if_bridge.c revision 1.164.4.3 1 /* $NetBSD: if_bridge.c,v 1.164.4.3 2025/05/15 18:01:48 martin 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 1.164.4.3 2025/05/15 18:01:48 martin Exp $");
84
85 #ifdef _KERNEL_OPT
86 #include "opt_bridge_ipf.h"
87 #include "opt_inet.h"
88 #include "opt_net_mpsafe.h"
89 #endif /* _KERNEL_OPT */
90
91 #include <sys/param.h>
92 #include <sys/kernel.h>
93 #include <sys/mbuf.h>
94 #include <sys/queue.h>
95 #include <sys/socket.h>
96 #include <sys/socketvar.h> /* for softnet_lock */
97 #include <sys/sockio.h>
98 #include <sys/systm.h>
99 #include <sys/proc.h>
100 #include <sys/pool.h>
101 #include <sys/kauth.h>
102 #include <sys/cpu.h>
103 #include <sys/cprng.h>
104 #include <sys/mutex.h>
105 #include <sys/kmem.h>
106
107 #include <net/bpf.h>
108 #include <net/if.h>
109 #include <net/if_dl.h>
110 #include <net/if_types.h>
111 #include <net/if_llc.h>
112
113 #include <net/if_ether.h>
114 #include <net/if_bridgevar.h>
115 #include <net/ether_sw_offload.h>
116
117 #if defined(BRIDGE_IPF)
118 /* Used for bridge_ip[6]_checkbasic */
119 #include <netinet/in.h>
120 #include <netinet/in_systm.h>
121 #include <netinet/ip.h>
122 #include <netinet/ip_var.h>
123 #include <netinet/ip_private.h> /* XXX */
124
125 #include <netinet/ip6.h>
126 #include <netinet6/in6_var.h>
127 #include <netinet6/ip6_var.h>
128 #include <netinet6/ip6_private.h> /* XXX */
129 #endif /* BRIDGE_IPF */
130
131 /*
132 * Size of the route hash table. Must be a power of two.
133 */
134 #ifndef BRIDGE_RTHASH_SIZE
135 #define BRIDGE_RTHASH_SIZE 1024
136 #endif
137
138 #define BRIDGE_RTHASH_MASK (BRIDGE_RTHASH_SIZE - 1)
139
140 #include "carp.h"
141 #if NCARP > 0
142 #include <netinet/in.h>
143 #include <netinet/in_var.h>
144 #include <netinet/ip_carp.h>
145 #endif
146
147 #include "ioconf.h"
148
149 __CTASSERT(sizeof(struct ifbifconf) == sizeof(struct ifbaconf));
150 __CTASSERT(offsetof(struct ifbifconf, ifbic_len) == offsetof(struct ifbaconf, ifbac_len));
151 __CTASSERT(offsetof(struct ifbifconf, ifbic_buf) == offsetof(struct ifbaconf, ifbac_buf));
152
153 /*
154 * Maximum number of addresses to cache.
155 */
156 #ifndef BRIDGE_RTABLE_MAX
157 #define BRIDGE_RTABLE_MAX 100
158 #endif
159
160 /*
161 * Spanning tree defaults.
162 */
163 #define BSTP_DEFAULT_MAX_AGE (20 * 256)
164 #define BSTP_DEFAULT_HELLO_TIME (2 * 256)
165 #define BSTP_DEFAULT_FORWARD_DELAY (15 * 256)
166 #define BSTP_DEFAULT_HOLD_TIME (1 * 256)
167 #define BSTP_DEFAULT_BRIDGE_PRIORITY 0x8000
168 #define BSTP_DEFAULT_PORT_PRIORITY 0x80
169 #define BSTP_DEFAULT_PATH_COST 55
170
171 /*
172 * Timeout (in seconds) for entries learned dynamically.
173 */
174 #ifndef BRIDGE_RTABLE_TIMEOUT
175 #define BRIDGE_RTABLE_TIMEOUT (20 * 60) /* same as ARP */
176 #endif
177
178 /*
179 * Number of seconds between walks of the route list.
180 */
181 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
182 #define BRIDGE_RTABLE_PRUNE_PERIOD (5 * 60)
183 #endif
184
185 #define BRIDGE_RT_LOCK(_sc) mutex_enter((_sc)->sc_rtlist_lock)
186 #define BRIDGE_RT_UNLOCK(_sc) mutex_exit((_sc)->sc_rtlist_lock)
187 #define BRIDGE_RT_LOCKED(_sc) mutex_owned((_sc)->sc_rtlist_lock)
188
189 #define BRIDGE_RT_PSZ_PERFORM(_sc) \
190 pserialize_perform((_sc)->sc_rtlist_psz)
191
192 #define BRIDGE_RT_RENTER(__s) do { __s = pserialize_read_enter(); } while (0)
193 #define BRIDGE_RT_REXIT(__s) do { pserialize_read_exit(__s); } while (0)
194
195 #define BRIDGE_RTLIST_READER_FOREACH(_brt, _sc) \
196 PSLIST_READER_FOREACH((_brt), &((_sc)->sc_rtlist), \
197 struct bridge_rtnode, brt_list)
198 #define BRIDGE_RTLIST_WRITER_FOREACH(_brt, _sc) \
199 PSLIST_WRITER_FOREACH((_brt), &((_sc)->sc_rtlist), \
200 struct bridge_rtnode, brt_list)
201 #define BRIDGE_RTLIST_WRITER_INSERT_HEAD(_sc, _brt) \
202 PSLIST_WRITER_INSERT_HEAD(&(_sc)->sc_rtlist, brt, brt_list)
203 #define BRIDGE_RTLIST_WRITER_REMOVE(_brt) \
204 PSLIST_WRITER_REMOVE((_brt), brt_list)
205
206 #define BRIDGE_RTHASH_READER_FOREACH(_brt, _sc, _hash) \
207 PSLIST_READER_FOREACH((_brt), &(_sc)->sc_rthash[(_hash)], \
208 struct bridge_rtnode, brt_hash)
209 #define BRIDGE_RTHASH_WRITER_FOREACH(_brt, _sc, _hash) \
210 PSLIST_WRITER_FOREACH((_brt), &(_sc)->sc_rthash[(_hash)], \
211 struct bridge_rtnode, brt_hash)
212 #define BRIDGE_RTHASH_WRITER_INSERT_HEAD(_sc, _hash, _brt) \
213 PSLIST_WRITER_INSERT_HEAD(&(_sc)->sc_rthash[(_hash)], brt, brt_hash)
214 #define BRIDGE_RTHASH_WRITER_INSERT_AFTER(_brt, _new) \
215 PSLIST_WRITER_INSERT_AFTER((_brt), (_new), brt_hash)
216 #define BRIDGE_RTHASH_WRITER_REMOVE(_brt) \
217 PSLIST_WRITER_REMOVE((_brt), brt_hash)
218
219 #ifdef NET_MPSAFE
220 #define DECLARE_LOCK_VARIABLE
221 #define ACQUIRE_GLOBAL_LOCKS() do { } while (0)
222 #define RELEASE_GLOBAL_LOCKS() do { } while (0)
223 #else
224 #define DECLARE_LOCK_VARIABLE int __s
225 #define ACQUIRE_GLOBAL_LOCKS() do { \
226 KERNEL_LOCK(1, NULL); \
227 mutex_enter(softnet_lock); \
228 __s = splsoftnet(); \
229 } while (0)
230 #define RELEASE_GLOBAL_LOCKS() do { \
231 splx(__s); \
232 mutex_exit(softnet_lock); \
233 KERNEL_UNLOCK_ONE(NULL); \
234 } while (0)
235 #endif
236
237 struct psref_class *bridge_psref_class __read_mostly;
238
239 int bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
240
241 static struct pool bridge_rtnode_pool;
242
243 static int bridge_clone_create(struct if_clone *, int);
244 static int bridge_clone_destroy(struct ifnet *);
245
246 static int bridge_ioctl(struct ifnet *, u_long, void *);
247 static int bridge_init(struct ifnet *);
248 static void bridge_stop(struct ifnet *, int);
249 static void bridge_start(struct ifnet *);
250
251 static void bridge_input(struct ifnet *, struct mbuf *);
252 static void bridge_forward(struct bridge_softc *, struct mbuf *);
253
254 static void bridge_timer(void *);
255
256 static void bridge_broadcast(struct bridge_softc *, struct ifnet *,
257 struct mbuf *);
258
259 static int bridge_rtupdate(struct bridge_softc *, const uint8_t *,
260 struct ifnet *, int, uint8_t);
261 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *);
262 static void bridge_rttrim(struct bridge_softc *);
263 static void bridge_rtage(struct bridge_softc *);
264 static void bridge_rtage_work(struct work *, void *);
265 static void bridge_rtflush(struct bridge_softc *, int);
266 static int bridge_rtdaddr(struct bridge_softc *, const uint8_t *);
267 static void bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp);
268
269 static void bridge_rtable_init(struct bridge_softc *);
270 static void bridge_rtable_fini(struct bridge_softc *);
271
272 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
273 const uint8_t *);
274 static int bridge_rtnode_insert(struct bridge_softc *,
275 struct bridge_rtnode *);
276 static void bridge_rtnode_remove(struct bridge_softc *,
277 struct bridge_rtnode *);
278 static void bridge_rtnode_destroy(struct bridge_rtnode *);
279
280 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
281 const char *name,
282 struct psref *);
283 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
284 struct ifnet *ifp,
285 struct psref *);
286 static void bridge_release_member(struct bridge_softc *, struct bridge_iflist *,
287 struct psref *);
288 static void bridge_delete_member(struct bridge_softc *,
289 struct bridge_iflist *);
290 static void bridge_acquire_member(struct bridge_softc *sc,
291 struct bridge_iflist *,
292 struct psref *);
293
294 static int bridge_ioctl_add(struct bridge_softc *, void *);
295 static int bridge_ioctl_del(struct bridge_softc *, void *);
296 static int bridge_ioctl_gifflags(struct bridge_softc *, void *);
297 static int bridge_ioctl_sifflags(struct bridge_softc *, void *);
298 static int bridge_ioctl_scache(struct bridge_softc *, void *);
299 static int bridge_ioctl_gcache(struct bridge_softc *, void *);
300 static int bridge_ioctl_gifs(struct bridge_softc *, void *);
301 static int bridge_ioctl_rts(struct bridge_softc *, void *);
302 static int bridge_ioctl_saddr(struct bridge_softc *, void *);
303 static int bridge_ioctl_sto(struct bridge_softc *, void *);
304 static int bridge_ioctl_gto(struct bridge_softc *, void *);
305 static int bridge_ioctl_daddr(struct bridge_softc *, void *);
306 static int bridge_ioctl_flush(struct bridge_softc *, void *);
307 static int bridge_ioctl_gpri(struct bridge_softc *, void *);
308 static int bridge_ioctl_spri(struct bridge_softc *, void *);
309 static int bridge_ioctl_ght(struct bridge_softc *, void *);
310 static int bridge_ioctl_sht(struct bridge_softc *, void *);
311 static int bridge_ioctl_gfd(struct bridge_softc *, void *);
312 static int bridge_ioctl_sfd(struct bridge_softc *, void *);
313 static int bridge_ioctl_gma(struct bridge_softc *, void *);
314 static int bridge_ioctl_sma(struct bridge_softc *, void *);
315 static int bridge_ioctl_sifprio(struct bridge_softc *, void *);
316 static int bridge_ioctl_sifcost(struct bridge_softc *, void *);
317 #if defined(BRIDGE_IPF)
318 static int bridge_ioctl_gfilt(struct bridge_softc *, void *);
319 static int bridge_ioctl_sfilt(struct bridge_softc *, void *);
320 static int bridge_ipf(void *, struct mbuf **, struct ifnet *, int);
321 static int bridge_ip_checkbasic(struct mbuf **mp);
322 # ifdef INET6
323 static int bridge_ip6_checkbasic(struct mbuf **mp);
324 # endif /* INET6 */
325 #endif /* BRIDGE_IPF */
326
327 struct bridge_control {
328 int (*bc_func)(struct bridge_softc *, void *);
329 int bc_argsize;
330 int bc_flags;
331 };
332
333 #define BC_F_COPYIN 0x01 /* copy arguments in */
334 #define BC_F_COPYOUT 0x02 /* copy arguments out */
335 #define BC_F_SUSER 0x04 /* do super-user check */
336 #define BC_F_XLATEIN 0x08 /* xlate arguments in */
337 #define BC_F_XLATEOUT 0x10 /* xlate arguments out */
338
339 static const struct bridge_control bridge_control_table[] = {
340 [BRDGADD] = {bridge_ioctl_add, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
341 [BRDGDEL] = {bridge_ioctl_del, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
342
343 [BRDGGIFFLGS] = {bridge_ioctl_gifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_COPYOUT},
344 [BRDGSIFFLGS] = {bridge_ioctl_sifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
345
346 [BRDGSCACHE] = {bridge_ioctl_scache, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
347 [BRDGGCACHE] = {bridge_ioctl_gcache, sizeof(struct ifbrparam), BC_F_COPYOUT},
348
349 [OBRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_COPYIN|BC_F_COPYOUT},
350 [OBRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_COPYIN|BC_F_COPYOUT},
351
352 [BRDGSADDR] = {bridge_ioctl_saddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
353
354 [BRDGSTO] = {bridge_ioctl_sto, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
355 [BRDGGTO] = {bridge_ioctl_gto, sizeof(struct ifbrparam), BC_F_COPYOUT},
356
357 [BRDGDADDR] = {bridge_ioctl_daddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
358
359 [BRDGFLUSH] = {bridge_ioctl_flush, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
360
361 [BRDGGPRI] = {bridge_ioctl_gpri, sizeof(struct ifbrparam), BC_F_COPYOUT},
362 [BRDGSPRI] = {bridge_ioctl_spri, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
363
364 [BRDGGHT] = {bridge_ioctl_ght, sizeof(struct ifbrparam), BC_F_COPYOUT},
365 [BRDGSHT] = {bridge_ioctl_sht, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
366
367 [BRDGGFD] = {bridge_ioctl_gfd, sizeof(struct ifbrparam), BC_F_COPYOUT},
368 [BRDGSFD] = {bridge_ioctl_sfd, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
369
370 [BRDGGMA] = {bridge_ioctl_gma, sizeof(struct ifbrparam), BC_F_COPYOUT},
371 [BRDGSMA] = {bridge_ioctl_sma, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
372
373 [BRDGSIFPRIO] = {bridge_ioctl_sifprio, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
374
375 [BRDGSIFCOST] = {bridge_ioctl_sifcost, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
376 #if defined(BRIDGE_IPF)
377 [BRDGGFILT] = {bridge_ioctl_gfilt, sizeof(struct ifbrparam), BC_F_COPYOUT},
378 [BRDGSFILT] = {bridge_ioctl_sfilt, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
379 #endif /* BRIDGE_IPF */
380 [BRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_XLATEIN|BC_F_XLATEOUT},
381 [BRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_XLATEIN|BC_F_XLATEOUT},
382 };
383
384 static const int bridge_control_table_size = __arraycount(bridge_control_table);
385
386 static struct if_clone bridge_cloner =
387 IF_CLONE_INITIALIZER("bridge", bridge_clone_create, bridge_clone_destroy);
388
389 /*
390 * bridgeattach:
391 *
392 * Pseudo-device attach routine.
393 */
394 void
395 bridgeattach(int n)
396 {
397
398 pool_init(&bridge_rtnode_pool, sizeof(struct bridge_rtnode),
399 0, 0, 0, "brtpl", NULL, IPL_NET);
400
401 bridge_psref_class = psref_class_create("bridge", IPL_SOFTNET);
402
403 if_clone_attach(&bridge_cloner);
404 }
405
406 /*
407 * bridge_clone_create:
408 *
409 * Create a new bridge instance.
410 */
411 static int
412 bridge_clone_create(struct if_clone *ifc, int unit)
413 {
414 struct bridge_softc *sc;
415 struct ifnet *ifp;
416 int error;
417
418 sc = kmem_zalloc(sizeof(*sc), KM_SLEEP);
419 ifp = &sc->sc_if;
420
421 sc->sc_brtmax = BRIDGE_RTABLE_MAX;
422 sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
423 sc->sc_bridge_max_age = BSTP_DEFAULT_MAX_AGE;
424 sc->sc_bridge_hello_time = BSTP_DEFAULT_HELLO_TIME;
425 sc->sc_bridge_forward_delay = BSTP_DEFAULT_FORWARD_DELAY;
426 sc->sc_bridge_priority = BSTP_DEFAULT_BRIDGE_PRIORITY;
427 sc->sc_hold_time = BSTP_DEFAULT_HOLD_TIME;
428 sc->sc_filter_flags = 0;
429
430 /* Initialize our routing table. */
431 bridge_rtable_init(sc);
432
433 error = workqueue_create(&sc->sc_rtage_wq, "bridge_rtage",
434 bridge_rtage_work, sc, PRI_SOFTNET, IPL_SOFTNET, WQ_MPSAFE);
435 if (error)
436 panic("%s: workqueue_create %d\n", __func__, error);
437
438 callout_init(&sc->sc_brcallout, CALLOUT_MPSAFE);
439 callout_init(&sc->sc_bstpcallout, CALLOUT_MPSAFE);
440
441 mutex_init(&sc->sc_iflist_psref.bip_lock, MUTEX_DEFAULT, IPL_NONE);
442 PSLIST_INIT(&sc->sc_iflist_psref.bip_iflist);
443 sc->sc_iflist_psref.bip_psz = pserialize_create();
444
445 if_initname(ifp, ifc->ifc_name, unit);
446 ifp->if_softc = sc;
447 ifp->if_extflags = IFEF_NO_LINK_STATE_CHANGE;
448 #ifdef NET_MPSAFE
449 ifp->if_extflags |= IFEF_MPSAFE;
450 #endif
451 ifp->if_mtu = ETHERMTU;
452 ifp->if_ioctl = bridge_ioctl;
453 ifp->if_output = bridge_output;
454 ifp->if_start = bridge_start;
455 ifp->if_stop = bridge_stop;
456 ifp->if_init = bridge_init;
457 ifp->if_type = IFT_BRIDGE;
458 ifp->if_addrlen = 0;
459 ifp->if_dlt = DLT_EN10MB;
460 ifp->if_hdrlen = ETHER_HDR_LEN;
461
462 error = if_initialize(ifp);
463 if (error != 0) {
464 pserialize_destroy(sc->sc_iflist_psref.bip_psz);
465 mutex_destroy(&sc->sc_iflist_psref.bip_lock);
466 callout_destroy(&sc->sc_brcallout);
467 callout_destroy(&sc->sc_bstpcallout);
468 workqueue_destroy(sc->sc_rtage_wq);
469 bridge_rtable_fini(sc);
470 kmem_free(sc, sizeof(*sc));
471
472 return error;
473 }
474 if_alloc_sadl(ifp);
475 if_register(ifp);
476
477 return 0;
478 }
479
480 /*
481 * bridge_clone_destroy:
482 *
483 * Destroy a bridge instance.
484 */
485 static int
486 bridge_clone_destroy(struct ifnet *ifp)
487 {
488 struct bridge_softc *sc = ifp->if_softc;
489 struct bridge_iflist *bif;
490
491 if ((ifp->if_flags & IFF_RUNNING) != 0)
492 bridge_stop(ifp, 1);
493
494 BRIDGE_LOCK(sc);
495 for (;;) {
496 bif = PSLIST_WRITER_FIRST(&sc->sc_iflist_psref.bip_iflist, struct bridge_iflist,
497 bif_next);
498 if (bif == NULL)
499 break;
500 bridge_delete_member(sc, bif);
501 }
502 PSLIST_DESTROY(&sc->sc_iflist_psref.bip_iflist);
503 BRIDGE_UNLOCK(sc);
504
505 if_detach(ifp);
506
507 /* Tear down the routing table. */
508 bridge_rtable_fini(sc);
509
510 pserialize_destroy(sc->sc_iflist_psref.bip_psz);
511 mutex_destroy(&sc->sc_iflist_psref.bip_lock);
512 callout_destroy(&sc->sc_brcallout);
513 callout_destroy(&sc->sc_bstpcallout);
514 workqueue_destroy(sc->sc_rtage_wq);
515 kmem_free(sc, sizeof(*sc));
516
517 return 0;
518 }
519
520 /*
521 * bridge_ioctl:
522 *
523 * Handle a control request from the operator.
524 */
525 static int
526 bridge_ioctl(struct ifnet *ifp, u_long cmd, void *data)
527 {
528 struct bridge_softc *sc = ifp->if_softc;
529 struct lwp *l = curlwp; /* XXX */
530 union {
531 struct ifbreq ifbreq;
532 struct ifbifconf ifbifconf;
533 struct ifbareq ifbareq;
534 struct ifbaconf ifbaconf;
535 struct ifbrparam ifbrparam;
536 } args;
537 struct ifdrv *ifd = (struct ifdrv *) data;
538 const struct bridge_control *bc = NULL; /* XXXGCC */
539 int s, error = 0;
540
541 /* Authorize command before calling splsoftnet(). */
542 switch (cmd) {
543 case SIOCGDRVSPEC:
544 case SIOCSDRVSPEC:
545 if (ifd->ifd_cmd >= bridge_control_table_size
546 || (bc = &bridge_control_table[ifd->ifd_cmd]) == NULL) {
547 error = EINVAL;
548 return error;
549 }
550
551 /* We only care about BC_F_SUSER at this point. */
552 if ((bc->bc_flags & BC_F_SUSER) == 0)
553 break;
554
555 error = kauth_authorize_network(l->l_cred,
556 KAUTH_NETWORK_INTERFACE_BRIDGE,
557 cmd == SIOCGDRVSPEC ?
558 KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_GETPRIV :
559 KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_SETPRIV,
560 ifd, NULL, NULL);
561 if (error)
562 return error;
563
564 break;
565 }
566
567 s = splsoftnet();
568
569 switch (cmd) {
570 case SIOCGDRVSPEC:
571 case SIOCSDRVSPEC:
572 KASSERT(bc != NULL);
573 if (cmd == SIOCGDRVSPEC &&
574 (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) == 0) {
575 error = EINVAL;
576 break;
577 }
578 else if (cmd == SIOCSDRVSPEC &&
579 (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) != 0) {
580 error = EINVAL;
581 break;
582 }
583
584 /* BC_F_SUSER is checked above, before splsoftnet(). */
585
586 if ((bc->bc_flags & (BC_F_XLATEIN|BC_F_XLATEOUT)) == 0
587 && (ifd->ifd_len != bc->bc_argsize
588 || ifd->ifd_len > sizeof(args))) {
589 error = EINVAL;
590 break;
591 }
592
593 memset(&args, 0, sizeof(args));
594 if (bc->bc_flags & BC_F_COPYIN) {
595 error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
596 if (error)
597 break;
598 } else if (bc->bc_flags & BC_F_XLATEIN) {
599 args.ifbifconf.ifbic_len = ifd->ifd_len;
600 args.ifbifconf.ifbic_buf = ifd->ifd_data;
601 }
602
603 error = (*bc->bc_func)(sc, &args);
604 if (error)
605 break;
606
607 if (bc->bc_flags & BC_F_COPYOUT) {
608 error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
609 } else if (bc->bc_flags & BC_F_XLATEOUT) {
610 ifd->ifd_len = args.ifbifconf.ifbic_len;
611 ifd->ifd_data = args.ifbifconf.ifbic_buf;
612 }
613 break;
614
615 case SIOCSIFFLAGS:
616 if ((error = ifioctl_common(ifp, cmd, data)) != 0)
617 break;
618 switch (ifp->if_flags & (IFF_UP|IFF_RUNNING)) {
619 case IFF_RUNNING:
620 /*
621 * If interface is marked down and it is running,
622 * then stop and disable it.
623 */
624 (*ifp->if_stop)(ifp, 1);
625 break;
626 case IFF_UP:
627 /*
628 * If interface is marked up and it is stopped, then
629 * start it.
630 */
631 error = (*ifp->if_init)(ifp);
632 break;
633 default:
634 break;
635 }
636 break;
637
638 case SIOCSIFMTU:
639 if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET)
640 error = 0;
641 break;
642
643 default:
644 error = ifioctl_common(ifp, cmd, data);
645 break;
646 }
647
648 splx(s);
649
650 return error;
651 }
652
653 /*
654 * bridge_lookup_member:
655 *
656 * Lookup a bridge member interface.
657 */
658 static struct bridge_iflist *
659 bridge_lookup_member(struct bridge_softc *sc, const char *name, struct psref *psref)
660 {
661 struct bridge_iflist *bif;
662 struct ifnet *ifp;
663 int s;
664
665 BRIDGE_PSZ_RENTER(s);
666
667 BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
668 ifp = bif->bif_ifp;
669 if (strcmp(ifp->if_xname, name) == 0)
670 break;
671 }
672 if (bif != NULL)
673 bridge_acquire_member(sc, bif, psref);
674
675 BRIDGE_PSZ_REXIT(s);
676
677 return bif;
678 }
679
680 /*
681 * bridge_lookup_member_if:
682 *
683 * Lookup a bridge member interface by ifnet*.
684 */
685 static struct bridge_iflist *
686 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp,
687 struct psref *psref)
688 {
689 struct bridge_iflist *bif;
690 int s;
691
692 BRIDGE_PSZ_RENTER(s);
693
694 bif = member_ifp->if_bridgeif;
695 if (bif != NULL) {
696 psref_acquire(psref, &bif->bif_psref,
697 bridge_psref_class);
698 }
699
700 BRIDGE_PSZ_REXIT(s);
701
702 return bif;
703 }
704
705 static void
706 bridge_acquire_member(struct bridge_softc *sc, struct bridge_iflist *bif,
707 struct psref *psref)
708 {
709
710 psref_acquire(psref, &bif->bif_psref, bridge_psref_class);
711 }
712
713 /*
714 * bridge_release_member:
715 *
716 * Release the specified member interface.
717 */
718 static void
719 bridge_release_member(struct bridge_softc *sc, struct bridge_iflist *bif,
720 struct psref *psref)
721 {
722
723 psref_release(psref, &bif->bif_psref, bridge_psref_class);
724 }
725
726 /*
727 * bridge_delete_member:
728 *
729 * Delete the specified member interface.
730 */
731 static void
732 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif)
733 {
734 struct ifnet *ifs = bif->bif_ifp;
735
736 KASSERT(BRIDGE_LOCKED(sc));
737
738 ifs->_if_input = ether_input;
739 ifs->if_bridge = NULL;
740 ifs->if_bridgeif = NULL;
741
742 PSLIST_WRITER_REMOVE(bif, bif_next);
743 BRIDGE_PSZ_PERFORM(sc);
744 BRIDGE_UNLOCK(sc);
745
746 switch (ifs->if_type) {
747 case IFT_ETHER:
748 case IFT_L2TP:
749 /*
750 * Take the interface out of promiscuous mode.
751 * Don't call it with holding a spin lock.
752 */
753 (void) ifpromisc(ifs, 0);
754 IFNET_LOCK(ifs);
755 (void) ether_disable_vlan_mtu(ifs);
756 IFNET_UNLOCK(ifs);
757 break;
758 default:
759 #ifdef DIAGNOSTIC
760 panic("%s: impossible", __func__);
761 #endif
762 break;
763 }
764
765 psref_target_destroy(&bif->bif_psref, bridge_psref_class);
766
767 PSLIST_ENTRY_DESTROY(bif, bif_next);
768 kmem_free(bif, sizeof(*bif));
769
770 BRIDGE_LOCK(sc);
771 }
772
773 /*
774 * bridge_calc_csum_flags:
775 *
776 * Calculate logical and b/w csum flags each member interface supports.
777 */
778 void
779 bridge_calc_csum_flags(struct bridge_softc *sc)
780 {
781 struct bridge_iflist *bif;
782 struct ifnet *ifs;
783 int flags = ~0;
784
785 BRIDGE_LOCK(sc);
786 BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
787 ifs = bif->bif_ifp;
788 flags &= ifs->if_csum_flags_tx;
789 }
790 sc->sc_csum_flags_tx = flags;
791 BRIDGE_UNLOCK(sc);
792 }
793
794 static int
795 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
796 {
797 struct ifbreq *req = arg;
798 struct bridge_iflist *bif = NULL;
799 struct ifnet *ifs;
800 int error = 0;
801 struct psref psref;
802
803 ifs = if_get(req->ifbr_ifsname, &psref);
804 if (ifs == NULL)
805 return ENOENT;
806
807 if (ifs->if_bridge == sc) {
808 error = EEXIST;
809 goto out;
810 }
811
812 if (ifs->if_bridge != NULL) {
813 error = EBUSY;
814 goto out;
815 }
816
817 if (ifs->_if_input != ether_input) {
818 error = EINVAL;
819 goto out;
820 }
821
822 /* FIXME: doesn't work with non-IFF_SIMPLEX interfaces */
823 if ((ifs->if_flags & IFF_SIMPLEX) == 0) {
824 error = EINVAL;
825 goto out;
826 }
827
828 bif = kmem_alloc(sizeof(*bif), KM_SLEEP);
829
830 switch (ifs->if_type) {
831 case IFT_ETHER:
832 if (sc->sc_if.if_mtu != ifs->if_mtu) {
833 error = EINVAL;
834 goto out;
835 }
836 /* FALLTHROUGH */
837 case IFT_L2TP:
838 IFNET_LOCK(ifs);
839 error = ether_enable_vlan_mtu(ifs);
840 IFNET_UNLOCK(ifs);
841 if (error > 0)
842 goto out;
843 /*
844 * Place the interface into promiscuous mode.
845 */
846 error = ifpromisc(ifs, 1);
847 if (error)
848 goto out;
849 break;
850 default:
851 error = EINVAL;
852 goto out;
853 }
854
855 bif->bif_ifp = ifs;
856 bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
857 bif->bif_priority = BSTP_DEFAULT_PORT_PRIORITY;
858 bif->bif_path_cost = BSTP_DEFAULT_PATH_COST;
859 PSLIST_ENTRY_INIT(bif, bif_next);
860 psref_target_init(&bif->bif_psref, bridge_psref_class);
861
862 BRIDGE_LOCK(sc);
863
864 ifs->if_bridge = sc;
865 ifs->if_bridgeif = bif;
866 PSLIST_WRITER_INSERT_HEAD(&sc->sc_iflist_psref.bip_iflist, bif, bif_next);
867 ifs->_if_input = bridge_input;
868
869 BRIDGE_UNLOCK(sc);
870
871 bridge_calc_csum_flags(sc);
872
873 if (sc->sc_if.if_flags & IFF_RUNNING)
874 bstp_initialization(sc);
875 else
876 bstp_stop(sc);
877
878 out:
879 if_put(ifs, &psref);
880 if (error) {
881 if (bif != NULL)
882 kmem_free(bif, sizeof(*bif));
883 }
884 return error;
885 }
886
887 static int
888 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
889 {
890 struct ifbreq *req = arg;
891 const char *name = req->ifbr_ifsname;
892 struct bridge_iflist *bif;
893 struct ifnet *ifs;
894
895 BRIDGE_LOCK(sc);
896
897 /*
898 * Don't use bridge_lookup_member. We want to get a member
899 * with bif_refs == 0.
900 */
901 BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) {
902 ifs = bif->bif_ifp;
903 if (strcmp(ifs->if_xname, name) == 0)
904 break;
905 }
906
907 if (bif == NULL) {
908 BRIDGE_UNLOCK(sc);
909 return ENOENT;
910 }
911
912 bridge_delete_member(sc, bif);
913
914 BRIDGE_UNLOCK(sc);
915
916 bridge_rtdelete(sc, ifs);
917 bridge_calc_csum_flags(sc);
918
919 if (sc->sc_if.if_flags & IFF_RUNNING)
920 bstp_initialization(sc);
921
922 return 0;
923 }
924
925 static int
926 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
927 {
928 struct ifbreq *req = arg;
929 struct bridge_iflist *bif;
930 struct psref psref;
931
932 bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
933 if (bif == NULL)
934 return ENOENT;
935
936 req->ifbr_ifsflags = bif->bif_flags;
937 req->ifbr_state = bif->bif_state;
938 req->ifbr_priority = bif->bif_priority;
939 req->ifbr_path_cost = bif->bif_path_cost;
940 req->ifbr_portno = bif->bif_ifp->if_index & 0xff;
941
942 bridge_release_member(sc, bif, &psref);
943
944 return 0;
945 }
946
947 static int
948 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
949 {
950 struct ifbreq *req = arg;
951 struct bridge_iflist *bif;
952 struct psref psref;
953
954 bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
955 if (bif == NULL)
956 return ENOENT;
957
958 if (req->ifbr_ifsflags & IFBIF_STP) {
959 switch (bif->bif_ifp->if_type) {
960 case IFT_ETHER:
961 case IFT_L2TP:
962 /* These can do spanning tree. */
963 break;
964
965 default:
966 /* Nothing else can. */
967 bridge_release_member(sc, bif, &psref);
968 return EINVAL;
969 }
970 }
971
972 bif->bif_flags = req->ifbr_ifsflags;
973
974 bridge_release_member(sc, bif, &psref);
975
976 if (sc->sc_if.if_flags & IFF_RUNNING)
977 bstp_initialization(sc);
978
979 return 0;
980 }
981
982 static int
983 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
984 {
985 struct ifbrparam *param = arg;
986
987 sc->sc_brtmax = param->ifbrp_csize;
988 bridge_rttrim(sc);
989
990 return 0;
991 }
992
993 static int
994 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
995 {
996 struct ifbrparam *param = arg;
997
998 param->ifbrp_csize = sc->sc_brtmax;
999
1000 return 0;
1001 }
1002
1003 static int
1004 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1005 {
1006 struct ifbifconf *bifc = arg;
1007 struct bridge_iflist *bif;
1008 struct ifbreq *breqs;
1009 int i, count, error = 0;
1010
1011 retry:
1012 BRIDGE_LOCK(sc);
1013 count = 0;
1014 BRIDGE_IFLIST_WRITER_FOREACH(bif, sc)
1015 count++;
1016 BRIDGE_UNLOCK(sc);
1017
1018 if (count == 0) {
1019 bifc->ifbic_len = 0;
1020 return 0;
1021 }
1022
1023 if (bifc->ifbic_len == 0 || bifc->ifbic_len < (sizeof(*breqs) * count)) {
1024 /* Tell that a larger buffer is needed */
1025 bifc->ifbic_len = sizeof(*breqs) * count;
1026 return 0;
1027 }
1028
1029 breqs = kmem_alloc(sizeof(*breqs) * count, KM_SLEEP);
1030
1031 BRIDGE_LOCK(sc);
1032
1033 i = 0;
1034 BRIDGE_IFLIST_WRITER_FOREACH(bif, sc)
1035 i++;
1036 if (i > count) {
1037 /*
1038 * The number of members has been increased.
1039 * We need more memory!
1040 */
1041 BRIDGE_UNLOCK(sc);
1042 kmem_free(breqs, sizeof(*breqs) * count);
1043 goto retry;
1044 }
1045
1046 i = 0;
1047 BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) {
1048 struct ifbreq *breq = &breqs[i++];
1049 memset(breq, 0, sizeof(*breq));
1050
1051 strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname,
1052 sizeof(breq->ifbr_ifsname));
1053 breq->ifbr_ifsflags = bif->bif_flags;
1054 breq->ifbr_state = bif->bif_state;
1055 breq->ifbr_priority = bif->bif_priority;
1056 breq->ifbr_path_cost = bif->bif_path_cost;
1057 breq->ifbr_portno = bif->bif_ifp->if_index & 0xff;
1058 }
1059
1060 /* Don't call copyout with holding the mutex */
1061 BRIDGE_UNLOCK(sc);
1062
1063 for (i = 0; i < count; i++) {
1064 error = copyout(&breqs[i], bifc->ifbic_req + i, sizeof(*breqs));
1065 if (error)
1066 break;
1067 }
1068 bifc->ifbic_len = sizeof(*breqs) * i;
1069
1070 kmem_free(breqs, sizeof(*breqs) * count);
1071
1072 return error;
1073 }
1074
1075 static int
1076 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1077 {
1078 struct ifbaconf *bac = arg;
1079 struct bridge_rtnode *brt;
1080 struct ifbareq bareq;
1081 int count = 0, error = 0, len;
1082
1083 if (bac->ifbac_len == 0)
1084 return 0;
1085
1086 BRIDGE_RT_LOCK(sc);
1087
1088 /* The passed buffer is not enough, tell a required size. */
1089 if (bac->ifbac_len < (sizeof(bareq) * sc->sc_brtcnt)) {
1090 count = sc->sc_brtcnt;
1091 goto out;
1092 }
1093
1094 len = bac->ifbac_len;
1095 BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) {
1096 if (len < sizeof(bareq))
1097 goto out;
1098 memset(&bareq, 0, sizeof(bareq));
1099 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1100 sizeof(bareq.ifba_ifsname));
1101 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1102 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
1103 bareq.ifba_expire = brt->brt_expire - time_uptime;
1104 } else
1105 bareq.ifba_expire = 0;
1106 bareq.ifba_flags = brt->brt_flags;
1107
1108 error = copyout(&bareq, bac->ifbac_req + count, sizeof(bareq));
1109 if (error)
1110 goto out;
1111 count++;
1112 len -= sizeof(bareq);
1113 }
1114 out:
1115 BRIDGE_RT_UNLOCK(sc);
1116
1117 bac->ifbac_len = sizeof(bareq) * count;
1118 return error;
1119 }
1120
1121 static int
1122 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1123 {
1124 struct ifbareq *req = arg;
1125 struct bridge_iflist *bif;
1126 int error;
1127 struct psref psref;
1128
1129 bif = bridge_lookup_member(sc, req->ifba_ifsname, &psref);
1130 if (bif == NULL)
1131 return ENOENT;
1132
1133 error = bridge_rtupdate(sc, req->ifba_dst, bif->bif_ifp, 1,
1134 req->ifba_flags);
1135
1136 bridge_release_member(sc, bif, &psref);
1137
1138 return error;
1139 }
1140
1141 static int
1142 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1143 {
1144 struct ifbrparam *param = arg;
1145
1146 sc->sc_brttimeout = param->ifbrp_ctime;
1147
1148 return 0;
1149 }
1150
1151 static int
1152 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1153 {
1154 struct ifbrparam *param = arg;
1155
1156 param->ifbrp_ctime = sc->sc_brttimeout;
1157
1158 return 0;
1159 }
1160
1161 static int
1162 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1163 {
1164 struct ifbareq *req = arg;
1165
1166 return (bridge_rtdaddr(sc, req->ifba_dst));
1167 }
1168
1169 static int
1170 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1171 {
1172 struct ifbreq *req = arg;
1173
1174 bridge_rtflush(sc, req->ifbr_ifsflags);
1175
1176 return 0;
1177 }
1178
1179 static int
1180 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1181 {
1182 struct ifbrparam *param = arg;
1183
1184 param->ifbrp_prio = sc->sc_bridge_priority;
1185
1186 return 0;
1187 }
1188
1189 static int
1190 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1191 {
1192 struct ifbrparam *param = arg;
1193
1194 sc->sc_bridge_priority = param->ifbrp_prio;
1195
1196 if (sc->sc_if.if_flags & IFF_RUNNING)
1197 bstp_initialization(sc);
1198
1199 return 0;
1200 }
1201
1202 static int
1203 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1204 {
1205 struct ifbrparam *param = arg;
1206
1207 param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8;
1208
1209 return 0;
1210 }
1211
1212 static int
1213 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1214 {
1215 struct ifbrparam *param = arg;
1216
1217 if (param->ifbrp_hellotime == 0)
1218 return EINVAL;
1219 sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8;
1220
1221 if (sc->sc_if.if_flags & IFF_RUNNING)
1222 bstp_initialization(sc);
1223
1224 return 0;
1225 }
1226
1227 static int
1228 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1229 {
1230 struct ifbrparam *param = arg;
1231
1232 param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8;
1233
1234 return 0;
1235 }
1236
1237 static int
1238 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1239 {
1240 struct ifbrparam *param = arg;
1241
1242 if (param->ifbrp_fwddelay == 0)
1243 return EINVAL;
1244 sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8;
1245
1246 if (sc->sc_if.if_flags & IFF_RUNNING)
1247 bstp_initialization(sc);
1248
1249 return 0;
1250 }
1251
1252 static int
1253 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1254 {
1255 struct ifbrparam *param = arg;
1256
1257 param->ifbrp_maxage = sc->sc_bridge_max_age >> 8;
1258
1259 return 0;
1260 }
1261
1262 static int
1263 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1264 {
1265 struct ifbrparam *param = arg;
1266
1267 if (param->ifbrp_maxage == 0)
1268 return EINVAL;
1269 sc->sc_bridge_max_age = param->ifbrp_maxage << 8;
1270
1271 if (sc->sc_if.if_flags & IFF_RUNNING)
1272 bstp_initialization(sc);
1273
1274 return 0;
1275 }
1276
1277 static int
1278 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1279 {
1280 struct ifbreq *req = arg;
1281 struct bridge_iflist *bif;
1282 struct psref psref;
1283
1284 bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
1285 if (bif == NULL)
1286 return ENOENT;
1287
1288 bif->bif_priority = req->ifbr_priority;
1289
1290 if (sc->sc_if.if_flags & IFF_RUNNING)
1291 bstp_initialization(sc);
1292
1293 bridge_release_member(sc, bif, &psref);
1294
1295 return 0;
1296 }
1297
1298 #if defined(BRIDGE_IPF)
1299 static int
1300 bridge_ioctl_gfilt(struct bridge_softc *sc, void *arg)
1301 {
1302 struct ifbrparam *param = arg;
1303
1304 param->ifbrp_filter = sc->sc_filter_flags;
1305
1306 return 0;
1307 }
1308
1309 static int
1310 bridge_ioctl_sfilt(struct bridge_softc *sc, void *arg)
1311 {
1312 struct ifbrparam *param = arg;
1313 uint32_t nflags, oflags;
1314
1315 if (param->ifbrp_filter & ~IFBF_FILT_MASK)
1316 return EINVAL;
1317
1318 nflags = param->ifbrp_filter;
1319 oflags = sc->sc_filter_flags;
1320
1321 if ((nflags & IFBF_FILT_USEIPF) && !(oflags & IFBF_FILT_USEIPF)) {
1322 pfil_add_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
1323 sc->sc_if.if_pfil);
1324 }
1325 if (!(nflags & IFBF_FILT_USEIPF) && (oflags & IFBF_FILT_USEIPF)) {
1326 pfil_remove_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
1327 sc->sc_if.if_pfil);
1328 }
1329
1330 sc->sc_filter_flags = nflags;
1331
1332 return 0;
1333 }
1334 #endif /* BRIDGE_IPF */
1335
1336 static int
1337 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1338 {
1339 struct ifbreq *req = arg;
1340 struct bridge_iflist *bif;
1341 struct psref psref;
1342
1343 bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
1344 if (bif == NULL)
1345 return ENOENT;
1346
1347 bif->bif_path_cost = req->ifbr_path_cost;
1348
1349 if (sc->sc_if.if_flags & IFF_RUNNING)
1350 bstp_initialization(sc);
1351
1352 bridge_release_member(sc, bif, &psref);
1353
1354 return 0;
1355 }
1356
1357 /*
1358 * bridge_ifdetach:
1359 *
1360 * Detach an interface from a bridge. Called when a member
1361 * interface is detaching.
1362 */
1363 void
1364 bridge_ifdetach(struct ifnet *ifp)
1365 {
1366 struct bridge_softc *sc = ifp->if_bridge;
1367 struct ifbreq breq;
1368
1369 /* ioctl_lock should prevent this from happening */
1370 KASSERT(sc != NULL);
1371
1372 memset(&breq, 0, sizeof(breq));
1373 strlcpy(breq.ifbr_ifsname, ifp->if_xname, sizeof(breq.ifbr_ifsname));
1374
1375 (void) bridge_ioctl_del(sc, &breq);
1376 }
1377
1378 /*
1379 * bridge_init:
1380 *
1381 * Initialize a bridge interface.
1382 */
1383 static int
1384 bridge_init(struct ifnet *ifp)
1385 {
1386 struct bridge_softc *sc = ifp->if_softc;
1387
1388 KASSERT((ifp->if_flags & IFF_RUNNING) == 0);
1389
1390 BRIDGE_LOCK(sc);
1391 sc->sc_stopping = false;
1392 BRIDGE_UNLOCK(sc);
1393
1394 callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1395 bridge_timer, sc);
1396 bstp_initialization(sc);
1397
1398 ifp->if_flags |= IFF_RUNNING;
1399 return 0;
1400 }
1401
1402 /*
1403 * bridge_stop:
1404 *
1405 * Stop the bridge interface.
1406 */
1407 static void
1408 bridge_stop(struct ifnet *ifp, int disable)
1409 {
1410 struct bridge_softc *sc = ifp->if_softc;
1411
1412 KASSERT((ifp->if_flags & IFF_RUNNING) != 0);
1413 ifp->if_flags &= ~IFF_RUNNING;
1414
1415 BRIDGE_LOCK(sc);
1416 sc->sc_stopping = true;
1417 BRIDGE_UNLOCK(sc);
1418
1419 callout_halt(&sc->sc_brcallout, NULL);
1420 workqueue_wait(sc->sc_rtage_wq, &sc->sc_rtage_wk);
1421 bstp_stop(sc);
1422 bridge_rtflush(sc, IFBF_FLUSHDYN);
1423 }
1424
1425 /*
1426 * bridge_enqueue:
1427 *
1428 * Enqueue a packet on a bridge member interface.
1429 */
1430 void
1431 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m,
1432 int runfilt)
1433 {
1434 int len, error;
1435 short mflags;
1436
1437 if (runfilt) {
1438 if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
1439 dst_ifp, PFIL_OUT) != 0) {
1440 if (m != NULL)
1441 m_freem(m);
1442 return;
1443 }
1444 if (m == NULL)
1445 return;
1446 }
1447
1448 #ifdef ALTQ
1449 KERNEL_LOCK(1, NULL);
1450 /*
1451 * If ALTQ is enabled on the member interface, do
1452 * classification; the queueing discipline might
1453 * not require classification, but might require
1454 * the address family/header pointer in the pktattr.
1455 */
1456 if (ALTQ_IS_ENABLED(&dst_ifp->if_snd)) {
1457 /* XXX IFT_ETHER */
1458 altq_etherclassify(&dst_ifp->if_snd, m);
1459 }
1460 KERNEL_UNLOCK_ONE(NULL);
1461 #endif /* ALTQ */
1462
1463 len = m->m_pkthdr.len;
1464 mflags = m->m_flags;
1465
1466 error = if_transmit_lock(dst_ifp, m);
1467 if (error) {
1468 /* mbuf is already freed */
1469 sc->sc_if.if_oerrors++;
1470 return;
1471 }
1472
1473 sc->sc_if.if_opackets++;
1474 sc->sc_if.if_obytes += len;
1475 if (mflags & M_MCAST)
1476 sc->sc_if.if_omcasts++;
1477 }
1478
1479 /*
1480 * bridge_output:
1481 *
1482 * Send output from a bridge member interface. This
1483 * performs the bridging function for locally originated
1484 * packets.
1485 *
1486 * The mbuf has the Ethernet header already attached. We must
1487 * enqueue or free the mbuf before returning.
1488 */
1489 int
1490 bridge_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa,
1491 const struct rtentry *rt)
1492 {
1493 struct ether_header *eh;
1494 struct ifnet *dst_if;
1495 struct bridge_softc *sc;
1496 struct mbuf *n;
1497 int s, bound;
1498
1499 /*
1500 * bridge_output() is called from ether_output(), furthermore
1501 * ifp argument doesn't point to bridge(4). So, don't assert
1502 * IFEF_MPSAFE here.
1503 */
1504
1505 if (m->m_len < ETHER_HDR_LEN) {
1506 m = m_pullup(m, ETHER_HDR_LEN);
1507 if (m == NULL)
1508 return 0;
1509 }
1510
1511 eh = mtod(m, struct ether_header *);
1512 sc = ifp->if_bridge;
1513
1514 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1515 if (memcmp(etherbroadcastaddr,
1516 eh->ether_dhost, ETHER_ADDR_LEN) == 0)
1517 m->m_flags |= M_BCAST;
1518 else
1519 m->m_flags |= M_MCAST;
1520 }
1521
1522 /*
1523 * If bridge is down, but the original output interface is up,
1524 * go ahead and send out that interface. Otherwise, the packet
1525 * is dropped below.
1526 */
1527 if (__predict_false(sc == NULL) ||
1528 (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
1529 dst_if = ifp;
1530 goto unicast_asis;
1531 }
1532
1533 /*
1534 * If the packet is a multicast, or we don't know a better way to
1535 * get there, send to all interfaces.
1536 */
1537 if ((m->m_flags & (M_MCAST | M_BCAST)) != 0)
1538 dst_if = NULL;
1539 else
1540 dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1541
1542 /*
1543 * In general, we need to handle TX offload in software before
1544 * enqueueing a packet. However, we can send it as is in the
1545 * cases of unicast via (1) the source interface, or (2) an
1546 * interface which supports the specified offload options.
1547 * For multicast or broadcast, send it as is only if (3) all
1548 * the member interfaces support the specified options.
1549 */
1550
1551 /*
1552 * Unicast via the source interface.
1553 */
1554 if (dst_if == ifp)
1555 goto unicast_asis;
1556
1557 /*
1558 * Unicast via other interface.
1559 */
1560 if (dst_if != NULL) {
1561 KASSERT(m->m_flags & M_PKTHDR);
1562 if (TX_OFFLOAD_SUPPORTED(dst_if->if_csum_flags_tx,
1563 m->m_pkthdr.csum_flags)) {
1564 /*
1565 * Unicast via an interface which supports the
1566 * specified offload options.
1567 */
1568 goto unicast_asis;
1569 }
1570
1571 /*
1572 * Handle TX offload in software. For TSO, a packet is
1573 * split into multiple chunks. Thus, the return value of
1574 * ether_sw_offload_tx() is mbuf queue consists of them.
1575 */
1576 m = ether_sw_offload_tx(ifp, m);
1577 if (m == NULL)
1578 return 0;
1579
1580 do {
1581 n = m->m_nextpkt;
1582 if ((dst_if->if_flags & IFF_RUNNING) == 0)
1583 m_freem(m);
1584 else
1585 bridge_enqueue(sc, dst_if, m, 0);
1586 m = n;
1587 } while (m != NULL);
1588
1589 return 0;
1590 }
1591
1592 /*
1593 * Multicast or broadcast.
1594 */
1595 if (TX_OFFLOAD_SUPPORTED(sc->sc_csum_flags_tx,
1596 m->m_pkthdr.csum_flags)) {
1597 /*
1598 * Specified TX offload options are supported by all
1599 * the member interfaces of this bridge.
1600 */
1601 m->m_nextpkt = NULL; /* XXX */
1602 } else {
1603 /*
1604 * Otherwise, handle TX offload in software.
1605 */
1606 m = ether_sw_offload_tx(ifp, m);
1607 if (m == NULL)
1608 return 0;
1609 }
1610
1611 /*
1612 * When we use pppoe over bridge, bridge_output() can be called
1613 * in a lwp context by pppoe_timeout_wk().
1614 */
1615 bound = curlwp_bind();
1616 do {
1617 /* XXX Should call bridge_broadcast, but there are locking
1618 * issues which need resolving first. */
1619 struct bridge_iflist *bif;
1620 struct mbuf *mc;
1621 bool used = false;
1622
1623 n = m->m_nextpkt;
1624
1625 BRIDGE_PSZ_RENTER(s);
1626 BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
1627 struct psref psref;
1628
1629 bridge_acquire_member(sc, bif, &psref);
1630 BRIDGE_PSZ_REXIT(s);
1631
1632 dst_if = bif->bif_ifp;
1633 if ((dst_if->if_flags & IFF_RUNNING) == 0)
1634 goto next;
1635
1636 /*
1637 * If this is not the original output interface,
1638 * and the interface is participating in spanning
1639 * tree, make sure the port is in a state that
1640 * allows forwarding.
1641 */
1642 if (dst_if != ifp &&
1643 (bif->bif_flags & IFBIF_STP) != 0) {
1644 switch (bif->bif_state) {
1645 case BSTP_IFSTATE_BLOCKING:
1646 case BSTP_IFSTATE_LISTENING:
1647 case BSTP_IFSTATE_DISABLED:
1648 goto next;
1649 }
1650 }
1651
1652 if (PSLIST_READER_NEXT(bif, struct bridge_iflist,
1653 bif_next) == NULL &&
1654 ((m->m_flags & (M_MCAST | M_BCAST)) == 0 ||
1655 dst_if == ifp))
1656 {
1657 used = true;
1658 mc = m;
1659 } else {
1660 mc = m_copypacket(m, M_DONTWAIT);
1661 if (mc == NULL) {
1662 sc->sc_if.if_oerrors++;
1663 goto next;
1664 }
1665 }
1666
1667 bridge_enqueue(sc, dst_if, mc, 0);
1668
1669 if ((m->m_flags & (M_MCAST | M_BCAST)) != 0 &&
1670 dst_if != ifp)
1671 {
1672 if (PSLIST_READER_NEXT(bif,
1673 struct bridge_iflist, bif_next) == NULL)
1674 {
1675 used = true;
1676 mc = m;
1677 } else {
1678 mc = m_copypacket(m, M_DONTWAIT);
1679 if (mc == NULL) {
1680 sc->sc_if.if_oerrors++;
1681 goto next;
1682 }
1683 }
1684
1685 m_set_rcvif(mc, dst_if);
1686 mc->m_flags &= ~M_PROMISC;
1687
1688 s = splsoftnet();
1689 KERNEL_LOCK_UNLESS_IFP_MPSAFE(dst_if);
1690 ether_input(dst_if, mc);
1691 KERNEL_UNLOCK_UNLESS_IFP_MPSAFE(dst_if);
1692 splx(s);
1693 }
1694
1695 next:
1696 BRIDGE_PSZ_RENTER(s);
1697 bridge_release_member(sc, bif, &psref);
1698
1699 /* Guarantee we don't re-enter the loop as we already
1700 * decided we're at the end. */
1701 if (used)
1702 break;
1703 }
1704 BRIDGE_PSZ_REXIT(s);
1705
1706 if (!used)
1707 m_freem(m);
1708
1709 m = n;
1710 } while (m != NULL);
1711 curlwp_bindx(bound);
1712
1713 return 0;
1714
1715 unicast_asis:
1716 /*
1717 * XXX Spanning tree consideration here?
1718 */
1719 if ((dst_if->if_flags & IFF_RUNNING) == 0)
1720 m_freem(m);
1721 else
1722 bridge_enqueue(sc, dst_if, m, 0);
1723 return 0;
1724 }
1725
1726 /*
1727 * bridge_start:
1728 *
1729 * Start output on a bridge.
1730 *
1731 * NOTE: This routine should never be called in this implementation.
1732 */
1733 static void
1734 bridge_start(struct ifnet *ifp)
1735 {
1736
1737 printf("%s: bridge_start() called\n", ifp->if_xname);
1738 }
1739
1740 /*
1741 * bridge_forward:
1742 *
1743 * The forwarding function of the bridge.
1744 */
1745 static void
1746 bridge_forward(struct bridge_softc *sc, struct mbuf *m)
1747 {
1748 struct bridge_iflist *bif;
1749 struct ifnet *src_if, *dst_if;
1750 struct ether_header *eh;
1751 struct psref psref;
1752 struct psref psref_src;
1753 DECLARE_LOCK_VARIABLE;
1754
1755 if ((sc->sc_if.if_flags & IFF_RUNNING) == 0)
1756 return;
1757
1758 src_if = m_get_rcvif_psref(m, &psref_src);
1759 if (src_if == NULL) {
1760 /* Interface is being destroyed? */
1761 m_freem(m);
1762 goto out;
1763 }
1764
1765 sc->sc_if.if_ipackets++;
1766 sc->sc_if.if_ibytes += m->m_pkthdr.len;
1767
1768 /*
1769 * Look up the bridge_iflist.
1770 */
1771 bif = bridge_lookup_member_if(sc, src_if, &psref);
1772 if (bif == NULL) {
1773 /* Interface is not a bridge member (anymore?) */
1774 m_freem(m);
1775 goto out;
1776 }
1777
1778 if (bif->bif_flags & IFBIF_STP) {
1779 switch (bif->bif_state) {
1780 case BSTP_IFSTATE_BLOCKING:
1781 case BSTP_IFSTATE_LISTENING:
1782 case BSTP_IFSTATE_DISABLED:
1783 m_freem(m);
1784 bridge_release_member(sc, bif, &psref);
1785 goto out;
1786 }
1787 }
1788
1789 eh = mtod(m, struct ether_header *);
1790
1791 /*
1792 * If the interface is learning, and the source
1793 * address is valid and not multicast, record
1794 * the address.
1795 */
1796 if ((bif->bif_flags & IFBIF_LEARNING) != 0 &&
1797 ETHER_IS_MULTICAST(eh->ether_shost) == 0 &&
1798 (eh->ether_shost[0] == 0 &&
1799 eh->ether_shost[1] == 0 &&
1800 eh->ether_shost[2] == 0 &&
1801 eh->ether_shost[3] == 0 &&
1802 eh->ether_shost[4] == 0 &&
1803 eh->ether_shost[5] == 0) == 0) {
1804 (void) bridge_rtupdate(sc, eh->ether_shost,
1805 src_if, 0, IFBAF_DYNAMIC);
1806 }
1807
1808 if ((bif->bif_flags & IFBIF_STP) != 0 &&
1809 bif->bif_state == BSTP_IFSTATE_LEARNING) {
1810 m_freem(m);
1811 bridge_release_member(sc, bif, &psref);
1812 goto out;
1813 }
1814
1815 bridge_release_member(sc, bif, &psref);
1816
1817 /*
1818 * At this point, the port either doesn't participate
1819 * in spanning tree or it is in the forwarding state.
1820 */
1821
1822 /*
1823 * If the packet is unicast, destined for someone on
1824 * "this" side of the bridge, drop it.
1825 */
1826 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
1827 dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1828 if (src_if == dst_if) {
1829 m_freem(m);
1830 goto out;
1831 }
1832 } else {
1833 /* ...forward it to all interfaces. */
1834 sc->sc_if.if_imcasts++;
1835 dst_if = NULL;
1836 }
1837
1838 if (pfil_run_hooks(sc->sc_if.if_pfil, &m, src_if, PFIL_IN) != 0) {
1839 if (m != NULL)
1840 m_freem(m);
1841 goto out;
1842 }
1843 if (m == NULL)
1844 goto out;
1845
1846 if (dst_if == NULL) {
1847 bridge_broadcast(sc, src_if, m);
1848 goto out;
1849 }
1850
1851 m_put_rcvif_psref(src_if, &psref_src);
1852 src_if = NULL;
1853
1854 /*
1855 * At this point, we're dealing with a unicast frame
1856 * going to a different interface.
1857 */
1858 if ((dst_if->if_flags & IFF_RUNNING) == 0) {
1859 m_freem(m);
1860 goto out;
1861 }
1862
1863 bif = bridge_lookup_member_if(sc, dst_if, &psref);
1864 if (bif == NULL) {
1865 /* Not a member of the bridge (anymore?) */
1866 m_freem(m);
1867 goto out;
1868 }
1869
1870 if (bif->bif_flags & IFBIF_STP) {
1871 switch (bif->bif_state) {
1872 case BSTP_IFSTATE_DISABLED:
1873 case BSTP_IFSTATE_BLOCKING:
1874 m_freem(m);
1875 bridge_release_member(sc, bif, &psref);
1876 goto out;
1877 }
1878 }
1879
1880 bridge_release_member(sc, bif, &psref);
1881
1882 /*
1883 * Before enqueueing this packet to the destination interface,
1884 * clear any in-bound checksum flags to prevent them from being
1885 * misused as out-bound flags.
1886 */
1887 m->m_pkthdr.csum_flags = 0;
1888
1889 ACQUIRE_GLOBAL_LOCKS();
1890 bridge_enqueue(sc, dst_if, m, 1);
1891 RELEASE_GLOBAL_LOCKS();
1892 out:
1893 if (src_if != NULL)
1894 m_put_rcvif_psref(src_if, &psref_src);
1895 return;
1896 }
1897
1898 static bool
1899 bstp_state_before_learning(struct bridge_iflist *bif)
1900 {
1901 if (bif->bif_flags & IFBIF_STP) {
1902 switch (bif->bif_state) {
1903 case BSTP_IFSTATE_BLOCKING:
1904 case BSTP_IFSTATE_LISTENING:
1905 case BSTP_IFSTATE_DISABLED:
1906 return true;
1907 }
1908 }
1909 return false;
1910 }
1911
1912 static bool
1913 bridge_ourether(struct bridge_iflist *bif, struct ether_header *eh, int src)
1914 {
1915 uint8_t *ether = src ? eh->ether_shost : eh->ether_dhost;
1916
1917 if (memcmp(CLLADDR(bif->bif_ifp->if_sadl), ether, ETHER_ADDR_LEN) == 0
1918 #if NCARP > 0
1919 || (bif->bif_ifp->if_carp &&
1920 carp_ourether(bif->bif_ifp->if_carp, eh, IFT_ETHER, src) != NULL)
1921 #endif /* NCARP > 0 */
1922 )
1923 return true;
1924
1925 return false;
1926 }
1927
1928 /*
1929 * bridge_input:
1930 *
1931 * Receive input from a member interface. Queue the packet for
1932 * bridging if it is not for us.
1933 */
1934 static void
1935 bridge_input(struct ifnet *ifp, struct mbuf *m)
1936 {
1937 struct bridge_softc *sc = ifp->if_bridge;
1938 struct bridge_iflist *bif;
1939 struct ether_header *eh;
1940 struct psref psref;
1941 int bound;
1942 DECLARE_LOCK_VARIABLE;
1943
1944 KASSERT(!cpu_intr_p());
1945
1946 if (__predict_false(sc == NULL) ||
1947 (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
1948 ACQUIRE_GLOBAL_LOCKS();
1949 ether_input(ifp, m);
1950 RELEASE_GLOBAL_LOCKS();
1951 return;
1952 }
1953
1954 bound = curlwp_bind();
1955 bif = bridge_lookup_member_if(sc, ifp, &psref);
1956 if (bif == NULL) {
1957 curlwp_bindx(bound);
1958 ACQUIRE_GLOBAL_LOCKS();
1959 ether_input(ifp, m);
1960 RELEASE_GLOBAL_LOCKS();
1961 return;
1962 }
1963
1964 eh = mtod(m, struct ether_header *);
1965
1966 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1967 if (memcmp(etherbroadcastaddr,
1968 eh->ether_dhost, ETHER_ADDR_LEN) == 0)
1969 m->m_flags |= M_BCAST;
1970 else
1971 m->m_flags |= M_MCAST;
1972 }
1973
1974 /*
1975 * A 'fast' path for packets addressed to interfaces that are
1976 * part of this bridge.
1977 */
1978 if (!(m->m_flags & (M_BCAST|M_MCAST)) &&
1979 !bstp_state_before_learning(bif)) {
1980 struct bridge_iflist *_bif;
1981 struct ifnet *_ifp = NULL;
1982 int s;
1983 struct psref _psref;
1984
1985 BRIDGE_PSZ_RENTER(s);
1986 BRIDGE_IFLIST_READER_FOREACH(_bif, sc) {
1987 /* It is destined for us. */
1988 if (bridge_ourether(_bif, eh, 0)) {
1989 bridge_acquire_member(sc, _bif, &_psref);
1990 BRIDGE_PSZ_REXIT(s);
1991 if (_bif->bif_flags & IFBIF_LEARNING)
1992 (void) bridge_rtupdate(sc,
1993 eh->ether_shost, ifp, 0, IFBAF_DYNAMIC);
1994 m_set_rcvif(m, _bif->bif_ifp);
1995 _ifp = _bif->bif_ifp;
1996 bridge_release_member(sc, _bif, &_psref);
1997 goto out;
1998 }
1999
2000 /* We just received a packet that we sent out. */
2001 if (bridge_ourether(_bif, eh, 1))
2002 break;
2003 }
2004 BRIDGE_PSZ_REXIT(s);
2005 out:
2006
2007 if (_bif != NULL) {
2008 bridge_release_member(sc, bif, &psref);
2009 curlwp_bindx(bound);
2010 if (_ifp != NULL) {
2011 m->m_flags &= ~M_PROMISC;
2012 ACQUIRE_GLOBAL_LOCKS();
2013 ether_input(_ifp, m);
2014 RELEASE_GLOBAL_LOCKS();
2015 } else
2016 m_freem(m);
2017 return;
2018 }
2019 }
2020
2021 /* Tap off 802.1D packets; they do not get forwarded. */
2022 if (bif->bif_flags & IFBIF_STP &&
2023 memcmp(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN) == 0) {
2024 bstp_input(sc, bif, m);
2025 bridge_release_member(sc, bif, &psref);
2026 curlwp_bindx(bound);
2027 return;
2028 }
2029
2030 /*
2031 * A normal switch would discard the packet here, but that's not what
2032 * we've done historically. This also prevents some obnoxious behaviour.
2033 */
2034 if (bstp_state_before_learning(bif)) {
2035 bridge_release_member(sc, bif, &psref);
2036 curlwp_bindx(bound);
2037 ACQUIRE_GLOBAL_LOCKS();
2038 ether_input(ifp, m);
2039 RELEASE_GLOBAL_LOCKS();
2040 return;
2041 }
2042
2043 bridge_release_member(sc, bif, &psref);
2044
2045 bridge_forward(sc, m);
2046
2047 curlwp_bindx(bound);
2048 }
2049
2050 /*
2051 * bridge_broadcast:
2052 *
2053 * Send a frame to all interfaces that are members of
2054 * the bridge, except for the one on which the packet
2055 * arrived.
2056 */
2057 static void
2058 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2059 struct mbuf *m)
2060 {
2061 struct bridge_iflist *bif;
2062 struct mbuf *mc;
2063 struct ifnet *dst_if;
2064 bool bmcast;
2065 int s;
2066 DECLARE_LOCK_VARIABLE;
2067
2068 bmcast = m->m_flags & (M_BCAST|M_MCAST);
2069
2070 BRIDGE_PSZ_RENTER(s);
2071 BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
2072 struct psref psref;
2073
2074 bridge_acquire_member(sc, bif, &psref);
2075 BRIDGE_PSZ_REXIT(s);
2076
2077 dst_if = bif->bif_ifp;
2078
2079 if (bif->bif_flags & IFBIF_STP) {
2080 switch (bif->bif_state) {
2081 case BSTP_IFSTATE_BLOCKING:
2082 case BSTP_IFSTATE_DISABLED:
2083 goto next;
2084 }
2085 }
2086
2087 if ((bif->bif_flags & IFBIF_DISCOVER) == 0 && !bmcast)
2088 goto next;
2089
2090 if ((dst_if->if_flags & IFF_RUNNING) == 0)
2091 goto next;
2092
2093 if (dst_if != src_if) {
2094 mc = m_copypacket(m, M_DONTWAIT);
2095 if (mc == NULL) {
2096 sc->sc_if.if_oerrors++;
2097 goto next;
2098 }
2099 /*
2100 * Before enqueueing this packet to the destination
2101 * interface, clear any in-bound checksum flags to
2102 * prevent them from being misused as out-bound flags.
2103 */
2104 mc->m_pkthdr.csum_flags = 0;
2105
2106 ACQUIRE_GLOBAL_LOCKS();
2107 bridge_enqueue(sc, dst_if, mc, 1);
2108 RELEASE_GLOBAL_LOCKS();
2109 }
2110
2111 if (bmcast) {
2112 mc = m_copypacket(m, M_DONTWAIT);
2113 if (mc == NULL) {
2114 sc->sc_if.if_oerrors++;
2115 goto next;
2116 }
2117
2118 m_set_rcvif(mc, dst_if);
2119 mc->m_flags &= ~M_PROMISC;
2120
2121 ACQUIRE_GLOBAL_LOCKS();
2122 ether_input(dst_if, mc);
2123 RELEASE_GLOBAL_LOCKS();
2124 }
2125 next:
2126 BRIDGE_PSZ_RENTER(s);
2127 bridge_release_member(sc, bif, &psref);
2128 }
2129 BRIDGE_PSZ_REXIT(s);
2130
2131 m_freem(m);
2132 }
2133
2134 static int
2135 bridge_rtalloc(struct bridge_softc *sc, const uint8_t *dst,
2136 struct bridge_rtnode **brtp)
2137 {
2138 struct bridge_rtnode *brt;
2139 int error;
2140
2141 if (sc->sc_brtcnt >= sc->sc_brtmax)
2142 return ENOSPC;
2143
2144 /*
2145 * Allocate a new bridge forwarding node, and
2146 * initialize the expiration time and Ethernet
2147 * address.
2148 */
2149 brt = pool_get(&bridge_rtnode_pool, PR_NOWAIT);
2150 if (brt == NULL)
2151 return ENOMEM;
2152
2153 memset(brt, 0, sizeof(*brt));
2154 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2155 brt->brt_flags = IFBAF_DYNAMIC;
2156 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2157 PSLIST_ENTRY_INIT(brt, brt_list);
2158 PSLIST_ENTRY_INIT(brt, brt_hash);
2159
2160 BRIDGE_RT_LOCK(sc);
2161 error = bridge_rtnode_insert(sc, brt);
2162 BRIDGE_RT_UNLOCK(sc);
2163
2164 if (error != 0) {
2165 pool_put(&bridge_rtnode_pool, brt);
2166 return error;
2167 }
2168
2169 *brtp = brt;
2170 return 0;
2171 }
2172
2173 /*
2174 * bridge_rtupdate:
2175 *
2176 * Add a bridge routing entry.
2177 */
2178 static int
2179 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
2180 struct ifnet *dst_if, int setflags, uint8_t flags)
2181 {
2182 struct bridge_rtnode *brt;
2183 int s;
2184
2185 again:
2186 /*
2187 * A route for this destination might already exist. If so,
2188 * update it, otherwise create a new one.
2189 */
2190 BRIDGE_RT_RENTER(s);
2191 brt = bridge_rtnode_lookup(sc, dst);
2192
2193 if (brt != NULL) {
2194 brt->brt_ifp = dst_if;
2195 if (setflags) {
2196 brt->brt_flags = flags;
2197 if (flags & IFBAF_STATIC)
2198 brt->brt_expire = 0;
2199 else
2200 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2201 } else {
2202 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2203 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2204 }
2205 }
2206 BRIDGE_RT_REXIT(s);
2207
2208 if (brt == NULL) {
2209 int r;
2210
2211 r = bridge_rtalloc(sc, dst, &brt);
2212 if (r != 0)
2213 return r;
2214 goto again;
2215 }
2216
2217 return 0;
2218 }
2219
2220 /*
2221 * bridge_rtlookup:
2222 *
2223 * Lookup the destination interface for an address.
2224 */
2225 static struct ifnet *
2226 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr)
2227 {
2228 struct bridge_rtnode *brt;
2229 struct ifnet *ifs = NULL;
2230 int s;
2231
2232 BRIDGE_RT_RENTER(s);
2233 brt = bridge_rtnode_lookup(sc, addr);
2234 if (brt != NULL)
2235 ifs = brt->brt_ifp;
2236 BRIDGE_RT_REXIT(s);
2237
2238 return ifs;
2239 }
2240
2241 typedef bool (*bridge_iterate_cb_t)
2242 (struct bridge_softc *, struct bridge_rtnode *, bool *, void *);
2243
2244 /*
2245 * bridge_rtlist_iterate_remove:
2246 *
2247 * It iterates on sc->sc_rtlist and removes rtnodes of it which func
2248 * callback judges to remove. Removals of rtnodes are done in a manner
2249 * of pserialize. To this end, all kmem_* operations are placed out of
2250 * mutexes.
2251 */
2252 static void
2253 bridge_rtlist_iterate_remove(struct bridge_softc *sc, bridge_iterate_cb_t func, void *arg)
2254 {
2255 struct bridge_rtnode *brt;
2256 struct bridge_rtnode **brt_list;
2257 int i, count;
2258
2259 retry:
2260 count = sc->sc_brtcnt;
2261 if (count == 0)
2262 return;
2263 brt_list = kmem_alloc(sizeof(*brt_list) * count, KM_SLEEP);
2264
2265 BRIDGE_RT_LOCK(sc);
2266 if (__predict_false(sc->sc_brtcnt > count)) {
2267 /* The rtnodes increased, we need more memory */
2268 BRIDGE_RT_UNLOCK(sc);
2269 kmem_free(brt_list, sizeof(*brt_list) * count);
2270 goto retry;
2271 }
2272
2273 i = 0;
2274 /*
2275 * We don't need to use a _SAFE variant here because we know
2276 * that a removed item keeps its next pointer as-is thanks to
2277 * pslist(9) and isn't freed in the loop.
2278 */
2279 BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) {
2280 bool need_break = false;
2281 if (func(sc, brt, &need_break, arg)) {
2282 bridge_rtnode_remove(sc, brt);
2283 brt_list[i++] = brt;
2284 }
2285 if (need_break)
2286 break;
2287 }
2288
2289 if (i > 0)
2290 BRIDGE_RT_PSZ_PERFORM(sc);
2291 BRIDGE_RT_UNLOCK(sc);
2292
2293 while (--i >= 0)
2294 bridge_rtnode_destroy(brt_list[i]);
2295
2296 kmem_free(brt_list, sizeof(*brt_list) * count);
2297 }
2298
2299 static bool
2300 bridge_rttrim0_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
2301 bool *need_break, void *arg)
2302 {
2303 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2304 /* Take into account of the subsequent removal */
2305 if ((sc->sc_brtcnt - 1) <= sc->sc_brtmax)
2306 *need_break = true;
2307 return true;
2308 } else
2309 return false;
2310 }
2311
2312 static void
2313 bridge_rttrim0(struct bridge_softc *sc)
2314 {
2315 bridge_rtlist_iterate_remove(sc, bridge_rttrim0_cb, NULL);
2316 }
2317
2318 /*
2319 * bridge_rttrim:
2320 *
2321 * Trim the routine table so that we have a number
2322 * of routing entries less than or equal to the
2323 * maximum number.
2324 */
2325 static void
2326 bridge_rttrim(struct bridge_softc *sc)
2327 {
2328
2329 /* Make sure we actually need to do this. */
2330 if (sc->sc_brtcnt <= sc->sc_brtmax)
2331 return;
2332
2333 /* Force an aging cycle; this might trim enough addresses. */
2334 bridge_rtage(sc);
2335 if (sc->sc_brtcnt <= sc->sc_brtmax)
2336 return;
2337
2338 bridge_rttrim0(sc);
2339
2340 return;
2341 }
2342
2343 /*
2344 * bridge_timer:
2345 *
2346 * Aging timer for the bridge.
2347 */
2348 static void
2349 bridge_timer(void *arg)
2350 {
2351 struct bridge_softc *sc = arg;
2352
2353 workqueue_enqueue(sc->sc_rtage_wq, &sc->sc_rtage_wk, NULL);
2354 }
2355
2356 static void
2357 bridge_rtage_work(struct work *wk, void *arg)
2358 {
2359 struct bridge_softc *sc = arg;
2360
2361 KASSERT(wk == &sc->sc_rtage_wk);
2362
2363 bridge_rtage(sc);
2364
2365 BRIDGE_LOCK(sc);
2366 if (!sc->sc_stopping) {
2367 callout_reset(&sc->sc_brcallout,
2368 bridge_rtable_prune_period * hz, bridge_timer, sc);
2369 }
2370 BRIDGE_UNLOCK(sc);
2371 }
2372
2373 static bool
2374 bridge_rtage_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
2375 bool *need_break, void *arg)
2376 {
2377 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2378 time_uptime >= brt->brt_expire)
2379 return true;
2380 else
2381 return false;
2382 }
2383
2384 /*
2385 * bridge_rtage:
2386 *
2387 * Perform an aging cycle.
2388 */
2389 static void
2390 bridge_rtage(struct bridge_softc *sc)
2391 {
2392 bridge_rtlist_iterate_remove(sc, bridge_rtage_cb, NULL);
2393 }
2394
2395
2396 static bool
2397 bridge_rtflush_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
2398 bool *need_break, void *arg)
2399 {
2400 int full = *(int*)arg;
2401
2402 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2403 return true;
2404 else
2405 return false;
2406 }
2407
2408 /*
2409 * bridge_rtflush:
2410 *
2411 * Remove all dynamic addresses from the bridge.
2412 */
2413 static void
2414 bridge_rtflush(struct bridge_softc *sc, int full)
2415 {
2416 bridge_rtlist_iterate_remove(sc, bridge_rtflush_cb, &full);
2417 }
2418
2419 /*
2420 * bridge_rtdaddr:
2421 *
2422 * Remove an address from the table.
2423 */
2424 static int
2425 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr)
2426 {
2427 struct bridge_rtnode *brt;
2428
2429 BRIDGE_RT_LOCK(sc);
2430 if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL) {
2431 BRIDGE_RT_UNLOCK(sc);
2432 return ENOENT;
2433 }
2434 bridge_rtnode_remove(sc, brt);
2435 BRIDGE_RT_PSZ_PERFORM(sc);
2436 BRIDGE_RT_UNLOCK(sc);
2437
2438 bridge_rtnode_destroy(brt);
2439
2440 return 0;
2441 }
2442
2443 /*
2444 * bridge_rtdelete:
2445 *
2446 * Delete routes to a speicifc member interface.
2447 */
2448 static void
2449 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp)
2450 {
2451 struct bridge_rtnode *brt;
2452
2453 /* XXX pserialize_perform for each entry is slow */
2454 again:
2455 BRIDGE_RT_LOCK(sc);
2456 BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) {
2457 if (brt->brt_ifp == ifp)
2458 break;
2459 }
2460 if (brt == NULL) {
2461 BRIDGE_RT_UNLOCK(sc);
2462 return;
2463 }
2464 bridge_rtnode_remove(sc, brt);
2465 BRIDGE_RT_PSZ_PERFORM(sc);
2466 BRIDGE_RT_UNLOCK(sc);
2467
2468 bridge_rtnode_destroy(brt);
2469
2470 goto again;
2471 }
2472
2473 /*
2474 * bridge_rtable_init:
2475 *
2476 * Initialize the route table for this bridge.
2477 */
2478 static void
2479 bridge_rtable_init(struct bridge_softc *sc)
2480 {
2481 int i;
2482
2483 sc->sc_rthash = kmem_alloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2484 KM_SLEEP);
2485
2486 for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2487 PSLIST_INIT(&sc->sc_rthash[i]);
2488
2489 sc->sc_rthash_key = cprng_fast32();
2490
2491 PSLIST_INIT(&sc->sc_rtlist);
2492
2493 sc->sc_rtlist_psz = pserialize_create();
2494 sc->sc_rtlist_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_SOFTNET);
2495 }
2496
2497 /*
2498 * bridge_rtable_fini:
2499 *
2500 * Deconstruct the route table for this bridge.
2501 */
2502 static void
2503 bridge_rtable_fini(struct bridge_softc *sc)
2504 {
2505
2506 kmem_free(sc->sc_rthash, sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE);
2507 mutex_obj_free(sc->sc_rtlist_lock);
2508 pserialize_destroy(sc->sc_rtlist_psz);
2509 }
2510
2511 /*
2512 * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2513 * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2514 */
2515 #define mix(a, b, c) \
2516 do { \
2517 a -= b; a -= c; a ^= (c >> 13); \
2518 b -= c; b -= a; b ^= (a << 8); \
2519 c -= a; c -= b; c ^= (b >> 13); \
2520 a -= b; a -= c; a ^= (c >> 12); \
2521 b -= c; b -= a; b ^= (a << 16); \
2522 c -= a; c -= b; c ^= (b >> 5); \
2523 a -= b; a -= c; a ^= (c >> 3); \
2524 b -= c; b -= a; b ^= (a << 10); \
2525 c -= a; c -= b; c ^= (b >> 15); \
2526 } while (/*CONSTCOND*/0)
2527
2528 static inline uint32_t
2529 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2530 {
2531 uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2532
2533 b += addr[5] << 8;
2534 b += addr[4];
2535 a += addr[3] << 24;
2536 a += addr[2] << 16;
2537 a += addr[1] << 8;
2538 a += addr[0];
2539
2540 mix(a, b, c);
2541
2542 return (c & BRIDGE_RTHASH_MASK);
2543 }
2544
2545 #undef mix
2546
2547 /*
2548 * bridge_rtnode_lookup:
2549 *
2550 * Look up a bridge route node for the specified destination.
2551 */
2552 static struct bridge_rtnode *
2553 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr)
2554 {
2555 struct bridge_rtnode *brt;
2556 uint32_t hash;
2557 int dir;
2558
2559 hash = bridge_rthash(sc, addr);
2560 BRIDGE_RTHASH_READER_FOREACH(brt, sc, hash) {
2561 dir = memcmp(addr, brt->brt_addr, ETHER_ADDR_LEN);
2562 if (dir == 0)
2563 return brt;
2564 if (dir > 0)
2565 return NULL;
2566 }
2567
2568 return NULL;
2569 }
2570
2571 /*
2572 * bridge_rtnode_insert:
2573 *
2574 * Insert the specified bridge node into the route table. We
2575 * assume the entry is not already in the table.
2576 */
2577 static int
2578 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2579 {
2580 struct bridge_rtnode *lbrt, *prev = NULL;
2581 uint32_t hash;
2582
2583 KASSERT(BRIDGE_RT_LOCKED(sc));
2584
2585 hash = bridge_rthash(sc, brt->brt_addr);
2586 BRIDGE_RTHASH_WRITER_FOREACH(lbrt, sc, hash) {
2587 int dir = memcmp(brt->brt_addr, lbrt->brt_addr, ETHER_ADDR_LEN);
2588 if (dir == 0)
2589 return EEXIST;
2590 if (dir > 0)
2591 break;
2592 prev = lbrt;
2593 }
2594 if (prev == NULL)
2595 BRIDGE_RTHASH_WRITER_INSERT_HEAD(sc, hash, brt);
2596 else
2597 BRIDGE_RTHASH_WRITER_INSERT_AFTER(prev, brt);
2598
2599 BRIDGE_RTLIST_WRITER_INSERT_HEAD(sc, brt);
2600 sc->sc_brtcnt++;
2601
2602 return 0;
2603 }
2604
2605 /*
2606 * bridge_rtnode_remove:
2607 *
2608 * Remove a bridge rtnode from the rthash and the rtlist of a bridge.
2609 */
2610 static void
2611 bridge_rtnode_remove(struct bridge_softc *sc, struct bridge_rtnode *brt)
2612 {
2613
2614 KASSERT(BRIDGE_RT_LOCKED(sc));
2615
2616 BRIDGE_RTHASH_WRITER_REMOVE(brt);
2617 BRIDGE_RTLIST_WRITER_REMOVE(brt);
2618 sc->sc_brtcnt--;
2619 }
2620
2621 /*
2622 * bridge_rtnode_destroy:
2623 *
2624 * Destroy a bridge rtnode.
2625 */
2626 static void
2627 bridge_rtnode_destroy(struct bridge_rtnode *brt)
2628 {
2629
2630 PSLIST_ENTRY_DESTROY(brt, brt_list);
2631 PSLIST_ENTRY_DESTROY(brt, brt_hash);
2632 pool_put(&bridge_rtnode_pool, brt);
2633 }
2634
2635 #if defined(BRIDGE_IPF)
2636 extern pfil_head_t *inet_pfil_hook; /* XXX */
2637 extern pfil_head_t *inet6_pfil_hook; /* XXX */
2638
2639 /*
2640 * Send bridge packets through IPF if they are one of the types IPF can deal
2641 * with, or if they are ARP or REVARP. (IPF will pass ARP and REVARP without
2642 * question.)
2643 */
2644 static int
2645 bridge_ipf(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
2646 {
2647 int snap, error;
2648 struct ether_header *eh1, eh2;
2649 struct llc llc1;
2650 uint16_t ether_type;
2651
2652 snap = 0;
2653 error = -1; /* Default error if not error == 0 */
2654 eh1 = mtod(*mp, struct ether_header *);
2655 ether_type = ntohs(eh1->ether_type);
2656
2657 /*
2658 * Check for SNAP/LLC.
2659 */
2660 if (ether_type < ETHERMTU) {
2661 struct llc *llc2 = (struct llc *)(eh1 + 1);
2662
2663 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
2664 llc2->llc_dsap == LLC_SNAP_LSAP &&
2665 llc2->llc_ssap == LLC_SNAP_LSAP &&
2666 llc2->llc_control == LLC_UI) {
2667 ether_type = htons(llc2->llc_un.type_snap.ether_type);
2668 snap = 1;
2669 }
2670 }
2671
2672 /*
2673 * If we're trying to filter bridge traffic, don't look at anything
2674 * other than IP and ARP traffic. If the filter doesn't understand
2675 * IPv6, don't allow IPv6 through the bridge either. This is lame
2676 * since if we really wanted, say, an AppleTalk filter, we are hosed,
2677 * but of course we don't have an AppleTalk filter to begin with.
2678 * (Note that since IPF doesn't understand ARP it will pass *ALL*
2679 * ARP traffic.)
2680 */
2681 switch (ether_type) {
2682 case ETHERTYPE_ARP:
2683 case ETHERTYPE_REVARP:
2684 return 0; /* Automatically pass */
2685 case ETHERTYPE_IP:
2686 # ifdef INET6
2687 case ETHERTYPE_IPV6:
2688 # endif /* INET6 */
2689 break;
2690 default:
2691 goto bad;
2692 }
2693
2694 /* Strip off the Ethernet header and keep a copy. */
2695 m_copydata(*mp, 0, ETHER_HDR_LEN, (void *) &eh2);
2696 m_adj(*mp, ETHER_HDR_LEN);
2697
2698 /* Strip off snap header, if present */
2699 if (snap) {
2700 m_copydata(*mp, 0, sizeof(struct llc), (void *) &llc1);
2701 m_adj(*mp, sizeof(struct llc));
2702 }
2703
2704 /*
2705 * Check basic packet sanity and run IPF through pfil.
2706 */
2707 KASSERT(!cpu_intr_p());
2708 switch (ether_type)
2709 {
2710 case ETHERTYPE_IP :
2711 error = bridge_ip_checkbasic(mp);
2712 if (error == 0)
2713 error = pfil_run_hooks(inet_pfil_hook, mp, ifp, dir);
2714 break;
2715 # ifdef INET6
2716 case ETHERTYPE_IPV6 :
2717 error = bridge_ip6_checkbasic(mp);
2718 if (error == 0)
2719 error = pfil_run_hooks(inet6_pfil_hook, mp, ifp, dir);
2720 break;
2721 # endif
2722 default :
2723 error = 0;
2724 break;
2725 }
2726
2727 if (*mp == NULL)
2728 return error;
2729 if (error != 0)
2730 goto bad;
2731
2732 error = -1;
2733
2734 /*
2735 * Finally, put everything back the way it was and return
2736 */
2737 if (snap) {
2738 M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
2739 if (*mp == NULL)
2740 return error;
2741 bcopy(&llc1, mtod(*mp, void *), sizeof(struct llc));
2742 }
2743
2744 M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
2745 if (*mp == NULL)
2746 return error;
2747 bcopy(&eh2, mtod(*mp, void *), ETHER_HDR_LEN);
2748
2749 return 0;
2750
2751 bad:
2752 m_freem(*mp);
2753 *mp = NULL;
2754 return error;
2755 }
2756
2757 /*
2758 * Perform basic checks on header size since
2759 * IPF assumes ip_input has already processed
2760 * it for it. Cut-and-pasted from ip_input.c.
2761 * Given how simple the IPv6 version is,
2762 * does the IPv4 version really need to be
2763 * this complicated?
2764 *
2765 * XXX Should we update ipstat here, or not?
2766 * XXX Right now we update ipstat but not
2767 * XXX csum_counter.
2768 */
2769 static int
2770 bridge_ip_checkbasic(struct mbuf **mp)
2771 {
2772 struct mbuf *m = *mp;
2773 struct ip *ip;
2774 int len, hlen;
2775
2776 if (*mp == NULL)
2777 return -1;
2778
2779 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
2780 if ((m = m_copyup(m, sizeof(struct ip),
2781 (max_linkhdr + 3) & ~3)) == NULL) {
2782 /* XXXJRT new stat, please */
2783 ip_statinc(IP_STAT_TOOSMALL);
2784 goto bad;
2785 }
2786 } else if (__predict_false(m->m_len < sizeof (struct ip))) {
2787 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
2788 ip_statinc(IP_STAT_TOOSMALL);
2789 goto bad;
2790 }
2791 }
2792 ip = mtod(m, struct ip *);
2793 if (ip == NULL) goto bad;
2794
2795 if (ip->ip_v != IPVERSION) {
2796 ip_statinc(IP_STAT_BADVERS);
2797 goto bad;
2798 }
2799 hlen = ip->ip_hl << 2;
2800 if (hlen < sizeof(struct ip)) { /* minimum header length */
2801 ip_statinc(IP_STAT_BADHLEN);
2802 goto bad;
2803 }
2804 if (hlen > m->m_len) {
2805 if ((m = m_pullup(m, hlen)) == 0) {
2806 ip_statinc(IP_STAT_BADHLEN);
2807 goto bad;
2808 }
2809 ip = mtod(m, struct ip *);
2810 if (ip == NULL) goto bad;
2811 }
2812
2813 switch (m->m_pkthdr.csum_flags &
2814 ((m_get_rcvif_NOMPSAFE(m)->if_csum_flags_rx & M_CSUM_IPv4) |
2815 M_CSUM_IPv4_BAD)) {
2816 case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
2817 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); */
2818 goto bad;
2819
2820 case M_CSUM_IPv4:
2821 /* Checksum was okay. */
2822 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); */
2823 break;
2824
2825 default:
2826 /* Must compute it ourselves. */
2827 /* INET_CSUM_COUNTER_INCR(&ip_swcsum); */
2828 if (in_cksum(m, hlen) != 0)
2829 goto bad;
2830 break;
2831 }
2832
2833 /* Retrieve the packet length. */
2834 len = ntohs(ip->ip_len);
2835
2836 /*
2837 * Check for additional length bogosity
2838 */
2839 if (len < hlen) {
2840 ip_statinc(IP_STAT_BADLEN);
2841 goto bad;
2842 }
2843
2844 /*
2845 * Check that the amount of data in the buffers
2846 * is as at least much as the IP header would have us expect.
2847 * Drop packet if shorter than we expect.
2848 */
2849 if (m->m_pkthdr.len < len) {
2850 ip_statinc(IP_STAT_TOOSHORT);
2851 goto bad;
2852 }
2853
2854 /* Checks out, proceed */
2855 *mp = m;
2856 return 0;
2857
2858 bad:
2859 *mp = m;
2860 return -1;
2861 }
2862
2863 # ifdef INET6
2864 /*
2865 * Same as above, but for IPv6.
2866 * Cut-and-pasted from ip6_input.c.
2867 * XXX Should we update ip6stat, or not?
2868 */
2869 static int
2870 bridge_ip6_checkbasic(struct mbuf **mp)
2871 {
2872 struct mbuf *m = *mp;
2873 struct ip6_hdr *ip6;
2874
2875 /*
2876 * If the IPv6 header is not aligned, slurp it up into a new
2877 * mbuf with space for link headers, in the event we forward
2878 * it. Otherwise, if it is aligned, make sure the entire base
2879 * IPv6 header is in the first mbuf of the chain.
2880 */
2881 if (IP6_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
2882 struct ifnet *inifp = m_get_rcvif_NOMPSAFE(m);
2883 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
2884 (max_linkhdr + 3) & ~3)) == NULL) {
2885 /* XXXJRT new stat, please */
2886 ip6_statinc(IP6_STAT_TOOSMALL);
2887 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
2888 goto bad;
2889 }
2890 } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
2891 struct ifnet *inifp = m_get_rcvif_NOMPSAFE(m);
2892 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
2893 ip6_statinc(IP6_STAT_TOOSMALL);
2894 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
2895 goto bad;
2896 }
2897 }
2898
2899 ip6 = mtod(m, struct ip6_hdr *);
2900
2901 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
2902 ip6_statinc(IP6_STAT_BADVERS);
2903 in6_ifstat_inc(m_get_rcvif_NOMPSAFE(m), ifs6_in_hdrerr);
2904 goto bad;
2905 }
2906
2907 /* Checks out, proceed */
2908 *mp = m;
2909 return 0;
2910
2911 bad:
2912 *mp = m;
2913 return -1;
2914 }
2915 # endif /* INET6 */
2916 #endif /* BRIDGE_IPF */
2917