1 /*
2 * Copyright (c) 2004-2024 Apple Inc. All rights reserved.
3 *
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5 *
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
14 *
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
17 *
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
25 *
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27 */
28
29 /* $NetBSD: if_bridge.c,v 1.31 2005/06/01 19:45:34 jdc Exp $ */
30 /*
31 * Copyright 2001 Wasabi Systems, Inc.
32 * All rights reserved.
33 *
34 * Written by Jason R. Thorpe for Wasabi Systems, Inc.
35 *
36 * Redistribution and use in source and binary forms, with or without
37 * modification, are permitted provided that the following conditions
38 * are met:
39 * 1. Redistributions of source code must retain the above copyright
40 * notice, this list of conditions and the following disclaimer.
41 * 2. Redistributions in binary form must reproduce the above copyright
42 * notice, this list of conditions and the following disclaimer in the
43 * documentation and/or other materials provided with the distribution.
44 * 3. All advertising materials mentioning features or use of this software
45 * must display the following acknowledgement:
46 * This product includes software developed for the NetBSD Project by
47 * Wasabi Systems, Inc.
48 * 4. The name of Wasabi Systems, Inc. may not be used to endorse
49 * or promote products derived from this software without specific prior
50 * written permission.
51 *
52 * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
53 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
54 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
55 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL WASABI SYSTEMS, INC
56 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
57 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
58 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
59 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
60 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
61 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
62 * POSSIBILITY OF SUCH DAMAGE.
63 */
64
65 /*
66 * Copyright (c) 1999, 2000 Jason L. Wright ([email protected])
67 * All rights reserved.
68 *
69 * Redistribution and use in source and binary forms, with or without
70 * modification, are permitted provided that the following conditions
71 * are met:
72 * 1. Redistributions of source code must retain the above copyright
73 * notice, this list of conditions and the following disclaimer.
74 * 2. Redistributions in binary form must reproduce the above copyright
75 * notice, this list of conditions and the following disclaimer in the
76 * documentation and/or other materials provided with the distribution.
77 *
78 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
79 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
80 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
81 * DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
82 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
83 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
84 * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
85 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
86 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
87 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
88 * POSSIBILITY OF SUCH DAMAGE.
89 *
90 * OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun Exp
91 */
92
93 /*
94 * Network interface bridge support.
95 *
96 * TODO:
97 *
98 * - Currently only supports Ethernet-like interfaces (Ethernet,
99 * 802.11, VLANs on Ethernet, etc.) Figure out a nice way
100 * to bridge other types of interfaces (FDDI-FDDI, and maybe
101 * consider heterogenous bridges).
102 *
103 * - GIF isn't handled due to the lack of IPPROTO_ETHERIP support.
104 */
105
106 #include <sys/cdefs.h>
107
108 #include <sys/param.h>
109 #include <sys/mbuf.h>
110 #include <sys/malloc.h>
111 #include <sys/protosw.h>
112 #include <sys/systm.h>
113 #include <sys/time.h>
114 #include <sys/socket.h> /* for net/if.h */
115 #include <sys/sockio.h>
116 #include <sys/kernel.h>
117 #include <sys/random.h>
118 #include <sys/syslog.h>
119 #include <sys/sysctl.h>
120 #include <sys/proc.h>
121 #include <sys/lock.h>
122 #include <sys/mcache.h>
123
124 #include <sys/kauth.h>
125
126 #include <kern/thread_call.h>
127
128 #include <libkern/libkern.h>
129
130 #include <kern/zalloc.h>
131
132 #if NBPFILTER > 0
133 #include <net/bpf.h>
134 #endif
135 #include <net/if.h>
136 #include <net/if_dl.h>
137 #include <net/if_types.h>
138 #include <net/if_var.h>
139 #include <net/if_media.h>
140 #include <net/net_api_stats.h>
141
142 #include <netinet/in.h> /* for struct arpcom */
143 #include <netinet/tcp.h> /* for struct tcphdr */
144 #include <netinet/in_systm.h>
145 #include <netinet/in_var.h>
146 #define _IP_VHL
147 #include <netinet/ip.h>
148 #include <netinet/ip_var.h>
149 #include <netinet/ip6.h>
150 #include <netinet6/ip6_var.h>
151 #include <netinet/if_ether.h> /* for struct arpcom */
152 #include <net/bridgestp.h>
153 #include <net/if_bridgevar.h>
154 #include <net/if_llc.h>
155 #if NVLAN > 0
156 #include <net/if_vlan_var.h>
157 #endif /* NVLAN > 0 */
158
159 #include <net/if_ether.h>
160 #include <net/dlil.h>
161 #include <net/kpi_interfacefilter.h>
162 #include <net/pfvar.h>
163
164 #include <net/route.h>
165 #include <dev/random/randomdev.h>
166
167 #include <netinet/bootp.h>
168 #include <netinet/dhcp.h>
169
170 #if SKYWALK
171 #include <skywalk/nexus/netif/nx_netif.h>
172 #endif /* SKYWALK */
173
174 #include <net/sockaddr_utils.h>
175 #include <net/mblist.h>
176
177 #include <os/log.h>
178
179 static struct in_addr inaddr_any = { .s_addr = INADDR_ANY };
180
181
182 #define __M_FLAGS_ARE_SET(m, flags) (((m)->m_flags & (flags)) != 0)
183 #define IS_BCAST(m) __M_FLAGS_ARE_SET(m, M_BCAST)
184 #define IS_MCAST(m) __M_FLAGS_ARE_SET(m, M_MCAST)
185 #define IS_BCAST_MCAST(m) __M_FLAGS_ARE_SET(m, M_BCAST | M_MCAST)
186
187 #define HTONS_ETHERTYPE_ARP htons(ETHERTYPE_ARP)
188 #define HTONS_ETHERTYPE_IP htons(ETHERTYPE_IP)
189 #define HTONS_ETHERTYPE_IPV6 htons(ETHERTYPE_IPV6)
190 #define HTONS_ARPHRD_ETHER htons(ARPHRD_ETHER)
191 #define HTONS_ARPOP_REQUEST htons(ARPOP_REQUEST)
192 #define HTONS_ARPOP_REPLY htons(ARPOP_REPLY)
193 #define HTONS_IPPORT_BOOTPC htons(IPPORT_BOOTPC)
194 #define HTONS_IPPORT_BOOTPS htons(IPPORT_BOOTPS)
195 #define HTONS_DHCP_FLAGS_BROADCAST htons(DHCP_FLAGS_BROADCAST)
196
197 /*
198 * if_bridge_debug, BR_DBGF_*
199 * - 'if_bridge_debug' is a bitmask of BR_DBGF_* flags that can be set
200 * to enable additional logs for the corresponding bridge function
201 * - "sysctl net.link.bridge.debug" controls the value of
202 * 'if_bridge_debug'
203 */
204 static uint32_t if_bridge_debug = 0;
205 #define BR_DBGF_LIFECYCLE 0x0001
206 #define BR_DBGF_INPUT 0x0002
207 #define BR_DBGF_OUTPUT 0x0004
208 #define BR_DBGF_RT_TABLE 0x0008
209 #define BR_DBGF_DELAYED_CALL 0x0010
210 #define BR_DBGF_IOCTL 0x0020
211 #define BR_DBGF_MBUF 0x0040
212 #define BR_DBGF_MCAST 0x0080
213 #define BR_DBGF_HOSTFILTER 0x0100
214 #define BR_DBGF_CHECKSUM 0x0200
215 #define BR_DBGF_MAC_NAT 0x0400
216 #define BR_DBGF_INPUT_LIST 0x0800
217
218 /*
219 * if_bridge_log_level
220 * - 'if_bridge_log_level' ensures that by default important logs are
221 * logged regardless of if_bridge_debug by comparing the log level
222 * in BRIDGE_LOG to if_bridge_log_level
223 * - use "sysctl net.link.bridge.log_level" controls the value of
224 * 'if_bridge_log_level'
225 * - the default value of 'if_bridge_log_level' is LOG_NOTICE; important
226 * logs must use LOG_NOTICE to ensure they appear by default
227 */
228 static int if_bridge_log_level = LOG_NOTICE;
229
230 #define BRIDGE_DBGF_ENABLED(__flag) ((if_bridge_debug & __flag) != 0)
231
232 /*
233 * BRIDGE_LOG, BRIDGE_LOG_SIMPLE
234 * - macros to generate the specified log conditionally based on
235 * the specified log level and debug flags
236 * - BRIDGE_LOG_SIMPLE does not include the function name in the log
237 */
238 #define BRIDGE_LOG(__level, __dbgf, __string, ...) \
239 do { \
240 if (__level <= if_bridge_log_level || \
241 BRIDGE_DBGF_ENABLED(__dbgf)) { \
242 os_log(OS_LOG_DEFAULT, "%s: " __string, \
243 __func__, ## __VA_ARGS__); \
244 } \
245 } while (0)
246 #define BRIDGE_LOG_SIMPLE(__level, __dbgf, __string, ...) \
247 do { \
248 if (__level <= if_bridge_log_level || \
249 BRIDGE_DBGF_ENABLED(__dbgf)) { \
250 os_log(OS_LOG_DEFAULT, __string, ## __VA_ARGS__); \
251 } \
252 } while (0)
253
254 #define _BRIDGE_LOCK(_sc) lck_mtx_lock(&(_sc)->sc_mtx)
255 #define _BRIDGE_UNLOCK(_sc) lck_mtx_unlock(&(_sc)->sc_mtx)
256 #define BRIDGE_LOCK_ASSERT_HELD(_sc) \
257 LCK_MTX_ASSERT(&(_sc)->sc_mtx, LCK_MTX_ASSERT_OWNED)
258 #define BRIDGE_LOCK_ASSERT_NOTHELD(_sc) \
259 LCK_MTX_ASSERT(&(_sc)->sc_mtx, LCK_MTX_ASSERT_NOTOWNED)
260
261 #define BRIDGE_LOCK_DEBUG 1
262 #if BRIDGE_LOCK_DEBUG
263
264 #define BR_LCKDBG_MAX 4
265
266 #define BRIDGE_LOCK(_sc) bridge_lock(_sc)
267 #define BRIDGE_UNLOCK(_sc) bridge_unlock(_sc)
268 #define BRIDGE_LOCK2REF(_sc, _err) _err = bridge_lock2ref(_sc)
269 #define BRIDGE_UNREF(_sc) bridge_unref(_sc)
270 #define BRIDGE_XLOCK(_sc) bridge_xlock(_sc)
271 #define BRIDGE_XDROP(_sc) bridge_xdrop(_sc)
272
273 #else /* !BRIDGE_LOCK_DEBUG */
274
275 #define BRIDGE_LOCK(_sc) _BRIDGE_LOCK(_sc)
276 #define BRIDGE_UNLOCK(_sc) _BRIDGE_UNLOCK(_sc)
277 #define BRIDGE_LOCK2REF(_sc, _err) do { \
278 BRIDGE_LOCK_ASSERT_HELD(_sc); \
279 if ((_sc)->sc_iflist_xcnt > 0) \
280 (_err) = EBUSY; \
281 else { \
282 (_sc)->sc_iflist_ref++; \
283 (_err) = 0; \
284 } \
285 _BRIDGE_UNLOCK(_sc); \
286 } while (0)
287 #define BRIDGE_UNREF(_sc) do { \
288 _BRIDGE_LOCK(_sc); \
289 (_sc)->sc_iflist_ref--; \
290 if (((_sc)->sc_iflist_xcnt > 0) && ((_sc)->sc_iflist_ref == 0)) { \
291 _BRIDGE_UNLOCK(_sc); \
292 wakeup(&(_sc)->sc_cv); \
293 } else \
294 _BRIDGE_UNLOCK(_sc); \
295 } while (0)
296 #define BRIDGE_XLOCK(_sc) do { \
297 BRIDGE_LOCK_ASSERT_HELD(_sc); \
298 (_sc)->sc_iflist_xcnt++; \
299 while ((_sc)->sc_iflist_ref > 0) \
300 msleep(&(_sc)->sc_cv, &(_sc)->sc_mtx, PZERO, \
301 "BRIDGE_XLOCK", NULL); \
302 } while (0)
303 #define BRIDGE_XDROP(_sc) do { \
304 BRIDGE_LOCK_ASSERT_HELD(_sc); \
305 (_sc)->sc_iflist_xcnt--; \
306 } while (0)
307
308 #endif /* BRIDGE_LOCK_DEBUG */
309
310 #define BRIDGE_BPF_TAP_IN(ifp, m) \
311 do { \
312 if (ifp->if_bpf != NULL) { \
313 bpf_tap_in(ifp, DLT_EN10MB, m, NULL, 0); \
314 } \
315 } while(0)
316
317 #define BRIDGE_BPF_TAP_OUT(ifp, m) \
318 do { \
319 if (ifp->if_bpf != NULL) { \
320 bpf_tap_out(ifp, DLT_EN10MB, m, NULL, 0); \
321 } \
322 } while(0)
323
324
325 /*
326 * Initial size of the route hash table. Must be a power of two.
327 */
328 #ifndef BRIDGE_RTHASH_SIZE
329 #define BRIDGE_RTHASH_SIZE 16
330 #endif
331
332 /*
333 * Maximum size of the routing hash table
334 */
335 #define BRIDGE_RTHASH_SIZE_MAX 2048
336
337 #define BRIDGE_RTHASH_MASK(sc) ((sc)->sc_rthash_size - 1)
338
339 /*
340 * Maximum number of addresses to cache.
341 */
342 #ifndef BRIDGE_RTABLE_MAX
343 #define BRIDGE_RTABLE_MAX 100
344 #endif
345
346 /*
347 * Timeout (in seconds) for entries learned dynamically.
348 */
349 #ifndef BRIDGE_RTABLE_TIMEOUT
350 #define BRIDGE_RTABLE_TIMEOUT (20 * 60) /* same as ARP */
351 #endif
352
353 /*
354 * Number of seconds between walks of the route list.
355 */
356 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
357 #define BRIDGE_RTABLE_PRUNE_PERIOD (5 * 60)
358 #endif
359
360 /*
361 * Number of MAC NAT entries
362 * - sized based on 16 clients (including MAC NAT interface)
363 * each with 4 addresses
364 */
365 #ifndef BRIDGE_MAC_NAT_ENTRY_MAX
366 #define BRIDGE_MAC_NAT_ENTRY_MAX 64
367 #endif /* BRIDGE_MAC_NAT_ENTRY_MAX */
368
369 /*
370 * List of capabilities to possibly mask on the member interface.
371 */
372 #define BRIDGE_IFCAPS_MASK (IFCAP_TSO | IFCAP_TXCSUM)
373 /*
374 * List of capabilities to disable on the member interface.
375 */
376 #define BRIDGE_IFCAPS_STRIP IFCAP_LRO
377
378 /*
379 * Bridge interface list entry.
380 */
381 struct bridge_iflist {
382 TAILQ_ENTRY(bridge_iflist) bif_next;
383 struct ifnet *bif_ifp; /* member if */
384 struct bstp_port bif_stp; /* STP state */
385 uint32_t bif_ifflags; /* member if flags */
386 int bif_savedcaps; /* saved capabilities */
387 uint32_t bif_addrmax; /* max # of addresses */
388 uint32_t bif_addrcnt; /* cur. # of addresses */
389 uint32_t bif_addrexceeded; /* # of address violations */
390
391 interface_filter_t bif_iff_ref;
392 struct bridge_softc *bif_sc;
393 uint32_t bif_flags;
394
395 /* host filter */
396 struct in_addr bif_hf_ipsrc;
397 uint8_t bif_hf_hwsrc[ETHER_ADDR_LEN];
398
399 struct ifbrmstats bif_stats;
400 };
401
402 static inline bool
bif_ifflags_are_set(struct bridge_iflist * bif,uint32_t flags)403 bif_ifflags_are_set(struct bridge_iflist * bif, uint32_t flags)
404 {
405 return (bif->bif_ifflags & flags) != 0;
406 }
407
408 static inline bool
bif_has_checksum_offload(struct bridge_iflist * bif)409 bif_has_checksum_offload(struct bridge_iflist * bif)
410 {
411 return bif_ifflags_are_set(bif, IFBIF_CHECKSUM_OFFLOAD);
412 }
413
414 static inline bool
bif_has_mac_nat(struct bridge_iflist * bif)415 bif_has_mac_nat(struct bridge_iflist * bif)
416 {
417 return bif_ifflags_are_set(bif, IFBIF_MAC_NAT);
418 }
419
420 /* fake errors to make the code clearer */
421 #define _EBADIP EJUSTRETURN
422 #define _EBADIPCHECKSUM EJUSTRETURN
423 #define _EBADIPV6 EJUSTRETURN
424 #define _EBADUDP EJUSTRETURN
425 #define _EBADTCP EJUSTRETURN
426 #define _EBADUDPCHECKSUM EJUSTRETURN
427 #define _EBADTCPCHECKSUM EJUSTRETURN
428
429 #define BIFF_PROMISC 0x01 /* promiscuous mode set */
430 #define BIFF_PROTO_ATTACHED 0x02 /* protocol attached */
431 #define BIFF_FILTER_ATTACHED 0x04 /* interface filter attached */
432 #define BIFF_MEDIA_ACTIVE 0x08 /* interface media active */
433 #define BIFF_HOST_FILTER 0x10 /* host filter enabled */
434 #define BIFF_HF_HWSRC 0x20 /* host filter source MAC is set */
435 #define BIFF_HF_IPSRC 0x40 /* host filter source IP is set */
436 #define BIFF_INPUT_BROADCAST 0x80 /* send broadcast packets in */
437 #define BIFF_IN_MEMBER_LIST 0x100 /* added to the member list */
438 #define BIFF_WIFI_INFRA 0x200 /* interface is Wi-Fi infra */
439 #define BIFF_ALL_MULTI 0x400 /* allmulti set */
440 #define BIFF_LRO_DISABLED 0x800 /* LRO was disabled */
441 #if SKYWALK
442 #define BIFF_FLOWSWITCH_ATTACHED 0x1000 /* we attached the flowswitch */
443 #define BIFF_NETAGENT_REMOVED 0x2000 /* we removed the netagent */
444 #endif /* SKYWALK */
445
446 /*
447 * mac_nat_entry
448 * - translates between an IP address and MAC address on a specific
449 * bridge interface member
450 */
451 struct mac_nat_entry {
452 LIST_ENTRY(mac_nat_entry) mne_list; /* list linkage */
453 struct bridge_iflist *mne_bif; /* originating interface */
454 unsigned long mne_expire; /* expiration time */
455 union {
456 struct in_addr mneu_ip; /* originating IPv4 address */
457 struct in6_addr mneu_ip6; /* originating IPv6 address */
458 } mne_u;
459 uint8_t mne_mac[ETHER_ADDR_LEN];
460 uint8_t mne_flags;
461 uint8_t mne_reserved;
462 };
463 #define mne_ip mne_u.mneu_ip
464 #define mne_ip6 mne_u.mneu_ip6
465
466 #define MNE_FLAGS_IPV6 0x01 /* IPv6 address */
467
468 LIST_HEAD(mac_nat_entry_list, mac_nat_entry);
469
470 /*
471 * mac_nat_record
472 * - used by bridge_mac_nat_output() to convey the translation that needs
473 * to take place in bridge_mac_nat_translate
474 * - holds enough information so that the translation can be done later
475 * when the destination interface is the MAC-NAT interface
476 */
477 struct mac_nat_record {
478 uint16_t mnr_ether_type;
479 union {
480 uint16_t mnru_arp_offset;
481 struct {
482 uint16_t mnruip_dhcp_flags;
483 uint16_t mnruip_udp_csum;
484 uint8_t mnruip_header_len;
485 } mnru_ip;
486 struct {
487 uint16_t mnruip6_icmp6_len;
488 uint16_t mnruip6_lladdr_offset;
489 uint8_t mnruip6_icmp6_type;
490 uint8_t mnruip6_header_len;
491 } mnru_ip6;
492 } mnr_u;
493 };
494
495 #define mnr_arp_offset mnr_u.mnru_arp_offset
496
497 #define mnr_ip_header_len mnr_u.mnru_ip.mnruip_header_len
498 #define mnr_ip_dhcp_flags mnr_u.mnru_ip.mnruip_dhcp_flags
499 #define mnr_ip_udp_csum mnr_u.mnru_ip.mnruip_udp_csum
500
501 #define mnr_ip6_icmp6_len mnr_u.mnru_ip6.mnruip6_icmp6_len
502 #define mnr_ip6_icmp6_type mnr_u.mnru_ip6.mnruip6_icmp6_type
503 #define mnr_ip6_header_len mnr_u.mnru_ip6.mnruip6_header_len
504 #define mnr_ip6_lladdr_offset mnr_u.mnru_ip6.mnruip6_lladdr_offset
505
506 /*
507 * Bridge route node.
508 */
509 struct bridge_rtnode {
510 LIST_ENTRY(bridge_rtnode) brt_hash; /* hash table linkage */
511 LIST_ENTRY(bridge_rtnode) brt_list; /* list linkage */
512 struct bridge_iflist *brt_dst; /* destination if */
513 unsigned long brt_expire; /* expiration time */
514 uint8_t brt_flags; /* address flags */
515 uint8_t brt_addr[ETHER_ADDR_LEN];
516 uint16_t brt_vlan; /* vlan id */
517 };
518
519 #define brt_ifp brt_dst->bif_ifp
520
521 /*
522 * Bridge delayed function call context
523 */
524 typedef void (*bridge_delayed_func_t)(struct bridge_softc *);
525
526 struct bridge_delayed_call {
527 struct bridge_softc *bdc_sc;
528 bridge_delayed_func_t bdc_func; /* Function to call */
529 struct timespec bdc_ts; /* Time to call */
530 u_int32_t bdc_flags;
531 thread_call_t bdc_thread_call;
532 };
533
534 #define BDCF_OUTSTANDING 0x01 /* Delayed call has been scheduled */
535 #define BDCF_CANCELLING 0x02 /* May be waiting for call completion */
536
537 /*
538 * Software state for each bridge.
539 */
540 LIST_HEAD(_bridge_rtnode_list, bridge_rtnode);
541
542 struct bridge_softc {
543 struct ifnet *sc_ifp; /* make this an interface */
544 uint32_t sc_flags;
545 LIST_ENTRY(bridge_softc) sc_list;
546 decl_lck_mtx_data(, sc_mtx);
547 struct _bridge_rtnode_list * __counted_by(sc_rthash_size) sc_rthash; /* our forwarding table */
548 struct _bridge_rtnode_list sc_rtlist; /* list version of above */
549 uint32_t sc_rthash_key; /* key for hash */
550 uint32_t sc_rthash_size; /* size of the hash table */
551 struct bridge_delayed_call sc_aging_timer;
552 struct bridge_delayed_call sc_resize_call;
553 TAILQ_HEAD(, bridge_iflist) sc_spanlist; /* span ports list */
554 struct bstp_state sc_stp; /* STP state */
555 void *sc_cv;
556 uint32_t sc_brtmax; /* max # of addresses */
557 uint32_t sc_brtcnt; /* cur. # of addresses */
558 uint32_t sc_brttimeout; /* rt timeout in seconds */
559 uint32_t sc_iflist_ref; /* refcount for sc_iflist */
560 uint32_t sc_iflist_xcnt; /* refcount for sc_iflist */
561 TAILQ_HEAD(, bridge_iflist) sc_iflist; /* member interface list */
562 uint32_t sc_brtexceeded; /* # of cache drops */
563 uint32_t sc_filter_flags; /* ipf and flags */
564 struct ifnet *sc_ifaddr; /* member mac copied from */
565 u_char sc_defaddr[6]; /* Default MAC address */
566 char sc_if_xname[IFNAMSIZ];
567
568 struct bridge_iflist *sc_mac_nat_bif; /* single MAC NAT interface */
569 struct mac_nat_entry_list sc_mne_list; /* MAC NAT IPv4 */
570 struct mac_nat_entry_list sc_mne_list_v6;/* MAC NAT IPv6 */
571 uint32_t sc_mne_max; /* max # of entries */
572 uint32_t sc_mne_count; /* cur. # of entries */
573 uint32_t sc_mne_allocation_failures;
574 #if BRIDGE_LOCK_DEBUG
575 /*
576 * Locking and unlocking calling history
577 */
578 void *lock_lr[BR_LCKDBG_MAX];
579 int next_lock_lr;
580 void *unlock_lr[BR_LCKDBG_MAX];
581 int next_unlock_lr;
582 #endif /* BRIDGE_LOCK_DEBUG */
583 };
584
585 #define SCF_DETACHING 0x01
586 #define SCF_RESIZING 0x02
587 #define SCF_MEDIA_ACTIVE 0x04
588 #define SCF_ADDRESS_ASSIGNED 0x08
589
590 typedef enum {
591 CHECKSUM_OPERATION_NONE = 0,
592 CHECKSUM_OPERATION_CLEAR_OFFLOAD = 1,
593 CHECKSUM_OPERATION_FINALIZE = 2,
594 CHECKSUM_OPERATION_COMPUTE = 3,
595 } ChecksumOperation;
596
597 typedef struct {
598 u_int ip_hlen; /* IP header length */
599 u_int ip_pay_len; /* length of payload (exclusive of ip_hlen) */
600 u_int ip_m0_len; /* bytes available at ip_hdr (without jumping mbufs) */
601 u_int ip_opt_len; /* IPv6 options headers length */
602 uint8_t ip_proto; /* IPPROTO_TCP, IPPROTO_UDP, etc. */
603 bool ip_is_ipv4;
604 bool ip_is_fragmented;
605 uint8_t *__sized_by(ip_m0_len) ip_hdr; /* pointer to IP header */
606 uint8_t *__indexable ip_proto_hdr; /* ptr to protocol header (TCP) */
607 } ip_packet_info, *ip_packet_info_t;
608
609 struct bridge_hostfilter_stats bridge_hostfilter_stats;
610
611 static LCK_GRP_DECLARE(bridge_lock_grp, "if_bridge");
612 #if BRIDGE_LOCK_DEBUG
613 static LCK_ATTR_DECLARE(bridge_lock_attr, 0, 0);
614 #else
615 static LCK_ATTR_DECLARE(bridge_lock_attr, LCK_ATTR_DEBUG, 0);
616 #endif
617 static LCK_MTX_DECLARE_ATTR(bridge_list_mtx, &bridge_lock_grp, &bridge_lock_attr);
618
619 static int bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
620
621 static KALLOC_TYPE_DEFINE(bridge_rtnode_pool, struct bridge_rtnode, NET_KT_DEFAULT);
622 static KALLOC_TYPE_DEFINE(bridge_mne_pool, struct mac_nat_entry, NET_KT_DEFAULT);
623
624 static int bridge_clone_create(struct if_clone *, uint32_t, void *);
625 static int bridge_clone_destroy(struct ifnet *);
626
627 static errno_t bridge_ioctl(struct ifnet *, u_long cmd, void *__sized_by(IOCPARM_LEN(cmd)));
628 #if HAS_IF_CAP
629 static void bridge_mutecaps(struct bridge_softc *);
630 static void bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
631 int);
632 #endif
633 static errno_t bridge_set_tso(struct bridge_softc *);
634 static void bridge_proto_attach_changed(struct ifnet *);
635 static int bridge_init(struct ifnet *);
636 static void bridge_ifstop(struct ifnet *, int);
637 static int bridge_output(struct ifnet *, struct mbuf *);
638 static void bridge_finalize_cksum(struct ifnet *, struct mbuf *);
639 static void bridge_start(struct ifnet *);
640 static errno_t bridge_input(struct ifnet *, mbuf_t *);
641 static errno_t bridge_iff_input(void *, ifnet_t, protocol_family_t,
642 mbuf_t *, char **);
643 static errno_t bridge_iff_output(void *, ifnet_t, protocol_family_t,
644 mbuf_t *);
645 static errno_t bridge_member_output(struct bridge_softc *sc, ifnet_t ifp,
646 mbuf_t *m);
647
648 static int bridge_enqueue(ifnet_t, struct ifnet *,
649 struct ifnet *, struct mbuf *, ChecksumOperation);
650 static void bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
651
652 static void bridge_forward(struct bridge_softc *, struct bridge_iflist *,
653 struct mbuf *);
654
655 static void bridge_aging_timer(struct bridge_softc *sc);
656
657 static void bridge_broadcast(struct bridge_softc *, struct bridge_iflist *,
658 struct mbuf *);
659 static void bridge_span(struct bridge_softc *, struct mbuf *);
660
661 static int bridge_rtupdate(struct bridge_softc *, const uint8_t[ETHER_ADDR_LEN],
662 uint16_t, struct bridge_iflist *, int, uint8_t);
663 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t[ETHER_ADDR_LEN],
664 uint16_t);
665 static void bridge_rttrim(struct bridge_softc *);
666 static void bridge_rtage(struct bridge_softc *);
667 static void bridge_rtflush(struct bridge_softc *, int);
668 static int bridge_rtdaddr(struct bridge_softc *, const uint8_t[ETHER_ADDR_LEN],
669 uint16_t);
670
671 static int bridge_rtable_init(struct bridge_softc *);
672 static void bridge_rtable_fini(struct bridge_softc *);
673
674 static void bridge_rthash_resize(struct bridge_softc *);
675
676 static int bridge_rtnode_addr_cmp(const uint8_t[ETHER_ADDR_LEN], const uint8_t[ETHER_ADDR_LEN]);
677 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
678 const uint8_t[ETHER_ADDR_LEN], uint16_t);
679 static int bridge_rtnode_hash(struct bridge_softc *,
680 struct bridge_rtnode *);
681 static int bridge_rtnode_insert(struct bridge_softc *,
682 struct bridge_rtnode *);
683 static void bridge_rtnode_destroy(struct bridge_softc *,
684 struct bridge_rtnode *);
685 #if BRIDGESTP
686 static void bridge_rtable_expire(struct ifnet *, int);
687 static void bridge_state_change(struct ifnet *, int);
688 #endif /* BRIDGESTP */
689
690 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
691 char * __sized_by(IFNAMSIZ) name);
692 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
693 struct ifnet *ifp);
694 static void bridge_delete_member(struct bridge_softc *,
695 struct bridge_iflist *);
696 static void bridge_delete_span(struct bridge_softc *,
697 struct bridge_iflist *);
698
699 static int bridge_ioctl_add(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
700 static int bridge_ioctl_del(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
701 static int bridge_ioctl_gifflags(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
702 static int bridge_ioctl_sifflags(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
703 static int bridge_ioctl_scache(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
704 static int bridge_ioctl_gcache(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
705 static int bridge_ioctl_gifs32(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
706 static int bridge_ioctl_gifs64(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
707 static int bridge_ioctl_rts32(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
708 static int bridge_ioctl_rts64(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
709 static int bridge_ioctl_saddr32(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
710 static int bridge_ioctl_saddr64(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
711 static int bridge_ioctl_sto(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
712 static int bridge_ioctl_gto(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
713 static int bridge_ioctl_daddr32(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
714 static int bridge_ioctl_daddr64(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
715 static int bridge_ioctl_flush(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
716 static int bridge_ioctl_gpri(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
717 static int bridge_ioctl_spri(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
718 static int bridge_ioctl_ght(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
719 static int bridge_ioctl_sht(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
720 static int bridge_ioctl_gfd(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
721 static int bridge_ioctl_sfd(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
722 static int bridge_ioctl_gma(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
723 static int bridge_ioctl_sma(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
724 static int bridge_ioctl_sifprio(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
725 static int bridge_ioctl_sifcost(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
726 static int bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
727 static int bridge_ioctl_addspan(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
728 static int bridge_ioctl_delspan(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
729 static int bridge_ioctl_gbparam32(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
730 static int bridge_ioctl_gbparam64(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
731 static int bridge_ioctl_grte(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
732 static int bridge_ioctl_gifsstp32(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
733 static int bridge_ioctl_gifsstp64(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
734 static int bridge_ioctl_sproto(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
735 static int bridge_ioctl_stxhc(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
736 static int bridge_ioctl_purge(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len);
737 static int bridge_ioctl_gfilt(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
738 static int bridge_ioctl_sfilt(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
739 static int bridge_ioctl_ghostfilter(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
740 static int bridge_ioctl_shostfilter(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
741 static int bridge_ioctl_gmnelist32(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
742 static int bridge_ioctl_gmnelist64(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
743 static int bridge_ioctl_gifstats32(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
744 static int bridge_ioctl_gifstats64(struct bridge_softc *, void *__sized_by(arg_len) arg, size_t arg_len);
745
746 static int bridge_pf(struct mbuf **, struct ifnet *,
747 uint32_t sc_filter_flags, bool input);
748 static int bridge_ip_checkbasic(struct mbuf **);
749 static int bridge_ip6_checkbasic(struct mbuf **);
750
751 static void bridge_detach(ifnet_t);
752 static void bridge_link_event(struct ifnet *, u_int32_t);
753 static void bridge_iflinkevent(struct ifnet *);
754 static u_int32_t bridge_updatelinkstatus(struct bridge_softc *);
755 static int interface_media_active(struct ifnet *);
756 static void bridge_schedule_delayed_call(struct bridge_delayed_call *);
757 static void bridge_cancel_delayed_call(struct bridge_delayed_call *);
758 static void bridge_cleanup_delayed_call(struct bridge_delayed_call *);
759 static int bridge_host_filter(struct bridge_iflist *, mbuf_t *);
760
761 static errno_t bridge_mac_nat_enable(struct bridge_softc *,
762 struct bridge_iflist *);
763 static void bridge_mac_nat_disable(struct bridge_softc *sc);
764 static void bridge_mac_nat_age_entries(struct bridge_softc *sc, unsigned long);
765 static void bridge_mac_nat_populate_entries(struct bridge_softc *sc);
766 static void bridge_mac_nat_flush_entries(struct bridge_softc *sc,
767 struct bridge_iflist *);
768 static mbuf_t bridge_mac_nat_input(struct bridge_softc *, ifnet_t, mbuf_t,
769 ifnet_t * dst_if);
770 static boolean_t bridge_mac_nat_output(struct bridge_softc *,
771 struct bridge_iflist *, mbuf_t *, struct mac_nat_record *);
772 static void bridge_mac_nat_translate(mbuf_t *, struct mac_nat_record *,
773 const char[ETHER_ADDR_LEN]);
774
775 static mblist bridge_mac_nat_input_list(struct bridge_softc *sc,
776 ifnet_t external_ifp, mbuf_t m, mbuf_t * forward_head);
777 static mbuf_t bridge_mac_nat_translate_list(struct bridge_softc * sc,
778 struct bridge_iflist *sbif, ifnet_t dst_if, mbuf_t m);
779 static mbuf_t bridge_mac_nat_copy_and_translate_list(struct bridge_softc * sc,
780 struct bridge_iflist *sbif, ifnet_t dst_if, mbuf_t m);
781
782 static mbuf_t bridge_pf_list(mbuf_t m, ifnet_t ifp,
783 uint32_t sc_filter_flags, bool input);
784
785 static bool is_broadcast_ip_packet(mbuf_t *);
786 static bool in_addr_is_ours(const struct in_addr);
787 static bool in6_addr_is_ours(const struct in6_addr *, uint32_t);
788
789 #define m_copypacket(m, how) m_copym(m, 0, M_COPYALL, how)
790
791 static mblist
792 gso_tcp(ifnet_t ifp, mbuf_t m, u_int mac_hlen, bool is_ipv4, bool is_tx);
793
794 static inline mblist
gso_tcp_receive(ifnet_t ifp,mbuf_t m,u_int mac_hlen,bool is_ipv4)795 gso_tcp_receive(ifnet_t ifp, mbuf_t m, u_int mac_hlen, bool is_ipv4)
796 {
797 return gso_tcp(ifp, m, mac_hlen, is_ipv4, false);
798 }
799
800 static inline mblist
gso_tcp_transmit(ifnet_t ifp,mbuf_t m,u_int mac_hlen,bool is_ipv4)801 gso_tcp_transmit(ifnet_t ifp, mbuf_t m, u_int mac_hlen, bool is_ipv4)
802 {
803 return gso_tcp(ifp, m, mac_hlen, is_ipv4, true);
804 }
805
806 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
807 #define VLANTAGOF(_m) 0
808
809 #define BSTP_ETHERADDR_RANGE_FIRST 0x00
810 #define BSTP_ETHERADDR_RANGE_LAST 0x0f
811
812 u_int8_t bstp_etheraddr[ETHER_ADDR_LEN] =
813 { 0x01, 0x80, 0xc2, 0x00, 0x00, BSTP_ETHERADDR_RANGE_FIRST };
814
815
816 static u_int8_t ethernulladdr[ETHER_ADDR_LEN] =
817 { 0x00, 0x00, 0x00, 0x00, 0x00, 0x00 };
818
819 #if BRIDGESTP
820 static struct bstp_cb_ops bridge_ops = {
821 .bcb_state = bridge_state_change,
822 .bcb_rtage = bridge_rtable_expire
823 };
824 #endif /* BRIDGESTP */
825
826 SYSCTL_DECL(_net_link);
827 SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW | CTLFLAG_LOCKED, 0,
828 "Bridge");
829
830 static int bridge_inherit_mac = 0; /* share MAC with first bridge member */
831 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac,
832 CTLFLAG_RW | CTLFLAG_LOCKED,
833 &bridge_inherit_mac, 0,
834 "Inherit MAC address from the first bridge member");
835
836 SYSCTL_INT(_net_link_bridge, OID_AUTO, rtable_prune_period,
837 CTLFLAG_RW | CTLFLAG_LOCKED,
838 &bridge_rtable_prune_period, 0,
839 "Interval between pruning of routing table");
840
841 static unsigned int bridge_rtable_hash_size_max = BRIDGE_RTHASH_SIZE_MAX;
842 SYSCTL_UINT(_net_link_bridge, OID_AUTO, rtable_hash_size_max,
843 CTLFLAG_RW | CTLFLAG_LOCKED,
844 &bridge_rtable_hash_size_max, 0,
845 "Maximum size of the routing hash table");
846
847 #if BRIDGE_DELAYED_CALLBACK_DEBUG
848 static int bridge_delayed_callback_delay = 0;
849 SYSCTL_INT(_net_link_bridge, OID_AUTO, delayed_callback_delay,
850 CTLFLAG_RW | CTLFLAG_LOCKED,
851 &bridge_delayed_callback_delay, 0,
852 "Delay before calling delayed function");
853 #endif
854
855 SYSCTL_STRUCT(_net_link_bridge, OID_AUTO,
856 hostfilterstats, CTLFLAG_RD | CTLFLAG_LOCKED,
857 &bridge_hostfilter_stats, bridge_hostfilter_stats, "");
858
859 #if BRIDGESTP
860 static int log_stp = 0; /* log STP state changes */
861 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp, CTLFLAG_RW,
862 &log_stp, 0, "Log STP state changes");
863 #endif /* BRIDGESTP */
864
865 struct bridge_control {
866 int (*bc_func)(struct bridge_softc *, void *__sized_by(arg_len) args, size_t arg_len);
867 unsigned int bc_argsize;
868 unsigned int bc_flags;
869 };
870
871 #define BC_F_COPYIN 0x01 /* copy arguments in */
872 #define BC_F_COPYOUT 0x02 /* copy arguments out */
873 #define BC_F_SUSER 0x04 /* do super-user check */
874
875 static const struct bridge_control bridge_control_table32[] = {
876 { .bc_func = bridge_ioctl_add, .bc_argsize = sizeof(struct ifbreq), /* 0 */
877 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
878 { .bc_func = bridge_ioctl_del, .bc_argsize = sizeof(struct ifbreq),
879 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
880
881 { .bc_func = bridge_ioctl_gifflags, .bc_argsize = sizeof(struct ifbreq),
882 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
883 { .bc_func = bridge_ioctl_sifflags, .bc_argsize = sizeof(struct ifbreq),
884 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
885
886 { .bc_func = bridge_ioctl_scache, .bc_argsize = sizeof(struct ifbrparam),
887 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
888 { .bc_func = bridge_ioctl_gcache, .bc_argsize = sizeof(struct ifbrparam),
889 .bc_flags = BC_F_COPYOUT },
890
891 { .bc_func = bridge_ioctl_gifs32, .bc_argsize = sizeof(struct ifbifconf32),
892 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
893 { .bc_func = bridge_ioctl_rts32, .bc_argsize = sizeof(struct ifbaconf32),
894 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
895
896 { .bc_func = bridge_ioctl_saddr32, .bc_argsize = sizeof(struct ifbareq32),
897 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
898
899 { .bc_func = bridge_ioctl_sto, .bc_argsize = sizeof(struct ifbrparam),
900 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
901 { .bc_func = bridge_ioctl_gto, .bc_argsize = sizeof(struct ifbrparam), /* 10 */
902 .bc_flags = BC_F_COPYOUT },
903
904 { .bc_func = bridge_ioctl_daddr32, .bc_argsize = sizeof(struct ifbareq32),
905 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
906
907 { .bc_func = bridge_ioctl_flush, .bc_argsize = sizeof(struct ifbreq),
908 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
909
910 { .bc_func = bridge_ioctl_gpri, .bc_argsize = sizeof(struct ifbrparam),
911 .bc_flags = BC_F_COPYOUT },
912 { .bc_func = bridge_ioctl_spri, .bc_argsize = sizeof(struct ifbrparam),
913 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
914
915 { .bc_func = bridge_ioctl_ght, .bc_argsize = sizeof(struct ifbrparam),
916 .bc_flags = BC_F_COPYOUT },
917 { .bc_func = bridge_ioctl_sht, .bc_argsize = sizeof(struct ifbrparam),
918 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
919
920 { .bc_func = bridge_ioctl_gfd, .bc_argsize = sizeof(struct ifbrparam),
921 .bc_flags = BC_F_COPYOUT },
922 { .bc_func = bridge_ioctl_sfd, .bc_argsize = sizeof(struct ifbrparam),
923 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
924
925 { .bc_func = bridge_ioctl_gma, .bc_argsize = sizeof(struct ifbrparam),
926 .bc_flags = BC_F_COPYOUT },
927 { .bc_func = bridge_ioctl_sma, .bc_argsize = sizeof(struct ifbrparam), /* 20 */
928 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
929
930 { .bc_func = bridge_ioctl_sifprio, .bc_argsize = sizeof(struct ifbreq),
931 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
932
933 { .bc_func = bridge_ioctl_sifcost, .bc_argsize = sizeof(struct ifbreq),
934 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
935
936 { .bc_func = bridge_ioctl_gfilt, .bc_argsize = sizeof(struct ifbrparam),
937 .bc_flags = BC_F_COPYOUT },
938 { .bc_func = bridge_ioctl_sfilt, .bc_argsize = sizeof(struct ifbrparam),
939 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
940
941 { .bc_func = bridge_ioctl_purge, .bc_argsize = sizeof(struct ifbreq),
942 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
943
944 { .bc_func = bridge_ioctl_addspan, .bc_argsize = sizeof(struct ifbreq),
945 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
946 { .bc_func = bridge_ioctl_delspan, .bc_argsize = sizeof(struct ifbreq),
947 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
948
949 { .bc_func = bridge_ioctl_gbparam32, .bc_argsize = sizeof(struct ifbropreq32),
950 .bc_flags = BC_F_COPYOUT },
951
952 { .bc_func = bridge_ioctl_grte, .bc_argsize = sizeof(struct ifbrparam),
953 .bc_flags = BC_F_COPYOUT },
954
955 { .bc_func = bridge_ioctl_gifsstp32, .bc_argsize = sizeof(struct ifbpstpconf32), /* 30 */
956 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
957
958 { .bc_func = bridge_ioctl_sproto, .bc_argsize = sizeof(struct ifbrparam),
959 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
960
961 { .bc_func = bridge_ioctl_stxhc, .bc_argsize = sizeof(struct ifbrparam),
962 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
963
964 { .bc_func = bridge_ioctl_sifmaxaddr, .bc_argsize = sizeof(struct ifbreq),
965 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
966
967 { .bc_func = bridge_ioctl_ghostfilter, .bc_argsize = sizeof(struct ifbrhostfilter),
968 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
969 { .bc_func = bridge_ioctl_shostfilter, .bc_argsize = sizeof(struct ifbrhostfilter),
970 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
971
972 { .bc_func = bridge_ioctl_gmnelist32,
973 .bc_argsize = sizeof(struct ifbrmnelist32),
974 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
975 { .bc_func = bridge_ioctl_gifstats32,
976 .bc_argsize = sizeof(struct ifbrmreq32),
977 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
978 };
979
980 static const struct bridge_control bridge_control_table64[] = {
981 { .bc_func = bridge_ioctl_add, .bc_argsize = sizeof(struct ifbreq), /* 0 */
982 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
983 { .bc_func = bridge_ioctl_del, .bc_argsize = sizeof(struct ifbreq),
984 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
985
986 { .bc_func = bridge_ioctl_gifflags, .bc_argsize = sizeof(struct ifbreq),
987 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
988 { .bc_func = bridge_ioctl_sifflags, .bc_argsize = sizeof(struct ifbreq),
989 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
990
991 { .bc_func = bridge_ioctl_scache, .bc_argsize = sizeof(struct ifbrparam),
992 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
993 { .bc_func = bridge_ioctl_gcache, .bc_argsize = sizeof(struct ifbrparam),
994 .bc_flags = BC_F_COPYOUT },
995
996 { .bc_func = bridge_ioctl_gifs64, .bc_argsize = sizeof(struct ifbifconf64),
997 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
998 { .bc_func = bridge_ioctl_rts64, .bc_argsize = sizeof(struct ifbaconf64),
999 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
1000
1001 { .bc_func = bridge_ioctl_saddr64, .bc_argsize = sizeof(struct ifbareq64),
1002 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1003
1004 { .bc_func = bridge_ioctl_sto, .bc_argsize = sizeof(struct ifbrparam),
1005 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1006 { .bc_func = bridge_ioctl_gto, .bc_argsize = sizeof(struct ifbrparam), /* 10 */
1007 .bc_flags = BC_F_COPYOUT },
1008
1009 { .bc_func = bridge_ioctl_daddr64, .bc_argsize = sizeof(struct ifbareq64),
1010 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1011
1012 { .bc_func = bridge_ioctl_flush, .bc_argsize = sizeof(struct ifbreq),
1013 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1014
1015 { .bc_func = bridge_ioctl_gpri, .bc_argsize = sizeof(struct ifbrparam),
1016 .bc_flags = BC_F_COPYOUT },
1017 { .bc_func = bridge_ioctl_spri, .bc_argsize = sizeof(struct ifbrparam),
1018 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1019
1020 { .bc_func = bridge_ioctl_ght, .bc_argsize = sizeof(struct ifbrparam),
1021 .bc_flags = BC_F_COPYOUT },
1022 { .bc_func = bridge_ioctl_sht, .bc_argsize = sizeof(struct ifbrparam),
1023 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1024
1025 { .bc_func = bridge_ioctl_gfd, .bc_argsize = sizeof(struct ifbrparam),
1026 .bc_flags = BC_F_COPYOUT },
1027 { .bc_func = bridge_ioctl_sfd, .bc_argsize = sizeof(struct ifbrparam),
1028 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1029
1030 { .bc_func = bridge_ioctl_gma, .bc_argsize = sizeof(struct ifbrparam),
1031 .bc_flags = BC_F_COPYOUT },
1032 { .bc_func = bridge_ioctl_sma, .bc_argsize = sizeof(struct ifbrparam), /* 20 */
1033 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1034
1035 { .bc_func = bridge_ioctl_sifprio, .bc_argsize = sizeof(struct ifbreq),
1036 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1037
1038 { .bc_func = bridge_ioctl_sifcost, .bc_argsize = sizeof(struct ifbreq),
1039 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1040
1041 { .bc_func = bridge_ioctl_gfilt, .bc_argsize = sizeof(struct ifbrparam),
1042 .bc_flags = BC_F_COPYOUT },
1043 { .bc_func = bridge_ioctl_sfilt, .bc_argsize = sizeof(struct ifbrparam),
1044 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1045
1046 { .bc_func = bridge_ioctl_purge, .bc_argsize = sizeof(struct ifbreq),
1047 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1048
1049 { .bc_func = bridge_ioctl_addspan, .bc_argsize = sizeof(struct ifbreq),
1050 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1051 { .bc_func = bridge_ioctl_delspan, .bc_argsize = sizeof(struct ifbreq),
1052 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1053
1054 { .bc_func = bridge_ioctl_gbparam64, .bc_argsize = sizeof(struct ifbropreq64),
1055 .bc_flags = BC_F_COPYOUT },
1056
1057 { .bc_func = bridge_ioctl_grte, .bc_argsize = sizeof(struct ifbrparam),
1058 .bc_flags = BC_F_COPYOUT },
1059
1060 { .bc_func = bridge_ioctl_gifsstp64, .bc_argsize = sizeof(struct ifbpstpconf64), /* 30 */
1061 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
1062
1063 { .bc_func = bridge_ioctl_sproto, .bc_argsize = sizeof(struct ifbrparam),
1064 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1065
1066 { .bc_func = bridge_ioctl_stxhc, .bc_argsize = sizeof(struct ifbrparam),
1067 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1068
1069 { .bc_func = bridge_ioctl_sifmaxaddr, .bc_argsize = sizeof(struct ifbreq),
1070 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1071
1072 { .bc_func = bridge_ioctl_ghostfilter, .bc_argsize = sizeof(struct ifbrhostfilter),
1073 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
1074 { .bc_func = bridge_ioctl_shostfilter, .bc_argsize = sizeof(struct ifbrhostfilter),
1075 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1076
1077 { .bc_func = bridge_ioctl_gmnelist64,
1078 .bc_argsize = sizeof(struct ifbrmnelist64),
1079 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
1080 { .bc_func = bridge_ioctl_gifstats64,
1081 .bc_argsize = sizeof(struct ifbrmreq64),
1082 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
1083 };
1084
1085 static const unsigned int bridge_control_table_size =
1086 sizeof(bridge_control_table32) / sizeof(bridge_control_table32[0]);
1087
1088 static LIST_HEAD(, bridge_softc) bridge_list =
1089 LIST_HEAD_INITIALIZER(bridge_list);
1090
1091 #define BRIDGENAME "bridge"
1092 #define BRIDGES_MAX IF_MAXUNIT
1093 #define BRIDGE_ZONE_MAX_ELEM MIN(IFNETS_MAX, BRIDGES_MAX)
1094
1095 static struct if_clone bridge_cloner =
1096 IF_CLONE_INITIALIZER(BRIDGENAME, bridge_clone_create, bridge_clone_destroy,
1097 0, BRIDGES_MAX);
1098
1099 static int if_bridge_txstart = 0;
1100 SYSCTL_INT(_net_link_bridge, OID_AUTO, txstart, CTLFLAG_RW | CTLFLAG_LOCKED,
1101 &if_bridge_txstart, 0, "Bridge interface uses TXSTART model");
1102
1103 SYSCTL_INT(_net_link_bridge, OID_AUTO, debug, CTLFLAG_RW | CTLFLAG_LOCKED,
1104 &if_bridge_debug, 0, "Bridge debug flags");
1105
1106 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_level,
1107 CTLFLAG_RW | CTLFLAG_LOCKED,
1108 &if_bridge_log_level, 0, "Bridge log level");
1109
1110 static int if_bridge_segmentation = 1;
1111 SYSCTL_INT(_net_link_bridge, OID_AUTO, segmentation,
1112 CTLFLAG_RW | CTLFLAG_LOCKED,
1113 &if_bridge_segmentation, 0, "Bridge interface enable segmentation");
1114
1115 static int if_bridge_output_skip_filters = 1;
1116 SYSCTL_INT(_net_link_bridge, OID_AUTO, output_skip_filters,
1117 CTLFLAG_RW | CTLFLAG_LOCKED,
1118 &if_bridge_output_skip_filters, 0, "Bridge skip output filters");
1119
1120 int bridge_enable_early_input = 1; /* DLIL early input */
1121 SYSCTL_INT(_net_link_bridge, OID_AUTO, enable_early_input,
1122 CTLFLAG_RW | CTLFLAG_LOCKED,
1123 &bridge_enable_early_input, 0,
1124 "Bridge enable early input");
1125
1126 #define BRIDGE_TSO_REDUCE_MSS_FORWARDING_MAX 256
1127 #define BRIDGE_TSO_REDUCE_MSS_FORWARDING_DEFAULT 110
1128 #define BRIDGE_TSO_REDUCE_MSS_TX_MAX 256
1129 #define BRIDGE_TSO_REDUCE_MSS_TX_DEFAULT 0
1130
1131 static u_int if_bridge_tso_reduce_mss_forwarding
1132 = BRIDGE_TSO_REDUCE_MSS_FORWARDING_DEFAULT;
1133 static u_int if_bridge_tso_reduce_mss_tx
1134 = BRIDGE_TSO_REDUCE_MSS_TX_DEFAULT;
1135
1136 static int
bridge_tso_reduce_mss(struct sysctl_req * req,u_int * val,u_int val_max)1137 bridge_tso_reduce_mss(struct sysctl_req *req, u_int * val, u_int val_max)
1138 {
1139 int changed;
1140 int error;
1141 u_int new_value;
1142
1143 error = sysctl_io_number(req, *val, sizeof(*val), &new_value,
1144 &changed);
1145 if (error == 0 && changed != 0) {
1146 if (new_value > val_max) {
1147 return EINVAL;
1148 }
1149 *val = new_value;
1150 }
1151 return error;
1152 }
1153
1154 static int
1155 bridge_tso_reduce_mss_forwarding_sysctl SYSCTL_HANDLER_ARGS
1156 {
1157 return bridge_tso_reduce_mss(req, &if_bridge_tso_reduce_mss_forwarding,
1158 BRIDGE_TSO_REDUCE_MSS_FORWARDING_MAX);
1159 }
1160
1161 SYSCTL_PROC(_net_link_bridge, OID_AUTO, tso_reduce_mss_forwarding,
1162 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,
1163 0, 0, bridge_tso_reduce_mss_forwarding_sysctl, "IU",
1164 "Bridge tso reduce mss when forwarding");
1165
1166 static int
1167 bridge_tso_reduce_mss_tx_sysctl SYSCTL_HANDLER_ARGS
1168 {
1169 return bridge_tso_reduce_mss(req, &if_bridge_tso_reduce_mss_tx,
1170 BRIDGE_TSO_REDUCE_MSS_TX_MAX);
1171 }
1172
1173 SYSCTL_PROC(_net_link_bridge, OID_AUTO, tso_reduce_mss_tx,
1174 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,
1175 0, 0, bridge_tso_reduce_mss_tx_sysctl, "IU",
1176 "Bridge tso reduce mss on transmit");
1177
1178 #if DEBUG || DEVELOPMENT
1179 /*
1180 * net.link.bridge.reduce_tso_mtu
1181 * - when non-zero, the bridge overrides the interface TSO MTU to a lower
1182 * value (i.e. 16K) to enable testing the "use GSO instead" path
1183 */
1184 static int if_bridge_reduce_tso_mtu = 0;
1185 SYSCTL_INT(_net_link_bridge, OID_AUTO, reduce_tso_mtu,
1186 CTLFLAG_RW | CTLFLAG_LOCKED,
1187 &if_bridge_reduce_tso_mtu, 0, "Bridge interface reduce TSO MTU");
1188
1189 #endif /* DEBUG || DEVELOPMENT */
1190
1191 static void brlog_ether_header(struct ether_header *);
1192 static void brlog_mbuf_data(mbuf_t, size_t, size_t);
1193 static void brlog_mbuf_pkthdr(mbuf_t, const char *, const char *);
1194 static void brlog_mbuf(mbuf_t, const char *, const char *);
1195 static void brlog_link(struct bridge_softc * sc);
1196
1197 #if BRIDGE_LOCK_DEBUG
1198 static void bridge_lock(struct bridge_softc *);
1199 static void bridge_unlock(struct bridge_softc *);
1200 static int bridge_lock2ref(struct bridge_softc *);
1201 static void bridge_unref(struct bridge_softc *);
1202 static void bridge_xlock(struct bridge_softc *);
1203 static void bridge_xdrop(struct bridge_softc *);
1204
1205 #define DECL_RETURN_ADDR(v) void * __single v = __unsafe_forge_single(void *, __builtin_return_address(0))
1206
1207 static void
bridge_lock(struct bridge_softc * sc)1208 bridge_lock(struct bridge_softc *sc)
1209 {
1210 DECL_RETURN_ADDR(lr_saved);
1211
1212 BRIDGE_LOCK_ASSERT_NOTHELD(sc);
1213
1214 _BRIDGE_LOCK(sc);
1215
1216 sc->lock_lr[sc->next_lock_lr] = lr_saved;
1217 sc->next_lock_lr = (sc->next_lock_lr + 1) % SO_LCKDBG_MAX;
1218 }
1219
1220 static void
bridge_unlock(struct bridge_softc * sc)1221 bridge_unlock(struct bridge_softc *sc)
1222 {
1223 DECL_RETURN_ADDR(lr_saved);
1224
1225 BRIDGE_LOCK_ASSERT_HELD(sc);
1226
1227 sc->unlock_lr[sc->next_unlock_lr] = lr_saved;
1228 sc->next_unlock_lr = (sc->next_unlock_lr + 1) % SO_LCKDBG_MAX;
1229
1230 _BRIDGE_UNLOCK(sc);
1231 }
1232
1233 static int
bridge_lock2ref(struct bridge_softc * sc)1234 bridge_lock2ref(struct bridge_softc *sc)
1235 {
1236 int error = 0;
1237 DECL_RETURN_ADDR(lr_saved);
1238
1239 BRIDGE_LOCK_ASSERT_HELD(sc);
1240
1241 if (sc->sc_iflist_xcnt > 0) {
1242 error = EBUSY;
1243 } else {
1244 sc->sc_iflist_ref++;
1245 }
1246
1247 sc->unlock_lr[sc->next_unlock_lr] = lr_saved;
1248 sc->next_unlock_lr = (sc->next_unlock_lr + 1) % SO_LCKDBG_MAX;
1249
1250 _BRIDGE_UNLOCK(sc);
1251
1252 return error;
1253 }
1254
1255 static void
bridge_unref(struct bridge_softc * sc)1256 bridge_unref(struct bridge_softc *sc)
1257 {
1258 DECL_RETURN_ADDR(lr_saved);
1259
1260 BRIDGE_LOCK_ASSERT_NOTHELD(sc);
1261
1262 _BRIDGE_LOCK(sc);
1263 sc->lock_lr[sc->next_lock_lr] = lr_saved;
1264 sc->next_lock_lr = (sc->next_lock_lr + 1) % SO_LCKDBG_MAX;
1265
1266 sc->sc_iflist_ref--;
1267
1268 sc->unlock_lr[sc->next_unlock_lr] = lr_saved;
1269 sc->next_unlock_lr = (sc->next_unlock_lr + 1) % SO_LCKDBG_MAX;
1270 if ((sc->sc_iflist_xcnt > 0) && (sc->sc_iflist_ref == 0)) {
1271 _BRIDGE_UNLOCK(sc);
1272 wakeup(&sc->sc_cv);
1273 } else {
1274 _BRIDGE_UNLOCK(sc);
1275 }
1276 }
1277
1278 static void
bridge_xlock(struct bridge_softc * sc)1279 bridge_xlock(struct bridge_softc *sc)
1280 {
1281 DECL_RETURN_ADDR(lr_saved);
1282
1283 BRIDGE_LOCK_ASSERT_HELD(sc);
1284
1285 sc->sc_iflist_xcnt++;
1286 while (sc->sc_iflist_ref > 0) {
1287 sc->unlock_lr[sc->next_unlock_lr] = lr_saved;
1288 sc->next_unlock_lr = (sc->next_unlock_lr + 1) % SO_LCKDBG_MAX;
1289
1290 msleep(&sc->sc_cv, &sc->sc_mtx, PZERO, "BRIDGE_XLOCK", NULL);
1291
1292 sc->lock_lr[sc->next_lock_lr] = lr_saved;
1293 sc->next_lock_lr = (sc->next_lock_lr + 1) % SO_LCKDBG_MAX;
1294 }
1295 }
1296
1297 #undef DECL_RETURN_ADDR
1298
1299 static void
bridge_xdrop(struct bridge_softc * sc)1300 bridge_xdrop(struct bridge_softc *sc)
1301 {
1302 BRIDGE_LOCK_ASSERT_HELD(sc);
1303
1304 sc->sc_iflist_xcnt--;
1305 }
1306
1307 #endif /* BRIDGE_LOCK_DEBUG */
1308
1309 static void
brlog_mbuf_pkthdr(mbuf_t m,const char * prefix,const char * suffix)1310 brlog_mbuf_pkthdr(mbuf_t m, const char *prefix, const char *suffix)
1311 {
1312 if (m) {
1313 BRIDGE_LOG_SIMPLE(LOG_NOTICE, 0,
1314 "%spktlen: %u rcvif: 0x%llx header: 0x%llx nextpkt: 0x%llx%s",
1315 prefix ? prefix : "", (unsigned int)mbuf_pkthdr_len(m),
1316 (uint64_t)VM_KERNEL_ADDRPERM(mbuf_pkthdr_rcvif(m)),
1317 (uint64_t)VM_KERNEL_ADDRPERM(mbuf_pkthdr_header(m)),
1318 (uint64_t)VM_KERNEL_ADDRPERM(mbuf_nextpkt(m)),
1319 suffix ? suffix : "");
1320 } else {
1321 BRIDGE_LOG_SIMPLE(LOG_NOTICE, 0, "%s<NULL>%s", prefix, suffix);
1322 }
1323 }
1324
1325 static void
brlog_mbuf(mbuf_t m,const char * prefix,const char * suffix)1326 brlog_mbuf(mbuf_t m, const char *prefix, const char *suffix)
1327 {
1328 if (m) {
1329 BRIDGE_LOG_SIMPLE(LOG_NOTICE, 0,
1330 "%s0x%llx type: %u flags: 0x%x len: %u data: 0x%llx "
1331 "maxlen: %u datastart: 0x%llx next: 0x%llx%s",
1332 prefix ? prefix : "", (uint64_t)VM_KERNEL_ADDRPERM(m),
1333 mbuf_type(m), mbuf_flags(m), (unsigned int)mbuf_len(m),
1334 (uint64_t)VM_KERNEL_ADDRPERM(mtod(m, void *)),
1335 (unsigned int)mbuf_maxlen(m),
1336 (uint64_t)VM_KERNEL_ADDRPERM(mbuf_datastart(m)),
1337 (uint64_t)VM_KERNEL_ADDRPERM(mbuf_next(m)),
1338 !suffix || (mbuf_flags(m) & MBUF_PKTHDR) ? "" : suffix);
1339 if ((mbuf_flags(m) & MBUF_PKTHDR)) {
1340 brlog_mbuf_pkthdr(m, "", suffix);
1341 }
1342 } else {
1343 BRIDGE_LOG_SIMPLE(LOG_NOTICE, 0, "%s<NULL>%s", prefix, suffix);
1344 }
1345 }
1346
1347 static void
brlog_mbuf_data(mbuf_t m,size_t offset,size_t len)1348 brlog_mbuf_data(mbuf_t m, size_t offset, size_t len)
1349 {
1350 mbuf_t n;
1351 size_t i, j;
1352 size_t pktlen, mlen, maxlen;
1353 unsigned char *ptr;
1354
1355 pktlen = mbuf_pkthdr_len(m);
1356
1357 if (offset > pktlen) {
1358 return;
1359 }
1360
1361 maxlen = (pktlen - offset > len) ? len : pktlen - offset;
1362 n = m;
1363 mlen = mbuf_len(n);
1364 ptr = mtod(n, unsigned char *);
1365 for (i = 0, j = 0; i < maxlen; i++, j++) {
1366 if (j >= mlen) {
1367 n = mbuf_next(n);
1368 if (n == 0) {
1369 break;
1370 }
1371 ptr = mtod(n, unsigned char *);
1372 mlen = mbuf_len(n);
1373 j = 0;
1374 }
1375 if (i >= offset) {
1376 BRIDGE_LOG_SIMPLE(LOG_NOTICE, 0,
1377 "%02x%s", ptr[j], i % 2 ? " " : "");
1378 }
1379 }
1380 }
1381
1382 static void
brlog_ether_header(struct ether_header * eh)1383 brlog_ether_header(struct ether_header *eh)
1384 {
1385 BRIDGE_LOG_SIMPLE(LOG_NOTICE, 0,
1386 "%02x:%02x:%02x:%02x:%02x:%02x > "
1387 "%02x:%02x:%02x:%02x:%02x:%02x 0x%04x ",
1388 eh->ether_shost[0], eh->ether_shost[1], eh->ether_shost[2],
1389 eh->ether_shost[3], eh->ether_shost[4], eh->ether_shost[5],
1390 eh->ether_dhost[0], eh->ether_dhost[1], eh->ether_dhost[2],
1391 eh->ether_dhost[3], eh->ether_dhost[4], eh->ether_dhost[5],
1392 ntohs(eh->ether_type));
1393 }
1394
1395 static char *
ether_ntop(char * __sized_by (len)buf,size_t len,const u_char ap[ETHER_ADDR_LEN])1396 ether_ntop(char * __sized_by(len) buf, size_t len, const u_char ap[ETHER_ADDR_LEN])
1397 {
1398 snprintf(buf, len, "%02x:%02x:%02x:%02x:%02x:%02x",
1399 ap[0], ap[1], ap[2], ap[3], ap[4], ap[5]);
1400
1401 return buf;
1402 }
1403
1404 static void
brlog_link(struct bridge_softc * sc)1405 brlog_link(struct bridge_softc * sc)
1406 {
1407 int i;
1408 uint32_t sdl_buffer[(offsetof(struct sockaddr_dl, sdl_data) +
1409 IFNAMSIZ + ETHER_ADDR_LEN)];
1410 struct sockaddr_dl *sdl = SDL((uint8_t*)&sdl_buffer); /* SDL requires byte pointer */
1411 const u_char * lladdr;
1412 char lladdr_str[48];
1413
1414 memset(sdl_buffer, 0, sizeof(sdl_buffer));
1415 sdl->sdl_family = AF_LINK;
1416 sdl->sdl_nlen = strbuflen(sc->sc_if_xname);
1417 sdl->sdl_alen = ETHER_ADDR_LEN;
1418 sdl->sdl_len = offsetof(struct sockaddr_dl, sdl_data);
1419 memcpy(sdl->sdl_data, sc->sc_if_xname, sdl->sdl_nlen);
1420 memcpy(LLADDR(sdl), sc->sc_defaddr, ETHER_ADDR_LEN);
1421 lladdr_str[0] = '\0';
1422 for (i = 0, lladdr = CONST_LLADDR(sdl);
1423 i < sdl->sdl_alen;
1424 i++, lladdr++) {
1425 char byte_str[4];
1426
1427 snprintf(byte_str, sizeof(byte_str), "%s%x", i ? ":" : "",
1428 *lladdr);
1429 strbufcat(lladdr_str, byte_str);
1430 }
1431 BRIDGE_LOG_SIMPLE(LOG_NOTICE, 0,
1432 "%s sdl len %d index %d family %d type 0x%x nlen %d alen %d"
1433 " slen %d addr %s", sc->sc_if_xname,
1434 sdl->sdl_len, sdl->sdl_index,
1435 sdl->sdl_family, sdl->sdl_type, sdl->sdl_nlen,
1436 sdl->sdl_alen, sdl->sdl_slen, lladdr_str);
1437 }
1438
1439
1440 /*
1441 * bridgeattach:
1442 *
1443 * Pseudo-device attach routine.
1444 */
1445 __private_extern__ int
bridgeattach(int n)1446 bridgeattach(int n)
1447 {
1448 #pragma unused(n)
1449 int error;
1450
1451 LIST_INIT(&bridge_list);
1452
1453 #if BRIDGESTP
1454 bstp_sys_init();
1455 #endif /* BRIDGESTP */
1456
1457 error = if_clone_attach(&bridge_cloner);
1458 if (error != 0) {
1459 BRIDGE_LOG(LOG_NOTICE, 0, "ifnet_clone_attach failed %d", error);
1460 }
1461 return error;
1462 }
1463
1464 static void
_mbuf_adjust_pkthdr_and_data(mbuf_t m,int len)1465 _mbuf_adjust_pkthdr_and_data(mbuf_t m, int len)
1466 {
1467 mbuf_setdata(m, mtodo(m, len), mbuf_len(m) - len);
1468 mbuf_pkthdr_adjustlen(m, -len);
1469 }
1470
1471 static errno_t
bridge_ifnet_set_attrs(struct ifnet * ifp)1472 bridge_ifnet_set_attrs(struct ifnet * ifp)
1473 {
1474 errno_t error;
1475
1476 error = ifnet_set_mtu(ifp, ETHERMTU);
1477 if (error != 0) {
1478 BRIDGE_LOG(LOG_NOTICE, 0, "ifnet_set_mtu failed %d", error);
1479 goto done;
1480 }
1481 error = ifnet_set_addrlen(ifp, ETHER_ADDR_LEN);
1482 if (error != 0) {
1483 BRIDGE_LOG(LOG_NOTICE, 0, "ifnet_set_addrlen failed %d", error);
1484 goto done;
1485 }
1486 error = ifnet_set_hdrlen(ifp, ETHER_HDR_LEN);
1487 if (error != 0) {
1488 BRIDGE_LOG(LOG_NOTICE, 0, "ifnet_set_hdrlen failed %d", error);
1489 goto done;
1490 }
1491 error = ifnet_set_flags(ifp,
1492 IFF_BROADCAST | IFF_SIMPLEX | IFF_NOTRAILERS | IFF_MULTICAST,
1493 0xffff);
1494
1495 if (error != 0) {
1496 BRIDGE_LOG(LOG_NOTICE, 0, "ifnet_set_flags failed %d", error);
1497 goto done;
1498 }
1499 done:
1500 return error;
1501 }
1502
1503 /*
1504 * bridge_clone_create:
1505 *
1506 * Create a new bridge instance.
1507 */
1508 static int
bridge_clone_create(struct if_clone * ifc,uint32_t unit,void * params)1509 bridge_clone_create(struct if_clone *ifc, uint32_t unit, void *params)
1510 {
1511 #pragma unused(params)
1512 ifnet_ref_t ifp = NULL;
1513 struct bridge_softc *sc = NULL;
1514 struct bridge_softc *sc2 = NULL;
1515 struct ifnet_init_eparams init_params;
1516 errno_t error = 0;
1517 uint8_t eth_hostid[ETHER_ADDR_LEN];
1518 int fb, retry, has_hostid;
1519
1520 sc = kalloc_type(struct bridge_softc, Z_WAITOK_ZERO_NOFAIL);
1521 lck_mtx_init(&sc->sc_mtx, &bridge_lock_grp, &bridge_lock_attr);
1522 sc->sc_brtmax = BRIDGE_RTABLE_MAX;
1523 sc->sc_mne_max = BRIDGE_MAC_NAT_ENTRY_MAX;
1524 sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
1525 sc->sc_filter_flags = 0;
1526
1527 TAILQ_INIT(&sc->sc_iflist);
1528
1529 /* use the interface name as the unique id for ifp recycle */
1530 snprintf(sc->sc_if_xname, sizeof(sc->sc_if_xname), "%s%d",
1531 ifc->ifc_name, unit);
1532 bzero(&init_params, sizeof(init_params));
1533 init_params.ver = IFNET_INIT_CURRENT_VERSION;
1534 init_params.len = sizeof(init_params);
1535 /* Initialize our routing table. */
1536 error = bridge_rtable_init(sc);
1537 if (error != 0) {
1538 BRIDGE_LOG(LOG_NOTICE, 0, "bridge_rtable_init failed %d", error);
1539 goto done;
1540 }
1541 TAILQ_INIT(&sc->sc_spanlist);
1542 if (if_bridge_txstart) {
1543 init_params.start = bridge_start;
1544 } else {
1545 init_params.flags = IFNET_INIT_LEGACY;
1546 init_params.output = bridge_output;
1547 }
1548 init_params.uniqueid_len = strbuflen(sc->sc_if_xname);
1549 init_params.uniqueid = sc->sc_if_xname;
1550 init_params.sndq_maxlen = IFQ_MAXLEN;
1551 init_params.name = __unsafe_null_terminated_from_indexable(ifc->ifc_name);
1552 init_params.unit = unit;
1553 init_params.family = IFNET_FAMILY_ETHERNET;
1554 init_params.type = IFT_BRIDGE;
1555 init_params.demux = ether_demux;
1556 init_params.add_proto = ether_add_proto;
1557 init_params.del_proto = ether_del_proto;
1558 init_params.check_multi = ether_check_multi;
1559 init_params.framer_extended = ether_frameout_extended;
1560 init_params.softc = sc;
1561 init_params.ioctl = bridge_ioctl;
1562 init_params.detach = bridge_detach;
1563 init_params.broadcast_addr = etherbroadcastaddr;
1564 init_params.broadcast_len = ETHER_ADDR_LEN;
1565
1566 error = ifnet_allocate_extended(&init_params, &ifp);
1567 if (error != 0) {
1568 BRIDGE_LOG(LOG_NOTICE, 0, "ifnet_allocate failed %d", error);
1569 goto done;
1570 }
1571 LIST_INIT(&sc->sc_mne_list);
1572 LIST_INIT(&sc->sc_mne_list_v6);
1573 sc->sc_ifp = ifp;
1574 error = bridge_ifnet_set_attrs(ifp);
1575 if (error != 0) {
1576 BRIDGE_LOG(LOG_NOTICE, 0, "bridge_ifnet_set_attrs failed %d",
1577 error);
1578 goto done;
1579 }
1580 /*
1581 * Generate an ethernet address with a locally administered address.
1582 *
1583 * Since we are using random ethernet addresses for the bridge, it is
1584 * possible that we might have address collisions, so make sure that
1585 * this hardware address isn't already in use on another bridge.
1586 * The first try uses the "hostid" and falls back to read_frandom();
1587 * for "hostid", we use the MAC address of the first-encountered
1588 * Ethernet-type interface that is currently configured.
1589 */
1590 fb = 0;
1591 has_hostid = (uuid_get_ethernet(ð_hostid[0]) == 0);
1592 for (retry = 1; retry != 0;) {
1593 if (fb || has_hostid == 0) {
1594 read_frandom(&sc->sc_defaddr, ETHER_ADDR_LEN);
1595 sc->sc_defaddr[0] &= ~1; /* clear multicast bit */
1596 sc->sc_defaddr[0] |= 2; /* set the LAA bit */
1597 } else {
1598 bcopy(ð_hostid[0], &sc->sc_defaddr,
1599 ETHER_ADDR_LEN);
1600 sc->sc_defaddr[0] &= ~1; /* clear multicast bit */
1601 sc->sc_defaddr[0] |= 2; /* set the LAA bit */
1602 sc->sc_defaddr[3] = /* stir it up a bit */
1603 ((sc->sc_defaddr[3] & 0x0f) << 4) |
1604 ((sc->sc_defaddr[3] & 0xf0) >> 4);
1605 /*
1606 * Mix in the LSB as it's actually pretty significant,
1607 * see rdar://14076061
1608 */
1609 sc->sc_defaddr[4] =
1610 (((sc->sc_defaddr[4] & 0x0f) << 4) |
1611 ((sc->sc_defaddr[4] & 0xf0) >> 4)) ^
1612 sc->sc_defaddr[5];
1613 sc->sc_defaddr[5] = ifp->if_unit & 0xff;
1614 }
1615
1616 fb = 1;
1617 retry = 0;
1618 lck_mtx_lock(&bridge_list_mtx);
1619 LIST_FOREACH(sc2, &bridge_list, sc_list) {
1620 if (_ether_cmp(sc->sc_defaddr,
1621 IF_LLADDR(sc2->sc_ifp)) == 0) {
1622 retry = 1;
1623 }
1624 }
1625 lck_mtx_unlock(&bridge_list_mtx);
1626 }
1627
1628 sc->sc_flags &= ~SCF_MEDIA_ACTIVE;
1629
1630 if (BRIDGE_DBGF_ENABLED(BR_DBGF_LIFECYCLE)) {
1631 brlog_link(sc);
1632 }
1633 error = ifnet_attach(ifp, NULL);
1634 if (error != 0) {
1635 BRIDGE_LOG(LOG_NOTICE, 0, "ifnet_attach failed %d", error);
1636 goto done;
1637 }
1638
1639 error = ifnet_set_lladdr_and_type(ifp, sc->sc_defaddr, ETHER_ADDR_LEN,
1640 IFT_ETHER);
1641 if (error != 0) {
1642 BRIDGE_LOG(LOG_NOTICE, 0, "ifnet_set_lladdr_and_type failed %d",
1643 error);
1644 goto done;
1645 }
1646
1647 ifnet_set_offload(ifp,
1648 IFNET_CSUM_IP | IFNET_CSUM_TCP | IFNET_CSUM_UDP |
1649 IFNET_CSUM_TCPIPV6 | IFNET_CSUM_UDPIPV6 | IFNET_MULTIPAGES);
1650 error = bridge_set_tso(sc);
1651 if (error != 0) {
1652 BRIDGE_LOG(LOG_NOTICE, 0, "bridge_set_tso failed %d", error);
1653 goto done;
1654 }
1655 #if BRIDGESTP
1656 bstp_attach(&sc->sc_stp, &bridge_ops);
1657 #endif /* BRIDGESTP */
1658
1659 lck_mtx_lock(&bridge_list_mtx);
1660 LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
1661 lck_mtx_unlock(&bridge_list_mtx);
1662
1663 /* attach as ethernet */
1664 error = bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header),
1665 NULL, NULL);
1666
1667 done:
1668 if (error != 0) {
1669 BRIDGE_LOG(LOG_NOTICE, 0, "failed error %d", error);
1670 /* TBD: Clean up: sc, sc_rthash etc */
1671 }
1672
1673 return error;
1674 }
1675
1676 /*
1677 * bridge_clone_destroy:
1678 *
1679 * Destroy a bridge instance.
1680 */
1681 static int
bridge_clone_destroy(struct ifnet * ifp)1682 bridge_clone_destroy(struct ifnet *ifp)
1683 {
1684 struct bridge_softc * __single sc = ifp->if_softc;
1685 struct bridge_iflist *bif;
1686 errno_t error;
1687
1688 BRIDGE_LOCK(sc);
1689 if ((sc->sc_flags & SCF_DETACHING)) {
1690 BRIDGE_UNLOCK(sc);
1691 return 0;
1692 }
1693 sc->sc_flags |= SCF_DETACHING;
1694
1695 bridge_ifstop(ifp, 1);
1696
1697 bridge_cancel_delayed_call(&sc->sc_resize_call);
1698
1699 bridge_cleanup_delayed_call(&sc->sc_resize_call);
1700 bridge_cleanup_delayed_call(&sc->sc_aging_timer);
1701
1702 error = ifnet_set_flags(ifp, 0, IFF_UP);
1703 if (error != 0) {
1704 BRIDGE_LOG(LOG_NOTICE, 0, "ifnet_set_flags failed %d", error);
1705 }
1706
1707 while ((bif = TAILQ_FIRST(&sc->sc_iflist)) != NULL) {
1708 bridge_delete_member(sc, bif);
1709 }
1710
1711 while ((bif = TAILQ_FIRST(&sc->sc_spanlist)) != NULL) {
1712 bridge_delete_span(sc, bif);
1713 }
1714 BRIDGE_UNLOCK(sc);
1715
1716 error = ifnet_detach(ifp);
1717 if (error != 0) {
1718 panic("%s (%d): ifnet_detach(%p) failed %d",
1719 __func__, __LINE__, ifp, error);
1720 }
1721 return 0;
1722 }
1723
1724 #define DRVSPEC do { \
1725 if (ifd->ifd_cmd >= bridge_control_table_size) { \
1726 error = EINVAL; \
1727 break; \
1728 } \
1729 bc = &bridge_control_table[ifd->ifd_cmd]; \
1730 \
1731 if (cmd == SIOCGDRVSPEC && \
1732 (bc->bc_flags & BC_F_COPYOUT) == 0) { \
1733 error = EINVAL; \
1734 break; \
1735 } else if (cmd == SIOCSDRVSPEC && \
1736 (bc->bc_flags & BC_F_COPYOUT) != 0) { \
1737 error = EINVAL; \
1738 break; \
1739 } \
1740 \
1741 if (bc->bc_flags & BC_F_SUSER) { \
1742 error = kauth_authorize_generic(kauth_cred_get(), \
1743 KAUTH_GENERIC_ISSUSER); \
1744 if (error) \
1745 break; \
1746 } \
1747 \
1748 if (ifd->ifd_len != bc->bc_argsize || \
1749 ifd->ifd_len > sizeof (args)) { \
1750 error = EINVAL; \
1751 break; \
1752 } \
1753 \
1754 bzero(&args, sizeof (args)); \
1755 if (bc->bc_flags & BC_F_COPYIN) { \
1756 error = copyin(ifd->ifd_data, &args, ifd->ifd_len); \
1757 if (error) \
1758 break; \
1759 } \
1760 \
1761 BRIDGE_LOCK(sc); \
1762 error = (*bc->bc_func)(sc, &args, sizeof(args)); \
1763 BRIDGE_UNLOCK(sc); \
1764 if (error) \
1765 break; \
1766 \
1767 if (bc->bc_flags & BC_F_COPYOUT) \
1768 error = copyout(&args, ifd->ifd_data, ifd->ifd_len); \
1769 } while (0)
1770
1771 static boolean_t
interface_needs_input_broadcast(struct ifnet * ifp)1772 interface_needs_input_broadcast(struct ifnet * ifp)
1773 {
1774 /*
1775 * Selectively enable input broadcast only when necessary.
1776 * The bridge interface itself attaches a fake protocol
1777 * so checking for at least two protocols means that the
1778 * interface is being used for something besides bridging
1779 * and needs to see broadcast packets from other members.
1780 */
1781 return if_get_protolist(ifp, NULL, 0) >= 2;
1782 }
1783
1784 static boolean_t
bif_set_input_broadcast(struct bridge_iflist * bif,boolean_t input_broadcast)1785 bif_set_input_broadcast(struct bridge_iflist * bif, boolean_t input_broadcast)
1786 {
1787 boolean_t old_input_broadcast;
1788
1789 old_input_broadcast = (bif->bif_flags & BIFF_INPUT_BROADCAST) != 0;
1790 if (input_broadcast) {
1791 bif->bif_flags |= BIFF_INPUT_BROADCAST;
1792 } else {
1793 bif->bif_flags &= ~BIFF_INPUT_BROADCAST;
1794 }
1795 return old_input_broadcast != input_broadcast;
1796 }
1797
1798 /*
1799 * bridge_ioctl:
1800 *
1801 * Handle a control request from the operator.
1802 */
1803 static errno_t
bridge_ioctl(struct ifnet * ifp,u_long cmd,void * __sized_by (IOCPARM_LEN (cmd))data)1804 bridge_ioctl(struct ifnet *ifp, u_long cmd, void *__sized_by(IOCPARM_LEN(cmd)) data)
1805 {
1806 struct bridge_softc * __single sc = ifp->if_softc;
1807 struct ifreq *ifr = (struct ifreq *)data;
1808 struct bridge_iflist *bif;
1809 int error = 0;
1810
1811 BRIDGE_LOCK_ASSERT_NOTHELD(sc);
1812
1813 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_IOCTL,
1814 "ifp %s cmd 0x%08lx (%c%c [%lu] %c %lu)",
1815 ifp->if_xname, cmd, (cmd & IOC_IN) ? 'I' : ' ',
1816 (cmd & IOC_OUT) ? 'O' : ' ', IOCPARM_LEN(cmd),
1817 (char)IOCGROUP(cmd), cmd & 0xff);
1818
1819 switch (cmd) {
1820 case SIOCAIFADDR_IN6_32:
1821 case SIOCAIFADDR_IN6_64:
1822 case SIOCSIFADDR:
1823 case SIOCAIFADDR:
1824 ifnet_set_flags(ifp, IFF_UP, IFF_UP);
1825 BRIDGE_LOCK(sc);
1826 sc->sc_flags |= SCF_ADDRESS_ASSIGNED;
1827 BRIDGE_UNLOCK(sc);
1828 BRIDGE_LOG(LOG_NOTICE, 0,
1829 "ifp %s has address", ifp->if_xname);
1830 break;
1831
1832 case SIOCGIFMEDIA32:
1833 case SIOCGIFMEDIA64: {
1834 // cast to 32bit version to work within bounds with 32bit userspace
1835 struct ifmediareq32 *ifmr = (struct ifmediareq32 *)data;
1836 user_addr_t user_addr;
1837
1838 user_addr = (cmd == SIOCGIFMEDIA64) ?
1839 ((struct ifmediareq64 *)data)->ifmu_ulist :
1840 CAST_USER_ADDR_T(((struct ifmediareq32 *)data)->ifmu_ulist);
1841
1842 ifmr->ifm_status = IFM_AVALID;
1843 ifmr->ifm_mask = 0;
1844 ifmr->ifm_count = 1;
1845
1846 BRIDGE_LOCK(sc);
1847 if (!(sc->sc_flags & SCF_DETACHING) &&
1848 (sc->sc_flags & SCF_MEDIA_ACTIVE)) {
1849 ifmr->ifm_status |= IFM_ACTIVE;
1850 ifmr->ifm_active = ifmr->ifm_current =
1851 IFM_ETHER | IFM_AUTO;
1852 } else {
1853 ifmr->ifm_active = ifmr->ifm_current = IFM_NONE;
1854 }
1855 BRIDGE_UNLOCK(sc);
1856
1857 if (user_addr != USER_ADDR_NULL) {
1858 error = copyout(&ifmr->ifm_current, user_addr,
1859 sizeof(int));
1860 }
1861 break;
1862 }
1863
1864 case SIOCADDMULTI:
1865 case SIOCDELMULTI:
1866 break;
1867
1868 case SIOCSDRVSPEC32:
1869 case SIOCGDRVSPEC32: {
1870 union {
1871 struct ifbreq ifbreq;
1872 struct ifbifconf32 ifbifconf;
1873 struct ifbareq32 ifbareq;
1874 struct ifbaconf32 ifbaconf;
1875 struct ifbrparam ifbrparam;
1876 struct ifbropreq32 ifbropreq;
1877 } args;
1878 struct ifdrv32 *ifd = (struct ifdrv32 *)data;
1879 const struct bridge_control *bridge_control_table =
1880 bridge_control_table32, *bc;
1881
1882 DRVSPEC;
1883
1884 break;
1885 }
1886 case SIOCSDRVSPEC64:
1887 case SIOCGDRVSPEC64: {
1888 union {
1889 struct ifbreq ifbreq;
1890 struct ifbifconf64 ifbifconf;
1891 struct ifbareq64 ifbareq;
1892 struct ifbaconf64 ifbaconf;
1893 struct ifbrparam ifbrparam;
1894 struct ifbropreq64 ifbropreq;
1895 } args;
1896 struct ifdrv64 *ifd = (struct ifdrv64 *)data;
1897 const struct bridge_control *bridge_control_table =
1898 bridge_control_table64, *bc;
1899
1900 DRVSPEC;
1901
1902 break;
1903 }
1904
1905 case SIOCSIFFLAGS:
1906 if (!(ifp->if_flags & IFF_UP) &&
1907 (ifp->if_flags & IFF_RUNNING)) {
1908 /*
1909 * If interface is marked down and it is running,
1910 * then stop and disable it.
1911 */
1912 BRIDGE_LOCK(sc);
1913 bridge_ifstop(ifp, 1);
1914 BRIDGE_UNLOCK(sc);
1915 } else if ((ifp->if_flags & IFF_UP) &&
1916 !(ifp->if_flags & IFF_RUNNING)) {
1917 /*
1918 * If interface is marked up and it is stopped, then
1919 * start it.
1920 */
1921 BRIDGE_LOCK(sc);
1922 error = bridge_init(ifp);
1923 BRIDGE_UNLOCK(sc);
1924 }
1925 break;
1926
1927 case SIOCSIFLLADDR:
1928 error = ifnet_set_lladdr(ifp, ifr->ifr_addr.sa_data,
1929 ifr->ifr_addr.sa_len);
1930 if (error != 0) {
1931 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_IOCTL,
1932 "%s SIOCSIFLLADDR error %d", ifp->if_xname,
1933 error);
1934 }
1935 break;
1936
1937 case SIOCSIFMTU:
1938 if (ifr->ifr_mtu < 576) {
1939 error = EINVAL;
1940 break;
1941 }
1942 BRIDGE_LOCK(sc);
1943 if (TAILQ_EMPTY(&sc->sc_iflist)) {
1944 sc->sc_ifp->if_mtu = ifr->ifr_mtu;
1945 BRIDGE_UNLOCK(sc);
1946 break;
1947 }
1948 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) {
1949 if (bif->bif_ifp->if_mtu != (unsigned)ifr->ifr_mtu) {
1950 BRIDGE_LOG(LOG_NOTICE, 0,
1951 "%s invalid MTU: %u(%s) != %d",
1952 sc->sc_ifp->if_xname,
1953 bif->bif_ifp->if_mtu,
1954 bif->bif_ifp->if_xname, ifr->ifr_mtu);
1955 error = EINVAL;
1956 break;
1957 }
1958 }
1959 if (!error) {
1960 sc->sc_ifp->if_mtu = ifr->ifr_mtu;
1961 }
1962 BRIDGE_UNLOCK(sc);
1963 break;
1964
1965 default:
1966 error = ether_ioctl(ifp, cmd, data);
1967 if (error != 0 && error != EOPNOTSUPP) {
1968 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_IOCTL,
1969 "ifp %s cmd 0x%08lx "
1970 "(%c%c [%lu] %c %lu) failed error: %d",
1971 ifp->if_xname, cmd,
1972 (cmd & IOC_IN) ? 'I' : ' ',
1973 (cmd & IOC_OUT) ? 'O' : ' ',
1974 IOCPARM_LEN(cmd), (char)IOCGROUP(cmd),
1975 cmd & 0xff, error);
1976 }
1977 break;
1978 }
1979 BRIDGE_LOCK_ASSERT_NOTHELD(sc);
1980
1981 return error;
1982 }
1983
1984 #if HAS_IF_CAP
1985 /*
1986 * bridge_mutecaps:
1987 *
1988 * Clear or restore unwanted capabilities on the member interface
1989 */
1990 static void
bridge_mutecaps(struct bridge_softc * sc)1991 bridge_mutecaps(struct bridge_softc *sc)
1992 {
1993 struct bridge_iflist *bif;
1994 int enabled, mask;
1995
1996 /* Initial bitmask of capabilities to test */
1997 mask = BRIDGE_IFCAPS_MASK;
1998
1999 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) {
2000 /* Every member must support it or its disabled */
2001 mask &= bif->bif_savedcaps;
2002 }
2003
2004 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) {
2005 enabled = bif->bif_ifp->if_capenable;
2006 enabled &= ~BRIDGE_IFCAPS_STRIP;
2007 /* strip off mask bits and enable them again if allowed */
2008 enabled &= ~BRIDGE_IFCAPS_MASK;
2009 enabled |= mask;
2010
2011 bridge_set_ifcap(sc, bif, enabled);
2012 }
2013 }
2014
2015 static void
bridge_set_ifcap(struct bridge_softc * sc,struct bridge_iflist * bif,int set)2016 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
2017 {
2018 struct ifnet *ifp = bif->bif_ifp;
2019 struct ifreq ifr;
2020 int error;
2021
2022 bzero(&ifr, sizeof(ifr));
2023 ifr.ifr_reqcap = set;
2024
2025 if (ifp->if_capenable != set) {
2026 IFF_LOCKGIANT(ifp);
2027 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
2028 IFF_UNLOCKGIANT(ifp);
2029 if (error) {
2030 BRIDGE_LOG(LOG_NOTICE, 0,
2031 "%s error setting interface capabilities on %s",
2032 sc->sc_ifp->if_xname, ifp->if_xname);
2033 }
2034 }
2035 }
2036 #endif /* HAS_IF_CAP */
2037
2038 static errno_t
siocsifcap(struct ifnet * ifp,uint32_t cap_enable)2039 siocsifcap(struct ifnet * ifp, uint32_t cap_enable)
2040 {
2041 struct ifreq ifr;
2042
2043 bzero(&ifr, sizeof(ifr));
2044 ifr.ifr_reqcap = cap_enable;
2045 return ifnet_ioctl(ifp, 0, SIOCSIFCAP, &ifr);
2046 }
2047
2048 static const char *
enable_disable_str(boolean_t enable)2049 enable_disable_str(boolean_t enable)
2050 {
2051 return (const char * __null_terminated)(enable ? "enable" : "disable");
2052 }
2053
2054 static boolean_t
bridge_set_lro(struct ifnet * ifp,boolean_t enable)2055 bridge_set_lro(struct ifnet * ifp, boolean_t enable)
2056 {
2057 uint32_t cap_enable;
2058 uint32_t cap_supported;
2059 boolean_t changed = FALSE;
2060 boolean_t lro_enabled;
2061
2062 cap_supported = ifnet_capabilities_supported(ifp);
2063 if ((cap_supported & IFCAP_LRO) == 0) {
2064 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_LIFECYCLE,
2065 "%s doesn't support LRO",
2066 ifp->if_xname);
2067 goto done;
2068 }
2069 cap_enable = ifnet_capabilities_enabled(ifp);
2070 lro_enabled = (cap_enable & IFCAP_LRO) != 0;
2071 if (lro_enabled != enable) {
2072 errno_t error;
2073
2074 if (enable) {
2075 cap_enable |= IFCAP_LRO;
2076 } else {
2077 cap_enable &= ~IFCAP_LRO;
2078 }
2079 error = siocsifcap(ifp, cap_enable);
2080 if (error != 0) {
2081 BRIDGE_LOG(LOG_NOTICE, 0,
2082 "%s %s failed (cap 0x%x) %d",
2083 ifp->if_xname,
2084 enable_disable_str(enable),
2085 cap_enable,
2086 error);
2087 } else {
2088 changed = TRUE;
2089 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_LIFECYCLE,
2090 "%s %s success (cap 0x%x)",
2091 ifp->if_xname,
2092 enable_disable_str(enable),
2093 cap_enable);
2094 }
2095 }
2096 done:
2097 return changed;
2098 }
2099
2100 static errno_t
bridge_set_tso(struct bridge_softc * sc)2101 bridge_set_tso(struct bridge_softc *sc)
2102 {
2103 struct bridge_iflist *bif;
2104 u_int32_t tso_v4_mtu;
2105 u_int32_t tso_v6_mtu;
2106 ifnet_offload_t offload;
2107 errno_t error = 0;
2108
2109 /* By default, support TSO */
2110 offload = sc->sc_ifp->if_hwassist | IFNET_TSO_IPV4 | IFNET_TSO_IPV6;
2111 tso_v4_mtu = IP_MAXPACKET;
2112 tso_v6_mtu = IP_MAXPACKET;
2113
2114 /* Use the lowest common denominator of the members */
2115 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) {
2116 ifnet_t ifp = bif->bif_ifp;
2117
2118 if (ifp == NULL) {
2119 continue;
2120 }
2121
2122 if (offload & IFNET_TSO_IPV4) {
2123 if (ifp->if_hwassist & IFNET_TSO_IPV4) {
2124 if (tso_v4_mtu > ifp->if_tso_v4_mtu) {
2125 tso_v4_mtu = ifp->if_tso_v4_mtu;
2126 }
2127 } else {
2128 offload &= ~IFNET_TSO_IPV4;
2129 tso_v4_mtu = 0;
2130 }
2131 }
2132 if (offload & IFNET_TSO_IPV6) {
2133 if (ifp->if_hwassist & IFNET_TSO_IPV6) {
2134 if (tso_v6_mtu > ifp->if_tso_v6_mtu) {
2135 tso_v6_mtu = ifp->if_tso_v6_mtu;
2136 }
2137 } else {
2138 offload &= ~IFNET_TSO_IPV6;
2139 tso_v6_mtu = 0;
2140 }
2141 }
2142 }
2143
2144 if (offload != sc->sc_ifp->if_hwassist) {
2145 error = ifnet_set_offload(sc->sc_ifp, offload);
2146 if (error != 0) {
2147 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_LIFECYCLE,
2148 "ifnet_set_offload(%s, 0x%x) failed %d",
2149 sc->sc_ifp->if_xname, offload, error);
2150 goto done;
2151 }
2152 /*
2153 * For ifnet_set_tso_mtu() sake, the TSO MTU must be at least
2154 * as large as the interface MTU
2155 */
2156 if (sc->sc_ifp->if_hwassist & IFNET_TSO_IPV4) {
2157 if (tso_v4_mtu < sc->sc_ifp->if_mtu) {
2158 tso_v4_mtu = sc->sc_ifp->if_mtu;
2159 }
2160 error = ifnet_set_tso_mtu(sc->sc_ifp, AF_INET,
2161 tso_v4_mtu);
2162 if (error != 0) {
2163 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_LIFECYCLE,
2164 "ifnet_set_tso_mtu(%s, "
2165 "AF_INET, %u) failed %d",
2166 sc->sc_ifp->if_xname,
2167 tso_v4_mtu, error);
2168 goto done;
2169 }
2170 }
2171 if (sc->sc_ifp->if_hwassist & IFNET_TSO_IPV6) {
2172 if (tso_v6_mtu < sc->sc_ifp->if_mtu) {
2173 tso_v6_mtu = sc->sc_ifp->if_mtu;
2174 }
2175 error = ifnet_set_tso_mtu(sc->sc_ifp, AF_INET6,
2176 tso_v6_mtu);
2177 if (error != 0) {
2178 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_LIFECYCLE,
2179 "ifnet_set_tso_mtu(%s, "
2180 "AF_INET6, %u) failed %d",
2181 sc->sc_ifp->if_xname,
2182 tso_v6_mtu, error);
2183 goto done;
2184 }
2185 }
2186 }
2187 done:
2188 return error;
2189 }
2190
2191 static const char *
sanitize_ifname(char * __sized_by (IFNAMSIZ)ifname)2192 sanitize_ifname(char * __sized_by(IFNAMSIZ) ifname)
2193 {
2194 ifname[IFNAMSIZ - 1] = '\0';
2195 return __unsafe_null_terminated_from_indexable(ifname, &ifname[IFNAMSIZ - 1]);
2196 }
2197
2198 /*
2199 * bridge_lookup_member:
2200 *
2201 * Lookup a bridge member interface.
2202 */
2203 static struct bridge_iflist *
bridge_lookup_member(struct bridge_softc * sc,char * __sized_by (IFNAMSIZ)name_unsanitized)2204 bridge_lookup_member(struct bridge_softc *sc, char * __sized_by(IFNAMSIZ) name_unsanitized)
2205 {
2206 struct bridge_iflist *bif;
2207 struct ifnet *ifp;
2208 const char * __null_terminated name = sanitize_ifname(name_unsanitized);
2209
2210 BRIDGE_LOCK_ASSERT_HELD(sc);
2211
2212 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) {
2213 ifp = bif->bif_ifp;
2214 if (strcmp(ifp->if_xname, name) == 0) {
2215 return bif;
2216 }
2217 }
2218
2219 return NULL;
2220 }
2221
2222 /*
2223 * bridge_lookup_member_if:
2224 *
2225 * Lookup a bridge member interface by ifnet*.
2226 */
2227 static struct bridge_iflist *
bridge_lookup_member_if(struct bridge_softc * sc,struct ifnet * member_ifp)2228 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
2229 {
2230 struct bridge_iflist *bif;
2231
2232 BRIDGE_LOCK_ASSERT_HELD(sc);
2233
2234 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) {
2235 if (bif->bif_ifp == member_ifp) {
2236 return bif;
2237 }
2238 }
2239
2240 return NULL;
2241 }
2242
2243 static errno_t
bridge_iff_input(void * cookie,ifnet_t ifp,protocol_family_t protocol,mbuf_t * data,char ** frame_ptr)2244 bridge_iff_input(void *cookie, ifnet_t ifp, protocol_family_t protocol,
2245 mbuf_t *data, char **frame_ptr)
2246 {
2247 #pragma unused(protocol)
2248 errno_t error = 0;
2249 struct bridge_iflist *bif = (struct bridge_iflist *)cookie;
2250 struct bridge_softc *sc = bif->bif_sc;
2251 int included = 0;
2252 size_t frmlen = 0;
2253 mbuf_t m = *data;
2254
2255 if ((m->m_flags & M_PROTO1)) {
2256 goto out;
2257 }
2258
2259 if (*frame_ptr >= (char *)mbuf_datastart(m) &&
2260 *frame_ptr <= mtod(m, char *)) {
2261 included = 1;
2262 frmlen = mtod(m, char *) - *frame_ptr;
2263 }
2264 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT,
2265 "%s from %s m 0x%llx data 0x%llx frame 0x%llx %s "
2266 "frmlen %lu", sc->sc_ifp->if_xname,
2267 ifp->if_xname, (uint64_t)VM_KERNEL_ADDRPERM(m),
2268 (uint64_t)VM_KERNEL_ADDRPERM(mtod(m, void *)),
2269 (uint64_t)VM_KERNEL_ADDRPERM(*frame_ptr),
2270 included ? "inside" : "outside", frmlen);
2271 if (BRIDGE_DBGF_ENABLED(BR_DBGF_MBUF)) {
2272 brlog_mbuf(m, "bridge_iff_input[", "");
2273 brlog_ether_header((struct ether_header *)
2274 (void *)*frame_ptr);
2275 brlog_mbuf_data(m, 0, 20);
2276 }
2277 if (included == 0) {
2278 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT, "frame_ptr outside mbuf");
2279 goto out;
2280 }
2281
2282 /* Move data pointer to start of frame to the link layer header */
2283 _mbuf_adjust_pkthdr_and_data(m, -frmlen);
2284
2285 /* make sure we can access the ethernet header */
2286 if (mbuf_pkthdr_len(m) < sizeof(struct ether_header)) {
2287 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT,
2288 "short frame %lu < %lu",
2289 mbuf_pkthdr_len(m), sizeof(struct ether_header));
2290 goto out;
2291 }
2292 if (mbuf_len(m) < sizeof(struct ether_header)) {
2293 error = mbuf_pullup(data, sizeof(struct ether_header));
2294 if (error != 0) {
2295 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT,
2296 "mbuf_pullup(%lu) failed %d",
2297 sizeof(struct ether_header),
2298 error);
2299 error = EJUSTRETURN;
2300 goto out;
2301 }
2302 if (m != *data) {
2303 m = *data;
2304 *frame_ptr = mtod(m, char *);
2305 }
2306 }
2307
2308 error = bridge_input(ifp, data);
2309
2310 /* Adjust packet back to original */
2311 if (error == 0) {
2312 /* bridge_input might have modified *data */
2313 if (*data != m) {
2314 m = *data;
2315 *frame_ptr = mtod(m, char *);
2316 }
2317 _mbuf_adjust_pkthdr_and_data(m, frmlen);
2318 }
2319
2320 if (BRIDGE_DBGF_ENABLED(BR_DBGF_MBUF) &&
2321 BRIDGE_DBGF_ENABLED(BR_DBGF_INPUT)) {
2322 brlog_mbuf(m, "bridge_iff_input]", "");
2323 }
2324
2325 out:
2326 BRIDGE_LOCK_ASSERT_NOTHELD(sc);
2327
2328 return error;
2329 }
2330
2331 static errno_t
bridge_iff_output(void * cookie,ifnet_t ifp,protocol_family_t protocol,mbuf_t * data)2332 bridge_iff_output(void *cookie, ifnet_t ifp, protocol_family_t protocol,
2333 mbuf_t *data)
2334 {
2335 #pragma unused(protocol)
2336 errno_t error = 0;
2337 struct bridge_iflist *bif = (struct bridge_iflist *)cookie;
2338 struct bridge_softc *sc = bif->bif_sc;
2339 mbuf_t m = *data;
2340
2341 if ((m->m_flags & M_PROTO1)) {
2342 goto out;
2343 }
2344 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_OUTPUT,
2345 "%s from %s m 0x%llx data 0x%llx",
2346 sc->sc_ifp->if_xname, ifp->if_xname,
2347 (uint64_t)VM_KERNEL_ADDRPERM(m),
2348 (uint64_t)VM_KERNEL_ADDRPERM(mtod(m, void *)));
2349
2350 error = bridge_member_output(sc, ifp, data);
2351 if (error != 0 && error != EJUSTRETURN) {
2352 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_OUTPUT,
2353 "bridge_member_output failed error %d",
2354 error);
2355 }
2356 out:
2357 BRIDGE_LOCK_ASSERT_NOTHELD(sc);
2358
2359 return error;
2360 }
2361
2362 static void
bridge_iff_event(void * cookie,ifnet_t ifp,protocol_family_t protocol,const struct kev_msg * event_msg)2363 bridge_iff_event(void *cookie, ifnet_t ifp, protocol_family_t protocol,
2364 const struct kev_msg *event_msg)
2365 {
2366 #pragma unused(protocol)
2367 struct bridge_iflist *bif = (struct bridge_iflist *)cookie;
2368 struct bridge_softc *sc = bif->bif_sc;
2369
2370 if (event_msg->vendor_code == KEV_VENDOR_APPLE &&
2371 event_msg->kev_class == KEV_NETWORK_CLASS &&
2372 event_msg->kev_subclass == KEV_DL_SUBCLASS) {
2373 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_LIFECYCLE,
2374 "%s event_code %u - %s",
2375 ifp->if_xname, event_msg->event_code,
2376 dlil_kev_dl_code_str(event_msg->event_code));
2377
2378 switch (event_msg->event_code) {
2379 case KEV_DL_LINK_OFF:
2380 case KEV_DL_LINK_ON: {
2381 bridge_iflinkevent(ifp);
2382 #if BRIDGESTP
2383 bstp_linkstate(ifp, event_msg->event_code);
2384 #endif /* BRIDGESTP */
2385 break;
2386 }
2387 case KEV_DL_SIFFLAGS: {
2388 if ((ifp->if_flags & IFF_UP) == 0) {
2389 break;
2390 }
2391 if ((bif->bif_flags & BIFF_PROMISC) == 0) {
2392 errno_t error;
2393
2394 error = ifnet_set_promiscuous(ifp, 1);
2395 if (error != 0) {
2396 BRIDGE_LOG(LOG_NOTICE, 0,
2397 "ifnet_set_promiscuous (%s)"
2398 " failed %d", ifp->if_xname,
2399 error);
2400 } else {
2401 bif->bif_flags |= BIFF_PROMISC;
2402 }
2403 }
2404 if ((bif->bif_flags & BIFF_WIFI_INFRA) != 0 &&
2405 (bif->bif_flags & BIFF_ALL_MULTI) == 0) {
2406 errno_t error;
2407
2408 error = if_allmulti(ifp, 1);
2409 if (error != 0) {
2410 BRIDGE_LOG(LOG_NOTICE, 0,
2411 "if_allmulti (%s)"
2412 " failed %d", ifp->if_xname,
2413 error);
2414 } else {
2415 bif->bif_flags |= BIFF_ALL_MULTI;
2416 #ifdef XNU_PLATFORM_AppleTVOS
2417 ip6_forwarding = 1;
2418 #endif /* XNU_PLATFORM_AppleTVOS */
2419 }
2420 }
2421 break;
2422 }
2423 case KEV_DL_IFCAP_CHANGED: {
2424 BRIDGE_LOCK(sc);
2425 bridge_set_tso(sc);
2426 BRIDGE_UNLOCK(sc);
2427 break;
2428 }
2429 case KEV_DL_PROTO_DETACHED:
2430 case KEV_DL_PROTO_ATTACHED: {
2431 bridge_proto_attach_changed(ifp);
2432 break;
2433 }
2434 default:
2435 break;
2436 }
2437 }
2438 }
2439
2440 /*
2441 * bridge_iff_detached:
2442 *
2443 * Called when our interface filter has been detached from a
2444 * member interface.
2445 */
2446 static void
bridge_iff_detached(void * cookie,ifnet_t ifp)2447 bridge_iff_detached(void *cookie, ifnet_t ifp)
2448 {
2449 #pragma unused(cookie)
2450 struct bridge_iflist *bif;
2451 struct bridge_softc * __single sc = ifp->if_bridge;
2452
2453 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_LIFECYCLE, "%s", ifp->if_xname);
2454
2455 /* Check if the interface is a bridge member */
2456 if (sc != NULL) {
2457 BRIDGE_LOCK(sc);
2458 bif = bridge_lookup_member_if(sc, ifp);
2459 if (bif != NULL) {
2460 bridge_delete_member(sc, bif);
2461 }
2462 BRIDGE_UNLOCK(sc);
2463 return;
2464 }
2465 /* Check if the interface is a span port */
2466 lck_mtx_lock(&bridge_list_mtx);
2467 LIST_FOREACH(sc, &bridge_list, sc_list) {
2468 BRIDGE_LOCK(sc);
2469 TAILQ_FOREACH(bif, &sc->sc_spanlist, bif_next)
2470 if (ifp == bif->bif_ifp) {
2471 bridge_delete_span(sc, bif);
2472 break;
2473 }
2474 BRIDGE_UNLOCK(sc);
2475 }
2476 lck_mtx_unlock(&bridge_list_mtx);
2477 }
2478
2479 static errno_t
bridge_proto_input(ifnet_t ifp,protocol_family_t protocol,mbuf_t packet,char * header)2480 bridge_proto_input(ifnet_t ifp, protocol_family_t protocol, mbuf_t packet,
2481 char *header)
2482 {
2483 #pragma unused(protocol, packet, header)
2484 BRIDGE_LOG(LOG_NOTICE, 0, "%s unexpected packet",
2485 ifp->if_xname);
2486 return 0;
2487 }
2488
2489 static int
bridge_attach_protocol(struct ifnet * ifp)2490 bridge_attach_protocol(struct ifnet *ifp)
2491 {
2492 int error;
2493 struct ifnet_attach_proto_param reg;
2494
2495 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_LIFECYCLE, "%s", ifp->if_xname);
2496 bzero(®, sizeof(reg));
2497 reg.input = bridge_proto_input;
2498
2499 error = ifnet_attach_protocol(ifp, PF_BRIDGE, ®);
2500 if (error) {
2501 BRIDGE_LOG(LOG_NOTICE, 0,
2502 "ifnet_attach_protocol(%s) failed, %d",
2503 ifp->if_xname, error);
2504 }
2505
2506 return error;
2507 }
2508
2509 static int
bridge_detach_protocol(struct ifnet * ifp)2510 bridge_detach_protocol(struct ifnet *ifp)
2511 {
2512 int error;
2513
2514 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_LIFECYCLE, "%s", ifp->if_xname);
2515 error = ifnet_detach_protocol(ifp, PF_BRIDGE);
2516 if (error) {
2517 BRIDGE_LOG(LOG_NOTICE, 0,
2518 "ifnet_detach_protocol(%s) failed, %d",
2519 ifp->if_xname, error);
2520 }
2521
2522 return error;
2523 }
2524
2525 /*
2526 * bridge_delete_member:
2527 *
2528 * Delete the specified member interface.
2529 */
2530 static void
bridge_delete_member(struct bridge_softc * sc,struct bridge_iflist * bif)2531 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif)
2532 {
2533 #if SKYWALK
2534 boolean_t add_netagent = FALSE;
2535 #endif /* SKYWALK */
2536 uint32_t bif_flags;
2537 struct ifnet *ifs = bif->bif_ifp, *bifp = sc->sc_ifp;
2538 int lladdr_changed = 0, error;
2539 uint8_t eaddr[ETHER_ADDR_LEN];
2540 u_int32_t event_code = 0;
2541
2542 BRIDGE_LOCK_ASSERT_HELD(sc);
2543 VERIFY(ifs != NULL);
2544
2545 /*
2546 * Remove the member from the list first so it cannot be found anymore
2547 * when we release the bridge lock below
2548 */
2549 if ((bif->bif_flags & BIFF_IN_MEMBER_LIST) != 0) {
2550 bif->bif_flags &= ~BIFF_IN_MEMBER_LIST;
2551 BRIDGE_XLOCK(sc);
2552 TAILQ_REMOVE(&sc->sc_iflist, bif, bif_next);
2553 BRIDGE_XDROP(sc);
2554 }
2555 if (sc->sc_mac_nat_bif != NULL) {
2556 if (bif == sc->sc_mac_nat_bif) {
2557 bridge_mac_nat_disable(sc);
2558 } else {
2559 bridge_mac_nat_flush_entries(sc, bif);
2560 }
2561 }
2562 #if BRIDGESTP
2563 if ((bif->bif_ifflags & IFBIF_STP) != 0) {
2564 bstp_disable(&bif->bif_stp);
2565 }
2566 #endif /* BRIDGESTP */
2567
2568 /*
2569 * If removing the interface that gave the bridge its mac address, set
2570 * the mac address of the bridge to the address of the next member, or
2571 * to its default address if no members are left.
2572 */
2573 if (bridge_inherit_mac && sc->sc_ifaddr == ifs) {
2574 ifnet_release(sc->sc_ifaddr);
2575 if (TAILQ_EMPTY(&sc->sc_iflist)) {
2576 bcopy(sc->sc_defaddr, eaddr, ETHER_ADDR_LEN);
2577 sc->sc_ifaddr = NULL;
2578 } else {
2579 struct ifnet *fif =
2580 TAILQ_FIRST(&sc->sc_iflist)->bif_ifp;
2581 bcopy(IF_LLADDR(fif), eaddr, ETHER_ADDR_LEN);
2582 sc->sc_ifaddr = fif;
2583 ifnet_reference(fif); /* for sc_ifaddr */
2584 }
2585 lladdr_changed = 1;
2586 }
2587
2588 #if HAS_IF_CAP
2589 bridge_mutecaps(sc); /* recalculate now this interface is removed */
2590 #endif /* HAS_IF_CAP */
2591
2592 error = bridge_set_tso(sc);
2593 if (error != 0) {
2594 BRIDGE_LOG(LOG_NOTICE, 0, "bridge_set_tso failed %d", error);
2595 }
2596
2597 bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
2598
2599 KASSERT(bif->bif_addrcnt == 0,
2600 ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
2601
2602 /*
2603 * Update link status of the bridge based on its remaining members
2604 */
2605 event_code = bridge_updatelinkstatus(sc);
2606 bif_flags = bif->bif_flags;
2607 BRIDGE_UNLOCK(sc);
2608
2609 /* only perform these steps if the interface is still attached */
2610 if (ifnet_is_attached(ifs, 1)) {
2611 #if SKYWALK
2612 add_netagent = (bif_flags & BIFF_NETAGENT_REMOVED) != 0;
2613
2614 if ((bif_flags & BIFF_FLOWSWITCH_ATTACHED) != 0) {
2615 ifnet_detach_flowswitch_nexus(ifs);
2616 }
2617 #endif /* SKYWALK */
2618 /* disable promiscuous mode */
2619 if ((bif_flags & BIFF_PROMISC) != 0) {
2620 (void) ifnet_set_promiscuous(ifs, 0);
2621 }
2622 /* disable all multi */
2623 if ((bif_flags & BIFF_ALL_MULTI) != 0) {
2624 (void)if_allmulti(ifs, 0);
2625 }
2626 #if HAS_IF_CAP
2627 /* re-enable any interface capabilities */
2628 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
2629 #endif
2630 /* detach bridge "protocol" */
2631 if ((bif_flags & BIFF_PROTO_ATTACHED) != 0) {
2632 (void)bridge_detach_protocol(ifs);
2633 }
2634 /* detach interface filter */
2635 if ((bif_flags & BIFF_FILTER_ATTACHED) != 0) {
2636 iflt_detach(bif->bif_iff_ref);
2637 }
2638 /* re-enable LRO */
2639 if ((bif_flags & BIFF_LRO_DISABLED) != 0) {
2640 (void)bridge_set_lro(ifs, TRUE);
2641 }
2642 ifnet_decr_iorefcnt(ifs);
2643 }
2644
2645 if (lladdr_changed &&
2646 (error = ifnet_set_lladdr(bifp, eaddr, ETHER_ADDR_LEN)) != 0) {
2647 BRIDGE_LOG(LOG_NOTICE, 0, "ifnet_set_lladdr failed %d", error);
2648 }
2649
2650 if (event_code != 0) {
2651 bridge_link_event(bifp, event_code);
2652 }
2653
2654 #if BRIDGESTP
2655 bstp_destroy(&bif->bif_stp); /* prepare to free */
2656 #endif /* BRIDGESTP */
2657
2658 kfree_type(struct bridge_iflist, bif);
2659 ifs->if_bridge = NULL;
2660 #if SKYWALK
2661 if (add_netagent && ifnet_is_attached(ifs, 1)) {
2662 (void)ifnet_add_netagent(ifs);
2663 ifnet_decr_iorefcnt(ifs);
2664 }
2665 #endif /* SKYWALK */
2666
2667 ifnet_release(ifs);
2668
2669 BRIDGE_LOCK(sc);
2670 }
2671
2672 /*
2673 * bridge_delete_span:
2674 *
2675 * Delete the specified span interface.
2676 */
2677 static void
bridge_delete_span(struct bridge_softc * sc,struct bridge_iflist * bif)2678 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
2679 {
2680 BRIDGE_LOCK_ASSERT_HELD(sc);
2681
2682 KASSERT(bif->bif_ifp->if_bridge == NULL,
2683 ("%s: not a span interface", __func__));
2684
2685 ifnet_release(bif->bif_ifp);
2686
2687 TAILQ_REMOVE(&sc->sc_spanlist, bif, bif_next);
2688 kfree_type(struct bridge_iflist, bif);
2689 }
2690
2691 static int
bridge_ioctl_add(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)2692 bridge_ioctl_add(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
2693 {
2694 struct ifbreq * __single req = arg;
2695 struct bridge_iflist *bif = NULL;
2696 struct ifnet *ifs, *bifp = sc->sc_ifp;
2697 int error = 0, lladdr_changed = 0;
2698 uint8_t eaddr[ETHER_ADDR_LEN];
2699 struct iff_filter iff;
2700 u_int32_t event_code = 0;
2701 boolean_t input_broadcast;
2702 int media_active;
2703 boolean_t wifi_infra = FALSE;
2704
2705 ifs = ifunit(sanitize_ifname(req->ifbr_ifsname));
2706 if (ifs == NULL) {
2707 return ENOENT;
2708 }
2709 if (ifs->if_ioctl == NULL) { /* must be supported */
2710 return EINVAL;
2711 }
2712
2713 if (IFNET_IS_INTCOPROC(ifs) || IFNET_IS_MANAGEMENT(ifs)) {
2714 return EINVAL;
2715 }
2716
2717 /* If it's in the span list, it can't be a member. */
2718 TAILQ_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2719 if (ifs == bif->bif_ifp) {
2720 return EBUSY;
2721 }
2722 }
2723
2724 if (ifs->if_bridge == sc) {
2725 return EEXIST;
2726 }
2727
2728 if (ifs->if_bridge != NULL) {
2729 return EBUSY;
2730 }
2731
2732 switch (ifs->if_type) {
2733 case IFT_ETHER:
2734 if (strcmp(ifs->if_name, "en") == 0 &&
2735 ifs->if_subfamily == IFNET_SUBFAMILY_WIFI &&
2736 (ifs->if_eflags & IFEF_IPV4_ROUTER) == 0) {
2737 /* XXX is there a better way to identify Wi-Fi STA? */
2738 wifi_infra = TRUE;
2739 }
2740 break;
2741 case IFT_L2VLAN:
2742 case IFT_IEEE8023ADLAG:
2743 break;
2744 default:
2745 return EINVAL;
2746 }
2747
2748 /* fail to add the interface if the MTU doesn't match */
2749 if (!TAILQ_EMPTY(&sc->sc_iflist) && sc->sc_ifp->if_mtu != ifs->if_mtu) {
2750 BRIDGE_LOG(LOG_NOTICE, 0, "%s invalid MTU for %s",
2751 sc->sc_ifp->if_xname,
2752 ifs->if_xname);
2753 return EINVAL;
2754 }
2755
2756 if (wifi_infra && sc->sc_mac_nat_bif != NULL) {
2757 /* there's already an interface that's doing MAC NAT */
2758 return EBUSY;
2759 }
2760
2761 /* prevent the interface from detaching while we add the member */
2762 if (!ifnet_is_attached(ifs, 1)) {
2763 return ENXIO;
2764 }
2765
2766 /* allocate a new member */
2767 bif = kalloc_type(struct bridge_iflist, Z_WAITOK | Z_ZERO | Z_NOFAIL);
2768 bif->bif_ifp = ifs;
2769 ifnet_reference(ifs);
2770 bif->bif_ifflags |= IFBIF_LEARNING | IFBIF_DISCOVER;
2771 #if HAS_IF_CAP
2772 bif->bif_savedcaps = ifs->if_capenable;
2773 #endif /* HAS_IF_CAP */
2774 bif->bif_sc = sc;
2775 if (wifi_infra) {
2776 (void)bridge_mac_nat_enable(sc, bif);
2777 }
2778
2779 /* Allow the first Ethernet member to define the MTU */
2780 if (TAILQ_EMPTY(&sc->sc_iflist)) {
2781 sc->sc_ifp->if_mtu = ifs->if_mtu;
2782 }
2783
2784 /*
2785 * Assign the interface's MAC address to the bridge if it's the first
2786 * member and the MAC address of the bridge has not been changed from
2787 * the default (randomly) generated one.
2788 */
2789 if (bridge_inherit_mac && TAILQ_EMPTY(&sc->sc_iflist) &&
2790 _ether_cmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr) == 0) {
2791 bcopy(IF_LLADDR(ifs), eaddr, ETHER_ADDR_LEN);
2792 sc->sc_ifaddr = ifs;
2793 ifnet_reference(ifs); /* for sc_ifaddr */
2794 lladdr_changed = 1;
2795 }
2796
2797 ifs->if_bridge = sc;
2798 #if BRIDGESTP
2799 bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
2800 #endif /* BRIDGESTP */
2801
2802 #if HAS_IF_CAP
2803 /* Set interface capabilities to the intersection set of all members */
2804 bridge_mutecaps(sc);
2805 #endif /* HAS_IF_CAP */
2806
2807 /*
2808 * Respect lock ordering with DLIL lock for the following operations
2809 */
2810 BRIDGE_UNLOCK(sc);
2811
2812 /* enable promiscuous mode */
2813 error = ifnet_set_promiscuous(ifs, 1);
2814 switch (error) {
2815 case 0:
2816 bif->bif_flags |= BIFF_PROMISC;
2817 break;
2818 case ENETDOWN:
2819 case EPWROFF:
2820 BRIDGE_LOG(LOG_NOTICE, 0,
2821 "ifnet_set_promiscuous(%s) failed %d, ignoring",
2822 ifs->if_xname, error);
2823 /* Ignore error when device is not up */
2824 error = 0;
2825 break;
2826 default:
2827 BRIDGE_LOG(LOG_NOTICE, 0,
2828 "ifnet_set_promiscuous(%s) failed %d",
2829 ifs->if_xname, error);
2830 BRIDGE_LOCK(sc);
2831 goto out;
2832 }
2833 if (wifi_infra) {
2834 int this_error;
2835
2836 /* Wi-Fi doesn't really support promiscuous, set allmulti */
2837 bif->bif_flags |= BIFF_WIFI_INFRA;
2838 this_error = if_allmulti(ifs, 1);
2839 if (this_error == 0) {
2840 bif->bif_flags |= BIFF_ALL_MULTI;
2841 #ifdef XNU_PLATFORM_AppleTVOS
2842 ip6_forwarding = 1;
2843 #endif /* XNU_PLATFORM_AppleTVOS */
2844 } else {
2845 BRIDGE_LOG(LOG_NOTICE, 0,
2846 "if_allmulti(%s) failed %d, ignoring",
2847 ifs->if_xname, this_error);
2848 }
2849 }
2850 #if SKYWALK
2851 /* ensure that the flowswitch is present for native interface */
2852 if (SKYWALK_NATIVE(ifs)) {
2853 if (ifnet_attach_flowswitch_nexus(ifs)) {
2854 bif->bif_flags |= BIFF_FLOWSWITCH_ATTACHED;
2855 }
2856 }
2857 /* remove the netagent on the flowswitch (rdar://75050182) */
2858 if (if_is_fsw_netagent_enabled()) {
2859 (void)ifnet_remove_netagent(ifs);
2860 bif->bif_flags |= BIFF_NETAGENT_REMOVED;
2861 }
2862 #endif /* SKYWALK */
2863
2864 /*
2865 * install an interface filter
2866 */
2867 memset(&iff, 0, sizeof(struct iff_filter));
2868 iff.iff_cookie = bif;
2869 iff.iff_name = "com.apple.kernel.bsd.net.if_bridge";
2870 iff.iff_input = bridge_iff_input;
2871 iff.iff_output = bridge_iff_output;
2872 iff.iff_event = bridge_iff_event;
2873 iff.iff_detached = bridge_iff_detached;
2874 error = dlil_attach_filter(ifs, &iff, &bif->bif_iff_ref,
2875 DLIL_IFF_TSO | DLIL_IFF_INTERNAL | DLIL_IFF_BRIDGE);
2876 if (error != 0) {
2877 BRIDGE_LOG(LOG_NOTICE, 0, "iflt_attach failed %d", error);
2878 BRIDGE_LOCK(sc);
2879 goto out;
2880 }
2881 bif->bif_flags |= BIFF_FILTER_ATTACHED;
2882
2883 /*
2884 * install a dummy "bridge" protocol
2885 */
2886 if ((error = bridge_attach_protocol(ifs)) != 0) {
2887 if (error != 0) {
2888 BRIDGE_LOG(LOG_NOTICE, 0,
2889 "bridge_attach_protocol failed %d", error);
2890 BRIDGE_LOCK(sc);
2891 goto out;
2892 }
2893 }
2894 bif->bif_flags |= BIFF_PROTO_ATTACHED;
2895
2896 if (lladdr_changed &&
2897 (error = ifnet_set_lladdr(bifp, eaddr, ETHER_ADDR_LEN)) != 0) {
2898 BRIDGE_LOG(LOG_NOTICE, 0, "ifnet_set_lladdr failed %d", error);
2899 }
2900
2901 media_active = interface_media_active(ifs);
2902
2903 /* disable LRO */
2904 if (bridge_set_lro(ifs, FALSE)) {
2905 bif->bif_flags |= BIFF_LRO_DISABLED;
2906 }
2907
2908 /*
2909 * No failures past this point. Add the member to the list.
2910 */
2911 BRIDGE_LOCK(sc);
2912 bif->bif_flags |= BIFF_IN_MEMBER_LIST;
2913 BRIDGE_XLOCK(sc);
2914 TAILQ_INSERT_TAIL(&sc->sc_iflist, bif, bif_next);
2915 BRIDGE_XDROP(sc);
2916
2917 /* cache the member link status */
2918 if (media_active != 0) {
2919 bif->bif_flags |= BIFF_MEDIA_ACTIVE;
2920 } else {
2921 bif->bif_flags &= ~BIFF_MEDIA_ACTIVE;
2922 }
2923
2924 /* the new member may change the link status of the bridge interface */
2925 event_code = bridge_updatelinkstatus(sc);
2926
2927 /* check whether we need input broadcast or not */
2928 input_broadcast = interface_needs_input_broadcast(ifs);
2929 bif_set_input_broadcast(bif, input_broadcast);
2930 BRIDGE_UNLOCK(sc);
2931
2932 if (event_code != 0) {
2933 bridge_link_event(bifp, event_code);
2934 }
2935 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_LIFECYCLE,
2936 "%s input broadcast %s", ifs->if_xname,
2937 input_broadcast ? "ENABLED" : "DISABLED");
2938
2939 BRIDGE_LOCK(sc);
2940 bridge_set_tso(sc);
2941
2942 out:
2943 /* allow the interface to detach */
2944 ifnet_decr_iorefcnt(ifs);
2945
2946 if (error != 0) {
2947 if (bif != NULL) {
2948 bridge_delete_member(sc, bif);
2949 }
2950 } else if (IFNET_IS_VMNET(ifs)) {
2951 INC_ATOMIC_INT64_LIM(net_api_stats.nas_vmnet_total);
2952 }
2953
2954 return error;
2955 }
2956
2957 static int
bridge_ioctl_del(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)2958 bridge_ioctl_del(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
2959 {
2960 struct ifbreq * __single req = arg;
2961 struct bridge_iflist *bif;
2962
2963 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
2964 if (bif == NULL) {
2965 return ENOENT;
2966 }
2967
2968 bridge_delete_member(sc, bif);
2969
2970 return 0;
2971 }
2972
2973 static int
bridge_ioctl_purge(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)2974 bridge_ioctl_purge(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
2975 {
2976 #pragma unused(sc, arg, arg_len)
2977 return 0;
2978 }
2979
2980 static int
bridge_ioctl_gifflags(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)2981 bridge_ioctl_gifflags(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
2982 {
2983 struct ifbreq * __single req = arg;
2984 struct bridge_iflist *bif;
2985
2986 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
2987 if (bif == NULL) {
2988 return ENOENT;
2989 }
2990
2991 struct bstp_port *bp;
2992
2993 bp = &bif->bif_stp;
2994 req->ifbr_state = bp->bp_state;
2995 req->ifbr_priority = bp->bp_priority;
2996 req->ifbr_path_cost = bp->bp_path_cost;
2997 req->ifbr_proto = bp->bp_protover;
2998 req->ifbr_role = bp->bp_role;
2999 req->ifbr_stpflags = bp->bp_flags;
3000 req->ifbr_ifsflags = bif->bif_ifflags;
3001
3002 /* Copy STP state options as flags */
3003 if (bp->bp_operedge) {
3004 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
3005 }
3006 if (bp->bp_flags & BSTP_PORT_AUTOEDGE) {
3007 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
3008 }
3009 if (bp->bp_ptp_link) {
3010 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
3011 }
3012 if (bp->bp_flags & BSTP_PORT_AUTOPTP) {
3013 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
3014 }
3015 if (bp->bp_flags & BSTP_PORT_ADMEDGE) {
3016 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
3017 }
3018 if (bp->bp_flags & BSTP_PORT_ADMCOST) {
3019 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
3020 }
3021
3022 req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
3023 req->ifbr_addrcnt = bif->bif_addrcnt;
3024 req->ifbr_addrmax = bif->bif_addrmax;
3025 req->ifbr_addrexceeded = bif->bif_addrexceeded;
3026
3027 return 0;
3028 }
3029
3030 static int
bridge_ioctl_sifflags(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3031 bridge_ioctl_sifflags(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3032 {
3033 struct ifbreq * __single req = arg;
3034 struct bridge_iflist *bif;
3035 #if BRIDGESTP
3036 struct bstp_port *bp;
3037 #endif /* BRIDGESTP */
3038 errno_t error;
3039 uint32_t ifsflags;
3040
3041 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
3042 if (bif == NULL) {
3043 return ENOENT;
3044 }
3045
3046 ifsflags = req->ifbr_ifsflags;
3047 if (ifsflags & IFBIF_SPAN) {
3048 /* SPAN is readonly */
3049 return EINVAL;
3050 }
3051
3052 #define CHECKSUM_OR_MAC_NAT (IFBIF_CHECKSUM_OFFLOAD | IFBIF_MAC_NAT)
3053 if ((ifsflags & CHECKSUM_OR_MAC_NAT) == CHECKSUM_OR_MAC_NAT) {
3054 /* can't specify both MAC-NAT and checksum offload */
3055 return EINVAL;
3056 }
3057 if ((ifsflags & IFBIF_MAC_NAT) != 0) {
3058 if ((bif->bif_flags & BIFF_HOST_FILTER) != 0) {
3059 /* no host filter with MAC-NAT */
3060 return EINVAL;
3061 }
3062 error = bridge_mac_nat_enable(sc, bif);
3063 if (error != 0) {
3064 return error;
3065 }
3066 } else if (sc->sc_mac_nat_bif == bif) {
3067 bridge_mac_nat_disable(sc);
3068 }
3069
3070 #if BRIDGESTP
3071 if (ifsflags & IFBIF_STP) {
3072 if ((bif->bif_ifflags & IFBIF_STP) == 0) {
3073 error = bstp_enable(&bif->bif_stp);
3074 if (error) {
3075 return error;
3076 }
3077 }
3078 } else {
3079 if ((bif->bif_ifflags & IFBIF_STP) != 0) {
3080 bstp_disable(&bif->bif_stp);
3081 }
3082 }
3083
3084 /* Pass on STP flags */
3085 bp = &bif->bif_stp;
3086 bstp_set_edge(bp, ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
3087 bstp_set_autoedge(bp, ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
3088 bstp_set_ptp(bp, ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
3089 bstp_set_autoptp(bp, ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
3090 #else /* !BRIDGESTP */
3091 if (ifsflags & IFBIF_STP) {
3092 return EOPNOTSUPP;
3093 }
3094 #endif /* !BRIDGESTP */
3095
3096 /* Save the bits relating to the bridge */
3097 bif->bif_ifflags = ifsflags & IFBIFMASK;
3098
3099 return 0;
3100 }
3101
3102 static int
bridge_ioctl_scache(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3103 bridge_ioctl_scache(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3104 {
3105 struct ifbrparam * __single param = arg;
3106
3107 sc->sc_brtmax = param->ifbrp_csize;
3108 bridge_rttrim(sc);
3109 return 0;
3110 }
3111
3112 static int
bridge_ioctl_gcache(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3113 bridge_ioctl_gcache(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3114 {
3115 struct ifbrparam * __single param = arg;
3116
3117 param->ifbrp_csize = sc->sc_brtmax;
3118
3119 return 0;
3120 }
3121
3122 #define BRIDGE_IOCTL_GIFS do { \
3123 struct bridge_iflist *bif; \
3124 struct ifbreq breq; \
3125 char *buf, *outbuf; \
3126 unsigned int count, buflen, len; \
3127 \
3128 count = 0; \
3129 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) \
3130 count++; \
3131 TAILQ_FOREACH(bif, &sc->sc_spanlist, bif_next) \
3132 count++; \
3133 \
3134 buflen = sizeof (breq) * count; \
3135 if (bifc->ifbic_len == 0) { \
3136 bifc->ifbic_len = buflen; \
3137 return (0); \
3138 } \
3139 BRIDGE_UNLOCK(sc); \
3140 outbuf = kalloc_data(buflen, Z_WAITOK | Z_ZERO); \
3141 BRIDGE_LOCK(sc); \
3142 \
3143 count = 0; \
3144 buf = outbuf; \
3145 len = min(bifc->ifbic_len, buflen); \
3146 bzero(&breq, sizeof (breq)); \
3147 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) { \
3148 if (len < sizeof (breq)) \
3149 break; \
3150 \
3151 snprintf(breq.ifbr_ifsname, sizeof (breq.ifbr_ifsname), \
3152 "%s", bif->bif_ifp->if_xname); \
3153 /* Fill in the ifbreq structure */ \
3154 error = bridge_ioctl_gifflags(sc, &breq, sizeof(breq)); \
3155 if (error) \
3156 break; \
3157 memcpy(buf, &breq, sizeof (breq)); \
3158 count++; \
3159 buf += sizeof (breq); \
3160 len -= sizeof (breq); \
3161 } \
3162 TAILQ_FOREACH(bif, &sc->sc_spanlist, bif_next) { \
3163 if (len < sizeof (breq)) \
3164 break; \
3165 \
3166 snprintf(breq.ifbr_ifsname, \
3167 sizeof (breq.ifbr_ifsname), \
3168 "%s", bif->bif_ifp->if_xname); \
3169 breq.ifbr_ifsflags = bif->bif_ifflags; \
3170 breq.ifbr_portno \
3171 = bif->bif_ifp->if_index & 0xfff; \
3172 memcpy(buf, &breq, sizeof (breq)); \
3173 count++; \
3174 buf += sizeof (breq); \
3175 len -= sizeof (breq); \
3176 } \
3177 \
3178 BRIDGE_UNLOCK(sc); \
3179 bifc->ifbic_len = sizeof (breq) * count; \
3180 if (bifc->ifbic_len > 0) { \
3181 error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);\
3182 } \
3183 BRIDGE_LOCK(sc); \
3184 kfree_data(outbuf, buflen); \
3185 } while (0)
3186
3187 static int
bridge_ioctl_gifs64(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3188 bridge_ioctl_gifs64(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3189 {
3190 struct ifbifconf64 * __single bifc = arg;
3191 int error = 0;
3192
3193 BRIDGE_IOCTL_GIFS;
3194
3195 return error;
3196 }
3197
3198 static int
bridge_ioctl_gifs32(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3199 bridge_ioctl_gifs32(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3200 {
3201 struct ifbifconf32 * __single bifc = arg;
3202 int error = 0;
3203
3204 BRIDGE_IOCTL_GIFS;
3205
3206 return error;
3207 }
3208
3209 #define BRIDGE_IOCTL_RTS do { \
3210 struct bridge_rtnode *brt; \
3211 char *buf; \
3212 char *outbuf = NULL; \
3213 unsigned int count, buflen, len; \
3214 unsigned long now; \
3215 \
3216 if (bac->ifbac_len == 0) \
3217 return (0); \
3218 \
3219 bzero(&bareq, sizeof (bareq)); \
3220 count = 0; \
3221 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) \
3222 count++; \
3223 buflen = sizeof (bareq) * count; \
3224 \
3225 BRIDGE_UNLOCK(sc); \
3226 outbuf = kalloc_data(buflen, Z_WAITOK | Z_ZERO); \
3227 BRIDGE_LOCK(sc); \
3228 \
3229 count = 0; \
3230 buf = outbuf; \
3231 len = min(bac->ifbac_len, buflen); \
3232 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) { \
3233 if (len < sizeof (bareq)) \
3234 goto out; \
3235 snprintf(bareq.ifba_ifsname, sizeof (bareq.ifba_ifsname), \
3236 "%s", brt->brt_ifp->if_xname); \
3237 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof (brt->brt_addr)); \
3238 bareq.ifba_vlan = brt->brt_vlan; \
3239 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) { \
3240 now = (unsigned long) net_uptime(); \
3241 if (now < brt->brt_expire) \
3242 bareq.ifba_expire = \
3243 brt->brt_expire - now; \
3244 } else \
3245 bareq.ifba_expire = 0; \
3246 bareq.ifba_flags = brt->brt_flags; \
3247 \
3248 memcpy(buf, &bareq, sizeof (bareq)); \
3249 count++; \
3250 buf += sizeof (bareq); \
3251 len -= sizeof (bareq); \
3252 } \
3253 out: \
3254 bac->ifbac_len = sizeof (bareq) * count; \
3255 if (outbuf != NULL) { \
3256 BRIDGE_UNLOCK(sc); \
3257 if (bac->ifbac_len > 0) { \
3258 error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);\
3259 } \
3260 kfree_data(outbuf, buflen); \
3261 BRIDGE_LOCK(sc); \
3262 } \
3263 return (error); \
3264 } while (0)
3265
3266 static int
bridge_ioctl_rts64(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3267 bridge_ioctl_rts64(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3268 {
3269 struct ifbaconf64 * __single bac = arg;
3270 struct ifbareq64 bareq;
3271 int error = 0;
3272
3273 BRIDGE_IOCTL_RTS;
3274 return error;
3275 }
3276
3277 static int
bridge_ioctl_rts32(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3278 bridge_ioctl_rts32(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3279 {
3280 struct ifbaconf32 * __single bac = arg;
3281 struct ifbareq32 bareq;
3282 int error = 0;
3283
3284 BRIDGE_IOCTL_RTS;
3285 return error;
3286 }
3287
3288 static int
bridge_ioctl_saddr32(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3289 bridge_ioctl_saddr32(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3290 {
3291 struct ifbareq32 * __single req = arg;
3292 struct bridge_iflist *bif;
3293 int error;
3294
3295 bif = bridge_lookup_member(sc, req->ifba_ifsname);
3296 if (bif == NULL) {
3297 return ENOENT;
3298 }
3299
3300 error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
3301 req->ifba_flags);
3302
3303 return error;
3304 }
3305
3306 static int
bridge_ioctl_saddr64(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3307 bridge_ioctl_saddr64(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3308 {
3309 struct ifbareq64 * __single req = arg;
3310 struct bridge_iflist *bif;
3311 int error;
3312
3313 bif = bridge_lookup_member(sc, req->ifba_ifsname);
3314 if (bif == NULL) {
3315 return ENOENT;
3316 }
3317
3318 error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
3319 req->ifba_flags);
3320
3321 return error;
3322 }
3323
3324 static int
bridge_ioctl_sto(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3325 bridge_ioctl_sto(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3326 {
3327 struct ifbrparam * __single param = arg;
3328
3329 sc->sc_brttimeout = param->ifbrp_ctime;
3330 return 0;
3331 }
3332
3333 static int
bridge_ioctl_gto(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3334 bridge_ioctl_gto(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3335 {
3336 struct ifbrparam * __single param = arg;
3337
3338 param->ifbrp_ctime = sc->sc_brttimeout;
3339 return 0;
3340 }
3341
3342 static int
bridge_ioctl_daddr32(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3343 bridge_ioctl_daddr32(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3344 {
3345 struct ifbareq32 * __single req = arg;
3346
3347 return bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan);
3348 }
3349
3350 static int
bridge_ioctl_daddr64(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3351 bridge_ioctl_daddr64(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3352 {
3353 struct ifbareq64 * __single req = arg;
3354
3355 return bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan);
3356 }
3357
3358 static int
bridge_ioctl_flush(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3359 bridge_ioctl_flush(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3360 {
3361 struct ifbreq * __single req = arg;
3362
3363 bridge_rtflush(sc, req->ifbr_ifsflags);
3364 return 0;
3365 }
3366
3367 static int
bridge_ioctl_gpri(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3368 bridge_ioctl_gpri(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3369 {
3370 struct ifbrparam * __single param = arg;
3371 struct bstp_state *bs = &sc->sc_stp;
3372
3373 param->ifbrp_prio = bs->bs_bridge_priority;
3374 return 0;
3375 }
3376
3377 static int
bridge_ioctl_spri(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3378 bridge_ioctl_spri(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3379 {
3380 #if BRIDGESTP
3381 struct ifbrparam *param = arg;
3382
3383 return bstp_set_priority(&sc->sc_stp, param->ifbrp_prio);
3384 #else /* !BRIDGESTP */
3385 #pragma unused(sc, arg)
3386 return EOPNOTSUPP;
3387 #endif /* !BRIDGESTP */
3388 }
3389
3390 static int
bridge_ioctl_ght(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3391 bridge_ioctl_ght(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3392 {
3393 struct ifbrparam * __single param = arg;
3394 struct bstp_state *bs = &sc->sc_stp;
3395
3396 param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
3397 return 0;
3398 }
3399
3400 static int
bridge_ioctl_sht(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3401 bridge_ioctl_sht(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3402 {
3403 #if BRIDGESTP
3404 struct ifbrparam *param = arg;
3405
3406 return bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime);
3407 #else /* !BRIDGESTP */
3408 #pragma unused(sc, arg)
3409 return EOPNOTSUPP;
3410 #endif /* !BRIDGESTP */
3411 }
3412
3413 static int
bridge_ioctl_gfd(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3414 bridge_ioctl_gfd(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3415 {
3416 struct ifbrparam * __single param;
3417 struct bstp_state *bs;
3418
3419 param = arg;
3420 bs = &sc->sc_stp;
3421 param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
3422 return 0;
3423 }
3424
3425 static int
bridge_ioctl_sfd(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3426 bridge_ioctl_sfd(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3427 {
3428 #if BRIDGESTP
3429 struct ifbrparam *param = arg;
3430
3431 return bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay);
3432 #else /* !BRIDGESTP */
3433 #pragma unused(sc, arg)
3434 return EOPNOTSUPP;
3435 #endif /* !BRIDGESTP */
3436 }
3437
3438 static int
bridge_ioctl_gma(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3439 bridge_ioctl_gma(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3440 {
3441 struct ifbrparam * __single param;
3442 struct bstp_state *bs;
3443
3444 param = arg;
3445 bs = &sc->sc_stp;
3446 param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
3447 return 0;
3448 }
3449
3450 static int
bridge_ioctl_sma(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3451 bridge_ioctl_sma(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3452 {
3453 #if BRIDGESTP
3454 struct ifbrparam *param = arg;
3455
3456 return bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage);
3457 #else /* !BRIDGESTP */
3458 #pragma unused(sc, arg)
3459 return EOPNOTSUPP;
3460 #endif /* !BRIDGESTP */
3461 }
3462
3463 static int
bridge_ioctl_sifprio(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3464 bridge_ioctl_sifprio(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3465 {
3466 #if BRIDGESTP
3467 struct ifbreq *req = arg;
3468 struct bridge_iflist *bif;
3469
3470 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
3471 if (bif == NULL) {
3472 return ENOENT;
3473 }
3474
3475 return bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority);
3476 #else /* !BRIDGESTP */
3477 #pragma unused(sc, arg)
3478 return EOPNOTSUPP;
3479 #endif /* !BRIDGESTP */
3480 }
3481
3482 static int
bridge_ioctl_sifcost(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3483 bridge_ioctl_sifcost(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3484 {
3485 #if BRIDGESTP
3486 struct ifbreq *req = arg;
3487 struct bridge_iflist *bif;
3488
3489 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
3490 if (bif == NULL) {
3491 return ENOENT;
3492 }
3493
3494 return bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost);
3495 #else /* !BRIDGESTP */
3496 #pragma unused(sc, arg)
3497 return EOPNOTSUPP;
3498 #endif /* !BRIDGESTP */
3499 }
3500
3501 static int
bridge_ioctl_gfilt(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3502 bridge_ioctl_gfilt(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3503 {
3504 struct ifbrparam * __single param = arg;
3505
3506 param->ifbrp_filter = sc->sc_filter_flags;
3507
3508 return 0;
3509 }
3510
3511 static int
bridge_ioctl_sfilt(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3512 bridge_ioctl_sfilt(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3513 {
3514 struct ifbrparam * __single param = arg;
3515
3516 if (param->ifbrp_filter & ~IFBF_FILT_MASK) {
3517 return EINVAL;
3518 }
3519
3520 if (param->ifbrp_filter & IFBF_FILT_USEIPF) {
3521 return EINVAL;
3522 }
3523
3524 sc->sc_filter_flags = param->ifbrp_filter;
3525
3526 return 0;
3527 }
3528
3529 static int
bridge_ioctl_sifmaxaddr(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3530 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3531 {
3532 struct ifbreq * __single req = arg;
3533 struct bridge_iflist *bif;
3534
3535 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
3536 if (bif == NULL) {
3537 return ENOENT;
3538 }
3539
3540 bif->bif_addrmax = req->ifbr_addrmax;
3541 return 0;
3542 }
3543
3544 static int
bridge_ioctl_addspan(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3545 bridge_ioctl_addspan(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3546 {
3547 struct ifbreq * __single req = arg;
3548 struct bridge_iflist *bif = NULL;
3549 struct ifnet *ifs;
3550
3551 ifs = ifunit(sanitize_ifname(req->ifbr_ifsname));
3552 if (ifs == NULL) {
3553 return ENOENT;
3554 }
3555
3556 if (IFNET_IS_INTCOPROC(ifs) || IFNET_IS_MANAGEMENT(ifs)) {
3557 return EINVAL;
3558 }
3559
3560 TAILQ_FOREACH(bif, &sc->sc_spanlist, bif_next)
3561 if (ifs == bif->bif_ifp) {
3562 return EBUSY;
3563 }
3564
3565 if (ifs->if_bridge != NULL) {
3566 return EBUSY;
3567 }
3568
3569 switch (ifs->if_type) {
3570 case IFT_ETHER:
3571 case IFT_L2VLAN:
3572 case IFT_IEEE8023ADLAG:
3573 break;
3574 default:
3575 return EINVAL;
3576 }
3577
3578 bif = kalloc_type(struct bridge_iflist, Z_WAITOK | Z_ZERO | Z_NOFAIL);
3579
3580 bif->bif_ifp = ifs;
3581 bif->bif_ifflags = IFBIF_SPAN;
3582
3583 ifnet_reference(bif->bif_ifp);
3584
3585 TAILQ_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
3586
3587 return 0;
3588 }
3589
3590 static int
bridge_ioctl_delspan(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3591 bridge_ioctl_delspan(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3592 {
3593 struct ifbreq * __single req = arg;
3594 struct bridge_iflist *bif;
3595 struct ifnet *ifs;
3596
3597 ifs = ifunit(sanitize_ifname(req->ifbr_ifsname));
3598 if (ifs == NULL) {
3599 return ENOENT;
3600 }
3601
3602 TAILQ_FOREACH(bif, &sc->sc_spanlist, bif_next)
3603 if (ifs == bif->bif_ifp) {
3604 break;
3605 }
3606
3607 if (bif == NULL) {
3608 return ENOENT;
3609 }
3610
3611 bridge_delete_span(sc, bif);
3612
3613 return 0;
3614 }
3615
3616 #define BRIDGE_IOCTL_GBPARAM do { \
3617 struct bstp_state *bs = &sc->sc_stp; \
3618 struct bstp_port *root_port; \
3619 \
3620 req->ifbop_maxage = bs->bs_bridge_max_age >> 8; \
3621 req->ifbop_hellotime = bs->bs_bridge_htime >> 8; \
3622 req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8; \
3623 \
3624 root_port = bs->bs_root_port; \
3625 if (root_port == NULL) \
3626 req->ifbop_root_port = 0; \
3627 else \
3628 req->ifbop_root_port = root_port->bp_ifp->if_index; \
3629 \
3630 req->ifbop_holdcount = bs->bs_txholdcount; \
3631 req->ifbop_priority = bs->bs_bridge_priority; \
3632 req->ifbop_protocol = bs->bs_protover; \
3633 req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost; \
3634 req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id; \
3635 req->ifbop_designated_root = bs->bs_root_pv.pv_root_id; \
3636 req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id; \
3637 req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec; \
3638 req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec; \
3639 } while (0)
3640
3641 static int
bridge_ioctl_gbparam32(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3642 bridge_ioctl_gbparam32(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3643 {
3644 struct ifbropreq32 * __single req = arg;
3645
3646 BRIDGE_IOCTL_GBPARAM;
3647 return 0;
3648 }
3649
3650 static int
bridge_ioctl_gbparam64(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3651 bridge_ioctl_gbparam64(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3652 {
3653 struct ifbropreq64 * __single req = arg;
3654
3655 BRIDGE_IOCTL_GBPARAM;
3656 return 0;
3657 }
3658
3659 static int
bridge_ioctl_grte(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3660 bridge_ioctl_grte(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3661 {
3662 struct ifbrparam * __single param = arg;
3663
3664 param->ifbrp_cexceeded = sc->sc_brtexceeded;
3665 return 0;
3666 }
3667
3668 #define BRIDGE_IOCTL_GIFSSTP do { \
3669 struct bridge_iflist *bif; \
3670 struct bstp_port *bp; \
3671 struct ifbpstpreq bpreq; \
3672 char *buf, *outbuf; \
3673 unsigned int count, buflen, len; \
3674 \
3675 count = 0; \
3676 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) { \
3677 if ((bif->bif_ifflags & IFBIF_STP) != 0) \
3678 count++; \
3679 } \
3680 \
3681 buflen = sizeof (bpreq) * count; \
3682 if (bifstp->ifbpstp_len == 0) { \
3683 bifstp->ifbpstp_len = buflen; \
3684 return (0); \
3685 } \
3686 \
3687 BRIDGE_UNLOCK(sc); \
3688 outbuf = kalloc_data(buflen, Z_WAITOK | Z_ZERO); \
3689 BRIDGE_LOCK(sc); \
3690 \
3691 count = 0; \
3692 buf = outbuf; \
3693 len = min(bifstp->ifbpstp_len, buflen); \
3694 bzero(&bpreq, sizeof (bpreq)); \
3695 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) { \
3696 if (len < sizeof (bpreq)) \
3697 break; \
3698 \
3699 if ((bif->bif_ifflags & IFBIF_STP) == 0) \
3700 continue; \
3701 \
3702 bp = &bif->bif_stp; \
3703 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff; \
3704 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions; \
3705 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost; \
3706 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id; \
3707 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id; \
3708 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id; \
3709 \
3710 memcpy(buf, &bpreq, sizeof (bpreq)); \
3711 count++; \
3712 buf += sizeof (bpreq); \
3713 len -= sizeof (bpreq); \
3714 } \
3715 \
3716 BRIDGE_UNLOCK(sc); \
3717 bifstp->ifbpstp_len = sizeof (bpreq) * count; \
3718 if (bifstp->ifbpstp_len > 0) { \
3719 error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);\
3720 } \
3721 BRIDGE_LOCK(sc); \
3722 kfree_data(outbuf, buflen); \
3723 return (error); \
3724 } while (0)
3725
3726 static int
bridge_ioctl_gifsstp32(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3727 bridge_ioctl_gifsstp32(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3728 {
3729 struct ifbpstpconf32 * __single bifstp = arg;
3730 int error = 0;
3731
3732 BRIDGE_IOCTL_GIFSSTP;
3733 return error;
3734 }
3735
3736 static int
bridge_ioctl_gifsstp64(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3737 bridge_ioctl_gifsstp64(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3738 {
3739 struct ifbpstpconf64 * __single bifstp = arg;
3740 int error = 0;
3741
3742 BRIDGE_IOCTL_GIFSSTP;
3743 return error;
3744 }
3745
3746 static int
bridge_ioctl_sproto(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3747 bridge_ioctl_sproto(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3748 {
3749 #if BRIDGESTP
3750 struct ifbrparam *param = arg;
3751
3752 return bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto);
3753 #else /* !BRIDGESTP */
3754 #pragma unused(sc, arg)
3755 return EOPNOTSUPP;
3756 #endif /* !BRIDGESTP */
3757 }
3758
3759 static int
bridge_ioctl_stxhc(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3760 bridge_ioctl_stxhc(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3761 {
3762 #if BRIDGESTP
3763 struct ifbrparam *param = arg;
3764
3765 return bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc);
3766 #else /* !BRIDGESTP */
3767 #pragma unused(sc, arg)
3768 return EOPNOTSUPP;
3769 #endif /* !BRIDGESTP */
3770 }
3771
3772
3773 static int
bridge_ioctl_ghostfilter(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3774 bridge_ioctl_ghostfilter(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3775 {
3776 struct ifbrhostfilter * __single req = arg;
3777 struct bridge_iflist *bif;
3778
3779 bif = bridge_lookup_member(sc, req->ifbrhf_ifsname);
3780 if (bif == NULL) {
3781 return ENOENT;
3782 }
3783
3784 bzero(req, sizeof(struct ifbrhostfilter));
3785 if (bif->bif_flags & BIFF_HOST_FILTER) {
3786 req->ifbrhf_flags |= IFBRHF_ENABLED;
3787 bcopy(bif->bif_hf_hwsrc, req->ifbrhf_hwsrca,
3788 ETHER_ADDR_LEN);
3789 req->ifbrhf_ipsrc = bif->bif_hf_ipsrc.s_addr;
3790 }
3791 return 0;
3792 }
3793
3794 static int
bridge_ioctl_shostfilter(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3795 bridge_ioctl_shostfilter(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3796 {
3797 struct ifbrhostfilter * __single req = arg;
3798 struct bridge_iflist *bif;
3799
3800 bif = bridge_lookup_member(sc, req->ifbrhf_ifsname);
3801 if (bif == NULL) {
3802 return ENOENT;
3803 }
3804 if (bif_has_mac_nat(bif)) {
3805 /* no host filter with MAC-NAT */
3806 return EINVAL;
3807 }
3808 if (req->ifbrhf_flags & IFBRHF_ENABLED) {
3809 bif->bif_flags |= BIFF_HOST_FILTER;
3810
3811 if (req->ifbrhf_flags & IFBRHF_HWSRC) {
3812 bcopy(req->ifbrhf_hwsrca, bif->bif_hf_hwsrc,
3813 ETHER_ADDR_LEN);
3814 if (bcmp(req->ifbrhf_hwsrca, ethernulladdr,
3815 ETHER_ADDR_LEN) != 0) {
3816 bif->bif_flags |= BIFF_HF_HWSRC;
3817 } else {
3818 bif->bif_flags &= ~BIFF_HF_HWSRC;
3819 }
3820 }
3821 if (req->ifbrhf_flags & IFBRHF_IPSRC) {
3822 bif->bif_hf_ipsrc.s_addr = req->ifbrhf_ipsrc;
3823 if (bif->bif_hf_ipsrc.s_addr != INADDR_ANY) {
3824 bif->bif_flags |= BIFF_HF_IPSRC;
3825 } else {
3826 bif->bif_flags &= ~BIFF_HF_IPSRC;
3827 }
3828 }
3829 } else {
3830 bif->bif_flags &= ~(BIFF_HOST_FILTER | BIFF_HF_HWSRC |
3831 BIFF_HF_IPSRC);
3832 bzero(bif->bif_hf_hwsrc, ETHER_ADDR_LEN);
3833 bif->bif_hf_ipsrc.s_addr = INADDR_ANY;
3834 }
3835
3836 return 0;
3837 }
3838
3839 static char *__indexable
bridge_mac_nat_entry_out(struct mac_nat_entry_list * list,unsigned int * count_p,char * __indexable buf,unsigned int * len_p)3840 bridge_mac_nat_entry_out(struct mac_nat_entry_list * list,
3841 unsigned int * count_p, char *__indexable buf,
3842 unsigned int * len_p)
3843 {
3844 unsigned int count = *count_p;
3845 struct ifbrmne ifbmne;
3846 unsigned int len = *len_p;
3847 struct mac_nat_entry *mne;
3848 unsigned long now;
3849
3850 bzero(&ifbmne, sizeof(ifbmne));
3851 LIST_FOREACH(mne, list, mne_list) {
3852 if (len < sizeof(ifbmne)) {
3853 break;
3854 }
3855 snprintf(ifbmne.ifbmne_ifname, sizeof(ifbmne.ifbmne_ifname),
3856 "%s", mne->mne_bif->bif_ifp->if_xname);
3857 memcpy(ifbmne.ifbmne_mac, mne->mne_mac,
3858 sizeof(ifbmne.ifbmne_mac));
3859 now = (unsigned long) net_uptime();
3860 if (now < mne->mne_expire) {
3861 ifbmne.ifbmne_expire = mne->mne_expire - now;
3862 } else {
3863 ifbmne.ifbmne_expire = 0;
3864 }
3865 if ((mne->mne_flags & MNE_FLAGS_IPV6) != 0) {
3866 ifbmne.ifbmne_af = AF_INET6;
3867 ifbmne.ifbmne_ip6_addr = mne->mne_ip6;
3868 } else {
3869 ifbmne.ifbmne_af = AF_INET;
3870 ifbmne.ifbmne_ip_addr = mne->mne_ip;
3871 }
3872 memcpy(buf, &ifbmne, sizeof(ifbmne));
3873 count++;
3874 buf += sizeof(ifbmne);
3875 len -= sizeof(ifbmne);
3876 }
3877 *count_p = count;
3878 *len_p = len;
3879 return buf;
3880 }
3881
3882 /*
3883 * bridge_ioctl_gmnelist()
3884 * Perform the get mac_nat_entry list ioctl.
3885 *
3886 * Note:
3887 * The struct ifbrmnelist32 and struct ifbrmnelist64 have the same
3888 * field size/layout except for the last field ifbml_buf, the user-supplied
3889 * buffer pointer. That is passed in separately via the 'user_addr'
3890 * parameter from the respective 32-bit or 64-bit ioctl routine.
3891 */
3892 static int
bridge_ioctl_gmnelist(struct bridge_softc * sc,struct ifbrmnelist32 * mnl,user_addr_t user_addr)3893 bridge_ioctl_gmnelist(struct bridge_softc *sc, struct ifbrmnelist32 *mnl,
3894 user_addr_t user_addr)
3895 {
3896 unsigned int count;
3897 char *buf;
3898 int error = 0;
3899 char *outbuf = NULL;
3900 struct mac_nat_entry *mne;
3901 unsigned int buflen;
3902 unsigned int len;
3903
3904 mnl->ifbml_elsize = sizeof(struct ifbrmne);
3905 count = 0;
3906 LIST_FOREACH(mne, &sc->sc_mne_list, mne_list) {
3907 count++;
3908 }
3909 LIST_FOREACH(mne, &sc->sc_mne_list_v6, mne_list) {
3910 count++;
3911 }
3912 buflen = sizeof(struct ifbrmne) * count;
3913 if (buflen == 0 || mnl->ifbml_len == 0) {
3914 mnl->ifbml_len = buflen;
3915 return error;
3916 }
3917 BRIDGE_UNLOCK(sc);
3918 outbuf = kalloc_data(buflen, Z_WAITOK | Z_ZERO);
3919 BRIDGE_LOCK(sc);
3920 count = 0;
3921 buf = outbuf;
3922 len = min(mnl->ifbml_len, buflen);
3923 buf = bridge_mac_nat_entry_out(&sc->sc_mne_list, &count, buf, &len);
3924 buf = bridge_mac_nat_entry_out(&sc->sc_mne_list_v6, &count, buf, &len);
3925 mnl->ifbml_len = count * sizeof(struct ifbrmne);
3926 BRIDGE_UNLOCK(sc);
3927 if (mnl->ifbml_len > 0) {
3928 error = copyout(outbuf, user_addr, mnl->ifbml_len);
3929 }
3930 kfree_data(outbuf, buflen);
3931 BRIDGE_LOCK(sc);
3932 return error;
3933 }
3934
3935 static int
bridge_ioctl_gmnelist64(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3936 bridge_ioctl_gmnelist64(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3937 {
3938 struct ifbrmnelist64 * __single mnl = arg;
3939
3940 return bridge_ioctl_gmnelist(sc, arg, mnl->ifbml_buf);
3941 }
3942
3943 static int
bridge_ioctl_gmnelist32(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3944 bridge_ioctl_gmnelist32(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3945 {
3946 struct ifbrmnelist32 * __single mnl = arg;
3947
3948 return bridge_ioctl_gmnelist(sc, arg,
3949 CAST_USER_ADDR_T(mnl->ifbml_buf));
3950 }
3951
3952 /*
3953 * bridge_ioctl_gifstats()
3954 * Return per-member stats.
3955 *
3956 * Note:
3957 * The ifbrmreq32 and ifbrmreq64 structures have the same
3958 * field size/layout except for the last field brmr_buf, the user-supplied
3959 * buffer pointer. That is passed in separately via the 'user_addr'
3960 * parameter from the respective 32-bit or 64-bit ioctl routine.
3961 */
3962 static int
bridge_ioctl_gifstats(struct bridge_softc * sc,struct ifbrmreq32 * mreq,user_addr_t user_addr)3963 bridge_ioctl_gifstats(struct bridge_softc *sc, struct ifbrmreq32 *mreq,
3964 user_addr_t user_addr)
3965 {
3966 struct bridge_iflist *bif;
3967 int error = 0;
3968 unsigned int buflen;
3969
3970 bif = bridge_lookup_member(sc, mreq->brmr_ifname);
3971 if (bif == NULL) {
3972 error = ENOENT;
3973 goto done;
3974 }
3975
3976 buflen = mreq->brmr_elsize = sizeof(struct ifbrmstats);
3977 if (buflen == 0 || mreq->brmr_len == 0) {
3978 mreq->brmr_len = buflen;
3979 goto done;
3980 }
3981 if (mreq->brmr_len != 0 && mreq->brmr_len < buflen) {
3982 error = ENOBUFS;
3983 goto done;
3984 }
3985 mreq->brmr_len = buflen;
3986 error = copyout(&bif->bif_stats, user_addr, buflen);
3987 done:
3988 return error;
3989 }
3990
3991 static int
bridge_ioctl_gifstats32(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)3992 bridge_ioctl_gifstats32(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
3993 {
3994 struct ifbrmreq32 * __single mreq = arg;
3995
3996 return bridge_ioctl_gifstats(sc, arg, mreq->brmr_buf);
3997 }
3998
3999 static int
bridge_ioctl_gifstats64(struct bridge_softc * sc,void * __sized_by (arg_len)arg,size_t arg_len __unused)4000 bridge_ioctl_gifstats64(struct bridge_softc *sc, void *__sized_by(arg_len) arg, size_t arg_len __unused)
4001 {
4002 struct ifbrmreq64 * __single mreq = arg;
4003
4004 return bridge_ioctl_gifstats(sc, arg, mreq->brmr_buf);
4005 }
4006
4007 /*
4008 * bridge_proto_attach_changed
4009 *
4010 * Called when protocol attachment on the interface changes.
4011 */
4012 static void
bridge_proto_attach_changed(struct ifnet * ifp)4013 bridge_proto_attach_changed(struct ifnet *ifp)
4014 {
4015 boolean_t changed = FALSE;
4016 struct bridge_iflist *bif;
4017 boolean_t input_broadcast;
4018 struct bridge_softc * __single sc = ifp->if_bridge;
4019
4020 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_LIFECYCLE, "%s", ifp->if_xname);
4021 if (sc == NULL) {
4022 return;
4023 }
4024 input_broadcast = interface_needs_input_broadcast(ifp);
4025 BRIDGE_LOCK(sc);
4026 bif = bridge_lookup_member_if(sc, ifp);
4027 if (bif != NULL) {
4028 changed = bif_set_input_broadcast(bif, input_broadcast);
4029 }
4030 BRIDGE_UNLOCK(sc);
4031 if (changed) {
4032 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_LIFECYCLE,
4033 "%s input broadcast %s", ifp->if_xname,
4034 input_broadcast ? "ENABLED" : "DISABLED");
4035 }
4036 return;
4037 }
4038
4039 /*
4040 * interface_media_active:
4041 *
4042 * Tells if an interface media is active.
4043 */
4044 static int
interface_media_active(struct ifnet * ifp)4045 interface_media_active(struct ifnet *ifp)
4046 {
4047 struct ifmediareq ifmr;
4048 int status = 0;
4049
4050 bzero(&ifmr, sizeof(ifmr));
4051 if (ifnet_ioctl(ifp, 0, SIOCGIFMEDIA, &ifmr) == 0) {
4052 if ((ifmr.ifm_status & IFM_AVALID) && ifmr.ifm_count > 0) {
4053 status = ifmr.ifm_status & IFM_ACTIVE ? 1 : 0;
4054 }
4055 }
4056
4057 return status;
4058 }
4059
4060 /*
4061 * bridge_updatelinkstatus:
4062 *
4063 * Update the media active status of the bridge based on the
4064 * media active status of its member.
4065 * If changed, return the corresponding onf/off link event.
4066 */
4067 static u_int32_t
bridge_updatelinkstatus(struct bridge_softc * sc)4068 bridge_updatelinkstatus(struct bridge_softc *sc)
4069 {
4070 struct bridge_iflist *bif;
4071 int active_member = 0;
4072 u_int32_t event_code = 0;
4073
4074 BRIDGE_LOCK_ASSERT_HELD(sc);
4075
4076 /*
4077 * Find out if we have an active interface
4078 */
4079 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) {
4080 if (bif->bif_flags & BIFF_MEDIA_ACTIVE) {
4081 active_member = 1;
4082 break;
4083 }
4084 }
4085
4086 if (active_member && !(sc->sc_flags & SCF_MEDIA_ACTIVE)) {
4087 sc->sc_flags |= SCF_MEDIA_ACTIVE;
4088 event_code = KEV_DL_LINK_ON;
4089 } else if (!active_member && (sc->sc_flags & SCF_MEDIA_ACTIVE)) {
4090 sc->sc_flags &= ~SCF_MEDIA_ACTIVE;
4091 event_code = KEV_DL_LINK_OFF;
4092 }
4093
4094 return event_code;
4095 }
4096
4097 /*
4098 * bridge_iflinkevent:
4099 */
4100 static void
bridge_iflinkevent(struct ifnet * ifp)4101 bridge_iflinkevent(struct ifnet *ifp)
4102 {
4103 struct bridge_softc * __single sc = ifp->if_bridge;
4104 struct bridge_iflist *bif;
4105 u_int32_t event_code = 0;
4106 int media_active;
4107
4108 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_LIFECYCLE, "%s", ifp->if_xname);
4109
4110 /* Check if the interface is a bridge member */
4111 if (sc == NULL) {
4112 return;
4113 }
4114
4115 media_active = interface_media_active(ifp);
4116 BRIDGE_LOCK(sc);
4117 bif = bridge_lookup_member_if(sc, ifp);
4118 if (bif != NULL) {
4119 if (media_active) {
4120 bif->bif_flags |= BIFF_MEDIA_ACTIVE;
4121 } else {
4122 bif->bif_flags &= ~BIFF_MEDIA_ACTIVE;
4123 }
4124 if (sc->sc_mac_nat_bif != NULL) {
4125 bridge_mac_nat_flush_entries(sc, bif);
4126 }
4127
4128 event_code = bridge_updatelinkstatus(sc);
4129 }
4130 BRIDGE_UNLOCK(sc);
4131
4132 if (event_code != 0) {
4133 bridge_link_event(sc->sc_ifp, event_code);
4134 }
4135 }
4136
4137 /*
4138 * bridge_delayed_callback:
4139 *
4140 * Makes a delayed call
4141 */
4142 static void
bridge_delayed_callback(void * param,__unused void * param2)4143 bridge_delayed_callback(void *param, __unused void *param2)
4144 {
4145 struct bridge_delayed_call *call = (struct bridge_delayed_call *)param;
4146 struct bridge_softc *sc = call->bdc_sc;
4147
4148 #if BRIDGE_DELAYED_CALLBACK_DEBUG
4149 if (bridge_delayed_callback_delay > 0) {
4150 struct timespec ts;
4151
4152 ts.tv_sec = bridge_delayed_callback_delay;
4153 ts.tv_nsec = 0;
4154
4155 BRIDGE_LOG(LOG_NOTICE, 0,
4156 "sleeping for %d seconds",
4157 bridge_delayed_callback_delay);
4158
4159 msleep(&bridge_delayed_callback_delay, NULL, PZERO,
4160 __func__, &ts);
4161
4162 BRIDGE_LOG(LOG_NOTICE, 0, "awoken");
4163 }
4164 #endif /* BRIDGE_DELAYED_CALLBACK_DEBUG */
4165
4166 BRIDGE_LOCK(sc);
4167
4168 #if BRIDGE_DELAYED_CALLBACK_DEBUG
4169 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_DELAYED_CALL,
4170 "%s call 0x%llx flags 0x%x",
4171 sc->sc_if_xname, (uint64_t)VM_KERNEL_ADDRPERM(call),
4172 call->bdc_flags);
4173 }
4174 #endif /* BRIDGE_DELAYED_CALLBACK_DEBUG */
4175
4176 if (call->bdc_flags & BDCF_CANCELLING) {
4177 wakeup(call);
4178 } else {
4179 if ((sc->sc_flags & SCF_DETACHING) == 0) {
4180 (*call->bdc_func)(sc);
4181 }
4182 }
4183 call->bdc_flags &= ~BDCF_OUTSTANDING;
4184 BRIDGE_UNLOCK(sc);
4185 }
4186
4187 /*
4188 * bridge_schedule_delayed_call:
4189 *
4190 * Schedule a function to be called on a separate thread
4191 * The actual call may be scheduled to run at a given time or ASAP.
4192 */
4193 static void
4194 bridge_schedule_delayed_call(struct bridge_delayed_call *call)
4195 {
4196 uint64_t deadline = 0;
4197 struct bridge_softc *sc = call->bdc_sc;
4198
4199 BRIDGE_LOCK_ASSERT_HELD(sc);
4200
4201 if ((sc->sc_flags & SCF_DETACHING) ||
4202 (call->bdc_flags & (BDCF_OUTSTANDING | BDCF_CANCELLING))) {
4203 return;
4204 }
4205
4206 if (call->bdc_ts.tv_sec || call->bdc_ts.tv_nsec) {
4207 nanoseconds_to_absolutetime(
4208 (uint64_t)call->bdc_ts.tv_sec * NSEC_PER_SEC +
4209 call->bdc_ts.tv_nsec, &deadline);
4210 clock_absolutetime_interval_to_deadline(deadline, &deadline);
4211 }
4212
4213 call->bdc_flags = BDCF_OUTSTANDING;
4214
4215 #if BRIDGE_DELAYED_CALLBACK_DEBUG
4216 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_DELAYED_CALL,
4217 "%s call 0x%llx flags 0x%x",
4218 sc->sc_if_xname, (uint64_t)VM_KERNEL_ADDRPERM(call),
4219 call->bdc_flags);
4220 }
4221 #endif /* BRIDGE_DELAYED_CALLBACK_DEBUG */
4222
4223 if (call->bdc_ts.tv_sec || call->bdc_ts.tv_nsec) {
4224 thread_call_func_delayed(
4225 (thread_call_func_t)bridge_delayed_callback,
4226 call, deadline);
4227 } else {
4228 if (call->bdc_thread_call == NULL) {
4229 call->bdc_thread_call = thread_call_allocate(
4230 (thread_call_func_t)bridge_delayed_callback,
4231 call);
4232 }
4233 thread_call_enter(call->bdc_thread_call);
4234 }
4235 }
4236
4237 /*
4238 * bridge_cancel_delayed_call:
4239 *
4240 * Cancel a queued or running delayed call.
4241 * If call is running, does not return until the call is done to
4242 * prevent race condition with the brigde interface getting destroyed
4243 */
4244 static void
4245 bridge_cancel_delayed_call(struct bridge_delayed_call *call)
4246 {
4247 boolean_t result;
4248 struct bridge_softc *sc = call->bdc_sc;
4249
4250 /*
4251 * The call was never scheduled
4252 */
4253 if (sc == NULL) {
4254 return;
4255 }
4256
4257 BRIDGE_LOCK_ASSERT_HELD(sc);
4258
4259 call->bdc_flags |= BDCF_CANCELLING;
4260
4261 while (call->bdc_flags & BDCF_OUTSTANDING) {
4262 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_DELAYED_CALL,
4263 "%s call 0x%llx flags 0x%x",
4264 sc->sc_if_xname, (uint64_t)VM_KERNEL_ADDRPERM(call),
4265 call->bdc_flags);
4266 result = thread_call_func_cancel(
4267 (thread_call_func_t)bridge_delayed_callback, call, FALSE);
4268
4269 if (result) {
4270 /*
4271 * We managed to dequeue the delayed call
4272 */
4273 call->bdc_flags &= ~BDCF_OUTSTANDING;
4274 } else {
4275 /*
4276 * Wait for delayed call do be done running
4277 */
4278 msleep(call, &sc->sc_mtx, PZERO, __func__, NULL);
4279 }
4280 }
4281 call->bdc_flags &= ~BDCF_CANCELLING;
4282 }
4283
4284 /*
4285 * bridge_cleanup_delayed_call:
4286 *
4287 * Dispose resource allocated for a delayed call
4288 * Assume the delayed call is not queued or running .
4289 */
4290 static void
4291 bridge_cleanup_delayed_call(struct bridge_delayed_call *call)
4292 {
4293 boolean_t result;
4294 struct bridge_softc *sc = call->bdc_sc;
4295
4296 /*
4297 * The call was never scheduled
4298 */
4299 if (sc == NULL) {
4300 return;
4301 }
4302
4303 BRIDGE_LOCK_ASSERT_HELD(sc);
4304
4305 VERIFY((call->bdc_flags & BDCF_OUTSTANDING) == 0);
4306 VERIFY((call->bdc_flags & BDCF_CANCELLING) == 0);
4307
4308 if (call->bdc_thread_call != NULL) {
4309 result = thread_call_free(call->bdc_thread_call);
4310 if (result == FALSE) {
4311 panic("%s thread_call_free() failed for call %p",
4312 __func__, call);
4313 }
4314 call->bdc_thread_call = NULL;
4315 }
4316 }
4317
4318 /*
4319 * bridge_init:
4320 *
4321 * Initialize a bridge interface.
4322 */
4323 static int
4324 bridge_init(struct ifnet *ifp)
4325 {
4326 struct bridge_softc *sc = (struct bridge_softc *)ifp->if_softc;
4327 errno_t error;
4328
4329 BRIDGE_LOCK_ASSERT_HELD(sc);
4330
4331 if ((ifnet_flags(ifp) & IFF_RUNNING)) {
4332 return 0;
4333 }
4334
4335 error = ifnet_set_flags(ifp, IFF_RUNNING, IFF_RUNNING);
4336
4337 /*
4338 * Calling bridge_aging_timer() is OK as there are no entries to
4339 * age so we're just going to arm the timer
4340 */
4341 bridge_aging_timer(sc);
4342 #if BRIDGESTP
4343 if (error == 0) {
4344 bstp_init(&sc->sc_stp); /* Initialize Spanning Tree */
4345 }
4346 #endif /* BRIDGESTP */
4347 return error;
4348 }
4349
4350 /*
4351 * bridge_ifstop:
4352 *
4353 * Stop the bridge interface.
4354 */
4355 static void
4356 bridge_ifstop(struct ifnet *ifp, int disable)
4357 {
4358 #pragma unused(disable)
4359 struct bridge_softc * __single sc = ifp->if_softc;
4360
4361 BRIDGE_LOCK_ASSERT_HELD(sc);
4362
4363 if ((ifnet_flags(ifp) & IFF_RUNNING) == 0) {
4364 return;
4365 }
4366
4367 bridge_cancel_delayed_call(&sc->sc_aging_timer);
4368
4369 #if BRIDGESTP
4370 bstp_stop(&sc->sc_stp);
4371 #endif /* BRIDGESTP */
4372
4373 bridge_rtflush(sc, IFBF_FLUSHDYN);
4374 (void) ifnet_set_flags(ifp, 0, IFF_RUNNING);
4375 }
4376
4377 /*
4378 * bridge_compute_cksum:
4379 *
4380 * If the packet has checksum flags, compare the hardware checksum
4381 * capabilities of the source and destination interfaces. If they
4382 * are the same, there's nothing to do. If they are different,
4383 * finalize the checksum so that it can be sent on the destination
4384 * interface.
4385 */
4386 static void
4387 bridge_compute_cksum(struct ifnet *src_if, struct ifnet *dst_if, struct mbuf *m)
4388 {
4389 uint32_t csum_flags;
4390 uint16_t dst_hw_csum;
4391 uint32_t did_sw = 0;
4392 struct ether_header *eh;
4393 uint16_t src_hw_csum;
4394
4395 if (src_if == dst_if) {
4396 return;
4397 }
4398 csum_flags = m->m_pkthdr.csum_flags & IF_HWASSIST_CSUM_MASK;
4399 if (csum_flags == 0) {
4400 /* no checksum offload */
4401 return;
4402 }
4403
4404 /*
4405 * if destination/source differ in checksum offload
4406 * capabilities, finalize/compute the checksum
4407 */
4408 dst_hw_csum = IF_HWASSIST_CSUM_FLAGS(dst_if->if_hwassist);
4409 src_hw_csum = IF_HWASSIST_CSUM_FLAGS(src_if->if_hwassist);
4410 if (dst_hw_csum == src_hw_csum) {
4411 return;
4412 }
4413 eh = mtod(m, struct ether_header *);
4414 switch (eh->ether_type) {
4415 case HTONS_ETHERTYPE_IP:
4416 did_sw = in_finalize_cksum(m, sizeof(*eh), csum_flags);
4417 break;
4418 case HTONS_ETHERTYPE_IPV6:
4419 did_sw = in6_finalize_cksum(m, sizeof(*eh), -1, -1, csum_flags);
4420 break;
4421 }
4422 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4423 "[%s -> %s] before 0x%x did 0x%x after 0x%x",
4424 src_if->if_xname, dst_if->if_xname, csum_flags, did_sw,
4425 m->m_pkthdr.csum_flags);
4426 }
4427
4428 static inline errno_t
4429 bridge_transmit(ifnet_t ifp, mbuf_t m)
4430 {
4431 struct flowadv adv = { .code = FADV_SUCCESS };
4432 errno_t error;
4433 int flags = DLIL_OUTPUT_FLAGS_RAW;
4434
4435 flags = (if_bridge_output_skip_filters != 0)
4436 ? (DLIL_OUTPUT_FLAGS_RAW | DLIL_OUTPUT_FLAGS_SKIP_IF_FILTERS)
4437 : DLIL_OUTPUT_FLAGS_RAW;
4438 error = dlil_output(ifp, 0, m, NULL, NULL, flags, &adv);
4439 if (error == 0) {
4440 if (adv.code == FADV_FLOW_CONTROLLED) {
4441 error = EQFULL;
4442 } else if (adv.code == FADV_SUSPENDED) {
4443 error = EQSUSPENDED;
4444 }
4445 }
4446 return error;
4447 }
4448
4449 static int
4450 get_last_ip6_hdr(struct mbuf *m, int off, int proto, int * nxtp,
4451 bool *is_fragmented)
4452 {
4453 int newoff;
4454
4455 *is_fragmented = false;
4456 while (1) {
4457 newoff = ip6_nexthdr(m, off, proto, nxtp);
4458 if (newoff < 0) {
4459 return off;
4460 } else if (newoff < off) {
4461 return -1; /* invalid */
4462 } else if (newoff == off) {
4463 return newoff;
4464 }
4465 off = newoff;
4466 proto = *nxtp;
4467 if (proto == IPPROTO_FRAGMENT) {
4468 *is_fragmented = true;
4469 }
4470 }
4471 }
4472
4473 #define __ATOMIC_INC(s) os_atomic_inc(&s, relaxed)
4474
4475 static int
4476 bridge_get_ip_proto(struct mbuf * * mp, u_int mac_hlen, bool is_ipv4,
4477 ip_packet_info_t info_p, struct bripstats * stats_p)
4478 {
4479 int error = 0;
4480 u_int hlen;
4481 u_int ip_hlen;
4482 u_int ip_pay_len;
4483 struct mbuf * m0 = *mp;
4484 int off;
4485 int opt_len = 0;
4486 int proto = 0;
4487
4488 bzero(info_p, sizeof(*info_p));
4489 if (is_ipv4) {
4490 struct ip * ip;
4491 u_int ip_total_len;
4492
4493 /* IPv4 */
4494 hlen = mac_hlen + sizeof(struct ip);
4495 if (m0->m_pkthdr.len < hlen) {
4496 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4497 "Short IP packet %d < %d",
4498 m0->m_pkthdr.len, hlen);
4499 error = _EBADIP;
4500 __ATOMIC_INC(stats_p->bips_bad_ip);
4501 goto done;
4502 }
4503 if (m0->m_len < hlen) {
4504 *mp = m0 = m_pullup(m0, hlen);
4505 if (m0 == NULL) {
4506 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4507 "m_pullup failed hlen %d",
4508 hlen);
4509 error = ENOBUFS;
4510 __ATOMIC_INC(stats_p->bips_bad_ip);
4511 goto done;
4512 }
4513 }
4514 ip = (struct ip *)mtodo(m0, mac_hlen);
4515 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
4516 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4517 "bad IP version");
4518 error = _EBADIP;
4519 __ATOMIC_INC(stats_p->bips_bad_ip);
4520 goto done;
4521 }
4522 ip_hlen = IP_VHL_HL(ip->ip_vhl) << 2;
4523 if (ip_hlen < sizeof(struct ip)) {
4524 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4525 "bad IP header length %d < %d",
4526 ip_hlen,
4527 (int)sizeof(struct ip));
4528 error = _EBADIP;
4529 __ATOMIC_INC(stats_p->bips_bad_ip);
4530 goto done;
4531 }
4532 hlen = mac_hlen + ip_hlen;
4533 if (m0->m_len < hlen) {
4534 *mp = m0 = m_pullup(m0, hlen);
4535 if (m0 == NULL) {
4536 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4537 "m_pullup failed hlen %d",
4538 hlen);
4539 error = ENOBUFS;
4540 __ATOMIC_INC(stats_p->bips_bad_ip);
4541 goto done;
4542 }
4543 ip = (struct ip *)mtodo(m0, mac_hlen);
4544 }
4545
4546 ip_total_len = ntohs(ip->ip_len);
4547 if (ip_total_len < ip_hlen) {
4548 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4549 "IP total len %d < header len %d",
4550 ip_total_len, ip_hlen);
4551 error = _EBADIP;
4552 __ATOMIC_INC(stats_p->bips_bad_ip);
4553 goto done;
4554 }
4555 if (ip_total_len > (m0->m_pkthdr.len - mac_hlen)) {
4556 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4557 "invalid IP payload length %d > %d",
4558 ip_total_len,
4559 (m0->m_pkthdr.len - mac_hlen));
4560 error = _EBADIP;
4561 __ATOMIC_INC(stats_p->bips_bad_ip);
4562 goto done;
4563 }
4564 ip_pay_len = ip_total_len - ip_hlen;
4565 info_p->ip_proto = ip->ip_p;
4566 info_p->ip_hdr = mtodo(m0, mac_hlen);
4567 info_p->ip_m0_len = m0->m_len - mac_hlen;
4568 info_p->ip_hlen = ip_hlen;
4569 #define FRAG_BITS (IP_OFFMASK | IP_MF)
4570 if ((ntohs(ip->ip_off) & FRAG_BITS) != 0) {
4571 info_p->ip_is_fragmented = true;
4572 }
4573 __ATOMIC_INC(stats_p->bips_ip);
4574 } else {
4575 struct ip6_hdr *ip6;
4576
4577 /* IPv6 */
4578 hlen = mac_hlen + sizeof(struct ip6_hdr);
4579 if (m0->m_pkthdr.len < hlen) {
4580 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4581 "short IPv6 packet %d < %d",
4582 m0->m_pkthdr.len, hlen);
4583 error = _EBADIPV6;
4584 __ATOMIC_INC(stats_p->bips_bad_ip6);
4585 goto done;
4586 }
4587 if (m0->m_len < hlen) {
4588 *mp = m0 = m_pullup(m0, hlen);
4589 if (m0 == NULL) {
4590 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4591 "m_pullup failed hlen %d",
4592 hlen);
4593 error = ENOBUFS;
4594 __ATOMIC_INC(stats_p->bips_bad_ip6);
4595 goto done;
4596 }
4597 }
4598 ip6 = (struct ip6_hdr *)(mtodo(m0, mac_hlen));
4599 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
4600 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4601 "bad IPv6 version");
4602 error = _EBADIPV6;
4603 __ATOMIC_INC(stats_p->bips_bad_ip6);
4604 goto done;
4605 }
4606 off = get_last_ip6_hdr(m0, mac_hlen, IPPROTO_IPV6, &proto,
4607 &info_p->ip_is_fragmented);
4608 if (off < 0 || m0->m_pkthdr.len < off) {
4609 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4610 "ip6_lasthdr() returned %d",
4611 off);
4612 error = _EBADIPV6;
4613 __ATOMIC_INC(stats_p->bips_bad_ip6);
4614 goto done;
4615 }
4616 ip_hlen = sizeof(*ip6);
4617 opt_len = off - mac_hlen - ip_hlen;
4618 if (opt_len < 0) {
4619 error = _EBADIPV6;
4620 __ATOMIC_INC(stats_p->bips_bad_ip6);
4621 goto done;
4622 }
4623 ip_pay_len = ntohs(ip6->ip6_plen);
4624 if (ip_pay_len > (m0->m_pkthdr.len - mac_hlen - ip_hlen)) {
4625 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4626 "invalid IPv6 payload length %d > %d",
4627 ip_pay_len,
4628 (m0->m_pkthdr.len - mac_hlen - ip_hlen));
4629 error = _EBADIPV6;
4630 __ATOMIC_INC(stats_p->bips_bad_ip6);
4631 goto done;
4632 }
4633 info_p->ip_proto = proto;
4634 info_p->ip_hdr = mtodo(m0, mac_hlen);
4635 info_p->ip_m0_len = m0->m_len - mac_hlen;
4636 info_p->ip_hlen = ip_hlen;
4637 __ATOMIC_INC(stats_p->bips_ip6);
4638 }
4639 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4640 "IPv%c proto %d ip %u pay %u opt %u pkt %u%s",
4641 is_ipv4 ? '4' : '6',
4642 proto, ip_hlen, ip_pay_len, opt_len,
4643 m0->m_pkthdr.len, info_p->ip_is_fragmented ? " frag" : "");
4644 info_p->ip_pay_len = ip_pay_len;
4645 info_p->ip_opt_len = opt_len;
4646 info_p->ip_is_ipv4 = is_ipv4;
4647 done:
4648 return error;
4649 }
4650
4651 static int
4652 bridge_get_tcp_header(struct mbuf * * mp, u_int mac_hlen, bool is_ipv4,
4653 ip_packet_info_t info_p, struct bripstats * stats_p)
4654 {
4655 int error;
4656 u_int hlen;
4657
4658 error = bridge_get_ip_proto(mp, mac_hlen, is_ipv4, info_p, stats_p);
4659 if (error != 0) {
4660 goto done;
4661 }
4662 if (info_p->ip_proto != IPPROTO_TCP) {
4663 /* not a TCP frame, not an error, just a bad guess */
4664 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4665 "non-TCP (%d) IPv%c frame %d bytes",
4666 info_p->ip_proto, is_ipv4 ? '4' : '6',
4667 (*mp)->m_pkthdr.len);
4668 goto done;
4669 }
4670 if (info_p->ip_is_fragmented) {
4671 /* both TSO and IP fragmentation don't make sense */
4672 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_CHECKSUM,
4673 "fragmented TSO packet?");
4674 __ATOMIC_INC(stats_p->bips_bad_tcp);
4675 error = _EBADTCP;
4676 goto done;
4677 }
4678 hlen = mac_hlen + info_p->ip_hlen + sizeof(struct tcphdr) +
4679 info_p->ip_opt_len;
4680 if ((*mp)->m_len < hlen) {
4681 *mp = m_pullup(*mp, hlen);
4682 if (*mp == NULL) {
4683 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4684 "m_pullup %d failed",
4685 hlen);
4686 __ATOMIC_INC(stats_p->bips_bad_tcp);
4687 error = _EBADTCP;
4688 goto done;
4689 }
4690 }
4691 info_p->ip_proto_hdr = info_p->ip_hdr + info_p->ip_hlen +
4692 info_p->ip_opt_len;
4693 done:
4694 return error;
4695 }
4696
4697 static inline void
4698 proto_csum_stats_increment(uint8_t proto, struct brcsumstats * stats_p)
4699 {
4700 if (proto == IPPROTO_TCP) {
4701 __ATOMIC_INC(stats_p->brcs_tcp_checksum);
4702 } else {
4703 __ATOMIC_INC(stats_p->brcs_udp_checksum);
4704 }
4705 return;
4706 }
4707
4708 #define ETHER_TYPE_FLAG_NONE 0x00
4709 #define ETHER_TYPE_FLAG_IPV4 0x01
4710 #define ETHER_TYPE_FLAG_IPV6 0x02
4711 #define ETHER_TYPE_FLAG_ARP 0x04
4712 #define ETHER_TYPE_FLAG_IP (ETHER_TYPE_FLAG_IPV4 | ETHER_TYPE_FLAG_IPV6)
4713 #define ETHER_TYPE_FLAG_IP_ARP (ETHER_TYPE_FLAG_IP | ETHER_TYPE_FLAG_ARP)
4714
4715 typedef uint8_t ether_type_flag_t;
4716
4717 static inline bool
4718 ether_type_flag_is_ip(ether_type_flag_t flag)
4719 {
4720 return (flag & ETHER_TYPE_FLAG_IP) != 0;
4721 }
4722
4723 static inline ether_type_flag_t
4724 ether_type_flag_get(uint16_t ether_type)
4725 {
4726 ether_type_flag_t flag = ETHER_TYPE_FLAG_NONE;
4727
4728 switch (ether_type) {
4729 case HTONS_ETHERTYPE_IP:
4730 flag = ETHER_TYPE_FLAG_IPV4;
4731 break;
4732 case HTONS_ETHERTYPE_IPV6:
4733 flag = ETHER_TYPE_FLAG_IPV6;
4734 break;
4735 case HTONS_ETHERTYPE_ARP:
4736 flag = ETHER_TYPE_FLAG_ARP;
4737 break;
4738 default:
4739 break;
4740 }
4741 return flag;
4742 }
4743
4744 static bool
4745 ether_header_type_is_ip(struct ether_header * eh, bool *is_ipv4)
4746 {
4747 uint16_t ether_type;
4748 bool is_ip = TRUE;
4749
4750 ether_type = ntohs(eh->ether_type);
4751 switch (ether_type) {
4752 case ETHERTYPE_IP:
4753 *is_ipv4 = TRUE;
4754 break;
4755 case ETHERTYPE_IPV6:
4756 *is_ipv4 = FALSE;
4757 break;
4758 default:
4759 is_ip = FALSE;
4760 break;
4761 }
4762 return is_ip;
4763 }
4764
4765 static errno_t
4766 bridge_verify_checksum(struct mbuf * * mp, struct ifbrmstats *stats_p)
4767 {
4768 struct brcsumstats *csum_stats_p;
4769 struct ether_header *eh;
4770 errno_t error = 0;
4771 ip_packet_info info;
4772 bool is_ipv4;
4773 struct mbuf * m;
4774 u_int mac_hlen = sizeof(struct ether_header);
4775 uint16_t sum;
4776 bool valid;
4777
4778 eh = mtod(*mp, struct ether_header *);
4779 if (!ether_header_type_is_ip(eh, &is_ipv4)) {
4780 goto done;
4781 }
4782 error = bridge_get_ip_proto(mp, mac_hlen, is_ipv4, &info,
4783 &stats_p->brms_out_ip);
4784 m = *mp;
4785 if (error != 0) {
4786 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4787 "bridge_get_ip_proto failed %d",
4788 error);
4789 goto done;
4790 }
4791 if (is_ipv4) {
4792 if ((m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) != 0) {
4793 /* hardware offloaded IP header checksum */
4794 valid = (m->m_pkthdr.csum_flags & CSUM_IP_VALID) != 0;
4795 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4796 "IP checksum HW %svalid",
4797 valid ? "" : "in");
4798 if (!valid) {
4799 __ATOMIC_INC(stats_p->brms_out_cksum_bad_hw.brcs_ip_checksum);
4800 error = _EBADIPCHECKSUM;
4801 goto done;
4802 }
4803 __ATOMIC_INC(stats_p->brms_out_cksum_good_hw.brcs_ip_checksum);
4804 } else {
4805 /* verify */
4806 sum = inet_cksum(m, 0, mac_hlen, info.ip_hlen);
4807 valid = (sum == 0);
4808 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4809 "IP checksum SW %svalid",
4810 valid ? "" : "in");
4811 if (!valid) {
4812 __ATOMIC_INC(stats_p->brms_out_cksum_bad.brcs_ip_checksum);
4813 error = _EBADIPCHECKSUM;
4814 goto done;
4815 }
4816 __ATOMIC_INC(stats_p->brms_out_cksum_good.brcs_ip_checksum);
4817 }
4818 }
4819 if (info.ip_is_fragmented) {
4820 /* can't verify checksum on fragmented packets */
4821 goto done;
4822 }
4823 switch (info.ip_proto) {
4824 case IPPROTO_TCP:
4825 __ATOMIC_INC(stats_p->brms_out_ip.bips_tcp);
4826 break;
4827 case IPPROTO_UDP:
4828 __ATOMIC_INC(stats_p->brms_out_ip.bips_udp);
4829 break;
4830 default:
4831 goto done;
4832 }
4833 /* check for hardware offloaded UDP/TCP checksum */
4834 #define HW_CSUM (CSUM_DATA_VALID | CSUM_PSEUDO_HDR)
4835 if ((m->m_pkthdr.csum_flags & HW_CSUM) == HW_CSUM) {
4836 /* checksum verified by hardware */
4837 valid = (m->m_pkthdr.csum_rx_val == 0xffff);
4838 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4839 "IPv%c %s checksum HW 0x%x %svalid",
4840 is_ipv4 ? '4' : '6',
4841 (info.ip_proto == IPPROTO_TCP)
4842 ? "TCP" : "UDP",
4843 m->m_pkthdr.csum_data,
4844 valid ? "" : "in" );
4845 if (!valid) {
4846 /* bad checksum */
4847 csum_stats_p = &stats_p->brms_out_cksum_bad_hw;
4848 error = (info.ip_proto == IPPROTO_TCP) ? _EBADTCPCHECKSUM
4849 : _EBADTCPCHECKSUM;
4850 } else {
4851 /* good checksum */
4852 csum_stats_p = &stats_p->brms_out_cksum_good_hw;
4853 }
4854 proto_csum_stats_increment(info.ip_proto, csum_stats_p);
4855 goto done;
4856 }
4857 /* adjust frame to skip mac-layer header */
4858 _mbuf_adjust_pkthdr_and_data(m, mac_hlen);
4859 if (is_ipv4) {
4860 sum = inet_cksum(m, info.ip_proto,
4861 info.ip_hlen,
4862 info.ip_pay_len);
4863 } else {
4864 sum = inet6_cksum(m, info.ip_proto,
4865 info.ip_hlen + info.ip_opt_len,
4866 info.ip_pay_len - info.ip_opt_len);
4867 }
4868 valid = (sum == 0);
4869 if (valid) {
4870 csum_stats_p = &stats_p->brms_out_cksum_good;
4871 } else {
4872 csum_stats_p = &stats_p->brms_out_cksum_bad;
4873 error = (info.ip_proto == IPPROTO_TCP)
4874 ? _EBADTCPCHECKSUM : _EBADUDPCHECKSUM;
4875 }
4876 proto_csum_stats_increment(info.ip_proto, csum_stats_p);
4877 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4878 "IPv%c %s checksum SW %svalid (0x%x) hlen %d paylen %d",
4879 is_ipv4 ? '4' : '6',
4880 (info.ip_proto == IPPROTO_TCP) ? "TCP" : "UDP",
4881 valid ? "" : "in",
4882 sum, info.ip_hlen, info.ip_pay_len);
4883 /* adjust frame back to start of mac-layer header */
4884 _mbuf_adjust_pkthdr_and_data(m, -mac_hlen);
4885
4886 done:
4887 return error;
4888 }
4889
4890 static mbuf_t
4891 bridge_verify_checksum_list(mbuf_t in_list, struct bridge_iflist * dbif)
4892 {
4893 mbuf_t next_packet;
4894 mblist ret;
4895
4896 mblist_init(&ret);
4897 for (mbuf_ref_t scan = in_list; scan != NULL; scan = next_packet) {
4898 errno_t error;
4899
4900 /* take packet out of the list */
4901 next_packet = scan->m_nextpkt;
4902 scan->m_nextpkt = NULL;
4903
4904 /* verify checksum */
4905 error = bridge_verify_checksum(&scan, &dbif->bif_stats);
4906 if (error != 0) {
4907 if (scan != NULL) {
4908 m_freem(scan);
4909 scan = NULL;
4910 }
4911 }
4912
4913 /* add it back to the list */
4914 if (scan != NULL) {
4915 mblist_append(&ret, scan);
4916 }
4917 }
4918 return ret.head;
4919 }
4920
4921
4922 static errno_t
4923 bridge_offload_checksum(struct mbuf * * mp, ip_packet_info * info_p,
4924 struct ifbrmstats * stats_p)
4925 {
4926 uint16_t * csum_p;
4927 errno_t error = 0;
4928 u_int hlen;
4929 struct mbuf * m0 = *mp;
4930 u_int mac_hlen = sizeof(struct ether_header);
4931 u_int pkt_hdr_len;
4932 struct tcphdr * tcp;
4933 u_int tcp_hlen;
4934 struct udphdr * udp;
4935
4936 if (info_p->ip_is_ipv4) {
4937 /* compute IP header checksum */
4938 struct ip *ip = (struct ip *)info_p->ip_hdr;
4939 ip->ip_sum = 0;
4940 ip->ip_sum = inet_cksum(m0, 0, mac_hlen, info_p->ip_hlen);
4941 __ATOMIC_INC(stats_p->brms_in_computed_cksum.brcs_ip_checksum);
4942 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4943 "IPv4 checksum 0x%x",
4944 ntohs(ip->ip_sum));
4945 }
4946 if (info_p->ip_is_fragmented) {
4947 /* can't compute checksum on fragmented packets */
4948 goto done;
4949 }
4950 pkt_hdr_len = m0->m_pkthdr.len;
4951 switch (info_p->ip_proto) {
4952 case IPPROTO_TCP:
4953 hlen = mac_hlen + info_p->ip_hlen + info_p->ip_opt_len
4954 + sizeof(struct tcphdr);
4955 if (m0->m_len < hlen) {
4956 *mp = m0 = m_pullup(m0, hlen);
4957 if (m0 == NULL) {
4958 __ATOMIC_INC(stats_p->brms_in_ip.bips_bad_tcp);
4959 error = _EBADTCP;
4960 goto done;
4961 }
4962 }
4963 tcp = (struct tcphdr *)(info_p->ip_hdr + info_p->ip_hlen
4964 + info_p->ip_opt_len);
4965 tcp_hlen = tcp->th_off << 2;
4966 hlen = mac_hlen + info_p->ip_hlen + info_p->ip_opt_len + tcp_hlen;
4967 if (hlen > pkt_hdr_len) {
4968 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4969 "bad tcp header length %u",
4970 tcp_hlen);
4971 __ATOMIC_INC(stats_p->brms_in_ip.bips_bad_tcp);
4972 error = _EBADTCP;
4973 goto done;
4974 }
4975 csum_p = &tcp->th_sum;
4976 __ATOMIC_INC(stats_p->brms_in_ip.bips_tcp);
4977 break;
4978 case IPPROTO_UDP:
4979 hlen = mac_hlen + info_p->ip_hlen + info_p->ip_opt_len + sizeof(*udp);
4980 if (m0->m_len < hlen) {
4981 *mp = m0 = m_pullup(m0, hlen);
4982 if (m0 == NULL) {
4983 __ATOMIC_INC(stats_p->brms_in_ip.bips_bad_udp);
4984 error = ENOBUFS;
4985 goto done;
4986 }
4987 }
4988 udp = (struct udphdr *)(info_p->ip_hdr + info_p->ip_hlen
4989 + info_p->ip_opt_len);
4990 csum_p = &udp->uh_sum;
4991 __ATOMIC_INC(stats_p->brms_in_ip.bips_udp);
4992 break;
4993 default:
4994 /* not TCP or UDP */
4995 goto done;
4996 }
4997 *csum_p = 0;
4998 /* adjust frame to skip mac-layer header */
4999 _mbuf_adjust_pkthdr_and_data(m0, mac_hlen);
5000 if (info_p->ip_is_ipv4) {
5001 *csum_p = inet_cksum(m0, info_p->ip_proto, info_p->ip_hlen,
5002 info_p->ip_pay_len);
5003 } else {
5004 *csum_p = inet6_cksum(m0, info_p->ip_proto,
5005 info_p->ip_hlen + info_p->ip_opt_len,
5006 info_p->ip_pay_len - info_p->ip_opt_len);
5007 }
5008 if (info_p->ip_proto == IPPROTO_UDP && *csum_p == 0) {
5009 /* RFC 1122 4.1.3.4 */
5010 *csum_p = 0xffff;
5011 }
5012 /* adjust frame back to start of mac-layer header */
5013 _mbuf_adjust_pkthdr_and_data(m0, -mac_hlen);
5014 proto_csum_stats_increment(info_p->ip_proto,
5015 &stats_p->brms_in_computed_cksum);
5016
5017 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
5018 "IPv%c %s set checksum 0x%x",
5019 info_p->ip_is_ipv4 ? '4' : '6',
5020 (info_p->ip_proto == IPPROTO_TCP) ? "TCP" : "UDP",
5021 ntohs(*csum_p));
5022 done:
5023 return error;
5024 }
5025
5026 static inline void
5027 bridge_handle_checksum_op(ifnet_t src_ifp, ifnet_t dst_ifp,
5028 mbuf_t m, ChecksumOperation cksum_op)
5029 {
5030 switch (cksum_op) {
5031 case CHECKSUM_OPERATION_CLEAR_OFFLOAD:
5032 m->m_pkthdr.csum_flags &= ~CSUM_TX_FLAGS;
5033 break;
5034 case CHECKSUM_OPERATION_FINALIZE:
5035 /* the checksum might not be correct, finalize now */
5036 VERIFY(dst_ifp != NULL);
5037 bridge_finalize_cksum(dst_ifp, m);
5038 break;
5039 case CHECKSUM_OPERATION_COMPUTE:
5040 VERIFY(dst_ifp != NULL && src_ifp != NULL);
5041 bridge_compute_cksum(src_ifp, dst_ifp, m);
5042 break;
5043 default:
5044 break;
5045 }
5046 return;
5047 }
5048
5049 static errno_t
5050 bridge_send(ifnet_t src_ifp, ifnet_t dst_ifp,
5051 mbuf_t m, ChecksumOperation cksum_op)
5052 {
5053 bridge_handle_checksum_op(src_ifp, dst_ifp, m, cksum_op);
5054 #if HAS_IF_CAP
5055 /*
5056 * If underlying interface can not do VLAN tag insertion itself
5057 * then attach a packet tag that holds it.
5058 */
5059 if ((m->m_flags & M_VLANTAG) &&
5060 (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
5061 m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
5062 if (m == NULL) {
5063 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_CHECKSUM,
5064 "%s: unable to prepend VLAN header",
5065 dst_ifp->if_xname);
5066 (void) ifnet_stat_increment_out(dst_ifp,
5067 0, 0, 1);
5068 return 0;
5069 }
5070 m->m_flags &= ~M_VLANTAG;
5071 }
5072 #endif /* HAS_IF_CAP */
5073 return bridge_transmit(dst_ifp, m);
5074 }
5075
5076 static errno_t
5077 bridge_send_tso(struct ifnet *dst_ifp, struct mbuf *m, bool is_ipv4)
5078 {
5079 errno_t error = 0;
5080 u_int mac_hlen;
5081 mblist segs;
5082
5083 mac_hlen = sizeof(struct ether_header);
5084
5085 #if HAS_IF_CAP
5086 /*
5087 * If underlying interface can not do VLAN tag insertion itself
5088 * then attach a packet tag that holds it.
5089 */
5090 if ((m->m_flags & M_VLANTAG) &&
5091 (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
5092 m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
5093 if (m == NULL) {
5094 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_CHECKSUM,
5095 "%s: unable to prepend VLAN header",
5096 dst_ifp->if_xname);
5097 (void) ifnet_stat_increment_out(dst_ifp,
5098 0, 0, 1);
5099 error = ENOBUFS;
5100 goto done;
5101 }
5102 m->m_flags &= ~M_VLANTAG;
5103 mac_hlen += ETHER_VLAN_ENCAP_LEN;
5104 }
5105 #endif /* HAS_IF_CAP */
5106 segs = gso_tcp_transmit(dst_ifp, m, mac_hlen, is_ipv4);
5107 if (segs.head != NULL) {
5108 error = bridge_transmit(dst_ifp, segs.head);
5109 }
5110 return error;
5111 }
5112
5113 static uint32_t
5114 get_if_tso_mtu(struct ifnet * ifp, bool is_ipv4)
5115 {
5116 uint32_t tso_mtu;
5117
5118 tso_mtu = is_ipv4 ? ifp->if_tso_v4_mtu : ifp->if_tso_v6_mtu;
5119 if (tso_mtu == 0) {
5120 tso_mtu = IP_MAXPACKET;
5121 }
5122
5123 #if DEBUG || DEVELOPMENT
5124 #define REDUCED_TSO_MTU (16 * 1024)
5125 if (if_bridge_reduce_tso_mtu != 0 && tso_mtu > REDUCED_TSO_MTU) {
5126 tso_mtu = REDUCED_TSO_MTU;
5127 }
5128 #endif /* DEBUG || DEVELOPMENT */
5129 return tso_mtu;
5130 }
5131
5132 /*
5133 * tso_hwassist:
5134 * - determine whether the destination interface supports TSO offload
5135 * - if the packet is already marked for offload and the hardware supports
5136 * it, just allow the packet to continue on
5137 * - if not, parse the packet headers to verify that this is a large TCP
5138 * packet requiring segmentation; if the hardware doesn't support it
5139 * set need_sw_tso; otherwise, mark the packet for TSO offload
5140 */
5141 static int
5142 tso_hwassist(struct mbuf **mp, bool is_ipv4, struct ifnet * ifp, u_int mac_hlen,
5143 bool * need_sw_tso, bool * is_large_tcp)
5144 {
5145 int error = 0;
5146 u_int32_t if_csum;
5147 u_int32_t if_tso;
5148 u_int32_t mbuf_tso;
5149 bool supports_cksum = false;
5150
5151 *need_sw_tso = false;
5152 *is_large_tcp = false;
5153 if (is_ipv4) {
5154 /*
5155 * Enable both TCP and IP offload if the hardware supports it.
5156 * If the hardware doesn't support TCP offload, supports_cksum
5157 * will be false so we won't set either offload.
5158 */
5159 if_csum = ifp->if_hwassist & (CSUM_TCP | CSUM_IP);
5160 supports_cksum = (if_csum & CSUM_TCP) != 0;
5161 if_tso = IFNET_TSO_IPV4;
5162 mbuf_tso = CSUM_TSO_IPV4;
5163 } else {
5164 supports_cksum = (ifp->if_hwassist & CSUM_TCPIPV6) != 0;
5165 if_csum = CSUM_TCPIPV6;
5166 if_tso = IFNET_TSO_IPV6;
5167 mbuf_tso = CSUM_TSO_IPV6;
5168 }
5169 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
5170 "%s: does%s support checksum 0x%x if_csum 0x%x",
5171 ifp->if_xname, supports_cksum ? "" : " not",
5172 ifp->if_hwassist, if_csum);
5173 if ((ifp->if_hwassist & if_tso) != 0 &&
5174 ((*mp)->m_pkthdr.csum_flags & mbuf_tso) != 0) {
5175 /* hardware TSO, mbuf already marked */
5176 } else {
5177 /* verify that this is a large TCP frame */
5178 uint32_t csum_flags;
5179 ip_packet_info info;
5180 int mss;
5181 uint32_t pkt_mtu;
5182 struct bripstats stats;
5183 struct tcphdr * tcp;
5184 uint32_t tso_mtu;
5185
5186 error = bridge_get_tcp_header(mp, mac_hlen, is_ipv4,
5187 &info, &stats);
5188 if (error != 0) {
5189 /* bad packet */
5190 goto done;
5191 }
5192 if (info.ip_proto_hdr == NULL) {
5193 /* not a TCP packet */
5194 goto done;
5195 }
5196 pkt_mtu = info.ip_hlen + info.ip_pay_len + info.ip_opt_len;
5197 if (pkt_mtu <= ifp->if_mtu) {
5198 /* not actually a large packet */
5199 goto done;
5200 }
5201 if ((ifp->if_hwassist & if_tso) == 0) {
5202 /* hardware does not support TSO, enable sw tso */
5203 *need_sw_tso = if_bridge_segmentation != 0;
5204 goto done;
5205 }
5206 tso_mtu = get_if_tso_mtu(ifp, is_ipv4);
5207 if (pkt_mtu > tso_mtu) {
5208 /* hardware can't segment this, enable sw tso */
5209 *need_sw_tso = if_bridge_segmentation != 0;
5210 goto done;
5211 }
5212
5213 /* use hardware TSO */
5214 (*mp)->m_pkthdr.pkt_proto = IPPROTO_TCP;
5215 tcp = (struct tcphdr *)info.ip_proto_hdr;
5216 mss = ifp->if_mtu - info.ip_hlen - info.ip_opt_len
5217 - (tcp->th_off << 2) - if_bridge_tso_reduce_mss_tx;
5218 assert(mss > 0);
5219 csum_flags = mbuf_tso;
5220 if (supports_cksum) {
5221 csum_flags |= if_csum;
5222 }
5223 (*mp)->m_pkthdr.tso_segsz = mss;
5224 (*mp)->m_pkthdr.csum_flags |= csum_flags;
5225 (*mp)->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum);
5226 *is_large_tcp = true;
5227 }
5228 done:
5229 return error;
5230 }
5231
5232 /*
5233 * bridge_enqueue:
5234 *
5235 * Enqueue a packet on a bridge member interface.
5236 *
5237 */
5238 static errno_t
5239 bridge_enqueue(ifnet_t bridge_ifp, struct ifnet *src_ifp,
5240 struct ifnet *dst_ifp, struct mbuf *m, ChecksumOperation cksum_op)
5241 {
5242 errno_t error = 0;
5243 int len;
5244
5245 VERIFY(dst_ifp != NULL);
5246
5247 for (struct mbuf *next_m = NULL; m != NULL; m = next_m) {
5248 bool need_sw_tso = false;
5249 bool is_ipv4 = false;
5250 bool is_large_pkt;
5251 errno_t _error = 0;
5252
5253 len = m->m_pkthdr.len;
5254 m->m_flags |= M_PROTO1; /* set to avoid loops */
5255 next_m = m->m_nextpkt;
5256 m->m_nextpkt = NULL;
5257 /*
5258 * Need to segment the packet if it is a large frame
5259 * and the destination interface does not support TSO.
5260 *
5261 * Note that with trailers, it's possible for a packet to
5262 * be large but not actually require segmentation.
5263 */
5264 is_large_pkt = (len > (bridge_ifp->if_mtu + ETHER_HDR_LEN));
5265 if (is_large_pkt) {
5266 struct ether_header *eh;
5267 bool is_large_tcp = false;
5268
5269 eh = mtod(m, struct ether_header *);
5270 if (ether_header_type_is_ip(eh, &is_ipv4)) {
5271 _error = tso_hwassist(&m, is_ipv4,
5272 dst_ifp, sizeof(struct ether_header),
5273 &need_sw_tso, &is_large_tcp);
5274 if (is_large_tcp) {
5275 cksum_op = CHECKSUM_OPERATION_NONE;
5276 }
5277 } else {
5278 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
5279 "large non IP packet");
5280 }
5281 }
5282 if (_error != 0) {
5283 if (m != NULL) {
5284 m_freem(m);
5285 }
5286 } else if (need_sw_tso) {
5287 _error = bridge_send_tso(dst_ifp, m, is_ipv4);
5288 } else {
5289 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
5290 "%s bridge_send(%s) len %d op %d",
5291 bridge_ifp->if_xname,
5292 dst_ifp->if_xname,
5293 len, cksum_op);
5294 _error = bridge_send(src_ifp, dst_ifp, m, cksum_op);
5295 }
5296
5297 /* Preserve first error value */
5298 if (error == 0 && _error != 0) {
5299 error = _error;
5300 }
5301 if (_error == 0) {
5302 (void) ifnet_stat_increment_out(bridge_ifp, 1, len, 0);
5303 } else {
5304 (void) ifnet_stat_increment_out(bridge_ifp, 0, 0, 1);
5305 }
5306 }
5307
5308 return error;
5309 }
5310
5311 static void
5312 bridge_enqueue_multi(ifnet_t bridge_ifp, ifnet_t src_if, ifnet_t dst_if,
5313 ether_type_flag_t etypef, mbuf_t in_list, ChecksumOperation orig_cksum_op)
5314 {
5315 mbuf_t next_packet;
5316 uint32_t out_errors = 0;
5317 mblist out_list;
5318
5319 VERIFY(dst_if != NULL);
5320
5321 mblist_init(&out_list);
5322 for (mbuf_ref_t scan = in_list; scan != NULL; scan = next_packet) {
5323 ChecksumOperation cksum_op = orig_cksum_op;
5324 errno_t error = 0;
5325 bool is_ipv4 = false;
5326 bool is_large_pkt;
5327 int len;
5328 bool need_sw_tso = false;
5329
5330 scan->m_flags |= M_PROTO1; /* set to avoid loops */
5331 next_packet = scan->m_nextpkt;
5332 scan->m_nextpkt = NULL;
5333 len = mbuf_pkthdr_len(scan);
5334
5335 /*
5336 * Need to segment the packet if it is a large frame
5337 * and the destination interface does not support TSO.
5338 *
5339 * Note that with trailers, it's possible for a packet to
5340 * be large but not actually require segmentation.
5341 */
5342 is_large_pkt = (len > (bridge_ifp->if_mtu + ETHER_HDR_LEN));
5343 if (is_large_pkt) {
5344 bool is_large_tcp = false;
5345
5346 if (ether_type_flag_is_ip(etypef)) {
5347 is_ipv4 = (etypef == ETHER_TYPE_FLAG_IPV4);
5348 error = tso_hwassist(&scan, is_ipv4,
5349 dst_if, sizeof(struct ether_header),
5350 &need_sw_tso, &is_large_tcp);
5351 if (is_large_tcp) {
5352 cksum_op = CHECKSUM_OPERATION_NONE;
5353 }
5354 } else {
5355 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
5356 "large non IP packet");
5357 }
5358 }
5359 if (error != 0) {
5360 if (scan != NULL) {
5361 m_freem(scan);
5362 scan = NULL;
5363 }
5364 out_errors++;
5365 } else if (need_sw_tso) {
5366 int mac_hlen = sizeof(struct ether_header);
5367 mblist segs;
5368
5369 /* segment packets, add to list */
5370 segs = gso_tcp_transmit(dst_if, scan, mac_hlen,
5371 is_ipv4);
5372 if (segs.head != NULL) {
5373 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
5374 "%s (%s) append gso #segs %u bytes %u",
5375 bridge_ifp->if_xname,
5376 dst_if->if_xname,
5377 segs.count, segs.bytes);
5378 mblist_append_list(&out_list, segs);
5379 } else {
5380 out_errors++;
5381 }
5382 } else {
5383 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
5384 "%s (%s) append %d bytes op %d",
5385 bridge_ifp->if_xname,
5386 dst_if->if_xname,
5387 len, cksum_op);
5388 bridge_handle_checksum_op(src_if, dst_if,
5389 scan, cksum_op);
5390 mblist_append(&out_list, scan);
5391 }
5392 }
5393 if (out_list.head != NULL) {
5394 errno_t error;
5395
5396 error = bridge_transmit(dst_if, out_list.head);
5397 if (error != 0) {
5398 out_errors++;
5399 }
5400 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
5401 "%s (%s) bridge_transmit packets %u bytes %u error %d",
5402 bridge_ifp->if_xname,
5403 dst_if->if_xname,
5404 out_list.count, out_list.bytes, error);
5405 }
5406 if (out_list.count != 0 || out_errors != 0) {
5407 ifnet_stat_increment_out(bridge_ifp, out_list.count,
5408 out_list.bytes, out_errors);
5409 }
5410 return;
5411 }
5412
5413 /*
5414 * bridge_member_output:
5415 *
5416 * Send output from a bridge member interface. This
5417 * performs the bridging function for locally originated
5418 * packets.
5419 *
5420 * The mbuf has the Ethernet header already attached.
5421 */
5422 static errno_t
5423 bridge_member_output(struct bridge_softc *sc, ifnet_t ifp, mbuf_t *data)
5424 {
5425 ifnet_t bridge_ifp;
5426 struct ether_header *eh;
5427 struct ifnet *dst_if;
5428 uint16_t vlan;
5429 struct bridge_iflist *mac_nat_bif;
5430 ifnet_t mac_nat_ifp;
5431 mbuf_t m = *data;
5432
5433 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_OUTPUT,
5434 "ifp %s", ifp->if_xname);
5435 if (m->m_len < ETHER_HDR_LEN) {
5436 m = m_pullup(m, ETHER_HDR_LEN);
5437 if (m == NULL) {
5438 *data = NULL;
5439 return EJUSTRETURN;
5440 }
5441 }
5442
5443 eh = mtod(m, struct ether_header *);
5444 vlan = VLANTAGOF(m);
5445
5446 BRIDGE_LOCK(sc);
5447 mac_nat_bif = sc->sc_mac_nat_bif;
5448 mac_nat_ifp = (mac_nat_bif != NULL) ? mac_nat_bif->bif_ifp : NULL;
5449 if (mac_nat_ifp == ifp) {
5450 /* record the IP address used by the MAC NAT interface */
5451 (void)bridge_mac_nat_output(sc, mac_nat_bif, data, NULL);
5452 m = *data;
5453 if (m == NULL) {
5454 /* packet was deallocated */
5455 BRIDGE_UNLOCK(sc);
5456 return EJUSTRETURN;
5457 }
5458 }
5459 bridge_ifp = sc->sc_ifp;
5460
5461 /*
5462 * APPLE MODIFICATION
5463 * If the packet is an 802.1X ethertype, then only send on the
5464 * original output interface.
5465 */
5466 if (eh->ether_type == htons(ETHERTYPE_PAE)) {
5467 dst_if = ifp;
5468 goto sendunicast;
5469 }
5470
5471 /*
5472 * If bridge is down, but the original output interface is up,
5473 * go ahead and send out that interface. Otherwise, the packet
5474 * is dropped below.
5475 */
5476 if ((bridge_ifp->if_flags & IFF_RUNNING) == 0) {
5477 dst_if = ifp;
5478 goto sendunicast;
5479 }
5480
5481 /*
5482 * If the packet is a multicast, or we don't know a better way to
5483 * get there, send to all interfaces.
5484 */
5485 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
5486 dst_if = NULL;
5487 } else {
5488 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
5489 }
5490 if (dst_if == NULL) {
5491 struct bridge_iflist *bif;
5492 struct mbuf *mc;
5493 errno_t error;
5494
5495
5496 bridge_span(sc, m);
5497
5498 BRIDGE_LOCK2REF(sc, error);
5499 if (error != 0) {
5500 m_freem(m);
5501 return EJUSTRETURN;
5502 }
5503
5504 /*
5505 * Duplicate and send the packet across all member interfaces
5506 * except the originating interface.
5507 */
5508 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) {
5509 dst_if = bif->bif_ifp;
5510 if (dst_if == ifp) {
5511 /* skip the originating interface */
5512 continue;
5513 }
5514 /* skip interface with inactive link status */
5515 if ((bif->bif_flags & BIFF_MEDIA_ACTIVE) == 0) {
5516 continue;
5517 }
5518
5519 /* skip interface that isn't running */
5520 if ((dst_if->if_flags & IFF_RUNNING) == 0) {
5521 continue;
5522 }
5523 /*
5524 * If the interface is participating in spanning
5525 * tree, make sure the port is in a state that
5526 * allows forwarding.
5527 */
5528 if ((bif->bif_ifflags & IFBIF_STP) &&
5529 bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
5530 continue;
5531 }
5532 /*
5533 * If the destination is the MAC NAT interface,
5534 * skip sending the packet. The packet can't be sent
5535 * if the source MAC is incorrect.
5536 */
5537 if (dst_if == mac_nat_ifp) {
5538 continue;
5539 }
5540
5541 /* make a deep copy to send on this member interface */
5542 mc = m_dup(m, M_DONTWAIT);
5543 if (mc == NULL) {
5544 (void)ifnet_stat_increment_out(bridge_ifp,
5545 0, 0, 1);
5546 continue;
5547 }
5548 (void)bridge_enqueue(bridge_ifp, ifp, dst_if,
5549 mc, CHECKSUM_OPERATION_COMPUTE);
5550 }
5551 BRIDGE_UNREF(sc);
5552
5553 if ((ifp->if_flags & IFF_RUNNING) == 0) {
5554 m_freem(m);
5555 return EJUSTRETURN;
5556 }
5557 /* allow packet to continue on the originating interface */
5558 return 0;
5559 }
5560
5561 sendunicast:
5562 /*
5563 * XXX Spanning tree consideration here?
5564 */
5565
5566 bridge_span(sc, m);
5567 if ((dst_if->if_flags & IFF_RUNNING) == 0) {
5568 m_freem(m);
5569 BRIDGE_UNLOCK(sc);
5570 return EJUSTRETURN;
5571 }
5572
5573 BRIDGE_UNLOCK(sc);
5574 if (dst_if == ifp) {
5575 /* allow packet to continue on the originating interface */
5576 return 0;
5577 }
5578 if (dst_if != mac_nat_ifp) {
5579 (void) bridge_enqueue(bridge_ifp, ifp, dst_if, m,
5580 CHECKSUM_OPERATION_COMPUTE);
5581 } else {
5582 /*
5583 * This is not the original output interface
5584 * and the destination is the MAC NAT interface.
5585 * Drop the packet because the packet can't be sent
5586 * if the source MAC is incorrect.
5587 */
5588 m_freem(m);
5589 }
5590 return EJUSTRETURN;
5591 }
5592
5593 /*
5594 * Output callback.
5595 *
5596 * This routine is called externally from above only when if_bridge_txstart
5597 * is disabled; otherwise it is called internally by bridge_start().
5598 */
5599 static int
5600 bridge_output(struct ifnet *ifp, struct mbuf *m)
5601 {
5602 struct bridge_softc * __single sc = ifnet_softc(ifp);
5603 struct ether_header *eh;
5604 struct ifnet *dst_if = NULL;
5605 int error = 0;
5606
5607 eh = mtod(m, struct ether_header *);
5608
5609 BRIDGE_LOCK(sc);
5610
5611 if (!IS_BCAST_MCAST(m)) {
5612 dst_if = bridge_rtlookup(sc, eh->ether_dhost, 0);
5613 }
5614
5615 (void) ifnet_stat_increment_out(ifp, 1, m->m_pkthdr.len, 0);
5616
5617 BRIDGE_BPF_TAP_OUT(ifp, m);
5618
5619 if (dst_if == NULL) {
5620 /* callee will unlock */
5621 bridge_broadcast(sc, NULL, m);
5622 } else {
5623 ifnet_t bridge_ifp;
5624
5625 bridge_ifp = sc->sc_ifp;
5626 BRIDGE_UNLOCK(sc);
5627
5628 error = bridge_enqueue(bridge_ifp, NULL, dst_if, m,
5629 CHECKSUM_OPERATION_FINALIZE);
5630 }
5631
5632 return error;
5633 }
5634
5635 static void
5636 bridge_finalize_cksum(struct ifnet *ifp, struct mbuf *m)
5637 {
5638 struct ether_header *eh;
5639 bool is_ipv4;
5640 uint32_t sw_csum, hwcap;
5641 uint32_t did_sw;
5642 uint32_t csum_flags;
5643
5644 eh = mtod(m, struct ether_header *);
5645 if (!ether_header_type_is_ip(eh, &is_ipv4)) {
5646 return;
5647 }
5648
5649 /* do in software what the hardware cannot */
5650 hwcap = (ifp->if_hwassist | CSUM_DATA_VALID);
5651 csum_flags = m->m_pkthdr.csum_flags;
5652 sw_csum = csum_flags & ~IF_HWASSIST_CSUM_FLAGS(hwcap);
5653 sw_csum &= IF_HWASSIST_CSUM_MASK;
5654
5655 if (is_ipv4) {
5656 if ((hwcap & CSUM_PARTIAL) && !(sw_csum & CSUM_DELAY_DATA) &&
5657 (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)) {
5658 if (m->m_pkthdr.csum_flags & CSUM_TCP) {
5659 uint16_t start =
5660 sizeof(*eh) + sizeof(struct ip);
5661 uint16_t ulpoff =
5662 m->m_pkthdr.csum_data & 0xffff;
5663 m->m_pkthdr.csum_flags |=
5664 (CSUM_DATA_VALID | CSUM_PARTIAL);
5665 m->m_pkthdr.csum_tx_stuff = (ulpoff + start);
5666 m->m_pkthdr.csum_tx_start = start;
5667 } else {
5668 sw_csum |= (CSUM_DELAY_DATA &
5669 m->m_pkthdr.csum_flags);
5670 }
5671 }
5672 did_sw = in_finalize_cksum(m, sizeof(*eh), sw_csum);
5673 } else {
5674 if ((hwcap & CSUM_PARTIAL) &&
5675 !(sw_csum & CSUM_DELAY_IPV6_DATA) &&
5676 (m->m_pkthdr.csum_flags & CSUM_DELAY_IPV6_DATA)) {
5677 if (m->m_pkthdr.csum_flags & CSUM_TCPIPV6) {
5678 uint16_t start =
5679 sizeof(*eh) + sizeof(struct ip6_hdr);
5680 uint16_t ulpoff =
5681 m->m_pkthdr.csum_data & 0xffff;
5682 m->m_pkthdr.csum_flags |=
5683 (CSUM_DATA_VALID | CSUM_PARTIAL);
5684 m->m_pkthdr.csum_tx_stuff = (ulpoff + start);
5685 m->m_pkthdr.csum_tx_start = start;
5686 } else {
5687 sw_csum |= (CSUM_DELAY_IPV6_DATA &
5688 m->m_pkthdr.csum_flags);
5689 }
5690 }
5691 did_sw = in6_finalize_cksum(m, sizeof(*eh), -1, -1, sw_csum);
5692 }
5693 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
5694 "[%s] before 0x%x hwcap 0x%x sw_csum 0x%x did 0x%x after 0x%x",
5695 ifp->if_xname, csum_flags, hwcap, sw_csum,
5696 did_sw, m->m_pkthdr.csum_flags);
5697 }
5698
5699 /*
5700 * bridge_start:
5701 *
5702 * Start output on a bridge.
5703 *
5704 * This routine is invoked by the start worker thread; because we never call
5705 * it directly, there is no need do deploy any serialization mechanism other
5706 * than what's already used by the worker thread, i.e. this is already single
5707 * threaded.
5708 *
5709 * This routine is called only when if_bridge_txstart is enabled.
5710 */
5711 static void
5712 bridge_start(struct ifnet *ifp)
5713 {
5714 mbuf_ref_t m;
5715
5716 for (;;) {
5717 if (ifnet_dequeue(ifp, &m) != 0) {
5718 break;
5719 }
5720
5721 (void) bridge_output(ifp, m);
5722 }
5723 }
5724
5725 /*
5726 * bridge_forward:
5727 *
5728 * The forwarding function of the bridge.
5729 *
5730 * NOTE: Releases the lock on return.
5731 */
5732 static void
5733 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
5734 struct mbuf *m)
5735 {
5736 struct bridge_iflist *dbif;
5737 ifnet_t bridge_ifp;
5738 struct ifnet *src_if, *dst_if;
5739 struct ether_header *eh;
5740 uint16_t vlan;
5741 uint8_t *dst;
5742 int error;
5743 struct mac_nat_record mnr;
5744 bool translate_mac = FALSE;
5745 uint32_t sc_filter_flags;
5746
5747 BRIDGE_LOCK_ASSERT_HELD(sc);
5748
5749 bridge_ifp = sc->sc_ifp;
5750 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_OUTPUT,
5751 "%s m 0x%llx", bridge_ifp->if_xname,
5752 (uint64_t)VM_KERNEL_ADDRPERM(m));
5753
5754 src_if = sbif->bif_ifp;
5755 VERIFY(src_if == m->m_pkthdr.rcvif);
5756
5757 (void) ifnet_stat_increment_in(bridge_ifp, 1, m->m_pkthdr.len, 0);
5758 vlan = VLANTAGOF(m);
5759
5760 eh = mtod(m, struct ether_header *);
5761 dst = eh->ether_dhost;
5762
5763 /* If the interface is learning, record the address. */
5764 if (sbif->bif_ifflags & IFBIF_LEARNING) {
5765 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
5766 sbif, 0, IFBAF_DYNAMIC);
5767 /*
5768 * If the interface has addresses limits then deny any source
5769 * that is not in the cache.
5770 */
5771 if (error && sbif->bif_addrmax) {
5772 goto drop;
5773 }
5774 }
5775
5776 if ((sbif->bif_ifflags & IFBIF_STP) != 0 &&
5777 sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING) {
5778 goto drop;
5779 }
5780
5781 /*
5782 * At this point, the port either doesn't participate
5783 * in spanning tree or it is in the forwarding state.
5784 */
5785
5786 /*
5787 * If the packet is unicast, destined for someone on
5788 * "this" side of the bridge, drop it.
5789 */
5790 if (!IS_BCAST_MCAST(m)) {
5791 /* unicast */
5792 dst_if = bridge_rtlookup(sc, dst, vlan);
5793 if (src_if == dst_if) {
5794 goto drop;
5795 }
5796 } else {
5797 /* broadcast/multicast */
5798
5799 /*
5800 * Check if it's a reserved multicast address, any address
5801 * listed in 802.1D section 7.12.6 may not be forwarded by the
5802 * bridge.
5803 * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
5804 */
5805 if (dst[0] == 0x01 && dst[1] == 0x80 &&
5806 dst[2] == 0xc2 && dst[3] == 0x00 &&
5807 dst[4] == 0x00 && dst[5] <= 0x0f) {
5808 goto drop;
5809 }
5810 /* ...forward it to all interfaces. */
5811 os_atomic_inc(&bridge_ifp->if_imcasts, relaxed);
5812 dst_if = NULL;
5813 }
5814
5815 /*
5816 * If we have a destination interface which is a member of our bridge,
5817 * OR this is a unicast packet, push it through the bpf(4) machinery.
5818 * For broadcast or multicast packets, don't bother because it will
5819 * be reinjected into ether_input. We do this before we pass the packets
5820 * through the pfil(9) framework, as it is possible that pfil(9) will
5821 * drop the packet, or possibly modify it, making it difficult to debug
5822 * firewall issues on the bridge.
5823 */
5824 #if NBPFILTER > 0
5825 if (eh->ether_type == htons(ETHERTYPE_RSN_PREAUTH) ||
5826 dst_if != NULL || !IS_BCAST_MCAST(m)) {
5827 m->m_pkthdr.rcvif = bridge_ifp;
5828 BRIDGE_BPF_TAP_IN(bridge_ifp, m);
5829 }
5830 #endif /* NBPFILTER */
5831
5832 if (dst_if == NULL) {
5833 /* bridge_broadcast will unlock */
5834 bridge_broadcast(sc, sbif, m);
5835 return;
5836 }
5837
5838 /*
5839 * Unicast.
5840 */
5841 /*
5842 * At this point, we're dealing with a unicast frame
5843 * going to a different interface.
5844 */
5845 if ((dst_if->if_flags & IFF_RUNNING) == 0) {
5846 goto drop;
5847 }
5848
5849 dbif = bridge_lookup_member_if(sc, dst_if);
5850 if (dbif == NULL) {
5851 /* Not a member of the bridge (anymore?) */
5852 goto drop;
5853 }
5854
5855 /* Private segments can not talk to each other */
5856 if (sbif->bif_ifflags & dbif->bif_ifflags & IFBIF_PRIVATE) {
5857 goto drop;
5858 }
5859
5860 if ((dbif->bif_ifflags & IFBIF_STP) &&
5861 dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
5862 goto drop;
5863 }
5864 if (dbif == sc->sc_mac_nat_bif) {
5865 /* determine how to translate the packet */
5866 translate_mac
5867 = bridge_mac_nat_output(sc, sbif, &m, &mnr);
5868 if (m == NULL) {
5869 /* packet was deallocated */
5870 BRIDGE_UNLOCK(sc);
5871 return;
5872 }
5873 } else if (bif_has_checksum_offload(dbif) &&
5874 !bif_has_checksum_offload(sbif)) {
5875 /*
5876 * If the destination interface has checksum offload enabled,
5877 * verify the checksum now, unless the source interface also has
5878 * checksum offload enabled. The checksum in that case has
5879 * already just been computed and verifying it is unnecessary.
5880 */
5881 error = bridge_verify_checksum(&m, &dbif->bif_stats);
5882 if (error != 0) {
5883 BRIDGE_UNLOCK(sc);
5884 if (m != NULL) {
5885 m_freem(m);
5886 }
5887 return;
5888 }
5889 }
5890
5891 sc_filter_flags = sc->sc_filter_flags;
5892
5893 BRIDGE_UNLOCK(sc);
5894
5895 if (translate_mac) {
5896 /* if we need to, translate the MAC address */
5897 bridge_mac_nat_translate(&m, &mnr, IF_LLADDR(dst_if));
5898 }
5899
5900 if (m != NULL && PF_IS_ENABLED && (sc_filter_flags & IFBF_FILT_MEMBER)) {
5901 if (bridge_pf(&m, dst_if, sc_filter_flags, false) != 0) {
5902 return;
5903 }
5904 if (m == NULL) {
5905 return;
5906 }
5907 }
5908
5909 /*
5910 * We're forwarding an inbound packet in which the checksum must
5911 * already have been computed and if required, verified.
5912 */
5913 if (m != NULL) {
5914 (void) bridge_enqueue(bridge_ifp, src_if, dst_if, m,
5915 CHECKSUM_OPERATION_CLEAR_OFFLOAD);
5916 }
5917 return;
5918
5919 drop:
5920 BRIDGE_UNLOCK(sc);
5921 m_freem(m);
5922 }
5923
5924 static void
5925 prepare_input_packet(ifnet_t ifp, mbuf_t m)
5926 {
5927 mbuf_pkthdr_setrcvif(m, ifp);
5928 mbuf_pkthdr_setheader(m, mtod(m, void *));
5929 /* adjust frame to skip mac-layer header */
5930 _mbuf_adjust_pkthdr_and_data(m, ETHER_HDR_LEN);
5931 }
5932
5933 static void
5934 inject_input_packet(ifnet_t ifp, mbuf_t m)
5935 {
5936 prepare_input_packet(ifp, m);
5937 m->m_flags |= M_PROTO1; /* set to avoid loops */
5938 dlil_input_packet_list(ifp, m);
5939 return;
5940 }
5941
5942 static void
5943 inject_input_packet_list(ifnet_t ifp, mbuf_t in_list, bool m_proto1)
5944 {
5945 for (mbuf_t scan = in_list; scan != NULL; scan = scan->m_nextpkt) {
5946 /* mark the packets as arriving on the interface */
5947 BRIDGE_BPF_TAP_IN(ifp, scan);
5948 if (m_proto1) {
5949 scan->m_flags |= M_PROTO1; /* set to avoid loops */
5950 }
5951 prepare_input_packet(ifp, scan);
5952 }
5953 dlil_input_packet_list(ifp, in_list);
5954 return;
5955 }
5956
5957 static void
5958 adjust_input_packet_list(mbuf_t in_list)
5959 {
5960 for (mbuf_t scan = in_list; scan != NULL; scan = scan->m_nextpkt) {
5961 mbuf_pkthdr_setheader(scan, mtod(scan, void *));
5962 _mbuf_adjust_pkthdr_and_data(scan, ETHER_HDR_LEN);
5963 }
5964 }
5965
5966 static bool
5967 in_addr_is_ours(struct in_addr ip)
5968 {
5969 struct in_ifaddr *ia;
5970 bool ours = false;
5971
5972 lck_rw_lock_shared(&in_ifaddr_rwlock);
5973 TAILQ_FOREACH(ia, INADDR_HASH(ip.s_addr), ia_hash) {
5974 if (ia->ia_addr.sin_addr.s_addr == ip.s_addr) {
5975 ours = true;
5976 break;
5977 }
5978 }
5979 lck_rw_done(&in_ifaddr_rwlock);
5980 return ours;
5981 }
5982
5983 static bool
5984 in6_addr_is_ours(const struct in6_addr * ip6_p, uint32_t ifscope)
5985 {
5986 struct in6_addr dst_ip;
5987 struct in6_ifaddr *ia6;
5988 bool ours = false;
5989
5990 if (in6_embedded_scope && IN6_IS_ADDR_LINKLOCAL(ip6_p)) {
5991 /* need to embed scope ID for comparison */
5992 bcopy(ip6_p, &dst_ip, sizeof(dst_ip));
5993 dst_ip.s6_addr16[1] = htons(ifscope);
5994 ip6_p = &dst_ip;
5995 }
5996 lck_rw_lock_shared(&in6_ifaddr_rwlock);
5997 TAILQ_FOREACH(ia6, IN6ADDR_HASH(ip6_p), ia6_hash) {
5998 if (in6_are_addr_equal_scoped(&ia6->ia_addr.sin6_addr, ip6_p,
5999 ia6->ia_addr.sin6_scope_id, ifscope)) {
6000 ours = true;
6001 break;
6002 }
6003 }
6004 lck_rw_done(&in6_ifaddr_rwlock);
6005 return ours;
6006 }
6007
6008
6009 static void
6010 bridge_interface_input(ifnet_t bridge_ifp, mbuf_t m, bool has_address)
6011 {
6012 bool do_segment;
6013 struct ether_header *eh;
6014 uint32_t in_bytes;
6015 uint32_t in_count;
6016 uint32_t in_errors = 0;
6017 bool is_ipv4;
6018 u_int mac_hlen;
6019 mblist seg;
6020
6021 /* segment large packets before sending them up */
6022 in_count = 1;
6023 in_bytes = mbuf_pkthdr_len(m);
6024 if (!has_address) {
6025 /* bridge has no address, so we're not forwarding */
6026 goto done;
6027 }
6028 if (if_bridge_segmentation == 0) {
6029 goto done;
6030 }
6031 if (in_bytes <= (bridge_ifp->if_mtu + ETHER_HDR_LEN)) {
6032 goto done;
6033 }
6034 eh = mtod(m, struct ether_header *);
6035 if (!ether_header_type_is_ip(eh, &is_ipv4)) {
6036 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
6037 "large non IPv4/IPv6 packet");
6038 goto done;
6039 }
6040 /* no need to segment if the packet will not be forwarded */
6041 if (is_ipv4) {
6042 do_segment = (ipforwarding != 0);
6043 } else {
6044 do_segment = (ip6_forwarding != 0);
6045 }
6046 if (!do_segment) {
6047 goto done;
6048 }
6049
6050 /*
6051 * We have a large IPv4/IPv6 TCP packet. Segment it if required.
6052 *
6053 * If gso_tcp_receive() returns non-NULL packets, the packet(s) are
6054 * ready to be passed up. If the destination is a local IP address,
6055 * the packet will not be segmented.
6056 */
6057 mac_hlen = sizeof(*eh);
6058 seg = gso_tcp_receive(bridge_ifp, m, mac_hlen, is_ipv4);
6059 if (seg.head == NULL) {
6060 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
6061 "gso_tcp returned no packets");
6062 in_errors++;
6063 in_bytes = in_count = 0;
6064 m = NULL;
6065 } else {
6066 m = seg.head;
6067 in_count = seg.count;
6068 in_bytes = seg.bytes;
6069 }
6070
6071 done:
6072 if (m != NULL) {
6073 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT,
6074 "%s packets %d bytes %d",
6075 bridge_ifp->if_xname, in_count, in_bytes);
6076 /* Mark the packets as arriving on the bridge interface */
6077 inject_input_packet_list(bridge_ifp, m, false);
6078 }
6079 ifnet_stat_increment_in(bridge_ifp, in_count, in_bytes, in_errors);
6080 return;
6081 }
6082
6083 static void
6084 bridge_interface_input_list(ifnet_t bridge_ifp, ether_type_flag_t etypef,
6085 mblist list, bool has_address)
6086 {
6087 bool do_segment = false;
6088 uint32_t in_errors = 0;
6089 bool is_ipv4;
6090 mblist in_list;
6091 u_int mac_hlen;
6092 mbuf_t next_packet;
6093
6094 if (has_address && if_bridge_segmentation != 0 &&
6095 ether_type_flag_is_ip(etypef)) {
6096 if (etypef == ETHER_TYPE_FLAG_IPV4) {
6097 is_ipv4 = true;
6098 do_segment = (ipforwarding != 0);
6099 } else {
6100 is_ipv4 = false;
6101 do_segment = (ip6_forwarding != 0);
6102 }
6103 }
6104 if (!do_segment) {
6105 in_list = list;
6106 goto done;
6107 }
6108
6109 mblist_init(&in_list);
6110 mac_hlen = sizeof(struct ether_header);
6111 for (mbuf_t scan = list.head; scan != NULL; scan = next_packet) {
6112 uint32_t len;
6113 mblist seg;
6114
6115 /* take it out of the list */
6116 next_packet = scan->m_nextpkt;
6117 scan->m_nextpkt = NULL;
6118
6119 /* check whether it might need segmentation */
6120 len = mbuf_pkthdr_len(scan);
6121 if (len <= (bridge_ifp->if_mtu + ETHER_HDR_LEN)) {
6122 mblist_append(&in_list, scan);
6123 } else {
6124 /* segment before sending up */
6125 seg = gso_tcp_receive(bridge_ifp, scan, mac_hlen,
6126 is_ipv4);
6127 if (seg.head == NULL) {
6128 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
6129 "gso_tcp returned no packets");
6130 in_errors++;
6131 } else {
6132 mblist_append_list(&in_list, seg);
6133 }
6134 }
6135 }
6136
6137 done:
6138 if (in_list.head != NULL) {
6139 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT,
6140 "%s packets %d bytes %d",
6141 bridge_ifp->if_xname,
6142 in_list.count, in_list.bytes);
6143 /* Mark the packets as arriving on the bridge interface */
6144 inject_input_packet_list(bridge_ifp, in_list.head, false);
6145 ifnet_stat_increment_in(bridge_ifp, in_list.count,
6146 in_list.bytes, in_errors);
6147 }
6148 return;
6149 }
6150
6151 /*
6152 * bridge_input:
6153 *
6154 * Filter input from a member interface. Queue the packet for
6155 * bridging if it is not for us.
6156 */
6157 errno_t
6158 bridge_input(struct ifnet *ifp, mbuf_t *data)
6159 {
6160 struct bridge_softc * __single sc = ifp->if_bridge;
6161 struct bridge_iflist *bif, *bif2;
6162 uint8_t *dhost;
6163 struct ether_header eh_in;
6164 bool is_ip = false;
6165 bool is_ipv4 = false;
6166 ifnet_t bridge_ifp;
6167 unsigned int mac_hlen = sizeof(struct ether_header);
6168 uint16_t vlan;
6169 errno_t error;
6170 ip_packet_info info;
6171 bool is_broadcast;
6172 bool is_ip_broadcast = false;
6173 bool is_ifp_mac = false;
6174 mbuf_t m;
6175 uint32_t sc_filter_flags;
6176 bool bridge_has_address;
6177
6178 if (sc == NULL) {
6179 return 0;
6180 }
6181 m = *data;
6182 is_broadcast = IS_BCAST_MCAST(m);
6183
6184 bridge_ifp = sc->sc_ifp;
6185 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT,
6186 "%s %sfrom %s m 0x%llx data 0x%llx",
6187 bridge_ifp->if_xname, is_broadcast ? "bcast_mcast " : "",
6188 ifp->if_xname,
6189 (uint64_t)VM_KERNEL_ADDRPERM(m),
6190 (uint64_t)VM_KERNEL_ADDRPERM(mtod(m, void *)));
6191 if ((bridge_ifp->if_flags & IFF_RUNNING) == 0) {
6192 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT,
6193 "%s not running passing along",
6194 bridge_ifp->if_xname);
6195 return 0;
6196 }
6197
6198 vlan = VLANTAGOF(m);
6199
6200 if (!is_broadcast) {
6201 /*
6202 * Need to clear the promiscuous flag otherwise it will be
6203 * dropped by DLIL after processing filters
6204 */
6205 if ((m->m_flags & M_PROMISC) != 0) {
6206 m->m_flags &= ~M_PROMISC;
6207 } else {
6208 is_ifp_mac = true;
6209 }
6210 }
6211
6212
6213 /* copy the ethernet header */
6214 eh_in = *(mtod(m, struct ether_header *));
6215 dhost = eh_in.ether_dhost;
6216
6217 is_ip = ether_header_type_is_ip(&eh_in, &is_ipv4);
6218
6219 sc_filter_flags = sc->sc_filter_flags;
6220 if (PF_IS_ENABLED && (sc_filter_flags & IFBF_FILT_MEMBER)) {
6221 error = bridge_pf(data, ifp, sc_filter_flags, true);
6222 m = *data;
6223 if (error != 0 || m == NULL) {
6224 return EJUSTRETURN;
6225 }
6226 }
6227
6228 BRIDGE_LOCK(sc);
6229 bif = bridge_lookup_member_if(sc, ifp);
6230 if (bif == NULL) {
6231 BRIDGE_UNLOCK(sc);
6232 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT,
6233 "%s bridge_lookup_member_if failed",
6234 bridge_ifp->if_xname);
6235 return 0;
6236 }
6237 if (is_ip && bif_has_checksum_offload(bif)) {
6238 error = bridge_get_ip_proto(data, mac_hlen, is_ipv4,
6239 &info, &bif->bif_stats.brms_in_ip);
6240 if (error != 0) {
6241 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_CHECKSUM,
6242 "%s(%s) bridge_get_ip_proto failed %d",
6243 bridge_ifp->if_xname,
6244 bif->bif_ifp->if_xname, error);
6245 } else {
6246 /* need to compute IP/UDP/TCP/checksums */
6247 error = bridge_offload_checksum(data, &info,
6248 &bif->bif_stats);
6249 if (error != 0) {
6250 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_CHECKSUM,
6251 "%s(%s) bridge_offload_checksum failed %d",
6252 bridge_ifp->if_xname,
6253 bif->bif_ifp->if_xname, error);
6254 }
6255 }
6256 if (error != 0) {
6257 BRIDGE_UNLOCK(sc);
6258 if (*data != NULL) {
6259 m_freem(*data);
6260 *data = NULL;
6261 }
6262 return EJUSTRETURN;
6263 }
6264 m = *data;
6265 }
6266
6267 if (bif->bif_flags & BIFF_HOST_FILTER) {
6268 error = bridge_host_filter(bif, data);
6269 if (error != 0) {
6270 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT,
6271 "%s bridge_host_filter failed",
6272 bif->bif_ifp->if_xname);
6273 BRIDGE_UNLOCK(sc);
6274 return EJUSTRETURN;
6275 }
6276 m = *data;
6277 }
6278
6279 /*
6280 * A link-layer unicast packet directed at the MAC-NAT interface
6281 * could be an IP broadcast packet. An IP broadcast packet needs to
6282 * be forwarded like regular link-layer broadcast.
6283 */
6284 if (sc->sc_mac_nat_bif == bif && is_ifp_mac && is_ip) {
6285 is_ip_broadcast = is_broadcast_ip_packet(data);
6286 m = *data;
6287 if (m == NULL) {
6288 BRIDGE_UNLOCK(sc);
6289 return EJUSTRETURN;
6290 }
6291 }
6292
6293 bridge_span(sc, m);
6294
6295 if (is_broadcast || is_ip_broadcast) {
6296 struct mbuf *mc;
6297
6298 if (IS_MCAST(m)) {
6299 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MCAST,
6300 " multicast: "
6301 "%02x:%02x:%02x:%02x:%02x:%02x",
6302 dhost[0], dhost[1],
6303 dhost[2], dhost[3],
6304 dhost[4], dhost[5]);
6305 /* Tap off 802.1D packets; they do not get forwarded. */
6306 if (_ether_cmp(dhost, bstp_etheraddr) == 0) {
6307 #if BRIDGESTP
6308 bstp_input(&bif->bif_stp, m);
6309 m = NULL;
6310 #else /* !BRIDGESTP */
6311 m_freem(m);
6312 *data = m = NULL;
6313 #endif /* !BRIDGESTP */
6314 if (m == NULL) {
6315 BRIDGE_UNLOCK(sc);
6316 return EJUSTRETURN;
6317 }
6318 }
6319 } else if (is_broadcast) {
6320 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MCAST,
6321 " broadcast: "
6322 "%02x:%02x:%02x:%02x:%02x:%02x",
6323 dhost[0], dhost[1],
6324 dhost[2], dhost[3],
6325 dhost[4], dhost[5]);
6326 }
6327
6328 if ((bif->bif_ifflags & IFBIF_STP) &&
6329 bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
6330 BRIDGE_UNLOCK(sc);
6331 return 0;
6332 }
6333 /*
6334 * Make a deep copy of the packet and enqueue the copy
6335 * for bridge processing.
6336 */
6337 mc = m_dup(m, M_DONTWAIT);
6338 if (mc == NULL) {
6339 BRIDGE_UNLOCK(sc);
6340 return 0;
6341 }
6342 /*
6343 * Perform the bridge forwarding function with the copy.
6344 *
6345 * Note that bridge_forward calls BRIDGE_UNLOCK
6346 */
6347 if (is_ip_broadcast) {
6348 struct ether_header *eh;
6349
6350 /* make the copy look like it is actually broadcast */
6351 mc->m_flags |= M_BCAST;
6352 eh = mtod(mc, struct ether_header *);
6353 bcopy(etherbroadcastaddr, eh->ether_dhost,
6354 ETHER_ADDR_LEN);
6355 }
6356
6357 bridge_forward(sc, bif, mc);
6358
6359 /*
6360 * Reinject the mbuf as arriving on the bridge so we have a
6361 * chance at claiming multicast packets. We can not loop back
6362 * here from ether_input as a bridge is never a member of a
6363 * bridge.
6364 */
6365 if (is_broadcast) {
6366 VERIFY(bridge_ifp->if_bridge == NULL);
6367 mc = m_dup(m, M_DONTWAIT);
6368 if (mc != NULL) {
6369 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MCAST,
6370 "%s mcast for us", bridge_ifp->if_xname);
6371 BRIDGE_BPF_TAP_IN(ifp, mc);
6372 prepare_input_packet(bridge_ifp, mc);
6373 ifnet_stat_increment_in(bridge_ifp, 1,
6374 mbuf_pkthdr_len(mc), 0);
6375 dlil_input_packet_list(bridge_ifp, mc);
6376 }
6377 }
6378
6379 /* Return the original packet for local processing. */
6380 return 0;
6381 }
6382
6383 if ((bif->bif_ifflags & IFBIF_STP) &&
6384 bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
6385 BRIDGE_UNLOCK(sc);
6386 return 0;
6387 }
6388
6389 #define GRAB_OUR_PACKETS(iface) \
6390 /* It is destined for us. */ \
6391 if (_ether_cmp(IF_LLADDR((iface)), dhost) == 0) { \
6392 BRIDGE_BPF_TAP_IN(iface, m); \
6393 if (bif->bif_ifflags & IFBIF_LEARNING) { \
6394 error = bridge_rtupdate(sc, eh_in.ether_shost, \
6395 vlan, bif, 0, IFBAF_DYNAMIC); \
6396 if (error && bif->bif_addrmax) { \
6397 BRIDGE_UNLOCK(sc); \
6398 m_freem(m); \
6399 return (EJUSTRETURN); \
6400 } \
6401 } \
6402 BRIDGE_UNLOCK(sc); \
6403 inject_input_packet(iface, m); \
6404 return (EJUSTRETURN); \
6405 } \
6406 \
6407 /* We just received a packet that we sent out. */ \
6408 if (_ether_cmp(IF_LLADDR((iface)), eh_in.ether_shost) == 0) { \
6409 BRIDGE_UNLOCK(sc); \
6410 m_freem(m); \
6411 return (EJUSTRETURN); \
6412 }
6413
6414 /*
6415 * Unicast.
6416 */
6417
6418 /* handle MAC-NAT if enabled */
6419 if (is_ifp_mac && sc->sc_mac_nat_bif == bif) {
6420 ifnet_ref_t dst_if = NULL;
6421
6422 m = *data = bridge_mac_nat_input(sc, ifp, m, &dst_if);
6423 if (m == NULL) {
6424 /* packet was deallocated */
6425 BRIDGE_UNLOCK(sc);
6426 return EJUSTRETURN;
6427 }
6428 if (dst_if != NULL) {
6429 /* need to forward the packet */
6430 BRIDGE_UNLOCK(sc);
6431 (void)bridge_enqueue(bridge_ifp, NULL,
6432 dst_if, m,
6433 CHECKSUM_OPERATION_CLEAR_OFFLOAD);
6434 return EJUSTRETURN;
6435 }
6436 }
6437
6438 /*
6439 * If the destination MAC matches the bridge interface, and
6440 * the bridge has an address assigned, or the address is
6441 * distinct from the receive interface, claim the packets
6442 * for the bridge interface.
6443 */
6444 bridge_has_address = (sc->sc_flags & SCF_ADDRESS_ASSIGNED) != 0;
6445 if (_ether_cmp(dhost, IF_LLADDR(bridge_ifp)) == 0 &&
6446 (bridge_has_address || _ether_cmp(dhost, IF_LLADDR(ifp)) != 0)) {
6447 /*
6448 * If the interface is learning, and the source
6449 * address is valid and not multicast, record
6450 * the address.
6451 */
6452 if (bif->bif_ifflags & IFBIF_LEARNING) {
6453 (void) bridge_rtupdate(sc, eh_in.ether_shost,
6454 vlan, bif, 0, IFBAF_DYNAMIC);
6455 }
6456 BRIDGE_UNLOCK(sc);
6457 bridge_interface_input(bridge_ifp, m, bridge_has_address);
6458 return EJUSTRETURN;
6459 }
6460
6461 /*
6462 * if the destination of the packet is for the MAC address of
6463 * the member interface itself, then we don't need to forward
6464 * it -- just pass it back. Note that it'll likely just be
6465 * dropped by the stack, but if something else is bound to
6466 * the interface directly (for example, the wireless stats
6467 * protocol -- although that actually uses BPF right now),
6468 * then it will consume the packet
6469 *
6470 * ALSO, note that we do this check AFTER checking for the
6471 * bridge's own MAC address, because the bridge may be
6472 * using the SAME MAC address as one of its interfaces
6473 */
6474 if (is_ifp_mac) {
6475 BRIDGE_UNLOCK(sc);
6476 return 0;
6477 }
6478
6479 /* Now check the remaining bridge members. */
6480 TAILQ_FOREACH(bif2, &sc->sc_iflist, bif_next) {
6481 if (bif2->bif_ifp != ifp) {
6482 GRAB_OUR_PACKETS(bif2->bif_ifp);
6483 }
6484 }
6485
6486 #undef GRAB_OUR_PACKETS
6487
6488 /*
6489 * Perform the bridge forwarding function.
6490 *
6491 * Note that bridge_forward calls BRIDGE_UNLOCK
6492 */
6493 bridge_forward(sc, bif, m);
6494
6495 return EJUSTRETURN;
6496 }
6497
6498 /*
6499 * bridge_broadcast:
6500 *
6501 * Send a frame to all interfaces that are members of
6502 * the bridge, except for the one on which the packet
6503 * arrived.
6504 *
6505 * NOTE: Releases the lock on return.
6506 */
6507 static void
6508 bridge_broadcast(struct bridge_softc *sc, struct bridge_iflist * sbif,
6509 struct mbuf *m)
6510 {
6511 ifnet_t bridge_ifp;
6512 struct bridge_iflist *dbif;
6513 struct ifnet * src_if;
6514 mbuf_ref_t mc;
6515 struct mbuf *mc_in;
6516 int error = 0, used = 0;
6517 ChecksumOperation cksum_op;
6518 struct mac_nat_record mnr;
6519 struct bridge_iflist *mac_nat_bif = sc->sc_mac_nat_bif;
6520 boolean_t translate_mac = FALSE;
6521 uint32_t sc_filter_flags;
6522 bool is_bcast_mcast;
6523
6524 bridge_ifp = sc->sc_ifp;
6525 if (sbif != NULL) {
6526 src_if = sbif->bif_ifp;
6527 cksum_op = CHECKSUM_OPERATION_CLEAR_OFFLOAD;
6528 if (mac_nat_bif != NULL && sbif != mac_nat_bif) {
6529 /* get the translation record */
6530 translate_mac
6531 = bridge_mac_nat_output(sc, sbif, &m, &mnr);
6532 if (m == NULL) {
6533 /* packet was deallocated */
6534 BRIDGE_UNLOCK(sc);
6535 return;
6536 }
6537 }
6538 } else {
6539 /*
6540 * sbif is NULL when the bridge interface calls
6541 * bridge_broadcast().
6542 */
6543 cksum_op = CHECKSUM_OPERATION_FINALIZE;
6544 src_if = NULL;
6545 }
6546
6547 BRIDGE_LOCK2REF(sc, error);
6548 if (error) {
6549 m_freem(m);
6550 return;
6551 }
6552 is_bcast_mcast = IS_BCAST_MCAST(m);
6553 sc_filter_flags = sc->sc_filter_flags;
6554 TAILQ_FOREACH(dbif, &sc->sc_iflist, bif_next) {
6555 ifnet_t dst_if;
6556
6557 dst_if = dbif->bif_ifp;
6558 if (dst_if == src_if) {
6559 /* skip the interface that the packet came in on */
6560 continue;
6561 }
6562
6563 /* Private segments can not talk to each other */
6564 if (sbif != NULL &&
6565 (sbif->bif_ifflags & dbif->bif_ifflags & IFBIF_PRIVATE)) {
6566 continue;
6567 }
6568
6569 if ((dbif->bif_ifflags & IFBIF_STP) &&
6570 dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
6571 continue;
6572 }
6573
6574 if ((dbif->bif_ifflags & IFBIF_DISCOVER) == 0 &&
6575 !is_bcast_mcast) {
6576 continue;
6577 }
6578
6579 if ((dst_if->if_flags & IFF_RUNNING) == 0) {
6580 continue;
6581 }
6582
6583 if ((dbif->bif_flags & BIFF_MEDIA_ACTIVE) == 0) {
6584 continue;
6585 }
6586
6587 if (TAILQ_NEXT(dbif, bif_next) == NULL) {
6588 mc = m;
6589 used = 1;
6590 } else {
6591 mc = m_dup(m, M_DONTWAIT);
6592 if (mc == NULL) {
6593 (void) ifnet_stat_increment_out(bridge_ifp,
6594 0, 0, 1);
6595 continue;
6596 }
6597 }
6598
6599 /*
6600 * If broadcast input is enabled, do so only if this
6601 * is an input packet.
6602 */
6603 if (sbif != NULL && is_bcast_mcast &&
6604 (dbif->bif_flags & BIFF_INPUT_BROADCAST) != 0) {
6605 mc_in = m_dup(mc, M_DONTWAIT);
6606 /* this could fail, but we continue anyways */
6607 } else {
6608 mc_in = NULL;
6609 }
6610
6611 /* out */
6612 if (translate_mac && mac_nat_bif == dbif) {
6613 /* translate the packet */
6614 bridge_mac_nat_translate(&mc, &mnr, IF_LLADDR(dst_if));
6615 }
6616
6617 if (mc != NULL && sbif != NULL &&
6618 PF_IS_ENABLED && (sc_filter_flags & IFBF_FILT_MEMBER)) {
6619 if (used == 0) {
6620 /* Keep the layer3 header aligned */
6621 int i = min(mc->m_pkthdr.len, max_protohdr);
6622 mc = m_copyup(mc, i, ETHER_ALIGN);
6623 if (mc == NULL) {
6624 (void) ifnet_stat_increment_out(
6625 sc->sc_ifp, 0, 0, 1);
6626 if (mc_in != NULL) {
6627 m_freem(mc_in);
6628 mc_in = NULL;
6629 }
6630 continue;
6631 }
6632 }
6633 if (bridge_pf(&mc, dst_if, sc_filter_flags, false) != 0) {
6634 if (mc_in != NULL) {
6635 m_freem(mc_in);
6636 mc_in = NULL;
6637 }
6638 continue;
6639 }
6640 if (mc == NULL) {
6641 if (mc_in != NULL) {
6642 m_freem(mc_in);
6643 mc_in = NULL;
6644 }
6645 continue;
6646 }
6647 }
6648
6649 if (mc != NULL) {
6650 /* verify checksum if necessary */
6651 if (bif_has_checksum_offload(dbif) && sbif != NULL &&
6652 !bif_has_checksum_offload(sbif)) {
6653 error = bridge_verify_checksum(&mc,
6654 &dbif->bif_stats);
6655 if (error != 0) {
6656 if (mc != NULL) {
6657 m_freem(mc);
6658 }
6659 mc = NULL;
6660 }
6661 }
6662 if (mc != NULL) {
6663 (void) bridge_enqueue(bridge_ifp,
6664 NULL, dst_if, mc, cksum_op);
6665 }
6666 }
6667
6668 /* in */
6669 if (mc_in == NULL) {
6670 continue;
6671 }
6672 BRIDGE_BPF_TAP_IN(dst_if, mc_in);
6673 prepare_input_packet(dst_if, mc_in);
6674 mc_in->m_flags |= M_PROTO1; /* set to avoid loops */
6675 dlil_input_packet_list(dst_if, mc_in);
6676 }
6677 if (used == 0) {
6678 m_freem(m);
6679 }
6680
6681
6682 BRIDGE_UNREF(sc);
6683 }
6684
6685 static mbuf_t
6686 copy_packet_list(mbuf_t m)
6687 {
6688 mblist ret;
6689 mbuf_t next_packet;
6690
6691 mblist_init(&ret);
6692 for (mbuf_t scan = m; scan != NULL; scan = next_packet) {
6693 mbuf_t copy_m;
6694
6695 /* take it out of the list */
6696 next_packet = scan->m_nextpkt;
6697 scan->m_nextpkt = NULL;
6698
6699 /* create a copy and add it to the new list */
6700 copy_m = m_dup(scan, M_DONTWAIT);
6701 if (copy_m != NULL) {
6702 mblist_append(&ret, copy_m);
6703 }
6704
6705 /* put it back in the original list */
6706 scan->m_nextpkt = next_packet;
6707 }
6708 return ret.head;
6709 }
6710
6711 /*
6712 * bridge_broadcast_list:
6713 *
6714 * Broadcast a list of packets to all members except `sbif`.
6715 * If `need_copy` is false, consumes `m` before returning.
6716 *
6717 * NOTE: Releases the lock on return.
6718 */
6719 static void
6720 bridge_broadcast_list(struct bridge_softc *sc, struct bridge_iflist * sbif,
6721 ether_type_flag_t etypef, mbuf_t m, bool need_copy)
6722 {
6723 bool bridge_has_address;
6724 ifnet_t bridge_ifp;
6725 struct bridge_iflist * dbif;
6726 bool is_bcast_mcast;
6727 errno_t error = 0;
6728 ChecksumOperation cksum_op;
6729 struct bridge_iflist * mac_nat_bif = sc->sc_mac_nat_bif;
6730 ifnet_t mac_nat_if = NULL;
6731 bool need_mac_nat = false;
6732 mbuf_t out_mac_nat = NULL;
6733 ifnet_t src_if;
6734 uint32_t sc_filter_flags;
6735 bool used = false;
6736
6737 bridge_ifp = sc->sc_ifp;
6738 if (sbif != NULL) {
6739 src_if = sbif->bif_ifp;
6740 cksum_op = CHECKSUM_OPERATION_CLEAR_OFFLOAD;
6741
6742 /*
6743 * If MAC-NAT is enabled and we'll be sending the packets
6744 * over it, verify that it is up and active before
6745 * deciding to make a translated copy.
6746 */
6747 if (mac_nat_bif != NULL && sbif != mac_nat_bif) {
6748 mac_nat_if = mac_nat_bif->bif_ifp;
6749 if ((mac_nat_if->if_flags & IFF_RUNNING) != 0 &&
6750 (mac_nat_bif->bif_flags & BIFF_MEDIA_ACTIVE) != 0) {
6751 need_mac_nat = true;
6752 }
6753 }
6754 } else {
6755 /*
6756 * sbif is NULL when the bridge interface calls
6757 * bridge_broadcast().
6758 */
6759 cksum_op = CHECKSUM_OPERATION_FINALIZE;
6760 src_if = NULL;
6761 }
6762
6763 /*
6764 * Create a translated copy for packets destined to MAC-NAT interface.
6765 */
6766 if (need_mac_nat) {
6767 out_mac_nat
6768 = bridge_mac_nat_copy_and_translate_list(sc, sbif,
6769 mac_nat_if, m);
6770 }
6771 sc_filter_flags = sc->sc_filter_flags;
6772 bridge_has_address = (sc->sc_flags & SCF_ADDRESS_ASSIGNED) != 0;
6773 BRIDGE_LOCK2REF(sc, error);
6774 if (error) {
6775 goto done;
6776 }
6777 is_bcast_mcast = IS_BCAST_MCAST(m);
6778
6779 /* make a copy for the bridge interface */
6780 if (is_bcast_mcast && bridge_has_address) {
6781 mbuf_t in_list;
6782
6783 in_list = copy_packet_list(m);
6784 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MCAST,
6785 "%s mcast for us in_m %p",
6786 bridge_ifp->if_xname, in_list);
6787 if (in_list != NULL) {
6788 inject_input_packet_list(bridge_ifp, in_list, false);
6789 }
6790 }
6791
6792 TAILQ_FOREACH(dbif, &sc->sc_iflist, bif_next) {
6793 ifnet_t dst_if;
6794 mbuf_t in_m = NULL;
6795 mbuf_t out_m = NULL;
6796
6797 dst_if = dbif->bif_ifp;
6798 if (dst_if == src_if) {
6799 /* skip the interface that the packet came in on */
6800 continue;
6801 }
6802
6803 /* Private segments can not talk to each other */
6804 if (sbif != NULL &&
6805 (sbif->bif_ifflags & dbif->bif_ifflags & IFBIF_PRIVATE)) {
6806 continue;
6807 }
6808
6809 if ((dbif->bif_ifflags & IFBIF_STP) &&
6810 dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
6811 continue;
6812 }
6813
6814 if ((dbif->bif_ifflags & IFBIF_DISCOVER) == 0 &&
6815 !is_bcast_mcast) {
6816 continue;
6817 }
6818
6819 if ((dst_if->if_flags & IFF_RUNNING) == 0) {
6820 continue;
6821 }
6822
6823 if ((dbif->bif_flags & BIFF_MEDIA_ACTIVE) == 0) {
6824 continue;
6825 }
6826 if (dbif == mac_nat_bif) {
6827 /* translated copy was created above, use that */
6828 out_m = out_mac_nat;
6829 out_mac_nat = NULL;
6830 } else if (!need_copy && TAILQ_NEXT(dbif, bif_next) == NULL) {
6831 /* consume `m` */
6832 out_m = m;
6833 used = true;
6834 } else {
6835 /* needs a copy */
6836 out_m = copy_packet_list(m);
6837 }
6838
6839 if (out_m == NULL) {
6840 ifnet_stat_increment_out(bridge_ifp, 0, 0, 1);
6841 continue;
6842 }
6843 /*
6844 * If broadcast input is enabled, do so only if this
6845 * is an input packet.
6846 */
6847 if (sbif != NULL && is_bcast_mcast &&
6848 (dbif->bif_flags & BIFF_INPUT_BROADCAST) != 0) {
6849 in_m = copy_packet_list(m);
6850 /* this could fail, but we continue anyways */
6851 } else {
6852 in_m = NULL;
6853 }
6854
6855 if (sbif != NULL &&
6856 PF_IS_ENABLED && (sc_filter_flags & IFBF_FILT_MEMBER)) {
6857 out_m = bridge_pf_list(out_m, dst_if,
6858 sc_filter_flags, false);
6859 }
6860 if (out_m != NULL) {
6861 /* verify checksum if necessary */
6862 if (sbif != NULL &&
6863 ether_type_flag_is_ip(etypef) &&
6864 bif_has_checksum_offload(dbif) &&
6865 !bif_has_checksum_offload(sbif)) {
6866 out_m = bridge_verify_checksum_list(out_m, dbif);
6867 }
6868 if (out_m != NULL) {
6869 bridge_enqueue_multi(bridge_ifp, src_if, dst_if,
6870 etypef, out_m, cksum_op);
6871 }
6872 }
6873
6874 /* in */
6875 if (in_m != NULL) {
6876 inject_input_packet_list(dst_if, in_m, true);
6877 }
6878 }
6879
6880 BRIDGE_UNREF(sc);
6881
6882 done:
6883 if (out_mac_nat != NULL) {
6884 m_freem_list(out_mac_nat);
6885 }
6886 if (!need_copy && !used) {
6887 m_freem_list(m);
6888 }
6889 return;
6890 }
6891
6892 static void
6893 bridge_forward_list(struct bridge_softc *sc, struct bridge_iflist * sbif,
6894 ifnet_t dst_if, ether_type_flag_t etypef, mbuf_t m)
6895 {
6896 ifnet_t bridge_ifp;
6897 struct bridge_iflist * dbif;
6898 uint32_t sc_filter_flags;
6899 ifnet_t src_if;
6900
6901 if ((dst_if->if_flags & IFF_RUNNING) == 0) {
6902 goto drop;
6903 }
6904 dbif = bridge_lookup_member_if(sc, dst_if);
6905 if (dbif == NULL) {
6906 /* Not a member of the bridge (anymore?) */
6907 goto drop;
6908 }
6909
6910 /* Private segments can not talk to each other */
6911 if ((sbif->bif_ifflags & dbif->bif_ifflags & IFBIF_PRIVATE) != 0) {
6912 goto drop;
6913 }
6914 if (dbif == sc->sc_mac_nat_bif) {
6915 /* translate the packets before forwarding them */
6916 if ((etypef & ETHER_TYPE_FLAG_IP_ARP) != 0) {
6917 m = bridge_mac_nat_translate_list(sc, sbif, dst_if, m);
6918 }
6919 } else if (ether_type_flag_is_ip(etypef) &&
6920 bif_has_checksum_offload(dbif) &&
6921 !bif_has_checksum_offload(sbif)) {
6922 /*
6923 * If the destination interface has checksum offload enabled,
6924 * verify the checksum now, unless the source interface also has
6925 * checksum offload enabled. The checksum in that case has
6926 * already just been computed and verifying it is unnecessary.
6927 */
6928 m = bridge_verify_checksum_list(m, dbif);
6929 }
6930 sc_filter_flags = sc->sc_filter_flags;
6931 bridge_ifp = sc->sc_ifp;
6932 src_if = sbif->bif_ifp;
6933 BRIDGE_UNLOCK(sc);
6934 if (PF_IS_ENABLED && (sc_filter_flags & IFBF_FILT_MEMBER)) {
6935 m = bridge_pf_list(m, dst_if, sc_filter_flags, false);
6936 }
6937
6938 /*
6939 * We're forwarding inbound packets for which the checksums must
6940 * already have been computed and if required, verified.
6941 */
6942 if (m != NULL) {
6943 bridge_enqueue_multi(bridge_ifp, src_if, dst_if, etypef, m,
6944 CHECKSUM_OPERATION_CLEAR_OFFLOAD);
6945 }
6946 return;
6947
6948 drop:
6949 BRIDGE_UNLOCK(sc);
6950 m_freem_list(m);
6951 return;
6952 }
6953
6954 /*
6955 * bridge_span:
6956 *
6957 * Duplicate a packet out one or more interfaces that are in span mode,
6958 * the original mbuf is unmodified.
6959 */
6960 static void
6961 bridge_span(struct bridge_softc *sc, struct mbuf *m)
6962 {
6963 struct bridge_iflist *bif;
6964 struct ifnet *dst_if;
6965 struct mbuf *mc;
6966
6967 if (TAILQ_EMPTY(&sc->sc_spanlist)) {
6968 return;
6969 }
6970
6971 TAILQ_FOREACH(bif, &sc->sc_spanlist, bif_next) {
6972 dst_if = bif->bif_ifp;
6973
6974 if ((dst_if->if_flags & IFF_RUNNING) == 0) {
6975 continue;
6976 }
6977
6978 mc = m_copypacket(m, M_DONTWAIT);
6979 if (mc == NULL) {
6980 (void) ifnet_stat_increment_out(sc->sc_ifp, 0, 0, 1);
6981 continue;
6982 }
6983
6984 (void) bridge_enqueue(sc->sc_ifp, NULL, dst_if, mc,
6985 CHECKSUM_OPERATION_NONE);
6986 }
6987 }
6988
6989 /*
6990 * bridge_rtupdate:
6991 *
6992 * Add a bridge routing entry.
6993 */
6994 static int
6995 bridge_rtupdate(struct bridge_softc *sc, const uint8_t dst[ETHER_ADDR_LEN], uint16_t vlan,
6996 struct bridge_iflist *bif, int setflags, uint8_t flags)
6997 {
6998 struct bridge_rtnode *brt;
6999 int error;
7000
7001 BRIDGE_LOCK_ASSERT_HELD(sc);
7002
7003 /* Check the source address is valid and not multicast. */
7004 if (ETHER_IS_MULTICAST(dst) ||
7005 (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
7006 dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0) {
7007 return EINVAL;
7008 }
7009
7010 /* 802.1p frames map to vlan 1 */
7011 if (vlan == 0) {
7012 vlan = 1;
7013 }
7014
7015 /*
7016 * A route for this destination might already exist. If so,
7017 * update it, otherwise create a new one.
7018 */
7019 if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
7020 if (sc->sc_brtcnt >= sc->sc_brtmax) {
7021 sc->sc_brtexceeded++;
7022 return ENOSPC;
7023 }
7024 /* Check per interface address limits (if enabled) */
7025 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
7026 bif->bif_addrexceeded++;
7027 return ENOSPC;
7028 }
7029
7030 /*
7031 * Allocate a new bridge forwarding node, and
7032 * initialize the expiration time and Ethernet
7033 * address.
7034 */
7035 brt = zalloc_noblock(bridge_rtnode_pool);
7036 if (brt == NULL) {
7037 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_RT_TABLE,
7038 "zalloc_nolock failed");
7039 return ENOMEM;
7040 }
7041 bzero(brt, sizeof(struct bridge_rtnode));
7042
7043 if (bif->bif_ifflags & IFBIF_STICKY) {
7044 brt->brt_flags = IFBAF_STICKY;
7045 } else {
7046 brt->brt_flags = IFBAF_DYNAMIC;
7047 }
7048
7049 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
7050 brt->brt_vlan = vlan;
7051
7052 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
7053 zfree(bridge_rtnode_pool, brt);
7054 return error;
7055 }
7056 brt->brt_dst = bif;
7057 bif->bif_addrcnt++;
7058 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_RT_TABLE,
7059 "added %02x:%02x:%02x:%02x:%02x:%02x "
7060 "on %s count %u hashsize %u",
7061 dst[0], dst[1], dst[2], dst[3], dst[4], dst[5],
7062 sc->sc_ifp->if_xname, sc->sc_brtcnt,
7063 sc->sc_rthash_size);
7064 }
7065
7066 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
7067 brt->brt_dst != bif) {
7068 brt->brt_dst->bif_addrcnt--;
7069 brt->brt_dst = bif;
7070 brt->brt_dst->bif_addrcnt++;
7071 }
7072
7073 if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
7074 unsigned long now;
7075
7076 now = (unsigned long) net_uptime();
7077 brt->brt_expire = now + sc->sc_brttimeout;
7078 }
7079 if (setflags) {
7080 brt->brt_flags = flags;
7081 }
7082
7083 return 0;
7084 }
7085
7086 /*
7087 * bridge_rtlookup:
7088 *
7089 * Lookup the destination interface for an address.
7090 */
7091 static struct ifnet *
7092 bridge_rtlookup(struct bridge_softc *sc, const uint8_t addr[ETHER_ADDR_LEN], uint16_t vlan)
7093 {
7094 struct bridge_rtnode *brt;
7095
7096 BRIDGE_LOCK_ASSERT_HELD(sc);
7097
7098 if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL) {
7099 return NULL;
7100 }
7101
7102 return brt->brt_ifp;
7103 }
7104
7105 /*
7106 * bridge_rttrim:
7107 *
7108 * Trim the routine table so that we have a number
7109 * of routing entries less than or equal to the
7110 * maximum number.
7111 */
7112 static void
7113 bridge_rttrim(struct bridge_softc *sc)
7114 {
7115 struct bridge_rtnode *brt, *nbrt;
7116
7117 BRIDGE_LOCK_ASSERT_HELD(sc);
7118
7119 /* Make sure we actually need to do this. */
7120 if (sc->sc_brtcnt <= sc->sc_brtmax) {
7121 return;
7122 }
7123
7124 /* Force an aging cycle; this might trim enough addresses. */
7125 bridge_rtage(sc);
7126 if (sc->sc_brtcnt <= sc->sc_brtmax) {
7127 return;
7128 }
7129
7130 LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
7131 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
7132 bridge_rtnode_destroy(sc, brt);
7133 if (sc->sc_brtcnt <= sc->sc_brtmax) {
7134 return;
7135 }
7136 }
7137 }
7138 }
7139
7140 /*
7141 * bridge_aging_timer:
7142 *
7143 * Aging periodic timer for the bridge routing table.
7144 */
7145 static void
7146 bridge_aging_timer(struct bridge_softc *sc)
7147 {
7148 BRIDGE_LOCK_ASSERT_HELD(sc);
7149
7150 bridge_rtage(sc);
7151 if ((sc->sc_ifp->if_flags & IFF_RUNNING) &&
7152 (sc->sc_flags & SCF_DETACHING) == 0) {
7153 sc->sc_aging_timer.bdc_sc = sc;
7154 sc->sc_aging_timer.bdc_func = bridge_aging_timer;
7155 sc->sc_aging_timer.bdc_ts.tv_sec = bridge_rtable_prune_period;
7156 bridge_schedule_delayed_call(&sc->sc_aging_timer);
7157 }
7158 }
7159
7160 /*
7161 * bridge_rtage:
7162 *
7163 * Perform an aging cycle.
7164 */
7165 static void
7166 bridge_rtage(struct bridge_softc *sc)
7167 {
7168 struct bridge_rtnode *brt, *nbrt;
7169 unsigned long now;
7170
7171 BRIDGE_LOCK_ASSERT_HELD(sc);
7172
7173 now = (unsigned long) net_uptime();
7174
7175 LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
7176 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
7177 if (now >= brt->brt_expire) {
7178 bridge_rtnode_destroy(sc, brt);
7179 }
7180 }
7181 }
7182 if (sc->sc_mac_nat_bif != NULL) {
7183 bridge_mac_nat_age_entries(sc, now);
7184 }
7185 }
7186
7187 /*
7188 * bridge_rtflush:
7189 *
7190 * Remove all dynamic addresses from the bridge.
7191 */
7192 static void
7193 bridge_rtflush(struct bridge_softc *sc, int full)
7194 {
7195 struct bridge_rtnode *brt, *nbrt;
7196
7197 BRIDGE_LOCK_ASSERT_HELD(sc);
7198
7199 LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
7200 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
7201 bridge_rtnode_destroy(sc, brt);
7202 }
7203 }
7204 }
7205
7206 /*
7207 * bridge_rtdaddr:
7208 *
7209 * Remove an address from the table.
7210 */
7211 static int
7212 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t addr[ETHER_ADDR_LEN], uint16_t vlan)
7213 {
7214 struct bridge_rtnode *brt;
7215 int found = 0;
7216
7217 BRIDGE_LOCK_ASSERT_HELD(sc);
7218
7219 /*
7220 * If vlan is zero then we want to delete for all vlans so the lookup
7221 * may return more than one.
7222 */
7223 while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
7224 bridge_rtnode_destroy(sc, brt);
7225 found = 1;
7226 }
7227
7228 return found ? 0 : ENOENT;
7229 }
7230
7231 /*
7232 * bridge_rtdelete:
7233 *
7234 * Delete routes to a specific member interface.
7235 */
7236 static void
7237 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
7238 {
7239 struct bridge_rtnode *brt, *nbrt;
7240
7241 BRIDGE_LOCK_ASSERT_HELD(sc);
7242
7243 LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
7244 if (brt->brt_ifp == ifp && (full ||
7245 (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)) {
7246 bridge_rtnode_destroy(sc, brt);
7247 }
7248 }
7249 }
7250
7251 /*
7252 * bridge_rtable_init:
7253 *
7254 * Initialize the route table for this bridge.
7255 */
7256 static int
7257 bridge_rtable_init(struct bridge_softc *sc)
7258 {
7259 u_int32_t i;
7260
7261 sc->sc_rthash = kalloc_type(struct _bridge_rtnode_list,
7262 BRIDGE_RTHASH_SIZE, Z_WAITOK_ZERO_NOFAIL);
7263 sc->sc_rthash_size = BRIDGE_RTHASH_SIZE;
7264
7265 for (i = 0; i < sc->sc_rthash_size; i++) {
7266 LIST_INIT(&sc->sc_rthash[i]);
7267 }
7268
7269 sc->sc_rthash_key = RandomULong();
7270
7271 LIST_INIT(&sc->sc_rtlist);
7272
7273 return 0;
7274 }
7275
7276 /*
7277 * bridge_rthash_delayed_resize:
7278 *
7279 * Resize the routing table hash on a delayed thread call.
7280 */
7281 static void
7282 bridge_rthash_delayed_resize(struct bridge_softc *sc)
7283 {
7284 u_int32_t new_rthash_size = 0;
7285 u_int32_t old_rthash_size = 0;
7286 struct _bridge_rtnode_list *new_rthash = NULL;
7287 struct _bridge_rtnode_list *old_rthash = NULL;
7288 u_int32_t i;
7289 struct bridge_rtnode *brt;
7290 int error = 0;
7291
7292 BRIDGE_LOCK_ASSERT_HELD(sc);
7293
7294 /*
7295 * Four entries per hash bucket is our ideal load factor
7296 */
7297 if (sc->sc_brtcnt < sc->sc_rthash_size * 4) {
7298 goto out;
7299 }
7300
7301 /*
7302 * Doubling the number of hash buckets may be too simplistic
7303 * especially when facing a spike of new entries
7304 */
7305 new_rthash_size = sc->sc_rthash_size * 2;
7306
7307 sc->sc_flags |= SCF_RESIZING;
7308 BRIDGE_UNLOCK(sc);
7309
7310 new_rthash = kalloc_type(struct _bridge_rtnode_list, new_rthash_size,
7311 Z_WAITOK | Z_ZERO);
7312
7313 BRIDGE_LOCK(sc);
7314 sc->sc_flags &= ~SCF_RESIZING;
7315
7316 if (new_rthash == NULL) {
7317 error = ENOMEM;
7318 goto out;
7319 }
7320 if ((sc->sc_flags & SCF_DETACHING)) {
7321 error = ENODEV;
7322 goto out;
7323 }
7324 /*
7325 * Fail safe from here on
7326 */
7327 old_rthash = sc->sc_rthash;
7328 old_rthash_size = sc->sc_rthash_size;
7329 sc->sc_rthash = new_rthash;
7330 sc->sc_rthash_size = new_rthash_size;
7331
7332 /*
7333 * Get a new key to force entries to be shuffled around to reduce
7334 * the likelihood they will land in the same buckets
7335 */
7336 sc->sc_rthash_key = RandomULong();
7337
7338 for (i = 0; i < sc->sc_rthash_size; i++) {
7339 LIST_INIT(&sc->sc_rthash[i]);
7340 }
7341
7342 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
7343 LIST_REMOVE(brt, brt_hash);
7344 (void) bridge_rtnode_hash(sc, brt);
7345 }
7346 out:
7347 if (error == 0) {
7348 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_RT_TABLE,
7349 "%s new size %u",
7350 sc->sc_ifp->if_xname, sc->sc_rthash_size);
7351 kfree_type(struct _bridge_rtnode_list, old_rthash_size, old_rthash);
7352 } else {
7353 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_RT_TABLE,
7354 "%s failed %d", sc->sc_ifp->if_xname, error);
7355 kfree_type(struct _bridge_rtnode_list, new_rthash_size, new_rthash);
7356 }
7357 }
7358
7359 /*
7360 * Resize the number of hash buckets based on the load factor
7361 * Currently only grow
7362 * Failing to resize the hash table is not fatal
7363 */
7364 static void
7365 bridge_rthash_resize(struct bridge_softc *sc)
7366 {
7367 BRIDGE_LOCK_ASSERT_HELD(sc);
7368
7369 if ((sc->sc_flags & SCF_DETACHING) || (sc->sc_flags & SCF_RESIZING)) {
7370 return;
7371 }
7372
7373 /*
7374 * Four entries per hash bucket is our ideal load factor
7375 */
7376 if (sc->sc_brtcnt < sc->sc_rthash_size * 4) {
7377 return;
7378 }
7379 /*
7380 * Hard limit on the size of the routing hash table
7381 */
7382 if (sc->sc_rthash_size >= bridge_rtable_hash_size_max) {
7383 return;
7384 }
7385
7386 sc->sc_resize_call.bdc_sc = sc;
7387 sc->sc_resize_call.bdc_func = bridge_rthash_delayed_resize;
7388 bridge_schedule_delayed_call(&sc->sc_resize_call);
7389 }
7390
7391 /*
7392 * bridge_rtable_fini:
7393 *
7394 * Deconstruct the route table for this bridge.
7395 */
7396 static void
7397 bridge_rtable_fini(struct bridge_softc *sc)
7398 {
7399 KASSERT(sc->sc_brtcnt == 0,
7400 ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
7401 kfree_type_counted_by(struct _bridge_rtnode_list, sc->sc_rthash_size,
7402 sc->sc_rthash);
7403 sc->sc_rthash = NULL;
7404 sc->sc_rthash_size = 0;
7405 }
7406
7407 /*
7408 * The following hash function is adapted from "Hash Functions" by Bob Jenkins
7409 * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
7410 */
7411 #define mix(a, b, c) \
7412 do { \
7413 a -= b; a -= c; a ^= (c >> 13); \
7414 b -= c; b -= a; b ^= (a << 8); \
7415 c -= a; c -= b; c ^= (b >> 13); \
7416 a -= b; a -= c; a ^= (c >> 12); \
7417 b -= c; b -= a; b ^= (a << 16); \
7418 c -= a; c -= b; c ^= (b >> 5); \
7419 a -= b; a -= c; a ^= (c >> 3); \
7420 b -= c; b -= a; b ^= (a << 10); \
7421 c -= a; c -= b; c ^= (b >> 15); \
7422 } while ( /*CONSTCOND*/ 0)
7423
7424 static __inline uint32_t
7425 bridge_rthash(struct bridge_softc *sc, const uint8_t addr[ETHER_ADDR_LEN])
7426 {
7427 uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
7428
7429 b += addr[5] << 8;
7430 b += addr[4];
7431 a += addr[3] << 24;
7432 a += addr[2] << 16;
7433 a += addr[1] << 8;
7434 a += addr[0];
7435
7436 mix(a, b, c);
7437
7438 return c & BRIDGE_RTHASH_MASK(sc);
7439 }
7440
7441 #undef mix
7442
7443 static int
7444 bridge_rtnode_addr_cmp(const uint8_t a[ETHER_ADDR_LEN], const uint8_t b[ETHER_ADDR_LEN])
7445 {
7446 int i, d;
7447
7448 for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
7449 d = ((int)a[i]) - ((int)b[i]);
7450 }
7451
7452 return d;
7453 }
7454
7455 /*
7456 * bridge_rtnode_lookup:
7457 *
7458 * Look up a bridge route node for the specified destination. Compare the
7459 * vlan id or if zero then just return the first match.
7460 */
7461 static struct bridge_rtnode *
7462 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t addr[ETHER_ADDR_LEN],
7463 uint16_t vlan)
7464 {
7465 struct bridge_rtnode *brt;
7466 uint32_t hash;
7467 int dir;
7468
7469 BRIDGE_LOCK_ASSERT_HELD(sc);
7470
7471 hash = bridge_rthash(sc, addr);
7472 LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
7473 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
7474 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0)) {
7475 return brt;
7476 }
7477 if (dir > 0) {
7478 return NULL;
7479 }
7480 }
7481
7482 return NULL;
7483 }
7484
7485 /*
7486 * bridge_rtnode_hash:
7487 *
7488 * Insert the specified bridge node into the route hash table.
7489 * This is used when adding a new node or to rehash when resizing
7490 * the hash table
7491 */
7492 static int
7493 bridge_rtnode_hash(struct bridge_softc *sc, struct bridge_rtnode *brt)
7494 {
7495 struct bridge_rtnode *lbrt;
7496 uint32_t hash;
7497 int dir;
7498
7499 BRIDGE_LOCK_ASSERT_HELD(sc);
7500
7501 hash = bridge_rthash(sc, brt->brt_addr);
7502
7503 lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
7504 if (lbrt == NULL) {
7505 LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
7506 goto out;
7507 }
7508
7509 do {
7510 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
7511 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan) {
7512 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_RT_TABLE,
7513 "%s EEXIST %02x:%02x:%02x:%02x:%02x:%02x",
7514 sc->sc_ifp->if_xname,
7515 brt->brt_addr[0], brt->brt_addr[1],
7516 brt->brt_addr[2], brt->brt_addr[3],
7517 brt->brt_addr[4], brt->brt_addr[5]);
7518 return EEXIST;
7519 }
7520 if (dir > 0) {
7521 LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
7522 goto out;
7523 }
7524 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
7525 LIST_INSERT_AFTER(lbrt, brt, brt_hash);
7526 goto out;
7527 }
7528 lbrt = LIST_NEXT(lbrt, brt_hash);
7529 } while (lbrt != NULL);
7530
7531 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_RT_TABLE,
7532 "%s impossible %02x:%02x:%02x:%02x:%02x:%02x",
7533 sc->sc_ifp->if_xname,
7534 brt->brt_addr[0], brt->brt_addr[1], brt->brt_addr[2],
7535 brt->brt_addr[3], brt->brt_addr[4], brt->brt_addr[5]);
7536 out:
7537 return 0;
7538 }
7539
7540 /*
7541 * bridge_rtnode_insert:
7542 *
7543 * Insert the specified bridge node into the route table. We
7544 * assume the entry is not already in the table.
7545 */
7546 static int
7547 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
7548 {
7549 int error;
7550
7551 error = bridge_rtnode_hash(sc, brt);
7552 if (error != 0) {
7553 return error;
7554 }
7555
7556 LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
7557 sc->sc_brtcnt++;
7558
7559 bridge_rthash_resize(sc);
7560
7561 return 0;
7562 }
7563
7564 /*
7565 * bridge_rtnode_destroy:
7566 *
7567 * Destroy a bridge rtnode.
7568 */
7569 static void
7570 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
7571 {
7572 BRIDGE_LOCK_ASSERT_HELD(sc);
7573
7574 LIST_REMOVE(brt, brt_hash);
7575
7576 LIST_REMOVE(brt, brt_list);
7577 sc->sc_brtcnt--;
7578 brt->brt_dst->bif_addrcnt--;
7579 zfree(bridge_rtnode_pool, brt);
7580 }
7581
7582 #if BRIDGESTP
7583 /*
7584 * bridge_rtable_expire:
7585 *
7586 * Set the expiry time for all routes on an interface.
7587 */
7588 static void
7589 bridge_rtable_expire(struct ifnet *ifp, int age)
7590 {
7591 struct bridge_softc *sc = ifp->if_bridge;
7592 struct bridge_rtnode *brt;
7593
7594 BRIDGE_LOCK(sc);
7595
7596 /*
7597 * If the age is zero then flush, otherwise set all the expiry times to
7598 * age for the interface
7599 */
7600 if (age == 0) {
7601 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
7602 } else {
7603 unsigned long now;
7604
7605 now = (unsigned long) net_uptime();
7606
7607 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
7608 /* Cap the expiry time to 'age' */
7609 if (brt->brt_ifp == ifp &&
7610 brt->brt_expire > now + age &&
7611 (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
7612 brt->brt_expire = now + age;
7613 }
7614 }
7615 }
7616 BRIDGE_UNLOCK(sc);
7617 }
7618
7619 /*
7620 * bridge_state_change:
7621 *
7622 * Callback from the bridgestp code when a port changes states.
7623 */
7624 static void
7625 bridge_state_change(struct ifnet *ifp, int state)
7626 {
7627 struct bridge_softc *sc = ifp->if_bridge;
7628 static const char *stpstates[] = {
7629 "disabled",
7630 "listening",
7631 "learning",
7632 "forwarding",
7633 "blocking",
7634 "discarding"
7635 };
7636
7637 if (log_stp) {
7638 log(LOG_NOTICE, "%s: state changed to %s on %s",
7639 sc->sc_ifp->if_xname,
7640 stpstates[state], ifp->if_xname);
7641 }
7642 }
7643 #endif /* BRIDGESTP */
7644
7645 /*
7646 * bridge_detach:
7647 *
7648 * Callback when interface has been detached.
7649 */
7650 static void
7651 bridge_detach(ifnet_t ifp)
7652 {
7653 struct bridge_softc *sc = (struct bridge_softc *)ifnet_softc(ifp);
7654
7655 #if BRIDGESTP
7656 bstp_detach(&sc->sc_stp);
7657 #endif /* BRIDGESTP */
7658
7659 /* Tear down the routing table. */
7660 bridge_rtable_fini(sc);
7661
7662 lck_mtx_lock(&bridge_list_mtx);
7663 LIST_REMOVE(sc, sc_list);
7664 lck_mtx_unlock(&bridge_list_mtx);
7665
7666 ifnet_release(ifp);
7667
7668 lck_mtx_destroy(&sc->sc_mtx, &bridge_lock_grp);
7669 kfree_type(struct bridge_softc, sc);
7670 }
7671
7672 /*
7673 * bridge_link_event:
7674 *
7675 * Report a data link event on an interface
7676 */
7677 static void
7678 bridge_link_event(struct ifnet *ifp, u_int32_t event_code)
7679 {
7680 struct event {
7681 u_int32_t ifnet_family;
7682 u_int32_t unit;
7683 char if_name[IFNAMSIZ];
7684 };
7685 _Alignas(struct kern_event_msg) char message[sizeof(struct kern_event_msg) + sizeof(struct event)] = { 0 };
7686 struct kern_event_msg *header = (struct kern_event_msg*)message;
7687 struct event *data = (struct event *)(message + KEV_MSG_HEADER_SIZE);
7688
7689 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_LIFECYCLE,
7690 "%s event_code %u - %s", ifp->if_xname,
7691 event_code, dlil_kev_dl_code_str(event_code));
7692 header->total_size = sizeof(message);
7693 header->vendor_code = KEV_VENDOR_APPLE;
7694 header->kev_class = KEV_NETWORK_CLASS;
7695 header->kev_subclass = KEV_DL_SUBCLASS;
7696 header->event_code = event_code;
7697 data->ifnet_family = ifnet_family(ifp);
7698 data->unit = (u_int32_t)ifnet_unit(ifp);
7699 strlcpy(data->if_name, ifnet_name(ifp), IFNAMSIZ);
7700 ifnet_event(ifp, header);
7701 }
7702
7703 #define BRIDGE_HF_DROP(reason, func, line) { \
7704 bridge_hostfilter_stats.reason++; \
7705 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_HOSTFILTER, \
7706 "%s.%d" #reason, func, line); \
7707 error = EINVAL; \
7708 }
7709
7710 /*
7711 * Make sure this is a DHCP or Bootp request that match the host filter
7712 */
7713 static int
7714 bridge_dhcp_filter(struct bridge_iflist *bif, struct mbuf *m, size_t offset)
7715 {
7716 int error = EINVAL;
7717 struct dhcp dhcp;
7718
7719 /*
7720 * Note: We use the dhcp structure because bootp structure definition
7721 * is larger and some vendors do not pad the request
7722 */
7723 error = mbuf_copydata(m, offset, sizeof(struct dhcp), &dhcp);
7724 if (error != 0) {
7725 BRIDGE_HF_DROP(brhf_dhcp_too_small, __func__, __LINE__);
7726 goto done;
7727 }
7728 if (dhcp.dp_op != BOOTREQUEST) {
7729 BRIDGE_HF_DROP(brhf_dhcp_bad_op, __func__, __LINE__);
7730 goto done;
7731 }
7732 /*
7733 * The hardware address must be an exact match
7734 */
7735 if (dhcp.dp_htype != ARPHRD_ETHER) {
7736 BRIDGE_HF_DROP(brhf_dhcp_bad_htype, __func__, __LINE__);
7737 goto done;
7738 }
7739 if (dhcp.dp_hlen != ETHER_ADDR_LEN) {
7740 BRIDGE_HF_DROP(brhf_dhcp_bad_hlen, __func__, __LINE__);
7741 goto done;
7742 }
7743 if (bcmp(dhcp.dp_chaddr, bif->bif_hf_hwsrc,
7744 ETHER_ADDR_LEN) != 0) {
7745 BRIDGE_HF_DROP(brhf_dhcp_bad_chaddr, __func__, __LINE__);
7746 goto done;
7747 }
7748 /*
7749 * Client address must match the host address or be not specified
7750 */
7751 if (dhcp.dp_ciaddr.s_addr != bif->bif_hf_ipsrc.s_addr &&
7752 dhcp.dp_ciaddr.s_addr != INADDR_ANY) {
7753 BRIDGE_HF_DROP(brhf_dhcp_bad_ciaddr, __func__, __LINE__);
7754 goto done;
7755 }
7756 error = 0;
7757 done:
7758 return error;
7759 }
7760
7761 static int
7762 bridge_host_filter_arp(struct bridge_iflist *bif, mbuf_t *data)
7763 {
7764 struct ether_arp *ea;
7765 struct ether_header *eh;
7766 int error = EINVAL;
7767 mbuf_t m = *data;
7768 size_t minlen = sizeof(struct ether_header) + sizeof(struct ether_arp);
7769
7770 /*
7771 * Make the Ethernet and ARP headers contiguous
7772 */
7773 if (mbuf_pkthdr_len(m) < minlen) {
7774 BRIDGE_HF_DROP(brhf_arp_too_small, __func__, __LINE__);
7775 goto done;
7776 }
7777 if (mbuf_len(m) < minlen && mbuf_pullup(data, minlen) != 0) {
7778 BRIDGE_HF_DROP(brhf_arp_pullup_failed,
7779 __func__, __LINE__);
7780 goto done;
7781 }
7782 m = *data;
7783
7784 /*
7785 * Restrict Ethernet protocols to ARP and IP/IPv6
7786 */
7787 eh = mtod(m, struct ether_header *);
7788 ea = (struct ether_arp *)(eh + 1);
7789 if (ea->arp_hrd != HTONS_ARPHRD_ETHER) {
7790 BRIDGE_HF_DROP(brhf_arp_bad_hw_type,
7791 __func__, __LINE__);
7792 goto done;
7793 }
7794 if (ea->arp_pro != HTONS_ETHERTYPE_IP) {
7795 BRIDGE_HF_DROP(brhf_arp_bad_pro_type,
7796 __func__, __LINE__);
7797 goto done;
7798 }
7799 /*
7800 * Verify the address lengths are correct
7801 */
7802 if (ea->arp_hln != ETHER_ADDR_LEN) {
7803 BRIDGE_HF_DROP(brhf_arp_bad_hw_len, __func__, __LINE__);
7804 goto done;
7805 }
7806 if (ea->arp_pln != sizeof(struct in_addr)) {
7807 BRIDGE_HF_DROP(brhf_arp_bad_pro_len,
7808 __func__, __LINE__);
7809 goto done;
7810 }
7811 /*
7812 * Allow only ARP request or ARP reply
7813 */
7814 if (ea->arp_op != HTONS_ARPOP_REQUEST &&
7815 ea->arp_op != HTONS_ARPOP_REPLY) {
7816 BRIDGE_HF_DROP(brhf_arp_bad_op, __func__, __LINE__);
7817 goto done;
7818 }
7819 if ((bif->bif_flags & BIFF_HF_HWSRC) != 0) {
7820 /*
7821 * Verify source hardware address matches
7822 */
7823 if (bcmp(ea->arp_sha, bif->bif_hf_hwsrc,
7824 ETHER_ADDR_LEN) != 0) {
7825 BRIDGE_HF_DROP(brhf_arp_bad_sha, __func__, __LINE__);
7826 goto done;
7827 }
7828 }
7829 if ((bif->bif_flags & BIFF_HF_IPSRC) != 0) {
7830 /*
7831 * Verify source protocol address:
7832 * May be null for an ARP probe
7833 */
7834 if (bcmp(ea->arp_spa, &bif->bif_hf_ipsrc.s_addr,
7835 sizeof(struct in_addr)) != 0 &&
7836 bcmp(ea->arp_spa, &inaddr_any,
7837 sizeof(struct in_addr)) != 0) {
7838 BRIDGE_HF_DROP(brhf_arp_bad_spa, __func__, __LINE__);
7839 goto done;
7840 }
7841 }
7842 bridge_hostfilter_stats.brhf_arp_ok += 1;
7843 error = 0;
7844 done:
7845 return error;
7846 }
7847
7848 static int
7849 bridge_host_filter(struct bridge_iflist *bif, mbuf_t *data)
7850 {
7851 int error = EINVAL;
7852 struct ether_header *eh;
7853 mbuf_t m = *data;
7854
7855 eh = mtod(m, struct ether_header *);
7856
7857 /*
7858 * Restrict the source hardware address
7859 */
7860 if ((bif->bif_flags & BIFF_HF_HWSRC) != 0 &&
7861 bcmp(eh->ether_shost, bif->bif_hf_hwsrc,
7862 ETHER_ADDR_LEN) != 0) {
7863 BRIDGE_HF_DROP(brhf_bad_ether_srchw_addr, __func__, __LINE__);
7864 goto done;
7865 }
7866
7867 /*
7868 * Restrict Ethernet protocols to ARP and IP/IPv6
7869 */
7870 if (eh->ether_type == htons(ETHERTYPE_ARP)) {
7871 error = bridge_host_filter_arp(bif, data);
7872 if (error != 0) {
7873 goto done;
7874 }
7875 } else if (eh->ether_type == htons(ETHERTYPE_IP)) {
7876 size_t minlen = sizeof(struct ether_header) + sizeof(struct ip);
7877 struct ip iphdr;
7878 size_t offset;
7879
7880 /*
7881 * Make the Ethernet and IP headers contiguous
7882 */
7883 if (mbuf_pkthdr_len(m) < minlen) {
7884 BRIDGE_HF_DROP(brhf_ip_too_small, __func__, __LINE__);
7885 goto done;
7886 }
7887 offset = sizeof(struct ether_header);
7888 error = mbuf_copydata(m, offset, sizeof(struct ip), &iphdr);
7889 if (error != 0) {
7890 BRIDGE_HF_DROP(brhf_ip_too_small, __func__, __LINE__);
7891 goto done;
7892 }
7893 if ((bif->bif_flags & BIFF_HF_IPSRC) != 0) {
7894 /*
7895 * Verify the source IP address
7896 */
7897 if (iphdr.ip_p == IPPROTO_UDP) {
7898 struct udphdr udp;
7899
7900 minlen += sizeof(struct udphdr);
7901 if (mbuf_pkthdr_len(m) < minlen) {
7902 BRIDGE_HF_DROP(brhf_ip_too_small,
7903 __func__, __LINE__);
7904 goto done;
7905 }
7906
7907 /*
7908 * Allow all zero addresses for DHCP requests
7909 */
7910 if (iphdr.ip_src.s_addr != bif->bif_hf_ipsrc.s_addr &&
7911 iphdr.ip_src.s_addr != INADDR_ANY) {
7912 BRIDGE_HF_DROP(brhf_ip_bad_srcaddr,
7913 __func__, __LINE__);
7914 goto done;
7915 }
7916 offset = sizeof(struct ether_header) +
7917 (IP_VHL_HL(iphdr.ip_vhl) << 2);
7918 error = mbuf_copydata(m, offset,
7919 sizeof(struct udphdr), &udp);
7920 if (error != 0) {
7921 BRIDGE_HF_DROP(brhf_ip_too_small,
7922 __func__, __LINE__);
7923 goto done;
7924 }
7925 /*
7926 * Either it's a Bootp/DHCP packet that we like or
7927 * it's a UDP packet from the host IP as source address
7928 */
7929 if (udp.uh_sport == htons(IPPORT_BOOTPC) &&
7930 udp.uh_dport == htons(IPPORT_BOOTPS)) {
7931 minlen += sizeof(struct dhcp);
7932 if (mbuf_pkthdr_len(m) < minlen) {
7933 BRIDGE_HF_DROP(brhf_ip_too_small,
7934 __func__, __LINE__);
7935 goto done;
7936 }
7937 offset += sizeof(struct udphdr);
7938 error = bridge_dhcp_filter(bif, m, offset);
7939 if (error != 0) {
7940 goto done;
7941 }
7942 } else if (iphdr.ip_src.s_addr == INADDR_ANY) {
7943 BRIDGE_HF_DROP(brhf_ip_bad_srcaddr,
7944 __func__, __LINE__);
7945 goto done;
7946 }
7947 } else if (iphdr.ip_src.s_addr != bif->bif_hf_ipsrc.s_addr) {
7948 assert(bif->bif_hf_ipsrc.s_addr != INADDR_ANY);
7949 BRIDGE_HF_DROP(brhf_ip_bad_srcaddr, __func__, __LINE__);
7950 goto done;
7951 }
7952 }
7953 /*
7954 * Allow only boring IP protocols
7955 */
7956 if (iphdr.ip_p != IPPROTO_TCP &&
7957 iphdr.ip_p != IPPROTO_UDP &&
7958 iphdr.ip_p != IPPROTO_ICMP &&
7959 iphdr.ip_p != IPPROTO_IGMP) {
7960 BRIDGE_HF_DROP(brhf_ip_bad_proto, __func__, __LINE__);
7961 goto done;
7962 }
7963 bridge_hostfilter_stats.brhf_ip_ok += 1;
7964 error = 0;
7965 } else if (eh->ether_type == htons(ETHERTYPE_IPV6)) {
7966 size_t minlen = sizeof(struct ether_header) + sizeof(struct ip6_hdr);
7967 struct ip6_hdr ip6hdr;
7968 size_t offset;
7969
7970 /*
7971 * Make the Ethernet and IP headers contiguous
7972 */
7973 if (mbuf_pkthdr_len(m) < minlen) {
7974 BRIDGE_HF_DROP(brhf_ip_too_small, __func__, __LINE__);
7975 goto done;
7976 }
7977 offset = sizeof(struct ether_header);
7978 error = mbuf_copydata(m, offset, sizeof(struct ip6_hdr), &ip6hdr);
7979 if (error != 0) {
7980 BRIDGE_HF_DROP(brhf_ip_too_small, __func__, __LINE__);
7981 goto done;
7982 }
7983 /*
7984 * Allow only boring IPv6 protocols
7985 */
7986 if (ip6hdr.ip6_nxt != IPPROTO_TCP &&
7987 ip6hdr.ip6_nxt != IPPROTO_UDP &&
7988 ip6hdr.ip6_nxt != IPPROTO_ICMPV6) {
7989 BRIDGE_HF_DROP(brhf_ip_bad_proto, __func__, __LINE__);
7990 goto done;
7991 }
7992 bridge_hostfilter_stats.brhf_ip_ok += 1;
7993 error = 0;
7994 } else {
7995 BRIDGE_HF_DROP(brhf_bad_ether_type, __func__, __LINE__);
7996 goto done;
7997 }
7998 done:
7999 if (error != 0) {
8000 if (BRIDGE_DBGF_ENABLED(BR_DBGF_HOSTFILTER)) {
8001 if (m) {
8002 brlog_mbuf_data(m, 0,
8003 sizeof(struct ether_header) +
8004 sizeof(struct ip));
8005 }
8006 }
8007
8008 if (m != NULL) {
8009 m_freem(m);
8010 *data = NULL;
8011 }
8012 }
8013 return error;
8014 }
8015
8016 /*
8017 * MAC NAT
8018 */
8019
8020 static errno_t
8021 bridge_mac_nat_enable(struct bridge_softc *sc, struct bridge_iflist *bif)
8022 {
8023 errno_t error = 0;
8024
8025 BRIDGE_LOCK_ASSERT_HELD(sc);
8026
8027 if (IFNET_IS_VMNET(bif->bif_ifp)) {
8028 error = EINVAL;
8029 goto done;
8030 }
8031 if (sc->sc_mac_nat_bif != NULL) {
8032 if (sc->sc_mac_nat_bif != bif) {
8033 error = EBUSY;
8034 }
8035 goto done;
8036 }
8037 sc->sc_mac_nat_bif = bif;
8038 bif->bif_ifflags |= IFBIF_MAC_NAT;
8039 bridge_mac_nat_populate_entries(sc);
8040
8041 done:
8042 return error;
8043 }
8044
8045 static void
8046 bridge_mac_nat_disable(struct bridge_softc *sc)
8047 {
8048 struct bridge_iflist *mac_nat_bif = sc->sc_mac_nat_bif;
8049
8050 assert(mac_nat_bif != NULL);
8051 bridge_mac_nat_flush_entries(sc, mac_nat_bif);
8052 mac_nat_bif->bif_ifflags &= ~IFBIF_MAC_NAT;
8053 sc->sc_mac_nat_bif = NULL;
8054 return;
8055 }
8056
8057 static void
8058 mac_nat_entry_print2(struct mac_nat_entry *mne,
8059 const char ifname[IFNAMSIZ], const char *msg1, const char *msg2)
8060 {
8061 int af;
8062 char etopbuf[24];
8063 char ntopbuf[MAX_IPv6_STR_LEN];
8064 const char *space;
8065
8066 af = ((mne->mne_flags & MNE_FLAGS_IPV6) != 0) ? AF_INET6 : AF_INET;
8067 ether_ntop(etopbuf, sizeof(etopbuf), mne->mne_mac);
8068 (void)inet_ntop(af, &mne->mne_u, ntopbuf, sizeof(ntopbuf));
8069 if (msg2 == NULL) {
8070 msg2 = "";
8071 space = "";
8072 } else {
8073 space = " ";
8074 }
8075 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8076 "%.*s %s%s%s %p (%s, %s, %s)", IFNAMSIZ, ifname, msg1, space, msg2, mne,
8077 mne->mne_bif->bif_ifp->if_xname, ntopbuf, etopbuf);
8078 }
8079
8080 static void
8081 mac_nat_entry_print(struct mac_nat_entry *mne,
8082 const char ifname[IFNAMSIZ], const char *msg)
8083 {
8084 mac_nat_entry_print2(mne, ifname, msg, NULL);
8085 }
8086
8087 static struct mac_nat_entry *
8088 bridge_lookup_mac_nat_entry_ipv4(const struct bridge_softc *sc, const struct in_addr *ip)
8089 {
8090 struct mac_nat_entry *mne;
8091 struct mac_nat_entry *ret_mne = NULL;
8092
8093 LIST_FOREACH(mne, &sc->sc_mne_list, mne_list) {
8094 if (mne->mne_ip.s_addr == ip->s_addr) {
8095 if (BRIDGE_DBGF_ENABLED(BR_DBGF_MAC_NAT)) {
8096 mac_nat_entry_print(mne, sc->sc_if_xname,
8097 "found");
8098 }
8099 ret_mne = mne;
8100 break;
8101 }
8102 }
8103
8104 return ret_mne;
8105 }
8106
8107 static struct mac_nat_entry *
8108 bridge_lookup_mac_nat_entry_ipv6(const struct bridge_softc *sc, const struct in6_addr *ip6)
8109 {
8110 struct mac_nat_entry *mne;
8111 struct mac_nat_entry *ret_mne = NULL;
8112
8113 LIST_FOREACH(mne, &sc->sc_mne_list_v6, mne_list) {
8114 if (IN6_ARE_ADDR_EQUAL(&mne->mne_ip6, ip6)) {
8115 if (BRIDGE_DBGF_ENABLED(BR_DBGF_MAC_NAT)) {
8116 mac_nat_entry_print(mne, sc->sc_if_xname,
8117 "found");
8118 }
8119 ret_mne = mne;
8120 break;
8121 }
8122 }
8123
8124 return ret_mne;
8125 }
8126
8127 static void
8128 bridge_destroy_mac_nat_entry(struct bridge_softc *sc,
8129 struct mac_nat_entry *mne, const char *reason)
8130 {
8131 LIST_REMOVE(mne, mne_list);
8132 if (BRIDGE_DBGF_ENABLED(BR_DBGF_MAC_NAT)) {
8133 mac_nat_entry_print(mne, sc->sc_if_xname, reason);
8134 }
8135 zfree(bridge_mne_pool, mne);
8136 sc->sc_mne_count--;
8137 }
8138
8139 static struct mac_nat_entry *
8140 bridge_create_mac_nat_entry_common(struct bridge_softc *sc,
8141 struct bridge_iflist *bif, const char eaddr[ETHER_ADDR_LEN])
8142 {
8143 struct mac_nat_entry *mne;
8144
8145 if (sc->sc_mne_count >= sc->sc_mne_max) {
8146 sc->sc_mne_allocation_failures++;
8147 return NULL;
8148 }
8149
8150 mne = zalloc_noblock(bridge_mne_pool);
8151 if (mne == NULL) {
8152 sc->sc_mne_allocation_failures++;
8153 return NULL;
8154 }
8155
8156 sc->sc_mne_count++;
8157 bzero(mne, sizeof(*mne));
8158 bcopy(eaddr, mne->mne_mac, sizeof(mne->mne_mac));
8159
8160 mne->mne_bif = bif;
8161 mne->mne_expire = (unsigned long)net_uptime() + sc->sc_brttimeout;
8162
8163 if (BRIDGE_DBGF_ENABLED(BR_DBGF_MAC_NAT)) {
8164 mac_nat_entry_print(mne, sc->sc_if_xname, "created");
8165 }
8166
8167 return mne;
8168 }
8169
8170 static struct mac_nat_entry *
8171 bridge_create_mac_nat_entry_ipv4(struct bridge_softc *sc,
8172 struct bridge_iflist *bif, const struct in_addr *ip, const char eaddr[ETHER_ADDR_LEN])
8173 {
8174 struct mac_nat_entry *mne;
8175
8176 mne = bridge_create_mac_nat_entry_common(sc, bif, eaddr);
8177 if (mne == NULL) {
8178 return NULL;
8179 }
8180
8181 bcopy(ip, &mne->mne_ip, sizeof(mne->mne_ip));
8182 LIST_INSERT_HEAD(&sc->sc_mne_list, mne, mne_list);
8183
8184 return mne;
8185 }
8186
8187 static struct mac_nat_entry *
8188 bridge_create_mac_nat_entry_ipv6(struct bridge_softc *sc,
8189 struct bridge_iflist *bif, const struct in6_addr *ip6, const char eaddr[ETHER_ADDR_LEN])
8190 {
8191 struct mac_nat_entry *mne;
8192
8193 mne = bridge_create_mac_nat_entry_common(sc, bif, eaddr);
8194 if (mne == NULL) {
8195 return NULL;
8196 }
8197
8198 bcopy(ip6, &mne->mne_ip6, sizeof(mne->mne_ip6));
8199 mne->mne_flags |= MNE_FLAGS_IPV6;
8200 LIST_INSERT_HEAD(&sc->sc_mne_list_v6, mne, mne_list);
8201
8202 return mne;
8203 }
8204
8205 static struct mac_nat_entry *
8206 bridge_update_mac_nat_entry_common(struct bridge_softc *sc, struct bridge_iflist *bif,
8207 struct mac_nat_entry *mne, const char eaddr[ETHER_ADDR_LEN])
8208 {
8209 struct bridge_iflist *mac_nat_bif = sc->sc_mac_nat_bif;
8210
8211 if (mne->mne_bif == mac_nat_bif) {
8212 /* the MAC NAT interface takes precedence */
8213 if (BRIDGE_DBGF_ENABLED(BR_DBGF_MAC_NAT)) {
8214 if (mne->mne_bif != bif) {
8215 mac_nat_entry_print2(mne,
8216 sc->sc_if_xname, "reject",
8217 bif->bif_ifp->if_xname);
8218 }
8219 }
8220 } else if (mne->mne_bif != bif) {
8221 const char *__null_terminated old_if = mne->mne_bif->bif_ifp->if_xname;
8222
8223 mne->mne_bif = bif;
8224 if (BRIDGE_DBGF_ENABLED(BR_DBGF_MAC_NAT)) {
8225 mac_nat_entry_print2(mne,
8226 sc->sc_if_xname, "replaced",
8227 old_if);
8228 }
8229 bcopy(eaddr, mne->mne_mac, sizeof(mne->mne_mac));
8230 }
8231
8232 mne->mne_expire = (unsigned long)net_uptime() + sc->sc_brttimeout;
8233
8234 return mne;
8235 }
8236
8237 static struct mac_nat_entry *
8238 bridge_update_mac_nat_entry_ipv4(struct bridge_softc *sc,
8239 struct bridge_iflist *bif, struct in_addr *ip, const char eaddr[ETHER_ADDR_LEN])
8240 {
8241 struct mac_nat_entry *mne;
8242
8243 mne = bridge_lookup_mac_nat_entry_ipv4(sc, ip);
8244 if (mne != NULL) {
8245 return bridge_update_mac_nat_entry_common(sc, bif, mne, eaddr);
8246 }
8247
8248 mne = bridge_create_mac_nat_entry_ipv4(sc, bif, ip, eaddr);
8249 return mne;
8250 }
8251
8252 static struct mac_nat_entry *
8253 bridge_update_mac_nat_entry_ipv6(struct bridge_softc *sc,
8254 struct bridge_iflist *bif, struct in6_addr *ip6, const char eaddr[ETHER_ADDR_LEN])
8255 {
8256 struct mac_nat_entry *mne;
8257
8258 mne = bridge_lookup_mac_nat_entry_ipv6(sc, ip6);
8259 if (mne != NULL) {
8260 return bridge_update_mac_nat_entry_common(sc, bif, mne, eaddr);
8261 }
8262
8263 mne = bridge_create_mac_nat_entry_ipv6(sc, bif, ip6, eaddr);
8264 return mne;
8265 }
8266
8267 static void
8268 bridge_mac_nat_flush_entries_common(struct bridge_softc *sc,
8269 struct mac_nat_entry_list *list, struct bridge_iflist *bif)
8270 {
8271 struct mac_nat_entry *mne;
8272 struct mac_nat_entry *tmne;
8273
8274 LIST_FOREACH_SAFE(mne, list, mne_list, tmne) {
8275 if (bif != NULL && mne->mne_bif != bif) {
8276 continue;
8277 }
8278 bridge_destroy_mac_nat_entry(sc, mne, "flushed");
8279 }
8280 }
8281
8282 /*
8283 * bridge_mac_nat_flush_entries:
8284 *
8285 * Flush MAC NAT entries for the specified member. Flush all entries if
8286 * the member is the one that requires MAC NAT, otherwise just flush the
8287 * ones for the specified member.
8288 */
8289 static void
8290 bridge_mac_nat_flush_entries(struct bridge_softc *sc, struct bridge_iflist * bif)
8291 {
8292 struct bridge_iflist *flush_bif;
8293
8294 flush_bif = (bif == sc->sc_mac_nat_bif) ? NULL : bif;
8295 bridge_mac_nat_flush_entries_common(sc, &sc->sc_mne_list, flush_bif);
8296 bridge_mac_nat_flush_entries_common(sc, &sc->sc_mne_list_v6, flush_bif);
8297 }
8298
8299 static void
8300 bridge_mac_nat_populate_entries(struct bridge_softc *sc)
8301 {
8302 errno_t error;
8303 ifnet_t ifp;
8304 uint16_t addresses_count = 0;
8305 ifaddr_t * __counted_by(addresses_count) list;
8306 struct bridge_iflist *mac_nat_bif = sc->sc_mac_nat_bif;
8307
8308 assert(mac_nat_bif != NULL);
8309 ifp = mac_nat_bif->bif_ifp;
8310 error = ifnet_get_address_list_family_with_count(ifp, &list, &addresses_count, 0);
8311 if (error != 0) {
8312 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
8313 "ifnet_get_address_list(%s) failed %d",
8314 ifp->if_xname, error);
8315 return;
8316 }
8317
8318 for (uint16_t i = 0; i < addresses_count; ++i) {
8319 sa_family_t af;
8320
8321 af = ifaddr_address_family(list[i]);
8322 switch (af) {
8323 case AF_INET: {
8324 struct sockaddr_in sin;
8325
8326 error = ifaddr_address(list[i], (struct sockaddr *)&sin, sizeof(sin));
8327 if (error != 0) {
8328 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
8329 "ifaddr_address failed %d",
8330 error);
8331 break;
8332 }
8333
8334 bridge_create_mac_nat_entry_ipv4(sc, mac_nat_bif, &sin.sin_addr, IF_LLADDR(ifp));
8335 break;
8336 }
8337
8338 case AF_INET6: {
8339 struct sockaddr_in6 sin6;
8340
8341 error = ifaddr_address(list[i], (struct sockaddr *)&sin6, sizeof(sin6));
8342 if (error != 0) {
8343 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
8344 "ifaddr_address failed %d",
8345 error);
8346 break;
8347 }
8348
8349 if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
8350 /* remove scope ID */
8351 sin6.sin6_addr.s6_addr16[1] = 0;
8352 }
8353
8354 bridge_create_mac_nat_entry_ipv6(sc, mac_nat_bif, &sin6.sin6_addr, IF_LLADDR(ifp));
8355 break;
8356 }
8357
8358 default:
8359 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
8360 "ifaddr_address_family unknown %d",
8361 af);
8362 break;
8363 }
8364 }
8365
8366 ifnet_address_list_free_counted_by(list, addresses_count);
8367 return;
8368 }
8369
8370 static void
8371 bridge_mac_nat_age_entries_common(struct bridge_softc *sc,
8372 struct mac_nat_entry_list *list, unsigned long now)
8373 {
8374 struct mac_nat_entry *mne;
8375 struct mac_nat_entry *tmne;
8376
8377 LIST_FOREACH_SAFE(mne, list, mne_list, tmne) {
8378 if (now >= mne->mne_expire) {
8379 bridge_destroy_mac_nat_entry(sc, mne, "aged out");
8380 }
8381 }
8382 }
8383
8384 static void
8385 bridge_mac_nat_age_entries(struct bridge_softc *sc, unsigned long now)
8386 {
8387 if (sc->sc_mac_nat_bif == NULL) {
8388 return;
8389 }
8390 bridge_mac_nat_age_entries_common(sc, &sc->sc_mne_list, now);
8391 bridge_mac_nat_age_entries_common(sc, &sc->sc_mne_list_v6, now);
8392 }
8393
8394 static const char *
8395 get_in_out_string(boolean_t is_output)
8396 {
8397 return (const char * __null_terminated)(is_output ? "OUT" : "IN");
8398 }
8399
8400 /*
8401 * is_valid_arp_packet:
8402 * Verify that this is a valid ARP packet.
8403 *
8404 * Returns TRUE if the packet is valid, FALSE otherwise.
8405 */
8406 static boolean_t
8407 is_valid_arp_packet(mbuf_t *data, bool is_output,
8408 struct ether_header **eh_p, struct ether_arp **ea_p)
8409 {
8410 struct ether_arp *ea;
8411 struct ether_header *eh;
8412 size_t minlen = sizeof(struct ether_header) + sizeof(struct ether_arp);
8413 boolean_t is_valid = FALSE;
8414 int flags = is_output ? BR_DBGF_OUTPUT : BR_DBGF_INPUT;
8415
8416 if (mbuf_pkthdr_len(*data) < minlen) {
8417 BRIDGE_LOG(LOG_DEBUG, flags,
8418 "ARP %s short frame %lu < %lu",
8419 get_in_out_string(is_output),
8420 mbuf_pkthdr_len(*data), minlen);
8421 goto done;
8422 }
8423 if (mbuf_len(*data) < minlen && mbuf_pullup(data, minlen) != 0) {
8424 BRIDGE_LOG(LOG_DEBUG, flags,
8425 "ARP %s size %lu mbuf_pullup fail",
8426 get_in_out_string(is_output),
8427 minlen);
8428 *data = NULL;
8429 goto done;
8430 }
8431
8432 /* validate ARP packet */
8433 eh = mtod(*data, struct ether_header *);
8434 ea = (struct ether_arp *)(eh + 1);
8435 if (ea->arp_hrd != HTONS_ARPHRD_ETHER) {
8436 BRIDGE_LOG(LOG_DEBUG, flags,
8437 "ARP %s htype not ethernet",
8438 get_in_out_string(is_output));
8439 goto done;
8440 }
8441 if (ea->arp_hln != ETHER_ADDR_LEN) {
8442 BRIDGE_LOG(LOG_DEBUG, flags,
8443 "ARP %s hlen not ethernet",
8444 get_in_out_string(is_output));
8445 goto done;
8446 }
8447 if (ea->arp_pro != HTONS_ETHERTYPE_IP) {
8448 BRIDGE_LOG(LOG_DEBUG, flags,
8449 "ARP %s ptype not IP",
8450 get_in_out_string(is_output));
8451 goto done;
8452 }
8453 if (ea->arp_pln != sizeof(struct in_addr)) {
8454 BRIDGE_LOG(LOG_DEBUG, flags,
8455 "ARP %s plen not IP",
8456 get_in_out_string(is_output));
8457 goto done;
8458 }
8459 is_valid = TRUE;
8460 *ea_p = ea;
8461 *eh_p = eh;
8462 done:
8463 return is_valid;
8464 }
8465
8466 static struct mac_nat_entry *
8467 bridge_mac_nat_arp_input(struct bridge_softc *sc, mbuf_t *data)
8468 {
8469 struct ether_arp * __single ea;
8470 struct ether_header * __single eh;
8471 struct mac_nat_entry *mne = NULL;
8472 u_short op;
8473 struct in_addr tpa;
8474
8475 if (!is_valid_arp_packet(data, FALSE, &eh, &ea)) {
8476 goto done;
8477 }
8478 op = ea->arp_op;
8479 switch (op) {
8480 case HTONS_ARPOP_REQUEST:
8481 case HTONS_ARPOP_REPLY:
8482 /* only care about REQUEST and REPLY */
8483 break;
8484 default:
8485 goto done;
8486 }
8487
8488 /* check the target IP address for a NAT entry */
8489 bcopy(ea->arp_tpa, &tpa, sizeof(tpa));
8490 if (tpa.s_addr != 0) {
8491 mne = bridge_lookup_mac_nat_entry_ipv4(sc, &tpa);
8492 }
8493 if (mne != NULL) {
8494 if (op == HTONS_ARPOP_REPLY) {
8495 /* translate the MAC address */
8496 if (BRIDGE_DBGF_ENABLED(BR_DBGF_MAC_NAT)) {
8497 char mac_src[24];
8498 char mac_dst[24];
8499
8500 ether_ntop(mac_src, sizeof(mac_src),
8501 ea->arp_tha);
8502 ether_ntop(mac_dst, sizeof(mac_dst),
8503 mne->mne_mac);
8504 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8505 "%s %s ARP %s -> %s",
8506 sc->sc_if_xname,
8507 mne->mne_bif->bif_ifp->if_xname,
8508 mac_src, mac_dst);
8509 }
8510 bcopy(mne->mne_mac, ea->arp_tha, sizeof(ea->arp_tha));
8511 }
8512 } else {
8513 /* handle conflicting ARP (sender matches mne) */
8514 struct in_addr spa;
8515
8516 bcopy(ea->arp_spa, &spa, sizeof(spa));
8517 if (spa.s_addr != 0 && spa.s_addr != tpa.s_addr) {
8518 /* check the source IP for a NAT entry */
8519 mne = bridge_lookup_mac_nat_entry_ipv4(sc, &spa);
8520 }
8521 }
8522
8523 done:
8524 return mne;
8525 }
8526
8527 static boolean_t
8528 bridge_mac_nat_arp_output(struct bridge_softc *sc,
8529 struct bridge_iflist *bif, mbuf_t *data, struct mac_nat_record *mnr)
8530 {
8531 struct ether_arp * __single ea;
8532 struct ether_header * __single eh;
8533 struct in_addr ip;
8534 struct mac_nat_entry *mne = NULL;
8535 u_short op;
8536 boolean_t translate = FALSE;
8537
8538 if (!is_valid_arp_packet(data, TRUE, &eh, &ea)) {
8539 goto done;
8540 }
8541 op = ea->arp_op;
8542 switch (op) {
8543 case HTONS_ARPOP_REQUEST:
8544 case HTONS_ARPOP_REPLY:
8545 /* only care about REQUEST and REPLY */
8546 break;
8547 default:
8548 goto done;
8549 }
8550
8551 bcopy(ea->arp_spa, &ip, sizeof(ip));
8552 if (ip.s_addr == 0) {
8553 goto done;
8554 }
8555 /* XXX validate IP address: no multicast/broadcast */
8556 mne = bridge_update_mac_nat_entry_ipv4(sc, bif, &ip,
8557 (const char *)ea->arp_sha);
8558 if (mnr != NULL && mne != NULL) {
8559 /* record the offset to do the replacement */
8560 translate = TRUE;
8561 mnr->mnr_arp_offset = (char *)ea->arp_sha - (char *)eh;
8562 }
8563
8564 done:
8565 return translate;
8566 }
8567
8568 #define ETHER_IPV4_HEADER_LEN (sizeof(struct ether_header) + \
8569 + sizeof(struct ip))
8570 static uint8_t * __indexable
8571 get_ether_ip_header_ptr(mbuf_t *data, boolean_t is_output)
8572 {
8573 uint8_t *header = NULL;
8574 int flags = is_output ? BR_DBGF_OUTPUT : BR_DBGF_INPUT;
8575 size_t minlen = ETHER_IPV4_HEADER_LEN;
8576
8577 if (mbuf_pkthdr_len(*data) < minlen) {
8578 BRIDGE_LOG(LOG_DEBUG, flags,
8579 "IP %s short frame %lu < %lu",
8580 get_in_out_string(is_output),
8581 mbuf_pkthdr_len(*data), minlen);
8582 goto done;
8583 }
8584 if (mbuf_len(*data) < minlen && mbuf_pullup(data, minlen) != 0) {
8585 BRIDGE_LOG(LOG_DEBUG, flags,
8586 "IP %s size %lu mbuf_pullup fail",
8587 get_in_out_string(is_output),
8588 minlen);
8589 *data = NULL;
8590 goto done;
8591 }
8592 header = mtod(*data, uint8_t *);
8593 done:
8594 return header;
8595 }
8596
8597 static bool
8598 is_broadcast_ip_packet(mbuf_t *data)
8599 {
8600 uint8_t *header;
8601 struct ether_header *eh;
8602 bool is_broadcast = FALSE;
8603
8604 eh = mtod(*data, struct ether_header *);
8605 switch (eh->ether_type) {
8606 case HTONS_ETHERTYPE_IP:
8607 header = get_ether_ip_header_ptr(data, FALSE);
8608 if (header != NULL) {
8609 struct in_addr dst;
8610 struct ip *iphdr;
8611
8612 iphdr = (struct ip *)(header + sizeof(struct ether_header));
8613 bcopy(&iphdr->ip_dst, &dst, sizeof(dst));
8614 is_broadcast = (dst.s_addr == INADDR_BROADCAST);
8615 }
8616 break;
8617 default:
8618 break;
8619 }
8620 return is_broadcast;
8621 }
8622
8623 static struct mac_nat_entry *
8624 bridge_mac_nat_ip_input(struct bridge_softc *sc, mbuf_t *data)
8625 {
8626 struct in_addr dst;
8627 uint8_t *header;
8628 struct ip *iphdr;
8629 struct mac_nat_entry *mne = NULL;
8630
8631 header = get_ether_ip_header_ptr(data, FALSE);
8632 if (header == NULL) {
8633 goto done;
8634 }
8635 iphdr = (struct ip *)(void *)(header + sizeof(struct ether_header));
8636 bcopy(&iphdr->ip_dst, &dst, sizeof(dst));
8637 /* XXX validate IP address */
8638 if (dst.s_addr == 0) {
8639 goto done;
8640 }
8641 mne = bridge_lookup_mac_nat_entry_ipv4(sc, &dst);
8642 done:
8643 return mne;
8644 }
8645
8646 static void
8647 bridge_mac_nat_udp_output(struct bridge_softc *sc,
8648 struct bridge_iflist *bif, mbuf_t m,
8649 uint8_t ip_header_len, struct mac_nat_record *mnr)
8650 {
8651 uint16_t dp_flags;
8652 errno_t error;
8653 size_t offset;
8654 struct udphdr udphdr;
8655
8656 /* copy the UDP header */
8657 offset = sizeof(struct ether_header) + ip_header_len;
8658 error = mbuf_copydata(m, offset, sizeof(struct udphdr), &udphdr);
8659 if (error != 0) {
8660 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8661 "mbuf_copydata udphdr failed %d",
8662 error);
8663 return;
8664 }
8665 if (udphdr.uh_sport != HTONS_IPPORT_BOOTPC ||
8666 udphdr.uh_dport != HTONS_IPPORT_BOOTPS) {
8667 /* not a BOOTP/DHCP packet */
8668 return;
8669 }
8670 /* check whether the broadcast bit is already set */
8671 offset += sizeof(struct udphdr) + offsetof(struct dhcp, dp_flags);
8672 error = mbuf_copydata(m, offset, sizeof(dp_flags), &dp_flags);
8673 if (error != 0) {
8674 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8675 "mbuf_copydata dp_flags failed %d",
8676 error);
8677 return;
8678 }
8679 if ((dp_flags & HTONS_DHCP_FLAGS_BROADCAST) != 0) {
8680 /* it's already set, nothing to do */
8681 return;
8682 }
8683 /* broadcast bit needs to be set */
8684 mnr->mnr_ip_dhcp_flags = dp_flags | htons(DHCP_FLAGS_BROADCAST);
8685 mnr->mnr_ip_header_len = ip_header_len;
8686 if (udphdr.uh_sum != 0) {
8687 uint16_t delta;
8688
8689 /* adjust checksum to take modified dp_flags into account */
8690 delta = dp_flags - mnr->mnr_ip_dhcp_flags;
8691 mnr->mnr_ip_udp_csum = udphdr.uh_sum + delta;
8692 }
8693 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8694 "%s %s DHCP dp_flags 0x%x UDP cksum 0x%x",
8695 sc->sc_if_xname,
8696 bif->bif_ifp->if_xname,
8697 ntohs(mnr->mnr_ip_dhcp_flags),
8698 ntohs(mnr->mnr_ip_udp_csum));
8699 return;
8700 }
8701
8702 static boolean_t
8703 bridge_mac_nat_ip_output(struct bridge_softc *sc,
8704 struct bridge_iflist *bif, mbuf_t *data, struct mac_nat_record *mnr)
8705 {
8706 #pragma unused(mnr)
8707 uint8_t *header;
8708 struct ether_header *eh;
8709 struct in_addr ip;
8710 struct ip *iphdr;
8711 uint8_t ip_header_len;
8712 struct mac_nat_entry *mne = NULL;
8713 boolean_t translate = FALSE;
8714
8715 header = get_ether_ip_header_ptr(data, TRUE);
8716 if (header == NULL) {
8717 goto done;
8718 }
8719
8720 eh = (struct ether_header *)header;
8721 iphdr = (struct ip *)(header + sizeof(*eh));
8722 ip_header_len = IP_VHL_HL(iphdr->ip_vhl) << 2;
8723 if (ip_header_len < sizeof(ip)) {
8724 /* bogus IP header */
8725 goto done;
8726 }
8727 bcopy(&iphdr->ip_src, &ip, sizeof(ip));
8728 /* XXX validate the source address */
8729 if (ip.s_addr != 0) {
8730 mne = bridge_update_mac_nat_entry_ipv4(sc, bif, &ip,
8731 (const char *)eh->ether_shost);
8732 }
8733 if (mnr != NULL) {
8734 if (ip.s_addr == 0 && iphdr->ip_p == IPPROTO_UDP) {
8735 /* handle DHCP must broadcast */
8736 bridge_mac_nat_udp_output(sc, bif, *data,
8737 ip_header_len, mnr);
8738 }
8739 translate = TRUE;
8740 }
8741 done:
8742 return translate;
8743 }
8744
8745 #define ETHER_IPV6_HEADER_LEN (sizeof(struct ether_header) + \
8746 + sizeof(struct ip6_hdr))
8747 static uint8_t * __indexable
8748 get_ether_ipv6_header_ptr(mbuf_t *data, size_t plen, boolean_t is_output)
8749 {
8750 uint8_t *header = NULL;
8751 int flags = is_output ? BR_DBGF_OUTPUT : BR_DBGF_INPUT;
8752 size_t minlen = ETHER_IPV6_HEADER_LEN + plen;
8753
8754 if (mbuf_pkthdr_len(*data) < minlen) {
8755 BRIDGE_LOG(LOG_DEBUG, flags,
8756 "IP %s short frame %lu < %lu",
8757 get_in_out_string(is_output),
8758 mbuf_pkthdr_len(*data), minlen);
8759 goto done;
8760 }
8761 if (mbuf_len(*data) < minlen && mbuf_pullup(data, minlen) != 0) {
8762 BRIDGE_LOG(LOG_DEBUG, flags,
8763 "IP %s size %lu mbuf_pullup fail",
8764 get_in_out_string(is_output),
8765 minlen);
8766 *data = NULL;
8767 goto done;
8768 }
8769 header = mtod(*data, uint8_t *);
8770 done:
8771 return header;
8772 }
8773
8774 #include <netinet/icmp6.h>
8775 #include <netinet6/nd6.h>
8776
8777 #define ETHER_ND_LLADDR_LEN (ETHER_ADDR_LEN + sizeof(struct nd_opt_hdr))
8778
8779 static void
8780 bridge_mac_nat_icmpv6_output(struct bridge_softc *sc,
8781 struct bridge_iflist *bif,
8782 mbuf_t *data, struct ip6_hdr *ip6h,
8783 struct in6_addr *saddrp,
8784 struct mac_nat_record *mnr)
8785 {
8786 uint8_t *header;
8787 struct ether_header *eh;
8788 struct icmp6_hdr *icmp6;
8789 uint8_t icmp6_type;
8790 uint32_t icmp6len;
8791 int lladdrlen = 0;
8792 char *lladdr = NULL;
8793 unsigned int off = sizeof(*ip6h);
8794
8795 icmp6len = (u_int32_t)ntohs(ip6h->ip6_plen);
8796 if (icmp6len < sizeof(*icmp6)) {
8797 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
8798 "short IPv6 payload length %d < %lu",
8799 icmp6len, sizeof(*icmp6));
8800 return;
8801 }
8802
8803 /* pullup IP6 header + ICMPv6 header */
8804 header = get_ether_ipv6_header_ptr(data, sizeof(*icmp6), TRUE);
8805 if (header == NULL) {
8806 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
8807 "failed to pullup icmp6 header");
8808 return;
8809 }
8810 eh = (struct ether_header *)header;
8811 ip6h = (struct ip6_hdr *)(header + sizeof(*eh));
8812 icmp6 = (struct icmp6_hdr *)(header + sizeof(*eh) + off);
8813 icmp6_type = icmp6->icmp6_type;
8814 switch (icmp6_type) {
8815 case ND_NEIGHBOR_SOLICIT:
8816 case ND_NEIGHBOR_ADVERT:
8817 case ND_ROUTER_ADVERT:
8818 case ND_ROUTER_SOLICIT:
8819 break;
8820 default:
8821 return;
8822 }
8823
8824 /* pullup IP6 header + payload */
8825 header = get_ether_ipv6_header_ptr(data, icmp6len, TRUE);
8826 if (header == NULL) {
8827 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
8828 "failed to pullup icmp6 + payload");
8829 return;
8830 }
8831 eh = (struct ether_header *)header;
8832 ip6h = (struct ip6_hdr *)(header + sizeof(*eh));
8833 icmp6 = (struct icmp6_hdr *)(header + sizeof(*eh) + off);
8834
8835 switch (icmp6_type) {
8836 case ND_NEIGHBOR_SOLICIT: {
8837 struct nd_neighbor_solicit *nd_ns;
8838 union nd_opts ndopts;
8839 boolean_t is_dad_probe;
8840 struct in6_addr taddr;
8841
8842 if (icmp6len < sizeof(*nd_ns)) {
8843 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8844 "short nd_ns %d < %lu",
8845 icmp6len, sizeof(*nd_ns));
8846 return;
8847 }
8848
8849 nd_ns = (struct nd_neighbor_solicit *)(void *)icmp6;
8850 bcopy(&nd_ns->nd_ns_target, &taddr, sizeof(taddr));
8851 if (IN6_IS_ADDR_MULTICAST(&taddr) ||
8852 IN6_IS_ADDR_UNSPECIFIED(&taddr)) {
8853 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8854 "invalid target ignored");
8855 return;
8856 }
8857
8858 /* parse options */
8859 nd6_option_init(nd_ns + 1, icmp6len - sizeof(*nd_ns), &ndopts);
8860 if (nd6_options(&ndopts) < 0) {
8861 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8862 "invalid ND6 NS option");
8863 return;
8864 }
8865 if (ndopts.nd_opts_src_lladdr != NULL) {
8866 ND_OPT_LLADDR(ndopts.nd_opts_src_lladdr, nd_opt_len,
8867 lladdr, lladdrlen);
8868 }
8869 is_dad_probe = IN6_IS_ADDR_UNSPECIFIED(saddrp);
8870 if (lladdr != NULL) {
8871 if (is_dad_probe) {
8872 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8873 "bad ND6 DAD packet");
8874 return;
8875 }
8876 if (lladdrlen != ETHER_ND_LLADDR_LEN) {
8877 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8878 "source lladdrlen %d != %lu",
8879 lladdrlen, ETHER_ND_LLADDR_LEN);
8880 return;
8881 }
8882 }
8883 if (is_dad_probe) {
8884 /* node is trying use taddr, create an mne for taddr */
8885 *saddrp = taddr;
8886 }
8887 break;
8888 }
8889 case ND_NEIGHBOR_ADVERT: {
8890 struct nd_neighbor_advert *nd_na;
8891 union nd_opts ndopts;
8892 struct in6_addr taddr;
8893
8894
8895 nd_na = (struct nd_neighbor_advert *)(void *)icmp6;
8896
8897 if (icmp6len < sizeof(*nd_na)) {
8898 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8899 "short nd_na %d < %lu",
8900 icmp6len, sizeof(*nd_na));
8901 return;
8902 }
8903
8904 bcopy(&nd_na->nd_na_target, &taddr, sizeof(taddr));
8905 if (IN6_IS_ADDR_MULTICAST(&taddr) ||
8906 IN6_IS_ADDR_UNSPECIFIED(&taddr)) {
8907 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8908 "invalid target ignored");
8909 return;
8910 }
8911
8912 /* parse options */
8913 nd6_option_init(nd_na + 1, icmp6len - sizeof(*nd_na), &ndopts);
8914 if (nd6_options(&ndopts) < 0) {
8915 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8916 "invalid ND6 NA option");
8917 return;
8918 }
8919 if (ndopts.nd_opts_tgt_lladdr == NULL) {
8920 /* target linklayer, nothing to do */
8921 return;
8922 }
8923
8924 ND_OPT_LLADDR(ndopts.nd_opts_tgt_lladdr, nd_opt_len, lladdr, lladdrlen);
8925 if (lladdrlen != ETHER_ND_LLADDR_LEN) {
8926 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8927 "target lladdrlen %d != %lu",
8928 lladdrlen, ETHER_ND_LLADDR_LEN);
8929 return;
8930 }
8931 break;
8932 }
8933 case ND_ROUTER_ADVERT:
8934 case ND_ROUTER_SOLICIT: {
8935 union nd_opts ndopts;
8936 uint32_t type_length;
8937 const char *description;
8938
8939 if (icmp6_type == ND_ROUTER_ADVERT) {
8940 type_length = sizeof(struct nd_router_advert);
8941 description = "RA";
8942 } else {
8943 type_length = sizeof(struct nd_router_solicit);
8944 description = "RS";
8945 }
8946 if (icmp6len < type_length) {
8947 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8948 "short ND6 %s %d < %d",
8949 description, icmp6len, type_length);
8950 return;
8951 }
8952
8953 /* parse options */
8954 nd6_option_init(((uint8_t *)icmp6) + type_length,
8955 icmp6len - type_length, &ndopts);
8956 if (nd6_options(&ndopts) < 0) {
8957 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8958 "invalid ND6 %s option", description);
8959 return;
8960 }
8961 if (ndopts.nd_opts_src_lladdr != NULL) {
8962 ND_OPT_LLADDR(ndopts.nd_opts_src_lladdr, nd_opt_len, lladdr, lladdrlen);
8963
8964 if (lladdrlen != ETHER_ND_LLADDR_LEN) {
8965 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8966 "source lladdrlen %d != %lu",
8967 lladdrlen, ETHER_ND_LLADDR_LEN);
8968 return;
8969 }
8970 }
8971 break;
8972 }
8973 default:
8974 break;
8975 }
8976
8977 if (lladdr != NULL) {
8978 mnr->mnr_ip6_lladdr_offset = (uint16_t)
8979 ((uintptr_t)lladdr - (uintptr_t)eh);
8980 mnr->mnr_ip6_icmp6_len = icmp6len;
8981 mnr->mnr_ip6_icmp6_type = icmp6_type;
8982 mnr->mnr_ip6_header_len = off;
8983 if (BRIDGE_DBGF_ENABLED(BR_DBGF_MAC_NAT)) {
8984 const char *str;
8985
8986 switch (mnr->mnr_ip6_icmp6_type) {
8987 case ND_ROUTER_ADVERT:
8988 str = "ROUTER ADVERT";
8989 break;
8990 case ND_ROUTER_SOLICIT:
8991 str = "ROUTER SOLICIT";
8992 break;
8993 case ND_NEIGHBOR_ADVERT:
8994 str = "NEIGHBOR ADVERT";
8995 break;
8996 case ND_NEIGHBOR_SOLICIT:
8997 str = "NEIGHBOR SOLICIT";
8998 break;
8999 default:
9000 str = "";
9001 break;
9002 }
9003 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
9004 "%s %s %s ip6len %d icmp6len %d lladdr offset %d",
9005 sc->sc_if_xname, bif->bif_ifp->if_xname, str,
9006 mnr->mnr_ip6_header_len,
9007 mnr->mnr_ip6_icmp6_len, mnr->mnr_ip6_lladdr_offset);
9008 }
9009 }
9010 }
9011
9012 static struct mac_nat_entry *
9013 bridge_mac_nat_ipv6_input(struct bridge_softc *sc, mbuf_t *data)
9014 {
9015 struct in6_addr dst;
9016 uint8_t *header;
9017 struct ether_header *eh;
9018 struct ip6_hdr *ip6h;
9019 struct mac_nat_entry *mne = NULL;
9020
9021 header = get_ether_ipv6_header_ptr(data, 0, FALSE);
9022 if (header == NULL) {
9023 goto done;
9024 }
9025 eh = (struct ether_header *)header;
9026 ip6h = (struct ip6_hdr *)(header + sizeof(*eh));
9027 bcopy(&ip6h->ip6_dst, &dst, sizeof(dst));
9028 /* XXX validate IPv6 address */
9029 if (IN6_IS_ADDR_UNSPECIFIED(&dst)) {
9030 goto done;
9031 }
9032 mne = bridge_lookup_mac_nat_entry_ipv6(sc, &dst);
9033
9034 done:
9035 return mne;
9036 }
9037
9038 static boolean_t
9039 bridge_mac_nat_ipv6_output(struct bridge_softc *sc,
9040 struct bridge_iflist *bif, mbuf_t *data, struct mac_nat_record *mnr)
9041 {
9042 uint8_t *header;
9043 struct ether_header *eh;
9044 ether_addr_t ether_shost;
9045 struct ip6_hdr *ip6h;
9046 struct in6_addr saddr;
9047 boolean_t translate;
9048
9049 translate = (bif == sc->sc_mac_nat_bif) ? FALSE : TRUE;
9050 header = get_ether_ipv6_header_ptr(data, 0, TRUE);
9051 if (header == NULL) {
9052 translate = FALSE;
9053 goto done;
9054 }
9055 eh = (struct ether_header *)header;
9056 bcopy(eh->ether_shost, ðer_shost, sizeof(ether_shost));
9057 ip6h = (struct ip6_hdr *)(header + sizeof(*eh));
9058 bcopy(&ip6h->ip6_src, &saddr, sizeof(saddr));
9059 if (mnr != NULL && ip6h->ip6_nxt == IPPROTO_ICMPV6) {
9060 bridge_mac_nat_icmpv6_output(sc, bif, data, ip6h, &saddr, mnr);
9061 }
9062 if (IN6_IS_ADDR_UNSPECIFIED(&saddr)) {
9063 goto done;
9064 }
9065 (void)bridge_update_mac_nat_entry_ipv6(sc, bif, &saddr,
9066 (const char *)ether_shost.octet);
9067
9068 done:
9069 return translate;
9070 }
9071
9072 /*
9073 * Function: bridge_mac_nat_input:
9074 *
9075 * Purpose:
9076 * Process a unicast packet arriving on the external interface `external_ifp`.
9077 *
9078 * If the packet is ARP, IPv4, or IPv6, lookup the address from the packet in
9079 * the mac_nat_entry table. If an entry is found, and the interface is
9080 * not `external_ifp`, replace the destination MAC address in the
9081 * ethernet header with the corresponding internal MAC address, and return
9082 * the interface via `*dst_if`.
9083 *
9084 * Returns:
9085 * NULL if the packet was deallocated during processing.
9086 *
9087 * Otherwise, returns non-NULL packet that should:
9088 * 1) if `*dst_if` is NULL, continue on as an input packet
9089 * over `external_ifp`, OR
9090 * 2) if `*dst_if` is not NULL, be delivered as an output packet
9091 * over `*dst_if`.
9092 */
9093 static mbuf_t
9094 bridge_mac_nat_input(struct bridge_softc *sc, ifnet_t external_ifp,
9095 mbuf_t m, ifnet_t * dst_if)
9096 {
9097 struct ether_header *eh;
9098 mbuf_t m0 = m;
9099 struct mac_nat_entry *mne = NULL;
9100
9101 BRIDGE_LOCK_ASSERT_HELD(sc);
9102 *dst_if = NULL;
9103 eh = mtod(m, struct ether_header *);
9104 switch (eh->ether_type) {
9105 case HTONS_ETHERTYPE_ARP:
9106 mne = bridge_mac_nat_arp_input(sc, &m);
9107 break;
9108 case HTONS_ETHERTYPE_IP:
9109 mne = bridge_mac_nat_ip_input(sc, &m);
9110 break;
9111 case HTONS_ETHERTYPE_IPV6:
9112 mne = bridge_mac_nat_ipv6_input(sc, &m);
9113 break;
9114 default:
9115 break;
9116 }
9117 if (m != NULL & mne != NULL) {
9118 *dst_if = mne->mne_bif->bif_ifp;
9119 if (*dst_if == external_ifp) {
9120 /* receive packet for ifp */
9121 *dst_if = NULL;
9122 } else {
9123 /* replace the destination MAC with internal one */
9124 if (m != m0) {
9125 /* it may have changed */
9126 eh = mtod(m, struct ether_header *);
9127 }
9128 bcopy(mne->mne_mac, eh->ether_dhost,
9129 sizeof(eh->ether_dhost));
9130 }
9131 }
9132 return m;
9133 }
9134
9135
9136 static mblist
9137 bridge_mac_nat_input_list(struct bridge_softc *sc, ifnet_t external_ifp,
9138 mbuf_t m, mbuf_t * forward_head)
9139 {
9140 mblist forward;
9141 mbuf_t next_packet;
9142 mblist ret;
9143
9144 mblist_init(&ret);
9145 mblist_init(&forward);
9146 for (mbuf_t scan = m; scan != NULL; scan = next_packet) {
9147 ifnet_ref_t dst_if;
9148
9149 /* take packet out of the list */
9150 next_packet = scan->m_nextpkt;
9151 scan->m_nextpkt = NULL;
9152
9153 scan = bridge_mac_nat_input(sc, external_ifp, scan, &dst_if);
9154 if (scan != NULL) {
9155 if (dst_if != NULL) {
9156 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
9157 "%s MAC-NAT input translate to %s",
9158 sc->sc_if_xname, dst_if->if_xname);
9159 /* use rcvif to store the egress interface */
9160 mbuf_pkthdr_setrcvif(scan, dst_if);
9161 /* add it to the forwarding list */
9162 mblist_append(&forward, scan);
9163 } else {
9164 /* add it to the "continue on as input" list */
9165 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
9166 "%s MAC-NAT input for %s",
9167 sc->sc_if_xname,
9168 external_ifp->if_xname);
9169 mblist_append(&ret, scan);
9170 }
9171 }
9172 }
9173 *forward_head = forward.head;
9174 return ret;
9175 }
9176
9177 /*
9178 * bridge_mac_nat_translate_list:
9179 * Process a list of packets destined to the MAC-NAT interface `dst_if`
9180 * from the bridge member `sbif`.
9181 *
9182 * For each packet in the list, update the MAC-NAT record, and if
9183 * translation is required, translate it.
9184 *
9185 * Returns the list of packets that should be delivered to the MAC-NAT
9186 * interface.
9187 */
9188 static mbuf_t
9189 bridge_mac_nat_translate_list(struct bridge_softc * sc,
9190 struct bridge_iflist *sbif, ifnet_t dst_if, mbuf_t m)
9191 {
9192 mbuf_t next_packet;
9193 mblist ret;
9194
9195 mblist_init(&ret);
9196 for (mbuf_ref_t scan = m; scan != NULL; scan = next_packet) {
9197 struct mac_nat_record mnr;
9198 bool translate_mac;
9199
9200 /* take packet out of the list */
9201 next_packet = scan->m_nextpkt;
9202 scan->m_nextpkt = NULL;
9203 translate_mac = bridge_mac_nat_output(sc, sbif, &scan, &mnr);
9204 if (scan != NULL) {
9205 if (translate_mac) {
9206 bridge_mac_nat_translate(&scan, &mnr,
9207 IF_LLADDR(dst_if));
9208 }
9209 if (scan != NULL) {
9210 /* add it back to the list */
9211 mblist_append(&ret, scan);
9212 }
9213 }
9214 }
9215 return ret.head;
9216 }
9217
9218 /*
9219 * bridge_mac_nat_copy_and_translate_list:
9220 * Same as bridge_mac_nat_translate_list() except that a copy of the
9221 * packet list is returned instead.
9222 *
9223 * The packet list `m` is left unaltered.
9224 */
9225 static mbuf_t
9226 bridge_mac_nat_copy_and_translate_list(struct bridge_softc * sc,
9227 struct bridge_iflist *sbif, ifnet_t dst_if, mbuf_t m)
9228 {
9229 mbuf_t next_packet;
9230 mblist ret;
9231
9232 mblist_init(&ret);
9233 for (mbuf_t scan = m; scan != NULL; scan = next_packet) {
9234 mbuf_ref_t mc = NULL;
9235 struct mac_nat_record mnr;
9236 bool translate_mac;
9237
9238 /* take packet out of the list, make a copy, put it back */
9239 next_packet = scan->m_nextpkt;
9240 scan->m_nextpkt = NULL;
9241 mc = m_dup(scan, M_DONTWAIT);
9242 scan->m_nextpkt = next_packet;
9243 if (mc == NULL) {
9244 continue;
9245 }
9246 translate_mac = bridge_mac_nat_output(sc, sbif, &mc, &mnr);
9247 if (mc != NULL) {
9248 if (translate_mac) {
9249 bridge_mac_nat_translate(&mc, &mnr,
9250 IF_LLADDR(dst_if));
9251 }
9252 if (mc != NULL) {
9253 /* add it to the new list */
9254 mblist_append(&ret, mc);
9255 }
9256 }
9257 }
9258 return ret.head;
9259 }
9260
9261 static void
9262 bridge_mac_nat_forward_list(ifnet_t bridge_ifp, ether_type_flag_t etypef,
9263 mbuf_t m)
9264 {
9265 int count = 0;
9266 ifnet_t dst_if;
9267 mblist list;
9268 int n_lists = 0;
9269 mbuf_t next_packet;
9270
9271 mblist_init(&list);
9272 for (mbuf_t scan = m; scan != NULL; scan = next_packet) {
9273 ifnet_t this_if;
9274
9275 next_packet = scan->m_nextpkt;
9276 this_if = mbuf_pkthdr_rcvif(scan);
9277 mbuf_pkthdr_setrcvif(scan, NULL);
9278 if (list.head == NULL) {
9279 /* start a new list */
9280 list.head = list.tail = scan;
9281 count = 1;
9282 dst_if = this_if;
9283 } else if (dst_if != this_if) {
9284 /* send up the previous chain */
9285 if (list.tail != NULL) {
9286 /* terminate the list */
9287 list.tail->m_nextpkt = NULL;
9288 }
9289 n_lists++;
9290 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
9291 "(%s): sublist %u pkts %u",
9292 dst_if->if_xname, n_lists, count);
9293 bridge_enqueue_multi(bridge_ifp, NULL,
9294 dst_if, etypef, list.head,
9295 CHECKSUM_OPERATION_CLEAR_OFFLOAD);
9296
9297 /* start new list */
9298 list.head = list.tail = scan;
9299 count = 1;
9300 dst_if = this_if;
9301 } else {
9302 count++;
9303 list.tail = scan;
9304 }
9305 if (next_packet == NULL) {
9306 /* last list */
9307 n_lists++;
9308 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
9309 "(%s): sublist %u pkts %u",
9310 dst_if->if_xname, n_lists, count);
9311 bridge_enqueue_multi(bridge_ifp, NULL,
9312 dst_if, etypef, list.head,
9313 CHECKSUM_OPERATION_CLEAR_OFFLOAD);
9314 }
9315 }
9316 return;
9317 }
9318
9319 /*
9320 * bridge_mac_nat_output:
9321 * Process a packet destined to the MAC NAT interface (sc_mac_nat_bif)
9322 * from the interface 'bif'.
9323 *
9324 * Create a mac_nat_entry containing the source IP address and MAC address
9325 * from the packet. Populate a mac_nat_record with information detailing
9326 * how to translate the packet. Translation takes place later by calling
9327 * `bridge_mac_nat_translate()`.
9328 *
9329 * If 'bif' == sc_mac_nat_bif, the stack over the MAC NAT
9330 * interface is generating an output packet. No translation is required in this
9331 * case, we just record the IP address used to prevent another bif from
9332 * claiming our IP address.
9333 *
9334 * Returns:
9335 * TRUE if the packet should be translated (*mnr updated as well),
9336 * FALSE otherwise.
9337 *
9338 * *data may be updated to point at a different mbuf chain or NULL if
9339 * the chain was deallocated during processing.
9340 */
9341
9342 static boolean_t
9343 bridge_mac_nat_output(struct bridge_softc *sc,
9344 struct bridge_iflist *bif, mbuf_t *data, struct mac_nat_record *mnr)
9345 {
9346 struct ether_header *eh;
9347 boolean_t translate = FALSE;
9348
9349 BRIDGE_LOCK_ASSERT_HELD(sc);
9350 assert(sc->sc_mac_nat_bif != NULL);
9351
9352 eh = mtod(*data, struct ether_header *);
9353 if (mnr != NULL) {
9354 bzero(mnr, sizeof(*mnr));
9355 mnr->mnr_ether_type = eh->ether_type;
9356 }
9357 switch (eh->ether_type) {
9358 case HTONS_ETHERTYPE_ARP:
9359 translate = bridge_mac_nat_arp_output(sc, bif, data, mnr);
9360 break;
9361 case HTONS_ETHERTYPE_IP:
9362 translate = bridge_mac_nat_ip_output(sc, bif, data, mnr);
9363 break;
9364 case HTONS_ETHERTYPE_IPV6:
9365 translate = bridge_mac_nat_ipv6_output(sc, bif, data, mnr);
9366 break;
9367 default:
9368 break;
9369 }
9370 return translate;
9371 }
9372
9373 static void
9374 bridge_mac_nat_arp_translate(mbuf_t *data, struct mac_nat_record *mnr,
9375 const char eaddr[ETHER_ADDR_LEN])
9376 {
9377 errno_t error;
9378
9379 if (mnr->mnr_arp_offset == 0) {
9380 return;
9381 }
9382 /* replace the source hardware address */
9383 error = mbuf_copyback(*data, mnr->mnr_arp_offset,
9384 ETHER_ADDR_LEN, eaddr,
9385 MBUF_DONTWAIT);
9386 if (error != 0) {
9387 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
9388 "mbuf_copyback failed");
9389 m_freem(*data);
9390 *data = NULL;
9391 }
9392 return;
9393 }
9394
9395 static void
9396 bridge_mac_nat_ip_translate(mbuf_t *data, struct mac_nat_record *mnr)
9397 {
9398 errno_t error;
9399 size_t offset;
9400
9401 if (mnr->mnr_ip_header_len == 0) {
9402 return;
9403 }
9404 /* update the UDP checksum */
9405 offset = sizeof(struct ether_header) + mnr->mnr_ip_header_len;
9406 error = mbuf_copyback(*data, offset + offsetof(struct udphdr, uh_sum),
9407 sizeof(mnr->mnr_ip_udp_csum),
9408 &mnr->mnr_ip_udp_csum,
9409 MBUF_DONTWAIT);
9410 if (error != 0) {
9411 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
9412 "mbuf_copyback uh_sum failed");
9413 m_freem(*data);
9414 *data = NULL;
9415 }
9416 /* update the DHCP must broadcast flag */
9417 offset += sizeof(struct udphdr);
9418 error = mbuf_copyback(*data, offset + offsetof(struct dhcp, dp_flags),
9419 sizeof(mnr->mnr_ip_dhcp_flags),
9420 &mnr->mnr_ip_dhcp_flags,
9421 MBUF_DONTWAIT);
9422 if (error != 0) {
9423 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
9424 "mbuf_copyback dp_flags failed");
9425 m_freem(*data);
9426 *data = NULL;
9427 }
9428 }
9429
9430 static void
9431 bridge_mac_nat_ipv6_translate(mbuf_t *data, struct mac_nat_record *mnr,
9432 const char eaddr[ETHER_ADDR_LEN])
9433 {
9434 uint16_t cksum;
9435 errno_t error;
9436 mbuf_t m = *data;
9437
9438 if (mnr->mnr_ip6_header_len == 0) {
9439 return;
9440 }
9441 switch (mnr->mnr_ip6_icmp6_type) {
9442 case ND_ROUTER_ADVERT:
9443 case ND_ROUTER_SOLICIT:
9444 case ND_NEIGHBOR_SOLICIT:
9445 case ND_NEIGHBOR_ADVERT:
9446 if (mnr->mnr_ip6_lladdr_offset == 0) {
9447 /* nothing to do */
9448 return;
9449 }
9450 break;
9451 default:
9452 return;
9453 }
9454
9455 /*
9456 * replace the lladdr
9457 */
9458 error = mbuf_copyback(m, mnr->mnr_ip6_lladdr_offset,
9459 ETHER_ADDR_LEN, eaddr,
9460 MBUF_DONTWAIT);
9461 if (error != 0) {
9462 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
9463 "mbuf_copyback lladdr failed");
9464 m_freem(m);
9465 *data = NULL;
9466 return;
9467 }
9468
9469 /*
9470 * recompute the icmp6 checksum
9471 */
9472
9473 /* skip past the ethernet header */
9474 _mbuf_adjust_pkthdr_and_data(m, ETHER_HDR_LEN);
9475
9476 #define CKSUM_OFFSET_ICMP6 offsetof(struct icmp6_hdr, icmp6_cksum)
9477 /* set the checksum to zero */
9478 cksum = 0;
9479 error = mbuf_copyback(m, mnr->mnr_ip6_header_len + CKSUM_OFFSET_ICMP6,
9480 sizeof(cksum), &cksum, MBUF_DONTWAIT);
9481 if (error != 0) {
9482 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
9483 "mbuf_copyback cksum=0 failed");
9484 m_freem(m);
9485 *data = NULL;
9486 return;
9487 }
9488 /* compute and set the new checksum */
9489 cksum = in6_cksum(m, IPPROTO_ICMPV6, mnr->mnr_ip6_header_len,
9490 mnr->mnr_ip6_icmp6_len);
9491 error = mbuf_copyback(m, mnr->mnr_ip6_header_len + CKSUM_OFFSET_ICMP6,
9492 sizeof(cksum), &cksum, MBUF_DONTWAIT);
9493 if (error != 0) {
9494 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
9495 "mbuf_copyback cksum failed");
9496 m_freem(m);
9497 *data = NULL;
9498 return;
9499 }
9500 /* restore the ethernet header */
9501 _mbuf_adjust_pkthdr_and_data(m, -ETHER_HDR_LEN);
9502 return;
9503 }
9504
9505 static void
9506 bridge_mac_nat_translate(mbuf_t *data, struct mac_nat_record *mnr,
9507 const char eaddr[ETHER_ADDR_LEN])
9508 {
9509 struct ether_header *eh;
9510
9511 /* replace the source ethernet address with the single MAC */
9512 eh = mtod(*data, struct ether_header *);
9513 bcopy(eaddr, eh->ether_shost, sizeof(eh->ether_shost));
9514 switch (mnr->mnr_ether_type) {
9515 case HTONS_ETHERTYPE_ARP:
9516 bridge_mac_nat_arp_translate(data, mnr, eaddr);
9517 break;
9518
9519 case HTONS_ETHERTYPE_IP:
9520 bridge_mac_nat_ip_translate(data, mnr);
9521 break;
9522
9523 case HTONS_ETHERTYPE_IPV6:
9524 bridge_mac_nat_ipv6_translate(data, mnr, eaddr);
9525 break;
9526
9527 default:
9528 break;
9529 }
9530 return;
9531 }
9532
9533 /*
9534 * bridge packet filtering
9535 */
9536
9537 /*
9538 * Perform basic checks on header size since
9539 * pfil assumes ip_input has already processed
9540 * it for it. Cut-and-pasted from ip_input.c.
9541 * Given how simple the IPv6 version is,
9542 * does the IPv4 version really need to be
9543 * this complicated?
9544 *
9545 * XXX Should we update ipstat here, or not?
9546 * XXX Right now we update ipstat but not
9547 * XXX csum_counter.
9548 */
9549 static int
9550 bridge_ip_checkbasic(struct mbuf **mp)
9551 {
9552 struct mbuf *m = *mp;
9553 struct ip *ip;
9554 int len, hlen;
9555 u_short sum;
9556
9557 if (*mp == NULL) {
9558 return -1;
9559 }
9560
9561 if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
9562 /* max_linkhdr is already rounded up to nearest 4-byte */
9563 if ((m = m_copyup(m, sizeof(struct ip),
9564 max_linkhdr)) == NULL) {
9565 /* XXXJRT new stat, please */
9566 ipstat.ips_toosmall++;
9567 goto bad;
9568 }
9569 } else if (OS_EXPECT((size_t)m->m_len < sizeof(struct ip), 0)) {
9570 if ((m = m_pullup(m, sizeof(struct ip))) == NULL) {
9571 ipstat.ips_toosmall++;
9572 goto bad;
9573 }
9574 }
9575 ip = mtod(m, struct ip *);
9576 if (ip == NULL) {
9577 goto bad;
9578 }
9579
9580 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
9581 ipstat.ips_badvers++;
9582 goto bad;
9583 }
9584 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
9585 if (hlen < (int)sizeof(struct ip)) { /* minimum header length */
9586 ipstat.ips_badhlen++;
9587 goto bad;
9588 }
9589 if (hlen > m->m_len) {
9590 if ((m = m_pullup(m, hlen)) == 0) {
9591 ipstat.ips_badhlen++;
9592 goto bad;
9593 }
9594 ip = mtod(m, struct ip *);
9595 if (ip == NULL) {
9596 goto bad;
9597 }
9598 }
9599
9600 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
9601 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
9602 } else {
9603 if (hlen == sizeof(struct ip)) {
9604 sum = in_cksum_hdr(ip);
9605 } else {
9606 sum = in_cksum(m, hlen);
9607 }
9608 }
9609 if (sum) {
9610 ipstat.ips_badsum++;
9611 goto bad;
9612 }
9613
9614 /* Retrieve the packet length. */
9615 len = ntohs(ip->ip_len);
9616
9617 /*
9618 * Check for additional length bogosity
9619 */
9620 if (len < hlen) {
9621 ipstat.ips_badlen++;
9622 goto bad;
9623 }
9624
9625 /*
9626 * Check that the amount of data in the buffers
9627 * is as at least much as the IP header would have us expect.
9628 * Drop packet if shorter than we expect.
9629 */
9630 if (m->m_pkthdr.len < len) {
9631 ipstat.ips_tooshort++;
9632 goto bad;
9633 }
9634
9635 /* Checks out, proceed */
9636 *mp = m;
9637 return 0;
9638
9639 bad:
9640 *mp = m;
9641 return -1;
9642 }
9643
9644 /*
9645 * Same as above, but for IPv6.
9646 * Cut-and-pasted from ip6_input.c.
9647 * XXX Should we update ip6stat, or not?
9648 */
9649 static int
9650 bridge_ip6_checkbasic(struct mbuf **mp)
9651 {
9652 struct mbuf *m = *mp;
9653 struct ip6_hdr *ip6;
9654
9655 /*
9656 * If the IPv6 header is not aligned, slurp it up into a new
9657 * mbuf with space for link headers, in the event we forward
9658 * it. Otherwise, if it is aligned, make sure the entire base
9659 * IPv6 header is in the first mbuf of the chain.
9660 */
9661 if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
9662 struct ifnet *inifp = m->m_pkthdr.rcvif;
9663 /* max_linkhdr is already rounded up to nearest 4-byte */
9664 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
9665 max_linkhdr)) == NULL) {
9666 /* XXXJRT new stat, please */
9667 ip6stat.ip6s_toosmall++;
9668 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
9669 goto bad;
9670 }
9671 } else if (OS_EXPECT((size_t)m->m_len < sizeof(struct ip6_hdr), 0)) {
9672 struct ifnet *inifp = m->m_pkthdr.rcvif;
9673 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
9674 ip6stat.ip6s_toosmall++;
9675 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
9676 goto bad;
9677 }
9678 }
9679
9680 ip6 = mtod(m, struct ip6_hdr *);
9681
9682 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
9683 ip6stat.ip6s_badvers++;
9684 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
9685 goto bad;
9686 }
9687
9688 /* Checks out, proceed */
9689 *mp = m;
9690 return 0;
9691
9692 bad:
9693 *mp = m;
9694 return -1;
9695 }
9696
9697 /*
9698 * the PF routines expect to be called from ip_input, so we
9699 * need to do and undo here some of the same processing.
9700 *
9701 * XXX : this is heavily inspired on bridge_pfil()
9702 */
9703 static int
9704 bridge_pf(struct mbuf **mp, struct ifnet *ifp, uint32_t sc_filter_flags,
9705 bool input)
9706 {
9707 /*
9708 * XXX : mpetit : heavily inspired by bridge_pfil()
9709 */
9710
9711 int snap, error, i, hlen;
9712 struct ether_header *eh1, eh2;
9713 struct ip *ip;
9714 struct llc llc1;
9715 u_int16_t ether_type;
9716
9717 snap = 0;
9718 error = -1; /* Default error if not error == 0 */
9719
9720 if ((sc_filter_flags & IFBF_FILT_MEMBER) == 0) {
9721 return 0; /* filtering is disabled */
9722 }
9723 i = min((*mp)->m_pkthdr.len, max_protohdr);
9724 if ((*mp)->m_len < i) {
9725 *mp = m_pullup(*mp, i);
9726 if (*mp == NULL) {
9727 BRIDGE_LOG(LOG_NOTICE, 0, "m_pullup failed");
9728 return -1;
9729 }
9730 }
9731
9732 eh1 = mtod(*mp, struct ether_header *);
9733 ether_type = ntohs(eh1->ether_type);
9734
9735 /*
9736 * Check for SNAP/LLC.
9737 */
9738 if (ether_type < ETHERMTU) {
9739 struct llc *llc2 = (struct llc *)(eh1 + 1);
9740
9741 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
9742 llc2->llc_dsap == LLC_SNAP_LSAP &&
9743 llc2->llc_ssap == LLC_SNAP_LSAP &&
9744 llc2->llc_control == LLC_UI) {
9745 ether_type = htons(llc2->llc_un.type_snap.ether_type);
9746 snap = 1;
9747 }
9748 }
9749
9750 /*
9751 * If we're trying to filter bridge traffic, don't look at anything
9752 * other than IP and ARP traffic. If the filter doesn't understand
9753 * IPv6, don't allow IPv6 through the bridge either. This is lame
9754 * since if we really wanted, say, an AppleTalk filter, we are hosed,
9755 * but of course we don't have an AppleTalk filter to begin with.
9756 * (Note that since pfil doesn't understand ARP it will pass *ALL*
9757 * ARP traffic.)
9758 */
9759 switch (ether_type) {
9760 case ETHERTYPE_ARP:
9761 case ETHERTYPE_REVARP:
9762 return 0; /* Automatically pass */
9763
9764 case ETHERTYPE_IP:
9765 case ETHERTYPE_IPV6:
9766 break;
9767 default:
9768 /*
9769 * Check to see if the user wants to pass non-ip
9770 * packets, these will not be checked by pf and
9771 * passed unconditionally so the default is to drop.
9772 */
9773 if ((sc_filter_flags & IFBF_FILT_ONLYIP)) {
9774 goto bad;
9775 }
9776 break;
9777 }
9778
9779 /* Strip off the Ethernet header and keep a copy. */
9780 m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t)&eh2);
9781 m_adj(*mp, ETHER_HDR_LEN);
9782
9783 /* Strip off snap header, if present */
9784 if (snap) {
9785 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t)&llc1);
9786 m_adj(*mp, sizeof(struct llc));
9787 }
9788
9789 /*
9790 * Check the IP header for alignment and errors
9791 */
9792 switch (ether_type) {
9793 case ETHERTYPE_IP:
9794 error = bridge_ip_checkbasic(mp);
9795 break;
9796 case ETHERTYPE_IPV6:
9797 error = bridge_ip6_checkbasic(mp);
9798 break;
9799 default:
9800 error = 0;
9801 break;
9802 }
9803 if (error) {
9804 goto bad;
9805 }
9806
9807 error = 0;
9808
9809 /*
9810 * Run the packet through pf rules
9811 */
9812 switch (ether_type) {
9813 case ETHERTYPE_IP:
9814 /*
9815 * before calling the firewall, swap fields the same as
9816 * IP does. here we assume the header is contiguous
9817 */
9818 ip = mtod(*mp, struct ip *);
9819
9820 ip->ip_len = ntohs(ip->ip_len);
9821 ip->ip_off = ntohs(ip->ip_off);
9822
9823 if (ifp != NULL) {
9824 error = pf_af_hook(ifp, 0, mp, AF_INET, input, NULL);
9825 }
9826
9827 if (*mp == NULL || error != 0) { /* filter may consume */
9828 break;
9829 }
9830
9831 /* Recalculate the ip checksum and restore byte ordering */
9832 ip = mtod(*mp, struct ip *);
9833 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
9834 if (hlen < (int)sizeof(struct ip)) {
9835 goto bad;
9836 }
9837 if (hlen > (*mp)->m_len) {
9838 if ((*mp = m_pullup(*mp, hlen)) == 0) {
9839 goto bad;
9840 }
9841 ip = mtod(*mp, struct ip *);
9842 if (ip == NULL) {
9843 goto bad;
9844 }
9845 }
9846 ip->ip_len = htons(ip->ip_len);
9847 ip->ip_off = htons(ip->ip_off);
9848 ip->ip_sum = 0;
9849 if (hlen == sizeof(struct ip)) {
9850 ip->ip_sum = in_cksum_hdr(ip);
9851 } else {
9852 ip->ip_sum = in_cksum(*mp, hlen);
9853 }
9854 break;
9855
9856 case ETHERTYPE_IPV6:
9857 if (ifp != NULL) {
9858 error = pf_af_hook(ifp, 0, mp, AF_INET6, input, NULL);
9859 }
9860
9861 if (*mp == NULL || error != 0) { /* filter may consume */
9862 break;
9863 }
9864 break;
9865 default:
9866 error = 0;
9867 break;
9868 }
9869
9870 if (*mp == NULL) {
9871 return error;
9872 }
9873 if (error != 0) {
9874 goto bad;
9875 }
9876
9877 error = -1;
9878
9879 /*
9880 * Finally, put everything back the way it was and return
9881 */
9882 if (snap) {
9883 M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT, 0);
9884 if (*mp == NULL) {
9885 return error;
9886 }
9887 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
9888 }
9889
9890 M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT, 0);
9891 if (*mp == NULL) {
9892 return error;
9893 }
9894 bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
9895
9896 return 0;
9897
9898 bad:
9899 m_freem(*mp);
9900 *mp = NULL;
9901 return error;
9902 }
9903
9904 #if BRIDGESTP
9905 static void
9906 bridge_bstp_input_list(struct bstp_port *bp, struct mbuf *head)
9907 {
9908 mbuf_t next_packet = NULL;
9909
9910 for (mbuf_t scan = head; scan != NULL; scan = next_packet) {
9911 next_packet = scan->m_nextpkt;
9912 scan->m_nextpkt = NULL;
9913 bstp_input(bp, scan);
9914 }
9915 }
9916 #endif /* BRIDGESTP */
9917
9918 static mblist
9919 bridge_filter_arp_list(struct bridge_iflist * bif, mbuf_t m)
9920 {
9921 mbuf_t next_packet = NULL;
9922 mblist ret;
9923
9924 mblist_init(&ret);
9925 for (mbuf_ref_t scan = m; scan != NULL; scan = next_packet) {
9926 errno_t error;
9927
9928 /* take packet out of the list */
9929 next_packet = scan->m_nextpkt;
9930 scan->m_nextpkt = NULL;
9931 /* filter the ARP packet */
9932 error = bridge_host_filter_arp(bif, &scan);
9933 if (error != 0 && scan != NULL) {
9934 if (BRIDGE_DBGF_ENABLED(BR_DBGF_HOSTFILTER)) {
9935 brlog_mbuf_data(scan, 0,
9936 sizeof(struct ether_header) +
9937 sizeof(struct ip));
9938 }
9939 m_freem(scan);
9940 scan = NULL;
9941 }
9942 if (scan != NULL) {
9943 /* add it to the list */
9944 mblist_append(&ret, scan);
9945 }
9946 }
9947 return ret;
9948 }
9949
9950 static mbuf_t
9951 bridge_filter_checksum(ifnet_t bridge_ifp, struct bridge_iflist * bif, mbuf_t m,
9952 bool is_ipv4, bool host_filter, bool checksum)
9953 {
9954 uint32_t dbgf = 0;
9955 errno_t error;
9956 ip_packet_info info;
9957 u_int mac_hlen = sizeof(struct ether_header);
9958
9959 if (host_filter) {
9960 dbgf |= BR_DBGF_HOSTFILTER;
9961 }
9962 if (checksum) {
9963 dbgf |= BR_DBGF_CHECKSUM;
9964 }
9965
9966 /* get the IP protocol header */
9967 error = bridge_get_ip_proto(&m, mac_hlen, is_ipv4, &info,
9968 &bif->bif_stats.brms_in_ip);
9969 if (error != 0) {
9970 BRIDGE_LOG(LOG_NOTICE, dbgf,
9971 "%s(%s) bridge_get_ip_proto failed %d",
9972 bridge_ifp->if_xname,
9973 bif->bif_ifp->if_xname, error);
9974 goto drop;
9975 }
9976 if (host_filter) {
9977 bool drop = true;
9978
9979 /* restrict IP protocols */
9980 switch (info.ip_proto) {
9981 case IPPROTO_ICMP:
9982 case IPPROTO_IGMP:
9983 drop = !is_ipv4;
9984 break;
9985 case IPPROTO_TCP:
9986 case IPPROTO_UDP:
9987 drop = false;
9988 break;
9989 case IPPROTO_ICMPV6:
9990 drop = is_ipv4;
9991 break;
9992 default:
9993 break;
9994 }
9995 if (drop) {
9996 BRIDGE_HF_DROP(brhf_ip_bad_proto, __func__, __LINE__);
9997 goto drop;
9998 }
9999 bridge_hostfilter_stats.brhf_ip_ok += 1;
10000 }
10001 if (checksum) {
10002 /* need to compute IP/UDP/TCP/checksums */
10003 error = bridge_offload_checksum(&m, &info, &bif->bif_stats);
10004 if (error != 0) {
10005 BRIDGE_LOG(LOG_NOTICE, dbgf,
10006 "%s(%s) bridge_offload_checksum failed %d",
10007 bridge_ifp->if_xname,
10008 bif->bif_ifp->if_xname, error);
10009 goto drop;
10010 }
10011 }
10012 return m;
10013
10014 drop:
10015 /* toss the packet */
10016 if (m != NULL) {
10017 if (host_filter &&
10018 BRIDGE_DBGF_ENABLED(BR_DBGF_HOSTFILTER)) {
10019 brlog_mbuf_data(m, 0,
10020 sizeof(struct ether_header) +
10021 sizeof(struct ip));
10022 }
10023 m_freem(m);
10024 m = NULL;
10025 }
10026 return NULL;
10027 }
10028
10029 static mblist
10030 bridge_filter_checksum_list(ifnet_t bridge_ifp, struct bridge_iflist * bif,
10031 mbuf_t in_list, ether_type_flag_t etypef, bool host_filter, bool checksum)
10032 {
10033 bool is_ipv4 = (etypef == ETHER_TYPE_FLAG_IPV4);
10034 mbuf_t next_packet = NULL;
10035 mblist ret;
10036
10037 mblist_init(&ret);
10038 for (mbuf_t scan = in_list; scan != NULL; scan = next_packet) {
10039 /* take packet out of the list */
10040 next_packet = scan->m_nextpkt;
10041 scan->m_nextpkt = NULL;
10042 scan = bridge_filter_checksum(bridge_ifp, bif,
10043 scan, is_ipv4, host_filter, checksum);
10044 if (scan != NULL) {
10045 /* add packet to the list */
10046 mblist_append(&ret, scan);
10047 }
10048 }
10049 return ret;
10050 }
10051
10052 static mbuf_t
10053 copy_broadcast_packet(mbuf_t m)
10054 {
10055 mbuf_t mc;
10056
10057 /* make a copy of the packet */
10058 mc = m_dup(m, M_DONTWAIT);
10059 if (mc != NULL) {
10060 struct ether_header *eh;
10061
10062 /* make copy look like it is broadcast */
10063 mc->m_flags |= M_BCAST;
10064 eh = mtod(mc, struct ether_header *);
10065 bcopy(etherbroadcastaddr, eh->ether_dhost, ETHER_ADDR_LEN);
10066 }
10067 return mc;
10068 }
10069
10070 static mblist
10071 bridge_find_broadcast_ipv4(mbuf_t in_list, mbuf_t * ip_bcast_head)
10072 {
10073 mblist ip_bcast;
10074 mbuf_t next_packet = NULL;
10075 mblist ret;
10076
10077 mblist_init(&ret);
10078 mblist_init(&ip_bcast);
10079 for (mbuf_ref_t scan = in_list; scan != NULL; scan = next_packet) {
10080 mbuf_t bcast_pkt = NULL;
10081 uint8_t *header;
10082
10083 /* take packet out of the list */
10084 next_packet = scan->m_nextpkt;
10085 scan->m_nextpkt = NULL;
10086
10087 header = get_ether_ip_header_ptr(&scan, FALSE);
10088 if (header != NULL) {
10089 struct in_addr dst;
10090 struct ip *iphdr;
10091
10092 iphdr = (struct ip *)(header + sizeof(struct ether_header));
10093 bcopy(&iphdr->ip_dst, &dst, sizeof(dst));
10094 if (dst.s_addr == INADDR_BROADCAST) {
10095 bcast_pkt = copy_broadcast_packet(scan);
10096 }
10097 }
10098 if (bcast_pkt != NULL) {
10099 /* add packet to broadcast list */
10100 mblist_append(&ip_bcast, bcast_pkt);
10101 }
10102 if (scan != NULL) {
10103 /* add packet back into the list */
10104 mblist_append(&ret, scan);
10105 }
10106 }
10107 *ip_bcast_head = ip_bcast.head;
10108 return ret;
10109 }
10110
10111 static ifnet_t
10112 bridge_find_member(struct bridge_softc * sc, uint8_t * lladdr,
10113 struct bridge_iflist * sbif)
10114 {
10115 struct bridge_iflist * bif;
10116
10117 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) {
10118 if (bif == sbif) {
10119 /* skip the input member */
10120 continue;
10121 }
10122 if (_ether_cmp(IF_LLADDR(bif->bif_ifp), lladdr) == 0) {
10123 return bif->bif_ifp;
10124 }
10125 }
10126 return NULL;
10127 }
10128
10129
10130 /*
10131 * Function: bridge_input_list
10132 *
10133 * Purpose:
10134 * Process a list of input packets through the bridge.
10135 * The caller ensures that all of the packets in the list
10136 * `list_head` .. `list_tail` have the same ethernet header.
10137 *
10138 * Returns:
10139 * Non-NULL head of the chain of packets that were not consumed/freed,
10140 * *tail_p set to the tail of that chain.
10141 *
10142 * NULL if all of the packets were consumed.
10143 */
10144 static mblist
10145 bridge_input_list(struct bridge_softc * sc, ifnet_t ifp,
10146 struct ether_header * eh_in_p, mblist list, bool is_promisc)
10147 {
10148 struct bridge_iflist * bif;
10149 bool bridge_has_address;
10150 ifnet_t bridge_ifp;
10151 bool checksum_offload;
10152 uint8_t * dhost;
10153 #if BRIDGESTP
10154 bool discarding = false;
10155 #endif /* BRIDGESTP */
10156 ifnet_t dst_if = NULL;
10157 errno_t error;
10158 ether_type_flag_t etypef;
10159 bool host_filter;
10160 bool host_filter_drop = false;
10161 mbuf_ref_t ip_bcast = NULL;
10162 bool is_bridge_mac = false;
10163 bool is_broadcast;
10164 bool is_ifp_mac;
10165 ifnet_t member_input = NULL;
10166 uint8_t * shost;
10167 uint16_t vlan;
10168
10169 if (ifp->if_bridge == NULL) {
10170 /* no longer part of bridge */
10171 goto done;
10172 }
10173 bridge_ifp = sc->sc_ifp;
10174 is_broadcast = IS_BCAST_MCAST(list.head);
10175 is_ifp_mac = (!is_broadcast && !is_promisc);
10176 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT,
10177 "%s from %s count %d head 0x%llx.0x%llx tail 0x%llx.0x%llx",
10178 bridge_ifp->if_xname, ifp->if_xname, list.count,
10179 (uint64_t)VM_KERNEL_ADDRPERM(list.head),
10180 (uint64_t)VM_KERNEL_ADDRPERM(mtod(list.head, void *)),
10181 (uint64_t)VM_KERNEL_ADDRPERM(list.tail),
10182 (uint64_t)VM_KERNEL_ADDRPERM(mtod(list.tail, void *)));
10183
10184 /* assume we'll return all packets */
10185 if ((bridge_ifp->if_flags & IFF_RUNNING) == 0) {
10186 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT,
10187 "%s not running passing along",
10188 bridge_ifp->if_xname);
10189 goto done;
10190 }
10191
10192 vlan = VLANTAGOF(m);
10193
10194 /* lookup the bridge member */
10195 BRIDGE_LOCK(sc);
10196 bif = bridge_lookup_member_if(sc, ifp);
10197 if (bif == NULL) {
10198 BRIDGE_UNLOCK(sc);
10199 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT,
10200 "%s bridge_lookup_member_if failed",
10201 bridge_ifp->if_xname);
10202 goto done;
10203 }
10204
10205 /*
10206 * host filter drops packets that:
10207 * - are not ARP, IPv4, or IPv6
10208 * - have incorrect source MAC address
10209 */
10210 host_filter = (bif->bif_flags & BIFF_HOST_FILTER) != 0;
10211 etypef = ether_type_flag_get(eh_in_p->ether_type);
10212 if (host_filter
10213 && (etypef & ETHER_TYPE_FLAG_IP_ARP) == 0) {
10214 /* ether type not one of ARP, IPv4, or IPv6 */
10215 BRIDGE_HF_DROP(brhf_bad_ether_type, __func__, __LINE__);
10216 host_filter_drop = true;
10217 } else if ((bif->bif_flags & BIFF_HF_HWSRC) != 0 &&
10218 bcmp(eh_in_p->ether_shost, bif->bif_hf_hwsrc, ETHER_ADDR_LEN)
10219 != 0) {
10220 /* only allow the single source MAC address */
10221 BRIDGE_HF_DROP(brhf_bad_ether_srchw_addr,
10222 __func__, __LINE__);
10223 host_filter_drop = true;
10224 }
10225 if (host_filter_drop) {
10226 BRIDGE_UNLOCK(sc);
10227 m_freem_list(list.head);
10228 list.head = list.tail = NULL;
10229 goto done;
10230 }
10231
10232 #if BRIDGESTP
10233 discarding = (bif->bif_ifflags & IFBIF_STP) != 0 &&
10234 bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING;
10235 #endif /* BRIDGESTP */
10236
10237 dhost = eh_in_p->ether_dhost;
10238 shost = eh_in_p->ether_shost;
10239 /*
10240 * Reserved multicast address listed in 802.1D section 7.12.6
10241 * must not be forwarded by the bridge.
10242 * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
10243 */
10244 if (is_broadcast) {
10245 if (IS_MCAST(list.head)) {
10246 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MCAST,
10247 " multicast: "
10248 "%02x:%02x:%02x:%02x:%02x:%02x",
10249 dhost[0], dhost[1],
10250 dhost[2], dhost[3],
10251 dhost[4], dhost[5]);
10252 }
10253 if (bcmp(dhost, bstp_etheraddr, (ETHER_ADDR_LEN - 1)) == 0) {
10254 if (dhost[5] == BSTP_ETHERADDR_RANGE_FIRST) {
10255 /* multicast for spanning tree */
10256 #if BRIDGESTP
10257 bridge_bstp_input_list(&bif->bif_stp, list.head);
10258 #else /* BRIDGESTP */
10259 m_freem_list(list.head);
10260 #endif /* BRIDGESTP */
10261 list.head = list.tail = NULL;
10262 BRIDGE_UNLOCK(sc);
10263 goto done;
10264 }
10265 if (dhost[5] <= BSTP_ETHERADDR_RANGE_LAST) {
10266 /* allow packet to continue up the stack */
10267 BRIDGE_UNLOCK(sc);
10268 goto done;
10269 }
10270 }
10271 /* broadcast to all members */
10272 os_atomic_add(&bridge_ifp->if_imcasts, list.count, relaxed);
10273 }
10274
10275 #if BRIDGESTP
10276 if (discarding) {
10277 BRIDGE_UNLOCK(sc);
10278 goto done;
10279 }
10280 #endif /* BRIDGESTP */
10281
10282 /* If the interface is learning, record the address. */
10283 if ((bif->bif_ifflags & IFBIF_LEARNING) != 0) {
10284 error = bridge_rtupdate(sc, shost, vlan, bif, 0, IFBAF_DYNAMIC);
10285 /*
10286 * If the interface has addresses limits then deny any source
10287 * that is not in the cache.
10288 */
10289 if (error != 0 && bif->bif_addrmax) {
10290 BRIDGE_UNLOCK(sc);
10291 goto done;
10292 }
10293 }
10294 #if BRIDGESTP
10295 if ((bif->bif_ifflags & IFBIF_STP) != 0 &&
10296 bif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING) {
10297 BRIDGE_UNLOCK(sc);
10298 goto done;
10299 }
10300 #endif /* BRIDGESTP */
10301
10302 /*
10303 * If the packet is not IP, let the host filter drop ARP packets.
10304 * Otherwise, if the host filter is enabled or we need to compute
10305 * checksums, do that.
10306 * Otherwise, if MAC-NAT is enabled and this is an IPv4 packet,
10307 * check for IPv4 broadcast packets. Accumulate those in a separate
10308 * list `ip_bcast`.
10309 */
10310 checksum_offload = bif_has_checksum_offload(bif);
10311 if (!ether_type_flag_is_ip(etypef)) {
10312 /* host filter process ARP */
10313 if (host_filter) {
10314 /* host filter check earlier means this must be ARP */
10315 VERIFY(etypef == ETHER_TYPE_FLAG_ARP);
10316 list = bridge_filter_arp_list(bif, list.head);
10317 if (list.head == NULL) {
10318 VERIFY(list.tail == NULL);
10319 BRIDGE_UNLOCK(sc);
10320 goto done;
10321 }
10322 }
10323 } else if (host_filter || checksum_offload) {
10324 /* host filter and/or checksum */
10325 list = bridge_filter_checksum_list(bridge_ifp, bif,
10326 list.head, etypef, host_filter, checksum_offload);
10327 if (list.head == NULL) {
10328 VERIFY(list.tail == NULL);
10329 BRIDGE_UNLOCK(sc);
10330 goto done;
10331 }
10332 } else if (is_ifp_mac && bif == sc->sc_mac_nat_bif &&
10333 etypef == ETHER_TYPE_FLAG_IPV4) {
10334 /* look for broadcast IPv4 packet */
10335 list = bridge_find_broadcast_ipv4(list.head, &ip_bcast);
10336 if (list.head == NULL && ip_bcast == NULL) {
10337 /* all packets were consumed */
10338 BRIDGE_UNLOCK(sc);
10339 goto done;
10340 }
10341 }
10342
10343 /*
10344 * If the destination MAC matches the bridge interface, and
10345 * the bridge has an address assigned, or the address is
10346 * distinct from the receive interface, claim the packets
10347 * for the bridge interface.
10348 */
10349 bridge_has_address = (sc->sc_flags & SCF_ADDRESS_ASSIGNED) != 0;
10350 if (!is_broadcast &&
10351 _ether_cmp(dhost, IF_LLADDR(bridge_ifp)) == 0 &&
10352 (bridge_has_address || _ether_cmp(dhost, IF_LLADDR(ifp)) != 0)) {
10353 is_bridge_mac = true;
10354 }
10355 if (is_ifp_mac) {
10356 /* unicast to the interface */
10357 if (sc->sc_mac_nat_bif == bif) {
10358 mbuf_ref_t forward = NULL;
10359
10360 if (list.head != NULL) {
10361 /* handle MAC-NAT if enabled */
10362 list = bridge_mac_nat_input_list(sc, ifp,
10363 list.head, &forward);
10364 }
10365 if (ip_bcast != NULL) {
10366 /* forward to all members except this one */
10367 /* bridge_broadcast_list unlocks */
10368 bridge_broadcast_list(sc, bif, etypef, ip_bcast,
10369 false);
10370 } else {
10371 BRIDGE_UNLOCK(sc);
10372 }
10373 if (forward != NULL) {
10374 bridge_mac_nat_forward_list(bridge_ifp, etypef,
10375 forward);
10376 }
10377 } else {
10378 BRIDGE_UNLOCK(sc);
10379 }
10380 /* unicast packets for this interface do not get forwarded */
10381 goto done;
10382 }
10383 if (is_bridge_mac || list.head == NULL) {
10384 BRIDGE_UNLOCK(sc);
10385 goto done;
10386 }
10387 if (!is_broadcast) {
10388 /* find where to send the packet */
10389 dst_if = bridge_rtlookup(sc, dhost, vlan);
10390 if (ifp == dst_if) {
10391 /* nothing to forward */
10392 BRIDGE_UNLOCK(sc);
10393 goto done;
10394 }
10395 if (dst_if == NULL) {
10396 /* if a member is the dhost, deliver as input */
10397 member_input = bridge_find_member(sc, dhost, bif);
10398 if (member_input != NULL) {
10399 /* grab packets destined to member */
10400 BRIDGE_UNLOCK(sc);
10401 goto done;
10402 }
10403 /* if a member is shost, there's a loop, drop it */
10404 if (bridge_find_member(sc, shost, bif) != NULL) {
10405 BRIDGE_UNLOCK(sc);
10406 m_freem_list(list.head);
10407 list.head = list.tail = NULL;
10408 goto done;
10409 }
10410 }
10411 }
10412 if (dst_if == NULL) {
10413 /* bridge_broadcast_list() copies list and unlocks */
10414 bridge_broadcast_list(sc, bif, etypef, list.head, true);
10415 } else {
10416 /* bridge_forward_list() consumes list and unlocks */
10417 bridge_forward_list(sc, bif, dst_if, etypef, list.head);
10418 list.head = list.tail = NULL;
10419 }
10420
10421 done:
10422 if (list.head != NULL) {
10423 if (member_input != NULL) {
10424 /* member gets the packets */
10425 inject_input_packet_list(member_input, list.head, true);
10426 list.head = list.tail = NULL;
10427 } else if (is_bridge_mac) {
10428 /* bridge consumes all the unicast packets */
10429 bridge_interface_input_list(bridge_ifp, etypef, list,
10430 bridge_has_address);
10431 list.head = list.tail = NULL;
10432 } else {
10433 adjust_input_packet_list(list.head);
10434 }
10435 }
10436 return list;
10437 }
10438
10439 static inline void
10440 update_mbuf_flags(struct ifnet * ifp, mbuf_t m, struct ether_header * eh)
10441 {
10442 /* duplicate some of the work done in ether_demux */
10443 if ((eh->ether_dhost[0] & 1) == 0) {
10444 if (_ether_cmp(eh->ether_dhost, IF_LLADDR(ifp)) != 0) {
10445 m->m_flags |= M_PROMISC;
10446 }
10447 } else {
10448 /* Check for broadcast */
10449 if (_ether_cmp(etherbroadcastaddr, eh->ether_dhost) == 0) {
10450 m->m_flags |= M_BCAST;
10451 } else {
10452 m->m_flags |= M_MCAST;
10453 }
10454 }
10455 if (m->m_flags & M_HASFCS) {
10456 /*
10457 * If the M_HASFCS is set by the driver we want to make sure
10458 * that we strip off the trailing FCS data before handing it
10459 * up the stack.
10460 */
10461 m_adj(m, -ETHER_CRC_LEN);
10462 m->m_flags &= ~M_HASFCS;
10463 }
10464 return;
10465 }
10466
10467 static mbuf_t
10468 bridge_pf_list(mbuf_t m, ifnet_t ifp, uint32_t sc_filter_flags, bool input)
10469 {
10470 mbuf_t next_packet = NULL;
10471 mblist ret;
10472
10473 mblist_init(&ret);
10474 for (mbuf_ref_t scan = m; scan != NULL; scan = next_packet) {
10475 next_packet = scan->m_nextpkt;
10476
10477 /* remove packet from list, and pass through PF */
10478 scan->m_nextpkt = NULL;
10479 MBUF_INPUT_CHECK(scan, ifp);
10480 bridge_pf(&scan, ifp, sc_filter_flags, input);
10481 if (scan != NULL) {
10482 /* add packet back to the list */
10483 mblist_append(&ret, scan);
10484 }
10485 }
10486 return ret.head;
10487 }
10488
10489 static inline bool
10490 get_and_clear_promisc(mbuf_t m)
10491 {
10492 bool is_promisc;
10493
10494 /*
10495 * Need to clear the promiscuous flag otherwise the packet will be
10496 * dropped by DLIL after processing filters
10497 */
10498 is_promisc = (mbuf_flags(m) & MBUF_PROMISC) != 0;
10499 if (is_promisc) {
10500 mbuf_setflags_mask(m, 0, MBUF_PROMISC);
10501 }
10502 return is_promisc;
10503 }
10504
10505 static inline bool
10506 bridge_check_frame_header(struct bridge_softc * sc, ifnet_t ifp, mbuf_t m)
10507 {
10508 bool included = false;
10509 char * __single header;
10510 size_t header_length = 0;
10511
10512 header = m->m_pkthdr.pkt_hdr;
10513 if (header >= (char *)mbuf_datastart(m) &&
10514 header <= mtod(m, char *)) {
10515 header_length = mtod(m, char *) - header;
10516 if (header_length >= ETHER_HDR_LEN) {
10517 included = true;
10518 }
10519 }
10520 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT,
10521 "%s from %s m 0x%llx data 0x%llx frame 0x%llx %s "
10522 "header length %lu", sc->sc_ifp->if_xname,
10523 ifp->if_xname, (uint64_t)VM_KERNEL_ADDRPERM(m),
10524 (uint64_t)VM_KERNEL_ADDRPERM(mtod(m, void *)),
10525 (uint64_t)VM_KERNEL_ADDRPERM(header),
10526 included ? "inside" : "outside", header_length);
10527 if (!included) {
10528 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT,
10529 "%s: frame_header outside mbuf", ifp->if_xname);
10530 }
10531 return included;
10532 }
10533
10534
10535 mbuf_t
10536 bridge_early_input(struct ifnet *ifp, mbuf_t in_list, u_int32_t cnt)
10537 {
10538 struct ether_header eh;
10539 mblist list;
10540 volatile bool list_is_promisc;
10541 int n_lists = 0;
10542 mbuf_t next_packet = NULL;
10543 mblist ret;
10544 struct bridge_softc * __single sc = ifp->if_bridge;
10545 uint32_t sc_filter_flags;
10546
10547 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT_LIST,
10548 "(%s): count %u", ifp->if_xname, cnt);
10549
10550 /* run packet list through PF first */
10551 sc_filter_flags = sc->sc_filter_flags;
10552 if (PF_IS_ENABLED && (sc_filter_flags & IFBF_FILT_MEMBER)) {
10553 in_list = bridge_pf_list(in_list, ifp, sc_filter_flags, true);
10554 }
10555
10556 /* form sublists with the same ethernet header */
10557 mblist_init(&list);
10558 mblist_init(&ret);
10559 for (mbuf_t scan = in_list; scan != NULL; scan = next_packet) {
10560 struct ether_header * eh_p;
10561 volatile bool is_promisc;
10562 mblist resid;
10563
10564 /* take it out of the list */
10565 next_packet = scan->m_nextpkt;
10566 scan->m_nextpkt = NULL;
10567
10568 /* don't loop the packet */
10569 if ((scan->m_flags & M_PROTO1) != 0) {
10570 mblist_append(&ret, scan);
10571 continue;
10572 }
10573 /* Check if this mbuf looks valid */
10574 MBUF_INPUT_CHECK(scan, ifp);
10575
10576 /* if the frame header isn't in the first mbuf, ignore */
10577 if (!bridge_check_frame_header(sc, ifp, scan)) {
10578 mblist_append(&ret, scan);
10579 continue;
10580 }
10581 eh_p = __unsafe_forge_single(struct ether_header *,
10582 scan->m_pkthdr.pkt_hdr);
10583 update_mbuf_flags(ifp, scan, eh_p);
10584
10585 /* set start back to include ether header */
10586 _mbuf_adjust_pkthdr_and_data(scan, -ETHER_HDR_LEN);
10587
10588 is_promisc = get_and_clear_promisc(scan);
10589 if (list.head == NULL) {
10590 /* start a new list */
10591 mblist_append(&list, scan);
10592 bcopy(eh_p, &eh, sizeof(eh));
10593 list_is_promisc = is_promisc;
10594 } else if (bcmp(eh_p, &eh, sizeof(eh)) != 0) {
10595 n_lists++;
10596 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT_LIST,
10597 "(%s): sublist %u pkts %u",
10598 ifp->if_xname, n_lists, list.count);
10599 if (BRIDGE_DBGF_ENABLED(BR_DBGF_INPUT_LIST)) {
10600 brlog_ether_header(&eh);
10601 }
10602 resid = bridge_input_list(sc, ifp, &eh, list,
10603 list_is_promisc);
10604 if (resid.head != NULL) {
10605 /* add to the packets to be returned */
10606 mblist_append_list(&ret, resid);
10607 }
10608 /* start new list */
10609 mblist_init(&list);
10610 mblist_append(&list, scan);
10611 list_is_promisc = is_promisc;
10612 bcopy(eh_p, &eh, sizeof(eh));
10613 } else {
10614 mblist_append(&list, scan);
10615 VERIFY(is_promisc == list_is_promisc);
10616 }
10617 if (next_packet == NULL) {
10618 /* last list */
10619 n_lists++;
10620 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT_LIST,
10621 "(%s): sublist %u pkts %u",
10622 ifp->if_xname, n_lists, list.count);
10623 if (BRIDGE_DBGF_ENABLED(BR_DBGF_INPUT_LIST)) {
10624 brlog_ether_header(&eh);
10625 }
10626 resid = bridge_input_list(sc, ifp, &eh, list,
10627 list_is_promisc);
10628 if (resid.head != NULL) {
10629 /* add to the packets to be returned */
10630 mblist_append_list(&ret, resid);
10631 }
10632 }
10633 }
10634 return ret.head;
10635 }
10636
10637 /*
10638 * Copyright (C) 2014, Stefano Garzarella - Universita` di Pisa.
10639 * All rights reserved.
10640 *
10641 * Redistribution and use in source and binary forms, with or without
10642 * modification, are permitted provided that the following conditions
10643 * are met:
10644 * 1. Redistributions of source code must retain the above copyright
10645 * notice, this list of conditions and the following disclaimer.
10646 * 2. Redistributions in binary form must reproduce the above copyright
10647 * notice, this list of conditions and the following disclaimer in the
10648 * documentation and/or other materials provided with the distribution.
10649 *
10650 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
10651 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
10652 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
10653 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
10654 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
10655 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
10656 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
10657 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
10658 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
10659 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
10660 * SUCH DAMAGE.
10661 */
10662
10663 /*
10664 * XXX-ste: Maybe this function must be moved into kern/uipc_mbuf.c
10665 *
10666 * Create a queue of packets/segments which fit the given mss + hdr_len.
10667 * m0 points to mbuf chain to be segmented.
10668 * This function splits the payload (m0-> m_pkthdr.len - hdr_len)
10669 * into segments of length MSS bytes and then copy the first hdr_len bytes
10670 * from m0 at the top of each segment.
10671 * If hdr2_buf is not NULL (hdr2_len is the buf length), it is copied
10672 * in each segment after the first hdr_len bytes
10673 *
10674 * Return the new queue with the segments on success, NULL on failure.
10675 * (the mbuf queue is freed in this case).
10676 */
10677
10678 static mblist
10679 m_seg(struct mbuf *m0, int hdr_len, int mss, char * hdr2_buf, int hdr2_len)
10680 {
10681 int off = 0, n, firstlen;
10682 struct mbuf *mseg;
10683 int total_len = m0->m_pkthdr.len;
10684 mblist ret;
10685
10686 mblist_init(&ret);
10687 mblist_append(&ret, m0);
10688
10689 /*
10690 * Segmentation useless
10691 */
10692 if (total_len <= hdr_len + mss) {
10693 n = 1;
10694 goto done;
10695 }
10696
10697 if (hdr2_buf == NULL || hdr2_len <= 0) {
10698 hdr2_buf = NULL;
10699 hdr2_len = 0;
10700 }
10701
10702 off = hdr_len + mss;
10703 firstlen = mss; /* first segment stored in the original mbuf */
10704
10705 for (n = 1; off < total_len; off += mss, n++) {
10706 struct mbuf *m;
10707 /*
10708 * Copy the header from the original packet
10709 * and create a new mbuf chain
10710 */
10711 if (MHLEN < hdr_len) {
10712 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
10713 } else {
10714 m = m_gethdr(M_NOWAIT, MT_DATA);
10715 }
10716
10717 if (m == NULL) {
10718 #ifdef GSO_DEBUG
10719 D("MGETHDR error\n");
10720 #endif
10721 goto err;
10722 }
10723
10724 m_copydata(m0, 0, hdr_len, mtod(m, caddr_t));
10725
10726 m->m_len = hdr_len;
10727 /*
10728 * if the optional header is present, copy it
10729 */
10730 if (hdr2_buf != NULL) {
10731 m_copyback(m, hdr_len, hdr2_len, hdr2_buf);
10732 }
10733
10734 m->m_flags |= (m0->m_flags & M_COPYFLAGS);
10735 if (off + mss >= total_len) { /* last segment */
10736 mss = total_len - off;
10737 }
10738 /*
10739 * Copy the payload from original packet
10740 */
10741 mseg = m_copym(m0, off, mss, M_NOWAIT);
10742 if (mseg == NULL) {
10743 m_freem(m);
10744 #ifdef GSO_DEBUG
10745 D("m_copym error\n");
10746 #endif
10747 goto err;
10748 }
10749 m_cat(m, mseg);
10750
10751 m->m_pkthdr.len = hdr_len + hdr2_len + mss;
10752 m->m_pkthdr.rcvif = m0->m_pkthdr.rcvif;
10753 /*
10754 * Copy the checksum flags and data (in_cksum() need this)
10755 */
10756 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
10757 m->m_pkthdr.csum_data = m0->m_pkthdr.csum_data;
10758 m->m_pkthdr.tso_segsz = m0->m_pkthdr.tso_segsz;
10759
10760 mblist_append(&ret, m);
10761 }
10762
10763 /*
10764 * Update first segment.
10765 * If the optional header is present, is necessary
10766 * to insert it into the first segment.
10767 */
10768 if (hdr2_buf == NULL) {
10769 m_adj(m0, hdr_len + firstlen - total_len);
10770 m0->m_pkthdr.len = hdr_len + firstlen;
10771 } else {
10772 mseg = m_copym(m0, hdr_len, firstlen, M_NOWAIT);
10773 if (mseg == NULL) {
10774 #ifdef GSO_DEBUG
10775 D("m_copym error\n");
10776 #endif
10777 goto err;
10778 }
10779 m_adj(m0, hdr_len - total_len);
10780 m_copyback(m0, hdr_len, hdr2_len, hdr2_buf);
10781 m_cat(m0, mseg);
10782 m0->m_pkthdr.len = hdr_len + hdr2_len + firstlen;
10783 }
10784
10785 done:
10786 return ret;
10787
10788 err:
10789 if (ret.head != NULL) {
10790 m_freem_list(ret.head);
10791 mblist_init(&ret);
10792 }
10793 return ret;
10794 }
10795
10796 /*
10797 * Wrappers of IPv4 checksum functions
10798 */
10799 static inline void
10800 gso_ipv4_data_cksum(struct mbuf *m, struct ip *ip, int mac_hlen)
10801 {
10802 m->m_data += mac_hlen;
10803 m->m_len -= mac_hlen;
10804 m->m_pkthdr.len -= mac_hlen;
10805 #if __FreeBSD_version < 1000000
10806 ip->ip_len = ntohs(ip->ip_len); /* needed for in_delayed_cksum() */
10807 #endif
10808
10809 in_delayed_cksum(m);
10810
10811 #if __FreeBSD_version < 1000000
10812 ip->ip_len = htons(ip->ip_len);
10813 #endif
10814 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
10815 m->m_len += mac_hlen;
10816 m->m_pkthdr.len += mac_hlen;
10817 m->m_data -= mac_hlen;
10818 }
10819
10820 static inline void
10821 gso_ipv4_hdr_cksum(struct mbuf *m, struct ip *ip, int mac_hlen, int ip_hlen)
10822 {
10823 m->m_data += mac_hlen;
10824
10825 ip->ip_sum = in_cksum(m, ip_hlen);
10826
10827 m->m_pkthdr.csum_flags &= ~CSUM_IP;
10828 m->m_data -= mac_hlen;
10829 }
10830
10831 /*
10832 * Structure that contains the state during the TCP segmentation
10833 */
10834 struct gso_ip_tcp_state {
10835 void (*update)
10836 (struct gso_ip_tcp_state*, struct mbuf*);
10837 void (*internal)
10838 (struct gso_ip_tcp_state*, struct mbuf*);
10839 u_int ip_m0_len;
10840 uint8_t * __counted_by(ip_m0_len) hdr;
10841 struct tcphdr *tcp;
10842 int mac_hlen;
10843 int ip_hlen;
10844 int tcp_hlen;
10845 int hlen;
10846 int pay_len;
10847 int sw_csum;
10848 uint32_t tcp_seq;
10849 uint16_t ip_id;
10850 boolean_t is_tx;
10851 };
10852
10853 /*
10854 * Update the pointers to TCP and IPv4 headers
10855 */
10856 static inline void
10857 gso_ipv4_tcp_update(struct gso_ip_tcp_state *state, struct mbuf *m)
10858 {
10859 state->hdr = mtodo(m, state->mac_hlen);
10860 state->ip_m0_len = m->m_len - state->mac_hlen;
10861 state->ip_hlen = state->ip_hlen;
10862 state->tcp = (struct tcphdr *)(state->hdr + state->ip_hlen);
10863 state->pay_len = m->m_pkthdr.len - state->hlen;
10864 }
10865
10866 /*
10867 * Set properly the TCP and IPv4 headers
10868 */
10869 static inline void
10870 gso_ipv4_tcp_internal(struct gso_ip_tcp_state *state, struct mbuf *m)
10871 {
10872 struct ip *ip;
10873 /*
10874 * Update IP header
10875 */
10876 ip = (struct ip *)state->hdr;
10877 ip->ip_id = htons((state->ip_id)++);
10878 ip->ip_len = htons(m->m_pkthdr.len - state->mac_hlen);
10879 /*
10880 * TCP Checksum
10881 */
10882 state->tcp->th_sum = 0;
10883 state->tcp->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr,
10884 htons(state->tcp_hlen + IPPROTO_TCP + state->pay_len));
10885 /*
10886 * Checksum HW not supported (TCP)
10887 */
10888 if (state->sw_csum & CSUM_DELAY_DATA) {
10889 gso_ipv4_data_cksum(m, ip, state->mac_hlen);
10890 }
10891
10892 state->tcp_seq += state->pay_len;
10893 /*
10894 * IP Checksum
10895 */
10896 ip->ip_sum = 0;
10897 /*
10898 * Checksum HW not supported (IP)
10899 */
10900 if (state->sw_csum & CSUM_IP) {
10901 gso_ipv4_hdr_cksum(m, ip, state->mac_hlen, state->ip_hlen);
10902 }
10903 }
10904
10905
10906 /*
10907 * Updates the pointers to TCP and IPv6 headers
10908 */
10909 static inline void
10910 gso_ipv6_tcp_update(struct gso_ip_tcp_state *state, struct mbuf *m)
10911 {
10912 state->hdr = mtodo(m, state->mac_hlen);
10913 state->ip_m0_len = m->m_len - state->mac_hlen;
10914 state->ip_hlen = state->ip_hlen;
10915 state->tcp = (struct tcphdr *)(state->hdr + state->ip_hlen);
10916 state->pay_len = m->m_pkthdr.len - state->hlen;
10917 }
10918
10919 /*
10920 * Sets properly the TCP and IPv6 headers
10921 */
10922 static inline void
10923 gso_ipv6_tcp_internal(struct gso_ip_tcp_state *state, struct mbuf *m)
10924 {
10925 struct ip6_hdr *ip6;
10926
10927 ip6 = (struct ip6_hdr *)state->hdr;
10928 ip6->ip6_plen = htons(m->m_pkthdr.len - state->mac_hlen - state->ip_hlen);
10929 /*
10930 * TCP Checksum
10931 */
10932 state->tcp->th_sum = 0;
10933 state->tcp->th_sum = in6_pseudo(&ip6->ip6_src, &ip6->ip6_dst,
10934 htonl(state->tcp_hlen + state->pay_len + IPPROTO_TCP));
10935 /*
10936 * Checksum HW not supported (TCP)
10937 */
10938 if (state->sw_csum & CSUM_DELAY_IPV6_DATA) {
10939 (void)in6_finalize_cksum(m, state->mac_hlen, -1, -1, state->sw_csum);
10940 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_IPV6_DATA;
10941 }
10942 state->tcp_seq += state->pay_len;
10943 }
10944
10945 /*
10946 * Init the state during the TCP segmentation
10947 */
10948 static void
10949 gso_ip_tcp_init_state(struct gso_ip_tcp_state *state, struct ifnet *ifp,
10950 bool is_ipv4, int mac_hlen, int ip_hlen,
10951 uint8_t *__counted_by(ip_m0_len) ip_hdr, u_int ip_m0_len,
10952 struct tcphdr * tcp_hdr)
10953 {
10954 #pragma unused(ifp)
10955
10956 state->hdr = ip_hdr;
10957 state->ip_m0_len = ip_m0_len;
10958 state->ip_hlen = ip_hlen;
10959 state->tcp = tcp_hdr;
10960 if (is_ipv4) {
10961 state->ip_id = ntohs(((struct ip *)state->hdr)->ip_id);
10962 state->update = gso_ipv4_tcp_update;
10963 state->internal = gso_ipv4_tcp_internal;
10964 state->sw_csum = CSUM_DELAY_DATA | CSUM_IP; /* XXX */
10965 } else {
10966 state->update = gso_ipv6_tcp_update;
10967 state->internal = gso_ipv6_tcp_internal;
10968 state->sw_csum = CSUM_DELAY_IPV6_DATA; /* XXX */
10969 }
10970 state->mac_hlen = mac_hlen;
10971 state->tcp_hlen = state->tcp->th_off << 2;
10972 state->hlen = mac_hlen + ip_hlen + state->tcp_hlen;
10973 state->tcp_seq = ntohl(state->tcp->th_seq);
10974 //state->sw_csum = m->m_pkthdr.csum_flags & ~IF_HWASSIST_CSUM_FLAGS(ifp->if_hwassist);
10975 return;
10976 }
10977
10978 /*
10979 * GSO on TCP/IP (v4 or v6)
10980 *
10981 * Segment the given mbuf and return the list of packets.
10982 *
10983 */
10984 static mblist
10985 gso_ip_tcp(ifnet_t ifp, mbuf_t m0, struct gso_ip_tcp_state *state, bool is_tx)
10986 {
10987 struct mbuf *m;
10988 int mss = 0;
10989 #ifdef GSO_STATS
10990 int total_len = m0->m_pkthdr.len;
10991 #endif /* GSO_STATS */
10992 mblist seg;
10993
10994 #if 1
10995 /* hack: we don't get MSS via packet metadata */
10996 u_int reduce_mss;
10997
10998 reduce_mss = is_tx ? if_bridge_tso_reduce_mss_tx
10999 : if_bridge_tso_reduce_mss_forwarding;
11000 mss = ifp->if_mtu - state->ip_hlen - state->tcp_hlen - reduce_mss;
11001 assert(mss > 0);
11002 #else
11003 if (m0->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) {/* TSO with GSO */
11004 mss = ifp->if_hw_tsomax - state->ip_hlen - state->tcp_hlen;
11005 } else {
11006 mss = m0->m_pkthdr.tso_segsz;
11007 }
11008 #endif
11009 seg = m_seg(m0, state->hlen, mss, 0, 0);
11010 if (seg.head == NULL || seg.head->m_nextpkt == NULL) {
11011 return seg;
11012 }
11013 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
11014 "%s %s mss %d nsegs %d",
11015 ifp->if_xname,
11016 is_tx ? "TX" : "RX",
11017 mss, seg.count);
11018 #ifdef GSO_STATS
11019 GSOSTAT_SET_MAX(tcp.gsos_max_mss, mss);
11020 GSOSTAT_SET_MIN(tcp.gsos_min_mss, mss);
11021 GSOSTAT_ADD(tcp.gsos_osegments, seg.count);
11022 #endif /* GSO_STATS */
11023
11024 /* first pkt */
11025 VERIFY(seg.head == m0);
11026 m = m0;
11027
11028 state->update(state, m);
11029
11030 do {
11031 state->tcp->th_flags &= ~(TH_FIN | TH_PUSH);
11032
11033 state->internal(state, m);
11034 m = m->m_nextpkt;
11035 state->update(state, m);
11036 state->tcp->th_flags &= ~TH_CWR;
11037 state->tcp->th_seq = htonl(state->tcp_seq);
11038 } while (m->m_nextpkt);
11039
11040 /* last pkt */
11041 state->internal(state, m);
11042
11043 #ifdef GSO_STATS
11044 if (!error) {
11045 GSOSTAT_INC(tcp.gsos_segmented);
11046 GSOSTAT_SET_MAX(tcp.gsos_maxsegmented, total_len);
11047 GSOSTAT_SET_MIN(tcp.gsos_minsegmented, total_len);
11048 GSOSTAT_ADD(tcp.gsos_totalbyteseg, total_len);
11049 }
11050 #endif /* GSO_STATS */
11051 return seg;
11052 }
11053
11054 /*
11055 * GSO for TCP/IPv[46]
11056 */
11057 static mblist
11058 gso_tcp(ifnet_t ifp, mbuf_t m, u_int mac_hlen, bool is_ipv4, bool is_tx)
11059 {
11060 int error;
11061 ip_packet_info info;
11062 uint32_t csum_flags;
11063 struct gso_ip_tcp_state state;
11064 struct bripstats stats; /* XXX ignored */
11065 struct tcphdr *tcp;
11066 mblist ret;
11067
11068 mblist_init(&ret);
11069 error = bridge_get_tcp_header(&m, mac_hlen, is_ipv4, &info, &stats);
11070 if (error != 0) {
11071 if (m != NULL) {
11072 m_freem(m);
11073 m = NULL;
11074 }
11075 goto no_segment;
11076 }
11077 if (info.ip_proto_hdr == NULL) {
11078 /* not actually a TCP packet, no segmentation */
11079 goto no_segment;
11080 }
11081 tcp = (struct tcphdr *)(void *)info.ip_proto_hdr;
11082 gso_ip_tcp_init_state(&state, ifp, is_ipv4, mac_hlen,
11083 info.ip_hlen + info.ip_opt_len, info.ip_hdr, info.ip_m0_len, tcp);
11084 if (is_ipv4) {
11085 csum_flags = CSUM_DELAY_DATA; /* XXX */
11086 if (!is_tx) {
11087 /* if RX to our local IP address, don't segment */
11088 struct in_addr dst_ip;
11089
11090 bcopy(&((struct ip *)state.hdr)->ip_dst, &dst_ip, sizeof(dst_ip));
11091 if (in_addr_is_ours(dst_ip)) {
11092 goto no_segment;
11093 }
11094 }
11095 } else {
11096 csum_flags = CSUM_DELAY_IPV6_DATA; /* XXX */
11097 if (!is_tx) {
11098 /* if RX to our local IP address, don't segment */
11099 if (in6_addr_is_ours(&((struct ip6_hdr *)state.hdr)->ip6_dst,
11100 ifp->if_index)) {
11101 /* local IP address, no need to segment */
11102 goto no_segment;
11103 }
11104 }
11105 }
11106 m->m_pkthdr.csum_flags = csum_flags;
11107 m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum);
11108 return gso_ip_tcp(ifp, m, &state, is_tx);
11109
11110 no_segment:
11111 if (m != NULL) {
11112 mblist_append(&ret, m);
11113 }
11114 return ret;
11115 }
11116