if_bridge.c revision 1.148.2.5 1 /* $NetBSD: if_bridge.c,v 1.148.2.5 2018/09/30 01:45:56 pgoyette 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 1.148.2.5 2018/09/30 01:45:56 pgoyette 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
116 #if defined(BRIDGE_IPF)
117 /* Used for bridge_ip[6]_checkbasic */
118 #include <netinet/in.h>
119 #include <netinet/in_systm.h>
120 #include <netinet/ip.h>
121 #include <netinet/ip_var.h>
122 #include <netinet/ip_private.h> /* XXX */
123
124 #include <netinet/ip6.h>
125 #include <netinet6/in6_var.h>
126 #include <netinet6/ip6_var.h>
127 #include <netinet6/ip6_private.h> /* XXX */
128 #endif /* BRIDGE_IPF */
129
130 /*
131 * Size of the route hash table. Must be a power of two.
132 */
133 #ifndef BRIDGE_RTHASH_SIZE
134 #define BRIDGE_RTHASH_SIZE 1024
135 #endif
136
137 #define BRIDGE_RTHASH_MASK (BRIDGE_RTHASH_SIZE - 1)
138
139 #include "carp.h"
140 #if NCARP > 0
141 #include <netinet/in.h>
142 #include <netinet/in_var.h>
143 #include <netinet/ip_carp.h>
144 #endif
145
146 #include "ioconf.h"
147
148 __CTASSERT(sizeof(struct ifbifconf) == sizeof(struct ifbaconf));
149 __CTASSERT(offsetof(struct ifbifconf, ifbic_len) == offsetof(struct ifbaconf, ifbac_len));
150 __CTASSERT(offsetof(struct ifbifconf, ifbic_buf) == offsetof(struct ifbaconf, ifbac_buf));
151
152 /*
153 * Maximum number of addresses to cache.
154 */
155 #ifndef BRIDGE_RTABLE_MAX
156 #define BRIDGE_RTABLE_MAX 100
157 #endif
158
159 /*
160 * Spanning tree defaults.
161 */
162 #define BSTP_DEFAULT_MAX_AGE (20 * 256)
163 #define BSTP_DEFAULT_HELLO_TIME (2 * 256)
164 #define BSTP_DEFAULT_FORWARD_DELAY (15 * 256)
165 #define BSTP_DEFAULT_HOLD_TIME (1 * 256)
166 #define BSTP_DEFAULT_BRIDGE_PRIORITY 0x8000
167 #define BSTP_DEFAULT_PORT_PRIORITY 0x80
168 #define BSTP_DEFAULT_PATH_COST 55
169
170 /*
171 * Timeout (in seconds) for entries learned dynamically.
172 */
173 #ifndef BRIDGE_RTABLE_TIMEOUT
174 #define BRIDGE_RTABLE_TIMEOUT (20 * 60) /* same as ARP */
175 #endif
176
177 /*
178 * Number of seconds between walks of the route list.
179 */
180 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
181 #define BRIDGE_RTABLE_PRUNE_PERIOD (5 * 60)
182 #endif
183
184 #define BRIDGE_RT_LOCK(_sc) mutex_enter((_sc)->sc_rtlist_lock)
185 #define BRIDGE_RT_UNLOCK(_sc) mutex_exit((_sc)->sc_rtlist_lock)
186 #define BRIDGE_RT_LOCKED(_sc) mutex_owned((_sc)->sc_rtlist_lock)
187
188 #define BRIDGE_RT_PSZ_PERFORM(_sc) \
189 pserialize_perform((_sc)->sc_rtlist_psz)
190
191 #define BRIDGE_RT_RENTER(__s) do { __s = pserialize_read_enter(); } while (0)
192 #define BRIDGE_RT_REXIT(__s) do { pserialize_read_exit(__s); } while (0)
193
194 #define BRIDGE_RTLIST_READER_FOREACH(_brt, _sc) \
195 PSLIST_READER_FOREACH((_brt), &((_sc)->sc_rtlist), \
196 struct bridge_rtnode, brt_list)
197 #define BRIDGE_RTLIST_WRITER_FOREACH(_brt, _sc) \
198 PSLIST_WRITER_FOREACH((_brt), &((_sc)->sc_rtlist), \
199 struct bridge_rtnode, brt_list)
200 #define BRIDGE_RTLIST_WRITER_INSERT_HEAD(_sc, _brt) \
201 PSLIST_WRITER_INSERT_HEAD(&(_sc)->sc_rtlist, brt, brt_list)
202 #define BRIDGE_RTLIST_WRITER_REMOVE(_brt) \
203 PSLIST_WRITER_REMOVE((_brt), brt_list)
204
205 #define BRIDGE_RTHASH_READER_FOREACH(_brt, _sc, _hash) \
206 PSLIST_READER_FOREACH((_brt), &(_sc)->sc_rthash[(_hash)], \
207 struct bridge_rtnode, brt_hash)
208 #define BRIDGE_RTHASH_WRITER_FOREACH(_brt, _sc, _hash) \
209 PSLIST_WRITER_FOREACH((_brt), &(_sc)->sc_rthash[(_hash)], \
210 struct bridge_rtnode, brt_hash)
211 #define BRIDGE_RTHASH_WRITER_INSERT_HEAD(_sc, _hash, _brt) \
212 PSLIST_WRITER_INSERT_HEAD(&(_sc)->sc_rthash[(_hash)], brt, brt_hash)
213 #define BRIDGE_RTHASH_WRITER_INSERT_AFTER(_brt, _new) \
214 PSLIST_WRITER_INSERT_AFTER((_brt), (_new), brt_hash)
215 #define BRIDGE_RTHASH_WRITER_REMOVE(_brt) \
216 PSLIST_WRITER_REMOVE((_brt), brt_hash)
217
218 #ifdef NET_MPSAFE
219 #define DECLARE_LOCK_VARIABLE
220 #define ACQUIRE_GLOBAL_LOCKS() do { } while (0)
221 #define RELEASE_GLOBAL_LOCKS() do { } while (0)
222 #else
223 #define DECLARE_LOCK_VARIABLE int __s
224 #define ACQUIRE_GLOBAL_LOCKS() do { \
225 KERNEL_LOCK(1, NULL); \
226 mutex_enter(softnet_lock); \
227 __s = splsoftnet(); \
228 } while (0)
229 #define RELEASE_GLOBAL_LOCKS() do { \
230 splx(__s); \
231 mutex_exit(softnet_lock); \
232 KERNEL_UNLOCK_ONE(NULL); \
233 } while (0)
234 #endif
235
236 struct psref_class *bridge_psref_class __read_mostly;
237
238 int bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
239
240 static struct pool bridge_rtnode_pool;
241
242 static int bridge_clone_create(struct if_clone *, int);
243 static int bridge_clone_destroy(struct ifnet *);
244
245 static int bridge_ioctl(struct ifnet *, u_long, void *);
246 static int bridge_init(struct ifnet *);
247 static void bridge_stop(struct ifnet *, int);
248 static void bridge_start(struct ifnet *);
249
250 static void bridge_input(struct ifnet *, struct mbuf *);
251 static void bridge_forward(struct bridge_softc *, struct mbuf *);
252
253 static void bridge_timer(void *);
254
255 static void bridge_broadcast(struct bridge_softc *, struct ifnet *,
256 struct mbuf *);
257
258 static int bridge_rtupdate(struct bridge_softc *, const uint8_t *,
259 struct ifnet *, int, uint8_t);
260 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *);
261 static void bridge_rttrim(struct bridge_softc *);
262 static void bridge_rtage(struct bridge_softc *);
263 static void bridge_rtage_work(struct work *, void *);
264 static void bridge_rtflush(struct bridge_softc *, int);
265 static int bridge_rtdaddr(struct bridge_softc *, const uint8_t *);
266 static void bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp);
267
268 static void bridge_rtable_init(struct bridge_softc *);
269 static void bridge_rtable_fini(struct bridge_softc *);
270
271 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
272 const uint8_t *);
273 static int bridge_rtnode_insert(struct bridge_softc *,
274 struct bridge_rtnode *);
275 static void bridge_rtnode_remove(struct bridge_softc *,
276 struct bridge_rtnode *);
277 static void bridge_rtnode_destroy(struct bridge_rtnode *);
278
279 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
280 const char *name,
281 struct psref *);
282 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
283 struct ifnet *ifp,
284 struct psref *);
285 static void bridge_release_member(struct bridge_softc *, struct bridge_iflist *,
286 struct psref *);
287 static void bridge_delete_member(struct bridge_softc *,
288 struct bridge_iflist *);
289 static void bridge_acquire_member(struct bridge_softc *sc,
290 struct bridge_iflist *,
291 struct psref *);
292
293 static int bridge_ioctl_add(struct bridge_softc *, void *);
294 static int bridge_ioctl_del(struct bridge_softc *, void *);
295 static int bridge_ioctl_gifflags(struct bridge_softc *, void *);
296 static int bridge_ioctl_sifflags(struct bridge_softc *, void *);
297 static int bridge_ioctl_scache(struct bridge_softc *, void *);
298 static int bridge_ioctl_gcache(struct bridge_softc *, void *);
299 static int bridge_ioctl_gifs(struct bridge_softc *, void *);
300 static int bridge_ioctl_rts(struct bridge_softc *, void *);
301 static int bridge_ioctl_saddr(struct bridge_softc *, void *);
302 static int bridge_ioctl_sto(struct bridge_softc *, void *);
303 static int bridge_ioctl_gto(struct bridge_softc *, void *);
304 static int bridge_ioctl_daddr(struct bridge_softc *, void *);
305 static int bridge_ioctl_flush(struct bridge_softc *, void *);
306 static int bridge_ioctl_gpri(struct bridge_softc *, void *);
307 static int bridge_ioctl_spri(struct bridge_softc *, void *);
308 static int bridge_ioctl_ght(struct bridge_softc *, void *);
309 static int bridge_ioctl_sht(struct bridge_softc *, void *);
310 static int bridge_ioctl_gfd(struct bridge_softc *, void *);
311 static int bridge_ioctl_sfd(struct bridge_softc *, void *);
312 static int bridge_ioctl_gma(struct bridge_softc *, void *);
313 static int bridge_ioctl_sma(struct bridge_softc *, void *);
314 static int bridge_ioctl_sifprio(struct bridge_softc *, void *);
315 static int bridge_ioctl_sifcost(struct bridge_softc *, void *);
316 #if defined(BRIDGE_IPF)
317 static int bridge_ioctl_gfilt(struct bridge_softc *, void *);
318 static int bridge_ioctl_sfilt(struct bridge_softc *, void *);
319 static int bridge_ipf(void *, struct mbuf **, struct ifnet *, int);
320 static int bridge_ip_checkbasic(struct mbuf **mp);
321 # ifdef INET6
322 static int bridge_ip6_checkbasic(struct mbuf **mp);
323 # endif /* INET6 */
324 #endif /* BRIDGE_IPF */
325
326 struct bridge_control {
327 int (*bc_func)(struct bridge_softc *, void *);
328 int bc_argsize;
329 int bc_flags;
330 };
331
332 #define BC_F_COPYIN 0x01 /* copy arguments in */
333 #define BC_F_COPYOUT 0x02 /* copy arguments out */
334 #define BC_F_SUSER 0x04 /* do super-user check */
335 #define BC_F_XLATEIN 0x08 /* xlate arguments in */
336 #define BC_F_XLATEOUT 0x10 /* xlate arguments out */
337
338 static const struct bridge_control bridge_control_table[] = {
339 [BRDGADD] = {bridge_ioctl_add, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
340 [BRDGDEL] = {bridge_ioctl_del, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
341
342 [BRDGGIFFLGS] = {bridge_ioctl_gifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_COPYOUT},
343 [BRDGSIFFLGS] = {bridge_ioctl_sifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
344
345 [BRDGSCACHE] = {bridge_ioctl_scache, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
346 [BRDGGCACHE] = {bridge_ioctl_gcache, sizeof(struct ifbrparam), BC_F_COPYOUT},
347
348 [OBRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_COPYIN|BC_F_COPYOUT},
349 [OBRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_COPYIN|BC_F_COPYOUT},
350
351 [BRDGSADDR] = {bridge_ioctl_saddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
352
353 [BRDGSTO] = {bridge_ioctl_sto, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
354 [BRDGGTO] = {bridge_ioctl_gto, sizeof(struct ifbrparam), BC_F_COPYOUT},
355
356 [BRDGDADDR] = {bridge_ioctl_daddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
357
358 [BRDGFLUSH] = {bridge_ioctl_flush, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
359
360 [BRDGGPRI] = {bridge_ioctl_gpri, sizeof(struct ifbrparam), BC_F_COPYOUT},
361 [BRDGSPRI] = {bridge_ioctl_spri, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
362
363 [BRDGGHT] = {bridge_ioctl_ght, sizeof(struct ifbrparam), BC_F_COPYOUT},
364 [BRDGSHT] = {bridge_ioctl_sht, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
365
366 [BRDGGFD] = {bridge_ioctl_gfd, sizeof(struct ifbrparam), BC_F_COPYOUT},
367 [BRDGSFD] = {bridge_ioctl_sfd, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
368
369 [BRDGGMA] = {bridge_ioctl_gma, sizeof(struct ifbrparam), BC_F_COPYOUT},
370 [BRDGSMA] = {bridge_ioctl_sma, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
371
372 [BRDGSIFPRIO] = {bridge_ioctl_sifprio, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
373
374 [BRDGSIFCOST] = {bridge_ioctl_sifcost, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
375 #if defined(BRIDGE_IPF)
376 [BRDGGFILT] = {bridge_ioctl_gfilt, sizeof(struct ifbrparam), BC_F_COPYOUT},
377 [BRDGSFILT] = {bridge_ioctl_sfilt, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
378 #endif /* BRIDGE_IPF */
379 [BRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_XLATEIN|BC_F_XLATEOUT},
380 [BRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_XLATEIN|BC_F_XLATEOUT},
381 };
382
383 static const int bridge_control_table_size = __arraycount(bridge_control_table);
384
385 static struct if_clone bridge_cloner =
386 IF_CLONE_INITIALIZER("bridge", bridge_clone_create, bridge_clone_destroy);
387
388 /*
389 * bridgeattach:
390 *
391 * Pseudo-device attach routine.
392 */
393 void
394 bridgeattach(int n)
395 {
396
397 pool_init(&bridge_rtnode_pool, sizeof(struct bridge_rtnode),
398 0, 0, 0, "brtpl", NULL, IPL_NET);
399
400 bridge_psref_class = psref_class_create("bridge", IPL_SOFTNET);
401
402 if_clone_attach(&bridge_cloner);
403 }
404
405 /*
406 * bridge_clone_create:
407 *
408 * Create a new bridge instance.
409 */
410 static int
411 bridge_clone_create(struct if_clone *ifc, int unit)
412 {
413 struct bridge_softc *sc;
414 struct ifnet *ifp;
415 int error;
416
417 sc = kmem_zalloc(sizeof(*sc), KM_SLEEP);
418 ifp = &sc->sc_if;
419
420 sc->sc_brtmax = BRIDGE_RTABLE_MAX;
421 sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
422 sc->sc_bridge_max_age = BSTP_DEFAULT_MAX_AGE;
423 sc->sc_bridge_hello_time = BSTP_DEFAULT_HELLO_TIME;
424 sc->sc_bridge_forward_delay = BSTP_DEFAULT_FORWARD_DELAY;
425 sc->sc_bridge_priority = BSTP_DEFAULT_BRIDGE_PRIORITY;
426 sc->sc_hold_time = BSTP_DEFAULT_HOLD_TIME;
427 sc->sc_filter_flags = 0;
428
429 /* Initialize our routing table. */
430 bridge_rtable_init(sc);
431
432 error = workqueue_create(&sc->sc_rtage_wq, "bridge_rtage",
433 bridge_rtage_work, sc, PRI_SOFTNET, IPL_SOFTNET, WQ_MPSAFE);
434 if (error)
435 panic("%s: workqueue_create %d\n", __func__, error);
436
437 callout_init(&sc->sc_brcallout, CALLOUT_MPSAFE);
438 callout_init(&sc->sc_bstpcallout, CALLOUT_MPSAFE);
439
440 mutex_init(&sc->sc_iflist_psref.bip_lock, MUTEX_DEFAULT, IPL_NONE);
441 PSLIST_INIT(&sc->sc_iflist_psref.bip_iflist);
442 sc->sc_iflist_psref.bip_psz = pserialize_create();
443
444 if_initname(ifp, ifc->ifc_name, unit);
445 ifp->if_softc = sc;
446 ifp->if_extflags = IFEF_NO_LINK_STATE_CHANGE;
447 #ifdef NET_MPSAFE
448 ifp->if_extflags |= IFEF_MPSAFE;
449 #endif
450 ifp->if_mtu = ETHERMTU;
451 ifp->if_ioctl = bridge_ioctl;
452 ifp->if_output = bridge_output;
453 ifp->if_start = bridge_start;
454 ifp->if_stop = bridge_stop;
455 ifp->if_init = bridge_init;
456 ifp->if_type = IFT_BRIDGE;
457 ifp->if_addrlen = 0;
458 ifp->if_dlt = DLT_EN10MB;
459 ifp->if_hdrlen = ETHER_HDR_LEN;
460
461 error = if_initialize(ifp);
462 if (error != 0) {
463 pserialize_destroy(sc->sc_iflist_psref.bip_psz);
464 mutex_destroy(&sc->sc_iflist_psref.bip_lock);
465 callout_destroy(&sc->sc_brcallout);
466 callout_destroy(&sc->sc_bstpcallout);
467 workqueue_destroy(sc->sc_rtage_wq);
468 bridge_rtable_fini(sc);
469 kmem_free(sc, sizeof(*sc));
470
471 return error;
472 }
473 if_alloc_sadl(ifp);
474 if_register(ifp);
475
476 return 0;
477 }
478
479 /*
480 * bridge_clone_destroy:
481 *
482 * Destroy a bridge instance.
483 */
484 static int
485 bridge_clone_destroy(struct ifnet *ifp)
486 {
487 struct bridge_softc *sc = ifp->if_softc;
488 struct bridge_iflist *bif;
489
490 if ((ifp->if_flags & IFF_RUNNING) != 0)
491 bridge_stop(ifp, 1);
492
493 BRIDGE_LOCK(sc);
494 for (;;) {
495 bif = PSLIST_WRITER_FIRST(&sc->sc_iflist_psref.bip_iflist, struct bridge_iflist,
496 bif_next);
497 if (bif == NULL)
498 break;
499 bridge_delete_member(sc, bif);
500 }
501 PSLIST_DESTROY(&sc->sc_iflist_psref.bip_iflist);
502 BRIDGE_UNLOCK(sc);
503
504 if_detach(ifp);
505
506 /* Tear down the routing table. */
507 bridge_rtable_fini(sc);
508
509 pserialize_destroy(sc->sc_iflist_psref.bip_psz);
510 mutex_destroy(&sc->sc_iflist_psref.bip_lock);
511 callout_destroy(&sc->sc_brcallout);
512 callout_destroy(&sc->sc_bstpcallout);
513 workqueue_destroy(sc->sc_rtage_wq);
514 kmem_free(sc, sizeof(*sc));
515
516 return 0;
517 }
518
519 /*
520 * bridge_ioctl:
521 *
522 * Handle a control request from the operator.
523 */
524 static int
525 bridge_ioctl(struct ifnet *ifp, u_long cmd, void *data)
526 {
527 struct bridge_softc *sc = ifp->if_softc;
528 struct lwp *l = curlwp; /* XXX */
529 union {
530 struct ifbreq ifbreq;
531 struct ifbifconf ifbifconf;
532 struct ifbareq ifbareq;
533 struct ifbaconf ifbaconf;
534 struct ifbrparam ifbrparam;
535 } args;
536 struct ifdrv *ifd = (struct ifdrv *) data;
537 const struct bridge_control *bc = NULL; /* XXXGCC */
538 int s, error = 0;
539
540 /* Authorize command before calling splsoftnet(). */
541 switch (cmd) {
542 case SIOCGDRVSPEC:
543 case SIOCSDRVSPEC:
544 if (ifd->ifd_cmd >= bridge_control_table_size
545 || (bc = &bridge_control_table[ifd->ifd_cmd]) == NULL) {
546 error = EINVAL;
547 return error;
548 }
549
550 /* We only care about BC_F_SUSER at this point. */
551 if ((bc->bc_flags & BC_F_SUSER) == 0)
552 break;
553
554 error = kauth_authorize_network(l->l_cred,
555 KAUTH_NETWORK_INTERFACE_BRIDGE,
556 cmd == SIOCGDRVSPEC ?
557 KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_GETPRIV :
558 KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_SETPRIV,
559 ifd, NULL, NULL);
560 if (error)
561 return error;
562
563 break;
564 }
565
566 s = splsoftnet();
567
568 switch (cmd) {
569 case SIOCGDRVSPEC:
570 case SIOCSDRVSPEC:
571 KASSERT(bc != NULL);
572 if (cmd == SIOCGDRVSPEC &&
573 (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) == 0) {
574 error = EINVAL;
575 break;
576 }
577 else if (cmd == SIOCSDRVSPEC &&
578 (bc->bc_flags & (BC_F_COPYOUT|BC_F_XLATEOUT)) != 0) {
579 error = EINVAL;
580 break;
581 }
582
583 /* BC_F_SUSER is checked above, before splsoftnet(). */
584
585 if ((bc->bc_flags & (BC_F_XLATEIN|BC_F_XLATEOUT)) == 0
586 && (ifd->ifd_len != bc->bc_argsize
587 || ifd->ifd_len > sizeof(args))) {
588 error = EINVAL;
589 break;
590 }
591
592 memset(&args, 0, sizeof(args));
593 if (bc->bc_flags & BC_F_COPYIN) {
594 error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
595 if (error)
596 break;
597 } else if (bc->bc_flags & BC_F_XLATEIN) {
598 args.ifbifconf.ifbic_len = ifd->ifd_len;
599 args.ifbifconf.ifbic_buf = ifd->ifd_data;
600 }
601
602 error = (*bc->bc_func)(sc, &args);
603 if (error)
604 break;
605
606 if (bc->bc_flags & BC_F_COPYOUT) {
607 error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
608 } else if (bc->bc_flags & BC_F_XLATEOUT) {
609 ifd->ifd_len = args.ifbifconf.ifbic_len;
610 ifd->ifd_data = args.ifbifconf.ifbic_buf;
611 }
612 break;
613
614 case SIOCSIFFLAGS:
615 if ((error = ifioctl_common(ifp, cmd, data)) != 0)
616 break;
617 switch (ifp->if_flags & (IFF_UP|IFF_RUNNING)) {
618 case IFF_RUNNING:
619 /*
620 * If interface is marked down and it is running,
621 * then stop and disable it.
622 */
623 (*ifp->if_stop)(ifp, 1);
624 break;
625 case IFF_UP:
626 /*
627 * If interface is marked up and it is stopped, then
628 * start it.
629 */
630 error = (*ifp->if_init)(ifp);
631 break;
632 default:
633 break;
634 }
635 break;
636
637 case SIOCSIFMTU:
638 if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET)
639 error = 0;
640 break;
641
642 default:
643 error = ifioctl_common(ifp, cmd, data);
644 break;
645 }
646
647 splx(s);
648
649 return error;
650 }
651
652 /*
653 * bridge_lookup_member:
654 *
655 * Lookup a bridge member interface.
656 */
657 static struct bridge_iflist *
658 bridge_lookup_member(struct bridge_softc *sc, const char *name, struct psref *psref)
659 {
660 struct bridge_iflist *bif;
661 struct ifnet *ifp;
662 int s;
663
664 BRIDGE_PSZ_RENTER(s);
665
666 BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
667 ifp = bif->bif_ifp;
668 if (strcmp(ifp->if_xname, name) == 0)
669 break;
670 }
671 if (bif != NULL)
672 bridge_acquire_member(sc, bif, psref);
673
674 BRIDGE_PSZ_REXIT(s);
675
676 return bif;
677 }
678
679 /*
680 * bridge_lookup_member_if:
681 *
682 * Lookup a bridge member interface by ifnet*.
683 */
684 static struct bridge_iflist *
685 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp,
686 struct psref *psref)
687 {
688 struct bridge_iflist *bif;
689 int s;
690
691 BRIDGE_PSZ_RENTER(s);
692
693 bif = member_ifp->if_bridgeif;
694 if (bif != NULL) {
695 psref_acquire(psref, &bif->bif_psref,
696 bridge_psref_class);
697 }
698
699 BRIDGE_PSZ_REXIT(s);
700
701 return bif;
702 }
703
704 static void
705 bridge_acquire_member(struct bridge_softc *sc, struct bridge_iflist *bif,
706 struct psref *psref)
707 {
708
709 psref_acquire(psref, &bif->bif_psref, bridge_psref_class);
710 }
711
712 /*
713 * bridge_release_member:
714 *
715 * Release the specified member interface.
716 */
717 static void
718 bridge_release_member(struct bridge_softc *sc, struct bridge_iflist *bif,
719 struct psref *psref)
720 {
721
722 psref_release(psref, &bif->bif_psref, bridge_psref_class);
723 }
724
725 /*
726 * bridge_delete_member:
727 *
728 * Delete the specified member interface.
729 */
730 static void
731 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif)
732 {
733 struct ifnet *ifs = bif->bif_ifp;
734
735 KASSERT(BRIDGE_LOCKED(sc));
736
737 ifs->_if_input = ether_input;
738 ifs->if_bridge = NULL;
739 ifs->if_bridgeif = NULL;
740
741 PSLIST_WRITER_REMOVE(bif, bif_next);
742 BRIDGE_PSZ_PERFORM(sc);
743 BRIDGE_UNLOCK(sc);
744
745 psref_target_destroy(&bif->bif_psref, bridge_psref_class);
746
747 PSLIST_ENTRY_DESTROY(bif, bif_next);
748 kmem_free(bif, sizeof(*bif));
749
750 BRIDGE_LOCK(sc);
751 }
752
753 static int
754 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
755 {
756 struct ifbreq *req = arg;
757 struct bridge_iflist *bif = NULL;
758 struct ifnet *ifs;
759 int error = 0;
760 struct psref psref;
761
762 ifs = if_get(req->ifbr_ifsname, &psref);
763 if (ifs == NULL)
764 return ENOENT;
765
766 if (ifs->if_bridge == sc) {
767 error = EEXIST;
768 goto out;
769 }
770
771 if (ifs->if_bridge != NULL) {
772 error = EBUSY;
773 goto out;
774 }
775
776 if (ifs->_if_input != ether_input) {
777 error = EINVAL;
778 goto out;
779 }
780
781 /* FIXME: doesn't work with non-IFF_SIMPLEX interfaces */
782 if ((ifs->if_flags & IFF_SIMPLEX) == 0) {
783 error = EINVAL;
784 goto out;
785 }
786
787 bif = kmem_alloc(sizeof(*bif), KM_SLEEP);
788
789 switch (ifs->if_type) {
790 case IFT_ETHER:
791 if (sc->sc_if.if_mtu != ifs->if_mtu) {
792 error = EINVAL;
793 goto out;
794 }
795 /* FALLTHROUGH */
796 case IFT_L2TP:
797 IFNET_LOCK(ifs);
798 error = ether_enable_vlan_mtu(ifs);
799 IFNET_UNLOCK(ifs);
800 if (error > 0)
801 goto out;
802 /*
803 * Place the interface into promiscuous mode.
804 */
805 error = ifpromisc(ifs, 1);
806 if (error)
807 goto out;
808 break;
809 default:
810 error = EINVAL;
811 goto out;
812 }
813
814 bif->bif_ifp = ifs;
815 bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
816 bif->bif_priority = BSTP_DEFAULT_PORT_PRIORITY;
817 bif->bif_path_cost = BSTP_DEFAULT_PATH_COST;
818 PSLIST_ENTRY_INIT(bif, bif_next);
819 psref_target_init(&bif->bif_psref, bridge_psref_class);
820
821 BRIDGE_LOCK(sc);
822
823 ifs->if_bridge = sc;
824 ifs->if_bridgeif = bif;
825 PSLIST_WRITER_INSERT_HEAD(&sc->sc_iflist_psref.bip_iflist, bif, bif_next);
826 ifs->_if_input = bridge_input;
827
828 BRIDGE_UNLOCK(sc);
829
830 if (sc->sc_if.if_flags & IFF_RUNNING)
831 bstp_initialization(sc);
832 else
833 bstp_stop(sc);
834
835 out:
836 if_put(ifs, &psref);
837 if (error) {
838 if (bif != NULL)
839 kmem_free(bif, sizeof(*bif));
840 }
841 return error;
842 }
843
844 static int
845 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
846 {
847 struct ifbreq *req = arg;
848 const char *name = req->ifbr_ifsname;
849 struct bridge_iflist *bif;
850 struct ifnet *ifs;
851
852 BRIDGE_LOCK(sc);
853
854 /*
855 * Don't use bridge_lookup_member. We want to get a member
856 * with bif_refs == 0.
857 */
858 BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) {
859 ifs = bif->bif_ifp;
860 if (strcmp(ifs->if_xname, name) == 0)
861 break;
862 }
863
864 if (bif == NULL) {
865 BRIDGE_UNLOCK(sc);
866 return ENOENT;
867 }
868
869 bridge_delete_member(sc, bif);
870
871 BRIDGE_UNLOCK(sc);
872
873 switch (ifs->if_type) {
874 case IFT_ETHER:
875 case IFT_L2TP:
876 /*
877 * Take the interface out of promiscuous mode.
878 * Don't call it with holding a spin lock.
879 */
880 (void) ifpromisc(ifs, 0);
881 IFNET_LOCK(ifs);
882 (void) ether_disable_vlan_mtu(ifs);
883 IFNET_UNLOCK(ifs);
884 break;
885 default:
886 #ifdef DIAGNOSTIC
887 panic("bridge_delete_member: impossible");
888 #endif
889 break;
890 }
891
892 bridge_rtdelete(sc, ifs);
893
894 if (sc->sc_if.if_flags & IFF_RUNNING)
895 bstp_initialization(sc);
896
897 return 0;
898 }
899
900 static int
901 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
902 {
903 struct ifbreq *req = arg;
904 struct bridge_iflist *bif;
905 struct psref psref;
906
907 bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
908 if (bif == NULL)
909 return ENOENT;
910
911 req->ifbr_ifsflags = bif->bif_flags;
912 req->ifbr_state = bif->bif_state;
913 req->ifbr_priority = bif->bif_priority;
914 req->ifbr_path_cost = bif->bif_path_cost;
915 req->ifbr_portno = bif->bif_ifp->if_index & 0xff;
916
917 bridge_release_member(sc, bif, &psref);
918
919 return 0;
920 }
921
922 static int
923 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
924 {
925 struct ifbreq *req = arg;
926 struct bridge_iflist *bif;
927 struct psref psref;
928
929 bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
930 if (bif == NULL)
931 return ENOENT;
932
933 if (req->ifbr_ifsflags & IFBIF_STP) {
934 switch (bif->bif_ifp->if_type) {
935 case IFT_ETHER:
936 case IFT_L2TP:
937 /* These can do spanning tree. */
938 break;
939
940 default:
941 /* Nothing else can. */
942 bridge_release_member(sc, bif, &psref);
943 return EINVAL;
944 }
945 }
946
947 bif->bif_flags = req->ifbr_ifsflags;
948
949 bridge_release_member(sc, bif, &psref);
950
951 if (sc->sc_if.if_flags & IFF_RUNNING)
952 bstp_initialization(sc);
953
954 return 0;
955 }
956
957 static int
958 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
959 {
960 struct ifbrparam *param = arg;
961
962 sc->sc_brtmax = param->ifbrp_csize;
963 bridge_rttrim(sc);
964
965 return 0;
966 }
967
968 static int
969 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
970 {
971 struct ifbrparam *param = arg;
972
973 param->ifbrp_csize = sc->sc_brtmax;
974
975 return 0;
976 }
977
978 static int
979 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
980 {
981 struct ifbifconf *bifc = arg;
982 struct bridge_iflist *bif;
983 struct ifbreq *breqs;
984 int i, count, error = 0;
985
986 retry:
987 BRIDGE_LOCK(sc);
988 count = 0;
989 BRIDGE_IFLIST_WRITER_FOREACH(bif, sc)
990 count++;
991 BRIDGE_UNLOCK(sc);
992
993 if (count == 0) {
994 bifc->ifbic_len = 0;
995 return 0;
996 }
997
998 if (bifc->ifbic_len == 0 || bifc->ifbic_len < (sizeof(*breqs) * count)) {
999 /* Tell that a larger buffer is needed */
1000 bifc->ifbic_len = sizeof(*breqs) * count;
1001 return 0;
1002 }
1003
1004 breqs = kmem_alloc(sizeof(*breqs) * count, KM_SLEEP);
1005
1006 BRIDGE_LOCK(sc);
1007
1008 i = 0;
1009 BRIDGE_IFLIST_WRITER_FOREACH(bif, sc)
1010 i++;
1011 if (i > count) {
1012 /*
1013 * The number of members has been increased.
1014 * We need more memory!
1015 */
1016 BRIDGE_UNLOCK(sc);
1017 kmem_free(breqs, sizeof(*breqs) * count);
1018 goto retry;
1019 }
1020
1021 i = 0;
1022 BRIDGE_IFLIST_WRITER_FOREACH(bif, sc) {
1023 struct ifbreq *breq = &breqs[i++];
1024 memset(breq, 0, sizeof(*breq));
1025
1026 strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname,
1027 sizeof(breq->ifbr_ifsname));
1028 breq->ifbr_ifsflags = bif->bif_flags;
1029 breq->ifbr_state = bif->bif_state;
1030 breq->ifbr_priority = bif->bif_priority;
1031 breq->ifbr_path_cost = bif->bif_path_cost;
1032 breq->ifbr_portno = bif->bif_ifp->if_index & 0xff;
1033 }
1034
1035 /* Don't call copyout with holding the mutex */
1036 BRIDGE_UNLOCK(sc);
1037
1038 for (i = 0; i < count; i++) {
1039 error = copyout(&breqs[i], bifc->ifbic_req + i, sizeof(*breqs));
1040 if (error)
1041 break;
1042 }
1043 bifc->ifbic_len = sizeof(*breqs) * i;
1044
1045 kmem_free(breqs, sizeof(*breqs) * count);
1046
1047 return error;
1048 }
1049
1050 static int
1051 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1052 {
1053 struct ifbaconf *bac = arg;
1054 struct bridge_rtnode *brt;
1055 struct ifbareq bareq;
1056 int count = 0, error = 0, len;
1057
1058 if (bac->ifbac_len == 0)
1059 return 0;
1060
1061 BRIDGE_RT_LOCK(sc);
1062
1063 len = bac->ifbac_len;
1064 BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) {
1065 if (len < sizeof(bareq))
1066 goto out;
1067 memset(&bareq, 0, sizeof(bareq));
1068 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1069 sizeof(bareq.ifba_ifsname));
1070 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1071 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
1072 bareq.ifba_expire = brt->brt_expire - time_uptime;
1073 } else
1074 bareq.ifba_expire = 0;
1075 bareq.ifba_flags = brt->brt_flags;
1076
1077 error = copyout(&bareq, bac->ifbac_req + count, sizeof(bareq));
1078 if (error)
1079 goto out;
1080 count++;
1081 len -= sizeof(bareq);
1082 }
1083 out:
1084 BRIDGE_RT_UNLOCK(sc);
1085
1086 bac->ifbac_len = sizeof(bareq) * count;
1087 return error;
1088 }
1089
1090 static int
1091 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1092 {
1093 struct ifbareq *req = arg;
1094 struct bridge_iflist *bif;
1095 int error;
1096 struct psref psref;
1097
1098 bif = bridge_lookup_member(sc, req->ifba_ifsname, &psref);
1099 if (bif == NULL)
1100 return ENOENT;
1101
1102 error = bridge_rtupdate(sc, req->ifba_dst, bif->bif_ifp, 1,
1103 req->ifba_flags);
1104
1105 bridge_release_member(sc, bif, &psref);
1106
1107 return error;
1108 }
1109
1110 static int
1111 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1112 {
1113 struct ifbrparam *param = arg;
1114
1115 sc->sc_brttimeout = param->ifbrp_ctime;
1116
1117 return 0;
1118 }
1119
1120 static int
1121 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1122 {
1123 struct ifbrparam *param = arg;
1124
1125 param->ifbrp_ctime = sc->sc_brttimeout;
1126
1127 return 0;
1128 }
1129
1130 static int
1131 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1132 {
1133 struct ifbareq *req = arg;
1134
1135 return (bridge_rtdaddr(sc, req->ifba_dst));
1136 }
1137
1138 static int
1139 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1140 {
1141 struct ifbreq *req = arg;
1142
1143 bridge_rtflush(sc, req->ifbr_ifsflags);
1144
1145 return 0;
1146 }
1147
1148 static int
1149 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1150 {
1151 struct ifbrparam *param = arg;
1152
1153 param->ifbrp_prio = sc->sc_bridge_priority;
1154
1155 return 0;
1156 }
1157
1158 static int
1159 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1160 {
1161 struct ifbrparam *param = arg;
1162
1163 sc->sc_bridge_priority = param->ifbrp_prio;
1164
1165 if (sc->sc_if.if_flags & IFF_RUNNING)
1166 bstp_initialization(sc);
1167
1168 return 0;
1169 }
1170
1171 static int
1172 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1173 {
1174 struct ifbrparam *param = arg;
1175
1176 param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8;
1177
1178 return 0;
1179 }
1180
1181 static int
1182 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1183 {
1184 struct ifbrparam *param = arg;
1185
1186 if (param->ifbrp_hellotime == 0)
1187 return EINVAL;
1188 sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8;
1189
1190 if (sc->sc_if.if_flags & IFF_RUNNING)
1191 bstp_initialization(sc);
1192
1193 return 0;
1194 }
1195
1196 static int
1197 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1198 {
1199 struct ifbrparam *param = arg;
1200
1201 param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8;
1202
1203 return 0;
1204 }
1205
1206 static int
1207 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1208 {
1209 struct ifbrparam *param = arg;
1210
1211 if (param->ifbrp_fwddelay == 0)
1212 return EINVAL;
1213 sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8;
1214
1215 if (sc->sc_if.if_flags & IFF_RUNNING)
1216 bstp_initialization(sc);
1217
1218 return 0;
1219 }
1220
1221 static int
1222 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1223 {
1224 struct ifbrparam *param = arg;
1225
1226 param->ifbrp_maxage = sc->sc_bridge_max_age >> 8;
1227
1228 return 0;
1229 }
1230
1231 static int
1232 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1233 {
1234 struct ifbrparam *param = arg;
1235
1236 if (param->ifbrp_maxage == 0)
1237 return EINVAL;
1238 sc->sc_bridge_max_age = param->ifbrp_maxage << 8;
1239
1240 if (sc->sc_if.if_flags & IFF_RUNNING)
1241 bstp_initialization(sc);
1242
1243 return 0;
1244 }
1245
1246 static int
1247 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1248 {
1249 struct ifbreq *req = arg;
1250 struct bridge_iflist *bif;
1251 struct psref psref;
1252
1253 bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
1254 if (bif == NULL)
1255 return ENOENT;
1256
1257 bif->bif_priority = req->ifbr_priority;
1258
1259 if (sc->sc_if.if_flags & IFF_RUNNING)
1260 bstp_initialization(sc);
1261
1262 bridge_release_member(sc, bif, &psref);
1263
1264 return 0;
1265 }
1266
1267 #if defined(BRIDGE_IPF)
1268 static int
1269 bridge_ioctl_gfilt(struct bridge_softc *sc, void *arg)
1270 {
1271 struct ifbrparam *param = arg;
1272
1273 param->ifbrp_filter = sc->sc_filter_flags;
1274
1275 return 0;
1276 }
1277
1278 static int
1279 bridge_ioctl_sfilt(struct bridge_softc *sc, void *arg)
1280 {
1281 struct ifbrparam *param = arg;
1282 uint32_t nflags, oflags;
1283
1284 if (param->ifbrp_filter & ~IFBF_FILT_MASK)
1285 return EINVAL;
1286
1287 nflags = param->ifbrp_filter;
1288 oflags = sc->sc_filter_flags;
1289
1290 if ((nflags & IFBF_FILT_USEIPF) && !(oflags & IFBF_FILT_USEIPF)) {
1291 pfil_add_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
1292 sc->sc_if.if_pfil);
1293 }
1294 if (!(nflags & IFBF_FILT_USEIPF) && (oflags & IFBF_FILT_USEIPF)) {
1295 pfil_remove_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
1296 sc->sc_if.if_pfil);
1297 }
1298
1299 sc->sc_filter_flags = nflags;
1300
1301 return 0;
1302 }
1303 #endif /* BRIDGE_IPF */
1304
1305 static int
1306 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1307 {
1308 struct ifbreq *req = arg;
1309 struct bridge_iflist *bif;
1310 struct psref psref;
1311
1312 bif = bridge_lookup_member(sc, req->ifbr_ifsname, &psref);
1313 if (bif == NULL)
1314 return ENOENT;
1315
1316 bif->bif_path_cost = req->ifbr_path_cost;
1317
1318 if (sc->sc_if.if_flags & IFF_RUNNING)
1319 bstp_initialization(sc);
1320
1321 bridge_release_member(sc, bif, &psref);
1322
1323 return 0;
1324 }
1325
1326 /*
1327 * bridge_ifdetach:
1328 *
1329 * Detach an interface from a bridge. Called when a member
1330 * interface is detaching.
1331 */
1332 void
1333 bridge_ifdetach(struct ifnet *ifp)
1334 {
1335 struct bridge_softc *sc = ifp->if_bridge;
1336 struct ifbreq breq;
1337
1338 /* ioctl_lock should prevent this from happening */
1339 KASSERT(sc != NULL);
1340
1341 memset(&breq, 0, sizeof(breq));
1342 strlcpy(breq.ifbr_ifsname, ifp->if_xname, sizeof(breq.ifbr_ifsname));
1343
1344 (void) bridge_ioctl_del(sc, &breq);
1345 }
1346
1347 /*
1348 * bridge_init:
1349 *
1350 * Initialize a bridge interface.
1351 */
1352 static int
1353 bridge_init(struct ifnet *ifp)
1354 {
1355 struct bridge_softc *sc = ifp->if_softc;
1356
1357 KASSERT((ifp->if_flags & IFF_RUNNING) == 0);
1358
1359 callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1360 bridge_timer, sc);
1361 bstp_initialization(sc);
1362
1363 ifp->if_flags |= IFF_RUNNING;
1364 return 0;
1365 }
1366
1367 /*
1368 * bridge_stop:
1369 *
1370 * Stop the bridge interface.
1371 */
1372 static void
1373 bridge_stop(struct ifnet *ifp, int disable)
1374 {
1375 struct bridge_softc *sc = ifp->if_softc;
1376
1377 KASSERT((ifp->if_flags & IFF_RUNNING) != 0);
1378 ifp->if_flags &= ~IFF_RUNNING;
1379
1380 callout_halt(&sc->sc_brcallout, NULL);
1381 workqueue_wait(sc->sc_rtage_wq, &sc->sc_rtage_wk);
1382 bstp_stop(sc);
1383 bridge_rtflush(sc, IFBF_FLUSHDYN);
1384 }
1385
1386 /*
1387 * bridge_enqueue:
1388 *
1389 * Enqueue a packet on a bridge member interface.
1390 */
1391 void
1392 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m,
1393 int runfilt)
1394 {
1395 int len, error;
1396 short mflags;
1397
1398 if (runfilt) {
1399 if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
1400 dst_ifp, PFIL_OUT) != 0) {
1401 if (m != NULL)
1402 m_freem(m);
1403 return;
1404 }
1405 if (m == NULL)
1406 return;
1407 }
1408
1409 #ifdef ALTQ
1410 KERNEL_LOCK(1, NULL);
1411 /*
1412 * If ALTQ is enabled on the member interface, do
1413 * classification; the queueing discipline might
1414 * not require classification, but might require
1415 * the address family/header pointer in the pktattr.
1416 */
1417 if (ALTQ_IS_ENABLED(&dst_ifp->if_snd)) {
1418 /* XXX IFT_ETHER */
1419 altq_etherclassify(&dst_ifp->if_snd, m);
1420 }
1421 KERNEL_UNLOCK_ONE(NULL);
1422 #endif /* ALTQ */
1423
1424 len = m->m_pkthdr.len;
1425 mflags = m->m_flags;
1426
1427 error = if_transmit_lock(dst_ifp, m);
1428 if (error) {
1429 /* mbuf is already freed */
1430 sc->sc_if.if_oerrors++;
1431 return;
1432 }
1433
1434 sc->sc_if.if_opackets++;
1435 sc->sc_if.if_obytes += len;
1436 if (mflags & M_MCAST)
1437 sc->sc_if.if_omcasts++;
1438 }
1439
1440 /*
1441 * bridge_output:
1442 *
1443 * Send output from a bridge member interface. This
1444 * performs the bridging function for locally originated
1445 * packets.
1446 *
1447 * The mbuf has the Ethernet header already attached. We must
1448 * enqueue or free the mbuf before returning.
1449 */
1450 int
1451 bridge_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa,
1452 const struct rtentry *rt)
1453 {
1454 struct ether_header *eh;
1455 struct ifnet *dst_if;
1456 struct bridge_softc *sc;
1457 int s;
1458
1459 /*
1460 * bridge_output() is called from ether_output(), furthermore
1461 * ifp argument doesn't point to bridge(4). So, don't assert
1462 * IFEF_MPSAFE here.
1463 */
1464
1465 if (m->m_len < ETHER_HDR_LEN) {
1466 m = m_pullup(m, ETHER_HDR_LEN);
1467 if (m == NULL)
1468 return 0;
1469 }
1470
1471 eh = mtod(m, struct ether_header *);
1472 sc = ifp->if_bridge;
1473
1474 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1475 if (memcmp(etherbroadcastaddr,
1476 eh->ether_dhost, ETHER_ADDR_LEN) == 0)
1477 m->m_flags |= M_BCAST;
1478 else
1479 m->m_flags |= M_MCAST;
1480 }
1481
1482 /*
1483 * If bridge is down, but the original output interface is up,
1484 * go ahead and send out that interface. Otherwise, the packet
1485 * is dropped below.
1486 */
1487 if (__predict_false(sc == NULL) ||
1488 (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
1489 dst_if = ifp;
1490 goto sendunicast;
1491 }
1492
1493 /*
1494 * If the packet is a multicast, or we don't know a better way to
1495 * get there, send to all interfaces.
1496 */
1497 if ((m->m_flags & (M_MCAST | M_BCAST)) != 0)
1498 dst_if = NULL;
1499 else
1500 dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1501 if (dst_if == NULL) {
1502 /* XXX Should call bridge_broadcast, but there are locking
1503 * issues which need resolving first. */
1504 struct bridge_iflist *bif;
1505 struct mbuf *mc;
1506 bool used = false;
1507
1508 BRIDGE_PSZ_RENTER(s);
1509 BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
1510 struct psref psref;
1511
1512 bridge_acquire_member(sc, bif, &psref);
1513 BRIDGE_PSZ_REXIT(s);
1514
1515 dst_if = bif->bif_ifp;
1516 if ((dst_if->if_flags & IFF_RUNNING) == 0)
1517 goto next;
1518
1519 /*
1520 * If this is not the original output interface,
1521 * and the interface is participating in spanning
1522 * tree, make sure the port is in a state that
1523 * allows forwarding.
1524 */
1525 if (dst_if != ifp &&
1526 (bif->bif_flags & IFBIF_STP) != 0) {
1527 switch (bif->bif_state) {
1528 case BSTP_IFSTATE_BLOCKING:
1529 case BSTP_IFSTATE_LISTENING:
1530 case BSTP_IFSTATE_DISABLED:
1531 goto next;
1532 }
1533 }
1534
1535 if (PSLIST_READER_NEXT(bif, struct bridge_iflist,
1536 bif_next) == NULL &&
1537 ((m->m_flags & (M_MCAST | M_BCAST)) == 0 ||
1538 dst_if == ifp))
1539 {
1540 used = true;
1541 mc = m;
1542 } else {
1543 mc = m_copypacket(m, M_DONTWAIT);
1544 if (mc == NULL) {
1545 sc->sc_if.if_oerrors++;
1546 goto next;
1547 }
1548 }
1549
1550 bridge_enqueue(sc, dst_if, mc, 0);
1551
1552 if ((m->m_flags & (M_MCAST | M_BCAST)) != 0 &&
1553 dst_if != ifp)
1554 {
1555 if (PSLIST_READER_NEXT(bif,
1556 struct bridge_iflist, bif_next) == NULL)
1557 {
1558 used = true;
1559 mc = m;
1560 } else {
1561 mc = m_copypacket(m, M_DONTWAIT);
1562 if (mc == NULL) {
1563 sc->sc_if.if_oerrors++;
1564 goto next;
1565 }
1566 }
1567
1568 m_set_rcvif(mc, dst_if);
1569 mc->m_flags &= ~M_PROMISC;
1570
1571 s = splsoftnet();
1572 KERNEL_LOCK_UNLESS_IFP_MPSAFE(dst_if);
1573 ether_input(dst_if, mc);
1574 KERNEL_UNLOCK_UNLESS_IFP_MPSAFE(dst_if);
1575 splx(s);
1576 }
1577
1578 next:
1579 BRIDGE_PSZ_RENTER(s);
1580 bridge_release_member(sc, bif, &psref);
1581
1582 /* Guarantee we don't re-enter the loop as we already
1583 * decided we're at the end. */
1584 if (used)
1585 break;
1586 }
1587 BRIDGE_PSZ_REXIT(s);
1588
1589 if (!used)
1590 m_freem(m);
1591 return 0;
1592 }
1593
1594 sendunicast:
1595 /*
1596 * XXX Spanning tree consideration here?
1597 */
1598
1599 if ((dst_if->if_flags & IFF_RUNNING) == 0) {
1600 m_freem(m);
1601 return 0;
1602 }
1603
1604 bridge_enqueue(sc, dst_if, m, 0);
1605
1606 return 0;
1607 }
1608
1609 /*
1610 * bridge_start:
1611 *
1612 * Start output on a bridge.
1613 *
1614 * NOTE: This routine should never be called in this implementation.
1615 */
1616 static void
1617 bridge_start(struct ifnet *ifp)
1618 {
1619
1620 printf("%s: bridge_start() called\n", ifp->if_xname);
1621 }
1622
1623 /*
1624 * bridge_forward:
1625 *
1626 * The forwarding function of the bridge.
1627 */
1628 static void
1629 bridge_forward(struct bridge_softc *sc, struct mbuf *m)
1630 {
1631 struct bridge_iflist *bif;
1632 struct ifnet *src_if, *dst_if;
1633 struct ether_header *eh;
1634 struct psref psref;
1635 struct psref psref_src;
1636 DECLARE_LOCK_VARIABLE;
1637
1638 if ((sc->sc_if.if_flags & IFF_RUNNING) == 0)
1639 return;
1640
1641 src_if = m_get_rcvif_psref(m, &psref_src);
1642 if (src_if == NULL) {
1643 /* Interface is being destroyed? */
1644 m_freem(m);
1645 goto out;
1646 }
1647
1648 sc->sc_if.if_ipackets++;
1649 sc->sc_if.if_ibytes += m->m_pkthdr.len;
1650
1651 /*
1652 * Look up the bridge_iflist.
1653 */
1654 bif = bridge_lookup_member_if(sc, src_if, &psref);
1655 if (bif == NULL) {
1656 /* Interface is not a bridge member (anymore?) */
1657 m_freem(m);
1658 goto out;
1659 }
1660
1661 if (bif->bif_flags & IFBIF_STP) {
1662 switch (bif->bif_state) {
1663 case BSTP_IFSTATE_BLOCKING:
1664 case BSTP_IFSTATE_LISTENING:
1665 case BSTP_IFSTATE_DISABLED:
1666 m_freem(m);
1667 bridge_release_member(sc, bif, &psref);
1668 goto out;
1669 }
1670 }
1671
1672 eh = mtod(m, struct ether_header *);
1673
1674 /*
1675 * If the interface is learning, and the source
1676 * address is valid and not multicast, record
1677 * the address.
1678 */
1679 if ((bif->bif_flags & IFBIF_LEARNING) != 0 &&
1680 ETHER_IS_MULTICAST(eh->ether_shost) == 0 &&
1681 (eh->ether_shost[0] == 0 &&
1682 eh->ether_shost[1] == 0 &&
1683 eh->ether_shost[2] == 0 &&
1684 eh->ether_shost[3] == 0 &&
1685 eh->ether_shost[4] == 0 &&
1686 eh->ether_shost[5] == 0) == 0) {
1687 (void) bridge_rtupdate(sc, eh->ether_shost,
1688 src_if, 0, IFBAF_DYNAMIC);
1689 }
1690
1691 if ((bif->bif_flags & IFBIF_STP) != 0 &&
1692 bif->bif_state == BSTP_IFSTATE_LEARNING) {
1693 m_freem(m);
1694 bridge_release_member(sc, bif, &psref);
1695 goto out;
1696 }
1697
1698 bridge_release_member(sc, bif, &psref);
1699
1700 /*
1701 * At this point, the port either doesn't participate
1702 * in spanning tree or it is in the forwarding state.
1703 */
1704
1705 /*
1706 * If the packet is unicast, destined for someone on
1707 * "this" side of the bridge, drop it.
1708 */
1709 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
1710 dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1711 if (src_if == dst_if) {
1712 m_freem(m);
1713 goto out;
1714 }
1715 } else {
1716 /* ...forward it to all interfaces. */
1717 sc->sc_if.if_imcasts++;
1718 dst_if = NULL;
1719 }
1720
1721 if (pfil_run_hooks(sc->sc_if.if_pfil, &m, src_if, PFIL_IN) != 0) {
1722 if (m != NULL)
1723 m_freem(m);
1724 goto out;
1725 }
1726 if (m == NULL)
1727 goto out;
1728
1729 if (dst_if == NULL) {
1730 bridge_broadcast(sc, src_if, m);
1731 goto out;
1732 }
1733
1734 m_put_rcvif_psref(src_if, &psref_src);
1735 src_if = NULL;
1736
1737 /*
1738 * At this point, we're dealing with a unicast frame
1739 * going to a different interface.
1740 */
1741 if ((dst_if->if_flags & IFF_RUNNING) == 0) {
1742 m_freem(m);
1743 goto out;
1744 }
1745
1746 bif = bridge_lookup_member_if(sc, dst_if, &psref);
1747 if (bif == NULL) {
1748 /* Not a member of the bridge (anymore?) */
1749 m_freem(m);
1750 goto out;
1751 }
1752
1753 if (bif->bif_flags & IFBIF_STP) {
1754 switch (bif->bif_state) {
1755 case BSTP_IFSTATE_DISABLED:
1756 case BSTP_IFSTATE_BLOCKING:
1757 m_freem(m);
1758 bridge_release_member(sc, bif, &psref);
1759 goto out;
1760 }
1761 }
1762
1763 bridge_release_member(sc, bif, &psref);
1764
1765 /*
1766 * Before enqueueing this packet to the destination interface,
1767 * clear any in-bound checksum flags to prevent them from being
1768 * misused as out-bound flags.
1769 */
1770 m->m_pkthdr.csum_flags = 0;
1771
1772 ACQUIRE_GLOBAL_LOCKS();
1773 bridge_enqueue(sc, dst_if, m, 1);
1774 RELEASE_GLOBAL_LOCKS();
1775 out:
1776 if (src_if != NULL)
1777 m_put_rcvif_psref(src_if, &psref_src);
1778 return;
1779 }
1780
1781 static bool
1782 bstp_state_before_learning(struct bridge_iflist *bif)
1783 {
1784 if (bif->bif_flags & IFBIF_STP) {
1785 switch (bif->bif_state) {
1786 case BSTP_IFSTATE_BLOCKING:
1787 case BSTP_IFSTATE_LISTENING:
1788 case BSTP_IFSTATE_DISABLED:
1789 return true;
1790 }
1791 }
1792 return false;
1793 }
1794
1795 static bool
1796 bridge_ourether(struct bridge_iflist *bif, struct ether_header *eh, int src)
1797 {
1798 uint8_t *ether = src ? eh->ether_shost : eh->ether_dhost;
1799
1800 if (memcmp(CLLADDR(bif->bif_ifp->if_sadl), ether, ETHER_ADDR_LEN) == 0
1801 #if NCARP > 0
1802 || (bif->bif_ifp->if_carp &&
1803 carp_ourether(bif->bif_ifp->if_carp, eh, IFT_ETHER, src) != NULL)
1804 #endif /* NCARP > 0 */
1805 )
1806 return true;
1807
1808 return false;
1809 }
1810
1811 /*
1812 * bridge_input:
1813 *
1814 * Receive input from a member interface. Queue the packet for
1815 * bridging if it is not for us.
1816 */
1817 static void
1818 bridge_input(struct ifnet *ifp, struct mbuf *m)
1819 {
1820 struct bridge_softc *sc = ifp->if_bridge;
1821 struct bridge_iflist *bif;
1822 struct ether_header *eh;
1823 struct psref psref;
1824 int bound;
1825 DECLARE_LOCK_VARIABLE;
1826
1827 KASSERT(!cpu_intr_p());
1828
1829 if (__predict_false(sc == NULL) ||
1830 (sc->sc_if.if_flags & IFF_RUNNING) == 0) {
1831 ACQUIRE_GLOBAL_LOCKS();
1832 ether_input(ifp, m);
1833 RELEASE_GLOBAL_LOCKS();
1834 return;
1835 }
1836
1837 bound = curlwp_bind();
1838 bif = bridge_lookup_member_if(sc, ifp, &psref);
1839 if (bif == NULL) {
1840 curlwp_bindx(bound);
1841 ACQUIRE_GLOBAL_LOCKS();
1842 ether_input(ifp, m);
1843 RELEASE_GLOBAL_LOCKS();
1844 return;
1845 }
1846
1847 eh = mtod(m, struct ether_header *);
1848
1849 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1850 if (memcmp(etherbroadcastaddr,
1851 eh->ether_dhost, ETHER_ADDR_LEN) == 0)
1852 m->m_flags |= M_BCAST;
1853 else
1854 m->m_flags |= M_MCAST;
1855 }
1856
1857 /*
1858 * A 'fast' path for packets addressed to interfaces that are
1859 * part of this bridge.
1860 */
1861 if (!(m->m_flags & (M_BCAST|M_MCAST)) &&
1862 !bstp_state_before_learning(bif)) {
1863 struct bridge_iflist *_bif;
1864 struct ifnet *_ifp = NULL;
1865 int s;
1866 struct psref _psref;
1867
1868 BRIDGE_PSZ_RENTER(s);
1869 BRIDGE_IFLIST_READER_FOREACH(_bif, sc) {
1870 /* It is destined for us. */
1871 if (bridge_ourether(_bif, eh, 0)) {
1872 bridge_acquire_member(sc, _bif, &_psref);
1873 BRIDGE_PSZ_REXIT(s);
1874 if (_bif->bif_flags & IFBIF_LEARNING)
1875 (void) bridge_rtupdate(sc,
1876 eh->ether_shost, ifp, 0, IFBAF_DYNAMIC);
1877 m_set_rcvif(m, _bif->bif_ifp);
1878 _ifp = _bif->bif_ifp;
1879 bridge_release_member(sc, _bif, &_psref);
1880 goto out;
1881 }
1882
1883 /* We just received a packet that we sent out. */
1884 if (bridge_ourether(_bif, eh, 1))
1885 break;
1886 }
1887 BRIDGE_PSZ_REXIT(s);
1888 out:
1889
1890 if (_bif != NULL) {
1891 bridge_release_member(sc, bif, &psref);
1892 curlwp_bindx(bound);
1893 if (_ifp != NULL) {
1894 m->m_flags &= ~M_PROMISC;
1895 ACQUIRE_GLOBAL_LOCKS();
1896 ether_input(_ifp, m);
1897 RELEASE_GLOBAL_LOCKS();
1898 } else
1899 m_freem(m);
1900 return;
1901 }
1902 }
1903
1904 /* Tap off 802.1D packets; they do not get forwarded. */
1905 if (bif->bif_flags & IFBIF_STP &&
1906 memcmp(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN) == 0) {
1907 bstp_input(sc, bif, m);
1908 bridge_release_member(sc, bif, &psref);
1909 curlwp_bindx(bound);
1910 return;
1911 }
1912
1913 /*
1914 * A normal switch would discard the packet here, but that's not what
1915 * we've done historically. This also prevents some obnoxious behaviour.
1916 */
1917 if (bstp_state_before_learning(bif)) {
1918 bridge_release_member(sc, bif, &psref);
1919 curlwp_bindx(bound);
1920 ACQUIRE_GLOBAL_LOCKS();
1921 ether_input(ifp, m);
1922 RELEASE_GLOBAL_LOCKS();
1923 return;
1924 }
1925
1926 bridge_release_member(sc, bif, &psref);
1927
1928 bridge_forward(sc, m);
1929
1930 curlwp_bindx(bound);
1931 }
1932
1933 /*
1934 * bridge_broadcast:
1935 *
1936 * Send a frame to all interfaces that are members of
1937 * the bridge, except for the one on which the packet
1938 * arrived.
1939 */
1940 static void
1941 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
1942 struct mbuf *m)
1943 {
1944 struct bridge_iflist *bif;
1945 struct mbuf *mc;
1946 struct ifnet *dst_if;
1947 bool bmcast;
1948 int s;
1949 DECLARE_LOCK_VARIABLE;
1950
1951 bmcast = m->m_flags & (M_BCAST|M_MCAST);
1952
1953 BRIDGE_PSZ_RENTER(s);
1954 BRIDGE_IFLIST_READER_FOREACH(bif, sc) {
1955 struct psref psref;
1956
1957 bridge_acquire_member(sc, bif, &psref);
1958 BRIDGE_PSZ_REXIT(s);
1959
1960 dst_if = bif->bif_ifp;
1961
1962 if (bif->bif_flags & IFBIF_STP) {
1963 switch (bif->bif_state) {
1964 case BSTP_IFSTATE_BLOCKING:
1965 case BSTP_IFSTATE_DISABLED:
1966 goto next;
1967 }
1968 }
1969
1970 if ((bif->bif_flags & IFBIF_DISCOVER) == 0 && !bmcast)
1971 goto next;
1972
1973 if ((dst_if->if_flags & IFF_RUNNING) == 0)
1974 goto next;
1975
1976 if (dst_if != src_if) {
1977 mc = m_copypacket(m, M_DONTWAIT);
1978 if (mc == NULL) {
1979 sc->sc_if.if_oerrors++;
1980 goto next;
1981 }
1982 /*
1983 * Before enqueueing this packet to the destination
1984 * interface, clear any in-bound checksum flags to
1985 * prevent them from being misused as out-bound flags.
1986 */
1987 mc->m_pkthdr.csum_flags = 0;
1988
1989 ACQUIRE_GLOBAL_LOCKS();
1990 bridge_enqueue(sc, dst_if, mc, 1);
1991 RELEASE_GLOBAL_LOCKS();
1992 }
1993
1994 if (bmcast) {
1995 mc = m_copypacket(m, M_DONTWAIT);
1996 if (mc == NULL) {
1997 sc->sc_if.if_oerrors++;
1998 goto next;
1999 }
2000
2001 m_set_rcvif(mc, dst_if);
2002 mc->m_flags &= ~M_PROMISC;
2003
2004 ACQUIRE_GLOBAL_LOCKS();
2005 ether_input(dst_if, mc);
2006 RELEASE_GLOBAL_LOCKS();
2007 }
2008 next:
2009 BRIDGE_PSZ_RENTER(s);
2010 bridge_release_member(sc, bif, &psref);
2011 }
2012 BRIDGE_PSZ_REXIT(s);
2013
2014 m_freem(m);
2015 }
2016
2017 static int
2018 bridge_rtalloc(struct bridge_softc *sc, const uint8_t *dst,
2019 struct bridge_rtnode **brtp)
2020 {
2021 struct bridge_rtnode *brt;
2022 int error;
2023
2024 if (sc->sc_brtcnt >= sc->sc_brtmax)
2025 return ENOSPC;
2026
2027 /*
2028 * Allocate a new bridge forwarding node, and
2029 * initialize the expiration time and Ethernet
2030 * address.
2031 */
2032 brt = pool_get(&bridge_rtnode_pool, PR_NOWAIT);
2033 if (brt == NULL)
2034 return ENOMEM;
2035
2036 memset(brt, 0, sizeof(*brt));
2037 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2038 brt->brt_flags = IFBAF_DYNAMIC;
2039 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2040 PSLIST_ENTRY_INIT(brt, brt_list);
2041 PSLIST_ENTRY_INIT(brt, brt_hash);
2042
2043 BRIDGE_RT_LOCK(sc);
2044 error = bridge_rtnode_insert(sc, brt);
2045 BRIDGE_RT_UNLOCK(sc);
2046
2047 if (error != 0) {
2048 pool_put(&bridge_rtnode_pool, brt);
2049 return error;
2050 }
2051
2052 *brtp = brt;
2053 return 0;
2054 }
2055
2056 /*
2057 * bridge_rtupdate:
2058 *
2059 * Add a bridge routing entry.
2060 */
2061 static int
2062 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
2063 struct ifnet *dst_if, int setflags, uint8_t flags)
2064 {
2065 struct bridge_rtnode *brt;
2066 int s;
2067
2068 again:
2069 /*
2070 * A route for this destination might already exist. If so,
2071 * update it, otherwise create a new one.
2072 */
2073 BRIDGE_RT_RENTER(s);
2074 brt = bridge_rtnode_lookup(sc, dst);
2075
2076 if (brt != NULL) {
2077 brt->brt_ifp = dst_if;
2078 if (setflags) {
2079 brt->brt_flags = flags;
2080 if (flags & IFBAF_STATIC)
2081 brt->brt_expire = 0;
2082 else
2083 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2084 } else {
2085 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2086 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2087 }
2088 }
2089 BRIDGE_RT_REXIT(s);
2090
2091 if (brt == NULL) {
2092 int r;
2093
2094 r = bridge_rtalloc(sc, dst, &brt);
2095 if (r != 0)
2096 return r;
2097 goto again;
2098 }
2099
2100 return 0;
2101 }
2102
2103 /*
2104 * bridge_rtlookup:
2105 *
2106 * Lookup the destination interface for an address.
2107 */
2108 static struct ifnet *
2109 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr)
2110 {
2111 struct bridge_rtnode *brt;
2112 struct ifnet *ifs = NULL;
2113 int s;
2114
2115 BRIDGE_RT_RENTER(s);
2116 brt = bridge_rtnode_lookup(sc, addr);
2117 if (brt != NULL)
2118 ifs = brt->brt_ifp;
2119 BRIDGE_RT_REXIT(s);
2120
2121 return ifs;
2122 }
2123
2124 typedef bool (*bridge_iterate_cb_t)
2125 (struct bridge_softc *, struct bridge_rtnode *, bool *, void *);
2126
2127 /*
2128 * bridge_rtlist_iterate_remove:
2129 *
2130 * It iterates on sc->sc_rtlist and removes rtnodes of it which func
2131 * callback judges to remove. Removals of rtnodes are done in a manner
2132 * of pserialize. To this end, all kmem_* operations are placed out of
2133 * mutexes.
2134 */
2135 static void
2136 bridge_rtlist_iterate_remove(struct bridge_softc *sc, bridge_iterate_cb_t func, void *arg)
2137 {
2138 struct bridge_rtnode *brt;
2139 struct bridge_rtnode **brt_list;
2140 int i, count;
2141
2142 retry:
2143 count = sc->sc_brtcnt;
2144 if (count == 0)
2145 return;
2146 brt_list = kmem_alloc(sizeof(*brt_list) * count, KM_SLEEP);
2147
2148 BRIDGE_RT_LOCK(sc);
2149 if (__predict_false(sc->sc_brtcnt > count)) {
2150 /* The rtnodes increased, we need more memory */
2151 BRIDGE_RT_UNLOCK(sc);
2152 kmem_free(brt_list, sizeof(*brt_list) * count);
2153 goto retry;
2154 }
2155
2156 i = 0;
2157 /*
2158 * We don't need to use a _SAFE variant here because we know
2159 * that a removed item keeps its next pointer as-is thanks to
2160 * pslist(9) and isn't freed in the loop.
2161 */
2162 BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) {
2163 bool need_break = false;
2164 if (func(sc, brt, &need_break, arg)) {
2165 bridge_rtnode_remove(sc, brt);
2166 brt_list[i++] = brt;
2167 }
2168 if (need_break)
2169 break;
2170 }
2171
2172 if (i > 0)
2173 BRIDGE_RT_PSZ_PERFORM(sc);
2174 BRIDGE_RT_UNLOCK(sc);
2175
2176 while (--i >= 0)
2177 bridge_rtnode_destroy(brt_list[i]);
2178
2179 kmem_free(brt_list, sizeof(*brt_list) * count);
2180 }
2181
2182 static bool
2183 bridge_rttrim0_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
2184 bool *need_break, void *arg)
2185 {
2186 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2187 /* Take into account of the subsequent removal */
2188 if ((sc->sc_brtcnt - 1) <= sc->sc_brtmax)
2189 *need_break = true;
2190 return true;
2191 } else
2192 return false;
2193 }
2194
2195 static void
2196 bridge_rttrim0(struct bridge_softc *sc)
2197 {
2198 bridge_rtlist_iterate_remove(sc, bridge_rttrim0_cb, NULL);
2199 }
2200
2201 /*
2202 * bridge_rttrim:
2203 *
2204 * Trim the routine table so that we have a number
2205 * of routing entries less than or equal to the
2206 * maximum number.
2207 */
2208 static void
2209 bridge_rttrim(struct bridge_softc *sc)
2210 {
2211
2212 /* Make sure we actually need to do this. */
2213 if (sc->sc_brtcnt <= sc->sc_brtmax)
2214 return;
2215
2216 /* Force an aging cycle; this might trim enough addresses. */
2217 bridge_rtage(sc);
2218 if (sc->sc_brtcnt <= sc->sc_brtmax)
2219 return;
2220
2221 bridge_rttrim0(sc);
2222
2223 return;
2224 }
2225
2226 /*
2227 * bridge_timer:
2228 *
2229 * Aging timer for the bridge.
2230 */
2231 static void
2232 bridge_timer(void *arg)
2233 {
2234 struct bridge_softc *sc = arg;
2235
2236 workqueue_enqueue(sc->sc_rtage_wq, &sc->sc_rtage_wk, NULL);
2237 }
2238
2239 static void
2240 bridge_rtage_work(struct work *wk, void *arg)
2241 {
2242 struct bridge_softc *sc = arg;
2243
2244 KASSERT(wk == &sc->sc_rtage_wk);
2245
2246 bridge_rtage(sc);
2247
2248 if (sc->sc_if.if_flags & IFF_RUNNING)
2249 callout_reset(&sc->sc_brcallout,
2250 bridge_rtable_prune_period * hz, bridge_timer, sc);
2251 }
2252
2253 static bool
2254 bridge_rtage_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
2255 bool *need_break, void *arg)
2256 {
2257 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2258 time_uptime >= brt->brt_expire)
2259 return true;
2260 else
2261 return false;
2262 }
2263
2264 /*
2265 * bridge_rtage:
2266 *
2267 * Perform an aging cycle.
2268 */
2269 static void
2270 bridge_rtage(struct bridge_softc *sc)
2271 {
2272 bridge_rtlist_iterate_remove(sc, bridge_rtage_cb, NULL);
2273 }
2274
2275
2276 static bool
2277 bridge_rtflush_cb(struct bridge_softc *sc, struct bridge_rtnode *brt,
2278 bool *need_break, void *arg)
2279 {
2280 int full = *(int*)arg;
2281
2282 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2283 return true;
2284 else
2285 return false;
2286 }
2287
2288 /*
2289 * bridge_rtflush:
2290 *
2291 * Remove all dynamic addresses from the bridge.
2292 */
2293 static void
2294 bridge_rtflush(struct bridge_softc *sc, int full)
2295 {
2296 bridge_rtlist_iterate_remove(sc, bridge_rtflush_cb, &full);
2297 }
2298
2299 /*
2300 * bridge_rtdaddr:
2301 *
2302 * Remove an address from the table.
2303 */
2304 static int
2305 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr)
2306 {
2307 struct bridge_rtnode *brt;
2308
2309 BRIDGE_RT_LOCK(sc);
2310 if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL) {
2311 BRIDGE_RT_UNLOCK(sc);
2312 return ENOENT;
2313 }
2314 bridge_rtnode_remove(sc, brt);
2315 BRIDGE_RT_PSZ_PERFORM(sc);
2316 BRIDGE_RT_UNLOCK(sc);
2317
2318 bridge_rtnode_destroy(brt);
2319
2320 return 0;
2321 }
2322
2323 /*
2324 * bridge_rtdelete:
2325 *
2326 * Delete routes to a speicifc member interface.
2327 */
2328 static void
2329 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp)
2330 {
2331 struct bridge_rtnode *brt;
2332
2333 /* XXX pserialize_perform for each entry is slow */
2334 again:
2335 BRIDGE_RT_LOCK(sc);
2336 BRIDGE_RTLIST_WRITER_FOREACH(brt, sc) {
2337 if (brt->brt_ifp == ifp)
2338 break;
2339 }
2340 if (brt == NULL) {
2341 BRIDGE_RT_UNLOCK(sc);
2342 return;
2343 }
2344 bridge_rtnode_remove(sc, brt);
2345 BRIDGE_RT_PSZ_PERFORM(sc);
2346 BRIDGE_RT_UNLOCK(sc);
2347
2348 bridge_rtnode_destroy(brt);
2349
2350 goto again;
2351 }
2352
2353 /*
2354 * bridge_rtable_init:
2355 *
2356 * Initialize the route table for this bridge.
2357 */
2358 static void
2359 bridge_rtable_init(struct bridge_softc *sc)
2360 {
2361 int i;
2362
2363 sc->sc_rthash = kmem_alloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2364 KM_SLEEP);
2365
2366 for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2367 PSLIST_INIT(&sc->sc_rthash[i]);
2368
2369 sc->sc_rthash_key = cprng_fast32();
2370
2371 PSLIST_INIT(&sc->sc_rtlist);
2372
2373 sc->sc_rtlist_psz = pserialize_create();
2374 sc->sc_rtlist_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_SOFTNET);
2375 }
2376
2377 /*
2378 * bridge_rtable_fini:
2379 *
2380 * Deconstruct the route table for this bridge.
2381 */
2382 static void
2383 bridge_rtable_fini(struct bridge_softc *sc)
2384 {
2385
2386 kmem_free(sc->sc_rthash, sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE);
2387 mutex_obj_free(sc->sc_rtlist_lock);
2388 pserialize_destroy(sc->sc_rtlist_psz);
2389 }
2390
2391 /*
2392 * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2393 * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2394 */
2395 #define mix(a, b, c) \
2396 do { \
2397 a -= b; a -= c; a ^= (c >> 13); \
2398 b -= c; b -= a; b ^= (a << 8); \
2399 c -= a; c -= b; c ^= (b >> 13); \
2400 a -= b; a -= c; a ^= (c >> 12); \
2401 b -= c; b -= a; b ^= (a << 16); \
2402 c -= a; c -= b; c ^= (b >> 5); \
2403 a -= b; a -= c; a ^= (c >> 3); \
2404 b -= c; b -= a; b ^= (a << 10); \
2405 c -= a; c -= b; c ^= (b >> 15); \
2406 } while (/*CONSTCOND*/0)
2407
2408 static inline uint32_t
2409 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2410 {
2411 uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2412
2413 b += addr[5] << 8;
2414 b += addr[4];
2415 a += addr[3] << 24;
2416 a += addr[2] << 16;
2417 a += addr[1] << 8;
2418 a += addr[0];
2419
2420 mix(a, b, c);
2421
2422 return (c & BRIDGE_RTHASH_MASK);
2423 }
2424
2425 #undef mix
2426
2427 /*
2428 * bridge_rtnode_lookup:
2429 *
2430 * Look up a bridge route node for the specified destination.
2431 */
2432 static struct bridge_rtnode *
2433 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr)
2434 {
2435 struct bridge_rtnode *brt;
2436 uint32_t hash;
2437 int dir;
2438
2439 hash = bridge_rthash(sc, addr);
2440 BRIDGE_RTHASH_READER_FOREACH(brt, sc, hash) {
2441 dir = memcmp(addr, brt->brt_addr, ETHER_ADDR_LEN);
2442 if (dir == 0)
2443 return brt;
2444 if (dir > 0)
2445 return NULL;
2446 }
2447
2448 return NULL;
2449 }
2450
2451 /*
2452 * bridge_rtnode_insert:
2453 *
2454 * Insert the specified bridge node into the route table. We
2455 * assume the entry is not already in the table.
2456 */
2457 static int
2458 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2459 {
2460 struct bridge_rtnode *lbrt, *prev = NULL;
2461 uint32_t hash;
2462
2463 KASSERT(BRIDGE_RT_LOCKED(sc));
2464
2465 hash = bridge_rthash(sc, brt->brt_addr);
2466 BRIDGE_RTHASH_WRITER_FOREACH(lbrt, sc, hash) {
2467 int dir = memcmp(brt->brt_addr, lbrt->brt_addr, ETHER_ADDR_LEN);
2468 if (dir == 0)
2469 return EEXIST;
2470 if (dir > 0)
2471 break;
2472 prev = lbrt;
2473 }
2474 if (prev == NULL)
2475 BRIDGE_RTHASH_WRITER_INSERT_HEAD(sc, hash, brt);
2476 else
2477 BRIDGE_RTHASH_WRITER_INSERT_AFTER(prev, brt);
2478
2479 BRIDGE_RTLIST_WRITER_INSERT_HEAD(sc, brt);
2480 sc->sc_brtcnt++;
2481
2482 return 0;
2483 }
2484
2485 /*
2486 * bridge_rtnode_remove:
2487 *
2488 * Remove a bridge rtnode from the rthash and the rtlist of a bridge.
2489 */
2490 static void
2491 bridge_rtnode_remove(struct bridge_softc *sc, struct bridge_rtnode *brt)
2492 {
2493
2494 KASSERT(BRIDGE_RT_LOCKED(sc));
2495
2496 BRIDGE_RTHASH_WRITER_REMOVE(brt);
2497 BRIDGE_RTLIST_WRITER_REMOVE(brt);
2498 sc->sc_brtcnt--;
2499 }
2500
2501 /*
2502 * bridge_rtnode_destroy:
2503 *
2504 * Destroy a bridge rtnode.
2505 */
2506 static void
2507 bridge_rtnode_destroy(struct bridge_rtnode *brt)
2508 {
2509
2510 PSLIST_ENTRY_DESTROY(brt, brt_list);
2511 PSLIST_ENTRY_DESTROY(brt, brt_hash);
2512 pool_put(&bridge_rtnode_pool, brt);
2513 }
2514
2515 #if defined(BRIDGE_IPF)
2516 extern pfil_head_t *inet_pfil_hook; /* XXX */
2517 extern pfil_head_t *inet6_pfil_hook; /* XXX */
2518
2519 /*
2520 * Send bridge packets through IPF if they are one of the types IPF can deal
2521 * with, or if they are ARP or REVARP. (IPF will pass ARP and REVARP without
2522 * question.)
2523 */
2524 static int
2525 bridge_ipf(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
2526 {
2527 int snap, error;
2528 struct ether_header *eh1, eh2;
2529 struct llc llc1;
2530 uint16_t ether_type;
2531
2532 snap = 0;
2533 error = -1; /* Default error if not error == 0 */
2534 eh1 = mtod(*mp, struct ether_header *);
2535 ether_type = ntohs(eh1->ether_type);
2536
2537 /*
2538 * Check for SNAP/LLC.
2539 */
2540 if (ether_type < ETHERMTU) {
2541 struct llc *llc2 = (struct llc *)(eh1 + 1);
2542
2543 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
2544 llc2->llc_dsap == LLC_SNAP_LSAP &&
2545 llc2->llc_ssap == LLC_SNAP_LSAP &&
2546 llc2->llc_control == LLC_UI) {
2547 ether_type = htons(llc2->llc_un.type_snap.ether_type);
2548 snap = 1;
2549 }
2550 }
2551
2552 /*
2553 * If we're trying to filter bridge traffic, don't look at anything
2554 * other than IP and ARP traffic. If the filter doesn't understand
2555 * IPv6, don't allow IPv6 through the bridge either. This is lame
2556 * since if we really wanted, say, an AppleTalk filter, we are hosed,
2557 * but of course we don't have an AppleTalk filter to begin with.
2558 * (Note that since IPF doesn't understand ARP it will pass *ALL*
2559 * ARP traffic.)
2560 */
2561 switch (ether_type) {
2562 case ETHERTYPE_ARP:
2563 case ETHERTYPE_REVARP:
2564 return 0; /* Automatically pass */
2565 case ETHERTYPE_IP:
2566 # ifdef INET6
2567 case ETHERTYPE_IPV6:
2568 # endif /* INET6 */
2569 break;
2570 default:
2571 goto bad;
2572 }
2573
2574 /* Strip off the Ethernet header and keep a copy. */
2575 m_copydata(*mp, 0, ETHER_HDR_LEN, (void *) &eh2);
2576 m_adj(*mp, ETHER_HDR_LEN);
2577
2578 /* Strip off snap header, if present */
2579 if (snap) {
2580 m_copydata(*mp, 0, sizeof(struct llc), (void *) &llc1);
2581 m_adj(*mp, sizeof(struct llc));
2582 }
2583
2584 /*
2585 * Check basic packet sanity and run IPF through pfil.
2586 */
2587 KASSERT(!cpu_intr_p());
2588 switch (ether_type)
2589 {
2590 case ETHERTYPE_IP :
2591 error = bridge_ip_checkbasic(mp);
2592 if (error == 0)
2593 error = pfil_run_hooks(inet_pfil_hook, mp, ifp, dir);
2594 break;
2595 # ifdef INET6
2596 case ETHERTYPE_IPV6 :
2597 error = bridge_ip6_checkbasic(mp);
2598 if (error == 0)
2599 error = pfil_run_hooks(inet6_pfil_hook, mp, ifp, dir);
2600 break;
2601 # endif
2602 default :
2603 error = 0;
2604 break;
2605 }
2606
2607 if (*mp == NULL)
2608 return error;
2609 if (error != 0)
2610 goto bad;
2611
2612 error = -1;
2613
2614 /*
2615 * Finally, put everything back the way it was and return
2616 */
2617 if (snap) {
2618 M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
2619 if (*mp == NULL)
2620 return error;
2621 bcopy(&llc1, mtod(*mp, void *), sizeof(struct llc));
2622 }
2623
2624 M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
2625 if (*mp == NULL)
2626 return error;
2627 bcopy(&eh2, mtod(*mp, void *), ETHER_HDR_LEN);
2628
2629 return 0;
2630
2631 bad:
2632 m_freem(*mp);
2633 *mp = NULL;
2634 return error;
2635 }
2636
2637 /*
2638 * Perform basic checks on header size since
2639 * IPF assumes ip_input has already processed
2640 * it for it. Cut-and-pasted from ip_input.c.
2641 * Given how simple the IPv6 version is,
2642 * does the IPv4 version really need to be
2643 * this complicated?
2644 *
2645 * XXX Should we update ipstat here, or not?
2646 * XXX Right now we update ipstat but not
2647 * XXX csum_counter.
2648 */
2649 static int
2650 bridge_ip_checkbasic(struct mbuf **mp)
2651 {
2652 struct mbuf *m = *mp;
2653 struct ip *ip;
2654 int len, hlen;
2655
2656 if (*mp == NULL)
2657 return -1;
2658
2659 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
2660 if ((m = m_copyup(m, sizeof(struct ip),
2661 (max_linkhdr + 3) & ~3)) == NULL) {
2662 /* XXXJRT new stat, please */
2663 ip_statinc(IP_STAT_TOOSMALL);
2664 goto bad;
2665 }
2666 } else if (__predict_false(m->m_len < sizeof (struct ip))) {
2667 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
2668 ip_statinc(IP_STAT_TOOSMALL);
2669 goto bad;
2670 }
2671 }
2672 ip = mtod(m, struct ip *);
2673 if (ip == NULL) goto bad;
2674
2675 if (ip->ip_v != IPVERSION) {
2676 ip_statinc(IP_STAT_BADVERS);
2677 goto bad;
2678 }
2679 hlen = ip->ip_hl << 2;
2680 if (hlen < sizeof(struct ip)) { /* minimum header length */
2681 ip_statinc(IP_STAT_BADHLEN);
2682 goto bad;
2683 }
2684 if (hlen > m->m_len) {
2685 if ((m = m_pullup(m, hlen)) == 0) {
2686 ip_statinc(IP_STAT_BADHLEN);
2687 goto bad;
2688 }
2689 ip = mtod(m, struct ip *);
2690 if (ip == NULL) goto bad;
2691 }
2692
2693 switch (m->m_pkthdr.csum_flags &
2694 ((m_get_rcvif_NOMPSAFE(m)->if_csum_flags_rx & M_CSUM_IPv4) |
2695 M_CSUM_IPv4_BAD)) {
2696 case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
2697 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); */
2698 goto bad;
2699
2700 case M_CSUM_IPv4:
2701 /* Checksum was okay. */
2702 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); */
2703 break;
2704
2705 default:
2706 /* Must compute it ourselves. */
2707 /* INET_CSUM_COUNTER_INCR(&ip_swcsum); */
2708 if (in_cksum(m, hlen) != 0)
2709 goto bad;
2710 break;
2711 }
2712
2713 /* Retrieve the packet length. */
2714 len = ntohs(ip->ip_len);
2715
2716 /*
2717 * Check for additional length bogosity
2718 */
2719 if (len < hlen) {
2720 ip_statinc(IP_STAT_BADLEN);
2721 goto bad;
2722 }
2723
2724 /*
2725 * Check that the amount of data in the buffers
2726 * is as at least much as the IP header would have us expect.
2727 * Drop packet if shorter than we expect.
2728 */
2729 if (m->m_pkthdr.len < len) {
2730 ip_statinc(IP_STAT_TOOSHORT);
2731 goto bad;
2732 }
2733
2734 /* Checks out, proceed */
2735 *mp = m;
2736 return 0;
2737
2738 bad:
2739 *mp = m;
2740 return -1;
2741 }
2742
2743 # ifdef INET6
2744 /*
2745 * Same as above, but for IPv6.
2746 * Cut-and-pasted from ip6_input.c.
2747 * XXX Should we update ip6stat, or not?
2748 */
2749 static int
2750 bridge_ip6_checkbasic(struct mbuf **mp)
2751 {
2752 struct mbuf *m = *mp;
2753 struct ip6_hdr *ip6;
2754
2755 /*
2756 * If the IPv6 header is not aligned, slurp it up into a new
2757 * mbuf with space for link headers, in the event we forward
2758 * it. Otherwise, if it is aligned, make sure the entire base
2759 * IPv6 header is in the first mbuf of the chain.
2760 */
2761 if (IP6_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
2762 struct ifnet *inifp = m_get_rcvif_NOMPSAFE(m);
2763 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
2764 (max_linkhdr + 3) & ~3)) == NULL) {
2765 /* XXXJRT new stat, please */
2766 ip6_statinc(IP6_STAT_TOOSMALL);
2767 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
2768 goto bad;
2769 }
2770 } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
2771 struct ifnet *inifp = m_get_rcvif_NOMPSAFE(m);
2772 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
2773 ip6_statinc(IP6_STAT_TOOSMALL);
2774 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
2775 goto bad;
2776 }
2777 }
2778
2779 ip6 = mtod(m, struct ip6_hdr *);
2780
2781 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
2782 ip6_statinc(IP6_STAT_BADVERS);
2783 in6_ifstat_inc(m_get_rcvif_NOMPSAFE(m), ifs6_in_hdrerr);
2784 goto bad;
2785 }
2786
2787 /* Checks out, proceed */
2788 *mp = m;
2789 return 0;
2790
2791 bad:
2792 *mp = m;
2793 return -1;
2794 }
2795 # endif /* INET6 */
2796 #endif /* BRIDGE_IPF */
2797