1 /* $NetBSD: if_bridge.c,v 1.31 2005/06/01 19:45:34 jdc Exp $ */
2
3 /*-
4 * SPDX-License-Identifier: BSD-4-Clause
5 *
6 * Copyright 2001 Wasabi Systems, Inc.
7 * All rights reserved.
8 *
9 * Written by Jason R. Thorpe for Wasabi Systems, Inc.
10 *
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
13 * are met:
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. All advertising materials mentioning features or use of this software
20 * must display the following acknowledgement:
21 * This product includes software developed for the NetBSD Project by
22 * Wasabi Systems, Inc.
23 * 4. The name of Wasabi Systems, Inc. may not be used to endorse
24 * or promote products derived from this software without specific prior
25 * written permission.
26 *
27 * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
28 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL WASABI SYSTEMS, INC
31 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37 * POSSIBILITY OF SUCH DAMAGE.
38 */
39
40 /*
41 * Copyright (c) 1999, 2000 Jason L. Wright (jason@thought.net)
42 * All rights reserved.
43 *
44 * Redistribution and use in source and binary forms, with or without
45 * modification, are permitted provided that the following conditions
46 * are met:
47 * 1. Redistributions of source code must retain the above copyright
48 * notice, this list of conditions and the following disclaimer.
49 * 2. Redistributions in binary form must reproduce the above copyright
50 * notice, this list of conditions and the following disclaimer in the
51 * documentation and/or other materials provided with the distribution.
52 *
53 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
54 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
55 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
56 * DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
57 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
58 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
59 * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
61 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
62 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
63 * POSSIBILITY OF SUCH DAMAGE.
64 *
65 * OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun Exp
66 */
67
68 /*
69 * Network interface bridge support.
70 *
71 * TODO:
72 *
73 * - Currently only supports Ethernet-like interfaces (Ethernet,
74 * 802.11, VLANs on Ethernet, etc.) Figure out a nice way
75 * to bridge other types of interfaces (maybe consider
76 * heterogeneous bridges).
77 */
78
79 #include <sys/cdefs.h>
80 #include "opt_inet.h"
81 #include "opt_inet6.h"
82
83 #include <sys/param.h>
84 #include <sys/eventhandler.h>
85 #include <sys/mbuf.h>
86 #include <sys/malloc.h>
87 #include <sys/protosw.h>
88 #include <sys/systm.h>
89 #include <sys/jail.h>
90 #include <sys/time.h>
91 #include <sys/socket.h> /* for net/if.h */
92 #include <sys/sockio.h>
93 #include <sys/ctype.h> /* string functions */
94 #include <sys/kernel.h>
95 #include <sys/random.h>
96 #include <sys/syslog.h>
97 #include <sys/sysctl.h>
98 #include <vm/uma.h>
99 #include <sys/module.h>
100 #include <sys/priv.h>
101 #include <sys/proc.h>
102 #include <sys/lock.h>
103 #include <sys/mutex.h>
104
105 #include <net/bpf.h>
106 #include <net/if.h>
107 #include <net/if_clone.h>
108 #include <net/if_dl.h>
109 #include <net/if_types.h>
110 #include <net/if_var.h>
111 #include <net/if_private.h>
112 #include <net/pfil.h>
113 #include <net/vnet.h>
114
115 #include <netinet/in.h>
116 #include <netinet/in_systm.h>
117 #include <netinet/in_var.h>
118 #include <netinet/ip.h>
119 #include <netinet/ip_var.h>
120 #ifdef INET6
121 #include <netinet/ip6.h>
122 #include <netinet6/ip6_var.h>
123 #include <netinet6/in6_ifattach.h>
124 #endif
125 #if defined(INET) || defined(INET6)
126 #include <netinet/ip_carp.h>
127 #endif
128 #include <machine/in_cksum.h>
129 #include <netinet/if_ether.h>
130 #include <net/bridgestp.h>
131 #include <net/if_bridgevar.h>
132 #include <net/if_llc.h>
133 #include <net/if_vlan_var.h>
134
135 #include <net/route.h>
136
137 /*
138 * At various points in the code we need to know if we're hooked into the INET
139 * and/or INET6 pfil. Define some macros to do that based on which IP versions
140 * are enabled in the kernel. This avoids littering the rest of the code with
141 * #ifnet INET6 to avoid referencing V_inet6_pfil_head.
142 */
143 #ifdef INET6
144 #define PFIL_HOOKED_IN_INET6 PFIL_HOOKED_IN(V_inet6_pfil_head)
145 #define PFIL_HOOKED_OUT_INET6 PFIL_HOOKED_OUT(V_inet6_pfil_head)
146 #else
147 #define PFIL_HOOKED_IN_INET6 false
148 #define PFIL_HOOKED_OUT_INET6 false
149 #endif
150
151 #ifdef INET
152 #define PFIL_HOOKED_IN_INET PFIL_HOOKED_IN(V_inet_pfil_head)
153 #define PFIL_HOOKED_OUT_INET PFIL_HOOKED_OUT(V_inet_pfil_head)
154 #else
155 #define PFIL_HOOKED_IN_INET false
156 #define PFIL_HOOKED_OUT_INET false
157 #endif
158
159 #define PFIL_HOOKED_IN_46 (PFIL_HOOKED_IN_INET6 || PFIL_HOOKED_IN_INET)
160 #define PFIL_HOOKED_OUT_46 (PFIL_HOOKED_OUT_INET6 || PFIL_HOOKED_OUT_INET)
161
162 /*
163 * Size of the route hash table. Must be a power of two.
164 */
165 #ifndef BRIDGE_RTHASH_SIZE
166 #define BRIDGE_RTHASH_SIZE 1024
167 #endif
168
169 #define BRIDGE_RTHASH_MASK (BRIDGE_RTHASH_SIZE - 1)
170
171 /*
172 * Default maximum number of addresses to cache.
173 */
174 #ifndef BRIDGE_RTABLE_MAX
175 #define BRIDGE_RTABLE_MAX 2000
176 #endif
177
178 /*
179 * Timeout (in seconds) for entries learned dynamically.
180 */
181 #ifndef BRIDGE_RTABLE_TIMEOUT
182 #define BRIDGE_RTABLE_TIMEOUT (20 * 60) /* same as ARP */
183 #endif
184
185 /*
186 * Number of seconds between walks of the route list.
187 */
188 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
189 #define BRIDGE_RTABLE_PRUNE_PERIOD (5 * 60)
190 #endif
191
192 /*
193 * List of capabilities to possibly mask on the member interface.
194 */
195 #define BRIDGE_IFCAPS_MASK (IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM|\
196 IFCAP_TXCSUM_IPV6)
197
198 /*
199 * List of capabilities to strip
200 */
201 #define BRIDGE_IFCAPS_STRIP IFCAP_LRO
202
203 /*
204 * Bridge locking
205 *
206 * The bridge relies heavily on the epoch(9) system to protect its data
207 * structures. This means we can safely use CK_LISTs while in NET_EPOCH, but we
208 * must ensure there is only one writer at a time.
209 *
210 * That is: for read accesses we only need to be in NET_EPOCH, but for write
211 * accesses we must hold:
212 *
213 * - BRIDGE_RT_LOCK, for any change to bridge_rtnodes
214 * - BRIDGE_LOCK, for any other change
215 *
216 * The BRIDGE_LOCK is a sleepable lock, because it is held across ioctl()
217 * calls to bridge member interfaces and these ioctl()s can sleep.
218 * The BRIDGE_RT_LOCK is a non-sleepable mutex, because it is sometimes
219 * required while we're in NET_EPOCH and then we're not allowed to sleep.
220 */
221 #define BRIDGE_LOCK_INIT(_sc) do { \
222 sx_init(&(_sc)->sc_sx, "if_bridge"); \
223 mtx_init(&(_sc)->sc_rt_mtx, "if_bridge rt", NULL, MTX_DEF); \
224 } while (0)
225 #define BRIDGE_LOCK_DESTROY(_sc) do { \
226 sx_destroy(&(_sc)->sc_sx); \
227 mtx_destroy(&(_sc)->sc_rt_mtx); \
228 } while (0)
229 #define BRIDGE_LOCK(_sc) sx_xlock(&(_sc)->sc_sx)
230 #define BRIDGE_UNLOCK(_sc) sx_xunlock(&(_sc)->sc_sx)
231 #define BRIDGE_LOCK_ASSERT(_sc) sx_assert(&(_sc)->sc_sx, SX_XLOCKED)
232 #define BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(_sc) \
233 MPASS(in_epoch(net_epoch_preempt) || sx_xlocked(&(_sc)->sc_sx))
234 #define BRIDGE_UNLOCK_ASSERT(_sc) sx_assert(&(_sc)->sc_sx, SX_UNLOCKED)
235 #define BRIDGE_RT_LOCK(_sc) mtx_lock(&(_sc)->sc_rt_mtx)
236 #define BRIDGE_RT_UNLOCK(_sc) mtx_unlock(&(_sc)->sc_rt_mtx)
237 #define BRIDGE_RT_LOCK_ASSERT(_sc) mtx_assert(&(_sc)->sc_rt_mtx, MA_OWNED)
238 #define BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(_sc) \
239 MPASS(in_epoch(net_epoch_preempt) || mtx_owned(&(_sc)->sc_rt_mtx))
240
241 /*
242 * Bridge interface list entry.
243 */
244 struct bridge_iflist {
245 CK_LIST_ENTRY(bridge_iflist) bif_next;
246 struct ifnet *bif_ifp; /* member if */
247 struct bstp_port bif_stp; /* STP state */
248 uint32_t bif_flags; /* member if flags */
249 int bif_savedcaps; /* saved capabilities */
250 uint32_t bif_addrmax; /* max # of addresses */
251 uint32_t bif_addrcnt; /* cur. # of addresses */
252 uint32_t bif_addrexceeded;/* # of address violations */
253 struct epoch_context bif_epoch_ctx;
254 };
255
256 /*
257 * Bridge route node.
258 */
259 struct bridge_rtnode {
260 CK_LIST_ENTRY(bridge_rtnode) brt_hash; /* hash table linkage */
261 CK_LIST_ENTRY(bridge_rtnode) brt_list; /* list linkage */
262 struct bridge_iflist *brt_dst; /* destination if */
263 unsigned long brt_expire; /* expiration time */
264 uint8_t brt_flags; /* address flags */
265 uint8_t brt_addr[ETHER_ADDR_LEN];
266 uint16_t brt_vlan; /* vlan id */
267 struct vnet *brt_vnet;
268 struct epoch_context brt_epoch_ctx;
269 };
270 #define brt_ifp brt_dst->bif_ifp
271
272 /*
273 * Software state for each bridge.
274 */
275 struct bridge_softc {
276 struct ifnet *sc_ifp; /* make this an interface */
277 LIST_ENTRY(bridge_softc) sc_list;
278 struct sx sc_sx;
279 struct mtx sc_rt_mtx;
280 uint32_t sc_brtmax; /* max # of addresses */
281 uint32_t sc_brtcnt; /* cur. # of addresses */
282 uint32_t sc_brttimeout; /* rt timeout in seconds */
283 struct callout sc_brcallout; /* bridge callout */
284 CK_LIST_HEAD(, bridge_iflist) sc_iflist; /* member interface list */
285 CK_LIST_HEAD(, bridge_rtnode) *sc_rthash; /* our forwarding table */
286 CK_LIST_HEAD(, bridge_rtnode) sc_rtlist; /* list version of above */
287 uint32_t sc_rthash_key; /* key for hash */
288 CK_LIST_HEAD(, bridge_iflist) sc_spanlist; /* span ports list */
289 struct bstp_state sc_stp; /* STP state */
290 uint32_t sc_brtexceeded; /* # of cache drops */
291 struct ifnet *sc_ifaddr; /* member mac copied from */
292 struct ether_addr sc_defaddr; /* Default MAC address */
293 if_input_fn_t sc_if_input; /* Saved copy of if_input */
294 struct epoch_context sc_epoch_ctx;
295 };
296
297 VNET_DEFINE_STATIC(struct sx, bridge_list_sx);
298 #define V_bridge_list_sx VNET(bridge_list_sx)
299 static eventhandler_tag bridge_detach_cookie;
300
301 int bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
302
303 VNET_DEFINE_STATIC(uma_zone_t, bridge_rtnode_zone);
304 #define V_bridge_rtnode_zone VNET(bridge_rtnode_zone)
305
306 static int bridge_clone_create(struct if_clone *, char *, size_t,
307 struct ifc_data *, struct ifnet **);
308 static int bridge_clone_destroy(struct if_clone *, struct ifnet *, uint32_t);
309
310 static int bridge_ioctl(struct ifnet *, u_long, caddr_t);
311 static void bridge_mutecaps(struct bridge_softc *);
312 static void bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
313 int);
314 static void bridge_ifdetach(void *arg __unused, struct ifnet *);
315 static void bridge_init(void *);
316 static void bridge_dummynet(struct mbuf *, struct ifnet *);
317 static void bridge_stop(struct ifnet *, int);
318 static int bridge_transmit(struct ifnet *, struct mbuf *);
319 #ifdef ALTQ
320 static void bridge_altq_start(if_t);
321 static int bridge_altq_transmit(if_t, struct mbuf *);
322 #endif
323 static void bridge_qflush(struct ifnet *);
324 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
325 static void bridge_inject(struct ifnet *, struct mbuf *);
326 static int bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
327 struct rtentry *);
328 static int bridge_enqueue(struct bridge_softc *, struct ifnet *,
329 struct mbuf *);
330 static void bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
331
332 static void bridge_forward(struct bridge_softc *, struct bridge_iflist *,
333 struct mbuf *m);
334
335 static void bridge_timer(void *);
336
337 static void bridge_broadcast(struct bridge_softc *, struct ifnet *,
338 struct mbuf *, int);
339 static void bridge_span(struct bridge_softc *, struct mbuf *);
340
341 static int bridge_rtupdate(struct bridge_softc *, const uint8_t *,
342 uint16_t, struct bridge_iflist *, int, uint8_t);
343 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
344 uint16_t);
345 static void bridge_rttrim(struct bridge_softc *);
346 static void bridge_rtage(struct bridge_softc *);
347 static void bridge_rtflush(struct bridge_softc *, int);
348 static int bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
349 uint16_t);
350
351 static void bridge_rtable_init(struct bridge_softc *);
352 static void bridge_rtable_fini(struct bridge_softc *);
353
354 static int bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
355 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
356 const uint8_t *, uint16_t);
357 static int bridge_rtnode_insert(struct bridge_softc *,
358 struct bridge_rtnode *);
359 static void bridge_rtnode_destroy(struct bridge_softc *,
360 struct bridge_rtnode *);
361 static void bridge_rtable_expire(struct ifnet *, int);
362 static void bridge_state_change(struct ifnet *, int);
363
364 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
365 const char *name);
366 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
367 struct ifnet *ifp);
368 static void bridge_delete_member(struct bridge_softc *,
369 struct bridge_iflist *, int);
370 static void bridge_delete_span(struct bridge_softc *,
371 struct bridge_iflist *);
372
373 static int bridge_ioctl_add(struct bridge_softc *, void *);
374 static int bridge_ioctl_del(struct bridge_softc *, void *);
375 static int bridge_ioctl_gifflags(struct bridge_softc *, void *);
376 static int bridge_ioctl_sifflags(struct bridge_softc *, void *);
377 static int bridge_ioctl_scache(struct bridge_softc *, void *);
378 static int bridge_ioctl_gcache(struct bridge_softc *, void *);
379 static int bridge_ioctl_gifs(struct bridge_softc *, void *);
380 static int bridge_ioctl_rts(struct bridge_softc *, void *);
381 static int bridge_ioctl_saddr(struct bridge_softc *, void *);
382 static int bridge_ioctl_sto(struct bridge_softc *, void *);
383 static int bridge_ioctl_gto(struct bridge_softc *, void *);
384 static int bridge_ioctl_daddr(struct bridge_softc *, void *);
385 static int bridge_ioctl_flush(struct bridge_softc *, void *);
386 static int bridge_ioctl_gpri(struct bridge_softc *, void *);
387 static int bridge_ioctl_spri(struct bridge_softc *, void *);
388 static int bridge_ioctl_ght(struct bridge_softc *, void *);
389 static int bridge_ioctl_sht(struct bridge_softc *, void *);
390 static int bridge_ioctl_gfd(struct bridge_softc *, void *);
391 static int bridge_ioctl_sfd(struct bridge_softc *, void *);
392 static int bridge_ioctl_gma(struct bridge_softc *, void *);
393 static int bridge_ioctl_sma(struct bridge_softc *, void *);
394 static int bridge_ioctl_sifprio(struct bridge_softc *, void *);
395 static int bridge_ioctl_sifcost(struct bridge_softc *, void *);
396 static int bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
397 static int bridge_ioctl_addspan(struct bridge_softc *, void *);
398 static int bridge_ioctl_delspan(struct bridge_softc *, void *);
399 static int bridge_ioctl_gbparam(struct bridge_softc *, void *);
400 static int bridge_ioctl_grte(struct bridge_softc *, void *);
401 static int bridge_ioctl_gifsstp(struct bridge_softc *, void *);
402 static int bridge_ioctl_sproto(struct bridge_softc *, void *);
403 static int bridge_ioctl_stxhc(struct bridge_softc *, void *);
404 static int bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
405 int);
406 #ifdef INET
407 static int bridge_ip_checkbasic(struct mbuf **mp);
408 static int bridge_fragment(struct ifnet *, struct mbuf **mp,
409 struct ether_header *, int, struct llc *);
410 #endif /* INET */
411 #ifdef INET6
412 static int bridge_ip6_checkbasic(struct mbuf **mp);
413 #endif /* INET6 */
414 static void bridge_linkstate(struct ifnet *ifp);
415 static void bridge_linkcheck(struct bridge_softc *sc);
416
417 /*
418 * Use the "null" value from IEEE 802.1Q-2014 Table 9-2
419 * to indicate untagged frames.
420 */
421 #define VLANTAGOF(_m) \
422 (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : DOT1Q_VID_NULL
423
424 static struct bstp_cb_ops bridge_ops = {
425 .bcb_state = bridge_state_change,
426 .bcb_rtage = bridge_rtable_expire
427 };
428
429 SYSCTL_DECL(_net_link);
430 static SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
431 "Bridge");
432
433 /* only pass IP[46] packets when pfil is enabled */
434 VNET_DEFINE_STATIC(int, pfil_onlyip) = 1;
435 #define V_pfil_onlyip VNET(pfil_onlyip)
436 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip,
437 CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_onlyip), 0,
438 "Only pass IP packets when pfil is enabled");
439
440 /* run pfil hooks on the bridge interface */
441 VNET_DEFINE_STATIC(int, pfil_bridge) = 0;
442 #define V_pfil_bridge VNET(pfil_bridge)
443 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge,
444 CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_bridge), 0,
445 "Packet filter on the bridge interface");
446
447 /* layer2 filter with ipfw */
448 VNET_DEFINE_STATIC(int, pfil_ipfw);
449 #define V_pfil_ipfw VNET(pfil_ipfw)
450
451 /* layer2 ARP filter with ipfw */
452 VNET_DEFINE_STATIC(int, pfil_ipfw_arp);
453 #define V_pfil_ipfw_arp VNET(pfil_ipfw_arp)
454 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp,
455 CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_ipfw_arp), 0,
456 "Filter ARP packets through IPFW layer2");
457
458 /* run pfil hooks on the member interface */
459 VNET_DEFINE_STATIC(int, pfil_member) = 0;
460 #define V_pfil_member VNET(pfil_member)
461 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member,
462 CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_member), 0,
463 "Packet filter on the member interface");
464
465 /* run pfil hooks on the physical interface for locally destined packets */
466 VNET_DEFINE_STATIC(int, pfil_local_phys);
467 #define V_pfil_local_phys VNET(pfil_local_phys)
468 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys,
469 CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_local_phys), 0,
470 "Packet filter on the physical interface for locally destined packets");
471
472 /* log STP state changes */
473 VNET_DEFINE_STATIC(int, log_stp);
474 #define V_log_stp VNET(log_stp)
475 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp,
476 CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(log_stp), 0,
477 "Log STP state changes");
478
479 /* share MAC with first bridge member */
480 VNET_DEFINE_STATIC(int, bridge_inherit_mac);
481 #define V_bridge_inherit_mac VNET(bridge_inherit_mac)
482 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac,
483 CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(bridge_inherit_mac), 0,
484 "Inherit MAC address from the first bridge member");
485
486 VNET_DEFINE_STATIC(int, allow_llz_overlap) = 0;
487 #define V_allow_llz_overlap VNET(allow_llz_overlap)
488 SYSCTL_INT(_net_link_bridge, OID_AUTO, allow_llz_overlap,
489 CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(allow_llz_overlap), 0,
490 "Allow overlap of link-local scope "
491 "zones of a bridge interface and the member interfaces");
492
493 /* log MAC address port flapping */
494 VNET_DEFINE_STATIC(bool, log_mac_flap) = true;
495 #define V_log_mac_flap VNET(log_mac_flap)
496 SYSCTL_BOOL(_net_link_bridge, OID_AUTO, log_mac_flap,
497 CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(log_mac_flap), true,
498 "Log MAC address port flapping");
499
500 VNET_DEFINE_STATIC(int, log_interval) = 5;
501 VNET_DEFINE_STATIC(int, log_count) = 0;
502 VNET_DEFINE_STATIC(struct timeval, log_last) = { 0 };
503
504 #define V_log_interval VNET(log_interval)
505 #define V_log_count VNET(log_count)
506 #define V_log_last VNET(log_last)
507
508 struct bridge_control {
509 int (*bc_func)(struct bridge_softc *, void *);
510 int bc_argsize;
511 int bc_flags;
512 };
513
514 #define BC_F_COPYIN 0x01 /* copy arguments in */
515 #define BC_F_COPYOUT 0x02 /* copy arguments out */
516 #define BC_F_SUSER 0x04 /* do super-user check */
517
518 static const struct bridge_control bridge_control_table[] = {
519 { bridge_ioctl_add, sizeof(struct ifbreq),
520 BC_F_COPYIN|BC_F_SUSER },
521 { bridge_ioctl_del, sizeof(struct ifbreq),
522 BC_F_COPYIN|BC_F_SUSER },
523
524 { bridge_ioctl_gifflags, sizeof(struct ifbreq),
525 BC_F_COPYIN|BC_F_COPYOUT },
526 { bridge_ioctl_sifflags, sizeof(struct ifbreq),
527 BC_F_COPYIN|BC_F_SUSER },
528
529 { bridge_ioctl_scache, sizeof(struct ifbrparam),
530 BC_F_COPYIN|BC_F_SUSER },
531 { bridge_ioctl_gcache, sizeof(struct ifbrparam),
532 BC_F_COPYOUT },
533
534 { bridge_ioctl_gifs, sizeof(struct ifbifconf),
535 BC_F_COPYIN|BC_F_COPYOUT },
536 { bridge_ioctl_rts, sizeof(struct ifbaconf),
537 BC_F_COPYIN|BC_F_COPYOUT },
538
539 { bridge_ioctl_saddr, sizeof(struct ifbareq),
540 BC_F_COPYIN|BC_F_SUSER },
541
542 { bridge_ioctl_sto, sizeof(struct ifbrparam),
543 BC_F_COPYIN|BC_F_SUSER },
544 { bridge_ioctl_gto, sizeof(struct ifbrparam),
545 BC_F_COPYOUT },
546
547 { bridge_ioctl_daddr, sizeof(struct ifbareq),
548 BC_F_COPYIN|BC_F_SUSER },
549
550 { bridge_ioctl_flush, sizeof(struct ifbreq),
551 BC_F_COPYIN|BC_F_SUSER },
552
553 { bridge_ioctl_gpri, sizeof(struct ifbrparam),
554 BC_F_COPYOUT },
555 { bridge_ioctl_spri, sizeof(struct ifbrparam),
556 BC_F_COPYIN|BC_F_SUSER },
557
558 { bridge_ioctl_ght, sizeof(struct ifbrparam),
559 BC_F_COPYOUT },
560 { bridge_ioctl_sht, sizeof(struct ifbrparam),
561 BC_F_COPYIN|BC_F_SUSER },
562
563 { bridge_ioctl_gfd, sizeof(struct ifbrparam),
564 BC_F_COPYOUT },
565 { bridge_ioctl_sfd, sizeof(struct ifbrparam),
566 BC_F_COPYIN|BC_F_SUSER },
567
568 { bridge_ioctl_gma, sizeof(struct ifbrparam),
569 BC_F_COPYOUT },
570 { bridge_ioctl_sma, sizeof(struct ifbrparam),
571 BC_F_COPYIN|BC_F_SUSER },
572
573 { bridge_ioctl_sifprio, sizeof(struct ifbreq),
574 BC_F_COPYIN|BC_F_SUSER },
575
576 { bridge_ioctl_sifcost, sizeof(struct ifbreq),
577 BC_F_COPYIN|BC_F_SUSER },
578
579 { bridge_ioctl_addspan, sizeof(struct ifbreq),
580 BC_F_COPYIN|BC_F_SUSER },
581 { bridge_ioctl_delspan, sizeof(struct ifbreq),
582 BC_F_COPYIN|BC_F_SUSER },
583
584 { bridge_ioctl_gbparam, sizeof(struct ifbropreq),
585 BC_F_COPYOUT },
586
587 { bridge_ioctl_grte, sizeof(struct ifbrparam),
588 BC_F_COPYOUT },
589
590 { bridge_ioctl_gifsstp, sizeof(struct ifbpstpconf),
591 BC_F_COPYIN|BC_F_COPYOUT },
592
593 { bridge_ioctl_sproto, sizeof(struct ifbrparam),
594 BC_F_COPYIN|BC_F_SUSER },
595
596 { bridge_ioctl_stxhc, sizeof(struct ifbrparam),
597 BC_F_COPYIN|BC_F_SUSER },
598
599 { bridge_ioctl_sifmaxaddr, sizeof(struct ifbreq),
600 BC_F_COPYIN|BC_F_SUSER },
601
602 };
603 static const int bridge_control_table_size = nitems(bridge_control_table);
604
605 VNET_DEFINE_STATIC(LIST_HEAD(, bridge_softc), bridge_list);
606 #define V_bridge_list VNET(bridge_list)
607 #define BRIDGE_LIST_LOCK_INIT(x) sx_init(&V_bridge_list_sx, \
608 "if_bridge list")
609 #define BRIDGE_LIST_LOCK_DESTROY(x) sx_destroy(&V_bridge_list_sx)
610 #define BRIDGE_LIST_LOCK(x) sx_xlock(&V_bridge_list_sx)
611 #define BRIDGE_LIST_UNLOCK(x) sx_xunlock(&V_bridge_list_sx)
612
613 VNET_DEFINE_STATIC(struct if_clone *, bridge_cloner);
614 #define V_bridge_cloner VNET(bridge_cloner)
615
616 static const char bridge_name[] = "bridge";
617
618 static void
vnet_bridge_init(const void * unused __unused)619 vnet_bridge_init(const void *unused __unused)
620 {
621
622 V_bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
623 sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
624 UMA_ALIGN_PTR, 0);
625 BRIDGE_LIST_LOCK_INIT();
626 LIST_INIT(&V_bridge_list);
627
628 struct if_clone_addreq req = {
629 .create_f = bridge_clone_create,
630 .destroy_f = bridge_clone_destroy,
631 .flags = IFC_F_AUTOUNIT,
632 };
633 V_bridge_cloner = ifc_attach_cloner(bridge_name, &req);
634 }
635 VNET_SYSINIT(vnet_bridge_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
636 vnet_bridge_init, NULL);
637
638 static void
vnet_bridge_uninit(const void * unused __unused)639 vnet_bridge_uninit(const void *unused __unused)
640 {
641
642 ifc_detach_cloner(V_bridge_cloner);
643 V_bridge_cloner = NULL;
644 BRIDGE_LIST_LOCK_DESTROY();
645
646 /* Callbacks may use the UMA zone. */
647 NET_EPOCH_DRAIN_CALLBACKS();
648
649 uma_zdestroy(V_bridge_rtnode_zone);
650 }
651 VNET_SYSUNINIT(vnet_bridge_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY,
652 vnet_bridge_uninit, NULL);
653
654 static int
bridge_modevent(module_t mod,int type,void * data)655 bridge_modevent(module_t mod, int type, void *data)
656 {
657
658 switch (type) {
659 case MOD_LOAD:
660 bridge_dn_p = bridge_dummynet;
661 bridge_detach_cookie = EVENTHANDLER_REGISTER(
662 ifnet_departure_event, bridge_ifdetach, NULL,
663 EVENTHANDLER_PRI_ANY);
664 break;
665 case MOD_UNLOAD:
666 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
667 bridge_detach_cookie);
668 bridge_dn_p = NULL;
669 break;
670 default:
671 return (EOPNOTSUPP);
672 }
673 return (0);
674 }
675
676 static moduledata_t bridge_mod = {
677 "if_bridge",
678 bridge_modevent,
679 0
680 };
681
682 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
683 MODULE_VERSION(if_bridge, 1);
684 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
685
686 /*
687 * handler for net.link.bridge.ipfw
688 */
689 static int
sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)690 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
691 {
692 int enable = V_pfil_ipfw;
693 int error;
694
695 error = sysctl_handle_int(oidp, &enable, 0, req);
696 enable &= 1;
697
698 if (enable != V_pfil_ipfw) {
699 V_pfil_ipfw = enable;
700
701 /*
702 * Disable pfil so that ipfw doesnt run twice, if the user
703 * really wants both then they can re-enable pfil_bridge and/or
704 * pfil_member. Also allow non-ip packets as ipfw can filter by
705 * layer2 type.
706 */
707 if (V_pfil_ipfw) {
708 V_pfil_onlyip = 0;
709 V_pfil_bridge = 0;
710 V_pfil_member = 0;
711 }
712 }
713
714 return (error);
715 }
716 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw,
717 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_VNET | CTLFLAG_NEEDGIANT,
718 &VNET_NAME(pfil_ipfw), 0, &sysctl_pfil_ipfw, "I",
719 "Layer2 filter with IPFW");
720
721 #ifdef VIMAGE
722 static void
bridge_reassign(struct ifnet * ifp,struct vnet * newvnet,char * arg)723 bridge_reassign(struct ifnet *ifp, struct vnet *newvnet, char *arg)
724 {
725 struct bridge_softc *sc = ifp->if_softc;
726 struct bridge_iflist *bif;
727
728 BRIDGE_LOCK(sc);
729
730 while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
731 bridge_delete_member(sc, bif, 0);
732
733 while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
734 bridge_delete_span(sc, bif);
735 }
736
737 BRIDGE_UNLOCK(sc);
738
739 ether_reassign(ifp, newvnet, arg);
740 }
741 #endif
742
743 /*
744 * bridge_clone_create:
745 *
746 * Create a new bridge instance.
747 */
748 static int
bridge_clone_create(struct if_clone * ifc,char * name,size_t len,struct ifc_data * ifd,struct ifnet ** ifpp)749 bridge_clone_create(struct if_clone *ifc, char *name, size_t len,
750 struct ifc_data *ifd, struct ifnet **ifpp)
751 {
752 struct bridge_softc *sc;
753 struct ifnet *ifp;
754
755 sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
756 ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
757
758 BRIDGE_LOCK_INIT(sc);
759 sc->sc_brtmax = BRIDGE_RTABLE_MAX;
760 sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
761
762 /* Initialize our routing table. */
763 bridge_rtable_init(sc);
764
765 callout_init_mtx(&sc->sc_brcallout, &sc->sc_rt_mtx, 0);
766
767 CK_LIST_INIT(&sc->sc_iflist);
768 CK_LIST_INIT(&sc->sc_spanlist);
769
770 ifp->if_softc = sc;
771 if_initname(ifp, bridge_name, ifd->unit);
772 ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
773 ifp->if_ioctl = bridge_ioctl;
774 #ifdef ALTQ
775 ifp->if_start = bridge_altq_start;
776 ifp->if_transmit = bridge_altq_transmit;
777 IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
778 ifp->if_snd.ifq_drv_maxlen = 0;
779 IFQ_SET_READY(&ifp->if_snd);
780 #else
781 ifp->if_transmit = bridge_transmit;
782 #endif
783 ifp->if_qflush = bridge_qflush;
784 ifp->if_init = bridge_init;
785 ifp->if_type = IFT_BRIDGE;
786
787 ether_gen_addr(ifp, &sc->sc_defaddr);
788
789 bstp_attach(&sc->sc_stp, &bridge_ops);
790 ether_ifattach(ifp, sc->sc_defaddr.octet);
791 /* Now undo some of the damage... */
792 ifp->if_baudrate = 0;
793 ifp->if_type = IFT_BRIDGE;
794 #ifdef VIMAGE
795 ifp->if_reassign = bridge_reassign;
796 #endif
797 sc->sc_if_input = ifp->if_input; /* ether_input */
798 ifp->if_input = bridge_inject;
799
800 /*
801 * Allow BRIDGE_INPUT() to pass in packets originating from the bridge
802 * itself via bridge_inject(). This is required for netmap but
803 * otherwise has no effect.
804 */
805 ifp->if_bridge_input = bridge_input;
806
807 BRIDGE_LIST_LOCK();
808 LIST_INSERT_HEAD(&V_bridge_list, sc, sc_list);
809 BRIDGE_LIST_UNLOCK();
810 *ifpp = ifp;
811
812 return (0);
813 }
814
815 static void
bridge_clone_destroy_cb(struct epoch_context * ctx)816 bridge_clone_destroy_cb(struct epoch_context *ctx)
817 {
818 struct bridge_softc *sc;
819
820 sc = __containerof(ctx, struct bridge_softc, sc_epoch_ctx);
821
822 BRIDGE_LOCK_DESTROY(sc);
823 free(sc, M_DEVBUF);
824 }
825
826 /*
827 * bridge_clone_destroy:
828 *
829 * Destroy a bridge instance.
830 */
831 static int
bridge_clone_destroy(struct if_clone * ifc,struct ifnet * ifp,uint32_t flags)832 bridge_clone_destroy(struct if_clone *ifc, struct ifnet *ifp, uint32_t flags)
833 {
834 struct bridge_softc *sc = ifp->if_softc;
835 struct bridge_iflist *bif;
836 struct epoch_tracker et;
837
838 BRIDGE_LOCK(sc);
839
840 bridge_stop(ifp, 1);
841 ifp->if_flags &= ~IFF_UP;
842
843 while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
844 bridge_delete_member(sc, bif, 0);
845
846 while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
847 bridge_delete_span(sc, bif);
848 }
849
850 /* Tear down the routing table. */
851 bridge_rtable_fini(sc);
852
853 BRIDGE_UNLOCK(sc);
854
855 NET_EPOCH_ENTER(et);
856
857 callout_drain(&sc->sc_brcallout);
858
859 BRIDGE_LIST_LOCK();
860 LIST_REMOVE(sc, sc_list);
861 BRIDGE_LIST_UNLOCK();
862
863 bstp_detach(&sc->sc_stp);
864 #ifdef ALTQ
865 IFQ_PURGE(&ifp->if_snd);
866 #endif
867 NET_EPOCH_EXIT(et);
868
869 ether_ifdetach(ifp);
870 if_free(ifp);
871
872 NET_EPOCH_CALL(bridge_clone_destroy_cb, &sc->sc_epoch_ctx);
873
874 return (0);
875 }
876
877 /*
878 * bridge_ioctl:
879 *
880 * Handle a control request from the operator.
881 */
882 static int
bridge_ioctl(struct ifnet * ifp,u_long cmd,caddr_t data)883 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
884 {
885 struct bridge_softc *sc = ifp->if_softc;
886 struct ifreq *ifr = (struct ifreq *)data;
887 struct bridge_iflist *bif;
888 struct thread *td = curthread;
889 union {
890 struct ifbreq ifbreq;
891 struct ifbifconf ifbifconf;
892 struct ifbareq ifbareq;
893 struct ifbaconf ifbaconf;
894 struct ifbrparam ifbrparam;
895 struct ifbropreq ifbropreq;
896 } args;
897 struct ifdrv *ifd = (struct ifdrv *) data;
898 const struct bridge_control *bc;
899 int error = 0, oldmtu;
900
901 BRIDGE_LOCK(sc);
902
903 switch (cmd) {
904 case SIOCADDMULTI:
905 case SIOCDELMULTI:
906 break;
907
908 case SIOCGDRVSPEC:
909 case SIOCSDRVSPEC:
910 if (ifd->ifd_cmd >= bridge_control_table_size) {
911 error = EINVAL;
912 break;
913 }
914 bc = &bridge_control_table[ifd->ifd_cmd];
915
916 if (cmd == SIOCGDRVSPEC &&
917 (bc->bc_flags & BC_F_COPYOUT) == 0) {
918 error = EINVAL;
919 break;
920 }
921 else if (cmd == SIOCSDRVSPEC &&
922 (bc->bc_flags & BC_F_COPYOUT) != 0) {
923 error = EINVAL;
924 break;
925 }
926
927 if (bc->bc_flags & BC_F_SUSER) {
928 error = priv_check(td, PRIV_NET_BRIDGE);
929 if (error)
930 break;
931 }
932
933 if (ifd->ifd_len != bc->bc_argsize ||
934 ifd->ifd_len > sizeof(args)) {
935 error = EINVAL;
936 break;
937 }
938
939 bzero(&args, sizeof(args));
940 if (bc->bc_flags & BC_F_COPYIN) {
941 error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
942 if (error)
943 break;
944 }
945
946 oldmtu = ifp->if_mtu;
947 error = (*bc->bc_func)(sc, &args);
948 if (error)
949 break;
950
951 /*
952 * Bridge MTU may change during addition of the first port.
953 * If it did, do network layer specific procedure.
954 */
955 if (ifp->if_mtu != oldmtu)
956 if_notifymtu(ifp);
957
958 if (bc->bc_flags & BC_F_COPYOUT)
959 error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
960
961 break;
962
963 case SIOCSIFFLAGS:
964 if (!(ifp->if_flags & IFF_UP) &&
965 (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
966 /*
967 * If interface is marked down and it is running,
968 * then stop and disable it.
969 */
970 bridge_stop(ifp, 1);
971 } else if ((ifp->if_flags & IFF_UP) &&
972 !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
973 /*
974 * If interface is marked up and it is stopped, then
975 * start it.
976 */
977 BRIDGE_UNLOCK(sc);
978 (*ifp->if_init)(sc);
979 BRIDGE_LOCK(sc);
980 }
981 break;
982
983 case SIOCSIFMTU:
984 oldmtu = sc->sc_ifp->if_mtu;
985
986 if (ifr->ifr_mtu < IF_MINMTU) {
987 error = EINVAL;
988 break;
989 }
990 if (CK_LIST_EMPTY(&sc->sc_iflist)) {
991 sc->sc_ifp->if_mtu = ifr->ifr_mtu;
992 break;
993 }
994 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
995 error = (*bif->bif_ifp->if_ioctl)(bif->bif_ifp,
996 SIOCSIFMTU, (caddr_t)ifr);
997 if (error != 0) {
998 log(LOG_NOTICE, "%s: invalid MTU: %u for"
999 " member %s\n", sc->sc_ifp->if_xname,
1000 ifr->ifr_mtu,
1001 bif->bif_ifp->if_xname);
1002 error = EINVAL;
1003 break;
1004 }
1005 }
1006 if (error) {
1007 /* Restore the previous MTU on all member interfaces. */
1008 ifr->ifr_mtu = oldmtu;
1009 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1010 (*bif->bif_ifp->if_ioctl)(bif->bif_ifp,
1011 SIOCSIFMTU, (caddr_t)ifr);
1012 }
1013 } else {
1014 sc->sc_ifp->if_mtu = ifr->ifr_mtu;
1015 }
1016 break;
1017 default:
1018 /*
1019 * drop the lock as ether_ioctl() will call bridge_start() and
1020 * cause the lock to be recursed.
1021 */
1022 BRIDGE_UNLOCK(sc);
1023 error = ether_ioctl(ifp, cmd, data);
1024 BRIDGE_LOCK(sc);
1025 break;
1026 }
1027
1028 BRIDGE_UNLOCK(sc);
1029
1030 return (error);
1031 }
1032
1033 /*
1034 * bridge_mutecaps:
1035 *
1036 * Clear or restore unwanted capabilities on the member interface
1037 */
1038 static void
bridge_mutecaps(struct bridge_softc * sc)1039 bridge_mutecaps(struct bridge_softc *sc)
1040 {
1041 struct bridge_iflist *bif;
1042 int enabled, mask;
1043
1044 BRIDGE_LOCK_ASSERT(sc);
1045
1046 /* Initial bitmask of capabilities to test */
1047 mask = BRIDGE_IFCAPS_MASK;
1048
1049 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1050 /* Every member must support it or its disabled */
1051 mask &= bif->bif_savedcaps;
1052 }
1053
1054 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1055 enabled = bif->bif_ifp->if_capenable;
1056 enabled &= ~BRIDGE_IFCAPS_STRIP;
1057 /* strip off mask bits and enable them again if allowed */
1058 enabled &= ~BRIDGE_IFCAPS_MASK;
1059 enabled |= mask;
1060 bridge_set_ifcap(sc, bif, enabled);
1061 }
1062 }
1063
1064 static void
bridge_set_ifcap(struct bridge_softc * sc,struct bridge_iflist * bif,int set)1065 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
1066 {
1067 struct ifnet *ifp = bif->bif_ifp;
1068 struct ifreq ifr;
1069 int error, mask, stuck;
1070
1071 bzero(&ifr, sizeof(ifr));
1072 ifr.ifr_reqcap = set;
1073
1074 if (ifp->if_capenable != set) {
1075 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
1076 if (error)
1077 if_printf(sc->sc_ifp,
1078 "error setting capabilities on %s: %d\n",
1079 ifp->if_xname, error);
1080 mask = BRIDGE_IFCAPS_MASK | BRIDGE_IFCAPS_STRIP;
1081 stuck = ifp->if_capenable & mask & ~set;
1082 if (stuck != 0)
1083 if_printf(sc->sc_ifp,
1084 "can't disable some capabilities on %s: 0x%x\n",
1085 ifp->if_xname, stuck);
1086 }
1087 }
1088
1089 /*
1090 * bridge_lookup_member:
1091 *
1092 * Lookup a bridge member interface.
1093 */
1094 static struct bridge_iflist *
bridge_lookup_member(struct bridge_softc * sc,const char * name)1095 bridge_lookup_member(struct bridge_softc *sc, const char *name)
1096 {
1097 struct bridge_iflist *bif;
1098 struct ifnet *ifp;
1099
1100 BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1101
1102 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1103 ifp = bif->bif_ifp;
1104 if (strcmp(ifp->if_xname, name) == 0)
1105 return (bif);
1106 }
1107
1108 return (NULL);
1109 }
1110
1111 /*
1112 * bridge_lookup_member_if:
1113 *
1114 * Lookup a bridge member interface by ifnet*.
1115 */
1116 static struct bridge_iflist *
bridge_lookup_member_if(struct bridge_softc * sc,struct ifnet * member_ifp)1117 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
1118 {
1119 struct bridge_iflist *bif;
1120
1121 BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1122
1123 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1124 if (bif->bif_ifp == member_ifp)
1125 return (bif);
1126 }
1127
1128 return (NULL);
1129 }
1130
1131 static void
bridge_delete_member_cb(struct epoch_context * ctx)1132 bridge_delete_member_cb(struct epoch_context *ctx)
1133 {
1134 struct bridge_iflist *bif;
1135
1136 bif = __containerof(ctx, struct bridge_iflist, bif_epoch_ctx);
1137
1138 free(bif, M_DEVBUF);
1139 }
1140
1141 /*
1142 * bridge_delete_member:
1143 *
1144 * Delete the specified member interface.
1145 */
1146 static void
bridge_delete_member(struct bridge_softc * sc,struct bridge_iflist * bif,int gone)1147 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
1148 int gone)
1149 {
1150 struct ifnet *ifs = bif->bif_ifp;
1151 struct ifnet *fif = NULL;
1152 struct bridge_iflist *bifl;
1153
1154 BRIDGE_LOCK_ASSERT(sc);
1155
1156 if (bif->bif_flags & IFBIF_STP)
1157 bstp_disable(&bif->bif_stp);
1158
1159 ifs->if_bridge = NULL;
1160 CK_LIST_REMOVE(bif, bif_next);
1161
1162 /*
1163 * If removing the interface that gave the bridge its mac address, set
1164 * the mac address of the bridge to the address of the next member, or
1165 * to its default address if no members are left.
1166 */
1167 if (V_bridge_inherit_mac && sc->sc_ifaddr == ifs) {
1168 if (CK_LIST_EMPTY(&sc->sc_iflist)) {
1169 bcopy(&sc->sc_defaddr,
1170 IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1171 sc->sc_ifaddr = NULL;
1172 } else {
1173 bifl = CK_LIST_FIRST(&sc->sc_iflist);
1174 fif = bifl->bif_ifp;
1175 bcopy(IF_LLADDR(fif),
1176 IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1177 sc->sc_ifaddr = fif;
1178 }
1179 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1180 }
1181
1182 bridge_linkcheck(sc);
1183 bridge_mutecaps(sc); /* recalcuate now this interface is removed */
1184 BRIDGE_RT_LOCK(sc);
1185 bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
1186 BRIDGE_RT_UNLOCK(sc);
1187 KASSERT(bif->bif_addrcnt == 0,
1188 ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
1189
1190 ifs->if_bridge_output = NULL;
1191 ifs->if_bridge_input = NULL;
1192 ifs->if_bridge_linkstate = NULL;
1193 if (!gone) {
1194 switch (ifs->if_type) {
1195 case IFT_ETHER:
1196 case IFT_L2VLAN:
1197 /*
1198 * Take the interface out of promiscuous mode, but only
1199 * if it was promiscuous in the first place. It might
1200 * not be if we're in the bridge_ioctl_add() error path.
1201 */
1202 if (ifs->if_flags & IFF_PROMISC)
1203 (void) ifpromisc(ifs, 0);
1204 break;
1205
1206 case IFT_GIF:
1207 break;
1208
1209 default:
1210 #ifdef DIAGNOSTIC
1211 panic("bridge_delete_member: impossible");
1212 #endif
1213 break;
1214 }
1215 /* reneable any interface capabilities */
1216 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
1217 }
1218 bstp_destroy(&bif->bif_stp); /* prepare to free */
1219
1220 NET_EPOCH_CALL(bridge_delete_member_cb, &bif->bif_epoch_ctx);
1221 }
1222
1223 /*
1224 * bridge_delete_span:
1225 *
1226 * Delete the specified span interface.
1227 */
1228 static void
bridge_delete_span(struct bridge_softc * sc,struct bridge_iflist * bif)1229 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1230 {
1231 BRIDGE_LOCK_ASSERT(sc);
1232
1233 KASSERT(bif->bif_ifp->if_bridge == NULL,
1234 ("%s: not a span interface", __func__));
1235
1236 CK_LIST_REMOVE(bif, bif_next);
1237
1238 NET_EPOCH_CALL(bridge_delete_member_cb, &bif->bif_epoch_ctx);
1239 }
1240
1241 static int
bridge_ioctl_add(struct bridge_softc * sc,void * arg)1242 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1243 {
1244 struct ifbreq *req = arg;
1245 struct bridge_iflist *bif = NULL;
1246 struct ifnet *ifs;
1247 int error = 0;
1248
1249 ifs = ifunit(req->ifbr_ifsname);
1250 if (ifs == NULL)
1251 return (ENOENT);
1252 if (ifs->if_ioctl == NULL) /* must be supported */
1253 return (EINVAL);
1254
1255 /* If it's in the span list, it can't be a member. */
1256 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1257 if (ifs == bif->bif_ifp)
1258 return (EBUSY);
1259
1260 if (ifs->if_bridge == sc)
1261 return (EEXIST);
1262
1263 if (ifs->if_bridge != NULL)
1264 return (EBUSY);
1265
1266 switch (ifs->if_type) {
1267 case IFT_ETHER:
1268 case IFT_L2VLAN:
1269 case IFT_GIF:
1270 /* permitted interface types */
1271 break;
1272 default:
1273 return (EINVAL);
1274 }
1275
1276 #ifdef INET6
1277 /*
1278 * Two valid inet6 addresses with link-local scope must not be
1279 * on the parent interface and the member interfaces at the
1280 * same time. This restriction is needed to prevent violation
1281 * of link-local scope zone. Attempts to add a member
1282 * interface which has inet6 addresses when the parent has
1283 * inet6 triggers removal of all inet6 addresses on the member
1284 * interface.
1285 */
1286
1287 /* Check if the parent interface has a link-local scope addr. */
1288 if (V_allow_llz_overlap == 0 &&
1289 in6ifa_llaonifp(sc->sc_ifp) != NULL) {
1290 /*
1291 * If any, remove all inet6 addresses from the member
1292 * interfaces.
1293 */
1294 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1295 if (in6ifa_llaonifp(bif->bif_ifp)) {
1296 in6_ifdetach(bif->bif_ifp);
1297 if_printf(sc->sc_ifp,
1298 "IPv6 addresses on %s have been removed "
1299 "before adding it as a member to prevent "
1300 "IPv6 address scope violation.\n",
1301 bif->bif_ifp->if_xname);
1302 }
1303 }
1304 if (in6ifa_llaonifp(ifs)) {
1305 in6_ifdetach(ifs);
1306 if_printf(sc->sc_ifp,
1307 "IPv6 addresses on %s have been removed "
1308 "before adding it as a member to prevent "
1309 "IPv6 address scope violation.\n",
1310 ifs->if_xname);
1311 }
1312 }
1313 #endif
1314 /* Allow the first Ethernet member to define the MTU */
1315 if (CK_LIST_EMPTY(&sc->sc_iflist))
1316 sc->sc_ifp->if_mtu = ifs->if_mtu;
1317 else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1318 struct ifreq ifr;
1319
1320 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "%s",
1321 ifs->if_xname);
1322 ifr.ifr_mtu = sc->sc_ifp->if_mtu;
1323
1324 error = (*ifs->if_ioctl)(ifs,
1325 SIOCSIFMTU, (caddr_t)&ifr);
1326 if (error != 0) {
1327 log(LOG_NOTICE, "%s: invalid MTU: %u for"
1328 " new member %s\n", sc->sc_ifp->if_xname,
1329 ifr.ifr_mtu,
1330 ifs->if_xname);
1331 return (EINVAL);
1332 }
1333 }
1334
1335 bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1336 if (bif == NULL)
1337 return (ENOMEM);
1338
1339 bif->bif_ifp = ifs;
1340 bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1341 bif->bif_savedcaps = ifs->if_capenable;
1342
1343 /*
1344 * Assign the interface's MAC address to the bridge if it's the first
1345 * member and the MAC address of the bridge has not been changed from
1346 * the default randomly generated one.
1347 */
1348 if (V_bridge_inherit_mac && CK_LIST_EMPTY(&sc->sc_iflist) &&
1349 !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr.octet, ETHER_ADDR_LEN)) {
1350 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1351 sc->sc_ifaddr = ifs;
1352 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1353 }
1354
1355 ifs->if_bridge = sc;
1356 ifs->if_bridge_output = bridge_output;
1357 ifs->if_bridge_input = bridge_input;
1358 ifs->if_bridge_linkstate = bridge_linkstate;
1359 bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1360 /*
1361 * XXX: XLOCK HERE!?!
1362 *
1363 * NOTE: insert_***HEAD*** should be safe for the traversals.
1364 */
1365 CK_LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1366
1367 /* Set interface capabilities to the intersection set of all members */
1368 bridge_mutecaps(sc);
1369 bridge_linkcheck(sc);
1370
1371 /* Place the interface into promiscuous mode */
1372 switch (ifs->if_type) {
1373 case IFT_ETHER:
1374 case IFT_L2VLAN:
1375 error = ifpromisc(ifs, 1);
1376 break;
1377 }
1378
1379 if (error)
1380 bridge_delete_member(sc, bif, 0);
1381 return (error);
1382 }
1383
1384 static int
bridge_ioctl_del(struct bridge_softc * sc,void * arg)1385 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1386 {
1387 struct ifbreq *req = arg;
1388 struct bridge_iflist *bif;
1389
1390 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1391 if (bif == NULL)
1392 return (ENOENT);
1393
1394 bridge_delete_member(sc, bif, 0);
1395
1396 return (0);
1397 }
1398
1399 static int
bridge_ioctl_gifflags(struct bridge_softc * sc,void * arg)1400 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1401 {
1402 struct ifbreq *req = arg;
1403 struct bridge_iflist *bif;
1404 struct bstp_port *bp;
1405
1406 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1407 if (bif == NULL)
1408 return (ENOENT);
1409
1410 bp = &bif->bif_stp;
1411 req->ifbr_ifsflags = bif->bif_flags;
1412 req->ifbr_state = bp->bp_state;
1413 req->ifbr_priority = bp->bp_priority;
1414 req->ifbr_path_cost = bp->bp_path_cost;
1415 req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1416 req->ifbr_proto = bp->bp_protover;
1417 req->ifbr_role = bp->bp_role;
1418 req->ifbr_stpflags = bp->bp_flags;
1419 req->ifbr_addrcnt = bif->bif_addrcnt;
1420 req->ifbr_addrmax = bif->bif_addrmax;
1421 req->ifbr_addrexceeded = bif->bif_addrexceeded;
1422
1423 /* Copy STP state options as flags */
1424 if (bp->bp_operedge)
1425 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1426 if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1427 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1428 if (bp->bp_ptp_link)
1429 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1430 if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1431 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1432 if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1433 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1434 if (bp->bp_flags & BSTP_PORT_ADMCOST)
1435 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1436 return (0);
1437 }
1438
1439 static int
bridge_ioctl_sifflags(struct bridge_softc * sc,void * arg)1440 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1441 {
1442 struct epoch_tracker et;
1443 struct ifbreq *req = arg;
1444 struct bridge_iflist *bif;
1445 struct bstp_port *bp;
1446 int error;
1447
1448 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1449 if (bif == NULL)
1450 return (ENOENT);
1451 bp = &bif->bif_stp;
1452
1453 if (req->ifbr_ifsflags & IFBIF_SPAN)
1454 /* SPAN is readonly */
1455 return (EINVAL);
1456
1457 NET_EPOCH_ENTER(et);
1458
1459 if (req->ifbr_ifsflags & IFBIF_STP) {
1460 if ((bif->bif_flags & IFBIF_STP) == 0) {
1461 error = bstp_enable(&bif->bif_stp);
1462 if (error) {
1463 NET_EPOCH_EXIT(et);
1464 return (error);
1465 }
1466 }
1467 } else {
1468 if ((bif->bif_flags & IFBIF_STP) != 0)
1469 bstp_disable(&bif->bif_stp);
1470 }
1471
1472 /* Pass on STP flags */
1473 bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1474 bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1475 bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1476 bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1477
1478 /* Save the bits relating to the bridge */
1479 bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1480
1481 NET_EPOCH_EXIT(et);
1482
1483 return (0);
1484 }
1485
1486 static int
bridge_ioctl_scache(struct bridge_softc * sc,void * arg)1487 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1488 {
1489 struct ifbrparam *param = arg;
1490
1491 sc->sc_brtmax = param->ifbrp_csize;
1492 bridge_rttrim(sc);
1493
1494 return (0);
1495 }
1496
1497 static int
bridge_ioctl_gcache(struct bridge_softc * sc,void * arg)1498 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1499 {
1500 struct ifbrparam *param = arg;
1501
1502 param->ifbrp_csize = sc->sc_brtmax;
1503
1504 return (0);
1505 }
1506
1507 static int
bridge_ioctl_gifs(struct bridge_softc * sc,void * arg)1508 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1509 {
1510 struct ifbifconf *bifc = arg;
1511 struct bridge_iflist *bif;
1512 struct ifbreq breq;
1513 char *buf, *outbuf;
1514 int count, buflen, len, error = 0;
1515
1516 count = 0;
1517 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1518 count++;
1519 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1520 count++;
1521
1522 buflen = sizeof(breq) * count;
1523 if (bifc->ifbic_len == 0) {
1524 bifc->ifbic_len = buflen;
1525 return (0);
1526 }
1527 outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1528 if (outbuf == NULL)
1529 return (ENOMEM);
1530
1531 count = 0;
1532 buf = outbuf;
1533 len = min(bifc->ifbic_len, buflen);
1534 bzero(&breq, sizeof(breq));
1535 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1536 if (len < sizeof(breq))
1537 break;
1538
1539 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1540 sizeof(breq.ifbr_ifsname));
1541 /* Fill in the ifbreq structure */
1542 error = bridge_ioctl_gifflags(sc, &breq);
1543 if (error)
1544 break;
1545 memcpy(buf, &breq, sizeof(breq));
1546 count++;
1547 buf += sizeof(breq);
1548 len -= sizeof(breq);
1549 }
1550 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1551 if (len < sizeof(breq))
1552 break;
1553
1554 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1555 sizeof(breq.ifbr_ifsname));
1556 breq.ifbr_ifsflags = bif->bif_flags;
1557 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1558 memcpy(buf, &breq, sizeof(breq));
1559 count++;
1560 buf += sizeof(breq);
1561 len -= sizeof(breq);
1562 }
1563
1564 bifc->ifbic_len = sizeof(breq) * count;
1565 error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1566 free(outbuf, M_TEMP);
1567 return (error);
1568 }
1569
1570 static int
bridge_ioctl_rts(struct bridge_softc * sc,void * arg)1571 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1572 {
1573 struct ifbaconf *bac = arg;
1574 struct bridge_rtnode *brt;
1575 struct ifbareq bareq;
1576 char *buf, *outbuf;
1577 int count, buflen, len, error = 0;
1578
1579 if (bac->ifbac_len == 0)
1580 return (0);
1581
1582 count = 0;
1583 CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1584 count++;
1585 buflen = sizeof(bareq) * count;
1586
1587 outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1588 if (outbuf == NULL)
1589 return (ENOMEM);
1590
1591 count = 0;
1592 buf = outbuf;
1593 len = min(bac->ifbac_len, buflen);
1594 bzero(&bareq, sizeof(bareq));
1595 CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1596 if (len < sizeof(bareq))
1597 goto out;
1598 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1599 sizeof(bareq.ifba_ifsname));
1600 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1601 bareq.ifba_vlan = brt->brt_vlan;
1602 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1603 time_uptime < brt->brt_expire)
1604 bareq.ifba_expire = brt->brt_expire - time_uptime;
1605 else
1606 bareq.ifba_expire = 0;
1607 bareq.ifba_flags = brt->brt_flags;
1608
1609 memcpy(buf, &bareq, sizeof(bareq));
1610 count++;
1611 buf += sizeof(bareq);
1612 len -= sizeof(bareq);
1613 }
1614 out:
1615 bac->ifbac_len = sizeof(bareq) * count;
1616 error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1617 free(outbuf, M_TEMP);
1618 return (error);
1619 }
1620
1621 static int
bridge_ioctl_saddr(struct bridge_softc * sc,void * arg)1622 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1623 {
1624 struct ifbareq *req = arg;
1625 struct bridge_iflist *bif;
1626 struct epoch_tracker et;
1627 int error;
1628
1629 NET_EPOCH_ENTER(et);
1630 bif = bridge_lookup_member(sc, req->ifba_ifsname);
1631 if (bif == NULL) {
1632 NET_EPOCH_EXIT(et);
1633 return (ENOENT);
1634 }
1635
1636 /* bridge_rtupdate() may acquire the lock. */
1637 error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1638 req->ifba_flags);
1639 NET_EPOCH_EXIT(et);
1640
1641 return (error);
1642 }
1643
1644 static int
bridge_ioctl_sto(struct bridge_softc * sc,void * arg)1645 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1646 {
1647 struct ifbrparam *param = arg;
1648
1649 sc->sc_brttimeout = param->ifbrp_ctime;
1650 return (0);
1651 }
1652
1653 static int
bridge_ioctl_gto(struct bridge_softc * sc,void * arg)1654 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1655 {
1656 struct ifbrparam *param = arg;
1657
1658 param->ifbrp_ctime = sc->sc_brttimeout;
1659 return (0);
1660 }
1661
1662 static int
bridge_ioctl_daddr(struct bridge_softc * sc,void * arg)1663 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1664 {
1665 struct ifbareq *req = arg;
1666 int vlan = req->ifba_vlan;
1667
1668 /* Userspace uses '0' to mean 'any vlan' */
1669 if (vlan == 0)
1670 vlan = DOT1Q_VID_RSVD_IMPL;
1671
1672 return (bridge_rtdaddr(sc, req->ifba_dst, vlan));
1673 }
1674
1675 static int
bridge_ioctl_flush(struct bridge_softc * sc,void * arg)1676 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1677 {
1678 struct ifbreq *req = arg;
1679
1680 BRIDGE_RT_LOCK(sc);
1681 bridge_rtflush(sc, req->ifbr_ifsflags);
1682 BRIDGE_RT_UNLOCK(sc);
1683
1684 return (0);
1685 }
1686
1687 static int
bridge_ioctl_gpri(struct bridge_softc * sc,void * arg)1688 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1689 {
1690 struct ifbrparam *param = arg;
1691 struct bstp_state *bs = &sc->sc_stp;
1692
1693 param->ifbrp_prio = bs->bs_bridge_priority;
1694 return (0);
1695 }
1696
1697 static int
bridge_ioctl_spri(struct bridge_softc * sc,void * arg)1698 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1699 {
1700 struct ifbrparam *param = arg;
1701
1702 return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1703 }
1704
1705 static int
bridge_ioctl_ght(struct bridge_softc * sc,void * arg)1706 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1707 {
1708 struct ifbrparam *param = arg;
1709 struct bstp_state *bs = &sc->sc_stp;
1710
1711 param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1712 return (0);
1713 }
1714
1715 static int
bridge_ioctl_sht(struct bridge_softc * sc,void * arg)1716 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1717 {
1718 struct ifbrparam *param = arg;
1719
1720 return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1721 }
1722
1723 static int
bridge_ioctl_gfd(struct bridge_softc * sc,void * arg)1724 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1725 {
1726 struct ifbrparam *param = arg;
1727 struct bstp_state *bs = &sc->sc_stp;
1728
1729 param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1730 return (0);
1731 }
1732
1733 static int
bridge_ioctl_sfd(struct bridge_softc * sc,void * arg)1734 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1735 {
1736 struct ifbrparam *param = arg;
1737
1738 return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1739 }
1740
1741 static int
bridge_ioctl_gma(struct bridge_softc * sc,void * arg)1742 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1743 {
1744 struct ifbrparam *param = arg;
1745 struct bstp_state *bs = &sc->sc_stp;
1746
1747 param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1748 return (0);
1749 }
1750
1751 static int
bridge_ioctl_sma(struct bridge_softc * sc,void * arg)1752 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1753 {
1754 struct ifbrparam *param = arg;
1755
1756 return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1757 }
1758
1759 static int
bridge_ioctl_sifprio(struct bridge_softc * sc,void * arg)1760 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1761 {
1762 struct ifbreq *req = arg;
1763 struct bridge_iflist *bif;
1764
1765 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1766 if (bif == NULL)
1767 return (ENOENT);
1768
1769 return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1770 }
1771
1772 static int
bridge_ioctl_sifcost(struct bridge_softc * sc,void * arg)1773 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1774 {
1775 struct ifbreq *req = arg;
1776 struct bridge_iflist *bif;
1777
1778 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1779 if (bif == NULL)
1780 return (ENOENT);
1781
1782 return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1783 }
1784
1785 static int
bridge_ioctl_sifmaxaddr(struct bridge_softc * sc,void * arg)1786 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1787 {
1788 struct ifbreq *req = arg;
1789 struct bridge_iflist *bif;
1790
1791 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1792 if (bif == NULL)
1793 return (ENOENT);
1794
1795 bif->bif_addrmax = req->ifbr_addrmax;
1796 return (0);
1797 }
1798
1799 static int
bridge_ioctl_addspan(struct bridge_softc * sc,void * arg)1800 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1801 {
1802 struct ifbreq *req = arg;
1803 struct bridge_iflist *bif = NULL;
1804 struct ifnet *ifs;
1805
1806 ifs = ifunit(req->ifbr_ifsname);
1807 if (ifs == NULL)
1808 return (ENOENT);
1809
1810 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1811 if (ifs == bif->bif_ifp)
1812 return (EBUSY);
1813
1814 if (ifs->if_bridge != NULL)
1815 return (EBUSY);
1816
1817 switch (ifs->if_type) {
1818 case IFT_ETHER:
1819 case IFT_GIF:
1820 case IFT_L2VLAN:
1821 break;
1822 default:
1823 return (EINVAL);
1824 }
1825
1826 bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1827 if (bif == NULL)
1828 return (ENOMEM);
1829
1830 bif->bif_ifp = ifs;
1831 bif->bif_flags = IFBIF_SPAN;
1832
1833 CK_LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1834
1835 return (0);
1836 }
1837
1838 static int
bridge_ioctl_delspan(struct bridge_softc * sc,void * arg)1839 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1840 {
1841 struct ifbreq *req = arg;
1842 struct bridge_iflist *bif;
1843 struct ifnet *ifs;
1844
1845 ifs = ifunit(req->ifbr_ifsname);
1846 if (ifs == NULL)
1847 return (ENOENT);
1848
1849 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1850 if (ifs == bif->bif_ifp)
1851 break;
1852
1853 if (bif == NULL)
1854 return (ENOENT);
1855
1856 bridge_delete_span(sc, bif);
1857
1858 return (0);
1859 }
1860
1861 static int
bridge_ioctl_gbparam(struct bridge_softc * sc,void * arg)1862 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1863 {
1864 struct ifbropreq *req = arg;
1865 struct bstp_state *bs = &sc->sc_stp;
1866 struct bstp_port *root_port;
1867
1868 req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1869 req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1870 req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1871
1872 root_port = bs->bs_root_port;
1873 if (root_port == NULL)
1874 req->ifbop_root_port = 0;
1875 else
1876 req->ifbop_root_port = root_port->bp_ifp->if_index;
1877
1878 req->ifbop_holdcount = bs->bs_txholdcount;
1879 req->ifbop_priority = bs->bs_bridge_priority;
1880 req->ifbop_protocol = bs->bs_protover;
1881 req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1882 req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1883 req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1884 req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1885 req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1886 req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1887
1888 return (0);
1889 }
1890
1891 static int
bridge_ioctl_grte(struct bridge_softc * sc,void * arg)1892 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1893 {
1894 struct ifbrparam *param = arg;
1895
1896 param->ifbrp_cexceeded = sc->sc_brtexceeded;
1897 return (0);
1898 }
1899
1900 static int
bridge_ioctl_gifsstp(struct bridge_softc * sc,void * arg)1901 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1902 {
1903 struct ifbpstpconf *bifstp = arg;
1904 struct bridge_iflist *bif;
1905 struct bstp_port *bp;
1906 struct ifbpstpreq bpreq;
1907 char *buf, *outbuf;
1908 int count, buflen, len, error = 0;
1909
1910 count = 0;
1911 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1912 if ((bif->bif_flags & IFBIF_STP) != 0)
1913 count++;
1914 }
1915
1916 buflen = sizeof(bpreq) * count;
1917 if (bifstp->ifbpstp_len == 0) {
1918 bifstp->ifbpstp_len = buflen;
1919 return (0);
1920 }
1921
1922 outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1923 if (outbuf == NULL)
1924 return (ENOMEM);
1925
1926 count = 0;
1927 buf = outbuf;
1928 len = min(bifstp->ifbpstp_len, buflen);
1929 bzero(&bpreq, sizeof(bpreq));
1930 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1931 if (len < sizeof(bpreq))
1932 break;
1933
1934 if ((bif->bif_flags & IFBIF_STP) == 0)
1935 continue;
1936
1937 bp = &bif->bif_stp;
1938 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1939 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1940 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1941 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1942 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1943 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1944
1945 memcpy(buf, &bpreq, sizeof(bpreq));
1946 count++;
1947 buf += sizeof(bpreq);
1948 len -= sizeof(bpreq);
1949 }
1950
1951 bifstp->ifbpstp_len = sizeof(bpreq) * count;
1952 error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1953 free(outbuf, M_TEMP);
1954 return (error);
1955 }
1956
1957 static int
bridge_ioctl_sproto(struct bridge_softc * sc,void * arg)1958 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1959 {
1960 struct ifbrparam *param = arg;
1961
1962 return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1963 }
1964
1965 static int
bridge_ioctl_stxhc(struct bridge_softc * sc,void * arg)1966 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1967 {
1968 struct ifbrparam *param = arg;
1969
1970 return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1971 }
1972
1973 /*
1974 * bridge_ifdetach:
1975 *
1976 * Detach an interface from a bridge. Called when a member
1977 * interface is detaching.
1978 */
1979 static void
bridge_ifdetach(void * arg __unused,struct ifnet * ifp)1980 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1981 {
1982 struct bridge_softc *sc = ifp->if_bridge;
1983 struct bridge_iflist *bif;
1984
1985 if (ifp->if_flags & IFF_RENAMING)
1986 return;
1987 if (V_bridge_cloner == NULL) {
1988 /*
1989 * This detach handler can be called after
1990 * vnet_bridge_uninit(). Just return in that case.
1991 */
1992 return;
1993 }
1994 /* Check if the interface is a bridge member */
1995 if (sc != NULL) {
1996 BRIDGE_LOCK(sc);
1997
1998 bif = bridge_lookup_member_if(sc, ifp);
1999 if (bif != NULL)
2000 bridge_delete_member(sc, bif, 1);
2001
2002 BRIDGE_UNLOCK(sc);
2003 return;
2004 }
2005
2006 /* Check if the interface is a span port */
2007 BRIDGE_LIST_LOCK();
2008 LIST_FOREACH(sc, &V_bridge_list, sc_list) {
2009 BRIDGE_LOCK(sc);
2010 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
2011 if (ifp == bif->bif_ifp) {
2012 bridge_delete_span(sc, bif);
2013 break;
2014 }
2015
2016 BRIDGE_UNLOCK(sc);
2017 }
2018 BRIDGE_LIST_UNLOCK();
2019 }
2020
2021 /*
2022 * bridge_init:
2023 *
2024 * Initialize a bridge interface.
2025 */
2026 static void
bridge_init(void * xsc)2027 bridge_init(void *xsc)
2028 {
2029 struct bridge_softc *sc = (struct bridge_softc *)xsc;
2030 struct ifnet *ifp = sc->sc_ifp;
2031
2032 if (ifp->if_drv_flags & IFF_DRV_RUNNING)
2033 return;
2034
2035 BRIDGE_LOCK(sc);
2036 callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
2037 bridge_timer, sc);
2038
2039 ifp->if_drv_flags |= IFF_DRV_RUNNING;
2040 bstp_init(&sc->sc_stp); /* Initialize Spanning Tree */
2041
2042 BRIDGE_UNLOCK(sc);
2043 }
2044
2045 /*
2046 * bridge_stop:
2047 *
2048 * Stop the bridge interface.
2049 */
2050 static void
bridge_stop(struct ifnet * ifp,int disable)2051 bridge_stop(struct ifnet *ifp, int disable)
2052 {
2053 struct bridge_softc *sc = ifp->if_softc;
2054
2055 BRIDGE_LOCK_ASSERT(sc);
2056
2057 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2058 return;
2059
2060 BRIDGE_RT_LOCK(sc);
2061 callout_stop(&sc->sc_brcallout);
2062
2063 bstp_stop(&sc->sc_stp);
2064
2065 bridge_rtflush(sc, IFBF_FLUSHDYN);
2066 BRIDGE_RT_UNLOCK(sc);
2067
2068 ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
2069 }
2070
2071 /*
2072 * bridge_enqueue:
2073 *
2074 * Enqueue a packet on a bridge member interface.
2075 *
2076 */
2077 static int
bridge_enqueue(struct bridge_softc * sc,struct ifnet * dst_ifp,struct mbuf * m)2078 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
2079 {
2080 int len, err = 0;
2081 short mflags;
2082 struct mbuf *m0;
2083
2084 /* We may be sending a fragment so traverse the mbuf */
2085 for (; m; m = m0) {
2086 m0 = m->m_nextpkt;
2087 m->m_nextpkt = NULL;
2088 len = m->m_pkthdr.len;
2089 mflags = m->m_flags;
2090
2091 /*
2092 * If underlying interface can not do VLAN tag insertion itself
2093 * then attach a packet tag that holds it.
2094 */
2095 if ((m->m_flags & M_VLANTAG) &&
2096 (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
2097 m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
2098 if (m == NULL) {
2099 if_printf(dst_ifp,
2100 "unable to prepend VLAN header\n");
2101 if_inc_counter(dst_ifp, IFCOUNTER_OERRORS, 1);
2102 continue;
2103 }
2104 m->m_flags &= ~M_VLANTAG;
2105 }
2106
2107 M_ASSERTPKTHDR(m); /* We shouldn't transmit mbuf without pkthdr */
2108 if ((err = dst_ifp->if_transmit(dst_ifp, m))) {
2109 int n;
2110
2111 for (m = m0, n = 1; m != NULL; m = m0, n++) {
2112 m0 = m->m_nextpkt;
2113 m_freem(m);
2114 }
2115 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, n);
2116 break;
2117 }
2118
2119 if_inc_counter(sc->sc_ifp, IFCOUNTER_OPACKETS, 1);
2120 if_inc_counter(sc->sc_ifp, IFCOUNTER_OBYTES, len);
2121 if (mflags & M_MCAST)
2122 if_inc_counter(sc->sc_ifp, IFCOUNTER_OMCASTS, 1);
2123 }
2124
2125 return (err);
2126 }
2127
2128 /*
2129 * bridge_dummynet:
2130 *
2131 * Receive a queued packet from dummynet and pass it on to the output
2132 * interface.
2133 *
2134 * The mbuf has the Ethernet header already attached.
2135 */
2136 static void
bridge_dummynet(struct mbuf * m,struct ifnet * ifp)2137 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
2138 {
2139 struct bridge_softc *sc;
2140
2141 sc = ifp->if_bridge;
2142
2143 /*
2144 * The packet didnt originate from a member interface. This should only
2145 * ever happen if a member interface is removed while packets are
2146 * queued for it.
2147 */
2148 if (sc == NULL) {
2149 m_freem(m);
2150 return;
2151 }
2152
2153 if (PFIL_HOOKED_OUT_46) {
2154 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
2155 return;
2156 if (m == NULL)
2157 return;
2158 }
2159
2160 bridge_enqueue(sc, ifp, m);
2161 }
2162
2163 /*
2164 * bridge_output:
2165 *
2166 * Send output from a bridge member interface. This
2167 * performs the bridging function for locally originated
2168 * packets.
2169 *
2170 * The mbuf has the Ethernet header already attached. We must
2171 * enqueue or free the mbuf before returning.
2172 */
2173 static int
bridge_output(struct ifnet * ifp,struct mbuf * m,struct sockaddr * sa,struct rtentry * rt)2174 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
2175 struct rtentry *rt)
2176 {
2177 struct ether_header *eh;
2178 struct ifnet *bifp, *dst_if;
2179 struct bridge_softc *sc;
2180 uint16_t vlan;
2181
2182 NET_EPOCH_ASSERT();
2183
2184 if (m->m_len < ETHER_HDR_LEN) {
2185 m = m_pullup(m, ETHER_HDR_LEN);
2186 if (m == NULL)
2187 return (0);
2188 }
2189
2190 eh = mtod(m, struct ether_header *);
2191 sc = ifp->if_bridge;
2192 vlan = VLANTAGOF(m);
2193
2194 bifp = sc->sc_ifp;
2195
2196 /*
2197 * If bridge is down, but the original output interface is up,
2198 * go ahead and send out that interface. Otherwise, the packet
2199 * is dropped below.
2200 */
2201 if ((bifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2202 dst_if = ifp;
2203 goto sendunicast;
2204 }
2205
2206 /*
2207 * If the packet is a multicast, or we don't know a better way to
2208 * get there, send to all interfaces.
2209 */
2210 if (ETHER_IS_MULTICAST(eh->ether_dhost))
2211 dst_if = NULL;
2212 else
2213 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
2214 /* Tap any traffic not passing back out the originating interface */
2215 if (dst_if != ifp)
2216 ETHER_BPF_MTAP(bifp, m);
2217 if (dst_if == NULL) {
2218 struct bridge_iflist *bif;
2219 struct mbuf *mc;
2220 int used = 0;
2221
2222 bridge_span(sc, m);
2223
2224 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
2225 dst_if = bif->bif_ifp;
2226
2227 if (dst_if->if_type == IFT_GIF)
2228 continue;
2229 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2230 continue;
2231
2232 /*
2233 * If this is not the original output interface,
2234 * and the interface is participating in spanning
2235 * tree, make sure the port is in a state that
2236 * allows forwarding.
2237 */
2238 if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
2239 bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2240 continue;
2241
2242 if (CK_LIST_NEXT(bif, bif_next) == NULL) {
2243 used = 1;
2244 mc = m;
2245 } else {
2246 mc = m_dup(m, M_NOWAIT);
2247 if (mc == NULL) {
2248 if_inc_counter(bifp, IFCOUNTER_OERRORS, 1);
2249 continue;
2250 }
2251 }
2252
2253 bridge_enqueue(sc, dst_if, mc);
2254 }
2255 if (used == 0)
2256 m_freem(m);
2257 return (0);
2258 }
2259
2260 sendunicast:
2261 /*
2262 * XXX Spanning tree consideration here?
2263 */
2264
2265 bridge_span(sc, m);
2266 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2267 m_freem(m);
2268 return (0);
2269 }
2270
2271 bridge_enqueue(sc, dst_if, m);
2272 return (0);
2273 }
2274
2275 /*
2276 * bridge_transmit:
2277 *
2278 * Do output on a bridge.
2279 *
2280 */
2281 static int
bridge_transmit(struct ifnet * ifp,struct mbuf * m)2282 bridge_transmit(struct ifnet *ifp, struct mbuf *m)
2283 {
2284 struct bridge_softc *sc;
2285 struct ether_header *eh;
2286 struct ifnet *dst_if;
2287 int error = 0;
2288
2289 sc = ifp->if_softc;
2290
2291 ETHER_BPF_MTAP(ifp, m);
2292
2293 eh = mtod(m, struct ether_header *);
2294
2295 if (((m->m_flags & (M_BCAST|M_MCAST)) == 0) &&
2296 (dst_if = bridge_rtlookup(sc, eh->ether_dhost, DOT1Q_VID_NULL)) !=
2297 NULL) {
2298 error = bridge_enqueue(sc, dst_if, m);
2299 } else
2300 bridge_broadcast(sc, ifp, m, 0);
2301
2302 return (error);
2303 }
2304
2305 #ifdef ALTQ
2306 static void
bridge_altq_start(if_t ifp)2307 bridge_altq_start(if_t ifp)
2308 {
2309 struct ifaltq *ifq = &ifp->if_snd;
2310 struct mbuf *m;
2311
2312 IFQ_LOCK(ifq);
2313 IFQ_DEQUEUE_NOLOCK(ifq, m);
2314 while (m != NULL) {
2315 bridge_transmit(ifp, m);
2316 IFQ_DEQUEUE_NOLOCK(ifq, m);
2317 }
2318 IFQ_UNLOCK(ifq);
2319 }
2320
2321 static int
bridge_altq_transmit(if_t ifp,struct mbuf * m)2322 bridge_altq_transmit(if_t ifp, struct mbuf *m)
2323 {
2324 int err;
2325
2326 if (ALTQ_IS_ENABLED(&ifp->if_snd)) {
2327 IFQ_ENQUEUE(&ifp->if_snd, m, err);
2328 if (err == 0)
2329 bridge_altq_start(ifp);
2330 } else
2331 err = bridge_transmit(ifp, m);
2332
2333 return (err);
2334 }
2335 #endif /* ALTQ */
2336
2337 /*
2338 * The ifp->if_qflush entry point for if_bridge(4) is no-op.
2339 */
2340 static void
bridge_qflush(struct ifnet * ifp __unused)2341 bridge_qflush(struct ifnet *ifp __unused)
2342 {
2343 }
2344
2345 /*
2346 * bridge_forward:
2347 *
2348 * The forwarding function of the bridge.
2349 *
2350 * NOTE: Releases the lock on return.
2351 */
2352 static void
bridge_forward(struct bridge_softc * sc,struct bridge_iflist * sbif,struct mbuf * m)2353 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2354 struct mbuf *m)
2355 {
2356 struct bridge_iflist *dbif;
2357 struct ifnet *src_if, *dst_if, *ifp;
2358 struct ether_header *eh;
2359 uint16_t vlan;
2360 uint8_t *dst;
2361 int error;
2362
2363 NET_EPOCH_ASSERT();
2364
2365 src_if = m->m_pkthdr.rcvif;
2366 ifp = sc->sc_ifp;
2367
2368 if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
2369 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2370 vlan = VLANTAGOF(m);
2371
2372 if ((sbif->bif_flags & IFBIF_STP) &&
2373 sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2374 goto drop;
2375
2376 eh = mtod(m, struct ether_header *);
2377 dst = eh->ether_dhost;
2378
2379 /* If the interface is learning, record the address. */
2380 if (sbif->bif_flags & IFBIF_LEARNING) {
2381 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2382 sbif, 0, IFBAF_DYNAMIC);
2383 /*
2384 * If the interface has addresses limits then deny any source
2385 * that is not in the cache.
2386 */
2387 if (error && sbif->bif_addrmax)
2388 goto drop;
2389 }
2390
2391 if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2392 sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2393 goto drop;
2394
2395 #ifdef DEV_NETMAP
2396 /*
2397 * Hand the packet to netmap only if it wasn't injected by netmap
2398 * itself.
2399 */
2400 if ((m->m_flags & M_BRIDGE_INJECT) == 0 &&
2401 (if_getcapenable(ifp) & IFCAP_NETMAP) != 0) {
2402 ifp->if_input(ifp, m);
2403 return;
2404 }
2405 m->m_flags &= ~M_BRIDGE_INJECT;
2406 #endif
2407
2408 /*
2409 * At this point, the port either doesn't participate
2410 * in spanning tree or it is in the forwarding state.
2411 */
2412
2413 /*
2414 * If the packet is unicast, destined for someone on
2415 * "this" side of the bridge, drop it.
2416 */
2417 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2418 dst_if = bridge_rtlookup(sc, dst, vlan);
2419 if (src_if == dst_if)
2420 goto drop;
2421 } else {
2422 /*
2423 * Check if its a reserved multicast address, any address
2424 * listed in 802.1D section 7.12.6 may not be forwarded by the
2425 * bridge.
2426 * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2427 */
2428 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2429 dst[2] == 0xc2 && dst[3] == 0x00 &&
2430 dst[4] == 0x00 && dst[5] <= 0x0f)
2431 goto drop;
2432
2433 /* ...forward it to all interfaces. */
2434 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
2435 dst_if = NULL;
2436 }
2437
2438 /*
2439 * If we have a destination interface which is a member of our bridge,
2440 * OR this is a unicast packet, push it through the bpf(4) machinery.
2441 * For broadcast or multicast packets, don't bother because it will
2442 * be reinjected into ether_input. We do this before we pass the packets
2443 * through the pfil(9) framework, as it is possible that pfil(9) will
2444 * drop the packet, or possibly modify it, making it difficult to debug
2445 * firewall issues on the bridge.
2446 */
2447 if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2448 ETHER_BPF_MTAP(ifp, m);
2449
2450 /* run the packet filter */
2451 if (PFIL_HOOKED_IN_46) {
2452 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2453 return;
2454 if (m == NULL)
2455 return;
2456 }
2457
2458 if (dst_if == NULL) {
2459 bridge_broadcast(sc, src_if, m, 1);
2460 return;
2461 }
2462
2463 /*
2464 * At this point, we're dealing with a unicast frame
2465 * going to a different interface.
2466 */
2467 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2468 goto drop;
2469
2470 dbif = bridge_lookup_member_if(sc, dst_if);
2471 if (dbif == NULL)
2472 /* Not a member of the bridge (anymore?) */
2473 goto drop;
2474
2475 /* Private segments can not talk to each other */
2476 if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2477 goto drop;
2478
2479 if ((dbif->bif_flags & IFBIF_STP) &&
2480 dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2481 goto drop;
2482
2483 if (PFIL_HOOKED_OUT_46) {
2484 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2485 return;
2486 if (m == NULL)
2487 return;
2488 }
2489
2490 bridge_enqueue(sc, dst_if, m);
2491 return;
2492
2493 drop:
2494 m_freem(m);
2495 }
2496
2497 /*
2498 * bridge_input:
2499 *
2500 * Receive input from a member interface. Queue the packet for
2501 * bridging if it is not for us.
2502 */
2503 static struct mbuf *
bridge_input(struct ifnet * ifp,struct mbuf * m)2504 bridge_input(struct ifnet *ifp, struct mbuf *m)
2505 {
2506 struct bridge_softc *sc;
2507 struct bridge_iflist *bif, *bif2;
2508 struct ifnet *bifp;
2509 struct ether_header *eh;
2510 struct mbuf *mc, *mc2;
2511 uint16_t vlan;
2512 int error;
2513
2514 NET_EPOCH_ASSERT();
2515
2516 eh = mtod(m, struct ether_header *);
2517 vlan = VLANTAGOF(m);
2518
2519 sc = ifp->if_bridge;
2520 if (sc == NULL) {
2521 /*
2522 * This packet originated from the bridge itself, so it must
2523 * have been transmitted by netmap. Derive the "source"
2524 * interface from the source address and drop the packet if the
2525 * source address isn't known.
2526 */
2527 KASSERT((m->m_flags & M_BRIDGE_INJECT) != 0,
2528 ("%s: ifnet %p missing a bridge softc", __func__, ifp));
2529 sc = if_getsoftc(ifp);
2530 ifp = bridge_rtlookup(sc, eh->ether_shost, vlan);
2531 if (ifp == NULL) {
2532 if_inc_counter(sc->sc_ifp, IFCOUNTER_IERRORS, 1);
2533 m_freem(m);
2534 return (NULL);
2535 }
2536 m->m_pkthdr.rcvif = ifp;
2537 }
2538 bifp = sc->sc_ifp;
2539 if ((bifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2540 return (m);
2541
2542 /*
2543 * Implement support for bridge monitoring. If this flag has been
2544 * set on this interface, discard the packet once we push it through
2545 * the bpf(4) machinery, but before we do, increment the byte and
2546 * packet counters associated with this interface.
2547 */
2548 if ((bifp->if_flags & IFF_MONITOR) != 0) {
2549 m->m_pkthdr.rcvif = bifp;
2550 ETHER_BPF_MTAP(bifp, m);
2551 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);
2552 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2553 m_freem(m);
2554 return (NULL);
2555 }
2556 bif = bridge_lookup_member_if(sc, ifp);
2557 if (bif == NULL) {
2558 return (m);
2559 }
2560
2561 bridge_span(sc, m);
2562
2563 if (m->m_flags & (M_BCAST|M_MCAST)) {
2564 /* Tap off 802.1D packets; they do not get forwarded. */
2565 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2566 ETHER_ADDR_LEN) == 0) {
2567 bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2568 return (NULL);
2569 }
2570
2571 if ((bif->bif_flags & IFBIF_STP) &&
2572 bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2573 return (m);
2574 }
2575
2576 /*
2577 * Make a deep copy of the packet and enqueue the copy
2578 * for bridge processing; return the original packet for
2579 * local processing.
2580 */
2581 mc = m_dup(m, M_NOWAIT);
2582 if (mc == NULL) {
2583 return (m);
2584 }
2585
2586 /* Perform the bridge forwarding function with the copy. */
2587 bridge_forward(sc, bif, mc);
2588
2589 #ifdef DEV_NETMAP
2590 /*
2591 * If netmap is enabled and has not already seen this packet,
2592 * then it will be consumed by bridge_forward().
2593 */
2594 if ((if_getcapenable(bifp) & IFCAP_NETMAP) != 0 &&
2595 (m->m_flags & M_BRIDGE_INJECT) == 0) {
2596 m_freem(m);
2597 return (NULL);
2598 }
2599 #endif
2600
2601 /*
2602 * Reinject the mbuf as arriving on the bridge so we have a
2603 * chance at claiming multicast packets. We can not loop back
2604 * here from ether_input as a bridge is never a member of a
2605 * bridge.
2606 */
2607 KASSERT(bifp->if_bridge == NULL,
2608 ("loop created in bridge_input"));
2609 mc2 = m_dup(m, M_NOWAIT);
2610 if (mc2 != NULL) {
2611 /* Keep the layer3 header aligned */
2612 int i = min(mc2->m_pkthdr.len, max_protohdr);
2613 mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2614 }
2615 if (mc2 != NULL) {
2616 mc2->m_pkthdr.rcvif = bifp;
2617 mc2->m_flags &= ~M_BRIDGE_INJECT;
2618 sc->sc_if_input(bifp, mc2);
2619 }
2620
2621 /* Return the original packet for local processing. */
2622 return (m);
2623 }
2624
2625 if ((bif->bif_flags & IFBIF_STP) &&
2626 bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2627 return (m);
2628 }
2629
2630 #if defined(INET) || defined(INET6)
2631 #define CARP_CHECK_WE_ARE_DST(iface) \
2632 ((iface)->if_carp && (*carp_forus_p)((iface), eh->ether_dhost))
2633 #define CARP_CHECK_WE_ARE_SRC(iface) \
2634 ((iface)->if_carp && (*carp_forus_p)((iface), eh->ether_shost))
2635 #else
2636 #define CARP_CHECK_WE_ARE_DST(iface) false
2637 #define CARP_CHECK_WE_ARE_SRC(iface) false
2638 #endif
2639
2640 #ifdef DEV_NETMAP
2641 #define GRAB_FOR_NETMAP(ifp, m) do { \
2642 if ((if_getcapenable(ifp) & IFCAP_NETMAP) != 0 && \
2643 ((m)->m_flags & M_BRIDGE_INJECT) == 0) { \
2644 (ifp)->if_input(ifp, m); \
2645 return (NULL); \
2646 } \
2647 } while (0)
2648 #else
2649 #define GRAB_FOR_NETMAP(ifp, m)
2650 #endif
2651
2652 #define GRAB_OUR_PACKETS(iface) \
2653 if ((iface)->if_type == IFT_GIF) \
2654 continue; \
2655 /* It is destined for us. */ \
2656 if (memcmp(IF_LLADDR(iface), eh->ether_dhost, ETHER_ADDR_LEN) == 0 || \
2657 CARP_CHECK_WE_ARE_DST(iface)) { \
2658 if (bif->bif_flags & IFBIF_LEARNING) { \
2659 error = bridge_rtupdate(sc, eh->ether_shost, \
2660 vlan, bif, 0, IFBAF_DYNAMIC); \
2661 if (error && bif->bif_addrmax) { \
2662 m_freem(m); \
2663 return (NULL); \
2664 } \
2665 } \
2666 m->m_pkthdr.rcvif = iface; \
2667 if ((iface) == ifp) { \
2668 /* Skip bridge processing... src == dest */ \
2669 return (m); \
2670 } \
2671 /* It's passing over or to the bridge, locally. */ \
2672 ETHER_BPF_MTAP(bifp, m); \
2673 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1); \
2674 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);\
2675 /* Hand the packet over to netmap if necessary. */ \
2676 GRAB_FOR_NETMAP(bifp, m); \
2677 /* Filter on the physical interface. */ \
2678 if (V_pfil_local_phys && PFIL_HOOKED_IN_46) { \
2679 if (bridge_pfil(&m, NULL, ifp, \
2680 PFIL_IN) != 0 || m == NULL) { \
2681 return (NULL); \
2682 } \
2683 } \
2684 if ((iface) != bifp) \
2685 ETHER_BPF_MTAP(iface, m); \
2686 return (m); \
2687 } \
2688 \
2689 /* We just received a packet that we sent out. */ \
2690 if (memcmp(IF_LLADDR(iface), eh->ether_shost, ETHER_ADDR_LEN) == 0 || \
2691 CARP_CHECK_WE_ARE_SRC(iface)) { \
2692 m_freem(m); \
2693 return (NULL); \
2694 }
2695
2696 /*
2697 * Unicast. Make sure it's not for the bridge.
2698 */
2699 do { GRAB_OUR_PACKETS(bifp) } while (0);
2700
2701 /*
2702 * Give a chance for ifp at first priority. This will help when the
2703 * packet comes through the interface like VLAN's with the same MACs
2704 * on several interfaces from the same bridge. This also will save
2705 * some CPU cycles in case the destination interface and the input
2706 * interface (eq ifp) are the same.
2707 */
2708 do { GRAB_OUR_PACKETS(ifp) } while (0);
2709
2710 /* Now check the all bridge members. */
2711 CK_LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2712 GRAB_OUR_PACKETS(bif2->bif_ifp)
2713 }
2714
2715 #undef CARP_CHECK_WE_ARE_DST
2716 #undef CARP_CHECK_WE_ARE_SRC
2717 #undef GRAB_FOR_NETMAP
2718 #undef GRAB_OUR_PACKETS
2719
2720 /* Perform the bridge forwarding function. */
2721 bridge_forward(sc, bif, m);
2722
2723 return (NULL);
2724 }
2725
2726 /*
2727 * Inject a packet back into the host ethernet stack. This will generally only
2728 * be used by netmap when an application writes to the host TX ring. The
2729 * M_BRIDGE_INJECT flag ensures that the packet is re-routed to the bridge
2730 * interface after ethernet processing.
2731 */
2732 static void
bridge_inject(struct ifnet * ifp,struct mbuf * m)2733 bridge_inject(struct ifnet *ifp, struct mbuf *m)
2734 {
2735 struct bridge_softc *sc;
2736
2737 KASSERT((if_getcapenable(ifp) & IFCAP_NETMAP) != 0,
2738 ("%s: iface %s is not running in netmap mode",
2739 __func__, if_name(ifp)));
2740 KASSERT((m->m_flags & M_BRIDGE_INJECT) == 0,
2741 ("%s: mbuf %p has M_BRIDGE_INJECT set", __func__, m));
2742
2743 m->m_flags |= M_BRIDGE_INJECT;
2744 sc = if_getsoftc(ifp);
2745 sc->sc_if_input(ifp, m);
2746 }
2747
2748 /*
2749 * bridge_broadcast:
2750 *
2751 * Send a frame to all interfaces that are members of
2752 * the bridge, except for the one on which the packet
2753 * arrived.
2754 *
2755 * NOTE: Releases the lock on return.
2756 */
2757 static void
bridge_broadcast(struct bridge_softc * sc,struct ifnet * src_if,struct mbuf * m,int runfilt)2758 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2759 struct mbuf *m, int runfilt)
2760 {
2761 struct bridge_iflist *dbif, *sbif;
2762 struct mbuf *mc;
2763 struct ifnet *dst_if;
2764 int used = 0, i;
2765
2766 NET_EPOCH_ASSERT();
2767
2768 sbif = bridge_lookup_member_if(sc, src_if);
2769
2770 /* Filter on the bridge interface before broadcasting */
2771 if (runfilt && PFIL_HOOKED_OUT_46) {
2772 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2773 return;
2774 if (m == NULL)
2775 return;
2776 }
2777
2778 CK_LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2779 dst_if = dbif->bif_ifp;
2780 if (dst_if == src_if)
2781 continue;
2782
2783 /* Private segments can not talk to each other */
2784 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2785 continue;
2786
2787 if ((dbif->bif_flags & IFBIF_STP) &&
2788 dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2789 continue;
2790
2791 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2792 (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2793 continue;
2794
2795 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2796 continue;
2797
2798 if (CK_LIST_NEXT(dbif, bif_next) == NULL) {
2799 mc = m;
2800 used = 1;
2801 } else {
2802 mc = m_dup(m, M_NOWAIT);
2803 if (mc == NULL) {
2804 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2805 continue;
2806 }
2807 }
2808
2809 /*
2810 * Filter on the output interface. Pass a NULL bridge interface
2811 * pointer so we do not redundantly filter on the bridge for
2812 * each interface we broadcast on.
2813 */
2814 if (runfilt && PFIL_HOOKED_OUT_46) {
2815 if (used == 0) {
2816 /* Keep the layer3 header aligned */
2817 i = min(mc->m_pkthdr.len, max_protohdr);
2818 mc = m_copyup(mc, i, ETHER_ALIGN);
2819 if (mc == NULL) {
2820 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2821 continue;
2822 }
2823 }
2824 if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2825 continue;
2826 if (mc == NULL)
2827 continue;
2828 }
2829
2830 bridge_enqueue(sc, dst_if, mc);
2831 }
2832 if (used == 0)
2833 m_freem(m);
2834 }
2835
2836 /*
2837 * bridge_span:
2838 *
2839 * Duplicate a packet out one or more interfaces that are in span mode,
2840 * the original mbuf is unmodified.
2841 */
2842 static void
bridge_span(struct bridge_softc * sc,struct mbuf * m)2843 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2844 {
2845 struct bridge_iflist *bif;
2846 struct ifnet *dst_if;
2847 struct mbuf *mc;
2848
2849 NET_EPOCH_ASSERT();
2850
2851 if (CK_LIST_EMPTY(&sc->sc_spanlist))
2852 return;
2853
2854 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2855 dst_if = bif->bif_ifp;
2856
2857 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2858 continue;
2859
2860 mc = m_dup(m, M_NOWAIT);
2861 if (mc == NULL) {
2862 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2863 continue;
2864 }
2865
2866 bridge_enqueue(sc, dst_if, mc);
2867 }
2868 }
2869
2870 /*
2871 * bridge_rtupdate:
2872 *
2873 * Add a bridge routing entry.
2874 */
2875 static int
bridge_rtupdate(struct bridge_softc * sc,const uint8_t * dst,uint16_t vlan,struct bridge_iflist * bif,int setflags,uint8_t flags)2876 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2877 struct bridge_iflist *bif, int setflags, uint8_t flags)
2878 {
2879 struct bridge_rtnode *brt;
2880 struct bridge_iflist *obif;
2881 int error;
2882
2883 BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
2884
2885 /* Check the source address is valid and not multicast. */
2886 if (ETHER_IS_MULTICAST(dst) ||
2887 (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2888 dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2889 return (EINVAL);
2890
2891 /*
2892 * A route for this destination might already exist. If so,
2893 * update it, otherwise create a new one.
2894 */
2895 if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2896 BRIDGE_RT_LOCK(sc);
2897
2898 /* Check again, now that we have the lock. There could have
2899 * been a race and we only want to insert this once. */
2900 if (bridge_rtnode_lookup(sc, dst, vlan) != NULL) {
2901 BRIDGE_RT_UNLOCK(sc);
2902 return (0);
2903 }
2904
2905 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2906 sc->sc_brtexceeded++;
2907 BRIDGE_RT_UNLOCK(sc);
2908 return (ENOSPC);
2909 }
2910 /* Check per interface address limits (if enabled) */
2911 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2912 bif->bif_addrexceeded++;
2913 BRIDGE_RT_UNLOCK(sc);
2914 return (ENOSPC);
2915 }
2916
2917 /*
2918 * Allocate a new bridge forwarding node, and
2919 * initialize the expiration time and Ethernet
2920 * address.
2921 */
2922 brt = uma_zalloc(V_bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2923 if (brt == NULL) {
2924 BRIDGE_RT_UNLOCK(sc);
2925 return (ENOMEM);
2926 }
2927 brt->brt_vnet = curvnet;
2928
2929 if (bif->bif_flags & IFBIF_STICKY)
2930 brt->brt_flags = IFBAF_STICKY;
2931 else
2932 brt->brt_flags = IFBAF_DYNAMIC;
2933
2934 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2935 brt->brt_vlan = vlan;
2936
2937 brt->brt_dst = bif;
2938 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2939 uma_zfree(V_bridge_rtnode_zone, brt);
2940 BRIDGE_RT_UNLOCK(sc);
2941 return (error);
2942 }
2943 bif->bif_addrcnt++;
2944
2945 BRIDGE_RT_UNLOCK(sc);
2946 }
2947
2948 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2949 (obif = brt->brt_dst) != bif) {
2950 MPASS(obif != NULL);
2951
2952 BRIDGE_RT_LOCK(sc);
2953 brt->brt_dst->bif_addrcnt--;
2954 brt->brt_dst = bif;
2955 brt->brt_dst->bif_addrcnt++;
2956 BRIDGE_RT_UNLOCK(sc);
2957
2958 if (V_log_mac_flap &&
2959 ppsratecheck(&V_log_last, &V_log_count, V_log_interval)) {
2960 log(LOG_NOTICE,
2961 "%s: mac address %6D vlan %d moved from %s to %s\n",
2962 sc->sc_ifp->if_xname,
2963 &brt->brt_addr[0], ":",
2964 brt->brt_vlan,
2965 obif->bif_ifp->if_xname,
2966 bif->bif_ifp->if_xname);
2967 }
2968 }
2969
2970 if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2971 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2972 if (setflags)
2973 brt->brt_flags = flags;
2974
2975 return (0);
2976 }
2977
2978 /*
2979 * bridge_rtlookup:
2980 *
2981 * Lookup the destination interface for an address.
2982 */
2983 static struct ifnet *
bridge_rtlookup(struct bridge_softc * sc,const uint8_t * addr,uint16_t vlan)2984 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2985 {
2986 struct bridge_rtnode *brt;
2987
2988 NET_EPOCH_ASSERT();
2989
2990 if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2991 return (NULL);
2992
2993 return (brt->brt_ifp);
2994 }
2995
2996 /*
2997 * bridge_rttrim:
2998 *
2999 * Trim the routine table so that we have a number
3000 * of routing entries less than or equal to the
3001 * maximum number.
3002 */
3003 static void
bridge_rttrim(struct bridge_softc * sc)3004 bridge_rttrim(struct bridge_softc *sc)
3005 {
3006 struct bridge_rtnode *brt, *nbrt;
3007
3008 NET_EPOCH_ASSERT();
3009 BRIDGE_RT_LOCK_ASSERT(sc);
3010
3011 /* Make sure we actually need to do this. */
3012 if (sc->sc_brtcnt <= sc->sc_brtmax)
3013 return;
3014
3015 /* Force an aging cycle; this might trim enough addresses. */
3016 bridge_rtage(sc);
3017 if (sc->sc_brtcnt <= sc->sc_brtmax)
3018 return;
3019
3020 CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3021 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
3022 bridge_rtnode_destroy(sc, brt);
3023 if (sc->sc_brtcnt <= sc->sc_brtmax)
3024 return;
3025 }
3026 }
3027 }
3028
3029 /*
3030 * bridge_timer:
3031 *
3032 * Aging timer for the bridge.
3033 */
3034 static void
bridge_timer(void * arg)3035 bridge_timer(void *arg)
3036 {
3037 struct bridge_softc *sc = arg;
3038
3039 BRIDGE_RT_LOCK_ASSERT(sc);
3040
3041 /* Destruction of rtnodes requires a proper vnet context */
3042 CURVNET_SET(sc->sc_ifp->if_vnet);
3043 bridge_rtage(sc);
3044
3045 if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
3046 callout_reset(&sc->sc_brcallout,
3047 bridge_rtable_prune_period * hz, bridge_timer, sc);
3048 CURVNET_RESTORE();
3049 }
3050
3051 /*
3052 * bridge_rtage:
3053 *
3054 * Perform an aging cycle.
3055 */
3056 static void
bridge_rtage(struct bridge_softc * sc)3057 bridge_rtage(struct bridge_softc *sc)
3058 {
3059 struct bridge_rtnode *brt, *nbrt;
3060
3061 BRIDGE_RT_LOCK_ASSERT(sc);
3062
3063 CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3064 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
3065 if (time_uptime >= brt->brt_expire)
3066 bridge_rtnode_destroy(sc, brt);
3067 }
3068 }
3069 }
3070
3071 /*
3072 * bridge_rtflush:
3073 *
3074 * Remove all dynamic addresses from the bridge.
3075 */
3076 static void
bridge_rtflush(struct bridge_softc * sc,int full)3077 bridge_rtflush(struct bridge_softc *sc, int full)
3078 {
3079 struct bridge_rtnode *brt, *nbrt;
3080
3081 BRIDGE_RT_LOCK_ASSERT(sc);
3082
3083 CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3084 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3085 bridge_rtnode_destroy(sc, brt);
3086 }
3087 }
3088
3089 /*
3090 * bridge_rtdaddr:
3091 *
3092 * Remove an address from the table.
3093 */
3094 static int
bridge_rtdaddr(struct bridge_softc * sc,const uint8_t * addr,uint16_t vlan)3095 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
3096 {
3097 struct bridge_rtnode *brt;
3098 int found = 0;
3099
3100 BRIDGE_RT_LOCK(sc);
3101
3102 /*
3103 * If vlan is DOT1Q_VID_RSVD_IMPL then we want to delete for all vlans
3104 * so the lookup may return more than one.
3105 */
3106 while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
3107 bridge_rtnode_destroy(sc, brt);
3108 found = 1;
3109 }
3110
3111 BRIDGE_RT_UNLOCK(sc);
3112
3113 return (found ? 0 : ENOENT);
3114 }
3115
3116 /*
3117 * bridge_rtdelete:
3118 *
3119 * Delete routes to a speicifc member interface.
3120 */
3121 static void
bridge_rtdelete(struct bridge_softc * sc,struct ifnet * ifp,int full)3122 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
3123 {
3124 struct bridge_rtnode *brt, *nbrt;
3125
3126 BRIDGE_RT_LOCK_ASSERT(sc);
3127
3128 CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3129 if (brt->brt_ifp == ifp && (full ||
3130 (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
3131 bridge_rtnode_destroy(sc, brt);
3132 }
3133 }
3134
3135 /*
3136 * bridge_rtable_init:
3137 *
3138 * Initialize the route table for this bridge.
3139 */
3140 static void
bridge_rtable_init(struct bridge_softc * sc)3141 bridge_rtable_init(struct bridge_softc *sc)
3142 {
3143 int i;
3144
3145 sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
3146 M_DEVBUF, M_WAITOK);
3147
3148 for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
3149 CK_LIST_INIT(&sc->sc_rthash[i]);
3150
3151 sc->sc_rthash_key = arc4random();
3152 CK_LIST_INIT(&sc->sc_rtlist);
3153 }
3154
3155 /*
3156 * bridge_rtable_fini:
3157 *
3158 * Deconstruct the route table for this bridge.
3159 */
3160 static void
bridge_rtable_fini(struct bridge_softc * sc)3161 bridge_rtable_fini(struct bridge_softc *sc)
3162 {
3163
3164 KASSERT(sc->sc_brtcnt == 0,
3165 ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
3166 free(sc->sc_rthash, M_DEVBUF);
3167 }
3168
3169 /*
3170 * The following hash function is adapted from "Hash Functions" by Bob Jenkins
3171 * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
3172 */
3173 #define mix(a, b, c) \
3174 do { \
3175 a -= b; a -= c; a ^= (c >> 13); \
3176 b -= c; b -= a; b ^= (a << 8); \
3177 c -= a; c -= b; c ^= (b >> 13); \
3178 a -= b; a -= c; a ^= (c >> 12); \
3179 b -= c; b -= a; b ^= (a << 16); \
3180 c -= a; c -= b; c ^= (b >> 5); \
3181 a -= b; a -= c; a ^= (c >> 3); \
3182 b -= c; b -= a; b ^= (a << 10); \
3183 c -= a; c -= b; c ^= (b >> 15); \
3184 } while (/*CONSTCOND*/0)
3185
3186 static __inline uint32_t
bridge_rthash(struct bridge_softc * sc,const uint8_t * addr)3187 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
3188 {
3189 uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
3190
3191 b += addr[5] << 8;
3192 b += addr[4];
3193 a += addr[3] << 24;
3194 a += addr[2] << 16;
3195 a += addr[1] << 8;
3196 a += addr[0];
3197
3198 mix(a, b, c);
3199
3200 return (c & BRIDGE_RTHASH_MASK);
3201 }
3202
3203 #undef mix
3204
3205 static int
bridge_rtnode_addr_cmp(const uint8_t * a,const uint8_t * b)3206 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
3207 {
3208 int i, d;
3209
3210 for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
3211 d = ((int)a[i]) - ((int)b[i]);
3212 }
3213
3214 return (d);
3215 }
3216
3217 /*
3218 * bridge_rtnode_lookup:
3219 *
3220 * Look up a bridge route node for the specified destination. Compare the
3221 * vlan id or if zero then just return the first match.
3222 */
3223 static struct bridge_rtnode *
bridge_rtnode_lookup(struct bridge_softc * sc,const uint8_t * addr,uint16_t vlan)3224 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
3225 {
3226 struct bridge_rtnode *brt;
3227 uint32_t hash;
3228 int dir;
3229
3230 BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(sc);
3231
3232 hash = bridge_rthash(sc, addr);
3233 CK_LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
3234 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
3235 if (dir == 0 && (brt->brt_vlan == vlan || vlan == DOT1Q_VID_RSVD_IMPL))
3236 return (brt);
3237 if (dir > 0)
3238 return (NULL);
3239 }
3240
3241 return (NULL);
3242 }
3243
3244 /*
3245 * bridge_rtnode_insert:
3246 *
3247 * Insert the specified bridge node into the route table. We
3248 * assume the entry is not already in the table.
3249 */
3250 static int
bridge_rtnode_insert(struct bridge_softc * sc,struct bridge_rtnode * brt)3251 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
3252 {
3253 struct bridge_rtnode *lbrt;
3254 uint32_t hash;
3255 int dir;
3256
3257 BRIDGE_RT_LOCK_ASSERT(sc);
3258
3259 hash = bridge_rthash(sc, brt->brt_addr);
3260
3261 lbrt = CK_LIST_FIRST(&sc->sc_rthash[hash]);
3262 if (lbrt == NULL) {
3263 CK_LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
3264 goto out;
3265 }
3266
3267 do {
3268 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
3269 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
3270 return (EEXIST);
3271 if (dir > 0) {
3272 CK_LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
3273 goto out;
3274 }
3275 if (CK_LIST_NEXT(lbrt, brt_hash) == NULL) {
3276 CK_LIST_INSERT_AFTER(lbrt, brt, brt_hash);
3277 goto out;
3278 }
3279 lbrt = CK_LIST_NEXT(lbrt, brt_hash);
3280 } while (lbrt != NULL);
3281
3282 #ifdef DIAGNOSTIC
3283 panic("bridge_rtnode_insert: impossible");
3284 #endif
3285
3286 out:
3287 CK_LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
3288 sc->sc_brtcnt++;
3289
3290 return (0);
3291 }
3292
3293 static void
bridge_rtnode_destroy_cb(struct epoch_context * ctx)3294 bridge_rtnode_destroy_cb(struct epoch_context *ctx)
3295 {
3296 struct bridge_rtnode *brt;
3297
3298 brt = __containerof(ctx, struct bridge_rtnode, brt_epoch_ctx);
3299
3300 CURVNET_SET(brt->brt_vnet);
3301 uma_zfree(V_bridge_rtnode_zone, brt);
3302 CURVNET_RESTORE();
3303 }
3304
3305 /*
3306 * bridge_rtnode_destroy:
3307 *
3308 * Destroy a bridge rtnode.
3309 */
3310 static void
bridge_rtnode_destroy(struct bridge_softc * sc,struct bridge_rtnode * brt)3311 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3312 {
3313 BRIDGE_RT_LOCK_ASSERT(sc);
3314
3315 CK_LIST_REMOVE(brt, brt_hash);
3316
3317 CK_LIST_REMOVE(brt, brt_list);
3318 sc->sc_brtcnt--;
3319 brt->brt_dst->bif_addrcnt--;
3320
3321 NET_EPOCH_CALL(bridge_rtnode_destroy_cb, &brt->brt_epoch_ctx);
3322 }
3323
3324 /*
3325 * bridge_rtable_expire:
3326 *
3327 * Set the expiry time for all routes on an interface.
3328 */
3329 static void
bridge_rtable_expire(struct ifnet * ifp,int age)3330 bridge_rtable_expire(struct ifnet *ifp, int age)
3331 {
3332 struct bridge_softc *sc = ifp->if_bridge;
3333 struct bridge_rtnode *brt;
3334
3335 CURVNET_SET(ifp->if_vnet);
3336 BRIDGE_RT_LOCK(sc);
3337
3338 /*
3339 * If the age is zero then flush, otherwise set all the expiry times to
3340 * age for the interface
3341 */
3342 if (age == 0)
3343 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
3344 else {
3345 CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
3346 /* Cap the expiry time to 'age' */
3347 if (brt->brt_ifp == ifp &&
3348 brt->brt_expire > time_uptime + age &&
3349 (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3350 brt->brt_expire = time_uptime + age;
3351 }
3352 }
3353 BRIDGE_RT_UNLOCK(sc);
3354 CURVNET_RESTORE();
3355 }
3356
3357 /*
3358 * bridge_state_change:
3359 *
3360 * Callback from the bridgestp code when a port changes states.
3361 */
3362 static void
bridge_state_change(struct ifnet * ifp,int state)3363 bridge_state_change(struct ifnet *ifp, int state)
3364 {
3365 struct bridge_softc *sc = ifp->if_bridge;
3366 static const char *stpstates[] = {
3367 "disabled",
3368 "listening",
3369 "learning",
3370 "forwarding",
3371 "blocking",
3372 "discarding"
3373 };
3374
3375 CURVNET_SET(ifp->if_vnet);
3376 if (V_log_stp)
3377 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
3378 sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
3379 CURVNET_RESTORE();
3380 }
3381
3382 /*
3383 * Send bridge packets through pfil if they are one of the types pfil can deal
3384 * with, or if they are ARP or REVARP. (pfil will pass ARP and REVARP without
3385 * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3386 * that interface.
3387 */
3388 static int
bridge_pfil(struct mbuf ** mp,struct ifnet * bifp,struct ifnet * ifp,int dir)3389 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3390 {
3391 int snap, error, i;
3392 struct ether_header *eh1, eh2;
3393 struct llc llc1;
3394 u_int16_t ether_type;
3395 pfil_return_t rv;
3396 #ifdef INET
3397 struct ip *ip = NULL;
3398 int hlen = 0;
3399 #endif
3400
3401 snap = 0;
3402 error = -1; /* Default error if not error == 0 */
3403
3404 #if 0
3405 /* we may return with the IP fields swapped, ensure its not shared */
3406 KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
3407 #endif
3408
3409 if (V_pfil_bridge == 0 && V_pfil_member == 0 && V_pfil_ipfw == 0)
3410 return (0); /* filtering is disabled */
3411
3412 i = min((*mp)->m_pkthdr.len, max_protohdr);
3413 if ((*mp)->m_len < i) {
3414 *mp = m_pullup(*mp, i);
3415 if (*mp == NULL) {
3416 printf("%s: m_pullup failed\n", __func__);
3417 return (-1);
3418 }
3419 }
3420
3421 eh1 = mtod(*mp, struct ether_header *);
3422 ether_type = ntohs(eh1->ether_type);
3423
3424 /*
3425 * Check for SNAP/LLC.
3426 */
3427 if (ether_type < ETHERMTU) {
3428 struct llc *llc2 = (struct llc *)(eh1 + 1);
3429
3430 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3431 llc2->llc_dsap == LLC_SNAP_LSAP &&
3432 llc2->llc_ssap == LLC_SNAP_LSAP &&
3433 llc2->llc_control == LLC_UI) {
3434 ether_type = htons(llc2->llc_un.type_snap.ether_type);
3435 snap = 1;
3436 }
3437 }
3438
3439 /*
3440 * If we're trying to filter bridge traffic, only look at traffic for
3441 * protocols available in the kernel (IPv4 and/or IPv6) to avoid
3442 * passing traffic for an unsupported protocol to the filter. This is
3443 * lame since if we really wanted, say, an AppleTalk filter, we are
3444 * hosed, but of course we don't have an AppleTalk filter to begin
3445 * with. (Note that since pfil doesn't understand ARP it will pass
3446 * *ALL* ARP traffic.)
3447 */
3448 switch (ether_type) {
3449 #ifdef INET
3450 case ETHERTYPE_ARP:
3451 case ETHERTYPE_REVARP:
3452 if (V_pfil_ipfw_arp == 0)
3453 return (0); /* Automatically pass */
3454
3455 /* FALLTHROUGH */
3456 case ETHERTYPE_IP:
3457 #endif
3458 #ifdef INET6
3459 case ETHERTYPE_IPV6:
3460 #endif /* INET6 */
3461 break;
3462
3463 default:
3464 /*
3465 * We get here if the packet isn't from a supported
3466 * protocol. Check to see if the user wants to pass
3467 * non-IP packets, these will not be checked by pfil(9)
3468 * and passed unconditionally so the default is to
3469 * drop.
3470 */
3471 if (V_pfil_onlyip)
3472 goto bad;
3473 }
3474
3475 /* Run the packet through pfil before stripping link headers */
3476 if (PFIL_HOOKED_OUT(V_link_pfil_head) && V_pfil_ipfw != 0 &&
3477 dir == PFIL_OUT && ifp != NULL) {
3478 switch (pfil_mbuf_out(V_link_pfil_head, mp, ifp, NULL)) {
3479 case PFIL_DROPPED:
3480 return (EACCES);
3481 case PFIL_CONSUMED:
3482 return (0);
3483 }
3484 }
3485
3486 /* Strip off the Ethernet header and keep a copy. */
3487 m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3488 m_adj(*mp, ETHER_HDR_LEN);
3489
3490 /* Strip off snap header, if present */
3491 if (snap) {
3492 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3493 m_adj(*mp, sizeof(struct llc));
3494 }
3495
3496 /*
3497 * Check the IP header for alignment and errors
3498 */
3499 if (dir == PFIL_IN) {
3500 switch (ether_type) {
3501 #ifdef INET
3502 case ETHERTYPE_IP:
3503 error = bridge_ip_checkbasic(mp);
3504 break;
3505 #endif
3506 #ifdef INET6
3507 case ETHERTYPE_IPV6:
3508 error = bridge_ip6_checkbasic(mp);
3509 break;
3510 #endif /* INET6 */
3511 default:
3512 error = 0;
3513 }
3514 if (error)
3515 goto bad;
3516 }
3517
3518 error = 0;
3519
3520 /*
3521 * Run the packet through pfil
3522 */
3523 rv = PFIL_PASS;
3524 switch (ether_type) {
3525 #ifdef INET
3526 case ETHERTYPE_IP:
3527 /*
3528 * Run pfil on the member interface and the bridge, both can
3529 * be skipped by clearing pfil_member or pfil_bridge.
3530 *
3531 * Keep the order:
3532 * in_if -> bridge_if -> out_if
3533 */
3534 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL && (rv =
3535 pfil_mbuf_out(V_inet_pfil_head, mp, bifp, NULL)) !=
3536 PFIL_PASS)
3537 break;
3538
3539 if (V_pfil_member && ifp != NULL) {
3540 rv = (dir == PFIL_OUT) ?
3541 pfil_mbuf_out(V_inet_pfil_head, mp, ifp, NULL) :
3542 pfil_mbuf_in(V_inet_pfil_head, mp, ifp, NULL);
3543 if (rv != PFIL_PASS)
3544 break;
3545 }
3546
3547 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv =
3548 pfil_mbuf_in(V_inet_pfil_head, mp, bifp, NULL)) !=
3549 PFIL_PASS)
3550 break;
3551
3552 /* check if we need to fragment the packet */
3553 /* bridge_fragment generates a mbuf chain of packets */
3554 /* that already include eth headers */
3555 if (V_pfil_member && ifp != NULL && dir == PFIL_OUT) {
3556 i = (*mp)->m_pkthdr.len;
3557 if (i > ifp->if_mtu) {
3558 error = bridge_fragment(ifp, mp, &eh2, snap,
3559 &llc1);
3560 return (error);
3561 }
3562 }
3563
3564 /* Recalculate the ip checksum. */
3565 ip = mtod(*mp, struct ip *);
3566 hlen = ip->ip_hl << 2;
3567 if (hlen < sizeof(struct ip))
3568 goto bad;
3569 if (hlen > (*mp)->m_len) {
3570 if ((*mp = m_pullup(*mp, hlen)) == NULL)
3571 goto bad;
3572 ip = mtod(*mp, struct ip *);
3573 if (ip == NULL)
3574 goto bad;
3575 }
3576 ip->ip_sum = 0;
3577 if (hlen == sizeof(struct ip))
3578 ip->ip_sum = in_cksum_hdr(ip);
3579 else
3580 ip->ip_sum = in_cksum(*mp, hlen);
3581
3582 break;
3583 #endif /* INET */
3584 #ifdef INET6
3585 case ETHERTYPE_IPV6:
3586 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL && (rv =
3587 pfil_mbuf_out(V_inet6_pfil_head, mp, bifp, NULL)) !=
3588 PFIL_PASS)
3589 break;
3590
3591 if (V_pfil_member && ifp != NULL) {
3592 rv = (dir == PFIL_OUT) ?
3593 pfil_mbuf_out(V_inet6_pfil_head, mp, ifp, NULL) :
3594 pfil_mbuf_in(V_inet6_pfil_head, mp, ifp, NULL);
3595 if (rv != PFIL_PASS)
3596 break;
3597 }
3598
3599 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv =
3600 pfil_mbuf_in(V_inet6_pfil_head, mp, bifp, NULL)) !=
3601 PFIL_PASS)
3602 break;
3603 break;
3604 #endif
3605 }
3606
3607 switch (rv) {
3608 case PFIL_CONSUMED:
3609 return (0);
3610 case PFIL_DROPPED:
3611 return (EACCES);
3612 default:
3613 break;
3614 }
3615
3616 error = -1;
3617
3618 /*
3619 * Finally, put everything back the way it was and return
3620 */
3621 if (snap) {
3622 M_PREPEND(*mp, sizeof(struct llc), M_NOWAIT);
3623 if (*mp == NULL)
3624 return (error);
3625 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3626 }
3627
3628 M_PREPEND(*mp, ETHER_HDR_LEN, M_NOWAIT);
3629 if (*mp == NULL)
3630 return (error);
3631 bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3632
3633 return (0);
3634
3635 bad:
3636 m_freem(*mp);
3637 *mp = NULL;
3638 return (error);
3639 }
3640
3641 #ifdef INET
3642 /*
3643 * Perform basic checks on header size since
3644 * pfil assumes ip_input has already processed
3645 * it for it. Cut-and-pasted from ip_input.c.
3646 * Given how simple the IPv6 version is,
3647 * does the IPv4 version really need to be
3648 * this complicated?
3649 *
3650 * XXX Should we update ipstat here, or not?
3651 * XXX Right now we update ipstat but not
3652 * XXX csum_counter.
3653 */
3654 static int
bridge_ip_checkbasic(struct mbuf ** mp)3655 bridge_ip_checkbasic(struct mbuf **mp)
3656 {
3657 struct mbuf *m = *mp;
3658 struct ip *ip;
3659 int len, hlen;
3660 u_short sum;
3661
3662 if (*mp == NULL)
3663 return (-1);
3664
3665 if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3666 if ((m = m_copyup(m, sizeof(struct ip),
3667 (max_linkhdr + 3) & ~3)) == NULL) {
3668 /* XXXJRT new stat, please */
3669 KMOD_IPSTAT_INC(ips_toosmall);
3670 goto bad;
3671 }
3672 } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3673 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3674 KMOD_IPSTAT_INC(ips_toosmall);
3675 goto bad;
3676 }
3677 }
3678 ip = mtod(m, struct ip *);
3679 if (ip == NULL) goto bad;
3680
3681 if (ip->ip_v != IPVERSION) {
3682 KMOD_IPSTAT_INC(ips_badvers);
3683 goto bad;
3684 }
3685 hlen = ip->ip_hl << 2;
3686 if (hlen < sizeof(struct ip)) { /* minimum header length */
3687 KMOD_IPSTAT_INC(ips_badhlen);
3688 goto bad;
3689 }
3690 if (hlen > m->m_len) {
3691 if ((m = m_pullup(m, hlen)) == NULL) {
3692 KMOD_IPSTAT_INC(ips_badhlen);
3693 goto bad;
3694 }
3695 ip = mtod(m, struct ip *);
3696 if (ip == NULL) goto bad;
3697 }
3698
3699 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3700 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3701 } else {
3702 if (hlen == sizeof(struct ip)) {
3703 sum = in_cksum_hdr(ip);
3704 } else {
3705 sum = in_cksum(m, hlen);
3706 }
3707 }
3708 if (sum) {
3709 KMOD_IPSTAT_INC(ips_badsum);
3710 goto bad;
3711 }
3712
3713 /* Retrieve the packet length. */
3714 len = ntohs(ip->ip_len);
3715
3716 /*
3717 * Check for additional length bogosity
3718 */
3719 if (len < hlen) {
3720 KMOD_IPSTAT_INC(ips_badlen);
3721 goto bad;
3722 }
3723
3724 /*
3725 * Check that the amount of data in the buffers
3726 * is as at least much as the IP header would have us expect.
3727 * Drop packet if shorter than we expect.
3728 */
3729 if (m->m_pkthdr.len < len) {
3730 KMOD_IPSTAT_INC(ips_tooshort);
3731 goto bad;
3732 }
3733
3734 /* Checks out, proceed */
3735 *mp = m;
3736 return (0);
3737
3738 bad:
3739 *mp = m;
3740 return (-1);
3741 }
3742 #endif /* INET */
3743
3744 #ifdef INET6
3745 /*
3746 * Same as above, but for IPv6.
3747 * Cut-and-pasted from ip6_input.c.
3748 * XXX Should we update ip6stat, or not?
3749 */
3750 static int
bridge_ip6_checkbasic(struct mbuf ** mp)3751 bridge_ip6_checkbasic(struct mbuf **mp)
3752 {
3753 struct mbuf *m = *mp;
3754 struct ip6_hdr *ip6;
3755
3756 /*
3757 * If the IPv6 header is not aligned, slurp it up into a new
3758 * mbuf with space for link headers, in the event we forward
3759 * it. Otherwise, if it is aligned, make sure the entire base
3760 * IPv6 header is in the first mbuf of the chain.
3761 */
3762 if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3763 struct ifnet *inifp = m->m_pkthdr.rcvif;
3764 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3765 (max_linkhdr + 3) & ~3)) == NULL) {
3766 /* XXXJRT new stat, please */
3767 IP6STAT_INC(ip6s_toosmall);
3768 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3769 goto bad;
3770 }
3771 } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3772 struct ifnet *inifp = m->m_pkthdr.rcvif;
3773 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3774 IP6STAT_INC(ip6s_toosmall);
3775 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3776 goto bad;
3777 }
3778 }
3779
3780 ip6 = mtod(m, struct ip6_hdr *);
3781
3782 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3783 IP6STAT_INC(ip6s_badvers);
3784 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3785 goto bad;
3786 }
3787
3788 /* Checks out, proceed */
3789 *mp = m;
3790 return (0);
3791
3792 bad:
3793 *mp = m;
3794 return (-1);
3795 }
3796 #endif /* INET6 */
3797
3798 #ifdef INET
3799 /*
3800 * bridge_fragment:
3801 *
3802 * Fragment mbuf chain in multiple packets and prepend ethernet header.
3803 */
3804 static int
bridge_fragment(struct ifnet * ifp,struct mbuf ** mp,struct ether_header * eh,int snap,struct llc * llc)3805 bridge_fragment(struct ifnet *ifp, struct mbuf **mp, struct ether_header *eh,
3806 int snap, struct llc *llc)
3807 {
3808 struct mbuf *m = *mp, *nextpkt = NULL, *mprev = NULL, *mcur = NULL;
3809 struct ip *ip;
3810 int error = -1;
3811
3812 if (m->m_len < sizeof(struct ip) &&
3813 (m = m_pullup(m, sizeof(struct ip))) == NULL)
3814 goto dropit;
3815 ip = mtod(m, struct ip *);
3816
3817 m->m_pkthdr.csum_flags |= CSUM_IP;
3818 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist);
3819 if (error)
3820 goto dropit;
3821
3822 /*
3823 * Walk the chain and re-add the Ethernet header for
3824 * each mbuf packet.
3825 */
3826 for (mcur = m; mcur; mcur = mcur->m_nextpkt) {
3827 nextpkt = mcur->m_nextpkt;
3828 mcur->m_nextpkt = NULL;
3829 if (snap) {
3830 M_PREPEND(mcur, sizeof(struct llc), M_NOWAIT);
3831 if (mcur == NULL) {
3832 error = ENOBUFS;
3833 if (mprev != NULL)
3834 mprev->m_nextpkt = nextpkt;
3835 goto dropit;
3836 }
3837 bcopy(llc, mtod(mcur, caddr_t),sizeof(struct llc));
3838 }
3839
3840 M_PREPEND(mcur, ETHER_HDR_LEN, M_NOWAIT);
3841 if (mcur == NULL) {
3842 error = ENOBUFS;
3843 if (mprev != NULL)
3844 mprev->m_nextpkt = nextpkt;
3845 goto dropit;
3846 }
3847 bcopy(eh, mtod(mcur, caddr_t), ETHER_HDR_LEN);
3848
3849 /*
3850 * The previous two M_PREPEND could have inserted one or two
3851 * mbufs in front so we have to update the previous packet's
3852 * m_nextpkt.
3853 */
3854 mcur->m_nextpkt = nextpkt;
3855 if (mprev != NULL)
3856 mprev->m_nextpkt = mcur;
3857 else {
3858 /* The first mbuf in the original chain needs to be
3859 * updated. */
3860 *mp = mcur;
3861 }
3862 mprev = mcur;
3863 }
3864
3865 KMOD_IPSTAT_INC(ips_fragmented);
3866 return (error);
3867
3868 dropit:
3869 for (mcur = *mp; mcur; mcur = m) { /* droping the full packet chain */
3870 m = mcur->m_nextpkt;
3871 m_freem(mcur);
3872 }
3873 return (error);
3874 }
3875 #endif /* INET */
3876
3877 static void
bridge_linkstate(struct ifnet * ifp)3878 bridge_linkstate(struct ifnet *ifp)
3879 {
3880 struct bridge_softc *sc = ifp->if_bridge;
3881 struct bridge_iflist *bif;
3882 struct epoch_tracker et;
3883
3884 NET_EPOCH_ENTER(et);
3885
3886 bif = bridge_lookup_member_if(sc, ifp);
3887 if (bif == NULL) {
3888 NET_EPOCH_EXIT(et);
3889 return;
3890 }
3891 bridge_linkcheck(sc);
3892
3893 bstp_linkstate(&bif->bif_stp);
3894
3895 NET_EPOCH_EXIT(et);
3896 }
3897
3898 static void
bridge_linkcheck(struct bridge_softc * sc)3899 bridge_linkcheck(struct bridge_softc *sc)
3900 {
3901 struct bridge_iflist *bif;
3902 int new_link, hasls;
3903
3904 BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
3905
3906 new_link = LINK_STATE_DOWN;
3907 hasls = 0;
3908 /* Our link is considered up if at least one of our ports is active */
3909 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
3910 if (bif->bif_ifp->if_capabilities & IFCAP_LINKSTATE)
3911 hasls++;
3912 if (bif->bif_ifp->if_link_state == LINK_STATE_UP) {
3913 new_link = LINK_STATE_UP;
3914 break;
3915 }
3916 }
3917 if (!CK_LIST_EMPTY(&sc->sc_iflist) && !hasls) {
3918 /* If no interfaces support link-state then we default to up */
3919 new_link = LINK_STATE_UP;
3920 }
3921 if_link_state_change(sc->sc_ifp, new_link);
3922 }
3923