if_bridge.c revision 1.89 1 /* $NetBSD: if_bridge.c,v 1.89 2014/07/23 04:09:48 ozaki-r Exp $ */
2
3 /*
4 * Copyright 2001 Wasabi Systems, Inc.
5 * All rights reserved.
6 *
7 * Written by Jason R. Thorpe for Wasabi Systems, Inc.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. All advertising materials mentioning features or use of this software
18 * must display the following acknowledgement:
19 * This product includes software developed for the NetBSD Project by
20 * Wasabi Systems, Inc.
21 * 4. The name of Wasabi Systems, Inc. may not be used to endorse
22 * or promote products derived from this software without specific prior
23 * written permission.
24 *
25 * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
27 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
28 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL WASABI SYSTEMS, INC
29 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
32 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
33 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
34 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
35 * POSSIBILITY OF SUCH DAMAGE.
36 */
37
38 /*
39 * Copyright (c) 1999, 2000 Jason L. Wright (jason (at) thought.net)
40 * All rights reserved.
41 *
42 * Redistribution and use in source and binary forms, with or without
43 * modification, are permitted provided that the following conditions
44 * are met:
45 * 1. Redistributions of source code must retain the above copyright
46 * notice, this list of conditions and the following disclaimer.
47 * 2. Redistributions in binary form must reproduce the above copyright
48 * notice, this list of conditions and the following disclaimer in the
49 * documentation and/or other materials provided with the distribution.
50 * 3. All advertising materials mentioning features or use of this software
51 * must display the following acknowledgement:
52 * This product includes software developed by Jason L. Wright
53 * 4. The name of the author may not be used to endorse or promote products
54 * derived from this software without specific prior written permission.
55 *
56 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
57 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
58 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
59 * DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
60 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
61 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
62 * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
63 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
64 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
65 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
66 * POSSIBILITY OF SUCH DAMAGE.
67 *
68 * OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun 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 #include <sys/cdefs.h>
83 __KERNEL_RCSID(0, "$NetBSD: if_bridge.c,v 1.89 2014/07/23 04:09:48 ozaki-r Exp $");
84
85 #ifdef _KERNEL_OPT
86 #include "opt_bridge_ipf.h"
87 #include "opt_inet.h"
88 #endif /* _KERNEL_OPT */
89
90 #include <sys/param.h>
91 #include <sys/kernel.h>
92 #include <sys/mbuf.h>
93 #include <sys/queue.h>
94 #include <sys/socket.h>
95 #include <sys/socketvar.h> /* for softnet_lock */
96 #include <sys/sockio.h>
97 #include <sys/systm.h>
98 #include <sys/proc.h>
99 #include <sys/pool.h>
100 #include <sys/kauth.h>
101 #include <sys/cpu.h>
102 #include <sys/cprng.h>
103 #include <sys/mutex.h>
104
105 #include <net/bpf.h>
106 #include <net/if.h>
107 #include <net/if_dl.h>
108 #include <net/if_types.h>
109 #include <net/if_llc.h>
110 #include <net/pktqueue.h>
111
112 #include <net/if_ether.h>
113 #include <net/if_bridgevar.h>
114
115 #if defined(BRIDGE_IPF)
116 /* Used for bridge_ip[6]_checkbasic */
117 #include <netinet/in.h>
118 #include <netinet/in_systm.h>
119 #include <netinet/ip.h>
120 #include <netinet/ip_var.h>
121 #include <netinet/ip_private.h> /* XXX */
122
123 #include <netinet/ip6.h>
124 #include <netinet6/in6_var.h>
125 #include <netinet6/ip6_var.h>
126 #include <netinet6/ip6_private.h> /* XXX */
127 #endif /* BRIDGE_IPF */
128
129 /*
130 * Size of the route hash table. Must be a power of two.
131 */
132 #ifndef BRIDGE_RTHASH_SIZE
133 #define BRIDGE_RTHASH_SIZE 1024
134 #endif
135
136 #define BRIDGE_RTHASH_MASK (BRIDGE_RTHASH_SIZE - 1)
137
138 #include "carp.h"
139 #if NCARP > 0
140 #include <netinet/in.h>
141 #include <netinet/in_var.h>
142 #include <netinet/ip_carp.h>
143 #endif
144
145 /*
146 * Maximum number of addresses to cache.
147 */
148 #ifndef BRIDGE_RTABLE_MAX
149 #define BRIDGE_RTABLE_MAX 100
150 #endif
151
152 /*
153 * Spanning tree defaults.
154 */
155 #define BSTP_DEFAULT_MAX_AGE (20 * 256)
156 #define BSTP_DEFAULT_HELLO_TIME (2 * 256)
157 #define BSTP_DEFAULT_FORWARD_DELAY (15 * 256)
158 #define BSTP_DEFAULT_HOLD_TIME (1 * 256)
159 #define BSTP_DEFAULT_BRIDGE_PRIORITY 0x8000
160 #define BSTP_DEFAULT_PORT_PRIORITY 0x80
161 #define BSTP_DEFAULT_PATH_COST 55
162
163 /*
164 * Timeout (in seconds) for entries learned dynamically.
165 */
166 #ifndef BRIDGE_RTABLE_TIMEOUT
167 #define BRIDGE_RTABLE_TIMEOUT (20 * 60) /* same as ARP */
168 #endif
169
170 /*
171 * Number of seconds between walks of the route list.
172 */
173 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
174 #define BRIDGE_RTABLE_PRUNE_PERIOD (5 * 60)
175 #endif
176
177 int bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
178
179 static struct pool bridge_rtnode_pool;
180
181 void bridgeattach(int);
182
183 static int bridge_clone_create(struct if_clone *, int);
184 static int bridge_clone_destroy(struct ifnet *);
185
186 static int bridge_ioctl(struct ifnet *, u_long, void *);
187 static int bridge_init(struct ifnet *);
188 static void bridge_stop(struct ifnet *, int);
189 static void bridge_start(struct ifnet *);
190
191 static void bridge_input(struct ifnet *, struct mbuf *);
192 static void bridge_forward(void *);
193
194 static void bridge_timer(void *);
195
196 static void bridge_broadcast(struct bridge_softc *, struct ifnet *,
197 struct mbuf *);
198
199 static int bridge_rtupdate(struct bridge_softc *, const uint8_t *,
200 struct ifnet *, int, uint8_t);
201 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *);
202 static void bridge_rttrim(struct bridge_softc *);
203 static void bridge_rtage(struct bridge_softc *);
204 static void bridge_rtflush(struct bridge_softc *, int);
205 static int bridge_rtdaddr(struct bridge_softc *, const uint8_t *);
206 static void bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp);
207
208 static int bridge_rtable_init(struct bridge_softc *);
209 static void bridge_rtable_fini(struct bridge_softc *);
210
211 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
212 const uint8_t *);
213 static int bridge_rtnode_insert(struct bridge_softc *,
214 struct bridge_rtnode *);
215 static void bridge_rtnode_destroy(struct bridge_softc *,
216 struct bridge_rtnode *);
217
218 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
219 const char *name);
220 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
221 struct ifnet *ifp);
222 static void bridge_release_member(struct bridge_softc *, struct bridge_iflist *);
223 static void bridge_delete_member(struct bridge_softc *,
224 struct bridge_iflist *);
225
226 static int bridge_ioctl_add(struct bridge_softc *, void *);
227 static int bridge_ioctl_del(struct bridge_softc *, void *);
228 static int bridge_ioctl_gifflags(struct bridge_softc *, void *);
229 static int bridge_ioctl_sifflags(struct bridge_softc *, void *);
230 static int bridge_ioctl_scache(struct bridge_softc *, void *);
231 static int bridge_ioctl_gcache(struct bridge_softc *, void *);
232 static int bridge_ioctl_gifs(struct bridge_softc *, void *);
233 static int bridge_ioctl_rts(struct bridge_softc *, void *);
234 static int bridge_ioctl_saddr(struct bridge_softc *, void *);
235 static int bridge_ioctl_sto(struct bridge_softc *, void *);
236 static int bridge_ioctl_gto(struct bridge_softc *, void *);
237 static int bridge_ioctl_daddr(struct bridge_softc *, void *);
238 static int bridge_ioctl_flush(struct bridge_softc *, void *);
239 static int bridge_ioctl_gpri(struct bridge_softc *, void *);
240 static int bridge_ioctl_spri(struct bridge_softc *, void *);
241 static int bridge_ioctl_ght(struct bridge_softc *, void *);
242 static int bridge_ioctl_sht(struct bridge_softc *, void *);
243 static int bridge_ioctl_gfd(struct bridge_softc *, void *);
244 static int bridge_ioctl_sfd(struct bridge_softc *, void *);
245 static int bridge_ioctl_gma(struct bridge_softc *, void *);
246 static int bridge_ioctl_sma(struct bridge_softc *, void *);
247 static int bridge_ioctl_sifprio(struct bridge_softc *, void *);
248 static int bridge_ioctl_sifcost(struct bridge_softc *, void *);
249 #if defined(BRIDGE_IPF)
250 static int bridge_ioctl_gfilt(struct bridge_softc *, void *);
251 static int bridge_ioctl_sfilt(struct bridge_softc *, void *);
252 static int bridge_ipf(void *, struct mbuf **, struct ifnet *, int);
253 static int bridge_ip_checkbasic(struct mbuf **mp);
254 # ifdef INET6
255 static int bridge_ip6_checkbasic(struct mbuf **mp);
256 # endif /* INET6 */
257 #endif /* BRIDGE_IPF */
258
259 static void bridge_sysctl_fwdq_setup(struct sysctllog **clog,
260 struct bridge_softc *sc);
261
262 struct bridge_control {
263 int (*bc_func)(struct bridge_softc *, void *);
264 int bc_argsize;
265 int bc_flags;
266 };
267
268 #define BC_F_COPYIN 0x01 /* copy arguments in */
269 #define BC_F_COPYOUT 0x02 /* copy arguments out */
270 #define BC_F_SUSER 0x04 /* do super-user check */
271
272 static const struct bridge_control bridge_control_table[] = {
273 [BRDGADD] = {bridge_ioctl_add, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
274 [BRDGDEL] = {bridge_ioctl_del, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
275
276 [BRDGGIFFLGS] = {bridge_ioctl_gifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_COPYOUT},
277 [BRDGSIFFLGS] = {bridge_ioctl_sifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
278
279 [BRDGSCACHE] = {bridge_ioctl_scache, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
280 [BRDGGCACHE] = {bridge_ioctl_gcache, sizeof(struct ifbrparam), BC_F_COPYOUT},
281
282 [BRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_COPYIN|BC_F_COPYOUT},
283 [BRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_COPYIN|BC_F_COPYOUT},
284
285 [BRDGSADDR] = {bridge_ioctl_saddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
286
287 [BRDGSTO] = {bridge_ioctl_sto, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
288 [BRDGGTO] = {bridge_ioctl_gto, sizeof(struct ifbrparam), BC_F_COPYOUT},
289
290 [BRDGDADDR] = {bridge_ioctl_daddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
291
292 [BRDGFLUSH] = {bridge_ioctl_flush, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
293
294 [BRDGGPRI] = {bridge_ioctl_gpri, sizeof(struct ifbrparam), BC_F_COPYOUT},
295 [BRDGSPRI] = {bridge_ioctl_spri, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
296
297 [BRDGGHT] = {bridge_ioctl_ght, sizeof(struct ifbrparam), BC_F_COPYOUT},
298 [BRDGSHT] = {bridge_ioctl_sht, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
299
300 [BRDGGFD] = {bridge_ioctl_gfd, sizeof(struct ifbrparam), BC_F_COPYOUT},
301 [BRDGSFD] = {bridge_ioctl_sfd, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
302
303 [BRDGGMA] = {bridge_ioctl_gma, sizeof(struct ifbrparam), BC_F_COPYOUT},
304 [BRDGSMA] = {bridge_ioctl_sma, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
305
306 [BRDGSIFPRIO] = {bridge_ioctl_sifprio, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
307
308 [BRDGSIFCOST] = {bridge_ioctl_sifcost, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
309 #if defined(BRIDGE_IPF)
310 [BRDGGFILT] = {bridge_ioctl_gfilt, sizeof(struct ifbrparam), BC_F_COPYOUT},
311 [BRDGSFILT] = {bridge_ioctl_sfilt, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
312 #endif /* BRIDGE_IPF */
313 };
314 static const int bridge_control_table_size = __arraycount(bridge_control_table);
315
316 static LIST_HEAD(, bridge_softc) bridge_list;
317 static kmutex_t bridge_list_lock;
318
319 static struct if_clone bridge_cloner =
320 IF_CLONE_INITIALIZER("bridge", bridge_clone_create, bridge_clone_destroy);
321
322 /*
323 * bridgeattach:
324 *
325 * Pseudo-device attach routine.
326 */
327 void
328 bridgeattach(int n)
329 {
330
331 pool_init(&bridge_rtnode_pool, sizeof(struct bridge_rtnode),
332 0, 0, 0, "brtpl", NULL, IPL_NET);
333
334 LIST_INIT(&bridge_list);
335 mutex_init(&bridge_list_lock, MUTEX_DEFAULT, IPL_NET);
336 if_clone_attach(&bridge_cloner);
337 }
338
339 /*
340 * bridge_clone_create:
341 *
342 * Create a new bridge instance.
343 */
344 static int
345 bridge_clone_create(struct if_clone *ifc, int unit)
346 {
347 struct bridge_softc *sc;
348 struct ifnet *ifp;
349
350 sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
351 ifp = &sc->sc_if;
352
353 sc->sc_brtmax = BRIDGE_RTABLE_MAX;
354 sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
355 sc->sc_bridge_max_age = BSTP_DEFAULT_MAX_AGE;
356 sc->sc_bridge_hello_time = BSTP_DEFAULT_HELLO_TIME;
357 sc->sc_bridge_forward_delay = BSTP_DEFAULT_FORWARD_DELAY;
358 sc->sc_bridge_priority = BSTP_DEFAULT_BRIDGE_PRIORITY;
359 sc->sc_hold_time = BSTP_DEFAULT_HOLD_TIME;
360 sc->sc_filter_flags = 0;
361
362 /* Initialize our routing table. */
363 bridge_rtable_init(sc);
364
365 callout_init(&sc->sc_brcallout, 0);
366 callout_init(&sc->sc_bstpcallout, 0);
367
368 LIST_INIT(&sc->sc_iflist);
369 #ifdef BRIDGE_MPSAFE
370 sc->sc_iflist_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
371 #else
372 sc->sc_iflist_lock = NULL;
373 #endif
374 cv_init(&sc->sc_iflist_cv, "if_bridge_cv");
375
376 if_initname(ifp, ifc->ifc_name, unit);
377 ifp->if_softc = sc;
378 ifp->if_mtu = ETHERMTU;
379 ifp->if_ioctl = bridge_ioctl;
380 ifp->if_output = bridge_output;
381 ifp->if_start = bridge_start;
382 ifp->if_stop = bridge_stop;
383 ifp->if_init = bridge_init;
384 ifp->if_type = IFT_BRIDGE;
385 ifp->if_addrlen = 0;
386 ifp->if_dlt = DLT_EN10MB;
387 ifp->if_hdrlen = ETHER_HDR_LEN;
388
389 sc->sc_fwd_pktq = pktq_create(IFQ_MAXLEN, bridge_forward, sc);
390 KASSERT(sc->sc_fwd_pktq != NULL);
391
392 bridge_sysctl_fwdq_setup(&ifp->if_sysctl_log, sc);
393
394 if_attach(ifp);
395
396 if_alloc_sadl(ifp);
397
398 mutex_enter(&bridge_list_lock);
399 LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
400 mutex_exit(&bridge_list_lock);
401
402 return (0);
403 }
404
405 /*
406 * bridge_clone_destroy:
407 *
408 * Destroy a bridge instance.
409 */
410 static int
411 bridge_clone_destroy(struct ifnet *ifp)
412 {
413 struct bridge_softc *sc = ifp->if_softc;
414 struct bridge_iflist *bif;
415 int s;
416
417 /* Must be called during IFF_RUNNING, i.e., before bridge_stop */
418 pktq_barrier(sc->sc_fwd_pktq);
419
420 s = splnet();
421
422 bridge_stop(ifp, 1);
423
424 BRIDGE_LOCK(sc);
425 while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
426 bridge_delete_member(sc, bif);
427 BRIDGE_UNLOCK(sc);
428
429 mutex_enter(&bridge_list_lock);
430 LIST_REMOVE(sc, sc_list);
431 mutex_exit(&bridge_list_lock);
432
433 splx(s);
434
435 if_detach(ifp);
436
437 /* Should be called after if_detach for safe */
438 pktq_flush(sc->sc_fwd_pktq);
439 pktq_destroy(sc->sc_fwd_pktq);
440
441 /* Tear down the routing table. */
442 bridge_rtable_fini(sc);
443
444 cv_destroy(&sc->sc_iflist_cv);
445 if (sc->sc_iflist_lock)
446 mutex_obj_free(sc->sc_iflist_lock);
447
448 free(sc, M_DEVBUF);
449
450 return (0);
451 }
452
453 static int
454 bridge_sysctl_fwdq_maxlen(SYSCTLFN_ARGS)
455 {
456 struct sysctlnode node = *rnode;
457 const struct bridge_softc *sc = node.sysctl_data;
458 return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), sc->sc_fwd_pktq);
459 }
460
461 #define SYSCTL_BRIDGE_PKTQ(cn, c) \
462 static int \
463 bridge_sysctl_fwdq_##cn(SYSCTLFN_ARGS) \
464 { \
465 struct sysctlnode node = *rnode; \
466 const struct bridge_softc *sc = node.sysctl_data; \
467 return sysctl_pktq_count(SYSCTLFN_CALL(rnode), \
468 sc->sc_fwd_pktq, c); \
469 }
470
471 SYSCTL_BRIDGE_PKTQ(items, PKTQ_NITEMS)
472 SYSCTL_BRIDGE_PKTQ(drops, PKTQ_DROPS)
473
474 static void
475 bridge_sysctl_fwdq_setup(struct sysctllog **clog, struct bridge_softc *sc)
476 {
477 const struct sysctlnode *cnode, *rnode;
478 sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL;
479 const char *ifname = sc->sc_if.if_xname;
480
481 len_func = bridge_sysctl_fwdq_items;
482 maxlen_func = bridge_sysctl_fwdq_maxlen;
483 drops_func = bridge_sysctl_fwdq_drops;
484
485 if (sysctl_createv(clog, 0, NULL, &rnode,
486 CTLFLAG_PERMANENT,
487 CTLTYPE_NODE, "interfaces",
488 SYSCTL_DESCR("Per-interface controls"),
489 NULL, 0, NULL, 0,
490 CTL_NET, CTL_CREATE, CTL_EOL) != 0)
491 goto bad;
492
493 if (sysctl_createv(clog, 0, &rnode, &rnode,
494 CTLFLAG_PERMANENT,
495 CTLTYPE_NODE, ifname,
496 SYSCTL_DESCR("Interface controls"),
497 NULL, 0, NULL, 0,
498 CTL_CREATE, CTL_EOL) != 0)
499 goto bad;
500
501 if (sysctl_createv(clog, 0, &rnode, &rnode,
502 CTLFLAG_PERMANENT,
503 CTLTYPE_NODE, "fwdq",
504 SYSCTL_DESCR("Protocol input queue controls"),
505 NULL, 0, NULL, 0,
506 CTL_CREATE, CTL_EOL) != 0)
507 goto bad;
508
509 if (sysctl_createv(clog, 0, &rnode, &cnode,
510 CTLFLAG_PERMANENT,
511 CTLTYPE_INT, "len",
512 SYSCTL_DESCR("Current forwarding queue length"),
513 len_func, 0, (void *)sc, 0,
514 CTL_CREATE, IFQCTL_LEN, CTL_EOL) != 0)
515 goto bad;
516
517 if (sysctl_createv(clog, 0, &rnode, &cnode,
518 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
519 CTLTYPE_INT, "maxlen",
520 SYSCTL_DESCR("Maximum allowed forwarding queue length"),
521 maxlen_func, 0, (void *)sc, 0,
522 CTL_CREATE, IFQCTL_MAXLEN, CTL_EOL) != 0)
523 goto bad;
524
525 if (sysctl_createv(clog, 0, &rnode, &cnode,
526 CTLFLAG_PERMANENT,
527 CTLTYPE_INT, "drops",
528 SYSCTL_DESCR("Packets dropped due to full forwarding queue"),
529 drops_func, 0, (void *)sc, 0,
530 CTL_CREATE, IFQCTL_DROPS, CTL_EOL) != 0)
531 goto bad;
532
533 return;
534 bad:
535 aprint_error("%s: could not attach sysctl nodes\n", ifname);
536 return;
537 }
538
539 /*
540 * bridge_ioctl:
541 *
542 * Handle a control request from the operator.
543 */
544 static int
545 bridge_ioctl(struct ifnet *ifp, u_long cmd, void *data)
546 {
547 struct bridge_softc *sc = ifp->if_softc;
548 struct lwp *l = curlwp; /* XXX */
549 union {
550 struct ifbreq ifbreq;
551 struct ifbifconf ifbifconf;
552 struct ifbareq ifbareq;
553 struct ifbaconf ifbaconf;
554 struct ifbrparam ifbrparam;
555 } args;
556 struct ifdrv *ifd = (struct ifdrv *) data;
557 const struct bridge_control *bc = NULL; /* XXXGCC */
558 int s, error = 0;
559
560 /* Authorize command before calling splnet(). */
561 switch (cmd) {
562 case SIOCGDRVSPEC:
563 case SIOCSDRVSPEC:
564 if (ifd->ifd_cmd >= bridge_control_table_size) {
565 error = EINVAL;
566 return error;
567 }
568
569 bc = &bridge_control_table[ifd->ifd_cmd];
570
571 /* We only care about BC_F_SUSER at this point. */
572 if ((bc->bc_flags & BC_F_SUSER) == 0)
573 break;
574
575 error = kauth_authorize_network(l->l_cred,
576 KAUTH_NETWORK_INTERFACE_BRIDGE,
577 cmd == SIOCGDRVSPEC ?
578 KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_GETPRIV :
579 KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_SETPRIV,
580 ifd, NULL, NULL);
581 if (error)
582 return (error);
583
584 break;
585 }
586
587 s = splnet();
588
589 switch (cmd) {
590 case SIOCGDRVSPEC:
591 case SIOCSDRVSPEC:
592 KASSERT(bc != NULL);
593 if (cmd == SIOCGDRVSPEC &&
594 (bc->bc_flags & BC_F_COPYOUT) == 0) {
595 error = EINVAL;
596 break;
597 }
598 else if (cmd == SIOCSDRVSPEC &&
599 (bc->bc_flags & BC_F_COPYOUT) != 0) {
600 error = EINVAL;
601 break;
602 }
603
604 /* BC_F_SUSER is checked above, before splnet(). */
605
606 if (ifd->ifd_len != bc->bc_argsize ||
607 ifd->ifd_len > sizeof(args)) {
608 error = EINVAL;
609 break;
610 }
611
612 memset(&args, 0, sizeof(args));
613 if (bc->bc_flags & BC_F_COPYIN) {
614 error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
615 if (error)
616 break;
617 }
618
619 error = (*bc->bc_func)(sc, &args);
620 if (error)
621 break;
622
623 if (bc->bc_flags & BC_F_COPYOUT)
624 error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
625
626 break;
627
628 case SIOCSIFFLAGS:
629 if ((error = ifioctl_common(ifp, cmd, data)) != 0)
630 break;
631 switch (ifp->if_flags & (IFF_UP|IFF_RUNNING)) {
632 case IFF_RUNNING:
633 /*
634 * If interface is marked down and it is running,
635 * then stop and disable it.
636 */
637 (*ifp->if_stop)(ifp, 1);
638 break;
639 case IFF_UP:
640 /*
641 * If interface is marked up and it is stopped, then
642 * start it.
643 */
644 error = (*ifp->if_init)(ifp);
645 break;
646 default:
647 break;
648 }
649 break;
650
651 case SIOCSIFMTU:
652 if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET)
653 error = 0;
654 break;
655
656 default:
657 error = ifioctl_common(ifp, cmd, data);
658 break;
659 }
660
661 splx(s);
662
663 return (error);
664 }
665
666 /*
667 * bridge_lookup_member:
668 *
669 * Lookup a bridge member interface.
670 */
671 static struct bridge_iflist *
672 bridge_lookup_member(struct bridge_softc *sc, const char *name)
673 {
674 struct bridge_iflist *bif;
675 struct ifnet *ifp;
676
677 BRIDGE_LOCK(sc);
678
679 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
680 ifp = bif->bif_ifp;
681 if (strcmp(ifp->if_xname, name) == 0)
682 break;
683 }
684
685 #ifdef BRIDGE_MPSAFE
686 if (bif != NULL) {
687 if (bif->bif_waiting)
688 bif = NULL;
689 else
690 atomic_inc_32(&bif->bif_refs);
691 }
692 #endif
693
694 BRIDGE_UNLOCK(sc);
695
696 return bif;
697 }
698
699 /*
700 * bridge_lookup_member_if:
701 *
702 * Lookup a bridge member interface by ifnet*.
703 */
704 static struct bridge_iflist *
705 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
706 {
707 struct bridge_iflist *bif;
708
709 BRIDGE_LOCK(sc);
710
711 bif = member_ifp->if_bridgeif;
712
713 #ifdef BRIDGE_MPSAFE
714 if (bif != NULL) {
715 if (bif->bif_waiting)
716 bif = NULL;
717 else
718 atomic_inc_32(&bif->bif_refs);
719 }
720 #endif
721
722 BRIDGE_UNLOCK(sc);
723
724 return bif;
725 }
726
727 /*
728 * bridge_release_member:
729 *
730 * Release the specified member interface.
731 */
732 static void
733 bridge_release_member(struct bridge_softc *sc, struct bridge_iflist *bif)
734 {
735 #ifdef BRIDGE_MPSAFE
736 atomic_dec_32(&bif->bif_refs);
737 membar_sync();
738 if (__predict_false(bif->bif_waiting && bif->bif_refs == 0)) {
739 BRIDGE_LOCK(sc);
740 cv_broadcast(&sc->sc_iflist_cv);
741 BRIDGE_UNLOCK(sc);
742 }
743 #else
744 (void)sc;
745 (void)bif;
746 #endif
747 }
748
749 /*
750 * bridge_delete_member:
751 *
752 * Delete the specified member interface.
753 */
754 static void
755 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif)
756 {
757 struct ifnet *ifs = bif->bif_ifp;
758
759 KASSERT(BRIDGE_LOCKED(sc));
760
761 ifs->if_input = ether_input;
762 ifs->if_bridge = NULL;
763 ifs->if_bridgeif = NULL;
764
765 LIST_REMOVE(bif, bif_next);
766
767 #ifdef BRIDGE_MPSAFE
768 bif->bif_waiting = true;
769 membar_sync();
770 while (bif->bif_refs > 0) {
771 aprint_debug("%s: cv_wait on iflist\n", __func__);
772 cv_wait(&sc->sc_iflist_cv, sc->sc_iflist_lock);
773 }
774 #endif
775
776 free(bif, M_DEVBUF);
777 }
778
779 static int
780 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
781 {
782 struct ifbreq *req = arg;
783 struct bridge_iflist *bif = NULL;
784 struct ifnet *ifs;
785 int error = 0;
786
787 ifs = ifunit(req->ifbr_ifsname);
788 if (ifs == NULL)
789 return (ENOENT);
790
791 if (sc->sc_if.if_mtu != ifs->if_mtu)
792 return (EINVAL);
793
794 if (ifs->if_bridge == sc)
795 return (EEXIST);
796
797 if (ifs->if_bridge != NULL)
798 return (EBUSY);
799
800 if (ifs->if_input != ether_input)
801 return EINVAL;
802
803 bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT);
804 if (bif == NULL)
805 return (ENOMEM);
806
807 switch (ifs->if_type) {
808 case IFT_ETHER:
809 /*
810 * Place the interface into promiscuous mode.
811 */
812 error = ifpromisc(ifs, 1);
813 if (error)
814 goto out;
815 break;
816 default:
817 error = EINVAL;
818 goto out;
819 }
820
821 bif->bif_ifp = ifs;
822 bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
823 bif->bif_priority = BSTP_DEFAULT_PORT_PRIORITY;
824 bif->bif_path_cost = BSTP_DEFAULT_PATH_COST;
825 bif->bif_refs = 0;
826 bif->bif_waiting = false;
827
828 BRIDGE_LOCK(sc);
829
830 ifs->if_bridge = sc;
831 ifs->if_bridgeif = bif;
832 LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
833 ifs->if_input = bridge_input;
834
835 BRIDGE_UNLOCK(sc);
836
837 if (sc->sc_if.if_flags & IFF_RUNNING)
838 bstp_initialization(sc);
839 else
840 bstp_stop(sc);
841
842 out:
843 if (error) {
844 if (bif != NULL)
845 free(bif, M_DEVBUF);
846 }
847 return (error);
848 }
849
850 static int
851 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
852 {
853 struct ifbreq *req = arg;
854 const char *name = req->ifbr_ifsname;
855 struct bridge_iflist *bif;
856 struct ifnet *ifs;
857
858 BRIDGE_LOCK(sc);
859
860 /*
861 * Don't use bridge_lookup_member. We want to get a member
862 * with bif_refs == 0.
863 */
864 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
865 ifs = bif->bif_ifp;
866 if (strcmp(ifs->if_xname, name) == 0)
867 break;
868 }
869
870 if (bif == NULL) {
871 BRIDGE_UNLOCK(sc);
872 return ENOENT;
873 }
874
875 bridge_delete_member(sc, bif);
876
877 BRIDGE_UNLOCK(sc);
878
879 switch (ifs->if_type) {
880 case IFT_ETHER:
881 /*
882 * Take the interface out of promiscuous mode.
883 * Don't call it with holding sc_iflist_lock.
884 */
885 (void) ifpromisc(ifs, 0);
886 break;
887 default:
888 #ifdef DIAGNOSTIC
889 panic("bridge_delete_member: impossible");
890 #endif
891 break;
892 }
893
894 bridge_rtdelete(sc, ifs);
895
896 if (sc->sc_if.if_flags & IFF_RUNNING)
897 bstp_initialization(sc);
898
899 return 0;
900 }
901
902 static int
903 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
904 {
905 struct ifbreq *req = arg;
906 struct bridge_iflist *bif;
907
908 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
909 if (bif == NULL)
910 return (ENOENT);
911
912 req->ifbr_ifsflags = bif->bif_flags;
913 req->ifbr_state = bif->bif_state;
914 req->ifbr_priority = bif->bif_priority;
915 req->ifbr_path_cost = bif->bif_path_cost;
916 req->ifbr_portno = bif->bif_ifp->if_index & 0xff;
917
918 bridge_release_member(sc, bif);
919
920 return (0);
921 }
922
923 static int
924 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
925 {
926 struct ifbreq *req = arg;
927 struct bridge_iflist *bif;
928
929 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
930 if (bif == NULL)
931 return (ENOENT);
932
933 if (req->ifbr_ifsflags & IFBIF_STP) {
934 switch (bif->bif_ifp->if_type) {
935 case IFT_ETHER:
936 /* These can do spanning tree. */
937 break;
938
939 default:
940 /* Nothing else can. */
941 bridge_release_member(sc, bif);
942 return (EINVAL);
943 }
944 }
945
946 bif->bif_flags = req->ifbr_ifsflags;
947
948 bridge_release_member(sc, bif);
949
950 if (sc->sc_if.if_flags & IFF_RUNNING)
951 bstp_initialization(sc);
952
953 return (0);
954 }
955
956 static int
957 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
958 {
959 struct ifbrparam *param = arg;
960
961 sc->sc_brtmax = param->ifbrp_csize;
962 bridge_rttrim(sc);
963
964 return (0);
965 }
966
967 static int
968 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
969 {
970 struct ifbrparam *param = arg;
971
972 param->ifbrp_csize = sc->sc_brtmax;
973
974 return (0);
975 }
976
977 static int
978 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
979 {
980 struct ifbifconf *bifc = arg;
981 struct bridge_iflist *bif;
982 struct ifbreq breq;
983 int count, len, error = 0;
984
985 BRIDGE_LOCK(sc);
986
987 count = 0;
988 LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
989 count++;
990
991 if (bifc->ifbic_len == 0) {
992 BRIDGE_UNLOCK(sc);
993 bifc->ifbic_len = sizeof(breq) * count;
994 return (0);
995 }
996
997 count = 0;
998 len = bifc->ifbic_len;
999 memset(&breq, 0, sizeof breq);
1000 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1001 if (len < sizeof(breq))
1002 break;
1003
1004 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1005 sizeof(breq.ifbr_ifsname));
1006 breq.ifbr_ifsflags = bif->bif_flags;
1007 breq.ifbr_state = bif->bif_state;
1008 breq.ifbr_priority = bif->bif_priority;
1009 breq.ifbr_path_cost = bif->bif_path_cost;
1010 breq.ifbr_portno = bif->bif_ifp->if_index & 0xff;
1011 error = copyout(&breq, bifc->ifbic_req + count, sizeof(breq));
1012 if (error)
1013 break;
1014 count++;
1015 len -= sizeof(breq);
1016 }
1017
1018 BRIDGE_UNLOCK(sc);
1019
1020 bifc->ifbic_len = sizeof(breq) * count;
1021 return (error);
1022 }
1023
1024 static int
1025 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1026 {
1027 struct ifbaconf *bac = arg;
1028 struct bridge_rtnode *brt;
1029 struct ifbareq bareq;
1030 int count = 0, error = 0, len;
1031
1032 if (bac->ifbac_len == 0)
1033 return (0);
1034
1035 mutex_enter(sc->sc_rtlist_lock);
1036
1037 len = bac->ifbac_len;
1038 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1039 if (len < sizeof(bareq))
1040 goto out;
1041 memset(&bareq, 0, sizeof(bareq));
1042 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1043 sizeof(bareq.ifba_ifsname));
1044 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1045 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
1046 bareq.ifba_expire = brt->brt_expire - time_uptime;
1047 } else
1048 bareq.ifba_expire = 0;
1049 bareq.ifba_flags = brt->brt_flags;
1050
1051 error = copyout(&bareq, bac->ifbac_req + count, sizeof(bareq));
1052 if (error)
1053 goto out;
1054 count++;
1055 len -= sizeof(bareq);
1056 }
1057 out:
1058 mutex_exit(sc->sc_rtlist_lock);
1059
1060 bac->ifbac_len = sizeof(bareq) * count;
1061 return (error);
1062 }
1063
1064 static int
1065 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1066 {
1067 struct ifbareq *req = arg;
1068 struct bridge_iflist *bif;
1069 int error;
1070
1071 bif = bridge_lookup_member(sc, req->ifba_ifsname);
1072 if (bif == NULL)
1073 return (ENOENT);
1074
1075 error = bridge_rtupdate(sc, req->ifba_dst, bif->bif_ifp, 1,
1076 req->ifba_flags);
1077
1078 bridge_release_member(sc, bif);
1079
1080 return (error);
1081 }
1082
1083 static int
1084 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1085 {
1086 struct ifbrparam *param = arg;
1087
1088 sc->sc_brttimeout = param->ifbrp_ctime;
1089
1090 return (0);
1091 }
1092
1093 static int
1094 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1095 {
1096 struct ifbrparam *param = arg;
1097
1098 param->ifbrp_ctime = sc->sc_brttimeout;
1099
1100 return (0);
1101 }
1102
1103 static int
1104 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1105 {
1106 struct ifbareq *req = arg;
1107
1108 return (bridge_rtdaddr(sc, req->ifba_dst));
1109 }
1110
1111 static int
1112 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1113 {
1114 struct ifbreq *req = arg;
1115
1116 bridge_rtflush(sc, req->ifbr_ifsflags);
1117
1118 return (0);
1119 }
1120
1121 static int
1122 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1123 {
1124 struct ifbrparam *param = arg;
1125
1126 param->ifbrp_prio = sc->sc_bridge_priority;
1127
1128 return (0);
1129 }
1130
1131 static int
1132 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1133 {
1134 struct ifbrparam *param = arg;
1135
1136 sc->sc_bridge_priority = param->ifbrp_prio;
1137
1138 if (sc->sc_if.if_flags & IFF_RUNNING)
1139 bstp_initialization(sc);
1140
1141 return (0);
1142 }
1143
1144 static int
1145 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1146 {
1147 struct ifbrparam *param = arg;
1148
1149 param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8;
1150
1151 return (0);
1152 }
1153
1154 static int
1155 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1156 {
1157 struct ifbrparam *param = arg;
1158
1159 if (param->ifbrp_hellotime == 0)
1160 return (EINVAL);
1161 sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8;
1162
1163 if (sc->sc_if.if_flags & IFF_RUNNING)
1164 bstp_initialization(sc);
1165
1166 return (0);
1167 }
1168
1169 static int
1170 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1171 {
1172 struct ifbrparam *param = arg;
1173
1174 param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8;
1175
1176 return (0);
1177 }
1178
1179 static int
1180 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1181 {
1182 struct ifbrparam *param = arg;
1183
1184 if (param->ifbrp_fwddelay == 0)
1185 return (EINVAL);
1186 sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8;
1187
1188 if (sc->sc_if.if_flags & IFF_RUNNING)
1189 bstp_initialization(sc);
1190
1191 return (0);
1192 }
1193
1194 static int
1195 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1196 {
1197 struct ifbrparam *param = arg;
1198
1199 param->ifbrp_maxage = sc->sc_bridge_max_age >> 8;
1200
1201 return (0);
1202 }
1203
1204 static int
1205 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1206 {
1207 struct ifbrparam *param = arg;
1208
1209 if (param->ifbrp_maxage == 0)
1210 return (EINVAL);
1211 sc->sc_bridge_max_age = param->ifbrp_maxage << 8;
1212
1213 if (sc->sc_if.if_flags & IFF_RUNNING)
1214 bstp_initialization(sc);
1215
1216 return (0);
1217 }
1218
1219 static int
1220 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1221 {
1222 struct ifbreq *req = arg;
1223 struct bridge_iflist *bif;
1224
1225 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1226 if (bif == NULL)
1227 return (ENOENT);
1228
1229 bif->bif_priority = req->ifbr_priority;
1230
1231 if (sc->sc_if.if_flags & IFF_RUNNING)
1232 bstp_initialization(sc);
1233
1234 bridge_release_member(sc, bif);
1235
1236 return (0);
1237 }
1238
1239 #if defined(BRIDGE_IPF)
1240 static int
1241 bridge_ioctl_gfilt(struct bridge_softc *sc, void *arg)
1242 {
1243 struct ifbrparam *param = arg;
1244
1245 param->ifbrp_filter = sc->sc_filter_flags;
1246
1247 return (0);
1248 }
1249
1250 static int
1251 bridge_ioctl_sfilt(struct bridge_softc *sc, void *arg)
1252 {
1253 struct ifbrparam *param = arg;
1254 uint32_t nflags, oflags;
1255
1256 if (param->ifbrp_filter & ~IFBF_FILT_MASK)
1257 return (EINVAL);
1258
1259 nflags = param->ifbrp_filter;
1260 oflags = sc->sc_filter_flags;
1261
1262 if ((nflags & IFBF_FILT_USEIPF) && !(oflags & IFBF_FILT_USEIPF)) {
1263 pfil_add_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
1264 sc->sc_if.if_pfil);
1265 }
1266 if (!(nflags & IFBF_FILT_USEIPF) && (oflags & IFBF_FILT_USEIPF)) {
1267 pfil_remove_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
1268 sc->sc_if.if_pfil);
1269 }
1270
1271 sc->sc_filter_flags = nflags;
1272
1273 return (0);
1274 }
1275 #endif /* BRIDGE_IPF */
1276
1277 static int
1278 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1279 {
1280 struct ifbreq *req = arg;
1281 struct bridge_iflist *bif;
1282
1283 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1284 if (bif == NULL)
1285 return (ENOENT);
1286
1287 bif->bif_path_cost = req->ifbr_path_cost;
1288
1289 if (sc->sc_if.if_flags & IFF_RUNNING)
1290 bstp_initialization(sc);
1291
1292 bridge_release_member(sc, bif);
1293
1294 return (0);
1295 }
1296
1297 /*
1298 * bridge_ifdetach:
1299 *
1300 * Detach an interface from a bridge. Called when a member
1301 * interface is detaching.
1302 */
1303 void
1304 bridge_ifdetach(struct ifnet *ifp)
1305 {
1306 struct bridge_softc *sc = ifp->if_bridge;
1307 struct ifbreq breq;
1308
1309 /* ioctl_lock should prevent this from happening */
1310 KASSERT(sc != NULL);
1311
1312 memset(&breq, 0, sizeof(breq));
1313 strlcpy(breq.ifbr_ifsname, ifp->if_xname, sizeof(breq.ifbr_ifsname));
1314
1315 (void) bridge_ioctl_del(sc, &breq);
1316 }
1317
1318 /*
1319 * bridge_init:
1320 *
1321 * Initialize a bridge interface.
1322 */
1323 static int
1324 bridge_init(struct ifnet *ifp)
1325 {
1326 struct bridge_softc *sc = ifp->if_softc;
1327
1328 if (ifp->if_flags & IFF_RUNNING)
1329 return (0);
1330
1331 callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1332 bridge_timer, sc);
1333
1334 ifp->if_flags |= IFF_RUNNING;
1335 bstp_initialization(sc);
1336 return (0);
1337 }
1338
1339 /*
1340 * bridge_stop:
1341 *
1342 * Stop the bridge interface.
1343 */
1344 static void
1345 bridge_stop(struct ifnet *ifp, int disable)
1346 {
1347 struct bridge_softc *sc = ifp->if_softc;
1348
1349 if ((ifp->if_flags & IFF_RUNNING) == 0)
1350 return;
1351
1352 callout_stop(&sc->sc_brcallout);
1353 bstp_stop(sc);
1354
1355 bridge_rtflush(sc, IFBF_FLUSHDYN);
1356
1357 ifp->if_flags &= ~IFF_RUNNING;
1358 }
1359
1360 /*
1361 * bridge_enqueue:
1362 *
1363 * Enqueue a packet on a bridge member interface.
1364 */
1365 void
1366 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m,
1367 int runfilt)
1368 {
1369 ALTQ_DECL(struct altq_pktattr pktattr;)
1370 int len, error;
1371 short mflags;
1372
1373 /*
1374 * Clear any in-bound checksum flags for this packet.
1375 */
1376 m->m_pkthdr.csum_flags = 0;
1377
1378 if (runfilt) {
1379 if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
1380 dst_ifp, PFIL_OUT) != 0) {
1381 if (m != NULL)
1382 m_freem(m);
1383 return;
1384 }
1385 if (m == NULL)
1386 return;
1387 }
1388
1389 #ifdef ALTQ
1390 /*
1391 * If ALTQ is enabled on the member interface, do
1392 * classification; the queueing discipline might
1393 * not require classification, but might require
1394 * the address family/header pointer in the pktattr.
1395 */
1396 if (ALTQ_IS_ENABLED(&dst_ifp->if_snd)) {
1397 /* XXX IFT_ETHER */
1398 altq_etherclassify(&dst_ifp->if_snd, m, &pktattr);
1399 }
1400 #endif /* ALTQ */
1401
1402 len = m->m_pkthdr.len;
1403 m->m_flags |= M_PROTO1;
1404 mflags = m->m_flags;
1405
1406 IFQ_ENQUEUE(&dst_ifp->if_snd, m, &pktattr, error);
1407
1408 if (error) {
1409 /* mbuf is already freed */
1410 sc->sc_if.if_oerrors++;
1411 return;
1412 }
1413
1414 sc->sc_if.if_opackets++;
1415 sc->sc_if.if_obytes += len;
1416
1417 dst_ifp->if_obytes += len;
1418
1419 if (mflags & M_MCAST) {
1420 sc->sc_if.if_omcasts++;
1421 dst_ifp->if_omcasts++;
1422 }
1423
1424 if ((dst_ifp->if_flags & IFF_OACTIVE) == 0)
1425 (*dst_ifp->if_start)(dst_ifp);
1426 }
1427
1428 /*
1429 * bridge_output:
1430 *
1431 * Send output from a bridge member interface. This
1432 * performs the bridging function for locally originated
1433 * packets.
1434 *
1435 * The mbuf has the Ethernet header already attached. We must
1436 * enqueue or free the mbuf before returning.
1437 */
1438 int
1439 bridge_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa,
1440 struct rtentry *rt)
1441 {
1442 struct ether_header *eh;
1443 struct ifnet *dst_if;
1444 struct bridge_softc *sc;
1445 #ifndef BRIDGE_MPSAFE
1446 int s;
1447 #endif
1448
1449 if (m->m_len < ETHER_HDR_LEN) {
1450 m = m_pullup(m, ETHER_HDR_LEN);
1451 if (m == NULL)
1452 return (0);
1453 }
1454
1455 eh = mtod(m, struct ether_header *);
1456 sc = ifp->if_bridge;
1457
1458 #ifndef BRIDGE_MPSAFE
1459 s = splnet();
1460 #endif
1461
1462 /*
1463 * If bridge is down, but the original output interface is up,
1464 * go ahead and send out that interface. Otherwise, the packet
1465 * is dropped below.
1466 */
1467 if (__predict_false(sc == NULL) ||
1468 (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
1469 dst_if = ifp;
1470 goto sendunicast;
1471 }
1472
1473 /*
1474 * If the packet is a multicast, or we don't know a better way to
1475 * get there, send to all interfaces.
1476 */
1477 if (ETHER_IS_MULTICAST(eh->ether_dhost))
1478 dst_if = NULL;
1479 else
1480 dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1481 if (dst_if == NULL) {
1482 struct bridge_iflist *bif;
1483 struct mbuf *mc;
1484 int used = 0;
1485
1486 BRIDGE_LOCK(sc);
1487
1488 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1489 dst_if = bif->bif_ifp;
1490 if ((dst_if->if_flags & IFF_RUNNING) == 0)
1491 continue;
1492
1493 /*
1494 * If this is not the original output interface,
1495 * and the interface is participating in spanning
1496 * tree, make sure the port is in a state that
1497 * allows forwarding.
1498 */
1499 if (dst_if != ifp &&
1500 (bif->bif_flags & IFBIF_STP) != 0) {
1501 switch (bif->bif_state) {
1502 case BSTP_IFSTATE_BLOCKING:
1503 case BSTP_IFSTATE_LISTENING:
1504 case BSTP_IFSTATE_DISABLED:
1505 continue;
1506 }
1507 }
1508
1509 if (LIST_NEXT(bif, bif_next) == NULL) {
1510 used = 1;
1511 mc = m;
1512 } else {
1513 mc = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1514 if (mc == NULL) {
1515 sc->sc_if.if_oerrors++;
1516 continue;
1517 }
1518 }
1519
1520 bridge_enqueue(sc, dst_if, mc, 0);
1521 }
1522
1523 BRIDGE_UNLOCK(sc);
1524
1525 if (used == 0)
1526 m_freem(m);
1527 #ifndef BRIDGE_MPSAFE
1528 splx(s);
1529 #endif
1530 return (0);
1531 }
1532
1533 sendunicast:
1534 /*
1535 * XXX Spanning tree consideration here?
1536 */
1537
1538 if ((dst_if->if_flags & IFF_RUNNING) == 0) {
1539 m_freem(m);
1540 #ifndef BRIDGE_MPSAFE
1541 splx(s);
1542 #endif
1543 return (0);
1544 }
1545
1546 bridge_enqueue(sc, dst_if, m, 0);
1547
1548 #ifndef BRIDGE_MPSAFE
1549 splx(s);
1550 #endif
1551 return (0);
1552 }
1553
1554 /*
1555 * bridge_start:
1556 *
1557 * Start output on a bridge.
1558 *
1559 * NOTE: This routine should never be called in this implementation.
1560 */
1561 static void
1562 bridge_start(struct ifnet *ifp)
1563 {
1564
1565 printf("%s: bridge_start() called\n", ifp->if_xname);
1566 }
1567
1568 /*
1569 * bridge_forward:
1570 *
1571 * The forwarding function of the bridge.
1572 */
1573 static void
1574 bridge_forward(void *v)
1575 {
1576 struct bridge_softc *sc = v;
1577 struct mbuf *m;
1578 struct bridge_iflist *bif;
1579 struct ifnet *src_if, *dst_if;
1580 struct ether_header *eh;
1581 #ifndef BRIDGE_MPSAFE
1582 int s;
1583
1584 KERNEL_LOCK(1, NULL);
1585 mutex_enter(softnet_lock);
1586 #endif
1587
1588 if ((sc->sc_if.if_flags & IFF_RUNNING) == 0) {
1589 #ifndef BRIDGE_MPSAFE
1590 mutex_exit(softnet_lock);
1591 KERNEL_UNLOCK_ONE(NULL);
1592 #endif
1593 return;
1594 }
1595
1596 #ifndef BRIDGE_MPSAFE
1597 s = splnet();
1598 #endif
1599 while ((m = pktq_dequeue(sc->sc_fwd_pktq)) != NULL) {
1600 src_if = m->m_pkthdr.rcvif;
1601
1602 sc->sc_if.if_ipackets++;
1603 sc->sc_if.if_ibytes += m->m_pkthdr.len;
1604
1605 /*
1606 * Look up the bridge_iflist.
1607 */
1608 bif = bridge_lookup_member_if(sc, src_if);
1609 if (bif == NULL) {
1610 /* Interface is not a bridge member (anymore?) */
1611 m_freem(m);
1612 continue;
1613 }
1614
1615 if (bif->bif_flags & IFBIF_STP) {
1616 switch (bif->bif_state) {
1617 case BSTP_IFSTATE_BLOCKING:
1618 case BSTP_IFSTATE_LISTENING:
1619 case BSTP_IFSTATE_DISABLED:
1620 m_freem(m);
1621 bridge_release_member(sc, bif);
1622 continue;
1623 }
1624 }
1625
1626 eh = mtod(m, struct ether_header *);
1627
1628 /*
1629 * If the interface is learning, and the source
1630 * address is valid and not multicast, record
1631 * the address.
1632 */
1633 if ((bif->bif_flags & IFBIF_LEARNING) != 0 &&
1634 ETHER_IS_MULTICAST(eh->ether_shost) == 0 &&
1635 (eh->ether_shost[0] == 0 &&
1636 eh->ether_shost[1] == 0 &&
1637 eh->ether_shost[2] == 0 &&
1638 eh->ether_shost[3] == 0 &&
1639 eh->ether_shost[4] == 0 &&
1640 eh->ether_shost[5] == 0) == 0) {
1641 (void) bridge_rtupdate(sc, eh->ether_shost,
1642 src_if, 0, IFBAF_DYNAMIC);
1643 }
1644
1645 if ((bif->bif_flags & IFBIF_STP) != 0 &&
1646 bif->bif_state == BSTP_IFSTATE_LEARNING) {
1647 m_freem(m);
1648 bridge_release_member(sc, bif);
1649 continue;
1650 }
1651
1652 bridge_release_member(sc, bif);
1653
1654 /*
1655 * At this point, the port either doesn't participate
1656 * in spanning tree or it is in the forwarding state.
1657 */
1658
1659 /*
1660 * If the packet is unicast, destined for someone on
1661 * "this" side of the bridge, drop it.
1662 */
1663 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
1664 dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1665 if (src_if == dst_if) {
1666 m_freem(m);
1667 continue;
1668 }
1669 } else {
1670 /* ...forward it to all interfaces. */
1671 sc->sc_if.if_imcasts++;
1672 dst_if = NULL;
1673 }
1674
1675 if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
1676 m->m_pkthdr.rcvif, PFIL_IN) != 0) {
1677 if (m != NULL)
1678 m_freem(m);
1679 continue;
1680 }
1681 if (m == NULL)
1682 continue;
1683
1684 if (dst_if == NULL) {
1685 bridge_broadcast(sc, src_if, m);
1686 continue;
1687 }
1688
1689 /*
1690 * At this point, we're dealing with a unicast frame
1691 * going to a different interface.
1692 */
1693 if ((dst_if->if_flags & IFF_RUNNING) == 0) {
1694 m_freem(m);
1695 continue;
1696 }
1697
1698 bif = bridge_lookup_member_if(sc, dst_if);
1699 if (bif == NULL) {
1700 /* Not a member of the bridge (anymore?) */
1701 m_freem(m);
1702 continue;
1703 }
1704
1705 if (bif->bif_flags & IFBIF_STP) {
1706 switch (bif->bif_state) {
1707 case BSTP_IFSTATE_DISABLED:
1708 case BSTP_IFSTATE_BLOCKING:
1709 m_freem(m);
1710 bridge_release_member(sc, bif);
1711 continue;
1712 }
1713 }
1714
1715 bridge_release_member(sc, bif);
1716
1717 bridge_enqueue(sc, dst_if, m, 1);
1718 }
1719 #ifndef BRIDGE_MPSAFE
1720 splx(s);
1721 mutex_exit(softnet_lock);
1722 KERNEL_UNLOCK_ONE(NULL);
1723 #endif
1724 }
1725
1726 static bool
1727 bstp_state_before_learning(struct bridge_iflist *bif)
1728 {
1729 if (bif->bif_flags & IFBIF_STP) {
1730 switch (bif->bif_state) {
1731 case BSTP_IFSTATE_BLOCKING:
1732 case BSTP_IFSTATE_LISTENING:
1733 case BSTP_IFSTATE_DISABLED:
1734 return true;
1735 }
1736 }
1737 return false;
1738 }
1739
1740 static bool
1741 bridge_ourether(struct bridge_iflist *bif, struct ether_header *eh, int src)
1742 {
1743 uint8_t *ether = src ? eh->ether_shost : eh->ether_dhost;
1744
1745 if (memcmp(CLLADDR(bif->bif_ifp->if_sadl), ether, ETHER_ADDR_LEN) == 0
1746 #if NCARP > 0
1747 || (bif->bif_ifp->if_carp &&
1748 carp_ourether(bif->bif_ifp->if_carp, eh, IFT_ETHER, src) != NULL)
1749 #endif /* NCARP > 0 */
1750 )
1751 return true;
1752
1753 return false;
1754 }
1755
1756 /*
1757 * bridge_input:
1758 *
1759 * Receive input from a member interface. Queue the packet for
1760 * bridging if it is not for us.
1761 */
1762 static void
1763 bridge_input(struct ifnet *ifp, struct mbuf *m)
1764 {
1765 struct bridge_softc *sc = ifp->if_bridge;
1766 struct bridge_iflist *bif;
1767 struct ether_header *eh;
1768
1769 if (__predict_false(sc == NULL) ||
1770 (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
1771 ether_input(ifp, m);
1772 return;
1773 }
1774
1775 bif = bridge_lookup_member_if(sc, ifp);
1776 if (bif == NULL) {
1777 ether_input(ifp, m);
1778 return;
1779 }
1780
1781 eh = mtod(m, struct ether_header *);
1782
1783 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1784 if (memcmp(etherbroadcastaddr,
1785 eh->ether_dhost, ETHER_ADDR_LEN) == 0)
1786 m->m_flags |= M_BCAST;
1787 else
1788 m->m_flags |= M_MCAST;
1789 }
1790
1791 /*
1792 * A 'fast' path for packets addressed to interfaces that are
1793 * part of this bridge.
1794 */
1795 if (!(m->m_flags & (M_BCAST|M_MCAST)) &&
1796 !bstp_state_before_learning(bif)) {
1797 struct bridge_iflist *_bif;
1798
1799 BRIDGE_LOCK(sc);
1800 LIST_FOREACH(_bif, &sc->sc_iflist, bif_next) {
1801 /* It is destined for us. */
1802 if (bridge_ourether(_bif, eh, 0)) {
1803 if (_bif->bif_flags & IFBIF_LEARNING)
1804 (void) bridge_rtupdate(sc,
1805 eh->ether_shost, ifp, 0, IFBAF_DYNAMIC);
1806 m->m_pkthdr.rcvif = _bif->bif_ifp;
1807 ether_input(_bif->bif_ifp, m);
1808 break;
1809 }
1810
1811 /* We just received a packet that we sent out. */
1812 if (bridge_ourether(_bif, eh, 1)) {
1813 m_freem(m);
1814 break;
1815 }
1816 }
1817 BRIDGE_UNLOCK(sc);
1818
1819 if (_bif != NULL) {
1820 bridge_release_member(sc, bif);
1821 return;
1822 }
1823 }
1824
1825 /* Tap off 802.1D packets; they do not get forwarded. */
1826 if (bif->bif_flags & IFBIF_STP &&
1827 memcmp(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN) == 0) {
1828 bstp_input(sc, bif, m);
1829 bridge_release_member(sc, bif);
1830 return;
1831 }
1832
1833 /*
1834 * A normal switch would discard the packet here, but that's not what
1835 * we've done historically. This also prevents some obnoxious behaviour.
1836 */
1837 if (bstp_state_before_learning(bif)) {
1838 ether_input(ifp, m);
1839 bridge_release_member(sc, bif);
1840 return;
1841 }
1842
1843 bridge_release_member(sc, bif);
1844
1845 /* Queue the packet for bridge forwarding. */
1846 if (__predict_false(!pktq_enqueue(sc->sc_fwd_pktq, m, 0)))
1847 m_freem(m);
1848 }
1849
1850 /*
1851 * bridge_broadcast:
1852 *
1853 * Send a frame to all interfaces that are members of
1854 * the bridge, except for the one on which the packet
1855 * arrived.
1856 */
1857 static void
1858 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
1859 struct mbuf *m)
1860 {
1861 struct bridge_iflist *bif;
1862 struct mbuf *mc;
1863 struct ifnet *dst_if;
1864 bool used, bmcast;
1865
1866 used = bmcast = m->m_flags & (M_BCAST|M_MCAST);
1867
1868 BRIDGE_LOCK(sc);
1869
1870 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1871 dst_if = bif->bif_ifp;
1872 if (dst_if == src_if)
1873 continue;
1874
1875 if (bif->bif_flags & IFBIF_STP) {
1876 switch (bif->bif_state) {
1877 case BSTP_IFSTATE_BLOCKING:
1878 case BSTP_IFSTATE_DISABLED:
1879 continue;
1880 }
1881 }
1882
1883 if ((bif->bif_flags & IFBIF_DISCOVER) == 0 && !bmcast)
1884 continue;
1885
1886 if ((dst_if->if_flags & IFF_RUNNING) == 0)
1887 continue;
1888
1889 if (!used && LIST_NEXT(bif, bif_next) == NULL) {
1890 mc = m;
1891 used = true;
1892 } else {
1893 mc = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
1894 if (mc == NULL) {
1895 sc->sc_if.if_oerrors++;
1896 continue;
1897 }
1898 }
1899
1900 bridge_enqueue(sc, dst_if, mc, 1);
1901 }
1902
1903 if (bmcast)
1904 ether_input(src_if, m);
1905 else if (!used)
1906 m_freem(m);
1907
1908 BRIDGE_UNLOCK(sc);
1909 }
1910
1911 /*
1912 * bridge_rtupdate:
1913 *
1914 * Add a bridge routing entry.
1915 */
1916 static int
1917 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
1918 struct ifnet *dst_if, int setflags, uint8_t flags)
1919 {
1920 struct bridge_rtnode *brt;
1921 int error = 0;
1922
1923 mutex_enter(sc->sc_rtlist_lock);
1924
1925 /*
1926 * A route for this destination might already exist. If so,
1927 * update it, otherwise create a new one.
1928 */
1929 if ((brt = bridge_rtnode_lookup(sc, dst)) == NULL) {
1930 if (sc->sc_brtcnt >= sc->sc_brtmax) {
1931 error = ENOSPC;
1932 goto out;
1933 }
1934
1935 /*
1936 * Allocate a new bridge forwarding node, and
1937 * initialize the expiration time and Ethernet
1938 * address.
1939 */
1940 brt = pool_get(&bridge_rtnode_pool, PR_NOWAIT);
1941 if (brt == NULL) {
1942 error = ENOMEM;
1943 goto out;
1944 }
1945
1946 memset(brt, 0, sizeof(*brt));
1947 brt->brt_expire = time_uptime + sc->sc_brttimeout;
1948 brt->brt_flags = IFBAF_DYNAMIC;
1949 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
1950
1951 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
1952 pool_put(&bridge_rtnode_pool, brt);
1953 goto out;
1954 }
1955 }
1956
1957 brt->brt_ifp = dst_if;
1958 if (setflags) {
1959 brt->brt_flags = flags;
1960 if (flags & IFBAF_STATIC)
1961 brt->brt_expire = 0;
1962 else
1963 brt->brt_expire = time_uptime + sc->sc_brttimeout;
1964 }
1965
1966 out:
1967 mutex_exit(sc->sc_rtlist_lock);
1968
1969 return error;
1970 }
1971
1972 /*
1973 * bridge_rtlookup:
1974 *
1975 * Lookup the destination interface for an address.
1976 */
1977 static struct ifnet *
1978 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr)
1979 {
1980 struct bridge_rtnode *brt;
1981 struct ifnet *ifs = NULL;
1982
1983 mutex_enter(sc->sc_rtlist_lock);
1984
1985 brt = bridge_rtnode_lookup(sc, addr);
1986 if (brt != NULL)
1987 ifs = brt->brt_ifp;
1988
1989 mutex_exit(sc->sc_rtlist_lock);
1990
1991 return ifs;
1992 }
1993
1994 /*
1995 * bridge_rttrim:
1996 *
1997 * Trim the routine table so that we have a number
1998 * of routing entries less than or equal to the
1999 * maximum number.
2000 */
2001 static void
2002 bridge_rttrim(struct bridge_softc *sc)
2003 {
2004 struct bridge_rtnode *brt, *nbrt;
2005
2006 mutex_enter(sc->sc_rtlist_lock);
2007
2008 /* Make sure we actually need to do this. */
2009 if (sc->sc_brtcnt <= sc->sc_brtmax)
2010 goto out;
2011
2012 /* Force an aging cycle; this might trim enough addresses. */
2013 bridge_rtage(sc);
2014 if (sc->sc_brtcnt <= sc->sc_brtmax)
2015 goto out;
2016
2017 for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
2018 nbrt = LIST_NEXT(brt, brt_list);
2019 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2020 bridge_rtnode_destroy(sc, brt);
2021 if (sc->sc_brtcnt <= sc->sc_brtmax)
2022 goto out;
2023 }
2024 }
2025
2026 out:
2027 mutex_exit(sc->sc_rtlist_lock);
2028
2029 return;
2030 }
2031
2032 /*
2033 * bridge_timer:
2034 *
2035 * Aging timer for the bridge.
2036 */
2037 static void
2038 bridge_timer(void *arg)
2039 {
2040 struct bridge_softc *sc = arg;
2041
2042 mutex_enter(sc->sc_rtlist_lock);
2043
2044 bridge_rtage(sc);
2045
2046 if (sc->sc_if.if_flags & IFF_RUNNING)
2047 callout_reset(&sc->sc_brcallout,
2048 bridge_rtable_prune_period * hz, bridge_timer, sc);
2049
2050 mutex_exit(sc->sc_rtlist_lock);
2051 }
2052
2053 /*
2054 * bridge_rtage:
2055 *
2056 * Perform an aging cycle.
2057 */
2058 static void
2059 bridge_rtage(struct bridge_softc *sc)
2060 {
2061 struct bridge_rtnode *brt, *nbrt;
2062
2063 KASSERT(mutex_owned(sc->sc_rtlist_lock));
2064
2065 for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
2066 nbrt = LIST_NEXT(brt, brt_list);
2067 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2068 if (time_uptime >= brt->brt_expire)
2069 bridge_rtnode_destroy(sc, brt);
2070 }
2071 }
2072 }
2073
2074 /*
2075 * bridge_rtflush:
2076 *
2077 * Remove all dynamic addresses from the bridge.
2078 */
2079 static void
2080 bridge_rtflush(struct bridge_softc *sc, int full)
2081 {
2082 struct bridge_rtnode *brt, *nbrt;
2083
2084 mutex_enter(sc->sc_rtlist_lock);
2085
2086 for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
2087 nbrt = LIST_NEXT(brt, brt_list);
2088 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2089 bridge_rtnode_destroy(sc, brt);
2090 }
2091
2092 mutex_exit(sc->sc_rtlist_lock);
2093 }
2094
2095 /*
2096 * bridge_rtdaddr:
2097 *
2098 * Remove an address from the table.
2099 */
2100 static int
2101 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr)
2102 {
2103 struct bridge_rtnode *brt;
2104 int error = 0;
2105
2106 mutex_enter(sc->sc_rtlist_lock);
2107
2108 if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL) {
2109 error = ENOENT;
2110 goto out;
2111 }
2112
2113 bridge_rtnode_destroy(sc, brt);
2114
2115 out:
2116 mutex_exit(sc->sc_rtlist_lock);
2117
2118 return error;
2119 }
2120
2121 /*
2122 * bridge_rtdelete:
2123 *
2124 * Delete routes to a speicifc member interface.
2125 */
2126 static void
2127 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp)
2128 {
2129 struct bridge_rtnode *brt, *nbrt;
2130
2131 mutex_enter(sc->sc_rtlist_lock);
2132
2133 for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
2134 nbrt = LIST_NEXT(brt, brt_list);
2135 if (brt->brt_ifp == ifp)
2136 bridge_rtnode_destroy(sc, brt);
2137 }
2138
2139 mutex_exit(sc->sc_rtlist_lock);
2140 }
2141
2142 /*
2143 * bridge_rtable_init:
2144 *
2145 * Initialize the route table for this bridge.
2146 */
2147 static int
2148 bridge_rtable_init(struct bridge_softc *sc)
2149 {
2150 int i;
2151
2152 sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2153 M_DEVBUF, M_NOWAIT);
2154 if (sc->sc_rthash == NULL)
2155 return (ENOMEM);
2156
2157 for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2158 LIST_INIT(&sc->sc_rthash[i]);
2159
2160 sc->sc_rthash_key = cprng_fast32();
2161
2162 LIST_INIT(&sc->sc_rtlist);
2163
2164 sc->sc_rtlist_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
2165
2166 return (0);
2167 }
2168
2169 /*
2170 * bridge_rtable_fini:
2171 *
2172 * Deconstruct the route table for this bridge.
2173 */
2174 static void
2175 bridge_rtable_fini(struct bridge_softc *sc)
2176 {
2177
2178 free(sc->sc_rthash, M_DEVBUF);
2179 if (sc->sc_rtlist_lock)
2180 mutex_obj_free(sc->sc_rtlist_lock);
2181 }
2182
2183 /*
2184 * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2185 * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2186 */
2187 #define mix(a, b, c) \
2188 do { \
2189 a -= b; a -= c; a ^= (c >> 13); \
2190 b -= c; b -= a; b ^= (a << 8); \
2191 c -= a; c -= b; c ^= (b >> 13); \
2192 a -= b; a -= c; a ^= (c >> 12); \
2193 b -= c; b -= a; b ^= (a << 16); \
2194 c -= a; c -= b; c ^= (b >> 5); \
2195 a -= b; a -= c; a ^= (c >> 3); \
2196 b -= c; b -= a; b ^= (a << 10); \
2197 c -= a; c -= b; c ^= (b >> 15); \
2198 } while (/*CONSTCOND*/0)
2199
2200 static inline uint32_t
2201 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2202 {
2203 uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2204
2205 b += addr[5] << 8;
2206 b += addr[4];
2207 a += addr[3] << 24;
2208 a += addr[2] << 16;
2209 a += addr[1] << 8;
2210 a += addr[0];
2211
2212 mix(a, b, c);
2213
2214 return (c & BRIDGE_RTHASH_MASK);
2215 }
2216
2217 #undef mix
2218
2219 /*
2220 * bridge_rtnode_lookup:
2221 *
2222 * Look up a bridge route node for the specified destination.
2223 */
2224 static struct bridge_rtnode *
2225 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr)
2226 {
2227 struct bridge_rtnode *brt;
2228 uint32_t hash;
2229 int dir;
2230
2231 KASSERT(mutex_owned(sc->sc_rtlist_lock));
2232
2233 hash = bridge_rthash(sc, addr);
2234 LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
2235 dir = memcmp(addr, brt->brt_addr, ETHER_ADDR_LEN);
2236 if (dir == 0)
2237 return (brt);
2238 if (dir > 0)
2239 return (NULL);
2240 }
2241
2242 return (NULL);
2243 }
2244
2245 /*
2246 * bridge_rtnode_insert:
2247 *
2248 * Insert the specified bridge node into the route table. We
2249 * assume the entry is not already in the table.
2250 */
2251 static int
2252 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2253 {
2254 struct bridge_rtnode *lbrt;
2255 uint32_t hash;
2256 int dir;
2257
2258 KASSERT(mutex_owned(sc->sc_rtlist_lock));
2259
2260 hash = bridge_rthash(sc, brt->brt_addr);
2261
2262 lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
2263 if (lbrt == NULL) {
2264 LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
2265 goto out;
2266 }
2267
2268 do {
2269 dir = memcmp(brt->brt_addr, lbrt->brt_addr, ETHER_ADDR_LEN);
2270 if (dir == 0)
2271 return (EEXIST);
2272 if (dir > 0) {
2273 LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
2274 goto out;
2275 }
2276 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
2277 LIST_INSERT_AFTER(lbrt, brt, brt_hash);
2278 goto out;
2279 }
2280 lbrt = LIST_NEXT(lbrt, brt_hash);
2281 } while (lbrt != NULL);
2282
2283 #ifdef DIAGNOSTIC
2284 panic("bridge_rtnode_insert: impossible");
2285 #endif
2286
2287 out:
2288 LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
2289 sc->sc_brtcnt++;
2290
2291 return (0);
2292 }
2293
2294 /*
2295 * bridge_rtnode_destroy:
2296 *
2297 * Destroy a bridge rtnode.
2298 */
2299 static void
2300 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
2301 {
2302
2303 KASSERT(mutex_owned(sc->sc_rtlist_lock));
2304
2305 LIST_REMOVE(brt, brt_hash);
2306
2307 LIST_REMOVE(brt, brt_list);
2308 sc->sc_brtcnt--;
2309 pool_put(&bridge_rtnode_pool, brt);
2310 }
2311
2312 #if defined(BRIDGE_IPF)
2313 extern pfil_head_t *inet_pfil_hook; /* XXX */
2314 extern pfil_head_t *inet6_pfil_hook; /* XXX */
2315
2316 /*
2317 * Send bridge packets through IPF if they are one of the types IPF can deal
2318 * with, or if they are ARP or REVARP. (IPF will pass ARP and REVARP without
2319 * question.)
2320 */
2321 static int
2322 bridge_ipf(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
2323 {
2324 int snap, error;
2325 struct ether_header *eh1, eh2;
2326 struct llc llc1;
2327 uint16_t ether_type;
2328
2329 snap = 0;
2330 error = -1; /* Default error if not error == 0 */
2331 eh1 = mtod(*mp, struct ether_header *);
2332 ether_type = ntohs(eh1->ether_type);
2333
2334 /*
2335 * Check for SNAP/LLC.
2336 */
2337 if (ether_type < ETHERMTU) {
2338 struct llc *llc2 = (struct llc *)(eh1 + 1);
2339
2340 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
2341 llc2->llc_dsap == LLC_SNAP_LSAP &&
2342 llc2->llc_ssap == LLC_SNAP_LSAP &&
2343 llc2->llc_control == LLC_UI) {
2344 ether_type = htons(llc2->llc_un.type_snap.ether_type);
2345 snap = 1;
2346 }
2347 }
2348
2349 /*
2350 * If we're trying to filter bridge traffic, don't look at anything
2351 * other than IP and ARP traffic. If the filter doesn't understand
2352 * IPv6, don't allow IPv6 through the bridge either. This is lame
2353 * since if we really wanted, say, an AppleTalk filter, we are hosed,
2354 * but of course we don't have an AppleTalk filter to begin with.
2355 * (Note that since IPF doesn't understand ARP it will pass *ALL*
2356 * ARP traffic.)
2357 */
2358 switch (ether_type) {
2359 case ETHERTYPE_ARP:
2360 case ETHERTYPE_REVARP:
2361 return 0; /* Automatically pass */
2362 case ETHERTYPE_IP:
2363 # ifdef INET6
2364 case ETHERTYPE_IPV6:
2365 # endif /* INET6 */
2366 break;
2367 default:
2368 goto bad;
2369 }
2370
2371 /* Strip off the Ethernet header and keep a copy. */
2372 m_copydata(*mp, 0, ETHER_HDR_LEN, (void *) &eh2);
2373 m_adj(*mp, ETHER_HDR_LEN);
2374
2375 /* Strip off snap header, if present */
2376 if (snap) {
2377 m_copydata(*mp, 0, sizeof(struct llc), (void *) &llc1);
2378 m_adj(*mp, sizeof(struct llc));
2379 }
2380
2381 /*
2382 * Check basic packet sanity and run IPF through pfil.
2383 */
2384 KASSERT(!cpu_intr_p());
2385 switch (ether_type)
2386 {
2387 case ETHERTYPE_IP :
2388 error = (dir == PFIL_IN) ? bridge_ip_checkbasic(mp) : 0;
2389 if (error == 0)
2390 error = pfil_run_hooks(inet_pfil_hook, mp, ifp, dir);
2391 break;
2392 # ifdef INET6
2393 case ETHERTYPE_IPV6 :
2394 error = (dir == PFIL_IN) ? bridge_ip6_checkbasic(mp) : 0;
2395 if (error == 0)
2396 error = pfil_run_hooks(inet6_pfil_hook, mp, ifp, dir);
2397 break;
2398 # endif
2399 default :
2400 error = 0;
2401 break;
2402 }
2403
2404 if (*mp == NULL)
2405 return error;
2406 if (error != 0)
2407 goto bad;
2408
2409 error = -1;
2410
2411 /*
2412 * Finally, put everything back the way it was and return
2413 */
2414 if (snap) {
2415 M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
2416 if (*mp == NULL)
2417 return error;
2418 bcopy(&llc1, mtod(*mp, void *), sizeof(struct llc));
2419 }
2420
2421 M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
2422 if (*mp == NULL)
2423 return error;
2424 bcopy(&eh2, mtod(*mp, void *), ETHER_HDR_LEN);
2425
2426 return 0;
2427
2428 bad:
2429 m_freem(*mp);
2430 *mp = NULL;
2431 return error;
2432 }
2433
2434 /*
2435 * Perform basic checks on header size since
2436 * IPF assumes ip_input has already processed
2437 * it for it. Cut-and-pasted from ip_input.c.
2438 * Given how simple the IPv6 version is,
2439 * does the IPv4 version really need to be
2440 * this complicated?
2441 *
2442 * XXX Should we update ipstat here, or not?
2443 * XXX Right now we update ipstat but not
2444 * XXX csum_counter.
2445 */
2446 static int
2447 bridge_ip_checkbasic(struct mbuf **mp)
2448 {
2449 struct mbuf *m = *mp;
2450 struct ip *ip;
2451 int len, hlen;
2452
2453 if (*mp == NULL)
2454 return -1;
2455
2456 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
2457 if ((m = m_copyup(m, sizeof(struct ip),
2458 (max_linkhdr + 3) & ~3)) == NULL) {
2459 /* XXXJRT new stat, please */
2460 ip_statinc(IP_STAT_TOOSMALL);
2461 goto bad;
2462 }
2463 } else if (__predict_false(m->m_len < sizeof (struct ip))) {
2464 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
2465 ip_statinc(IP_STAT_TOOSMALL);
2466 goto bad;
2467 }
2468 }
2469 ip = mtod(m, struct ip *);
2470 if (ip == NULL) goto bad;
2471
2472 if (ip->ip_v != IPVERSION) {
2473 ip_statinc(IP_STAT_BADVERS);
2474 goto bad;
2475 }
2476 hlen = ip->ip_hl << 2;
2477 if (hlen < sizeof(struct ip)) { /* minimum header length */
2478 ip_statinc(IP_STAT_BADHLEN);
2479 goto bad;
2480 }
2481 if (hlen > m->m_len) {
2482 if ((m = m_pullup(m, hlen)) == 0) {
2483 ip_statinc(IP_STAT_BADHLEN);
2484 goto bad;
2485 }
2486 ip = mtod(m, struct ip *);
2487 if (ip == NULL) goto bad;
2488 }
2489
2490 switch (m->m_pkthdr.csum_flags &
2491 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) |
2492 M_CSUM_IPv4_BAD)) {
2493 case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
2494 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); */
2495 goto bad;
2496
2497 case M_CSUM_IPv4:
2498 /* Checksum was okay. */
2499 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); */
2500 break;
2501
2502 default:
2503 /* Must compute it ourselves. */
2504 /* INET_CSUM_COUNTER_INCR(&ip_swcsum); */
2505 if (in_cksum(m, hlen) != 0)
2506 goto bad;
2507 break;
2508 }
2509
2510 /* Retrieve the packet length. */
2511 len = ntohs(ip->ip_len);
2512
2513 /*
2514 * Check for additional length bogosity
2515 */
2516 if (len < hlen) {
2517 ip_statinc(IP_STAT_BADLEN);
2518 goto bad;
2519 }
2520
2521 /*
2522 * Check that the amount of data in the buffers
2523 * is as at least much as the IP header would have us expect.
2524 * Drop packet if shorter than we expect.
2525 */
2526 if (m->m_pkthdr.len < len) {
2527 ip_statinc(IP_STAT_TOOSHORT);
2528 goto bad;
2529 }
2530
2531 /* Checks out, proceed */
2532 *mp = m;
2533 return 0;
2534
2535 bad:
2536 *mp = m;
2537 return -1;
2538 }
2539
2540 # ifdef INET6
2541 /*
2542 * Same as above, but for IPv6.
2543 * Cut-and-pasted from ip6_input.c.
2544 * XXX Should we update ip6stat, or not?
2545 */
2546 static int
2547 bridge_ip6_checkbasic(struct mbuf **mp)
2548 {
2549 struct mbuf *m = *mp;
2550 struct ip6_hdr *ip6;
2551
2552 /*
2553 * If the IPv6 header is not aligned, slurp it up into a new
2554 * mbuf with space for link headers, in the event we forward
2555 * it. Otherwise, if it is aligned, make sure the entire base
2556 * IPv6 header is in the first mbuf of the chain.
2557 */
2558 if (IP6_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
2559 struct ifnet *inifp = m->m_pkthdr.rcvif;
2560 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
2561 (max_linkhdr + 3) & ~3)) == NULL) {
2562 /* XXXJRT new stat, please */
2563 ip6_statinc(IP6_STAT_TOOSMALL);
2564 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
2565 goto bad;
2566 }
2567 } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
2568 struct ifnet *inifp = m->m_pkthdr.rcvif;
2569 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
2570 ip6_statinc(IP6_STAT_TOOSMALL);
2571 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
2572 goto bad;
2573 }
2574 }
2575
2576 ip6 = mtod(m, struct ip6_hdr *);
2577
2578 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
2579 ip6_statinc(IP6_STAT_BADVERS);
2580 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
2581 goto bad;
2582 }
2583
2584 /* Checks out, proceed */
2585 *mp = m;
2586 return 0;
2587
2588 bad:
2589 *mp = m;
2590 return -1;
2591 }
2592 # endif /* INET6 */
2593 #endif /* BRIDGE_IPF */
2594