if_bridge.c revision 1.92 1 /* $NetBSD: if_bridge.c,v 1.92 2014/12/22 09:42:45 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 1.92 2014/12/22 09:42:45 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 /* FIXME: doesn't work with non-IFF_SIMPLEX interfaces */
804 if ((ifs->if_flags & IFF_SIMPLEX) == 0)
805 return EINVAL;
806
807 bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT);
808 if (bif == NULL)
809 return (ENOMEM);
810
811 switch (ifs->if_type) {
812 case IFT_ETHER:
813 /*
814 * Place the interface into promiscuous mode.
815 */
816 error = ifpromisc(ifs, 1);
817 if (error)
818 goto out;
819 break;
820 default:
821 error = EINVAL;
822 goto out;
823 }
824
825 bif->bif_ifp = ifs;
826 bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
827 bif->bif_priority = BSTP_DEFAULT_PORT_PRIORITY;
828 bif->bif_path_cost = BSTP_DEFAULT_PATH_COST;
829 bif->bif_refs = 0;
830 bif->bif_waiting = false;
831
832 BRIDGE_LOCK(sc);
833
834 ifs->if_bridge = sc;
835 ifs->if_bridgeif = bif;
836 LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
837 ifs->if_input = bridge_input;
838
839 BRIDGE_UNLOCK(sc);
840
841 if (sc->sc_if.if_flags & IFF_RUNNING)
842 bstp_initialization(sc);
843 else
844 bstp_stop(sc);
845
846 out:
847 if (error) {
848 if (bif != NULL)
849 free(bif, M_DEVBUF);
850 }
851 return (error);
852 }
853
854 static int
855 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
856 {
857 struct ifbreq *req = arg;
858 const char *name = req->ifbr_ifsname;
859 struct bridge_iflist *bif;
860 struct ifnet *ifs;
861
862 BRIDGE_LOCK(sc);
863
864 /*
865 * Don't use bridge_lookup_member. We want to get a member
866 * with bif_refs == 0.
867 */
868 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
869 ifs = bif->bif_ifp;
870 if (strcmp(ifs->if_xname, name) == 0)
871 break;
872 }
873
874 if (bif == NULL) {
875 BRIDGE_UNLOCK(sc);
876 return ENOENT;
877 }
878
879 bridge_delete_member(sc, bif);
880
881 BRIDGE_UNLOCK(sc);
882
883 switch (ifs->if_type) {
884 case IFT_ETHER:
885 /*
886 * Take the interface out of promiscuous mode.
887 * Don't call it with holding sc_iflist_lock.
888 */
889 (void) ifpromisc(ifs, 0);
890 break;
891 default:
892 #ifdef DIAGNOSTIC
893 panic("bridge_delete_member: impossible");
894 #endif
895 break;
896 }
897
898 bridge_rtdelete(sc, ifs);
899
900 if (sc->sc_if.if_flags & IFF_RUNNING)
901 bstp_initialization(sc);
902
903 return 0;
904 }
905
906 static int
907 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
908 {
909 struct ifbreq *req = arg;
910 struct bridge_iflist *bif;
911
912 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
913 if (bif == NULL)
914 return (ENOENT);
915
916 req->ifbr_ifsflags = bif->bif_flags;
917 req->ifbr_state = bif->bif_state;
918 req->ifbr_priority = bif->bif_priority;
919 req->ifbr_path_cost = bif->bif_path_cost;
920 req->ifbr_portno = bif->bif_ifp->if_index & 0xff;
921
922 bridge_release_member(sc, bif);
923
924 return (0);
925 }
926
927 static int
928 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
929 {
930 struct ifbreq *req = arg;
931 struct bridge_iflist *bif;
932
933 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
934 if (bif == NULL)
935 return (ENOENT);
936
937 if (req->ifbr_ifsflags & IFBIF_STP) {
938 switch (bif->bif_ifp->if_type) {
939 case IFT_ETHER:
940 /* These can do spanning tree. */
941 break;
942
943 default:
944 /* Nothing else can. */
945 bridge_release_member(sc, bif);
946 return (EINVAL);
947 }
948 }
949
950 bif->bif_flags = req->ifbr_ifsflags;
951
952 bridge_release_member(sc, bif);
953
954 if (sc->sc_if.if_flags & IFF_RUNNING)
955 bstp_initialization(sc);
956
957 return (0);
958 }
959
960 static int
961 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
962 {
963 struct ifbrparam *param = arg;
964
965 sc->sc_brtmax = param->ifbrp_csize;
966 bridge_rttrim(sc);
967
968 return (0);
969 }
970
971 static int
972 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
973 {
974 struct ifbrparam *param = arg;
975
976 param->ifbrp_csize = sc->sc_brtmax;
977
978 return (0);
979 }
980
981 static int
982 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
983 {
984 struct ifbifconf *bifc = arg;
985 struct bridge_iflist *bif;
986 struct ifbreq *breqs;
987 int i, count, error = 0;
988
989 BRIDGE_LOCK(sc);
990
991 count = 0;
992 LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
993 count++;
994
995 if (bifc->ifbic_len == 0 || bifc->ifbic_len < (sizeof(*breqs) * count)) {
996 BRIDGE_UNLOCK(sc);
997 /* Tell that a larger buffer is needed */
998 bifc->ifbic_len = sizeof(*breqs) * count;
999 return 0;
1000 }
1001
1002 breqs = malloc(sizeof(*breqs) * count, M_DEVBUF, M_NOWAIT);
1003
1004 i = 0;
1005 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1006 struct ifbreq *breq = &breqs[i++];
1007 memset(breq, 0, sizeof(*breq));
1008
1009 strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname,
1010 sizeof(breq->ifbr_ifsname));
1011 breq->ifbr_ifsflags = bif->bif_flags;
1012 breq->ifbr_state = bif->bif_state;
1013 breq->ifbr_priority = bif->bif_priority;
1014 breq->ifbr_path_cost = bif->bif_path_cost;
1015 breq->ifbr_portno = bif->bif_ifp->if_index & 0xff;
1016 }
1017
1018 /* Don't call copyout with holding the mutex */
1019 BRIDGE_UNLOCK(sc);
1020
1021 for (i = 0; i < count; i++) {
1022 error = copyout(&breqs[i], bifc->ifbic_req + i, sizeof(*breqs));
1023 if (error)
1024 break;
1025 }
1026 bifc->ifbic_len = sizeof(*breqs) * i;
1027
1028 free(breqs, M_DEVBUF);
1029
1030 return error;
1031 }
1032
1033 static int
1034 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1035 {
1036 struct ifbaconf *bac = arg;
1037 struct bridge_rtnode *brt;
1038 struct ifbareq bareq;
1039 int count = 0, error = 0, len;
1040
1041 if (bac->ifbac_len == 0)
1042 return (0);
1043
1044 mutex_enter(sc->sc_rtlist_lock);
1045
1046 len = bac->ifbac_len;
1047 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1048 if (len < sizeof(bareq))
1049 goto out;
1050 memset(&bareq, 0, sizeof(bareq));
1051 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1052 sizeof(bareq.ifba_ifsname));
1053 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1054 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
1055 bareq.ifba_expire = brt->brt_expire - time_uptime;
1056 } else
1057 bareq.ifba_expire = 0;
1058 bareq.ifba_flags = brt->brt_flags;
1059
1060 error = copyout(&bareq, bac->ifbac_req + count, sizeof(bareq));
1061 if (error)
1062 goto out;
1063 count++;
1064 len -= sizeof(bareq);
1065 }
1066 out:
1067 mutex_exit(sc->sc_rtlist_lock);
1068
1069 bac->ifbac_len = sizeof(bareq) * count;
1070 return (error);
1071 }
1072
1073 static int
1074 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1075 {
1076 struct ifbareq *req = arg;
1077 struct bridge_iflist *bif;
1078 int error;
1079
1080 bif = bridge_lookup_member(sc, req->ifba_ifsname);
1081 if (bif == NULL)
1082 return (ENOENT);
1083
1084 error = bridge_rtupdate(sc, req->ifba_dst, bif->bif_ifp, 1,
1085 req->ifba_flags);
1086
1087 bridge_release_member(sc, bif);
1088
1089 return (error);
1090 }
1091
1092 static int
1093 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1094 {
1095 struct ifbrparam *param = arg;
1096
1097 sc->sc_brttimeout = param->ifbrp_ctime;
1098
1099 return (0);
1100 }
1101
1102 static int
1103 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1104 {
1105 struct ifbrparam *param = arg;
1106
1107 param->ifbrp_ctime = sc->sc_brttimeout;
1108
1109 return (0);
1110 }
1111
1112 static int
1113 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1114 {
1115 struct ifbareq *req = arg;
1116
1117 return (bridge_rtdaddr(sc, req->ifba_dst));
1118 }
1119
1120 static int
1121 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1122 {
1123 struct ifbreq *req = arg;
1124
1125 bridge_rtflush(sc, req->ifbr_ifsflags);
1126
1127 return (0);
1128 }
1129
1130 static int
1131 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1132 {
1133 struct ifbrparam *param = arg;
1134
1135 param->ifbrp_prio = sc->sc_bridge_priority;
1136
1137 return (0);
1138 }
1139
1140 static int
1141 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1142 {
1143 struct ifbrparam *param = arg;
1144
1145 sc->sc_bridge_priority = param->ifbrp_prio;
1146
1147 if (sc->sc_if.if_flags & IFF_RUNNING)
1148 bstp_initialization(sc);
1149
1150 return (0);
1151 }
1152
1153 static int
1154 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1155 {
1156 struct ifbrparam *param = arg;
1157
1158 param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8;
1159
1160 return (0);
1161 }
1162
1163 static int
1164 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1165 {
1166 struct ifbrparam *param = arg;
1167
1168 if (param->ifbrp_hellotime == 0)
1169 return (EINVAL);
1170 sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8;
1171
1172 if (sc->sc_if.if_flags & IFF_RUNNING)
1173 bstp_initialization(sc);
1174
1175 return (0);
1176 }
1177
1178 static int
1179 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1180 {
1181 struct ifbrparam *param = arg;
1182
1183 param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8;
1184
1185 return (0);
1186 }
1187
1188 static int
1189 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1190 {
1191 struct ifbrparam *param = arg;
1192
1193 if (param->ifbrp_fwddelay == 0)
1194 return (EINVAL);
1195 sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8;
1196
1197 if (sc->sc_if.if_flags & IFF_RUNNING)
1198 bstp_initialization(sc);
1199
1200 return (0);
1201 }
1202
1203 static int
1204 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1205 {
1206 struct ifbrparam *param = arg;
1207
1208 param->ifbrp_maxage = sc->sc_bridge_max_age >> 8;
1209
1210 return (0);
1211 }
1212
1213 static int
1214 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1215 {
1216 struct ifbrparam *param = arg;
1217
1218 if (param->ifbrp_maxage == 0)
1219 return (EINVAL);
1220 sc->sc_bridge_max_age = param->ifbrp_maxage << 8;
1221
1222 if (sc->sc_if.if_flags & IFF_RUNNING)
1223 bstp_initialization(sc);
1224
1225 return (0);
1226 }
1227
1228 static int
1229 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1230 {
1231 struct ifbreq *req = arg;
1232 struct bridge_iflist *bif;
1233
1234 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1235 if (bif == NULL)
1236 return (ENOENT);
1237
1238 bif->bif_priority = req->ifbr_priority;
1239
1240 if (sc->sc_if.if_flags & IFF_RUNNING)
1241 bstp_initialization(sc);
1242
1243 bridge_release_member(sc, bif);
1244
1245 return (0);
1246 }
1247
1248 #if defined(BRIDGE_IPF)
1249 static int
1250 bridge_ioctl_gfilt(struct bridge_softc *sc, void *arg)
1251 {
1252 struct ifbrparam *param = arg;
1253
1254 param->ifbrp_filter = sc->sc_filter_flags;
1255
1256 return (0);
1257 }
1258
1259 static int
1260 bridge_ioctl_sfilt(struct bridge_softc *sc, void *arg)
1261 {
1262 struct ifbrparam *param = arg;
1263 uint32_t nflags, oflags;
1264
1265 if (param->ifbrp_filter & ~IFBF_FILT_MASK)
1266 return (EINVAL);
1267
1268 nflags = param->ifbrp_filter;
1269 oflags = sc->sc_filter_flags;
1270
1271 if ((nflags & IFBF_FILT_USEIPF) && !(oflags & IFBF_FILT_USEIPF)) {
1272 pfil_add_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
1273 sc->sc_if.if_pfil);
1274 }
1275 if (!(nflags & IFBF_FILT_USEIPF) && (oflags & IFBF_FILT_USEIPF)) {
1276 pfil_remove_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
1277 sc->sc_if.if_pfil);
1278 }
1279
1280 sc->sc_filter_flags = nflags;
1281
1282 return (0);
1283 }
1284 #endif /* BRIDGE_IPF */
1285
1286 static int
1287 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1288 {
1289 struct ifbreq *req = arg;
1290 struct bridge_iflist *bif;
1291
1292 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1293 if (bif == NULL)
1294 return (ENOENT);
1295
1296 bif->bif_path_cost = req->ifbr_path_cost;
1297
1298 if (sc->sc_if.if_flags & IFF_RUNNING)
1299 bstp_initialization(sc);
1300
1301 bridge_release_member(sc, bif);
1302
1303 return (0);
1304 }
1305
1306 /*
1307 * bridge_ifdetach:
1308 *
1309 * Detach an interface from a bridge. Called when a member
1310 * interface is detaching.
1311 */
1312 void
1313 bridge_ifdetach(struct ifnet *ifp)
1314 {
1315 struct bridge_softc *sc = ifp->if_bridge;
1316 struct ifbreq breq;
1317
1318 /* ioctl_lock should prevent this from happening */
1319 KASSERT(sc != NULL);
1320
1321 memset(&breq, 0, sizeof(breq));
1322 strlcpy(breq.ifbr_ifsname, ifp->if_xname, sizeof(breq.ifbr_ifsname));
1323
1324 (void) bridge_ioctl_del(sc, &breq);
1325 }
1326
1327 /*
1328 * bridge_init:
1329 *
1330 * Initialize a bridge interface.
1331 */
1332 static int
1333 bridge_init(struct ifnet *ifp)
1334 {
1335 struct bridge_softc *sc = ifp->if_softc;
1336
1337 if (ifp->if_flags & IFF_RUNNING)
1338 return (0);
1339
1340 callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1341 bridge_timer, sc);
1342
1343 ifp->if_flags |= IFF_RUNNING;
1344 bstp_initialization(sc);
1345 return (0);
1346 }
1347
1348 /*
1349 * bridge_stop:
1350 *
1351 * Stop the bridge interface.
1352 */
1353 static void
1354 bridge_stop(struct ifnet *ifp, int disable)
1355 {
1356 struct bridge_softc *sc = ifp->if_softc;
1357
1358 if ((ifp->if_flags & IFF_RUNNING) == 0)
1359 return;
1360
1361 callout_stop(&sc->sc_brcallout);
1362 bstp_stop(sc);
1363
1364 bridge_rtflush(sc, IFBF_FLUSHDYN);
1365
1366 ifp->if_flags &= ~IFF_RUNNING;
1367 }
1368
1369 /*
1370 * bridge_enqueue:
1371 *
1372 * Enqueue a packet on a bridge member interface.
1373 */
1374 void
1375 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m,
1376 int runfilt)
1377 {
1378 ALTQ_DECL(struct altq_pktattr pktattr;)
1379 int len, error;
1380 short mflags;
1381
1382 /*
1383 * Clear any in-bound checksum flags for this packet.
1384 */
1385 m->m_pkthdr.csum_flags = 0;
1386
1387 if (runfilt) {
1388 if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
1389 dst_ifp, PFIL_OUT) != 0) {
1390 if (m != NULL)
1391 m_freem(m);
1392 return;
1393 }
1394 if (m == NULL)
1395 return;
1396 }
1397
1398 #ifdef ALTQ
1399 /*
1400 * If ALTQ is enabled on the member interface, do
1401 * classification; the queueing discipline might
1402 * not require classification, but might require
1403 * the address family/header pointer in the pktattr.
1404 */
1405 if (ALTQ_IS_ENABLED(&dst_ifp->if_snd)) {
1406 /* XXX IFT_ETHER */
1407 altq_etherclassify(&dst_ifp->if_snd, m, &pktattr);
1408 }
1409 #endif /* ALTQ */
1410
1411 len = m->m_pkthdr.len;
1412 m->m_flags |= M_PROTO1;
1413 mflags = m->m_flags;
1414
1415 IFQ_ENQUEUE(&dst_ifp->if_snd, m, &pktattr, error);
1416
1417 if (error) {
1418 /* mbuf is already freed */
1419 sc->sc_if.if_oerrors++;
1420 return;
1421 }
1422
1423 sc->sc_if.if_opackets++;
1424 sc->sc_if.if_obytes += len;
1425
1426 dst_ifp->if_obytes += len;
1427
1428 if (mflags & M_MCAST) {
1429 sc->sc_if.if_omcasts++;
1430 dst_ifp->if_omcasts++;
1431 }
1432
1433 if ((dst_ifp->if_flags & IFF_OACTIVE) == 0)
1434 (*dst_ifp->if_start)(dst_ifp);
1435 }
1436
1437 /*
1438 * bridge_output:
1439 *
1440 * Send output from a bridge member interface. This
1441 * performs the bridging function for locally originated
1442 * packets.
1443 *
1444 * The mbuf has the Ethernet header already attached. We must
1445 * enqueue or free the mbuf before returning.
1446 */
1447 int
1448 bridge_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa,
1449 struct rtentry *rt)
1450 {
1451 struct ether_header *eh;
1452 struct ifnet *dst_if;
1453 struct bridge_softc *sc;
1454 #ifndef BRIDGE_MPSAFE
1455 int s;
1456 #endif
1457
1458 if (m->m_len < ETHER_HDR_LEN) {
1459 m = m_pullup(m, ETHER_HDR_LEN);
1460 if (m == NULL)
1461 return (0);
1462 }
1463
1464 eh = mtod(m, struct ether_header *);
1465 sc = ifp->if_bridge;
1466
1467 #ifndef BRIDGE_MPSAFE
1468 s = splnet();
1469 #endif
1470
1471 /*
1472 * If bridge is down, but the original output interface is up,
1473 * go ahead and send out that interface. Otherwise, the packet
1474 * is dropped below.
1475 */
1476 if (__predict_false(sc == NULL) ||
1477 (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
1478 dst_if = ifp;
1479 goto sendunicast;
1480 }
1481
1482 /*
1483 * If the packet is a multicast, or we don't know a better way to
1484 * get there, send to all interfaces.
1485 */
1486 if (ETHER_IS_MULTICAST(eh->ether_dhost))
1487 dst_if = NULL;
1488 else
1489 dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1490 if (dst_if == NULL) {
1491 struct bridge_iflist *bif;
1492 struct mbuf *mc;
1493 int used = 0;
1494
1495 BRIDGE_LOCK(sc);
1496
1497 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1498 dst_if = bif->bif_ifp;
1499 if ((dst_if->if_flags & IFF_RUNNING) == 0)
1500 continue;
1501
1502 /*
1503 * If this is not the original output interface,
1504 * and the interface is participating in spanning
1505 * tree, make sure the port is in a state that
1506 * allows forwarding.
1507 */
1508 if (dst_if != ifp &&
1509 (bif->bif_flags & IFBIF_STP) != 0) {
1510 switch (bif->bif_state) {
1511 case BSTP_IFSTATE_BLOCKING:
1512 case BSTP_IFSTATE_LISTENING:
1513 case BSTP_IFSTATE_DISABLED:
1514 continue;
1515 }
1516 }
1517
1518 if (LIST_NEXT(bif, bif_next) == NULL) {
1519 used = 1;
1520 mc = m;
1521 } else {
1522 mc = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1523 if (mc == NULL) {
1524 sc->sc_if.if_oerrors++;
1525 continue;
1526 }
1527 }
1528
1529 bridge_enqueue(sc, dst_if, mc, 0);
1530 }
1531
1532 BRIDGE_UNLOCK(sc);
1533
1534 if (used == 0)
1535 m_freem(m);
1536 #ifndef BRIDGE_MPSAFE
1537 splx(s);
1538 #endif
1539 return (0);
1540 }
1541
1542 sendunicast:
1543 /*
1544 * XXX Spanning tree consideration here?
1545 */
1546
1547 if ((dst_if->if_flags & IFF_RUNNING) == 0) {
1548 m_freem(m);
1549 #ifndef BRIDGE_MPSAFE
1550 splx(s);
1551 #endif
1552 return (0);
1553 }
1554
1555 bridge_enqueue(sc, dst_if, m, 0);
1556
1557 #ifndef BRIDGE_MPSAFE
1558 splx(s);
1559 #endif
1560 return (0);
1561 }
1562
1563 /*
1564 * bridge_start:
1565 *
1566 * Start output on a bridge.
1567 *
1568 * NOTE: This routine should never be called in this implementation.
1569 */
1570 static void
1571 bridge_start(struct ifnet *ifp)
1572 {
1573
1574 printf("%s: bridge_start() called\n", ifp->if_xname);
1575 }
1576
1577 /*
1578 * bridge_forward:
1579 *
1580 * The forwarding function of the bridge.
1581 */
1582 static void
1583 bridge_forward(void *v)
1584 {
1585 struct bridge_softc *sc = v;
1586 struct mbuf *m;
1587 struct bridge_iflist *bif;
1588 struct ifnet *src_if, *dst_if;
1589 struct ether_header *eh;
1590 #ifndef BRIDGE_MPSAFE
1591 int s;
1592
1593 KERNEL_LOCK(1, NULL);
1594 mutex_enter(softnet_lock);
1595 #endif
1596
1597 if ((sc->sc_if.if_flags & IFF_RUNNING) == 0) {
1598 #ifndef BRIDGE_MPSAFE
1599 mutex_exit(softnet_lock);
1600 KERNEL_UNLOCK_ONE(NULL);
1601 #endif
1602 return;
1603 }
1604
1605 #ifndef BRIDGE_MPSAFE
1606 s = splnet();
1607 #endif
1608 while ((m = pktq_dequeue(sc->sc_fwd_pktq)) != NULL) {
1609 src_if = m->m_pkthdr.rcvif;
1610
1611 sc->sc_if.if_ipackets++;
1612 sc->sc_if.if_ibytes += m->m_pkthdr.len;
1613
1614 /*
1615 * Look up the bridge_iflist.
1616 */
1617 bif = bridge_lookup_member_if(sc, src_if);
1618 if (bif == NULL) {
1619 /* Interface is not a bridge member (anymore?) */
1620 m_freem(m);
1621 continue;
1622 }
1623
1624 if (bif->bif_flags & IFBIF_STP) {
1625 switch (bif->bif_state) {
1626 case BSTP_IFSTATE_BLOCKING:
1627 case BSTP_IFSTATE_LISTENING:
1628 case BSTP_IFSTATE_DISABLED:
1629 m_freem(m);
1630 bridge_release_member(sc, bif);
1631 continue;
1632 }
1633 }
1634
1635 eh = mtod(m, struct ether_header *);
1636
1637 /*
1638 * If the interface is learning, and the source
1639 * address is valid and not multicast, record
1640 * the address.
1641 */
1642 if ((bif->bif_flags & IFBIF_LEARNING) != 0 &&
1643 ETHER_IS_MULTICAST(eh->ether_shost) == 0 &&
1644 (eh->ether_shost[0] == 0 &&
1645 eh->ether_shost[1] == 0 &&
1646 eh->ether_shost[2] == 0 &&
1647 eh->ether_shost[3] == 0 &&
1648 eh->ether_shost[4] == 0 &&
1649 eh->ether_shost[5] == 0) == 0) {
1650 (void) bridge_rtupdate(sc, eh->ether_shost,
1651 src_if, 0, IFBAF_DYNAMIC);
1652 }
1653
1654 if ((bif->bif_flags & IFBIF_STP) != 0 &&
1655 bif->bif_state == BSTP_IFSTATE_LEARNING) {
1656 m_freem(m);
1657 bridge_release_member(sc, bif);
1658 continue;
1659 }
1660
1661 bridge_release_member(sc, bif);
1662
1663 /*
1664 * At this point, the port either doesn't participate
1665 * in spanning tree or it is in the forwarding state.
1666 */
1667
1668 /*
1669 * If the packet is unicast, destined for someone on
1670 * "this" side of the bridge, drop it.
1671 */
1672 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
1673 dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1674 if (src_if == dst_if) {
1675 m_freem(m);
1676 continue;
1677 }
1678 } else {
1679 /* ...forward it to all interfaces. */
1680 sc->sc_if.if_imcasts++;
1681 dst_if = NULL;
1682 }
1683
1684 if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
1685 m->m_pkthdr.rcvif, PFIL_IN) != 0) {
1686 if (m != NULL)
1687 m_freem(m);
1688 continue;
1689 }
1690 if (m == NULL)
1691 continue;
1692
1693 if (dst_if == NULL) {
1694 bridge_broadcast(sc, src_if, m);
1695 continue;
1696 }
1697
1698 /*
1699 * At this point, we're dealing with a unicast frame
1700 * going to a different interface.
1701 */
1702 if ((dst_if->if_flags & IFF_RUNNING) == 0) {
1703 m_freem(m);
1704 continue;
1705 }
1706
1707 bif = bridge_lookup_member_if(sc, dst_if);
1708 if (bif == NULL) {
1709 /* Not a member of the bridge (anymore?) */
1710 m_freem(m);
1711 continue;
1712 }
1713
1714 if (bif->bif_flags & IFBIF_STP) {
1715 switch (bif->bif_state) {
1716 case BSTP_IFSTATE_DISABLED:
1717 case BSTP_IFSTATE_BLOCKING:
1718 m_freem(m);
1719 bridge_release_member(sc, bif);
1720 continue;
1721 }
1722 }
1723
1724 bridge_release_member(sc, bif);
1725
1726 bridge_enqueue(sc, dst_if, m, 1);
1727 }
1728 #ifndef BRIDGE_MPSAFE
1729 splx(s);
1730 mutex_exit(softnet_lock);
1731 KERNEL_UNLOCK_ONE(NULL);
1732 #endif
1733 }
1734
1735 static bool
1736 bstp_state_before_learning(struct bridge_iflist *bif)
1737 {
1738 if (bif->bif_flags & IFBIF_STP) {
1739 switch (bif->bif_state) {
1740 case BSTP_IFSTATE_BLOCKING:
1741 case BSTP_IFSTATE_LISTENING:
1742 case BSTP_IFSTATE_DISABLED:
1743 return true;
1744 }
1745 }
1746 return false;
1747 }
1748
1749 static bool
1750 bridge_ourether(struct bridge_iflist *bif, struct ether_header *eh, int src)
1751 {
1752 uint8_t *ether = src ? eh->ether_shost : eh->ether_dhost;
1753
1754 if (memcmp(CLLADDR(bif->bif_ifp->if_sadl), ether, ETHER_ADDR_LEN) == 0
1755 #if NCARP > 0
1756 || (bif->bif_ifp->if_carp &&
1757 carp_ourether(bif->bif_ifp->if_carp, eh, IFT_ETHER, src) != NULL)
1758 #endif /* NCARP > 0 */
1759 )
1760 return true;
1761
1762 return false;
1763 }
1764
1765 /*
1766 * bridge_input:
1767 *
1768 * Receive input from a member interface. Queue the packet for
1769 * bridging if it is not for us.
1770 */
1771 static void
1772 bridge_input(struct ifnet *ifp, struct mbuf *m)
1773 {
1774 struct bridge_softc *sc = ifp->if_bridge;
1775 struct bridge_iflist *bif;
1776 struct ether_header *eh;
1777
1778 if (__predict_false(sc == NULL) ||
1779 (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
1780 ether_input(ifp, m);
1781 return;
1782 }
1783
1784 bif = bridge_lookup_member_if(sc, ifp);
1785 if (bif == NULL) {
1786 ether_input(ifp, m);
1787 return;
1788 }
1789
1790 eh = mtod(m, struct ether_header *);
1791
1792 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1793 if (memcmp(etherbroadcastaddr,
1794 eh->ether_dhost, ETHER_ADDR_LEN) == 0)
1795 m->m_flags |= M_BCAST;
1796 else
1797 m->m_flags |= M_MCAST;
1798 }
1799
1800 /*
1801 * A 'fast' path for packets addressed to interfaces that are
1802 * part of this bridge.
1803 */
1804 if (!(m->m_flags & (M_BCAST|M_MCAST)) &&
1805 !bstp_state_before_learning(bif)) {
1806 struct bridge_iflist *_bif;
1807 struct ifnet *_ifp = NULL;
1808
1809 BRIDGE_LOCK(sc);
1810 LIST_FOREACH(_bif, &sc->sc_iflist, bif_next) {
1811 /* It is destined for us. */
1812 if (bridge_ourether(_bif, eh, 0)) {
1813 if (_bif->bif_flags & IFBIF_LEARNING)
1814 (void) bridge_rtupdate(sc,
1815 eh->ether_shost, ifp, 0, IFBAF_DYNAMIC);
1816 _ifp = m->m_pkthdr.rcvif = _bif->bif_ifp;
1817 break;
1818 }
1819
1820 /* We just received a packet that we sent out. */
1821 if (bridge_ourether(_bif, eh, 1))
1822 break;
1823 }
1824 BRIDGE_UNLOCK(sc);
1825
1826 if (_bif != NULL) {
1827 bridge_release_member(sc, bif);
1828 if (_ifp != NULL)
1829 ether_input(_ifp, m);
1830 else
1831 m_freem(m);
1832 return;
1833 }
1834 }
1835
1836 /* Tap off 802.1D packets; they do not get forwarded. */
1837 if (bif->bif_flags & IFBIF_STP &&
1838 memcmp(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN) == 0) {
1839 bstp_input(sc, bif, m);
1840 bridge_release_member(sc, bif);
1841 return;
1842 }
1843
1844 /*
1845 * A normal switch would discard the packet here, but that's not what
1846 * we've done historically. This also prevents some obnoxious behaviour.
1847 */
1848 if (bstp_state_before_learning(bif)) {
1849 bridge_release_member(sc, bif);
1850 ether_input(ifp, m);
1851 return;
1852 }
1853
1854 bridge_release_member(sc, bif);
1855
1856 /* Queue the packet for bridge forwarding. */
1857 if (__predict_false(!pktq_enqueue(sc->sc_fwd_pktq, m, 0)))
1858 m_freem(m);
1859 }
1860
1861 /*
1862 * bridge_broadcast:
1863 *
1864 * Send a frame to all interfaces that are members of
1865 * the bridge, except for the one on which the packet
1866 * arrived.
1867 */
1868 static void
1869 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
1870 struct mbuf *m)
1871 {
1872 struct bridge_iflist *bif;
1873 struct mbuf *mc;
1874 struct ifnet *dst_if;
1875 bool used, bmcast;
1876
1877 used = bmcast = m->m_flags & (M_BCAST|M_MCAST);
1878
1879 BRIDGE_LOCK(sc);
1880
1881 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1882 dst_if = bif->bif_ifp;
1883 if (dst_if == src_if)
1884 continue;
1885
1886 if (bif->bif_flags & IFBIF_STP) {
1887 switch (bif->bif_state) {
1888 case BSTP_IFSTATE_BLOCKING:
1889 case BSTP_IFSTATE_DISABLED:
1890 continue;
1891 }
1892 }
1893
1894 if ((bif->bif_flags & IFBIF_DISCOVER) == 0 && !bmcast)
1895 continue;
1896
1897 if ((dst_if->if_flags & IFF_RUNNING) == 0)
1898 continue;
1899
1900 if (!used && LIST_NEXT(bif, bif_next) == NULL) {
1901 mc = m;
1902 used = true;
1903 } else {
1904 mc = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
1905 if (mc == NULL) {
1906 sc->sc_if.if_oerrors++;
1907 continue;
1908 }
1909 }
1910
1911 bridge_enqueue(sc, dst_if, mc, 1);
1912 }
1913 BRIDGE_UNLOCK(sc);
1914
1915 if (bmcast)
1916 ether_input(src_if, m);
1917 else if (!used)
1918 m_freem(m);
1919 }
1920
1921 /*
1922 * bridge_rtupdate:
1923 *
1924 * Add a bridge routing entry.
1925 */
1926 static int
1927 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
1928 struct ifnet *dst_if, int setflags, uint8_t flags)
1929 {
1930 struct bridge_rtnode *brt;
1931 int error = 0;
1932
1933 mutex_enter(sc->sc_rtlist_lock);
1934
1935 /*
1936 * A route for this destination might already exist. If so,
1937 * update it, otherwise create a new one.
1938 */
1939 if ((brt = bridge_rtnode_lookup(sc, dst)) == NULL) {
1940 if (sc->sc_brtcnt >= sc->sc_brtmax) {
1941 error = ENOSPC;
1942 goto out;
1943 }
1944
1945 /*
1946 * Allocate a new bridge forwarding node, and
1947 * initialize the expiration time and Ethernet
1948 * address.
1949 */
1950 brt = pool_get(&bridge_rtnode_pool, PR_NOWAIT);
1951 if (brt == NULL) {
1952 error = ENOMEM;
1953 goto out;
1954 }
1955
1956 memset(brt, 0, sizeof(*brt));
1957 brt->brt_expire = time_uptime + sc->sc_brttimeout;
1958 brt->brt_flags = IFBAF_DYNAMIC;
1959 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
1960
1961 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
1962 pool_put(&bridge_rtnode_pool, brt);
1963 goto out;
1964 }
1965 }
1966
1967 brt->brt_ifp = dst_if;
1968 if (setflags) {
1969 brt->brt_flags = flags;
1970 if (flags & IFBAF_STATIC)
1971 brt->brt_expire = 0;
1972 else
1973 brt->brt_expire = time_uptime + sc->sc_brttimeout;
1974 }
1975
1976 out:
1977 mutex_exit(sc->sc_rtlist_lock);
1978
1979 return error;
1980 }
1981
1982 /*
1983 * bridge_rtlookup:
1984 *
1985 * Lookup the destination interface for an address.
1986 */
1987 static struct ifnet *
1988 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr)
1989 {
1990 struct bridge_rtnode *brt;
1991 struct ifnet *ifs = NULL;
1992
1993 mutex_enter(sc->sc_rtlist_lock);
1994
1995 brt = bridge_rtnode_lookup(sc, addr);
1996 if (brt != NULL)
1997 ifs = brt->brt_ifp;
1998
1999 mutex_exit(sc->sc_rtlist_lock);
2000
2001 return ifs;
2002 }
2003
2004 /*
2005 * bridge_rttrim:
2006 *
2007 * Trim the routine table so that we have a number
2008 * of routing entries less than or equal to the
2009 * maximum number.
2010 */
2011 static void
2012 bridge_rttrim(struct bridge_softc *sc)
2013 {
2014 struct bridge_rtnode *brt, *nbrt;
2015
2016 mutex_enter(sc->sc_rtlist_lock);
2017
2018 /* Make sure we actually need to do this. */
2019 if (sc->sc_brtcnt <= sc->sc_brtmax)
2020 goto out;
2021
2022 /* Force an aging cycle; this might trim enough addresses. */
2023 bridge_rtage(sc);
2024 if (sc->sc_brtcnt <= sc->sc_brtmax)
2025 goto out;
2026
2027 for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
2028 nbrt = LIST_NEXT(brt, brt_list);
2029 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2030 bridge_rtnode_destroy(sc, brt);
2031 if (sc->sc_brtcnt <= sc->sc_brtmax)
2032 goto out;
2033 }
2034 }
2035
2036 out:
2037 mutex_exit(sc->sc_rtlist_lock);
2038
2039 return;
2040 }
2041
2042 /*
2043 * bridge_timer:
2044 *
2045 * Aging timer for the bridge.
2046 */
2047 static void
2048 bridge_timer(void *arg)
2049 {
2050 struct bridge_softc *sc = arg;
2051
2052 mutex_enter(sc->sc_rtlist_lock);
2053
2054 bridge_rtage(sc);
2055
2056 if (sc->sc_if.if_flags & IFF_RUNNING)
2057 callout_reset(&sc->sc_brcallout,
2058 bridge_rtable_prune_period * hz, bridge_timer, sc);
2059
2060 mutex_exit(sc->sc_rtlist_lock);
2061 }
2062
2063 /*
2064 * bridge_rtage:
2065 *
2066 * Perform an aging cycle.
2067 */
2068 static void
2069 bridge_rtage(struct bridge_softc *sc)
2070 {
2071 struct bridge_rtnode *brt, *nbrt;
2072
2073 KASSERT(mutex_owned(sc->sc_rtlist_lock));
2074
2075 for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
2076 nbrt = LIST_NEXT(brt, brt_list);
2077 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2078 if (time_uptime >= brt->brt_expire)
2079 bridge_rtnode_destroy(sc, brt);
2080 }
2081 }
2082 }
2083
2084 /*
2085 * bridge_rtflush:
2086 *
2087 * Remove all dynamic addresses from the bridge.
2088 */
2089 static void
2090 bridge_rtflush(struct bridge_softc *sc, int full)
2091 {
2092 struct bridge_rtnode *brt, *nbrt;
2093
2094 mutex_enter(sc->sc_rtlist_lock);
2095
2096 for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
2097 nbrt = LIST_NEXT(brt, brt_list);
2098 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2099 bridge_rtnode_destroy(sc, brt);
2100 }
2101
2102 mutex_exit(sc->sc_rtlist_lock);
2103 }
2104
2105 /*
2106 * bridge_rtdaddr:
2107 *
2108 * Remove an address from the table.
2109 */
2110 static int
2111 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr)
2112 {
2113 struct bridge_rtnode *brt;
2114 int error = 0;
2115
2116 mutex_enter(sc->sc_rtlist_lock);
2117
2118 if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL) {
2119 error = ENOENT;
2120 goto out;
2121 }
2122
2123 bridge_rtnode_destroy(sc, brt);
2124
2125 out:
2126 mutex_exit(sc->sc_rtlist_lock);
2127
2128 return error;
2129 }
2130
2131 /*
2132 * bridge_rtdelete:
2133 *
2134 * Delete routes to a speicifc member interface.
2135 */
2136 static void
2137 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp)
2138 {
2139 struct bridge_rtnode *brt, *nbrt;
2140
2141 mutex_enter(sc->sc_rtlist_lock);
2142
2143 for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
2144 nbrt = LIST_NEXT(brt, brt_list);
2145 if (brt->brt_ifp == ifp)
2146 bridge_rtnode_destroy(sc, brt);
2147 }
2148
2149 mutex_exit(sc->sc_rtlist_lock);
2150 }
2151
2152 /*
2153 * bridge_rtable_init:
2154 *
2155 * Initialize the route table for this bridge.
2156 */
2157 static int
2158 bridge_rtable_init(struct bridge_softc *sc)
2159 {
2160 int i;
2161
2162 sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2163 M_DEVBUF, M_NOWAIT);
2164 if (sc->sc_rthash == NULL)
2165 return (ENOMEM);
2166
2167 for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2168 LIST_INIT(&sc->sc_rthash[i]);
2169
2170 sc->sc_rthash_key = cprng_fast32();
2171
2172 LIST_INIT(&sc->sc_rtlist);
2173
2174 sc->sc_rtlist_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
2175
2176 return (0);
2177 }
2178
2179 /*
2180 * bridge_rtable_fini:
2181 *
2182 * Deconstruct the route table for this bridge.
2183 */
2184 static void
2185 bridge_rtable_fini(struct bridge_softc *sc)
2186 {
2187
2188 free(sc->sc_rthash, M_DEVBUF);
2189 if (sc->sc_rtlist_lock)
2190 mutex_obj_free(sc->sc_rtlist_lock);
2191 }
2192
2193 /*
2194 * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2195 * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2196 */
2197 #define mix(a, b, c) \
2198 do { \
2199 a -= b; a -= c; a ^= (c >> 13); \
2200 b -= c; b -= a; b ^= (a << 8); \
2201 c -= a; c -= b; c ^= (b >> 13); \
2202 a -= b; a -= c; a ^= (c >> 12); \
2203 b -= c; b -= a; b ^= (a << 16); \
2204 c -= a; c -= b; c ^= (b >> 5); \
2205 a -= b; a -= c; a ^= (c >> 3); \
2206 b -= c; b -= a; b ^= (a << 10); \
2207 c -= a; c -= b; c ^= (b >> 15); \
2208 } while (/*CONSTCOND*/0)
2209
2210 static inline uint32_t
2211 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2212 {
2213 uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2214
2215 b += addr[5] << 8;
2216 b += addr[4];
2217 a += addr[3] << 24;
2218 a += addr[2] << 16;
2219 a += addr[1] << 8;
2220 a += addr[0];
2221
2222 mix(a, b, c);
2223
2224 return (c & BRIDGE_RTHASH_MASK);
2225 }
2226
2227 #undef mix
2228
2229 /*
2230 * bridge_rtnode_lookup:
2231 *
2232 * Look up a bridge route node for the specified destination.
2233 */
2234 static struct bridge_rtnode *
2235 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr)
2236 {
2237 struct bridge_rtnode *brt;
2238 uint32_t hash;
2239 int dir;
2240
2241 KASSERT(mutex_owned(sc->sc_rtlist_lock));
2242
2243 hash = bridge_rthash(sc, addr);
2244 LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
2245 dir = memcmp(addr, brt->brt_addr, ETHER_ADDR_LEN);
2246 if (dir == 0)
2247 return (brt);
2248 if (dir > 0)
2249 return (NULL);
2250 }
2251
2252 return (NULL);
2253 }
2254
2255 /*
2256 * bridge_rtnode_insert:
2257 *
2258 * Insert the specified bridge node into the route table. We
2259 * assume the entry is not already in the table.
2260 */
2261 static int
2262 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2263 {
2264 struct bridge_rtnode *lbrt;
2265 uint32_t hash;
2266 int dir;
2267
2268 KASSERT(mutex_owned(sc->sc_rtlist_lock));
2269
2270 hash = bridge_rthash(sc, brt->brt_addr);
2271
2272 lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
2273 if (lbrt == NULL) {
2274 LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
2275 goto out;
2276 }
2277
2278 do {
2279 dir = memcmp(brt->brt_addr, lbrt->brt_addr, ETHER_ADDR_LEN);
2280 if (dir == 0)
2281 return (EEXIST);
2282 if (dir > 0) {
2283 LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
2284 goto out;
2285 }
2286 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
2287 LIST_INSERT_AFTER(lbrt, brt, brt_hash);
2288 goto out;
2289 }
2290 lbrt = LIST_NEXT(lbrt, brt_hash);
2291 } while (lbrt != NULL);
2292
2293 #ifdef DIAGNOSTIC
2294 panic("bridge_rtnode_insert: impossible");
2295 #endif
2296
2297 out:
2298 LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
2299 sc->sc_brtcnt++;
2300
2301 return (0);
2302 }
2303
2304 /*
2305 * bridge_rtnode_destroy:
2306 *
2307 * Destroy a bridge rtnode.
2308 */
2309 static void
2310 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
2311 {
2312
2313 KASSERT(mutex_owned(sc->sc_rtlist_lock));
2314
2315 LIST_REMOVE(brt, brt_hash);
2316
2317 LIST_REMOVE(brt, brt_list);
2318 sc->sc_brtcnt--;
2319 pool_put(&bridge_rtnode_pool, brt);
2320 }
2321
2322 #if defined(BRIDGE_IPF)
2323 extern pfil_head_t *inet_pfil_hook; /* XXX */
2324 extern pfil_head_t *inet6_pfil_hook; /* XXX */
2325
2326 /*
2327 * Send bridge packets through IPF if they are one of the types IPF can deal
2328 * with, or if they are ARP or REVARP. (IPF will pass ARP and REVARP without
2329 * question.)
2330 */
2331 static int
2332 bridge_ipf(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
2333 {
2334 int snap, error;
2335 struct ether_header *eh1, eh2;
2336 struct llc llc1;
2337 uint16_t ether_type;
2338
2339 snap = 0;
2340 error = -1; /* Default error if not error == 0 */
2341 eh1 = mtod(*mp, struct ether_header *);
2342 ether_type = ntohs(eh1->ether_type);
2343
2344 /*
2345 * Check for SNAP/LLC.
2346 */
2347 if (ether_type < ETHERMTU) {
2348 struct llc *llc2 = (struct llc *)(eh1 + 1);
2349
2350 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
2351 llc2->llc_dsap == LLC_SNAP_LSAP &&
2352 llc2->llc_ssap == LLC_SNAP_LSAP &&
2353 llc2->llc_control == LLC_UI) {
2354 ether_type = htons(llc2->llc_un.type_snap.ether_type);
2355 snap = 1;
2356 }
2357 }
2358
2359 /*
2360 * If we're trying to filter bridge traffic, don't look at anything
2361 * other than IP and ARP traffic. If the filter doesn't understand
2362 * IPv6, don't allow IPv6 through the bridge either. This is lame
2363 * since if we really wanted, say, an AppleTalk filter, we are hosed,
2364 * but of course we don't have an AppleTalk filter to begin with.
2365 * (Note that since IPF doesn't understand ARP it will pass *ALL*
2366 * ARP traffic.)
2367 */
2368 switch (ether_type) {
2369 case ETHERTYPE_ARP:
2370 case ETHERTYPE_REVARP:
2371 return 0; /* Automatically pass */
2372 case ETHERTYPE_IP:
2373 # ifdef INET6
2374 case ETHERTYPE_IPV6:
2375 # endif /* INET6 */
2376 break;
2377 default:
2378 goto bad;
2379 }
2380
2381 /* Strip off the Ethernet header and keep a copy. */
2382 m_copydata(*mp, 0, ETHER_HDR_LEN, (void *) &eh2);
2383 m_adj(*mp, ETHER_HDR_LEN);
2384
2385 /* Strip off snap header, if present */
2386 if (snap) {
2387 m_copydata(*mp, 0, sizeof(struct llc), (void *) &llc1);
2388 m_adj(*mp, sizeof(struct llc));
2389 }
2390
2391 /*
2392 * Check basic packet sanity and run IPF through pfil.
2393 */
2394 KASSERT(!cpu_intr_p());
2395 switch (ether_type)
2396 {
2397 case ETHERTYPE_IP :
2398 error = (dir == PFIL_IN) ? bridge_ip_checkbasic(mp) : 0;
2399 if (error == 0)
2400 error = pfil_run_hooks(inet_pfil_hook, mp, ifp, dir);
2401 break;
2402 # ifdef INET6
2403 case ETHERTYPE_IPV6 :
2404 error = (dir == PFIL_IN) ? bridge_ip6_checkbasic(mp) : 0;
2405 if (error == 0)
2406 error = pfil_run_hooks(inet6_pfil_hook, mp, ifp, dir);
2407 break;
2408 # endif
2409 default :
2410 error = 0;
2411 break;
2412 }
2413
2414 if (*mp == NULL)
2415 return error;
2416 if (error != 0)
2417 goto bad;
2418
2419 error = -1;
2420
2421 /*
2422 * Finally, put everything back the way it was and return
2423 */
2424 if (snap) {
2425 M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
2426 if (*mp == NULL)
2427 return error;
2428 bcopy(&llc1, mtod(*mp, void *), sizeof(struct llc));
2429 }
2430
2431 M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
2432 if (*mp == NULL)
2433 return error;
2434 bcopy(&eh2, mtod(*mp, void *), ETHER_HDR_LEN);
2435
2436 return 0;
2437
2438 bad:
2439 m_freem(*mp);
2440 *mp = NULL;
2441 return error;
2442 }
2443
2444 /*
2445 * Perform basic checks on header size since
2446 * IPF assumes ip_input has already processed
2447 * it for it. Cut-and-pasted from ip_input.c.
2448 * Given how simple the IPv6 version is,
2449 * does the IPv4 version really need to be
2450 * this complicated?
2451 *
2452 * XXX Should we update ipstat here, or not?
2453 * XXX Right now we update ipstat but not
2454 * XXX csum_counter.
2455 */
2456 static int
2457 bridge_ip_checkbasic(struct mbuf **mp)
2458 {
2459 struct mbuf *m = *mp;
2460 struct ip *ip;
2461 int len, hlen;
2462
2463 if (*mp == NULL)
2464 return -1;
2465
2466 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
2467 if ((m = m_copyup(m, sizeof(struct ip),
2468 (max_linkhdr + 3) & ~3)) == NULL) {
2469 /* XXXJRT new stat, please */
2470 ip_statinc(IP_STAT_TOOSMALL);
2471 goto bad;
2472 }
2473 } else if (__predict_false(m->m_len < sizeof (struct ip))) {
2474 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
2475 ip_statinc(IP_STAT_TOOSMALL);
2476 goto bad;
2477 }
2478 }
2479 ip = mtod(m, struct ip *);
2480 if (ip == NULL) goto bad;
2481
2482 if (ip->ip_v != IPVERSION) {
2483 ip_statinc(IP_STAT_BADVERS);
2484 goto bad;
2485 }
2486 hlen = ip->ip_hl << 2;
2487 if (hlen < sizeof(struct ip)) { /* minimum header length */
2488 ip_statinc(IP_STAT_BADHLEN);
2489 goto bad;
2490 }
2491 if (hlen > m->m_len) {
2492 if ((m = m_pullup(m, hlen)) == 0) {
2493 ip_statinc(IP_STAT_BADHLEN);
2494 goto bad;
2495 }
2496 ip = mtod(m, struct ip *);
2497 if (ip == NULL) goto bad;
2498 }
2499
2500 switch (m->m_pkthdr.csum_flags &
2501 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) |
2502 M_CSUM_IPv4_BAD)) {
2503 case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
2504 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); */
2505 goto bad;
2506
2507 case M_CSUM_IPv4:
2508 /* Checksum was okay. */
2509 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); */
2510 break;
2511
2512 default:
2513 /* Must compute it ourselves. */
2514 /* INET_CSUM_COUNTER_INCR(&ip_swcsum); */
2515 if (in_cksum(m, hlen) != 0)
2516 goto bad;
2517 break;
2518 }
2519
2520 /* Retrieve the packet length. */
2521 len = ntohs(ip->ip_len);
2522
2523 /*
2524 * Check for additional length bogosity
2525 */
2526 if (len < hlen) {
2527 ip_statinc(IP_STAT_BADLEN);
2528 goto bad;
2529 }
2530
2531 /*
2532 * Check that the amount of data in the buffers
2533 * is as at least much as the IP header would have us expect.
2534 * Drop packet if shorter than we expect.
2535 */
2536 if (m->m_pkthdr.len < len) {
2537 ip_statinc(IP_STAT_TOOSHORT);
2538 goto bad;
2539 }
2540
2541 /* Checks out, proceed */
2542 *mp = m;
2543 return 0;
2544
2545 bad:
2546 *mp = m;
2547 return -1;
2548 }
2549
2550 # ifdef INET6
2551 /*
2552 * Same as above, but for IPv6.
2553 * Cut-and-pasted from ip6_input.c.
2554 * XXX Should we update ip6stat, or not?
2555 */
2556 static int
2557 bridge_ip6_checkbasic(struct mbuf **mp)
2558 {
2559 struct mbuf *m = *mp;
2560 struct ip6_hdr *ip6;
2561
2562 /*
2563 * If the IPv6 header is not aligned, slurp it up into a new
2564 * mbuf with space for link headers, in the event we forward
2565 * it. Otherwise, if it is aligned, make sure the entire base
2566 * IPv6 header is in the first mbuf of the chain.
2567 */
2568 if (IP6_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
2569 struct ifnet *inifp = m->m_pkthdr.rcvif;
2570 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
2571 (max_linkhdr + 3) & ~3)) == NULL) {
2572 /* XXXJRT new stat, please */
2573 ip6_statinc(IP6_STAT_TOOSMALL);
2574 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
2575 goto bad;
2576 }
2577 } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
2578 struct ifnet *inifp = m->m_pkthdr.rcvif;
2579 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
2580 ip6_statinc(IP6_STAT_TOOSMALL);
2581 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
2582 goto bad;
2583 }
2584 }
2585
2586 ip6 = mtod(m, struct ip6_hdr *);
2587
2588 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
2589 ip6_statinc(IP6_STAT_BADVERS);
2590 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
2591 goto bad;
2592 }
2593
2594 /* Checks out, proceed */
2595 *mp = m;
2596 return 0;
2597
2598 bad:
2599 *mp = m;
2600 return -1;
2601 }
2602 # endif /* INET6 */
2603 #endif /* BRIDGE_IPF */
2604