890          * If bridge is down, but the original output interface is up,
1891          * go ahead and send out that interface.  Otherwise, the packet
1892          * is dropped below.
1893          */
1894         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1895                 dst_if = ifp;
1896                 goto sendunicast;
1897         }
1898
1899         /*
1900          * If the packet is a multicast, or we don't know a better way to
1901          * get there, send to all interfaces.
1902          */
1903         if (ETHER_IS_MULTICAST(eh->ether_dhost))
1904                 dst_if = NULL;
1905         else
1906                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
1907         if (dst_if == NULL) {
1908                 struct bridge_iflist *bif;
1909                 struct mbuf *mc;
1910                 int error = 0, used = 0;
1911
1912                 bridge_span(sc, m);
1913
1914                 BRIDGE_LOCK2REF(sc, error);
1915                 if (error) {
1916                         m_freem(m);
1917                         return (0);
1918                 }
1919
1920                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1921                         dst_if = bif->bif_ifp;
1922
1923                         if (dst_if->if_type == IFT_GIF)
1924                                 continue;
1925                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
1926                                 continue;
1927
1928                         /*
1929                          * If this is not the original output interface,
1930                          * and the interface is participating in spanning
1931                          * tree, make sure the port is in a state that
1932                          * allows forwarding.
1933                          */
1934                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
1935                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
1936                                 continue;
1937
1938                         if (LIST_NEXT(bif, bif_next) == NULL) {
1939                                 used = 1;
1940                                 mc = m;
1941                         } else {
1942                                 mc = m_copypacket(m, M_DONTWAIT);
1943                                 if (mc == NULL) {
1944                                         sc->sc_ifp->if_oerrors++;
1945                                         continue;
1946                                 }
1947                         }
1948
1949                         bridge_enqueue(sc, dst_if, mc);
1950                 }
1951                 if (used == 0)
1952                         m_freem(m);
1953                 BRIDGE_UNREF(sc);
1954                 return (0);
1955         }
1956
1957 sendunicast:
1958         /*
1959          * XXX Spanning tree consideration here?
1960          */
1961
1962         bridge_span(sc, m);
1963         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1964                 m_freem(m);
1965                 BRIDGE_UNLOCK(sc);
1966                 return (0);
1967         }
1968
1969         BRIDGE_UNLOCK(sc);
1970         bridge_enqueue(sc, dst_if, m);
1971         return (0);
1972 }
1973
1974 /*
1975  * bridge_start:
1976  *
1977  *      Start output on a bridge.
1978  *
1979  */
1980 static void
1981 bridge_start(struct ifnet *ifp)
1982 {
1983         struct bridge_softc *sc;
1984         struct mbuf *m;
1985         struct ether_header *eh;
1986         struct ifnet *dst_if;
1987
1988         sc = ifp->if_softc;
1989
1990         ifp->if_drv_flags |= IFF_DRV_OACTIVE;
1991         for (;;) {
1992                 IFQ_DEQUEUE(&ifp->if_snd, m);
1993                 if (m == 0)
1994                         break;
1995                 ETHER_BPF_MTAP(ifp, m);
1996
1997                 eh = mtod(m, struct ether_header *);
1998                 dst_if = NULL;
1999
2000                 BRIDGE_LOCK(sc);
2001                 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2002                         dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1);
2003                 }
2004
2005                 if (dst_if == NULL)
2006                         bridge_broadcast(sc, ifp, m, 0);
2007                 else {
2008                         BRIDGE_UNLOCK(sc);
2009                         bridge_enqueue(sc, dst_if, m);
2010                 }
2011         }
2012         ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
2013 }
2014
2015 /*
2016  * bridge_forward:
2017  *
2018  *      The forwarding function of the bridge.
2019  *
2020  *      NOTE: Releases the lock on return.
2021  */
2022 static void
2023 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2024     struct mbuf *m)
2025 {
2026         struct bridge_iflist *dbif;
2027         struct ifnet *src_if, *dst_if, *ifp;
2028         struct ether_header *eh;
2029         uint16_t vlan;
2030         uint8_t *dst;
2031         int error;
2032
2033         src_if = m->m_pkthdr.rcvif;
2034         ifp = sc->sc_ifp;
2035
2036         ifp->if_ipackets++;
2037         ifp->if_ibytes += m->m_pkthdr.len;
2038         vlan = VLANTAGOF(m);
2039
2040         if ((sbif->bif_flags & IFBIF_STP) &&
2041             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2042                 goto drop;
2043
2044         eh = mtod(m, struct ether_header *);
2045         dst = eh->ether_dhost;
2046
2047         /* If the interface is learning, record the address. */
2048         if (sbif->bif_flags & IFBIF_LEARNING) {
2049                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2050                     sbif, 0, IFBAF_DYNAMIC);
2051                 /*
2052                  * If the interface has addresses limits then deny any source
2053                  * that is not in the cache.
2054                  */
2055                 if (error && sbif->bif_addrmax)
2056                         goto drop;
2057         }
2058
2059         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2060             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2061                 goto drop;
2062
2063         /*
2064          * At this point, the port either doesn't participate
2065          * in spanning tree or it is in the forwarding state.
2066          */
2067
2068         /*
2069          * If the packet is unicast, destined for someone on
2070          * "this" side of the bridge, drop it.
2071          */
2072         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2073                 dst_if = bridge_rtlookup(sc, dst, vlan);
2074                 if (src_if == dst_if)
2075                         goto drop;
2076         } else {
2077                 /*
2078                  * Check if its a reserved multicast address, any address
2079                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2080                  * bridge.
2081                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2082                  */
2083                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2084                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2085                     dst[4] == 0x00 && dst[5] <= 0x0f)
2086                         goto drop;
2087
2088                 /* ...forward it to all interfaces. */
2089                 ifp->if_imcasts++;
2090                 dst_if = NULL;
2091         }
2092
2093         /*
2094          * If we have a destination interface which is a member of our bridge,
2095          * OR this is a unicast packet, push it through the bpf(4) machinery.
2096          * For broadcast or multicast packets, don't bother because it will
2097          * be reinjected into ether_input. We do this before we pass the packets
2098          * through the pfil(9) framework, as it is possible that pfil(9) will
2099          * drop the packet, or possibly modify it, making it difficult to debug
2100          * firewall issues on the bridge.
2101          */
2102         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2103                 ETHER_BPF_MTAP(ifp, m);
2104
2105         /* run the packet filter */
2106         if (PFIL_HOOKED(&V_inet_pfil_hook)
2107 #ifdef INET6
2108             || PFIL_HOOKED(&V_inet6_pfil_hook)
2109 #endif
2110             ) {
2111                 BRIDGE_UNLOCK(sc);
2112                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2113                         return;
2114                 if (m == NULL)
2115                         return;
2116                 BRIDGE_LOCK(sc);
2117         }
2118
2119         if (dst_if == NULL) {
2120                 bridge_broadcast(sc, src_if, m, 1);
2121                 return;
2122         }
2123
2124         /*
2125          * At this point, we're dealing with a unicast frame
2126          * going to a different interface.
2127          */
2128         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2129                 goto drop;
2130
2131         dbif = bridge_lookup_member_if(sc, dst_if);
2132         if (dbif == NULL)
2133                 /* Not a member of the bridge (anymore?) */
2134                 goto drop;
2135
2136         /* Private segments can not talk to each other */
2137         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2138                 goto drop;
2139
2140         if ((dbif->bif_flags & IFBIF_STP) &&
2141             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2142                 goto drop;
2143
2144         BRIDGE_UNLOCK(sc);
2145
2146         if (PFIL_HOOKED(&V_inet_pfil_hook)
2147 #ifdef INET6
2148             || PFIL_HOOKED(&V_inet6_pfil_hook)
2149 #endif
2150             ) {
2151                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2152                         return;
2153                 if (m == NULL)
2154                         return;
2155         }
2156
2157         bridge_enqueue(sc, dst_if, m);
2158         return;
2159
2160 drop:
2161         BRIDGE_UNLOCK(sc);
2162         m_freem(m);
2163 }
2164
2165 /*
2166  * bridge_input:
2167  *
2168  *      Receive input from a member interface.  Queue the packet for
2169  *      bridging if it is not for us.
2170  */
2171 static struct mbuf *
2172 bridge_input(struct ifnet *ifp, struct mbuf *m)
2173 {
2174         struct bridge_softc *sc = ifp->if_bridge;
2175         struct bridge_iflist *bif, *bif2;
2176         struct ifnet *bifp;
2177         struct ether_header *eh;
2178         struct mbuf *mc, *mc2;
2179         uint16_t vlan;
2180         int error;
2181
2182         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2183                 return (m);
2184
2185         bifp = sc->sc_ifp;
2186         vlan = VLANTAGOF(m);
2187
2188         /*
2189          * Implement support for bridge monitoring. If this flag has been
2190          * set on this interface, discard the packet once we push it through
2191          * the bpf(4) machinery, but before we do, increment the byte and
2192          * packet counters associated with this interface.
2193          */
2194         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2195                 m->m_pkthdr.rcvif  = bifp;
2196                 ETHER_BPF_MTAP(bifp, m);
2197                 bifp->if_ipackets++;
2198                 bifp->if_ibytes += m->m_pkthdr.len;
2199                 m_freem(m);
2200                 return (NULL);
2201         }
2202         BRIDGE_LOCK(sc);
2203         bif = bridge_lookup_member_if(sc, ifp);
2204         if (bif == NULL) {
2205                 BRIDGE_UNLOCK(sc);
2206                 return (m);
2207         }
2208
2209         eh = mtod(m, struct ether_header *);
2210
2211         bridge_span(sc, m);
2212
2213         if (m->m_flags & (M_BCAST|M_MCAST)) {
2214                 /* Tap off 802.1D packets; they do not get forwarded. */
2215                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2216                     ETHER_ADDR_LEN) == 0) {
2217                         bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2218                         BRIDGE_UNLOCK(sc);
2219                         return (NULL);
2220                 }
2221
2222                 if ((bif->bif_flags & IFBIF_STP) &&
2223                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2224                         BRIDGE_UNLOCK(sc);
2225                         return (m);
2226                 }
2227
2228                 /*
2229                  * Make a deep copy of the packet and enqueue the copy
2230                  * for bridge processing; return the original packet for
2231                  * local processing.
2232                  */
2233                 mc = m_dup(m, M_DONTWAIT);
2234                 if (mc == NULL) {
2235                         BRIDGE_UNLOCK(sc);
2236                         return (m);
2237                 }
2238
2239                 /* Perform the bridge forwarding function with the copy. */
2240                 bridge_forward(sc, bif, mc);
2241
2242                 /*
2243                  * Reinject the mbuf as arriving on the bridge so we have a
2244                  * chance at claiming multicast packets. We can not loop back
2245                  * here from ether_input as a bridge is never a member of a
2246                  * bridge.
2247                  */
2248                 KASSERT(bifp->if_bridge == NULL,
2249                     ("loop created in bridge_input"));
2250                 mc2 = m_dup(m, M_DONTWAIT);
2251                 if (mc2 != NULL) {
2252                         /* Keep the layer3 header aligned */
2253                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2254                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2255                 }
2256                 if (mc2 != NULL) {
2257                         mc2->m_pkthdr.rcvif = bifp;
2258                         (*bifp->if_input)(bifp, mc2);
2259                 }
2260
2261                 /* Return the original packet for local processing. */
2262                 return (m);
2263         }
2264
2265         if ((bif->bif_flags & IFBIF_STP) &&
2266             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2267                 BRIDGE_UNLOCK(sc);
2268                 return (m);
2269         }
2270
2271 #if (defined(INET) || defined(INET6))
2272 #   define OR_CARP_CHECK_WE_ARE_DST(iface) \
2273         || ((iface)->if_carp \
2274             && (*carp_forus_p)((iface), eh->ether_dhost))
2275 #   define OR_CARP_CHECK_WE_ARE_SRC(iface) \
2276         || ((iface)->if_carp \
2277             && (*carp_forus_p)((iface), eh->ether_shost))
2278 #else
2279 #   define OR_CARP_CHECK_WE_ARE_DST(iface)
2280 #   define OR_CARP_CHECK_WE_ARE_SRC(iface)
2281 #endif
2282
2283 #ifdef INET6
2284 #   define OR_PFIL_HOOKED_INET6 \
2285         || PFIL_HOOKED(&V_inet6_pfil_hook)
2286 #else
2287 #   define OR_PFIL_HOOKED_INET6
2288 #endif
2289
2290 #define GRAB_OUR_PACKETS(iface) \
2291         if ((iface)->if_type == IFT_GIF) \
2292                 continue; \
2293         /* It is destined for us. */ \
2294         if (memcmp(IF_LLADDR((iface)), eh->ether_dhost,  ETHER_ADDR_LEN) == 0 \
2295             OR_CARP_CHECK_WE_ARE_DST((iface))                           \
2296             ) {                                                         \
2297                 if ((iface)->if_type == IFT_BRIDGE) {                   \
2298                         ETHER_BPF_MTAP(iface, m);                       \
2299                         iface->if_ipackets++;                           \
2300                         /* Filter on the physical interface. */         \
2301                         if (pfil_local_phys &&                          \
2302                             (PFIL_HOOKED(&V_inet_pfil_hook)             \
2303                              OR_PFIL_HOOKED_INET6)) {                   \
2304                                 if (bridge_pfil(&m, NULL, ifp,          \
2305                                     PFIL_IN) != 0 || m == NULL) {       \
2306                                         BRIDGE_UNLOCK(sc);              \
2307                                         return (NULL);                  \
2308                                 }                                       \
2309                         }                                               \
2310                 }                                                       \
2311                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2312                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2313                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2314                         if (error && bif->bif_addrmax) {                \
2315                                 BRIDGE_UNLOCK(sc);                      \
2316                                 m_freem(m);                             \
2317                                 return (NULL);                          \
2318                         }                                               \
2319                 }                                                       \
2320                 m->m_pkthdr.rcvif = iface;                              \
2321                 BRIDGE_UNLOCK(sc);                                      \
2322                 return (m);                                             \
2323         }                                                               \
2324                                                                         \
2325         /* We just received a packet that we sent out. */               \
2326         if (memcmp(IF_LLADDR((iface)), eh->ether_shost, ETHER_ADDR_LEN) == 0 \
2327             OR_CARP_CHECK_WE_ARE_SRC((iface))                   \
2328             ) {                                                         \
2329                 BRIDGE_UNLOCK(sc);                                      \
2330                 m_freem(m);                                             \
2331                 return (NULL);                                          \
2332         }
2333
2334         /*
2335          * Unicast.  Make sure it's not for the bridge.
2336          */
2337         do { GRAB_OUR_PACKETS(bifp) } while (0);
2338
2339         /*
2340          * Give a chance for ifp at first priority. This will help when the
2341          * packet comes through the interface like VLAN's with the same MACs
2342          * on several interfaces from the same bridge. This also will save
2343          * some CPU cycles in case the destination interface and the input
2344          * interface (eq ifp) are the same.
2345          */
2346         do { GRAB_OUR_PACKETS(ifp) } while (0);
2347
2348         /* Now check the all bridge members. */
2349         LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2350                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2351         }
2352
2353 #undef OR_CARP_CHECK_WE_ARE_DST
2354 #undef OR_CARP_CHECK_WE_ARE_SRC
2355 #undef OR_PFIL_HOOKED_INET6
2356 #undef GRAB_OUR_PACKETS
2357
2358         /* Perform the bridge forwarding function. */
2359         bridge_forward(sc, bif, m);
2360
2361         return (NULL);
2362 }
2363
2364 /*
2365  * bridge_broadcast:
2366  *
2367  *      Send a frame to all interfaces that are members of
2368  *      the bridge, except for the one on which the packet
2369  *      arrived.
2370  *
2371  *      NOTE: Releases the lock on return.
2372  */
2373 static void
2374 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2375     struct mbuf *m, int runfilt)
2376 {
2377         struct bridge_iflist *dbif, *sbif;
2378         struct mbuf *mc;
2379         struct ifnet *dst_if;
2380         int error = 0, used = 0, i;
2381
2382         sbif = bridge_lookup_member_if(sc, src_if);
2383
2384         BRIDGE_LOCK2REF(sc, error);
2385         if (error) {
2386                 m_freem(m);
2387                 return;
2388         }
2389
2390         /* Filter on the bridge interface before broadcasting */
2391         if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2392 #ifdef INET6
2393             || PFIL_HOOKED(&V_inet6_pfil_hook)
2394 #endif
2395             )) {
2396                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2397                         goto out;
2398                 if (m == NULL)
2399                         goto out;
2400         }
2401
2402         LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2403                 dst_if = dbif->bif_ifp;
2404                 if (dst_if == src_if)
2405                         continue;
2406
2407                 /* Private segments can not talk to each other */
2408                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2409                         continue;
2410
2411                 if ((dbif->bif_flags & IFBIF_STP) &&
2412                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2413                         continue;
2414
2415                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2416                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2417                         continue;
2418
2419                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2420                         continue;
2421
2422                 if (LIST_NEXT(dbif, bif_next) == NULL) {
2423                         mc = m;
2424                         used = 1;
2425                 } else {
2426                         mc = m_dup(m, M_DONTWAIT);
2427                         if (mc == NULL) {
2428                                 sc->sc_ifp->if_oerrors++;
2429                                 continue;
2430                         }
2431                 }
2432
2433                 /*
2434                  * Filter on the output interface. Pass a NULL bridge interface
2435                  * pointer so we do not redundantly filter on the bridge for
2436                  * each interface we broadcast on.
2437                  */
2438                 if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2439 #ifdef INET6
2440                     || PFIL_HOOKED(&V_inet6_pfil_hook)
2441 #endif
2442                     )) {
2443                         if (used == 0) {
2444                                 /* Keep the layer3 header aligned */
2445                                 i = min(mc->m_pkthdr.len, max_protohdr);
2446                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2447                                 if (mc == NULL) {
2448                                         sc->sc_ifp->if_oerrors++;
2449                                         continue;
2450                                 }
2451                         }
2452                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2453                                 continue;
2454                         if (mc == NULL)
2455                                 continue;
2456                 }
2457
2458                 bridge_enqueue(sc, dst_if, mc);
2459         }
2460         if (used == 0)
2461                 m_freem(m);
2462
2463 out:
2464         BRIDGE_UNREF(sc);
2465 }
2466
2467 /*
2468  * bridge_span:
2469  *
2470  *      Duplicate a packet out one or more interfaces that are in span mode,
2471  *      the original mbuf is unmodified.
2472  */
2473 static void
2474 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2475 {
2476         struct bridge_iflist *bif;
2477         struct ifnet *dst_if;
2478         struct mbuf *mc;
2479
2480         if (LIST_EMPTY(&sc->sc_spanlist))
2481                 return;
2482
2483         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2484                 dst_if = bif->bif_ifp;
2485
2486                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2487                         continue;
2488
2489                 mc = m_copypacket(m, M_DONTWAIT);
2490                 if (mc == NULL) {
2491                         sc->sc_ifp->if_oerrors++;
2492                         continue;
2493                 }
2494
2495                 bridge_enqueue(sc, dst_if, mc);
2496         }
2497 }
2498
2499 /*
2500  * bridge_rtupdate:
2501  *
2502  *      Add a bridge routing entry.
2503  */
2504 static int
2505 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2506     struct bridge_iflist *bif, int setflags, uint8_t flags)
2507 {
2508         struct bridge_rtnode *brt;
2509         int error;
2510
2511         BRIDGE_LOCK_ASSERT(sc);
2512
2513         /* Check the source address is valid and not multicast. */
2514         if (ETHER_IS_MULTICAST(dst) ||
2515             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2516              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2517                 return (EINVAL);
2518
2519         /* 802.1p frames map to vlan 1 */
2520         if (vlan == 0)
2521                 vlan = 1;
2522
2523         /*
2524          * A route for this destination might already exist.  If so,
2525          * update it, otherwise create a new one.
2526          */
2527         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2528                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2529                         sc->sc_brtexceeded++;
2530                         return (ENOSPC);
2531                 }
2532                 /* Check per interface address limits (if enabled) */
2533                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2534                         bif->bif_addrexceeded++;
2535                         return (ENOSPC);
2536                 }
2537
2538                 /*
2539                  * Allocate a new bridge forwarding node, and
2540                  * initialize the expiration time and Ethernet
2541                  * address.
2542                  */
2543                 brt = uma_zalloc(bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2544                 if (brt == NULL)
2545                         return (ENOMEM);
2546
2547                 if (bif->bif_flags & IFBIF_STICKY)
2548                         brt->brt_flags = IFBAF_STICKY;
2549                 else
2550                         brt->brt_flags = IFBAF_DYNAMIC;
2551
2552                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2553                 brt->brt_vlan = vlan;
2554
2555                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2556                         uma_zfree(bridge_rtnode_zone, brt);
2557                         return (error);
2558                 }
2559                 brt->brt_dst = bif;
2560                 bif->bif_addrcnt++;
2561         }
2562
2563         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2564             brt->brt_dst != bif) {
2565                 brt->brt_dst->bif_addrcnt--;
2566                 brt->brt_dst = bif;
2567                 brt->brt_dst->bif_addrcnt++;
2568         }
2569
2570         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2571                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2572         if (setflags)
2573                 brt->brt_flags = flags;
2574
2575         return (0);
2576 }
2577
2578 /*
2579  * bridge_rtlookup:
2580  *
2581  *      Lookup the destination interface for an address.
2582  */
2583 static struct ifnet *
2584 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2585 {
2586         struct bridge_rtnode *brt;
2587
2588         BRIDGE_LOCK_ASSERT(sc);
2589
2590         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2591                 return (NULL);
2592
2593         return (brt->brt_ifp);
2594 }
2595
2596 /*
2597  * bridge_rttrim:
2598  *
2599  *      Trim the routine table so that we have a number
2600  *      of routing entries less than or equal to the
2601  *      maximum number.
2602  */
2603 static void
2604 bridge_rttrim(struct bridge_softc *sc)
2605 {
2606         struct bridge_rtnode *brt, *nbrt;
2607
2608         BRIDGE_LOCK_ASSERT(sc);
2609
2610         /* Make sure we actually need to do this. */
2611         if (sc->sc_brtcnt <= sc->sc_brtmax)
2612                 return;
2613
2614         /* Force an aging cycle; this might trim enough addresses. */
2615         bridge_rtage(sc);
2616         if (sc->sc_brtcnt <= sc->sc_brtmax)
2617                 return;
2618
2619         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2620                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2621                         bridge_rtnode_destroy(sc, brt);
2622                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2623                                 return;
2624                 }
2625         }
2626 }
2627
2628 /*
2629  * bridge_timer:
2630  *
2631  *      Aging timer for the bridge.
2632  */
2633 static void
2634 bridge_timer(void *arg)
2635 {
2636         struct bridge_softc *sc = arg;
2637
2638         BRIDGE_LOCK_ASSERT(sc);
2639
2640         bridge_rtage(sc);
2641
2642         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2643                 callout_reset(&sc->sc_brcallout,
2644                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2645 }
2646
2647 /*
2648  * bridge_rtage:
2649  *
2650  *      Perform an aging cycle.
2651  */
2652 static void
2653 bridge_rtage(struct bridge_softc *sc)
2654 {
2655         struct bridge_rtnode *brt, *nbrt;
2656
2657         BRIDGE_LOCK_ASSERT(sc);
2658
2659         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2660                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2661                         if (time_uptime >= brt->brt_expire)
2662                                 bridge_rtnode_destroy(sc, brt);
2663                 }
2664         }
2665 }
2666
2667 /*
2668  * bridge_rtflush:
2669  *
2670  *      Remove all dynamic addresses from the bridge.
2671  */
2672 static void
2673 bridge_rtflush(struct bridge_softc *sc, int full)
2674 {
2675         struct bridge_rtnode *brt, *nbrt;
2676
2677         BRIDGE_LOCK_ASSERT(sc);
2678
2679         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2680                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2681                         bridge_rtnode_destroy(sc, brt);
2682         }
2683 }
2684
2685 /*
2686  * bridge_rtdaddr:
2687  *
2688  *      Remove an address from the table.
2689  */
2690 static int
2691 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2692 {
2693         struct bridge_rtnode *brt;
2694         int found = 0;
2695
2696         BRIDGE_LOCK_ASSERT(sc);
2697
2698         /*
2699          * If vlan is zero then we want to delete for all vlans so the lookup
2700          * may return more than one.
2701          */
2702         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
2703                 bridge_rtnode_destroy(sc, brt);
2704                 found = 1;
2705         }
2706
2707         return (found ? 0 : ENOENT);
2708 }
2709
2710 /*
2711  * bridge_rtdelete:
2712  *
2713  *      Delete routes to a speicifc member interface.
2714  */
2715 static void
2716 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2717 {
2718         struct bridge_rtnode *brt, *nbrt;
2719
2720         BRIDGE_LOCK_ASSERT(sc);
2721
2722         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2723                 if (brt->brt_ifp == ifp && (full ||
2724                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
2725                         bridge_rtnode_destroy(sc, brt);
2726         }
2727 }
2728
2729 /*
2730  * bridge_rtable_init:
2731  *
2732  *      Initialize the route table for this bridge.
2733  */
2734 static int
2735 bridge_rtable_init(struct bridge_softc *sc)
2736 {
2737         int i;
2738
2739         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2740             M_DEVBUF, M_NOWAIT);
2741         if (sc->sc_rthash == NULL)
2742                 return (ENOMEM);
2743
2744         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2745                 LIST_INIT(&sc->sc_rthash[i]);
2746
2747         sc->sc_rthash_key = arc4random();
2748
2749         LIST_INIT(&sc->sc_rtlist);
2750
2751         return (0);
2752 }
2753
2754 /*
2755  * bridge_rtable_fini:
2756  *
2757  *      Deconstruct the route table for this bridge.
2758  */
2759 static void
2760 bridge_rtable_fini(struct bridge_softc *sc)
2761 {
2762
2763         KASSERT(sc->sc_brtcnt == 0,
2764             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
2765         free(sc->sc_rthash, M_DEVBUF);
2766 }
2767
2768 /*
2769  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2770  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2771  */
2772 #define mix(a, b, c)                                                    \
2773 do {                                                                    \
2774         a -= b; a -= c; a ^= (c >> 13);                                 \
2775         b -= c; b -= a; b ^= (a << 8);                                  \
2776         c -= a; c -= b; c ^= (b >> 13);                                 \
2777         a -= b; a -= c; a ^= (c >> 12);                                 \
2778         b -= c; b -= a; b ^= (a << 16);                                 \
2779         c -= a; c -= b; c ^= (b >> 5);                                  \
2780         a -= b; a -= c; a ^= (c >> 3);                                  \
2781         b -= c; b -= a; b ^= (a << 10);                                 \
2782         c -= a; c -= b; c ^= (b >> 15);                                 \
2783 } while (/*CONSTCOND*/0)
2784
2785 static __inline uint32_t
2786 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2787 {
2788         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2789
2790         b += addr[5] << 8;
2791         b += addr[4];
2792         a += addr[3] << 24;
2793         a += addr[2] << 16;
2794         a += addr[1] << 8;
2795         a += addr[0];
2796
2797         mix(a, b, c);
2798
2799         return (c & BRIDGE_RTHASH_MASK);
2800 }
2801
2802 #undef mix
2803
2804 static int
2805 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
2806 {
2807         int i, d;
2808
2809         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
2810                 d = ((int)a[i]) - ((int)b[i]);
2811         }
2812
2813         return (d);
2814 }
2815
2816 /*
2817  * bridge_rtnode_lookup:
2818  *
2819  *      Look up a bridge route node for the specified destination. Compare the
2820  *      vlan id or if zero then just return the first match.
2821  */
2822 static struct bridge_rtnode *
2823 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2824 {
2825         struct bridge_rtnode *brt;
2826         uint32_t hash;
2827         int dir;
2828
2829         BRIDGE_LOCK_ASSERT(sc);
2830
2831         hash = bridge_rthash(sc, addr);
2832         LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
2833                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
2834                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
2835                         return (brt);
2836                 if (dir > 0)
2837                         return (NULL);
2838         }
2839
2840         return (NULL);
2841 }
2842
2843 /*
2844  * bridge_rtnode_insert:
2845  *
2846  *      Insert the specified bridge node into the route table.  We
2847  *      assume the entry is not already in the table.
2848  */
2849 static int
2850 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2851 {
2852         struct bridge_rtnode *lbrt;
2853         uint32_t hash;
2854         int dir;
2855
2856         BRIDGE_LOCK_ASSERT(sc);
2857
2858         hash = bridge_rthash(sc, brt->brt_addr);
2859
2860         lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
2861         if (lbrt == NULL) {
2862                 LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
2863                 goto out;
2864         }
2865
2866         do {
2867                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
2868                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
2869                         return (EEXIST);
2870                 if (dir > 0) {
2871                         LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
2872                         goto out;
2873                 }
2874                 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
2875                         LIST_INSERT_AFTER(lbrt, brt, brt_hash);
2876                         goto out;
2877                 }
2878                 lbrt = LIST_NEXT(lbrt, brt_hash);
2879         } while (lbrt != NULL);
2880
2881 #ifdef DIAGNOSTIC
2882         panic("bridge_rtnode_insert: impossible");
2883 #endif
2884
2885 out:
2886         LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
2887         sc->sc_brtcnt++;
2888
2889         return (0);
2890 }
2891
2892 /*
2893  * bridge_rtnode_destroy:
2894  *
2895  *      Destroy a bridge rtnode.
2896  */
2897 static void
2898 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
2899 {
2900         BRIDGE_LOCK_ASSERT(sc);
2901
2902         LIST_REMOVE(brt, brt_hash);
2903
2904         LIST_REMOVE(brt, brt_list);
2905         sc->sc_brtcnt--;
2906         brt->brt_dst->bif_addrcnt--;
2907         uma_zfree(bridge_rtnode_zone, brt);
2908 }
2909
2910 /*
2911  * bridge_rtable_expire:
2912  *
2913  *      Set the expiry time for all routes on an interface.
2914  */
2915 static void
2916 bridge_rtable_expire(struct ifnet *ifp, int age)
2917 {
2918         struct bridge_softc *sc = ifp->if_bridge;
2919         struct bridge_rtnode *brt;
2920
2921         BRIDGE_LOCK(sc);
2922
2923         /*
2924          * If the age is zero then flush, otherwise set all the expiry times to
2925          * age for the interface
2926          */
2927         if (age == 0)
2928                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
2929         else {
2930                 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
2931                         /* Cap the expiry time to 'age' */
2932                         if (brt->brt_ifp == ifp &&
2933                             brt->brt_expire > time_uptime + age &&
2934                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2935                                 brt->brt_expire = time_uptime + age;
2936                 }
2937         }
2938         BRIDGE_UNLOCK(sc);
2939 }
2940
2941 /*
2942  * bridge_state_change:
2943  *
2944  *      Callback from the bridgestp code when a port changes states.
2945  */
2946 static void
2947 bridge_state_change(struct ifnet *ifp, int state)
2948 {
2949         struct bridge_softc *sc = ifp->if_bridge;
2950         static const char *stpstates[] = {
2951                 "disabled",
2952                 "listening",
2953                 "learning",
2954                 "forwarding",
2955                 "blocking",
2956                 "discarding"
2957         };
2958
2959         if (log_stp)
2960                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
2961                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
2962 }
2963
2964 /*
2965  * Send bridge packets through pfil if they are one of the types pfil can deal
2966  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
2967  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
2968  * that interface.
2969  */
2970 static int
2971 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
2972 {
2973         int snap, error, i, hlen;
2974         struct ether_header *eh1, eh2;
2975         struct ip_fw_args args;
2976         struct ip *ip;
2977         struct llc llc1;
2978         u_int16_t ether_type;
2979
2980         snap = 0;
2981         error = -1;     /* Default error if not error == 0 */
2982
2983 #if 0
2984         /* we may return with the IP fields swapped, ensure its not shared */
2985         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
2986 #endif
2987
2988         if (pfil_bridge == 0 && pfil_member == 0 && pfil_ipfw == 0)
2989                 return (0); /* filtering is disabled */
2990
2991         i = min((*mp)->m_pkthdr.len, max_protohdr);
2992         if ((*mp)->m_len < i) {
2993             *mp = m_pullup(*mp, i);
2994             if (*mp == NULL) {
2995                 printf("%s: m_pullup failed\n", __func__);
2996                 return (-1);
2997             }
2998         }
2999
3000         eh1 = mtod(*mp, struct ether_header *);
3001         ether_type = ntohs(eh1->ether_type);
3002
3003         /*
3004          * Check for SNAP/LLC.
3005          */
3006         if (ether_type < ETHERMTU) {
3007                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3008
3009                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3010                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3011                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3012                     llc2->llc_control == LLC_UI) {
3013                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3014                         snap = 1;
3015                 }
3016         }
3017
3018         /*
3019          * If we're trying to filter bridge traffic, don't look at anything
3020          * other than IP and ARP traffic.  If the filter doesn't understand
3021          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3022          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3023          * but of course we don't have an AppleTalk filter to begin with.
3024          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3025          * ARP traffic.)
3026          */
3027         switch (ether_type) {
3028                 case ETHERTYPE_ARP:
3029                 case ETHERTYPE_REVARP:
3030                         if (pfil_ipfw_arp == 0)
3031                                 return (0); /* Automatically pass */
3032                         break;
3033
3034                 case ETHERTYPE_IP:
3035 #ifdef INET6
3036                 case ETHERTYPE_IPV6:
3037 #endif /* INET6 */
3038                         break;
3039                 default:
3040                         /*
3041                          * Check to see if the user wants to pass non-ip
3042                          * packets, these will not be checked by pfil(9) and
3043                          * passed unconditionally so the default is to drop.
3044                          */
3045                         if (pfil_onlyip)
3046                                 goto bad;
3047         }
3048
3049         /* Strip off the Ethernet header and keep a copy. */
3050         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3051         m_adj(*mp, ETHER_HDR_LEN);
3052
3053         /* Strip off snap header, if present */
3054         if (snap) {
3055                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3056                 m_adj(*mp, sizeof(struct llc));
3057         }
3058
3059         /*
3060          * Check the IP header for alignment and errors
3061          */
3062         if (dir == PFIL_IN) {
3063                 switch (ether_type) {
3064                         case ETHERTYPE_IP:
3065                                 error = bridge_ip_checkbasic(mp);
3066                                 break;
3067 #ifdef INET6
3068                         case ETHERTYPE_IPV6:
3069                                 error = bridge_ip6_checkbasic(mp);
3070                                 break;
3071 #endif /* INET6 */
3072                         default:
3073                                 error = 0;
3074                 }
3075                 if (error)
3076                         goto bad;
3077         }
3078
3079         /* XXX this section is also in if_ethersubr.c */
3080         // XXX PFIL_OUT or DIR_OUT ?
3081         if (V_ip_fw_chk_ptr && pfil_ipfw != 0 &&
3082                         dir == PFIL_OUT && ifp != NULL) {
3083                 struct m_tag *mtag;
3084
3085                 error = -1;
3086                 /* fetch the start point from existing tags, if any */
3087                 mtag = m_tag_locate(*mp, MTAG_IPFW_RULE, 0, NULL);
3088                 if (mtag == NULL) {
3089                         args.rule.slot = 0;
3090                 } else {
3091                         struct ipfw_rule_ref *r;
3092
3093                         /* XXX can we free the tag after use ? */
3094                         mtag->m_tag_id = PACKET_TAG_NONE;
3095                         r = (struct ipfw_rule_ref *)(mtag + 1);
3096                         /* packet already partially processed ? */
3097                         if (r->info & IPFW_ONEPASS)
3098                                 goto ipfwpass;
3099                         args.rule = *r;
3100                 }
3101
3102                 args.m = *mp;
3103                 args.oif = ifp;
3104                 args.next_hop = NULL;
3105                 args.next_hop6 = NULL;
3106                 args.eh = &eh2;
3107                 args.inp = NULL;        /* used by ipfw uid/gid/jail rules */
3108                 i = V_ip_fw_chk_ptr(&args);
3109                 *mp = args.m;
3110
3111                 if (*mp == NULL)
3112                         return (error);
3113
3114                 if (ip_dn_io_ptr && (i == IP_FW_DUMMYNET)) {
3115
3116                         /* put the Ethernet header back on */
3117                         M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
3118                         if (*mp == NULL)
3119                                 return (error);
3120                         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3121
3122                         /*
3123                          * Pass the pkt to dummynet, which consumes it. The
3124                          * packet will return to us via bridge_dummynet().
3125                          */
3126                         args.oif = ifp;
3127                         ip_dn_io_ptr(mp, DIR_FWD | PROTO_IFB, &args);
3128                         return (error);
3129                 }
3130
3131                 if (i != IP_FW_PASS) /* drop */
3132                         goto bad;
3133         }
3134
3135 ipfwpass:
3136         error = 0;
3137
3138         /*
3139          * Run the packet through pfil
3140          */
3141         switch (ether_type) {
3142         case ETHERTYPE_IP:
3143                 /*
3144                  * before calling the firewall, swap fields the same as
3145                  * IP does. here we assume the header is contiguous
3146                  */
3147                 ip = mtod(*mp, struct ip *);
3148
3149                 ip->ip_len = ntohs(ip->ip_len);
3150                 ip->ip_off = ntohs(ip->ip_off);
3151
3152                 /*
3153                  * Run pfil on the member interface and the bridge, both can
3154                  * be skipped by clearing pfil_member or pfil_bridge.
3155                  *
3156                  * Keep the order:
3157                  *   in_if -> bridge_if -> out_if
3158                  */
3159                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3160                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3161                                         dir, NULL);
3162
3163                 if (*mp == NULL || error != 0) /* filter may consume */
3164                         break;
3165
3166                 if (pfil_member && ifp != NULL)
3167                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp,
3168                                         dir, NULL);
3169
3170                 if (*mp == NULL || error != 0) /* filter may consume */
3171                         break;
3172
3173                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3174                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3175                                         dir, NULL);
3176
3177                 if (*mp == NULL || error != 0) /* filter may consume */
3178                         break;
3179
3180                 /* check if we need to fragment the packet */
3181                 if (pfil_member && ifp != NULL && dir == PFIL_OUT) {
3182                         i = (*mp)->m_pkthdr.len;
3183                         if (i > ifp->if_mtu) {
3184                                 error = bridge_fragment(ifp, *mp, &eh2, snap,
3185                                             &llc1);
3186                                 return (error);
3187                         }
3188                 }
3189
3190                 /* Recalculate the ip checksum and restore byte ordering */
3191                 ip = mtod(*mp, struct ip *);
3192                 hlen = ip->ip_hl << 2;
3193                 if (hlen < sizeof(struct ip))
3194                         goto bad;
3195                 if (hlen > (*mp)->m_len) {
3196                         if ((*mp = m_pullup(*mp, hlen)) == 0)
3197                                 goto bad;
3198                         ip = mtod(*mp, struct ip *);
3199                         if (ip == NULL)
3200                                 goto bad;
3201                 }
3202                 ip->ip_len = htons(ip->ip_len);
3203                 ip->ip_off = htons(ip->ip_off);
3204                 ip->ip_sum = 0;
3205                 if (hlen == sizeof(struct ip))
3206                         ip->ip_sum = in_cksum_hdr(ip);
3207                 else
3208                         ip->ip_sum = in_cksum(*mp, hlen);
3209
3210                 break;
3211 #ifdef INET6
3212         case ETHERTYPE_IPV6:
3213                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3214                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3215                                         dir, NULL);
3216
3217                 if (*mp == NULL || error != 0) /* filter may consume */
3218                         break;
3219
3220                 if (pfil_member && ifp != NULL)
3221                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, ifp,
3222                                         dir, NULL);
3223
3224                 if (*mp == NULL || error != 0) /* filter may consume */
3225                         break;
3226
3227                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3228                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3229                                         dir, NULL);
3230                 break;
3231 #endif
3232         default:
3233                 error = 0;
3234                 break;
3235         }
3236
3237         if (*mp == NULL)
3238                 return (error);
3239         if (error != 0)
3240                 goto bad;
3241
3242         error = -1;
3243
3244         /*
3245          * Finally, put everything back the way it was and return
3246          */
3247         if (snap) {
3248                 M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
3249                 if (*mp == NULL)
3250                         return (error);
3251                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3252         }
3253
3254         M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
3255         if (*mp == NULL)
3256                 return (error);
3257         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3258
3259         return (0);
3260
3261 bad:
3262         m_freem(*mp);
3263         *mp = NULL;
3264         return (error);
3265 }
3266
3267 /*
3268  * Perform basic checks on header size since
3269  * pfil assumes ip_input has already processed
3270  * it for it.  Cut-and-pasted from ip_input.c.
3271  * Given how simple the IPv6 version is,
3272  * does the IPv4 version really need to be
3273  * this complicated?
3274  *
3275  * XXX Should we update ipstat here, or not?
3276  * XXX Right now we update ipstat but not
3277  * XXX csum_counter.
3278  */
3279 static int
3280 bridge_ip_checkbasic(struct mbuf **mp)
3281 {
3282         struct mbuf *m = *mp;
3283         struct ip *ip;
3284         int len, hlen;
3285         u_short sum;
3286
3287         if (*mp == NULL)
3288                 return (-1);
3289
3290         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3291                 if ((m = m_copyup(m, sizeof(struct ip),
3292                         (max_linkhdr + 3) & ~3)) == NULL) {
3293                         /* XXXJRT new stat, please */
3294                         KMOD_IPSTAT_INC(ips_toosmall);
3295                         goto bad;
3296                 }
3297         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3298                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3299                         KMOD_IPSTAT_INC(ips_toosmall);
3300                         goto bad;
3301                 }
3302         }
3303         ip = mtod(m, struct ip *);
3304         if (ip == NULL) goto bad;
3305
3306         if (ip->ip_v != IPVERSION) {
3307                 KMOD_IPSTAT_INC(ips_badvers);
3308                 goto bad;
3309         }
3310         hlen = ip->ip_hl << 2;
3311         if (hlen < sizeof(struct ip)) { /* minimum header length */
3312                 KMOD_IPSTAT_INC(ips_badhlen);
3313                 goto bad;
3314         }
3315         if (hlen > m->m_len) {
3316                 if ((m = m_pullup(m, hlen)) == 0) {
3317                         KMOD_IPSTAT_INC(ips_badhlen);
3318                         goto bad;
3319                 }
3320                 ip = mtod(m, struct ip *);
3321                 if (ip == NULL) goto bad;
3322         }
3323
3324         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3325                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3326         } else {
3327                 if (hlen == sizeof(struct ip)) {
3328                         sum = in_cksum_hdr(ip);
3329                 } else {
3330                         sum = in_cksum(m, hlen);
3331                 }
3332         }
3333         if (sum) {
3334                 KMOD_IPSTAT_INC(ips_badsum);
3335                 goto bad;
3336         }
3337
3338         /* Retrieve the packet length. */
3339         len = ntohs(ip->ip_len);
3340
3341         /*
3342          * Check for additional length bogosity
3343          */
3344         if (len < hlen) {
3345                 KMOD_IPSTAT_INC(ips_badlen);
3346                 goto bad;
3347         }
3348
3349         /*
3350          * Check that the amount of data in the buffers
3351          * is as at least much as the IP header would have us expect.
3352          * Drop packet if shorter than we expect.
3353          */
3354         if (m->m_pkthdr.len < len) {
3355                 KMOD_IPSTAT_INC(ips_tooshort);
3356                 goto bad;
3357         }
3358
3359         /* Checks out, proceed */
3360         *mp = m;
3361         return (0);
3362
3363 bad:
3364         *mp = m;
3365         return (-1);
3366 }
3367
3368 #ifdef INET6
3369 /*
3370  * Same as above, but for IPv6.
3371  * Cut-and-pasted from ip6_input.c.
3372  * XXX Should we update ip6stat, or not?
3373  */
3374 static int
3375 bridge_ip6_checkbasic(struct mbuf **mp)
3376 {
3377         struct mbuf *m = *mp;
3378         struct ip6_hdr *ip6;
3379
3380         /*
3381          * If the IPv6 header is not aligned, slurp it up into a new
3382          * mbuf with space for link headers, in the event we forward
3383          * it.  Otherwise, if it is aligned, make sure the entire base
3384          * IPv6 header is in the first mbuf of the chain.
3385          */
3386         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3387                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3388                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3389                             (max_linkhdr + 3) & ~3)) == NULL) {
3390                         /* XXXJRT new stat, please */
3391                         V_ip6stat.ip6s_toosmall++;
3392                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3393                         goto bad;
3394                 }
3395         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3396                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3397                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3398                         V_ip6stat.ip6s_toosmall++;
3399                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3400                         goto bad;
3401                 }
3402         }
3403
3404         ip6 = mtod(m, struct ip6_hdr *);
3405
3406         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3407                 V_ip6stat.ip6s_badvers++;
3408                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3409                 goto bad;
3410         }
3411
3412         /* Checks out, proceed */
3413         *mp = m;
3414         return (0);
3415
3416 bad:
3417         *mp = m;
3418         return (-1);
3419 }
3420 #endif /* INET6 */
3421
3422 /*
3423  * bridge_fragment:
3424  *
3425  *      Return a fragmented mbuf chain.
3426  */
3427 static int
3428 bridge_fragment(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh,
3429     int snap, struct llc *llc)
3430 {
3431         struct mbuf *m0;
3432         struct ip *ip;
3433         int error = -1;
3434
3435         if (m->m_len < sizeof(struct ip) &&
3436             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3437                 goto out;
3438         ip = mtod(m, struct ip *);
3439
3440         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist,
3441                     CSUM_DELAY_IP);
3442         if (error)
3443                 goto out;
3444
3445         /* walk the chain and re-add the Ethernet header */
3446         for (m0 = m; m0; m0 = m0->m_nextpkt) {
3447                 if (error == 0) {
3448                         if (snap) {
3449                                 M_PREPEND(m0, sizeof(struct llc), M_DONTWAIT);
3450                                 if (m0 == NULL) {
3451                                         error = ENOBUFS;
3452                                         continue;
3453                                 }
3454                                 bcopy(llc, mtod(m0, caddr_t),
3455                                     sizeof(struct llc));
3456                         }
3457                         M_PREPEND(m0, ETHER_HDR_LEN, M_DONTWAIT);
3458                         if (m0 == NULL) {
3459                                 error = ENOBUFS;
3460                                 continue;
3461                         }
3462                         bcopy(eh, mtod(m0, caddr_t), ETHER_HDR_LEN);
3463                 } else 
3464                         m_freem(m);
3465         }
3466
3467         if (error == 0)
3468                 KMOD_IPSTAT_INC(ips_fragmented);
3469
3470         return (error);
3471
3472 out:
3473         if (m != NULL)
3474                 m_freem(m);
3475         return (error);
3476 }
3477
3478 static void
3479 bridge_linkstate(struct ifnet *ifp)
3480 {
3481         struct bridge_softc *sc = ifp->if_bridge;
3482         struct bridge_iflist *bif, *bif2;
3483         int new_link, hasls;
3484
3485         BRIDGE_LOCK(sc);
3486         bif = bridge_lookup_member_if(sc, ifp);
3487         if (bif == NULL) {
3488                 BRIDGE_UNLOCK(sc);
3489                 return;
3490         }
3491         new_link = LINK_STATE_DOWN;
3492         hasls = 0;
3493         /* Our link is considered up if at least one of our ports is active */
3494         LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
3495                 if (bif2->bif_ifp->if_capabilities & IFCAP_LINKSTATE)
3496                         hasls++;
3497                 if (bif2->bif_ifp->if_link_state == LINK_STATE_UP) {
3498                         new_link = LINK_STATE_UP;
3499                         break;
3500                 }
3501         }
3502         if (!LIST_EMPTY(&sc->sc_iflist) && !hasls) {
3503                 /* If no interfaces support link-state then we default to up */
3504                 new_link = LINK_STATE_UP;
3505         }
3506         if_link_state_change(sc->sc_ifp, new_link);
3507         BRIDGE_UNLOCK(sc);
3508
3509         bstp_linkstate(&bif->bif_stp);
3510 }
3511