]> CyberLeo.Net >> Repos - FreeBSD/stable/9.git/blob - sys/net/if_bridge.c
MFC r232014,r232030,r232070
[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
337 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
338 #define VLANTAGOF(_m)   \
339     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : 1
340
341 static struct bstp_cb_ops bridge_ops = {
342         .bcb_state = bridge_state_change,
343         .bcb_rtage = bridge_rtable_expire
344 };
345
346 SYSCTL_DECL(_net_link);
347 SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge");
348
349 static int pfil_onlyip = 1; /* only pass IP[46] packets when pfil is enabled */
350 static int pfil_bridge = 1; /* run pfil hooks on the bridge interface */
351 static int pfil_member = 1; /* run pfil hooks on the member interface */
352 static int pfil_ipfw = 0;   /* layer2 filter with ipfw */
353 static int pfil_ipfw_arp = 0;   /* layer2 filter with ipfw */
354 static int pfil_local_phys = 0; /* run pfil hooks on the physical interface for
355                                    locally destined packets */
356 static int log_stp   = 0;   /* log STP state changes */
357 static int bridge_inherit_mac = 0;   /* share MAC with first bridge member */
358 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip, CTLFLAG_RW,
359     &pfil_onlyip, 0, "Only pass IP packets when pfil is enabled");
360 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp, CTLFLAG_RW,
361     &pfil_ipfw_arp, 0, "Filter ARP packets through IPFW layer2");
362 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge, CTLFLAG_RW,
363     &pfil_bridge, 0, "Packet filter on the bridge interface");
364 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member, CTLFLAG_RW,
365     &pfil_member, 0, "Packet filter on the member interface");
366 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys, CTLFLAG_RW,
367     &pfil_local_phys, 0,
368     "Packet filter on the physical interface for locally destined packets");
369 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp, CTLFLAG_RW,
370     &log_stp, 0, "Log STP state changes");
371 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac, CTLFLAG_RW,
372     &bridge_inherit_mac, 0,
373     "Inherit MAC address from the first bridge member");
374
375 struct bridge_control {
376         int     (*bc_func)(struct bridge_softc *, void *);
377         int     bc_argsize;
378         int     bc_flags;
379 };
380
381 #define BC_F_COPYIN             0x01    /* copy arguments in */
382 #define BC_F_COPYOUT            0x02    /* copy arguments out */
383 #define BC_F_SUSER              0x04    /* do super-user check */
384
385 const struct bridge_control bridge_control_table[] = {
386         { bridge_ioctl_add,             sizeof(struct ifbreq),
387           BC_F_COPYIN|BC_F_SUSER },
388         { bridge_ioctl_del,             sizeof(struct ifbreq),
389           BC_F_COPYIN|BC_F_SUSER },
390
391         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
392           BC_F_COPYIN|BC_F_COPYOUT },
393         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
394           BC_F_COPYIN|BC_F_SUSER },
395
396         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
397           BC_F_COPYIN|BC_F_SUSER },
398         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
399           BC_F_COPYOUT },
400
401         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
402           BC_F_COPYIN|BC_F_COPYOUT },
403         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
404           BC_F_COPYIN|BC_F_COPYOUT },
405
406         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
407           BC_F_COPYIN|BC_F_SUSER },
408
409         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
410           BC_F_COPYIN|BC_F_SUSER },
411         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
412           BC_F_COPYOUT },
413
414         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
415           BC_F_COPYIN|BC_F_SUSER },
416
417         { bridge_ioctl_flush,           sizeof(struct ifbreq),
418           BC_F_COPYIN|BC_F_SUSER },
419
420         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
421           BC_F_COPYOUT },
422         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
423           BC_F_COPYIN|BC_F_SUSER },
424
425         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
426           BC_F_COPYOUT },
427         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
428           BC_F_COPYIN|BC_F_SUSER },
429
430         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
431           BC_F_COPYOUT },
432         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
433           BC_F_COPYIN|BC_F_SUSER },
434
435         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
436           BC_F_COPYOUT },
437         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
438           BC_F_COPYIN|BC_F_SUSER },
439
440         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
441           BC_F_COPYIN|BC_F_SUSER },
442
443         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
444           BC_F_COPYIN|BC_F_SUSER },
445
446         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
447           BC_F_COPYIN|BC_F_SUSER },
448         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
449           BC_F_COPYIN|BC_F_SUSER },
450
451         { bridge_ioctl_gbparam,         sizeof(struct ifbropreq),
452           BC_F_COPYOUT },
453
454         { bridge_ioctl_grte,            sizeof(struct ifbrparam),
455           BC_F_COPYOUT },
456
457         { bridge_ioctl_gifsstp,         sizeof(struct ifbpstpconf),
458           BC_F_COPYIN|BC_F_COPYOUT },
459
460         { bridge_ioctl_sproto,          sizeof(struct ifbrparam),
461           BC_F_COPYIN|BC_F_SUSER },
462
463         { bridge_ioctl_stxhc,           sizeof(struct ifbrparam),
464           BC_F_COPYIN|BC_F_SUSER },
465
466         { bridge_ioctl_sifmaxaddr,      sizeof(struct ifbreq),
467           BC_F_COPYIN|BC_F_SUSER },
468
469 };
470 const int bridge_control_table_size =
471     sizeof(bridge_control_table) / sizeof(bridge_control_table[0]);
472
473 LIST_HEAD(, bridge_softc) bridge_list;
474
475 IFC_SIMPLE_DECLARE(bridge, 0);
476
477 static int
478 bridge_modevent(module_t mod, int type, void *data)
479 {
480
481         switch (type) {
482         case MOD_LOAD:
483                 mtx_init(&bridge_list_mtx, "if_bridge list", NULL, MTX_DEF);
484                 if_clone_attach(&bridge_cloner);
485                 bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
486                     sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
487                     UMA_ALIGN_PTR, 0);
488                 LIST_INIT(&bridge_list);
489                 bridge_input_p = bridge_input;
490                 bridge_output_p = bridge_output;
491                 bridge_dn_p = bridge_dummynet;
492                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
493                     ifnet_departure_event, bridge_ifdetach, NULL,
494                     EVENTHANDLER_PRI_ANY);
495                 break;
496         case MOD_UNLOAD:
497                 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
498                     bridge_detach_cookie);
499                 if_clone_detach(&bridge_cloner);
500                 uma_zdestroy(bridge_rtnode_zone);
501                 bridge_input_p = NULL;
502                 bridge_output_p = NULL;
503                 bridge_dn_p = NULL;
504                 mtx_destroy(&bridge_list_mtx);
505                 break;
506         default:
507                 return (EOPNOTSUPP);
508         }
509         return (0);
510 }
511
512 static moduledata_t bridge_mod = {
513         "if_bridge",
514         bridge_modevent,
515         0
516 };
517
518 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
519 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
520
521 /*
522  * handler for net.link.bridge.pfil_ipfw
523  */
524 static int
525 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
526 {
527         int enable = pfil_ipfw;
528         int error;
529
530         error = sysctl_handle_int(oidp, &enable, 0, req);
531         enable = (enable) ? 1 : 0;
532
533         if (enable != pfil_ipfw) {
534                 pfil_ipfw = enable;
535
536                 /*
537                  * Disable pfil so that ipfw doesnt run twice, if the user
538                  * really wants both then they can re-enable pfil_bridge and/or
539                  * pfil_member. Also allow non-ip packets as ipfw can filter by
540                  * layer2 type.
541                  */
542                 if (pfil_ipfw) {
543                         pfil_onlyip = 0;
544                         pfil_bridge = 0;
545                         pfil_member = 0;
546                 }
547         }
548
549         return (error);
550 }
551 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw, CTLTYPE_INT|CTLFLAG_RW,
552             &pfil_ipfw, 0, &sysctl_pfil_ipfw, "I", "Layer2 filter with IPFW");
553
554 /*
555  * bridge_clone_create:
556  *
557  *      Create a new bridge instance.
558  */
559 static int
560 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t params)
561 {
562         struct bridge_softc *sc, *sc2;
563         struct ifnet *bifp, *ifp;
564         int fb, retry;
565         unsigned long hostid;
566
567         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
568         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
569         if (ifp == NULL) {
570                 free(sc, M_DEVBUF);
571                 return (ENOSPC);
572         }
573
574         BRIDGE_LOCK_INIT(sc);
575         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
576         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
577
578         /* Initialize our routing table. */
579         bridge_rtable_init(sc);
580
581         callout_init_mtx(&sc->sc_brcallout, &sc->sc_mtx, 0);
582
583         LIST_INIT(&sc->sc_iflist);
584         LIST_INIT(&sc->sc_spanlist);
585
586         ifp->if_softc = sc;
587         if_initname(ifp, ifc->ifc_name, unit);
588         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
589         ifp->if_ioctl = bridge_ioctl;
590         ifp->if_start = bridge_start;
591         ifp->if_init = bridge_init;
592         ifp->if_type = IFT_BRIDGE;
593         IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
594         ifp->if_snd.ifq_drv_maxlen = ifqmaxlen;
595         IFQ_SET_READY(&ifp->if_snd);
596
597         /*
598          * Generate an ethernet address with a locally administered address.
599          *
600          * Since we are using random ethernet addresses for the bridge, it is
601          * possible that we might have address collisions, so make sure that
602          * this hardware address isn't already in use on another bridge.
603          * The first try uses the hostid and falls back to arc4rand().
604          */
605         fb = 0;
606         getcredhostid(curthread->td_ucred, &hostid);
607         for (retry = 1; retry != 0;) {
608                 if (fb || hostid == 0) {
609                         arc4rand(sc->sc_defaddr, ETHER_ADDR_LEN, 1);
610                         sc->sc_defaddr[0] &= ~1;/* clear multicast bit */
611                         sc->sc_defaddr[0] |= 2; /* set the LAA bit */
612                 } else {
613                         sc->sc_defaddr[0] = 0x2;
614                         sc->sc_defaddr[1] = (hostid >> 24) & 0xff;
615                         sc->sc_defaddr[2] = (hostid >> 16) & 0xff;
616                         sc->sc_defaddr[3] = (hostid >> 8 ) & 0xff;
617                         sc->sc_defaddr[4] =  hostid        & 0xff;
618                         sc->sc_defaddr[5] = ifp->if_dunit & 0xff;
619                 }
620
621                 fb = 1;
622                 retry = 0;
623                 mtx_lock(&bridge_list_mtx);
624                 LIST_FOREACH(sc2, &bridge_list, sc_list) {
625                         bifp = sc2->sc_ifp;
626                         if (memcmp(sc->sc_defaddr,
627                             IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0)
628                                 retry = 1;
629                 }
630                 mtx_unlock(&bridge_list_mtx);
631         }
632
633         bstp_attach(&sc->sc_stp, &bridge_ops);
634         ether_ifattach(ifp, sc->sc_defaddr);
635         /* Now undo some of the damage... */
636         ifp->if_baudrate = 0;
637         ifp->if_type = IFT_BRIDGE;
638
639         mtx_lock(&bridge_list_mtx);
640         LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
641         mtx_unlock(&bridge_list_mtx);
642
643         return (0);
644 }
645
646 /*
647  * bridge_clone_destroy:
648  *
649  *      Destroy a bridge instance.
650  */
651 static void
652 bridge_clone_destroy(struct ifnet *ifp)
653 {
654         struct bridge_softc *sc = ifp->if_softc;
655         struct bridge_iflist *bif;
656
657         BRIDGE_LOCK(sc);
658
659         bridge_stop(ifp, 1);
660         ifp->if_flags &= ~IFF_UP;
661
662         while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
663                 bridge_delete_member(sc, bif, 0);
664
665         while ((bif = LIST_FIRST(&sc->sc_spanlist)) != NULL) {
666                 bridge_delete_span(sc, bif);
667         }
668
669         BRIDGE_UNLOCK(sc);
670
671         callout_drain(&sc->sc_brcallout);
672
673         mtx_lock(&bridge_list_mtx);
674         LIST_REMOVE(sc, sc_list);
675         mtx_unlock(&bridge_list_mtx);
676
677         bstp_detach(&sc->sc_stp);
678         ether_ifdetach(ifp);
679         if_free_type(ifp, IFT_ETHER);
680
681         /* Tear down the routing table. */
682         bridge_rtable_fini(sc);
683
684         BRIDGE_LOCK_DESTROY(sc);
685         free(sc, M_DEVBUF);
686 }
687
688 /*
689  * bridge_ioctl:
690  *
691  *      Handle a control request from the operator.
692  */
693 static int
694 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
695 {
696         struct bridge_softc *sc = ifp->if_softc;
697         struct ifreq *ifr = (struct ifreq *)data;
698         struct bridge_iflist *bif;
699         struct thread *td = curthread;
700         union {
701                 struct ifbreq ifbreq;
702                 struct ifbifconf ifbifconf;
703                 struct ifbareq ifbareq;
704                 struct ifbaconf ifbaconf;
705                 struct ifbrparam ifbrparam;
706                 struct ifbropreq ifbropreq;
707         } args;
708         struct ifdrv *ifd = (struct ifdrv *) data;
709         const struct bridge_control *bc;
710         int error = 0;
711
712         switch (cmd) {
713
714         case SIOCADDMULTI:
715         case SIOCDELMULTI:
716                 break;
717
718         case SIOCGDRVSPEC:
719         case SIOCSDRVSPEC:
720                 if (ifd->ifd_cmd >= bridge_control_table_size) {
721                         error = EINVAL;
722                         break;
723                 }
724                 bc = &bridge_control_table[ifd->ifd_cmd];
725
726                 if (cmd == SIOCGDRVSPEC &&
727                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
728                         error = EINVAL;
729                         break;
730                 }
731                 else if (cmd == SIOCSDRVSPEC &&
732                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
733                         error = EINVAL;
734                         break;
735                 }
736
737                 if (bc->bc_flags & BC_F_SUSER) {
738                         error = priv_check(td, PRIV_NET_BRIDGE);
739                         if (error)
740                                 break;
741                 }
742
743                 if (ifd->ifd_len != bc->bc_argsize ||
744                     ifd->ifd_len > sizeof(args)) {
745                         error = EINVAL;
746                         break;
747                 }
748
749                 bzero(&args, sizeof(args));
750                 if (bc->bc_flags & BC_F_COPYIN) {
751                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
752                         if (error)
753                                 break;
754                 }
755
756                 BRIDGE_LOCK(sc);
757                 error = (*bc->bc_func)(sc, &args);
758                 BRIDGE_UNLOCK(sc);
759                 if (error)
760                         break;
761
762                 if (bc->bc_flags & BC_F_COPYOUT)
763                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
764
765                 break;
766
767         case SIOCSIFFLAGS:
768                 if (!(ifp->if_flags & IFF_UP) &&
769                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
770                         /*
771                          * If interface is marked down and it is running,
772                          * then stop and disable it.
773                          */
774                         BRIDGE_LOCK(sc);
775                         bridge_stop(ifp, 1);
776                         BRIDGE_UNLOCK(sc);
777                 } else if ((ifp->if_flags & IFF_UP) &&
778                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
779                         /*
780                          * If interface is marked up and it is stopped, then
781                          * start it.
782                          */
783                         (*ifp->if_init)(sc);
784                 }
785                 break;
786
787         case SIOCSIFMTU:
788                 if (ifr->ifr_mtu < 576) {
789                         error = EINVAL;
790                         break;
791                 }
792                 if (LIST_EMPTY(&sc->sc_iflist)) {
793                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
794                         break;
795                 }
796                 BRIDGE_LOCK(sc);
797                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
798                         if (bif->bif_ifp->if_mtu != ifr->ifr_mtu) {
799                                 log(LOG_NOTICE, "%s: invalid MTU: %lu(%s)"
800                                     " != %d\n", sc->sc_ifp->if_xname,
801                                     bif->bif_ifp->if_mtu,
802                                     bif->bif_ifp->if_xname, ifr->ifr_mtu);
803                                 error = EINVAL;
804                                 break;
805                         }
806                 }
807                 if (!error)
808                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
809                 BRIDGE_UNLOCK(sc);
810                 break;
811         default:
812                 /*
813                  * drop the lock as ether_ioctl() will call bridge_start() and
814                  * cause the lock to be recursed.
815                  */
816                 error = ether_ioctl(ifp, cmd, data);
817                 break;
818         }
819
820         return (error);
821 }
822
823 /*
824  * bridge_mutecaps:
825  *
826  *      Clear or restore unwanted capabilities on the member interface
827  */
828 static void
829 bridge_mutecaps(struct bridge_softc *sc)
830 {
831         struct bridge_iflist *bif;
832         int enabled, mask;
833
834         /* Initial bitmask of capabilities to test */
835         mask = BRIDGE_IFCAPS_MASK;
836
837         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
838                 /* Every member must support it or its disabled */
839                 mask &= bif->bif_savedcaps;
840         }
841
842         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
843                 enabled = bif->bif_ifp->if_capenable;
844                 enabled &= ~BRIDGE_IFCAPS_STRIP;
845                 /* strip off mask bits and enable them again if allowed */
846                 enabled &= ~BRIDGE_IFCAPS_MASK;
847                 enabled |= mask;
848                 bridge_set_ifcap(sc, bif, enabled);
849         }
850
851 }
852
853 static void
854 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
855 {
856         struct ifnet *ifp = bif->bif_ifp;
857         struct ifreq ifr;
858         int error;
859
860         bzero(&ifr, sizeof(ifr));
861         ifr.ifr_reqcap = set;
862
863         if (ifp->if_capenable != set) {
864                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
865                 if (error)
866                         if_printf(sc->sc_ifp,
867                             "error setting interface capabilities on %s\n",
868                             ifp->if_xname);
869         }
870 }
871
872 /*
873  * bridge_lookup_member:
874  *
875  *      Lookup a bridge member interface.
876  */
877 static struct bridge_iflist *
878 bridge_lookup_member(struct bridge_softc *sc, const char *name)
879 {
880         struct bridge_iflist *bif;
881         struct ifnet *ifp;
882
883         BRIDGE_LOCK_ASSERT(sc);
884
885         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
886                 ifp = bif->bif_ifp;
887                 if (strcmp(ifp->if_xname, name) == 0)
888                         return (bif);
889         }
890
891         return (NULL);
892 }
893
894 /*
895  * bridge_lookup_member_if:
896  *
897  *      Lookup a bridge member interface by ifnet*.
898  */
899 static struct bridge_iflist *
900 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
901 {
902         struct bridge_iflist *bif;
903
904         BRIDGE_LOCK_ASSERT(sc);
905
906         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
907                 if (bif->bif_ifp == member_ifp)
908                         return (bif);
909         }
910
911         return (NULL);
912 }
913
914 /*
915  * bridge_delete_member:
916  *
917  *      Delete the specified member interface.
918  */
919 static void
920 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
921     int gone)
922 {
923         struct ifnet *ifs = bif->bif_ifp;
924         struct ifnet *fif = NULL;
925
926         BRIDGE_LOCK_ASSERT(sc);
927
928         if (bif->bif_flags & IFBIF_STP)
929                 bstp_disable(&bif->bif_stp);
930
931         ifs->if_bridge = NULL;
932         BRIDGE_XLOCK(sc);
933         LIST_REMOVE(bif, bif_next);
934         BRIDGE_XDROP(sc);
935
936         /*
937          * If removing the interface that gave the bridge its mac address, set
938          * the mac address of the bridge to the address of the next member, or
939          * to its default address if no members are left.
940          */
941         if (bridge_inherit_mac && sc->sc_ifaddr == ifs) {
942                 if (LIST_EMPTY(&sc->sc_iflist)) {
943                         bcopy(sc->sc_defaddr,
944                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
945                         sc->sc_ifaddr = NULL;
946                 } else {
947                         fif = LIST_FIRST(&sc->sc_iflist)->bif_ifp;
948                         bcopy(IF_LLADDR(fif),
949                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
950                         sc->sc_ifaddr = fif;
951                 }
952                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
953         }
954
955         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
956         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
957         KASSERT(bif->bif_addrcnt == 0,
958             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
959
960         BRIDGE_UNLOCK(sc);
961         if (!gone) {
962                 switch (ifs->if_type) {
963                 case IFT_ETHER:
964                 case IFT_L2VLAN:
965                         /*
966                          * Take the interface out of promiscuous mode.
967                          */
968                         (void) ifpromisc(ifs, 0);
969                         break;
970
971                 case IFT_GIF:
972                         break;
973
974                 default:
975 #ifdef DIAGNOSTIC
976                         panic("bridge_delete_member: impossible");
977 #endif
978                         break;
979                 }
980                 /* reneable any interface capabilities */
981                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
982         }
983         bstp_destroy(&bif->bif_stp);    /* prepare to free */
984         BRIDGE_LOCK(sc);
985         free(bif, M_DEVBUF);
986 }
987
988 /*
989  * bridge_delete_span:
990  *
991  *      Delete the specified span interface.
992  */
993 static void
994 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
995 {
996         BRIDGE_LOCK_ASSERT(sc);
997
998         KASSERT(bif->bif_ifp->if_bridge == NULL,
999             ("%s: not a span interface", __func__));
1000
1001         LIST_REMOVE(bif, bif_next);
1002         free(bif, M_DEVBUF);
1003 }
1004
1005 static int
1006 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1007 {
1008         struct ifbreq *req = arg;
1009         struct bridge_iflist *bif = NULL;
1010         struct ifnet *ifs;
1011         int error = 0;
1012
1013         ifs = ifunit(req->ifbr_ifsname);
1014         if (ifs == NULL)
1015                 return (ENOENT);
1016         if (ifs->if_ioctl == NULL)      /* must be supported */
1017                 return (EINVAL);
1018
1019         /* If it's in the span list, it can't be a member. */
1020         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1021                 if (ifs == bif->bif_ifp)
1022                         return (EBUSY);
1023
1024         if (ifs->if_bridge == sc)
1025                 return (EEXIST);
1026
1027         if (ifs->if_bridge != NULL)
1028                 return (EBUSY);
1029
1030         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1031         if (bif == NULL)
1032                 return (ENOMEM);
1033
1034         bif->bif_ifp = ifs;
1035         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1036         bif->bif_savedcaps = ifs->if_capenable;
1037
1038         switch (ifs->if_type) {
1039         case IFT_ETHER:
1040         case IFT_L2VLAN:
1041         case IFT_GIF:
1042                 /* permitted interface types */
1043                 break;
1044         default:
1045                 error = EINVAL;
1046                 goto out;
1047         }
1048
1049         /* Allow the first Ethernet member to define the MTU */
1050         if (LIST_EMPTY(&sc->sc_iflist))
1051                 sc->sc_ifp->if_mtu = ifs->if_mtu;
1052         else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1053                 if_printf(sc->sc_ifp, "invalid MTU: %lu(%s) != %lu\n",
1054                     ifs->if_mtu, ifs->if_xname, sc->sc_ifp->if_mtu);
1055                 error = EINVAL;
1056                 goto out;
1057         }
1058
1059         /*
1060          * Assign the interface's MAC address to the bridge if it's the first
1061          * member and the MAC address of the bridge has not been changed from
1062          * the default randomly generated one.
1063          */
1064         if (bridge_inherit_mac && LIST_EMPTY(&sc->sc_iflist) &&
1065             !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr, ETHER_ADDR_LEN)) {
1066                 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1067                 sc->sc_ifaddr = ifs;
1068                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1069         }
1070
1071         ifs->if_bridge = sc;
1072         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1073         /*
1074          * XXX: XLOCK HERE!?!
1075          *
1076          * NOTE: insert_***HEAD*** should be safe for the traversals.
1077          */
1078         LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1079
1080         /* Set interface capabilities to the intersection set of all members */
1081         bridge_mutecaps(sc);
1082
1083         switch (ifs->if_type) {
1084         case IFT_ETHER:
1085         case IFT_L2VLAN:
1086                 /*
1087                  * Place the interface into promiscuous mode.
1088                  */
1089                 BRIDGE_UNLOCK(sc);
1090                 error = ifpromisc(ifs, 1);
1091                 BRIDGE_LOCK(sc);
1092                 break;
1093         }
1094         if (error)
1095                 bridge_delete_member(sc, bif, 0);
1096 out:
1097         if (error) {
1098                 if (bif != NULL)
1099                         free(bif, M_DEVBUF);
1100         }
1101         return (error);
1102 }
1103
1104 static int
1105 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1106 {
1107         struct ifbreq *req = arg;
1108         struct bridge_iflist *bif;
1109
1110         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1111         if (bif == NULL)
1112                 return (ENOENT);
1113
1114         bridge_delete_member(sc, bif, 0);
1115
1116         return (0);
1117 }
1118
1119 static int
1120 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1121 {
1122         struct ifbreq *req = arg;
1123         struct bridge_iflist *bif;
1124         struct bstp_port *bp;
1125
1126         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1127         if (bif == NULL)
1128                 return (ENOENT);
1129
1130         bp = &bif->bif_stp;
1131         req->ifbr_ifsflags = bif->bif_flags;
1132         req->ifbr_state = bp->bp_state;
1133         req->ifbr_priority = bp->bp_priority;
1134         req->ifbr_path_cost = bp->bp_path_cost;
1135         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1136         req->ifbr_proto = bp->bp_protover;
1137         req->ifbr_role = bp->bp_role;
1138         req->ifbr_stpflags = bp->bp_flags;
1139         req->ifbr_addrcnt = bif->bif_addrcnt;
1140         req->ifbr_addrmax = bif->bif_addrmax;
1141         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1142
1143         /* Copy STP state options as flags */
1144         if (bp->bp_operedge)
1145                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1146         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1147                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1148         if (bp->bp_ptp_link)
1149                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1150         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1151                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1152         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1153                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1154         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1155                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1156         return (0);
1157 }
1158
1159 static int
1160 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1161 {
1162         struct ifbreq *req = arg;
1163         struct bridge_iflist *bif;
1164         struct bstp_port *bp;
1165         int error;
1166
1167         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1168         if (bif == NULL)
1169                 return (ENOENT);
1170         bp = &bif->bif_stp;
1171
1172         if (req->ifbr_ifsflags & IFBIF_SPAN)
1173                 /* SPAN is readonly */
1174                 return (EINVAL);
1175
1176         if (req->ifbr_ifsflags & IFBIF_STP) {
1177                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1178                         error = bstp_enable(&bif->bif_stp);
1179                         if (error)
1180                                 return (error);
1181                 }
1182         } else {
1183                 if ((bif->bif_flags & IFBIF_STP) != 0)
1184                         bstp_disable(&bif->bif_stp);
1185         }
1186
1187         /* Pass on STP flags */
1188         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1189         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1190         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1191         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1192
1193         /* Save the bits relating to the bridge */
1194         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1195
1196         return (0);
1197 }
1198
1199 static int
1200 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1201 {
1202         struct ifbrparam *param = arg;
1203
1204         sc->sc_brtmax = param->ifbrp_csize;
1205         bridge_rttrim(sc);
1206
1207         return (0);
1208 }
1209
1210 static int
1211 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1212 {
1213         struct ifbrparam *param = arg;
1214
1215         param->ifbrp_csize = sc->sc_brtmax;
1216
1217         return (0);
1218 }
1219
1220 static int
1221 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1222 {
1223         struct ifbifconf *bifc = arg;
1224         struct bridge_iflist *bif;
1225         struct ifbreq breq;
1226         char *buf, *outbuf;
1227         int count, buflen, len, error = 0;
1228
1229         count = 0;
1230         LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1231                 count++;
1232         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1233                 count++;
1234
1235         buflen = sizeof(breq) * count;
1236         if (bifc->ifbic_len == 0) {
1237                 bifc->ifbic_len = buflen;
1238                 return (0);
1239         }
1240         BRIDGE_UNLOCK(sc);
1241         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1242         BRIDGE_LOCK(sc);
1243
1244         count = 0;
1245         buf = outbuf;
1246         len = min(bifc->ifbic_len, buflen);
1247         bzero(&breq, sizeof(breq));
1248         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1249                 if (len < sizeof(breq))
1250                         break;
1251
1252                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1253                     sizeof(breq.ifbr_ifsname));
1254                 /* Fill in the ifbreq structure */
1255                 error = bridge_ioctl_gifflags(sc, &breq);
1256                 if (error)
1257                         break;
1258                 memcpy(buf, &breq, sizeof(breq));
1259                 count++;
1260                 buf += sizeof(breq);
1261                 len -= sizeof(breq);
1262         }
1263         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1264                 if (len < sizeof(breq))
1265                         break;
1266
1267                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1268                     sizeof(breq.ifbr_ifsname));
1269                 breq.ifbr_ifsflags = bif->bif_flags;
1270                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1271                 memcpy(buf, &breq, sizeof(breq));
1272                 count++;
1273                 buf += sizeof(breq);
1274                 len -= sizeof(breq);
1275         }
1276
1277         BRIDGE_UNLOCK(sc);
1278         bifc->ifbic_len = sizeof(breq) * count;
1279         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1280         BRIDGE_LOCK(sc);
1281         free(outbuf, M_TEMP);
1282         return (error);
1283 }
1284
1285 static int
1286 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1287 {
1288         struct ifbaconf *bac = arg;
1289         struct bridge_rtnode *brt;
1290         struct ifbareq bareq;
1291         char *buf, *outbuf;
1292         int count, buflen, len, error = 0;
1293
1294         if (bac->ifbac_len == 0)
1295                 return (0);
1296
1297         count = 0;
1298         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1299                 count++;
1300         buflen = sizeof(bareq) * count;
1301
1302         BRIDGE_UNLOCK(sc);
1303         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1304         BRIDGE_LOCK(sc);
1305
1306         count = 0;
1307         buf = outbuf;
1308         len = min(bac->ifbac_len, buflen);
1309         bzero(&bareq, sizeof(bareq));
1310         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1311                 if (len < sizeof(bareq))
1312                         goto out;
1313                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1314                     sizeof(bareq.ifba_ifsname));
1315                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1316                 bareq.ifba_vlan = brt->brt_vlan;
1317                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1318                                 time_uptime < brt->brt_expire)
1319                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1320                 else
1321                         bareq.ifba_expire = 0;
1322                 bareq.ifba_flags = brt->brt_flags;
1323
1324                 memcpy(buf, &bareq, sizeof(bareq));
1325                 count++;
1326                 buf += sizeof(bareq);
1327                 len -= sizeof(bareq);
1328         }
1329 out:
1330         BRIDGE_UNLOCK(sc);
1331         bac->ifbac_len = sizeof(bareq) * count;
1332         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1333         BRIDGE_LOCK(sc);
1334         free(outbuf, M_TEMP);
1335         return (error);
1336 }
1337
1338 static int
1339 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1340 {
1341         struct ifbareq *req = arg;
1342         struct bridge_iflist *bif;
1343         int error;
1344
1345         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1346         if (bif == NULL)
1347                 return (ENOENT);
1348
1349         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1350             req->ifba_flags);
1351
1352         return (error);
1353 }
1354
1355 static int
1356 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1357 {
1358         struct ifbrparam *param = arg;
1359
1360         sc->sc_brttimeout = param->ifbrp_ctime;
1361         return (0);
1362 }
1363
1364 static int
1365 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1366 {
1367         struct ifbrparam *param = arg;
1368
1369         param->ifbrp_ctime = sc->sc_brttimeout;
1370         return (0);
1371 }
1372
1373 static int
1374 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1375 {
1376         struct ifbareq *req = arg;
1377
1378         return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1379 }
1380
1381 static int
1382 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1383 {
1384         struct ifbreq *req = arg;
1385
1386         bridge_rtflush(sc, req->ifbr_ifsflags);
1387         return (0);
1388 }
1389
1390 static int
1391 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1392 {
1393         struct ifbrparam *param = arg;
1394         struct bstp_state *bs = &sc->sc_stp;
1395
1396         param->ifbrp_prio = bs->bs_bridge_priority;
1397         return (0);
1398 }
1399
1400 static int
1401 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1402 {
1403         struct ifbrparam *param = arg;
1404
1405         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1406 }
1407
1408 static int
1409 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1410 {
1411         struct ifbrparam *param = arg;
1412         struct bstp_state *bs = &sc->sc_stp;
1413
1414         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1415         return (0);
1416 }
1417
1418 static int
1419 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1420 {
1421         struct ifbrparam *param = arg;
1422
1423         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1424 }
1425
1426 static int
1427 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1428 {
1429         struct ifbrparam *param = arg;
1430         struct bstp_state *bs = &sc->sc_stp;
1431
1432         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1433         return (0);
1434 }
1435
1436 static int
1437 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1438 {
1439         struct ifbrparam *param = arg;
1440
1441         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1442 }
1443
1444 static int
1445 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1446 {
1447         struct ifbrparam *param = arg;
1448         struct bstp_state *bs = &sc->sc_stp;
1449
1450         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1451         return (0);
1452 }
1453
1454 static int
1455 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1456 {
1457         struct ifbrparam *param = arg;
1458
1459         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1460 }
1461
1462 static int
1463 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1464 {
1465         struct ifbreq *req = arg;
1466         struct bridge_iflist *bif;
1467
1468         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1469         if (bif == NULL)
1470                 return (ENOENT);
1471
1472         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1473 }
1474
1475 static int
1476 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1477 {
1478         struct ifbreq *req = arg;
1479         struct bridge_iflist *bif;
1480
1481         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1482         if (bif == NULL)
1483                 return (ENOENT);
1484
1485         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1486 }
1487
1488 static int
1489 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1490 {
1491         struct ifbreq *req = arg;
1492         struct bridge_iflist *bif;
1493
1494         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1495         if (bif == NULL)
1496                 return (ENOENT);
1497
1498         bif->bif_addrmax = req->ifbr_addrmax;
1499         return (0);
1500 }
1501
1502 static int
1503 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1504 {
1505         struct ifbreq *req = arg;
1506         struct bridge_iflist *bif = NULL;
1507         struct ifnet *ifs;
1508
1509         ifs = ifunit(req->ifbr_ifsname);
1510         if (ifs == NULL)
1511                 return (ENOENT);
1512
1513         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1514                 if (ifs == bif->bif_ifp)
1515                         return (EBUSY);
1516
1517         if (ifs->if_bridge != NULL)
1518                 return (EBUSY);
1519
1520         switch (ifs->if_type) {
1521                 case IFT_ETHER:
1522                 case IFT_GIF:
1523                 case IFT_L2VLAN:
1524                         break;
1525                 default:
1526                         return (EINVAL);
1527         }
1528
1529         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1530         if (bif == NULL)
1531                 return (ENOMEM);
1532
1533         bif->bif_ifp = ifs;
1534         bif->bif_flags = IFBIF_SPAN;
1535
1536         LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1537
1538         return (0);
1539 }
1540
1541 static int
1542 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1543 {
1544         struct ifbreq *req = arg;
1545         struct bridge_iflist *bif;
1546         struct ifnet *ifs;
1547
1548         ifs = ifunit(req->ifbr_ifsname);
1549         if (ifs == NULL)
1550                 return (ENOENT);
1551
1552         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1553                 if (ifs == bif->bif_ifp)
1554                         break;
1555
1556         if (bif == NULL)
1557                 return (ENOENT);
1558
1559         bridge_delete_span(sc, bif);
1560
1561         return (0);
1562 }
1563
1564 static int
1565 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1566 {
1567         struct ifbropreq *req = arg;
1568         struct bstp_state *bs = &sc->sc_stp;
1569         struct bstp_port *root_port;
1570
1571         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1572         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1573         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1574
1575         root_port = bs->bs_root_port;
1576         if (root_port == NULL)
1577                 req->ifbop_root_port = 0;
1578         else
1579                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1580
1581         req->ifbop_holdcount = bs->bs_txholdcount;
1582         req->ifbop_priority = bs->bs_bridge_priority;
1583         req->ifbop_protocol = bs->bs_protover;
1584         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1585         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1586         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1587         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1588         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1589         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1590
1591         return (0);
1592 }
1593
1594 static int
1595 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1596 {
1597         struct ifbrparam *param = arg;
1598
1599         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1600         return (0);
1601 }
1602
1603 static int
1604 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1605 {
1606         struct ifbpstpconf *bifstp = arg;
1607         struct bridge_iflist *bif;
1608         struct bstp_port *bp;
1609         struct ifbpstpreq bpreq;
1610         char *buf, *outbuf;
1611         int count, buflen, len, error = 0;
1612
1613         count = 0;
1614         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1615                 if ((bif->bif_flags & IFBIF_STP) != 0)
1616                         count++;
1617         }
1618
1619         buflen = sizeof(bpreq) * count;
1620         if (bifstp->ifbpstp_len == 0) {
1621                 bifstp->ifbpstp_len = buflen;
1622                 return (0);
1623         }
1624
1625         BRIDGE_UNLOCK(sc);
1626         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1627         BRIDGE_LOCK(sc);
1628
1629         count = 0;
1630         buf = outbuf;
1631         len = min(bifstp->ifbpstp_len, buflen);
1632         bzero(&bpreq, sizeof(bpreq));
1633         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1634                 if (len < sizeof(bpreq))
1635                         break;
1636
1637                 if ((bif->bif_flags & IFBIF_STP) == 0)
1638                         continue;
1639
1640                 bp = &bif->bif_stp;
1641                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1642                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1643                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1644                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1645                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1646                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1647
1648                 memcpy(buf, &bpreq, sizeof(bpreq));
1649                 count++;
1650                 buf += sizeof(bpreq);
1651                 len -= sizeof(bpreq);
1652         }
1653
1654         BRIDGE_UNLOCK(sc);
1655         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1656         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1657         BRIDGE_LOCK(sc);
1658         free(outbuf, M_TEMP);
1659         return (error);
1660 }
1661
1662 static int
1663 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1664 {
1665         struct ifbrparam *param = arg;
1666
1667         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1668 }
1669
1670 static int
1671 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1672 {
1673         struct ifbrparam *param = arg;
1674
1675         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1676 }
1677
1678 /*
1679  * bridge_ifdetach:
1680  *
1681  *      Detach an interface from a bridge.  Called when a member
1682  *      interface is detaching.
1683  */
1684 static void
1685 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1686 {
1687         struct bridge_softc *sc = ifp->if_bridge;
1688         struct bridge_iflist *bif;
1689
1690         /* Check if the interface is a bridge member */
1691         if (sc != NULL) {
1692                 BRIDGE_LOCK(sc);
1693
1694                 bif = bridge_lookup_member_if(sc, ifp);
1695                 if (bif != NULL)
1696                         bridge_delete_member(sc, bif, 1);
1697
1698                 BRIDGE_UNLOCK(sc);
1699                 return;
1700         }
1701
1702         /* Check if the interface is a span port */
1703         mtx_lock(&bridge_list_mtx);
1704         LIST_FOREACH(sc, &bridge_list, sc_list) {
1705                 BRIDGE_LOCK(sc);
1706                 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1707                         if (ifp == bif->bif_ifp) {
1708                                 bridge_delete_span(sc, bif);
1709                                 break;
1710                         }
1711
1712                 BRIDGE_UNLOCK(sc);
1713         }
1714         mtx_unlock(&bridge_list_mtx);
1715 }
1716
1717 /*
1718  * bridge_init:
1719  *
1720  *      Initialize a bridge interface.
1721  */
1722 static void
1723 bridge_init(void *xsc)
1724 {
1725         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1726         struct ifnet *ifp = sc->sc_ifp;
1727
1728         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1729                 return;
1730
1731         BRIDGE_LOCK(sc);
1732         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1733             bridge_timer, sc);
1734
1735         ifp->if_drv_flags |= IFF_DRV_RUNNING;
1736         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
1737
1738         BRIDGE_UNLOCK(sc);
1739 }
1740
1741 /*
1742  * bridge_stop:
1743  *
1744  *      Stop the bridge interface.
1745  */
1746 static void
1747 bridge_stop(struct ifnet *ifp, int disable)
1748 {
1749         struct bridge_softc *sc = ifp->if_softc;
1750
1751         BRIDGE_LOCK_ASSERT(sc);
1752
1753         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1754                 return;
1755
1756         callout_stop(&sc->sc_brcallout);
1757         bstp_stop(&sc->sc_stp);
1758
1759         bridge_rtflush(sc, IFBF_FLUSHDYN);
1760
1761         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1762 }
1763
1764 /*
1765  * bridge_enqueue:
1766  *
1767  *      Enqueue a packet on a bridge member interface.
1768  *
1769  */
1770 static void
1771 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
1772 {
1773         int len, err = 0;
1774         short mflags;
1775         struct mbuf *m0;
1776
1777         len = m->m_pkthdr.len;
1778         mflags = m->m_flags;
1779
1780         /* We may be sending a fragment so traverse the mbuf */
1781         for (; m; m = m0) {
1782                 m0 = m->m_nextpkt;
1783                 m->m_nextpkt = NULL;
1784
1785                 /*
1786                  * If underlying interface can not do VLAN tag insertion itself
1787                  * then attach a packet tag that holds it.
1788                  */
1789                 if ((m->m_flags & M_VLANTAG) &&
1790                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
1791                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
1792                         if (m == NULL) {
1793                                 if_printf(dst_ifp,
1794                                     "unable to prepend VLAN header\n");
1795                                 dst_ifp->if_oerrors++;
1796                                 continue;
1797                         }
1798                         m->m_flags &= ~M_VLANTAG;
1799                 }
1800
1801                 if (err == 0)
1802                         dst_ifp->if_transmit(dst_ifp, m);
1803         }
1804
1805         if (err == 0) {
1806                 sc->sc_ifp->if_opackets++;
1807                 sc->sc_ifp->if_obytes += len;
1808                 if (mflags & M_MCAST)
1809                         sc->sc_ifp->if_omcasts++;
1810         }
1811 }
1812
1813 /*
1814  * bridge_dummynet:
1815  *
1816  *      Receive a queued packet from dummynet and pass it on to the output
1817  *      interface.
1818  *
1819  *      The mbuf has the Ethernet header already attached.
1820  */
1821 static void
1822 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
1823 {
1824         struct bridge_softc *sc;
1825
1826         sc = ifp->if_bridge;
1827
1828         /*
1829          * The packet didnt originate from a member interface. This should only
1830          * ever happen if a member interface is removed while packets are
1831          * queued for it.
1832          */
1833         if (sc == NULL) {
1834                 m_freem(m);
1835                 return;
1836         }
1837
1838         if (PFIL_HOOKED(&V_inet_pfil_hook)
1839 #ifdef INET6
1840             || PFIL_HOOKED(&V_inet6_pfil_hook)
1841 #endif
1842             ) {
1843                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
1844                         return;
1845                 if (m == NULL)
1846                         return;
1847         }
1848
1849         bridge_enqueue(sc, ifp, m);
1850 }
1851
1852 /*
1853  * bridge_output:
1854  *
1855  *      Send output from a bridge member interface.  This
1856  *      performs the bridging function for locally originated
1857  *      packets.
1858  *
1859  *      The mbuf has the Ethernet header already attached.  We must
1860  *      enqueue or free the mbuf before returning.
1861  */
1862 static int
1863 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
1864     struct rtentry *rt)
1865 {
1866         struct ether_header *eh;
1867         struct ifnet *dst_if;
1868         struct bridge_softc *sc;
1869         uint16_t vlan;
1870
1871         if (m->m_len < ETHER_HDR_LEN) {
1872                 m = m_pullup(m, ETHER_HDR_LEN);
1873                 if (m == NULL)
1874                         return (0);
1875         }
1876
1877         eh = mtod(m, struct ether_header *);
1878         sc = ifp->if_bridge;
1879         vlan = VLANTAGOF(m);
1880
1881         BRIDGE_LOCK(sc);
1882
1883         /*
1884          * If bridge is down, but the original output interface is up,
1885          * go ahead and send out that interface.  Otherwise, the packet
1886          * is dropped below.
1887          */
1888         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1889                 dst_if = ifp;
1890                 goto sendunicast;
1891         }
1892
1893         /*
1894          * If the packet is a multicast, or we don't know a better way to
1895          * get there, send to all interfaces.
1896          */
1897         if (ETHER_IS_MULTICAST(eh->ether_dhost))
1898                 dst_if = NULL;
1899         else
1900                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
1901         if (dst_if == NULL) {
1902                 struct bridge_iflist *bif;
1903                 struct mbuf *mc;
1904                 int error = 0, used = 0;
1905
1906                 bridge_span(sc, m);
1907
1908                 BRIDGE_LOCK2REF(sc, error);
1909                 if (error) {
1910                         m_freem(m);
1911                         return (0);
1912                 }
1913
1914                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1915                         dst_if = bif->bif_ifp;
1916
1917                         if (dst_if->if_type == IFT_GIF)
1918                                 continue;
1919                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
1920                                 continue;
1921
1922                         /*
1923                          * If this is not the original output interface,
1924                          * and the interface is participating in spanning
1925                          * tree, make sure the port is in a state that
1926                          * allows forwarding.
1927                          */
1928                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
1929                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
1930                                 continue;
1931
1932                         if (LIST_NEXT(bif, bif_next) == NULL) {
1933                                 used = 1;
1934                                 mc = m;
1935                         } else {
1936                                 mc = m_copypacket(m, M_DONTWAIT);
1937                                 if (mc == NULL) {
1938                                         sc->sc_ifp->if_oerrors++;
1939                                         continue;
1940                                 }
1941                         }
1942
1943                         bridge_enqueue(sc, dst_if, mc);
1944                 }
1945                 if (used == 0)
1946                         m_freem(m);
1947                 BRIDGE_UNREF(sc);
1948                 return (0);
1949         }
1950
1951 sendunicast:
1952         /*
1953          * XXX Spanning tree consideration here?
1954          */
1955
1956         bridge_span(sc, m);
1957         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1958                 m_freem(m);
1959                 BRIDGE_UNLOCK(sc);
1960                 return (0);
1961         }
1962
1963         BRIDGE_UNLOCK(sc);
1964         bridge_enqueue(sc, dst_if, m);
1965         return (0);
1966 }
1967
1968 /*
1969  * bridge_start:
1970  *
1971  *      Start output on a bridge.
1972  *
1973  */
1974 static void
1975 bridge_start(struct ifnet *ifp)
1976 {
1977         struct bridge_softc *sc;
1978         struct mbuf *m;
1979         struct ether_header *eh;
1980         struct ifnet *dst_if;
1981
1982         sc = ifp->if_softc;
1983
1984         ifp->if_drv_flags |= IFF_DRV_OACTIVE;
1985         for (;;) {
1986                 IFQ_DEQUEUE(&ifp->if_snd, m);
1987                 if (m == 0)
1988                         break;
1989                 ETHER_BPF_MTAP(ifp, m);
1990
1991                 eh = mtod(m, struct ether_header *);
1992                 dst_if = NULL;
1993
1994                 BRIDGE_LOCK(sc);
1995                 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
1996                         dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1);
1997                 }
1998
1999                 if (dst_if == NULL)
2000                         bridge_broadcast(sc, ifp, m, 0);
2001                 else {
2002                         BRIDGE_UNLOCK(sc);
2003                         bridge_enqueue(sc, dst_if, m);
2004                 }
2005         }
2006         ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
2007 }
2008
2009 /*
2010  * bridge_forward:
2011  *
2012  *      The forwarding function of the bridge.
2013  *
2014  *      NOTE: Releases the lock on return.
2015  */
2016 static void
2017 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2018     struct mbuf *m)
2019 {
2020         struct bridge_iflist *dbif;
2021         struct ifnet *src_if, *dst_if, *ifp;
2022         struct ether_header *eh;
2023         uint16_t vlan;
2024         uint8_t *dst;
2025         int error;
2026
2027         src_if = m->m_pkthdr.rcvif;
2028         ifp = sc->sc_ifp;
2029
2030         ifp->if_ipackets++;
2031         ifp->if_ibytes += m->m_pkthdr.len;
2032         vlan = VLANTAGOF(m);
2033
2034         if ((sbif->bif_flags & IFBIF_STP) &&
2035             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2036                 goto drop;
2037
2038         eh = mtod(m, struct ether_header *);
2039         dst = eh->ether_dhost;
2040
2041         /* If the interface is learning, record the address. */
2042         if (sbif->bif_flags & IFBIF_LEARNING) {
2043                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2044                     sbif, 0, IFBAF_DYNAMIC);
2045                 /*
2046                  * If the interface has addresses limits then deny any source
2047                  * that is not in the cache.
2048                  */
2049                 if (error && sbif->bif_addrmax)
2050                         goto drop;
2051         }
2052
2053         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2054             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2055                 goto drop;
2056
2057         /*
2058          * At this point, the port either doesn't participate
2059          * in spanning tree or it is in the forwarding state.
2060          */
2061
2062         /*
2063          * If the packet is unicast, destined for someone on
2064          * "this" side of the bridge, drop it.
2065          */
2066         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2067                 dst_if = bridge_rtlookup(sc, dst, vlan);
2068                 if (src_if == dst_if)
2069                         goto drop;
2070         } else {
2071                 /*
2072                  * Check if its a reserved multicast address, any address
2073                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2074                  * bridge.
2075                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2076                  */
2077                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2078                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2079                     dst[4] == 0x00 && dst[5] <= 0x0f)
2080                         goto drop;
2081
2082                 /* ...forward it to all interfaces. */
2083                 ifp->if_imcasts++;
2084                 dst_if = NULL;
2085         }
2086
2087         /*
2088          * If we have a destination interface which is a member of our bridge,
2089          * OR this is a unicast packet, push it through the bpf(4) machinery.
2090          * For broadcast or multicast packets, don't bother because it will
2091          * be reinjected into ether_input. We do this before we pass the packets
2092          * through the pfil(9) framework, as it is possible that pfil(9) will
2093          * drop the packet, or possibly modify it, making it difficult to debug
2094          * firewall issues on the bridge.
2095          */
2096         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2097                 ETHER_BPF_MTAP(ifp, m);
2098
2099         /* run the packet filter */
2100         if (PFIL_HOOKED(&V_inet_pfil_hook)
2101 #ifdef INET6
2102             || PFIL_HOOKED(&V_inet6_pfil_hook)
2103 #endif
2104             ) {
2105                 BRIDGE_UNLOCK(sc);
2106                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2107                         return;
2108                 if (m == NULL)
2109                         return;
2110                 BRIDGE_LOCK(sc);
2111         }
2112
2113         if (dst_if == NULL) {
2114                 bridge_broadcast(sc, src_if, m, 1);
2115                 return;
2116         }
2117
2118         /*
2119          * At this point, we're dealing with a unicast frame
2120          * going to a different interface.
2121          */
2122         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2123                 goto drop;
2124
2125         dbif = bridge_lookup_member_if(sc, dst_if);
2126         if (dbif == NULL)
2127                 /* Not a member of the bridge (anymore?) */
2128                 goto drop;
2129
2130         /* Private segments can not talk to each other */
2131         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2132                 goto drop;
2133
2134         if ((dbif->bif_flags & IFBIF_STP) &&
2135             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2136                 goto drop;
2137
2138         BRIDGE_UNLOCK(sc);
2139
2140         if (PFIL_HOOKED(&V_inet_pfil_hook)
2141 #ifdef INET6
2142             || PFIL_HOOKED(&V_inet6_pfil_hook)
2143 #endif
2144             ) {
2145                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2146                         return;
2147                 if (m == NULL)
2148                         return;
2149         }
2150
2151         bridge_enqueue(sc, dst_if, m);
2152         return;
2153
2154 drop:
2155         BRIDGE_UNLOCK(sc);
2156         m_freem(m);
2157 }
2158
2159 /*
2160  * bridge_input:
2161  *
2162  *      Receive input from a member interface.  Queue the packet for
2163  *      bridging if it is not for us.
2164  */
2165 static struct mbuf *
2166 bridge_input(struct ifnet *ifp, struct mbuf *m)
2167 {
2168         struct bridge_softc *sc = ifp->if_bridge;
2169         struct bridge_iflist *bif, *bif2;
2170         struct ifnet *bifp;
2171         struct ether_header *eh;
2172         struct mbuf *mc, *mc2;
2173         uint16_t vlan;
2174         int error;
2175
2176         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2177                 return (m);
2178
2179         bifp = sc->sc_ifp;
2180         vlan = VLANTAGOF(m);
2181
2182         /*
2183          * Implement support for bridge monitoring. If this flag has been
2184          * set on this interface, discard the packet once we push it through
2185          * the bpf(4) machinery, but before we do, increment the byte and
2186          * packet counters associated with this interface.
2187          */
2188         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2189                 m->m_pkthdr.rcvif  = bifp;
2190                 ETHER_BPF_MTAP(bifp, m);
2191                 bifp->if_ipackets++;
2192                 bifp->if_ibytes += m->m_pkthdr.len;
2193                 m_freem(m);
2194                 return (NULL);
2195         }
2196         BRIDGE_LOCK(sc);
2197         bif = bridge_lookup_member_if(sc, ifp);
2198         if (bif == NULL) {
2199                 BRIDGE_UNLOCK(sc);
2200                 return (m);
2201         }
2202
2203         eh = mtod(m, struct ether_header *);
2204
2205         bridge_span(sc, m);
2206
2207         if (m->m_flags & (M_BCAST|M_MCAST)) {
2208                 /* Tap off 802.1D packets; they do not get forwarded. */
2209                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2210                     ETHER_ADDR_LEN) == 0) {
2211                         bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2212                         BRIDGE_UNLOCK(sc);
2213                         return (NULL);
2214                 }
2215
2216                 if ((bif->bif_flags & IFBIF_STP) &&
2217                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2218                         BRIDGE_UNLOCK(sc);
2219                         return (m);
2220                 }
2221
2222                 /*
2223                  * Make a deep copy of the packet and enqueue the copy
2224                  * for bridge processing; return the original packet for
2225                  * local processing.
2226                  */
2227                 mc = m_dup(m, M_DONTWAIT);
2228                 if (mc == NULL) {
2229                         BRIDGE_UNLOCK(sc);
2230                         return (m);
2231                 }
2232
2233                 /* Perform the bridge forwarding function with the copy. */
2234                 bridge_forward(sc, bif, mc);
2235
2236                 /*
2237                  * Reinject the mbuf as arriving on the bridge so we have a
2238                  * chance at claiming multicast packets. We can not loop back
2239                  * here from ether_input as a bridge is never a member of a
2240                  * bridge.
2241                  */
2242                 KASSERT(bifp->if_bridge == NULL,
2243                     ("loop created in bridge_input"));
2244                 mc2 = m_dup(m, M_DONTWAIT);
2245                 if (mc2 != NULL) {
2246                         /* Keep the layer3 header aligned */
2247                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2248                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2249                 }
2250                 if (mc2 != NULL) {
2251                         mc2->m_pkthdr.rcvif = bifp;
2252                         (*bifp->if_input)(bifp, mc2);
2253                 }
2254
2255                 /* Return the original packet for local processing. */
2256                 return (m);
2257         }
2258
2259         if ((bif->bif_flags & IFBIF_STP) &&
2260             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2261                 BRIDGE_UNLOCK(sc);
2262                 return (m);
2263         }
2264
2265 #if (defined(INET) || defined(INET6))
2266 #   define OR_CARP_CHECK_WE_ARE_DST(iface) \
2267         || ((iface)->if_carp \
2268             && (*carp_forus_p)((iface), eh->ether_dhost))
2269 #   define OR_CARP_CHECK_WE_ARE_SRC(iface) \
2270         || ((iface)->if_carp \
2271             && (*carp_forus_p)((iface), eh->ether_shost))
2272 #else
2273 #   define OR_CARP_CHECK_WE_ARE_DST(iface)
2274 #   define OR_CARP_CHECK_WE_ARE_SRC(iface)
2275 #endif
2276
2277 #ifdef INET6
2278 #   define OR_PFIL_HOOKED_INET6 \
2279         || PFIL_HOOKED(&V_inet6_pfil_hook)
2280 #else
2281 #   define OR_PFIL_HOOKED_INET6
2282 #endif
2283
2284 #define GRAB_OUR_PACKETS(iface) \
2285         if ((iface)->if_type == IFT_GIF) \
2286                 continue; \
2287         /* It is destined for us. */ \
2288         if (memcmp(IF_LLADDR((iface)), eh->ether_dhost,  ETHER_ADDR_LEN) == 0 \
2289             OR_CARP_CHECK_WE_ARE_DST((iface))                           \
2290             ) {                                                         \
2291                 if ((iface)->if_type == IFT_BRIDGE) {                   \
2292                         ETHER_BPF_MTAP(iface, m);                       \
2293                         iface->if_ipackets++;                           \
2294                         /* Filter on the physical interface. */         \
2295                         if (pfil_local_phys &&                          \
2296                             (PFIL_HOOKED(&V_inet_pfil_hook)             \
2297                              OR_PFIL_HOOKED_INET6)) {                   \
2298                                 if (bridge_pfil(&m, NULL, ifp,          \
2299                                     PFIL_IN) != 0 || m == NULL) {       \
2300                                         BRIDGE_UNLOCK(sc);              \
2301                                         return (NULL);                  \
2302                                 }                                       \
2303                         }                                               \
2304                 }                                                       \
2305                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2306                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2307                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2308                         if (error && bif->bif_addrmax) {                \
2309                                 BRIDGE_UNLOCK(sc);                      \
2310                                 m_freem(m);                             \
2311                                 return (NULL);                          \
2312                         }                                               \
2313                 }                                                       \
2314                 m->m_pkthdr.rcvif = iface;                              \
2315                 BRIDGE_UNLOCK(sc);                                      \
2316                 return (m);                                             \
2317         }                                                               \
2318                                                                         \
2319         /* We just received a packet that we sent out. */               \
2320         if (memcmp(IF_LLADDR((iface)), eh->ether_shost, ETHER_ADDR_LEN) == 0 \
2321             OR_CARP_CHECK_WE_ARE_SRC((iface))                   \
2322             ) {                                                         \
2323                 BRIDGE_UNLOCK(sc);                                      \
2324                 m_freem(m);                                             \
2325                 return (NULL);                                          \
2326         }
2327
2328         /*
2329          * Unicast.  Make sure it's not for the bridge.
2330          */
2331         do { GRAB_OUR_PACKETS(bifp) } while (0);
2332
2333         /*
2334          * Give a chance for ifp at first priority. This will help when the
2335          * packet comes through the interface like VLAN's with the same MACs
2336          * on several interfaces from the same bridge. This also will save
2337          * some CPU cycles in case the destination interface and the input
2338          * interface (eq ifp) are the same.
2339          */
2340         do { GRAB_OUR_PACKETS(ifp) } while (0);
2341
2342         /* Now check the all bridge members. */
2343         LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2344                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2345         }
2346
2347 #undef OR_CARP_CHECK_WE_ARE_DST
2348 #undef OR_CARP_CHECK_WE_ARE_SRC
2349 #undef OR_PFIL_HOOKED_INET6
2350 #undef GRAB_OUR_PACKETS
2351
2352         /* Perform the bridge forwarding function. */
2353         bridge_forward(sc, bif, m);
2354
2355         return (NULL);
2356 }
2357
2358 /*
2359  * bridge_broadcast:
2360  *
2361  *      Send a frame to all interfaces that are members of
2362  *      the bridge, except for the one on which the packet
2363  *      arrived.
2364  *
2365  *      NOTE: Releases the lock on return.
2366  */
2367 static void
2368 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2369     struct mbuf *m, int runfilt)
2370 {
2371         struct bridge_iflist *dbif, *sbif;
2372         struct mbuf *mc;
2373         struct ifnet *dst_if;
2374         int error = 0, used = 0, i;
2375
2376         sbif = bridge_lookup_member_if(sc, src_if);
2377
2378         BRIDGE_LOCK2REF(sc, error);
2379         if (error) {
2380                 m_freem(m);
2381                 return;
2382         }
2383
2384         /* Filter on the bridge interface before broadcasting */
2385         if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2386 #ifdef INET6
2387             || PFIL_HOOKED(&V_inet6_pfil_hook)
2388 #endif
2389             )) {
2390                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2391                         goto out;
2392                 if (m == NULL)
2393                         goto out;
2394         }
2395
2396         LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2397                 dst_if = dbif->bif_ifp;
2398                 if (dst_if == src_if)
2399                         continue;
2400
2401                 /* Private segments can not talk to each other */
2402                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2403                         continue;
2404
2405                 if ((dbif->bif_flags & IFBIF_STP) &&
2406                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2407                         continue;
2408
2409                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2410                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2411                         continue;
2412
2413                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2414                         continue;
2415
2416                 if (LIST_NEXT(dbif, bif_next) == NULL) {
2417                         mc = m;
2418                         used = 1;
2419                 } else {
2420                         mc = m_dup(m, M_DONTWAIT);
2421                         if (mc == NULL) {
2422                                 sc->sc_ifp->if_oerrors++;
2423                                 continue;
2424                         }
2425                 }
2426
2427                 /*
2428                  * Filter on the output interface. Pass a NULL bridge interface
2429                  * pointer so we do not redundantly filter on the bridge for
2430                  * each interface we broadcast on.
2431                  */
2432                 if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2433 #ifdef INET6
2434                     || PFIL_HOOKED(&V_inet6_pfil_hook)
2435 #endif
2436                     )) {
2437                         if (used == 0) {
2438                                 /* Keep the layer3 header aligned */
2439                                 i = min(mc->m_pkthdr.len, max_protohdr);
2440                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2441                                 if (mc == NULL) {
2442                                         sc->sc_ifp->if_oerrors++;
2443                                         continue;
2444                                 }
2445                         }
2446                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2447                                 continue;
2448                         if (mc == NULL)
2449                                 continue;
2450                 }
2451
2452                 bridge_enqueue(sc, dst_if, mc);
2453         }
2454         if (used == 0)
2455                 m_freem(m);
2456
2457 out:
2458         BRIDGE_UNREF(sc);
2459 }
2460
2461 /*
2462  * bridge_span:
2463  *
2464  *      Duplicate a packet out one or more interfaces that are in span mode,
2465  *      the original mbuf is unmodified.
2466  */
2467 static void
2468 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2469 {
2470         struct bridge_iflist *bif;
2471         struct ifnet *dst_if;
2472         struct mbuf *mc;
2473
2474         if (LIST_EMPTY(&sc->sc_spanlist))
2475                 return;
2476
2477         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2478                 dst_if = bif->bif_ifp;
2479
2480                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2481                         continue;
2482
2483                 mc = m_copypacket(m, M_DONTWAIT);
2484                 if (mc == NULL) {
2485                         sc->sc_ifp->if_oerrors++;
2486                         continue;
2487                 }
2488
2489                 bridge_enqueue(sc, dst_if, mc);
2490         }
2491 }
2492
2493 /*
2494  * bridge_rtupdate:
2495  *
2496  *      Add a bridge routing entry.
2497  */
2498 static int
2499 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2500     struct bridge_iflist *bif, int setflags, uint8_t flags)
2501 {
2502         struct bridge_rtnode *brt;
2503         int error;
2504
2505         BRIDGE_LOCK_ASSERT(sc);
2506
2507         /* Check the source address is valid and not multicast. */
2508         if (ETHER_IS_MULTICAST(dst) ||
2509             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2510              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2511                 return (EINVAL);
2512
2513         /* 802.1p frames map to vlan 1 */
2514         if (vlan == 0)
2515                 vlan = 1;
2516
2517         /*
2518          * A route for this destination might already exist.  If so,
2519          * update it, otherwise create a new one.
2520          */
2521         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2522                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2523                         sc->sc_brtexceeded++;
2524                         return (ENOSPC);
2525                 }
2526                 /* Check per interface address limits (if enabled) */
2527                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2528                         bif->bif_addrexceeded++;
2529                         return (ENOSPC);
2530                 }
2531
2532                 /*
2533                  * Allocate a new bridge forwarding node, and
2534                  * initialize the expiration time and Ethernet
2535                  * address.
2536                  */
2537                 brt = uma_zalloc(bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2538                 if (brt == NULL)
2539                         return (ENOMEM);
2540
2541                 if (bif->bif_flags & IFBIF_STICKY)
2542                         brt->brt_flags = IFBAF_STICKY;
2543                 else
2544                         brt->brt_flags = IFBAF_DYNAMIC;
2545
2546                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2547                 brt->brt_vlan = vlan;
2548
2549                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2550                         uma_zfree(bridge_rtnode_zone, brt);
2551                         return (error);
2552                 }
2553                 brt->brt_dst = bif;
2554                 bif->bif_addrcnt++;
2555         }
2556
2557         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2558             brt->brt_dst != bif) {
2559                 brt->brt_dst->bif_addrcnt--;
2560                 brt->brt_dst = bif;
2561                 brt->brt_dst->bif_addrcnt++;
2562         }
2563
2564         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2565                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2566         if (setflags)
2567                 brt->brt_flags = flags;
2568
2569         return (0);
2570 }
2571
2572 /*
2573  * bridge_rtlookup:
2574  *
2575  *      Lookup the destination interface for an address.
2576  */
2577 static struct ifnet *
2578 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2579 {
2580         struct bridge_rtnode *brt;
2581
2582         BRIDGE_LOCK_ASSERT(sc);
2583
2584         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2585                 return (NULL);
2586
2587         return (brt->brt_ifp);
2588 }
2589
2590 /*
2591  * bridge_rttrim:
2592  *
2593  *      Trim the routine table so that we have a number
2594  *      of routing entries less than or equal to the
2595  *      maximum number.
2596  */
2597 static void
2598 bridge_rttrim(struct bridge_softc *sc)
2599 {
2600         struct bridge_rtnode *brt, *nbrt;
2601
2602         BRIDGE_LOCK_ASSERT(sc);
2603
2604         /* Make sure we actually need to do this. */
2605         if (sc->sc_brtcnt <= sc->sc_brtmax)
2606                 return;
2607
2608         /* Force an aging cycle; this might trim enough addresses. */
2609         bridge_rtage(sc);
2610         if (sc->sc_brtcnt <= sc->sc_brtmax)
2611                 return;
2612
2613         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2614                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2615                         bridge_rtnode_destroy(sc, brt);
2616                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2617                                 return;
2618                 }
2619         }
2620 }
2621
2622 /*
2623  * bridge_timer:
2624  *
2625  *      Aging timer for the bridge.
2626  */
2627 static void
2628 bridge_timer(void *arg)
2629 {
2630         struct bridge_softc *sc = arg;
2631
2632         BRIDGE_LOCK_ASSERT(sc);
2633
2634         bridge_rtage(sc);
2635
2636         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2637                 callout_reset(&sc->sc_brcallout,
2638                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2639 }
2640
2641 /*
2642  * bridge_rtage:
2643  *
2644  *      Perform an aging cycle.
2645  */
2646 static void
2647 bridge_rtage(struct bridge_softc *sc)
2648 {
2649         struct bridge_rtnode *brt, *nbrt;
2650
2651         BRIDGE_LOCK_ASSERT(sc);
2652
2653         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2654                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2655                         if (time_uptime >= brt->brt_expire)
2656                                 bridge_rtnode_destroy(sc, brt);
2657                 }
2658         }
2659 }
2660
2661 /*
2662  * bridge_rtflush:
2663  *
2664  *      Remove all dynamic addresses from the bridge.
2665  */
2666 static void
2667 bridge_rtflush(struct bridge_softc *sc, int full)
2668 {
2669         struct bridge_rtnode *brt, *nbrt;
2670
2671         BRIDGE_LOCK_ASSERT(sc);
2672
2673         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2674                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2675                         bridge_rtnode_destroy(sc, brt);
2676         }
2677 }
2678
2679 /*
2680  * bridge_rtdaddr:
2681  *
2682  *      Remove an address from the table.
2683  */
2684 static int
2685 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2686 {
2687         struct bridge_rtnode *brt;
2688         int found = 0;
2689
2690         BRIDGE_LOCK_ASSERT(sc);
2691
2692         /*
2693          * If vlan is zero then we want to delete for all vlans so the lookup
2694          * may return more than one.
2695          */
2696         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
2697                 bridge_rtnode_destroy(sc, brt);
2698                 found = 1;
2699         }
2700
2701         return (found ? 0 : ENOENT);
2702 }
2703
2704 /*
2705  * bridge_rtdelete:
2706  *
2707  *      Delete routes to a speicifc member interface.
2708  */
2709 static void
2710 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2711 {
2712         struct bridge_rtnode *brt, *nbrt;
2713
2714         BRIDGE_LOCK_ASSERT(sc);
2715
2716         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2717                 if (brt->brt_ifp == ifp && (full ||
2718                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
2719                         bridge_rtnode_destroy(sc, brt);
2720         }
2721 }
2722
2723 /*
2724  * bridge_rtable_init:
2725  *
2726  *      Initialize the route table for this bridge.
2727  */
2728 static int
2729 bridge_rtable_init(struct bridge_softc *sc)
2730 {
2731         int i;
2732
2733         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2734             M_DEVBUF, M_NOWAIT);
2735         if (sc->sc_rthash == NULL)
2736                 return (ENOMEM);
2737
2738         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2739                 LIST_INIT(&sc->sc_rthash[i]);
2740
2741         sc->sc_rthash_key = arc4random();
2742
2743         LIST_INIT(&sc->sc_rtlist);
2744
2745         return (0);
2746 }
2747
2748 /*
2749  * bridge_rtable_fini:
2750  *
2751  *      Deconstruct the route table for this bridge.
2752  */
2753 static void
2754 bridge_rtable_fini(struct bridge_softc *sc)
2755 {
2756
2757         KASSERT(sc->sc_brtcnt == 0,
2758             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
2759         free(sc->sc_rthash, M_DEVBUF);
2760 }
2761
2762 /*
2763  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2764  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2765  */
2766 #define mix(a, b, c)                                                    \
2767 do {                                                                    \
2768         a -= b; a -= c; a ^= (c >> 13);                                 \
2769         b -= c; b -= a; b ^= (a << 8);                                  \
2770         c -= a; c -= b; c ^= (b >> 13);                                 \
2771         a -= b; a -= c; a ^= (c >> 12);                                 \
2772         b -= c; b -= a; b ^= (a << 16);                                 \
2773         c -= a; c -= b; c ^= (b >> 5);                                  \
2774         a -= b; a -= c; a ^= (c >> 3);                                  \
2775         b -= c; b -= a; b ^= (a << 10);                                 \
2776         c -= a; c -= b; c ^= (b >> 15);                                 \
2777 } while (/*CONSTCOND*/0)
2778
2779 static __inline uint32_t
2780 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2781 {
2782         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2783
2784         b += addr[5] << 8;
2785         b += addr[4];
2786         a += addr[3] << 24;
2787         a += addr[2] << 16;
2788         a += addr[1] << 8;
2789         a += addr[0];
2790
2791         mix(a, b, c);
2792
2793         return (c & BRIDGE_RTHASH_MASK);
2794 }
2795
2796 #undef mix
2797
2798 static int
2799 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
2800 {
2801         int i, d;
2802
2803         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
2804                 d = ((int)a[i]) - ((int)b[i]);
2805         }
2806
2807         return (d);
2808 }
2809
2810 /*
2811  * bridge_rtnode_lookup:
2812  *
2813  *      Look up a bridge route node for the specified destination. Compare the
2814  *      vlan id or if zero then just return the first match.
2815  */
2816 static struct bridge_rtnode *
2817 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2818 {
2819         struct bridge_rtnode *brt;
2820         uint32_t hash;
2821         int dir;
2822
2823         BRIDGE_LOCK_ASSERT(sc);
2824
2825         hash = bridge_rthash(sc, addr);
2826         LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
2827                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
2828                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
2829                         return (brt);
2830                 if (dir > 0)
2831                         return (NULL);
2832         }
2833
2834         return (NULL);
2835 }
2836
2837 /*
2838  * bridge_rtnode_insert:
2839  *
2840  *      Insert the specified bridge node into the route table.  We
2841  *      assume the entry is not already in the table.
2842  */
2843 static int
2844 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2845 {
2846         struct bridge_rtnode *lbrt;
2847         uint32_t hash;
2848         int dir;
2849
2850         BRIDGE_LOCK_ASSERT(sc);
2851
2852         hash = bridge_rthash(sc, brt->brt_addr);
2853
2854         lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
2855         if (lbrt == NULL) {
2856                 LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
2857                 goto out;
2858         }
2859
2860         do {
2861                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
2862                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
2863                         return (EEXIST);
2864                 if (dir > 0) {
2865                         LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
2866                         goto out;
2867                 }
2868                 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
2869                         LIST_INSERT_AFTER(lbrt, brt, brt_hash);
2870                         goto out;
2871                 }
2872                 lbrt = LIST_NEXT(lbrt, brt_hash);
2873         } while (lbrt != NULL);
2874
2875 #ifdef DIAGNOSTIC
2876         panic("bridge_rtnode_insert: impossible");
2877 #endif
2878
2879 out:
2880         LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
2881         sc->sc_brtcnt++;
2882
2883         return (0);
2884 }
2885
2886 /*
2887  * bridge_rtnode_destroy:
2888  *
2889  *      Destroy a bridge rtnode.
2890  */
2891 static void
2892 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
2893 {
2894         BRIDGE_LOCK_ASSERT(sc);
2895
2896         LIST_REMOVE(brt, brt_hash);
2897
2898         LIST_REMOVE(brt, brt_list);
2899         sc->sc_brtcnt--;
2900         brt->brt_dst->bif_addrcnt--;
2901         uma_zfree(bridge_rtnode_zone, brt);
2902 }
2903
2904 /*
2905  * bridge_rtable_expire:
2906  *
2907  *      Set the expiry time for all routes on an interface.
2908  */
2909 static void
2910 bridge_rtable_expire(struct ifnet *ifp, int age)
2911 {
2912         struct bridge_softc *sc = ifp->if_bridge;
2913         struct bridge_rtnode *brt;
2914
2915         BRIDGE_LOCK(sc);
2916
2917         /*
2918          * If the age is zero then flush, otherwise set all the expiry times to
2919          * age for the interface
2920          */
2921         if (age == 0)
2922                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
2923         else {
2924                 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
2925                         /* Cap the expiry time to 'age' */
2926                         if (brt->brt_ifp == ifp &&
2927                             brt->brt_expire > time_uptime + age &&
2928                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2929                                 brt->brt_expire = time_uptime + age;
2930                 }
2931         }
2932         BRIDGE_UNLOCK(sc);
2933 }
2934
2935 /*
2936  * bridge_state_change:
2937  *
2938  *      Callback from the bridgestp code when a port changes states.
2939  */
2940 static void
2941 bridge_state_change(struct ifnet *ifp, int state)
2942 {
2943         struct bridge_softc *sc = ifp->if_bridge;
2944         static const char *stpstates[] = {
2945                 "disabled",
2946                 "listening",
2947                 "learning",
2948                 "forwarding",
2949                 "blocking",
2950                 "discarding"
2951         };
2952
2953         if (log_stp)
2954                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
2955                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
2956 }
2957
2958 /*
2959  * Send bridge packets through pfil if they are one of the types pfil can deal
2960  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
2961  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
2962  * that interface.
2963  */
2964 static int
2965 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
2966 {
2967         int snap, error, i, hlen;
2968         struct ether_header *eh1, eh2;
2969         struct ip_fw_args args;
2970         struct ip *ip;
2971         struct llc llc1;
2972         u_int16_t ether_type;
2973
2974         snap = 0;
2975         error = -1;     /* Default error if not error == 0 */
2976
2977 #if 0
2978         /* we may return with the IP fields swapped, ensure its not shared */
2979         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
2980 #endif
2981
2982         if (pfil_bridge == 0 && pfil_member == 0 && pfil_ipfw == 0)
2983                 return (0); /* filtering is disabled */
2984
2985         i = min((*mp)->m_pkthdr.len, max_protohdr);
2986         if ((*mp)->m_len < i) {
2987             *mp = m_pullup(*mp, i);
2988             if (*mp == NULL) {
2989                 printf("%s: m_pullup failed\n", __func__);
2990                 return (-1);
2991             }
2992         }
2993
2994         eh1 = mtod(*mp, struct ether_header *);
2995         ether_type = ntohs(eh1->ether_type);
2996
2997         /*
2998          * Check for SNAP/LLC.
2999          */
3000         if (ether_type < ETHERMTU) {
3001                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3002
3003                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3004                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3005                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3006                     llc2->llc_control == LLC_UI) {
3007                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3008                         snap = 1;
3009                 }
3010         }
3011
3012         /*
3013          * If we're trying to filter bridge traffic, don't look at anything
3014          * other than IP and ARP traffic.  If the filter doesn't understand
3015          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3016          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3017          * but of course we don't have an AppleTalk filter to begin with.
3018          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3019          * ARP traffic.)
3020          */
3021         switch (ether_type) {
3022                 case ETHERTYPE_ARP:
3023                 case ETHERTYPE_REVARP:
3024                         if (pfil_ipfw_arp == 0)
3025                                 return (0); /* Automatically pass */
3026                         break;
3027
3028                 case ETHERTYPE_IP:
3029 #ifdef INET6
3030                 case ETHERTYPE_IPV6:
3031 #endif /* INET6 */
3032                         break;
3033                 default:
3034                         /*
3035                          * Check to see if the user wants to pass non-ip
3036                          * packets, these will not be checked by pfil(9) and
3037                          * passed unconditionally so the default is to drop.
3038                          */
3039                         if (pfil_onlyip)
3040                                 goto bad;
3041         }
3042
3043         /* Strip off the Ethernet header and keep a copy. */
3044         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3045         m_adj(*mp, ETHER_HDR_LEN);
3046
3047         /* Strip off snap header, if present */
3048         if (snap) {
3049                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3050                 m_adj(*mp, sizeof(struct llc));
3051         }
3052
3053         /*
3054          * Check the IP header for alignment and errors
3055          */
3056         if (dir == PFIL_IN) {
3057                 switch (ether_type) {
3058                         case ETHERTYPE_IP:
3059                                 error = bridge_ip_checkbasic(mp);
3060                                 break;
3061 #ifdef INET6
3062                         case ETHERTYPE_IPV6:
3063                                 error = bridge_ip6_checkbasic(mp);
3064                                 break;
3065 #endif /* INET6 */
3066                         default:
3067                                 error = 0;
3068                 }
3069                 if (error)
3070                         goto bad;
3071         }
3072
3073         /* XXX this section is also in if_ethersubr.c */
3074         // XXX PFIL_OUT or DIR_OUT ?
3075         if (V_ip_fw_chk_ptr && pfil_ipfw != 0 &&
3076                         dir == PFIL_OUT && ifp != NULL) {
3077                 struct m_tag *mtag;
3078
3079                 error = -1;
3080                 /* fetch the start point from existing tags, if any */
3081                 mtag = m_tag_locate(*mp, MTAG_IPFW_RULE, 0, NULL);
3082                 if (mtag == NULL) {
3083                         args.rule.slot = 0;
3084                 } else {
3085                         struct ipfw_rule_ref *r;
3086
3087                         /* XXX can we free the tag after use ? */
3088                         mtag->m_tag_id = PACKET_TAG_NONE;
3089                         r = (struct ipfw_rule_ref *)(mtag + 1);
3090                         /* packet already partially processed ? */
3091                         if (r->info & IPFW_ONEPASS)
3092                                 goto ipfwpass;
3093                         args.rule = *r;
3094                 }
3095
3096                 args.m = *mp;
3097                 args.oif = ifp;
3098                 args.next_hop = NULL;
3099                 args.next_hop6 = NULL;
3100                 args.eh = &eh2;
3101                 args.inp = NULL;        /* used by ipfw uid/gid/jail rules */
3102                 i = V_ip_fw_chk_ptr(&args);
3103                 *mp = args.m;
3104
3105                 if (*mp == NULL)
3106                         return (error);
3107
3108                 if (ip_dn_io_ptr && (i == IP_FW_DUMMYNET)) {
3109
3110                         /* put the Ethernet header back on */
3111                         M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
3112                         if (*mp == NULL)
3113                                 return (error);
3114                         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3115
3116                         /*
3117                          * Pass the pkt to dummynet, which consumes it. The
3118                          * packet will return to us via bridge_dummynet().
3119                          */
3120                         args.oif = ifp;
3121                         ip_dn_io_ptr(mp, DIR_FWD | PROTO_IFB, &args);
3122                         return (error);
3123                 }
3124
3125                 if (i != IP_FW_PASS) /* drop */
3126                         goto bad;
3127         }
3128
3129 ipfwpass:
3130         error = 0;
3131
3132         /*
3133          * Run the packet through pfil
3134          */
3135         switch (ether_type) {
3136         case ETHERTYPE_IP:
3137                 /*
3138                  * before calling the firewall, swap fields the same as
3139                  * IP does. here we assume the header is contiguous
3140                  */
3141                 ip = mtod(*mp, struct ip *);
3142
3143                 ip->ip_len = ntohs(ip->ip_len);
3144                 ip->ip_off = ntohs(ip->ip_off);
3145
3146                 /*
3147                  * Run pfil on the member interface and the bridge, both can
3148                  * be skipped by clearing pfil_member or pfil_bridge.
3149                  *
3150                  * Keep the order:
3151                  *   in_if -> bridge_if -> out_if
3152                  */
3153                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3154                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3155                                         dir, NULL);
3156
3157                 if (*mp == NULL || error != 0) /* filter may consume */
3158                         break;
3159
3160                 if (pfil_member && ifp != NULL)
3161                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp,
3162                                         dir, NULL);
3163
3164                 if (*mp == NULL || error != 0) /* filter may consume */
3165                         break;
3166
3167                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3168                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3169                                         dir, NULL);
3170
3171                 if (*mp == NULL || error != 0) /* filter may consume */
3172                         break;
3173
3174                 /* check if we need to fragment the packet */
3175                 if (pfil_member && ifp != NULL && dir == PFIL_OUT) {
3176                         i = (*mp)->m_pkthdr.len;
3177                         if (i > ifp->if_mtu) {
3178                                 error = bridge_fragment(ifp, *mp, &eh2, snap,
3179                                             &llc1);
3180                                 return (error);
3181                         }
3182                 }
3183
3184                 /* Recalculate the ip checksum and restore byte ordering */
3185                 ip = mtod(*mp, struct ip *);
3186                 hlen = ip->ip_hl << 2;
3187                 if (hlen < sizeof(struct ip))
3188                         goto bad;
3189                 if (hlen > (*mp)->m_len) {
3190                         if ((*mp = m_pullup(*mp, hlen)) == 0)
3191                                 goto bad;
3192                         ip = mtod(*mp, struct ip *);
3193                         if (ip == NULL)
3194                                 goto bad;
3195                 }
3196                 ip->ip_len = htons(ip->ip_len);
3197                 ip->ip_off = htons(ip->ip_off);
3198                 ip->ip_sum = 0;
3199                 if (hlen == sizeof(struct ip))
3200                         ip->ip_sum = in_cksum_hdr(ip);
3201                 else
3202                         ip->ip_sum = in_cksum(*mp, hlen);
3203
3204                 break;
3205 #ifdef INET6
3206         case ETHERTYPE_IPV6:
3207                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3208                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3209                                         dir, NULL);
3210
3211                 if (*mp == NULL || error != 0) /* filter may consume */
3212                         break;
3213
3214                 if (pfil_member && ifp != NULL)
3215                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, ifp,
3216                                         dir, NULL);
3217
3218                 if (*mp == NULL || error != 0) /* filter may consume */
3219                         break;
3220
3221                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3222                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3223                                         dir, NULL);
3224                 break;
3225 #endif
3226         default:
3227                 error = 0;
3228                 break;
3229         }
3230
3231         if (*mp == NULL)
3232                 return (error);
3233         if (error != 0)
3234                 goto bad;
3235
3236         error = -1;
3237
3238         /*
3239          * Finally, put everything back the way it was and return
3240          */
3241         if (snap) {
3242                 M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
3243                 if (*mp == NULL)
3244                         return (error);
3245                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3246         }
3247
3248         M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
3249         if (*mp == NULL)
3250                 return (error);
3251         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3252
3253         return (0);
3254
3255 bad:
3256         m_freem(*mp);
3257         *mp = NULL;
3258         return (error);
3259 }
3260
3261 /*
3262  * Perform basic checks on header size since
3263  * pfil assumes ip_input has already processed
3264  * it for it.  Cut-and-pasted from ip_input.c.
3265  * Given how simple the IPv6 version is,
3266  * does the IPv4 version really need to be
3267  * this complicated?
3268  *
3269  * XXX Should we update ipstat here, or not?
3270  * XXX Right now we update ipstat but not
3271  * XXX csum_counter.
3272  */
3273 static int
3274 bridge_ip_checkbasic(struct mbuf **mp)
3275 {
3276         struct mbuf *m = *mp;
3277         struct ip *ip;
3278         int len, hlen;
3279         u_short sum;
3280
3281         if (*mp == NULL)
3282                 return (-1);
3283
3284         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3285                 if ((m = m_copyup(m, sizeof(struct ip),
3286                         (max_linkhdr + 3) & ~3)) == NULL) {
3287                         /* XXXJRT new stat, please */
3288                         KMOD_IPSTAT_INC(ips_toosmall);
3289                         goto bad;
3290                 }
3291         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3292                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3293                         KMOD_IPSTAT_INC(ips_toosmall);
3294                         goto bad;
3295                 }
3296         }
3297         ip = mtod(m, struct ip *);
3298         if (ip == NULL) goto bad;
3299
3300         if (ip->ip_v != IPVERSION) {
3301                 KMOD_IPSTAT_INC(ips_badvers);
3302                 goto bad;
3303         }
3304         hlen = ip->ip_hl << 2;
3305         if (hlen < sizeof(struct ip)) { /* minimum header length */
3306                 KMOD_IPSTAT_INC(ips_badhlen);
3307                 goto bad;
3308         }
3309         if (hlen > m->m_len) {
3310                 if ((m = m_pullup(m, hlen)) == 0) {
3311                         KMOD_IPSTAT_INC(ips_badhlen);
3312                         goto bad;
3313                 }
3314                 ip = mtod(m, struct ip *);
3315                 if (ip == NULL) goto bad;
3316         }
3317
3318         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3319                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3320         } else {
3321                 if (hlen == sizeof(struct ip)) {
3322                         sum = in_cksum_hdr(ip);
3323                 } else {
3324                         sum = in_cksum(m, hlen);
3325                 }
3326         }
3327         if (sum) {
3328                 KMOD_IPSTAT_INC(ips_badsum);
3329                 goto bad;
3330         }
3331
3332         /* Retrieve the packet length. */
3333         len = ntohs(ip->ip_len);
3334
3335         /*
3336          * Check for additional length bogosity
3337          */
3338         if (len < hlen) {
3339                 KMOD_IPSTAT_INC(ips_badlen);
3340                 goto bad;
3341         }
3342
3343         /*
3344          * Check that the amount of data in the buffers
3345          * is as at least much as the IP header would have us expect.
3346          * Drop packet if shorter than we expect.
3347          */
3348         if (m->m_pkthdr.len < len) {
3349                 KMOD_IPSTAT_INC(ips_tooshort);
3350                 goto bad;
3351         }
3352
3353         /* Checks out, proceed */
3354         *mp = m;
3355         return (0);
3356
3357 bad:
3358         *mp = m;
3359         return (-1);
3360 }
3361
3362 #ifdef INET6
3363 /*
3364  * Same as above, but for IPv6.
3365  * Cut-and-pasted from ip6_input.c.
3366  * XXX Should we update ip6stat, or not?
3367  */
3368 static int
3369 bridge_ip6_checkbasic(struct mbuf **mp)
3370 {
3371         struct mbuf *m = *mp;
3372         struct ip6_hdr *ip6;
3373
3374         /*
3375          * If the IPv6 header is not aligned, slurp it up into a new
3376          * mbuf with space for link headers, in the event we forward
3377          * it.  Otherwise, if it is aligned, make sure the entire base
3378          * IPv6 header is in the first mbuf of the chain.
3379          */
3380         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3381                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3382                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3383                             (max_linkhdr + 3) & ~3)) == NULL) {
3384                         /* XXXJRT new stat, please */
3385                         V_ip6stat.ip6s_toosmall++;
3386                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3387                         goto bad;
3388                 }
3389         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3390                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3391                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3392                         V_ip6stat.ip6s_toosmall++;
3393                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3394                         goto bad;
3395                 }
3396         }
3397
3398         ip6 = mtod(m, struct ip6_hdr *);
3399
3400         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3401                 V_ip6stat.ip6s_badvers++;
3402                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3403                 goto bad;
3404         }
3405
3406         /* Checks out, proceed */
3407         *mp = m;
3408         return (0);
3409
3410 bad:
3411         *mp = m;
3412         return (-1);
3413 }
3414 #endif /* INET6 */
3415
3416 /*
3417  * bridge_fragment:
3418  *
3419  *      Return a fragmented mbuf chain.
3420  */
3421 static int
3422 bridge_fragment(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh,
3423     int snap, struct llc *llc)
3424 {
3425         struct mbuf *m0;
3426         struct ip *ip;
3427         int error = -1;
3428
3429         if (m->m_len < sizeof(struct ip) &&
3430             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3431                 goto out;
3432         ip = mtod(m, struct ip *);
3433
3434         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist,
3435                     CSUM_DELAY_IP);
3436         if (error)
3437                 goto out;
3438
3439         /* walk the chain and re-add the Ethernet header */
3440         for (m0 = m; m0; m0 = m0->m_nextpkt) {
3441                 if (error == 0) {
3442                         if (snap) {
3443                                 M_PREPEND(m0, sizeof(struct llc), M_DONTWAIT);
3444                                 if (m0 == NULL) {
3445                                         error = ENOBUFS;
3446                                         continue;
3447                                 }
3448                                 bcopy(llc, mtod(m0, caddr_t),
3449                                     sizeof(struct llc));
3450                         }
3451                         M_PREPEND(m0, ETHER_HDR_LEN, M_DONTWAIT);
3452                         if (m0 == NULL) {
3453                                 error = ENOBUFS;
3454                                 continue;
3455                         }
3456                         bcopy(eh, mtod(m0, caddr_t), ETHER_HDR_LEN);
3457                 } else 
3458                         m_freem(m);
3459         }
3460
3461         if (error == 0)
3462                 KMOD_IPSTAT_INC(ips_fragmented);
3463
3464         return (error);
3465
3466 out:
3467         if (m != NULL)
3468                 m_freem(m);
3469         return (error);
3470 }