kernel - More if_bridge work + misc fixes
[dragonfly.git] / sys / net / bridge / if_bridge.c
1 /*
2  * Copyright 2001 Wasabi Systems, Inc.
3  * All rights reserved.
4  *
5  * Written by Jason R. Thorpe for Wasabi Systems, Inc.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. All advertising materials mentioning features or use of this software
16  *    must display the following acknowledgement:
17  *      This product includes software developed for the NetBSD Project by
18  *      Wasabi Systems, Inc.
19  * 4. The name of Wasabi Systems, Inc. may not be used to endorse
20  *    or promote products derived from this software without specific prior
21  *    written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
25  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
26  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL WASABI SYSTEMS, INC
27  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
30  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
31  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
32  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
33  * POSSIBILITY OF SUCH DAMAGE.
34  */
35
36 /*
37  * Copyright (c) 1999, 2000 Jason L. Wright (jason@thought.net)
38  * All rights reserved.
39  *
40  * Redistribution and use in source and binary forms, with or without
41  * modification, are permitted provided that the following conditions
42  * are met:
43  * 1. Redistributions of source code must retain the above copyright
44  *    notice, this list of conditions and the following disclaimer.
45  * 2. Redistributions in binary form must reproduce the above copyright
46  *    notice, this list of conditions and the following disclaimer in the
47  *    documentation and/or other materials provided with the distribution.
48  * 3. All advertising materials mentioning features or use of this software
49  *    must display the following acknowledgement:
50  *      This product includes software developed by Jason L. Wright
51  * 4. The name of the author may not be used to endorse or promote products
52  *    derived from this software without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
55  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
56  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
57  * DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
58  * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
59  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
60  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
62  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
63  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
64  * POSSIBILITY OF SUCH DAMAGE.
65  *
66  * $OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun Exp $
67  * $NetBSD: if_bridge.c,v 1.31 2005/06/01 19:45:34 jdc Exp $
68  * $FreeBSD: src/sys/net/if_bridge.c,v 1.26 2005/10/13 23:05:55 thompsa Exp $
69  */
70
71 /*
72  * Network interface bridge support.
73  *
74  * TODO:
75  *
76  *      - Currently only supports Ethernet-like interfaces (Ethernet,
77  *        802.11, VLANs on Ethernet, etc.)  Figure out a nice way
78  *        to bridge other types of interfaces (FDDI-FDDI, and maybe
79  *        consider heterogenous bridges).
80  *
81  *
82  * Bridge's route information is duplicated to each CPUs:
83  *
84  *      CPU0          CPU1          CPU2          CPU3
85  * +-----------+ +-----------+ +-----------+ +-----------+
86  * |  rtnode   | |  rtnode   | |  rtnode   | |  rtnode   |
87  * |           | |           | |           | |           |
88  * | dst eaddr | | dst eaddr | | dst eaddr | | dst eaddr |
89  * +-----------+ +-----------+ +-----------+ +-----------+
90  *       |         |                     |         |
91  *       |         |                     |         |
92  *       |         |     +----------+    |         |
93  *       |         |     |  rtinfo  |    |         |
94  *       |         +---->|          |<---+         |
95  *       |               |  flags   |              |
96  *       +-------------->|  timeout |<-------------+
97  *                       |  dst_ifp |
98  *                       +----------+
99  *
100  * We choose to put timeout and dst_ifp into shared part, so updating
101  * them will be cheaper than using message forwarding.  Also there is
102  * not need to use spinlock to protect the updating: timeout and dst_ifp
103  * is not related and specific field's updating order has no importance.
104  * The cache pollution by the share part should not be heavy: in a stable
105  * setup, dst_ifp probably will be not changed in rtnode's life time,
106  * while timeout is refreshed once per second; most of the time, timeout
107  * and dst_ifp are read-only accessed.
108  *
109  *
110  * Bridge route information installation on bridge_input path:
111  *
112  *      CPU0           CPU1         CPU2          CPU3
113  *
114  *                               tcp_thread2
115  *                                    |
116  *                                alloc nmsg
117  *                    snd nmsg        |
118  *                    w/o rtinfo      |
119  *      ifnet0<-----------------------+
120  *        |                           :
121  *    lookup dst                      :
122  *   rtnode exists?(Y)free nmsg       :
123  *        |(N)                        :
124  *        |
125  *  alloc rtinfo
126  *  alloc rtnode
127  * install rtnode
128  *        |
129  *        +---------->ifnet1
130  *        : fwd nmsg    |
131  *        : w/ rtinfo   |
132  *        :             |
133  *        :             |
134  *                 alloc rtnode
135  *               (w/ nmsg's rtinfo)
136  *                install rtnode
137  *                      |
138  *                      +---------->ifnet2
139  *                      : fwd nmsg    |
140  *                      : w/ rtinfo   |
141  *                      :             |
142  *                      :         same as ifnet1
143  *                                    |
144  *                                    +---------->ifnet3
145  *                                    : fwd nmsg    |
146  *                                    : w/ rtinfo   |
147  *                                    :             |
148  *                                    :         same as ifnet1
149  *                                               free nmsg
150  *                                                  :
151  *                                                  :
152  *
153  * The netmsgs forwarded between protocol threads and ifnet threads are
154  * allocated with (M_WAITOK|M_NULLOK), so it will not fail under most
155  * cases (route information is too precious to be not installed :).
156  * Since multiple threads may try to install route information for the
157  * same dst eaddr, we look up route information in ifnet0.  However, this
158  * looking up only need to be performed on ifnet0, which is the start
159  * point of the route information installation process.
160  *
161  *
162  * Bridge route information deleting/flushing:
163  *
164  *  CPU0            CPU1             CPU2             CPU3
165  *
166  * netisr0
167  *   |
168  * find suitable rtnodes,
169  * mark their rtinfo dead
170  *   |
171  *   | domsg <------------------------------------------+
172  *   |                                                  | replymsg
173  *   |                                                  |
174  *   V     fwdmsg           fwdmsg           fwdmsg     |
175  * ifnet0 --------> ifnet1 --------> ifnet2 --------> ifnet3
176  * delete rtnodes   delete rtnodes   delete rtnodes   delete rtnodes
177  * w/ dead rtinfo   w/ dead rtinfo   w/ dead rtinfo   w/ dead rtinfo
178  *                                                    free dead rtinfos
179  *
180  * All deleting/flushing operations are serialized by netisr0, so each
181  * operation only reaps the route information marked dead by itself.
182  *
183  *
184  * Bridge route information adding/deleting/flushing:
185  * Since all operation is serialized by the fixed message flow between
186  * ifnet threads, it is not possible to create corrupted per-cpu route
187  * information.
188  *
189  *
190  *
191  * Percpu member interface list iteration with blocking operation:
192  * Since one bridge could only delete one member interface at a time and
193  * the deleted member interface is not freed after netmsg_service_sync(),
194  * following way is used to make sure that even if the certain member
195  * interface is ripped from the percpu list during the blocking operation,
196  * the iteration still could keep going:
197  *
198  * LIST_FOREACH_MUTABLE(bif, sc->sc_iflists[mycpuid], bif_next, nbif) {
199  *     blocking operation;
200  *     blocking operation;
201  *     ...
202  *     ...
203  *     if (nbif != NULL && !nbif->bif_onlist) {
204  *         KKASSERT(bif->bif_onlist);
205  *         nbif = LIST_NEXT(bif, bif_next);
206  *     }
207  * }
208  *
209  * As mentioned above only one member interface could be unlinked from the
210  * percpu member interface list, so either bif or nbif may be not on the list,
211  * but _not_ both.  To keep the list iteration, we don't care about bif, but
212  * only nbif.  Since removed member interface will only be freed after we
213  * finish our work, it is safe to access any field in an unlinked bif (here
214  * bif_onlist).  If nbif is no longer on the list, then bif must be on the
215  * list, so we change nbif to the next element of bif and keep going.
216  */
217
218 #include "opt_inet.h"
219 #include "opt_inet6.h"
220
221 #include <sys/param.h>
222 #include <sys/mbuf.h>
223 #include <sys/malloc.h>
224 #include <sys/protosw.h>
225 #include <sys/systm.h>
226 #include <sys/time.h>
227 #include <sys/socket.h> /* for net/if.h */
228 #include <sys/sockio.h>
229 #include <sys/ctype.h>  /* string functions */
230 #include <sys/kernel.h>
231 #include <sys/random.h>
232 #include <sys/sysctl.h>
233 #include <sys/module.h>
234 #include <sys/proc.h>
235 #include <sys/priv.h>
236 #include <sys/lock.h>
237 #include <sys/thread.h>
238 #include <sys/thread2.h>
239 #include <sys/mpipe.h>
240
241 #include <net/bpf.h>
242 #include <net/if.h>
243 #include <net/if_dl.h>
244 #include <net/if_types.h>
245 #include <net/if_var.h>
246 #include <net/pfil.h>
247 #include <net/ifq_var.h>
248 #include <net/if_clone.h>
249
250 #include <netinet/in.h> /* for struct arpcom */
251 #include <netinet/in_systm.h>
252 #include <netinet/in_var.h>
253 #include <netinet/ip.h>
254 #include <netinet/ip_var.h>
255 #ifdef INET6
256 #include <netinet/ip6.h>
257 #include <netinet6/ip6_var.h>
258 #endif
259 #include <netinet/if_ether.h> /* for struct arpcom */
260 #include <net/bridge/if_bridgevar.h>
261 #include <net/if_llc.h>
262 #include <net/netmsg2.h>
263
264 #include <net/route.h>
265 #include <sys/in_cksum.h>
266
267 /*
268  * Size of the route hash table.  Must be a power of two.
269  */
270 #ifndef BRIDGE_RTHASH_SIZE
271 #define BRIDGE_RTHASH_SIZE              1024
272 #endif
273
274 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
275
276 /*
277  * Maximum number of addresses to cache.
278  */
279 #ifndef BRIDGE_RTABLE_MAX
280 #define BRIDGE_RTABLE_MAX               100
281 #endif
282
283 /*
284  * Spanning tree defaults.
285  */
286 #define BSTP_DEFAULT_MAX_AGE            (20 * 256)
287 #define BSTP_DEFAULT_HELLO_TIME         (2 * 256)
288 #define BSTP_DEFAULT_FORWARD_DELAY      (15 * 256)
289 #define BSTP_DEFAULT_HOLD_TIME          (1 * 256)
290 #define BSTP_DEFAULT_BRIDGE_PRIORITY    0x8000
291 #define BSTP_DEFAULT_PORT_PRIORITY      0x80
292 #define BSTP_DEFAULT_PATH_COST          55
293
294 /*
295  * Timeout (in seconds) for entries learned dynamically.
296  */
297 #ifndef BRIDGE_RTABLE_TIMEOUT
298 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
299 #endif
300
301 /*
302  * Number of seconds between walks of the route list.
303  */
304 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
305 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
306 #endif
307
308 /*
309  * List of capabilities to mask on the member interface.
310  */
311 #define BRIDGE_IFCAPS_MASK              IFCAP_TXCSUM
312
313 typedef int     (*bridge_ctl_t)(struct bridge_softc *, void *);
314
315 struct netmsg_brctl {
316         struct netmsg_base      base;
317         bridge_ctl_t            bc_func;
318         struct bridge_softc     *bc_sc;
319         void                    *bc_arg;
320 };
321
322 struct netmsg_brsaddr {
323         struct netmsg_base      base;
324         struct bridge_softc     *br_softc;
325         struct ifnet            *br_dst_if;
326         struct bridge_rtinfo    *br_rtinfo;
327         int                     br_setflags;
328         uint8_t                 br_dst[ETHER_ADDR_LEN];
329         uint8_t                 br_flags;
330 };
331
332 struct netmsg_braddbif {
333         struct netmsg_base      base;
334         struct bridge_softc     *br_softc;
335         struct bridge_ifinfo    *br_bif_info;
336         struct ifnet            *br_bif_ifp;
337 };
338
339 struct netmsg_brdelbif {
340         struct netmsg_base      base;
341         struct bridge_softc     *br_softc;
342         struct bridge_ifinfo    *br_bif_info;
343         struct bridge_iflist_head *br_bif_list;
344 };
345
346 struct netmsg_brsflags {
347         struct netmsg_base      base;
348         struct bridge_softc     *br_softc;
349         struct bridge_ifinfo    *br_bif_info;
350         uint32_t                br_bif_flags;
351 };
352
353 eventhandler_tag        bridge_detach_cookie = NULL;
354
355 extern  struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
356 extern  int (*bridge_output_p)(struct ifnet *, struct mbuf *);
357 extern  void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
358
359 static int      bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
360
361 static int      bridge_clone_create(struct if_clone *, int, caddr_t);
362 static int      bridge_clone_destroy(struct ifnet *);
363
364 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t, struct ucred *);
365 static void     bridge_mutecaps(struct bridge_ifinfo *, struct ifnet *, int);
366 static void     bridge_ifdetach(void *, struct ifnet *);
367 static void     bridge_init(void *);
368 static void     bridge_stop(struct ifnet *);
369 static void     bridge_start(struct ifnet *);
370 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
371 static int      bridge_output(struct ifnet *, struct mbuf *);
372
373 static void     bridge_forward(struct bridge_softc *, struct mbuf *m);
374
375 static void     bridge_timer_handler(netmsg_t);
376 static void     bridge_timer(void *);
377
378 static void     bridge_start_bcast(struct bridge_softc *, struct mbuf *);
379 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
380                     struct mbuf *);
381 static void     bridge_span(struct bridge_softc *, struct mbuf *);
382
383 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
384                     struct ifnet *, uint8_t);
385 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *);
386 static void     bridge_rtreap(struct bridge_softc *);
387 static void     bridge_rtreap_async(struct bridge_softc *);
388 static void     bridge_rttrim(struct bridge_softc *);
389 static int      bridge_rtage_finddead(struct bridge_softc *);
390 static void     bridge_rtage(struct bridge_softc *);
391 static void     bridge_rtflush(struct bridge_softc *, int);
392 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *);
393 static int      bridge_rtsaddr(struct bridge_softc *, const uint8_t *,
394                     struct ifnet *, uint8_t);
395 static void     bridge_rtmsg_sync(struct bridge_softc *sc);
396 static void     bridge_rtreap_handler(netmsg_t);
397 static void     bridge_rtinstall_handler(netmsg_t);
398 static int      bridge_rtinstall_oncpu(struct bridge_softc *, const uint8_t *,
399                     struct ifnet *, int, uint8_t, struct bridge_rtinfo **);
400
401 static void     bridge_rtable_init(struct bridge_softc *);
402 static void     bridge_rtable_fini(struct bridge_softc *);
403
404 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
405 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
406                     const uint8_t *);
407 static void     bridge_rtnode_insert(struct bridge_softc *,
408                     struct bridge_rtnode *);
409 static void     bridge_rtnode_destroy(struct bridge_softc *,
410                     struct bridge_rtnode *);
411
412 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
413                     const char *name);
414 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
415                     struct ifnet *ifp);
416 static struct bridge_iflist *bridge_lookup_member_ifinfo(struct bridge_softc *,
417                     struct bridge_ifinfo *);
418 static void     bridge_delete_member(struct bridge_softc *,
419                     struct bridge_iflist *, int);
420 static void     bridge_delete_span(struct bridge_softc *,
421                     struct bridge_iflist *);
422
423 static int      bridge_control(struct bridge_softc *, u_long,
424                                bridge_ctl_t, void *);
425 static int      bridge_ioctl_init(struct bridge_softc *, void *);
426 static int      bridge_ioctl_stop(struct bridge_softc *, void *);
427 static int      bridge_ioctl_add(struct bridge_softc *, void *);
428 static int      bridge_ioctl_del(struct bridge_softc *, void *);
429 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
430 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
431 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
432 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
433 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
434 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
435 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
436 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
437 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
438 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
439 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
440 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
441 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
442 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
443 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
444 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
445 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
446 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
447 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
448 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
449 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
450 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
451 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
452 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
453                     int);
454 static int      bridge_ip_checkbasic(struct mbuf **mp);
455 #ifdef INET6
456 static int      bridge_ip6_checkbasic(struct mbuf **mp);
457 #endif /* INET6 */
458 static int      bridge_fragment(struct ifnet *, struct mbuf *,
459                     struct ether_header *, int, struct llc *);
460 static void     bridge_enqueue_handler(netmsg_t);
461 static void     bridge_handoff(struct ifnet *, struct mbuf *, int);
462
463 static void     bridge_del_bif_handler(netmsg_t);
464 static void     bridge_add_bif_handler(netmsg_t);
465 static void     bridge_set_bifflags_handler(netmsg_t);
466 static void     bridge_del_bif(struct bridge_softc *, struct bridge_ifinfo *,
467                     struct bridge_iflist_head *);
468 static void     bridge_add_bif(struct bridge_softc *, struct bridge_ifinfo *,
469                     struct ifnet *);
470 static void     bridge_set_bifflags(struct bridge_softc *,
471                     struct bridge_ifinfo *, uint32_t);
472
473 SYSCTL_DECL(_net_link);
474 SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge");
475
476 static int pfil_onlyip = 1; /* only pass IP[46] packets when pfil is enabled */
477 static int pfil_bridge = 1; /* run pfil hooks on the bridge interface */
478 static int pfil_member = 1; /* run pfil hooks on the member interface */
479 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip, CTLFLAG_RW,
480     &pfil_onlyip, 0, "Only pass IP packets when pfil is enabled");
481 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge, CTLFLAG_RW,
482     &pfil_bridge, 0, "Packet filter on the bridge interface");
483 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member, CTLFLAG_RW,
484     &pfil_member, 0, "Packet filter on the member interface");
485
486 struct bridge_control_arg {
487         union {
488                 struct ifbreq ifbreq;
489                 struct ifbifconf ifbifconf;
490                 struct ifbareq ifbareq;
491                 struct ifbaconf ifbaconf;
492                 struct ifbrparam ifbrparam;
493         } bca_u;
494         int     bca_len;
495         void    *bca_uptr;
496         void    *bca_kptr;
497 };
498
499 struct bridge_control {
500         bridge_ctl_t    bc_func;
501         int             bc_argsize;
502         int             bc_flags;
503 };
504
505 #define BC_F_COPYIN             0x01    /* copy arguments in */
506 #define BC_F_COPYOUT            0x02    /* copy arguments out */
507 #define BC_F_SUSER              0x04    /* do super-user check */
508
509 const struct bridge_control bridge_control_table[] = {
510         { bridge_ioctl_add,             sizeof(struct ifbreq),
511           BC_F_COPYIN|BC_F_SUSER },
512         { bridge_ioctl_del,             sizeof(struct ifbreq),
513           BC_F_COPYIN|BC_F_SUSER },
514
515         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
516           BC_F_COPYIN|BC_F_COPYOUT },
517         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
518           BC_F_COPYIN|BC_F_SUSER },
519
520         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
521           BC_F_COPYIN|BC_F_SUSER },
522         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
523           BC_F_COPYOUT },
524
525         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
526           BC_F_COPYIN|BC_F_COPYOUT },
527         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
528           BC_F_COPYIN|BC_F_COPYOUT },
529
530         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
531           BC_F_COPYIN|BC_F_SUSER },
532
533         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
534           BC_F_COPYIN|BC_F_SUSER },
535         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
536           BC_F_COPYOUT },
537
538         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
539           BC_F_COPYIN|BC_F_SUSER },
540
541         { bridge_ioctl_flush,           sizeof(struct ifbreq),
542           BC_F_COPYIN|BC_F_SUSER },
543
544         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
545           BC_F_COPYOUT },
546         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
547           BC_F_COPYIN|BC_F_SUSER },
548
549         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
550           BC_F_COPYOUT },
551         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
552           BC_F_COPYIN|BC_F_SUSER },
553
554         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
555           BC_F_COPYOUT },
556         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
557           BC_F_COPYIN|BC_F_SUSER },
558
559         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
560           BC_F_COPYOUT },
561         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
562           BC_F_COPYIN|BC_F_SUSER },
563
564         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
565           BC_F_COPYIN|BC_F_SUSER },
566
567         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
568           BC_F_COPYIN|BC_F_SUSER },
569
570         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
571           BC_F_COPYIN|BC_F_SUSER },
572         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
573           BC_F_COPYIN|BC_F_SUSER },
574 };
575 static const int bridge_control_table_size = NELEM(bridge_control_table);
576
577 LIST_HEAD(, bridge_softc) bridge_list;
578
579 struct if_clone bridge_cloner = IF_CLONE_INITIALIZER("bridge",
580                                 bridge_clone_create,
581                                 bridge_clone_destroy, 0, IF_MAXUNIT);
582
583 static int
584 bridge_modevent(module_t mod, int type, void *data)
585 {
586         switch (type) {
587         case MOD_LOAD:
588                 LIST_INIT(&bridge_list);
589                 if_clone_attach(&bridge_cloner);
590                 bridge_input_p = bridge_input;
591                 bridge_output_p = bridge_output;
592                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
593                     ifnet_detach_event, bridge_ifdetach, NULL,
594                     EVENTHANDLER_PRI_ANY);
595 #if notyet
596                 bstp_linkstate_p = bstp_linkstate;
597 #endif
598                 break;
599         case MOD_UNLOAD:
600                 if (!LIST_EMPTY(&bridge_list))
601                         return (EBUSY);
602                 EVENTHANDLER_DEREGISTER(ifnet_detach_event,
603                     bridge_detach_cookie);
604                 if_clone_detach(&bridge_cloner);
605                 bridge_input_p = NULL;
606                 bridge_output_p = NULL;
607 #if notyet
608                 bstp_linkstate_p = NULL;
609 #endif
610                 break;
611         default:
612                 return (EOPNOTSUPP);
613         }
614         return (0);
615 }
616
617 static moduledata_t bridge_mod = {
618         "if_bridge",
619         bridge_modevent,
620         0
621 };
622
623 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
624
625
626 /*
627  * bridge_clone_create:
628  *
629  *      Create a new bridge instance.
630  */
631 static int
632 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t param __unused)
633 {
634         struct bridge_softc *sc;
635         struct ifnet *ifp;
636         u_char eaddr[6];
637         int cpu, rnd;
638
639         sc = kmalloc(sizeof(*sc), M_DEVBUF, M_WAITOK | M_ZERO);
640         ifp = sc->sc_ifp = &sc->sc_if;
641
642         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
643         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
644         sc->sc_bridge_max_age = BSTP_DEFAULT_MAX_AGE;
645         sc->sc_bridge_hello_time = BSTP_DEFAULT_HELLO_TIME;
646         sc->sc_bridge_forward_delay = BSTP_DEFAULT_FORWARD_DELAY;
647         sc->sc_bridge_priority = BSTP_DEFAULT_BRIDGE_PRIORITY;
648         sc->sc_hold_time = BSTP_DEFAULT_HOLD_TIME;
649
650         /* Initialize our routing table. */
651         bridge_rtable_init(sc);
652
653         callout_init(&sc->sc_brcallout);
654         netmsg_init(&sc->sc_brtimemsg, NULL, &netisr_adone_rport,
655                     MSGF_DROPABLE, bridge_timer_handler);
656         sc->sc_brtimemsg.lmsg.u.ms_resultp = sc;
657
658         callout_init(&sc->sc_bstpcallout);
659         netmsg_init(&sc->sc_bstptimemsg, NULL, &netisr_adone_rport,
660                     MSGF_DROPABLE, bstp_tick_handler);
661         sc->sc_bstptimemsg.lmsg.u.ms_resultp = sc;
662
663         /* Initialize per-cpu member iface lists */
664         sc->sc_iflists = kmalloc(sizeof(*sc->sc_iflists) * ncpus,
665                                  M_DEVBUF, M_WAITOK);
666         for (cpu = 0; cpu < ncpus; ++cpu)
667                 LIST_INIT(&sc->sc_iflists[cpu]);
668
669         LIST_INIT(&sc->sc_spanlist);
670
671         ifp->if_softc = sc;
672         if_initname(ifp, ifc->ifc_name, unit);
673         ifp->if_mtu = ETHERMTU;
674         ifp->if_flags = IFF_BROADCAST | IFF_MULTICAST;
675         ifp->if_ioctl = bridge_ioctl;
676         ifp->if_start = bridge_start;
677         ifp->if_init = bridge_init;
678         ifp->if_type = IFT_BRIDGE;
679         ifq_set_maxlen(&ifp->if_snd, ifqmaxlen);
680         ifq_set_ready(&ifp->if_snd);
681         ifp->if_hdrlen = ETHER_HDR_LEN;
682
683         /*
684          * Generate a random ethernet address and use the private AC:DE:48
685          * OUI code.
686          */
687         rnd = karc4random();
688         bcopy(&rnd, &eaddr[0], 4); /* ETHER_ADDR_LEN == 6 */
689         rnd = karc4random();
690         bcopy(&rnd, &eaddr[2], 4); /* ETHER_ADDR_LEN == 6 */
691
692         eaddr[0] &= ~1; /* clear multicast bit */
693         eaddr[0] |= 2;  /* set the LAA bit */
694
695         ether_ifattach(ifp, eaddr, NULL);
696         /* Now undo some of the damage... */
697         ifp->if_baudrate = 0;
698         ifp->if_type = IFT_BRIDGE;
699
700         crit_enter();   /* XXX MP */
701         LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
702         crit_exit();
703
704         return (0);
705 }
706
707 static void
708 bridge_delete_dispatch(netmsg_t msg)
709 {
710         struct bridge_softc *sc = msg->lmsg.u.ms_resultp;
711         struct ifnet *bifp = sc->sc_ifp;
712         struct bridge_iflist *bif;
713
714         ifnet_serialize_all(bifp);
715
716         while ((bif = LIST_FIRST(&sc->sc_iflists[mycpuid])) != NULL)
717                 bridge_delete_member(sc, bif, 0);
718
719         while ((bif = LIST_FIRST(&sc->sc_spanlist)) != NULL)
720                 bridge_delete_span(sc, bif);
721
722         ifnet_deserialize_all(bifp);
723
724         lwkt_replymsg(&msg->lmsg, 0);
725 }
726
727 /*
728  * bridge_clone_destroy:
729  *
730  *      Destroy a bridge instance.
731  */
732 static int
733 bridge_clone_destroy(struct ifnet *ifp)
734 {
735         struct bridge_softc *sc = ifp->if_softc;
736         struct netmsg_base msg;
737
738         ifnet_serialize_all(ifp);
739
740         bridge_stop(ifp);
741         ifp->if_flags &= ~IFF_UP;
742
743         ifnet_deserialize_all(ifp);
744
745         netmsg_init(&msg, NULL, &curthread->td_msgport,
746                     0, bridge_delete_dispatch);
747         msg.lmsg.u.ms_resultp = sc;
748         lwkt_domsg(BRIDGE_CFGPORT, &msg.lmsg, 0);
749
750         crit_enter();   /* XXX MP */
751         LIST_REMOVE(sc, sc_list);
752         crit_exit();
753
754         ether_ifdetach(ifp);
755
756         /* Tear down the routing table. */
757         bridge_rtable_fini(sc);
758
759         /* Free per-cpu member iface lists */
760         kfree(sc->sc_iflists, M_DEVBUF);
761
762         kfree(sc, M_DEVBUF);
763
764         return 0;
765 }
766
767 /*
768  * bridge_ioctl:
769  *
770  *      Handle a control request from the operator.
771  */
772 static int
773 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data, struct ucred *cr)
774 {
775         struct bridge_softc *sc = ifp->if_softc;
776         struct bridge_control_arg args;
777         struct ifdrv *ifd = (struct ifdrv *) data;
778         const struct bridge_control *bc;
779         int error = 0;
780
781         ASSERT_IFNET_SERIALIZED_ALL(ifp);
782
783         switch (cmd) {
784         case SIOCADDMULTI:
785         case SIOCDELMULTI:
786                 break;
787
788         case SIOCGDRVSPEC:
789         case SIOCSDRVSPEC:
790                 if (ifd->ifd_cmd >= bridge_control_table_size) {
791                         error = EINVAL;
792                         break;
793                 }
794                 bc = &bridge_control_table[ifd->ifd_cmd];
795
796                 if (cmd == SIOCGDRVSPEC &&
797                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
798                         error = EINVAL;
799                         break;
800                 } else if (cmd == SIOCSDRVSPEC &&
801                            (bc->bc_flags & BC_F_COPYOUT)) {
802                         error = EINVAL;
803                         break;
804                 }
805
806                 if (bc->bc_flags & BC_F_SUSER) {
807                         error = priv_check_cred(cr, PRIV_ROOT, NULL_CRED_OKAY);
808                         if (error)
809                                 break;
810                 }
811
812                 if (ifd->ifd_len != bc->bc_argsize ||
813                     ifd->ifd_len > sizeof(args.bca_u)) {
814                         error = EINVAL;
815                         break;
816                 }
817
818                 memset(&args, 0, sizeof(args));
819                 if (bc->bc_flags & BC_F_COPYIN) {
820                         error = copyin(ifd->ifd_data, &args.bca_u,
821                                        ifd->ifd_len);
822                         if (error)
823                                 break;
824                 }
825
826                 error = bridge_control(sc, cmd, bc->bc_func, &args);
827                 if (error) {
828                         KKASSERT(args.bca_len == 0 && args.bca_kptr == NULL);
829                         break;
830                 }
831
832                 if (bc->bc_flags & BC_F_COPYOUT) {
833                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
834                         if (args.bca_len != 0) {
835                                 KKASSERT(args.bca_kptr != NULL);
836                                 if (!error) {
837                                         error = copyout(args.bca_kptr,
838                                                 args.bca_uptr, args.bca_len);
839                                 }
840                                 kfree(args.bca_kptr, M_TEMP);
841                         } else {
842                                 KKASSERT(args.bca_kptr == NULL);
843                         }
844                 } else {
845                         KKASSERT(args.bca_len == 0 && args.bca_kptr == NULL);
846                 }
847                 break;
848
849         case SIOCSIFFLAGS:
850                 if (!(ifp->if_flags & IFF_UP) &&
851                     (ifp->if_flags & IFF_RUNNING)) {
852                         /*
853                          * If interface is marked down and it is running,
854                          * then stop it.
855                          */
856                         bridge_stop(ifp);
857                 } else if ((ifp->if_flags & IFF_UP) &&
858                     !(ifp->if_flags & IFF_RUNNING)) {
859                         /*
860                          * If interface is marked up and it is stopped, then
861                          * start it.
862                          */
863                         ifp->if_init(sc);
864                 }
865                 break;
866
867         case SIOCSIFMTU:
868                 /* Do not allow the MTU to be changed on the bridge */
869                 error = EINVAL;
870                 break;
871
872         default:
873                 error = ether_ioctl(ifp, cmd, data);
874                 break;
875         }
876         return (error);
877 }
878
879 /*
880  * bridge_mutecaps:
881  *
882  *      Clear or restore unwanted capabilities on the member interface
883  */
884 static void
885 bridge_mutecaps(struct bridge_ifinfo *bif_info, struct ifnet *ifp, int mute)
886 {
887         struct ifreq ifr;
888         int error;
889
890         if (ifp->if_ioctl == NULL)
891                 return;
892
893         bzero(&ifr, sizeof(ifr));
894         ifr.ifr_reqcap = ifp->if_capenable;
895
896         if (mute) {
897                 /* mask off and save capabilities */
898                 bif_info->bifi_mutecap = ifr.ifr_reqcap & BRIDGE_IFCAPS_MASK;
899                 if (bif_info->bifi_mutecap != 0)
900                         ifr.ifr_reqcap &= ~BRIDGE_IFCAPS_MASK;
901         } else {
902                 /* restore muted capabilities */
903                 ifr.ifr_reqcap |= bif_info->bifi_mutecap;
904         }
905
906         if (bif_info->bifi_mutecap != 0) {
907                 ifnet_serialize_all(ifp);
908                 error = ifp->if_ioctl(ifp, SIOCSIFCAP, (caddr_t)&ifr, NULL);
909                 ifnet_deserialize_all(ifp);
910         }
911 }
912
913 /*
914  * bridge_lookup_member:
915  *
916  *      Lookup a bridge member interface.
917  */
918 static struct bridge_iflist *
919 bridge_lookup_member(struct bridge_softc *sc, const char *name)
920 {
921         struct bridge_iflist *bif;
922
923         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
924                 if (strcmp(bif->bif_ifp->if_xname, name) == 0)
925                         return (bif);
926         }
927         return (NULL);
928 }
929
930 /*
931  * bridge_lookup_member_if:
932  *
933  *      Lookup a bridge member interface by ifnet*.
934  */
935 static struct bridge_iflist *
936 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
937 {
938         struct bridge_iflist *bif;
939
940         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
941                 if (bif->bif_ifp == member_ifp)
942                         return (bif);
943         }
944         return (NULL);
945 }
946
947 /*
948  * bridge_lookup_member_ifinfo:
949  *
950  *      Lookup a bridge member interface by bridge_ifinfo.
951  */
952 static struct bridge_iflist *
953 bridge_lookup_member_ifinfo(struct bridge_softc *sc,
954                             struct bridge_ifinfo *bif_info)
955 {
956         struct bridge_iflist *bif;
957
958         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
959                 if (bif->bif_info == bif_info)
960                         return (bif);
961         }
962         return (NULL);
963 }
964
965 /*
966  * bridge_delete_member:
967  *
968  *      Delete the specified member interface.
969  */
970 static void
971 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
972     int gone)
973 {
974         struct ifnet *ifs = bif->bif_ifp;
975         struct ifnet *bifp = sc->sc_ifp;
976         struct bridge_ifinfo *bif_info = bif->bif_info;
977         struct bridge_iflist_head saved_bifs;
978
979         ASSERT_IFNET_SERIALIZED_ALL(bifp);
980         KKASSERT(bif_info != NULL);
981
982         ifs->if_bridge = NULL;
983
984         /*
985          * Release bridge interface's serializer:
986          * - To avoid possible dead lock.
987          * - Various sync operation will block the current thread.
988          */
989         ifnet_deserialize_all(bifp);
990
991         if (!gone) {
992                 switch (ifs->if_type) {
993                 case IFT_ETHER:
994                 case IFT_L2VLAN:
995                         /*
996                          * Take the interface out of promiscuous mode.
997                          */
998                         ifpromisc(ifs, 0);
999                         bridge_mutecaps(bif_info, ifs, 0);
1000                         break;
1001
1002                 case IFT_GIF:
1003                         break;
1004
1005                 default:
1006                         panic("bridge_delete_member: impossible");
1007                         break;
1008                 }
1009         }
1010
1011         /*
1012          * Remove bifs from percpu linked list.
1013          *
1014          * Removed bifs are not freed immediately, instead,
1015          * they are saved in saved_bifs.  They will be freed
1016          * after we make sure that no one is accessing them,
1017          * i.e. after following netmsg_service_sync()
1018          */
1019         LIST_INIT(&saved_bifs);
1020         bridge_del_bif(sc, bif_info, &saved_bifs);
1021
1022         /*
1023          * Make sure that all protocol threads:
1024          * o  see 'ifs' if_bridge is changed
1025          * o  know that bif is removed from the percpu linked list
1026          */
1027         netmsg_service_sync();
1028
1029         /*
1030          * Free the removed bifs
1031          */
1032         KKASSERT(!LIST_EMPTY(&saved_bifs));
1033         while ((bif = LIST_FIRST(&saved_bifs)) != NULL) {
1034                 LIST_REMOVE(bif, bif_next);
1035                 kfree(bif, M_DEVBUF);
1036         }
1037
1038         /* See the comment in bridge_ioctl_stop() */
1039         bridge_rtmsg_sync(sc);
1040         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL | IFBF_FLUSHSYNC);
1041
1042         ifnet_serialize_all(bifp);
1043
1044         if (bifp->if_flags & IFF_RUNNING)
1045                 bstp_initialization(sc);
1046
1047         /*
1048          * Free the bif_info after bstp_initialization(), so that
1049          * bridge_softc.sc_root_port will not reference a dangling
1050          * pointer.
1051          */
1052         kfree(bif_info, M_DEVBUF);
1053 }
1054
1055 /*
1056  * bridge_delete_span:
1057  *
1058  *      Delete the specified span interface.
1059  */
1060 static void
1061 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1062 {
1063         KASSERT(bif->bif_ifp->if_bridge == NULL,
1064             ("%s: not a span interface", __func__));
1065
1066         LIST_REMOVE(bif, bif_next);
1067         kfree(bif, M_DEVBUF);
1068 }
1069
1070 static int
1071 bridge_ioctl_init(struct bridge_softc *sc, void *arg __unused)
1072 {
1073         struct ifnet *ifp = sc->sc_ifp;
1074
1075         if (ifp->if_flags & IFF_RUNNING)
1076                 return 0;
1077
1078         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1079             bridge_timer, sc);
1080
1081         ifp->if_flags |= IFF_RUNNING;
1082         bstp_initialization(sc);
1083         return 0;
1084 }
1085
1086 static int
1087 bridge_ioctl_stop(struct bridge_softc *sc, void *arg __unused)
1088 {
1089         struct ifnet *ifp = sc->sc_ifp;
1090         struct lwkt_msg *lmsg;
1091
1092         if ((ifp->if_flags & IFF_RUNNING) == 0)
1093                 return 0;
1094
1095         callout_stop(&sc->sc_brcallout);
1096
1097         crit_enter();
1098         lmsg = &sc->sc_brtimemsg.lmsg;
1099         if ((lmsg->ms_flags & MSGF_DONE) == 0) {
1100                 /* Pending to be processed; drop it */
1101                 lwkt_dropmsg(lmsg);
1102         }
1103         crit_exit();
1104
1105         bstp_stop(sc);
1106
1107         ifp->if_flags &= ~IFF_RUNNING;
1108
1109         ifnet_deserialize_all(ifp);
1110
1111         /* Let everyone know that we are stopped */
1112         netmsg_service_sync();
1113
1114         /*
1115          * Sync ifnetX msgports in the order we forward rtnode
1116          * installation message.  This is used to make sure that
1117          * all rtnode installation messages sent by bridge_rtupdate()
1118          * during above netmsg_service_sync() are flushed.
1119          */
1120         bridge_rtmsg_sync(sc);
1121         bridge_rtflush(sc, IFBF_FLUSHDYN | IFBF_FLUSHSYNC);
1122
1123         ifnet_serialize_all(ifp);
1124         return 0;
1125 }
1126
1127 static int
1128 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1129 {
1130         struct ifbreq *req = arg;
1131         struct bridge_iflist *bif;
1132         struct bridge_ifinfo *bif_info;
1133         struct ifnet *ifs, *bifp;
1134         int error = 0;
1135
1136         bifp = sc->sc_ifp;
1137         ASSERT_IFNET_SERIALIZED_ALL(bifp);
1138
1139         ifs = ifunit(req->ifbr_ifsname);
1140         if (ifs == NULL)
1141                 return (ENOENT);
1142
1143         /* If it's in the span list, it can't be a member. */
1144         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1145                 if (ifs == bif->bif_ifp)
1146                         return (EBUSY);
1147
1148         /* Allow the first Ethernet member to define the MTU */
1149         if (ifs->if_type != IFT_GIF) {
1150                 if (LIST_EMPTY(&sc->sc_iflists[mycpuid])) {
1151                         bifp->if_mtu = ifs->if_mtu;
1152                 } else if (bifp->if_mtu != ifs->if_mtu) {
1153                         if_printf(bifp, "invalid MTU for %s\n", ifs->if_xname);
1154                         return (EINVAL);
1155                 }
1156         }
1157
1158         if (ifs->if_bridge == sc)
1159                 return (EEXIST);
1160
1161         if (ifs->if_bridge != NULL)
1162                 return (EBUSY);
1163
1164         bif_info = kmalloc(sizeof(*bif_info), M_DEVBUF, M_WAITOK | M_ZERO);
1165         bif_info->bifi_priority = BSTP_DEFAULT_PORT_PRIORITY;
1166         bif_info->bifi_path_cost = BSTP_DEFAULT_PATH_COST;
1167         bif_info->bifi_ifp = ifs;
1168
1169         /*
1170          * Release bridge interface's serializer:
1171          * - To avoid possible dead lock.
1172          * - Various sync operation will block the current thread.
1173          */
1174         ifnet_deserialize_all(bifp);
1175
1176         switch (ifs->if_type) {
1177         case IFT_ETHER:
1178         case IFT_L2VLAN:
1179                 /*
1180                  * Place the interface into promiscuous mode.
1181                  */
1182                 error = ifpromisc(ifs, 1);
1183                 if (error) {
1184                         ifnet_serialize_all(bifp);
1185                         goto out;
1186                 }
1187                 bridge_mutecaps(bif_info, ifs, 1);
1188                 break;
1189
1190         case IFT_GIF: /* :^) */
1191                 break;
1192
1193         default:
1194                 error = EINVAL;
1195                 ifnet_serialize_all(bifp);
1196                 goto out;
1197         }
1198
1199         /*
1200          * Add bifs to percpu linked lists
1201          */
1202         bridge_add_bif(sc, bif_info, ifs);
1203
1204         ifnet_serialize_all(bifp);
1205
1206         if (bifp->if_flags & IFF_RUNNING)
1207                 bstp_initialization(sc);
1208         else
1209                 bstp_stop(sc);
1210
1211         /*
1212          * Everything has been setup, so let the member interface
1213          * deliver packets to this bridge on its input/output path.
1214          */
1215         ifs->if_bridge = sc;
1216 out:
1217         if (error) {
1218                 if (bif_info != NULL)
1219                         kfree(bif_info, M_DEVBUF);
1220         }
1221         return (error);
1222 }
1223
1224 static int
1225 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1226 {
1227         struct ifbreq *req = arg;
1228         struct bridge_iflist *bif;
1229
1230         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1231         if (bif == NULL)
1232                 return (ENOENT);
1233
1234         bridge_delete_member(sc, bif, 0);
1235
1236         return (0);
1237 }
1238
1239 static int
1240 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1241 {
1242         struct ifbreq *req = arg;
1243         struct bridge_iflist *bif;
1244
1245         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1246         if (bif == NULL)
1247                 return (ENOENT);
1248
1249         req->ifbr_ifsflags = bif->bif_flags;
1250         req->ifbr_state = bif->bif_state;
1251         req->ifbr_priority = bif->bif_priority;
1252         req->ifbr_path_cost = bif->bif_path_cost;
1253         req->ifbr_portno = bif->bif_ifp->if_index & 0xff;
1254
1255         return (0);
1256 }
1257
1258 static int
1259 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1260 {
1261         struct ifbreq *req = arg;
1262         struct bridge_iflist *bif;
1263         struct ifnet *bifp = sc->sc_ifp;
1264
1265         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1266         if (bif == NULL)
1267                 return (ENOENT);
1268
1269         if (req->ifbr_ifsflags & IFBIF_SPAN) {
1270                 /* SPAN is readonly */
1271                 return (EINVAL);
1272         }
1273
1274         if (req->ifbr_ifsflags & IFBIF_STP) {
1275                 switch (bif->bif_ifp->if_type) {
1276                 case IFT_ETHER:
1277                         /* These can do spanning tree. */
1278                         break;
1279
1280                 default:
1281                         /* Nothing else can. */
1282                         return (EINVAL);
1283                 }
1284         }
1285
1286         ifnet_deserialize_all(bifp);
1287         bridge_set_bifflags(sc, bif->bif_info, req->ifbr_ifsflags);
1288         ifnet_serialize_all(bifp);
1289
1290         if (bifp->if_flags & IFF_RUNNING)
1291                 bstp_initialization(sc);
1292
1293         return (0);
1294 }
1295
1296 static int
1297 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1298 {
1299         struct ifbrparam *param = arg;
1300         struct ifnet *ifp = sc->sc_ifp;
1301
1302         sc->sc_brtmax = param->ifbrp_csize;
1303
1304         ifnet_deserialize_all(ifp);
1305         bridge_rttrim(sc);
1306         ifnet_serialize_all(ifp);
1307
1308         return (0);
1309 }
1310
1311 static int
1312 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1313 {
1314         struct ifbrparam *param = arg;
1315
1316         param->ifbrp_csize = sc->sc_brtmax;
1317
1318         return (0);
1319 }
1320
1321 static int
1322 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1323 {
1324         struct bridge_control_arg *bc_arg = arg;
1325         struct ifbifconf *bifc = arg;
1326         struct bridge_iflist *bif;
1327         struct ifbreq *breq;
1328         int count, len;
1329
1330         count = 0;
1331         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next)
1332                 count++;
1333         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1334                 count++;
1335
1336         if (bifc->ifbic_len == 0) {
1337                 bifc->ifbic_len = sizeof(*breq) * count;
1338                 return 0;
1339         } else if (count == 0 || bifc->ifbic_len < sizeof(*breq)) {
1340                 bifc->ifbic_len = 0;
1341                 return 0;
1342         }
1343
1344         len = min(bifc->ifbic_len, sizeof(*breq) * count);
1345         KKASSERT(len >= sizeof(*breq));
1346
1347         breq = kmalloc(len, M_TEMP, M_WAITOK | M_NULLOK | M_ZERO);
1348         if (breq == NULL) {
1349                 bifc->ifbic_len = 0;
1350                 return ENOMEM;
1351         }
1352         bc_arg->bca_kptr = breq;
1353
1354         count = 0;
1355         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
1356                 if (len < sizeof(*breq))
1357                         break;
1358
1359                 strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname,
1360                         sizeof(breq->ifbr_ifsname));
1361                 breq->ifbr_ifsflags = bif->bif_flags;
1362                 breq->ifbr_state = bif->bif_state;
1363                 breq->ifbr_priority = bif->bif_priority;
1364                 breq->ifbr_path_cost = bif->bif_path_cost;
1365                 breq->ifbr_portno = bif->bif_ifp->if_index & 0xff;
1366                 breq++;
1367                 count++;
1368                 len -= sizeof(*breq);
1369         }
1370         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1371                 if (len < sizeof(*breq))
1372                         break;
1373
1374                 strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname,
1375                         sizeof(breq->ifbr_ifsname));
1376                 breq->ifbr_ifsflags = bif->bif_flags;
1377                 breq->ifbr_portno = bif->bif_ifp->if_index & 0xff;
1378                 breq++;
1379                 count++;
1380                 len -= sizeof(*breq);
1381         }
1382
1383         bifc->ifbic_len = sizeof(*breq) * count;
1384         KKASSERT(bifc->ifbic_len > 0);
1385
1386         bc_arg->bca_len = bifc->ifbic_len;
1387         bc_arg->bca_uptr = bifc->ifbic_req;
1388         return 0;
1389 }
1390
1391 static int
1392 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1393 {
1394         struct bridge_control_arg *bc_arg = arg;
1395         struct ifbaconf *bac = arg;
1396         struct bridge_rtnode *brt;
1397         struct ifbareq *bareq;
1398         int count, len;
1399
1400         count = 0;
1401         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list)
1402                 count++;
1403
1404         if (bac->ifbac_len == 0) {
1405                 bac->ifbac_len = sizeof(*bareq) * count;
1406                 return 0;
1407         } else if (count == 0 || bac->ifbac_len < sizeof(*bareq)) {
1408                 bac->ifbac_len = 0;
1409                 return 0;
1410         }
1411
1412         len = min(bac->ifbac_len, sizeof(*bareq) * count);
1413         KKASSERT(len >= sizeof(*bareq));
1414
1415         bareq = kmalloc(len, M_TEMP, M_WAITOK | M_NULLOK | M_ZERO);
1416         if (bareq == NULL) {
1417                 bac->ifbac_len = 0;
1418                 return ENOMEM;
1419         }
1420         bc_arg->bca_kptr = bareq;
1421
1422         count = 0;
1423         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
1424                 struct bridge_rtinfo *bri = brt->brt_info;
1425                 unsigned long expire;
1426
1427                 if (len < sizeof(*bareq))
1428                         break;
1429
1430                 strlcpy(bareq->ifba_ifsname, bri->bri_ifp->if_xname,
1431                         sizeof(bareq->ifba_ifsname));
1432                 memcpy(bareq->ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1433                 expire = bri->bri_expire;
1434                 if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1435                     time_second < expire)
1436                         bareq->ifba_expire = expire - time_second;
1437                 else
1438                         bareq->ifba_expire = 0;
1439                 bareq->ifba_flags = bri->bri_flags;
1440                 bareq++;
1441                 count++;
1442                 len -= sizeof(*bareq);
1443         }
1444
1445         bac->ifbac_len = sizeof(*bareq) * count;
1446         KKASSERT(bac->ifbac_len > 0);
1447
1448         bc_arg->bca_len = bac->ifbac_len;
1449         bc_arg->bca_uptr = bac->ifbac_req;
1450         return 0;
1451 }
1452
1453 static int
1454 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1455 {
1456         struct ifbareq *req = arg;
1457         struct bridge_iflist *bif;
1458         struct ifnet *ifp = sc->sc_ifp;
1459         int error;
1460
1461         ASSERT_IFNET_SERIALIZED_ALL(ifp);
1462
1463         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1464         if (bif == NULL)
1465                 return (ENOENT);
1466
1467         ifnet_deserialize_all(ifp);
1468         error = bridge_rtsaddr(sc, req->ifba_dst, bif->bif_ifp,
1469                                req->ifba_flags);
1470         ifnet_serialize_all(ifp);
1471         return (error);
1472 }
1473
1474 static int
1475 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1476 {
1477         struct ifbrparam *param = arg;
1478
1479         sc->sc_brttimeout = param->ifbrp_ctime;
1480
1481         return (0);
1482 }
1483
1484 static int
1485 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1486 {
1487         struct ifbrparam *param = arg;
1488
1489         param->ifbrp_ctime = sc->sc_brttimeout;
1490
1491         return (0);
1492 }
1493
1494 static int
1495 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1496 {
1497         struct ifbareq *req = arg;
1498         struct ifnet *ifp = sc->sc_ifp;
1499         int error;
1500
1501         ifnet_deserialize_all(ifp);
1502         error = bridge_rtdaddr(sc, req->ifba_dst);
1503         ifnet_serialize_all(ifp);
1504         return error;
1505 }
1506
1507 static int
1508 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1509 {
1510         struct ifbreq *req = arg;
1511         struct ifnet *ifp = sc->sc_ifp;
1512
1513         ifnet_deserialize_all(ifp);
1514         bridge_rtflush(sc, req->ifbr_ifsflags | IFBF_FLUSHSYNC);
1515         ifnet_serialize_all(ifp);
1516
1517         return (0);
1518 }
1519
1520 static int
1521 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1522 {
1523         struct ifbrparam *param = arg;
1524
1525         param->ifbrp_prio = sc->sc_bridge_priority;
1526
1527         return (0);
1528 }
1529
1530 static int
1531 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1532 {
1533         struct ifbrparam *param = arg;
1534
1535         sc->sc_bridge_priority = param->ifbrp_prio;
1536
1537         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1538                 bstp_initialization(sc);
1539
1540         return (0);
1541 }
1542
1543 static int
1544 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1545 {
1546         struct ifbrparam *param = arg;
1547
1548         param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8;
1549
1550         return (0);
1551 }
1552
1553 static int
1554 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1555 {
1556         struct ifbrparam *param = arg;
1557
1558         if (param->ifbrp_hellotime == 0)
1559                 return (EINVAL);
1560         sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8;
1561
1562         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1563                 bstp_initialization(sc);
1564
1565         return (0);
1566 }
1567
1568 static int
1569 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1570 {
1571         struct ifbrparam *param = arg;
1572
1573         param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8;
1574
1575         return (0);
1576 }
1577
1578 static int
1579 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1580 {
1581         struct ifbrparam *param = arg;
1582
1583         if (param->ifbrp_fwddelay == 0)
1584                 return (EINVAL);
1585         sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8;
1586
1587         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1588                 bstp_initialization(sc);
1589
1590         return (0);
1591 }
1592
1593 static int
1594 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1595 {
1596         struct ifbrparam *param = arg;
1597
1598         param->ifbrp_maxage = sc->sc_bridge_max_age >> 8;
1599
1600         return (0);
1601 }
1602
1603 static int
1604 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1605 {
1606         struct ifbrparam *param = arg;
1607
1608         if (param->ifbrp_maxage == 0)
1609                 return (EINVAL);
1610         sc->sc_bridge_max_age = param->ifbrp_maxage << 8;
1611
1612         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1613                 bstp_initialization(sc);
1614
1615         return (0);
1616 }
1617
1618 static int
1619 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1620 {
1621         struct ifbreq *req = arg;
1622         struct bridge_iflist *bif;
1623
1624         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1625         if (bif == NULL)
1626                 return (ENOENT);
1627
1628         bif->bif_priority = req->ifbr_priority;
1629
1630         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1631                 bstp_initialization(sc);
1632
1633         return (0);
1634 }
1635
1636 static int
1637 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1638 {
1639         struct ifbreq *req = arg;
1640         struct bridge_iflist *bif;
1641
1642         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1643         if (bif == NULL)
1644                 return (ENOENT);
1645
1646         bif->bif_path_cost = req->ifbr_path_cost;
1647
1648         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1649                 bstp_initialization(sc);
1650
1651         return (0);
1652 }
1653
1654 static int
1655 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1656 {
1657         struct ifbreq *req = arg;
1658         struct bridge_iflist *bif;
1659         struct ifnet *ifs;
1660
1661         ifs = ifunit(req->ifbr_ifsname);
1662         if (ifs == NULL)
1663                 return (ENOENT);
1664
1665         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1666                 if (ifs == bif->bif_ifp)
1667                         return (EBUSY);
1668
1669         if (ifs->if_bridge != NULL)
1670                 return (EBUSY);
1671
1672         switch (ifs->if_type) {
1673         case IFT_ETHER:
1674         case IFT_GIF:
1675         case IFT_L2VLAN:
1676                 break;
1677
1678         default:
1679                 return (EINVAL);
1680         }
1681
1682         bif = kmalloc(sizeof(*bif), M_DEVBUF, M_WAITOK | M_ZERO);
1683         bif->bif_ifp = ifs;
1684         bif->bif_flags = IFBIF_SPAN;
1685         /* NOTE: span bif does not need bridge_ifinfo */
1686
1687         LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1688
1689         sc->sc_span = 1;
1690
1691         return (0);
1692 }
1693
1694 static int
1695 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1696 {
1697         struct ifbreq *req = arg;
1698         struct bridge_iflist *bif;
1699         struct ifnet *ifs;
1700
1701         ifs = ifunit(req->ifbr_ifsname);
1702         if (ifs == NULL)
1703                 return (ENOENT);
1704
1705         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1706                 if (ifs == bif->bif_ifp)
1707                         break;
1708
1709         if (bif == NULL)
1710                 return (ENOENT);
1711
1712         bridge_delete_span(sc, bif);
1713
1714         if (LIST_EMPTY(&sc->sc_spanlist))
1715                 sc->sc_span = 0;
1716
1717         return (0);
1718 }
1719
1720 static void
1721 bridge_ifdetach_dispatch(netmsg_t msg)
1722 {
1723         struct ifnet *ifp, *bifp;
1724         struct bridge_softc *sc;
1725         struct bridge_iflist *bif;
1726
1727         ifp = msg->lmsg.u.ms_resultp;
1728         sc = ifp->if_bridge;
1729
1730         /* Check if the interface is a bridge member */
1731         if (sc != NULL) {
1732                 bifp = sc->sc_ifp;
1733
1734                 ifnet_serialize_all(bifp);
1735
1736                 bif = bridge_lookup_member_if(sc, ifp);
1737                 if (bif != NULL) {
1738                         bridge_delete_member(sc, bif, 1);
1739                 } else {
1740                         /* XXX Why bif will be NULL? */
1741                 }
1742
1743                 ifnet_deserialize_all(bifp);
1744                 goto reply;
1745         }
1746
1747         crit_enter();   /* XXX MP */
1748
1749         /* Check if the interface is a span port */
1750         LIST_FOREACH(sc, &bridge_list, sc_list) {
1751                 bifp = sc->sc_ifp;
1752
1753                 ifnet_serialize_all(bifp);
1754
1755                 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1756                         if (ifp == bif->bif_ifp) {
1757                                 bridge_delete_span(sc, bif);
1758                                 break;
1759                         }
1760
1761                 ifnet_deserialize_all(bifp);
1762         }
1763
1764         crit_exit();
1765
1766 reply:
1767         lwkt_replymsg(&msg->lmsg, 0);
1768 }
1769
1770 /*
1771  * bridge_ifdetach:
1772  *
1773  *      Detach an interface from a bridge.  Called when a member
1774  *      interface is detaching.
1775  */
1776 static void
1777 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1778 {
1779         struct netmsg_base msg;
1780
1781         netmsg_init(&msg, NULL, &curthread->td_msgport,
1782                     0, bridge_ifdetach_dispatch);
1783         msg.lmsg.u.ms_resultp = ifp;
1784
1785         lwkt_domsg(BRIDGE_CFGPORT, &msg.lmsg, 0);
1786 }
1787
1788 /*
1789  * bridge_init:
1790  *
1791  *      Initialize a bridge interface.
1792  */
1793 static void
1794 bridge_init(void *xsc)
1795 {
1796         bridge_control(xsc, SIOCSIFFLAGS, bridge_ioctl_init, NULL);
1797 }
1798
1799 /*
1800  * bridge_stop:
1801  *
1802  *      Stop the bridge interface.
1803  */
1804 static void
1805 bridge_stop(struct ifnet *ifp)
1806 {
1807         bridge_control(ifp->if_softc, SIOCSIFFLAGS, bridge_ioctl_stop, NULL);
1808 }
1809
1810 /*
1811  * bridge_enqueue:
1812  *
1813  *      Enqueue a packet on a bridge member interface.
1814  *
1815  */
1816 void
1817 bridge_enqueue(struct ifnet *dst_ifp, struct mbuf *m)
1818 {
1819         struct netmsg_packet *nmp;
1820
1821         nmp = &m->m_hdr.mh_netmsg;
1822         netmsg_init(&nmp->base, NULL, &netisr_apanic_rport,
1823                     0, bridge_enqueue_handler);
1824         nmp->nm_packet = m;
1825         nmp->base.lmsg.u.ms_resultp = dst_ifp;
1826
1827         lwkt_sendmsg(ifnet_portfn(mycpu->gd_cpuid), &nmp->base.lmsg);
1828 }
1829
1830 /*
1831  * bridge_output:
1832  *
1833  *      Send output from a bridge member interface.  This
1834  *      performs the bridging function for locally originated
1835  *      packets.
1836  *
1837  *      The mbuf has the Ethernet header already attached.  We must
1838  *      enqueue or free the mbuf before returning.
1839  */
1840 static int
1841 bridge_output(struct ifnet *ifp, struct mbuf *m)
1842 {
1843         struct bridge_softc *sc = ifp->if_bridge;
1844         struct ether_header *eh;
1845         struct ifnet *dst_if, *bifp;
1846         int from_us;
1847
1848         ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
1849
1850         /*
1851          * Make sure that we are still a member of a bridge interface.
1852          */
1853         if (sc == NULL) {
1854                 m_freem(m);
1855                 return (0);
1856         }
1857         bifp = sc->sc_ifp;
1858
1859         if (m->m_len < ETHER_HDR_LEN) {
1860                 m = m_pullup(m, ETHER_HDR_LEN);
1861                 if (m == NULL)
1862                         return (0);
1863         }
1864         eh = mtod(m, struct ether_header *);
1865
1866         if (memcmp(eh->ether_dhost, IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0)
1867                 from_us = 1;
1868         else
1869                 from_us = 0;
1870
1871         /*
1872          * If bridge is down, but the original output interface is up,
1873          * go ahead and send out that interface.  Otherwise, the packet
1874          * is dropped below.
1875          */
1876         if ((bifp->if_flags & IFF_RUNNING) == 0) {
1877                 dst_if = ifp;
1878                 goto sendunicast;
1879         }
1880
1881         /*
1882          * If the packet is a multicast, or we don't know a better way to
1883          * get there, send to all interfaces.
1884          */
1885         if (ETHER_IS_MULTICAST(eh->ether_dhost))
1886                 dst_if = NULL;
1887         else
1888                 dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1889         if (dst_if == NULL) {
1890                 struct bridge_iflist *bif, *nbif;
1891                 struct mbuf *mc;
1892                 int used = 0;
1893
1894                 if (sc->sc_span)
1895                         bridge_span(sc, m);
1896
1897                 LIST_FOREACH_MUTABLE(bif, &sc->sc_iflists[mycpuid],
1898                                      bif_next, nbif) {
1899                         dst_if = bif->bif_ifp;
1900                         if ((dst_if->if_flags & IFF_RUNNING) == 0)
1901                                 continue;
1902
1903                         /*
1904                          * If this is not the original output interface,
1905                          * and the interface is participating in spanning
1906                          * tree, make sure the port is in a state that
1907                          * allows forwarding.
1908                          */
1909                         if (dst_if != ifp &&
1910                             (bif->bif_flags & IFBIF_STP) != 0) {
1911                                 switch (bif->bif_state) {
1912                                 case BSTP_IFSTATE_BLOCKING:
1913                                 case BSTP_IFSTATE_LISTENING:
1914                                 case BSTP_IFSTATE_DISABLED:
1915                                         continue;
1916                                 }
1917                         }
1918
1919                         if (LIST_NEXT(bif, bif_next) == NULL) {
1920                                 used = 1;
1921                                 mc = m;
1922                         } else {
1923                                 mc = m_copypacket(m, MB_DONTWAIT);
1924                                 if (mc == NULL) {
1925                                         bifp->if_oerrors++;
1926                                         continue;
1927                                 }
1928                         }
1929
1930                         /*
1931                          * If the packet is 'from' us override ether_shost.
1932                          */
1933                         bridge_handoff(dst_if, mc, from_us);
1934
1935                         if (nbif != NULL && !nbif->bif_onlist) {
1936                                 KKASSERT(bif->bif_onlist);
1937                                 nbif = LIST_NEXT(bif, bif_next);
1938                         }
1939                 }
1940                 if (used == 0)
1941                         m_freem(m);
1942                 return (0);
1943         }
1944
1945 sendunicast:
1946         /*
1947          * XXX Spanning tree consideration here?
1948          */
1949         if (sc->sc_span)
1950                 bridge_span(sc, m);
1951         if ((dst_if->if_flags & IFF_RUNNING) == 0)
1952                 m_freem(m);
1953         else
1954                 bridge_handoff(dst_if, m, from_us);
1955         return (0);
1956 }
1957
1958 /*
1959  * bridge_start:
1960  *
1961  *      Start output on a bridge.
1962  */
1963 static void
1964 bridge_start(struct ifnet *ifp)
1965 {
1966         struct bridge_softc *sc = ifp->if_softc;
1967
1968         ASSERT_IFNET_SERIALIZED_TX(ifp);
1969
1970         ifp->if_flags |= IFF_OACTIVE;
1971         for (;;) {
1972                 struct ifnet *dst_if = NULL;
1973                 struct ether_header *eh;
1974                 struct mbuf *m;
1975
1976                 m = ifq_dequeue(&ifp->if_snd, NULL);
1977                 if (m == NULL)
1978                         break;
1979
1980                 if (m->m_len < sizeof(*eh)) {
1981                         m = m_pullup(m, sizeof(*eh));
1982                         if (m == NULL) {
1983                                 ifp->if_oerrors++;
1984                                 continue;
1985                         }
1986                 }
1987                 eh = mtod(m, struct ether_header *);
1988
1989                 BPF_MTAP(ifp, m);
1990                 ifp->if_opackets++;
1991
1992                 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0)
1993                         dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1994
1995                 if (dst_if == NULL)
1996                         bridge_start_bcast(sc, m);
1997                 else
1998                         bridge_enqueue(dst_if, m);
1999         }
2000         ifp->if_flags &= ~IFF_OACTIVE;
2001 }
2002
2003 /*
2004  * bridge_forward:
2005  *
2006  *      Forward packets received on a bridge interface via the input
2007  *      path.
2008  *
2009  *      The forwarding function of the bridge.
2010  */
2011 static void
2012 bridge_forward(struct bridge_softc *sc, struct mbuf *m)
2013 {
2014         struct bridge_iflist *bif;
2015         struct ifnet *src_if, *dst_if, *ifp;
2016         struct ether_header *eh;
2017
2018         src_if = m->m_pkthdr.rcvif;
2019         ifp = sc->sc_ifp;
2020
2021         ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
2022
2023         ifp->if_ipackets++;
2024         ifp->if_ibytes += m->m_pkthdr.len;
2025
2026         /*
2027          * Look up the bridge_iflist.
2028          */
2029         bif = bridge_lookup_member_if(sc, src_if);
2030         if (bif == NULL) {
2031                 /* Interface is not a bridge member (anymore?) */
2032                 m_freem(m);
2033                 return;
2034         }
2035
2036         if (bif->bif_flags & IFBIF_STP) {
2037                 switch (bif->bif_state) {
2038                 case BSTP_IFSTATE_BLOCKING:
2039                 case BSTP_IFSTATE_LISTENING:
2040                 case BSTP_IFSTATE_DISABLED:
2041                         m_freem(m);
2042                         return;
2043                 }
2044         }
2045
2046         eh = mtod(m, struct ether_header *);
2047
2048         /*
2049          * If the interface is learning, and the source
2050          * address is valid and not multicast, record
2051          * the address.
2052          */
2053         if ((bif->bif_flags & IFBIF_LEARNING) != 0 &&
2054             ETHER_IS_MULTICAST(eh->ether_shost) == 0 &&
2055             (eh->ether_shost[0] == 0 &&
2056              eh->ether_shost[1] == 0 &&
2057              eh->ether_shost[2] == 0 &&
2058              eh->ether_shost[3] == 0 &&
2059              eh->ether_shost[4] == 0 &&
2060              eh->ether_shost[5] == 0) == 0)
2061                 bridge_rtupdate(sc, eh->ether_shost, src_if, IFBAF_DYNAMIC);
2062
2063         if ((bif->bif_flags & IFBIF_STP) != 0 &&
2064             bif->bif_state == BSTP_IFSTATE_LEARNING) {
2065                 m_freem(m);
2066                 return;
2067         }
2068
2069         /*
2070          * At this point, the port either doesn't participate
2071          * in spanning tree or it is in the forwarding state.
2072          */
2073
2074         /*
2075          * If the packet is unicast, destined for someone on
2076          * "this" side of the bridge, drop it.
2077          */
2078         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2079                 dst_if = bridge_rtlookup(sc, eh->ether_dhost);
2080                 if (src_if == dst_if) {
2081                         m_freem(m);
2082                         return;
2083                 }
2084         } else {
2085                 /* ...forward it to all interfaces. */
2086                 ifp->if_imcasts++;
2087                 dst_if = NULL;
2088         }
2089
2090         if (dst_if == NULL) {
2091                 bridge_broadcast(sc, src_if, m);
2092                 return;
2093         }
2094
2095         /*
2096          * At this point, we're dealing with a unicast frame
2097          * going to a different interface.
2098          */
2099         if ((dst_if->if_flags & IFF_RUNNING) == 0) {
2100                 m_freem(m);
2101                 return;
2102         }
2103         bif = bridge_lookup_member_if(sc, dst_if);
2104         if (bif == NULL) {
2105                 /* Not a member of the bridge (anymore?) */
2106                 m_freem(m);
2107                 return;
2108         }
2109
2110         if (bif->bif_flags & IFBIF_STP) {
2111                 switch (bif->bif_state) {
2112                 case BSTP_IFSTATE_DISABLED:
2113                 case BSTP_IFSTATE_BLOCKING:
2114                         m_freem(m);
2115                         return;
2116                 }
2117         }
2118
2119         if (inet_pfil_hook.ph_hashooks > 0
2120 #ifdef INET6
2121             || inet6_pfil_hook.ph_hashooks > 0
2122 #endif
2123             ) {
2124                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2125                         return;
2126                 if (m == NULL)
2127                         return;
2128
2129                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2130                         return;
2131                 if (m == NULL)
2132                         return;
2133         }
2134         bridge_handoff(dst_if, m, 0);
2135 }
2136
2137 /*
2138  * bridge_input:
2139  *
2140  *      Receive input from a member interface.  Queue the packet for
2141  *      bridging if it is not for us.
2142  */
2143 static struct mbuf *
2144 bridge_input(struct ifnet *ifp, struct mbuf *m)
2145 {
2146         struct bridge_softc *sc = ifp->if_bridge;
2147         struct bridge_iflist *bif;
2148         struct ifnet *bifp, *new_ifp;
2149         struct ether_header *eh;
2150         struct mbuf *mc, *mc2;
2151
2152         ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
2153
2154         /*
2155          * Make sure that we are still a member of a bridge interface.
2156          */
2157         if (sc == NULL)
2158                 return m;
2159
2160         new_ifp = NULL;
2161         bifp = sc->sc_ifp;
2162
2163         if ((bifp->if_flags & IFF_RUNNING) == 0)
2164                 goto out;
2165
2166         /*
2167          * Implement support for bridge monitoring.  If this flag has been
2168          * set on this interface, discard the packet once we push it through
2169          * the bpf(4) machinery, but before we do, increment various counters
2170          * associated with this bridge.
2171          */
2172         if (bifp->if_flags & IFF_MONITOR) {
2173                 /* Change input interface to this bridge */
2174                 m->m_pkthdr.rcvif = bifp;
2175
2176                 BPF_MTAP(bifp, m);
2177
2178                 /* Update bridge's ifnet statistics */
2179                 bifp->if_ipackets++;
2180                 bifp->if_ibytes += m->m_pkthdr.len;
2181                 if (m->m_flags & (M_MCAST | M_BCAST))
2182                         bifp->if_imcasts++;
2183
2184                 m_freem(m);
2185                 m = NULL;
2186                 goto out;
2187         }
2188
2189         /*
2190          * Handle the ether_header
2191          */
2192         eh = mtod(m, struct ether_header *);
2193         m->m_pkthdr.fw_flags |= BRIDGE_MBUF_TAGGED;
2194         bcopy(eh, &m->m_pkthdr.br.ether, sizeof(*eh));
2195
2196         if (memcmp(eh->ether_dhost, IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0) {
2197                 /*
2198                  * If the packet is for us, set the packets source as the
2199                  * bridge, and return the packet back to ifnet.if_input for
2200                  * local processing.
2201                  */
2202                 KASSERT(bifp->if_bridge == NULL,
2203                         ("loop created in bridge_input"));
2204                 if (pfil_member != 0) {
2205                         if (inet_pfil_hook.ph_hashooks > 0
2206 #ifdef INET6
2207                             || inet6_pfil_hook.ph_hashooks > 0
2208 #endif
2209                         ) {
2210                                 if (bridge_pfil(&m, NULL, ifp, PFIL_IN) != 0)
2211                                         goto out;
2212                                 if (m == NULL)
2213                                         goto out;
2214                         }
2215                 }
2216                 new_ifp = bifp;
2217                 goto out;
2218         }
2219
2220         /*
2221          * Tap all packets arriving on the bridge, no matter if
2222          * they are local destinations or not.  In is in.
2223          */
2224         BPF_MTAP(bifp, m);
2225
2226         bif = bridge_lookup_member_if(sc, ifp);
2227         if (bif == NULL)
2228                 goto out;
2229
2230         if (sc->sc_span)
2231                 bridge_span(sc, m);
2232
2233         if (m->m_flags & (M_BCAST | M_MCAST)) {
2234                 /* Tap off 802.1D packets; they do not get forwarded. */
2235                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2236                     ETHER_ADDR_LEN) == 0) {
2237                         ifnet_serialize_all(bifp);
2238                         bstp_input(sc, bif, m);
2239                         ifnet_deserialize_all(bifp);
2240
2241                         /* m is freed by bstp_input */
2242                         m = NULL;
2243                         goto out;
2244                 }
2245
2246                 if (bif->bif_flags & IFBIF_STP) {
2247                         switch (bif->bif_state) {
2248                         case BSTP_IFSTATE_BLOCKING:
2249                         case BSTP_IFSTATE_LISTENING:
2250                         case BSTP_IFSTATE_DISABLED:
2251                                 goto out;
2252                         }
2253                 }
2254
2255                 /*
2256                  * Make a deep copy of the packet and enqueue the copy
2257                  * for bridge processing; return the original packet for
2258                  * local processing.
2259                  */
2260                 mc = m_dup(m, MB_DONTWAIT);
2261                 if (mc == NULL)
2262                         goto out;
2263
2264                 bridge_forward(sc, mc);
2265
2266                 /*
2267                  * Reinject the mbuf as arriving on the bridge so we have a
2268                  * chance at claiming multicast packets. We can not loop back
2269                  * here from ether_input as a bridge is never a member of a
2270                  * bridge.
2271                  */
2272                 KASSERT(bifp->if_bridge == NULL,
2273                         ("loop created in bridge_input"));
2274                 mc2 = m_dup(m, MB_DONTWAIT);
2275 #ifdef notyet
2276                 if (mc2 != NULL) {
2277                         /* Keep the layer3 header aligned */
2278                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2279                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2280                 }
2281 #endif
2282                 if (mc2 != NULL) {
2283                         /*
2284                          * Don't tap to bpf(4) again; we have already done
2285                          * the tapping.
2286                          *
2287                          * Leave m_pkthdr.rcvif alone, so ARP replies are
2288                          * processed as coming in on the correct interface.
2289                          */
2290                         ether_reinput_oncpu(bifp, mc2, REINPUT_KEEPRCVIF);
2291                 }
2292
2293                 /* Return the original packet for local processing. */
2294                 goto out;
2295         }
2296
2297         if (bif->bif_flags & IFBIF_STP) {
2298                 switch (bif->bif_state) {
2299                 case BSTP_IFSTATE_BLOCKING:
2300                 case BSTP_IFSTATE_LISTENING:
2301                 case BSTP_IFSTATE_DISABLED:
2302                         goto out;
2303                 }
2304         }
2305
2306         /*
2307          * Unicast.  Make sure it's not for us.
2308          *
2309          * This loop is MPSAFE; the only blocking operation (bridge_rtupdate)
2310          * is followed by breaking out of the loop.
2311          */
2312         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
2313                 if (bif->bif_ifp->if_type != IFT_ETHER)
2314                         continue;
2315
2316                 /*
2317                  * It is destined for us.  Reinput on the same interface
2318                  * it came in on so things like ARP responses get assigned
2319                  * to the correct member (the incoming interface) and not
2320                  * to the member which happens to have the matching dhost.
2321                  */
2322                 if (memcmp(IF_LLADDR(bif->bif_ifp), eh->ether_dhost,
2323                     ETHER_ADDR_LEN) == 0) {
2324                         if (bif->bif_ifp != ifp) {
2325                                 /* XXX loop prevention */
2326                                 m->m_flags |= M_ETHER_BRIDGED;
2327                                 new_ifp = bif->bif_ifp;
2328                         }
2329                         if (bif->bif_flags & IFBIF_LEARNING) {
2330                                 bridge_rtupdate(sc, eh->ether_shost,
2331                                                 ifp, IFBAF_DYNAMIC);
2332                         }
2333                         goto out;
2334                 }
2335
2336                 /* We just received a packet that we sent out. */
2337                 if (memcmp(IF_LLADDR(bif->bif_ifp), eh->ether_shost,
2338                     ETHER_ADDR_LEN) == 0) {
2339                         m_freem(m);
2340                         m = NULL;
2341                         goto out;
2342                 }
2343         }
2344
2345         /* Perform the bridge forwarding function. */
2346         bridge_forward(sc, m);
2347         m = NULL;
2348
2349         /*
2350          * Leave m_pkthdr.rcvif alone, so ARP replies are
2351          * processed as coming in on the correct interface.
2352          */
2353 out:
2354         if (new_ifp != NULL) {
2355                 ether_reinput_oncpu(new_ifp, m,
2356                                     REINPUT_KEEPRCVIF|REINPUT_RUNBPF);
2357                 m = NULL;
2358         }
2359         return (m);
2360 }
2361
2362 /*
2363  * bridge_start_bcast:
2364  *
2365  *      Broadcast the packet sent from bridge to all member
2366  *      interfaces.
2367  *      This is a simplified version of bridge_broadcast(), however,
2368  *      this function expects caller to hold bridge's serializer.
2369  */
2370 static void
2371 bridge_start_bcast(struct bridge_softc *sc, struct mbuf *m)
2372 {
2373         struct bridge_iflist *bif;
2374         struct mbuf *mc;
2375         struct ifnet *dst_if, *bifp;
2376         int used = 0;
2377
2378         bifp = sc->sc_ifp;
2379         ASSERT_IFNET_SERIALIZED_ALL(bifp);
2380
2381         /*
2382          * Following loop is MPSAFE; nothing is blocking
2383          * in the loop body.
2384          */
2385         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
2386                 dst_if = bif->bif_ifp;
2387
2388                 if (bif->bif_flags & IFBIF_STP) {
2389                         switch (bif->bif_state) {
2390                         case BSTP_IFSTATE_BLOCKING:
2391                         case BSTP_IFSTATE_DISABLED:
2392                                 continue;
2393                         }
2394                 }
2395
2396                 if ((bif->bif_flags & IFBIF_DISCOVER) == 0 &&
2397                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2398                         continue;
2399
2400                 if ((dst_if->if_flags & IFF_RUNNING) == 0)
2401                         continue;
2402
2403                 if (LIST_NEXT(bif, bif_next) == NULL) {
2404                         mc = m;
2405                         used = 1;
2406                 } else {
2407                         mc = m_copypacket(m, MB_DONTWAIT);
2408                         if (mc == NULL) {
2409                                 bifp->if_oerrors++;
2410                                 continue;
2411                         }
2412                 }
2413                 bridge_enqueue(dst_if, mc);
2414         }
2415         if (used == 0)
2416                 m_freem(m);
2417 }
2418
2419 /*
2420  * bridge_broadcast:
2421  *
2422  *      Send a frame to all interfaces that are members of
2423  *      the bridge, except for the one on which the packet
2424  *      arrived.
2425  */
2426 static void
2427 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2428                  struct mbuf *m)
2429 {
2430         struct bridge_iflist *bif, *nbif;
2431         struct ether_header *eh;
2432         struct mbuf *mc;
2433         struct ifnet *dst_if, *bifp;
2434         int used = 0;
2435         int from_us;
2436
2437         bifp = sc->sc_ifp;
2438         ASSERT_IFNET_NOT_SERIALIZED_ALL(bifp);
2439
2440         eh = mtod(m, struct ether_header *);
2441         if (memcmp(eh->ether_dhost, IF_LLADDR(src_if), ETHER_ADDR_LEN) == 0)
2442                 from_us = 1;
2443         else
2444                 from_us = 0;
2445
2446         if (inet_pfil_hook.ph_hashooks > 0
2447 #ifdef INET6
2448             || inet6_pfil_hook.ph_hashooks > 0
2449 #endif
2450             ) {
2451                 if (bridge_pfil(&m, bifp, src_if, PFIL_IN) != 0)
2452                         return;
2453                 if (m == NULL)
2454                         return;
2455
2456                 /* Filter on the bridge interface before broadcasting */
2457                 if (bridge_pfil(&m, bifp, NULL, PFIL_OUT) != 0)
2458                         return;
2459                 if (m == NULL)
2460                         return;
2461         }
2462
2463         LIST_FOREACH_MUTABLE(bif, &sc->sc_iflists[mycpuid], bif_next, nbif) {
2464                 dst_if = bif->bif_ifp;
2465                 if (dst_if == src_if)
2466                         continue;
2467
2468                 if (bif->bif_flags & IFBIF_STP) {
2469                         switch (bif->bif_state) {
2470                         case BSTP_IFSTATE_BLOCKING:
2471                         case BSTP_IFSTATE_DISABLED:
2472                                 continue;
2473                         }
2474                 }
2475
2476                 if ((bif->bif_flags & IFBIF_DISCOVER) == 0 &&
2477                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2478                         continue;
2479
2480                 if ((dst_if->if_flags & IFF_RUNNING) == 0)
2481                         continue;
2482
2483                 if (LIST_NEXT(bif, bif_next) == NULL) {
2484                         mc = m;
2485                         used = 1;
2486                 } else {
2487                         mc = m_copypacket(m, MB_DONTWAIT);
2488                         if (mc == NULL) {
2489                                 sc->sc_ifp->if_oerrors++;
2490                                 continue;
2491                         }
2492                 }
2493
2494                 /*
2495                  * Filter on the output interface.  Pass a NULL bridge
2496                  * interface pointer so we do not redundantly filter on
2497                  * the bridge for each interface we broadcast on.
2498                  */
2499                 if (inet_pfil_hook.ph_hashooks > 0
2500 #ifdef INET6
2501                     || inet6_pfil_hook.ph_hashooks > 0
2502 #endif
2503                     ) {
2504                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2505                                 continue;
2506                         if (mc == NULL)
2507                                 continue;
2508                 }
2509                 bridge_handoff(dst_if, mc, from_us);
2510
2511                 if (nbif != NULL && !nbif->bif_onlist) {
2512                         KKASSERT(bif->bif_onlist);
2513                         nbif = LIST_NEXT(bif, bif_next);
2514                 }
2515         }
2516         if (used == 0)
2517                 m_freem(m);
2518 }
2519
2520 /*
2521  * bridge_span:
2522  *
2523  *      Duplicate a packet out one or more interfaces that are in span mode,
2524  *      the original mbuf is unmodified.
2525  */
2526 static void
2527 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2528 {
2529         struct bridge_iflist *bif;
2530         struct ifnet *dst_if, *bifp;
2531         struct mbuf *mc;
2532
2533         bifp = sc->sc_ifp;
2534         ifnet_serialize_all(bifp);
2535
2536         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2537                 dst_if = bif->bif_ifp;
2538
2539                 if ((dst_if->if_flags & IFF_RUNNING) == 0)
2540                         continue;
2541
2542                 mc = m_copypacket(m, MB_DONTWAIT);
2543                 if (mc == NULL) {
2544                         sc->sc_ifp->if_oerrors++;
2545                         continue;
2546                 }
2547                 bridge_enqueue(dst_if, mc);
2548         }
2549
2550         ifnet_deserialize_all(bifp);
2551 }
2552
2553 static void
2554 bridge_rtmsg_sync_handler(netmsg_t msg)
2555 {
2556         ifnet_forwardmsg(&msg->lmsg, mycpuid + 1);
2557 }
2558
2559 static void
2560 bridge_rtmsg_sync(struct bridge_softc *sc)
2561 {
2562         struct netmsg_base msg;
2563
2564         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
2565
2566         netmsg_init(&msg, NULL, &curthread->td_msgport,
2567                     0, bridge_rtmsg_sync_handler);
2568         ifnet_domsg(&msg.lmsg, 0);
2569 }
2570
2571 static __inline void
2572 bridge_rtinfo_update(struct bridge_rtinfo *bri, struct ifnet *dst_if,
2573                      int setflags, uint8_t flags, uint32_t timeo)
2574 {
2575         if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2576             bri->bri_ifp != dst_if)
2577                 bri->bri_ifp = dst_if;
2578         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2579             bri->bri_expire != time_second + timeo)
2580                 bri->bri_expire = time_second + timeo;
2581         if (setflags)
2582                 bri->bri_flags = flags;
2583 }
2584
2585 static int
2586 bridge_rtinstall_oncpu(struct bridge_softc *sc, const uint8_t *dst,
2587                        struct ifnet *dst_if, int setflags, uint8_t flags,
2588                        struct bridge_rtinfo **bri0)
2589 {
2590         struct bridge_rtnode *brt;
2591         struct bridge_rtinfo *bri;
2592
2593         if (mycpuid == 0) {
2594                 brt = bridge_rtnode_lookup(sc, dst);
2595                 if (brt != NULL) {
2596                         /*
2597                          * rtnode for 'dst' already exists.  We inform the
2598                          * caller about this by leaving bri0 as NULL.  The
2599                          * caller will terminate the intallation upon getting
2600                          * NULL bri0.  However, we still need to update the
2601                          * rtinfo.
2602                          */
2603                         KKASSERT(*bri0 == NULL);
2604
2605                         /* Update rtinfo */
2606                         bridge_rtinfo_update(brt->brt_info, dst_if, setflags,
2607                                              flags, sc->sc_brttimeout);
2608                         return 0;
2609                 }
2610
2611                 /*
2612                  * We only need to check brtcnt on CPU0, since if limit
2613                  * is to be exceeded, ENOSPC is returned.  Caller knows
2614                  * this and will terminate the installation.
2615                  */
2616                 if (sc->sc_brtcnt >= sc->sc_brtmax)
2617                         return ENOSPC;
2618
2619                 KKASSERT(*bri0 == NULL);
2620                 bri = kmalloc(sizeof(struct bridge_rtinfo), M_DEVBUF,
2621                                   M_WAITOK | M_ZERO);
2622                 *bri0 = bri;
2623
2624                 /* Setup rtinfo */
2625                 bri->bri_flags = IFBAF_DYNAMIC;
2626                 bridge_rtinfo_update(bri, dst_if, setflags, flags,
2627                                      sc->sc_brttimeout);
2628         } else {
2629                 bri = *bri0;
2630                 KKASSERT(bri != NULL);
2631         }
2632
2633         brt = kmalloc(sizeof(struct bridge_rtnode), M_DEVBUF,
2634                       M_WAITOK | M_ZERO);
2635         memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2636         brt->brt_info = bri;
2637
2638         bridge_rtnode_insert(sc, brt);
2639         return 0;
2640 }
2641
2642 static void
2643 bridge_rtinstall_handler(netmsg_t msg)
2644 {
2645         struct netmsg_brsaddr *brmsg = (struct netmsg_brsaddr *)msg;
2646         int error;
2647
2648         error = bridge_rtinstall_oncpu(brmsg->br_softc,
2649                                        brmsg->br_dst, brmsg->br_dst_if,
2650                                        brmsg->br_setflags, brmsg->br_flags,
2651                                        &brmsg->br_rtinfo);
2652         if (error) {
2653                 KKASSERT(mycpuid == 0 && brmsg->br_rtinfo == NULL);
2654                 lwkt_replymsg(&brmsg->base.lmsg, error);
2655                 return;
2656         } else if (brmsg->br_rtinfo == NULL) {
2657                 /* rtnode already exists for 'dst' */
2658                 KKASSERT(mycpuid == 0);
2659                 lwkt_replymsg(&brmsg->base.lmsg, 0);
2660                 return;
2661         }
2662         ifnet_forwardmsg(&brmsg->base.lmsg, mycpuid + 1);
2663 }
2664
2665 /*
2666  * bridge_rtupdate:
2667  *
2668  *      Add/Update a bridge routing entry.
2669  */
2670 static int
2671 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
2672                 struct ifnet *dst_if, uint8_t flags)
2673 {
2674         struct bridge_rtnode *brt;
2675
2676         /*
2677          * A route for this destination might already exist.  If so,
2678          * update it, otherwise create a new one.
2679          */
2680         if ((brt = bridge_rtnode_lookup(sc, dst)) == NULL) {
2681                 struct netmsg_brsaddr *brmsg;
2682
2683                 if (sc->sc_brtcnt >= sc->sc_brtmax)
2684                         return ENOSPC;
2685
2686                 brmsg = kmalloc(sizeof(*brmsg), M_LWKTMSG, M_WAITOK | M_NULLOK);
2687                 if (brmsg == NULL)
2688                         return ENOMEM;
2689
2690                 netmsg_init(&brmsg->base, NULL, &netisr_afree_rport,
2691                             0, bridge_rtinstall_handler);
2692                 memcpy(brmsg->br_dst, dst, ETHER_ADDR_LEN);
2693                 brmsg->br_dst_if = dst_if;
2694                 brmsg->br_flags = flags;
2695                 brmsg->br_setflags = 0;
2696                 brmsg->br_softc = sc;
2697                 brmsg->br_rtinfo = NULL;
2698
2699                 ifnet_sendmsg(&brmsg->base.lmsg, 0);
2700                 return 0;
2701         }
2702         bridge_rtinfo_update(brt->brt_info, dst_if, 0, flags,
2703                              sc->sc_brttimeout);
2704         return 0;
2705 }
2706
2707 static int
2708 bridge_rtsaddr(struct bridge_softc *sc, const uint8_t *dst,
2709                struct ifnet *dst_if, uint8_t flags)
2710 {
2711         struct netmsg_brsaddr brmsg;
2712
2713         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
2714
2715         netmsg_init(&brmsg.base, NULL, &curthread->td_msgport,
2716                     0, bridge_rtinstall_handler);
2717         memcpy(brmsg.br_dst, dst, ETHER_ADDR_LEN);
2718         brmsg.br_dst_if = dst_if;
2719         brmsg.br_flags = flags;
2720         brmsg.br_setflags = 1;
2721         brmsg.br_softc = sc;
2722         brmsg.br_rtinfo = NULL;
2723
2724         return ifnet_domsg(&brmsg.base.lmsg, 0);
2725 }
2726
2727 /*
2728  * bridge_rtlookup:
2729  *
2730  *      Lookup the destination interface for an address.
2731  */
2732 static struct ifnet *
2733 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr)
2734 {
2735         struct bridge_rtnode *brt;
2736
2737         if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL)
2738                 return NULL;
2739         return brt->brt_info->bri_ifp;
2740 }
2741
2742 static void
2743 bridge_rtreap_handler(netmsg_t msg)
2744 {
2745         struct bridge_softc *sc = msg->lmsg.u.ms_resultp;
2746         struct bridge_rtnode *brt, *nbrt;
2747
2748         LIST_FOREACH_MUTABLE(brt, &sc->sc_rtlists[mycpuid], brt_list, nbrt) {
2749                 if (brt->brt_info->bri_dead)
2750                         bridge_rtnode_destroy(sc, brt);
2751         }
2752         ifnet_forwardmsg(&msg->lmsg, mycpuid + 1);
2753 }
2754
2755 static void
2756 bridge_rtreap(struct bridge_softc *sc)
2757 {
2758         struct netmsg_base msg;
2759
2760         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
2761
2762         netmsg_init(&msg, NULL, &curthread->td_msgport,
2763                     0, bridge_rtreap_handler);
2764         msg.lmsg.u.ms_resultp = sc;
2765
2766         ifnet_domsg(&msg.lmsg, 0);
2767 }
2768
2769 static void
2770 bridge_rtreap_async(struct bridge_softc *sc)
2771 {
2772         struct netmsg_base *msg;
2773
2774         msg = kmalloc(sizeof(*msg), M_LWKTMSG, M_WAITOK);
2775
2776         netmsg_init(msg, NULL, &netisr_afree_rport,
2777                     0, bridge_rtreap_handler);
2778         msg->lmsg.u.ms_resultp = sc;
2779
2780         ifnet_sendmsg(&msg->lmsg, 0);
2781 }
2782
2783 /*
2784  * bridge_rttrim:
2785  *
2786  *      Trim the routine table so that we have a number
2787  *      of routing entries less than or equal to the
2788  *      maximum number.
2789  */
2790 static void
2791 bridge_rttrim(struct bridge_softc *sc)
2792 {
2793         struct bridge_rtnode *brt;
2794         int dead;
2795
2796         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
2797
2798         /* Make sure we actually need to do this. */
2799         if (sc->sc_brtcnt <= sc->sc_brtmax)
2800                 return;
2801
2802         /*
2803          * Find out how many rtnodes are dead
2804          */
2805         dead = bridge_rtage_finddead(sc);
2806         KKASSERT(dead <= sc->sc_brtcnt);
2807
2808         if (sc->sc_brtcnt - dead <= sc->sc_brtmax) {
2809                 /* Enough dead rtnodes are found */
2810                 bridge_rtreap(sc);
2811                 return;
2812         }
2813
2814         /*
2815          * Kill some dynamic rtnodes to meet the brtmax
2816          */
2817         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
2818                 struct bridge_rtinfo *bri = brt->brt_info;
2819
2820                 if (bri->bri_dead) {
2821                         /*
2822                          * We have counted this rtnode in
2823                          * bridge_rtage_finddead()
2824                          */
2825                         continue;
2826                 }
2827
2828                 if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2829                         bri->bri_dead = 1;
2830                         ++dead;
2831                         KKASSERT(dead <= sc->sc_brtcnt);
2832
2833                         if (sc->sc_brtcnt - dead <= sc->sc_brtmax) {
2834                                 /* Enough rtnodes are collected */
2835                                 break;
2836                         }
2837                 }
2838         }
2839         if (dead)
2840                 bridge_rtreap(sc);
2841 }
2842
2843 /*
2844  * bridge_timer:
2845  *
2846  *      Aging timer for the bridge.
2847  */
2848 static void
2849 bridge_timer(void *arg)
2850 {
2851         struct bridge_softc *sc = arg;
2852         struct netmsg_base *msg;
2853
2854         KKASSERT(mycpuid == BRIDGE_CFGCPU);
2855
2856         crit_enter();
2857
2858         if (callout_pending(&sc->sc_brcallout) ||
2859             !callout_active(&sc->sc_brcallout)) {
2860                 crit_exit();
2861                 return;
2862         }
2863         callout_deactivate(&sc->sc_brcallout);
2864
2865         msg = &sc->sc_brtimemsg;
2866         KKASSERT(msg->lmsg.ms_flags & MSGF_DONE);
2867         lwkt_sendmsg(BRIDGE_CFGPORT, &msg->lmsg);
2868
2869         crit_exit();
2870 }
2871
2872 static void
2873 bridge_timer_handler(netmsg_t msg)
2874 {
2875         struct bridge_softc *sc = msg->lmsg.u.ms_resultp;
2876
2877         KKASSERT(&curthread->td_msgport == BRIDGE_CFGPORT);
2878
2879         crit_enter();
2880         /* Reply ASAP */
2881         lwkt_replymsg(&msg->lmsg, 0);
2882         crit_exit();
2883
2884         bridge_rtage(sc);
2885         if (sc->sc_ifp->if_flags & IFF_RUNNING) {
2886                 callout_reset(&sc->sc_brcallout,
2887                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2888         }
2889 }
2890
2891 static int
2892 bridge_rtage_finddead(struct bridge_softc *sc)
2893 {
2894         struct bridge_rtnode *brt;
2895         int dead = 0;
2896
2897         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
2898                 struct bridge_rtinfo *bri = brt->brt_info;
2899
2900                 if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2901                     time_second >= bri->bri_expire) {
2902                         bri->bri_dead = 1;
2903                         ++dead;
2904                         KKASSERT(dead <= sc->sc_brtcnt);
2905                 }
2906         }
2907         return dead;
2908 }
2909
2910 /*
2911  * bridge_rtage:
2912  *
2913  *      Perform an aging cycle.
2914  */
2915 static void
2916 bridge_rtage(struct bridge_softc *sc)
2917 {
2918         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
2919
2920         if (bridge_rtage_finddead(sc))
2921                 bridge_rtreap(sc);
2922 }
2923
2924 /*
2925  * bridge_rtflush:
2926  *
2927  *      Remove all dynamic addresses from the bridge.
2928  */
2929 static void
2930 bridge_rtflush(struct bridge_softc *sc, int bf)
2931 {
2932         struct bridge_rtnode *brt;
2933         int reap;
2934
2935         reap = 0;
2936         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
2937                 struct bridge_rtinfo *bri = brt->brt_info;
2938
2939                 if ((bf & IFBF_FLUSHALL) ||
2940                     (bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2941                         bri->bri_dead = 1;
2942                         reap = 1;
2943                 }
2944         }
2945         if (reap) {
2946                 if (bf & IFBF_FLUSHSYNC)
2947                         bridge_rtreap(sc);
2948                 else
2949                         bridge_rtreap_async(sc);
2950         }
2951 }
2952
2953 /*
2954  * bridge_rtdaddr:
2955  *
2956  *      Remove an address from the table.
2957  */
2958 static int
2959 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr)
2960 {
2961         struct bridge_rtnode *brt;
2962
2963         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
2964
2965         if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL)
2966                 return (ENOENT);
2967
2968         /* TODO: add a cheaper delete operation */
2969         brt->brt_info->bri_dead = 1;
2970         bridge_rtreap(sc);
2971         return (0);
2972 }
2973
2974 /*
2975  * bridge_rtdelete:
2976  *
2977  *      Delete routes to a speicifc member interface.
2978  */
2979 void
2980 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int bf)
2981 {
2982         struct bridge_rtnode *brt;
2983         int reap;
2984
2985         reap = 0;
2986         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
2987                 struct bridge_rtinfo *bri = brt->brt_info;
2988
2989                 if (bri->bri_ifp == ifp &&
2990                     ((bf & IFBF_FLUSHALL) ||
2991                      (bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)) {
2992                         bri->bri_dead = 1;
2993                         reap = 1;
2994                 }
2995         }
2996         if (reap) {
2997                 if (bf & IFBF_FLUSHSYNC)
2998                         bridge_rtreap(sc);
2999                 else
3000                         bridge_rtreap_async(sc);
3001         }
3002 }
3003
3004 /*
3005  * bridge_rtable_init:
3006  *
3007  *      Initialize the route table for this bridge.
3008  */
3009 static void
3010 bridge_rtable_init(struct bridge_softc *sc)
3011 {
3012         int cpu;
3013
3014         /*
3015          * Initialize per-cpu hash tables
3016          */
3017         sc->sc_rthashs = kmalloc(sizeof(*sc->sc_rthashs) * ncpus,
3018                                  M_DEVBUF, M_WAITOK);
3019         for (cpu = 0; cpu < ncpus; ++cpu) {
3020                 int i;
3021
3022                 sc->sc_rthashs[cpu] =
3023                 kmalloc(sizeof(struct bridge_rtnode_head) * BRIDGE_RTHASH_SIZE,
3024                         M_DEVBUF, M_WAITOK);
3025
3026                 for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
3027                         LIST_INIT(&sc->sc_rthashs[cpu][i]);
3028         }
3029         sc->sc_rthash_key = karc4random();
3030
3031         /*
3032          * Initialize per-cpu lists
3033          */
3034         sc->sc_rtlists = kmalloc(sizeof(struct bridge_rtnode_head) * ncpus,
3035                                  M_DEVBUF, M_WAITOK);
3036         for (cpu = 0; cpu < ncpus; ++cpu)
3037                 LIST_INIT(&sc->sc_rtlists[cpu]);
3038 }
3039
3040 /*
3041  * bridge_rtable_fini:
3042  *
3043  *      Deconstruct the route table for this bridge.
3044  */
3045 static void
3046 bridge_rtable_fini(struct bridge_softc *sc)
3047 {
3048         int cpu;
3049
3050         /*
3051          * Free per-cpu hash tables
3052          */
3053         for (cpu = 0; cpu < ncpus; ++cpu)
3054                 kfree(sc->sc_rthashs[cpu], M_DEVBUF);
3055         kfree(sc->sc_rthashs, M_DEVBUF);
3056
3057         /*
3058          * Free per-cpu lists
3059          */
3060         kfree(sc->sc_rtlists, M_DEVBUF);
3061 }
3062
3063 /*
3064  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
3065  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
3066  */
3067 #define mix(a, b, c)                                                    \
3068 do {                                                                    \
3069         a -= b; a -= c; a ^= (c >> 13);                                 \
3070         b -= c; b -= a; b ^= (a << 8);                                  \
3071         c -= a; c -= b; c ^= (b >> 13);                                 \
3072         a -= b; a -= c; a ^= (c >> 12);                                 \
3073         b -= c; b -= a; b ^= (a << 16);                                 \
3074         c -= a; c -= b; c ^= (b >> 5);                                  \
3075         a -= b; a -= c; a ^= (c >> 3);                                  \
3076         b -= c; b -= a; b ^= (a << 10);                                 \
3077         c -= a; c -= b; c ^= (b >> 15);                                 \
3078 } while (/*CONSTCOND*/0)
3079
3080 static __inline uint32_t
3081 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
3082 {
3083         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
3084
3085         b += addr[5] << 8;
3086         b += addr[4];
3087         a += addr[3] << 24;
3088         a += addr[2] << 16;
3089         a += addr[1] << 8;
3090         a += addr[0];
3091
3092         mix(a, b, c);
3093
3094         return (c & BRIDGE_RTHASH_MASK);
3095 }
3096
3097 #undef mix
3098
3099 static int
3100 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
3101 {
3102         int i, d;
3103
3104         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
3105                 d = ((int)a[i]) - ((int)b[i]);
3106         }
3107
3108         return (d);
3109 }
3110
3111 /*
3112  * bridge_rtnode_lookup:
3113  *
3114  *      Look up a bridge route node for the specified destination.
3115  */
3116 static struct bridge_rtnode *
3117 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr)
3118 {
3119         struct bridge_rtnode *brt;
3120         uint32_t hash;
3121         int dir;
3122
3123         hash = bridge_rthash(sc, addr);
3124         LIST_FOREACH(brt, &sc->sc_rthashs[mycpuid][hash], brt_hash) {
3125                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
3126                 if (dir == 0)
3127                         return (brt);
3128                 if (dir > 0)
3129                         return (NULL);
3130         }
3131
3132         return (NULL);
3133 }
3134
3135 /*
3136  * bridge_rtnode_insert:
3137  *
3138  *      Insert the specified bridge node into the route table.
3139  *      Caller has to make sure that rtnode does not exist.
3140  */
3141 static void
3142 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
3143 {
3144         struct bridge_rtnode *lbrt;
3145         uint32_t hash;
3146         int dir;
3147
3148         hash = bridge_rthash(sc, brt->brt_addr);
3149
3150         lbrt = LIST_FIRST(&sc->sc_rthashs[mycpuid][hash]);
3151         if (lbrt == NULL) {
3152                 LIST_INSERT_HEAD(&sc->sc_rthashs[mycpuid][hash], brt, brt_hash);
3153                 goto out;
3154         }
3155
3156         do {
3157                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
3158                 KASSERT(dir != 0, ("rtnode already exist\n"));
3159
3160                 if (dir > 0) {
3161                         LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
3162                         goto out;
3163                 }
3164                 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
3165                         LIST_INSERT_AFTER(lbrt, brt, brt_hash);
3166                         goto out;
3167                 }
3168                 lbrt = LIST_NEXT(lbrt, brt_hash);
3169         } while (lbrt != NULL);
3170
3171         panic("no suitable position found for rtnode\n");
3172 out:
3173         LIST_INSERT_HEAD(&sc->sc_rtlists[mycpuid], brt, brt_list);
3174         if (mycpuid == 0) {
3175                 /*
3176                  * Update the brtcnt.
3177                  * We only need to do it once and we do it on CPU0.
3178                  */
3179                 sc->sc_brtcnt++;
3180         }
3181 }
3182
3183 /*
3184  * bridge_rtnode_destroy:
3185  *
3186  *      Destroy a bridge rtnode.
3187  */
3188 static void
3189 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3190 {
3191         LIST_REMOVE(brt, brt_hash);
3192         LIST_REMOVE(brt, brt_list);
3193
3194         if (mycpuid + 1 == ncpus) {
3195                 /* Free rtinfo associated with rtnode on the last cpu */
3196                 kfree(brt->brt_info, M_DEVBUF);
3197         }
3198         kfree(brt, M_DEVBUF);
3199
3200         if (mycpuid == 0) {
3201                 /* Update brtcnt only on CPU0 */
3202                 sc->sc_brtcnt--;
3203         }
3204 }
3205
3206 static __inline int
3207 bridge_post_pfil(struct mbuf *m)
3208 {
3209         if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED)
3210                 return EOPNOTSUPP;
3211
3212         /* Not yet */
3213         if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED)
3214                 return EOPNOTSUPP;
3215
3216         return 0;
3217 }
3218
3219 /*
3220  * Send bridge packets through pfil if they are one of the types pfil can deal
3221  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3222  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3223  * that interface.
3224  */
3225 static int
3226 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3227 {
3228         int snap, error, i, hlen;
3229         struct ether_header *eh1, eh2;
3230         struct ip *ip;
3231         struct llc llc1;
3232         u_int16_t ether_type;
3233
3234         snap = 0;
3235         error = -1;     /* Default error if not error == 0 */
3236
3237         if (pfil_bridge == 0 && pfil_member == 0)
3238                 return (0); /* filtering is disabled */
3239
3240         i = min((*mp)->m_pkthdr.len, max_protohdr);
3241         if ((*mp)->m_len < i) {
3242                 *mp = m_pullup(*mp, i);
3243                 if (*mp == NULL) {
3244                         kprintf("%s: m_pullup failed\n", __func__);
3245                         return (-1);
3246                 }
3247         }
3248
3249         eh1 = mtod(*mp, struct ether_header *);
3250         ether_type = ntohs(eh1->ether_type);
3251
3252         /*
3253          * Check for SNAP/LLC.
3254          */
3255         if (ether_type < ETHERMTU) {
3256                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3257
3258                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3259                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3260                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3261                     llc2->llc_control == LLC_UI) {
3262                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3263                         snap = 1;
3264                 }
3265         }
3266
3267         /*
3268          * If we're trying to filter bridge traffic, don't look at anything
3269          * other than IP and ARP traffic.  If the filter doesn't understand
3270          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3271          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3272          * but of course we don't have an AppleTalk filter to begin with.
3273          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3274          * ARP traffic.)
3275          */
3276         switch (ether_type) {
3277         case ETHERTYPE_ARP:
3278         case ETHERTYPE_REVARP:
3279                 return (0); /* Automatically pass */
3280
3281         case ETHERTYPE_IP:
3282 #ifdef INET6
3283         case ETHERTYPE_IPV6:
3284 #endif /* INET6 */
3285                 break;
3286
3287         default:
3288                 /*
3289                  * Check to see if the user wants to pass non-ip
3290                  * packets, these will not be checked by pfil(9)
3291                  * and passed unconditionally so the default is to drop.
3292                  */
3293                 if (pfil_onlyip)
3294                         goto bad;
3295         }
3296
3297         /* Strip off the Ethernet header and keep a copy. */
3298         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3299         m_adj(*mp, ETHER_HDR_LEN);
3300
3301         /* Strip off snap header, if present */
3302         if (snap) {
3303                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3304                 m_adj(*mp, sizeof(struct llc));
3305         }
3306
3307         /*
3308          * Check the IP header for alignment and errors
3309          */
3310         if (dir == PFIL_IN) {
3311                 switch (ether_type) {
3312                 case ETHERTYPE_IP:
3313                         error = bridge_ip_checkbasic(mp);
3314                         break;
3315 #ifdef INET6
3316                 case ETHERTYPE_IPV6:
3317                         error = bridge_ip6_checkbasic(mp);
3318                         break;
3319 #endif /* INET6 */
3320                 default:
3321                         error = 0;
3322                 }
3323                 if (error)
3324                         goto bad;
3325         }
3326
3327         error = 0;
3328
3329         /*
3330          * Run the packet through pfil
3331          */
3332         switch (ether_type) {
3333         case ETHERTYPE_IP:
3334                 /*
3335                  * before calling the firewall, swap fields the same as
3336                  * IP does. here we assume the header is contiguous
3337                  */
3338                 ip = mtod(*mp, struct ip *);
3339
3340                 ip->ip_len = ntohs(ip->ip_len);
3341                 ip->ip_off = ntohs(ip->ip_off);
3342
3343                 /*
3344                  * Run pfil on the member interface and the bridge, both can
3345                  * be skipped by clearing pfil_member or pfil_bridge.
3346                  *
3347                  * Keep the order:
3348                  *   in_if -> bridge_if -> out_if
3349                  */
3350                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL) {
3351                         error = pfil_run_hooks(&inet_pfil_hook, mp, bifp, dir);
3352                         if (*mp == NULL || error != 0) /* filter may consume */
3353                                 break;
3354                         error = bridge_post_pfil(*mp);
3355                         if (error)
3356                                 break;
3357                 }
3358
3359                 if (pfil_member && ifp != NULL) {
3360                         error = pfil_run_hooks(&inet_pfil_hook, mp, ifp, dir);
3361                         if (*mp == NULL || error != 0) /* filter may consume */
3362                                 break;
3363                         error = bridge_post_pfil(*mp);
3364                         if (error)
3365                                 break;
3366                 }
3367
3368                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL) {
3369                         error = pfil_run_hooks(&inet_pfil_hook, mp, bifp, dir);
3370                         if (*mp == NULL || error != 0) /* filter may consume */
3371                                 break;
3372                         error = bridge_post_pfil(*mp);
3373                         if (error)
3374                                 break;
3375                 }
3376
3377                 /* check if we need to fragment the packet */
3378                 if (pfil_member && ifp != NULL && dir == PFIL_OUT) {
3379                         i = (*mp)->m_pkthdr.len;
3380                         if (i > ifp->if_mtu) {
3381                                 error = bridge_fragment(ifp, *mp, &eh2, snap,
3382                                             &llc1);
3383                                 return (error);
3384                         }
3385                 }
3386
3387                 /* Recalculate the ip checksum and restore byte ordering */
3388                 ip = mtod(*mp, struct ip *);
3389                 hlen = ip->ip_hl << 2;
3390                 if (hlen < sizeof(struct ip))
3391                         goto bad;
3392                 if (hlen > (*mp)->m_len) {
3393                         if ((*mp = m_pullup(*mp, hlen)) == 0)
3394                                 goto bad;
3395                         ip = mtod(*mp, struct ip *);
3396                         if (ip == NULL)
3397                                 goto bad;
3398                 }
3399                 ip->ip_len = htons(ip->ip_len);
3400                 ip->ip_off = htons(ip->ip_off);
3401                 ip->ip_sum = 0;
3402                 if (hlen == sizeof(struct ip))
3403                         ip->ip_sum = in_cksum_hdr(ip);
3404                 else
3405                         ip->ip_sum = in_cksum(*mp, hlen);
3406
3407                 break;
3408 #ifdef INET6
3409         case ETHERTYPE_IPV6:
3410                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3411                         error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp,
3412                                         dir);
3413
3414                 if (*mp == NULL || error != 0) /* filter may consume */
3415                         break;
3416
3417                 if (pfil_member && ifp != NULL)
3418                         error = pfil_run_hooks(&inet6_pfil_hook, mp, ifp,
3419                                         dir);
3420
3421                 if (*mp == NULL || error != 0) /* filter may consume */
3422                         break;
3423
3424                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3425                         error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp,
3426                                         dir);
3427                 break;
3428 #endif
3429         default:
3430                 error = 0;
3431                 break;
3432         }
3433
3434         if (*mp == NULL)
3435                 return (error);
3436         if (error != 0)
3437                 goto bad;
3438
3439         error = -1;
3440
3441         /*
3442          * Finally, put everything back the way it was and return
3443          */
3444         if (snap) {
3445                 M_PREPEND(*mp, sizeof(struct llc), MB_DONTWAIT);
3446                 if (*mp == NULL)
3447                         return (error);
3448                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3449         }
3450
3451         M_PREPEND(*mp, ETHER_HDR_LEN, MB_DONTWAIT);
3452         if (*mp == NULL)
3453                 return (error);
3454         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3455
3456         return (0);
3457
3458 bad:
3459         m_freem(*mp);
3460         *mp = NULL;
3461         return (error);
3462 }
3463
3464 /*
3465  * Perform basic checks on header size since
3466  * pfil assumes ip_input has already processed
3467  * it for it.  Cut-and-pasted from ip_input.c.
3468  * Given how simple the IPv6 version is,
3469  * does the IPv4 version really need to be
3470  * this complicated?
3471  *
3472  * XXX Should we update ipstat here, or not?
3473  * XXX Right now we update ipstat but not
3474  * XXX csum_counter.
3475  */
3476 static int
3477 bridge_ip_checkbasic(struct mbuf **mp)
3478 {
3479         struct mbuf *m = *mp;
3480         struct ip *ip;
3481         int len, hlen;
3482         u_short sum;
3483
3484         if (*mp == NULL)
3485                 return (-1);
3486 #if notyet
3487         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3488                 if ((m = m_copyup(m, sizeof(struct ip),
3489                         (max_linkhdr + 3) & ~3)) == NULL) {
3490                         /* XXXJRT new stat, please */
3491                         ipstat.ips_toosmall++;
3492                         goto bad;
3493                 }
3494         } else
3495 #endif
3496 #ifndef __predict_false
3497 #define __predict_false(x) x
3498 #endif
3499          if (__predict_false(m->m_len < sizeof (struct ip))) {
3500                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3501                         ipstat.ips_toosmall++;
3502                         goto bad;
3503                 }
3504         }
3505         ip = mtod(m, struct ip *);
3506         if (ip == NULL) goto bad;
3507
3508         if (ip->ip_v != IPVERSION) {
3509                 ipstat.ips_badvers++;
3510                 goto bad;
3511         }
3512         hlen = ip->ip_hl << 2;
3513         if (hlen < sizeof(struct ip)) { /* minimum header length */
3514                 ipstat.ips_badhlen++;
3515                 goto bad;
3516         }
3517         if (hlen > m->m_len) {
3518                 if ((m = m_pullup(m, hlen)) == 0) {
3519                         ipstat.ips_badhlen++;
3520                         goto bad;
3521                 }
3522                 ip = mtod(m, struct ip *);
3523                 if (ip == NULL) goto bad;
3524         }
3525
3526         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3527                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3528         } else {
3529                 if (hlen == sizeof(struct ip)) {
3530                         sum = in_cksum_hdr(ip);
3531                 } else {
3532                         sum = in_cksum(m, hlen);
3533                 }
3534         }
3535         if (sum) {
3536                 ipstat.ips_badsum++;
3537                 goto bad;
3538         }
3539
3540         /* Retrieve the packet length. */
3541         len = ntohs(ip->ip_len);
3542
3543         /*
3544          * Check for additional length bogosity
3545          */
3546         if (len < hlen) {
3547                 ipstat.ips_badlen++;
3548                 goto bad;
3549         }
3550
3551         /*
3552          * Check that the amount of data in the buffers
3553          * is as at least much as the IP header would have us expect.
3554          * Drop packet if shorter than we expect.
3555          */
3556         if (m->m_pkthdr.len < len) {
3557                 ipstat.ips_tooshort++;
3558                 goto bad;
3559         }
3560
3561         /* Checks out, proceed */
3562         *mp = m;
3563         return (0);
3564
3565 bad:
3566         *mp = m;
3567         return (-1);
3568 }
3569
3570 #ifdef INET6
3571 /*
3572  * Same as above, but for IPv6.
3573  * Cut-and-pasted from ip6_input.c.
3574  * XXX Should we update ip6stat, or not?
3575  */
3576 static int
3577 bridge_ip6_checkbasic(struct mbuf **mp)
3578 {
3579         struct mbuf *m = *mp;
3580         struct ip6_hdr *ip6;
3581
3582         /*
3583          * If the IPv6 header is not aligned, slurp it up into a new
3584          * mbuf with space for link headers, in the event we forward
3585          * it.  Otherwise, if it is aligned, make sure the entire base
3586          * IPv6 header is in the first mbuf of the chain.
3587          */
3588 #if notyet
3589         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3590                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3591                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3592                             (max_linkhdr + 3) & ~3)) == NULL) {
3593                         /* XXXJRT new stat, please */
3594                         ip6stat.ip6s_toosmall++;
3595                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3596                         goto bad;
3597                 }
3598         } else
3599 #endif
3600         if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3601                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3602                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3603                         ip6stat.ip6s_toosmall++;
3604                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3605                         goto bad;
3606                 }
3607         }
3608
3609         ip6 = mtod(m, struct ip6_hdr *);
3610
3611         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3612                 ip6stat.ip6s_badvers++;
3613                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3614                 goto bad;
3615         }
3616
3617         /* Checks out, proceed */
3618         *mp = m;
3619         return (0);
3620
3621 bad:
3622         *mp = m;
3623         return (-1);
3624 }
3625 #endif /* INET6 */
3626
3627 /*
3628  * bridge_fragment:
3629  *
3630  *      Return a fragmented mbuf chain.
3631  */
3632 static int
3633 bridge_fragment(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh,
3634     int snap, struct llc *llc)
3635 {
3636         struct mbuf *m0;
3637         struct ip *ip;
3638         int error = -1;
3639
3640         if (m->m_len < sizeof(struct ip) &&
3641             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3642                 goto out;
3643         ip = mtod(m, struct ip *);
3644
3645         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist,
3646                     CSUM_DELAY_IP);
3647         if (error)
3648                 goto out;
3649
3650         /* walk the chain and re-add the Ethernet header */
3651         for (m0 = m; m0; m0 = m0->m_nextpkt) {
3652                 if (error == 0) {
3653                         if (snap) {
3654                                 M_PREPEND(m0, sizeof(struct llc), MB_DONTWAIT);
3655                                 if (m0 == NULL) {
3656                                         error = ENOBUFS;
3657                                         continue;
3658                                 }
3659                                 bcopy(llc, mtod(m0, caddr_t),
3660                                     sizeof(struct llc));
3661                         }
3662                         M_PREPEND(m0, ETHER_HDR_LEN, MB_DONTWAIT);
3663                         if (m0 == NULL) {
3664                                 error = ENOBUFS;
3665                                 continue;
3666                         }
3667                         bcopy(eh, mtod(m0, caddr_t), ETHER_HDR_LEN);
3668                 } else 
3669                         m_freem(m);
3670         }
3671
3672         if (error == 0)
3673                 ipstat.ips_fragmented++;
3674
3675         return (error);
3676
3677 out:
3678         if (m != NULL)
3679                 m_freem(m);
3680         return (error);
3681 }
3682
3683 static void
3684 bridge_enqueue_handler(netmsg_t msg)
3685 {
3686         struct netmsg_packet *nmp;
3687         struct ifnet *dst_ifp;
3688         struct mbuf *m;
3689
3690         nmp = &msg->packet;
3691         m = nmp->nm_packet;
3692         dst_ifp = nmp->base.lmsg.u.ms_resultp;
3693
3694         bridge_handoff(dst_ifp, m, 1);
3695 }
3696
3697 static void
3698 bridge_handoff(struct ifnet *dst_ifp, struct mbuf *m, int from_us)
3699 {
3700         struct mbuf *m0;
3701         struct ifnet *bifp;
3702
3703         bifp = ((struct bridge_softc *)dst_ifp->if_bridge)->sc_ifp;
3704
3705         /* We may be sending a fragment so traverse the mbuf */
3706         for (; m; m = m0) {
3707                 struct altq_pktattr pktattr;
3708
3709                 m0 = m->m_nextpkt;
3710                 m->m_nextpkt = NULL;
3711
3712                 /*
3713                  * If being sent from our host override ether_shost
3714                  * so any replies go the correct interface.  This is
3715                  * mandatory or ARP replies will wind up on the wrong
3716                  * interface.
3717                  *
3718                  * Otherwise if we are in transparent mode
3719                  */
3720                 if (from_us) {
3721                         m_copyback(m,
3722                                    offsetof(struct ether_header, ether_shost),
3723                                    ETHER_ADDR_LEN, IF_LLADDR(dst_ifp));
3724                 } else if ((bifp->if_flags & IFF_LINK0) &&
3725                            (m->m_pkthdr.fw_flags & BRIDGE_MBUF_TAGGED)) {
3726                         m_copyback(m,
3727                                    offsetof(struct ether_header, ether_shost),
3728                                    ETHER_ADDR_LEN,
3729                                    m->m_pkthdr.br.ether.ether_shost);
3730                 }
3731
3732                 if (ifq_is_enabled(&dst_ifp->if_snd))
3733                         altq_etherclassify(&dst_ifp->if_snd, m, &pktattr);
3734
3735                 ifq_dispatch(dst_ifp, m, &pktattr);
3736         }
3737 }
3738
3739 static void
3740 bridge_control_dispatch(netmsg_t msg)
3741 {
3742         struct netmsg_brctl *bc_msg = (struct netmsg_brctl *)msg;
3743         struct ifnet *bifp = bc_msg->bc_sc->sc_ifp;
3744         int error;
3745
3746         ifnet_serialize_all(bifp);
3747         error = bc_msg->bc_func(bc_msg->bc_sc, bc_msg->bc_arg);
3748         ifnet_deserialize_all(bifp);
3749
3750         lwkt_replymsg(&bc_msg->base.lmsg, error);
3751 }
3752
3753 static int
3754 bridge_control(struct bridge_softc *sc, u_long cmd,
3755                bridge_ctl_t bc_func, void *bc_arg)
3756 {
3757         struct ifnet *bifp = sc->sc_ifp;
3758         struct netmsg_brctl bc_msg;
3759         int error;
3760
3761         ASSERT_IFNET_SERIALIZED_ALL(bifp);
3762
3763         bzero(&bc_msg, sizeof(bc_msg));
3764
3765         netmsg_init(&bc_msg.base, NULL, &curthread->td_msgport,
3766                     0, bridge_control_dispatch);
3767         bc_msg.bc_func = bc_func;
3768         bc_msg.bc_sc = sc;
3769         bc_msg.bc_arg = bc_arg;
3770
3771         ifnet_deserialize_all(bifp);
3772         error = lwkt_domsg(BRIDGE_CFGPORT, &bc_msg.base.lmsg, 0);
3773         ifnet_serialize_all(bifp);
3774         return error;
3775 }
3776
3777 static void
3778 bridge_add_bif_handler(netmsg_t msg)
3779 {
3780         struct netmsg_braddbif *amsg = (struct netmsg_braddbif *)msg;
3781         struct bridge_softc *sc;
3782         struct bridge_iflist *bif;
3783
3784         sc = amsg->br_softc;
3785
3786         bif = kmalloc(sizeof(*bif), M_DEVBUF, M_WAITOK | M_ZERO);
3787         bif->bif_ifp = amsg->br_bif_ifp;
3788         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
3789         bif->bif_onlist = 1;
3790         bif->bif_info = amsg->br_bif_info;
3791
3792         LIST_INSERT_HEAD(&sc->sc_iflists[mycpuid], bif, bif_next);
3793
3794         ifnet_forwardmsg(&amsg->base.lmsg, mycpuid + 1);
3795 }
3796
3797 static void
3798 bridge_add_bif(struct bridge_softc *sc, struct bridge_ifinfo *bif_info,
3799                struct ifnet *ifp)
3800 {
3801         struct netmsg_braddbif amsg;
3802
3803         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
3804
3805         netmsg_init(&amsg.base, NULL, &curthread->td_msgport,
3806                     0, bridge_add_bif_handler);
3807         amsg.br_softc = sc;
3808         amsg.br_bif_info = bif_info;
3809         amsg.br_bif_ifp = ifp;
3810
3811         ifnet_domsg(&amsg.base.lmsg, 0);
3812 }
3813
3814 static void
3815 bridge_del_bif_handler(netmsg_t msg)
3816 {
3817         struct netmsg_brdelbif *dmsg = (struct netmsg_brdelbif *)msg;
3818         struct bridge_softc *sc;
3819         struct bridge_iflist *bif;
3820
3821         sc = dmsg->br_softc;
3822
3823         /*
3824          * Locate the bif associated with the br_bif_info
3825          * on the current CPU
3826          */
3827         bif = bridge_lookup_member_ifinfo(sc, dmsg->br_bif_info);
3828         KKASSERT(bif != NULL && bif->bif_onlist);
3829
3830         /* Remove the bif from the current CPU's iflist */
3831         bif->bif_onlist = 0;
3832         LIST_REMOVE(bif, bif_next);
3833
3834         /* Save the removed bif for later freeing */
3835         LIST_INSERT_HEAD(dmsg->br_bif_list, bif, bif_next);
3836
3837         ifnet_forwardmsg(&dmsg->base.lmsg, mycpuid + 1);
3838 }
3839
3840 static void
3841 bridge_del_bif(struct bridge_softc *sc, struct bridge_ifinfo *bif_info,
3842                struct bridge_iflist_head *saved_bifs)
3843 {
3844         struct netmsg_brdelbif dmsg;
3845
3846         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
3847
3848         netmsg_init(&dmsg.base, NULL, &curthread->td_msgport,
3849                     0, bridge_del_bif_handler);
3850         dmsg.br_softc = sc;
3851         dmsg.br_bif_info = bif_info;
3852         dmsg.br_bif_list = saved_bifs;
3853
3854         ifnet_domsg(&dmsg.base.lmsg, 0);
3855 }
3856
3857 static void
3858 bridge_set_bifflags_handler(netmsg_t msg)
3859 {
3860         struct netmsg_brsflags *smsg = (struct netmsg_brsflags *)msg;
3861         struct bridge_softc *sc;
3862         struct bridge_iflist *bif;
3863
3864         sc = smsg->br_softc;
3865
3866         /*
3867          * Locate the bif associated with the br_bif_info
3868          * on the current CPU
3869          */
3870         bif = bridge_lookup_member_ifinfo(sc, smsg->br_bif_info);
3871         KKASSERT(bif != NULL && bif->bif_onlist);
3872
3873         bif->bif_flags = smsg->br_bif_flags;
3874
3875         ifnet_forwardmsg(&smsg->base.lmsg, mycpuid + 1);
3876 }
3877
3878 static void
3879 bridge_set_bifflags(struct bridge_softc *sc, struct bridge_ifinfo *bif_info,
3880                     uint32_t bif_flags)
3881 {
3882         struct netmsg_brsflags smsg;
3883
3884         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
3885
3886         netmsg_init(&smsg.base, NULL, &curthread->td_msgport,
3887                     0, bridge_set_bifflags_handler);
3888         smsg.br_softc = sc;
3889         smsg.br_bif_info = bif_info;
3890         smsg.br_bif_flags = bif_flags;
3891
3892         ifnet_domsg(&smsg.base.lmsg, 0);
3893 }