1 /*
2 * Copyright (c) 2004-2023 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 #include <net/pfvar.h>
142
143 #include <netinet/in.h> /* for struct arpcom */
144 #include <netinet/tcp.h> /* for struct tcphdr */
145 #include <netinet/in_systm.h>
146 #include <netinet/in_var.h>
147 #define _IP_VHL
148 #include <netinet/ip.h>
149 #include <netinet/ip_var.h>
150 #include <netinet/ip6.h>
151 #include <netinet6/ip6_var.h>
152 #ifdef DEV_CARP
153 #include <netinet/ip_carp.h>
154 #endif
155 #include <netinet/if_ether.h> /* for struct arpcom */
156 #include <net/bridgestp.h>
157 #include <net/if_bridgevar.h>
158 #include <net/if_llc.h>
159 #if NVLAN > 0
160 #include <net/if_vlan_var.h>
161 #endif /* NVLAN > 0 */
162
163 #include <net/if_ether.h>
164 #include <net/dlil.h>
165 #include <net/kpi_interfacefilter.h>
166
167 #include <net/route.h>
168 #include <dev/random/randomdev.h>
169
170 #include <netinet/bootp.h>
171 #include <netinet/dhcp.h>
172
173 #if SKYWALK
174 #include <skywalk/nexus/netif/nx_netif.h>
175 #endif /* SKYWALK */
176
177 #include <os/log.h>
178
179 /*
180 * if_bridge_debug, BR_DBGF_*
181 * - 'if_bridge_debug' is a bitmask of BR_DBGF_* flags that can be set
182 * to enable additional logs for the corresponding bridge function
183 * - "sysctl net.link.bridge.debug" controls the value of
184 * 'if_bridge_debug'
185 */
186 static uint32_t if_bridge_debug = 0;
187 #define BR_DBGF_LIFECYCLE 0x0001
188 #define BR_DBGF_INPUT 0x0002
189 #define BR_DBGF_OUTPUT 0x0004
190 #define BR_DBGF_RT_TABLE 0x0008
191 #define BR_DBGF_DELAYED_CALL 0x0010
192 #define BR_DBGF_IOCTL 0x0020
193 #define BR_DBGF_MBUF 0x0040
194 #define BR_DBGF_MCAST 0x0080
195 #define BR_DBGF_HOSTFILTER 0x0100
196 #define BR_DBGF_CHECKSUM 0x0200
197 #define BR_DBGF_MAC_NAT 0x0400
198
199 /*
200 * if_bridge_log_level
201 * - 'if_bridge_log_level' ensures that by default important logs are
202 * logged regardless of if_bridge_debug by comparing the log level
203 * in BRIDGE_LOG to if_bridge_log_level
204 * - use "sysctl net.link.bridge.log_level" controls the value of
205 * 'if_bridge_log_level'
206 * - the default value of 'if_bridge_log_level' is LOG_NOTICE; important
207 * logs must use LOG_NOTICE to ensure they appear by default
208 */
209 static int if_bridge_log_level = LOG_NOTICE;
210
211 #define BRIDGE_DBGF_ENABLED(__flag) ((if_bridge_debug & __flag) != 0)
212
213 /*
214 * BRIDGE_LOG, BRIDGE_LOG_SIMPLE
215 * - macros to generate the specified log conditionally based on
216 * the specified log level and debug flags
217 * - BRIDGE_LOG_SIMPLE does not include the function name in the log
218 */
219 #define BRIDGE_LOG(__level, __dbgf, __string, ...) \
220 do { \
221 if (__level <= if_bridge_log_level || \
222 BRIDGE_DBGF_ENABLED(__dbgf)) { \
223 os_log(OS_LOG_DEFAULT, "%s: " __string, \
224 __func__, ## __VA_ARGS__); \
225 } \
226 } while (0)
227 #define BRIDGE_LOG_SIMPLE(__level, __dbgf, __string, ...) \
228 do { \
229 if (__level <= if_bridge_log_level || \
230 BRIDGE_DBGF_ENABLED(__dbgf)) { \
231 os_log(OS_LOG_DEFAULT, __string, ## __VA_ARGS__); \
232 } \
233 } while (0)
234
235 #define _BRIDGE_LOCK(_sc) lck_mtx_lock(&(_sc)->sc_mtx)
236 #define _BRIDGE_UNLOCK(_sc) lck_mtx_unlock(&(_sc)->sc_mtx)
237 #define BRIDGE_LOCK_ASSERT_HELD(_sc) \
238 LCK_MTX_ASSERT(&(_sc)->sc_mtx, LCK_MTX_ASSERT_OWNED)
239 #define BRIDGE_LOCK_ASSERT_NOTHELD(_sc) \
240 LCK_MTX_ASSERT(&(_sc)->sc_mtx, LCK_MTX_ASSERT_NOTOWNED)
241
242 #define BRIDGE_LOCK_DEBUG 1
243 #if BRIDGE_LOCK_DEBUG
244
245 #define BR_LCKDBG_MAX 4
246
247 #define BRIDGE_LOCK(_sc) bridge_lock(_sc)
248 #define BRIDGE_UNLOCK(_sc) bridge_unlock(_sc)
249 #define BRIDGE_LOCK2REF(_sc, _err) _err = bridge_lock2ref(_sc)
250 #define BRIDGE_UNREF(_sc) bridge_unref(_sc)
251 #define BRIDGE_XLOCK(_sc) bridge_xlock(_sc)
252 #define BRIDGE_XDROP(_sc) bridge_xdrop(_sc)
253
254 #else /* !BRIDGE_LOCK_DEBUG */
255
256 #define BRIDGE_LOCK(_sc) _BRIDGE_LOCK(_sc)
257 #define BRIDGE_UNLOCK(_sc) _BRIDGE_UNLOCK(_sc)
258 #define BRIDGE_LOCK2REF(_sc, _err) do { \
259 BRIDGE_LOCK_ASSERT_HELD(_sc); \
260 if ((_sc)->sc_iflist_xcnt > 0) \
261 (_err) = EBUSY; \
262 else { \
263 (_sc)->sc_iflist_ref++; \
264 (_err) = 0; \
265 } \
266 _BRIDGE_UNLOCK(_sc); \
267 } while (0)
268 #define BRIDGE_UNREF(_sc) do { \
269 _BRIDGE_LOCK(_sc); \
270 (_sc)->sc_iflist_ref--; \
271 if (((_sc)->sc_iflist_xcnt > 0) && ((_sc)->sc_iflist_ref == 0)) { \
272 _BRIDGE_UNLOCK(_sc); \
273 wakeup(&(_sc)->sc_cv); \
274 } else \
275 _BRIDGE_UNLOCK(_sc); \
276 } while (0)
277 #define BRIDGE_XLOCK(_sc) do { \
278 BRIDGE_LOCK_ASSERT_HELD(_sc); \
279 (_sc)->sc_iflist_xcnt++; \
280 while ((_sc)->sc_iflist_ref > 0) \
281 msleep(&(_sc)->sc_cv, &(_sc)->sc_mtx, PZERO, \
282 "BRIDGE_XLOCK", NULL); \
283 } while (0)
284 #define BRIDGE_XDROP(_sc) do { \
285 BRIDGE_LOCK_ASSERT_HELD(_sc); \
286 (_sc)->sc_iflist_xcnt--; \
287 } while (0)
288
289 #endif /* BRIDGE_LOCK_DEBUG */
290
291 #if NBPFILTER > 0
292 #define BRIDGE_BPF_MTAP_INPUT(sc, m) \
293 if (sc->sc_bpf_input != NULL) \
294 bridge_bpf_input(sc->sc_ifp, m, __func__, __LINE__)
295 #else /* NBPFILTER */
296 #define BRIDGE_BPF_MTAP_INPUT(ifp, m)
297 #endif /* NBPFILTER */
298
299 /*
300 * Initial size of the route hash table. Must be a power of two.
301 */
302 #ifndef BRIDGE_RTHASH_SIZE
303 #define BRIDGE_RTHASH_SIZE 16
304 #endif
305
306 /*
307 * Maximum size of the routing hash table
308 */
309 #define BRIDGE_RTHASH_SIZE_MAX 2048
310
311 #define BRIDGE_RTHASH_MASK(sc) ((sc)->sc_rthash_size - 1)
312
313 /*
314 * Maximum number of addresses to cache.
315 */
316 #ifndef BRIDGE_RTABLE_MAX
317 #define BRIDGE_RTABLE_MAX 100
318 #endif
319
320
321 /*
322 * Timeout (in seconds) for entries learned dynamically.
323 */
324 #ifndef BRIDGE_RTABLE_TIMEOUT
325 #define BRIDGE_RTABLE_TIMEOUT (20 * 60) /* same as ARP */
326 #endif
327
328 /*
329 * Number of seconds between walks of the route list.
330 */
331 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
332 #define BRIDGE_RTABLE_PRUNE_PERIOD (5 * 60)
333 #endif
334
335 /*
336 * Number of MAC NAT entries
337 * - sized based on 16 clients (including MAC NAT interface)
338 * each with 4 addresses
339 */
340 #ifndef BRIDGE_MAC_NAT_ENTRY_MAX
341 #define BRIDGE_MAC_NAT_ENTRY_MAX 64
342 #endif /* BRIDGE_MAC_NAT_ENTRY_MAX */
343
344 /*
345 * List of capabilities to possibly mask on the member interface.
346 */
347 #define BRIDGE_IFCAPS_MASK (IFCAP_TSO | IFCAP_TXCSUM)
348 /*
349 * List of capabilities to disable on the member interface.
350 */
351 #define BRIDGE_IFCAPS_STRIP IFCAP_LRO
352
353 /*
354 * Bridge interface list entry.
355 */
356 struct bridge_iflist {
357 TAILQ_ENTRY(bridge_iflist) bif_next;
358 struct ifnet *bif_ifp; /* member if */
359 struct bstp_port bif_stp; /* STP state */
360 uint32_t bif_ifflags; /* member if flags */
361 int bif_savedcaps; /* saved capabilities */
362 uint32_t bif_addrmax; /* max # of addresses */
363 uint32_t bif_addrcnt; /* cur. # of addresses */
364 uint32_t bif_addrexceeded; /* # of address violations */
365
366 interface_filter_t bif_iff_ref;
367 struct bridge_softc *bif_sc;
368 uint32_t bif_flags;
369
370 /* host filter */
371 struct in_addr bif_hf_ipsrc;
372 uint8_t bif_hf_hwsrc[ETHER_ADDR_LEN];
373
374 struct ifbrmstats bif_stats;
375 };
376
377 static inline bool
bif_ifflags_are_set(struct bridge_iflist * bif,uint32_t flags)378 bif_ifflags_are_set(struct bridge_iflist * bif, uint32_t flags)
379 {
380 return (bif->bif_ifflags & flags) == flags;
381 }
382
383 static inline bool
bif_has_checksum_offload(struct bridge_iflist * bif)384 bif_has_checksum_offload(struct bridge_iflist * bif)
385 {
386 return bif_ifflags_are_set(bif, IFBIF_CHECKSUM_OFFLOAD);
387 }
388
389 /* fake errors to make the code clearer */
390 #define _EBADIP EJUSTRETURN
391 #define _EBADIPCHECKSUM EJUSTRETURN
392 #define _EBADIPV6 EJUSTRETURN
393 #define _EBADUDP EJUSTRETURN
394 #define _EBADTCP EJUSTRETURN
395 #define _EBADUDPCHECKSUM EJUSTRETURN
396 #define _EBADTCPCHECKSUM EJUSTRETURN
397
398 #define BIFF_PROMISC 0x01 /* promiscuous mode set */
399 #define BIFF_PROTO_ATTACHED 0x02 /* protocol attached */
400 #define BIFF_FILTER_ATTACHED 0x04 /* interface filter attached */
401 #define BIFF_MEDIA_ACTIVE 0x08 /* interface media active */
402 #define BIFF_HOST_FILTER 0x10 /* host filter enabled */
403 #define BIFF_HF_HWSRC 0x20 /* host filter source MAC is set */
404 #define BIFF_HF_IPSRC 0x40 /* host filter source IP is set */
405 #define BIFF_INPUT_BROADCAST 0x80 /* send broadcast packets in */
406 #define BIFF_IN_MEMBER_LIST 0x100 /* added to the member list */
407 #define BIFF_WIFI_INFRA 0x200 /* interface is Wi-Fi infra */
408 #define BIFF_ALL_MULTI 0x400 /* allmulti set */
409 #if SKYWALK
410 #define BIFF_FLOWSWITCH_ATTACHED 0x1000 /* we attached the flowswitch */
411 #define BIFF_NETAGENT_REMOVED 0x2000 /* we removed the netagent */
412 #endif /* SKYWALK */
413
414 /*
415 * mac_nat_entry
416 * - translates between an IP address and MAC address on a specific
417 * bridge interface member
418 */
419 struct mac_nat_entry {
420 LIST_ENTRY(mac_nat_entry) mne_list; /* list linkage */
421 struct bridge_iflist *mne_bif; /* originating interface */
422 unsigned long mne_expire; /* expiration time */
423 union {
424 struct in_addr mneu_ip; /* originating IPv4 address */
425 struct in6_addr mneu_ip6; /* originating IPv6 address */
426 } mne_u;
427 uint8_t mne_mac[ETHER_ADDR_LEN];
428 uint8_t mne_flags;
429 uint8_t mne_reserved;
430 };
431 #define mne_ip mne_u.mneu_ip
432 #define mne_ip6 mne_u.mneu_ip6
433
434 #define MNE_FLAGS_IPV6 0x01 /* IPv6 address */
435
436 LIST_HEAD(mac_nat_entry_list, mac_nat_entry);
437
438 /*
439 * mac_nat_record
440 * - used by bridge_mac_nat_output() to convey the translation that needs
441 * to take place in bridge_mac_nat_translate
442 * - holds enough information so that the translation can be done later without
443 * holding the bridge lock
444 */
445 struct mac_nat_record {
446 uint16_t mnr_ether_type;
447 union {
448 uint16_t mnru_arp_offset;
449 struct {
450 uint16_t mnruip_dhcp_flags;
451 uint16_t mnruip_udp_csum;
452 uint8_t mnruip_header_len;
453 } mnru_ip;
454 struct {
455 uint16_t mnruip6_icmp6_len;
456 uint16_t mnruip6_lladdr_offset;
457 uint8_t mnruip6_icmp6_type;
458 uint8_t mnruip6_header_len;
459 } mnru_ip6;
460 } mnr_u;
461 };
462
463 #define mnr_arp_offset mnr_u.mnru_arp_offset
464
465 #define mnr_ip_header_len mnr_u.mnru_ip.mnruip_header_len
466 #define mnr_ip_dhcp_flags mnr_u.mnru_ip.mnruip_dhcp_flags
467 #define mnr_ip_udp_csum mnr_u.mnru_ip.mnruip_udp_csum
468
469 #define mnr_ip6_icmp6_len mnr_u.mnru_ip6.mnruip6_icmp6_len
470 #define mnr_ip6_icmp6_type mnr_u.mnru_ip6.mnruip6_icmp6_type
471 #define mnr_ip6_header_len mnr_u.mnru_ip6.mnruip6_header_len
472 #define mnr_ip6_lladdr_offset mnr_u.mnru_ip6.mnruip6_lladdr_offset
473
474 /*
475 * Bridge route node.
476 */
477 struct bridge_rtnode {
478 LIST_ENTRY(bridge_rtnode) brt_hash; /* hash table linkage */
479 LIST_ENTRY(bridge_rtnode) brt_list; /* list linkage */
480 struct bridge_iflist *brt_dst; /* destination if */
481 unsigned long brt_expire; /* expiration time */
482 uint8_t brt_flags; /* address flags */
483 uint8_t brt_addr[ETHER_ADDR_LEN];
484 uint16_t brt_vlan; /* vlan id */
485
486 };
487 #define brt_ifp brt_dst->bif_ifp
488
489 /*
490 * Bridge delayed function call context
491 */
492 typedef void (*bridge_delayed_func_t)(struct bridge_softc *);
493
494 struct bridge_delayed_call {
495 struct bridge_softc *bdc_sc;
496 bridge_delayed_func_t bdc_func; /* Function to call */
497 struct timespec bdc_ts; /* Time to call */
498 u_int32_t bdc_flags;
499 thread_call_t bdc_thread_call;
500 };
501
502 #define BDCF_OUTSTANDING 0x01 /* Delayed call has been scheduled */
503 #define BDCF_CANCELLING 0x02 /* May be waiting for call completion */
504
505 /*
506 * Software state for each bridge.
507 */
508 LIST_HEAD(_bridge_rtnode_list, bridge_rtnode);
509
510 struct bridge_softc {
511 struct ifnet *sc_ifp; /* make this an interface */
512 u_int32_t sc_flags;
513 LIST_ENTRY(bridge_softc) sc_list;
514 decl_lck_mtx_data(, sc_mtx);
515 struct _bridge_rtnode_list *sc_rthash; /* our forwarding table */
516 struct _bridge_rtnode_list sc_rtlist; /* list version of above */
517 uint32_t sc_rthash_key; /* key for hash */
518 uint32_t sc_rthash_size; /* size of the hash table */
519 struct bridge_delayed_call sc_aging_timer;
520 struct bridge_delayed_call sc_resize_call;
521 TAILQ_HEAD(, bridge_iflist) sc_spanlist; /* span ports list */
522 struct bstp_state sc_stp; /* STP state */
523 bpf_packet_func sc_bpf_input;
524 bpf_packet_func sc_bpf_output;
525 void *sc_cv;
526 uint32_t sc_brtmax; /* max # of addresses */
527 uint32_t sc_brtcnt; /* cur. # of addresses */
528 uint32_t sc_brttimeout; /* rt timeout in seconds */
529 uint32_t sc_iflist_ref; /* refcount for sc_iflist */
530 uint32_t sc_iflist_xcnt; /* refcount for sc_iflist */
531 TAILQ_HEAD(, bridge_iflist) sc_iflist; /* member interface list */
532 uint32_t sc_brtexceeded; /* # of cache drops */
533 uint32_t sc_filter_flags; /* ipf and flags */
534 struct ifnet *sc_ifaddr; /* member mac copied from */
535 u_char sc_defaddr[6]; /* Default MAC address */
536 char sc_if_xname[IFNAMSIZ];
537
538 struct bridge_iflist *sc_mac_nat_bif; /* single MAC NAT interface */
539 struct mac_nat_entry_list sc_mne_list; /* MAC NAT IPv4 */
540 struct mac_nat_entry_list sc_mne_list_v6;/* MAC NAT IPv6 */
541 uint32_t sc_mne_max; /* max # of entries */
542 uint32_t sc_mne_count; /* cur. # of entries */
543 uint32_t sc_mne_allocation_failures;
544 #if BRIDGE_LOCK_DEBUG
545 /*
546 * Locking and unlocking calling history
547 */
548 void *lock_lr[BR_LCKDBG_MAX];
549 int next_lock_lr;
550 void *unlock_lr[BR_LCKDBG_MAX];
551 int next_unlock_lr;
552 #endif /* BRIDGE_LOCK_DEBUG */
553 };
554
555 #define SCF_DETACHING 0x01
556 #define SCF_RESIZING 0x02
557 #define SCF_MEDIA_ACTIVE 0x04
558
559 typedef enum {
560 CHECKSUM_OPERATION_NONE = 0,
561 CHECKSUM_OPERATION_CLEAR_OFFLOAD = 1,
562 CHECKSUM_OPERATION_FINALIZE = 2,
563 CHECKSUM_OPERATION_COMPUTE = 3,
564 } ChecksumOperation;
565
566 union iphdr {
567 struct ip *ip;
568 struct ip6_hdr *ip6;
569 void * ptr;
570 };
571
572 typedef struct {
573 u_int ip_hlen; /* IP header length */
574 u_int ip_pay_len; /* length of payload (exclusive of ip_hlen) */
575 u_int ip_opt_len; /* IPv6 options headers length */
576 uint8_t ip_proto; /* IPPROTO_TCP, IPPROTO_UDP, etc. */
577 bool ip_is_ipv4;
578 bool ip_is_fragmented;
579 union iphdr ip_hdr; /* pointer to IP header */
580 void * ip_proto_hdr; /* ptr to protocol header (TCP) */
581 } ip_packet_info, *ip_packet_info_t;
582
583 struct bridge_hostfilter_stats bridge_hostfilter_stats;
584
585 static LCK_GRP_DECLARE(bridge_lock_grp, "if_bridge");
586 #if BRIDGE_LOCK_DEBUG
587 static LCK_ATTR_DECLARE(bridge_lock_attr, 0, 0);
588 #else
589 static LCK_ATTR_DECLARE(bridge_lock_attr, LCK_ATTR_DEBUG, 0);
590 #endif
591 static LCK_MTX_DECLARE_ATTR(bridge_list_mtx, &bridge_lock_grp, &bridge_lock_attr);
592
593 static int bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
594
595 static KALLOC_TYPE_DEFINE(bridge_rtnode_pool, struct bridge_rtnode, NET_KT_DEFAULT);
596 static KALLOC_TYPE_DEFINE(bridge_mne_pool, struct mac_nat_entry, NET_KT_DEFAULT);
597
598 static int bridge_clone_create(struct if_clone *, uint32_t, void *);
599 static int bridge_clone_destroy(struct ifnet *);
600
601 static errno_t bridge_ioctl(struct ifnet *, u_long, void *);
602 #if HAS_IF_CAP
603 static void bridge_mutecaps(struct bridge_softc *);
604 static void bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
605 int);
606 #endif
607 static errno_t bridge_set_tso(struct bridge_softc *);
608 static void bridge_proto_attach_changed(struct ifnet *);
609 static int bridge_init(struct ifnet *);
610 #if HAS_BRIDGE_DUMMYNET
611 static void bridge_dummynet(struct mbuf *, struct ifnet *);
612 #endif
613 static void bridge_ifstop(struct ifnet *, int);
614 static int bridge_output(struct ifnet *, struct mbuf *);
615 static void bridge_finalize_cksum(struct ifnet *, struct mbuf *);
616 static void bridge_start(struct ifnet *);
617 static errno_t bridge_input(struct ifnet *, mbuf_t *);
618 static errno_t bridge_iff_input(void *, ifnet_t, protocol_family_t,
619 mbuf_t *, char **);
620 static errno_t bridge_iff_output(void *, ifnet_t, protocol_family_t,
621 mbuf_t *);
622 static errno_t bridge_member_output(struct bridge_softc *sc, ifnet_t ifp,
623 mbuf_t *m);
624
625 static int bridge_enqueue(ifnet_t, struct ifnet *,
626 struct ifnet *, struct mbuf *, ChecksumOperation);
627 static void bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
628
629 static void bridge_forward(struct bridge_softc *, struct bridge_iflist *,
630 struct mbuf *);
631
632 static void bridge_aging_timer(struct bridge_softc *sc);
633
634 static void bridge_broadcast(struct bridge_softc *, struct bridge_iflist *,
635 struct mbuf *, int);
636 static void bridge_span(struct bridge_softc *, struct mbuf *);
637
638 static int bridge_rtupdate(struct bridge_softc *, const uint8_t *,
639 uint16_t, struct bridge_iflist *, int, uint8_t);
640 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
641 uint16_t);
642 static void bridge_rttrim(struct bridge_softc *);
643 static void bridge_rtage(struct bridge_softc *);
644 static void bridge_rtflush(struct bridge_softc *, int);
645 static int bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
646 uint16_t);
647
648 static int bridge_rtable_init(struct bridge_softc *);
649 static void bridge_rtable_fini(struct bridge_softc *);
650
651 static void bridge_rthash_resize(struct bridge_softc *);
652
653 static int bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
654 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
655 const uint8_t *, uint16_t);
656 static int bridge_rtnode_hash(struct bridge_softc *,
657 struct bridge_rtnode *);
658 static int bridge_rtnode_insert(struct bridge_softc *,
659 struct bridge_rtnode *);
660 static void bridge_rtnode_destroy(struct bridge_softc *,
661 struct bridge_rtnode *);
662 #if BRIDGESTP
663 static void bridge_rtable_expire(struct ifnet *, int);
664 static void bridge_state_change(struct ifnet *, int);
665 #endif /* BRIDGESTP */
666
667 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
668 const char *name);
669 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
670 struct ifnet *ifp);
671 static void bridge_delete_member(struct bridge_softc *,
672 struct bridge_iflist *);
673 static void bridge_delete_span(struct bridge_softc *,
674 struct bridge_iflist *);
675
676 static int bridge_ioctl_add(struct bridge_softc *, void *);
677 static int bridge_ioctl_del(struct bridge_softc *, void *);
678 static int bridge_ioctl_gifflags(struct bridge_softc *, void *);
679 static int bridge_ioctl_sifflags(struct bridge_softc *, void *);
680 static int bridge_ioctl_scache(struct bridge_softc *, void *);
681 static int bridge_ioctl_gcache(struct bridge_softc *, void *);
682 static int bridge_ioctl_gifs32(struct bridge_softc *, void *);
683 static int bridge_ioctl_gifs64(struct bridge_softc *, void *);
684 static int bridge_ioctl_rts32(struct bridge_softc *, void *);
685 static int bridge_ioctl_rts64(struct bridge_softc *, void *);
686 static int bridge_ioctl_saddr32(struct bridge_softc *, void *);
687 static int bridge_ioctl_saddr64(struct bridge_softc *, void *);
688 static int bridge_ioctl_sto(struct bridge_softc *, void *);
689 static int bridge_ioctl_gto(struct bridge_softc *, void *);
690 static int bridge_ioctl_daddr32(struct bridge_softc *, void *);
691 static int bridge_ioctl_daddr64(struct bridge_softc *, void *);
692 static int bridge_ioctl_flush(struct bridge_softc *, void *);
693 static int bridge_ioctl_gpri(struct bridge_softc *, void *);
694 static int bridge_ioctl_spri(struct bridge_softc *, void *);
695 static int bridge_ioctl_ght(struct bridge_softc *, void *);
696 static int bridge_ioctl_sht(struct bridge_softc *, void *);
697 static int bridge_ioctl_gfd(struct bridge_softc *, void *);
698 static int bridge_ioctl_sfd(struct bridge_softc *, void *);
699 static int bridge_ioctl_gma(struct bridge_softc *, void *);
700 static int bridge_ioctl_sma(struct bridge_softc *, void *);
701 static int bridge_ioctl_sifprio(struct bridge_softc *, void *);
702 static int bridge_ioctl_sifcost(struct bridge_softc *, void *);
703 static int bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
704 static int bridge_ioctl_addspan(struct bridge_softc *, void *);
705 static int bridge_ioctl_delspan(struct bridge_softc *, void *);
706 static int bridge_ioctl_gbparam32(struct bridge_softc *, void *);
707 static int bridge_ioctl_gbparam64(struct bridge_softc *, void *);
708 static int bridge_ioctl_grte(struct bridge_softc *, void *);
709 static int bridge_ioctl_gifsstp32(struct bridge_softc *, void *);
710 static int bridge_ioctl_gifsstp64(struct bridge_softc *, void *);
711 static int bridge_ioctl_sproto(struct bridge_softc *, void *);
712 static int bridge_ioctl_stxhc(struct bridge_softc *, void *);
713 static int bridge_ioctl_purge(struct bridge_softc *sc, void *);
714 static int bridge_ioctl_gfilt(struct bridge_softc *, void *);
715 static int bridge_ioctl_sfilt(struct bridge_softc *, void *);
716 static int bridge_ioctl_ghostfilter(struct bridge_softc *, void *);
717 static int bridge_ioctl_shostfilter(struct bridge_softc *, void *);
718 static int bridge_ioctl_gmnelist32(struct bridge_softc *, void *);
719 static int bridge_ioctl_gmnelist64(struct bridge_softc *, void *);
720 static int bridge_ioctl_gifstats32(struct bridge_softc *, void *);
721 static int bridge_ioctl_gifstats64(struct bridge_softc *, void *);
722
723 static int bridge_pf(struct mbuf **, struct ifnet *, uint32_t sc_filter_flags, int input);
724 static int bridge_ip_checkbasic(struct mbuf **);
725 static int bridge_ip6_checkbasic(struct mbuf **);
726
727 static errno_t bridge_set_bpf_tap(ifnet_t, bpf_tap_mode, bpf_packet_func);
728 static errno_t bridge_bpf_input(ifnet_t, struct mbuf *, const char *, int);
729 static errno_t bridge_bpf_output(ifnet_t, struct mbuf *);
730
731 static void bridge_detach(ifnet_t);
732 static void bridge_link_event(struct ifnet *, u_int32_t);
733 static void bridge_iflinkevent(struct ifnet *);
734 static u_int32_t bridge_updatelinkstatus(struct bridge_softc *);
735 static int interface_media_active(struct ifnet *);
736 static void bridge_schedule_delayed_call(struct bridge_delayed_call *);
737 static void bridge_cancel_delayed_call(struct bridge_delayed_call *);
738 static void bridge_cleanup_delayed_call(struct bridge_delayed_call *);
739 static int bridge_host_filter(struct bridge_iflist *, mbuf_t *);
740
741 static errno_t bridge_mac_nat_enable(struct bridge_softc *,
742 struct bridge_iflist *);
743 static void bridge_mac_nat_disable(struct bridge_softc *sc);
744 static void bridge_mac_nat_age_entries(struct bridge_softc *sc, unsigned long);
745 static void bridge_mac_nat_populate_entries(struct bridge_softc *sc);
746 static void bridge_mac_nat_flush_entries(struct bridge_softc *sc,
747 struct bridge_iflist *);
748 static ifnet_t bridge_mac_nat_input(struct bridge_softc *, mbuf_t *,
749 boolean_t *);
750 static boolean_t bridge_mac_nat_output(struct bridge_softc *,
751 struct bridge_iflist *, mbuf_t *, struct mac_nat_record *);
752 static void bridge_mac_nat_translate(mbuf_t *, struct mac_nat_record *,
753 const caddr_t);
754 static bool is_broadcast_ip_packet(mbuf_t *);
755 static bool in_addr_is_ours(const struct in_addr);
756 static bool in6_addr_is_ours(const struct in6_addr *, uint32_t);
757
758 #define m_copypacket(m, how) m_copym(m, 0, M_COPYALL, how)
759
760 static int
761 gso_tcp(struct ifnet *ifp, struct mbuf **mp, u_int mac_hlen, bool is_ipv4,
762 boolean_t is_tx);
763
764 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
765 #define VLANTAGOF(_m) 0
766
767 u_int8_t bstp_etheraddr[ETHER_ADDR_LEN] =
768 { 0x01, 0x80, 0xc2, 0x00, 0x00, 0x00 };
769
770 static u_int8_t ethernulladdr[ETHER_ADDR_LEN] =
771 { 0x00, 0x00, 0x00, 0x00, 0x00, 0x00 };
772
773 #if BRIDGESTP
774 static struct bstp_cb_ops bridge_ops = {
775 .bcb_state = bridge_state_change,
776 .bcb_rtage = bridge_rtable_expire
777 };
778 #endif /* BRIDGESTP */
779
780 SYSCTL_DECL(_net_link);
781 SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW | CTLFLAG_LOCKED, 0,
782 "Bridge");
783
784 static int bridge_inherit_mac = 0; /* share MAC with first bridge member */
785 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac,
786 CTLFLAG_RW | CTLFLAG_LOCKED,
787 &bridge_inherit_mac, 0,
788 "Inherit MAC address from the first bridge member");
789
790 SYSCTL_INT(_net_link_bridge, OID_AUTO, rtable_prune_period,
791 CTLFLAG_RW | CTLFLAG_LOCKED,
792 &bridge_rtable_prune_period, 0,
793 "Interval between pruning of routing table");
794
795 static unsigned int bridge_rtable_hash_size_max = BRIDGE_RTHASH_SIZE_MAX;
796 SYSCTL_UINT(_net_link_bridge, OID_AUTO, rtable_hash_size_max,
797 CTLFLAG_RW | CTLFLAG_LOCKED,
798 &bridge_rtable_hash_size_max, 0,
799 "Maximum size of the routing hash table");
800
801 #if BRIDGE_DELAYED_CALLBACK_DEBUG
802 static int bridge_delayed_callback_delay = 0;
803 SYSCTL_INT(_net_link_bridge, OID_AUTO, delayed_callback_delay,
804 CTLFLAG_RW | CTLFLAG_LOCKED,
805 &bridge_delayed_callback_delay, 0,
806 "Delay before calling delayed function");
807 #endif
808
809 SYSCTL_STRUCT(_net_link_bridge, OID_AUTO,
810 hostfilterstats, CTLFLAG_RD | CTLFLAG_LOCKED,
811 &bridge_hostfilter_stats, bridge_hostfilter_stats, "");
812
813 #if BRIDGESTP
814 static int log_stp = 0; /* log STP state changes */
815 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp, CTLFLAG_RW,
816 &log_stp, 0, "Log STP state changes");
817 #endif /* BRIDGESTP */
818
819 struct bridge_control {
820 int (*bc_func)(struct bridge_softc *, void *);
821 unsigned int bc_argsize;
822 unsigned int bc_flags;
823 };
824
825 #define VMNET_TAG "com.apple.vmnet"
826 #define VMNET_LOCAL_TAG VMNET_TAG ".local"
827 #define VMNET_BROADCAST_TAG VMNET_TAG ".broadcast"
828 #define VMNET_MULTICAST_TAG VMNET_TAG ".multicast"
829
830 static u_int16_t vmnet_tag;
831 static u_int16_t vmnet_local_tag;
832 static u_int16_t vmnet_broadcast_tag;
833 static u_int16_t vmnet_multicast_tag;
834
835 static u_int16_t
allocate_pf_tag(char * name)836 allocate_pf_tag(char * name)
837 {
838 u_int16_t tag;
839
840 tag = pf_tagname2tag_ext(name);
841 BRIDGE_LOG(LOG_NOTICE, 0, "%s %d", name, tag);
842 return tag;
843 }
844
845 static void
allocate_vmnet_pf_tags(void)846 allocate_vmnet_pf_tags(void)
847 {
848 /* allocate tags to use with PF */
849 if (vmnet_tag == 0) {
850 vmnet_tag = allocate_pf_tag(VMNET_TAG);
851 }
852 if (vmnet_local_tag == 0) {
853 vmnet_local_tag = allocate_pf_tag(VMNET_LOCAL_TAG);
854 }
855 if (vmnet_broadcast_tag == 0) {
856 vmnet_broadcast_tag = allocate_pf_tag(VMNET_BROADCAST_TAG);
857 }
858 if (vmnet_multicast_tag == 0) {
859 vmnet_multicast_tag = allocate_pf_tag(VMNET_MULTICAST_TAG);
860 }
861 }
862
863 #define BC_F_COPYIN 0x01 /* copy arguments in */
864 #define BC_F_COPYOUT 0x02 /* copy arguments out */
865 #define BC_F_SUSER 0x04 /* do super-user check */
866
867 static const struct bridge_control bridge_control_table32[] = {
868 { .bc_func = bridge_ioctl_add, .bc_argsize = sizeof(struct ifbreq), /* 0 */
869 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
870 { .bc_func = bridge_ioctl_del, .bc_argsize = sizeof(struct ifbreq),
871 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
872
873 { .bc_func = bridge_ioctl_gifflags, .bc_argsize = sizeof(struct ifbreq),
874 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
875 { .bc_func = bridge_ioctl_sifflags, .bc_argsize = sizeof(struct ifbreq),
876 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
877
878 { .bc_func = bridge_ioctl_scache, .bc_argsize = sizeof(struct ifbrparam),
879 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
880 { .bc_func = bridge_ioctl_gcache, .bc_argsize = sizeof(struct ifbrparam),
881 .bc_flags = BC_F_COPYOUT },
882
883 { .bc_func = bridge_ioctl_gifs32, .bc_argsize = sizeof(struct ifbifconf32),
884 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
885 { .bc_func = bridge_ioctl_rts32, .bc_argsize = sizeof(struct ifbaconf32),
886 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
887
888 { .bc_func = bridge_ioctl_saddr32, .bc_argsize = sizeof(struct ifbareq32),
889 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
890
891 { .bc_func = bridge_ioctl_sto, .bc_argsize = sizeof(struct ifbrparam),
892 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
893 { .bc_func = bridge_ioctl_gto, .bc_argsize = sizeof(struct ifbrparam), /* 10 */
894 .bc_flags = BC_F_COPYOUT },
895
896 { .bc_func = bridge_ioctl_daddr32, .bc_argsize = sizeof(struct ifbareq32),
897 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
898
899 { .bc_func = bridge_ioctl_flush, .bc_argsize = sizeof(struct ifbreq),
900 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
901
902 { .bc_func = bridge_ioctl_gpri, .bc_argsize = sizeof(struct ifbrparam),
903 .bc_flags = BC_F_COPYOUT },
904 { .bc_func = bridge_ioctl_spri, .bc_argsize = sizeof(struct ifbrparam),
905 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
906
907 { .bc_func = bridge_ioctl_ght, .bc_argsize = sizeof(struct ifbrparam),
908 .bc_flags = BC_F_COPYOUT },
909 { .bc_func = bridge_ioctl_sht, .bc_argsize = sizeof(struct ifbrparam),
910 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
911
912 { .bc_func = bridge_ioctl_gfd, .bc_argsize = sizeof(struct ifbrparam),
913 .bc_flags = BC_F_COPYOUT },
914 { .bc_func = bridge_ioctl_sfd, .bc_argsize = sizeof(struct ifbrparam),
915 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
916
917 { .bc_func = bridge_ioctl_gma, .bc_argsize = sizeof(struct ifbrparam),
918 .bc_flags = BC_F_COPYOUT },
919 { .bc_func = bridge_ioctl_sma, .bc_argsize = sizeof(struct ifbrparam), /* 20 */
920 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
921
922 { .bc_func = bridge_ioctl_sifprio, .bc_argsize = sizeof(struct ifbreq),
923 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
924
925 { .bc_func = bridge_ioctl_sifcost, .bc_argsize = sizeof(struct ifbreq),
926 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
927
928 { .bc_func = bridge_ioctl_gfilt, .bc_argsize = sizeof(struct ifbrparam),
929 .bc_flags = BC_F_COPYOUT },
930 { .bc_func = bridge_ioctl_sfilt, .bc_argsize = sizeof(struct ifbrparam),
931 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
932
933 { .bc_func = bridge_ioctl_purge, .bc_argsize = sizeof(struct ifbreq),
934 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
935
936 { .bc_func = bridge_ioctl_addspan, .bc_argsize = sizeof(struct ifbreq),
937 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
938 { .bc_func = bridge_ioctl_delspan, .bc_argsize = sizeof(struct ifbreq),
939 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
940
941 { .bc_func = bridge_ioctl_gbparam32, .bc_argsize = sizeof(struct ifbropreq32),
942 .bc_flags = BC_F_COPYOUT },
943
944 { .bc_func = bridge_ioctl_grte, .bc_argsize = sizeof(struct ifbrparam),
945 .bc_flags = BC_F_COPYOUT },
946
947 { .bc_func = bridge_ioctl_gifsstp32, .bc_argsize = sizeof(struct ifbpstpconf32), /* 30 */
948 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
949
950 { .bc_func = bridge_ioctl_sproto, .bc_argsize = sizeof(struct ifbrparam),
951 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
952
953 { .bc_func = bridge_ioctl_stxhc, .bc_argsize = sizeof(struct ifbrparam),
954 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
955
956 { .bc_func = bridge_ioctl_sifmaxaddr, .bc_argsize = sizeof(struct ifbreq),
957 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
958
959 { .bc_func = bridge_ioctl_ghostfilter, .bc_argsize = sizeof(struct ifbrhostfilter),
960 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
961 { .bc_func = bridge_ioctl_shostfilter, .bc_argsize = sizeof(struct ifbrhostfilter),
962 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
963
964 { .bc_func = bridge_ioctl_gmnelist32,
965 .bc_argsize = sizeof(struct ifbrmnelist32),
966 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
967 { .bc_func = bridge_ioctl_gifstats32,
968 .bc_argsize = sizeof(struct ifbrmreq32),
969 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
970 };
971
972 static const struct bridge_control bridge_control_table64[] = {
973 { .bc_func = bridge_ioctl_add, .bc_argsize = sizeof(struct ifbreq), /* 0 */
974 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
975 { .bc_func = bridge_ioctl_del, .bc_argsize = sizeof(struct ifbreq),
976 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
977
978 { .bc_func = bridge_ioctl_gifflags, .bc_argsize = sizeof(struct ifbreq),
979 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
980 { .bc_func = bridge_ioctl_sifflags, .bc_argsize = sizeof(struct ifbreq),
981 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
982
983 { .bc_func = bridge_ioctl_scache, .bc_argsize = sizeof(struct ifbrparam),
984 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
985 { .bc_func = bridge_ioctl_gcache, .bc_argsize = sizeof(struct ifbrparam),
986 .bc_flags = BC_F_COPYOUT },
987
988 { .bc_func = bridge_ioctl_gifs64, .bc_argsize = sizeof(struct ifbifconf64),
989 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
990 { .bc_func = bridge_ioctl_rts64, .bc_argsize = sizeof(struct ifbaconf64),
991 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
992
993 { .bc_func = bridge_ioctl_saddr64, .bc_argsize = sizeof(struct ifbareq64),
994 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
995
996 { .bc_func = bridge_ioctl_sto, .bc_argsize = sizeof(struct ifbrparam),
997 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
998 { .bc_func = bridge_ioctl_gto, .bc_argsize = sizeof(struct ifbrparam), /* 10 */
999 .bc_flags = BC_F_COPYOUT },
1000
1001 { .bc_func = bridge_ioctl_daddr64, .bc_argsize = sizeof(struct ifbareq64),
1002 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1003
1004 { .bc_func = bridge_ioctl_flush, .bc_argsize = sizeof(struct ifbreq),
1005 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1006
1007 { .bc_func = bridge_ioctl_gpri, .bc_argsize = sizeof(struct ifbrparam),
1008 .bc_flags = BC_F_COPYOUT },
1009 { .bc_func = bridge_ioctl_spri, .bc_argsize = sizeof(struct ifbrparam),
1010 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1011
1012 { .bc_func = bridge_ioctl_ght, .bc_argsize = sizeof(struct ifbrparam),
1013 .bc_flags = BC_F_COPYOUT },
1014 { .bc_func = bridge_ioctl_sht, .bc_argsize = sizeof(struct ifbrparam),
1015 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1016
1017 { .bc_func = bridge_ioctl_gfd, .bc_argsize = sizeof(struct ifbrparam),
1018 .bc_flags = BC_F_COPYOUT },
1019 { .bc_func = bridge_ioctl_sfd, .bc_argsize = sizeof(struct ifbrparam),
1020 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1021
1022 { .bc_func = bridge_ioctl_gma, .bc_argsize = sizeof(struct ifbrparam),
1023 .bc_flags = BC_F_COPYOUT },
1024 { .bc_func = bridge_ioctl_sma, .bc_argsize = sizeof(struct ifbrparam), /* 20 */
1025 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1026
1027 { .bc_func = bridge_ioctl_sifprio, .bc_argsize = sizeof(struct ifbreq),
1028 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1029
1030 { .bc_func = bridge_ioctl_sifcost, .bc_argsize = sizeof(struct ifbreq),
1031 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1032
1033 { .bc_func = bridge_ioctl_gfilt, .bc_argsize = sizeof(struct ifbrparam),
1034 .bc_flags = BC_F_COPYOUT },
1035 { .bc_func = bridge_ioctl_sfilt, .bc_argsize = sizeof(struct ifbrparam),
1036 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1037
1038 { .bc_func = bridge_ioctl_purge, .bc_argsize = sizeof(struct ifbreq),
1039 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1040
1041 { .bc_func = bridge_ioctl_addspan, .bc_argsize = sizeof(struct ifbreq),
1042 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1043 { .bc_func = bridge_ioctl_delspan, .bc_argsize = sizeof(struct ifbreq),
1044 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1045
1046 { .bc_func = bridge_ioctl_gbparam64, .bc_argsize = sizeof(struct ifbropreq64),
1047 .bc_flags = BC_F_COPYOUT },
1048
1049 { .bc_func = bridge_ioctl_grte, .bc_argsize = sizeof(struct ifbrparam),
1050 .bc_flags = BC_F_COPYOUT },
1051
1052 { .bc_func = bridge_ioctl_gifsstp64, .bc_argsize = sizeof(struct ifbpstpconf64), /* 30 */
1053 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
1054
1055 { .bc_func = bridge_ioctl_sproto, .bc_argsize = sizeof(struct ifbrparam),
1056 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1057
1058 { .bc_func = bridge_ioctl_stxhc, .bc_argsize = sizeof(struct ifbrparam),
1059 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1060
1061 { .bc_func = bridge_ioctl_sifmaxaddr, .bc_argsize = sizeof(struct ifbreq),
1062 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1063
1064 { .bc_func = bridge_ioctl_ghostfilter, .bc_argsize = sizeof(struct ifbrhostfilter),
1065 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
1066 { .bc_func = bridge_ioctl_shostfilter, .bc_argsize = sizeof(struct ifbrhostfilter),
1067 .bc_flags = BC_F_COPYIN | BC_F_SUSER },
1068
1069 { .bc_func = bridge_ioctl_gmnelist64,
1070 .bc_argsize = sizeof(struct ifbrmnelist64),
1071 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
1072 { .bc_func = bridge_ioctl_gifstats64,
1073 .bc_argsize = sizeof(struct ifbrmreq64),
1074 .bc_flags = BC_F_COPYIN | BC_F_COPYOUT },
1075 };
1076
1077 static const unsigned int bridge_control_table_size =
1078 sizeof(bridge_control_table32) / sizeof(bridge_control_table32[0]);
1079
1080 static LIST_HEAD(, bridge_softc) bridge_list =
1081 LIST_HEAD_INITIALIZER(bridge_list);
1082
1083 #define BRIDGENAME "bridge"
1084 #define BRIDGES_MAX IF_MAXUNIT
1085 #define BRIDGE_ZONE_MAX_ELEM MIN(IFNETS_MAX, BRIDGES_MAX)
1086
1087 static struct if_clone bridge_cloner =
1088 IF_CLONE_INITIALIZER(BRIDGENAME, bridge_clone_create, bridge_clone_destroy,
1089 0, BRIDGES_MAX);
1090
1091 static int if_bridge_txstart = 0;
1092 SYSCTL_INT(_net_link_bridge, OID_AUTO, txstart, CTLFLAG_RW | CTLFLAG_LOCKED,
1093 &if_bridge_txstart, 0, "Bridge interface uses TXSTART model");
1094
1095 SYSCTL_INT(_net_link_bridge, OID_AUTO, debug, CTLFLAG_RW | CTLFLAG_LOCKED,
1096 &if_bridge_debug, 0, "Bridge debug flags");
1097
1098 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_level,
1099 CTLFLAG_RW | CTLFLAG_LOCKED,
1100 &if_bridge_log_level, 0, "Bridge log level");
1101
1102 static int if_bridge_segmentation = 1;
1103 SYSCTL_INT(_net_link_bridge, OID_AUTO, segmentation,
1104 CTLFLAG_RW | CTLFLAG_LOCKED,
1105 &if_bridge_segmentation, 0, "Bridge interface enable segmentation");
1106
1107 static int if_bridge_vmnet_pf_tagging = 1;
1108 SYSCTL_INT(_net_link_bridge, OID_AUTO, vmnet_pf_tagging,
1109 CTLFLAG_RW | CTLFLAG_LOCKED,
1110 &if_bridge_segmentation, 0, "Bridge interface enable vmnet PF tagging");
1111
1112 #define BRIDGE_TSO_REDUCE_MSS_FORWARDING_MAX 256
1113 #define BRIDGE_TSO_REDUCE_MSS_FORWARDING_DEFAULT 110
1114 #define BRIDGE_TSO_REDUCE_MSS_TX_MAX 256
1115 #define BRIDGE_TSO_REDUCE_MSS_TX_DEFAULT 0
1116
1117 static u_int if_bridge_tso_reduce_mss_forwarding
1118 = BRIDGE_TSO_REDUCE_MSS_FORWARDING_DEFAULT;
1119 static u_int if_bridge_tso_reduce_mss_tx
1120 = BRIDGE_TSO_REDUCE_MSS_TX_DEFAULT;
1121
1122 static int
bridge_tso_reduce_mss(struct sysctl_req * req,u_int * val,u_int val_max)1123 bridge_tso_reduce_mss(struct sysctl_req *req, u_int * val, u_int val_max)
1124 {
1125 int changed;
1126 int error;
1127 u_int new_value;
1128
1129 error = sysctl_io_number(req, *val, sizeof(*val), &new_value,
1130 &changed);
1131 if (error == 0 && changed != 0) {
1132 if (new_value > val_max) {
1133 return EINVAL;
1134 }
1135 *val = new_value;
1136 }
1137 return error;
1138 }
1139
1140 static int
1141 bridge_tso_reduce_mss_forwarding_sysctl SYSCTL_HANDLER_ARGS
1142 {
1143 return bridge_tso_reduce_mss(req, &if_bridge_tso_reduce_mss_forwarding,
1144 BRIDGE_TSO_REDUCE_MSS_FORWARDING_MAX);
1145 }
1146
1147 SYSCTL_PROC(_net_link_bridge, OID_AUTO, tso_reduce_mss_forwarding,
1148 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,
1149 0, 0, bridge_tso_reduce_mss_forwarding_sysctl, "IU",
1150 "Bridge tso reduce mss when forwarding");
1151
1152 static int
1153 bridge_tso_reduce_mss_tx_sysctl SYSCTL_HANDLER_ARGS
1154 {
1155 return bridge_tso_reduce_mss(req, &if_bridge_tso_reduce_mss_tx,
1156 BRIDGE_TSO_REDUCE_MSS_TX_MAX);
1157 }
1158
1159 SYSCTL_PROC(_net_link_bridge, OID_AUTO, tso_reduce_mss_tx,
1160 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,
1161 0, 0, bridge_tso_reduce_mss_tx_sysctl, "IU",
1162 "Bridge tso reduce mss on transmit");
1163
1164
1165 #if DEBUG || DEVELOPMENT
1166 #define BRIDGE_FORCE_ONE 0x00000001
1167 #define BRIDGE_FORCE_TWO 0x00000002
1168 static u_int32_t if_bridge_force_errors = 0;
1169 SYSCTL_INT(_net_link_bridge, OID_AUTO, force_errors,
1170 CTLFLAG_RW | CTLFLAG_LOCKED,
1171 &if_bridge_force_errors, 0, "Bridge interface force errors");
1172 static inline bool
bridge_error_is_forced(u_int32_t flags)1173 bridge_error_is_forced(u_int32_t flags)
1174 {
1175 return (if_bridge_force_errors & flags) != 0;
1176 }
1177
1178 #define BRIDGE_ERROR_GET_FORCED(__is_forced, __flags) \
1179 do { \
1180 __is_forced = bridge_error_is_forced(__flags); \
1181 if (__is_forced) { \
1182 BRIDGE_LOG(LOG_NOTICE, 0, "0x%x forced", __flags); \
1183 } \
1184 } while (0)
1185 #endif /* DEBUG || DEVELOPMENT */
1186
1187
1188 static void brlog_ether_header(struct ether_header *);
1189 static void brlog_mbuf_data(mbuf_t, size_t, size_t);
1190 static void brlog_mbuf_pkthdr(mbuf_t, const char *, const char *);
1191 static void brlog_mbuf(mbuf_t, const char *, const char *);
1192 static void brlog_link(struct bridge_softc * sc);
1193
1194 #if BRIDGE_LOCK_DEBUG
1195 static void bridge_lock(struct bridge_softc *);
1196 static void bridge_unlock(struct bridge_softc *);
1197 static int bridge_lock2ref(struct bridge_softc *);
1198 static void bridge_unref(struct bridge_softc *);
1199 static void bridge_xlock(struct bridge_softc *);
1200 static void bridge_xdrop(struct bridge_softc *);
1201
1202 static void
bridge_lock(struct bridge_softc * sc)1203 bridge_lock(struct bridge_softc *sc)
1204 {
1205 void *lr_saved = __builtin_return_address(0);
1206
1207 BRIDGE_LOCK_ASSERT_NOTHELD(sc);
1208
1209 _BRIDGE_LOCK(sc);
1210
1211 sc->lock_lr[sc->next_lock_lr] = lr_saved;
1212 sc->next_lock_lr = (sc->next_lock_lr + 1) % SO_LCKDBG_MAX;
1213 }
1214
1215 static void
bridge_unlock(struct bridge_softc * sc)1216 bridge_unlock(struct bridge_softc *sc)
1217 {
1218 void *lr_saved = __builtin_return_address(0);
1219
1220 BRIDGE_LOCK_ASSERT_HELD(sc);
1221
1222 sc->unlock_lr[sc->next_unlock_lr] = lr_saved;
1223 sc->next_unlock_lr = (sc->next_unlock_lr + 1) % SO_LCKDBG_MAX;
1224
1225 _BRIDGE_UNLOCK(sc);
1226 }
1227
1228 static int
bridge_lock2ref(struct bridge_softc * sc)1229 bridge_lock2ref(struct bridge_softc *sc)
1230 {
1231 int error = 0;
1232 void *lr_saved = __builtin_return_address(0);
1233
1234 BRIDGE_LOCK_ASSERT_HELD(sc);
1235
1236 if (sc->sc_iflist_xcnt > 0) {
1237 error = EBUSY;
1238 } else {
1239 sc->sc_iflist_ref++;
1240 }
1241
1242 sc->unlock_lr[sc->next_unlock_lr] = lr_saved;
1243 sc->next_unlock_lr = (sc->next_unlock_lr + 1) % SO_LCKDBG_MAX;
1244
1245 _BRIDGE_UNLOCK(sc);
1246
1247 return error;
1248 }
1249
1250 static void
bridge_unref(struct bridge_softc * sc)1251 bridge_unref(struct bridge_softc *sc)
1252 {
1253 void *lr_saved = __builtin_return_address(0);
1254
1255 BRIDGE_LOCK_ASSERT_NOTHELD(sc);
1256
1257 _BRIDGE_LOCK(sc);
1258 sc->lock_lr[sc->next_lock_lr] = lr_saved;
1259 sc->next_lock_lr = (sc->next_lock_lr + 1) % SO_LCKDBG_MAX;
1260
1261 sc->sc_iflist_ref--;
1262
1263 sc->unlock_lr[sc->next_unlock_lr] = lr_saved;
1264 sc->next_unlock_lr = (sc->next_unlock_lr + 1) % SO_LCKDBG_MAX;
1265 if ((sc->sc_iflist_xcnt > 0) && (sc->sc_iflist_ref == 0)) {
1266 _BRIDGE_UNLOCK(sc);
1267 wakeup(&sc->sc_cv);
1268 } else {
1269 _BRIDGE_UNLOCK(sc);
1270 }
1271 }
1272
1273 static void
bridge_xlock(struct bridge_softc * sc)1274 bridge_xlock(struct bridge_softc *sc)
1275 {
1276 void *lr_saved = __builtin_return_address(0);
1277
1278 BRIDGE_LOCK_ASSERT_HELD(sc);
1279
1280 sc->sc_iflist_xcnt++;
1281 while (sc->sc_iflist_ref > 0) {
1282 sc->unlock_lr[sc->next_unlock_lr] = lr_saved;
1283 sc->next_unlock_lr = (sc->next_unlock_lr + 1) % SO_LCKDBG_MAX;
1284
1285 msleep(&sc->sc_cv, &sc->sc_mtx, PZERO, "BRIDGE_XLOCK", NULL);
1286
1287 sc->lock_lr[sc->next_lock_lr] = lr_saved;
1288 sc->next_lock_lr = (sc->next_lock_lr + 1) % SO_LCKDBG_MAX;
1289 }
1290 }
1291
1292 static void
bridge_xdrop(struct bridge_softc * sc)1293 bridge_xdrop(struct bridge_softc *sc)
1294 {
1295 BRIDGE_LOCK_ASSERT_HELD(sc);
1296
1297 sc->sc_iflist_xcnt--;
1298 }
1299
1300 #endif /* BRIDGE_LOCK_DEBUG */
1301
1302 static void
brlog_mbuf_pkthdr(mbuf_t m,const char * prefix,const char * suffix)1303 brlog_mbuf_pkthdr(mbuf_t m, const char *prefix, const char *suffix)
1304 {
1305 if (m) {
1306 BRIDGE_LOG_SIMPLE(LOG_NOTICE, 0,
1307 "%spktlen: %u rcvif: 0x%llx header: 0x%llx nextpkt: 0x%llx%s",
1308 prefix ? prefix : "", (unsigned int)mbuf_pkthdr_len(m),
1309 (uint64_t)VM_KERNEL_ADDRPERM(mbuf_pkthdr_rcvif(m)),
1310 (uint64_t)VM_KERNEL_ADDRPERM(mbuf_pkthdr_header(m)),
1311 (uint64_t)VM_KERNEL_ADDRPERM(mbuf_nextpkt(m)),
1312 suffix ? suffix : "");
1313 } else {
1314 BRIDGE_LOG_SIMPLE(LOG_NOTICE, 0, "%s<NULL>%s", prefix, suffix);
1315 }
1316 }
1317
1318 static void
brlog_mbuf(mbuf_t m,const char * prefix,const char * suffix)1319 brlog_mbuf(mbuf_t m, const char *prefix, const char *suffix)
1320 {
1321 if (m) {
1322 BRIDGE_LOG_SIMPLE(LOG_NOTICE, 0,
1323 "%s0x%llx type: %u flags: 0x%x len: %u data: 0x%llx "
1324 "maxlen: %u datastart: 0x%llx next: 0x%llx%s",
1325 prefix ? prefix : "", (uint64_t)VM_KERNEL_ADDRPERM(m),
1326 mbuf_type(m), mbuf_flags(m), (unsigned int)mbuf_len(m),
1327 (uint64_t)VM_KERNEL_ADDRPERM(mbuf_data(m)),
1328 (unsigned int)mbuf_maxlen(m),
1329 (uint64_t)VM_KERNEL_ADDRPERM(mbuf_datastart(m)),
1330 (uint64_t)VM_KERNEL_ADDRPERM(mbuf_next(m)),
1331 !suffix || (mbuf_flags(m) & MBUF_PKTHDR) ? "" : suffix);
1332 if ((mbuf_flags(m) & MBUF_PKTHDR)) {
1333 brlog_mbuf_pkthdr(m, "", suffix);
1334 }
1335 } else {
1336 BRIDGE_LOG_SIMPLE(LOG_NOTICE, 0, "%s<NULL>%s", prefix, suffix);
1337 }
1338 }
1339
1340 static void
brlog_mbuf_data(mbuf_t m,size_t offset,size_t len)1341 brlog_mbuf_data(mbuf_t m, size_t offset, size_t len)
1342 {
1343 mbuf_t n;
1344 size_t i, j;
1345 size_t pktlen, mlen, maxlen;
1346 unsigned char *ptr;
1347
1348 pktlen = mbuf_pkthdr_len(m);
1349
1350 if (offset > pktlen) {
1351 return;
1352 }
1353
1354 maxlen = (pktlen - offset > len) ? len : pktlen - offset;
1355 n = m;
1356 mlen = mbuf_len(n);
1357 ptr = mbuf_data(n);
1358 for (i = 0, j = 0; i < maxlen; i++, j++) {
1359 if (j >= mlen) {
1360 n = mbuf_next(n);
1361 if (n == 0) {
1362 break;
1363 }
1364 ptr = mbuf_data(n);
1365 mlen = mbuf_len(n);
1366 j = 0;
1367 }
1368 if (i >= offset) {
1369 BRIDGE_LOG_SIMPLE(LOG_NOTICE, 0,
1370 "%02x%s", ptr[j], i % 2 ? " " : "");
1371 }
1372 }
1373 }
1374
1375 static void
brlog_ether_header(struct ether_header * eh)1376 brlog_ether_header(struct ether_header *eh)
1377 {
1378 BRIDGE_LOG_SIMPLE(LOG_NOTICE, 0,
1379 "%02x:%02x:%02x:%02x:%02x:%02x > "
1380 "%02x:%02x:%02x:%02x:%02x:%02x 0x%04x ",
1381 eh->ether_shost[0], eh->ether_shost[1], eh->ether_shost[2],
1382 eh->ether_shost[3], eh->ether_shost[4], eh->ether_shost[5],
1383 eh->ether_dhost[0], eh->ether_dhost[1], eh->ether_dhost[2],
1384 eh->ether_dhost[3], eh->ether_dhost[4], eh->ether_dhost[5],
1385 ntohs(eh->ether_type));
1386 }
1387
1388 static char *
ether_ntop(char * buf,size_t len,const u_char * ap)1389 ether_ntop(char *buf, size_t len, const u_char *ap)
1390 {
1391 snprintf(buf, len, "%02x:%02x:%02x:%02x:%02x:%02x",
1392 ap[0], ap[1], ap[2], ap[3], ap[4], ap[5]);
1393
1394 return buf;
1395 }
1396
1397 static void
brlog_link(struct bridge_softc * sc)1398 brlog_link(struct bridge_softc * sc)
1399 {
1400 int i;
1401 uint32_t sdl_buffer[offsetof(struct sockaddr_dl, sdl_data) +
1402 IFNAMSIZ + ETHER_ADDR_LEN];
1403 struct sockaddr_dl *sdl = (struct sockaddr_dl *)sdl_buffer;
1404 const u_char * lladdr;
1405 char lladdr_str[48];
1406
1407 memset(sdl, 0, sizeof(sdl_buffer));
1408 sdl->sdl_family = AF_LINK;
1409 sdl->sdl_nlen = strlen(sc->sc_if_xname);
1410 sdl->sdl_alen = ETHER_ADDR_LEN;
1411 sdl->sdl_len = offsetof(struct sockaddr_dl, sdl_data);
1412 memcpy(sdl->sdl_data, sc->sc_if_xname, sdl->sdl_nlen);
1413 memcpy(LLADDR(sdl), sc->sc_defaddr, ETHER_ADDR_LEN);
1414 lladdr_str[0] = '\0';
1415 for (i = 0, lladdr = CONST_LLADDR(sdl);
1416 i < sdl->sdl_alen;
1417 i++, lladdr++) {
1418 char byte_str[4];
1419
1420 snprintf(byte_str, sizeof(byte_str), "%s%x", i ? ":" : "",
1421 *lladdr);
1422 strlcat(lladdr_str, byte_str, sizeof(lladdr_str));
1423 }
1424 BRIDGE_LOG_SIMPLE(LOG_NOTICE, 0,
1425 "%s sdl len %d index %d family %d type 0x%x nlen %d alen %d"
1426 " slen %d addr %s", sc->sc_if_xname,
1427 sdl->sdl_len, sdl->sdl_index,
1428 sdl->sdl_family, sdl->sdl_type, sdl->sdl_nlen,
1429 sdl->sdl_alen, sdl->sdl_slen, lladdr_str);
1430 }
1431
1432
1433 /*
1434 * bridgeattach:
1435 *
1436 * Pseudo-device attach routine.
1437 */
1438 __private_extern__ int
bridgeattach(int n)1439 bridgeattach(int n)
1440 {
1441 #pragma unused(n)
1442 int error;
1443
1444 LIST_INIT(&bridge_list);
1445
1446 #if BRIDGESTP
1447 bstp_sys_init();
1448 #endif /* BRIDGESTP */
1449
1450 error = if_clone_attach(&bridge_cloner);
1451 if (error != 0) {
1452 BRIDGE_LOG(LOG_NOTICE, 0, "ifnet_clone_attach failed %d", error);
1453 }
1454 return error;
1455 }
1456
1457
1458 static errno_t
bridge_ifnet_set_attrs(struct ifnet * ifp)1459 bridge_ifnet_set_attrs(struct ifnet * ifp)
1460 {
1461 errno_t error;
1462
1463 error = ifnet_set_mtu(ifp, ETHERMTU);
1464 if (error != 0) {
1465 BRIDGE_LOG(LOG_NOTICE, 0, "ifnet_set_mtu failed %d", error);
1466 goto done;
1467 }
1468 error = ifnet_set_addrlen(ifp, ETHER_ADDR_LEN);
1469 if (error != 0) {
1470 BRIDGE_LOG(LOG_NOTICE, 0, "ifnet_set_addrlen failed %d", error);
1471 goto done;
1472 }
1473 error = ifnet_set_hdrlen(ifp, ETHER_HDR_LEN);
1474 if (error != 0) {
1475 BRIDGE_LOG(LOG_NOTICE, 0, "ifnet_set_hdrlen failed %d", error);
1476 goto done;
1477 }
1478 error = ifnet_set_flags(ifp,
1479 IFF_BROADCAST | IFF_SIMPLEX | IFF_NOTRAILERS | IFF_MULTICAST,
1480 0xffff);
1481
1482 if (error != 0) {
1483 BRIDGE_LOG(LOG_NOTICE, 0, "ifnet_set_flags failed %d", error);
1484 goto done;
1485 }
1486 done:
1487 return error;
1488 }
1489
1490 /*
1491 * bridge_clone_create:
1492 *
1493 * Create a new bridge instance.
1494 */
1495 static int
bridge_clone_create(struct if_clone * ifc,uint32_t unit,void * params)1496 bridge_clone_create(struct if_clone *ifc, uint32_t unit, void *params)
1497 {
1498 #pragma unused(params)
1499 struct ifnet *ifp = NULL;
1500 struct bridge_softc *sc = NULL;
1501 struct bridge_softc *sc2 = NULL;
1502 struct ifnet_init_eparams init_params;
1503 errno_t error = 0;
1504 uint8_t eth_hostid[ETHER_ADDR_LEN];
1505 int fb, retry, has_hostid;
1506
1507 sc = kalloc_type(struct bridge_softc, Z_WAITOK_ZERO_NOFAIL);
1508 lck_mtx_init(&sc->sc_mtx, &bridge_lock_grp, &bridge_lock_attr);
1509 sc->sc_brtmax = BRIDGE_RTABLE_MAX;
1510 sc->sc_mne_max = BRIDGE_MAC_NAT_ENTRY_MAX;
1511 sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
1512 sc->sc_filter_flags = 0;
1513
1514 TAILQ_INIT(&sc->sc_iflist);
1515
1516 /* use the interface name as the unique id for ifp recycle */
1517 snprintf(sc->sc_if_xname, sizeof(sc->sc_if_xname), "%s%d",
1518 ifc->ifc_name, unit);
1519 bzero(&init_params, sizeof(init_params));
1520 init_params.ver = IFNET_INIT_CURRENT_VERSION;
1521 init_params.len = sizeof(init_params);
1522 /* Initialize our routing table. */
1523 error = bridge_rtable_init(sc);
1524 if (error != 0) {
1525 BRIDGE_LOG(LOG_NOTICE, 0, "bridge_rtable_init failed %d", error);
1526 goto done;
1527 }
1528 TAILQ_INIT(&sc->sc_spanlist);
1529 if (if_bridge_txstart) {
1530 init_params.start = bridge_start;
1531 } else {
1532 init_params.flags = IFNET_INIT_LEGACY;
1533 init_params.output = bridge_output;
1534 }
1535 init_params.set_bpf_tap = bridge_set_bpf_tap;
1536 init_params.uniqueid = sc->sc_if_xname;
1537 init_params.uniqueid_len = strlen(sc->sc_if_xname);
1538 init_params.sndq_maxlen = IFQ_MAXLEN;
1539 init_params.name = ifc->ifc_name;
1540 init_params.unit = unit;
1541 init_params.family = IFNET_FAMILY_ETHERNET;
1542 init_params.type = IFT_BRIDGE;
1543 init_params.demux = ether_demux;
1544 init_params.add_proto = ether_add_proto;
1545 init_params.del_proto = ether_del_proto;
1546 init_params.check_multi = ether_check_multi;
1547 init_params.framer_extended = ether_frameout_extended;
1548 init_params.softc = sc;
1549 init_params.ioctl = bridge_ioctl;
1550 init_params.detach = bridge_detach;
1551 init_params.broadcast_addr = etherbroadcastaddr;
1552 init_params.broadcast_len = ETHER_ADDR_LEN;
1553
1554 error = ifnet_allocate_extended(&init_params, &ifp);
1555 if (error != 0) {
1556 BRIDGE_LOG(LOG_NOTICE, 0, "ifnet_allocate failed %d", error);
1557 goto done;
1558 }
1559 LIST_INIT(&sc->sc_mne_list);
1560 LIST_INIT(&sc->sc_mne_list_v6);
1561 sc->sc_ifp = ifp;
1562 error = bridge_ifnet_set_attrs(ifp);
1563 if (error != 0) {
1564 BRIDGE_LOG(LOG_NOTICE, 0, "bridge_ifnet_set_attrs failed %d",
1565 error);
1566 goto done;
1567 }
1568 /*
1569 * Generate an ethernet address with a locally administered address.
1570 *
1571 * Since we are using random ethernet addresses for the bridge, it is
1572 * possible that we might have address collisions, so make sure that
1573 * this hardware address isn't already in use on another bridge.
1574 * The first try uses the "hostid" and falls back to read_frandom();
1575 * for "hostid", we use the MAC address of the first-encountered
1576 * Ethernet-type interface that is currently configured.
1577 */
1578 fb = 0;
1579 has_hostid = (uuid_get_ethernet(ð_hostid[0]) == 0);
1580 for (retry = 1; retry != 0;) {
1581 if (fb || has_hostid == 0) {
1582 read_frandom(&sc->sc_defaddr, ETHER_ADDR_LEN);
1583 sc->sc_defaddr[0] &= ~1; /* clear multicast bit */
1584 sc->sc_defaddr[0] |= 2; /* set the LAA bit */
1585 } else {
1586 bcopy(ð_hostid[0], &sc->sc_defaddr,
1587 ETHER_ADDR_LEN);
1588 sc->sc_defaddr[0] &= ~1; /* clear multicast bit */
1589 sc->sc_defaddr[0] |= 2; /* set the LAA bit */
1590 sc->sc_defaddr[3] = /* stir it up a bit */
1591 ((sc->sc_defaddr[3] & 0x0f) << 4) |
1592 ((sc->sc_defaddr[3] & 0xf0) >> 4);
1593 /*
1594 * Mix in the LSB as it's actually pretty significant,
1595 * see rdar://14076061
1596 */
1597 sc->sc_defaddr[4] =
1598 (((sc->sc_defaddr[4] & 0x0f) << 4) |
1599 ((sc->sc_defaddr[4] & 0xf0) >> 4)) ^
1600 sc->sc_defaddr[5];
1601 sc->sc_defaddr[5] = ifp->if_unit & 0xff;
1602 }
1603
1604 fb = 1;
1605 retry = 0;
1606 lck_mtx_lock(&bridge_list_mtx);
1607 LIST_FOREACH(sc2, &bridge_list, sc_list) {
1608 if (_ether_cmp(sc->sc_defaddr,
1609 IF_LLADDR(sc2->sc_ifp)) == 0) {
1610 retry = 1;
1611 }
1612 }
1613 lck_mtx_unlock(&bridge_list_mtx);
1614 }
1615
1616 sc->sc_flags &= ~SCF_MEDIA_ACTIVE;
1617
1618 if (BRIDGE_DBGF_ENABLED(BR_DBGF_LIFECYCLE)) {
1619 brlog_link(sc);
1620 }
1621 error = ifnet_attach(ifp, NULL);
1622 if (error != 0) {
1623 BRIDGE_LOG(LOG_NOTICE, 0, "ifnet_attach failed %d", error);
1624 goto done;
1625 }
1626
1627 error = ifnet_set_lladdr_and_type(ifp, sc->sc_defaddr, ETHER_ADDR_LEN,
1628 IFT_ETHER);
1629 if (error != 0) {
1630 BRIDGE_LOG(LOG_NOTICE, 0, "ifnet_set_lladdr_and_type failed %d",
1631 error);
1632 goto done;
1633 }
1634
1635 ifnet_set_offload(ifp,
1636 IFNET_CSUM_IP | IFNET_CSUM_TCP | IFNET_CSUM_UDP |
1637 IFNET_CSUM_TCPIPV6 | IFNET_CSUM_UDPIPV6 | IFNET_MULTIPAGES);
1638 error = bridge_set_tso(sc);
1639 if (error != 0) {
1640 BRIDGE_LOG(LOG_NOTICE, 0, "bridge_set_tso failed %d", error);
1641 goto done;
1642 }
1643 #if BRIDGESTP
1644 bstp_attach(&sc->sc_stp, &bridge_ops);
1645 #endif /* BRIDGESTP */
1646
1647 lck_mtx_lock(&bridge_list_mtx);
1648 LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
1649 lck_mtx_unlock(&bridge_list_mtx);
1650
1651 /* attach as ethernet */
1652 error = bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header),
1653 NULL, NULL);
1654
1655 done:
1656 if (error != 0) {
1657 BRIDGE_LOG(LOG_NOTICE, 0, "failed error %d", error);
1658 /* TBD: Clean up: sc, sc_rthash etc */
1659 }
1660
1661 return error;
1662 }
1663
1664 /*
1665 * bridge_clone_destroy:
1666 *
1667 * Destroy a bridge instance.
1668 */
1669 static int
bridge_clone_destroy(struct ifnet * ifp)1670 bridge_clone_destroy(struct ifnet *ifp)
1671 {
1672 struct bridge_softc *sc = ifp->if_softc;
1673 struct bridge_iflist *bif;
1674 errno_t error;
1675
1676 BRIDGE_LOCK(sc);
1677 if ((sc->sc_flags & SCF_DETACHING)) {
1678 BRIDGE_UNLOCK(sc);
1679 return 0;
1680 }
1681 sc->sc_flags |= SCF_DETACHING;
1682
1683 bridge_ifstop(ifp, 1);
1684
1685 bridge_cancel_delayed_call(&sc->sc_resize_call);
1686
1687 bridge_cleanup_delayed_call(&sc->sc_resize_call);
1688 bridge_cleanup_delayed_call(&sc->sc_aging_timer);
1689
1690 error = ifnet_set_flags(ifp, 0, IFF_UP);
1691 if (error != 0) {
1692 BRIDGE_LOG(LOG_NOTICE, 0, "ifnet_set_flags failed %d", error);
1693 }
1694
1695 while ((bif = TAILQ_FIRST(&sc->sc_iflist)) != NULL) {
1696 bridge_delete_member(sc, bif);
1697 }
1698
1699 while ((bif = TAILQ_FIRST(&sc->sc_spanlist)) != NULL) {
1700 bridge_delete_span(sc, bif);
1701 }
1702 BRIDGE_UNLOCK(sc);
1703
1704 error = ifnet_detach(ifp);
1705 if (error != 0) {
1706 panic("%s (%d): ifnet_detach(%p) failed %d",
1707 __func__, __LINE__, ifp, error);
1708 }
1709 return 0;
1710 }
1711
1712 #define DRVSPEC do { \
1713 if (ifd->ifd_cmd >= bridge_control_table_size) { \
1714 error = EINVAL; \
1715 break; \
1716 } \
1717 bc = &bridge_control_table[ifd->ifd_cmd]; \
1718 \
1719 if (cmd == SIOCGDRVSPEC && \
1720 (bc->bc_flags & BC_F_COPYOUT) == 0) { \
1721 error = EINVAL; \
1722 break; \
1723 } else if (cmd == SIOCSDRVSPEC && \
1724 (bc->bc_flags & BC_F_COPYOUT) != 0) { \
1725 error = EINVAL; \
1726 break; \
1727 } \
1728 \
1729 if (bc->bc_flags & BC_F_SUSER) { \
1730 error = kauth_authorize_generic(kauth_cred_get(), \
1731 KAUTH_GENERIC_ISSUSER); \
1732 if (error) \
1733 break; \
1734 } \
1735 \
1736 if (ifd->ifd_len != bc->bc_argsize || \
1737 ifd->ifd_len > sizeof (args)) { \
1738 error = EINVAL; \
1739 break; \
1740 } \
1741 \
1742 bzero(&args, sizeof (args)); \
1743 if (bc->bc_flags & BC_F_COPYIN) { \
1744 error = copyin(ifd->ifd_data, &args, ifd->ifd_len); \
1745 if (error) \
1746 break; \
1747 } \
1748 \
1749 BRIDGE_LOCK(sc); \
1750 error = (*bc->bc_func)(sc, &args); \
1751 BRIDGE_UNLOCK(sc); \
1752 if (error) \
1753 break; \
1754 \
1755 if (bc->bc_flags & BC_F_COPYOUT) \
1756 error = copyout(&args, ifd->ifd_data, ifd->ifd_len); \
1757 } while (0)
1758
1759 static boolean_t
interface_needs_input_broadcast(struct ifnet * ifp)1760 interface_needs_input_broadcast(struct ifnet * ifp)
1761 {
1762 /*
1763 * Selectively enable input broadcast only when necessary.
1764 * The bridge interface itself attaches a fake protocol
1765 * so checking for at least two protocols means that the
1766 * interface is being used for something besides bridging
1767 * and needs to see broadcast packets from other members.
1768 */
1769 return if_get_protolist(ifp, NULL, 0) >= 2;
1770 }
1771
1772 static boolean_t
bif_set_input_broadcast(struct bridge_iflist * bif,boolean_t input_broadcast)1773 bif_set_input_broadcast(struct bridge_iflist * bif, boolean_t input_broadcast)
1774 {
1775 boolean_t old_input_broadcast;
1776
1777 old_input_broadcast = (bif->bif_flags & BIFF_INPUT_BROADCAST) != 0;
1778 if (input_broadcast) {
1779 bif->bif_flags |= BIFF_INPUT_BROADCAST;
1780 } else {
1781 bif->bif_flags &= ~BIFF_INPUT_BROADCAST;
1782 }
1783 return old_input_broadcast != input_broadcast;
1784 }
1785
1786 /*
1787 * bridge_ioctl:
1788 *
1789 * Handle a control request from the operator.
1790 */
1791 static errno_t
bridge_ioctl(struct ifnet * ifp,u_long cmd,void * data)1792 bridge_ioctl(struct ifnet *ifp, u_long cmd, void *data)
1793 {
1794 struct bridge_softc *sc = ifp->if_softc;
1795 struct ifreq *ifr = (struct ifreq *)data;
1796 struct bridge_iflist *bif;
1797 int error = 0;
1798
1799 BRIDGE_LOCK_ASSERT_NOTHELD(sc);
1800
1801 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_IOCTL,
1802 "ifp %s cmd 0x%08lx (%c%c [%lu] %c %lu)",
1803 ifp->if_xname, cmd, (cmd & IOC_IN) ? 'I' : ' ',
1804 (cmd & IOC_OUT) ? 'O' : ' ', IOCPARM_LEN(cmd),
1805 (char)IOCGROUP(cmd), cmd & 0xff);
1806
1807 switch (cmd) {
1808 case SIOCSIFADDR:
1809 case SIOCAIFADDR:
1810 ifnet_set_flags(ifp, IFF_UP, IFF_UP);
1811 break;
1812
1813 case SIOCGIFMEDIA32:
1814 case SIOCGIFMEDIA64: {
1815 struct ifmediareq *ifmr = (struct ifmediareq *)data;
1816 user_addr_t user_addr;
1817
1818 user_addr = (cmd == SIOCGIFMEDIA64) ?
1819 ((struct ifmediareq64 *)ifmr)->ifmu_ulist :
1820 CAST_USER_ADDR_T(((struct ifmediareq32 *)ifmr)->ifmu_ulist);
1821
1822 ifmr->ifm_status = IFM_AVALID;
1823 ifmr->ifm_mask = 0;
1824 ifmr->ifm_count = 1;
1825
1826 BRIDGE_LOCK(sc);
1827 if (!(sc->sc_flags & SCF_DETACHING) &&
1828 (sc->sc_flags & SCF_MEDIA_ACTIVE)) {
1829 ifmr->ifm_status |= IFM_ACTIVE;
1830 ifmr->ifm_active = ifmr->ifm_current =
1831 IFM_ETHER | IFM_AUTO;
1832 } else {
1833 ifmr->ifm_active = ifmr->ifm_current = IFM_NONE;
1834 }
1835 BRIDGE_UNLOCK(sc);
1836
1837 if (user_addr != USER_ADDR_NULL) {
1838 error = copyout(&ifmr->ifm_current, user_addr,
1839 sizeof(int));
1840 }
1841 break;
1842 }
1843
1844 case SIOCADDMULTI:
1845 case SIOCDELMULTI:
1846 break;
1847
1848 case SIOCSDRVSPEC32:
1849 case SIOCGDRVSPEC32: {
1850 union {
1851 struct ifbreq ifbreq;
1852 struct ifbifconf32 ifbifconf;
1853 struct ifbareq32 ifbareq;
1854 struct ifbaconf32 ifbaconf;
1855 struct ifbrparam ifbrparam;
1856 struct ifbropreq32 ifbropreq;
1857 } args;
1858 struct ifdrv32 *ifd = (struct ifdrv32 *)data;
1859 const struct bridge_control *bridge_control_table =
1860 bridge_control_table32, *bc;
1861
1862 DRVSPEC;
1863
1864 break;
1865 }
1866 case SIOCSDRVSPEC64:
1867 case SIOCGDRVSPEC64: {
1868 union {
1869 struct ifbreq ifbreq;
1870 struct ifbifconf64 ifbifconf;
1871 struct ifbareq64 ifbareq;
1872 struct ifbaconf64 ifbaconf;
1873 struct ifbrparam ifbrparam;
1874 struct ifbropreq64 ifbropreq;
1875 } args;
1876 struct ifdrv64 *ifd = (struct ifdrv64 *)data;
1877 const struct bridge_control *bridge_control_table =
1878 bridge_control_table64, *bc;
1879
1880 DRVSPEC;
1881
1882 break;
1883 }
1884
1885 case SIOCSIFFLAGS:
1886 if (!(ifp->if_flags & IFF_UP) &&
1887 (ifp->if_flags & IFF_RUNNING)) {
1888 /*
1889 * If interface is marked down and it is running,
1890 * then stop and disable it.
1891 */
1892 BRIDGE_LOCK(sc);
1893 bridge_ifstop(ifp, 1);
1894 BRIDGE_UNLOCK(sc);
1895 } else if ((ifp->if_flags & IFF_UP) &&
1896 !(ifp->if_flags & IFF_RUNNING)) {
1897 /*
1898 * If interface is marked up and it is stopped, then
1899 * start it.
1900 */
1901 BRIDGE_LOCK(sc);
1902 error = bridge_init(ifp);
1903 BRIDGE_UNLOCK(sc);
1904 }
1905 break;
1906
1907 case SIOCSIFLLADDR:
1908 error = ifnet_set_lladdr(ifp, ifr->ifr_addr.sa_data,
1909 ifr->ifr_addr.sa_len);
1910 if (error != 0) {
1911 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_IOCTL,
1912 "%s SIOCSIFLLADDR error %d", ifp->if_xname,
1913 error);
1914 }
1915 break;
1916
1917 case SIOCSIFMTU:
1918 if (ifr->ifr_mtu < 576) {
1919 error = EINVAL;
1920 break;
1921 }
1922 BRIDGE_LOCK(sc);
1923 if (TAILQ_EMPTY(&sc->sc_iflist)) {
1924 sc->sc_ifp->if_mtu = ifr->ifr_mtu;
1925 BRIDGE_UNLOCK(sc);
1926 break;
1927 }
1928 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) {
1929 if (bif->bif_ifp->if_mtu != (unsigned)ifr->ifr_mtu) {
1930 BRIDGE_LOG(LOG_NOTICE, 0,
1931 "%s invalid MTU: %u(%s) != %d",
1932 sc->sc_ifp->if_xname,
1933 bif->bif_ifp->if_mtu,
1934 bif->bif_ifp->if_xname, ifr->ifr_mtu);
1935 error = EINVAL;
1936 break;
1937 }
1938 }
1939 if (!error) {
1940 sc->sc_ifp->if_mtu = ifr->ifr_mtu;
1941 }
1942 BRIDGE_UNLOCK(sc);
1943 break;
1944
1945 default:
1946 error = ether_ioctl(ifp, cmd, data);
1947 if (error != 0 && error != EOPNOTSUPP) {
1948 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_IOCTL,
1949 "ifp %s cmd 0x%08lx "
1950 "(%c%c [%lu] %c %lu) failed error: %d",
1951 ifp->if_xname, cmd,
1952 (cmd & IOC_IN) ? 'I' : ' ',
1953 (cmd & IOC_OUT) ? 'O' : ' ',
1954 IOCPARM_LEN(cmd), (char)IOCGROUP(cmd),
1955 cmd & 0xff, error);
1956 }
1957 break;
1958 }
1959 BRIDGE_LOCK_ASSERT_NOTHELD(sc);
1960
1961 return error;
1962 }
1963
1964 #if HAS_IF_CAP
1965 /*
1966 * bridge_mutecaps:
1967 *
1968 * Clear or restore unwanted capabilities on the member interface
1969 */
1970 static void
bridge_mutecaps(struct bridge_softc * sc)1971 bridge_mutecaps(struct bridge_softc *sc)
1972 {
1973 struct bridge_iflist *bif;
1974 int enabled, mask;
1975
1976 /* Initial bitmask of capabilities to test */
1977 mask = BRIDGE_IFCAPS_MASK;
1978
1979 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) {
1980 /* Every member must support it or its disabled */
1981 mask &= bif->bif_savedcaps;
1982 }
1983
1984 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) {
1985 enabled = bif->bif_ifp->if_capenable;
1986 enabled &= ~BRIDGE_IFCAPS_STRIP;
1987 /* strip off mask bits and enable them again if allowed */
1988 enabled &= ~BRIDGE_IFCAPS_MASK;
1989 enabled |= mask;
1990
1991 bridge_set_ifcap(sc, bif, enabled);
1992 }
1993 }
1994
1995 static void
bridge_set_ifcap(struct bridge_softc * sc,struct bridge_iflist * bif,int set)1996 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
1997 {
1998 struct ifnet *ifp = bif->bif_ifp;
1999 struct ifreq ifr;
2000 int error;
2001
2002 bzero(&ifr, sizeof(ifr));
2003 ifr.ifr_reqcap = set;
2004
2005 if (ifp->if_capenable != set) {
2006 IFF_LOCKGIANT(ifp);
2007 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
2008 IFF_UNLOCKGIANT(ifp);
2009 if (error) {
2010 BRIDGE_LOG(LOG_NOTICE, 0,
2011 "%s error setting interface capabilities on %s",
2012 sc->sc_ifp->if_xname, ifp->if_xname);
2013 }
2014 }
2015 }
2016 #endif /* HAS_IF_CAP */
2017
2018 static errno_t
bridge_set_tso(struct bridge_softc * sc)2019 bridge_set_tso(struct bridge_softc *sc)
2020 {
2021 struct bridge_iflist *bif;
2022 u_int32_t tso_v4_mtu;
2023 u_int32_t tso_v6_mtu;
2024 ifnet_offload_t offload;
2025 errno_t error = 0;
2026
2027 /* By default, support TSO */
2028 offload = sc->sc_ifp->if_hwassist | IFNET_TSO_IPV4 | IFNET_TSO_IPV6;
2029 tso_v4_mtu = IP_MAXPACKET;
2030 tso_v6_mtu = IP_MAXPACKET;
2031
2032 /* Use the lowest common denominator of the members */
2033 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) {
2034 ifnet_t ifp = bif->bif_ifp;
2035
2036 if (ifp == NULL) {
2037 continue;
2038 }
2039
2040 if (offload & IFNET_TSO_IPV4) {
2041 if (ifp->if_hwassist & IFNET_TSO_IPV4) {
2042 if (tso_v4_mtu > ifp->if_tso_v4_mtu) {
2043 tso_v4_mtu = ifp->if_tso_v4_mtu;
2044 }
2045 } else {
2046 offload &= ~IFNET_TSO_IPV4;
2047 tso_v4_mtu = 0;
2048 }
2049 }
2050 if (offload & IFNET_TSO_IPV6) {
2051 if (ifp->if_hwassist & IFNET_TSO_IPV6) {
2052 if (tso_v6_mtu > ifp->if_tso_v6_mtu) {
2053 tso_v6_mtu = ifp->if_tso_v6_mtu;
2054 }
2055 } else {
2056 offload &= ~IFNET_TSO_IPV6;
2057 tso_v6_mtu = 0;
2058 }
2059 }
2060 }
2061
2062 if (offload != sc->sc_ifp->if_hwassist) {
2063 error = ifnet_set_offload(sc->sc_ifp, offload);
2064 if (error != 0) {
2065 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_LIFECYCLE,
2066 "ifnet_set_offload(%s, 0x%x) failed %d",
2067 sc->sc_ifp->if_xname, offload, error);
2068 goto done;
2069 }
2070 /*
2071 * For ifnet_set_tso_mtu() sake, the TSO MTU must be at least
2072 * as large as the interface MTU
2073 */
2074 if (sc->sc_ifp->if_hwassist & IFNET_TSO_IPV4) {
2075 if (tso_v4_mtu < sc->sc_ifp->if_mtu) {
2076 tso_v4_mtu = sc->sc_ifp->if_mtu;
2077 }
2078 error = ifnet_set_tso_mtu(sc->sc_ifp, AF_INET,
2079 tso_v4_mtu);
2080 if (error != 0) {
2081 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_LIFECYCLE,
2082 "ifnet_set_tso_mtu(%s, "
2083 "AF_INET, %u) failed %d",
2084 sc->sc_ifp->if_xname,
2085 tso_v4_mtu, error);
2086 goto done;
2087 }
2088 }
2089 if (sc->sc_ifp->if_hwassist & IFNET_TSO_IPV6) {
2090 if (tso_v6_mtu < sc->sc_ifp->if_mtu) {
2091 tso_v6_mtu = sc->sc_ifp->if_mtu;
2092 }
2093 error = ifnet_set_tso_mtu(sc->sc_ifp, AF_INET6,
2094 tso_v6_mtu);
2095 if (error != 0) {
2096 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_LIFECYCLE,
2097 "ifnet_set_tso_mtu(%s, "
2098 "AF_INET6, %u) failed %d",
2099 sc->sc_ifp->if_xname,
2100 tso_v6_mtu, error);
2101 goto done;
2102 }
2103 }
2104 }
2105 done:
2106 return error;
2107 }
2108
2109 /*
2110 * bridge_lookup_member:
2111 *
2112 * Lookup a bridge member interface.
2113 */
2114 static struct bridge_iflist *
bridge_lookup_member(struct bridge_softc * sc,const char * name)2115 bridge_lookup_member(struct bridge_softc *sc, const char *name)
2116 {
2117 struct bridge_iflist *bif;
2118 struct ifnet *ifp;
2119
2120 BRIDGE_LOCK_ASSERT_HELD(sc);
2121
2122 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) {
2123 ifp = bif->bif_ifp;
2124 if (strcmp(ifp->if_xname, name) == 0) {
2125 return bif;
2126 }
2127 }
2128
2129 return NULL;
2130 }
2131
2132 /*
2133 * bridge_lookup_member_if:
2134 *
2135 * Lookup a bridge member interface by ifnet*.
2136 */
2137 static struct bridge_iflist *
bridge_lookup_member_if(struct bridge_softc * sc,struct ifnet * member_ifp)2138 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
2139 {
2140 struct bridge_iflist *bif;
2141
2142 BRIDGE_LOCK_ASSERT_HELD(sc);
2143
2144 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) {
2145 if (bif->bif_ifp == member_ifp) {
2146 return bif;
2147 }
2148 }
2149
2150 return NULL;
2151 }
2152
2153 static errno_t
bridge_iff_input(void * cookie,ifnet_t ifp,protocol_family_t protocol,mbuf_t * data,char ** frame_ptr)2154 bridge_iff_input(void *cookie, ifnet_t ifp, protocol_family_t protocol,
2155 mbuf_t *data, char **frame_ptr)
2156 {
2157 #pragma unused(protocol)
2158 errno_t error = 0;
2159 struct bridge_iflist *bif = (struct bridge_iflist *)cookie;
2160 struct bridge_softc *sc = bif->bif_sc;
2161 int included = 0;
2162 size_t frmlen = 0;
2163 mbuf_t m = *data;
2164
2165 if ((m->m_flags & M_PROTO1)) {
2166 goto out;
2167 }
2168
2169 if (*frame_ptr >= (char *)mbuf_datastart(m) &&
2170 *frame_ptr <= (char *)mbuf_data(m)) {
2171 included = 1;
2172 frmlen = (char *)mbuf_data(m) - *frame_ptr;
2173 }
2174 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT,
2175 "%s from %s m 0x%llx data 0x%llx frame 0x%llx %s "
2176 "frmlen %lu", sc->sc_ifp->if_xname,
2177 ifp->if_xname, (uint64_t)VM_KERNEL_ADDRPERM(m),
2178 (uint64_t)VM_KERNEL_ADDRPERM(mbuf_data(m)),
2179 (uint64_t)VM_KERNEL_ADDRPERM(*frame_ptr),
2180 included ? "inside" : "outside", frmlen);
2181 if (BRIDGE_DBGF_ENABLED(BR_DBGF_MBUF)) {
2182 brlog_mbuf(m, "bridge_iff_input[", "");
2183 brlog_ether_header((struct ether_header *)
2184 (void *)*frame_ptr);
2185 brlog_mbuf_data(m, 0, 20);
2186 }
2187 if (included == 0) {
2188 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT, "frame_ptr outside mbuf");
2189 goto out;
2190 }
2191
2192 /* Move data pointer to start of frame to the link layer header */
2193 (void) mbuf_setdata(m, (char *)mbuf_data(m) - frmlen,
2194 mbuf_len(m) + frmlen);
2195 (void) mbuf_pkthdr_adjustlen(m, frmlen);
2196
2197 /* make sure we can access the ethernet header */
2198 if (mbuf_pkthdr_len(m) < sizeof(struct ether_header)) {
2199 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT,
2200 "short frame %lu < %lu",
2201 mbuf_pkthdr_len(m), sizeof(struct ether_header));
2202 goto out;
2203 }
2204 if (mbuf_len(m) < sizeof(struct ether_header)) {
2205 error = mbuf_pullup(data, sizeof(struct ether_header));
2206 if (error != 0) {
2207 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT,
2208 "mbuf_pullup(%lu) failed %d",
2209 sizeof(struct ether_header),
2210 error);
2211 error = EJUSTRETURN;
2212 goto out;
2213 }
2214 if (m != *data) {
2215 m = *data;
2216 *frame_ptr = mbuf_data(m);
2217 }
2218 }
2219
2220 error = bridge_input(ifp, data);
2221
2222 /* Adjust packet back to original */
2223 if (error == 0) {
2224 /* bridge_input might have modified *data */
2225 if (*data != m) {
2226 m = *data;
2227 *frame_ptr = mbuf_data(m);
2228 }
2229 (void) mbuf_setdata(m, (char *)mbuf_data(m) + frmlen,
2230 mbuf_len(m) - frmlen);
2231 (void) mbuf_pkthdr_adjustlen(m, -frmlen);
2232 }
2233
2234 if (BRIDGE_DBGF_ENABLED(BR_DBGF_MBUF) &&
2235 BRIDGE_DBGF_ENABLED(BR_DBGF_INPUT)) {
2236 brlog_mbuf(m, "bridge_iff_input]", "");
2237 }
2238
2239 out:
2240 BRIDGE_LOCK_ASSERT_NOTHELD(sc);
2241
2242 return error;
2243 }
2244
2245 static errno_t
bridge_iff_output(void * cookie,ifnet_t ifp,protocol_family_t protocol,mbuf_t * data)2246 bridge_iff_output(void *cookie, ifnet_t ifp, protocol_family_t protocol,
2247 mbuf_t *data)
2248 {
2249 #pragma unused(protocol)
2250 errno_t error = 0;
2251 struct bridge_iflist *bif = (struct bridge_iflist *)cookie;
2252 struct bridge_softc *sc = bif->bif_sc;
2253 mbuf_t m = *data;
2254
2255 if ((m->m_flags & M_PROTO1)) {
2256 goto out;
2257 }
2258 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_OUTPUT,
2259 "%s from %s m 0x%llx data 0x%llx",
2260 sc->sc_ifp->if_xname, ifp->if_xname,
2261 (uint64_t)VM_KERNEL_ADDRPERM(m),
2262 (uint64_t)VM_KERNEL_ADDRPERM(mbuf_data(m)));
2263
2264 error = bridge_member_output(sc, ifp, data);
2265 if (error != 0 && error != EJUSTRETURN) {
2266 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_OUTPUT,
2267 "bridge_member_output failed error %d",
2268 error);
2269 }
2270 out:
2271 BRIDGE_LOCK_ASSERT_NOTHELD(sc);
2272
2273 return error;
2274 }
2275
2276 static void
bridge_iff_event(void * cookie,ifnet_t ifp,protocol_family_t protocol,const struct kev_msg * event_msg)2277 bridge_iff_event(void *cookie, ifnet_t ifp, protocol_family_t protocol,
2278 const struct kev_msg *event_msg)
2279 {
2280 #pragma unused(protocol)
2281 struct bridge_iflist *bif = (struct bridge_iflist *)cookie;
2282 struct bridge_softc *sc = bif->bif_sc;
2283
2284 if (event_msg->vendor_code == KEV_VENDOR_APPLE &&
2285 event_msg->kev_class == KEV_NETWORK_CLASS &&
2286 event_msg->kev_subclass == KEV_DL_SUBCLASS) {
2287 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_LIFECYCLE,
2288 "%s event_code %u - %s",
2289 ifp->if_xname, event_msg->event_code,
2290 dlil_kev_dl_code_str(event_msg->event_code));
2291
2292 switch (event_msg->event_code) {
2293 case KEV_DL_LINK_OFF:
2294 case KEV_DL_LINK_ON: {
2295 bridge_iflinkevent(ifp);
2296 #if BRIDGESTP
2297 bstp_linkstate(ifp, event_msg->event_code);
2298 #endif /* BRIDGESTP */
2299 break;
2300 }
2301 case KEV_DL_SIFFLAGS: {
2302 if ((ifp->if_flags & IFF_UP) == 0) {
2303 break;
2304 }
2305 if ((bif->bif_flags & BIFF_PROMISC) == 0) {
2306 errno_t error;
2307
2308 error = ifnet_set_promiscuous(ifp, 1);
2309 if (error != 0) {
2310 BRIDGE_LOG(LOG_NOTICE, 0,
2311 "ifnet_set_promiscuous (%s)"
2312 " failed %d", ifp->if_xname,
2313 error);
2314 } else {
2315 bif->bif_flags |= BIFF_PROMISC;
2316 }
2317 }
2318 if ((bif->bif_flags & BIFF_WIFI_INFRA) != 0 &&
2319 (bif->bif_flags & BIFF_ALL_MULTI) == 0) {
2320 errno_t error;
2321
2322 error = if_allmulti(ifp, 1);
2323 if (error != 0) {
2324 BRIDGE_LOG(LOG_NOTICE, 0,
2325 "if_allmulti (%s)"
2326 " failed %d", ifp->if_xname,
2327 error);
2328 } else {
2329 bif->bif_flags |= BIFF_ALL_MULTI;
2330 #ifdef XNU_PLATFORM_AppleTVOS
2331 ip6_forwarding = 1;
2332 #endif /* XNU_PLATFORM_AppleTVOS */
2333 }
2334 }
2335 break;
2336 }
2337 case KEV_DL_IFCAP_CHANGED: {
2338 BRIDGE_LOCK(sc);
2339 bridge_set_tso(sc);
2340 BRIDGE_UNLOCK(sc);
2341 break;
2342 }
2343 case KEV_DL_PROTO_DETACHED:
2344 case KEV_DL_PROTO_ATTACHED: {
2345 bridge_proto_attach_changed(ifp);
2346 break;
2347 }
2348 default:
2349 break;
2350 }
2351 }
2352 }
2353
2354 /*
2355 * bridge_iff_detached:
2356 *
2357 * Called when our interface filter has been detached from a
2358 * member interface.
2359 */
2360 static void
bridge_iff_detached(void * cookie,ifnet_t ifp)2361 bridge_iff_detached(void *cookie, ifnet_t ifp)
2362 {
2363 #pragma unused(cookie)
2364 struct bridge_iflist *bif;
2365 struct bridge_softc *sc = ifp->if_bridge;
2366
2367 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_LIFECYCLE, "%s", ifp->if_xname);
2368
2369 /* Check if the interface is a bridge member */
2370 if (sc != NULL) {
2371 BRIDGE_LOCK(sc);
2372 bif = bridge_lookup_member_if(sc, ifp);
2373 if (bif != NULL) {
2374 bridge_delete_member(sc, bif);
2375 }
2376 BRIDGE_UNLOCK(sc);
2377 return;
2378 }
2379 /* Check if the interface is a span port */
2380 lck_mtx_lock(&bridge_list_mtx);
2381 LIST_FOREACH(sc, &bridge_list, sc_list) {
2382 BRIDGE_LOCK(sc);
2383 TAILQ_FOREACH(bif, &sc->sc_spanlist, bif_next)
2384 if (ifp == bif->bif_ifp) {
2385 bridge_delete_span(sc, bif);
2386 break;
2387 }
2388 BRIDGE_UNLOCK(sc);
2389 }
2390 lck_mtx_unlock(&bridge_list_mtx);
2391 }
2392
2393 static errno_t
bridge_proto_input(ifnet_t ifp,protocol_family_t protocol,mbuf_t packet,char * header)2394 bridge_proto_input(ifnet_t ifp, protocol_family_t protocol, mbuf_t packet,
2395 char *header)
2396 {
2397 #pragma unused(protocol, packet, header)
2398 BRIDGE_LOG(LOG_NOTICE, 0, "%s unexpected packet",
2399 ifp->if_xname);
2400 return 0;
2401 }
2402
2403 static int
bridge_attach_protocol(struct ifnet * ifp)2404 bridge_attach_protocol(struct ifnet *ifp)
2405 {
2406 int error;
2407 struct ifnet_attach_proto_param reg;
2408
2409 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_LIFECYCLE, "%s", ifp->if_xname);
2410 bzero(®, sizeof(reg));
2411 reg.input = bridge_proto_input;
2412
2413 error = ifnet_attach_protocol(ifp, PF_BRIDGE, ®);
2414 if (error) {
2415 BRIDGE_LOG(LOG_NOTICE, 0,
2416 "ifnet_attach_protocol(%s) failed, %d",
2417 ifp->if_xname, error);
2418 }
2419
2420 return error;
2421 }
2422
2423 static int
bridge_detach_protocol(struct ifnet * ifp)2424 bridge_detach_protocol(struct ifnet *ifp)
2425 {
2426 int error;
2427
2428 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_LIFECYCLE, "%s", ifp->if_xname);
2429 error = ifnet_detach_protocol(ifp, PF_BRIDGE);
2430 if (error) {
2431 BRIDGE_LOG(LOG_NOTICE, 0,
2432 "ifnet_detach_protocol(%s) failed, %d",
2433 ifp->if_xname, error);
2434 }
2435
2436 return error;
2437 }
2438
2439 /*
2440 * bridge_delete_member:
2441 *
2442 * Delete the specified member interface.
2443 */
2444 static void
bridge_delete_member(struct bridge_softc * sc,struct bridge_iflist * bif)2445 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif)
2446 {
2447 uint32_t bif_flags;
2448 struct ifnet *ifs = bif->bif_ifp, *bifp = sc->sc_ifp;
2449 int lladdr_changed = 0, error;
2450 uint8_t eaddr[ETHER_ADDR_LEN];
2451 u_int32_t event_code = 0;
2452
2453 BRIDGE_LOCK_ASSERT_HELD(sc);
2454 VERIFY(ifs != NULL);
2455
2456 /*
2457 * Remove the member from the list first so it cannot be found anymore
2458 * when we release the bridge lock below
2459 */
2460 if ((bif->bif_flags & BIFF_IN_MEMBER_LIST) != 0) {
2461 BRIDGE_XLOCK(sc);
2462 TAILQ_REMOVE(&sc->sc_iflist, bif, bif_next);
2463 BRIDGE_XDROP(sc);
2464 }
2465 if (sc->sc_mac_nat_bif != NULL) {
2466 if (bif == sc->sc_mac_nat_bif) {
2467 bridge_mac_nat_disable(sc);
2468 } else {
2469 bridge_mac_nat_flush_entries(sc, bif);
2470 }
2471 }
2472 #if BRIDGESTP
2473 if ((bif->bif_ifflags & IFBIF_STP) != 0) {
2474 bstp_disable(&bif->bif_stp);
2475 }
2476 #endif /* BRIDGESTP */
2477
2478 /*
2479 * If removing the interface that gave the bridge its mac address, set
2480 * the mac address of the bridge to the address of the next member, or
2481 * to its default address if no members are left.
2482 */
2483 if (bridge_inherit_mac && sc->sc_ifaddr == ifs) {
2484 ifnet_release(sc->sc_ifaddr);
2485 if (TAILQ_EMPTY(&sc->sc_iflist)) {
2486 bcopy(sc->sc_defaddr, eaddr, ETHER_ADDR_LEN);
2487 sc->sc_ifaddr = NULL;
2488 } else {
2489 struct ifnet *fif =
2490 TAILQ_FIRST(&sc->sc_iflist)->bif_ifp;
2491 bcopy(IF_LLADDR(fif), eaddr, ETHER_ADDR_LEN);
2492 sc->sc_ifaddr = fif;
2493 ifnet_reference(fif); /* for sc_ifaddr */
2494 }
2495 lladdr_changed = 1;
2496 }
2497
2498 #if HAS_IF_CAP
2499 bridge_mutecaps(sc); /* recalculate now this interface is removed */
2500 #endif /* HAS_IF_CAP */
2501
2502 error = bridge_set_tso(sc);
2503 if (error != 0) {
2504 BRIDGE_LOG(LOG_NOTICE, 0, "bridge_set_tso failed %d", error);
2505 }
2506
2507 bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
2508
2509 KASSERT(bif->bif_addrcnt == 0,
2510 ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
2511
2512 /*
2513 * Update link status of the bridge based on its remaining members
2514 */
2515 event_code = bridge_updatelinkstatus(sc);
2516 bif_flags = bif->bif_flags;
2517 BRIDGE_UNLOCK(sc);
2518
2519 /* only perform these steps if the interface is still attached */
2520 if (ifnet_is_attached(ifs, 1)) {
2521 #if SKYWALK
2522 if ((bif_flags & BIFF_NETAGENT_REMOVED) != 0) {
2523 ifnet_add_netagent(ifs);
2524 }
2525 if ((bif_flags & BIFF_FLOWSWITCH_ATTACHED) != 0) {
2526 ifnet_detach_flowswitch_nexus(ifs);
2527 }
2528 #endif /* SKYWALK */
2529 /* disable promiscuous mode */
2530 if ((bif_flags & BIFF_PROMISC) != 0) {
2531 (void) ifnet_set_promiscuous(ifs, 0);
2532 }
2533 /* disable all multi */
2534 if ((bif_flags & BIFF_ALL_MULTI) != 0) {
2535 (void)if_allmulti(ifs, 0);
2536 }
2537 #if HAS_IF_CAP
2538 /* re-enable any interface capabilities */
2539 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
2540 #endif
2541 /* detach bridge "protocol" */
2542 if ((bif_flags & BIFF_PROTO_ATTACHED) != 0) {
2543 (void)bridge_detach_protocol(ifs);
2544 }
2545 /* detach interface filter */
2546 if ((bif_flags & BIFF_FILTER_ATTACHED) != 0) {
2547 iflt_detach(bif->bif_iff_ref);
2548 }
2549 ifnet_decr_iorefcnt(ifs);
2550 }
2551
2552 if (lladdr_changed &&
2553 (error = ifnet_set_lladdr(bifp, eaddr, ETHER_ADDR_LEN)) != 0) {
2554 BRIDGE_LOG(LOG_NOTICE, 0, "ifnet_set_lladdr failed %d", error);
2555 }
2556
2557 if (event_code != 0) {
2558 bridge_link_event(bifp, event_code);
2559 }
2560
2561 #if BRIDGESTP
2562 bstp_destroy(&bif->bif_stp); /* prepare to free */
2563 #endif /* BRIDGESTP */
2564
2565 kfree_type(struct bridge_iflist, bif);
2566 ifs->if_bridge = NULL;
2567 ifnet_release(ifs);
2568
2569 BRIDGE_LOCK(sc);
2570 }
2571
2572 /*
2573 * bridge_delete_span:
2574 *
2575 * Delete the specified span interface.
2576 */
2577 static void
bridge_delete_span(struct bridge_softc * sc,struct bridge_iflist * bif)2578 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
2579 {
2580 BRIDGE_LOCK_ASSERT_HELD(sc);
2581
2582 KASSERT(bif->bif_ifp->if_bridge == NULL,
2583 ("%s: not a span interface", __func__));
2584
2585 ifnet_release(bif->bif_ifp);
2586
2587 TAILQ_REMOVE(&sc->sc_spanlist, bif, bif_next);
2588 kfree_type(struct bridge_iflist, bif);
2589 }
2590
2591 static int
bridge_ioctl_add(struct bridge_softc * sc,void * arg)2592 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
2593 {
2594 struct ifbreq *req = arg;
2595 struct bridge_iflist *bif = NULL;
2596 struct ifnet *ifs, *bifp = sc->sc_ifp;
2597 int error = 0, lladdr_changed = 0;
2598 uint8_t eaddr[ETHER_ADDR_LEN];
2599 struct iff_filter iff;
2600 u_int32_t event_code = 0;
2601 boolean_t input_broadcast;
2602 boolean_t wifi_infra = FALSE;
2603
2604 ifs = ifunit(req->ifbr_ifsname);
2605 if (ifs == NULL) {
2606 return ENOENT;
2607 }
2608 if (ifs->if_ioctl == NULL) { /* must be supported */
2609 return EINVAL;
2610 }
2611
2612 if (IFNET_IS_INTCOPROC(ifs)) {
2613 return EINVAL;
2614 }
2615
2616 /* If it's in the span list, it can't be a member. */
2617 TAILQ_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2618 if (ifs == bif->bif_ifp) {
2619 return EBUSY;
2620 }
2621 }
2622
2623 if (ifs->if_bridge == sc) {
2624 return EEXIST;
2625 }
2626
2627 if (ifs->if_bridge != NULL) {
2628 return EBUSY;
2629 }
2630
2631 switch (ifs->if_type) {
2632 case IFT_ETHER:
2633 if (strcmp(ifs->if_name, "en") == 0 &&
2634 ifs->if_subfamily == IFNET_SUBFAMILY_WIFI &&
2635 (ifs->if_eflags & IFEF_IPV4_ROUTER) == 0) {
2636 /* XXX is there a better way to identify Wi-Fi STA? */
2637 wifi_infra = TRUE;
2638 }
2639 break;
2640 case IFT_L2VLAN:
2641 case IFT_IEEE8023ADLAG:
2642 break;
2643 case IFT_GIF:
2644 /* currently not supported */
2645 /* FALLTHRU */
2646 default:
2647 return EINVAL;
2648 }
2649
2650 /* fail to add the interface if the MTU doesn't match */
2651 if (!TAILQ_EMPTY(&sc->sc_iflist) && sc->sc_ifp->if_mtu != ifs->if_mtu) {
2652 BRIDGE_LOG(LOG_NOTICE, 0, "%s invalid MTU for %s",
2653 sc->sc_ifp->if_xname,
2654 ifs->if_xname);
2655 return EINVAL;
2656 }
2657
2658 /* there's already an interface that's doing MAC NAT */
2659 if (wifi_infra && sc->sc_mac_nat_bif != NULL) {
2660 return EBUSY;
2661 }
2662
2663 /* prevent the interface from detaching while we add the member */
2664 if (!ifnet_is_attached(ifs, 1)) {
2665 return ENXIO;
2666 }
2667
2668 /* allocate a new member */
2669 bif = kalloc_type(struct bridge_iflist, Z_WAITOK | Z_ZERO | Z_NOFAIL);
2670 bif->bif_ifp = ifs;
2671 ifnet_reference(ifs);
2672 bif->bif_ifflags |= IFBIF_LEARNING | IFBIF_DISCOVER;
2673 #if HAS_IF_CAP
2674 bif->bif_savedcaps = ifs->if_capenable;
2675 #endif /* HAS_IF_CAP */
2676 bif->bif_sc = sc;
2677 if (wifi_infra) {
2678 (void)bridge_mac_nat_enable(sc, bif);
2679 }
2680
2681 if (IFNET_IS_VMNET(ifs)) {
2682 allocate_vmnet_pf_tags();
2683 }
2684 /* Allow the first Ethernet member to define the MTU */
2685 if (TAILQ_EMPTY(&sc->sc_iflist)) {
2686 sc->sc_ifp->if_mtu = ifs->if_mtu;
2687 }
2688
2689 /*
2690 * Assign the interface's MAC address to the bridge if it's the first
2691 * member and the MAC address of the bridge has not been changed from
2692 * the default (randomly) generated one.
2693 */
2694 if (bridge_inherit_mac && TAILQ_EMPTY(&sc->sc_iflist) &&
2695 _ether_cmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr) == 0) {
2696 bcopy(IF_LLADDR(ifs), eaddr, ETHER_ADDR_LEN);
2697 sc->sc_ifaddr = ifs;
2698 ifnet_reference(ifs); /* for sc_ifaddr */
2699 lladdr_changed = 1;
2700 }
2701
2702 ifs->if_bridge = sc;
2703 #if BRIDGESTP
2704 bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
2705 #endif /* BRIDGESTP */
2706
2707 #if HAS_IF_CAP
2708 /* Set interface capabilities to the intersection set of all members */
2709 bridge_mutecaps(sc);
2710 #endif /* HAS_IF_CAP */
2711
2712
2713 /*
2714 * Respect lock ordering with DLIL lock for the following operations
2715 */
2716 BRIDGE_UNLOCK(sc);
2717
2718 /* enable promiscuous mode */
2719 error = ifnet_set_promiscuous(ifs, 1);
2720 switch (error) {
2721 case 0:
2722 bif->bif_flags |= BIFF_PROMISC;
2723 break;
2724 case ENETDOWN:
2725 case EPWROFF:
2726 BRIDGE_LOG(LOG_NOTICE, 0,
2727 "ifnet_set_promiscuous(%s) failed %d, ignoring",
2728 ifs->if_xname, error);
2729 /* Ignore error when device is not up */
2730 error = 0;
2731 break;
2732 default:
2733 BRIDGE_LOG(LOG_NOTICE, 0,
2734 "ifnet_set_promiscuous(%s) failed %d",
2735 ifs->if_xname, error);
2736 BRIDGE_LOCK(sc);
2737 goto out;
2738 }
2739 if (wifi_infra) {
2740 int this_error;
2741
2742 /* Wi-Fi doesn't really support promiscuous, set allmulti */
2743 bif->bif_flags |= BIFF_WIFI_INFRA;
2744 this_error = if_allmulti(ifs, 1);
2745 if (this_error == 0) {
2746 bif->bif_flags |= BIFF_ALL_MULTI;
2747 #ifdef XNU_PLATFORM_AppleTVOS
2748 ip6_forwarding = 1;
2749 #endif /* XNU_PLATFORM_AppleTVOS */
2750 } else {
2751 BRIDGE_LOG(LOG_NOTICE, 0,
2752 "if_allmulti(%s) failed %d, ignoring",
2753 ifs->if_xname, this_error);
2754 }
2755 }
2756 #if SKYWALK
2757 /* ensure that the flowswitch is present for native interface */
2758 if (SKYWALK_NATIVE(ifs)) {
2759 if (ifnet_attach_flowswitch_nexus(ifs)) {
2760 bif->bif_flags |= BIFF_FLOWSWITCH_ATTACHED;
2761 }
2762 }
2763 /* remove the netagent on the flowswitch (rdar://75050182) */
2764 if (ifnet_remove_netagent(ifs)) {
2765 bif->bif_flags |= BIFF_NETAGENT_REMOVED;
2766 }
2767 #endif /* SKYWALK */
2768
2769 /*
2770 * install an interface filter
2771 */
2772 memset(&iff, 0, sizeof(struct iff_filter));
2773 iff.iff_cookie = bif;
2774 iff.iff_name = "com.apple.kernel.bsd.net.if_bridge";
2775 iff.iff_input = bridge_iff_input;
2776 iff.iff_output = bridge_iff_output;
2777 iff.iff_event = bridge_iff_event;
2778 iff.iff_detached = bridge_iff_detached;
2779 error = dlil_attach_filter(ifs, &iff, &bif->bif_iff_ref,
2780 DLIL_IFF_TSO | DLIL_IFF_INTERNAL);
2781 if (error != 0) {
2782 BRIDGE_LOG(LOG_NOTICE, 0, "iflt_attach failed %d", error);
2783 BRIDGE_LOCK(sc);
2784 goto out;
2785 }
2786 bif->bif_flags |= BIFF_FILTER_ATTACHED;
2787
2788 /*
2789 * install a dummy "bridge" protocol
2790 */
2791 if ((error = bridge_attach_protocol(ifs)) != 0) {
2792 if (error != 0) {
2793 BRIDGE_LOG(LOG_NOTICE, 0,
2794 "bridge_attach_protocol failed %d", error);
2795 BRIDGE_LOCK(sc);
2796 goto out;
2797 }
2798 }
2799 bif->bif_flags |= BIFF_PROTO_ATTACHED;
2800
2801 if (lladdr_changed &&
2802 (error = ifnet_set_lladdr(bifp, eaddr, ETHER_ADDR_LEN)) != 0) {
2803 BRIDGE_LOG(LOG_NOTICE, 0, "ifnet_set_lladdr failed %d", error);
2804 }
2805
2806 /*
2807 * No failures past this point. Add the member to the list.
2808 */
2809 BRIDGE_LOCK(sc);
2810 bif->bif_flags |= BIFF_IN_MEMBER_LIST;
2811 BRIDGE_XLOCK(sc);
2812 TAILQ_INSERT_TAIL(&sc->sc_iflist, bif, bif_next);
2813 BRIDGE_XDROP(sc);
2814
2815 /* cache the member link status */
2816 if (interface_media_active(ifs)) {
2817 bif->bif_flags |= BIFF_MEDIA_ACTIVE;
2818 } else {
2819 bif->bif_flags &= ~BIFF_MEDIA_ACTIVE;
2820 }
2821
2822 /* the new member may change the link status of the bridge interface */
2823 event_code = bridge_updatelinkstatus(sc);
2824
2825 /* check whether we need input broadcast or not */
2826 input_broadcast = interface_needs_input_broadcast(ifs);
2827 bif_set_input_broadcast(bif, input_broadcast);
2828 BRIDGE_UNLOCK(sc);
2829
2830 if (event_code != 0) {
2831 bridge_link_event(bifp, event_code);
2832 }
2833 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_LIFECYCLE,
2834 "%s input broadcast %s", ifs->if_xname,
2835 input_broadcast ? "ENABLED" : "DISABLED");
2836
2837 BRIDGE_LOCK(sc);
2838 bridge_set_tso(sc);
2839
2840 out:
2841 /* allow the interface to detach */
2842 ifnet_decr_iorefcnt(ifs);
2843
2844 if (error != 0) {
2845 if (bif != NULL) {
2846 bridge_delete_member(sc, bif);
2847 }
2848 } else if (IFNET_IS_VMNET(ifs)) {
2849 INC_ATOMIC_INT64_LIM(net_api_stats.nas_vmnet_total);
2850 }
2851
2852 return error;
2853 }
2854
2855 static int
bridge_ioctl_del(struct bridge_softc * sc,void * arg)2856 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
2857 {
2858 struct ifbreq *req = arg;
2859 struct bridge_iflist *bif;
2860
2861 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
2862 if (bif == NULL) {
2863 return ENOENT;
2864 }
2865
2866 bridge_delete_member(sc, bif);
2867
2868 return 0;
2869 }
2870
2871 static int
bridge_ioctl_purge(struct bridge_softc * sc,void * arg)2872 bridge_ioctl_purge(struct bridge_softc *sc, void *arg)
2873 {
2874 #pragma unused(sc, arg)
2875 return 0;
2876 }
2877
2878 static int
bridge_ioctl_gifflags(struct bridge_softc * sc,void * arg)2879 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
2880 {
2881 struct ifbreq *req = arg;
2882 struct bridge_iflist *bif;
2883
2884 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
2885 if (bif == NULL) {
2886 return ENOENT;
2887 }
2888
2889 struct bstp_port *bp;
2890
2891 bp = &bif->bif_stp;
2892 req->ifbr_state = bp->bp_state;
2893 req->ifbr_priority = bp->bp_priority;
2894 req->ifbr_path_cost = bp->bp_path_cost;
2895 req->ifbr_proto = bp->bp_protover;
2896 req->ifbr_role = bp->bp_role;
2897 req->ifbr_stpflags = bp->bp_flags;
2898 req->ifbr_ifsflags = bif->bif_ifflags;
2899
2900 /* Copy STP state options as flags */
2901 if (bp->bp_operedge) {
2902 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
2903 }
2904 if (bp->bp_flags & BSTP_PORT_AUTOEDGE) {
2905 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
2906 }
2907 if (bp->bp_ptp_link) {
2908 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
2909 }
2910 if (bp->bp_flags & BSTP_PORT_AUTOPTP) {
2911 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
2912 }
2913 if (bp->bp_flags & BSTP_PORT_ADMEDGE) {
2914 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
2915 }
2916 if (bp->bp_flags & BSTP_PORT_ADMCOST) {
2917 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
2918 }
2919
2920 req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
2921 req->ifbr_addrcnt = bif->bif_addrcnt;
2922 req->ifbr_addrmax = bif->bif_addrmax;
2923 req->ifbr_addrexceeded = bif->bif_addrexceeded;
2924
2925 return 0;
2926 }
2927
2928 static int
bridge_ioctl_sifflags(struct bridge_softc * sc,void * arg)2929 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
2930 {
2931 struct ifbreq *req = arg;
2932 struct bridge_iflist *bif;
2933 #if BRIDGESTP
2934 struct bstp_port *bp;
2935 int error;
2936 #endif /* BRIDGESTP */
2937
2938 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
2939 if (bif == NULL) {
2940 return ENOENT;
2941 }
2942
2943 if (req->ifbr_ifsflags & IFBIF_SPAN) {
2944 /* SPAN is readonly */
2945 return EINVAL;
2946 }
2947 #define _EXCLUSIVE_FLAGS (IFBIF_CHECKSUM_OFFLOAD | IFBIF_MAC_NAT)
2948 if ((req->ifbr_ifsflags & _EXCLUSIVE_FLAGS) == _EXCLUSIVE_FLAGS) {
2949 /* can't specify both MAC-NAT and checksum offload */
2950 return EINVAL;
2951 }
2952 if ((req->ifbr_ifsflags & IFBIF_MAC_NAT) != 0) {
2953 errno_t error;
2954
2955 error = bridge_mac_nat_enable(sc, bif);
2956 if (error != 0) {
2957 return error;
2958 }
2959 } else if (sc->sc_mac_nat_bif == bif) {
2960 bridge_mac_nat_disable(sc);
2961 }
2962
2963
2964 #if BRIDGESTP
2965 if (req->ifbr_ifsflags & IFBIF_STP) {
2966 if ((bif->bif_ifflags & IFBIF_STP) == 0) {
2967 error = bstp_enable(&bif->bif_stp);
2968 if (error) {
2969 return error;
2970 }
2971 }
2972 } else {
2973 if ((bif->bif_ifflags & IFBIF_STP) != 0) {
2974 bstp_disable(&bif->bif_stp);
2975 }
2976 }
2977
2978 /* Pass on STP flags */
2979 bp = &bif->bif_stp;
2980 bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
2981 bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
2982 bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
2983 bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
2984 #else /* !BRIDGESTP */
2985 if (req->ifbr_ifsflags & IFBIF_STP) {
2986 return EOPNOTSUPP;
2987 }
2988 #endif /* !BRIDGESTP */
2989
2990 /* Save the bits relating to the bridge */
2991 bif->bif_ifflags = req->ifbr_ifsflags & IFBIFMASK;
2992
2993
2994 return 0;
2995 }
2996
2997 static int
bridge_ioctl_scache(struct bridge_softc * sc,void * arg)2998 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
2999 {
3000 struct ifbrparam *param = arg;
3001
3002 sc->sc_brtmax = param->ifbrp_csize;
3003 bridge_rttrim(sc);
3004 return 0;
3005 }
3006
3007 static int
bridge_ioctl_gcache(struct bridge_softc * sc,void * arg)3008 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
3009 {
3010 struct ifbrparam *param = arg;
3011
3012 param->ifbrp_csize = sc->sc_brtmax;
3013
3014 return 0;
3015 }
3016
3017 #define BRIDGE_IOCTL_GIFS do { \
3018 struct bridge_iflist *bif; \
3019 struct ifbreq breq; \
3020 char *buf, *outbuf; \
3021 unsigned int count, buflen, len; \
3022 \
3023 count = 0; \
3024 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) \
3025 count++; \
3026 TAILQ_FOREACH(bif, &sc->sc_spanlist, bif_next) \
3027 count++; \
3028 \
3029 buflen = sizeof (breq) * count; \
3030 if (bifc->ifbic_len == 0) { \
3031 bifc->ifbic_len = buflen; \
3032 return (0); \
3033 } \
3034 BRIDGE_UNLOCK(sc); \
3035 outbuf = (char *)kalloc_data(buflen, Z_WAITOK | Z_ZERO); \
3036 BRIDGE_LOCK(sc); \
3037 \
3038 count = 0; \
3039 buf = outbuf; \
3040 len = min(bifc->ifbic_len, buflen); \
3041 bzero(&breq, sizeof (breq)); \
3042 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) { \
3043 if (len < sizeof (breq)) \
3044 break; \
3045 \
3046 snprintf(breq.ifbr_ifsname, sizeof (breq.ifbr_ifsname), \
3047 "%s", bif->bif_ifp->if_xname); \
3048 /* Fill in the ifbreq structure */ \
3049 error = bridge_ioctl_gifflags(sc, &breq); \
3050 if (error) \
3051 break; \
3052 memcpy(buf, &breq, sizeof (breq)); \
3053 count++; \
3054 buf += sizeof (breq); \
3055 len -= sizeof (breq); \
3056 } \
3057 TAILQ_FOREACH(bif, &sc->sc_spanlist, bif_next) { \
3058 if (len < sizeof (breq)) \
3059 break; \
3060 \
3061 snprintf(breq.ifbr_ifsname, \
3062 sizeof (breq.ifbr_ifsname), \
3063 "%s", bif->bif_ifp->if_xname); \
3064 breq.ifbr_ifsflags = bif->bif_ifflags; \
3065 breq.ifbr_portno \
3066 = bif->bif_ifp->if_index & 0xfff; \
3067 memcpy(buf, &breq, sizeof (breq)); \
3068 count++; \
3069 buf += sizeof (breq); \
3070 len -= sizeof (breq); \
3071 } \
3072 \
3073 BRIDGE_UNLOCK(sc); \
3074 bifc->ifbic_len = sizeof (breq) * count; \
3075 error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len); \
3076 BRIDGE_LOCK(sc); \
3077 kfree_data(outbuf, buflen); \
3078 } while (0)
3079
3080 static int
bridge_ioctl_gifs64(struct bridge_softc * sc,void * arg)3081 bridge_ioctl_gifs64(struct bridge_softc *sc, void *arg)
3082 {
3083 struct ifbifconf64 *bifc = arg;
3084 int error = 0;
3085
3086 BRIDGE_IOCTL_GIFS;
3087
3088 return error;
3089 }
3090
3091 static int
bridge_ioctl_gifs32(struct bridge_softc * sc,void * arg)3092 bridge_ioctl_gifs32(struct bridge_softc *sc, void *arg)
3093 {
3094 struct ifbifconf32 *bifc = arg;
3095 int error = 0;
3096
3097 BRIDGE_IOCTL_GIFS;
3098
3099 return error;
3100 }
3101
3102 #define BRIDGE_IOCTL_RTS do { \
3103 struct bridge_rtnode *brt; \
3104 char *buf; \
3105 char *outbuf = NULL; \
3106 unsigned int count, buflen, len; \
3107 unsigned long now; \
3108 \
3109 if (bac->ifbac_len == 0) \
3110 return (0); \
3111 \
3112 bzero(&bareq, sizeof (bareq)); \
3113 count = 0; \
3114 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) \
3115 count++; \
3116 buflen = sizeof (bareq) * count; \
3117 \
3118 BRIDGE_UNLOCK(sc); \
3119 outbuf = (char *)kalloc_data(buflen, Z_WAITOK | Z_ZERO); \
3120 BRIDGE_LOCK(sc); \
3121 \
3122 count = 0; \
3123 buf = outbuf; \
3124 len = min(bac->ifbac_len, buflen); \
3125 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) { \
3126 if (len < sizeof (bareq)) \
3127 goto out; \
3128 snprintf(bareq.ifba_ifsname, sizeof (bareq.ifba_ifsname), \
3129 "%s", brt->brt_ifp->if_xname); \
3130 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof (brt->brt_addr)); \
3131 bareq.ifba_vlan = brt->brt_vlan; \
3132 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) { \
3133 now = (unsigned long) net_uptime(); \
3134 if (now < brt->brt_expire) \
3135 bareq.ifba_expire = \
3136 brt->brt_expire - now; \
3137 } else \
3138 bareq.ifba_expire = 0; \
3139 bareq.ifba_flags = brt->brt_flags; \
3140 \
3141 memcpy(buf, &bareq, sizeof (bareq)); \
3142 count++; \
3143 buf += sizeof (bareq); \
3144 len -= sizeof (bareq); \
3145 } \
3146 out: \
3147 bac->ifbac_len = sizeof (bareq) * count; \
3148 if (outbuf != NULL) { \
3149 BRIDGE_UNLOCK(sc); \
3150 error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len); \
3151 kfree_data(outbuf, buflen); \
3152 BRIDGE_LOCK(sc); \
3153 } \
3154 return (error); \
3155 } while (0)
3156
3157 static int
bridge_ioctl_rts64(struct bridge_softc * sc,void * arg)3158 bridge_ioctl_rts64(struct bridge_softc *sc, void *arg)
3159 {
3160 struct ifbaconf64 *bac = arg;
3161 struct ifbareq64 bareq;
3162 int error = 0;
3163
3164 BRIDGE_IOCTL_RTS;
3165 return error;
3166 }
3167
3168 static int
bridge_ioctl_rts32(struct bridge_softc * sc,void * arg)3169 bridge_ioctl_rts32(struct bridge_softc *sc, void *arg)
3170 {
3171 struct ifbaconf32 *bac = arg;
3172 struct ifbareq32 bareq;
3173 int error = 0;
3174
3175 BRIDGE_IOCTL_RTS;
3176 return error;
3177 }
3178
3179 static int
bridge_ioctl_saddr32(struct bridge_softc * sc,void * arg)3180 bridge_ioctl_saddr32(struct bridge_softc *sc, void *arg)
3181 {
3182 struct ifbareq32 *req = arg;
3183 struct bridge_iflist *bif;
3184 int error;
3185
3186 bif = bridge_lookup_member(sc, req->ifba_ifsname);
3187 if (bif == NULL) {
3188 return ENOENT;
3189 }
3190
3191 error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
3192 req->ifba_flags);
3193
3194 return error;
3195 }
3196
3197 static int
bridge_ioctl_saddr64(struct bridge_softc * sc,void * arg)3198 bridge_ioctl_saddr64(struct bridge_softc *sc, void *arg)
3199 {
3200 struct ifbareq64 *req = arg;
3201 struct bridge_iflist *bif;
3202 int error;
3203
3204 bif = bridge_lookup_member(sc, req->ifba_ifsname);
3205 if (bif == NULL) {
3206 return ENOENT;
3207 }
3208
3209 error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
3210 req->ifba_flags);
3211
3212 return error;
3213 }
3214
3215 static int
bridge_ioctl_sto(struct bridge_softc * sc,void * arg)3216 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
3217 {
3218 struct ifbrparam *param = arg;
3219
3220 sc->sc_brttimeout = param->ifbrp_ctime;
3221 return 0;
3222 }
3223
3224 static int
bridge_ioctl_gto(struct bridge_softc * sc,void * arg)3225 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
3226 {
3227 struct ifbrparam *param = arg;
3228
3229 param->ifbrp_ctime = sc->sc_brttimeout;
3230 return 0;
3231 }
3232
3233 static int
bridge_ioctl_daddr32(struct bridge_softc * sc,void * arg)3234 bridge_ioctl_daddr32(struct bridge_softc *sc, void *arg)
3235 {
3236 struct ifbareq32 *req = arg;
3237
3238 return bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan);
3239 }
3240
3241 static int
bridge_ioctl_daddr64(struct bridge_softc * sc,void * arg)3242 bridge_ioctl_daddr64(struct bridge_softc *sc, void *arg)
3243 {
3244 struct ifbareq64 *req = arg;
3245
3246 return bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan);
3247 }
3248
3249 static int
bridge_ioctl_flush(struct bridge_softc * sc,void * arg)3250 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
3251 {
3252 struct ifbreq *req = arg;
3253
3254 bridge_rtflush(sc, req->ifbr_ifsflags);
3255 return 0;
3256 }
3257
3258 static int
bridge_ioctl_gpri(struct bridge_softc * sc,void * arg)3259 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
3260 {
3261 struct ifbrparam *param = arg;
3262 struct bstp_state *bs = &sc->sc_stp;
3263
3264 param->ifbrp_prio = bs->bs_bridge_priority;
3265 return 0;
3266 }
3267
3268 static int
bridge_ioctl_spri(struct bridge_softc * sc,void * arg)3269 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
3270 {
3271 #if BRIDGESTP
3272 struct ifbrparam *param = arg;
3273
3274 return bstp_set_priority(&sc->sc_stp, param->ifbrp_prio);
3275 #else /* !BRIDGESTP */
3276 #pragma unused(sc, arg)
3277 return EOPNOTSUPP;
3278 #endif /* !BRIDGESTP */
3279 }
3280
3281 static int
bridge_ioctl_ght(struct bridge_softc * sc,void * arg)3282 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
3283 {
3284 struct ifbrparam *param = arg;
3285 struct bstp_state *bs = &sc->sc_stp;
3286
3287 param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
3288 return 0;
3289 }
3290
3291 static int
bridge_ioctl_sht(struct bridge_softc * sc,void * arg)3292 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
3293 {
3294 #if BRIDGESTP
3295 struct ifbrparam *param = arg;
3296
3297 return bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime);
3298 #else /* !BRIDGESTP */
3299 #pragma unused(sc, arg)
3300 return EOPNOTSUPP;
3301 #endif /* !BRIDGESTP */
3302 }
3303
3304 static int
bridge_ioctl_gfd(struct bridge_softc * sc,void * arg)3305 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
3306 {
3307 struct ifbrparam *param;
3308 struct bstp_state *bs;
3309
3310 param = arg;
3311 bs = &sc->sc_stp;
3312 param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
3313 return 0;
3314 }
3315
3316 static int
bridge_ioctl_sfd(struct bridge_softc * sc,void * arg)3317 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
3318 {
3319 #if BRIDGESTP
3320 struct ifbrparam *param = arg;
3321
3322 return bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay);
3323 #else /* !BRIDGESTP */
3324 #pragma unused(sc, arg)
3325 return EOPNOTSUPP;
3326 #endif /* !BRIDGESTP */
3327 }
3328
3329 static int
bridge_ioctl_gma(struct bridge_softc * sc,void * arg)3330 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
3331 {
3332 struct ifbrparam *param;
3333 struct bstp_state *bs;
3334
3335 param = arg;
3336 bs = &sc->sc_stp;
3337 param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
3338 return 0;
3339 }
3340
3341 static int
bridge_ioctl_sma(struct bridge_softc * sc,void * arg)3342 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
3343 {
3344 #if BRIDGESTP
3345 struct ifbrparam *param = arg;
3346
3347 return bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage);
3348 #else /* !BRIDGESTP */
3349 #pragma unused(sc, arg)
3350 return EOPNOTSUPP;
3351 #endif /* !BRIDGESTP */
3352 }
3353
3354 static int
bridge_ioctl_sifprio(struct bridge_softc * sc,void * arg)3355 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
3356 {
3357 #if BRIDGESTP
3358 struct ifbreq *req = arg;
3359 struct bridge_iflist *bif;
3360
3361 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
3362 if (bif == NULL) {
3363 return ENOENT;
3364 }
3365
3366 return bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority);
3367 #else /* !BRIDGESTP */
3368 #pragma unused(sc, arg)
3369 return EOPNOTSUPP;
3370 #endif /* !BRIDGESTP */
3371 }
3372
3373 static int
bridge_ioctl_sifcost(struct bridge_softc * sc,void * arg)3374 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
3375 {
3376 #if BRIDGESTP
3377 struct ifbreq *req = arg;
3378 struct bridge_iflist *bif;
3379
3380 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
3381 if (bif == NULL) {
3382 return ENOENT;
3383 }
3384
3385 return bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost);
3386 #else /* !BRIDGESTP */
3387 #pragma unused(sc, arg)
3388 return EOPNOTSUPP;
3389 #endif /* !BRIDGESTP */
3390 }
3391
3392 static int
bridge_ioctl_gfilt(struct bridge_softc * sc,void * arg)3393 bridge_ioctl_gfilt(struct bridge_softc *sc, void *arg)
3394 {
3395 struct ifbrparam *param = arg;
3396
3397 param->ifbrp_filter = sc->sc_filter_flags;
3398
3399 return 0;
3400 }
3401
3402 static int
bridge_ioctl_sfilt(struct bridge_softc * sc,void * arg)3403 bridge_ioctl_sfilt(struct bridge_softc *sc, void *arg)
3404 {
3405 struct ifbrparam *param = arg;
3406
3407 if (param->ifbrp_filter & ~IFBF_FILT_MASK) {
3408 return EINVAL;
3409 }
3410
3411 if (param->ifbrp_filter & IFBF_FILT_USEIPF) {
3412 return EINVAL;
3413 }
3414
3415 sc->sc_filter_flags = param->ifbrp_filter;
3416
3417 return 0;
3418 }
3419
3420 static int
bridge_ioctl_sifmaxaddr(struct bridge_softc * sc,void * arg)3421 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
3422 {
3423 struct ifbreq *req = arg;
3424 struct bridge_iflist *bif;
3425
3426 bif = bridge_lookup_member(sc, req->ifbr_ifsname);
3427 if (bif == NULL) {
3428 return ENOENT;
3429 }
3430
3431 bif->bif_addrmax = req->ifbr_addrmax;
3432 return 0;
3433 }
3434
3435 static int
bridge_ioctl_addspan(struct bridge_softc * sc,void * arg)3436 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
3437 {
3438 struct ifbreq *req = arg;
3439 struct bridge_iflist *bif = NULL;
3440 struct ifnet *ifs;
3441
3442 ifs = ifunit(req->ifbr_ifsname);
3443 if (ifs == NULL) {
3444 return ENOENT;
3445 }
3446
3447 if (IFNET_IS_INTCOPROC(ifs)) {
3448 return EINVAL;
3449 }
3450
3451 TAILQ_FOREACH(bif, &sc->sc_spanlist, bif_next)
3452 if (ifs == bif->bif_ifp) {
3453 return EBUSY;
3454 }
3455
3456 if (ifs->if_bridge != NULL) {
3457 return EBUSY;
3458 }
3459
3460 switch (ifs->if_type) {
3461 case IFT_ETHER:
3462 case IFT_L2VLAN:
3463 case IFT_IEEE8023ADLAG:
3464 break;
3465 case IFT_GIF:
3466 /* currently not supported */
3467 /* FALLTHRU */
3468 default:
3469 return EINVAL;
3470 }
3471
3472 bif = kalloc_type(struct bridge_iflist, Z_WAITOK | Z_ZERO | Z_NOFAIL);
3473
3474 bif->bif_ifp = ifs;
3475 bif->bif_ifflags = IFBIF_SPAN;
3476
3477 ifnet_reference(bif->bif_ifp);
3478
3479 TAILQ_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
3480
3481 return 0;
3482 }
3483
3484 static int
bridge_ioctl_delspan(struct bridge_softc * sc,void * arg)3485 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
3486 {
3487 struct ifbreq *req = arg;
3488 struct bridge_iflist *bif;
3489 struct ifnet *ifs;
3490
3491 ifs = ifunit(req->ifbr_ifsname);
3492 if (ifs == NULL) {
3493 return ENOENT;
3494 }
3495
3496 TAILQ_FOREACH(bif, &sc->sc_spanlist, bif_next)
3497 if (ifs == bif->bif_ifp) {
3498 break;
3499 }
3500
3501 if (bif == NULL) {
3502 return ENOENT;
3503 }
3504
3505 bridge_delete_span(sc, bif);
3506
3507 return 0;
3508 }
3509
3510 #define BRIDGE_IOCTL_GBPARAM do { \
3511 struct bstp_state *bs = &sc->sc_stp; \
3512 struct bstp_port *root_port; \
3513 \
3514 req->ifbop_maxage = bs->bs_bridge_max_age >> 8; \
3515 req->ifbop_hellotime = bs->bs_bridge_htime >> 8; \
3516 req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8; \
3517 \
3518 root_port = bs->bs_root_port; \
3519 if (root_port == NULL) \
3520 req->ifbop_root_port = 0; \
3521 else \
3522 req->ifbop_root_port = root_port->bp_ifp->if_index; \
3523 \
3524 req->ifbop_holdcount = bs->bs_txholdcount; \
3525 req->ifbop_priority = bs->bs_bridge_priority; \
3526 req->ifbop_protocol = bs->bs_protover; \
3527 req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost; \
3528 req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id; \
3529 req->ifbop_designated_root = bs->bs_root_pv.pv_root_id; \
3530 req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id; \
3531 req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec; \
3532 req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec; \
3533 } while (0)
3534
3535 static int
bridge_ioctl_gbparam32(struct bridge_softc * sc,void * arg)3536 bridge_ioctl_gbparam32(struct bridge_softc *sc, void *arg)
3537 {
3538 struct ifbropreq32 *req = arg;
3539
3540 BRIDGE_IOCTL_GBPARAM;
3541 return 0;
3542 }
3543
3544 static int
bridge_ioctl_gbparam64(struct bridge_softc * sc,void * arg)3545 bridge_ioctl_gbparam64(struct bridge_softc *sc, void *arg)
3546 {
3547 struct ifbropreq64 *req = arg;
3548
3549 BRIDGE_IOCTL_GBPARAM;
3550 return 0;
3551 }
3552
3553 static int
bridge_ioctl_grte(struct bridge_softc * sc,void * arg)3554 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
3555 {
3556 struct ifbrparam *param = arg;
3557
3558 param->ifbrp_cexceeded = sc->sc_brtexceeded;
3559 return 0;
3560 }
3561
3562 #define BRIDGE_IOCTL_GIFSSTP do { \
3563 struct bridge_iflist *bif; \
3564 struct bstp_port *bp; \
3565 struct ifbpstpreq bpreq; \
3566 char *buf, *outbuf; \
3567 unsigned int count, buflen, len; \
3568 \
3569 count = 0; \
3570 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) { \
3571 if ((bif->bif_ifflags & IFBIF_STP) != 0) \
3572 count++; \
3573 } \
3574 \
3575 buflen = sizeof (bpreq) * count; \
3576 if (bifstp->ifbpstp_len == 0) { \
3577 bifstp->ifbpstp_len = buflen; \
3578 return (0); \
3579 } \
3580 \
3581 BRIDGE_UNLOCK(sc); \
3582 outbuf = (char *)kalloc_data(buflen, Z_WAITOK | Z_ZERO); \
3583 BRIDGE_LOCK(sc); \
3584 \
3585 count = 0; \
3586 buf = outbuf; \
3587 len = min(bifstp->ifbpstp_len, buflen); \
3588 bzero(&bpreq, sizeof (bpreq)); \
3589 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) { \
3590 if (len < sizeof (bpreq)) \
3591 break; \
3592 \
3593 if ((bif->bif_ifflags & IFBIF_STP) == 0) \
3594 continue; \
3595 \
3596 bp = &bif->bif_stp; \
3597 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff; \
3598 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions; \
3599 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost; \
3600 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id; \
3601 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id; \
3602 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id; \
3603 \
3604 memcpy(buf, &bpreq, sizeof (bpreq)); \
3605 count++; \
3606 buf += sizeof (bpreq); \
3607 len -= sizeof (bpreq); \
3608 } \
3609 \
3610 BRIDGE_UNLOCK(sc); \
3611 bifstp->ifbpstp_len = sizeof (bpreq) * count; \
3612 error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len); \
3613 BRIDGE_LOCK(sc); \
3614 kfree_data(outbuf, buflen); \
3615 return (error); \
3616 } while (0)
3617
3618 static int
bridge_ioctl_gifsstp32(struct bridge_softc * sc,void * arg)3619 bridge_ioctl_gifsstp32(struct bridge_softc *sc, void *arg)
3620 {
3621 struct ifbpstpconf32 *bifstp = arg;
3622 int error = 0;
3623
3624 BRIDGE_IOCTL_GIFSSTP;
3625 return error;
3626 }
3627
3628 static int
bridge_ioctl_gifsstp64(struct bridge_softc * sc,void * arg)3629 bridge_ioctl_gifsstp64(struct bridge_softc *sc, void *arg)
3630 {
3631 struct ifbpstpconf64 *bifstp = arg;
3632 int error = 0;
3633
3634 BRIDGE_IOCTL_GIFSSTP;
3635 return error;
3636 }
3637
3638 static int
bridge_ioctl_sproto(struct bridge_softc * sc,void * arg)3639 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
3640 {
3641 #if BRIDGESTP
3642 struct ifbrparam *param = arg;
3643
3644 return bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto);
3645 #else /* !BRIDGESTP */
3646 #pragma unused(sc, arg)
3647 return EOPNOTSUPP;
3648 #endif /* !BRIDGESTP */
3649 }
3650
3651 static int
bridge_ioctl_stxhc(struct bridge_softc * sc,void * arg)3652 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
3653 {
3654 #if BRIDGESTP
3655 struct ifbrparam *param = arg;
3656
3657 return bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc);
3658 #else /* !BRIDGESTP */
3659 #pragma unused(sc, arg)
3660 return EOPNOTSUPP;
3661 #endif /* !BRIDGESTP */
3662 }
3663
3664
3665 static int
bridge_ioctl_ghostfilter(struct bridge_softc * sc,void * arg)3666 bridge_ioctl_ghostfilter(struct bridge_softc *sc, void *arg)
3667 {
3668 struct ifbrhostfilter *req = arg;
3669 struct bridge_iflist *bif;
3670
3671 bif = bridge_lookup_member(sc, req->ifbrhf_ifsname);
3672 if (bif == NULL) {
3673 return ENOENT;
3674 }
3675
3676 bzero(req, sizeof(struct ifbrhostfilter));
3677 if (bif->bif_flags & BIFF_HOST_FILTER) {
3678 req->ifbrhf_flags |= IFBRHF_ENABLED;
3679 bcopy(bif->bif_hf_hwsrc, req->ifbrhf_hwsrca,
3680 ETHER_ADDR_LEN);
3681 req->ifbrhf_ipsrc = bif->bif_hf_ipsrc.s_addr;
3682 }
3683 return 0;
3684 }
3685
3686 static int
bridge_ioctl_shostfilter(struct bridge_softc * sc,void * arg)3687 bridge_ioctl_shostfilter(struct bridge_softc *sc, void *arg)
3688 {
3689 struct ifbrhostfilter *req = arg;
3690 struct bridge_iflist *bif;
3691
3692 bif = bridge_lookup_member(sc, req->ifbrhf_ifsname);
3693 if (bif == NULL) {
3694 return ENOENT;
3695 }
3696
3697 if (req->ifbrhf_flags & IFBRHF_ENABLED) {
3698 bif->bif_flags |= BIFF_HOST_FILTER;
3699
3700 if (req->ifbrhf_flags & IFBRHF_HWSRC) {
3701 bcopy(req->ifbrhf_hwsrca, bif->bif_hf_hwsrc,
3702 ETHER_ADDR_LEN);
3703 if (bcmp(req->ifbrhf_hwsrca, ethernulladdr,
3704 ETHER_ADDR_LEN) != 0) {
3705 bif->bif_flags |= BIFF_HF_HWSRC;
3706 } else {
3707 bif->bif_flags &= ~BIFF_HF_HWSRC;
3708 }
3709 }
3710 if (req->ifbrhf_flags & IFBRHF_IPSRC) {
3711 bif->bif_hf_ipsrc.s_addr = req->ifbrhf_ipsrc;
3712 if (bif->bif_hf_ipsrc.s_addr != INADDR_ANY) {
3713 bif->bif_flags |= BIFF_HF_IPSRC;
3714 } else {
3715 bif->bif_flags &= ~BIFF_HF_IPSRC;
3716 }
3717 }
3718 } else {
3719 bif->bif_flags &= ~(BIFF_HOST_FILTER | BIFF_HF_HWSRC |
3720 BIFF_HF_IPSRC);
3721 bzero(bif->bif_hf_hwsrc, ETHER_ADDR_LEN);
3722 bif->bif_hf_ipsrc.s_addr = INADDR_ANY;
3723 }
3724
3725 return 0;
3726 }
3727
3728 static char *
bridge_mac_nat_entry_out(struct mac_nat_entry_list * list,unsigned int * count_p,char * buf,unsigned int * len_p)3729 bridge_mac_nat_entry_out(struct mac_nat_entry_list * list,
3730 unsigned int * count_p, char *buf, unsigned int *len_p)
3731 {
3732 unsigned int count = *count_p;
3733 struct ifbrmne ifbmne;
3734 unsigned int len = *len_p;
3735 struct mac_nat_entry *mne;
3736 unsigned long now;
3737
3738 bzero(&ifbmne, sizeof(ifbmne));
3739 LIST_FOREACH(mne, list, mne_list) {
3740 if (len < sizeof(ifbmne)) {
3741 break;
3742 }
3743 snprintf(ifbmne.ifbmne_ifname, sizeof(ifbmne.ifbmne_ifname),
3744 "%s", mne->mne_bif->bif_ifp->if_xname);
3745 memcpy(ifbmne.ifbmne_mac, mne->mne_mac,
3746 sizeof(ifbmne.ifbmne_mac));
3747 now = (unsigned long) net_uptime();
3748 if (now < mne->mne_expire) {
3749 ifbmne.ifbmne_expire = mne->mne_expire - now;
3750 } else {
3751 ifbmne.ifbmne_expire = 0;
3752 }
3753 if ((mne->mne_flags & MNE_FLAGS_IPV6) != 0) {
3754 ifbmne.ifbmne_af = AF_INET6;
3755 ifbmne.ifbmne_ip6_addr = mne->mne_ip6;
3756 } else {
3757 ifbmne.ifbmne_af = AF_INET;
3758 ifbmne.ifbmne_ip_addr = mne->mne_ip;
3759 }
3760 memcpy(buf, &ifbmne, sizeof(ifbmne));
3761 count++;
3762 buf += sizeof(ifbmne);
3763 len -= sizeof(ifbmne);
3764 }
3765 *count_p = count;
3766 *len_p = len;
3767 return buf;
3768 }
3769
3770 /*
3771 * bridge_ioctl_gmnelist()
3772 * Perform the get mac_nat_entry list ioctl.
3773 *
3774 * Note:
3775 * The struct ifbrmnelist32 and struct ifbrmnelist64 have the same
3776 * field size/layout except for the last field ifbml_buf, the user-supplied
3777 * buffer pointer. That is passed in separately via the 'user_addr'
3778 * parameter from the respective 32-bit or 64-bit ioctl routine.
3779 */
3780 static int
bridge_ioctl_gmnelist(struct bridge_softc * sc,struct ifbrmnelist32 * mnl,user_addr_t user_addr)3781 bridge_ioctl_gmnelist(struct bridge_softc *sc, struct ifbrmnelist32 *mnl,
3782 user_addr_t user_addr)
3783 {
3784 unsigned int count;
3785 char *buf;
3786 int error = 0;
3787 char *outbuf = NULL;
3788 struct mac_nat_entry *mne;
3789 unsigned int buflen;
3790 unsigned int len;
3791
3792 mnl->ifbml_elsize = sizeof(struct ifbrmne);
3793 count = 0;
3794 LIST_FOREACH(mne, &sc->sc_mne_list, mne_list) {
3795 count++;
3796 }
3797 LIST_FOREACH(mne, &sc->sc_mne_list_v6, mne_list) {
3798 count++;
3799 }
3800 buflen = sizeof(struct ifbrmne) * count;
3801 if (buflen == 0 || mnl->ifbml_len == 0) {
3802 mnl->ifbml_len = buflen;
3803 return error;
3804 }
3805 BRIDGE_UNLOCK(sc);
3806 outbuf = (char *)kalloc_data(buflen, Z_WAITOK | Z_ZERO);
3807 BRIDGE_LOCK(sc);
3808 count = 0;
3809 buf = outbuf;
3810 len = min(mnl->ifbml_len, buflen);
3811 buf = bridge_mac_nat_entry_out(&sc->sc_mne_list, &count, buf, &len);
3812 buf = bridge_mac_nat_entry_out(&sc->sc_mne_list_v6, &count, buf, &len);
3813 mnl->ifbml_len = count * sizeof(struct ifbrmne);
3814 BRIDGE_UNLOCK(sc);
3815 error = copyout(outbuf, user_addr, mnl->ifbml_len);
3816 kfree_data(outbuf, buflen);
3817 BRIDGE_LOCK(sc);
3818 return error;
3819 }
3820
3821 static int
bridge_ioctl_gmnelist64(struct bridge_softc * sc,void * arg)3822 bridge_ioctl_gmnelist64(struct bridge_softc *sc, void *arg)
3823 {
3824 struct ifbrmnelist64 *mnl = arg;
3825
3826 return bridge_ioctl_gmnelist(sc, arg, mnl->ifbml_buf);
3827 }
3828
3829 static int
bridge_ioctl_gmnelist32(struct bridge_softc * sc,void * arg)3830 bridge_ioctl_gmnelist32(struct bridge_softc *sc, void *arg)
3831 {
3832 struct ifbrmnelist32 *mnl = arg;
3833
3834 return bridge_ioctl_gmnelist(sc, arg,
3835 CAST_USER_ADDR_T(mnl->ifbml_buf));
3836 }
3837
3838 /*
3839 * bridge_ioctl_gifstats()
3840 * Return per-member stats.
3841 *
3842 * Note:
3843 * The ifbrmreq32 and ifbrmreq64 structures have the same
3844 * field size/layout except for the last field brmr_buf, the user-supplied
3845 * buffer pointer. That is passed in separately via the 'user_addr'
3846 * parameter from the respective 32-bit or 64-bit ioctl routine.
3847 */
3848 static int
bridge_ioctl_gifstats(struct bridge_softc * sc,struct ifbrmreq32 * mreq,user_addr_t user_addr)3849 bridge_ioctl_gifstats(struct bridge_softc *sc, struct ifbrmreq32 *mreq,
3850 user_addr_t user_addr)
3851 {
3852 struct bridge_iflist *bif;
3853 int error = 0;
3854 unsigned int buflen;
3855
3856 bif = bridge_lookup_member(sc, mreq->brmr_ifname);
3857 if (bif == NULL) {
3858 error = ENOENT;
3859 goto done;
3860 }
3861
3862 buflen = mreq->brmr_elsize = sizeof(struct ifbrmstats);
3863 if (buflen == 0 || mreq->brmr_len == 0) {
3864 mreq->brmr_len = buflen;
3865 goto done;
3866 }
3867 if (mreq->brmr_len != 0 && mreq->brmr_len < buflen) {
3868 error = ENOBUFS;
3869 goto done;
3870 }
3871 mreq->brmr_len = buflen;
3872 error = copyout(&bif->bif_stats, user_addr, buflen);
3873 done:
3874 return error;
3875 }
3876
3877 static int
bridge_ioctl_gifstats32(struct bridge_softc * sc,void * arg)3878 bridge_ioctl_gifstats32(struct bridge_softc *sc, void *arg)
3879 {
3880 struct ifbrmreq32 *mreq = arg;
3881
3882 return bridge_ioctl_gifstats(sc, arg, mreq->brmr_buf);
3883 }
3884
3885 static int
bridge_ioctl_gifstats64(struct bridge_softc * sc,void * arg)3886 bridge_ioctl_gifstats64(struct bridge_softc *sc, void *arg)
3887 {
3888 struct ifbrmreq64 *mreq = arg;
3889
3890 return bridge_ioctl_gifstats(sc, arg, mreq->brmr_buf);
3891 }
3892
3893 /*
3894 * bridge_proto_attach_changed
3895 *
3896 * Called when protocol attachment on the interface changes.
3897 */
3898 static void
bridge_proto_attach_changed(struct ifnet * ifp)3899 bridge_proto_attach_changed(struct ifnet *ifp)
3900 {
3901 boolean_t changed = FALSE;
3902 struct bridge_iflist *bif;
3903 boolean_t input_broadcast;
3904 struct bridge_softc *sc = ifp->if_bridge;
3905
3906 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_LIFECYCLE, "%s", ifp->if_xname);
3907 if (sc == NULL) {
3908 return;
3909 }
3910 input_broadcast = interface_needs_input_broadcast(ifp);
3911 BRIDGE_LOCK(sc);
3912 bif = bridge_lookup_member_if(sc, ifp);
3913 if (bif != NULL) {
3914 changed = bif_set_input_broadcast(bif, input_broadcast);
3915 }
3916 BRIDGE_UNLOCK(sc);
3917 if (changed) {
3918 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_LIFECYCLE,
3919 "%s input broadcast %s", ifp->if_xname,
3920 input_broadcast ? "ENABLED" : "DISABLED");
3921 }
3922 return;
3923 }
3924
3925 /*
3926 * interface_media_active:
3927 *
3928 * Tells if an interface media is active.
3929 */
3930 static int
interface_media_active(struct ifnet * ifp)3931 interface_media_active(struct ifnet *ifp)
3932 {
3933 struct ifmediareq ifmr;
3934 int status = 0;
3935
3936 bzero(&ifmr, sizeof(ifmr));
3937 if (ifnet_ioctl(ifp, 0, SIOCGIFMEDIA, &ifmr) == 0) {
3938 if ((ifmr.ifm_status & IFM_AVALID) && ifmr.ifm_count > 0) {
3939 status = ifmr.ifm_status & IFM_ACTIVE ? 1 : 0;
3940 }
3941 }
3942
3943 return status;
3944 }
3945
3946 /*
3947 * bridge_updatelinkstatus:
3948 *
3949 * Update the media active status of the bridge based on the
3950 * media active status of its member.
3951 * If changed, return the corresponding onf/off link event.
3952 */
3953 static u_int32_t
bridge_updatelinkstatus(struct bridge_softc * sc)3954 bridge_updatelinkstatus(struct bridge_softc *sc)
3955 {
3956 struct bridge_iflist *bif;
3957 int active_member = 0;
3958 u_int32_t event_code = 0;
3959
3960 BRIDGE_LOCK_ASSERT_HELD(sc);
3961
3962 /*
3963 * Find out if we have an active interface
3964 */
3965 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) {
3966 if (bif->bif_flags & BIFF_MEDIA_ACTIVE) {
3967 active_member = 1;
3968 break;
3969 }
3970 }
3971
3972 if (active_member && !(sc->sc_flags & SCF_MEDIA_ACTIVE)) {
3973 sc->sc_flags |= SCF_MEDIA_ACTIVE;
3974 event_code = KEV_DL_LINK_ON;
3975 } else if (!active_member && (sc->sc_flags & SCF_MEDIA_ACTIVE)) {
3976 sc->sc_flags &= ~SCF_MEDIA_ACTIVE;
3977 event_code = KEV_DL_LINK_OFF;
3978 }
3979
3980 return event_code;
3981 }
3982
3983 /*
3984 * bridge_iflinkevent:
3985 */
3986 static void
bridge_iflinkevent(struct ifnet * ifp)3987 bridge_iflinkevent(struct ifnet *ifp)
3988 {
3989 struct bridge_softc *sc = ifp->if_bridge;
3990 struct bridge_iflist *bif;
3991 u_int32_t event_code = 0;
3992 int media_active;
3993
3994 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_LIFECYCLE, "%s", ifp->if_xname);
3995
3996 /* Check if the interface is a bridge member */
3997 if (sc == NULL) {
3998 return;
3999 }
4000
4001 media_active = interface_media_active(ifp);
4002 BRIDGE_LOCK(sc);
4003 bif = bridge_lookup_member_if(sc, ifp);
4004 if (bif != NULL) {
4005 if (media_active) {
4006 bif->bif_flags |= BIFF_MEDIA_ACTIVE;
4007 } else {
4008 bif->bif_flags &= ~BIFF_MEDIA_ACTIVE;
4009 }
4010 if (sc->sc_mac_nat_bif != NULL) {
4011 bridge_mac_nat_flush_entries(sc, bif);
4012 }
4013
4014 event_code = bridge_updatelinkstatus(sc);
4015 }
4016 BRIDGE_UNLOCK(sc);
4017
4018 if (event_code != 0) {
4019 bridge_link_event(sc->sc_ifp, event_code);
4020 }
4021 }
4022
4023 /*
4024 * bridge_delayed_callback:
4025 *
4026 * Makes a delayed call
4027 */
4028 static void
bridge_delayed_callback(void * param,__unused void * param2)4029 bridge_delayed_callback(void *param, __unused void *param2)
4030 {
4031 struct bridge_delayed_call *call = (struct bridge_delayed_call *)param;
4032 struct bridge_softc *sc = call->bdc_sc;
4033
4034 #if BRIDGE_DELAYED_CALLBACK_DEBUG
4035 if (bridge_delayed_callback_delay > 0) {
4036 struct timespec ts;
4037
4038 ts.tv_sec = bridge_delayed_callback_delay;
4039 ts.tv_nsec = 0;
4040
4041 BRIDGE_LOG(LOG_NOTICE, 0,
4042 "sleeping for %d seconds",
4043 bridge_delayed_callback_delay);
4044
4045 msleep(&bridge_delayed_callback_delay, NULL, PZERO,
4046 __func__, &ts);
4047
4048 BRIDGE_LOG(LOG_NOTICE, 0, "awoken");
4049 }
4050 #endif /* BRIDGE_DELAYED_CALLBACK_DEBUG */
4051
4052 BRIDGE_LOCK(sc);
4053
4054 #if BRIDGE_DELAYED_CALLBACK_DEBUG
4055 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_DELAYED_CALL,
4056 "%s call 0x%llx flags 0x%x",
4057 sc->sc_if_xname, (uint64_t)VM_KERNEL_ADDRPERM(call),
4058 call->bdc_flags);
4059 }
4060 #endif /* BRIDGE_DELAYED_CALLBACK_DEBUG */
4061
4062 if (call->bdc_flags & BDCF_CANCELLING) {
4063 wakeup(call);
4064 } else {
4065 if ((sc->sc_flags & SCF_DETACHING) == 0) {
4066 (*call->bdc_func)(sc);
4067 }
4068 }
4069 call->bdc_flags &= ~BDCF_OUTSTANDING;
4070 BRIDGE_UNLOCK(sc);
4071 }
4072
4073 /*
4074 * bridge_schedule_delayed_call:
4075 *
4076 * Schedule a function to be called on a separate thread
4077 * The actual call may be scheduled to run at a given time or ASAP.
4078 */
4079 static void
4080 bridge_schedule_delayed_call(struct bridge_delayed_call *call)
4081 {
4082 uint64_t deadline = 0;
4083 struct bridge_softc *sc = call->bdc_sc;
4084
4085 BRIDGE_LOCK_ASSERT_HELD(sc);
4086
4087 if ((sc->sc_flags & SCF_DETACHING) ||
4088 (call->bdc_flags & (BDCF_OUTSTANDING | BDCF_CANCELLING))) {
4089 return;
4090 }
4091
4092 if (call->bdc_ts.tv_sec || call->bdc_ts.tv_nsec) {
4093 nanoseconds_to_absolutetime(
4094 (uint64_t)call->bdc_ts.tv_sec * NSEC_PER_SEC +
4095 call->bdc_ts.tv_nsec, &deadline);
4096 clock_absolutetime_interval_to_deadline(deadline, &deadline);
4097 }
4098
4099 call->bdc_flags = BDCF_OUTSTANDING;
4100
4101 #if BRIDGE_DELAYED_CALLBACK_DEBUG
4102 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_DELAYED_CALL,
4103 "%s call 0x%llx flags 0x%x",
4104 sc->sc_if_xname, (uint64_t)VM_KERNEL_ADDRPERM(call),
4105 call->bdc_flags);
4106 }
4107 #endif /* BRIDGE_DELAYED_CALLBACK_DEBUG */
4108
4109 if (call->bdc_ts.tv_sec || call->bdc_ts.tv_nsec) {
4110 thread_call_func_delayed(
4111 (thread_call_func_t)bridge_delayed_callback,
4112 call, deadline);
4113 } else {
4114 if (call->bdc_thread_call == NULL) {
4115 call->bdc_thread_call = thread_call_allocate(
4116 (thread_call_func_t)bridge_delayed_callback,
4117 call);
4118 }
4119 thread_call_enter(call->bdc_thread_call);
4120 }
4121 }
4122
4123 /*
4124 * bridge_cancel_delayed_call:
4125 *
4126 * Cancel a queued or running delayed call.
4127 * If call is running, does not return until the call is done to
4128 * prevent race condition with the brigde interface getting destroyed
4129 */
4130 static void
4131 bridge_cancel_delayed_call(struct bridge_delayed_call *call)
4132 {
4133 boolean_t result;
4134 struct bridge_softc *sc = call->bdc_sc;
4135
4136 /*
4137 * The call was never scheduled
4138 */
4139 if (sc == NULL) {
4140 return;
4141 }
4142
4143 BRIDGE_LOCK_ASSERT_HELD(sc);
4144
4145 call->bdc_flags |= BDCF_CANCELLING;
4146
4147 while (call->bdc_flags & BDCF_OUTSTANDING) {
4148 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_DELAYED_CALL,
4149 "%s call 0x%llx flags 0x%x",
4150 sc->sc_if_xname, (uint64_t)VM_KERNEL_ADDRPERM(call),
4151 call->bdc_flags);
4152 result = thread_call_func_cancel(
4153 (thread_call_func_t)bridge_delayed_callback, call, FALSE);
4154
4155 if (result) {
4156 /*
4157 * We managed to dequeue the delayed call
4158 */
4159 call->bdc_flags &= ~BDCF_OUTSTANDING;
4160 } else {
4161 /*
4162 * Wait for delayed call do be done running
4163 */
4164 msleep(call, &sc->sc_mtx, PZERO, __func__, NULL);
4165 }
4166 }
4167 call->bdc_flags &= ~BDCF_CANCELLING;
4168 }
4169
4170 /*
4171 * bridge_cleanup_delayed_call:
4172 *
4173 * Dispose resource allocated for a delayed call
4174 * Assume the delayed call is not queued or running .
4175 */
4176 static void
4177 bridge_cleanup_delayed_call(struct bridge_delayed_call *call)
4178 {
4179 boolean_t result;
4180 struct bridge_softc *sc = call->bdc_sc;
4181
4182 /*
4183 * The call was never scheduled
4184 */
4185 if (sc == NULL) {
4186 return;
4187 }
4188
4189 BRIDGE_LOCK_ASSERT_HELD(sc);
4190
4191 VERIFY((call->bdc_flags & BDCF_OUTSTANDING) == 0);
4192 VERIFY((call->bdc_flags & BDCF_CANCELLING) == 0);
4193
4194 if (call->bdc_thread_call != NULL) {
4195 result = thread_call_free(call->bdc_thread_call);
4196 if (result == FALSE) {
4197 panic("%s thread_call_free() failed for call %p",
4198 __func__, call);
4199 }
4200 call->bdc_thread_call = NULL;
4201 }
4202 }
4203
4204 /*
4205 * bridge_init:
4206 *
4207 * Initialize a bridge interface.
4208 */
4209 static int
4210 bridge_init(struct ifnet *ifp)
4211 {
4212 struct bridge_softc *sc = (struct bridge_softc *)ifp->if_softc;
4213 errno_t error;
4214
4215 BRIDGE_LOCK_ASSERT_HELD(sc);
4216
4217 if ((ifnet_flags(ifp) & IFF_RUNNING)) {
4218 return 0;
4219 }
4220
4221 error = ifnet_set_flags(ifp, IFF_RUNNING, IFF_RUNNING);
4222
4223 /*
4224 * Calling bridge_aging_timer() is OK as there are no entries to
4225 * age so we're just going to arm the timer
4226 */
4227 bridge_aging_timer(sc);
4228 #if BRIDGESTP
4229 if (error == 0) {
4230 bstp_init(&sc->sc_stp); /* Initialize Spanning Tree */
4231 }
4232 #endif /* BRIDGESTP */
4233 return error;
4234 }
4235
4236 /*
4237 * bridge_ifstop:
4238 *
4239 * Stop the bridge interface.
4240 */
4241 static void
4242 bridge_ifstop(struct ifnet *ifp, int disable)
4243 {
4244 #pragma unused(disable)
4245 struct bridge_softc *sc = ifp->if_softc;
4246
4247 BRIDGE_LOCK_ASSERT_HELD(sc);
4248
4249 if ((ifnet_flags(ifp) & IFF_RUNNING) == 0) {
4250 return;
4251 }
4252
4253 bridge_cancel_delayed_call(&sc->sc_aging_timer);
4254
4255 #if BRIDGESTP
4256 bstp_stop(&sc->sc_stp);
4257 #endif /* BRIDGESTP */
4258
4259 bridge_rtflush(sc, IFBF_FLUSHDYN);
4260 (void) ifnet_set_flags(ifp, 0, IFF_RUNNING);
4261 }
4262
4263 /*
4264 * bridge_compute_cksum:
4265 *
4266 * If the packet has checksum flags, compare the hardware checksum
4267 * capabilities of the source and destination interfaces. If they
4268 * are the same, there's nothing to do. If they are different,
4269 * finalize the checksum so that it can be sent on the destination
4270 * interface.
4271 */
4272 static void
4273 bridge_compute_cksum(struct ifnet *src_if, struct ifnet *dst_if, struct mbuf *m)
4274 {
4275 uint32_t csum_flags;
4276 uint16_t dst_hw_csum;
4277 uint32_t did_sw = 0;
4278 struct ether_header *eh;
4279 uint16_t src_hw_csum;
4280
4281 if (src_if == dst_if) {
4282 return;
4283 }
4284 csum_flags = m->m_pkthdr.csum_flags & IF_HWASSIST_CSUM_MASK;
4285 if (csum_flags == 0) {
4286 /* no checksum offload */
4287 return;
4288 }
4289
4290 /*
4291 * if destination/source differ in checksum offload
4292 * capabilities, finalize/compute the checksum
4293 */
4294 dst_hw_csum = IF_HWASSIST_CSUM_FLAGS(dst_if->if_hwassist);
4295 src_hw_csum = IF_HWASSIST_CSUM_FLAGS(src_if->if_hwassist);
4296 if (dst_hw_csum == src_hw_csum) {
4297 return;
4298 }
4299 eh = mtod(m, struct ether_header *);
4300 switch (ntohs(eh->ether_type)) {
4301 case ETHERTYPE_IP:
4302 did_sw = in_finalize_cksum(m, sizeof(*eh), csum_flags);
4303 break;
4304 case ETHERTYPE_IPV6:
4305 did_sw = in6_finalize_cksum(m, sizeof(*eh), -1, -1, csum_flags);
4306 break;
4307 }
4308 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4309 "[%s -> %s] before 0x%x did 0x%x after 0x%x",
4310 src_if->if_xname, dst_if->if_xname, csum_flags, did_sw,
4311 m->m_pkthdr.csum_flags);
4312 }
4313
4314 static errno_t
4315 bridge_transmit(struct ifnet * ifp, struct mbuf *m)
4316 {
4317 struct flowadv adv = { .code = FADV_SUCCESS };
4318 errno_t error;
4319
4320 error = dlil_output(ifp, 0, m, NULL, NULL, 1, &adv);
4321 if (error == 0) {
4322 if (adv.code == FADV_FLOW_CONTROLLED) {
4323 error = EQFULL;
4324 } else if (adv.code == FADV_SUSPENDED) {
4325 error = EQSUSPENDED;
4326 }
4327 }
4328 return error;
4329 }
4330
4331 static int
4332 get_last_ip6_hdr(struct mbuf *m, int off, int proto, int * nxtp,
4333 bool *is_fragmented)
4334 {
4335 int newoff;
4336
4337 *is_fragmented = false;
4338 while (1) {
4339 newoff = ip6_nexthdr(m, off, proto, nxtp);
4340 if (newoff < 0) {
4341 return off;
4342 } else if (newoff < off) {
4343 return -1; /* invalid */
4344 } else if (newoff == off) {
4345 return newoff;
4346 }
4347 off = newoff;
4348 proto = *nxtp;
4349 if (proto == IPPROTO_FRAGMENT) {
4350 *is_fragmented = true;
4351 }
4352 }
4353 }
4354
4355 static int
4356 bridge_get_ip_proto(struct mbuf * * mp, u_int mac_hlen, bool is_ipv4,
4357 ip_packet_info_t info_p, struct bripstats * stats_p)
4358 {
4359 int error = 0;
4360 u_int hlen;
4361 u_int ip_hlen;
4362 u_int ip_pay_len;
4363 struct mbuf * m0 = *mp;
4364 int off;
4365 int opt_len = 0;
4366 int proto = 0;
4367
4368 bzero(info_p, sizeof(*info_p));
4369 if (is_ipv4) {
4370 struct ip * ip;
4371 u_int ip_total_len;
4372
4373 /* IPv4 */
4374 hlen = mac_hlen + sizeof(struct ip);
4375 if (m0->m_pkthdr.len < hlen) {
4376 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4377 "Short IP packet %d < %d",
4378 m0->m_pkthdr.len, hlen);
4379 error = _EBADIP;
4380 stats_p->bips_bad_ip++;
4381 goto done;
4382 }
4383 if (m0->m_len < hlen) {
4384 *mp = m0 = m_pullup(m0, hlen);
4385 if (m0 == NULL) {
4386 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4387 "m_pullup failed hlen %d",
4388 hlen);
4389 error = ENOBUFS;
4390 stats_p->bips_bad_ip++;
4391 goto done;
4392 }
4393 }
4394 ip = (struct ip *)(void *)(mtod(m0, uint8_t *) + mac_hlen);
4395 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
4396 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4397 "bad IP version");
4398 error = _EBADIP;
4399 stats_p->bips_bad_ip++;
4400 goto done;
4401 }
4402 ip_hlen = IP_VHL_HL(ip->ip_vhl) << 2;
4403 if (ip_hlen < sizeof(struct ip)) {
4404 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4405 "bad IP header length %d < %d",
4406 ip_hlen,
4407 (int)sizeof(struct ip));
4408 error = _EBADIP;
4409 stats_p->bips_bad_ip++;
4410 goto done;
4411 }
4412 hlen = mac_hlen + ip_hlen;
4413 if (m0->m_len < hlen) {
4414 *mp = m0 = m_pullup(m0, hlen);
4415 if (m0 == NULL) {
4416 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4417 "m_pullup failed hlen %d",
4418 hlen);
4419 error = ENOBUFS;
4420 stats_p->bips_bad_ip++;
4421 goto done;
4422 }
4423 }
4424
4425 ip_total_len = ntohs(ip->ip_len);
4426 if (ip_total_len < ip_hlen) {
4427 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4428 "IP total len %d < header len %d",
4429 ip_total_len, ip_hlen);
4430 error = _EBADIP;
4431 stats_p->bips_bad_ip++;
4432 goto done;
4433 }
4434 if (ip_total_len > (m0->m_pkthdr.len - mac_hlen)) {
4435 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4436 "invalid IP payload length %d > %d",
4437 ip_total_len,
4438 (m0->m_pkthdr.len - mac_hlen));
4439 error = _EBADIP;
4440 stats_p->bips_bad_ip++;
4441 goto done;
4442 }
4443 ip_pay_len = ip_total_len - ip_hlen;
4444 info_p->ip_proto = ip->ip_p;
4445 info_p->ip_hdr.ip = ip;
4446 #define FRAG_BITS (IP_OFFMASK | IP_MF)
4447 if ((ntohs(ip->ip_off) & FRAG_BITS) != 0) {
4448 info_p->ip_is_fragmented = true;
4449 }
4450 stats_p->bips_ip++;
4451 } else {
4452 struct ip6_hdr *ip6;
4453
4454 /* IPv6 */
4455 hlen = mac_hlen + sizeof(struct ip6_hdr);
4456 if (m0->m_pkthdr.len < hlen) {
4457 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4458 "short IPv6 packet %d < %d",
4459 m0->m_pkthdr.len, hlen);
4460 error = _EBADIPV6;
4461 stats_p->bips_bad_ip6++;
4462 goto done;
4463 }
4464 if (m0->m_len < hlen) {
4465 *mp = m0 = m_pullup(m0, hlen);
4466 if (m0 == NULL) {
4467 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4468 "m_pullup failed hlen %d",
4469 hlen);
4470 error = ENOBUFS;
4471 stats_p->bips_bad_ip6++;
4472 goto done;
4473 }
4474 }
4475 ip6 = (struct ip6_hdr *)(mtod(m0, uint8_t *) + mac_hlen);
4476 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
4477 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4478 "bad IPv6 version");
4479 error = _EBADIPV6;
4480 stats_p->bips_bad_ip6++;
4481 goto done;
4482 }
4483 off = get_last_ip6_hdr(m0, mac_hlen, IPPROTO_IPV6, &proto,
4484 &info_p->ip_is_fragmented);
4485 if (off < 0 || m0->m_pkthdr.len < off) {
4486 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4487 "ip6_lasthdr() returned %d",
4488 off);
4489 error = _EBADIPV6;
4490 stats_p->bips_bad_ip6++;
4491 goto done;
4492 }
4493 ip_hlen = sizeof(*ip6);
4494 opt_len = off - mac_hlen - ip_hlen;
4495 if (opt_len < 0) {
4496 error = _EBADIPV6;
4497 stats_p->bips_bad_ip6++;
4498 goto done;
4499 }
4500 info_p->ip_proto = proto;
4501 info_p->ip_hdr.ip6 = ip6;
4502 ip_pay_len = ntohs(ip6->ip6_plen);
4503 if (ip_pay_len > (m0->m_pkthdr.len - mac_hlen - ip_hlen)) {
4504 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4505 "invalid IPv6 payload length %d > %d",
4506 ip_pay_len,
4507 (m0->m_pkthdr.len - mac_hlen - ip_hlen));
4508 error = _EBADIPV6;
4509 stats_p->bips_bad_ip6++;
4510 goto done;
4511 }
4512 stats_p->bips_ip6++;
4513 }
4514 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4515 "IPv%c proto %d ip %u pay %u opt %u pkt %u%s",
4516 is_ipv4 ? '4' : '6',
4517 proto, ip_hlen, ip_pay_len, opt_len,
4518 m0->m_pkthdr.len, info_p->ip_is_fragmented ? " frag" : "");
4519 info_p->ip_hlen = ip_hlen;
4520 info_p->ip_pay_len = ip_pay_len;
4521 info_p->ip_opt_len = opt_len;
4522 info_p->ip_is_ipv4 = is_ipv4;
4523 done:
4524 return error;
4525 }
4526
4527 static int
4528 bridge_get_tcp_header(struct mbuf * * mp, u_int mac_hlen, bool is_ipv4,
4529 ip_packet_info_t info_p, struct bripstats * stats_p)
4530 {
4531 int error;
4532 u_int hlen;
4533
4534 error = bridge_get_ip_proto(mp, mac_hlen, is_ipv4, info_p, stats_p);
4535 if (error != 0) {
4536 goto done;
4537 }
4538 if (info_p->ip_proto != IPPROTO_TCP) {
4539 /* not a TCP frame, not an error, just a bad guess */
4540 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4541 "non-TCP (%d) IPv%c frame %d bytes",
4542 info_p->ip_proto, is_ipv4 ? '4' : '6',
4543 (*mp)->m_pkthdr.len);
4544 goto done;
4545 }
4546 if (info_p->ip_is_fragmented) {
4547 /* both TSO and IP fragmentation don't make sense */
4548 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_CHECKSUM,
4549 "fragmented TSO packet?");
4550 stats_p->bips_bad_tcp++;
4551 error = _EBADTCP;
4552 goto done;
4553 }
4554 hlen = mac_hlen + info_p->ip_hlen + sizeof(struct tcphdr) +
4555 info_p->ip_opt_len;
4556 if ((*mp)->m_len < hlen) {
4557 *mp = m_pullup(*mp, hlen);
4558 if (*mp == NULL) {
4559 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4560 "m_pullup %d failed",
4561 hlen);
4562 stats_p->bips_bad_tcp++;
4563 error = _EBADTCP;
4564 goto done;
4565 }
4566 }
4567 info_p->ip_proto_hdr = ((caddr_t)info_p->ip_hdr.ptr) +
4568 info_p->ip_hlen + info_p->ip_opt_len;
4569 done:
4570 return error;
4571 }
4572
4573 static inline void
4574 proto_csum_stats_increment(uint8_t proto, struct brcsumstats * stats_p)
4575 {
4576 if (proto == IPPROTO_TCP) {
4577 stats_p->brcs_tcp_checksum++;
4578 } else {
4579 stats_p->brcs_udp_checksum++;
4580 }
4581 return;
4582 }
4583
4584 static bool
4585 ether_header_type_is_ip(struct ether_header * eh, bool *is_ipv4)
4586 {
4587 uint16_t ether_type;
4588 bool is_ip = TRUE;
4589
4590 ether_type = ntohs(eh->ether_type);
4591 switch (ether_type) {
4592 case ETHERTYPE_IP:
4593 *is_ipv4 = TRUE;
4594 break;
4595 case ETHERTYPE_IPV6:
4596 *is_ipv4 = FALSE;
4597 break;
4598 default:
4599 is_ip = FALSE;
4600 break;
4601 }
4602 return is_ip;
4603 }
4604
4605 static errno_t
4606 bridge_verify_checksum(struct mbuf * * mp, struct ifbrmstats *stats_p)
4607 {
4608 struct brcsumstats *csum_stats_p;
4609 struct ether_header *eh;
4610 errno_t error = 0;
4611 ip_packet_info info;
4612 bool is_ipv4;
4613 struct mbuf * m;
4614 u_int mac_hlen = sizeof(struct ether_header);
4615 uint16_t sum;
4616 bool valid;
4617
4618 eh = mtod(*mp, struct ether_header *);
4619 if (!ether_header_type_is_ip(eh, &is_ipv4)) {
4620 goto done;
4621 }
4622 error = bridge_get_ip_proto(mp, mac_hlen, is_ipv4, &info,
4623 &stats_p->brms_out_ip);
4624 m = *mp;
4625 if (error != 0) {
4626 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4627 "bridge_get_ip_proto failed %d",
4628 error);
4629 goto done;
4630 }
4631 if (is_ipv4) {
4632 if ((m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) != 0) {
4633 /* hardware offloaded IP header checksum */
4634 valid = (m->m_pkthdr.csum_flags & CSUM_IP_VALID) != 0;
4635 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4636 "IP checksum HW %svalid",
4637 valid ? "" : "in");
4638 if (!valid) {
4639 stats_p->brms_out_cksum_bad_hw.brcs_ip_checksum++;
4640 error = _EBADIPCHECKSUM;
4641 goto done;
4642 }
4643 stats_p->brms_out_cksum_good_hw.brcs_ip_checksum++;
4644 } else {
4645 /* verify */
4646 sum = inet_cksum(m, 0, mac_hlen, info.ip_hlen);
4647 valid = (sum == 0);
4648 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4649 "IP checksum SW %svalid",
4650 valid ? "" : "in");
4651 if (!valid) {
4652 stats_p->brms_out_cksum_bad.brcs_ip_checksum++;
4653 error = _EBADIPCHECKSUM;
4654 goto done;
4655 }
4656 stats_p->brms_out_cksum_good.brcs_ip_checksum++;
4657 }
4658 }
4659 if (info.ip_is_fragmented) {
4660 /* can't verify checksum on fragmented packets */
4661 goto done;
4662 }
4663 switch (info.ip_proto) {
4664 case IPPROTO_TCP:
4665 stats_p->brms_out_ip.bips_tcp++;
4666 break;
4667 case IPPROTO_UDP:
4668 stats_p->brms_out_ip.bips_udp++;
4669 break;
4670 default:
4671 goto done;
4672 }
4673 /* check for hardware offloaded UDP/TCP checksum */
4674 #define HW_CSUM (CSUM_DATA_VALID | CSUM_PSEUDO_HDR)
4675 if ((m->m_pkthdr.csum_flags & HW_CSUM) == HW_CSUM) {
4676 /* checksum verified by hardware */
4677 valid = (m->m_pkthdr.csum_rx_val == 0xffff);
4678 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4679 "IPv%c %s checksum HW 0x%x %svalid",
4680 is_ipv4 ? '4' : '6',
4681 (info.ip_proto == IPPROTO_TCP)
4682 ? "TCP" : "UDP",
4683 m->m_pkthdr.csum_data,
4684 valid ? "" : "in" );
4685 if (!valid) {
4686 /* bad checksum */
4687 csum_stats_p = &stats_p->brms_out_cksum_bad_hw;
4688 error = (info.ip_proto == IPPROTO_TCP) ? _EBADTCPCHECKSUM
4689 : _EBADTCPCHECKSUM;
4690 } else {
4691 /* good checksum */
4692 csum_stats_p = &stats_p->brms_out_cksum_good_hw;
4693 }
4694 proto_csum_stats_increment(info.ip_proto, csum_stats_p);
4695 goto done;
4696 }
4697 m->m_data += mac_hlen;
4698 m->m_len -= mac_hlen;
4699 m->m_pkthdr.len -= mac_hlen;
4700 if (is_ipv4) {
4701 sum = inet_cksum(m, info.ip_proto,
4702 info.ip_hlen,
4703 info.ip_pay_len);
4704 } else {
4705 sum = inet6_cksum(m, info.ip_proto,
4706 info.ip_hlen + info.ip_opt_len,
4707 info.ip_pay_len - info.ip_opt_len);
4708 }
4709 valid = (sum == 0);
4710 if (valid) {
4711 csum_stats_p = &stats_p->brms_out_cksum_good;
4712 } else {
4713 csum_stats_p = &stats_p->brms_out_cksum_bad;
4714 error = (info.ip_proto == IPPROTO_TCP)
4715 ? _EBADTCPCHECKSUM : _EBADUDPCHECKSUM;
4716 }
4717 proto_csum_stats_increment(info.ip_proto, csum_stats_p);
4718 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4719 "IPv%c %s checksum SW %svalid (0x%x) hlen %d paylen %d",
4720 is_ipv4 ? '4' : '6',
4721 (info.ip_proto == IPPROTO_TCP) ? "TCP" : "UDP",
4722 valid ? "" : "in",
4723 sum, info.ip_hlen, info.ip_pay_len);
4724 m->m_data -= mac_hlen;
4725 m->m_len += mac_hlen;
4726 m->m_pkthdr.len += mac_hlen;
4727 done:
4728 return error;
4729 }
4730
4731 static errno_t
4732 bridge_offload_checksum(struct mbuf * * mp, ip_packet_info * info_p,
4733 struct ifbrmstats * stats_p)
4734 {
4735 uint16_t * csum_p;
4736 errno_t error = 0;
4737 u_int hlen;
4738 struct mbuf * m0 = *mp;
4739 u_int mac_hlen = sizeof(struct ether_header);
4740 u_int pkt_hdr_len;
4741 struct tcphdr * tcp;
4742 u_int tcp_hlen;
4743 struct udphdr * udp;
4744
4745 if (info_p->ip_is_ipv4) {
4746 /* compute IP header checksum */
4747 info_p->ip_hdr.ip->ip_sum = 0;
4748 info_p->ip_hdr.ip->ip_sum = inet_cksum(m0, 0, mac_hlen,
4749 info_p->ip_hlen);
4750 stats_p->brms_in_computed_cksum.brcs_ip_checksum++;
4751 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4752 "IPv4 checksum 0x%x",
4753 ntohs(info_p->ip_hdr.ip->ip_sum));
4754 }
4755 if (info_p->ip_is_fragmented) {
4756 /* can't compute checksum on fragmented packets */
4757 goto done;
4758 }
4759 pkt_hdr_len = m0->m_pkthdr.len;
4760 switch (info_p->ip_proto) {
4761 case IPPROTO_TCP:
4762 hlen = mac_hlen + info_p->ip_hlen + info_p->ip_opt_len
4763 + sizeof(struct tcphdr);
4764 if (m0->m_len < hlen) {
4765 *mp = m0 = m_pullup(m0, hlen);
4766 if (m0 == NULL) {
4767 stats_p->brms_in_ip.bips_bad_tcp++;
4768 error = _EBADTCP;
4769 goto done;
4770 }
4771 }
4772 tcp = (struct tcphdr *)(void *)
4773 ((caddr_t)info_p->ip_hdr.ptr + info_p->ip_hlen
4774 + info_p->ip_opt_len);
4775 tcp_hlen = tcp->th_off << 2;
4776 hlen = mac_hlen + info_p->ip_hlen + info_p->ip_opt_len + tcp_hlen;
4777 if (hlen > pkt_hdr_len) {
4778 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4779 "bad tcp header length %u",
4780 tcp_hlen);
4781 stats_p->brms_in_ip.bips_bad_tcp++;
4782 error = _EBADTCP;
4783 goto done;
4784 }
4785 csum_p = &tcp->th_sum;
4786 stats_p->brms_in_ip.bips_tcp++;
4787 break;
4788 case IPPROTO_UDP:
4789 hlen = mac_hlen + info_p->ip_hlen + info_p->ip_opt_len + sizeof(*udp);
4790 if (m0->m_len < hlen) {
4791 *mp = m0 = m_pullup(m0, hlen);
4792 if (m0 == NULL) {
4793 stats_p->brms_in_ip.bips_bad_udp++;
4794 error = ENOBUFS;
4795 goto done;
4796 }
4797 }
4798 udp = (struct udphdr *)(void *)
4799 ((caddr_t)info_p->ip_hdr.ptr + info_p->ip_hlen
4800 + info_p->ip_opt_len);
4801 csum_p = &udp->uh_sum;
4802 stats_p->brms_in_ip.bips_udp++;
4803 break;
4804 default:
4805 /* not TCP or UDP */
4806 goto done;
4807 }
4808 *csum_p = 0;
4809 m0->m_data += mac_hlen;
4810 m0->m_len -= mac_hlen;
4811 m0->m_pkthdr.len -= mac_hlen;
4812 if (info_p->ip_is_ipv4) {
4813 *csum_p = inet_cksum(m0, info_p->ip_proto, info_p->ip_hlen,
4814 info_p->ip_pay_len);
4815 } else {
4816 *csum_p = inet6_cksum(m0, info_p->ip_proto,
4817 info_p->ip_hlen + info_p->ip_opt_len,
4818 info_p->ip_pay_len - info_p->ip_opt_len);
4819 }
4820 if (info_p->ip_proto == IPPROTO_UDP && *csum_p == 0) {
4821 /* RFC 1122 4.1.3.4 */
4822 *csum_p = 0xffff;
4823 }
4824 m0->m_data -= mac_hlen;
4825 m0->m_len += mac_hlen;
4826 m0->m_pkthdr.len += mac_hlen;
4827 proto_csum_stats_increment(info_p->ip_proto,
4828 &stats_p->brms_in_computed_cksum);
4829
4830 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4831 "IPv%c %s set checksum 0x%x",
4832 info_p->ip_is_ipv4 ? '4' : '6',
4833 (info_p->ip_proto == IPPROTO_TCP) ? "TCP" : "UDP",
4834 ntohs(*csum_p));
4835 done:
4836 return error;
4837 }
4838
4839 static errno_t
4840 bridge_send(struct ifnet *src_ifp,
4841 struct ifnet *dst_ifp, struct mbuf *m, ChecksumOperation cksum_op)
4842 {
4843 switch (cksum_op) {
4844 case CHECKSUM_OPERATION_CLEAR_OFFLOAD:
4845 m->m_pkthdr.csum_flags &= ~CSUM_TX_FLAGS;
4846 break;
4847 case CHECKSUM_OPERATION_FINALIZE:
4848 /* the checksum might not be correct, finalize now */
4849 bridge_finalize_cksum(dst_ifp, m);
4850 break;
4851 case CHECKSUM_OPERATION_COMPUTE:
4852 bridge_compute_cksum(src_ifp, dst_ifp, m);
4853 break;
4854 default:
4855 break;
4856 }
4857 #if HAS_IF_CAP
4858 /*
4859 * If underlying interface can not do VLAN tag insertion itself
4860 * then attach a packet tag that holds it.
4861 */
4862 if ((m->m_flags & M_VLANTAG) &&
4863 (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
4864 m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
4865 if (m == NULL) {
4866 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_CHECKSUM,
4867 "%s: unable to prepend VLAN header",
4868 dst_ifp->if_xname);
4869 (void) ifnet_stat_increment_out(dst_ifp,
4870 0, 0, 1);
4871 return 0;
4872 }
4873 m->m_flags &= ~M_VLANTAG;
4874 }
4875 #endif /* HAS_IF_CAP */
4876 return bridge_transmit(dst_ifp, m);
4877 }
4878
4879 static errno_t
4880 bridge_send_tso(struct ifnet *dst_ifp, struct mbuf *m, bool is_ipv4)
4881 {
4882 errno_t error;
4883 u_int mac_hlen;
4884
4885 mac_hlen = sizeof(struct ether_header);
4886
4887 #if HAS_IF_CAP
4888 /*
4889 * If underlying interface can not do VLAN tag insertion itself
4890 * then attach a packet tag that holds it.
4891 */
4892 if ((m->m_flags & M_VLANTAG) &&
4893 (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
4894 m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
4895 if (m == NULL) {
4896 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_CHECKSUM,
4897 "%s: unable to prepend VLAN header",
4898 dst_ifp->if_xname);
4899 (void) ifnet_stat_increment_out(dst_ifp,
4900 0, 0, 1);
4901 error = ENOBUFS;
4902 goto done;
4903 }
4904 m->m_flags &= ~M_VLANTAG;
4905 mac_hlen += ETHER_VLAN_ENCAP_LEN;
4906 }
4907 #endif /* HAS_IF_CAP */
4908 error = gso_tcp(dst_ifp, &m, mac_hlen, is_ipv4, TRUE);
4909 return error;
4910 }
4911
4912 /*
4913 * tso_hwassist:
4914 * - determine whether the destination interface supports TSO offload
4915 * - if the packet is already marked for offload and the hardware supports
4916 * it, just allow the packet to continue on
4917 * - if not, parse the packet headers to verify that this is a large TCP
4918 * packet requiring segmentation; if the hardware doesn't support it
4919 * set need_sw_tso; otherwise, mark the packet for TSO offload
4920 */
4921 static int
4922 tso_hwassist(struct mbuf **mp, bool is_ipv4, struct ifnet * ifp, u_int mac_hlen,
4923 bool * need_sw_tso, bool * is_large_tcp)
4924 {
4925 int error = 0;
4926 u_int32_t if_csum;
4927 u_int32_t if_tso;
4928 u_int32_t mbuf_tso;
4929 bool supports_cksum = false;
4930
4931 *need_sw_tso = false;
4932 *is_large_tcp = false;
4933 if (is_ipv4) {
4934 /*
4935 * Enable both TCP and IP offload if the hardware supports it.
4936 * If the hardware doesn't support TCP offload, supports_cksum
4937 * will be false so we won't set either offload.
4938 */
4939 if_csum = ifp->if_hwassist & (CSUM_TCP | CSUM_IP);
4940 supports_cksum = (if_csum & CSUM_TCP) != 0;
4941 if_tso = IFNET_TSO_IPV4;
4942 mbuf_tso = CSUM_TSO_IPV4;
4943 } else {
4944 supports_cksum = (ifp->if_hwassist & CSUM_TCPIPV6) != 0;
4945 if_csum = CSUM_TCPIPV6;
4946 if_tso = IFNET_TSO_IPV6;
4947 mbuf_tso = CSUM_TSO_IPV6;
4948 }
4949 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
4950 "%s: does%s support checksum 0x%x if_csum 0x%x",
4951 ifp->if_xname, supports_cksum ? "" : " not",
4952 ifp->if_hwassist, if_csum);
4953 if ((ifp->if_hwassist & if_tso) != 0 &&
4954 ((*mp)->m_pkthdr.csum_flags & mbuf_tso) != 0) {
4955 /* hardware TSO, mbuf already marked */
4956 } else {
4957 /* verify that this is a large TCP frame */
4958 uint32_t csum_flags;
4959 ip_packet_info info;
4960 int mss;
4961 struct bripstats stats;
4962 struct tcphdr * tcp;
4963
4964 error = bridge_get_tcp_header(mp, mac_hlen, is_ipv4,
4965 &info, &stats);
4966 if (error != 0) {
4967 /* bad packet */
4968 goto done;
4969 }
4970 if ((info.ip_hlen + info.ip_pay_len + info.ip_opt_len) <=
4971 ifp->if_mtu) {
4972 /* not actually a large packet */
4973 goto done;
4974 }
4975 if (info.ip_proto_hdr == NULL) {
4976 /* not a TCP packet */
4977 goto done;
4978 }
4979 if ((ifp->if_hwassist & if_tso) == 0) {
4980 /* hardware does not support TSO, enable sw tso */
4981 *need_sw_tso = if_bridge_segmentation != 0;
4982 goto done;
4983 }
4984 /* use hardware TSO */
4985 (*mp)->m_pkthdr.pkt_proto = IPPROTO_TCP;
4986 tcp = (struct tcphdr *)info.ip_proto_hdr;
4987 mss = ifp->if_mtu - info.ip_hlen - info.ip_opt_len
4988 - (tcp->th_off << 2) - if_bridge_tso_reduce_mss_tx;
4989 assert(mss > 0);
4990 csum_flags = mbuf_tso;
4991 if (supports_cksum) {
4992 csum_flags |= if_csum;
4993 }
4994 (*mp)->m_pkthdr.tso_segsz = mss;
4995 (*mp)->m_pkthdr.csum_flags |= csum_flags;
4996 (*mp)->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum);
4997 *is_large_tcp = true;
4998 }
4999 done:
5000 return error;
5001 }
5002
5003 /*
5004 * bridge_enqueue:
5005 *
5006 * Enqueue a packet on a bridge member interface.
5007 *
5008 */
5009 static errno_t
5010 bridge_enqueue(ifnet_t bridge_ifp, struct ifnet *src_ifp,
5011 struct ifnet *dst_ifp, struct mbuf *m, ChecksumOperation cksum_op)
5012 {
5013 errno_t error = 0;
5014 int len;
5015
5016 VERIFY(dst_ifp != NULL);
5017
5018 /*
5019 * We may be sending a fragment so traverse the mbuf
5020 *
5021 * NOTE: bridge_fragment() is called only when PFIL_HOOKS is enabled.
5022 */
5023 for (struct mbuf *next_m = NULL; m != NULL; m = next_m) {
5024 bool need_sw_tso = false;
5025 bool is_ipv4 = false;
5026 bool is_large_pkt;
5027 errno_t _error = 0;
5028
5029 len = m->m_pkthdr.len;
5030 m->m_flags |= M_PROTO1; /* set to avoid loops */
5031 next_m = m->m_nextpkt;
5032 m->m_nextpkt = NULL;
5033 /*
5034 * Need to segment the packet if it is a large frame
5035 * and the destination interface does not support TSO.
5036 *
5037 * Note that with trailers, it's possible for a packet to
5038 * be large but not actually require segmentation.
5039 */
5040 is_large_pkt = (len > (bridge_ifp->if_mtu + ETHER_HDR_LEN));
5041 if (is_large_pkt) {
5042 struct ether_header *eh;
5043 bool is_large_tcp = false;
5044
5045 eh = mtod(m, struct ether_header *);
5046 if (ether_header_type_is_ip(eh, &is_ipv4)) {
5047 _error = tso_hwassist(&m, is_ipv4,
5048 dst_ifp, sizeof(struct ether_header),
5049 &need_sw_tso, &is_large_tcp);
5050 if (is_large_tcp) {
5051 cksum_op = CHECKSUM_OPERATION_NONE;
5052 }
5053 } else {
5054 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
5055 "large non IP packet");
5056 }
5057 }
5058 if (_error != 0) {
5059 if (m != NULL) {
5060 m_freem(m);
5061 }
5062 } else if (need_sw_tso) {
5063 _error = bridge_send_tso(dst_ifp, m, is_ipv4);
5064 } else {
5065 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
5066 "%s bridge_send(%s) len %d op %d",
5067 bridge_ifp->if_xname,
5068 dst_ifp->if_xname,
5069 len, cksum_op);
5070 _error = bridge_send(src_ifp, dst_ifp, m, cksum_op);
5071 }
5072
5073 /* Preserve first error value */
5074 if (error == 0 && _error != 0) {
5075 error = _error;
5076 }
5077 if (_error == 0) {
5078 (void) ifnet_stat_increment_out(bridge_ifp, 1, len, 0);
5079 } else {
5080 (void) ifnet_stat_increment_out(bridge_ifp, 0, 0, 1);
5081 }
5082 }
5083
5084 return error;
5085 }
5086
5087 #if HAS_BRIDGE_DUMMYNET
5088 /*
5089 * bridge_dummynet:
5090 *
5091 * Receive a queued packet from dummynet and pass it on to the output
5092 * interface.
5093 *
5094 * The mbuf has the Ethernet header already attached.
5095 */
5096 static void
5097 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
5098 {
5099 struct bridge_softc *sc;
5100
5101 sc = ifp->if_bridge;
5102
5103 /*
5104 * The packet didn't originate from a member interface. This should only
5105 * ever happen if a member interface is removed while packets are
5106 * queued for it.
5107 */
5108 if (sc == NULL) {
5109 m_freem(m);
5110 return;
5111 }
5112
5113 if (PFIL_HOOKED(&inet_pfil_hook) || PFIL_HOOKED_INET6) {
5114 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0) {
5115 return;
5116 }
5117 if (m == NULL) {
5118 return;
5119 }
5120 }
5121 (void) bridge_enqueue(sc->sc_ifp, NULL, ifp, m, CHECKSUM_OPERATION_NONE);
5122 }
5123
5124 #endif /* HAS_BRIDGE_DUMMYNET */
5125
5126 /*
5127 * bridge_member_output:
5128 *
5129 * Send output from a bridge member interface. This
5130 * performs the bridging function for locally originated
5131 * packets.
5132 *
5133 * The mbuf has the Ethernet header already attached.
5134 */
5135 static errno_t
5136 bridge_member_output(struct bridge_softc *sc, ifnet_t ifp, mbuf_t *data)
5137 {
5138 ifnet_t bridge_ifp;
5139 struct ether_header *eh;
5140 struct ifnet *dst_if;
5141 uint16_t vlan;
5142 struct bridge_iflist *mac_nat_bif;
5143 ifnet_t mac_nat_ifp;
5144 mbuf_t m = *data;
5145
5146 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_OUTPUT,
5147 "ifp %s", ifp->if_xname);
5148 if (m->m_len < ETHER_HDR_LEN) {
5149 m = m_pullup(m, ETHER_HDR_LEN);
5150 if (m == NULL) {
5151 *data = NULL;
5152 return EJUSTRETURN;
5153 }
5154 }
5155
5156 eh = mtod(m, struct ether_header *);
5157 vlan = VLANTAGOF(m);
5158
5159 BRIDGE_LOCK(sc);
5160 mac_nat_bif = sc->sc_mac_nat_bif;
5161 mac_nat_ifp = (mac_nat_bif != NULL) ? mac_nat_bif->bif_ifp : NULL;
5162 if (mac_nat_ifp == ifp) {
5163 /* record the IP address used by the MAC NAT interface */
5164 (void)bridge_mac_nat_output(sc, mac_nat_bif, data, NULL);
5165 m = *data;
5166 if (m == NULL) {
5167 /* packet was deallocated */
5168 BRIDGE_UNLOCK(sc);
5169 return EJUSTRETURN;
5170 }
5171 }
5172 bridge_ifp = sc->sc_ifp;
5173
5174 /*
5175 * APPLE MODIFICATION
5176 * If the packet is an 802.1X ethertype, then only send on the
5177 * original output interface.
5178 */
5179 if (eh->ether_type == htons(ETHERTYPE_PAE)) {
5180 dst_if = ifp;
5181 goto sendunicast;
5182 }
5183
5184 /*
5185 * If bridge is down, but the original output interface is up,
5186 * go ahead and send out that interface. Otherwise, the packet
5187 * is dropped below.
5188 */
5189 if ((bridge_ifp->if_flags & IFF_RUNNING) == 0) {
5190 dst_if = ifp;
5191 goto sendunicast;
5192 }
5193
5194 /*
5195 * If the packet is a multicast, or we don't know a better way to
5196 * get there, send to all interfaces.
5197 */
5198 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
5199 dst_if = NULL;
5200 } else {
5201 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
5202 }
5203 if (dst_if == NULL) {
5204 struct bridge_iflist *bif;
5205 struct mbuf *mc;
5206 errno_t error;
5207
5208
5209 bridge_span(sc, m);
5210
5211 BRIDGE_LOCK2REF(sc, error);
5212 if (error != 0) {
5213 m_freem(m);
5214 return EJUSTRETURN;
5215 }
5216
5217 /*
5218 * Duplicate and send the packet across all member interfaces
5219 * except the originating interface.
5220 */
5221 TAILQ_FOREACH(bif, &sc->sc_iflist, bif_next) {
5222 dst_if = bif->bif_ifp;
5223 if (dst_if == ifp) {
5224 /* skip the originating interface */
5225 continue;
5226 }
5227 /* skip interface with inactive link status */
5228 if ((bif->bif_flags & BIFF_MEDIA_ACTIVE) == 0) {
5229 continue;
5230 }
5231 #if 0
5232 if (dst_if->if_type == IFT_GIF) {
5233 continue;
5234 }
5235 #endif
5236 /* skip interface that isn't running */
5237 if ((dst_if->if_flags & IFF_RUNNING) == 0) {
5238 continue;
5239 }
5240 /*
5241 * If the interface is participating in spanning
5242 * tree, make sure the port is in a state that
5243 * allows forwarding.
5244 */
5245 if ((bif->bif_ifflags & IFBIF_STP) &&
5246 bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
5247 continue;
5248 }
5249 /*
5250 * If the destination is the MAC NAT interface,
5251 * skip sending the packet. The packet can't be sent
5252 * if the source MAC is incorrect.
5253 */
5254 if (dst_if == mac_nat_ifp) {
5255 continue;
5256 }
5257
5258 /* make a deep copy to send on this member interface */
5259 mc = m_dup(m, M_DONTWAIT);
5260 if (mc == NULL) {
5261 (void)ifnet_stat_increment_out(bridge_ifp,
5262 0, 0, 1);
5263 continue;
5264 }
5265 (void)bridge_enqueue(bridge_ifp, ifp, dst_if,
5266 mc, CHECKSUM_OPERATION_COMPUTE);
5267 }
5268 BRIDGE_UNREF(sc);
5269
5270 if ((ifp->if_flags & IFF_RUNNING) == 0) {
5271 m_freem(m);
5272 return EJUSTRETURN;
5273 }
5274 /* allow packet to continue on the originating interface */
5275 return 0;
5276 }
5277
5278 sendunicast:
5279 /*
5280 * XXX Spanning tree consideration here?
5281 */
5282
5283 bridge_span(sc, m);
5284 if ((dst_if->if_flags & IFF_RUNNING) == 0) {
5285 m_freem(m);
5286 BRIDGE_UNLOCK(sc);
5287 return EJUSTRETURN;
5288 }
5289
5290 BRIDGE_UNLOCK(sc);
5291 if (dst_if == ifp) {
5292 /* allow packet to continue on the originating interface */
5293 return 0;
5294 }
5295 if (dst_if != mac_nat_ifp) {
5296 (void) bridge_enqueue(bridge_ifp, ifp, dst_if, m,
5297 CHECKSUM_OPERATION_COMPUTE);
5298 } else {
5299 /*
5300 * This is not the original output interface
5301 * and the destination is the MAC NAT interface.
5302 * Drop the packet because the packet can't be sent
5303 * if the source MAC is incorrect.
5304 */
5305 m_freem(m);
5306 }
5307 return EJUSTRETURN;
5308 }
5309
5310 /*
5311 * Output callback.
5312 *
5313 * This routine is called externally from above only when if_bridge_txstart
5314 * is disabled; otherwise it is called internally by bridge_start().
5315 */
5316 static int
5317 bridge_output(struct ifnet *ifp, struct mbuf *m)
5318 {
5319 struct bridge_softc *sc = ifnet_softc(ifp);
5320 struct ether_header *eh;
5321 struct ifnet *dst_if = NULL;
5322 int error = 0;
5323
5324 eh = mtod(m, struct ether_header *);
5325
5326 BRIDGE_LOCK(sc);
5327
5328 if (!(m->m_flags & (M_BCAST | M_MCAST))) {
5329 dst_if = bridge_rtlookup(sc, eh->ether_dhost, 0);
5330 }
5331
5332 (void) ifnet_stat_increment_out(ifp, 1, m->m_pkthdr.len, 0);
5333
5334 #if NBPFILTER > 0
5335 if (sc->sc_bpf_output) {
5336 bridge_bpf_output(ifp, m);
5337 }
5338 #endif
5339
5340 if (dst_if == NULL) {
5341 /* callee will unlock */
5342 bridge_broadcast(sc, NULL, m, 0);
5343 } else {
5344 ifnet_t bridge_ifp;
5345
5346 bridge_ifp = sc->sc_ifp;
5347 BRIDGE_UNLOCK(sc);
5348
5349 error = bridge_enqueue(bridge_ifp, NULL, dst_if, m,
5350 CHECKSUM_OPERATION_FINALIZE);
5351 }
5352
5353 return error;
5354 }
5355
5356 static void
5357 bridge_finalize_cksum(struct ifnet *ifp, struct mbuf *m)
5358 {
5359 struct ether_header *eh;
5360 bool is_ipv4;
5361 uint32_t sw_csum, hwcap;
5362 uint32_t did_sw;
5363 uint32_t csum_flags;
5364
5365 eh = mtod(m, struct ether_header *);
5366 if (!ether_header_type_is_ip(eh, &is_ipv4)) {
5367 return;
5368 }
5369
5370 /* do in software what the hardware cannot */
5371 hwcap = (ifp->if_hwassist | CSUM_DATA_VALID);
5372 csum_flags = m->m_pkthdr.csum_flags;
5373 sw_csum = csum_flags & ~IF_HWASSIST_CSUM_FLAGS(hwcap);
5374 sw_csum &= IF_HWASSIST_CSUM_MASK;
5375
5376 if (is_ipv4) {
5377 if ((hwcap & CSUM_PARTIAL) && !(sw_csum & CSUM_DELAY_DATA) &&
5378 (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)) {
5379 if (m->m_pkthdr.csum_flags & CSUM_TCP) {
5380 uint16_t start =
5381 sizeof(*eh) + sizeof(struct ip);
5382 uint16_t ulpoff =
5383 m->m_pkthdr.csum_data & 0xffff;
5384 m->m_pkthdr.csum_flags |=
5385 (CSUM_DATA_VALID | CSUM_PARTIAL);
5386 m->m_pkthdr.csum_tx_stuff = (ulpoff + start);
5387 m->m_pkthdr.csum_tx_start = start;
5388 } else {
5389 sw_csum |= (CSUM_DELAY_DATA &
5390 m->m_pkthdr.csum_flags);
5391 }
5392 }
5393 did_sw = in_finalize_cksum(m, sizeof(*eh), sw_csum);
5394 } else {
5395 if ((hwcap & CSUM_PARTIAL) &&
5396 !(sw_csum & CSUM_DELAY_IPV6_DATA) &&
5397 (m->m_pkthdr.csum_flags & CSUM_DELAY_IPV6_DATA)) {
5398 if (m->m_pkthdr.csum_flags & CSUM_TCPIPV6) {
5399 uint16_t start =
5400 sizeof(*eh) + sizeof(struct ip6_hdr);
5401 uint16_t ulpoff =
5402 m->m_pkthdr.csum_data & 0xffff;
5403 m->m_pkthdr.csum_flags |=
5404 (CSUM_DATA_VALID | CSUM_PARTIAL);
5405 m->m_pkthdr.csum_tx_stuff = (ulpoff + start);
5406 m->m_pkthdr.csum_tx_start = start;
5407 } else {
5408 sw_csum |= (CSUM_DELAY_IPV6_DATA &
5409 m->m_pkthdr.csum_flags);
5410 }
5411 }
5412 did_sw = in6_finalize_cksum(m, sizeof(*eh), -1, -1, sw_csum);
5413 }
5414 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
5415 "[%s] before 0x%x hwcap 0x%x sw_csum 0x%x did 0x%x after 0x%x",
5416 ifp->if_xname, csum_flags, hwcap, sw_csum,
5417 did_sw, m->m_pkthdr.csum_flags);
5418 }
5419
5420 /*
5421 * bridge_start:
5422 *
5423 * Start output on a bridge.
5424 *
5425 * This routine is invoked by the start worker thread; because we never call
5426 * it directly, there is no need do deploy any serialization mechanism other
5427 * than what's already used by the worker thread, i.e. this is already single
5428 * threaded.
5429 *
5430 * This routine is called only when if_bridge_txstart is enabled.
5431 */
5432 static void
5433 bridge_start(struct ifnet *ifp)
5434 {
5435 struct mbuf *m;
5436
5437 for (;;) {
5438 if (ifnet_dequeue(ifp, &m) != 0) {
5439 break;
5440 }
5441
5442 (void) bridge_output(ifp, m);
5443 }
5444 }
5445
5446 /*
5447 * bridge_forward:
5448 *
5449 * The forwarding function of the bridge.
5450 *
5451 * NOTE: Releases the lock on return.
5452 */
5453 static void
5454 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
5455 struct mbuf *m)
5456 {
5457 struct bridge_iflist *dbif;
5458 ifnet_t bridge_ifp;
5459 struct ifnet *src_if, *dst_if;
5460 struct ether_header *eh;
5461 uint16_t vlan;
5462 uint8_t *dst;
5463 int error;
5464 struct mac_nat_record mnr;
5465 bool translate_mac = FALSE;
5466 uint32_t sc_filter_flags = 0;
5467
5468 BRIDGE_LOCK_ASSERT_HELD(sc);
5469
5470 bridge_ifp = sc->sc_ifp;
5471 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_OUTPUT,
5472 "%s m 0x%llx", bridge_ifp->if_xname,
5473 (uint64_t)VM_KERNEL_ADDRPERM(m));
5474
5475 src_if = m->m_pkthdr.rcvif;
5476 if (src_if != sbif->bif_ifp) {
5477 const char * src_if_name;
5478
5479 src_if_name = (src_if != NULL) ? src_if->if_xname : "?";
5480 BRIDGE_LOG(LOG_NOTICE, 0,
5481 "src_if %s != bif_ifp %s",
5482 src_if_name, sbif->bif_ifp->if_xname);
5483 goto drop;
5484 }
5485
5486 (void) ifnet_stat_increment_in(bridge_ifp, 1, m->m_pkthdr.len, 0);
5487 vlan = VLANTAGOF(m);
5488
5489
5490 if ((sbif->bif_ifflags & IFBIF_STP) &&
5491 sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
5492 goto drop;
5493 }
5494
5495 eh = mtod(m, struct ether_header *);
5496 dst = eh->ether_dhost;
5497
5498 /* If the interface is learning, record the address. */
5499 if (sbif->bif_ifflags & IFBIF_LEARNING) {
5500 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
5501 sbif, 0, IFBAF_DYNAMIC);
5502 /*
5503 * If the interface has addresses limits then deny any source
5504 * that is not in the cache.
5505 */
5506 if (error && sbif->bif_addrmax) {
5507 goto drop;
5508 }
5509 }
5510
5511 if ((sbif->bif_ifflags & IFBIF_STP) != 0 &&
5512 sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING) {
5513 goto drop;
5514 }
5515
5516 /*
5517 * At this point, the port either doesn't participate
5518 * in spanning tree or it is in the forwarding state.
5519 */
5520
5521 /*
5522 * If the packet is unicast, destined for someone on
5523 * "this" side of the bridge, drop it.
5524 */
5525 if ((m->m_flags & (M_BCAST | M_MCAST)) == 0) {
5526 /* unicast */
5527 dst_if = bridge_rtlookup(sc, dst, vlan);
5528 if (src_if == dst_if) {
5529 goto drop;
5530 }
5531 } else {
5532 /* broadcast/multicast */
5533
5534 /*
5535 * Check if its a reserved multicast address, any address
5536 * listed in 802.1D section 7.12.6 may not be forwarded by the
5537 * bridge.
5538 * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
5539 */
5540 if (dst[0] == 0x01 && dst[1] == 0x80 &&
5541 dst[2] == 0xc2 && dst[3] == 0x00 &&
5542 dst[4] == 0x00 && dst[5] <= 0x0f) {
5543 goto drop;
5544 }
5545
5546
5547 /* ...forward it to all interfaces. */
5548 atomic_add_64(&bridge_ifp->if_imcasts, 1);
5549 dst_if = NULL;
5550 }
5551
5552 /*
5553 * If we have a destination interface which is a member of our bridge,
5554 * OR this is a unicast packet, push it through the bpf(4) machinery.
5555 * For broadcast or multicast packets, don't bother because it will
5556 * be reinjected into ether_input. We do this before we pass the packets
5557 * through the pfil(9) framework, as it is possible that pfil(9) will
5558 * drop the packet, or possibly modify it, making it difficult to debug
5559 * firewall issues on the bridge.
5560 */
5561 #if NBPFILTER > 0
5562 if (eh->ether_type == htons(ETHERTYPE_RSN_PREAUTH) ||
5563 dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0) {
5564 m->m_pkthdr.rcvif = bridge_ifp;
5565 BRIDGE_BPF_MTAP_INPUT(sc, m);
5566 }
5567 #endif /* NBPFILTER */
5568
5569 if (dst_if == NULL) {
5570 /* bridge_broadcast will unlock */
5571 bridge_broadcast(sc, sbif, m, 1);
5572 return;
5573 }
5574
5575 /*
5576 * Unicast.
5577 */
5578 /*
5579 * At this point, we're dealing with a unicast frame
5580 * going to a different interface.
5581 */
5582 if ((dst_if->if_flags & IFF_RUNNING) == 0) {
5583 goto drop;
5584 }
5585
5586 dbif = bridge_lookup_member_if(sc, dst_if);
5587 if (dbif == NULL) {
5588 /* Not a member of the bridge (anymore?) */
5589 goto drop;
5590 }
5591
5592 /* Private segments can not talk to each other */
5593 if (sbif->bif_ifflags & dbif->bif_ifflags & IFBIF_PRIVATE) {
5594 goto drop;
5595 }
5596
5597 if ((dbif->bif_ifflags & IFBIF_STP) &&
5598 dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
5599 goto drop;
5600 }
5601
5602 #if HAS_DHCPRA_MASK
5603 /* APPLE MODIFICATION <rdar:6985737> */
5604 if ((dst_if->if_extflags & IFEXTF_DHCPRA_MASK) != 0) {
5605 m = ip_xdhcpra_output(dst_if, m);
5606 if (!m) {
5607 ++bridge_ifp.if_xdhcpra;
5608 BRIDGE_UNLOCK(sc);
5609 return;
5610 }
5611 }
5612 #endif /* HAS_DHCPRA_MASK */
5613
5614 if (dbif == sc->sc_mac_nat_bif) {
5615 /* determine how to translate the packet */
5616 translate_mac
5617 = bridge_mac_nat_output(sc, sbif, &m, &mnr);
5618 if (m == NULL) {
5619 /* packet was deallocated */
5620 BRIDGE_UNLOCK(sc);
5621 return;
5622 }
5623 } else if (bif_has_checksum_offload(dbif) &&
5624 !bif_has_checksum_offload(sbif)) {
5625 /*
5626 * If the destination interface has checksum offload enabled,
5627 * verify the checksum now, unless the source interface also has
5628 * checksum offload enabled. The checksum in that case has
5629 * already just been computed and verifying it is unnecessary.
5630 */
5631 error = bridge_verify_checksum(&m, &dbif->bif_stats);
5632 if (error != 0) {
5633 BRIDGE_UNLOCK(sc);
5634 if (m != NULL) {
5635 m_freem(m);
5636 }
5637 return;
5638 }
5639 }
5640
5641 sc_filter_flags = sc->sc_filter_flags;
5642
5643 BRIDGE_UNLOCK(sc);
5644 if (PF_IS_ENABLED && (sc_filter_flags & IFBF_FILT_MEMBER)) {
5645 if (bridge_pf(&m, dst_if, sc_filter_flags, FALSE) != 0) {
5646 return;
5647 }
5648 if (m == NULL) {
5649 return;
5650 }
5651 }
5652
5653 /* if we need to, translate the MAC address */
5654 if (translate_mac) {
5655 bridge_mac_nat_translate(&m, &mnr, IF_LLADDR(dst_if));
5656 }
5657 /*
5658 * We're forwarding an inbound packet in which the checksum must
5659 * already have been computed and if required, verified.
5660 */
5661 if (m != NULL) {
5662 (void) bridge_enqueue(bridge_ifp, src_if, dst_if, m,
5663 CHECKSUM_OPERATION_CLEAR_OFFLOAD);
5664 }
5665 return;
5666
5667 drop:
5668 BRIDGE_UNLOCK(sc);
5669 m_freem(m);
5670 }
5671
5672 static void
5673 inject_input_packet(ifnet_t ifp, mbuf_t m)
5674 {
5675 mbuf_pkthdr_setrcvif(m, ifp);
5676 mbuf_pkthdr_setheader(m, mbuf_data(m));
5677 mbuf_setdata(m, (char *)mbuf_data(m) + ETHER_HDR_LEN,
5678 mbuf_len(m) - ETHER_HDR_LEN);
5679 mbuf_pkthdr_adjustlen(m, -ETHER_HDR_LEN);
5680 m->m_flags |= M_PROTO1; /* set to avoid loops */
5681 dlil_input_packet_list(ifp, m);
5682 return;
5683 }
5684
5685 static bool
5686 in_addr_is_ours(struct in_addr ip)
5687 {
5688 struct in_ifaddr *ia;
5689 bool ours = false;
5690
5691 lck_rw_lock_shared(&in_ifaddr_rwlock);
5692 TAILQ_FOREACH(ia, INADDR_HASH(ip.s_addr), ia_hash) {
5693 if (IA_SIN(ia)->sin_addr.s_addr == ip.s_addr) {
5694 ours = true;
5695 break;
5696 }
5697 }
5698 lck_rw_done(&in_ifaddr_rwlock);
5699 return ours;
5700 }
5701
5702 static bool
5703 in6_addr_is_ours(const struct in6_addr * ip6_p, uint32_t ifscope)
5704 {
5705 struct in6_ifaddr *ia6;
5706 bool ours = false;
5707
5708 if (in6_embedded_scope && IN6_IS_ADDR_LINKLOCAL(ip6_p)) {
5709 struct in6_addr dst_ip;
5710
5711 /* need to embed scope ID for comparison */
5712 bcopy(ip6_p, &dst_ip, sizeof(dst_ip));
5713 dst_ip.s6_addr16[1] = htons(ifscope);
5714 ip6_p = &dst_ip;
5715 }
5716 lck_rw_lock_shared(&in6_ifaddr_rwlock);
5717 TAILQ_FOREACH(ia6, IN6ADDR_HASH(ip6_p), ia6_hash) {
5718 if (in6_are_addr_equal_scoped(&ia6->ia_addr.sin6_addr, ip6_p,
5719 ia6->ia_addr.sin6_scope_id, ifscope)) {
5720 ours = true;
5721 break;
5722 }
5723 }
5724 lck_rw_done(&in6_ifaddr_rwlock);
5725 return ours;
5726 }
5727
5728 static void
5729 bridge_interface_input(ifnet_t bridge_ifp, mbuf_t m,
5730 bpf_packet_func bpf_input_func)
5731 {
5732 size_t byte_count;
5733 struct ether_header *eh;
5734 errno_t error;
5735 bool is_ipv4;
5736 int len;
5737 u_int mac_hlen;
5738 int pkt_count;
5739
5740 /* segment large packets before sending them up */
5741 if (if_bridge_segmentation == 0) {
5742 goto done;
5743 }
5744 len = m->m_pkthdr.len;
5745 if (len <= (bridge_ifp->if_mtu + ETHER_HDR_LEN)) {
5746 goto done;
5747 }
5748 eh = mtod(m, struct ether_header *);
5749 if (!ether_header_type_is_ip(eh, &is_ipv4)) {
5750 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
5751 "large non IPv4/IPv6 packet");
5752 goto done;
5753 }
5754
5755 /*
5756 * We have a large IPv4/IPv6 TCP packet. Segment it if required.
5757 *
5758 * If gso_tcp() returns success (0), the packet(s) are
5759 * ready to be passed up. If the destination is a local IP address,
5760 * the packet will be passed up as a large, single packet.
5761 *
5762 * If gso_tcp() returns an error, the packet has already
5763 * been freed.
5764 */
5765 mac_hlen = sizeof(*eh);
5766 error = gso_tcp(bridge_ifp, &m, mac_hlen, is_ipv4, FALSE);
5767 if (error != 0) {
5768 return;
5769 }
5770
5771 done:
5772 pkt_count = 0;
5773 byte_count = 0;
5774 for (mbuf_t scan = m; scan != NULL; scan = scan->m_nextpkt) {
5775 /* Mark the packet as arriving on the bridge interface */
5776 mbuf_pkthdr_setrcvif(scan, bridge_ifp);
5777 mbuf_pkthdr_setheader(scan, mbuf_data(scan));
5778 if (bpf_input_func != NULL) {
5779 (*bpf_input_func)(bridge_ifp, scan);
5780 }
5781 mbuf_setdata(scan, (char *)mbuf_data(scan) + ETHER_HDR_LEN,
5782 mbuf_len(scan) - ETHER_HDR_LEN);
5783 mbuf_pkthdr_adjustlen(scan, -ETHER_HDR_LEN);
5784 byte_count += mbuf_pkthdr_len(scan);
5785 pkt_count++;
5786 }
5787 (void)ifnet_stat_increment_in(bridge_ifp, pkt_count, byte_count, 0);
5788 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT,
5789 "%s %d packet(s) %ld bytes",
5790 bridge_ifp->if_xname, pkt_count, byte_count);
5791 dlil_input_packet_list(bridge_ifp, m);
5792 return;
5793 }
5794
5795 static bool
5796 is_our_ip(ip_packet_info_t info_p, uint32_t ifscope)
5797 {
5798 bool ours;
5799
5800 if (info_p->ip_is_ipv4) {
5801 struct in_addr dst_ip;
5802
5803 bcopy(&info_p->ip_hdr.ip->ip_dst, &dst_ip, sizeof(dst_ip));
5804 ours = in_addr_is_ours(dst_ip);
5805 } else {
5806 ours = in6_addr_is_ours(&info_p->ip_hdr.ip6->ip6_dst, ifscope);
5807 }
5808 return ours;
5809 }
5810
5811 static inline errno_t
5812 bridge_vmnet_tag_input(ifnet_t bridge_ifp, ifnet_t ifp,
5813 const u_char * ether_dhost, mbuf_t *mp,
5814 bool is_broadcast, bool is_ip, bool is_ipv4,
5815 ip_packet_info * info_p, struct bripstats * stats_p,
5816 bool *info_initialized)
5817 {
5818 errno_t error = 0;
5819 bool is_local = false;
5820 struct pf_mtag *pf_mtag;
5821 u_int16_t tag = vmnet_tag;
5822
5823 *info_initialized = false;
5824 if (is_broadcast) {
5825 if (_ether_cmp(ether_dhost, etherbroadcastaddr) == 0) {
5826 tag = vmnet_broadcast_tag;
5827 } else {
5828 tag = vmnet_multicast_tag;
5829 }
5830 } else if (is_ip) {
5831 unsigned int mac_hlen = sizeof(struct ether_header);
5832
5833 bzero(stats_p, sizeof(*stats_p));
5834 *info_initialized = true;
5835 error = bridge_get_ip_proto(mp, mac_hlen, is_ipv4, info_p,
5836 stats_p);
5837 if (error != 0) {
5838 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_INPUT,
5839 "%s(%s) bridge_get_ip_proto failed %d",
5840 bridge_ifp->if_xname,
5841 ifp->if_xname, error);
5842 if (*mp == NULL) {
5843 return EJUSTRETURN;
5844 }
5845 } else {
5846 is_local = is_our_ip(info_p, bridge_ifp->if_index);
5847 if (is_local) {
5848 tag = vmnet_local_tag;
5849 }
5850 }
5851 }
5852 pf_mtag = pf_get_mtag(*mp);
5853 if (pf_mtag != NULL) {
5854 pf_mtag->pftag_tag = tag;
5855 }
5856 #if DEBUG || DEVELOPMENT
5857 {
5858 bool forced;
5859
5860 BRIDGE_ERROR_GET_FORCED(forced, BRIDGE_FORCE_ONE);
5861 if (forced) {
5862 m_freem(*mp);
5863 *mp = NULL;
5864 error = EJUSTRETURN;
5865 goto done;
5866 }
5867 BRIDGE_ERROR_GET_FORCED(forced, BRIDGE_FORCE_TWO);
5868 if (forced) {
5869 error = _EBADIP;
5870 goto done;
5871 }
5872 }
5873 done:
5874 #endif /* DEBUG || DEVELOPMENT */
5875 return error;
5876 }
5877
5878 static void
5879 bripstats_apply(struct bripstats *dst_p, const struct bripstats *src_p)
5880 {
5881 dst_p->bips_ip += src_p->bips_ip;
5882 dst_p->bips_ip6 += src_p->bips_ip6;
5883 dst_p->bips_udp += src_p->bips_udp;
5884 dst_p->bips_tcp += src_p->bips_tcp;
5885
5886 dst_p->bips_bad_ip += src_p->bips_bad_ip;
5887 dst_p->bips_bad_ip6 += src_p->bips_bad_ip6;
5888 dst_p->bips_bad_udp += src_p->bips_bad_udp;
5889 dst_p->bips_bad_tcp += src_p->bips_bad_tcp;
5890 }
5891
5892 static void
5893 bridge_bripstats_apply(ifnet_t ifp, const struct bripstats *stats_p)
5894 {
5895 struct bridge_iflist *bif;
5896 struct bridge_softc *sc = ifp->if_bridge;
5897
5898 BRIDGE_LOCK(sc);
5899 bif = bridge_lookup_member_if(sc, ifp);
5900 if (bif == NULL) {
5901 goto done;
5902 }
5903 if (!bif_has_checksum_offload(bif)) {
5904 goto done;
5905 }
5906 bripstats_apply(&bif->bif_stats.brms_in_ip, stats_p);
5907
5908 done:
5909 BRIDGE_UNLOCK(sc);
5910 return;
5911 }
5912
5913 /*
5914 * bridge_input:
5915 *
5916 * Filter input from a member interface. Queue the packet for
5917 * bridging if it is not for us.
5918 */
5919 errno_t
5920 bridge_input(struct ifnet *ifp, mbuf_t *data)
5921 {
5922 struct bridge_softc *sc = ifp->if_bridge;
5923 struct bridge_iflist *bif, *bif2;
5924 struct ether_header eh_in;
5925 bool is_ip = false;
5926 bool is_ipv4 = false;
5927 ifnet_t bridge_ifp;
5928 struct mbuf *mc, *mc2;
5929 unsigned int mac_hlen = sizeof(struct ether_header);
5930 uint16_t vlan;
5931 errno_t error;
5932 ip_packet_info info;
5933 struct bripstats stats;
5934 bool info_initialized = false;
5935 errno_t ip_packet_error = 0;
5936 bool is_broadcast;
5937 bool is_ip_broadcast = false;
5938 bool is_ifp_mac = false;
5939 mbuf_t m = *data;
5940 uint32_t sc_filter_flags = 0;
5941
5942 bridge_ifp = sc->sc_ifp;
5943 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT,
5944 "%s from %s m 0x%llx data 0x%llx",
5945 bridge_ifp->if_xname, ifp->if_xname,
5946 (uint64_t)VM_KERNEL_ADDRPERM(m),
5947 (uint64_t)VM_KERNEL_ADDRPERM(mbuf_data(m)));
5948 if ((sc->sc_ifp->if_flags & IFF_RUNNING) == 0) {
5949 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT,
5950 "%s not running passing along",
5951 bridge_ifp->if_xname);
5952 return 0;
5953 }
5954
5955 vlan = VLANTAGOF(m);
5956
5957 #ifdef IFF_MONITOR
5958 /*
5959 * Implement support for bridge monitoring. If this flag has been
5960 * set on this interface, discard the packet once we push it through
5961 * the bpf(4) machinery, but before we do, increment the byte and
5962 * packet counters associated with this interface.
5963 */
5964 if ((bridge_ifp->if_flags & IFF_MONITOR) != 0) {
5965 m->m_pkthdr.rcvif = bridge_ifp;
5966 BRIDGE_BPF_MTAP_INPUT(sc, m);
5967 (void) ifnet_stat_increment_in(bridge_ifp, 1, m->m_pkthdr.len, 0);
5968 *data = NULL;
5969 m_freem(m);
5970 return EJUSTRETURN;
5971 }
5972 #endif /* IFF_MONITOR */
5973
5974 is_broadcast = (m->m_flags & (M_BCAST | M_MCAST)) != 0;
5975
5976 /*
5977 * Need to clear the promiscuous flag otherwise it will be
5978 * dropped by DLIL after processing filters
5979 */
5980 if ((mbuf_flags(m) & MBUF_PROMISC)) {
5981 mbuf_setflags_mask(m, 0, MBUF_PROMISC);
5982 }
5983
5984 /* copy the ethernet header */
5985 eh_in = *(mtod(m, struct ether_header *));
5986
5987 is_ip = ether_header_type_is_ip(&eh_in, &is_ipv4);
5988
5989 if (if_bridge_vmnet_pf_tagging != 0 && IFNET_IS_VMNET(ifp)) {
5990 /* tag packets coming from VMNET interfaces */
5991 ip_packet_error = bridge_vmnet_tag_input(bridge_ifp, ifp,
5992 eh_in.ether_dhost, data, is_broadcast, is_ip, is_ipv4,
5993 &info, &stats, &info_initialized);
5994 m = *data;
5995 if (m == NULL) {
5996 bridge_bripstats_apply(ifp, &stats);
5997 return EJUSTRETURN;
5998 }
5999 }
6000
6001 sc_filter_flags = sc->sc_filter_flags;
6002 if (PF_IS_ENABLED && (sc_filter_flags & IFBF_FILT_MEMBER)) {
6003 error = bridge_pf(data, ifp, sc_filter_flags, TRUE);
6004 m = *data;
6005 if (error != 0 || m == NULL) {
6006 return EJUSTRETURN;
6007 }
6008 }
6009
6010 BRIDGE_LOCK(sc);
6011 bif = bridge_lookup_member_if(sc, ifp);
6012 if (bif == NULL) {
6013 BRIDGE_UNLOCK(sc);
6014 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT,
6015 "%s bridge_lookup_member_if failed",
6016 bridge_ifp->if_xname);
6017 return 0;
6018 }
6019 if (is_ip && bif_has_checksum_offload(bif)) {
6020 if (info_initialized) {
6021 bripstats_apply(&bif->bif_stats.brms_in_ip, &stats);
6022 } else {
6023 error = bridge_get_ip_proto(data, mac_hlen, is_ipv4,
6024 &info, &bif->bif_stats.brms_in_ip);
6025 if (error != 0) {
6026 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_CHECKSUM,
6027 "%s(%s) bridge_get_ip_proto failed %d",
6028 bridge_ifp->if_xname,
6029 bif->bif_ifp->if_xname, error);
6030 ip_packet_error = error;
6031 }
6032 }
6033 if (ip_packet_error == 0) {
6034 /* need to compute IP/UDP/TCP/checksums */
6035 error = bridge_offload_checksum(data, &info,
6036 &bif->bif_stats);
6037 if (error != 0) {
6038 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_CHECKSUM,
6039 "%s(%s) bridge_offload_checksum failed %d",
6040 bridge_ifp->if_xname,
6041 bif->bif_ifp->if_xname, error);
6042 ip_packet_error = error;
6043 }
6044 }
6045 if (ip_packet_error != 0) {
6046 BRIDGE_UNLOCK(sc);
6047 if (*data != NULL) {
6048 m_freem(*data);
6049 *data = NULL;
6050 }
6051 return EJUSTRETURN;
6052 }
6053 m = *data;
6054 }
6055
6056 if (bif->bif_flags & BIFF_HOST_FILTER) {
6057 error = bridge_host_filter(bif, data);
6058 if (error != 0) {
6059 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_INPUT,
6060 "%s bridge_host_filter failed",
6061 bif->bif_ifp->if_xname);
6062 BRIDGE_UNLOCK(sc);
6063 return EJUSTRETURN;
6064 }
6065 m = *data;
6066 }
6067
6068 if (!is_broadcast &&
6069 _ether_cmp(eh_in.ether_dhost, IF_LLADDR(ifp)) == 0) {
6070 /* the packet is unicast to the interface's MAC address */
6071 if (is_ip && sc->sc_mac_nat_bif == bif) {
6072 /* doing MAC-NAT, check if destination is IP broadcast */
6073 is_ip_broadcast = is_broadcast_ip_packet(data);
6074 if (*data == NULL) {
6075 BRIDGE_UNLOCK(sc);
6076 return EJUSTRETURN;
6077 }
6078 m = *data;
6079 }
6080 if (!is_ip_broadcast) {
6081 is_ifp_mac = TRUE;
6082 }
6083 }
6084
6085 bridge_span(sc, m);
6086
6087 if (is_broadcast || is_ip_broadcast) {
6088 if (is_broadcast && (m->m_flags & M_MCAST) != 0) {
6089 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MCAST,
6090 " multicast: "
6091 "%02x:%02x:%02x:%02x:%02x:%02x",
6092 eh_in.ether_dhost[0], eh_in.ether_dhost[1],
6093 eh_in.ether_dhost[2], eh_in.ether_dhost[3],
6094 eh_in.ether_dhost[4], eh_in.ether_dhost[5]);
6095 }
6096 /* Tap off 802.1D packets; they do not get forwarded. */
6097 if (is_broadcast &&
6098 _ether_cmp(eh_in.ether_dhost, bstp_etheraddr) == 0) {
6099 #if BRIDGESTP
6100 m = bstp_input(&bif->bif_stp, ifp, m);
6101 #else /* !BRIDGESTP */
6102 m_freem(m);
6103 m = NULL;
6104 #endif /* !BRIDGESTP */
6105 if (m == NULL) {
6106 BRIDGE_UNLOCK(sc);
6107 return EJUSTRETURN;
6108 }
6109 }
6110
6111 if ((bif->bif_ifflags & IFBIF_STP) &&
6112 bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
6113 BRIDGE_UNLOCK(sc);
6114 return 0;
6115 }
6116
6117 /*
6118 * Make a deep copy of the packet and enqueue the copy
6119 * for bridge processing.
6120 */
6121 mc = m_dup(m, M_DONTWAIT);
6122 if (mc == NULL) {
6123 BRIDGE_UNLOCK(sc);
6124 return 0;
6125 }
6126
6127 /*
6128 * Perform the bridge forwarding function with the copy.
6129 *
6130 * Note that bridge_forward calls BRIDGE_UNLOCK
6131 */
6132 if (is_ip_broadcast) {
6133 struct ether_header *eh;
6134
6135 /* make the copy look like it is actually broadcast */
6136 mc->m_flags |= M_BCAST;
6137 eh = mtod(mc, struct ether_header *);
6138 bcopy(etherbroadcastaddr, eh->ether_dhost,
6139 ETHER_ADDR_LEN);
6140 }
6141 bridge_forward(sc, bif, mc);
6142
6143 /*
6144 * Reinject the mbuf as arriving on the bridge so we have a
6145 * chance at claiming multicast packets. We can not loop back
6146 * here from ether_input as a bridge is never a member of a
6147 * bridge.
6148 */
6149 VERIFY(bridge_ifp->if_bridge == NULL);
6150 mc2 = m_dup(m, M_DONTWAIT);
6151 if (mc2 != NULL) {
6152 /* Keep the layer3 header aligned */
6153 int i = min(mc2->m_pkthdr.len, max_protohdr);
6154 mc2 = m_copyup(mc2, i, ETHER_ALIGN);
6155 }
6156 if (mc2 != NULL) {
6157 /* mark packet as arriving on the bridge */
6158 mc2->m_pkthdr.rcvif = bridge_ifp;
6159 mc2->m_pkthdr.pkt_hdr = mbuf_data(mc2);
6160 BRIDGE_BPF_MTAP_INPUT(sc, mc2);
6161 (void) mbuf_setdata(mc2,
6162 (char *)mbuf_data(mc2) + ETHER_HDR_LEN,
6163 mbuf_len(mc2) - ETHER_HDR_LEN);
6164 (void) mbuf_pkthdr_adjustlen(mc2, -ETHER_HDR_LEN);
6165 (void) ifnet_stat_increment_in(bridge_ifp, 1,
6166 mbuf_pkthdr_len(mc2), 0);
6167 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MCAST,
6168 "%s mcast for us", bridge_ifp->if_xname);
6169 dlil_input_packet_list(bridge_ifp, mc2);
6170 }
6171
6172 /* Return the original packet for local processing. */
6173 return 0;
6174 }
6175
6176 if ((bif->bif_ifflags & IFBIF_STP) &&
6177 bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
6178 BRIDGE_UNLOCK(sc);
6179 return 0;
6180 }
6181
6182 #ifdef DEV_CARP
6183 #define CARP_CHECK_WE_ARE_DST(iface) \
6184 ((iface)->if_carp &&\
6185 carp_forus((iface)->if_carp, eh_in.ether_dhost))
6186 #define CARP_CHECK_WE_ARE_SRC(iface) \
6187 ((iface)->if_carp &&\
6188 carp_forus((iface)->if_carp, eh_in.ether_shost))
6189 #else
6190 #define CARP_CHECK_WE_ARE_DST(iface) 0
6191 #define CARP_CHECK_WE_ARE_SRC(iface) 0
6192 #endif
6193
6194 #define PFIL_HOOKED_INET6 PFIL_HOOKED(&inet6_pfil_hook)
6195
6196 #define PFIL_PHYS(sc, ifp, m)
6197
6198 #define GRAB_OUR_PACKETS(iface) \
6199 if ((iface)->if_type == IFT_GIF) \
6200 continue; \
6201 /* It is destined for us. */ \
6202 if (_ether_cmp(IF_LLADDR((iface)), eh_in.ether_dhost) == 0 || \
6203 CARP_CHECK_WE_ARE_DST((iface))) { \
6204 if ((iface)->if_type == IFT_BRIDGE) { \
6205 BRIDGE_BPF_MTAP_INPUT(sc, m); \
6206 /* Filter on the physical interface. */ \
6207 PFIL_PHYS(sc, iface, m); \
6208 } else { \
6209 bpf_tap_in(iface, DLT_EN10MB, m, NULL, 0); \
6210 } \
6211 if (bif->bif_ifflags & IFBIF_LEARNING) { \
6212 error = bridge_rtupdate(sc, eh_in.ether_shost, \
6213 vlan, bif, 0, IFBAF_DYNAMIC); \
6214 if (error && bif->bif_addrmax) { \
6215 BRIDGE_UNLOCK(sc); \
6216 m_freem(m); \
6217 return (EJUSTRETURN); \
6218 } \
6219 } \
6220 BRIDGE_UNLOCK(sc); \
6221 inject_input_packet(iface, m); \
6222 return (EJUSTRETURN); \
6223 } \
6224 \
6225 /* We just received a packet that we sent out. */ \
6226 if (_ether_cmp(IF_LLADDR((iface)), eh_in.ether_shost) == 0 || \
6227 CARP_CHECK_WE_ARE_SRC((iface))) { \
6228 BRIDGE_UNLOCK(sc); \
6229 m_freem(m); \
6230 return (EJUSTRETURN); \
6231 }
6232
6233 /*
6234 * Unicast.
6235 */
6236
6237 /* handle MAC-NAT if enabled */
6238 if (is_ifp_mac && sc->sc_mac_nat_bif == bif) {
6239 ifnet_t dst_if;
6240 boolean_t is_input = FALSE;
6241
6242 dst_if = bridge_mac_nat_input(sc, data, &is_input);
6243 m = *data;
6244 if (dst_if == ifp) {
6245 /* our input packet */
6246 } else if (dst_if != NULL || m == NULL) {
6247 BRIDGE_UNLOCK(sc);
6248 if (dst_if != NULL) {
6249 ASSERT(m != NULL);
6250 if (is_input) {
6251 inject_input_packet(dst_if, m);
6252 } else {
6253 (void)bridge_enqueue(bridge_ifp, NULL,
6254 dst_if, m,
6255 CHECKSUM_OPERATION_CLEAR_OFFLOAD);
6256 }
6257 }
6258 return EJUSTRETURN;
6259 }
6260 }
6261
6262 /*
6263 * If the packet is for the bridge, pass it up for local processing.
6264 */
6265 if (_ether_cmp(eh_in.ether_dhost, IF_LLADDR(bridge_ifp)) == 0 ||
6266 CARP_CHECK_WE_ARE_DST(bridge_ifp)) {
6267 bpf_packet_func bpf_input_func = sc->sc_bpf_input;
6268
6269 /*
6270 * If the interface is learning, and the source
6271 * address is valid and not multicast, record
6272 * the address.
6273 */
6274 if (bif->bif_ifflags & IFBIF_LEARNING) {
6275 (void) bridge_rtupdate(sc, eh_in.ether_shost,
6276 vlan, bif, 0, IFBAF_DYNAMIC);
6277 }
6278 BRIDGE_UNLOCK(sc);
6279
6280 bridge_interface_input(bridge_ifp, m, bpf_input_func);
6281 return EJUSTRETURN;
6282 }
6283
6284 /*
6285 * if the destination of the packet is for the MAC address of
6286 * the member interface itself, then we don't need to forward
6287 * it -- just pass it back. Note that it'll likely just be
6288 * dropped by the stack, but if something else is bound to
6289 * the interface directly (for example, the wireless stats
6290 * protocol -- although that actually uses BPF right now),
6291 * then it will consume the packet
6292 *
6293 * ALSO, note that we do this check AFTER checking for the
6294 * bridge's own MAC address, because the bridge may be
6295 * using the SAME MAC address as one of its interfaces
6296 */
6297 if (is_ifp_mac) {
6298
6299 #ifdef VERY_VERY_VERY_DIAGNOSTIC
6300 BRIDGE_LOG(LOG_NOTICE, 0,
6301 "not forwarding packet bound for member interface");
6302 #endif
6303
6304 BRIDGE_UNLOCK(sc);
6305 return 0;
6306 }
6307
6308 /* Now check the remaining bridge members. */
6309 TAILQ_FOREACH(bif2, &sc->sc_iflist, bif_next) {
6310 if (bif2->bif_ifp != ifp) {
6311 GRAB_OUR_PACKETS(bif2->bif_ifp);
6312 }
6313 }
6314
6315 #undef CARP_CHECK_WE_ARE_DST
6316 #undef CARP_CHECK_WE_ARE_SRC
6317 #undef GRAB_OUR_PACKETS
6318
6319 /*
6320 * Perform the bridge forwarding function.
6321 *
6322 * Note that bridge_forward calls BRIDGE_UNLOCK
6323 */
6324 bridge_forward(sc, bif, m);
6325
6326 return EJUSTRETURN;
6327 }
6328
6329 /*
6330 * bridge_broadcast:
6331 *
6332 * Send a frame to all interfaces that are members of
6333 * the bridge, except for the one on which the packet
6334 * arrived.
6335 *
6336 * NOTE: Releases the lock on return.
6337 */
6338 static void
6339 bridge_broadcast(struct bridge_softc *sc, struct bridge_iflist * sbif,
6340 struct mbuf *m, int runfilt)
6341 {
6342 ifnet_t bridge_ifp;
6343 struct bridge_iflist *dbif;
6344 struct ifnet * src_if;
6345 struct mbuf *mc;
6346 struct mbuf *mc_in;
6347 struct ifnet *dst_if;
6348 int error = 0, used = 0;
6349 boolean_t bridge_if_out;
6350 ChecksumOperation cksum_op;
6351 struct mac_nat_record mnr;
6352 struct bridge_iflist *mac_nat_bif = sc->sc_mac_nat_bif;
6353 boolean_t translate_mac = FALSE;
6354 uint32_t sc_filter_flags = 0;
6355
6356 bridge_ifp = sc->sc_ifp;
6357 if (sbif != NULL) {
6358 bridge_if_out = FALSE;
6359 src_if = sbif->bif_ifp;
6360 cksum_op = CHECKSUM_OPERATION_CLEAR_OFFLOAD;
6361 if (mac_nat_bif != NULL && sbif != mac_nat_bif) {
6362 /* get the translation record while holding the lock */
6363 translate_mac
6364 = bridge_mac_nat_output(sc, sbif, &m, &mnr);
6365 if (m == NULL) {
6366 /* packet was deallocated */
6367 BRIDGE_UNLOCK(sc);
6368 return;
6369 }
6370 }
6371 } else {
6372 /*
6373 * sbif is NULL when the bridge interface calls
6374 * bridge_broadcast().
6375 */
6376 bridge_if_out = TRUE;
6377 cksum_op = CHECKSUM_OPERATION_FINALIZE;
6378 sbif = NULL;
6379 src_if = NULL;
6380 }
6381
6382 BRIDGE_LOCK2REF(sc, error);
6383 if (error) {
6384 m_freem(m);
6385 return;
6386 }
6387
6388 TAILQ_FOREACH(dbif, &sc->sc_iflist, bif_next) {
6389 dst_if = dbif->bif_ifp;
6390 if (dst_if == src_if) {
6391 /* skip the interface that the packet came in on */
6392 continue;
6393 }
6394
6395 /* Private segments can not talk to each other */
6396 if (sbif != NULL &&
6397 (sbif->bif_ifflags & dbif->bif_ifflags & IFBIF_PRIVATE)) {
6398 continue;
6399 }
6400
6401 if ((dbif->bif_ifflags & IFBIF_STP) &&
6402 dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
6403 continue;
6404 }
6405
6406 if ((dbif->bif_ifflags & IFBIF_DISCOVER) == 0 &&
6407 (m->m_flags & (M_BCAST | M_MCAST)) == 0) {
6408 continue;
6409 }
6410
6411 if ((dst_if->if_flags & IFF_RUNNING) == 0) {
6412 continue;
6413 }
6414
6415 if (!(dbif->bif_flags & BIFF_MEDIA_ACTIVE)) {
6416 continue;
6417 }
6418
6419 if (TAILQ_NEXT(dbif, bif_next) == NULL) {
6420 mc = m;
6421 used = 1;
6422 } else {
6423 mc = m_dup(m, M_DONTWAIT);
6424 if (mc == NULL) {
6425 (void) ifnet_stat_increment_out(bridge_ifp,
6426 0, 0, 1);
6427 continue;
6428 }
6429 }
6430
6431 /*
6432 * If broadcast input is enabled, do so only if this
6433 * is an input packet.
6434 */
6435 if (!bridge_if_out &&
6436 (dbif->bif_flags & BIFF_INPUT_BROADCAST) != 0) {
6437 mc_in = m_dup(mc, M_DONTWAIT);
6438 /* this could fail, but we continue anyways */
6439 } else {
6440 mc_in = NULL;
6441 }
6442
6443 /* out */
6444 if (translate_mac && mac_nat_bif == dbif) {
6445 /* translate the packet without holding the lock */
6446 bridge_mac_nat_translate(&mc, &mnr, IF_LLADDR(dst_if));
6447 }
6448
6449 sc_filter_flags = sc->sc_filter_flags;
6450 if (runfilt &&
6451 PF_IS_ENABLED && (sc_filter_flags & IFBF_FILT_MEMBER)) {
6452 if (used == 0) {
6453 /* Keep the layer3 header aligned */
6454 int i = min(mc->m_pkthdr.len, max_protohdr);
6455 mc = m_copyup(mc, i, ETHER_ALIGN);
6456 if (mc == NULL) {
6457 (void) ifnet_stat_increment_out(
6458 sc->sc_ifp, 0, 0, 1);
6459 if (mc_in != NULL) {
6460 m_freem(mc_in);
6461 mc_in = NULL;
6462 }
6463 continue;
6464 }
6465 }
6466 if (bridge_pf(&mc, dst_if, sc_filter_flags, FALSE) != 0) {
6467 if (mc_in != NULL) {
6468 m_freem(mc_in);
6469 mc_in = NULL;
6470 }
6471 continue;
6472 }
6473 if (mc == NULL) {
6474 if (mc_in != NULL) {
6475 m_freem(mc_in);
6476 mc_in = NULL;
6477 }
6478 continue;
6479 }
6480 }
6481
6482 if (mc != NULL) {
6483 /* verify checksum if necessary */
6484 if (bif_has_checksum_offload(dbif) && sbif != NULL &&
6485 !bif_has_checksum_offload(sbif)) {
6486 error = bridge_verify_checksum(&mc,
6487 &dbif->bif_stats);
6488 if (error != 0) {
6489 if (mc != NULL) {
6490 m_freem(mc);
6491 }
6492 mc = NULL;
6493 }
6494 }
6495 if (mc != NULL) {
6496 (void) bridge_enqueue(bridge_ifp,
6497 NULL, dst_if, mc, cksum_op);
6498 }
6499 }
6500
6501 /* in */
6502 if (mc_in == NULL) {
6503 continue;
6504 }
6505 bpf_tap_in(dst_if, DLT_EN10MB, mc_in, NULL, 0);
6506 mbuf_pkthdr_setrcvif(mc_in, dst_if);
6507 mbuf_pkthdr_setheader(mc_in, mbuf_data(mc_in));
6508 mbuf_setdata(mc_in, (char *)mbuf_data(mc_in) + ETHER_HDR_LEN,
6509 mbuf_len(mc_in) - ETHER_HDR_LEN);
6510 mbuf_pkthdr_adjustlen(mc_in, -ETHER_HDR_LEN);
6511 mc_in->m_flags |= M_PROTO1; /* set to avoid loops */
6512 dlil_input_packet_list(dst_if, mc_in);
6513 }
6514 if (used == 0) {
6515 m_freem(m);
6516 }
6517
6518
6519 BRIDGE_UNREF(sc);
6520 }
6521
6522 /*
6523 * bridge_span:
6524 *
6525 * Duplicate a packet out one or more interfaces that are in span mode,
6526 * the original mbuf is unmodified.
6527 */
6528 static void
6529 bridge_span(struct bridge_softc *sc, struct mbuf *m)
6530 {
6531 struct bridge_iflist *bif;
6532 struct ifnet *dst_if;
6533 struct mbuf *mc;
6534
6535 if (TAILQ_EMPTY(&sc->sc_spanlist)) {
6536 return;
6537 }
6538
6539 TAILQ_FOREACH(bif, &sc->sc_spanlist, bif_next) {
6540 dst_if = bif->bif_ifp;
6541
6542 if ((dst_if->if_flags & IFF_RUNNING) == 0) {
6543 continue;
6544 }
6545
6546 mc = m_copypacket(m, M_DONTWAIT);
6547 if (mc == NULL) {
6548 (void) ifnet_stat_increment_out(sc->sc_ifp, 0, 0, 1);
6549 continue;
6550 }
6551
6552 (void) bridge_enqueue(sc->sc_ifp, NULL, dst_if, mc,
6553 CHECKSUM_OPERATION_NONE);
6554 }
6555 }
6556
6557
6558 /*
6559 * bridge_rtupdate:
6560 *
6561 * Add a bridge routing entry.
6562 */
6563 static int
6564 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
6565 struct bridge_iflist *bif, int setflags, uint8_t flags)
6566 {
6567 struct bridge_rtnode *brt;
6568 int error;
6569
6570 BRIDGE_LOCK_ASSERT_HELD(sc);
6571
6572 /* Check the source address is valid and not multicast. */
6573 if (ETHER_IS_MULTICAST(dst) ||
6574 (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
6575 dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0) {
6576 return EINVAL;
6577 }
6578
6579
6580 /* 802.1p frames map to vlan 1 */
6581 if (vlan == 0) {
6582 vlan = 1;
6583 }
6584
6585 /*
6586 * A route for this destination might already exist. If so,
6587 * update it, otherwise create a new one.
6588 */
6589 if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
6590 if (sc->sc_brtcnt >= sc->sc_brtmax) {
6591 sc->sc_brtexceeded++;
6592 return ENOSPC;
6593 }
6594 /* Check per interface address limits (if enabled) */
6595 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
6596 bif->bif_addrexceeded++;
6597 return ENOSPC;
6598 }
6599
6600 /*
6601 * Allocate a new bridge forwarding node, and
6602 * initialize the expiration time and Ethernet
6603 * address.
6604 */
6605 brt = zalloc_noblock(bridge_rtnode_pool);
6606 if (brt == NULL) {
6607 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_RT_TABLE,
6608 "zalloc_nolock failed");
6609 return ENOMEM;
6610 }
6611 bzero(brt, sizeof(struct bridge_rtnode));
6612
6613 if (bif->bif_ifflags & IFBIF_STICKY) {
6614 brt->brt_flags = IFBAF_STICKY;
6615 } else {
6616 brt->brt_flags = IFBAF_DYNAMIC;
6617 }
6618
6619 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
6620 brt->brt_vlan = vlan;
6621
6622
6623 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
6624 zfree(bridge_rtnode_pool, brt);
6625 return error;
6626 }
6627 brt->brt_dst = bif;
6628 bif->bif_addrcnt++;
6629 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_RT_TABLE,
6630 "added %02x:%02x:%02x:%02x:%02x:%02x "
6631 "on %s count %u hashsize %u",
6632 dst[0], dst[1], dst[2], dst[3], dst[4], dst[5],
6633 sc->sc_ifp->if_xname, sc->sc_brtcnt,
6634 sc->sc_rthash_size);
6635 }
6636
6637 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
6638 brt->brt_dst != bif) {
6639 brt->brt_dst->bif_addrcnt--;
6640 brt->brt_dst = bif;
6641 brt->brt_dst->bif_addrcnt++;
6642 }
6643
6644 if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
6645 unsigned long now;
6646
6647 now = (unsigned long) net_uptime();
6648 brt->brt_expire = now + sc->sc_brttimeout;
6649 }
6650 if (setflags) {
6651 brt->brt_flags = flags;
6652 }
6653
6654
6655 return 0;
6656 }
6657
6658 /*
6659 * bridge_rtlookup:
6660 *
6661 * Lookup the destination interface for an address.
6662 */
6663 static struct ifnet *
6664 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
6665 {
6666 struct bridge_rtnode *brt;
6667
6668 BRIDGE_LOCK_ASSERT_HELD(sc);
6669
6670 if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL) {
6671 return NULL;
6672 }
6673
6674 return brt->brt_ifp;
6675 }
6676
6677 /*
6678 * bridge_rttrim:
6679 *
6680 * Trim the routine table so that we have a number
6681 * of routing entries less than or equal to the
6682 * maximum number.
6683 */
6684 static void
6685 bridge_rttrim(struct bridge_softc *sc)
6686 {
6687 struct bridge_rtnode *brt, *nbrt;
6688
6689 BRIDGE_LOCK_ASSERT_HELD(sc);
6690
6691 /* Make sure we actually need to do this. */
6692 if (sc->sc_brtcnt <= sc->sc_brtmax) {
6693 return;
6694 }
6695
6696 /* Force an aging cycle; this might trim enough addresses. */
6697 bridge_rtage(sc);
6698 if (sc->sc_brtcnt <= sc->sc_brtmax) {
6699 return;
6700 }
6701
6702 LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
6703 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
6704 bridge_rtnode_destroy(sc, brt);
6705 if (sc->sc_brtcnt <= sc->sc_brtmax) {
6706 return;
6707 }
6708 }
6709 }
6710 }
6711
6712 /*
6713 * bridge_aging_timer:
6714 *
6715 * Aging periodic timer for the bridge routing table.
6716 */
6717 static void
6718 bridge_aging_timer(struct bridge_softc *sc)
6719 {
6720 BRIDGE_LOCK_ASSERT_HELD(sc);
6721
6722 bridge_rtage(sc);
6723 if ((sc->sc_ifp->if_flags & IFF_RUNNING) &&
6724 (sc->sc_flags & SCF_DETACHING) == 0) {
6725 sc->sc_aging_timer.bdc_sc = sc;
6726 sc->sc_aging_timer.bdc_func = bridge_aging_timer;
6727 sc->sc_aging_timer.bdc_ts.tv_sec = bridge_rtable_prune_period;
6728 bridge_schedule_delayed_call(&sc->sc_aging_timer);
6729 }
6730 }
6731
6732 /*
6733 * bridge_rtage:
6734 *
6735 * Perform an aging cycle.
6736 */
6737 static void
6738 bridge_rtage(struct bridge_softc *sc)
6739 {
6740 struct bridge_rtnode *brt, *nbrt;
6741 unsigned long now;
6742
6743 BRIDGE_LOCK_ASSERT_HELD(sc);
6744
6745 now = (unsigned long) net_uptime();
6746
6747 LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
6748 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
6749 if (now >= brt->brt_expire) {
6750 bridge_rtnode_destroy(sc, brt);
6751 }
6752 }
6753 }
6754 if (sc->sc_mac_nat_bif != NULL) {
6755 bridge_mac_nat_age_entries(sc, now);
6756 }
6757 }
6758
6759 /*
6760 * bridge_rtflush:
6761 *
6762 * Remove all dynamic addresses from the bridge.
6763 */
6764 static void
6765 bridge_rtflush(struct bridge_softc *sc, int full)
6766 {
6767 struct bridge_rtnode *brt, *nbrt;
6768
6769 BRIDGE_LOCK_ASSERT_HELD(sc);
6770
6771 LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
6772 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
6773 bridge_rtnode_destroy(sc, brt);
6774 }
6775 }
6776 }
6777
6778 /*
6779 * bridge_rtdaddr:
6780 *
6781 * Remove an address from the table.
6782 */
6783 static int
6784 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
6785 {
6786 struct bridge_rtnode *brt;
6787 int found = 0;
6788
6789 BRIDGE_LOCK_ASSERT_HELD(sc);
6790
6791 /*
6792 * If vlan is zero then we want to delete for all vlans so the lookup
6793 * may return more than one.
6794 */
6795 while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
6796 bridge_rtnode_destroy(sc, brt);
6797 found = 1;
6798 }
6799
6800 return found ? 0 : ENOENT;
6801 }
6802
6803 /*
6804 * bridge_rtdelete:
6805 *
6806 * Delete routes to a specific member interface.
6807 */
6808 static void
6809 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
6810 {
6811 struct bridge_rtnode *brt, *nbrt;
6812
6813 BRIDGE_LOCK_ASSERT_HELD(sc);
6814
6815 LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
6816 if (brt->brt_ifp == ifp && (full ||
6817 (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)) {
6818 bridge_rtnode_destroy(sc, brt);
6819 }
6820 }
6821 }
6822
6823 /*
6824 * bridge_rtable_init:
6825 *
6826 * Initialize the route table for this bridge.
6827 */
6828 static int
6829 bridge_rtable_init(struct bridge_softc *sc)
6830 {
6831 u_int32_t i;
6832
6833 sc->sc_rthash = kalloc_type(struct _bridge_rtnode_list,
6834 BRIDGE_RTHASH_SIZE, Z_WAITOK_ZERO_NOFAIL);
6835 sc->sc_rthash_size = BRIDGE_RTHASH_SIZE;
6836
6837 for (i = 0; i < sc->sc_rthash_size; i++) {
6838 LIST_INIT(&sc->sc_rthash[i]);
6839 }
6840
6841 sc->sc_rthash_key = RandomULong();
6842
6843 LIST_INIT(&sc->sc_rtlist);
6844
6845 return 0;
6846 }
6847
6848 /*
6849 * bridge_rthash_delayed_resize:
6850 *
6851 * Resize the routing table hash on a delayed thread call.
6852 */
6853 static void
6854 bridge_rthash_delayed_resize(struct bridge_softc *sc)
6855 {
6856 u_int32_t new_rthash_size = 0;
6857 u_int32_t old_rthash_size = 0;
6858 struct _bridge_rtnode_list *new_rthash = NULL;
6859 struct _bridge_rtnode_list *old_rthash = NULL;
6860 u_int32_t i;
6861 struct bridge_rtnode *brt;
6862 int error = 0;
6863
6864 BRIDGE_LOCK_ASSERT_HELD(sc);
6865
6866 /*
6867 * Four entries per hash bucket is our ideal load factor
6868 */
6869 if (sc->sc_brtcnt < sc->sc_rthash_size * 4) {
6870 goto out;
6871 }
6872
6873 /*
6874 * Doubling the number of hash buckets may be too simplistic
6875 * especially when facing a spike of new entries
6876 */
6877 new_rthash_size = sc->sc_rthash_size * 2;
6878
6879 sc->sc_flags |= SCF_RESIZING;
6880 BRIDGE_UNLOCK(sc);
6881
6882 new_rthash = kalloc_type(struct _bridge_rtnode_list, new_rthash_size,
6883 Z_WAITOK | Z_ZERO);
6884
6885 BRIDGE_LOCK(sc);
6886 sc->sc_flags &= ~SCF_RESIZING;
6887
6888 if (new_rthash == NULL) {
6889 error = ENOMEM;
6890 goto out;
6891 }
6892 if ((sc->sc_flags & SCF_DETACHING)) {
6893 error = ENODEV;
6894 goto out;
6895 }
6896 /*
6897 * Fail safe from here on
6898 */
6899 old_rthash = sc->sc_rthash;
6900 old_rthash_size = sc->sc_rthash_size;
6901 sc->sc_rthash = new_rthash;
6902 sc->sc_rthash_size = new_rthash_size;
6903
6904 /*
6905 * Get a new key to force entries to be shuffled around to reduce
6906 * the likelihood they will land in the same buckets
6907 */
6908 sc->sc_rthash_key = RandomULong();
6909
6910 for (i = 0; i < sc->sc_rthash_size; i++) {
6911 LIST_INIT(&sc->sc_rthash[i]);
6912 }
6913
6914 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
6915 LIST_REMOVE(brt, brt_hash);
6916 (void) bridge_rtnode_hash(sc, brt);
6917 }
6918 out:
6919 if (error == 0) {
6920 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_RT_TABLE,
6921 "%s new size %u",
6922 sc->sc_ifp->if_xname, sc->sc_rthash_size);
6923 kfree_type(struct _bridge_rtnode_list, old_rthash_size, old_rthash);
6924 } else {
6925 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_RT_TABLE,
6926 "%s failed %d", sc->sc_ifp->if_xname, error);
6927 kfree_type(struct _bridge_rtnode_list, new_rthash_size, new_rthash);
6928 }
6929 }
6930
6931 /*
6932 * Resize the number of hash buckets based on the load factor
6933 * Currently only grow
6934 * Failing to resize the hash table is not fatal
6935 */
6936 static void
6937 bridge_rthash_resize(struct bridge_softc *sc)
6938 {
6939 BRIDGE_LOCK_ASSERT_HELD(sc);
6940
6941 if ((sc->sc_flags & SCF_DETACHING) || (sc->sc_flags & SCF_RESIZING)) {
6942 return;
6943 }
6944
6945 /*
6946 * Four entries per hash bucket is our ideal load factor
6947 */
6948 if (sc->sc_brtcnt < sc->sc_rthash_size * 4) {
6949 return;
6950 }
6951 /*
6952 * Hard limit on the size of the routing hash table
6953 */
6954 if (sc->sc_rthash_size >= bridge_rtable_hash_size_max) {
6955 return;
6956 }
6957
6958 sc->sc_resize_call.bdc_sc = sc;
6959 sc->sc_resize_call.bdc_func = bridge_rthash_delayed_resize;
6960 bridge_schedule_delayed_call(&sc->sc_resize_call);
6961 }
6962
6963 /*
6964 * bridge_rtable_fini:
6965 *
6966 * Deconstruct the route table for this bridge.
6967 */
6968 static void
6969 bridge_rtable_fini(struct bridge_softc *sc)
6970 {
6971 KASSERT(sc->sc_brtcnt == 0,
6972 ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
6973 kfree_type(struct _bridge_rtnode_list, sc->sc_rthash_size,
6974 sc->sc_rthash);
6975 sc->sc_rthash = NULL;
6976 sc->sc_rthash_size = 0;
6977 }
6978
6979 /*
6980 * The following hash function is adapted from "Hash Functions" by Bob Jenkins
6981 * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
6982 */
6983 #define mix(a, b, c) \
6984 do { \
6985 a -= b; a -= c; a ^= (c >> 13); \
6986 b -= c; b -= a; b ^= (a << 8); \
6987 c -= a; c -= b; c ^= (b >> 13); \
6988 a -= b; a -= c; a ^= (c >> 12); \
6989 b -= c; b -= a; b ^= (a << 16); \
6990 c -= a; c -= b; c ^= (b >> 5); \
6991 a -= b; a -= c; a ^= (c >> 3); \
6992 b -= c; b -= a; b ^= (a << 10); \
6993 c -= a; c -= b; c ^= (b >> 15); \
6994 } while ( /*CONSTCOND*/ 0)
6995
6996 static __inline uint32_t
6997 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
6998 {
6999 uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
7000
7001 b += addr[5] << 8;
7002 b += addr[4];
7003 a += addr[3] << 24;
7004 a += addr[2] << 16;
7005 a += addr[1] << 8;
7006 a += addr[0];
7007
7008 mix(a, b, c);
7009
7010 return c & BRIDGE_RTHASH_MASK(sc);
7011 }
7012
7013 #undef mix
7014
7015 static int
7016 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
7017 {
7018 int i, d;
7019
7020 for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
7021 d = ((int)a[i]) - ((int)b[i]);
7022 }
7023
7024 return d;
7025 }
7026
7027 /*
7028 * bridge_rtnode_lookup:
7029 *
7030 * Look up a bridge route node for the specified destination. Compare the
7031 * vlan id or if zero then just return the first match.
7032 */
7033 static struct bridge_rtnode *
7034 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr,
7035 uint16_t vlan)
7036 {
7037 struct bridge_rtnode *brt;
7038 uint32_t hash;
7039 int dir;
7040
7041 BRIDGE_LOCK_ASSERT_HELD(sc);
7042
7043 hash = bridge_rthash(sc, addr);
7044 LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
7045 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
7046 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0)) {
7047 return brt;
7048 }
7049 if (dir > 0) {
7050 return NULL;
7051 }
7052 }
7053
7054 return NULL;
7055 }
7056
7057 /*
7058 * bridge_rtnode_hash:
7059 *
7060 * Insert the specified bridge node into the route hash table.
7061 * This is used when adding a new node or to rehash when resizing
7062 * the hash table
7063 */
7064 static int
7065 bridge_rtnode_hash(struct bridge_softc *sc, struct bridge_rtnode *brt)
7066 {
7067 struct bridge_rtnode *lbrt;
7068 uint32_t hash;
7069 int dir;
7070
7071 BRIDGE_LOCK_ASSERT_HELD(sc);
7072
7073 hash = bridge_rthash(sc, brt->brt_addr);
7074
7075 lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
7076 if (lbrt == NULL) {
7077 LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
7078 goto out;
7079 }
7080
7081 do {
7082 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
7083 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan) {
7084 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_RT_TABLE,
7085 "%s EEXIST %02x:%02x:%02x:%02x:%02x:%02x",
7086 sc->sc_ifp->if_xname,
7087 brt->brt_addr[0], brt->brt_addr[1],
7088 brt->brt_addr[2], brt->brt_addr[3],
7089 brt->brt_addr[4], brt->brt_addr[5]);
7090 return EEXIST;
7091 }
7092 if (dir > 0) {
7093 LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
7094 goto out;
7095 }
7096 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
7097 LIST_INSERT_AFTER(lbrt, brt, brt_hash);
7098 goto out;
7099 }
7100 lbrt = LIST_NEXT(lbrt, brt_hash);
7101 } while (lbrt != NULL);
7102
7103 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_RT_TABLE,
7104 "%s impossible %02x:%02x:%02x:%02x:%02x:%02x",
7105 sc->sc_ifp->if_xname,
7106 brt->brt_addr[0], brt->brt_addr[1], brt->brt_addr[2],
7107 brt->brt_addr[3], brt->brt_addr[4], brt->brt_addr[5]);
7108 out:
7109 return 0;
7110 }
7111
7112 /*
7113 * bridge_rtnode_insert:
7114 *
7115 * Insert the specified bridge node into the route table. We
7116 * assume the entry is not already in the table.
7117 */
7118 static int
7119 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
7120 {
7121 int error;
7122
7123 error = bridge_rtnode_hash(sc, brt);
7124 if (error != 0) {
7125 return error;
7126 }
7127
7128 LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
7129 sc->sc_brtcnt++;
7130
7131 bridge_rthash_resize(sc);
7132
7133 return 0;
7134 }
7135
7136 /*
7137 * bridge_rtnode_destroy:
7138 *
7139 * Destroy a bridge rtnode.
7140 */
7141 static void
7142 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
7143 {
7144 BRIDGE_LOCK_ASSERT_HELD(sc);
7145
7146 LIST_REMOVE(brt, brt_hash);
7147
7148 LIST_REMOVE(brt, brt_list);
7149 sc->sc_brtcnt--;
7150 brt->brt_dst->bif_addrcnt--;
7151 zfree(bridge_rtnode_pool, brt);
7152 }
7153
7154 #if BRIDGESTP
7155 /*
7156 * bridge_rtable_expire:
7157 *
7158 * Set the expiry time for all routes on an interface.
7159 */
7160 static void
7161 bridge_rtable_expire(struct ifnet *ifp, int age)
7162 {
7163 struct bridge_softc *sc = ifp->if_bridge;
7164 struct bridge_rtnode *brt;
7165
7166 BRIDGE_LOCK(sc);
7167
7168 /*
7169 * If the age is zero then flush, otherwise set all the expiry times to
7170 * age for the interface
7171 */
7172 if (age == 0) {
7173 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
7174 } else {
7175 unsigned long now;
7176
7177 now = (unsigned long) net_uptime();
7178
7179 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
7180 /* Cap the expiry time to 'age' */
7181 if (brt->brt_ifp == ifp &&
7182 brt->brt_expire > now + age &&
7183 (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
7184 brt->brt_expire = now + age;
7185 }
7186 }
7187 }
7188 BRIDGE_UNLOCK(sc);
7189 }
7190
7191 /*
7192 * bridge_state_change:
7193 *
7194 * Callback from the bridgestp code when a port changes states.
7195 */
7196 static void
7197 bridge_state_change(struct ifnet *ifp, int state)
7198 {
7199 struct bridge_softc *sc = ifp->if_bridge;
7200 static const char *stpstates[] = {
7201 "disabled",
7202 "listening",
7203 "learning",
7204 "forwarding",
7205 "blocking",
7206 "discarding"
7207 };
7208
7209 if (log_stp) {
7210 log(LOG_NOTICE, "%s: state changed to %s on %s",
7211 sc->sc_ifp->if_xname,
7212 stpstates[state], ifp->if_xname);
7213 }
7214 }
7215 #endif /* BRIDGESTP */
7216
7217 /*
7218 * bridge_set_bpf_tap:
7219 *
7220 * Sets ups the BPF callbacks.
7221 */
7222 static errno_t
7223 bridge_set_bpf_tap(ifnet_t ifp, bpf_tap_mode mode, bpf_packet_func bpf_callback)
7224 {
7225 struct bridge_softc *sc = (struct bridge_softc *)ifnet_softc(ifp);
7226
7227 /* TBD locking */
7228 if (sc == NULL || (sc->sc_flags & SCF_DETACHING)) {
7229 return ENODEV;
7230 }
7231 switch (mode) {
7232 case BPF_TAP_DISABLE:
7233 sc->sc_bpf_input = sc->sc_bpf_output = NULL;
7234 break;
7235
7236 case BPF_TAP_INPUT:
7237 sc->sc_bpf_input = bpf_callback;
7238 break;
7239
7240 case BPF_TAP_OUTPUT:
7241 sc->sc_bpf_output = bpf_callback;
7242 break;
7243
7244 case BPF_TAP_INPUT_OUTPUT:
7245 sc->sc_bpf_input = sc->sc_bpf_output = bpf_callback;
7246 break;
7247
7248 default:
7249 break;
7250 }
7251
7252 return 0;
7253 }
7254
7255 /*
7256 * bridge_detach:
7257 *
7258 * Callback when interface has been detached.
7259 */
7260 static void
7261 bridge_detach(ifnet_t ifp)
7262 {
7263 struct bridge_softc *sc = (struct bridge_softc *)ifnet_softc(ifp);
7264
7265 #if BRIDGESTP
7266 bstp_detach(&sc->sc_stp);
7267 #endif /* BRIDGESTP */
7268
7269 /* Tear down the routing table. */
7270 bridge_rtable_fini(sc);
7271
7272 lck_mtx_lock(&bridge_list_mtx);
7273 LIST_REMOVE(sc, sc_list);
7274 lck_mtx_unlock(&bridge_list_mtx);
7275
7276 ifnet_release(ifp);
7277
7278 lck_mtx_destroy(&sc->sc_mtx, &bridge_lock_grp);
7279 kfree_type(struct bridge_softc, sc);
7280 }
7281
7282 /*
7283 * bridge_bpf_input:
7284 *
7285 * Invoke the input BPF callback if enabled
7286 */
7287 static errno_t
7288 bridge_bpf_input(ifnet_t ifp, struct mbuf *m, const char * func, int line)
7289 {
7290 struct bridge_softc *sc = (struct bridge_softc *)ifnet_softc(ifp);
7291 bpf_packet_func input_func = sc->sc_bpf_input;
7292
7293 if (input_func != NULL) {
7294 if (mbuf_pkthdr_rcvif(m) != ifp) {
7295 BRIDGE_LOG(LOG_NOTICE, 0,
7296 "%s.%d: rcvif: 0x%llx != ifp 0x%llx", func, line,
7297 (uint64_t)VM_KERNEL_ADDRPERM(mbuf_pkthdr_rcvif(m)),
7298 (uint64_t)VM_KERNEL_ADDRPERM(ifp));
7299 }
7300 (*input_func)(ifp, m);
7301 }
7302 return 0;
7303 }
7304
7305 /*
7306 * bridge_bpf_output:
7307 *
7308 * Invoke the output BPF callback if enabled
7309 */
7310 static errno_t
7311 bridge_bpf_output(ifnet_t ifp, struct mbuf *m)
7312 {
7313 struct bridge_softc *sc = (struct bridge_softc *)ifnet_softc(ifp);
7314 bpf_packet_func output_func = sc->sc_bpf_output;
7315
7316 if (output_func != NULL) {
7317 (*output_func)(ifp, m);
7318 }
7319 return 0;
7320 }
7321
7322 /*
7323 * bridge_link_event:
7324 *
7325 * Report a data link event on an interface
7326 */
7327 static void
7328 bridge_link_event(struct ifnet *ifp, u_int32_t event_code)
7329 {
7330 struct event {
7331 u_int32_t ifnet_family;
7332 u_int32_t unit;
7333 char if_name[IFNAMSIZ];
7334 };
7335 _Alignas(struct kern_event_msg) char message[sizeof(struct kern_event_msg) + sizeof(struct event)] = { 0 };
7336 struct kern_event_msg *header = (struct kern_event_msg*)message;
7337 struct event *data = (struct event *)(header + 1);
7338
7339 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_LIFECYCLE,
7340 "%s event_code %u - %s", ifp->if_xname,
7341 event_code, dlil_kev_dl_code_str(event_code));
7342 header->total_size = sizeof(message);
7343 header->vendor_code = KEV_VENDOR_APPLE;
7344 header->kev_class = KEV_NETWORK_CLASS;
7345 header->kev_subclass = KEV_DL_SUBCLASS;
7346 header->event_code = event_code;
7347 data->ifnet_family = ifnet_family(ifp);
7348 data->unit = (u_int32_t)ifnet_unit(ifp);
7349 strlcpy(data->if_name, ifnet_name(ifp), IFNAMSIZ);
7350 ifnet_event(ifp, header);
7351 }
7352
7353 #define BRIDGE_HF_DROP(reason, func, line) { \
7354 bridge_hostfilter_stats.reason++; \
7355 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_HOSTFILTER, \
7356 "%s.%d" #reason, func, line); \
7357 error = EINVAL; \
7358 }
7359
7360 /*
7361 * Make sure this is a DHCP or Bootp request that match the host filter
7362 */
7363 static int
7364 bridge_dhcp_filter(struct bridge_iflist *bif, struct mbuf *m, size_t offset)
7365 {
7366 int error = EINVAL;
7367 struct dhcp dhcp;
7368
7369 /*
7370 * Note: We use the dhcp structure because bootp structure definition
7371 * is larger and some vendors do not pad the request
7372 */
7373 error = mbuf_copydata(m, offset, sizeof(struct dhcp), &dhcp);
7374 if (error != 0) {
7375 BRIDGE_HF_DROP(brhf_dhcp_too_small, __func__, __LINE__);
7376 goto done;
7377 }
7378 if (dhcp.dp_op != BOOTREQUEST) {
7379 BRIDGE_HF_DROP(brhf_dhcp_bad_op, __func__, __LINE__);
7380 goto done;
7381 }
7382 /*
7383 * The hardware address must be an exact match
7384 */
7385 if (dhcp.dp_htype != ARPHRD_ETHER) {
7386 BRIDGE_HF_DROP(brhf_dhcp_bad_htype, __func__, __LINE__);
7387 goto done;
7388 }
7389 if (dhcp.dp_hlen != ETHER_ADDR_LEN) {
7390 BRIDGE_HF_DROP(brhf_dhcp_bad_hlen, __func__, __LINE__);
7391 goto done;
7392 }
7393 if (bcmp(dhcp.dp_chaddr, bif->bif_hf_hwsrc,
7394 ETHER_ADDR_LEN) != 0) {
7395 BRIDGE_HF_DROP(brhf_dhcp_bad_chaddr, __func__, __LINE__);
7396 goto done;
7397 }
7398 /*
7399 * Client address must match the host address or be not specified
7400 */
7401 if (dhcp.dp_ciaddr.s_addr != bif->bif_hf_ipsrc.s_addr &&
7402 dhcp.dp_ciaddr.s_addr != INADDR_ANY) {
7403 BRIDGE_HF_DROP(brhf_dhcp_bad_ciaddr, __func__, __LINE__);
7404 goto done;
7405 }
7406 error = 0;
7407 done:
7408 return error;
7409 }
7410
7411 static int
7412 bridge_host_filter(struct bridge_iflist *bif, mbuf_t *data)
7413 {
7414 int error = EINVAL;
7415 struct ether_header *eh;
7416 static struct in_addr inaddr_any = { .s_addr = INADDR_ANY };
7417 mbuf_t m = *data;
7418
7419 eh = mtod(m, struct ether_header *);
7420
7421 /*
7422 * Restrict the source hardware address
7423 */
7424 if ((bif->bif_flags & BIFF_HF_HWSRC) == 0 ||
7425 bcmp(eh->ether_shost, bif->bif_hf_hwsrc,
7426 ETHER_ADDR_LEN) != 0) {
7427 BRIDGE_HF_DROP(brhf_bad_ether_srchw_addr, __func__, __LINE__);
7428 goto done;
7429 }
7430
7431 /*
7432 * Restrict Ethernet protocols to ARP and IP
7433 */
7434 if (eh->ether_type == htons(ETHERTYPE_ARP)) {
7435 struct ether_arp *ea;
7436 size_t minlen = sizeof(struct ether_header) +
7437 sizeof(struct ether_arp);
7438
7439 /*
7440 * Make the Ethernet and ARP headers contiguous
7441 */
7442 if (mbuf_pkthdr_len(m) < minlen) {
7443 BRIDGE_HF_DROP(brhf_arp_too_small, __func__, __LINE__);
7444 goto done;
7445 }
7446 if (mbuf_len(m) < minlen && mbuf_pullup(data, minlen) != 0) {
7447 BRIDGE_HF_DROP(brhf_arp_pullup_failed,
7448 __func__, __LINE__);
7449 goto done;
7450 }
7451 m = *data;
7452
7453 /*
7454 * Verify this is an ethernet/ip arp
7455 */
7456 eh = mtod(m, struct ether_header *);
7457 ea = (struct ether_arp *)(eh + 1);
7458 if (ea->arp_hrd != htons(ARPHRD_ETHER)) {
7459 BRIDGE_HF_DROP(brhf_arp_bad_hw_type,
7460 __func__, __LINE__);
7461 goto done;
7462 }
7463 if (ea->arp_pro != htons(ETHERTYPE_IP)) {
7464 BRIDGE_HF_DROP(brhf_arp_bad_pro_type,
7465 __func__, __LINE__);
7466 goto done;
7467 }
7468 /*
7469 * Verify the address lengths are correct
7470 */
7471 if (ea->arp_hln != ETHER_ADDR_LEN) {
7472 BRIDGE_HF_DROP(brhf_arp_bad_hw_len, __func__, __LINE__);
7473 goto done;
7474 }
7475 if (ea->arp_pln != sizeof(struct in_addr)) {
7476 BRIDGE_HF_DROP(brhf_arp_bad_pro_len,
7477 __func__, __LINE__);
7478 goto done;
7479 }
7480
7481 /*
7482 * Allow only ARP request or ARP reply
7483 */
7484 if (ea->arp_op != htons(ARPOP_REQUEST) &&
7485 ea->arp_op != htons(ARPOP_REPLY)) {
7486 BRIDGE_HF_DROP(brhf_arp_bad_op, __func__, __LINE__);
7487 goto done;
7488 }
7489 /*
7490 * Verify source hardware address matches
7491 */
7492 if (bcmp(ea->arp_sha, bif->bif_hf_hwsrc,
7493 ETHER_ADDR_LEN) != 0) {
7494 BRIDGE_HF_DROP(brhf_arp_bad_sha, __func__, __LINE__);
7495 goto done;
7496 }
7497 /*
7498 * Verify source protocol address:
7499 * May be null for an ARP probe
7500 */
7501 if (bcmp(ea->arp_spa, &bif->bif_hf_ipsrc.s_addr,
7502 sizeof(struct in_addr)) != 0 &&
7503 bcmp(ea->arp_spa, &inaddr_any,
7504 sizeof(struct in_addr)) != 0) {
7505 BRIDGE_HF_DROP(brhf_arp_bad_spa, __func__, __LINE__);
7506 goto done;
7507 }
7508 bridge_hostfilter_stats.brhf_arp_ok += 1;
7509 error = 0;
7510 } else if (eh->ether_type == htons(ETHERTYPE_IP)) {
7511 size_t minlen = sizeof(struct ether_header) + sizeof(struct ip);
7512 struct ip iphdr;
7513 size_t offset;
7514
7515 /*
7516 * Make the Ethernet and IP headers contiguous
7517 */
7518 if (mbuf_pkthdr_len(m) < minlen) {
7519 BRIDGE_HF_DROP(brhf_ip_too_small, __func__, __LINE__);
7520 goto done;
7521 }
7522 offset = sizeof(struct ether_header);
7523 error = mbuf_copydata(m, offset, sizeof(struct ip), &iphdr);
7524 if (error != 0) {
7525 BRIDGE_HF_DROP(brhf_ip_too_small, __func__, __LINE__);
7526 goto done;
7527 }
7528 /*
7529 * Verify the source IP address
7530 */
7531 if (iphdr.ip_p == IPPROTO_UDP) {
7532 struct udphdr udp;
7533
7534 minlen += sizeof(struct udphdr);
7535 if (mbuf_pkthdr_len(m) < minlen) {
7536 BRIDGE_HF_DROP(brhf_ip_too_small,
7537 __func__, __LINE__);
7538 goto done;
7539 }
7540
7541 /*
7542 * Allow all zero addresses for DHCP requests
7543 */
7544 if (iphdr.ip_src.s_addr != bif->bif_hf_ipsrc.s_addr &&
7545 iphdr.ip_src.s_addr != INADDR_ANY) {
7546 BRIDGE_HF_DROP(brhf_ip_bad_srcaddr,
7547 __func__, __LINE__);
7548 goto done;
7549 }
7550 offset = sizeof(struct ether_header) +
7551 (IP_VHL_HL(iphdr.ip_vhl) << 2);
7552 error = mbuf_copydata(m, offset,
7553 sizeof(struct udphdr), &udp);
7554 if (error != 0) {
7555 BRIDGE_HF_DROP(brhf_ip_too_small,
7556 __func__, __LINE__);
7557 goto done;
7558 }
7559 /*
7560 * Either it's a Bootp/DHCP packet that we like or
7561 * it's a UDP packet from the host IP as source address
7562 */
7563 if (udp.uh_sport == htons(IPPORT_BOOTPC) &&
7564 udp.uh_dport == htons(IPPORT_BOOTPS)) {
7565 minlen += sizeof(struct dhcp);
7566 if (mbuf_pkthdr_len(m) < minlen) {
7567 BRIDGE_HF_DROP(brhf_ip_too_small,
7568 __func__, __LINE__);
7569 goto done;
7570 }
7571 offset += sizeof(struct udphdr);
7572 error = bridge_dhcp_filter(bif, m, offset);
7573 if (error != 0) {
7574 goto done;
7575 }
7576 } else if (iphdr.ip_src.s_addr == INADDR_ANY) {
7577 BRIDGE_HF_DROP(brhf_ip_bad_srcaddr,
7578 __func__, __LINE__);
7579 goto done;
7580 }
7581 } else if (iphdr.ip_src.s_addr != bif->bif_hf_ipsrc.s_addr ||
7582 bif->bif_hf_ipsrc.s_addr == INADDR_ANY) {
7583 BRIDGE_HF_DROP(brhf_ip_bad_srcaddr, __func__, __LINE__);
7584 goto done;
7585 }
7586 /*
7587 * Allow only boring IP protocols
7588 */
7589 if (iphdr.ip_p != IPPROTO_TCP &&
7590 iphdr.ip_p != IPPROTO_UDP &&
7591 iphdr.ip_p != IPPROTO_ICMP &&
7592 iphdr.ip_p != IPPROTO_ESP &&
7593 iphdr.ip_p != IPPROTO_AH &&
7594 iphdr.ip_p != IPPROTO_GRE) {
7595 BRIDGE_HF_DROP(brhf_ip_bad_proto, __func__, __LINE__);
7596 goto done;
7597 }
7598 bridge_hostfilter_stats.brhf_ip_ok += 1;
7599 error = 0;
7600 } else {
7601 BRIDGE_HF_DROP(brhf_bad_ether_type, __func__, __LINE__);
7602 goto done;
7603 }
7604 done:
7605 if (error != 0) {
7606 if (BRIDGE_DBGF_ENABLED(BR_DBGF_HOSTFILTER)) {
7607 if (m) {
7608 brlog_mbuf_data(m, 0,
7609 sizeof(struct ether_header) +
7610 sizeof(struct ip));
7611 }
7612 }
7613
7614 if (m != NULL) {
7615 m_freem(m);
7616 }
7617 }
7618 return error;
7619 }
7620
7621 /*
7622 * MAC NAT
7623 */
7624
7625 static errno_t
7626 bridge_mac_nat_enable(struct bridge_softc *sc, struct bridge_iflist *bif)
7627 {
7628 errno_t error = 0;
7629
7630 BRIDGE_LOCK_ASSERT_HELD(sc);
7631
7632 if (IFNET_IS_VMNET(bif->bif_ifp)) {
7633 error = EINVAL;
7634 goto done;
7635 }
7636 if (sc->sc_mac_nat_bif != NULL) {
7637 if (sc->sc_mac_nat_bif != bif) {
7638 error = EBUSY;
7639 }
7640 goto done;
7641 }
7642 sc->sc_mac_nat_bif = bif;
7643 bif->bif_ifflags |= IFBIF_MAC_NAT;
7644 bridge_mac_nat_populate_entries(sc);
7645
7646 done:
7647 return error;
7648 }
7649
7650 static void
7651 bridge_mac_nat_disable(struct bridge_softc *sc)
7652 {
7653 struct bridge_iflist *mac_nat_bif = sc->sc_mac_nat_bif;
7654
7655 assert(mac_nat_bif != NULL);
7656 bridge_mac_nat_flush_entries(sc, mac_nat_bif);
7657 mac_nat_bif->bif_ifflags &= ~IFBIF_MAC_NAT;
7658 sc->sc_mac_nat_bif = NULL;
7659 return;
7660 }
7661
7662 static void
7663 mac_nat_entry_print2(struct mac_nat_entry *mne,
7664 char *ifname, const char *msg1, const char *msg2)
7665 {
7666 int af;
7667 char etopbuf[24];
7668 char ntopbuf[MAX_IPv6_STR_LEN];
7669 const char *space;
7670
7671 af = ((mne->mne_flags & MNE_FLAGS_IPV6) != 0) ? AF_INET6 : AF_INET;
7672 ether_ntop(etopbuf, sizeof(etopbuf), mne->mne_mac);
7673 (void)inet_ntop(af, &mne->mne_u, ntopbuf, sizeof(ntopbuf));
7674 if (msg2 == NULL) {
7675 msg2 = "";
7676 space = "";
7677 } else {
7678 space = " ";
7679 }
7680 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
7681 "%s %s%s%s %p (%s, %s, %s)",
7682 ifname, msg1, space, msg2, mne, mne->mne_bif->bif_ifp->if_xname,
7683 ntopbuf, etopbuf);
7684 }
7685
7686 static void
7687 mac_nat_entry_print(struct mac_nat_entry *mne,
7688 char *ifname, const char *msg)
7689 {
7690 mac_nat_entry_print2(mne, ifname, msg, NULL);
7691 }
7692
7693 static struct mac_nat_entry *
7694 bridge_lookup_mac_nat_entry(struct bridge_softc *sc, int af, void * ip)
7695 {
7696 struct mac_nat_entry *mne;
7697 struct mac_nat_entry *ret_mne = NULL;
7698
7699 if (af == AF_INET) {
7700 in_addr_t s_addr = ((struct in_addr *)ip)->s_addr;
7701
7702 LIST_FOREACH(mne, &sc->sc_mne_list, mne_list) {
7703 if (mne->mne_ip.s_addr == s_addr) {
7704 if (BRIDGE_DBGF_ENABLED(BR_DBGF_MAC_NAT)) {
7705 mac_nat_entry_print(mne, sc->sc_if_xname,
7706 "found");
7707 }
7708 ret_mne = mne;
7709 break;
7710 }
7711 }
7712 } else {
7713 const struct in6_addr *ip6 = (const struct in6_addr *)ip;
7714
7715 LIST_FOREACH(mne, &sc->sc_mne_list_v6, mne_list) {
7716 if (IN6_ARE_ADDR_EQUAL(&mne->mne_ip6, ip6)) {
7717 if (BRIDGE_DBGF_ENABLED(BR_DBGF_MAC_NAT)) {
7718 mac_nat_entry_print(mne, sc->sc_if_xname,
7719 "found");
7720 }
7721 ret_mne = mne;
7722 break;
7723 }
7724 }
7725 }
7726 return ret_mne;
7727 }
7728
7729 static void
7730 bridge_destroy_mac_nat_entry(struct bridge_softc *sc,
7731 struct mac_nat_entry *mne, const char *reason)
7732 {
7733 LIST_REMOVE(mne, mne_list);
7734 if (BRIDGE_DBGF_ENABLED(BR_DBGF_MAC_NAT)) {
7735 mac_nat_entry_print(mne, sc->sc_if_xname, reason);
7736 }
7737 zfree(bridge_mne_pool, mne);
7738 sc->sc_mne_count--;
7739 }
7740
7741 static struct mac_nat_entry *
7742 bridge_create_mac_nat_entry(struct bridge_softc *sc,
7743 struct bridge_iflist *bif, int af, const void *ip, uint8_t *eaddr)
7744 {
7745 struct mac_nat_entry_list *list;
7746 struct mac_nat_entry *mne;
7747
7748 if (sc->sc_mne_count >= sc->sc_mne_max) {
7749 sc->sc_mne_allocation_failures++;
7750 return NULL;
7751 }
7752 mne = zalloc_noblock(bridge_mne_pool);
7753 if (mne == NULL) {
7754 sc->sc_mne_allocation_failures++;
7755 return NULL;
7756 }
7757 sc->sc_mne_count++;
7758 bzero(mne, sizeof(*mne));
7759 bcopy(eaddr, mne->mne_mac, sizeof(mne->mne_mac));
7760 mne->mne_bif = bif;
7761 if (af == AF_INET) {
7762 bcopy(ip, &mne->mne_ip, sizeof(mne->mne_ip));
7763 list = &sc->sc_mne_list;
7764 } else {
7765 bcopy(ip, &mne->mne_ip6, sizeof(mne->mne_ip6));
7766 mne->mne_flags |= MNE_FLAGS_IPV6;
7767 list = &sc->sc_mne_list_v6;
7768 }
7769 LIST_INSERT_HEAD(list, mne, mne_list);
7770 mne->mne_expire = (unsigned long)net_uptime() + sc->sc_brttimeout;
7771 if (BRIDGE_DBGF_ENABLED(BR_DBGF_MAC_NAT)) {
7772 mac_nat_entry_print(mne, sc->sc_if_xname, "created");
7773 }
7774 return mne;
7775 }
7776
7777 static struct mac_nat_entry *
7778 bridge_update_mac_nat_entry(struct bridge_softc *sc,
7779 struct bridge_iflist *bif, int af, void *ip, uint8_t *eaddr)
7780 {
7781 struct mac_nat_entry *mne;
7782
7783 mne = bridge_lookup_mac_nat_entry(sc, af, ip);
7784 if (mne != NULL) {
7785 struct bridge_iflist *mac_nat_bif = sc->sc_mac_nat_bif;
7786
7787 if (mne->mne_bif == mac_nat_bif) {
7788 /* the MAC NAT interface takes precedence */
7789 if (BRIDGE_DBGF_ENABLED(BR_DBGF_MAC_NAT)) {
7790 if (mne->mne_bif != bif) {
7791 mac_nat_entry_print2(mne,
7792 sc->sc_if_xname, "reject",
7793 bif->bif_ifp->if_xname);
7794 }
7795 }
7796 } else if (mne->mne_bif != bif) {
7797 const char *old_if = mne->mne_bif->bif_ifp->if_xname;
7798
7799 mne->mne_bif = bif;
7800 if (BRIDGE_DBGF_ENABLED(BR_DBGF_MAC_NAT)) {
7801 mac_nat_entry_print2(mne,
7802 sc->sc_if_xname, "replaced",
7803 old_if);
7804 }
7805 bcopy(eaddr, mne->mne_mac, sizeof(mne->mne_mac));
7806 }
7807 mne->mne_expire = (unsigned long)net_uptime() +
7808 sc->sc_brttimeout;
7809 } else {
7810 mne = bridge_create_mac_nat_entry(sc, bif, af, ip, eaddr);
7811 }
7812 return mne;
7813 }
7814
7815 static void
7816 bridge_mac_nat_flush_entries_common(struct bridge_softc *sc,
7817 struct mac_nat_entry_list *list, struct bridge_iflist *bif)
7818 {
7819 struct mac_nat_entry *mne;
7820 struct mac_nat_entry *tmne;
7821
7822 LIST_FOREACH_SAFE(mne, list, mne_list, tmne) {
7823 if (bif != NULL && mne->mne_bif != bif) {
7824 continue;
7825 }
7826 bridge_destroy_mac_nat_entry(sc, mne, "flushed");
7827 }
7828 }
7829
7830 /*
7831 * bridge_mac_nat_flush_entries:
7832 *
7833 * Flush MAC NAT entries for the specified member. Flush all entries if
7834 * the member is the one that requires MAC NAT, otherwise just flush the
7835 * ones for the specified member.
7836 */
7837 static void
7838 bridge_mac_nat_flush_entries(struct bridge_softc *sc, struct bridge_iflist * bif)
7839 {
7840 struct bridge_iflist *flush_bif;
7841
7842 flush_bif = (bif == sc->sc_mac_nat_bif) ? NULL : bif;
7843 bridge_mac_nat_flush_entries_common(sc, &sc->sc_mne_list, flush_bif);
7844 bridge_mac_nat_flush_entries_common(sc, &sc->sc_mne_list_v6, flush_bif);
7845 }
7846
7847 static void
7848 bridge_mac_nat_populate_entries(struct bridge_softc *sc)
7849 {
7850 errno_t error;
7851 ifnet_t ifp;
7852 ifaddr_t *list;
7853 struct bridge_iflist *mac_nat_bif = sc->sc_mac_nat_bif;
7854
7855 assert(mac_nat_bif != NULL);
7856 ifp = mac_nat_bif->bif_ifp;
7857 error = ifnet_get_address_list(ifp, &list);
7858 if (error != 0) {
7859 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
7860 "ifnet_get_address_list(%s) failed %d",
7861 ifp->if_xname, error);
7862 return;
7863 }
7864 for (ifaddr_t *scan = list; *scan != NULL; scan++) {
7865 sa_family_t af;
7866 void *ip;
7867
7868 union {
7869 struct sockaddr sa;
7870 struct sockaddr_in sin;
7871 struct sockaddr_in6 sin6;
7872 } u;
7873 af = ifaddr_address_family(*scan);
7874 switch (af) {
7875 case AF_INET:
7876 case AF_INET6:
7877 error = ifaddr_address(*scan, &u.sa, sizeof(u));
7878 if (error != 0) {
7879 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
7880 "ifaddr_address failed %d",
7881 error);
7882 break;
7883 }
7884 if (af == AF_INET) {
7885 ip = (void *)&u.sin.sin_addr;
7886 } else {
7887 if (IN6_IS_ADDR_LINKLOCAL(&u.sin6.sin6_addr)) {
7888 /* remove scope ID */
7889 u.sin6.sin6_addr.s6_addr16[1] = 0;
7890 }
7891 ip = (void *)&u.sin6.sin6_addr;
7892 }
7893 bridge_create_mac_nat_entry(sc, mac_nat_bif, af, ip,
7894 (uint8_t *)IF_LLADDR(ifp));
7895 break;
7896 default:
7897 break;
7898 }
7899 }
7900 ifnet_free_address_list(list);
7901 return;
7902 }
7903
7904 static void
7905 bridge_mac_nat_age_entries_common(struct bridge_softc *sc,
7906 struct mac_nat_entry_list *list, unsigned long now)
7907 {
7908 struct mac_nat_entry *mne;
7909 struct mac_nat_entry *tmne;
7910
7911 LIST_FOREACH_SAFE(mne, list, mne_list, tmne) {
7912 if (now >= mne->mne_expire) {
7913 bridge_destroy_mac_nat_entry(sc, mne, "aged out");
7914 }
7915 }
7916 }
7917
7918 static void
7919 bridge_mac_nat_age_entries(struct bridge_softc *sc, unsigned long now)
7920 {
7921 if (sc->sc_mac_nat_bif == NULL) {
7922 return;
7923 }
7924 bridge_mac_nat_age_entries_common(sc, &sc->sc_mne_list, now);
7925 bridge_mac_nat_age_entries_common(sc, &sc->sc_mne_list_v6, now);
7926 }
7927
7928 static const char *
7929 get_in_out_string(boolean_t is_output)
7930 {
7931 return is_output ? "OUT" : "IN";
7932 }
7933
7934 /*
7935 * is_valid_arp_packet:
7936 * Verify that this is a valid ARP packet.
7937 *
7938 * Returns TRUE if the packet is valid, FALSE otherwise.
7939 */
7940 static boolean_t
7941 is_valid_arp_packet(mbuf_t *data, boolean_t is_output,
7942 struct ether_header **eh_p, struct ether_arp **ea_p)
7943 {
7944 struct ether_arp *ea;
7945 struct ether_header *eh;
7946 size_t minlen = sizeof(struct ether_header) + sizeof(struct ether_arp);
7947 boolean_t is_valid = FALSE;
7948 int flags = is_output ? BR_DBGF_OUTPUT : BR_DBGF_INPUT;
7949
7950 if (mbuf_pkthdr_len(*data) < minlen) {
7951 BRIDGE_LOG(LOG_DEBUG, flags,
7952 "ARP %s short frame %lu < %lu",
7953 get_in_out_string(is_output),
7954 mbuf_pkthdr_len(*data), minlen);
7955 goto done;
7956 }
7957 if (mbuf_len(*data) < minlen && mbuf_pullup(data, minlen) != 0) {
7958 BRIDGE_LOG(LOG_DEBUG, flags,
7959 "ARP %s size %lu mbuf_pullup fail",
7960 get_in_out_string(is_output),
7961 minlen);
7962 *data = NULL;
7963 goto done;
7964 }
7965
7966 /* validate ARP packet */
7967 eh = mtod(*data, struct ether_header *);
7968 ea = (struct ether_arp *)(eh + 1);
7969 if (ntohs(ea->arp_hrd) != ARPHRD_ETHER) {
7970 BRIDGE_LOG(LOG_DEBUG, flags,
7971 "ARP %s htype not ethernet",
7972 get_in_out_string(is_output));
7973 goto done;
7974 }
7975 if (ea->arp_hln != ETHER_ADDR_LEN) {
7976 BRIDGE_LOG(LOG_DEBUG, flags,
7977 "ARP %s hlen not ethernet",
7978 get_in_out_string(is_output));
7979 goto done;
7980 }
7981 if (ntohs(ea->arp_pro) != ETHERTYPE_IP) {
7982 BRIDGE_LOG(LOG_DEBUG, flags,
7983 "ARP %s ptype not IP",
7984 get_in_out_string(is_output));
7985 goto done;
7986 }
7987 if (ea->arp_pln != sizeof(struct in_addr)) {
7988 BRIDGE_LOG(LOG_DEBUG, flags,
7989 "ARP %s plen not IP",
7990 get_in_out_string(is_output));
7991 goto done;
7992 }
7993 is_valid = TRUE;
7994 *ea_p = ea;
7995 *eh_p = eh;
7996 done:
7997 return is_valid;
7998 }
7999
8000 static struct mac_nat_entry *
8001 bridge_mac_nat_arp_input(struct bridge_softc *sc, mbuf_t *data)
8002 {
8003 struct ether_arp *ea;
8004 struct ether_header *eh;
8005 struct mac_nat_entry *mne = NULL;
8006 u_short op;
8007 struct in_addr tpa;
8008
8009 if (!is_valid_arp_packet(data, FALSE, &eh, &ea)) {
8010 goto done;
8011 }
8012 op = ntohs(ea->arp_op);
8013 switch (op) {
8014 case ARPOP_REQUEST:
8015 case ARPOP_REPLY:
8016 /* only care about REQUEST and REPLY */
8017 break;
8018 default:
8019 goto done;
8020 }
8021
8022 /* check the target IP address for a NAT entry */
8023 bcopy(ea->arp_tpa, &tpa, sizeof(tpa));
8024 if (tpa.s_addr != 0) {
8025 mne = bridge_lookup_mac_nat_entry(sc, AF_INET, &tpa);
8026 }
8027 if (mne != NULL) {
8028 if (op == ARPOP_REPLY) {
8029 /* translate the MAC address */
8030 if (BRIDGE_DBGF_ENABLED(BR_DBGF_MAC_NAT)) {
8031 char mac_src[24];
8032 char mac_dst[24];
8033
8034 ether_ntop(mac_src, sizeof(mac_src),
8035 ea->arp_tha);
8036 ether_ntop(mac_dst, sizeof(mac_dst),
8037 mne->mne_mac);
8038 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8039 "%s %s ARP %s -> %s",
8040 sc->sc_if_xname,
8041 mne->mne_bif->bif_ifp->if_xname,
8042 mac_src, mac_dst);
8043 }
8044 bcopy(mne->mne_mac, ea->arp_tha, sizeof(ea->arp_tha));
8045 }
8046 } else {
8047 /* handle conflicting ARP (sender matches mne) */
8048 struct in_addr spa;
8049
8050 bcopy(ea->arp_spa, &spa, sizeof(spa));
8051 if (spa.s_addr != 0 && spa.s_addr != tpa.s_addr) {
8052 /* check the source IP for a NAT entry */
8053 mne = bridge_lookup_mac_nat_entry(sc, AF_INET, &spa);
8054 }
8055 }
8056
8057 done:
8058 return mne;
8059 }
8060
8061 static boolean_t
8062 bridge_mac_nat_arp_output(struct bridge_softc *sc,
8063 struct bridge_iflist *bif, mbuf_t *data, struct mac_nat_record *mnr)
8064 {
8065 struct ether_arp *ea;
8066 struct ether_header *eh;
8067 struct in_addr ip;
8068 struct mac_nat_entry *mne = NULL;
8069 u_short op;
8070 boolean_t translate = FALSE;
8071
8072 if (!is_valid_arp_packet(data, TRUE, &eh, &ea)) {
8073 goto done;
8074 }
8075 op = ntohs(ea->arp_op);
8076 switch (op) {
8077 case ARPOP_REQUEST:
8078 case ARPOP_REPLY:
8079 /* only care about REQUEST and REPLY */
8080 break;
8081 default:
8082 goto done;
8083 }
8084
8085 bcopy(ea->arp_spa, &ip, sizeof(ip));
8086 if (ip.s_addr == 0) {
8087 goto done;
8088 }
8089 /* XXX validate IP address: no multicast/broadcast */
8090 mne = bridge_update_mac_nat_entry(sc, bif, AF_INET, &ip, ea->arp_sha);
8091 if (mnr != NULL && mne != NULL) {
8092 /* record the offset to do the replacement */
8093 translate = TRUE;
8094 mnr->mnr_arp_offset = (char *)ea->arp_sha - (char *)eh;
8095 }
8096
8097 done:
8098 return translate;
8099 }
8100
8101 #define ETHER_IPV4_HEADER_LEN (sizeof(struct ether_header) + \
8102 + sizeof(struct ip))
8103 static struct ether_header *
8104 get_ether_ip_header(mbuf_t *data, boolean_t is_output)
8105 {
8106 struct ether_header *eh = NULL;
8107 int flags = is_output ? BR_DBGF_OUTPUT : BR_DBGF_INPUT;
8108 size_t minlen = ETHER_IPV4_HEADER_LEN;
8109
8110 if (mbuf_pkthdr_len(*data) < minlen) {
8111 BRIDGE_LOG(LOG_DEBUG, flags,
8112 "IP %s short frame %lu < %lu",
8113 get_in_out_string(is_output),
8114 mbuf_pkthdr_len(*data), minlen);
8115 goto done;
8116 }
8117 if (mbuf_len(*data) < minlen && mbuf_pullup(data, minlen) != 0) {
8118 BRIDGE_LOG(LOG_DEBUG, flags,
8119 "IP %s size %lu mbuf_pullup fail",
8120 get_in_out_string(is_output),
8121 minlen);
8122 *data = NULL;
8123 goto done;
8124 }
8125 eh = mtod(*data, struct ether_header *);
8126 done:
8127 return eh;
8128 }
8129
8130 static bool
8131 is_broadcast_ip_packet(mbuf_t *data)
8132 {
8133 struct ether_header *eh;
8134 uint16_t ether_type;
8135 bool is_broadcast = FALSE;
8136
8137 eh = mtod(*data, struct ether_header *);
8138 ether_type = ntohs(eh->ether_type);
8139 switch (ether_type) {
8140 case ETHERTYPE_IP:
8141 eh = get_ether_ip_header(data, FALSE);
8142 if (eh != NULL) {
8143 struct in_addr dst;
8144 struct ip *iphdr;
8145
8146 iphdr = (struct ip *)(void *)(eh + 1);
8147 bcopy(&iphdr->ip_dst, &dst, sizeof(dst));
8148 is_broadcast = (dst.s_addr == INADDR_BROADCAST);
8149 }
8150 break;
8151 default:
8152 break;
8153 }
8154 return is_broadcast;
8155 }
8156
8157 static struct mac_nat_entry *
8158 bridge_mac_nat_ip_input(struct bridge_softc *sc, mbuf_t *data)
8159 {
8160 struct in_addr dst;
8161 struct ether_header *eh;
8162 struct ip *iphdr;
8163 struct mac_nat_entry *mne = NULL;
8164
8165 eh = get_ether_ip_header(data, FALSE);
8166 if (eh == NULL) {
8167 goto done;
8168 }
8169 iphdr = (struct ip *)(void *)(eh + 1);
8170 bcopy(&iphdr->ip_dst, &dst, sizeof(dst));
8171 /* XXX validate IP address */
8172 if (dst.s_addr == 0) {
8173 goto done;
8174 }
8175 mne = bridge_lookup_mac_nat_entry(sc, AF_INET, &dst);
8176 done:
8177 return mne;
8178 }
8179
8180 static void
8181 bridge_mac_nat_udp_output(struct bridge_softc *sc,
8182 struct bridge_iflist *bif, mbuf_t m,
8183 uint8_t ip_header_len, struct mac_nat_record *mnr)
8184 {
8185 uint16_t dp_flags;
8186 errno_t error;
8187 size_t offset;
8188 struct udphdr udphdr;
8189
8190 /* copy the UDP header */
8191 offset = sizeof(struct ether_header) + ip_header_len;
8192 error = mbuf_copydata(m, offset, sizeof(struct udphdr), &udphdr);
8193 if (error != 0) {
8194 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8195 "mbuf_copydata udphdr failed %d",
8196 error);
8197 return;
8198 }
8199 if (ntohs(udphdr.uh_sport) != IPPORT_BOOTPC ||
8200 ntohs(udphdr.uh_dport) != IPPORT_BOOTPS) {
8201 /* not a BOOTP/DHCP packet */
8202 return;
8203 }
8204 /* check whether the broadcast bit is already set */
8205 offset += sizeof(struct udphdr) + offsetof(struct dhcp, dp_flags);
8206 error = mbuf_copydata(m, offset, sizeof(dp_flags), &dp_flags);
8207 if (error != 0) {
8208 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8209 "mbuf_copydata dp_flags failed %d",
8210 error);
8211 return;
8212 }
8213 if ((ntohs(dp_flags) & DHCP_FLAGS_BROADCAST) != 0) {
8214 /* it's already set, nothing to do */
8215 return;
8216 }
8217 /* broadcast bit needs to be set */
8218 mnr->mnr_ip_dhcp_flags = dp_flags | htons(DHCP_FLAGS_BROADCAST);
8219 mnr->mnr_ip_header_len = ip_header_len;
8220 if (udphdr.uh_sum != 0) {
8221 uint16_t delta;
8222
8223 /* adjust checksum to take modified dp_flags into account */
8224 delta = dp_flags - mnr->mnr_ip_dhcp_flags;
8225 mnr->mnr_ip_udp_csum = udphdr.uh_sum + delta;
8226 }
8227 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8228 "%s %s DHCP dp_flags 0x%x UDP cksum 0x%x",
8229 sc->sc_if_xname,
8230 bif->bif_ifp->if_xname,
8231 ntohs(mnr->mnr_ip_dhcp_flags),
8232 ntohs(mnr->mnr_ip_udp_csum));
8233 return;
8234 }
8235
8236 static boolean_t
8237 bridge_mac_nat_ip_output(struct bridge_softc *sc,
8238 struct bridge_iflist *bif, mbuf_t *data, struct mac_nat_record *mnr)
8239 {
8240 #pragma unused(mnr)
8241 struct ether_header *eh;
8242 struct in_addr ip;
8243 struct ip *iphdr;
8244 uint8_t ip_header_len;
8245 struct mac_nat_entry *mne = NULL;
8246 boolean_t translate = FALSE;
8247
8248 eh = get_ether_ip_header(data, TRUE);
8249 if (eh == NULL) {
8250 goto done;
8251 }
8252 iphdr = (struct ip *)(void *)(eh + 1);
8253 ip_header_len = IP_VHL_HL(iphdr->ip_vhl) << 2;
8254 if (ip_header_len < sizeof(ip)) {
8255 /* bogus IP header */
8256 goto done;
8257 }
8258 bcopy(&iphdr->ip_src, &ip, sizeof(ip));
8259 /* XXX validate the source address */
8260 if (ip.s_addr != 0) {
8261 mne = bridge_update_mac_nat_entry(sc, bif, AF_INET, &ip,
8262 eh->ether_shost);
8263 }
8264 if (mnr != NULL) {
8265 if (iphdr->ip_p == IPPROTO_UDP) {
8266 /* handle DHCP must broadcast */
8267 bridge_mac_nat_udp_output(sc, bif, *data,
8268 ip_header_len, mnr);
8269 }
8270 translate = TRUE;
8271 }
8272 done:
8273 return translate;
8274 }
8275
8276 #define ETHER_IPV6_HEADER_LEN (sizeof(struct ether_header) + \
8277 + sizeof(struct ip6_hdr))
8278 static struct ether_header *
8279 get_ether_ipv6_header(mbuf_t *data, size_t plen, boolean_t is_output)
8280 {
8281 struct ether_header *eh = NULL;
8282 int flags = is_output ? BR_DBGF_OUTPUT : BR_DBGF_INPUT;
8283 size_t minlen = ETHER_IPV6_HEADER_LEN + plen;
8284
8285 if (mbuf_pkthdr_len(*data) < minlen) {
8286 BRIDGE_LOG(LOG_DEBUG, flags,
8287 "IP %s short frame %lu < %lu",
8288 get_in_out_string(is_output),
8289 mbuf_pkthdr_len(*data), minlen);
8290 goto done;
8291 }
8292 if (mbuf_len(*data) < minlen && mbuf_pullup(data, minlen) != 0) {
8293 BRIDGE_LOG(LOG_DEBUG, flags,
8294 "IP %s size %lu mbuf_pullup fail",
8295 get_in_out_string(is_output),
8296 minlen);
8297 *data = NULL;
8298 goto done;
8299 }
8300 eh = mtod(*data, struct ether_header *);
8301 done:
8302 return eh;
8303 }
8304
8305 #include <netinet/icmp6.h>
8306 #include <netinet6/nd6.h>
8307
8308 #define ETHER_ND_LLADDR_LEN (ETHER_ADDR_LEN + sizeof(struct nd_opt_hdr))
8309
8310 static void
8311 bridge_mac_nat_icmpv6_output(struct bridge_softc *sc,
8312 struct bridge_iflist *bif,
8313 mbuf_t *data, struct ip6_hdr *ip6h,
8314 struct in6_addr *saddrp,
8315 struct mac_nat_record *mnr)
8316 {
8317 struct ether_header *eh;
8318 struct icmp6_hdr *icmp6;
8319 uint8_t icmp6_type;
8320 uint32_t icmp6len;
8321 int lladdrlen = 0;
8322 char *lladdr = NULL;
8323 unsigned int off = sizeof(*ip6h);
8324
8325 icmp6len = (u_int32_t)ntohs(ip6h->ip6_plen);
8326 if (icmp6len < sizeof(*icmp6)) {
8327 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
8328 "short IPv6 payload length %d < %lu",
8329 icmp6len, sizeof(*icmp6));
8330 return;
8331 }
8332
8333 /* pullup IP6 header + ICMPv6 header */
8334 eh = get_ether_ipv6_header(data, sizeof(*icmp6), TRUE);
8335 if (eh == NULL) {
8336 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
8337 "failed to pullup icmp6 header");
8338 return;
8339 }
8340 ip6h = (struct ip6_hdr *)(void *)(eh + 1);
8341 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6h + off);
8342 icmp6_type = icmp6->icmp6_type;
8343 switch (icmp6_type) {
8344 case ND_NEIGHBOR_SOLICIT:
8345 case ND_NEIGHBOR_ADVERT:
8346 case ND_ROUTER_ADVERT:
8347 case ND_ROUTER_SOLICIT:
8348 break;
8349 default:
8350 return;
8351 }
8352
8353 /* pullup IP6 header + payload */
8354 eh = get_ether_ipv6_header(data, icmp6len, TRUE);
8355 if (eh == NULL) {
8356 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
8357 "failed to pullup icmp6 + payload");
8358 return;
8359 }
8360 ip6h = (struct ip6_hdr *)(void *)(eh + 1);
8361 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6h + off);
8362 switch (icmp6_type) {
8363 case ND_NEIGHBOR_SOLICIT: {
8364 struct nd_neighbor_solicit *nd_ns;
8365 union nd_opts ndopts;
8366 boolean_t is_dad_probe;
8367 struct in6_addr taddr;
8368
8369 if (icmp6len < sizeof(*nd_ns)) {
8370 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8371 "short nd_ns %d < %lu",
8372 icmp6len, sizeof(*nd_ns));
8373 return;
8374 }
8375
8376 nd_ns = (struct nd_neighbor_solicit *)(void *)icmp6;
8377 bcopy(&nd_ns->nd_ns_target, &taddr, sizeof(taddr));
8378 if (IN6_IS_ADDR_MULTICAST(&taddr) ||
8379 IN6_IS_ADDR_UNSPECIFIED(&taddr)) {
8380 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8381 "invalid target ignored");
8382 return;
8383 }
8384 /* parse options */
8385 nd6_option_init(nd_ns + 1, icmp6len - sizeof(*nd_ns), &ndopts);
8386 if (nd6_options(&ndopts) < 0) {
8387 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8388 "invalid ND6 NS option");
8389 return;
8390 }
8391 if (ndopts.nd_opts_src_lladdr != NULL) {
8392 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
8393 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
8394 }
8395 is_dad_probe = IN6_IS_ADDR_UNSPECIFIED(saddrp);
8396 if (lladdr != NULL) {
8397 if (is_dad_probe) {
8398 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8399 "bad ND6 DAD packet");
8400 return;
8401 }
8402 if (lladdrlen != ETHER_ND_LLADDR_LEN) {
8403 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8404 "source lladdrlen %d != %lu",
8405 lladdrlen, ETHER_ND_LLADDR_LEN);
8406 return;
8407 }
8408 }
8409 if (is_dad_probe) {
8410 /* node is trying use taddr, create an mne for taddr */
8411 *saddrp = taddr;
8412 }
8413 break;
8414 }
8415 case ND_NEIGHBOR_ADVERT: {
8416 struct nd_neighbor_advert *nd_na;
8417 union nd_opts ndopts;
8418 struct in6_addr taddr;
8419
8420
8421 nd_na = (struct nd_neighbor_advert *)(void *)icmp6;
8422
8423 if (icmp6len < sizeof(*nd_na)) {
8424 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8425 "short nd_na %d < %lu",
8426 icmp6len, sizeof(*nd_na));
8427 return;
8428 }
8429
8430 bcopy(&nd_na->nd_na_target, &taddr, sizeof(taddr));
8431 if (IN6_IS_ADDR_MULTICAST(&taddr) ||
8432 IN6_IS_ADDR_UNSPECIFIED(&taddr)) {
8433 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8434 "invalid target ignored");
8435 return;
8436 }
8437 /* parse options */
8438 nd6_option_init(nd_na + 1, icmp6len - sizeof(*nd_na), &ndopts);
8439 if (nd6_options(&ndopts) < 0) {
8440 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8441 "invalid ND6 NA option");
8442 return;
8443 }
8444 if (ndopts.nd_opts_tgt_lladdr == NULL) {
8445 /* target linklayer, nothing to do */
8446 return;
8447 }
8448 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
8449 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
8450 if (lladdrlen != ETHER_ND_LLADDR_LEN) {
8451 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8452 "target lladdrlen %d != %lu",
8453 lladdrlen, ETHER_ND_LLADDR_LEN);
8454 return;
8455 }
8456 break;
8457 }
8458 case ND_ROUTER_ADVERT:
8459 case ND_ROUTER_SOLICIT: {
8460 union nd_opts ndopts;
8461 uint32_t type_length;
8462 const char *description;
8463
8464 if (icmp6_type == ND_ROUTER_ADVERT) {
8465 type_length = sizeof(struct nd_router_advert);
8466 description = "RA";
8467 } else {
8468 type_length = sizeof(struct nd_router_solicit);
8469 description = "RS";
8470 }
8471 if (icmp6len < type_length) {
8472 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8473 "short ND6 %s %d < %d",
8474 description, icmp6len, type_length);
8475 return;
8476 }
8477 /* parse options */
8478 nd6_option_init(((uint8_t *)icmp6) + type_length,
8479 icmp6len - type_length, &ndopts);
8480 if (nd6_options(&ndopts) < 0) {
8481 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8482 "invalid ND6 %s option", description);
8483 return;
8484 }
8485 if (ndopts.nd_opts_src_lladdr != NULL) {
8486 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
8487 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
8488 if (lladdrlen != ETHER_ND_LLADDR_LEN) {
8489 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8490 "source lladdrlen %d != %lu",
8491 lladdrlen, ETHER_ND_LLADDR_LEN);
8492 return;
8493 }
8494 }
8495 break;
8496 }
8497 default:
8498 break;
8499 }
8500 if (lladdr != NULL) {
8501 mnr->mnr_ip6_lladdr_offset = (uint16_t)
8502 ((uintptr_t)lladdr - (uintptr_t)eh);
8503 mnr->mnr_ip6_icmp6_len = icmp6len;
8504 mnr->mnr_ip6_icmp6_type = icmp6_type;
8505 mnr->mnr_ip6_header_len = off;
8506 if (BRIDGE_DBGF_ENABLED(BR_DBGF_MAC_NAT)) {
8507 const char *str;
8508
8509 switch (mnr->mnr_ip6_icmp6_type) {
8510 case ND_ROUTER_ADVERT:
8511 str = "ROUTER ADVERT";
8512 break;
8513 case ND_ROUTER_SOLICIT:
8514 str = "ROUTER SOLICIT";
8515 break;
8516 case ND_NEIGHBOR_ADVERT:
8517 str = "NEIGHBOR ADVERT";
8518 break;
8519 case ND_NEIGHBOR_SOLICIT:
8520 str = "NEIGHBOR SOLICIT";
8521 break;
8522 default:
8523 str = "";
8524 break;
8525 }
8526 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_MAC_NAT,
8527 "%s %s %s ip6len %d icmp6len %d lladdr offset %d",
8528 sc->sc_if_xname, bif->bif_ifp->if_xname, str,
8529 mnr->mnr_ip6_header_len,
8530 mnr->mnr_ip6_icmp6_len, mnr->mnr_ip6_lladdr_offset);
8531 }
8532 }
8533 }
8534
8535 static struct mac_nat_entry *
8536 bridge_mac_nat_ipv6_input(struct bridge_softc *sc, mbuf_t *data)
8537 {
8538 struct in6_addr dst;
8539 struct ether_header *eh;
8540 struct ip6_hdr *ip6h;
8541 struct mac_nat_entry *mne = NULL;
8542
8543 eh = get_ether_ipv6_header(data, 0, FALSE);
8544 if (eh == NULL) {
8545 goto done;
8546 }
8547 ip6h = (struct ip6_hdr *)(void *)(eh + 1);
8548 bcopy(&ip6h->ip6_dst, &dst, sizeof(dst));
8549 /* XXX validate IPv6 address */
8550 if (IN6_IS_ADDR_UNSPECIFIED(&dst)) {
8551 goto done;
8552 }
8553 mne = bridge_lookup_mac_nat_entry(sc, AF_INET6, &dst);
8554
8555 done:
8556 return mne;
8557 }
8558
8559 static boolean_t
8560 bridge_mac_nat_ipv6_output(struct bridge_softc *sc,
8561 struct bridge_iflist *bif, mbuf_t *data, struct mac_nat_record *mnr)
8562 {
8563 struct ether_header *eh;
8564 ether_addr_t ether_shost;
8565 struct ip6_hdr *ip6h;
8566 struct in6_addr saddr;
8567 boolean_t translate;
8568
8569 translate = (bif == sc->sc_mac_nat_bif) ? FALSE : TRUE;
8570 eh = get_ether_ipv6_header(data, 0, TRUE);
8571 if (eh == NULL) {
8572 translate = FALSE;
8573 goto done;
8574 }
8575 bcopy(eh->ether_shost, ðer_shost, sizeof(ether_shost));
8576 ip6h = (struct ip6_hdr *)(void *)(eh + 1);
8577 bcopy(&ip6h->ip6_src, &saddr, sizeof(saddr));
8578 if (mnr != NULL && ip6h->ip6_nxt == IPPROTO_ICMPV6) {
8579 bridge_mac_nat_icmpv6_output(sc, bif, data, ip6h, &saddr, mnr);
8580 }
8581 if (IN6_IS_ADDR_UNSPECIFIED(&saddr)) {
8582 goto done;
8583 }
8584 (void)bridge_update_mac_nat_entry(sc, bif, AF_INET6, &saddr,
8585 ether_shost.octet);
8586
8587 done:
8588 return translate;
8589 }
8590
8591 /*
8592 * bridge_mac_nat_input:
8593 * Process a packet arriving on the MAC NAT interface (sc_mac_nat_bif).
8594 * This interface is the "external" interface with respect to NAT.
8595 * The interface is only capable of receiving a single MAC address
8596 * (e.g. a Wi-Fi STA interface).
8597 *
8598 * When a packet arrives on the external interface, look up the destination
8599 * IP address in the mac_nat_entry table. If there is a match, *is_input
8600 * is set to TRUE if it's for the MAC NAT interface, otherwise *is_input
8601 * is set to FALSE and translate the MAC address if necessary.
8602 *
8603 * Returns:
8604 * The internal interface to direct the packet to, or NULL if the packet
8605 * should not be redirected.
8606 *
8607 * *data may be updated to point at a different mbuf chain, or set to NULL
8608 * if the chain was deallocated during processing.
8609 */
8610 static ifnet_t
8611 bridge_mac_nat_input(struct bridge_softc *sc, mbuf_t *data,
8612 boolean_t *is_input)
8613 {
8614 ifnet_t dst_if = NULL;
8615 struct ether_header *eh;
8616 uint16_t ether_type;
8617 boolean_t is_unicast;
8618 mbuf_t m = *data;
8619 struct mac_nat_entry *mne = NULL;
8620
8621 BRIDGE_LOCK_ASSERT_HELD(sc);
8622 *is_input = FALSE;
8623 assert(sc->sc_mac_nat_bif != NULL);
8624 is_unicast = ((m->m_flags & (M_BCAST | M_MCAST)) == 0);
8625 eh = mtod(m, struct ether_header *);
8626 ether_type = ntohs(eh->ether_type);
8627 switch (ether_type) {
8628 case ETHERTYPE_ARP:
8629 mne = bridge_mac_nat_arp_input(sc, data);
8630 break;
8631 case ETHERTYPE_IP:
8632 if (is_unicast) {
8633 mne = bridge_mac_nat_ip_input(sc, data);
8634 }
8635 break;
8636 case ETHERTYPE_IPV6:
8637 if (is_unicast) {
8638 mne = bridge_mac_nat_ipv6_input(sc, data);
8639 }
8640 break;
8641 default:
8642 break;
8643 }
8644 if (mne != NULL) {
8645 if (is_unicast) {
8646 if (m != *data) {
8647 /* it may have changed */
8648 eh = mtod(*data, struct ether_header *);
8649 }
8650 bcopy(mne->mne_mac, eh->ether_dhost,
8651 sizeof(eh->ether_dhost));
8652 }
8653 dst_if = mne->mne_bif->bif_ifp;
8654 *is_input = (mne->mne_bif == sc->sc_mac_nat_bif);
8655 }
8656 return dst_if;
8657 }
8658
8659 /*
8660 * bridge_mac_nat_output:
8661 * Process a packet destined to the MAC NAT interface (sc_mac_nat_bif)
8662 * from the interface 'bif'.
8663 *
8664 * Create a mac_nat_entry containing the source IP address and MAC address
8665 * from the packet. Populate a mac_nat_record with information detailing
8666 * how to translate the packet. Translation takes place later when
8667 * the bridge lock is no longer held.
8668 *
8669 * If 'bif' == sc_mac_nat_bif, the stack over the MAC NAT
8670 * interface is generating an output packet. No translation is required in this
8671 * case, we just record the IP address used to prevent another bif from
8672 * claiming our IP address.
8673 *
8674 * Returns:
8675 * TRUE if the packet should be translated (*mnr updated as well),
8676 * FALSE otherwise.
8677 *
8678 * *data may be updated to point at a different mbuf chain or NULL if
8679 * the chain was deallocated during processing.
8680 */
8681
8682 static boolean_t
8683 bridge_mac_nat_output(struct bridge_softc *sc,
8684 struct bridge_iflist *bif, mbuf_t *data, struct mac_nat_record *mnr)
8685 {
8686 struct ether_header *eh;
8687 uint16_t ether_type;
8688 boolean_t translate = FALSE;
8689
8690 BRIDGE_LOCK_ASSERT_HELD(sc);
8691 assert(sc->sc_mac_nat_bif != NULL);
8692
8693 eh = mtod(*data, struct ether_header *);
8694 ether_type = ntohs(eh->ether_type);
8695 if (mnr != NULL) {
8696 bzero(mnr, sizeof(*mnr));
8697 mnr->mnr_ether_type = ether_type;
8698 }
8699 switch (ether_type) {
8700 case ETHERTYPE_ARP:
8701 translate = bridge_mac_nat_arp_output(sc, bif, data, mnr);
8702 break;
8703 case ETHERTYPE_IP:
8704 translate = bridge_mac_nat_ip_output(sc, bif, data, mnr);
8705 break;
8706 case ETHERTYPE_IPV6:
8707 translate = bridge_mac_nat_ipv6_output(sc, bif, data, mnr);
8708 break;
8709 default:
8710 break;
8711 }
8712 return translate;
8713 }
8714
8715 static void
8716 bridge_mac_nat_arp_translate(mbuf_t *data, struct mac_nat_record *mnr,
8717 const caddr_t eaddr)
8718 {
8719 errno_t error;
8720
8721 if (mnr->mnr_arp_offset == 0) {
8722 return;
8723 }
8724 /* replace the source hardware address */
8725 error = mbuf_copyback(*data, mnr->mnr_arp_offset,
8726 ETHER_ADDR_LEN, eaddr,
8727 MBUF_DONTWAIT);
8728 if (error != 0) {
8729 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
8730 "mbuf_copyback failed");
8731 m_freem(*data);
8732 *data = NULL;
8733 }
8734 return;
8735 }
8736
8737 static void
8738 bridge_mac_nat_ip_translate(mbuf_t *data, struct mac_nat_record *mnr)
8739 {
8740 errno_t error;
8741 size_t offset;
8742
8743 if (mnr->mnr_ip_header_len == 0) {
8744 return;
8745 }
8746 /* update the UDP checksum */
8747 offset = sizeof(struct ether_header) + mnr->mnr_ip_header_len;
8748 error = mbuf_copyback(*data, offset + offsetof(struct udphdr, uh_sum),
8749 sizeof(mnr->mnr_ip_udp_csum),
8750 &mnr->mnr_ip_udp_csum,
8751 MBUF_DONTWAIT);
8752 if (error != 0) {
8753 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
8754 "mbuf_copyback uh_sum failed");
8755 m_freem(*data);
8756 *data = NULL;
8757 }
8758 /* update the DHCP must broadcast flag */
8759 offset += sizeof(struct udphdr);
8760 error = mbuf_copyback(*data, offset + offsetof(struct dhcp, dp_flags),
8761 sizeof(mnr->mnr_ip_dhcp_flags),
8762 &mnr->mnr_ip_dhcp_flags,
8763 MBUF_DONTWAIT);
8764 if (error != 0) {
8765 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
8766 "mbuf_copyback dp_flags failed");
8767 m_freem(*data);
8768 *data = NULL;
8769 }
8770 }
8771
8772 static void
8773 bridge_mac_nat_ipv6_translate(mbuf_t *data, struct mac_nat_record *mnr,
8774 const caddr_t eaddr)
8775 {
8776 uint16_t cksum;
8777 errno_t error;
8778 mbuf_t m = *data;
8779
8780 if (mnr->mnr_ip6_header_len == 0) {
8781 return;
8782 }
8783 switch (mnr->mnr_ip6_icmp6_type) {
8784 case ND_ROUTER_ADVERT:
8785 case ND_ROUTER_SOLICIT:
8786 case ND_NEIGHBOR_SOLICIT:
8787 case ND_NEIGHBOR_ADVERT:
8788 if (mnr->mnr_ip6_lladdr_offset == 0) {
8789 /* nothing to do */
8790 return;
8791 }
8792 break;
8793 default:
8794 return;
8795 }
8796
8797 /*
8798 * replace the lladdr
8799 */
8800 error = mbuf_copyback(m, mnr->mnr_ip6_lladdr_offset,
8801 ETHER_ADDR_LEN, eaddr,
8802 MBUF_DONTWAIT);
8803 if (error != 0) {
8804 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
8805 "mbuf_copyback lladdr failed");
8806 m_freem(m);
8807 *data = NULL;
8808 return;
8809 }
8810
8811 /*
8812 * recompute the icmp6 checksum
8813 */
8814
8815 /* skip past the ethernet header */
8816 mbuf_setdata(m, (char *)mbuf_data(m) + ETHER_HDR_LEN,
8817 mbuf_len(m) - ETHER_HDR_LEN);
8818 mbuf_pkthdr_adjustlen(m, -ETHER_HDR_LEN);
8819
8820 #define CKSUM_OFFSET_ICMP6 offsetof(struct icmp6_hdr, icmp6_cksum)
8821 /* set the checksum to zero */
8822 cksum = 0;
8823 error = mbuf_copyback(m, mnr->mnr_ip6_header_len + CKSUM_OFFSET_ICMP6,
8824 sizeof(cksum), &cksum, MBUF_DONTWAIT);
8825 if (error != 0) {
8826 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
8827 "mbuf_copyback cksum=0 failed");
8828 m_freem(m);
8829 *data = NULL;
8830 return;
8831 }
8832 /* compute and set the new checksum */
8833 cksum = in6_cksum(m, IPPROTO_ICMPV6, mnr->mnr_ip6_header_len,
8834 mnr->mnr_ip6_icmp6_len);
8835 error = mbuf_copyback(m, mnr->mnr_ip6_header_len + CKSUM_OFFSET_ICMP6,
8836 sizeof(cksum), &cksum, MBUF_DONTWAIT);
8837 if (error != 0) {
8838 BRIDGE_LOG(LOG_NOTICE, BR_DBGF_MAC_NAT,
8839 "mbuf_copyback cksum failed");
8840 m_freem(m);
8841 *data = NULL;
8842 return;
8843 }
8844 /* restore the ethernet header */
8845 mbuf_setdata(m, (char *)mbuf_data(m) - ETHER_HDR_LEN,
8846 mbuf_len(m) + ETHER_HDR_LEN);
8847 mbuf_pkthdr_adjustlen(m, ETHER_HDR_LEN);
8848 return;
8849 }
8850
8851 static void
8852 bridge_mac_nat_translate(mbuf_t *data, struct mac_nat_record *mnr,
8853 const caddr_t eaddr)
8854 {
8855 struct ether_header *eh;
8856
8857 /* replace the source ethernet address with the single MAC */
8858 eh = mtod(*data, struct ether_header *);
8859 bcopy(eaddr, eh->ether_shost, sizeof(eh->ether_shost));
8860 switch (mnr->mnr_ether_type) {
8861 case ETHERTYPE_ARP:
8862 bridge_mac_nat_arp_translate(data, mnr, eaddr);
8863 break;
8864
8865 case ETHERTYPE_IP:
8866 bridge_mac_nat_ip_translate(data, mnr);
8867 break;
8868
8869 case ETHERTYPE_IPV6:
8870 bridge_mac_nat_ipv6_translate(data, mnr, eaddr);
8871 break;
8872
8873 default:
8874 break;
8875 }
8876 return;
8877 }
8878
8879 /*
8880 * bridge packet filtering
8881 */
8882
8883 /*
8884 * Perform basic checks on header size since
8885 * pfil assumes ip_input has already processed
8886 * it for it. Cut-and-pasted from ip_input.c.
8887 * Given how simple the IPv6 version is,
8888 * does the IPv4 version really need to be
8889 * this complicated?
8890 *
8891 * XXX Should we update ipstat here, or not?
8892 * XXX Right now we update ipstat but not
8893 * XXX csum_counter.
8894 */
8895 static int
8896 bridge_ip_checkbasic(struct mbuf **mp)
8897 {
8898 struct mbuf *m = *mp;
8899 struct ip *ip;
8900 int len, hlen;
8901 u_short sum;
8902
8903 if (*mp == NULL) {
8904 return -1;
8905 }
8906
8907 if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
8908 /* max_linkhdr is already rounded up to nearest 4-byte */
8909 if ((m = m_copyup(m, sizeof(struct ip),
8910 max_linkhdr)) == NULL) {
8911 /* XXXJRT new stat, please */
8912 ipstat.ips_toosmall++;
8913 goto bad;
8914 }
8915 } else if (OS_EXPECT((size_t)m->m_len < sizeof(struct ip), 0)) {
8916 if ((m = m_pullup(m, sizeof(struct ip))) == NULL) {
8917 ipstat.ips_toosmall++;
8918 goto bad;
8919 }
8920 }
8921 ip = mtod(m, struct ip *);
8922 if (ip == NULL) {
8923 goto bad;
8924 }
8925
8926 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
8927 ipstat.ips_badvers++;
8928 goto bad;
8929 }
8930 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
8931 if (hlen < (int)sizeof(struct ip)) { /* minimum header length */
8932 ipstat.ips_badhlen++;
8933 goto bad;
8934 }
8935 if (hlen > m->m_len) {
8936 if ((m = m_pullup(m, hlen)) == 0) {
8937 ipstat.ips_badhlen++;
8938 goto bad;
8939 }
8940 ip = mtod(m, struct ip *);
8941 if (ip == NULL) {
8942 goto bad;
8943 }
8944 }
8945
8946 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
8947 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
8948 } else {
8949 if (hlen == sizeof(struct ip)) {
8950 sum = in_cksum_hdr(ip);
8951 } else {
8952 sum = in_cksum(m, hlen);
8953 }
8954 }
8955 if (sum) {
8956 ipstat.ips_badsum++;
8957 goto bad;
8958 }
8959
8960 /* Retrieve the packet length. */
8961 len = ntohs(ip->ip_len);
8962
8963 /*
8964 * Check for additional length bogosity
8965 */
8966 if (len < hlen) {
8967 ipstat.ips_badlen++;
8968 goto bad;
8969 }
8970
8971 /*
8972 * Check that the amount of data in the buffers
8973 * is as at least much as the IP header would have us expect.
8974 * Drop packet if shorter than we expect.
8975 */
8976 if (m->m_pkthdr.len < len) {
8977 ipstat.ips_tooshort++;
8978 goto bad;
8979 }
8980
8981 /* Checks out, proceed */
8982 *mp = m;
8983 return 0;
8984
8985 bad:
8986 *mp = m;
8987 return -1;
8988 }
8989
8990 /*
8991 * Same as above, but for IPv6.
8992 * Cut-and-pasted from ip6_input.c.
8993 * XXX Should we update ip6stat, or not?
8994 */
8995 static int
8996 bridge_ip6_checkbasic(struct mbuf **mp)
8997 {
8998 struct mbuf *m = *mp;
8999 struct ip6_hdr *ip6;
9000
9001 /*
9002 * If the IPv6 header is not aligned, slurp it up into a new
9003 * mbuf with space for link headers, in the event we forward
9004 * it. Otherwise, if it is aligned, make sure the entire base
9005 * IPv6 header is in the first mbuf of the chain.
9006 */
9007 if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
9008 struct ifnet *inifp = m->m_pkthdr.rcvif;
9009 /* max_linkhdr is already rounded up to nearest 4-byte */
9010 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
9011 max_linkhdr)) == NULL) {
9012 /* XXXJRT new stat, please */
9013 ip6stat.ip6s_toosmall++;
9014 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
9015 goto bad;
9016 }
9017 } else if (OS_EXPECT((size_t)m->m_len < sizeof(struct ip6_hdr), 0)) {
9018 struct ifnet *inifp = m->m_pkthdr.rcvif;
9019 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
9020 ip6stat.ip6s_toosmall++;
9021 in6_ifstat_inc(inifp, ifs6_in_hdrerr);
9022 goto bad;
9023 }
9024 }
9025
9026 ip6 = mtod(m, struct ip6_hdr *);
9027
9028 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
9029 ip6stat.ip6s_badvers++;
9030 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
9031 goto bad;
9032 }
9033
9034 /* Checks out, proceed */
9035 *mp = m;
9036 return 0;
9037
9038 bad:
9039 *mp = m;
9040 return -1;
9041 }
9042
9043 /*
9044 * the PF routines expect to be called from ip_input, so we
9045 * need to do and undo here some of the same processing.
9046 *
9047 * XXX : this is heavily inspired on bridge_pfil()
9048 */
9049 static int
9050 bridge_pf(struct mbuf **mp, struct ifnet *ifp, uint32_t sc_filter_flags,
9051 int input)
9052 {
9053 /*
9054 * XXX : mpetit : heavily inspired by bridge_pfil()
9055 */
9056
9057 int snap, error, i, hlen;
9058 struct ether_header *eh1, eh2;
9059 struct ip *ip;
9060 struct llc llc1;
9061 u_int16_t ether_type;
9062
9063 snap = 0;
9064 error = -1; /* Default error if not error == 0 */
9065
9066 if ((sc_filter_flags & IFBF_FILT_MEMBER) == 0) {
9067 return 0; /* filtering is disabled */
9068 }
9069 i = min((*mp)->m_pkthdr.len, max_protohdr);
9070 if ((*mp)->m_len < i) {
9071 *mp = m_pullup(*mp, i);
9072 if (*mp == NULL) {
9073 BRIDGE_LOG(LOG_NOTICE, 0, "m_pullup failed");
9074 return -1;
9075 }
9076 }
9077
9078 eh1 = mtod(*mp, struct ether_header *);
9079 ether_type = ntohs(eh1->ether_type);
9080
9081 /*
9082 * Check for SNAP/LLC.
9083 */
9084 if (ether_type < ETHERMTU) {
9085 struct llc *llc2 = (struct llc *)(eh1 + 1);
9086
9087 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
9088 llc2->llc_dsap == LLC_SNAP_LSAP &&
9089 llc2->llc_ssap == LLC_SNAP_LSAP &&
9090 llc2->llc_control == LLC_UI) {
9091 ether_type = htons(llc2->llc_un.type_snap.ether_type);
9092 snap = 1;
9093 }
9094 }
9095
9096 /*
9097 * If we're trying to filter bridge traffic, don't look at anything
9098 * other than IP and ARP traffic. If the filter doesn't understand
9099 * IPv6, don't allow IPv6 through the bridge either. This is lame
9100 * since if we really wanted, say, an AppleTalk filter, we are hosed,
9101 * but of course we don't have an AppleTalk filter to begin with.
9102 * (Note that since pfil doesn't understand ARP it will pass *ALL*
9103 * ARP traffic.)
9104 */
9105 switch (ether_type) {
9106 case ETHERTYPE_ARP:
9107 case ETHERTYPE_REVARP:
9108 return 0; /* Automatically pass */
9109
9110 case ETHERTYPE_IP:
9111 case ETHERTYPE_IPV6:
9112 break;
9113 default:
9114 /*
9115 * Check to see if the user wants to pass non-ip
9116 * packets, these will not be checked by pf and
9117 * passed unconditionally so the default is to drop.
9118 */
9119 if ((sc_filter_flags & IFBF_FILT_ONLYIP)) {
9120 goto bad;
9121 }
9122 break;
9123 }
9124
9125 /* Strip off the Ethernet header and keep a copy. */
9126 m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t)&eh2);
9127 m_adj(*mp, ETHER_HDR_LEN);
9128
9129 /* Strip off snap header, if present */
9130 if (snap) {
9131 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t)&llc1);
9132 m_adj(*mp, sizeof(struct llc));
9133 }
9134
9135 /*
9136 * Check the IP header for alignment and errors
9137 */
9138 switch (ether_type) {
9139 case ETHERTYPE_IP:
9140 error = bridge_ip_checkbasic(mp);
9141 break;
9142 case ETHERTYPE_IPV6:
9143 error = bridge_ip6_checkbasic(mp);
9144 break;
9145 default:
9146 error = 0;
9147 break;
9148 }
9149 if (error) {
9150 goto bad;
9151 }
9152
9153 error = 0;
9154
9155 /*
9156 * Run the packet through pf rules
9157 */
9158 switch (ether_type) {
9159 case ETHERTYPE_IP:
9160 /*
9161 * before calling the firewall, swap fields the same as
9162 * IP does. here we assume the header is contiguous
9163 */
9164 ip = mtod(*mp, struct ip *);
9165
9166 ip->ip_len = ntohs(ip->ip_len);
9167 ip->ip_off = ntohs(ip->ip_off);
9168
9169 if (ifp != NULL) {
9170 error = pf_af_hook(ifp, 0, mp, AF_INET, input, NULL);
9171 }
9172
9173 if (*mp == NULL || error != 0) { /* filter may consume */
9174 break;
9175 }
9176
9177 /* Recalculate the ip checksum and restore byte ordering */
9178 ip = mtod(*mp, struct ip *);
9179 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
9180 if (hlen < (int)sizeof(struct ip)) {
9181 goto bad;
9182 }
9183 if (hlen > (*mp)->m_len) {
9184 if ((*mp = m_pullup(*mp, hlen)) == 0) {
9185 goto bad;
9186 }
9187 ip = mtod(*mp, struct ip *);
9188 if (ip == NULL) {
9189 goto bad;
9190 }
9191 }
9192 ip->ip_len = htons(ip->ip_len);
9193 ip->ip_off = htons(ip->ip_off);
9194 ip->ip_sum = 0;
9195 if (hlen == sizeof(struct ip)) {
9196 ip->ip_sum = in_cksum_hdr(ip);
9197 } else {
9198 ip->ip_sum = in_cksum(*mp, hlen);
9199 }
9200 break;
9201
9202 case ETHERTYPE_IPV6:
9203 if (ifp != NULL) {
9204 error = pf_af_hook(ifp, 0, mp, AF_INET6, input, NULL);
9205 }
9206
9207 if (*mp == NULL || error != 0) { /* filter may consume */
9208 break;
9209 }
9210 break;
9211 default:
9212 error = 0;
9213 break;
9214 }
9215
9216 if (*mp == NULL) {
9217 return error;
9218 }
9219 if (error != 0) {
9220 goto bad;
9221 }
9222
9223 error = -1;
9224
9225 /*
9226 * Finally, put everything back the way it was and return
9227 */
9228 if (snap) {
9229 M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT, 0);
9230 if (*mp == NULL) {
9231 return error;
9232 }
9233 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
9234 }
9235
9236 M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT, 0);
9237 if (*mp == NULL) {
9238 return error;
9239 }
9240 bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
9241
9242 return 0;
9243
9244 bad:
9245 m_freem(*mp);
9246 *mp = NULL;
9247 return error;
9248 }
9249
9250 /*
9251 * Copyright (C) 2014, Stefano Garzarella - Universita` di Pisa.
9252 * All rights reserved.
9253 *
9254 * Redistribution and use in source and binary forms, with or without
9255 * modification, are permitted provided that the following conditions
9256 * are met:
9257 * 1. Redistributions of source code must retain the above copyright
9258 * notice, this list of conditions and the following disclaimer.
9259 * 2. Redistributions in binary form must reproduce the above copyright
9260 * notice, this list of conditions and the following disclaimer in the
9261 * documentation and/or other materials provided with the distribution.
9262 *
9263 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
9264 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
9265 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
9266 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
9267 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
9268 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
9269 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
9270 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
9271 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
9272 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
9273 * SUCH DAMAGE.
9274 */
9275
9276 /*
9277 * XXX-ste: Maybe this function must be moved into kern/uipc_mbuf.c
9278 *
9279 * Create a queue of packets/segments which fit the given mss + hdr_len.
9280 * m0 points to mbuf chain to be segmented.
9281 * This function splits the payload (m0-> m_pkthdr.len - hdr_len)
9282 * into segments of length MSS bytes and then copy the first hdr_len bytes
9283 * from m0 at the top of each segment.
9284 * If hdr2_buf is not NULL (hdr2_len is the buf length), it is copied
9285 * in each segment after the first hdr_len bytes
9286 *
9287 * Return the new queue with the segments on success, NULL on failure.
9288 * (the mbuf queue is freed in this case).
9289 * nsegs contains the number of segments generated.
9290 */
9291
9292 static struct mbuf *
9293 m_seg(struct mbuf *m0, int hdr_len, int mss, int *nsegs,
9294 char * hdr2_buf, int hdr2_len)
9295 {
9296 int off = 0, n, firstlen;
9297 struct mbuf **mnext, *mseg;
9298 int total_len = m0->m_pkthdr.len;
9299
9300 /*
9301 * Segmentation useless
9302 */
9303 if (total_len <= hdr_len + mss) {
9304 return m0;
9305 }
9306
9307 if (hdr2_buf == NULL || hdr2_len <= 0) {
9308 hdr2_buf = NULL;
9309 hdr2_len = 0;
9310 }
9311
9312 off = hdr_len + mss;
9313 firstlen = mss; /* first segment stored in the original mbuf */
9314
9315 mnext = &(m0->m_nextpkt); /* pointer to next packet */
9316
9317 for (n = 1; off < total_len; off += mss, n++) {
9318 struct mbuf *m;
9319 /*
9320 * Copy the header from the original packet
9321 * and create a new mbuf chain
9322 */
9323 if (MHLEN < hdr_len) {
9324 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
9325 } else {
9326 m = m_gethdr(M_NOWAIT, MT_DATA);
9327 }
9328
9329 if (m == NULL) {
9330 #ifdef GSO_DEBUG
9331 D("MGETHDR error\n");
9332 #endif
9333 goto err;
9334 }
9335
9336 m_copydata(m0, 0, hdr_len, mtod(m, caddr_t));
9337
9338 m->m_len = hdr_len;
9339 /*
9340 * if the optional header is present, copy it
9341 */
9342 if (hdr2_buf != NULL) {
9343 m_copyback(m, hdr_len, hdr2_len, hdr2_buf);
9344 }
9345
9346 m->m_flags |= (m0->m_flags & M_COPYFLAGS);
9347 if (off + mss >= total_len) { /* last segment */
9348 mss = total_len - off;
9349 }
9350 /*
9351 * Copy the payload from original packet
9352 */
9353 mseg = m_copym(m0, off, mss, M_NOWAIT);
9354 if (mseg == NULL) {
9355 m_freem(m);
9356 #ifdef GSO_DEBUG
9357 D("m_copym error\n");
9358 #endif
9359 goto err;
9360 }
9361 m_cat(m, mseg);
9362
9363 m->m_pkthdr.len = hdr_len + hdr2_len + mss;
9364 m->m_pkthdr.rcvif = m0->m_pkthdr.rcvif;
9365 /*
9366 * Copy the checksum flags and data (in_cksum() need this)
9367 */
9368 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
9369 m->m_pkthdr.csum_data = m0->m_pkthdr.csum_data;
9370 m->m_pkthdr.tso_segsz = m0->m_pkthdr.tso_segsz;
9371
9372 *mnext = m;
9373 mnext = &(m->m_nextpkt);
9374 }
9375
9376 /*
9377 * Update first segment.
9378 * If the optional header is present, is necessary
9379 * to insert it into the first segment.
9380 */
9381 if (hdr2_buf == NULL) {
9382 m_adj(m0, hdr_len + firstlen - total_len);
9383 m0->m_pkthdr.len = hdr_len + firstlen;
9384 } else {
9385 mseg = m_copym(m0, hdr_len, firstlen, M_NOWAIT);
9386 if (mseg == NULL) {
9387 #ifdef GSO_DEBUG
9388 D("m_copym error\n");
9389 #endif
9390 goto err;
9391 }
9392 m_adj(m0, hdr_len - total_len);
9393 m_copyback(m0, hdr_len, hdr2_len, hdr2_buf);
9394 m_cat(m0, mseg);
9395 m0->m_pkthdr.len = hdr_len + hdr2_len + firstlen;
9396 }
9397
9398 if (nsegs != NULL) {
9399 *nsegs = n;
9400 }
9401 return m0;
9402 err:
9403 while (m0 != NULL) {
9404 mseg = m0->m_nextpkt;
9405 m0->m_nextpkt = NULL;
9406 m_freem(m0);
9407 m0 = mseg;
9408 }
9409 return NULL;
9410 }
9411
9412 /*
9413 * Wrappers of IPv4 checksum functions
9414 */
9415 static inline void
9416 gso_ipv4_data_cksum(struct mbuf *m, struct ip *ip, int mac_hlen)
9417 {
9418 m->m_data += mac_hlen;
9419 m->m_len -= mac_hlen;
9420 m->m_pkthdr.len -= mac_hlen;
9421 #if __FreeBSD_version < 1000000
9422 ip->ip_len = ntohs(ip->ip_len); /* needed for in_delayed_cksum() */
9423 #endif
9424
9425 in_delayed_cksum(m);
9426
9427 #if __FreeBSD_version < 1000000
9428 ip->ip_len = htons(ip->ip_len);
9429 #endif
9430 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
9431 m->m_len += mac_hlen;
9432 m->m_pkthdr.len += mac_hlen;
9433 m->m_data -= mac_hlen;
9434 }
9435
9436 static inline void
9437 gso_ipv4_hdr_cksum(struct mbuf *m, struct ip *ip, int mac_hlen, int ip_hlen)
9438 {
9439 m->m_data += mac_hlen;
9440
9441 ip->ip_sum = in_cksum(m, ip_hlen);
9442
9443 m->m_pkthdr.csum_flags &= ~CSUM_IP;
9444 m->m_data -= mac_hlen;
9445 }
9446
9447 /*
9448 * Structure that contains the state during the TCP segmentation
9449 */
9450 struct gso_ip_tcp_state {
9451 void (*update)
9452 (struct gso_ip_tcp_state*, struct mbuf*);
9453 void (*internal)
9454 (struct gso_ip_tcp_state*, struct mbuf*);
9455 union iphdr hdr;
9456 struct tcphdr *tcp;
9457 int mac_hlen;
9458 int ip_hlen;
9459 int tcp_hlen;
9460 int hlen;
9461 int pay_len;
9462 int sw_csum;
9463 uint32_t tcp_seq;
9464 uint16_t ip_id;
9465 boolean_t is_tx;
9466 };
9467
9468 /*
9469 * Update the pointers to TCP and IPv4 headers
9470 */
9471 static inline void
9472 gso_ipv4_tcp_update(struct gso_ip_tcp_state *state, struct mbuf *m)
9473 {
9474 state->hdr.ip = (struct ip *)(void *)(mtod(m, uint8_t *) + state->mac_hlen);
9475 state->tcp = (struct tcphdr *)(void *)((caddr_t)(state->hdr.ip) + state->ip_hlen);
9476 state->pay_len = m->m_pkthdr.len - state->hlen;
9477 }
9478
9479 /*
9480 * Set properly the TCP and IPv4 headers
9481 */
9482 static inline void
9483 gso_ipv4_tcp_internal(struct gso_ip_tcp_state *state, struct mbuf *m)
9484 {
9485 /*
9486 * Update IP header
9487 */
9488 state->hdr.ip->ip_id = htons((state->ip_id)++);
9489 state->hdr.ip->ip_len = htons(m->m_pkthdr.len - state->mac_hlen);
9490 /*
9491 * TCP Checksum
9492 */
9493 state->tcp->th_sum = 0;
9494 state->tcp->th_sum = in_pseudo(state->hdr.ip->ip_src.s_addr,
9495 state->hdr.ip->ip_dst.s_addr,
9496 htons(state->tcp_hlen + IPPROTO_TCP + state->pay_len));
9497 /*
9498 * Checksum HW not supported (TCP)
9499 */
9500 if (state->sw_csum & CSUM_DELAY_DATA) {
9501 gso_ipv4_data_cksum(m, state->hdr.ip, state->mac_hlen);
9502 }
9503
9504 state->tcp_seq += state->pay_len;
9505 /*
9506 * IP Checksum
9507 */
9508 state->hdr.ip->ip_sum = 0;
9509 /*
9510 * Checksum HW not supported (IP)
9511 */
9512 if (state->sw_csum & CSUM_IP) {
9513 gso_ipv4_hdr_cksum(m, state->hdr.ip, state->mac_hlen, state->ip_hlen);
9514 }
9515 }
9516
9517
9518 /*
9519 * Updates the pointers to TCP and IPv6 headers
9520 */
9521 static inline void
9522 gso_ipv6_tcp_update(struct gso_ip_tcp_state *state, struct mbuf *m)
9523 {
9524 state->hdr.ip6 = (struct ip6_hdr *)(mtod(m, uint8_t *) + state->mac_hlen);
9525 state->tcp = (struct tcphdr *)(void *)((caddr_t)(state->hdr.ip6) + state->ip_hlen);
9526 state->pay_len = m->m_pkthdr.len - state->hlen;
9527 }
9528
9529 /*
9530 * Sets properly the TCP and IPv6 headers
9531 */
9532 static inline void
9533 gso_ipv6_tcp_internal(struct gso_ip_tcp_state *state, struct mbuf *m)
9534 {
9535 state->hdr.ip6->ip6_plen = htons(m->m_pkthdr.len -
9536 state->mac_hlen - state->ip_hlen);
9537 /*
9538 * TCP Checksum
9539 */
9540 state->tcp->th_sum = 0;
9541 state->tcp->th_sum = in6_pseudo(&state->hdr.ip6->ip6_src,
9542 &state->hdr.ip6->ip6_dst,
9543 htonl(state->tcp_hlen + state->pay_len + IPPROTO_TCP));
9544 /*
9545 * Checksum HW not supported (TCP)
9546 */
9547 if (state->sw_csum & CSUM_DELAY_IPV6_DATA) {
9548 (void)in6_finalize_cksum(m, state->mac_hlen, -1, -1, state->sw_csum);
9549 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_IPV6_DATA;
9550 }
9551 state->tcp_seq += state->pay_len;
9552 }
9553
9554 /*
9555 * Init the state during the TCP segmentation
9556 */
9557 static void
9558 gso_ip_tcp_init_state(struct gso_ip_tcp_state *state, struct ifnet *ifp,
9559 bool is_ipv4, int mac_hlen, int ip_hlen,
9560 void * ip_hdr, struct tcphdr * tcp_hdr)
9561 {
9562 #pragma unused(ifp)
9563
9564 state->hdr.ptr = ip_hdr;
9565 state->tcp = tcp_hdr;
9566 if (is_ipv4) {
9567 state->ip_id = ntohs(state->hdr.ip->ip_id);
9568 state->update = gso_ipv4_tcp_update;
9569 state->internal = gso_ipv4_tcp_internal;
9570 state->sw_csum = CSUM_DELAY_DATA | CSUM_IP; /* XXX */
9571 } else {
9572 state->update = gso_ipv6_tcp_update;
9573 state->internal = gso_ipv6_tcp_internal;
9574 state->sw_csum = CSUM_DELAY_IPV6_DATA; /* XXX */
9575 }
9576 state->mac_hlen = mac_hlen;
9577 state->ip_hlen = ip_hlen;
9578 state->tcp_hlen = state->tcp->th_off << 2;
9579 state->hlen = mac_hlen + ip_hlen + state->tcp_hlen;
9580 state->tcp_seq = ntohl(state->tcp->th_seq);
9581 //state->sw_csum = m->m_pkthdr.csum_flags & ~IF_HWASSIST_CSUM_FLAGS(ifp->if_hwassist);
9582 return;
9583 }
9584
9585 /*
9586 * GSO on TCP/IP (v4 or v6)
9587 *
9588 * If is_tx is TRUE, segmented packets are transmitted after they are
9589 * segmented.
9590 *
9591 * If is_tx is FALSE, the segmented packets are returned as a chain in *mp.
9592 */
9593 static int
9594 gso_ip_tcp(struct ifnet *ifp, struct mbuf **mp, struct gso_ip_tcp_state *state,
9595 boolean_t is_tx)
9596 {
9597 struct mbuf *m, *m_tx;
9598 int error = 0;
9599 int mss = 0;
9600 int nsegs = 0;
9601 struct mbuf *m0 = *mp;
9602 #ifdef GSO_STATS
9603 int total_len = m0->m_pkthdr.len;
9604 #endif /* GSO_STATS */
9605
9606 #if 1
9607 u_int reduce_mss;
9608
9609 reduce_mss = is_tx ? if_bridge_tso_reduce_mss_tx
9610 : if_bridge_tso_reduce_mss_forwarding;
9611 mss = ifp->if_mtu - state->ip_hlen - state->tcp_hlen - reduce_mss;
9612 assert(mss > 0);
9613 #else
9614 if (m0->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) {/* TSO with GSO */
9615 mss = ifp->if_hw_tsomax - state->ip_hlen - state->tcp_hlen;
9616 } else {
9617 mss = m0->m_pkthdr.tso_segsz;
9618 }
9619 #endif
9620
9621 *mp = m0 = m_seg(m0, state->hlen, mss, &nsegs, 0, 0);
9622 if (m0 == NULL) {
9623 return ENOBUFS; /* XXX ok? */
9624 }
9625 BRIDGE_LOG(LOG_DEBUG, BR_DBGF_CHECKSUM,
9626 "%s %s mss %d nsegs %d",
9627 ifp->if_xname,
9628 is_tx ? "TX" : "RX",
9629 mss, nsegs);
9630 /*
9631 * XXX-ste: can this happen?
9632 */
9633 if (m0->m_nextpkt == NULL) {
9634 #ifdef GSO_DEBUG
9635 D("only 1 segment");
9636 #endif
9637 if (is_tx) {
9638 error = bridge_transmit(ifp, m0);
9639 }
9640 return error;
9641 }
9642 #ifdef GSO_STATS
9643 GSOSTAT_SET_MAX(tcp.gsos_max_mss, mss);
9644 GSOSTAT_SET_MIN(tcp.gsos_min_mss, mss);
9645 GSOSTAT_ADD(tcp.gsos_osegments, nsegs);
9646 #endif /* GSO_STATS */
9647
9648 /* first pkt */
9649 m = m0;
9650
9651 state->update(state, m);
9652
9653 do {
9654 state->tcp->th_flags &= ~(TH_FIN | TH_PUSH);
9655
9656 state->internal(state, m);
9657 m_tx = m;
9658 m = m->m_nextpkt;
9659 if (is_tx) {
9660 m_tx->m_nextpkt = NULL;
9661 if ((error = bridge_transmit(ifp, m_tx)) != 0) {
9662 /*
9663 * XXX: If a segment can not be sent, discard the following
9664 * segments and propagate the error to the upper levels.
9665 * In this way the TCP retransmits all the initial packet.
9666 */
9667 #ifdef GSO_DEBUG
9668 D("if_transmit error\n");
9669 #endif
9670 goto err;
9671 }
9672 }
9673 state->update(state, m);
9674
9675 state->tcp->th_flags &= ~TH_CWR;
9676 state->tcp->th_seq = htonl(state->tcp_seq);
9677 } while (m->m_nextpkt);
9678
9679 /* last pkt */
9680 state->internal(state, m);
9681
9682 if (is_tx) {
9683 error = bridge_transmit(ifp, m);
9684 #ifdef GSO_DEBUG
9685 if (error) {
9686 D("last if_transmit error\n");
9687 D("error - type = %d \n", error);
9688 }
9689 #endif
9690 }
9691 #ifdef GSO_STATS
9692 if (!error) {
9693 GSOSTAT_INC(tcp.gsos_segmented);
9694 GSOSTAT_SET_MAX(tcp.gsos_maxsegmented, total_len);
9695 GSOSTAT_SET_MIN(tcp.gsos_minsegmented, total_len);
9696 GSOSTAT_ADD(tcp.gsos_totalbyteseg, total_len);
9697 }
9698 #endif /* GSO_STATS */
9699 return error;
9700
9701 err:
9702 #ifdef GSO_DEBUG
9703 D("error - type = %d \n", error);
9704 #endif
9705 while (m != NULL) {
9706 m_tx = m->m_nextpkt;
9707 m->m_nextpkt = NULL;
9708 m_freem(m);
9709 m = m_tx;
9710 }
9711 return error;
9712 }
9713
9714 /*
9715 * GSO for TCP/IPv[46]
9716 */
9717 static int
9718 gso_tcp(struct ifnet *ifp, struct mbuf **mp, u_int mac_hlen, bool is_ipv4,
9719 boolean_t is_tx)
9720 {
9721 int error;
9722 ip_packet_info info;
9723 uint32_t csum_flags;
9724 struct gso_ip_tcp_state state;
9725 struct bripstats stats; /* XXX ignored */
9726 struct tcphdr *tcp;
9727
9728 if (!is_tx && ipforwarding == 0) {
9729 /* no need to segment if the packet will not be forwarded */
9730 return 0;
9731 }
9732 error = bridge_get_tcp_header(mp, mac_hlen, is_ipv4, &info, &stats);
9733 if (error != 0) {
9734 if (*mp != NULL) {
9735 m_freem(*mp);
9736 *mp = NULL;
9737 }
9738 return error;
9739 }
9740 if (info.ip_proto_hdr == NULL) {
9741 /* not a TCP packet */
9742 return 0;
9743 }
9744 tcp = (struct tcphdr *)(void *)info.ip_proto_hdr;
9745 gso_ip_tcp_init_state(&state, ifp, is_ipv4, mac_hlen,
9746 info.ip_hlen, info.ip_hdr.ptr, tcp);
9747 if (is_ipv4) {
9748 csum_flags = CSUM_DELAY_DATA; /* XXX */
9749 if (!is_tx) {
9750 /* if RX to our local IP address, don't segment */
9751 struct in_addr dst_ip;
9752
9753 bcopy(&state.hdr.ip->ip_dst, &dst_ip, sizeof(dst_ip));
9754 if (in_addr_is_ours(dst_ip)) {
9755 return 0;
9756 }
9757 }
9758 } else {
9759 csum_flags = CSUM_DELAY_IPV6_DATA; /* XXX */
9760 if (!is_tx) {
9761 /* if RX to our local IP address, don't segment */
9762 if (in6_addr_is_ours(&state.hdr.ip6->ip6_dst,
9763 ifp->if_index)) {
9764 /* local IP address, no need to segment */
9765 return 0;
9766 }
9767 }
9768 }
9769 (*mp)->m_pkthdr.csum_flags = csum_flags;
9770 (*mp)->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum);
9771 return gso_ip_tcp(ifp, mp, &state, is_tx);
9772 }
9773