Home | History | Annotate | Line # | Download | only in net
if_vlan.c revision 1.135
      1 /*	$NetBSD: if_vlan.c,v 1.135 2019/04/26 11:51:56 pgoyette Exp $	*/
      2 
      3 /*
      4  * Copyright (c) 2000, 2001 The NetBSD Foundation, Inc.
      5  * All rights reserved.
      6  *
      7  * This code is derived from software contributed to The NetBSD Foundation
      8  * by Andrew Doran, and by Jason R. Thorpe of Zembu Labs, Inc.
      9  *
     10  * Redistribution and use in source and binary forms, with or without
     11  * modification, are permitted provided that the following conditions
     12  * are met:
     13  * 1. Redistributions of source code must retain the above copyright
     14  *    notice, this list of conditions and the following disclaimer.
     15  * 2. Redistributions in binary form must reproduce the above copyright
     16  *    notice, this list of conditions and the following disclaimer in the
     17  *    documentation and/or other materials provided with the distribution.
     18  *
     19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     29  * POSSIBILITY OF SUCH DAMAGE.
     30  */
     31 
     32 /*
     33  * Copyright 1998 Massachusetts Institute of Technology
     34  *
     35  * Permission to use, copy, modify, and distribute this software and
     36  * its documentation for any purpose and without fee is hereby
     37  * granted, provided that both the above copyright notice and this
     38  * permission notice appear in all copies, that both the above
     39  * copyright notice and this permission notice appear in all
     40  * supporting documentation, and that the name of M.I.T. not be used
     41  * in advertising or publicity pertaining to distribution of the
     42  * software without specific, written prior permission.  M.I.T. makes
     43  * no representations about the suitability of this software for any
     44  * purpose.  It is provided "as is" without express or implied
     45  * warranty.
     46  *
     47  * THIS SOFTWARE IS PROVIDED BY M.I.T. ``AS IS''.  M.I.T. DISCLAIMS
     48  * ALL EXPRESS OR IMPLIED WARRANTIES WITH REGARD TO THIS SOFTWARE,
     49  * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
     50  * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. IN NO EVENT
     51  * SHALL M.I.T. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
     52  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
     53  * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF
     54  * USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
     55  * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
     56  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
     57  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     58  * SUCH DAMAGE.
     59  *
     60  * from FreeBSD: if_vlan.c,v 1.16 2000/03/26 15:21:40 charnier Exp
     61  * via OpenBSD: if_vlan.c,v 1.4 2000/05/15 19:15:00 chris Exp
     62  */
     63 
     64 /*
     65  * if_vlan.c - pseudo-device driver for IEEE 802.1Q virtual LANs.  Might be
     66  * extended some day to also handle IEEE 802.1P priority tagging.  This is
     67  * sort of sneaky in the implementation, since we need to pretend to be
     68  * enough of an Ethernet implementation to make ARP work.  The way we do
     69  * this is by telling everyone that we are an Ethernet interface, and then
     70  * catch the packets that ether_output() left on our output queue when it
     71  * calls if_start(), rewrite them for use by the real outgoing interface,
     72  * and ask it to send them.
     73  *
     74  * TODO:
     75  *
     76  *	- Need some way to notify vlan interfaces when the parent
     77  *	  interface changes MTU.
     78  */
     79 
     80 #include <sys/cdefs.h>
     81 __KERNEL_RCSID(0, "$NetBSD: if_vlan.c,v 1.135 2019/04/26 11:51:56 pgoyette Exp $");
     82 
     83 #ifdef _KERNEL_OPT
     84 #include "opt_inet.h"
     85 #include "opt_net_mpsafe.h"
     86 #endif
     87 
     88 #include <sys/param.h>
     89 #include <sys/systm.h>
     90 #include <sys/kernel.h>
     91 #include <sys/mbuf.h>
     92 #include <sys/queue.h>
     93 #include <sys/socket.h>
     94 #include <sys/sockio.h>
     95 #include <sys/systm.h>
     96 #include <sys/proc.h>
     97 #include <sys/kauth.h>
     98 #include <sys/mutex.h>
     99 #include <sys/kmem.h>
    100 #include <sys/cpu.h>
    101 #include <sys/pserialize.h>
    102 #include <sys/psref.h>
    103 #include <sys/pslist.h>
    104 #include <sys/atomic.h>
    105 #include <sys/device.h>
    106 #include <sys/module.h>
    107 
    108 #include <net/bpf.h>
    109 #include <net/if.h>
    110 #include <net/if_dl.h>
    111 #include <net/if_types.h>
    112 #include <net/if_ether.h>
    113 #include <net/if_vlanvar.h>
    114 
    115 #ifdef INET
    116 #include <netinet/in.h>
    117 #include <netinet/if_inarp.h>
    118 #endif
    119 #ifdef INET6
    120 #include <netinet6/in6_ifattach.h>
    121 #include <netinet6/in6_var.h>
    122 #endif
    123 
    124 #include "ioconf.h"
    125 
    126 struct vlan_mc_entry {
    127 	LIST_ENTRY(vlan_mc_entry)	mc_entries;
    128 	/*
    129 	 * A key to identify this entry.  The mc_addr below can't be
    130 	 * used since multiple sockaddr may mapped into the same
    131 	 * ether_multi (e.g., AF_UNSPEC).
    132 	 */
    133 	union {
    134 		struct ether_multi	*mcu_enm;
    135 	} mc_u;
    136 	struct sockaddr_storage		mc_addr;
    137 };
    138 
    139 #define	mc_enm		mc_u.mcu_enm
    140 
    141 
    142 struct ifvlan_linkmib {
    143 	struct ifvlan *ifvm_ifvlan;
    144 	const struct vlan_multisw *ifvm_msw;
    145 	int	ifvm_encaplen;	/* encapsulation length */
    146 	int	ifvm_mtufudge;	/* MTU fudged by this much */
    147 	int	ifvm_mintu;	/* min transmission unit */
    148 	uint16_t ifvm_proto;	/* encapsulation ethertype */
    149 	uint16_t ifvm_tag;	/* tag to apply on packets */
    150 	struct ifnet *ifvm_p;		/* parent interface of this vlan */
    151 
    152 	struct psref_target ifvm_psref;
    153 };
    154 
    155 struct ifvlan {
    156 	union {
    157 		struct ethercom ifvu_ec;
    158 	} ifv_u;
    159 	struct ifvlan_linkmib *ifv_mib;	/*
    160 					 * reader must use vlan_getref_linkmib()
    161 					 * instead of direct dereference
    162 					 */
    163 	kmutex_t ifv_lock;		/* writer lock for ifv_mib */
    164 	pserialize_t ifv_psz;
    165 
    166 	LIST_HEAD(__vlan_mchead, vlan_mc_entry) ifv_mc_listhead;
    167 	LIST_ENTRY(ifvlan) ifv_list;
    168 	struct pslist_entry ifv_hash;
    169 	int ifv_flags;
    170 };
    171 
    172 #define	IFVF_PROMISC	0x01		/* promiscuous mode enabled */
    173 
    174 #define	ifv_ec		ifv_u.ifvu_ec
    175 
    176 #define	ifv_if		ifv_ec.ec_if
    177 
    178 #define	ifv_msw		ifv_mib.ifvm_msw
    179 #define	ifv_encaplen	ifv_mib.ifvm_encaplen
    180 #define	ifv_mtufudge	ifv_mib.ifvm_mtufudge
    181 #define	ifv_mintu	ifv_mib.ifvm_mintu
    182 #define	ifv_tag		ifv_mib.ifvm_tag
    183 
    184 struct vlan_multisw {
    185 	int	(*vmsw_addmulti)(struct ifvlan *, struct ifreq *);
    186 	int	(*vmsw_delmulti)(struct ifvlan *, struct ifreq *);
    187 	void	(*vmsw_purgemulti)(struct ifvlan *);
    188 };
    189 
    190 static int	vlan_ether_addmulti(struct ifvlan *, struct ifreq *);
    191 static int	vlan_ether_delmulti(struct ifvlan *, struct ifreq *);
    192 static void	vlan_ether_purgemulti(struct ifvlan *);
    193 
    194 const struct vlan_multisw vlan_ether_multisw = {
    195 	.vmsw_addmulti = vlan_ether_addmulti,
    196 	.vmsw_delmulti = vlan_ether_delmulti,
    197 	.vmsw_purgemulti = vlan_ether_purgemulti,
    198 };
    199 
    200 static int	vlan_clone_create(struct if_clone *, int);
    201 static int	vlan_clone_destroy(struct ifnet *);
    202 static int	vlan_config(struct ifvlan *, struct ifnet *,
    203     uint16_t);
    204 static int	vlan_ioctl(struct ifnet *, u_long, void *);
    205 static void	vlan_start(struct ifnet *);
    206 static int	vlan_transmit(struct ifnet *, struct mbuf *);
    207 static void	vlan_unconfig(struct ifnet *);
    208 static int	vlan_unconfig_locked(struct ifvlan *,
    209     struct ifvlan_linkmib *);
    210 static void	vlan_hash_init(void);
    211 static int	vlan_hash_fini(void);
    212 static int	vlan_tag_hash(uint16_t, u_long);
    213 static struct ifvlan_linkmib*	vlan_getref_linkmib(struct ifvlan *,
    214     struct psref *);
    215 static void	vlan_putref_linkmib(struct ifvlan_linkmib *,
    216     struct psref *);
    217 static void	vlan_linkmib_update(struct ifvlan *,
    218     struct ifvlan_linkmib *);
    219 static struct ifvlan_linkmib*	vlan_lookup_tag_psref(struct ifnet *,
    220     uint16_t, struct psref *);
    221 
    222 LIST_HEAD(vlan_ifvlist, ifvlan);
    223 static struct {
    224 	kmutex_t lock;
    225 	struct vlan_ifvlist list;
    226 } ifv_list __cacheline_aligned;
    227 
    228 
    229 #if !defined(VLAN_TAG_HASH_SIZE)
    230 #define VLAN_TAG_HASH_SIZE 32
    231 #endif
    232 static struct {
    233 	kmutex_t lock;
    234 	struct pslist_head *lists;
    235 	u_long mask;
    236 } ifv_hash __cacheline_aligned = {
    237 	.lists = NULL,
    238 	.mask = 0,
    239 };
    240 
    241 pserialize_t vlan_psz __read_mostly;
    242 static struct psref_class *ifvm_psref_class __read_mostly;
    243 
    244 struct if_clone vlan_cloner =
    245     IF_CLONE_INITIALIZER("vlan", vlan_clone_create, vlan_clone_destroy);
    246 
    247 /* Used to pad ethernet frames with < ETHER_MIN_LEN bytes */
    248 static char vlan_zero_pad_buff[ETHER_MIN_LEN];
    249 
    250 static inline int
    251 vlan_safe_ifpromisc(struct ifnet *ifp, int pswitch)
    252 {
    253 	int e;
    254 
    255 	KERNEL_LOCK_UNLESS_NET_MPSAFE();
    256 	e = ifpromisc(ifp, pswitch);
    257 	KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
    258 
    259 	return e;
    260 }
    261 
    262 static inline int
    263 vlan_safe_ifpromisc_locked(struct ifnet *ifp, int pswitch)
    264 {
    265 	int e;
    266 
    267 	KERNEL_LOCK_UNLESS_NET_MPSAFE();
    268 	e = ifpromisc_locked(ifp, pswitch);
    269 	KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
    270 
    271 	return e;
    272 }
    273 
    274 void
    275 vlanattach(int n)
    276 {
    277 
    278 	/*
    279 	 * Nothing to do here, initialization is handled by the
    280 	 * module initialization code in vlaninit() below.
    281 	 */
    282 }
    283 
    284 static void
    285 vlaninit(void)
    286 {
    287 	mutex_init(&ifv_list.lock, MUTEX_DEFAULT, IPL_NONE);
    288 	LIST_INIT(&ifv_list.list);
    289 
    290 	mutex_init(&ifv_hash.lock, MUTEX_DEFAULT, IPL_NONE);
    291 	vlan_psz = pserialize_create();
    292 	ifvm_psref_class = psref_class_create("vlanlinkmib", IPL_SOFTNET);
    293 	if_clone_attach(&vlan_cloner);
    294 
    295 	vlan_hash_init();
    296 	MODULE_HOOK_SET(if_vlan_vlan_input_hook, "vlan_inp", vlan_input);
    297 }
    298 
    299 static int
    300 vlandetach(void)
    301 {
    302 	bool is_empty;
    303 	int error;
    304 
    305 	mutex_enter(&ifv_list.lock);
    306 	is_empty = LIST_EMPTY(&ifv_list.list);
    307 	mutex_exit(&ifv_list.lock);
    308 
    309 	if (!is_empty)
    310 		return EBUSY;
    311 
    312 	error = vlan_hash_fini();
    313 	if (error != 0)
    314 		return error;
    315 
    316 	if_clone_detach(&vlan_cloner);
    317 	psref_class_destroy(ifvm_psref_class);
    318 	pserialize_destroy(vlan_psz);
    319 	mutex_destroy(&ifv_hash.lock);
    320 	mutex_destroy(&ifv_list.lock);
    321 
    322 	MODULE_HOOK_UNSET(if_vlan_vlan_input_hook);
    323 	return 0;
    324 }
    325 
    326 static void
    327 vlan_reset_linkname(struct ifnet *ifp)
    328 {
    329 
    330 	/*
    331 	 * We start out with a "802.1Q VLAN" type and zero-length
    332 	 * addresses.  When we attach to a parent interface, we
    333 	 * inherit its type, address length, address, and data link
    334 	 * type.
    335 	 */
    336 
    337 	ifp->if_type = IFT_L2VLAN;
    338 	ifp->if_addrlen = 0;
    339 	ifp->if_dlt = DLT_NULL;
    340 	if_alloc_sadl(ifp);
    341 }
    342 
    343 static int
    344 vlan_clone_create(struct if_clone *ifc, int unit)
    345 {
    346 	struct ifvlan *ifv;
    347 	struct ifnet *ifp;
    348 	struct ifvlan_linkmib *mib;
    349 	int rv;
    350 
    351 	ifv = malloc(sizeof(struct ifvlan), M_DEVBUF, M_WAITOK|M_ZERO);
    352 	mib = kmem_zalloc(sizeof(struct ifvlan_linkmib), KM_SLEEP);
    353 	ifp = &ifv->ifv_if;
    354 	LIST_INIT(&ifv->ifv_mc_listhead);
    355 
    356 	mib->ifvm_ifvlan = ifv;
    357 	mib->ifvm_p = NULL;
    358 	psref_target_init(&mib->ifvm_psref, ifvm_psref_class);
    359 
    360 	mutex_init(&ifv->ifv_lock, MUTEX_DEFAULT, IPL_NONE);
    361 	ifv->ifv_psz = pserialize_create();
    362 	ifv->ifv_mib = mib;
    363 
    364 	mutex_enter(&ifv_list.lock);
    365 	LIST_INSERT_HEAD(&ifv_list.list, ifv, ifv_list);
    366 	mutex_exit(&ifv_list.lock);
    367 
    368 	if_initname(ifp, ifc->ifc_name, unit);
    369 	ifp->if_softc = ifv;
    370 	ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
    371 	ifp->if_extflags = IFEF_NO_LINK_STATE_CHANGE;
    372 #ifdef NET_MPSAFE
    373 	ifp->if_extflags |= IFEF_MPSAFE;
    374 #endif
    375 	ifp->if_start = vlan_start;
    376 	ifp->if_transmit = vlan_transmit;
    377 	ifp->if_ioctl = vlan_ioctl;
    378 	IFQ_SET_READY(&ifp->if_snd);
    379 
    380 	rv = if_initialize(ifp);
    381 	if (rv != 0) {
    382 		aprint_error("%s: if_initialize failed(%d)\n", ifp->if_xname,
    383 		    rv);
    384 		goto fail;
    385 	}
    386 
    387 	vlan_reset_linkname(ifp);
    388 	if_register(ifp);
    389 	return 0;
    390 
    391 fail:
    392 	mutex_enter(&ifv_list.lock);
    393 	LIST_REMOVE(ifv, ifv_list);
    394 	mutex_exit(&ifv_list.lock);
    395 
    396 	mutex_destroy(&ifv->ifv_lock);
    397 	psref_target_destroy(&ifv->ifv_mib->ifvm_psref, ifvm_psref_class);
    398 	kmem_free(ifv->ifv_mib, sizeof(struct ifvlan_linkmib));
    399 	free(ifv, M_DEVBUF);
    400 
    401 	return rv;
    402 }
    403 
    404 static int
    405 vlan_clone_destroy(struct ifnet *ifp)
    406 {
    407 	struct ifvlan *ifv = ifp->if_softc;
    408 
    409 	mutex_enter(&ifv_list.lock);
    410 	LIST_REMOVE(ifv, ifv_list);
    411 	mutex_exit(&ifv_list.lock);
    412 
    413 	IFNET_LOCK(ifp);
    414 	vlan_unconfig(ifp);
    415 	IFNET_UNLOCK(ifp);
    416 	if_detach(ifp);
    417 
    418 	psref_target_destroy(&ifv->ifv_mib->ifvm_psref, ifvm_psref_class);
    419 	kmem_free(ifv->ifv_mib, sizeof(struct ifvlan_linkmib));
    420 	pserialize_destroy(ifv->ifv_psz);
    421 	mutex_destroy(&ifv->ifv_lock);
    422 	free(ifv, M_DEVBUF);
    423 
    424 	return 0;
    425 }
    426 
    427 /*
    428  * Configure a VLAN interface.
    429  */
    430 static int
    431 vlan_config(struct ifvlan *ifv, struct ifnet *p, uint16_t tag)
    432 {
    433 	struct ifnet *ifp = &ifv->ifv_if;
    434 	struct ifvlan_linkmib *nmib = NULL;
    435 	struct ifvlan_linkmib *omib = NULL;
    436 	struct ifvlan_linkmib *checkmib;
    437 	struct psref_target *nmib_psref = NULL;
    438 	const uint16_t vid = EVL_VLANOFTAG(tag);
    439 	int error = 0;
    440 	int idx;
    441 	bool omib_cleanup = false;
    442 	struct psref psref;
    443 
    444 	/* VLAN ID 0 and 4095 are reserved in the spec */
    445 	if ((vid == 0) || (vid == 0xfff))
    446 		return EINVAL;
    447 
    448 	nmib = kmem_alloc(sizeof(*nmib), KM_SLEEP);
    449 	mutex_enter(&ifv->ifv_lock);
    450 	omib = ifv->ifv_mib;
    451 
    452 	if (omib->ifvm_p != NULL) {
    453 		error = EBUSY;
    454 		goto done;
    455 	}
    456 
    457 	/* Duplicate check */
    458 	checkmib = vlan_lookup_tag_psref(p, vid, &psref);
    459 	if (checkmib != NULL) {
    460 		vlan_putref_linkmib(checkmib, &psref);
    461 		error = EEXIST;
    462 		goto done;
    463 	}
    464 
    465 	*nmib = *omib;
    466 	nmib_psref = &nmib->ifvm_psref;
    467 
    468 	psref_target_init(nmib_psref, ifvm_psref_class);
    469 
    470 	switch (p->if_type) {
    471 	case IFT_ETHER:
    472 	    {
    473 		struct ethercom *ec = (void *)p;
    474 		nmib->ifvm_msw = &vlan_ether_multisw;
    475 		nmib->ifvm_encaplen = ETHER_VLAN_ENCAP_LEN;
    476 		nmib->ifvm_mintu = ETHERMIN;
    477 
    478 		if (ec->ec_nvlans++ == 0) {
    479 			IFNET_LOCK(p);
    480 			error = ether_enable_vlan_mtu(p);
    481 			IFNET_UNLOCK(p);
    482 			if (error >= 0) {
    483 				if (error) {
    484 					ec->ec_nvlans--;
    485 					goto done;
    486 				}
    487 				nmib->ifvm_mtufudge = 0;
    488 			} else {
    489 				/*
    490 				 * Fudge the MTU by the encapsulation size. This
    491 				 * makes us incompatible with strictly compliant
    492 				 * 802.1Q implementations, but allows us to use
    493 				 * the feature with other NetBSD
    494 				 * implementations, which might still be useful.
    495 				 */
    496 				nmib->ifvm_mtufudge = nmib->ifvm_encaplen;
    497 			}
    498 			error = 0;
    499 		}
    500 
    501 		/*
    502 		 * If the parent interface can do hardware-assisted
    503 		 * VLAN encapsulation, then propagate its hardware-
    504 		 * assisted checksumming flags and tcp segmentation
    505 		 * offload.
    506 		 */
    507 		if (ec->ec_capabilities & ETHERCAP_VLAN_HWTAGGING) {
    508 			ec->ec_capenable |= ETHERCAP_VLAN_HWTAGGING;
    509 			ifp->if_capabilities = p->if_capabilities &
    510 			    (IFCAP_TSOv4 | IFCAP_TSOv6 |
    511 			     IFCAP_CSUM_IPv4_Tx|IFCAP_CSUM_IPv4_Rx|
    512 			     IFCAP_CSUM_TCPv4_Tx|IFCAP_CSUM_TCPv4_Rx|
    513 			     IFCAP_CSUM_UDPv4_Tx|IFCAP_CSUM_UDPv4_Rx|
    514 			     IFCAP_CSUM_TCPv6_Tx|IFCAP_CSUM_TCPv6_Rx|
    515 			     IFCAP_CSUM_UDPv6_Tx|IFCAP_CSUM_UDPv6_Rx);
    516 		}
    517 
    518 		/*
    519 		 * We inherit the parent's Ethernet address.
    520 		 */
    521 		ether_ifattach(ifp, CLLADDR(p->if_sadl));
    522 		ifp->if_hdrlen = sizeof(struct ether_vlan_header); /* XXX? */
    523 		break;
    524 	    }
    525 
    526 	default:
    527 		error = EPROTONOSUPPORT;
    528 		goto done;
    529 	}
    530 
    531 	nmib->ifvm_p = p;
    532 	nmib->ifvm_tag = vid;
    533 	ifv->ifv_if.if_mtu = p->if_mtu - nmib->ifvm_mtufudge;
    534 	ifv->ifv_if.if_flags = p->if_flags &
    535 	    (IFF_UP | IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST);
    536 
    537 	/*
    538 	 * Inherit the if_type from the parent.  This allows us
    539 	 * to participate in bridges of that type.
    540 	 */
    541 	ifv->ifv_if.if_type = p->if_type;
    542 
    543 	PSLIST_ENTRY_INIT(ifv, ifv_hash);
    544 	idx = vlan_tag_hash(vid, ifv_hash.mask);
    545 
    546 	mutex_enter(&ifv_hash.lock);
    547 	PSLIST_WRITER_INSERT_HEAD(&ifv_hash.lists[idx], ifv, ifv_hash);
    548 	mutex_exit(&ifv_hash.lock);
    549 
    550 	vlan_linkmib_update(ifv, nmib);
    551 	nmib = NULL;
    552 	nmib_psref = NULL;
    553 	omib_cleanup = true;
    554 
    555 done:
    556 	mutex_exit(&ifv->ifv_lock);
    557 
    558 	if (nmib_psref)
    559 		psref_target_destroy(nmib_psref, ifvm_psref_class);
    560 	if (nmib)
    561 		kmem_free(nmib, sizeof(*nmib));
    562 	if (omib_cleanup)
    563 		kmem_free(omib, sizeof(*omib));
    564 
    565 	return error;
    566 }
    567 
    568 /*
    569  * Unconfigure a VLAN interface.
    570  */
    571 static void
    572 vlan_unconfig(struct ifnet *ifp)
    573 {
    574 	struct ifvlan *ifv = ifp->if_softc;
    575 	struct ifvlan_linkmib *nmib = NULL;
    576 	int error;
    577 
    578 	KASSERT(IFNET_LOCKED(ifp));
    579 
    580 	nmib = kmem_alloc(sizeof(*nmib), KM_SLEEP);
    581 
    582 	mutex_enter(&ifv->ifv_lock);
    583 	error = vlan_unconfig_locked(ifv, nmib);
    584 	mutex_exit(&ifv->ifv_lock);
    585 
    586 	if (error)
    587 		kmem_free(nmib, sizeof(*nmib));
    588 }
    589 static int
    590 vlan_unconfig_locked(struct ifvlan *ifv, struct ifvlan_linkmib *nmib)
    591 {
    592 	struct ifnet *p;
    593 	struct ifnet *ifp = &ifv->ifv_if;
    594 	struct psref_target *nmib_psref = NULL;
    595 	struct ifvlan_linkmib *omib;
    596 	int error = 0;
    597 
    598 	KASSERT(IFNET_LOCKED(ifp));
    599 	KASSERT(mutex_owned(&ifv->ifv_lock));
    600 
    601 	ifp->if_flags &= ~(IFF_UP|IFF_RUNNING);
    602 
    603 	omib = ifv->ifv_mib;
    604 	p = omib->ifvm_p;
    605 
    606 	if (p == NULL) {
    607 		error = -1;
    608 		goto done;
    609 	}
    610 
    611 	*nmib = *omib;
    612 	nmib_psref = &nmib->ifvm_psref;
    613 	psref_target_init(nmib_psref, ifvm_psref_class);
    614 
    615 	/*
    616  	 * Since the interface is being unconfigured, we need to empty the
    617 	 * list of multicast groups that we may have joined while we were
    618 	 * alive and remove them from the parent's list also.
    619 	 */
    620 	(*nmib->ifvm_msw->vmsw_purgemulti)(ifv);
    621 
    622 	/* Disconnect from parent. */
    623 	switch (p->if_type) {
    624 	case IFT_ETHER:
    625 	    {
    626 		struct ethercom *ec = (void *)p;
    627 		if (--ec->ec_nvlans == 0) {
    628 			IFNET_LOCK(p);
    629 			(void) ether_disable_vlan_mtu(p);
    630 			IFNET_UNLOCK(p);
    631 		}
    632 
    633 		/* XXX ether_ifdetach must not be called with IFNET_LOCK */
    634 		mutex_exit(&ifv->ifv_lock);
    635 		IFNET_UNLOCK(ifp);
    636 		ether_ifdetach(ifp);
    637 		IFNET_LOCK(ifp);
    638 		mutex_enter(&ifv->ifv_lock);
    639 
    640 		/* if_free_sadl must be called with IFNET_LOCK */
    641 		if_free_sadl(ifp, 1);
    642 
    643 		/* Restore vlan_ioctl overwritten by ether_ifdetach */
    644 		ifp->if_ioctl = vlan_ioctl;
    645 		vlan_reset_linkname(ifp);
    646 		break;
    647 	    }
    648 
    649 	default:
    650 		panic("%s: impossible", __func__);
    651 	}
    652 
    653 	nmib->ifvm_p = NULL;
    654 	ifv->ifv_if.if_mtu = 0;
    655 	ifv->ifv_flags = 0;
    656 
    657 	mutex_enter(&ifv_hash.lock);
    658 	PSLIST_WRITER_REMOVE(ifv, ifv_hash);
    659 	pserialize_perform(vlan_psz);
    660 	mutex_exit(&ifv_hash.lock);
    661 	PSLIST_ENTRY_DESTROY(ifv, ifv_hash);
    662 
    663 	vlan_linkmib_update(ifv, nmib);
    664 
    665 	mutex_exit(&ifv->ifv_lock);
    666 
    667 	nmib_psref = NULL;
    668 	kmem_free(omib, sizeof(*omib));
    669 
    670 #ifdef INET6
    671 	KERNEL_LOCK_UNLESS_NET_MPSAFE();
    672 	/* To delete v6 link local addresses */
    673 	if (in6_present)
    674 		in6_ifdetach(ifp);
    675 	KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
    676 #endif
    677 
    678 	if ((ifp->if_flags & IFF_PROMISC) != 0)
    679 		vlan_safe_ifpromisc_locked(ifp, 0);
    680 	if_down_locked(ifp);
    681 	ifp->if_capabilities = 0;
    682 	mutex_enter(&ifv->ifv_lock);
    683 done:
    684 
    685 	if (nmib_psref)
    686 		psref_target_destroy(nmib_psref, ifvm_psref_class);
    687 
    688 	return error;
    689 }
    690 
    691 static void
    692 vlan_hash_init(void)
    693 {
    694 
    695 	ifv_hash.lists = hashinit(VLAN_TAG_HASH_SIZE, HASH_PSLIST, true,
    696 	    &ifv_hash.mask);
    697 }
    698 
    699 static int
    700 vlan_hash_fini(void)
    701 {
    702 	int i;
    703 
    704 	mutex_enter(&ifv_hash.lock);
    705 
    706 	for (i = 0; i < ifv_hash.mask + 1; i++) {
    707 		if (PSLIST_WRITER_FIRST(&ifv_hash.lists[i], struct ifvlan,
    708 		    ifv_hash) != NULL) {
    709 			mutex_exit(&ifv_hash.lock);
    710 			return EBUSY;
    711 		}
    712 	}
    713 
    714 	for (i = 0; i < ifv_hash.mask + 1; i++)
    715 		PSLIST_DESTROY(&ifv_hash.lists[i]);
    716 
    717 	mutex_exit(&ifv_hash.lock);
    718 
    719 	hashdone(ifv_hash.lists, HASH_PSLIST, ifv_hash.mask);
    720 
    721 	ifv_hash.lists = NULL;
    722 	ifv_hash.mask = 0;
    723 
    724 	return 0;
    725 }
    726 
    727 static int
    728 vlan_tag_hash(uint16_t tag, u_long mask)
    729 {
    730 	uint32_t hash;
    731 
    732 	hash = (tag >> 8) ^ tag;
    733 	hash = (hash >> 2) ^ hash;
    734 
    735 	return hash & mask;
    736 }
    737 
    738 static struct ifvlan_linkmib *
    739 vlan_getref_linkmib(struct ifvlan *sc, struct psref *psref)
    740 {
    741 	struct ifvlan_linkmib *mib;
    742 	int s;
    743 
    744 	s = pserialize_read_enter();
    745 	mib = sc->ifv_mib;
    746 	if (mib == NULL) {
    747 		pserialize_read_exit(s);
    748 		return NULL;
    749 	}
    750 	membar_datadep_consumer();
    751 	psref_acquire(psref, &mib->ifvm_psref, ifvm_psref_class);
    752 	pserialize_read_exit(s);
    753 
    754 	return mib;
    755 }
    756 
    757 static void
    758 vlan_putref_linkmib(struct ifvlan_linkmib *mib, struct psref *psref)
    759 {
    760 	if (mib == NULL)
    761 		return;
    762 	psref_release(psref, &mib->ifvm_psref, ifvm_psref_class);
    763 }
    764 
    765 static struct ifvlan_linkmib *
    766 vlan_lookup_tag_psref(struct ifnet *ifp, uint16_t tag, struct psref *psref)
    767 {
    768 	int idx;
    769 	int s;
    770 	struct ifvlan *sc;
    771 
    772 	idx = vlan_tag_hash(tag, ifv_hash.mask);
    773 
    774 	s = pserialize_read_enter();
    775 	PSLIST_READER_FOREACH(sc, &ifv_hash.lists[idx], struct ifvlan,
    776 	    ifv_hash) {
    777 		struct ifvlan_linkmib *mib = sc->ifv_mib;
    778 		if (mib == NULL)
    779 			continue;
    780 		if (mib->ifvm_tag != tag)
    781 			continue;
    782 		if (mib->ifvm_p != ifp)
    783 			continue;
    784 
    785 		psref_acquire(psref, &mib->ifvm_psref, ifvm_psref_class);
    786 		pserialize_read_exit(s);
    787 		return mib;
    788 	}
    789 	pserialize_read_exit(s);
    790 	return NULL;
    791 }
    792 
    793 static void
    794 vlan_linkmib_update(struct ifvlan *ifv, struct ifvlan_linkmib *nmib)
    795 {
    796 	struct ifvlan_linkmib *omib = ifv->ifv_mib;
    797 
    798 	KASSERT(mutex_owned(&ifv->ifv_lock));
    799 
    800 	membar_producer();
    801 	ifv->ifv_mib = nmib;
    802 
    803 	pserialize_perform(ifv->ifv_psz);
    804 	psref_target_destroy(&omib->ifvm_psref, ifvm_psref_class);
    805 }
    806 
    807 /*
    808  * Called when a parent interface is detaching; destroy any VLAN
    809  * configuration for the parent interface.
    810  */
    811 void
    812 vlan_ifdetach(struct ifnet *p)
    813 {
    814 	struct ifvlan *ifv;
    815 	struct ifvlan_linkmib *mib, **nmibs;
    816 	struct psref psref;
    817 	int error;
    818 	int bound;
    819 	int i, cnt = 0;
    820 
    821 	bound = curlwp_bind();
    822 
    823 	mutex_enter(&ifv_list.lock);
    824 	LIST_FOREACH(ifv, &ifv_list.list, ifv_list) {
    825 		mib = vlan_getref_linkmib(ifv, &psref);
    826 		if (mib == NULL)
    827 			continue;
    828 
    829 		if (mib->ifvm_p == p)
    830 			cnt++;
    831 
    832 		vlan_putref_linkmib(mib, &psref);
    833 	}
    834 	mutex_exit(&ifv_list.lock);
    835 
    836 	if (cnt == 0) {
    837 		curlwp_bindx(bound);
    838 		return;
    839 	}
    840 
    841 	/*
    842 	 * The value of "cnt" does not increase while ifv_list.lock
    843 	 * and ifv->ifv_lock are released here, because the parent
    844 	 * interface is detaching.
    845 	 */
    846 	nmibs = kmem_alloc(sizeof(*nmibs) * cnt, KM_SLEEP);
    847 	for (i = 0; i < cnt; i++) {
    848 		nmibs[i] = kmem_alloc(sizeof(*nmibs[i]), KM_SLEEP);
    849 	}
    850 
    851 	mutex_enter(&ifv_list.lock);
    852 
    853 	i = 0;
    854 	LIST_FOREACH(ifv, &ifv_list.list, ifv_list) {
    855 		struct ifnet *ifp = &ifv->ifv_if;
    856 
    857 		/* IFNET_LOCK must be held before ifv_lock. */
    858 		IFNET_LOCK(ifp);
    859 		mutex_enter(&ifv->ifv_lock);
    860 
    861 		/* XXX ifv_mib = NULL? */
    862 		if (ifv->ifv_mib->ifvm_p == p) {
    863 			KASSERTMSG(i < cnt, "no memory for unconfig, parent=%s",
    864 			    p->if_xname);
    865 			error = vlan_unconfig_locked(ifv, nmibs[i]);
    866 			if (!error) {
    867 				nmibs[i] = NULL;
    868 				i++;
    869 			}
    870 
    871 		}
    872 
    873 		mutex_exit(&ifv->ifv_lock);
    874 		IFNET_UNLOCK(ifp);
    875 	}
    876 
    877 	mutex_exit(&ifv_list.lock);
    878 
    879 	curlwp_bindx(bound);
    880 
    881 	for (i = 0; i < cnt; i++) {
    882 		if (nmibs[i])
    883 			kmem_free(nmibs[i], sizeof(*nmibs[i]));
    884 	}
    885 
    886 	kmem_free(nmibs, sizeof(*nmibs) * cnt);
    887 
    888 	return;
    889 }
    890 
    891 static int
    892 vlan_set_promisc(struct ifnet *ifp)
    893 {
    894 	struct ifvlan *ifv = ifp->if_softc;
    895 	struct ifvlan_linkmib *mib;
    896 	struct psref psref;
    897 	int error = 0;
    898 	int bound;
    899 
    900 	bound = curlwp_bind();
    901 	mib = vlan_getref_linkmib(ifv, &psref);
    902 	if (mib == NULL) {
    903 		curlwp_bindx(bound);
    904 		return EBUSY;
    905 	}
    906 
    907 	if ((ifp->if_flags & IFF_PROMISC) != 0) {
    908 		if ((ifv->ifv_flags & IFVF_PROMISC) == 0) {
    909 			error = vlan_safe_ifpromisc(mib->ifvm_p, 1);
    910 			if (error == 0)
    911 				ifv->ifv_flags |= IFVF_PROMISC;
    912 		}
    913 	} else {
    914 		if ((ifv->ifv_flags & IFVF_PROMISC) != 0) {
    915 			error = vlan_safe_ifpromisc(mib->ifvm_p, 0);
    916 			if (error == 0)
    917 				ifv->ifv_flags &= ~IFVF_PROMISC;
    918 		}
    919 	}
    920 	vlan_putref_linkmib(mib, &psref);
    921 	curlwp_bindx(bound);
    922 
    923 	return error;
    924 }
    925 
    926 static int
    927 vlan_ioctl(struct ifnet *ifp, u_long cmd, void *data)
    928 {
    929 	struct lwp *l = curlwp;
    930 	struct ifvlan *ifv = ifp->if_softc;
    931 	struct ifaddr *ifa = (struct ifaddr *) data;
    932 	struct ifreq *ifr = (struct ifreq *) data;
    933 	struct ifnet *pr;
    934 	struct ifcapreq *ifcr;
    935 	struct vlanreq vlr;
    936 	struct ifvlan_linkmib *mib;
    937 	struct psref psref;
    938 	int error = 0;
    939 	int bound;
    940 
    941 	switch (cmd) {
    942 	case SIOCSIFMTU:
    943 		bound = curlwp_bind();
    944 		mib = vlan_getref_linkmib(ifv, &psref);
    945 		if (mib == NULL) {
    946 			curlwp_bindx(bound);
    947 			error = EBUSY;
    948 			break;
    949 		}
    950 
    951 		if (mib->ifvm_p == NULL) {
    952 			vlan_putref_linkmib(mib, &psref);
    953 			curlwp_bindx(bound);
    954 			error = EINVAL;
    955 		} else if (
    956 		    ifr->ifr_mtu > (mib->ifvm_p->if_mtu - mib->ifvm_mtufudge) ||
    957 		    ifr->ifr_mtu < (mib->ifvm_mintu - mib->ifvm_mtufudge)) {
    958 			vlan_putref_linkmib(mib, &psref);
    959 			curlwp_bindx(bound);
    960 			error = EINVAL;
    961 		} else {
    962 			vlan_putref_linkmib(mib, &psref);
    963 			curlwp_bindx(bound);
    964 
    965 			error = ifioctl_common(ifp, cmd, data);
    966 			if (error == ENETRESET)
    967 				error = 0;
    968 		}
    969 
    970 		break;
    971 
    972 	case SIOCSETVLAN:
    973 		if ((error = kauth_authorize_network(l->l_cred,
    974 		    KAUTH_NETWORK_INTERFACE,
    975 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
    976 		    NULL)) != 0)
    977 			break;
    978 		if ((error = copyin(ifr->ifr_data, &vlr, sizeof(vlr))) != 0)
    979 			break;
    980 
    981 		if (vlr.vlr_parent[0] == '\0') {
    982 			bound = curlwp_bind();
    983 			mib = vlan_getref_linkmib(ifv, &psref);
    984 			if (mib == NULL) {
    985 				curlwp_bindx(bound);
    986 				error = EBUSY;
    987 				break;
    988 			}
    989 
    990 			if (mib->ifvm_p != NULL &&
    991 			    (ifp->if_flags & IFF_PROMISC) != 0)
    992 				error = vlan_safe_ifpromisc(mib->ifvm_p, 0);
    993 
    994 			vlan_putref_linkmib(mib, &psref);
    995 			curlwp_bindx(bound);
    996 
    997 			vlan_unconfig(ifp);
    998 			break;
    999 		}
   1000 		if (vlr.vlr_tag != EVL_VLANOFTAG(vlr.vlr_tag)) {
   1001 			error = EINVAL;		 /* check for valid tag */
   1002 			break;
   1003 		}
   1004 		if ((pr = ifunit(vlr.vlr_parent)) == NULL) {
   1005 			error = ENOENT;
   1006 			break;
   1007 		}
   1008 		error = vlan_config(ifv, pr, vlr.vlr_tag);
   1009 		if (error != 0) {
   1010 			break;
   1011 		}
   1012 
   1013 		/* Update promiscuous mode, if necessary. */
   1014 		vlan_set_promisc(ifp);
   1015 
   1016 		ifp->if_flags |= IFF_RUNNING;
   1017 		break;
   1018 
   1019 	case SIOCGETVLAN:
   1020 		memset(&vlr, 0, sizeof(vlr));
   1021 		bound = curlwp_bind();
   1022 		mib = vlan_getref_linkmib(ifv, &psref);
   1023 		if (mib == NULL) {
   1024 			curlwp_bindx(bound);
   1025 			error = EBUSY;
   1026 			break;
   1027 		}
   1028 		if (mib->ifvm_p != NULL) {
   1029 			snprintf(vlr.vlr_parent, sizeof(vlr.vlr_parent), "%s",
   1030 			    mib->ifvm_p->if_xname);
   1031 			vlr.vlr_tag = mib->ifvm_tag;
   1032 		}
   1033 		vlan_putref_linkmib(mib, &psref);
   1034 		curlwp_bindx(bound);
   1035 		error = copyout(&vlr, ifr->ifr_data, sizeof(vlr));
   1036 		break;
   1037 
   1038 	case SIOCSIFFLAGS:
   1039 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
   1040 			break;
   1041 		/*
   1042 		 * For promiscuous mode, we enable promiscuous mode on
   1043 		 * the parent if we need promiscuous on the VLAN interface.
   1044 		 */
   1045 		bound = curlwp_bind();
   1046 		mib = vlan_getref_linkmib(ifv, &psref);
   1047 		if (mib == NULL) {
   1048 			curlwp_bindx(bound);
   1049 			error = EBUSY;
   1050 			break;
   1051 		}
   1052 
   1053 		if (mib->ifvm_p != NULL)
   1054 			error = vlan_set_promisc(ifp);
   1055 		vlan_putref_linkmib(mib, &psref);
   1056 		curlwp_bindx(bound);
   1057 		break;
   1058 
   1059 	case SIOCADDMULTI:
   1060 		mutex_enter(&ifv->ifv_lock);
   1061 		mib = ifv->ifv_mib;
   1062 		if (mib == NULL) {
   1063 			error = EBUSY;
   1064 			mutex_exit(&ifv->ifv_lock);
   1065 			break;
   1066 		}
   1067 
   1068 		error = (mib->ifvm_p != NULL) ?
   1069 		    (*mib->ifvm_msw->vmsw_addmulti)(ifv, ifr) : EINVAL;
   1070 		mib = NULL;
   1071 		mutex_exit(&ifv->ifv_lock);
   1072 		break;
   1073 
   1074 	case SIOCDELMULTI:
   1075 		mutex_enter(&ifv->ifv_lock);
   1076 		mib = ifv->ifv_mib;
   1077 		if (mib == NULL) {
   1078 			error = EBUSY;
   1079 			mutex_exit(&ifv->ifv_lock);
   1080 			break;
   1081 		}
   1082 		error = (mib->ifvm_p != NULL) ?
   1083 		    (*mib->ifvm_msw->vmsw_delmulti)(ifv, ifr) : EINVAL;
   1084 		mib = NULL;
   1085 		mutex_exit(&ifv->ifv_lock);
   1086 		break;
   1087 
   1088 	case SIOCSIFCAP:
   1089 		ifcr = data;
   1090 		/* make sure caps are enabled on parent */
   1091 		bound = curlwp_bind();
   1092 		mib = vlan_getref_linkmib(ifv, &psref);
   1093 		if (mib == NULL) {
   1094 			curlwp_bindx(bound);
   1095 			error = EBUSY;
   1096 			break;
   1097 		}
   1098 
   1099 		if (mib->ifvm_p == NULL) {
   1100 			vlan_putref_linkmib(mib, &psref);
   1101 			curlwp_bindx(bound);
   1102 			error = EINVAL;
   1103 			break;
   1104 		}
   1105 		if ((mib->ifvm_p->if_capenable & ifcr->ifcr_capenable) !=
   1106 		    ifcr->ifcr_capenable) {
   1107 			vlan_putref_linkmib(mib, &psref);
   1108 			curlwp_bindx(bound);
   1109 			error = EINVAL;
   1110 			break;
   1111 		}
   1112 
   1113 		vlan_putref_linkmib(mib, &psref);
   1114 		curlwp_bindx(bound);
   1115 
   1116 		if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET)
   1117 			error = 0;
   1118 		break;
   1119 	case SIOCINITIFADDR:
   1120 		bound = curlwp_bind();
   1121 		mib = vlan_getref_linkmib(ifv, &psref);
   1122 		if (mib == NULL) {
   1123 			curlwp_bindx(bound);
   1124 			error = EBUSY;
   1125 			break;
   1126 		}
   1127 
   1128 		if (mib->ifvm_p == NULL) {
   1129 			error = EINVAL;
   1130 			vlan_putref_linkmib(mib, &psref);
   1131 			curlwp_bindx(bound);
   1132 			break;
   1133 		}
   1134 		vlan_putref_linkmib(mib, &psref);
   1135 		curlwp_bindx(bound);
   1136 
   1137 		ifp->if_flags |= IFF_UP;
   1138 #ifdef INET
   1139 		if (ifa->ifa_addr->sa_family == AF_INET)
   1140 			arp_ifinit(ifp, ifa);
   1141 #endif
   1142 		break;
   1143 
   1144 	default:
   1145 		error = ether_ioctl(ifp, cmd, data);
   1146 	}
   1147 
   1148 	return error;
   1149 }
   1150 
   1151 static int
   1152 vlan_ether_addmulti(struct ifvlan *ifv, struct ifreq *ifr)
   1153 {
   1154 	const struct sockaddr *sa = ifreq_getaddr(SIOCADDMULTI, ifr);
   1155 	struct vlan_mc_entry *mc;
   1156 	uint8_t addrlo[ETHER_ADDR_LEN], addrhi[ETHER_ADDR_LEN];
   1157 	struct ifvlan_linkmib *mib;
   1158 	int error;
   1159 
   1160 	KASSERT(mutex_owned(&ifv->ifv_lock));
   1161 
   1162 	if (sa->sa_len > sizeof(struct sockaddr_storage))
   1163 		return EINVAL;
   1164 
   1165 	error = ether_addmulti(sa, &ifv->ifv_ec);
   1166 	if (error != ENETRESET)
   1167 		return error;
   1168 
   1169 	/*
   1170 	 * This is a new multicast address.  We have to tell parent
   1171 	 * about it.  Also, remember this multicast address so that
   1172 	 * we can delete it on unconfigure.
   1173 	 */
   1174 	mc = malloc(sizeof(struct vlan_mc_entry), M_DEVBUF, M_NOWAIT);
   1175 	if (mc == NULL) {
   1176 		error = ENOMEM;
   1177 		goto alloc_failed;
   1178 	}
   1179 
   1180 	/*
   1181 	 * Since ether_addmulti() returned ENETRESET, the following two
   1182 	 * statements shouldn't fail. Here ifv_ec is implicitly protected
   1183 	 * by the ifv_lock lock.
   1184 	 */
   1185 	error = ether_multiaddr(sa, addrlo, addrhi);
   1186 	KASSERT(error == 0);
   1187 
   1188 	ETHER_LOCK(&ifv->ifv_ec);
   1189 	mc->mc_enm = ether_lookup_multi(addrlo, addrhi, &ifv->ifv_ec);
   1190 	ETHER_UNLOCK(&ifv->ifv_ec);
   1191 
   1192 	KASSERT(mc->mc_enm != NULL);
   1193 
   1194 	memcpy(&mc->mc_addr, sa, sa->sa_len);
   1195 	LIST_INSERT_HEAD(&ifv->ifv_mc_listhead, mc, mc_entries);
   1196 
   1197 	mib = ifv->ifv_mib;
   1198 
   1199 	KERNEL_LOCK_UNLESS_IFP_MPSAFE(mib->ifvm_p);
   1200 	IFNET_LOCK(mib->ifvm_p);
   1201 	error = if_mcast_op(mib->ifvm_p, SIOCADDMULTI, sa);
   1202 	IFNET_UNLOCK(mib->ifvm_p);
   1203 	KERNEL_UNLOCK_UNLESS_IFP_MPSAFE(mib->ifvm_p);
   1204 
   1205 	if (error != 0)
   1206 		goto ioctl_failed;
   1207 	return error;
   1208 
   1209 ioctl_failed:
   1210 	LIST_REMOVE(mc, mc_entries);
   1211 	free(mc, M_DEVBUF);
   1212 
   1213 alloc_failed:
   1214 	(void)ether_delmulti(sa, &ifv->ifv_ec);
   1215 	return error;
   1216 }
   1217 
   1218 static int
   1219 vlan_ether_delmulti(struct ifvlan *ifv, struct ifreq *ifr)
   1220 {
   1221 	const struct sockaddr *sa = ifreq_getaddr(SIOCDELMULTI, ifr);
   1222 	struct ether_multi *enm;
   1223 	struct vlan_mc_entry *mc;
   1224 	struct ifvlan_linkmib *mib;
   1225 	uint8_t addrlo[ETHER_ADDR_LEN], addrhi[ETHER_ADDR_LEN];
   1226 	int error;
   1227 
   1228 	KASSERT(mutex_owned(&ifv->ifv_lock));
   1229 
   1230 	/*
   1231 	 * Find a key to lookup vlan_mc_entry.  We have to do this
   1232 	 * before calling ether_delmulti for obvious reasons.
   1233 	 */
   1234 	if ((error = ether_multiaddr(sa, addrlo, addrhi)) != 0)
   1235 		return error;
   1236 
   1237 	ETHER_LOCK(&ifv->ifv_ec);
   1238 	enm = ether_lookup_multi(addrlo, addrhi, &ifv->ifv_ec);
   1239 	ETHER_UNLOCK(&ifv->ifv_ec);
   1240 	if (enm == NULL)
   1241 		return EINVAL;
   1242 
   1243 	LIST_FOREACH(mc, &ifv->ifv_mc_listhead, mc_entries) {
   1244 		if (mc->mc_enm == enm)
   1245 			break;
   1246 	}
   1247 
   1248 	/* We woun't delete entries we didn't add */
   1249 	if (mc == NULL)
   1250 		return EINVAL;
   1251 
   1252 	error = ether_delmulti(sa, &ifv->ifv_ec);
   1253 	if (error != ENETRESET)
   1254 		return error;
   1255 
   1256 	/* We no longer use this multicast address.  Tell parent so. */
   1257 	mib = ifv->ifv_mib;
   1258 	IFNET_LOCK(mib->ifvm_p);
   1259 	error = if_mcast_op(mib->ifvm_p, SIOCDELMULTI, sa);
   1260 	IFNET_UNLOCK(mib->ifvm_p);
   1261 
   1262 	if (error == 0) {
   1263 		/* And forget about this address. */
   1264 		LIST_REMOVE(mc, mc_entries);
   1265 		free(mc, M_DEVBUF);
   1266 	} else {
   1267 		(void)ether_addmulti(sa, &ifv->ifv_ec);
   1268 	}
   1269 
   1270 	return error;
   1271 }
   1272 
   1273 /*
   1274  * Delete any multicast address we have asked to add from parent
   1275  * interface.  Called when the vlan is being unconfigured.
   1276  */
   1277 static void
   1278 vlan_ether_purgemulti(struct ifvlan *ifv)
   1279 {
   1280 	struct vlan_mc_entry *mc;
   1281 	struct ifvlan_linkmib *mib;
   1282 
   1283 	KASSERT(mutex_owned(&ifv->ifv_lock));
   1284 	mib = ifv->ifv_mib;
   1285 	if (mib == NULL) {
   1286 		return;
   1287 	}
   1288 
   1289 	while ((mc = LIST_FIRST(&ifv->ifv_mc_listhead)) != NULL) {
   1290 		IFNET_LOCK(mib->ifvm_p);
   1291 		(void)if_mcast_op(mib->ifvm_p, SIOCDELMULTI,
   1292 		    sstocsa(&mc->mc_addr));
   1293 		IFNET_UNLOCK(mib->ifvm_p);
   1294 		LIST_REMOVE(mc, mc_entries);
   1295 		free(mc, M_DEVBUF);
   1296 	}
   1297 }
   1298 
   1299 static void
   1300 vlan_start(struct ifnet *ifp)
   1301 {
   1302 	struct ifvlan *ifv = ifp->if_softc;
   1303 	struct ifnet *p;
   1304 	struct ethercom *ec;
   1305 	struct mbuf *m;
   1306 	struct ifvlan_linkmib *mib;
   1307 	struct psref psref;
   1308 	int error;
   1309 
   1310 	mib = vlan_getref_linkmib(ifv, &psref);
   1311 	if (mib == NULL)
   1312 		return;
   1313 	p = mib->ifvm_p;
   1314 	ec = (void *)mib->ifvm_p;
   1315 
   1316 	ifp->if_flags |= IFF_OACTIVE;
   1317 
   1318 	for (;;) {
   1319 		IFQ_DEQUEUE(&ifp->if_snd, m);
   1320 		if (m == NULL)
   1321 			break;
   1322 
   1323 #ifdef ALTQ
   1324 		/*
   1325 		 * KERNEL_LOCK is required for ALTQ even if NET_MPSAFE is
   1326 		 * defined.
   1327 		 */
   1328 		KERNEL_LOCK(1, NULL);
   1329 		/*
   1330 		 * If ALTQ is enabled on the parent interface, do
   1331 		 * classification; the queueing discipline might
   1332 		 * not require classification, but might require
   1333 		 * the address family/header pointer in the pktattr.
   1334 		 */
   1335 		if (ALTQ_IS_ENABLED(&p->if_snd)) {
   1336 			switch (p->if_type) {
   1337 			case IFT_ETHER:
   1338 				altq_etherclassify(&p->if_snd, m);
   1339 				break;
   1340 			default:
   1341 				panic("%s: impossible (altq)", __func__);
   1342 			}
   1343 		}
   1344 		KERNEL_UNLOCK_ONE(NULL);
   1345 #endif /* ALTQ */
   1346 
   1347 		bpf_mtap(ifp, m, BPF_D_OUT);
   1348 		/*
   1349 		 * If the parent can insert the tag itself, just mark
   1350 		 * the tag in the mbuf header.
   1351 		 */
   1352 		if (ec->ec_capabilities & ETHERCAP_VLAN_HWTAGGING) {
   1353 			vlan_set_tag(m, mib->ifvm_tag);
   1354 		} else {
   1355 			/*
   1356 			 * insert the tag ourselves
   1357 			 */
   1358 			M_PREPEND(m, mib->ifvm_encaplen, M_DONTWAIT);
   1359 			if (m == NULL) {
   1360 				printf("%s: unable to prepend encap header",
   1361 				    p->if_xname);
   1362 				ifp->if_oerrors++;
   1363 				continue;
   1364 			}
   1365 
   1366 			switch (p->if_type) {
   1367 			case IFT_ETHER:
   1368 			    {
   1369 				struct ether_vlan_header *evl;
   1370 
   1371 				if (m->m_len < sizeof(struct ether_vlan_header))
   1372 					m = m_pullup(m,
   1373 					    sizeof(struct ether_vlan_header));
   1374 				if (m == NULL) {
   1375 					printf("%s: unable to pullup encap "
   1376 					    "header", p->if_xname);
   1377 					ifp->if_oerrors++;
   1378 					continue;
   1379 				}
   1380 
   1381 				/*
   1382 				 * Transform the Ethernet header into an
   1383 				 * Ethernet header with 802.1Q encapsulation.
   1384 				 */
   1385 				memmove(mtod(m, void *),
   1386 				    mtod(m, char *) + mib->ifvm_encaplen,
   1387 				    sizeof(struct ether_header));
   1388 				evl = mtod(m, struct ether_vlan_header *);
   1389 				evl->evl_proto = evl->evl_encap_proto;
   1390 				evl->evl_encap_proto = htons(ETHERTYPE_VLAN);
   1391 				evl->evl_tag = htons(mib->ifvm_tag);
   1392 
   1393 				/*
   1394 				 * To cater for VLAN-aware layer 2 ethernet
   1395 				 * switches which may need to strip the tag
   1396 				 * before forwarding the packet, make sure
   1397 				 * the packet+tag is at least 68 bytes long.
   1398 				 * This is necessary because our parent will
   1399 				 * only pad to 64 bytes (ETHER_MIN_LEN) and
   1400 				 * some switches will not pad by themselves
   1401 				 * after deleting a tag.
   1402 				 */
   1403 				const size_t min_data_len = ETHER_MIN_LEN -
   1404 				    ETHER_CRC_LEN + ETHER_VLAN_ENCAP_LEN;
   1405 				if (m->m_pkthdr.len < min_data_len) {
   1406 					m_copyback(m, m->m_pkthdr.len,
   1407 					    min_data_len - m->m_pkthdr.len,
   1408 					    vlan_zero_pad_buff);
   1409 				}
   1410 				break;
   1411 			    }
   1412 
   1413 			default:
   1414 				panic("%s: impossible", __func__);
   1415 			}
   1416 		}
   1417 
   1418 		if ((p->if_flags & IFF_RUNNING) == 0) {
   1419 			m_freem(m);
   1420 			continue;
   1421 		}
   1422 
   1423 		error = if_transmit_lock(p, m);
   1424 		if (error) {
   1425 			/* mbuf is already freed */
   1426 			ifp->if_oerrors++;
   1427 			continue;
   1428 		}
   1429 		ifp->if_opackets++;
   1430 	}
   1431 
   1432 	ifp->if_flags &= ~IFF_OACTIVE;
   1433 
   1434 	/* Remove reference to mib before release */
   1435 	vlan_putref_linkmib(mib, &psref);
   1436 }
   1437 
   1438 static int
   1439 vlan_transmit(struct ifnet *ifp, struct mbuf *m)
   1440 {
   1441 	struct ifvlan *ifv = ifp->if_softc;
   1442 	struct ifnet *p;
   1443 	struct ethercom *ec;
   1444 	struct ifvlan_linkmib *mib;
   1445 	struct psref psref;
   1446 	int error;
   1447 	size_t pktlen = m->m_pkthdr.len;
   1448 	bool mcast = (m->m_flags & M_MCAST) != 0;
   1449 
   1450 	mib = vlan_getref_linkmib(ifv, &psref);
   1451 	if (mib == NULL) {
   1452 		m_freem(m);
   1453 		return ENETDOWN;
   1454 	}
   1455 
   1456 	p = mib->ifvm_p;
   1457 	ec = (void *)mib->ifvm_p;
   1458 
   1459 	bpf_mtap(ifp, m, BPF_D_OUT);
   1460 
   1461 	if ((error = pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_OUT)) != 0)
   1462 		goto out;
   1463 	if (m == NULL)
   1464 		goto out;
   1465 
   1466 	/*
   1467 	 * If the parent can insert the tag itself, just mark
   1468 	 * the tag in the mbuf header.
   1469 	 */
   1470 	if (ec->ec_capabilities & ETHERCAP_VLAN_HWTAGGING) {
   1471 		vlan_set_tag(m, mib->ifvm_tag);
   1472 	} else {
   1473 		/*
   1474 		 * insert the tag ourselves
   1475 		 */
   1476 		M_PREPEND(m, mib->ifvm_encaplen, M_DONTWAIT);
   1477 		if (m == NULL) {
   1478 			printf("%s: unable to prepend encap header",
   1479 			    p->if_xname);
   1480 			ifp->if_oerrors++;
   1481 			error = ENOBUFS;
   1482 			goto out;
   1483 		}
   1484 
   1485 		switch (p->if_type) {
   1486 		case IFT_ETHER:
   1487 		    {
   1488 			struct ether_vlan_header *evl;
   1489 
   1490 			if (m->m_len < sizeof(struct ether_vlan_header))
   1491 				m = m_pullup(m,
   1492 				    sizeof(struct ether_vlan_header));
   1493 			if (m == NULL) {
   1494 				printf("%s: unable to pullup encap "
   1495 				    "header", p->if_xname);
   1496 				ifp->if_oerrors++;
   1497 				error = ENOBUFS;
   1498 				goto out;
   1499 			}
   1500 
   1501 			/*
   1502 			 * Transform the Ethernet header into an
   1503 			 * Ethernet header with 802.1Q encapsulation.
   1504 			 */
   1505 			memmove(mtod(m, void *),
   1506 			    mtod(m, char *) + mib->ifvm_encaplen,
   1507 			    sizeof(struct ether_header));
   1508 			evl = mtod(m, struct ether_vlan_header *);
   1509 			evl->evl_proto = evl->evl_encap_proto;
   1510 			evl->evl_encap_proto = htons(ETHERTYPE_VLAN);
   1511 			evl->evl_tag = htons(mib->ifvm_tag);
   1512 
   1513 			/*
   1514 			 * To cater for VLAN-aware layer 2 ethernet
   1515 			 * switches which may need to strip the tag
   1516 			 * before forwarding the packet, make sure
   1517 			 * the packet+tag is at least 68 bytes long.
   1518 			 * This is necessary because our parent will
   1519 			 * only pad to 64 bytes (ETHER_MIN_LEN) and
   1520 			 * some switches will not pad by themselves
   1521 			 * after deleting a tag.
   1522 			 */
   1523 			const size_t min_data_len = ETHER_MIN_LEN -
   1524 			    ETHER_CRC_LEN + ETHER_VLAN_ENCAP_LEN;
   1525 			if (m->m_pkthdr.len < min_data_len) {
   1526 				m_copyback(m, m->m_pkthdr.len,
   1527 				    min_data_len - m->m_pkthdr.len,
   1528 				    vlan_zero_pad_buff);
   1529 			}
   1530 			break;
   1531 		    }
   1532 
   1533 		default:
   1534 			panic("%s: impossible", __func__);
   1535 		}
   1536 	}
   1537 
   1538 	if ((p->if_flags & IFF_RUNNING) == 0) {
   1539 		m_freem(m);
   1540 		error = ENETDOWN;
   1541 		goto out;
   1542 	}
   1543 
   1544 	error = if_transmit_lock(p, m);
   1545 	if (error) {
   1546 		/* mbuf is already freed */
   1547 		ifp->if_oerrors++;
   1548 	} else {
   1549 
   1550 		ifp->if_opackets++;
   1551 		ifp->if_obytes += pktlen;
   1552 		if (mcast)
   1553 			ifp->if_omcasts++;
   1554 	}
   1555 
   1556 out:
   1557 	/* Remove reference to mib before release */
   1558 	vlan_putref_linkmib(mib, &psref);
   1559 	return error;
   1560 }
   1561 
   1562 /*
   1563  * Given an Ethernet frame, find a valid vlan interface corresponding to the
   1564  * given source interface and tag, then run the real packet through the
   1565  * parent's input routine.
   1566  */
   1567 void
   1568 vlan_input(struct ifnet *ifp, struct mbuf *m)
   1569 {
   1570 	struct ifvlan *ifv;
   1571 	uint16_t vid;
   1572 	struct ifvlan_linkmib *mib;
   1573 	struct psref psref;
   1574 	bool have_vtag;
   1575 
   1576 	have_vtag = vlan_has_tag(m);
   1577 	if (have_vtag) {
   1578 		vid = EVL_VLANOFTAG(vlan_get_tag(m));
   1579 		m->m_flags &= ~M_VLANTAG;
   1580 	} else {
   1581 		struct ether_vlan_header *evl;
   1582 
   1583 		if (ifp->if_type != IFT_ETHER) {
   1584 			panic("%s: impossible", __func__);
   1585 		}
   1586 
   1587 		if (m->m_len < sizeof(struct ether_vlan_header) &&
   1588 		    (m = m_pullup(m,
   1589 		     sizeof(struct ether_vlan_header))) == NULL) {
   1590 			printf("%s: no memory for VLAN header, "
   1591 			    "dropping packet.\n", ifp->if_xname);
   1592 			return;
   1593 		}
   1594 		evl = mtod(m, struct ether_vlan_header *);
   1595 		KASSERT(ntohs(evl->evl_encap_proto) == ETHERTYPE_VLAN);
   1596 
   1597 		vid = EVL_VLANOFTAG(ntohs(evl->evl_tag));
   1598 
   1599 		/*
   1600 		 * Restore the original ethertype.  We'll remove
   1601 		 * the encapsulation after we've found the vlan
   1602 		 * interface corresponding to the tag.
   1603 		 */
   1604 		evl->evl_encap_proto = evl->evl_proto;
   1605 	}
   1606 
   1607 	mib = vlan_lookup_tag_psref(ifp, vid, &psref);
   1608 	if (mib == NULL) {
   1609 		m_freem(m);
   1610 		ifp->if_noproto++;
   1611 		return;
   1612 	}
   1613 	KASSERT(mib->ifvm_encaplen == ETHER_VLAN_ENCAP_LEN);
   1614 
   1615 	ifv = mib->ifvm_ifvlan;
   1616 	if ((ifv->ifv_if.if_flags & (IFF_UP|IFF_RUNNING)) !=
   1617 	    (IFF_UP|IFF_RUNNING)) {
   1618 		m_freem(m);
   1619 		ifp->if_noproto++;
   1620 		goto out;
   1621 	}
   1622 
   1623 	/*
   1624 	 * Now, remove the encapsulation header.  The original
   1625 	 * header has already been fixed up above.
   1626 	 */
   1627 	if (!have_vtag) {
   1628 		memmove(mtod(m, char *) + mib->ifvm_encaplen,
   1629 		    mtod(m, void *), sizeof(struct ether_header));
   1630 		m_adj(m, mib->ifvm_encaplen);
   1631 	}
   1632 
   1633 	m_set_rcvif(m, &ifv->ifv_if);
   1634 	ifv->ifv_if.if_ipackets++;
   1635 
   1636 	if (pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_IN) != 0)
   1637 		goto out;
   1638 	if (m == NULL)
   1639 		goto out;
   1640 
   1641 	m->m_flags &= ~M_PROMISC;
   1642 	if_input(&ifv->ifv_if, m);
   1643 out:
   1644 	vlan_putref_linkmib(mib, &psref);
   1645 }
   1646 
   1647 /*
   1648  * Module infrastructure
   1649  */
   1650 #include "if_module.h"
   1651 
   1652 IF_MODULE(MODULE_CLASS_DRIVER, vlan, NULL)
   1653