xref: /freebsd/sys/net/if.c (revision 08aba0aec7b7f676ccc3f7886f59f277d668d5b4)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 2010 Bjoern A. Zeeb <bz@FreeBSD.org>
5  * Copyright (c) 1980, 1986, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  *	@(#)if.c	8.5 (Berkeley) 1/9/95
33  * $FreeBSD$
34  */
35 
36 #include "opt_bpf.h"
37 #include "opt_inet6.h"
38 #include "opt_inet.h"
39 #include "opt_ddb.h"
40 
41 #include <sys/param.h>
42 #include <sys/capsicum.h>
43 #include <sys/conf.h>
44 #include <sys/eventhandler.h>
45 #include <sys/malloc.h>
46 #include <sys/domainset.h>
47 #include <sys/sbuf.h>
48 #include <sys/bus.h>
49 #include <sys/epoch.h>
50 #include <sys/mbuf.h>
51 #include <sys/systm.h>
52 #include <sys/priv.h>
53 #include <sys/proc.h>
54 #include <sys/socket.h>
55 #include <sys/socketvar.h>
56 #include <sys/protosw.h>
57 #include <sys/kernel.h>
58 #include <sys/lock.h>
59 #include <sys/refcount.h>
60 #include <sys/module.h>
61 #include <sys/nv.h>
62 #include <sys/rwlock.h>
63 #include <sys/sockio.h>
64 #include <sys/syslog.h>
65 #include <sys/sysctl.h>
66 #include <sys/sysent.h>
67 #include <sys/taskqueue.h>
68 #include <sys/domain.h>
69 #include <sys/jail.h>
70 #include <sys/priv.h>
71 
72 #ifdef DDB
73 #include <ddb/ddb.h>
74 #endif
75 
76 #include <machine/stdarg.h>
77 #include <vm/uma.h>
78 
79 #include <net/bpf.h>
80 #include <net/ethernet.h>
81 #include <net/if.h>
82 #include <net/if_arp.h>
83 #include <net/if_clone.h>
84 #include <net/if_dl.h>
85 #include <net/if_types.h>
86 #include <net/if_var.h>
87 #include <net/if_media.h>
88 #include <net/if_mib.h>
89 #include <net/if_vlan_var.h>
90 #include <net/radix.h>
91 #include <net/route.h>
92 #include <net/route/route_ctl.h>
93 #include <net/vnet.h>
94 
95 #if defined(INET) || defined(INET6)
96 #include <net/ethernet.h>
97 #include <netinet/in.h>
98 #include <netinet/in_var.h>
99 #include <netinet/ip.h>
100 #include <netinet/ip_carp.h>
101 #ifdef INET
102 #include <net/debugnet.h>
103 #include <netinet/if_ether.h>
104 #endif /* INET */
105 #ifdef INET6
106 #include <netinet6/in6_var.h>
107 #include <netinet6/in6_ifattach.h>
108 #endif /* INET6 */
109 #endif /* INET || INET6 */
110 
111 #include <security/mac/mac_framework.h>
112 
113 /*
114  * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name
115  * and ifr_ifru when it is used in SIOCGIFCONF.
116  */
117 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) ==
118     offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru");
119 
120 __read_mostly epoch_t net_epoch_preempt;
121 #ifdef COMPAT_FREEBSD32
122 #include <sys/mount.h>
123 #include <compat/freebsd32/freebsd32.h>
124 
125 struct ifreq_buffer32 {
126 	uint32_t	length;		/* (size_t) */
127 	uint32_t	buffer;		/* (void *) */
128 };
129 
130 /*
131  * Interface request structure used for socket
132  * ioctl's.  All interface ioctl's must have parameter
133  * definitions which begin with ifr_name.  The
134  * remainder may be interface specific.
135  */
136 struct ifreq32 {
137 	char	ifr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
138 	union {
139 		struct sockaddr	ifru_addr;
140 		struct sockaddr	ifru_dstaddr;
141 		struct sockaddr	ifru_broadaddr;
142 		struct ifreq_buffer32 ifru_buffer;
143 		short		ifru_flags[2];
144 		short		ifru_index;
145 		int		ifru_jid;
146 		int		ifru_metric;
147 		int		ifru_mtu;
148 		int		ifru_phys;
149 		int		ifru_media;
150 		uint32_t	ifru_data;
151 		int		ifru_cap[2];
152 		u_int		ifru_fib;
153 		u_char		ifru_vlan_pcp;
154 	} ifr_ifru;
155 };
156 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32));
157 CTASSERT(__offsetof(struct ifreq, ifr_ifru) ==
158     __offsetof(struct ifreq32, ifr_ifru));
159 
160 struct ifconf32 {
161 	int32_t	ifc_len;
162 	union {
163 		uint32_t	ifcu_buf;
164 		uint32_t	ifcu_req;
165 	} ifc_ifcu;
166 };
167 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
168 
169 struct ifdrv32 {
170 	char		ifd_name[IFNAMSIZ];
171 	uint32_t	ifd_cmd;
172 	uint32_t	ifd_len;
173 	uint32_t	ifd_data;
174 };
175 #define SIOCSDRVSPEC32	_IOC_NEWTYPE(SIOCSDRVSPEC, struct ifdrv32)
176 #define SIOCGDRVSPEC32	_IOC_NEWTYPE(SIOCGDRVSPEC, struct ifdrv32)
177 
178 struct ifgroupreq32 {
179 	char	ifgr_name[IFNAMSIZ];
180 	u_int	ifgr_len;
181 	union {
182 		char		ifgru_group[IFNAMSIZ];
183 		uint32_t	ifgru_groups;
184 	} ifgr_ifgru;
185 };
186 #define	SIOCAIFGROUP32	_IOC_NEWTYPE(SIOCAIFGROUP, struct ifgroupreq32)
187 #define	SIOCGIFGROUP32	_IOC_NEWTYPE(SIOCGIFGROUP, struct ifgroupreq32)
188 #define	SIOCDIFGROUP32	_IOC_NEWTYPE(SIOCDIFGROUP, struct ifgroupreq32)
189 #define	SIOCGIFGMEMB32	_IOC_NEWTYPE(SIOCGIFGMEMB, struct ifgroupreq32)
190 
191 struct ifmediareq32 {
192 	char		ifm_name[IFNAMSIZ];
193 	int		ifm_current;
194 	int		ifm_mask;
195 	int		ifm_status;
196 	int		ifm_active;
197 	int		ifm_count;
198 	uint32_t	ifm_ulist;	/* (int *) */
199 };
200 #define	SIOCGIFMEDIA32	_IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32)
201 #define	SIOCGIFXMEDIA32	_IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32)
202 #endif /* COMPAT_FREEBSD32 */
203 
204 union ifreq_union {
205 	struct ifreq	ifr;
206 #ifdef COMPAT_FREEBSD32
207 	struct ifreq32	ifr32;
208 #endif
209 };
210 
211 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
212     "Link layers");
213 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
214     "Generic link-management");
215 
216 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
217     &ifqmaxlen, 0, "max send queue size");
218 
219 /* Log link state change events */
220 static int log_link_state_change = 1;
221 
222 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
223 	&log_link_state_change, 0,
224 	"log interface link state change events");
225 
226 /* Log promiscuous mode change events */
227 static int log_promisc_mode_change = 1;
228 
229 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
230 	&log_promisc_mode_change, 1,
231 	"log promiscuous mode change events");
232 
233 /* Interface description */
234 static unsigned int ifdescr_maxlen = 1024;
235 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
236 	&ifdescr_maxlen, 0,
237 	"administrative maximum length for interface description");
238 
239 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
240 
241 /* global sx for non-critical path ifdescr */
242 static struct sx ifdescr_sx;
243 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
244 
245 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
246 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
247 /* These are external hooks for CARP. */
248 void	(*carp_linkstate_p)(struct ifnet *ifp);
249 void	(*carp_demote_adj_p)(int, char *);
250 int	(*carp_master_p)(struct ifaddr *);
251 #if defined(INET) || defined(INET6)
252 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
253 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
254     const struct sockaddr *sa);
255 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
256 int	(*carp_attach_p)(struct ifaddr *, int);
257 void	(*carp_detach_p)(struct ifaddr *, bool);
258 #endif
259 #ifdef INET
260 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
261 #endif
262 #ifdef INET6
263 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
264 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
265     const struct in6_addr *taddr);
266 #endif
267 
268 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
269 
270 /*
271  * XXX: Style; these should be sorted alphabetically, and unprototyped
272  * static functions should be prototyped. Currently they are sorted by
273  * declaration order.
274  */
275 static void	if_attachdomain(void *);
276 static void	if_attachdomain1(struct ifnet *);
277 static int	ifconf(u_long, caddr_t);
278 static void	if_input_default(struct ifnet *, struct mbuf *);
279 static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
280 static void	if_route(struct ifnet *, int flag, int fam);
281 static int	if_setflag(struct ifnet *, int, int, int *, int);
282 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
283 static void	if_unroute(struct ifnet *, int flag, int fam);
284 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
285 static void	do_link_state_change(void *, int);
286 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
287 static int	if_getgroupmembers(struct ifgroupreq *);
288 static void	if_delgroups(struct ifnet *);
289 static void	if_attach_internal(struct ifnet *, bool);
290 static int	if_detach_internal(struct ifnet *, bool);
291 static void	if_siocaddmulti(void *, int);
292 static void	if_link_ifnet(struct ifnet *);
293 static bool	if_unlink_ifnet(struct ifnet *, bool);
294 #ifdef VIMAGE
295 static int	if_vmove(struct ifnet *, struct vnet *);
296 #endif
297 
298 #ifdef INET6
299 /*
300  * XXX: declare here to avoid to include many inet6 related files..
301  * should be more generalized?
302  */
303 extern void	nd6_setmtu(struct ifnet *);
304 #endif
305 
306 /* ipsec helper hooks */
307 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
308 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
309 
310 int	ifqmaxlen = IFQ_MAXLEN;
311 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
312 VNET_DEFINE(struct ifgrouphead, ifg_head);
313 
314 /* Table of ifnet by index. */
315 static int if_index;
316 static int if_indexlim = 8;
317 static struct ifindex_entry {
318 	struct ifnet	*ife_ifnet;
319 	uint16_t	ife_gencnt;
320 } *ifindex_table;
321 
322 SYSCTL_NODE(_net_link_generic, IFMIB_SYSTEM, system,
323     CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
324     "Variables global to all interfaces");
325 static int
326 sysctl_ifcount(SYSCTL_HANDLER_ARGS)
327 {
328 	int rv = 0;
329 
330 	IFNET_RLOCK();
331 	for (int i = 1; i <= if_index; i++)
332 		if (ifindex_table[i].ife_ifnet != NULL &&
333 		    ifindex_table[i].ife_ifnet->if_vnet == curvnet)
334 			rv = i;
335 	IFNET_RUNLOCK();
336 
337 	return (sysctl_handle_int(oidp, &rv, 0, req));
338 }
339 SYSCTL_PROC(_net_link_generic_system, IFMIB_IFCOUNT, ifcount,
340     CTLTYPE_INT | CTLFLAG_VNET | CTLFLAG_RD, NULL, 0, sysctl_ifcount, "I",
341     "Maximum known interface index");
342 
343 /*
344  * The global network interface list (V_ifnet) and related state (such as
345  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock.
346  * This may be acquired to stabilise the list, or we may rely on NET_EPOCH.
347  */
348 struct sx ifnet_sxlock;
349 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
350 
351 struct sx ifnet_detach_sxlock;
352 SX_SYSINIT_FLAGS(ifnet_detach, &ifnet_detach_sxlock, "ifnet_detach_sx",
353     SX_RECURSE);
354 
355 #ifdef VIMAGE
356 #define	VNET_IS_SHUTTING_DOWN(_vnet)					\
357     ((_vnet)->vnet_shutdown && (_vnet)->vnet_state < SI_SUB_VNET_DONE)
358 #endif
359 
360 static	if_com_alloc_t *if_com_alloc[256];
361 static	if_com_free_t *if_com_free[256];
362 
363 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
364 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
365 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
366 
367 struct ifnet *
368 ifnet_byindex(u_int idx)
369 {
370 	struct ifnet *ifp;
371 
372 	NET_EPOCH_ASSERT();
373 
374 	if (__predict_false(idx > if_index))
375 		return (NULL);
376 
377 	ifp = ck_pr_load_ptr(&ifindex_table[idx].ife_ifnet);
378 
379 	if (curvnet != NULL && ifp != NULL && ifp->if_vnet != curvnet)
380 		ifp = NULL;
381 
382 	return (ifp);
383 }
384 
385 struct ifnet *
386 ifnet_byindex_ref(u_int idx)
387 {
388 	struct ifnet *ifp;
389 
390 	ifp = ifnet_byindex(idx);
391 	if (ifp == NULL || (ifp->if_flags & IFF_DYING))
392 		return (NULL);
393 	if (!if_try_ref(ifp))
394 		return (NULL);
395 	return (ifp);
396 }
397 
398 struct ifnet *
399 ifnet_byindexgen(uint16_t idx, uint16_t gen)
400 {
401 	struct ifnet *ifp;
402 
403 	NET_EPOCH_ASSERT();
404 
405 	if (__predict_false(idx > if_index))
406 		return (NULL);
407 
408 	ifp = ck_pr_load_ptr(&ifindex_table[idx].ife_ifnet);
409 
410 	if (ifindex_table[idx].ife_gencnt == gen)
411 		return (ifp);
412 	else
413 		return (NULL);
414 }
415 
416 /*
417  * Network interface utility routines.
418  *
419  * Routines with ifa_ifwith* names take sockaddr *'s as
420  * parameters.
421  */
422 
423 static void
424 if_init(void *arg __unused)
425 {
426 
427 	ifindex_table = malloc(if_indexlim * sizeof(*ifindex_table),
428 	    M_IFNET, M_WAITOK | M_ZERO);
429 }
430 SYSINIT(if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, if_init, NULL);
431 
432 static void
433 vnet_if_init(const void *unused __unused)
434 {
435 
436 	CK_STAILQ_INIT(&V_ifnet);
437 	CK_STAILQ_INIT(&V_ifg_head);
438 	vnet_if_clone_init();
439 }
440 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
441     NULL);
442 
443 static void
444 if_link_ifnet(struct ifnet *ifp)
445 {
446 
447 	IFNET_WLOCK();
448 	CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
449 #ifdef VIMAGE
450 	curvnet->vnet_ifcnt++;
451 #endif
452 	IFNET_WUNLOCK();
453 }
454 
455 static bool
456 if_unlink_ifnet(struct ifnet *ifp, bool vmove)
457 {
458 	struct ifnet *iter;
459 	int found = 0;
460 
461 	IFNET_WLOCK();
462 	CK_STAILQ_FOREACH(iter, &V_ifnet, if_link)
463 		if (iter == ifp) {
464 			CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link);
465 			if (!vmove)
466 				ifp->if_flags |= IFF_DYING;
467 			found = 1;
468 			break;
469 		}
470 #ifdef VIMAGE
471 	curvnet->vnet_ifcnt--;
472 #endif
473 	IFNET_WUNLOCK();
474 
475 	return (found);
476 }
477 
478 #ifdef VIMAGE
479 static void
480 vnet_if_return(const void *unused __unused)
481 {
482 	struct ifnet *ifp, *nifp;
483 	struct ifnet **pending;
484 	int found __diagused;
485 	int i;
486 
487 	i = 0;
488 
489 	/*
490 	 * We need to protect our access to the V_ifnet tailq. Ordinarily we'd
491 	 * enter NET_EPOCH, but that's not possible, because if_vmove() calls
492 	 * if_detach_internal(), which waits for NET_EPOCH callbacks to
493 	 * complete. We can't do that from within NET_EPOCH.
494 	 *
495 	 * However, we can also use the IFNET_xLOCK, which is the V_ifnet
496 	 * read/write lock. We cannot hold the lock as we call if_vmove()
497 	 * though, as that presents LOR w.r.t ifnet_sx, in_multi_sx and iflib
498 	 * ctx lock.
499 	 */
500 	IFNET_WLOCK();
501 
502 	pending = malloc(sizeof(struct ifnet *) * curvnet->vnet_ifcnt,
503 	    M_IFNET, M_WAITOK | M_ZERO);
504 
505 	/* Return all inherited interfaces to their parent vnets. */
506 	CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
507 		if (ifp->if_home_vnet != ifp->if_vnet) {
508 			found = if_unlink_ifnet(ifp, true);
509 			MPASS(found);
510 
511 			pending[i++] = ifp;
512 		}
513 	}
514 	IFNET_WUNLOCK();
515 
516 	for (int j = 0; j < i; j++) {
517 		sx_xlock(&ifnet_detach_sxlock);
518 		if_vmove(pending[j], pending[j]->if_home_vnet);
519 		sx_xunlock(&ifnet_detach_sxlock);
520 	}
521 
522 	free(pending, M_IFNET);
523 }
524 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
525     vnet_if_return, NULL);
526 #endif
527 
528 /*
529  * Allocate a struct ifnet and an index for an interface.  A layer 2
530  * common structure will also be allocated if an allocation routine is
531  * registered for the passed type.
532  */
533 static struct ifnet *
534 if_alloc_domain(u_char type, int numa_domain)
535 {
536 	struct ifnet *ifp;
537 	u_short idx;
538 
539 	KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large"));
540 	if (numa_domain == IF_NODOM)
541 		ifp = malloc(sizeof(struct ifnet), M_IFNET,
542 		    M_WAITOK | M_ZERO);
543 	else
544 		ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET,
545 		    DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO);
546 	ifp->if_type = type;
547 	ifp->if_alloctype = type;
548 	ifp->if_numa_domain = numa_domain;
549 #ifdef VIMAGE
550 	ifp->if_vnet = curvnet;
551 #endif
552 	if (if_com_alloc[type] != NULL) {
553 		ifp->if_l2com = if_com_alloc[type](type, ifp);
554 		KASSERT(ifp->if_l2com, ("%s: if_com_alloc[%u] failed", __func__,
555 		    type));
556 	}
557 
558 	IF_ADDR_LOCK_INIT(ifp);
559 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
560 	TASK_INIT(&ifp->if_addmultitask, 0, if_siocaddmulti, ifp);
561 	ifp->if_afdata_initialized = 0;
562 	IF_AFDATA_LOCK_INIT(ifp);
563 	CK_STAILQ_INIT(&ifp->if_addrhead);
564 	CK_STAILQ_INIT(&ifp->if_multiaddrs);
565 	CK_STAILQ_INIT(&ifp->if_groups);
566 #ifdef MAC
567 	mac_ifnet_init(ifp);
568 #endif
569 	ifq_init(&ifp->if_snd, ifp);
570 
571 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
572 	for (int i = 0; i < IFCOUNTERS; i++)
573 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
574 	ifp->if_get_counter = if_get_counter_default;
575 	ifp->if_pcp = IFNET_PCP_NONE;
576 
577 	/* Allocate an ifindex array entry. */
578 	IFNET_WLOCK();
579 	/*
580 	 * Try to find an empty slot below if_index.  If we fail, take the
581 	 * next slot.
582 	 */
583 	for (idx = 1; idx <= if_index; idx++) {
584 		if (ifindex_table[idx].ife_ifnet == NULL)
585 			break;
586 	}
587 
588 	/* Catch if_index overflow. */
589 	if (idx >= if_indexlim) {
590 		struct ifindex_entry *new, *old;
591 		int newlim;
592 
593 		newlim = if_indexlim * 2;
594 		new = malloc(newlim * sizeof(*new), M_IFNET, M_WAITOK | M_ZERO);
595 		memcpy(new, ifindex_table, if_indexlim * sizeof(*new));
596 		old = ifindex_table;
597 		ck_pr_store_ptr(&ifindex_table, new);
598 		if_indexlim = newlim;
599 		epoch_wait_preempt(net_epoch_preempt);
600 		free(old, M_IFNET);
601 	}
602 	if (idx > if_index)
603 		if_index = idx;
604 
605 	ifp->if_index = idx;
606 	ifp->if_idxgen = ifindex_table[idx].ife_gencnt;
607 	ck_pr_store_ptr(&ifindex_table[idx].ife_ifnet, ifp);
608 	IFNET_WUNLOCK();
609 
610 	return (ifp);
611 }
612 
613 struct ifnet *
614 if_alloc_dev(u_char type, device_t dev)
615 {
616 	int numa_domain;
617 
618 	if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0)
619 		return (if_alloc_domain(type, IF_NODOM));
620 	return (if_alloc_domain(type, numa_domain));
621 }
622 
623 struct ifnet *
624 if_alloc(u_char type)
625 {
626 
627 	return (if_alloc_domain(type, IF_NODOM));
628 }
629 /*
630  * Do the actual work of freeing a struct ifnet, and layer 2 common
631  * structure.  This call is made when the network epoch guarantees
632  * us that nobody holds a pointer to the interface.
633  */
634 static void
635 if_free_deferred(epoch_context_t ctx)
636 {
637 	struct ifnet *ifp = __containerof(ctx, struct ifnet, if_epoch_ctx);
638 
639 	KASSERT((ifp->if_flags & IFF_DYING),
640 	    ("%s: interface not dying", __func__));
641 
642 	if (if_com_free[ifp->if_alloctype] != NULL)
643 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
644 		    ifp->if_alloctype);
645 
646 #ifdef MAC
647 	mac_ifnet_destroy(ifp);
648 #endif /* MAC */
649 	IF_AFDATA_DESTROY(ifp);
650 	IF_ADDR_LOCK_DESTROY(ifp);
651 	ifq_delete(&ifp->if_snd);
652 
653 	for (int i = 0; i < IFCOUNTERS; i++)
654 		counter_u64_free(ifp->if_counters[i]);
655 
656 	free(ifp->if_description, M_IFDESCR);
657 	free(ifp->if_hw_addr, M_IFADDR);
658 	free(ifp, M_IFNET);
659 }
660 
661 /*
662  * Deregister an interface and free the associated storage.
663  */
664 void
665 if_free(struct ifnet *ifp)
666 {
667 
668 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
669 
670 	/*
671 	 * XXXGL: An interface index is really an alias to ifp pointer.
672 	 * Why would we clear the alias now, and not in the deferred
673 	 * context?  Indeed there is nothing wrong with some network
674 	 * thread obtaining ifp via ifnet_byindex() inside the network
675 	 * epoch and then dereferencing ifp while we perform if_free(),
676 	 * and after if_free() finished, too.
677 	 *
678 	 * This early index freeing was important back when ifindex was
679 	 * virtualized and interface would outlive the vnet.
680 	 */
681 	IFNET_WLOCK();
682 	MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
683 	ck_pr_store_ptr(&ifindex_table[ifp->if_index].ife_ifnet, NULL);
684 	ifindex_table[ifp->if_index].ife_gencnt++;
685 	while (if_index > 0 && ifindex_table[if_index].ife_ifnet == NULL)
686 		if_index--;
687 	IFNET_WUNLOCK();
688 
689 	if (refcount_release(&ifp->if_refcount))
690 		NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx);
691 }
692 
693 /*
694  * Interfaces to keep an ifnet type-stable despite the possibility of the
695  * driver calling if_free().  If there are additional references, we defer
696  * freeing the underlying data structure.
697  */
698 void
699 if_ref(struct ifnet *ifp)
700 {
701 	u_int old __diagused;
702 
703 	/* We don't assert the ifnet list lock here, but arguably should. */
704 	old = refcount_acquire(&ifp->if_refcount);
705 	KASSERT(old > 0, ("%s: ifp %p has 0 refs", __func__, ifp));
706 }
707 
708 bool
709 if_try_ref(struct ifnet *ifp)
710 {
711 	NET_EPOCH_ASSERT();
712 	return (refcount_acquire_if_not_zero(&ifp->if_refcount));
713 }
714 
715 void
716 if_rele(struct ifnet *ifp)
717 {
718 
719 	if (!refcount_release(&ifp->if_refcount))
720 		return;
721 	NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx);
722 }
723 
724 void
725 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
726 {
727 
728 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
729 
730 	if (ifq->ifq_maxlen == 0)
731 		ifq->ifq_maxlen = ifqmaxlen;
732 
733 	ifq->altq_type = 0;
734 	ifq->altq_disc = NULL;
735 	ifq->altq_flags &= ALTQF_CANTCHANGE;
736 	ifq->altq_tbr  = NULL;
737 	ifq->altq_ifp  = ifp;
738 }
739 
740 void
741 ifq_delete(struct ifaltq *ifq)
742 {
743 	mtx_destroy(&ifq->ifq_mtx);
744 }
745 
746 /*
747  * Perform generic interface initialization tasks and attach the interface
748  * to the list of "active" interfaces.  If vmove flag is set on entry
749  * to if_attach_internal(), perform only a limited subset of initialization
750  * tasks, given that we are moving from one vnet to another an ifnet which
751  * has already been fully initialized.
752  *
753  * Note that if_detach_internal() removes group membership unconditionally
754  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
755  * Thus, when if_vmove() is applied to a cloned interface, group membership
756  * is lost while a cloned one always joins a group whose name is
757  * ifc->ifc_name.  To recover this after if_detach_internal() and
758  * if_attach_internal(), the cloner should be specified to
759  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
760  * attempts to join a group whose name is ifc->ifc_name.
761  *
762  * XXX:
763  *  - The decision to return void and thus require this function to
764  *    succeed is questionable.
765  *  - We should probably do more sanity checking.  For instance we don't
766  *    do anything to insure if_xname is unique or non-empty.
767  */
768 void
769 if_attach(struct ifnet *ifp)
770 {
771 
772 	if_attach_internal(ifp, false);
773 }
774 
775 /*
776  * Compute the least common TSO limit.
777  */
778 void
779 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
780 {
781 	/*
782 	 * 1) If there is no limit currently, take the limit from
783 	 * the network adapter.
784 	 *
785 	 * 2) If the network adapter has a limit below the current
786 	 * limit, apply it.
787 	 */
788 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
789 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
790 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
791 	}
792 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
793 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
794 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
795 	}
796 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
797 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
798 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
799 	}
800 }
801 
802 /*
803  * Update TSO limit of a network adapter.
804  *
805  * Returns zero if no change. Else non-zero.
806  */
807 int
808 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
809 {
810 	int retval = 0;
811 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
812 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
813 		retval++;
814 	}
815 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
816 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
817 		retval++;
818 	}
819 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
820 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
821 		retval++;
822 	}
823 	return (retval);
824 }
825 
826 static void
827 if_attach_internal(struct ifnet *ifp, bool vmove)
828 {
829 	unsigned socksize, ifasize;
830 	int namelen, masklen;
831 	struct sockaddr_dl *sdl;
832 	struct ifaddr *ifa;
833 
834 	MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
835 
836 #ifdef VIMAGE
837 	ifp->if_vnet = curvnet;
838 	if (ifp->if_home_vnet == NULL)
839 		ifp->if_home_vnet = curvnet;
840 #endif
841 
842 	if_addgroup(ifp, IFG_ALL);
843 
844 #ifdef VIMAGE
845 	/* Restore group membership for cloned interface. */
846 	if (vmove)
847 		if_clone_restoregroup(ifp);
848 #endif
849 
850 	getmicrotime(&ifp->if_lastchange);
851 	ifp->if_epoch = time_uptime;
852 
853 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
854 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
855 	    ("transmit and qflush must both either be set or both be NULL"));
856 	if (ifp->if_transmit == NULL) {
857 		ifp->if_transmit = if_transmit;
858 		ifp->if_qflush = if_qflush;
859 	}
860 	if (ifp->if_input == NULL)
861 		ifp->if_input = if_input_default;
862 
863 	if (ifp->if_requestencap == NULL)
864 		ifp->if_requestencap = if_requestencap_default;
865 
866 	if (!vmove) {
867 #ifdef MAC
868 		mac_ifnet_create(ifp);
869 #endif
870 
871 		/*
872 		 * Create a Link Level name for this device.
873 		 */
874 		namelen = strlen(ifp->if_xname);
875 		/*
876 		 * Always save enough space for any possiable name so we
877 		 * can do a rename in place later.
878 		 */
879 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
880 		socksize = masklen + ifp->if_addrlen;
881 		if (socksize < sizeof(*sdl))
882 			socksize = sizeof(*sdl);
883 		socksize = roundup2(socksize, sizeof(long));
884 		ifasize = sizeof(*ifa) + 2 * socksize;
885 		ifa = ifa_alloc(ifasize, M_WAITOK);
886 		sdl = (struct sockaddr_dl *)(ifa + 1);
887 		sdl->sdl_len = socksize;
888 		sdl->sdl_family = AF_LINK;
889 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
890 		sdl->sdl_nlen = namelen;
891 		sdl->sdl_index = ifp->if_index;
892 		sdl->sdl_type = ifp->if_type;
893 		ifp->if_addr = ifa;
894 		ifa->ifa_ifp = ifp;
895 		ifa->ifa_addr = (struct sockaddr *)sdl;
896 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
897 		ifa->ifa_netmask = (struct sockaddr *)sdl;
898 		sdl->sdl_len = masklen;
899 		while (namelen != 0)
900 			sdl->sdl_data[--namelen] = 0xff;
901 		CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
902 		/* Reliably crash if used uninitialized. */
903 		ifp->if_broadcastaddr = NULL;
904 
905 		if (ifp->if_type == IFT_ETHER) {
906 			ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
907 			    M_WAITOK | M_ZERO);
908 		}
909 
910 #if defined(INET) || defined(INET6)
911 		/* Use defaults for TSO, if nothing is set */
912 		if (ifp->if_hw_tsomax == 0 &&
913 		    ifp->if_hw_tsomaxsegcount == 0 &&
914 		    ifp->if_hw_tsomaxsegsize == 0) {
915 			/*
916 			 * The TSO defaults needs to be such that an
917 			 * NFS mbuf list of 35 mbufs totalling just
918 			 * below 64K works and that a chain of mbufs
919 			 * can be defragged into at most 32 segments:
920 			 */
921 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
922 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
923 			ifp->if_hw_tsomaxsegcount = 35;
924 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
925 
926 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
927 			if (ifp->if_capabilities & IFCAP_TSO) {
928 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
929 				    ifp->if_hw_tsomax,
930 				    ifp->if_hw_tsomaxsegcount,
931 				    ifp->if_hw_tsomaxsegsize);
932 			}
933 		}
934 #endif
935 	}
936 #ifdef VIMAGE
937 	else {
938 		/*
939 		 * Update the interface index in the link layer address
940 		 * of the interface.
941 		 */
942 		for (ifa = ifp->if_addr; ifa != NULL;
943 		    ifa = CK_STAILQ_NEXT(ifa, ifa_link)) {
944 			if (ifa->ifa_addr->sa_family == AF_LINK) {
945 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
946 				sdl->sdl_index = ifp->if_index;
947 			}
948 		}
949 	}
950 #endif
951 
952 	if_link_ifnet(ifp);
953 
954 	if (domain_init_status >= 2)
955 		if_attachdomain1(ifp);
956 
957 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
958 	if (IS_DEFAULT_VNET(curvnet))
959 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
960 }
961 
962 static void
963 if_epochalloc(void *dummy __unused)
964 {
965 
966 	net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT);
967 }
968 SYSINIT(ifepochalloc, SI_SUB_EPOCH, SI_ORDER_ANY, if_epochalloc, NULL);
969 
970 static void
971 if_attachdomain(void *dummy)
972 {
973 	struct ifnet *ifp;
974 
975 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
976 		if_attachdomain1(ifp);
977 }
978 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
979     if_attachdomain, NULL);
980 
981 static void
982 if_attachdomain1(struct ifnet *ifp)
983 {
984 	struct domain *dp;
985 
986 	/*
987 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
988 	 * cannot lock ifp->if_afdata initialization, entirely.
989 	 */
990 	IF_AFDATA_LOCK(ifp);
991 	if (ifp->if_afdata_initialized >= domain_init_status) {
992 		IF_AFDATA_UNLOCK(ifp);
993 		log(LOG_WARNING, "%s called more than once on %s\n",
994 		    __func__, ifp->if_xname);
995 		return;
996 	}
997 	ifp->if_afdata_initialized = domain_init_status;
998 	IF_AFDATA_UNLOCK(ifp);
999 
1000 	/* address family dependent data region */
1001 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
1002 	for (dp = domains; dp; dp = dp->dom_next) {
1003 		if (dp->dom_ifattach)
1004 			ifp->if_afdata[dp->dom_family] =
1005 			    (*dp->dom_ifattach)(ifp);
1006 	}
1007 }
1008 
1009 /*
1010  * Remove any unicast or broadcast network addresses from an interface.
1011  */
1012 void
1013 if_purgeaddrs(struct ifnet *ifp)
1014 {
1015 	struct ifaddr *ifa;
1016 
1017 #ifdef INET6
1018 	/*
1019 	 * Need to leave multicast addresses of proxy NDP llentries
1020 	 * before in6_purgeifaddr() because the llentries are keys
1021 	 * for in6_multi objects of proxy NDP entries.
1022 	 * in6_purgeifaddr()s clean up llentries including proxy NDPs
1023 	 * then we would lose the keys if they are called earlier.
1024 	 */
1025 	in6_purge_proxy_ndp(ifp);
1026 #endif
1027 	while (1) {
1028 		struct epoch_tracker et;
1029 
1030 		NET_EPOCH_ENTER(et);
1031 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1032 			if (ifa->ifa_addr->sa_family != AF_LINK)
1033 				break;
1034 		}
1035 		NET_EPOCH_EXIT(et);
1036 
1037 		if (ifa == NULL)
1038 			break;
1039 #ifdef INET
1040 		/* XXX: Ugly!! ad hoc just for INET */
1041 		if (ifa->ifa_addr->sa_family == AF_INET) {
1042 			struct ifaliasreq ifr;
1043 
1044 			bzero(&ifr, sizeof(ifr));
1045 			ifr.ifra_addr = *ifa->ifa_addr;
1046 			if (ifa->ifa_dstaddr)
1047 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
1048 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
1049 			    NULL) == 0)
1050 				continue;
1051 		}
1052 #endif /* INET */
1053 #ifdef INET6
1054 		if (ifa->ifa_addr->sa_family == AF_INET6) {
1055 			in6_purgeifaddr((struct in6_ifaddr *)ifa);
1056 			/* ifp_addrhead is already updated */
1057 			continue;
1058 		}
1059 #endif /* INET6 */
1060 		IF_ADDR_WLOCK(ifp);
1061 		CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1062 		IF_ADDR_WUNLOCK(ifp);
1063 		ifa_free(ifa);
1064 	}
1065 }
1066 
1067 /*
1068  * Remove any multicast network addresses from an interface when an ifnet
1069  * is going away.
1070  */
1071 static void
1072 if_purgemaddrs(struct ifnet *ifp)
1073 {
1074 	struct ifmultiaddr *ifma;
1075 
1076 	IF_ADDR_WLOCK(ifp);
1077 	while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) {
1078 		ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs);
1079 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
1080 		if_delmulti_locked(ifp, ifma, 1);
1081 	}
1082 	IF_ADDR_WUNLOCK(ifp);
1083 }
1084 
1085 /*
1086  * Detach an interface, removing it from the list of "active" interfaces.
1087  * If vmove flag is set on entry to if_detach_internal(), perform only a
1088  * limited subset of cleanup tasks, given that we are moving an ifnet from
1089  * one vnet to another, where it must be fully operational.
1090  *
1091  * XXXRW: There are some significant questions about event ordering, and
1092  * how to prevent things from starting to use the interface during detach.
1093  */
1094 void
1095 if_detach(struct ifnet *ifp)
1096 {
1097 	bool found;
1098 
1099 	CURVNET_SET_QUIET(ifp->if_vnet);
1100 	found = if_unlink_ifnet(ifp, false);
1101 	if (found) {
1102 		sx_xlock(&ifnet_detach_sxlock);
1103 		if_detach_internal(ifp, false);
1104 		sx_xunlock(&ifnet_detach_sxlock);
1105 	}
1106 	CURVNET_RESTORE();
1107 }
1108 
1109 /*
1110  * The vmove flag, if set, indicates that we are called from a callpath
1111  * that is moving an interface to a different vnet instance.
1112  *
1113  * The shutdown flag, if set, indicates that we are called in the
1114  * process of shutting down a vnet instance.  Currently only the
1115  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
1116  * on a vnet instance shutdown without this flag being set, e.g., when
1117  * the cloned interfaces are destoyed as first thing of teardown.
1118  */
1119 static int
1120 if_detach_internal(struct ifnet *ifp, bool vmove)
1121 {
1122 	struct ifaddr *ifa;
1123 	int i;
1124 	struct domain *dp;
1125 #ifdef VIMAGE
1126 	bool shutdown;
1127 
1128 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1129 #endif
1130 
1131 	/*
1132 	 * At this point we know the interface still was on the ifnet list
1133 	 * and we removed it so we are in a stable state.
1134 	 */
1135 	epoch_wait_preempt(net_epoch_preempt);
1136 
1137 	/*
1138 	 * Ensure all pending EPOCH(9) callbacks have been executed. This
1139 	 * fixes issues about late destruction of multicast options
1140 	 * which lead to leave group calls, which in turn access the
1141 	 * belonging ifnet structure:
1142 	 */
1143 	NET_EPOCH_DRAIN_CALLBACKS();
1144 
1145 	/*
1146 	 * In any case (destroy or vmove) detach us from the groups
1147 	 * and remove/wait for pending events on the taskq.
1148 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
1149 	 */
1150 	if_delgroups(ifp);
1151 
1152 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1153 	taskqueue_drain(taskqueue_swi, &ifp->if_addmultitask);
1154 
1155 	if_down(ifp);
1156 
1157 #ifdef VIMAGE
1158 	/*
1159 	 * On VNET shutdown abort here as the stack teardown will do all
1160 	 * the work top-down for us.
1161 	 */
1162 	if (shutdown) {
1163 		/* Give interface users the chance to clean up. */
1164 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1165 
1166 		/*
1167 		 * In case of a vmove we are done here without error.
1168 		 * If we would signal an error it would lead to the same
1169 		 * abort as if we did not find the ifnet anymore.
1170 		 * if_detach() calls us in void context and does not care
1171 		 * about an early abort notification, so life is splendid :)
1172 		 */
1173 		goto finish_vnet_shutdown;
1174 	}
1175 #endif
1176 
1177 	/*
1178 	 * At this point we are not tearing down a VNET and are either
1179 	 * going to destroy or vmove the interface and have to cleanup
1180 	 * accordingly.
1181 	 */
1182 
1183 	/*
1184 	 * Remove routes and flush queues.
1185 	 */
1186 #ifdef ALTQ
1187 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1188 		altq_disable(&ifp->if_snd);
1189 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1190 		altq_detach(&ifp->if_snd);
1191 #endif
1192 
1193 	if_purgeaddrs(ifp);
1194 
1195 #ifdef INET
1196 	in_ifdetach(ifp);
1197 #endif
1198 
1199 #ifdef INET6
1200 	/*
1201 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
1202 	 * before removing routing entries below, since IPv6 interface direct
1203 	 * routes are expected to be removed by the IPv6-specific kernel API.
1204 	 * Otherwise, the kernel will detect some inconsistency and bark it.
1205 	 */
1206 	in6_ifdetach(ifp);
1207 #endif
1208 	if_purgemaddrs(ifp);
1209 
1210 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1211 	if (IS_DEFAULT_VNET(curvnet))
1212 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1213 
1214 	if (!vmove) {
1215 		/*
1216 		 * Prevent further calls into the device driver via ifnet.
1217 		 */
1218 		if_dead(ifp);
1219 
1220 		/*
1221 		 * Clean up all addresses.
1222 		 */
1223 		IF_ADDR_WLOCK(ifp);
1224 		if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) {
1225 			ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
1226 			CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1227 			IF_ADDR_WUNLOCK(ifp);
1228 			ifa_free(ifa);
1229 		} else
1230 			IF_ADDR_WUNLOCK(ifp);
1231 	}
1232 
1233 	rt_flushifroutes(ifp);
1234 
1235 #ifdef VIMAGE
1236 finish_vnet_shutdown:
1237 #endif
1238 	/*
1239 	 * We cannot hold the lock over dom_ifdetach calls as they might
1240 	 * sleep, for example trying to drain a callout, thus open up the
1241 	 * theoretical race with re-attaching.
1242 	 */
1243 	IF_AFDATA_LOCK(ifp);
1244 	i = ifp->if_afdata_initialized;
1245 	ifp->if_afdata_initialized = 0;
1246 	IF_AFDATA_UNLOCK(ifp);
1247 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1248 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1249 			(*dp->dom_ifdetach)(ifp,
1250 			    ifp->if_afdata[dp->dom_family]);
1251 			ifp->if_afdata[dp->dom_family] = NULL;
1252 		}
1253 	}
1254 
1255 	return (0);
1256 }
1257 
1258 #ifdef VIMAGE
1259 /*
1260  * if_vmove() performs a limited version of if_detach() in current
1261  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1262  */
1263 static int
1264 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1265 {
1266 #ifdef DEV_BPF
1267 	u_int bif_dlt, bif_hdrlen;
1268 #endif
1269 	int rc;
1270 
1271 #ifdef DEV_BPF
1272  	/*
1273 	 * if_detach_internal() will call the eventhandler to notify
1274 	 * interface departure.  That will detach if_bpf.  We need to
1275 	 * safe the dlt and hdrlen so we can re-attach it later.
1276 	 */
1277 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1278 #endif
1279 
1280 	/*
1281 	 * Detach from current vnet, but preserve LLADDR info, do not
1282 	 * mark as dead etc. so that the ifnet can be reattached later.
1283 	 * If we cannot find it, we lost the race to someone else.
1284 	 */
1285 	rc = if_detach_internal(ifp, true);
1286 	if (rc != 0)
1287 		return (rc);
1288 
1289 	/*
1290 	 * Perform interface-specific reassignment tasks, if provided by
1291 	 * the driver.
1292 	 */
1293 	if (ifp->if_reassign != NULL)
1294 		ifp->if_reassign(ifp, new_vnet, NULL);
1295 
1296 	/*
1297 	 * Switch to the context of the target vnet.
1298 	 */
1299 	CURVNET_SET_QUIET(new_vnet);
1300 	if_attach_internal(ifp, true);
1301 
1302 #ifdef DEV_BPF
1303 	if (ifp->if_bpf == NULL)
1304 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1305 #endif
1306 
1307 	CURVNET_RESTORE();
1308 	return (0);
1309 }
1310 
1311 /*
1312  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1313  */
1314 static int
1315 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1316 {
1317 	struct prison *pr;
1318 	struct ifnet *difp;
1319 	int error;
1320 	bool found __diagused;
1321 	bool shutdown;
1322 
1323 	MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
1324 
1325 	/* Try to find the prison within our visibility. */
1326 	sx_slock(&allprison_lock);
1327 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1328 	sx_sunlock(&allprison_lock);
1329 	if (pr == NULL)
1330 		return (ENXIO);
1331 	prison_hold_locked(pr);
1332 	mtx_unlock(&pr->pr_mtx);
1333 
1334 	/* Do not try to move the iface from and to the same prison. */
1335 	if (pr->pr_vnet == ifp->if_vnet) {
1336 		prison_free(pr);
1337 		return (EEXIST);
1338 	}
1339 
1340 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1341 	/* XXX Lock interfaces to avoid races. */
1342 	CURVNET_SET_QUIET(pr->pr_vnet);
1343 	difp = ifunit(ifname);
1344 	if (difp != NULL) {
1345 		CURVNET_RESTORE();
1346 		prison_free(pr);
1347 		return (EEXIST);
1348 	}
1349 	sx_xlock(&ifnet_detach_sxlock);
1350 
1351 	/* Make sure the VNET is stable. */
1352 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1353 	if (shutdown) {
1354 		sx_xunlock(&ifnet_detach_sxlock);
1355 		CURVNET_RESTORE();
1356 		prison_free(pr);
1357 		return (EBUSY);
1358 	}
1359 	CURVNET_RESTORE();
1360 
1361 	found = if_unlink_ifnet(ifp, true);
1362 	if (! found) {
1363 		sx_xunlock(&ifnet_detach_sxlock);
1364 		CURVNET_RESTORE();
1365 		prison_free(pr);
1366 		return (ENODEV);
1367 	}
1368 
1369 	/* Move the interface into the child jail/vnet. */
1370 	error = if_vmove(ifp, pr->pr_vnet);
1371 
1372 	/* Report the new if_xname back to the userland on success. */
1373 	if (error == 0)
1374 		sprintf(ifname, "%s", ifp->if_xname);
1375 
1376 	sx_xunlock(&ifnet_detach_sxlock);
1377 
1378 	prison_free(pr);
1379 	return (error);
1380 }
1381 
1382 static int
1383 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1384 {
1385 	struct prison *pr;
1386 	struct vnet *vnet_dst;
1387 	struct ifnet *ifp;
1388 	int error, found __diagused;
1389  	bool shutdown;
1390 
1391 	/* Try to find the prison within our visibility. */
1392 	sx_slock(&allprison_lock);
1393 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1394 	sx_sunlock(&allprison_lock);
1395 	if (pr == NULL)
1396 		return (ENXIO);
1397 	prison_hold_locked(pr);
1398 	mtx_unlock(&pr->pr_mtx);
1399 
1400 	/* Make sure the named iface exists in the source prison/vnet. */
1401 	CURVNET_SET(pr->pr_vnet);
1402 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1403 	if (ifp == NULL) {
1404 		CURVNET_RESTORE();
1405 		prison_free(pr);
1406 		return (ENXIO);
1407 	}
1408 
1409 	/* Do not try to move the iface from and to the same prison. */
1410 	vnet_dst = TD_TO_VNET(td);
1411 	if (vnet_dst == ifp->if_vnet) {
1412 		CURVNET_RESTORE();
1413 		prison_free(pr);
1414 		return (EEXIST);
1415 	}
1416 
1417 	/* Make sure the VNET is stable. */
1418 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1419 	if (shutdown) {
1420 		CURVNET_RESTORE();
1421 		prison_free(pr);
1422 		return (EBUSY);
1423 	}
1424 
1425 	/* Get interface back from child jail/vnet. */
1426 	found = if_unlink_ifnet(ifp, true);
1427 	MPASS(found);
1428 	sx_xlock(&ifnet_detach_sxlock);
1429 	error = if_vmove(ifp, vnet_dst);
1430 	sx_xunlock(&ifnet_detach_sxlock);
1431 	CURVNET_RESTORE();
1432 
1433 	/* Report the new if_xname back to the userland on success. */
1434 	if (error == 0)
1435 		sprintf(ifname, "%s", ifp->if_xname);
1436 
1437 	prison_free(pr);
1438 	return (error);
1439 }
1440 #endif /* VIMAGE */
1441 
1442 /*
1443  * Add a group to an interface
1444  */
1445 int
1446 if_addgroup(struct ifnet *ifp, const char *groupname)
1447 {
1448 	struct ifg_list		*ifgl;
1449 	struct ifg_group	*ifg = NULL;
1450 	struct ifg_member	*ifgm;
1451 	int 			 new = 0;
1452 
1453 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1454 	    groupname[strlen(groupname) - 1] <= '9')
1455 		return (EINVAL);
1456 
1457 	IFNET_WLOCK();
1458 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1459 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1460 			IFNET_WUNLOCK();
1461 			return (EEXIST);
1462 		}
1463 
1464 	if ((ifgl = malloc(sizeof(*ifgl), M_TEMP, M_NOWAIT)) == NULL) {
1465 	    	IFNET_WUNLOCK();
1466 		return (ENOMEM);
1467 	}
1468 
1469 	if ((ifgm = malloc(sizeof(*ifgm), M_TEMP, M_NOWAIT)) == NULL) {
1470 		free(ifgl, M_TEMP);
1471 		IFNET_WUNLOCK();
1472 		return (ENOMEM);
1473 	}
1474 
1475 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1476 		if (!strcmp(ifg->ifg_group, groupname))
1477 			break;
1478 
1479 	if (ifg == NULL) {
1480 		if ((ifg = malloc(sizeof(*ifg), M_TEMP, M_NOWAIT)) == NULL) {
1481 			free(ifgl, M_TEMP);
1482 			free(ifgm, M_TEMP);
1483 			IFNET_WUNLOCK();
1484 			return (ENOMEM);
1485 		}
1486 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1487 		ifg->ifg_refcnt = 0;
1488 		CK_STAILQ_INIT(&ifg->ifg_members);
1489 		CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1490 		new = 1;
1491 	}
1492 
1493 	ifg->ifg_refcnt++;
1494 	ifgl->ifgl_group = ifg;
1495 	ifgm->ifgm_ifp = ifp;
1496 
1497 	IF_ADDR_WLOCK(ifp);
1498 	CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1499 	CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1500 	IF_ADDR_WUNLOCK(ifp);
1501 
1502 	IFNET_WUNLOCK();
1503 
1504 	if (new)
1505 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1506 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1507 
1508 	return (0);
1509 }
1510 
1511 /*
1512  * Helper function to remove a group out of an interface.  Expects the global
1513  * ifnet lock to be write-locked, and drops it before returning.
1514  */
1515 static void
1516 _if_delgroup_locked(struct ifnet *ifp, struct ifg_list *ifgl,
1517     const char *groupname)
1518 {
1519 	struct ifg_member *ifgm;
1520 	bool freeifgl;
1521 
1522 	IFNET_WLOCK_ASSERT();
1523 
1524 	IF_ADDR_WLOCK(ifp);
1525 	CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1526 	IF_ADDR_WUNLOCK(ifp);
1527 
1528 	CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) {
1529 		if (ifgm->ifgm_ifp == ifp) {
1530 			CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1531 			    ifg_member, ifgm_next);
1532 			break;
1533 		}
1534 	}
1535 
1536 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1537 		CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group,
1538 		    ifg_next);
1539 		freeifgl = true;
1540 	} else {
1541 		freeifgl = false;
1542 	}
1543 	IFNET_WUNLOCK();
1544 
1545 	epoch_wait_preempt(net_epoch_preempt);
1546 	if (freeifgl) {
1547 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1548 		free(ifgl->ifgl_group, M_TEMP);
1549 	}
1550 	free(ifgm, M_TEMP);
1551 	free(ifgl, M_TEMP);
1552 
1553 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1554 }
1555 
1556 /*
1557  * Remove a group from an interface
1558  */
1559 int
1560 if_delgroup(struct ifnet *ifp, const char *groupname)
1561 {
1562 	struct ifg_list *ifgl;
1563 
1564 	IFNET_WLOCK();
1565 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1566 		if (strcmp(ifgl->ifgl_group->ifg_group, groupname) == 0)
1567 			break;
1568 	if (ifgl == NULL) {
1569 		IFNET_WUNLOCK();
1570 		return (ENOENT);
1571 	}
1572 
1573 	_if_delgroup_locked(ifp, ifgl, groupname);
1574 
1575 	return (0);
1576 }
1577 
1578 /*
1579  * Remove an interface from all groups
1580  */
1581 static void
1582 if_delgroups(struct ifnet *ifp)
1583 {
1584 	struct ifg_list *ifgl;
1585 	char groupname[IFNAMSIZ];
1586 
1587 	IFNET_WLOCK();
1588 	while ((ifgl = CK_STAILQ_FIRST(&ifp->if_groups)) != NULL) {
1589 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1590 		_if_delgroup_locked(ifp, ifgl, groupname);
1591 		IFNET_WLOCK();
1592 	}
1593 	IFNET_WUNLOCK();
1594 }
1595 
1596 /*
1597  * Stores all groups from an interface in memory pointed to by ifgr.
1598  */
1599 static int
1600 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp)
1601 {
1602 	int			 len, error;
1603 	struct ifg_list		*ifgl;
1604 	struct ifg_req		 ifgrq, *ifgp;
1605 
1606 	NET_EPOCH_ASSERT();
1607 
1608 	if (ifgr->ifgr_len == 0) {
1609 		CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1610 			ifgr->ifgr_len += sizeof(struct ifg_req);
1611 		return (0);
1612 	}
1613 
1614 	len = ifgr->ifgr_len;
1615 	ifgp = ifgr->ifgr_groups;
1616 	/* XXX: wire */
1617 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1618 		if (len < sizeof(ifgrq))
1619 			return (EINVAL);
1620 		bzero(&ifgrq, sizeof ifgrq);
1621 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1622 		    sizeof(ifgrq.ifgrq_group));
1623 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req))))
1624 			return (error);
1625 		len -= sizeof(ifgrq);
1626 		ifgp++;
1627 	}
1628 
1629 	return (0);
1630 }
1631 
1632 /*
1633  * Stores all members of a group in memory pointed to by igfr
1634  */
1635 static int
1636 if_getgroupmembers(struct ifgroupreq *ifgr)
1637 {
1638 	struct ifg_group	*ifg;
1639 	struct ifg_member	*ifgm;
1640 	struct ifg_req		 ifgrq, *ifgp;
1641 	int			 len, error;
1642 
1643 	IFNET_RLOCK();
1644 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1645 		if (strcmp(ifg->ifg_group, ifgr->ifgr_name) == 0)
1646 			break;
1647 	if (ifg == NULL) {
1648 		IFNET_RUNLOCK();
1649 		return (ENOENT);
1650 	}
1651 
1652 	if (ifgr->ifgr_len == 0) {
1653 		CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1654 			ifgr->ifgr_len += sizeof(ifgrq);
1655 		IFNET_RUNLOCK();
1656 		return (0);
1657 	}
1658 
1659 	len = ifgr->ifgr_len;
1660 	ifgp = ifgr->ifgr_groups;
1661 	CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1662 		if (len < sizeof(ifgrq)) {
1663 			IFNET_RUNLOCK();
1664 			return (EINVAL);
1665 		}
1666 		bzero(&ifgrq, sizeof ifgrq);
1667 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1668 		    sizeof(ifgrq.ifgrq_member));
1669 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1670 			IFNET_RUNLOCK();
1671 			return (error);
1672 		}
1673 		len -= sizeof(ifgrq);
1674 		ifgp++;
1675 	}
1676 	IFNET_RUNLOCK();
1677 
1678 	return (0);
1679 }
1680 
1681 /*
1682  * Return counter values from counter(9)s stored in ifnet.
1683  */
1684 uint64_t
1685 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1686 {
1687 
1688 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1689 
1690 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1691 }
1692 
1693 /*
1694  * Increase an ifnet counter. Usually used for counters shared
1695  * between the stack and a driver, but function supports them all.
1696  */
1697 void
1698 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1699 {
1700 
1701 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1702 
1703 	counter_u64_add(ifp->if_counters[cnt], inc);
1704 }
1705 
1706 /*
1707  * Copy data from ifnet to userland API structure if_data.
1708  */
1709 void
1710 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1711 {
1712 
1713 	ifd->ifi_type = ifp->if_type;
1714 	ifd->ifi_physical = 0;
1715 	ifd->ifi_addrlen = ifp->if_addrlen;
1716 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1717 	ifd->ifi_link_state = ifp->if_link_state;
1718 	ifd->ifi_vhid = 0;
1719 	ifd->ifi_datalen = sizeof(struct if_data);
1720 	ifd->ifi_mtu = ifp->if_mtu;
1721 	ifd->ifi_metric = ifp->if_metric;
1722 	ifd->ifi_baudrate = ifp->if_baudrate;
1723 	ifd->ifi_hwassist = ifp->if_hwassist;
1724 	ifd->ifi_epoch = ifp->if_epoch;
1725 	ifd->ifi_lastchange = ifp->if_lastchange;
1726 
1727 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1728 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1729 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1730 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1731 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1732 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1733 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1734 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1735 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1736 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1737 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1738 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1739 }
1740 
1741 /*
1742  * Initialization, destruction and refcounting functions for ifaddrs.
1743  */
1744 struct ifaddr *
1745 ifa_alloc(size_t size, int flags)
1746 {
1747 	struct ifaddr *ifa;
1748 
1749 	KASSERT(size >= sizeof(struct ifaddr),
1750 	    ("%s: invalid size %zu", __func__, size));
1751 
1752 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1753 	if (ifa == NULL)
1754 		return (NULL);
1755 
1756 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1757 		goto fail;
1758 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1759 		goto fail;
1760 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1761 		goto fail;
1762 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1763 		goto fail;
1764 
1765 	refcount_init(&ifa->ifa_refcnt, 1);
1766 
1767 	return (ifa);
1768 
1769 fail:
1770 	/* free(NULL) is okay */
1771 	counter_u64_free(ifa->ifa_opackets);
1772 	counter_u64_free(ifa->ifa_ipackets);
1773 	counter_u64_free(ifa->ifa_obytes);
1774 	counter_u64_free(ifa->ifa_ibytes);
1775 	free(ifa, M_IFADDR);
1776 
1777 	return (NULL);
1778 }
1779 
1780 void
1781 ifa_ref(struct ifaddr *ifa)
1782 {
1783 	u_int old __diagused;
1784 
1785 	old = refcount_acquire(&ifa->ifa_refcnt);
1786 	KASSERT(old > 0, ("%s: ifa %p has 0 refs", __func__, ifa));
1787 }
1788 
1789 int
1790 ifa_try_ref(struct ifaddr *ifa)
1791 {
1792 
1793 	NET_EPOCH_ASSERT();
1794 	return (refcount_acquire_if_not_zero(&ifa->ifa_refcnt));
1795 }
1796 
1797 static void
1798 ifa_destroy(epoch_context_t ctx)
1799 {
1800 	struct ifaddr *ifa;
1801 
1802 	ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx);
1803 	counter_u64_free(ifa->ifa_opackets);
1804 	counter_u64_free(ifa->ifa_ipackets);
1805 	counter_u64_free(ifa->ifa_obytes);
1806 	counter_u64_free(ifa->ifa_ibytes);
1807 	free(ifa, M_IFADDR);
1808 }
1809 
1810 void
1811 ifa_free(struct ifaddr *ifa)
1812 {
1813 
1814 	if (refcount_release(&ifa->ifa_refcnt))
1815 		NET_EPOCH_CALL(ifa_destroy, &ifa->ifa_epoch_ctx);
1816 }
1817 
1818 /*
1819  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1820  * structs used to represent other address families, it is necessary
1821  * to perform a different comparison.
1822  */
1823 
1824 #define	sa_dl_equal(a1, a2)	\
1825 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1826 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1827 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1828 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1829 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1830 
1831 /*
1832  * Locate an interface based on a complete address.
1833  */
1834 /*ARGSUSED*/
1835 struct ifaddr *
1836 ifa_ifwithaddr(const struct sockaddr *addr)
1837 {
1838 	struct ifnet *ifp;
1839 	struct ifaddr *ifa;
1840 
1841 	NET_EPOCH_ASSERT();
1842 
1843 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1844 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1845 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1846 				continue;
1847 			if (sa_equal(addr, ifa->ifa_addr)) {
1848 				goto done;
1849 			}
1850 			/* IP6 doesn't have broadcast */
1851 			if ((ifp->if_flags & IFF_BROADCAST) &&
1852 			    ifa->ifa_broadaddr &&
1853 			    ifa->ifa_broadaddr->sa_len != 0 &&
1854 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1855 				goto done;
1856 			}
1857 		}
1858 	}
1859 	ifa = NULL;
1860 done:
1861 	return (ifa);
1862 }
1863 
1864 int
1865 ifa_ifwithaddr_check(const struct sockaddr *addr)
1866 {
1867 	struct epoch_tracker et;
1868 	int rc;
1869 
1870 	NET_EPOCH_ENTER(et);
1871 	rc = (ifa_ifwithaddr(addr) != NULL);
1872 	NET_EPOCH_EXIT(et);
1873 	return (rc);
1874 }
1875 
1876 /*
1877  * Locate an interface based on the broadcast address.
1878  */
1879 /* ARGSUSED */
1880 struct ifaddr *
1881 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1882 {
1883 	struct ifnet *ifp;
1884 	struct ifaddr *ifa;
1885 
1886 	NET_EPOCH_ASSERT();
1887 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1888 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1889 			continue;
1890 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1891 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1892 				continue;
1893 			if ((ifp->if_flags & IFF_BROADCAST) &&
1894 			    ifa->ifa_broadaddr &&
1895 			    ifa->ifa_broadaddr->sa_len != 0 &&
1896 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1897 				goto done;
1898 			}
1899 		}
1900 	}
1901 	ifa = NULL;
1902 done:
1903 	return (ifa);
1904 }
1905 
1906 /*
1907  * Locate the point to point interface with a given destination address.
1908  */
1909 /*ARGSUSED*/
1910 struct ifaddr *
1911 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1912 {
1913 	struct ifnet *ifp;
1914 	struct ifaddr *ifa;
1915 
1916 	NET_EPOCH_ASSERT();
1917 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1918 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1919 			continue;
1920 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1921 			continue;
1922 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1923 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1924 				continue;
1925 			if (ifa->ifa_dstaddr != NULL &&
1926 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1927 				goto done;
1928 			}
1929 		}
1930 	}
1931 	ifa = NULL;
1932 done:
1933 	return (ifa);
1934 }
1935 
1936 /*
1937  * Find an interface on a specific network.  If many, choice
1938  * is most specific found.
1939  */
1940 struct ifaddr *
1941 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
1942 {
1943 	struct ifnet *ifp;
1944 	struct ifaddr *ifa;
1945 	struct ifaddr *ifa_maybe = NULL;
1946 	u_int af = addr->sa_family;
1947 	const char *addr_data = addr->sa_data, *cplim;
1948 
1949 	NET_EPOCH_ASSERT();
1950 	/*
1951 	 * AF_LINK addresses can be looked up directly by their index number,
1952 	 * so do that if we can.
1953 	 */
1954 	if (af == AF_LINK) {
1955 		ifp = ifnet_byindex(
1956 		    ((const struct sockaddr_dl *)addr)->sdl_index);
1957 		return (ifp ? ifp->if_addr : NULL);
1958 	}
1959 
1960 	/*
1961 	 * Scan though each interface, looking for ones that have addresses
1962 	 * in this address family and the requested fib.
1963 	 */
1964 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1965 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1966 			continue;
1967 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1968 			const char *cp, *cp2, *cp3;
1969 
1970 			if (ifa->ifa_addr->sa_family != af)
1971 next:				continue;
1972 			if (af == AF_INET &&
1973 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1974 				/*
1975 				 * This is a bit broken as it doesn't
1976 				 * take into account that the remote end may
1977 				 * be a single node in the network we are
1978 				 * looking for.
1979 				 * The trouble is that we don't know the
1980 				 * netmask for the remote end.
1981 				 */
1982 				if (ifa->ifa_dstaddr != NULL &&
1983 				    sa_equal(addr, ifa->ifa_dstaddr)) {
1984 					goto done;
1985 				}
1986 			} else {
1987 				/*
1988 				 * Scan all the bits in the ifa's address.
1989 				 * If a bit dissagrees with what we are
1990 				 * looking for, mask it with the netmask
1991 				 * to see if it really matters.
1992 				 * (A byte at a time)
1993 				 */
1994 				if (ifa->ifa_netmask == 0)
1995 					continue;
1996 				cp = addr_data;
1997 				cp2 = ifa->ifa_addr->sa_data;
1998 				cp3 = ifa->ifa_netmask->sa_data;
1999 				cplim = ifa->ifa_netmask->sa_len
2000 					+ (char *)ifa->ifa_netmask;
2001 				while (cp3 < cplim)
2002 					if ((*cp++ ^ *cp2++) & *cp3++)
2003 						goto next; /* next address! */
2004 				/*
2005 				 * If the netmask of what we just found
2006 				 * is more specific than what we had before
2007 				 * (if we had one), or if the virtual status
2008 				 * of new prefix is better than of the old one,
2009 				 * then remember the new one before continuing
2010 				 * to search for an even better one.
2011 				 */
2012 				if (ifa_maybe == NULL ||
2013 				    ifa_preferred(ifa_maybe, ifa) ||
2014 				    rn_refines((caddr_t)ifa->ifa_netmask,
2015 				    (caddr_t)ifa_maybe->ifa_netmask)) {
2016 					ifa_maybe = ifa;
2017 				}
2018 			}
2019 		}
2020 	}
2021 	ifa = ifa_maybe;
2022 	ifa_maybe = NULL;
2023 done:
2024 	return (ifa);
2025 }
2026 
2027 /*
2028  * Find an interface address specific to an interface best matching
2029  * a given address.
2030  */
2031 struct ifaddr *
2032 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
2033 {
2034 	struct ifaddr *ifa;
2035 	const char *cp, *cp2, *cp3;
2036 	char *cplim;
2037 	struct ifaddr *ifa_maybe = NULL;
2038 	u_int af = addr->sa_family;
2039 
2040 	if (af >= AF_MAX)
2041 		return (NULL);
2042 
2043 	NET_EPOCH_ASSERT();
2044 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2045 		if (ifa->ifa_addr->sa_family != af)
2046 			continue;
2047 		if (ifa_maybe == NULL)
2048 			ifa_maybe = ifa;
2049 		if (ifa->ifa_netmask == 0) {
2050 			if (sa_equal(addr, ifa->ifa_addr) ||
2051 			    (ifa->ifa_dstaddr &&
2052 			    sa_equal(addr, ifa->ifa_dstaddr)))
2053 				goto done;
2054 			continue;
2055 		}
2056 		if (ifp->if_flags & IFF_POINTOPOINT) {
2057 			if (sa_equal(addr, ifa->ifa_dstaddr))
2058 				goto done;
2059 		} else {
2060 			cp = addr->sa_data;
2061 			cp2 = ifa->ifa_addr->sa_data;
2062 			cp3 = ifa->ifa_netmask->sa_data;
2063 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2064 			for (; cp3 < cplim; cp3++)
2065 				if ((*cp++ ^ *cp2++) & *cp3)
2066 					break;
2067 			if (cp3 == cplim)
2068 				goto done;
2069 		}
2070 	}
2071 	ifa = ifa_maybe;
2072 done:
2073 	return (ifa);
2074 }
2075 
2076 /*
2077  * See whether new ifa is better than current one:
2078  * 1) A non-virtual one is preferred over virtual.
2079  * 2) A virtual in master state preferred over any other state.
2080  *
2081  * Used in several address selecting functions.
2082  */
2083 int
2084 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2085 {
2086 
2087 	return (cur->ifa_carp && (!next->ifa_carp ||
2088 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2089 }
2090 
2091 struct sockaddr_dl *
2092 link_alloc_sdl(size_t size, int flags)
2093 {
2094 
2095 	return (malloc(size, M_TEMP, flags));
2096 }
2097 
2098 void
2099 link_free_sdl(struct sockaddr *sa)
2100 {
2101 	free(sa, M_TEMP);
2102 }
2103 
2104 /*
2105  * Fills in given sdl with interface basic info.
2106  * Returns pointer to filled sdl.
2107  */
2108 struct sockaddr_dl *
2109 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2110 {
2111 	struct sockaddr_dl *sdl;
2112 
2113 	sdl = (struct sockaddr_dl *)paddr;
2114 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2115 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2116 	sdl->sdl_family = AF_LINK;
2117 	sdl->sdl_index = ifp->if_index;
2118 	sdl->sdl_type = iftype;
2119 
2120 	return (sdl);
2121 }
2122 
2123 /*
2124  * Mark an interface down and notify protocols of
2125  * the transition.
2126  */
2127 static void
2128 if_unroute(struct ifnet *ifp, int flag, int fam)
2129 {
2130 
2131 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2132 
2133 	ifp->if_flags &= ~flag;
2134 	getmicrotime(&ifp->if_lastchange);
2135 	ifp->if_qflush(ifp);
2136 
2137 	if (ifp->if_carp)
2138 		(*carp_linkstate_p)(ifp);
2139 	rt_ifmsg(ifp);
2140 }
2141 
2142 /*
2143  * Mark an interface up and notify protocols of
2144  * the transition.
2145  */
2146 static void
2147 if_route(struct ifnet *ifp, int flag, int fam)
2148 {
2149 
2150 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2151 
2152 	ifp->if_flags |= flag;
2153 	getmicrotime(&ifp->if_lastchange);
2154 	if (ifp->if_carp)
2155 		(*carp_linkstate_p)(ifp);
2156 	rt_ifmsg(ifp);
2157 #ifdef INET6
2158 	in6_if_up(ifp);
2159 #endif
2160 }
2161 
2162 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2163 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2164 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2165 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2166 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2167 int	(*vlan_pcp_p)(struct ifnet *, uint16_t *);
2168 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2169 void	*(*vlan_cookie_p)(struct ifnet *);
2170 
2171 /*
2172  * Handle a change in the interface link state. To avoid LORs
2173  * between driver lock and upper layer locks, as well as possible
2174  * recursions, we post event to taskqueue, and all job
2175  * is done in static do_link_state_change().
2176  */
2177 void
2178 if_link_state_change(struct ifnet *ifp, int link_state)
2179 {
2180 	/* Return if state hasn't changed. */
2181 	if (ifp->if_link_state == link_state)
2182 		return;
2183 
2184 	ifp->if_link_state = link_state;
2185 
2186 	/* XXXGL: reference ifp? */
2187 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2188 }
2189 
2190 static void
2191 do_link_state_change(void *arg, int pending)
2192 {
2193 	struct ifnet *ifp;
2194 	int link_state;
2195 
2196 	ifp = arg;
2197 	link_state = ifp->if_link_state;
2198 
2199 	CURVNET_SET(ifp->if_vnet);
2200 	rt_ifmsg(ifp);
2201 	if (ifp->if_vlantrunk != NULL)
2202 		(*vlan_link_state_p)(ifp);
2203 
2204 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2205 	    ifp->if_l2com != NULL)
2206 		(*ng_ether_link_state_p)(ifp, link_state);
2207 	if (ifp->if_carp)
2208 		(*carp_linkstate_p)(ifp);
2209 	if (ifp->if_bridge)
2210 		ifp->if_bridge_linkstate(ifp);
2211 	if (ifp->if_lagg)
2212 		(*lagg_linkstate_p)(ifp, link_state);
2213 
2214 	if (IS_DEFAULT_VNET(curvnet))
2215 		devctl_notify("IFNET", ifp->if_xname,
2216 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2217 		    NULL);
2218 	if (pending > 1)
2219 		if_printf(ifp, "%d link states coalesced\n", pending);
2220 	if (log_link_state_change)
2221 		if_printf(ifp, "link state changed to %s\n",
2222 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2223 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2224 	CURVNET_RESTORE();
2225 }
2226 
2227 /*
2228  * Mark an interface down and notify protocols of
2229  * the transition.
2230  */
2231 void
2232 if_down(struct ifnet *ifp)
2233 {
2234 
2235 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2236 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2237 }
2238 
2239 /*
2240  * Mark an interface up and notify protocols of
2241  * the transition.
2242  */
2243 void
2244 if_up(struct ifnet *ifp)
2245 {
2246 
2247 	if_route(ifp, IFF_UP, AF_UNSPEC);
2248 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2249 }
2250 
2251 /*
2252  * Flush an interface queue.
2253  */
2254 void
2255 if_qflush(struct ifnet *ifp)
2256 {
2257 	struct mbuf *m, *n;
2258 	struct ifaltq *ifq;
2259 
2260 	ifq = &ifp->if_snd;
2261 	IFQ_LOCK(ifq);
2262 #ifdef ALTQ
2263 	if (ALTQ_IS_ENABLED(ifq))
2264 		ALTQ_PURGE(ifq);
2265 #endif
2266 	n = ifq->ifq_head;
2267 	while ((m = n) != NULL) {
2268 		n = m->m_nextpkt;
2269 		m_freem(m);
2270 	}
2271 	ifq->ifq_head = 0;
2272 	ifq->ifq_tail = 0;
2273 	ifq->ifq_len = 0;
2274 	IFQ_UNLOCK(ifq);
2275 }
2276 
2277 /*
2278  * Map interface name to interface structure pointer, with or without
2279  * returning a reference.
2280  */
2281 struct ifnet *
2282 ifunit_ref(const char *name)
2283 {
2284 	struct epoch_tracker et;
2285 	struct ifnet *ifp;
2286 
2287 	NET_EPOCH_ENTER(et);
2288 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2289 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2290 		    !(ifp->if_flags & IFF_DYING))
2291 			break;
2292 	}
2293 	if (ifp != NULL) {
2294 		if_ref(ifp);
2295 		MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
2296 	}
2297 
2298 	NET_EPOCH_EXIT(et);
2299 	return (ifp);
2300 }
2301 
2302 struct ifnet *
2303 ifunit(const char *name)
2304 {
2305 	struct epoch_tracker et;
2306 	struct ifnet *ifp;
2307 
2308 	NET_EPOCH_ENTER(et);
2309 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2310 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2311 			break;
2312 	}
2313 	NET_EPOCH_EXIT(et);
2314 	return (ifp);
2315 }
2316 
2317 void *
2318 ifr_buffer_get_buffer(void *data)
2319 {
2320 	union ifreq_union *ifrup;
2321 
2322 	ifrup = data;
2323 #ifdef COMPAT_FREEBSD32
2324 	if (SV_CURPROC_FLAG(SV_ILP32))
2325 		return ((void *)(uintptr_t)
2326 		    ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2327 #endif
2328 	return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2329 }
2330 
2331 static void
2332 ifr_buffer_set_buffer_null(void *data)
2333 {
2334 	union ifreq_union *ifrup;
2335 
2336 	ifrup = data;
2337 #ifdef COMPAT_FREEBSD32
2338 	if (SV_CURPROC_FLAG(SV_ILP32))
2339 		ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2340 	else
2341 #endif
2342 		ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2343 }
2344 
2345 size_t
2346 ifr_buffer_get_length(void *data)
2347 {
2348 	union ifreq_union *ifrup;
2349 
2350 	ifrup = data;
2351 #ifdef COMPAT_FREEBSD32
2352 	if (SV_CURPROC_FLAG(SV_ILP32))
2353 		return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2354 #endif
2355 	return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2356 }
2357 
2358 static void
2359 ifr_buffer_set_length(void *data, size_t len)
2360 {
2361 	union ifreq_union *ifrup;
2362 
2363 	ifrup = data;
2364 #ifdef COMPAT_FREEBSD32
2365 	if (SV_CURPROC_FLAG(SV_ILP32))
2366 		ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2367 	else
2368 #endif
2369 		ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2370 }
2371 
2372 void *
2373 ifr_data_get_ptr(void *ifrp)
2374 {
2375 	union ifreq_union *ifrup;
2376 
2377 	ifrup = ifrp;
2378 #ifdef COMPAT_FREEBSD32
2379 	if (SV_CURPROC_FLAG(SV_ILP32))
2380 		return ((void *)(uintptr_t)
2381 		    ifrup->ifr32.ifr_ifru.ifru_data);
2382 #endif
2383 		return (ifrup->ifr.ifr_ifru.ifru_data);
2384 }
2385 
2386 struct ifcap_nv_bit_name {
2387 	int cap_bit;
2388 	const char *cap_name;
2389 };
2390 #define CAPNV(x) {.cap_bit = IFCAP_##x, \
2391     .cap_name = __CONCAT(IFCAP_, __CONCAT(x, _NAME)) }
2392 const struct ifcap_nv_bit_name ifcap_nv_bit_names[] = {
2393 	CAPNV(RXCSUM),
2394 	CAPNV(TXCSUM),
2395 	CAPNV(NETCONS),
2396 	CAPNV(VLAN_MTU),
2397 	CAPNV(VLAN_HWTAGGING),
2398 	CAPNV(JUMBO_MTU),
2399 	CAPNV(POLLING),
2400 	CAPNV(VLAN_HWCSUM),
2401 	CAPNV(TSO4),
2402 	CAPNV(TSO6),
2403 	CAPNV(LRO),
2404 	CAPNV(WOL_UCAST),
2405 	CAPNV(WOL_MCAST),
2406 	CAPNV(WOL_MAGIC),
2407 	CAPNV(TOE4),
2408 	CAPNV(TOE6),
2409 	CAPNV(VLAN_HWFILTER),
2410 	CAPNV(VLAN_HWTSO),
2411 	CAPNV(LINKSTATE),
2412 	CAPNV(NETMAP),
2413 	CAPNV(RXCSUM_IPV6),
2414 	CAPNV(TXCSUM_IPV6),
2415 	CAPNV(HWSTATS),
2416 	CAPNV(TXRTLMT),
2417 	CAPNV(HWRXTSTMP),
2418 	CAPNV(MEXTPG),
2419 	CAPNV(TXTLS4),
2420 	CAPNV(TXTLS6),
2421 	CAPNV(VXLAN_HWCSUM),
2422 	CAPNV(VXLAN_HWTSO),
2423 	CAPNV(TXTLS_RTLMT),
2424 	{0, NULL}
2425 };
2426 #define CAP2NV(x) {.cap_bit = IFCAP2_##x, \
2427     .cap_name = __CONCAT(IFCAP2_, __CONCAT(x, _NAME)) }
2428 const struct ifcap_nv_bit_name ifcap2_nv_bit_names[] = {
2429 	CAP2NV(RXTLS4),
2430 	CAP2NV(RXTLS6),
2431 	{0, NULL}
2432 };
2433 #undef CAPNV
2434 #undef CAP2NV
2435 
2436 int
2437 if_capnv_to_capint(const nvlist_t *nv, int *old_cap,
2438     const struct ifcap_nv_bit_name *nn, bool all)
2439 {
2440 	int i, res;
2441 
2442 	res = 0;
2443 	for (i = 0; nn[i].cap_name != NULL; i++) {
2444 		if (nvlist_exists_bool(nv, nn[i].cap_name)) {
2445 			if (all || nvlist_get_bool(nv, nn[i].cap_name))
2446 				res |= nn[i].cap_bit;
2447 		} else {
2448 			res |= *old_cap & nn[i].cap_bit;
2449 		}
2450 	}
2451 	return (res);
2452 }
2453 
2454 void
2455 if_capint_to_capnv(nvlist_t *nv, const struct ifcap_nv_bit_name *nn,
2456     int ifr_cap, int ifr_req)
2457 {
2458 	int i;
2459 
2460 	for (i = 0; nn[i].cap_name != NULL; i++) {
2461 		if ((nn[i].cap_bit & ifr_cap) != 0) {
2462 			nvlist_add_bool(nv, nn[i].cap_name,
2463 			    (nn[i].cap_bit & ifr_req) != 0);
2464 		}
2465 	}
2466 }
2467 
2468 /*
2469  * Hardware specific interface ioctls.
2470  */
2471 int
2472 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2473 {
2474 	struct ifreq *ifr;
2475 	int error = 0, do_ifup = 0;
2476 	int new_flags, temp_flags;
2477 	size_t namelen, onamelen;
2478 	size_t descrlen, nvbuflen;
2479 	char *descrbuf, *odescrbuf;
2480 	char new_name[IFNAMSIZ];
2481 	char old_name[IFNAMSIZ], strbuf[IFNAMSIZ + 8];
2482 	struct ifaddr *ifa;
2483 	struct sockaddr_dl *sdl;
2484 	void *buf;
2485 	nvlist_t *nvcap;
2486 	struct siocsifcapnv_driver_data drv_ioctl_data;
2487 
2488 	ifr = (struct ifreq *)data;
2489 	switch (cmd) {
2490 	case SIOCGIFINDEX:
2491 		ifr->ifr_index = ifp->if_index;
2492 		break;
2493 
2494 	case SIOCGIFFLAGS:
2495 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2496 		ifr->ifr_flags = temp_flags & 0xffff;
2497 		ifr->ifr_flagshigh = temp_flags >> 16;
2498 		break;
2499 
2500 	case SIOCGIFCAP:
2501 		ifr->ifr_reqcap = ifp->if_capabilities;
2502 		ifr->ifr_curcap = ifp->if_capenable;
2503 		break;
2504 
2505 	case SIOCGIFCAPNV:
2506 		if ((ifp->if_capabilities & IFCAP_NV) == 0) {
2507 			error = EINVAL;
2508 			break;
2509 		}
2510 		buf = NULL;
2511 		nvcap = nvlist_create(0);
2512 		for (;;) {
2513 			if_capint_to_capnv(nvcap, ifcap_nv_bit_names,
2514 			    ifp->if_capabilities, ifp->if_capenable);
2515 			if_capint_to_capnv(nvcap, ifcap2_nv_bit_names,
2516 			    ifp->if_capabilities2, ifp->if_capenable2);
2517 			error = (*ifp->if_ioctl)(ifp, SIOCGIFCAPNV,
2518 			    __DECONST(caddr_t, nvcap));
2519 			if (error != 0) {
2520 				if_printf(ifp,
2521 			    "SIOCGIFCAPNV driver mistake: nvlist error %d\n",
2522 				    error);
2523 				break;
2524 			}
2525 			buf = nvlist_pack(nvcap, &nvbuflen);
2526 			if (buf == NULL) {
2527 				error = nvlist_error(nvcap);
2528 				if (error == 0)
2529 					error = EDOOFUS;
2530 				break;
2531 			}
2532 			if (nvbuflen > ifr->ifr_cap_nv.buf_length) {
2533 				ifr->ifr_cap_nv.length = nvbuflen;
2534 				ifr->ifr_cap_nv.buffer = NULL;
2535 				error = EFBIG;
2536 				break;
2537 			}
2538 			ifr->ifr_cap_nv.length = nvbuflen;
2539 			error = copyout(buf, ifr->ifr_cap_nv.buffer, nvbuflen);
2540 			break;
2541 		}
2542 		free(buf, M_NVLIST);
2543 		nvlist_destroy(nvcap);
2544 		break;
2545 
2546 	case SIOCGIFDATA:
2547 	{
2548 		struct if_data ifd;
2549 
2550 		/* Ensure uninitialised padding is not leaked. */
2551 		memset(&ifd, 0, sizeof(ifd));
2552 
2553 		if_data_copy(ifp, &ifd);
2554 		error = copyout(&ifd, ifr_data_get_ptr(ifr), sizeof(ifd));
2555 		break;
2556 	}
2557 
2558 #ifdef MAC
2559 	case SIOCGIFMAC:
2560 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2561 		break;
2562 #endif
2563 
2564 	case SIOCGIFMETRIC:
2565 		ifr->ifr_metric = ifp->if_metric;
2566 		break;
2567 
2568 	case SIOCGIFMTU:
2569 		ifr->ifr_mtu = ifp->if_mtu;
2570 		break;
2571 
2572 	case SIOCGIFPHYS:
2573 		/* XXXGL: did this ever worked? */
2574 		ifr->ifr_phys = 0;
2575 		break;
2576 
2577 	case SIOCGIFDESCR:
2578 		error = 0;
2579 		sx_slock(&ifdescr_sx);
2580 		if (ifp->if_description == NULL)
2581 			error = ENOMSG;
2582 		else {
2583 			/* space for terminating nul */
2584 			descrlen = strlen(ifp->if_description) + 1;
2585 			if (ifr_buffer_get_length(ifr) < descrlen)
2586 				ifr_buffer_set_buffer_null(ifr);
2587 			else
2588 				error = copyout(ifp->if_description,
2589 				    ifr_buffer_get_buffer(ifr), descrlen);
2590 			ifr_buffer_set_length(ifr, descrlen);
2591 		}
2592 		sx_sunlock(&ifdescr_sx);
2593 		break;
2594 
2595 	case SIOCSIFDESCR:
2596 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2597 		if (error)
2598 			return (error);
2599 
2600 		/*
2601 		 * Copy only (length-1) bytes to make sure that
2602 		 * if_description is always nul terminated.  The
2603 		 * length parameter is supposed to count the
2604 		 * terminating nul in.
2605 		 */
2606 		if (ifr_buffer_get_length(ifr) > ifdescr_maxlen)
2607 			return (ENAMETOOLONG);
2608 		else if (ifr_buffer_get_length(ifr) == 0)
2609 			descrbuf = NULL;
2610 		else {
2611 			descrbuf = malloc(ifr_buffer_get_length(ifr),
2612 			    M_IFDESCR, M_WAITOK | M_ZERO);
2613 			error = copyin(ifr_buffer_get_buffer(ifr), descrbuf,
2614 			    ifr_buffer_get_length(ifr) - 1);
2615 			if (error) {
2616 				free(descrbuf, M_IFDESCR);
2617 				break;
2618 			}
2619 		}
2620 
2621 		sx_xlock(&ifdescr_sx);
2622 		odescrbuf = ifp->if_description;
2623 		ifp->if_description = descrbuf;
2624 		sx_xunlock(&ifdescr_sx);
2625 
2626 		getmicrotime(&ifp->if_lastchange);
2627 		free(odescrbuf, M_IFDESCR);
2628 		break;
2629 
2630 	case SIOCGIFFIB:
2631 		ifr->ifr_fib = ifp->if_fib;
2632 		break;
2633 
2634 	case SIOCSIFFIB:
2635 		error = priv_check(td, PRIV_NET_SETIFFIB);
2636 		if (error)
2637 			return (error);
2638 		if (ifr->ifr_fib >= rt_numfibs)
2639 			return (EINVAL);
2640 
2641 		ifp->if_fib = ifr->ifr_fib;
2642 		break;
2643 
2644 	case SIOCSIFFLAGS:
2645 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2646 		if (error)
2647 			return (error);
2648 		/*
2649 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2650 		 * check, so we don't need special handling here yet.
2651 		 */
2652 		new_flags = (ifr->ifr_flags & 0xffff) |
2653 		    (ifr->ifr_flagshigh << 16);
2654 		if (ifp->if_flags & IFF_UP &&
2655 		    (new_flags & IFF_UP) == 0) {
2656 			if_down(ifp);
2657 		} else if (new_flags & IFF_UP &&
2658 		    (ifp->if_flags & IFF_UP) == 0) {
2659 			do_ifup = 1;
2660 		}
2661 		/* See if permanently promiscuous mode bit is about to flip */
2662 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2663 			if (new_flags & IFF_PPROMISC)
2664 				ifp->if_flags |= IFF_PROMISC;
2665 			else if (ifp->if_pcount == 0)
2666 				ifp->if_flags &= ~IFF_PROMISC;
2667 			if (log_promisc_mode_change)
2668                                 if_printf(ifp, "permanently promiscuous mode %s\n",
2669                                     ((new_flags & IFF_PPROMISC) ?
2670                                      "enabled" : "disabled"));
2671 		}
2672 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2673 			(new_flags &~ IFF_CANTCHANGE);
2674 		if (ifp->if_ioctl) {
2675 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2676 		}
2677 		if (do_ifup)
2678 			if_up(ifp);
2679 		getmicrotime(&ifp->if_lastchange);
2680 		break;
2681 
2682 	case SIOCSIFCAP:
2683 		error = priv_check(td, PRIV_NET_SETIFCAP);
2684 		if (error != 0)
2685 			return (error);
2686 		if (ifp->if_ioctl == NULL)
2687 			return (EOPNOTSUPP);
2688 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2689 			return (EINVAL);
2690 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2691 		if (error == 0)
2692 			getmicrotime(&ifp->if_lastchange);
2693 		break;
2694 
2695 	case SIOCSIFCAPNV:
2696 		error = priv_check(td, PRIV_NET_SETIFCAP);
2697 		if (error != 0)
2698 			return (error);
2699 		if (ifp->if_ioctl == NULL)
2700 			return (EOPNOTSUPP);
2701 		if ((ifp->if_capabilities & IFCAP_NV) == 0)
2702 			return (EINVAL);
2703 		if (ifr->ifr_cap_nv.length > IFR_CAP_NV_MAXBUFSIZE)
2704 			return (EINVAL);
2705 		nvcap = NULL;
2706 		buf = malloc(ifr->ifr_cap_nv.length, M_TEMP, M_WAITOK);
2707 		for (;;) {
2708 			error = copyin(ifr->ifr_cap_nv.buffer, buf,
2709 			    ifr->ifr_cap_nv.length);
2710 			if (error != 0)
2711 				break;
2712 			nvcap = nvlist_unpack(buf, ifr->ifr_cap_nv.length, 0);
2713 			if (nvcap == NULL) {
2714 				error = EINVAL;
2715 				break;
2716 			}
2717 			drv_ioctl_data.reqcap = if_capnv_to_capint(nvcap,
2718 			    &ifp->if_capenable, ifcap_nv_bit_names, false);
2719 			if ((drv_ioctl_data.reqcap &
2720 			    ~ifp->if_capabilities) != 0) {
2721 				error = EINVAL;
2722 				break;
2723 			}
2724 			drv_ioctl_data.reqcap2 = if_capnv_to_capint(nvcap,
2725 			    &ifp->if_capenable2, ifcap2_nv_bit_names, false);
2726 			if ((drv_ioctl_data.reqcap2 &
2727 			    ~ifp->if_capabilities2) != 0) {
2728 				error = EINVAL;
2729 				break;
2730 			}
2731 			drv_ioctl_data.nvcap = nvcap;
2732 			error = (*ifp->if_ioctl)(ifp, SIOCSIFCAPNV,
2733 			    (caddr_t)&drv_ioctl_data);
2734 			break;
2735 		}
2736 		nvlist_destroy(nvcap);
2737 		free(buf, M_TEMP);
2738 		if (error == 0)
2739 			getmicrotime(&ifp->if_lastchange);
2740 		break;
2741 
2742 #ifdef MAC
2743 	case SIOCSIFMAC:
2744 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2745 		break;
2746 #endif
2747 
2748 	case SIOCSIFNAME:
2749 		error = priv_check(td, PRIV_NET_SETIFNAME);
2750 		if (error)
2751 			return (error);
2752 		error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ,
2753 		    NULL);
2754 		if (error != 0)
2755 			return (error);
2756 		if (new_name[0] == '\0')
2757 			return (EINVAL);
2758 		if (strcmp(new_name, ifp->if_xname) == 0)
2759 			break;
2760 		if (ifunit(new_name) != NULL)
2761 			return (EEXIST);
2762 
2763 		/*
2764 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2765 		 * and there are numerous other races with the
2766 		 * ifunit() checks not being atomic with namespace
2767 		 * changes (renames, vmoves, if_attach, etc).
2768 		 */
2769 		ifp->if_flags |= IFF_RENAMING;
2770 
2771 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2772 
2773 		if_printf(ifp, "changing name to '%s'\n", new_name);
2774 
2775 		IF_ADDR_WLOCK(ifp);
2776 		strlcpy(old_name, ifp->if_xname, sizeof(old_name));
2777 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2778 		ifa = ifp->if_addr;
2779 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2780 		namelen = strlen(new_name);
2781 		onamelen = sdl->sdl_nlen;
2782 		/*
2783 		 * Move the address if needed.  This is safe because we
2784 		 * allocate space for a name of length IFNAMSIZ when we
2785 		 * create this in if_attach().
2786 		 */
2787 		if (namelen != onamelen) {
2788 			bcopy(sdl->sdl_data + onamelen,
2789 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2790 		}
2791 		bcopy(new_name, sdl->sdl_data, namelen);
2792 		sdl->sdl_nlen = namelen;
2793 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2794 		bzero(sdl->sdl_data, onamelen);
2795 		while (namelen != 0)
2796 			sdl->sdl_data[--namelen] = 0xff;
2797 		IF_ADDR_WUNLOCK(ifp);
2798 
2799 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2800 
2801 		ifp->if_flags &= ~IFF_RENAMING;
2802 
2803 		snprintf(strbuf, sizeof(strbuf), "name=%s", new_name);
2804 		devctl_notify("IFNET", old_name, "RENAME", strbuf);
2805 		break;
2806 
2807 #ifdef VIMAGE
2808 	case SIOCSIFVNET:
2809 		error = priv_check(td, PRIV_NET_SETIFVNET);
2810 		if (error)
2811 			return (error);
2812 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2813 		break;
2814 #endif
2815 
2816 	case SIOCSIFMETRIC:
2817 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2818 		if (error)
2819 			return (error);
2820 		ifp->if_metric = ifr->ifr_metric;
2821 		getmicrotime(&ifp->if_lastchange);
2822 		break;
2823 
2824 	case SIOCSIFPHYS:
2825 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2826 		if (error)
2827 			return (error);
2828 		if (ifp->if_ioctl == NULL)
2829 			return (EOPNOTSUPP);
2830 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2831 		if (error == 0)
2832 			getmicrotime(&ifp->if_lastchange);
2833 		break;
2834 
2835 	case SIOCSIFMTU:
2836 	{
2837 		u_long oldmtu = ifp->if_mtu;
2838 
2839 		error = priv_check(td, PRIV_NET_SETIFMTU);
2840 		if (error)
2841 			return (error);
2842 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2843 			return (EINVAL);
2844 		if (ifp->if_ioctl == NULL)
2845 			return (EOPNOTSUPP);
2846 		/* Disallow MTU changes on bridge member interfaces. */
2847 		if (ifp->if_bridge)
2848 			return (EOPNOTSUPP);
2849 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2850 		if (error == 0) {
2851 			getmicrotime(&ifp->if_lastchange);
2852 			rt_ifmsg(ifp);
2853 #ifdef INET
2854 			DEBUGNET_NOTIFY_MTU(ifp);
2855 #endif
2856 		}
2857 		/*
2858 		 * If the link MTU changed, do network layer specific procedure.
2859 		 */
2860 		if (ifp->if_mtu != oldmtu) {
2861 #ifdef INET6
2862 			nd6_setmtu(ifp);
2863 #endif
2864 			rt_updatemtu(ifp);
2865 		}
2866 		break;
2867 	}
2868 
2869 	case SIOCADDMULTI:
2870 	case SIOCDELMULTI:
2871 		if (cmd == SIOCADDMULTI)
2872 			error = priv_check(td, PRIV_NET_ADDMULTI);
2873 		else
2874 			error = priv_check(td, PRIV_NET_DELMULTI);
2875 		if (error)
2876 			return (error);
2877 
2878 		/* Don't allow group membership on non-multicast interfaces. */
2879 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2880 			return (EOPNOTSUPP);
2881 
2882 		/* Don't let users screw up protocols' entries. */
2883 		if (ifr->ifr_addr.sa_family != AF_LINK)
2884 			return (EINVAL);
2885 
2886 		if (cmd == SIOCADDMULTI) {
2887 			struct epoch_tracker et;
2888 			struct ifmultiaddr *ifma;
2889 
2890 			/*
2891 			 * Userland is only permitted to join groups once
2892 			 * via the if_addmulti() KPI, because it cannot hold
2893 			 * struct ifmultiaddr * between calls. It may also
2894 			 * lose a race while we check if the membership
2895 			 * already exists.
2896 			 */
2897 			NET_EPOCH_ENTER(et);
2898 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2899 			NET_EPOCH_EXIT(et);
2900 			if (ifma != NULL)
2901 				error = EADDRINUSE;
2902 			else
2903 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2904 		} else {
2905 			error = if_delmulti(ifp, &ifr->ifr_addr);
2906 		}
2907 		if (error == 0)
2908 			getmicrotime(&ifp->if_lastchange);
2909 		break;
2910 
2911 	case SIOCSIFPHYADDR:
2912 	case SIOCDIFPHYADDR:
2913 #ifdef INET6
2914 	case SIOCSIFPHYADDR_IN6:
2915 #endif
2916 	case SIOCSIFMEDIA:
2917 	case SIOCSIFGENERIC:
2918 		error = priv_check(td, PRIV_NET_HWIOCTL);
2919 		if (error)
2920 			return (error);
2921 		if (ifp->if_ioctl == NULL)
2922 			return (EOPNOTSUPP);
2923 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2924 		if (error == 0)
2925 			getmicrotime(&ifp->if_lastchange);
2926 		break;
2927 
2928 	case SIOCGIFSTATUS:
2929 	case SIOCGIFPSRCADDR:
2930 	case SIOCGIFPDSTADDR:
2931 	case SIOCGIFMEDIA:
2932 	case SIOCGIFXMEDIA:
2933 	case SIOCGIFGENERIC:
2934 	case SIOCGIFRSSKEY:
2935 	case SIOCGIFRSSHASH:
2936 	case SIOCGIFDOWNREASON:
2937 		if (ifp->if_ioctl == NULL)
2938 			return (EOPNOTSUPP);
2939 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2940 		break;
2941 
2942 	case SIOCSIFLLADDR:
2943 		error = priv_check(td, PRIV_NET_SETLLADDR);
2944 		if (error)
2945 			return (error);
2946 		error = if_setlladdr(ifp,
2947 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2948 		break;
2949 
2950 	case SIOCGHWADDR:
2951 		error = if_gethwaddr(ifp, ifr);
2952 		break;
2953 
2954 	case SIOCAIFGROUP:
2955 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2956 		if (error)
2957 			return (error);
2958 		error = if_addgroup(ifp,
2959 		    ((struct ifgroupreq *)data)->ifgr_group);
2960 		if (error != 0)
2961 			return (error);
2962 		break;
2963 
2964 	case SIOCGIFGROUP:
2965 	{
2966 		struct epoch_tracker et;
2967 
2968 		NET_EPOCH_ENTER(et);
2969 		error = if_getgroup((struct ifgroupreq *)data, ifp);
2970 		NET_EPOCH_EXIT(et);
2971 		break;
2972 	}
2973 
2974 	case SIOCDIFGROUP:
2975 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2976 		if (error)
2977 			return (error);
2978 		error = if_delgroup(ifp,
2979 		    ((struct ifgroupreq *)data)->ifgr_group);
2980 		if (error != 0)
2981 			return (error);
2982 		break;
2983 
2984 	default:
2985 		error = ENOIOCTL;
2986 		break;
2987 	}
2988 	return (error);
2989 }
2990 
2991 /*
2992  * Interface ioctls.
2993  */
2994 int
2995 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2996 {
2997 #ifdef COMPAT_FREEBSD32
2998 	union {
2999 		struct ifconf ifc;
3000 		struct ifdrv ifd;
3001 		struct ifgroupreq ifgr;
3002 		struct ifmediareq ifmr;
3003 	} thunk;
3004 	u_long saved_cmd;
3005 	struct ifconf32 *ifc32;
3006 	struct ifdrv32 *ifd32;
3007 	struct ifgroupreq32 *ifgr32;
3008 	struct ifmediareq32 *ifmr32;
3009 #endif
3010 	struct ifnet *ifp;
3011 	struct ifreq *ifr;
3012 	int error;
3013 	int oif_flags;
3014 #ifdef VIMAGE
3015 	bool shutdown;
3016 #endif
3017 
3018 	CURVNET_SET(so->so_vnet);
3019 #ifdef VIMAGE
3020 	/* Make sure the VNET is stable. */
3021 	shutdown = VNET_IS_SHUTTING_DOWN(so->so_vnet);
3022 	if (shutdown) {
3023 		CURVNET_RESTORE();
3024 		return (EBUSY);
3025 	}
3026 #endif
3027 
3028 #ifdef COMPAT_FREEBSD32
3029 	saved_cmd = cmd;
3030 	switch (cmd) {
3031 	case SIOCGIFCONF32:
3032 		ifc32 = (struct ifconf32 *)data;
3033 		thunk.ifc.ifc_len = ifc32->ifc_len;
3034 		thunk.ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
3035 		data = (caddr_t)&thunk.ifc;
3036 		cmd = SIOCGIFCONF;
3037 		break;
3038 	case SIOCGDRVSPEC32:
3039 	case SIOCSDRVSPEC32:
3040 		ifd32 = (struct ifdrv32 *)data;
3041 		memcpy(thunk.ifd.ifd_name, ifd32->ifd_name,
3042 		    sizeof(thunk.ifd.ifd_name));
3043 		thunk.ifd.ifd_cmd = ifd32->ifd_cmd;
3044 		thunk.ifd.ifd_len = ifd32->ifd_len;
3045 		thunk.ifd.ifd_data = PTRIN(ifd32->ifd_data);
3046 		data = (caddr_t)&thunk.ifd;
3047 		cmd = _IOC_NEWTYPE(cmd, struct ifdrv);
3048 		break;
3049 	case SIOCAIFGROUP32:
3050 	case SIOCGIFGROUP32:
3051 	case SIOCDIFGROUP32:
3052 	case SIOCGIFGMEMB32:
3053 		ifgr32 = (struct ifgroupreq32 *)data;
3054 		memcpy(thunk.ifgr.ifgr_name, ifgr32->ifgr_name,
3055 		    sizeof(thunk.ifgr.ifgr_name));
3056 		thunk.ifgr.ifgr_len = ifgr32->ifgr_len;
3057 		switch (cmd) {
3058 		case SIOCAIFGROUP32:
3059 		case SIOCDIFGROUP32:
3060 			memcpy(thunk.ifgr.ifgr_group, ifgr32->ifgr_group,
3061 			    sizeof(thunk.ifgr.ifgr_group));
3062 			break;
3063 		case SIOCGIFGROUP32:
3064 		case SIOCGIFGMEMB32:
3065 			thunk.ifgr.ifgr_groups = PTRIN(ifgr32->ifgr_groups);
3066 			break;
3067 		}
3068 		data = (caddr_t)&thunk.ifgr;
3069 		cmd = _IOC_NEWTYPE(cmd, struct ifgroupreq);
3070 		break;
3071 	case SIOCGIFMEDIA32:
3072 	case SIOCGIFXMEDIA32:
3073 		ifmr32 = (struct ifmediareq32 *)data;
3074 		memcpy(thunk.ifmr.ifm_name, ifmr32->ifm_name,
3075 		    sizeof(thunk.ifmr.ifm_name));
3076 		thunk.ifmr.ifm_current = ifmr32->ifm_current;
3077 		thunk.ifmr.ifm_mask = ifmr32->ifm_mask;
3078 		thunk.ifmr.ifm_status = ifmr32->ifm_status;
3079 		thunk.ifmr.ifm_active = ifmr32->ifm_active;
3080 		thunk.ifmr.ifm_count = ifmr32->ifm_count;
3081 		thunk.ifmr.ifm_ulist = PTRIN(ifmr32->ifm_ulist);
3082 		data = (caddr_t)&thunk.ifmr;
3083 		cmd = _IOC_NEWTYPE(cmd, struct ifmediareq);
3084 		break;
3085 	}
3086 #endif
3087 
3088 	switch (cmd) {
3089 	case SIOCGIFCONF:
3090 		error = ifconf(cmd, data);
3091 		goto out_noref;
3092 	}
3093 
3094 	ifr = (struct ifreq *)data;
3095 	switch (cmd) {
3096 #ifdef VIMAGE
3097 	case SIOCSIFRVNET:
3098 		error = priv_check(td, PRIV_NET_SETIFVNET);
3099 		if (error == 0)
3100 			error = if_vmove_reclaim(td, ifr->ifr_name,
3101 			    ifr->ifr_jid);
3102 		goto out_noref;
3103 #endif
3104 	case SIOCIFCREATE:
3105 	case SIOCIFCREATE2:
3106 		error = priv_check(td, PRIV_NET_IFCREATE);
3107 		if (error == 0)
3108 			error = if_clone_create(ifr->ifr_name,
3109 			    sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ?
3110 			    ifr_data_get_ptr(ifr) : NULL);
3111 		goto out_noref;
3112 	case SIOCIFDESTROY:
3113 		error = priv_check(td, PRIV_NET_IFDESTROY);
3114 
3115 		if (error == 0) {
3116 			sx_xlock(&ifnet_detach_sxlock);
3117 			error = if_clone_destroy(ifr->ifr_name);
3118 			sx_xunlock(&ifnet_detach_sxlock);
3119 		}
3120 		goto out_noref;
3121 
3122 	case SIOCIFGCLONERS:
3123 		error = if_clone_list((struct if_clonereq *)data);
3124 		goto out_noref;
3125 
3126 	case SIOCGIFGMEMB:
3127 		error = if_getgroupmembers((struct ifgroupreq *)data);
3128 		goto out_noref;
3129 
3130 #if defined(INET) || defined(INET6)
3131 	case SIOCSVH:
3132 	case SIOCGVH:
3133 		if (carp_ioctl_p == NULL)
3134 			error = EPROTONOSUPPORT;
3135 		else
3136 			error = (*carp_ioctl_p)(ifr, cmd, td);
3137 		goto out_noref;
3138 #endif
3139 	}
3140 
3141 	ifp = ifunit_ref(ifr->ifr_name);
3142 	if (ifp == NULL) {
3143 		error = ENXIO;
3144 		goto out_noref;
3145 	}
3146 
3147 	error = ifhwioctl(cmd, ifp, data, td);
3148 	if (error != ENOIOCTL)
3149 		goto out_ref;
3150 
3151 	oif_flags = ifp->if_flags;
3152 	if (so->so_proto == NULL) {
3153 		error = EOPNOTSUPP;
3154 		goto out_ref;
3155 	}
3156 
3157 	/*
3158 	 * Pass the request on to the socket control method, and if the
3159 	 * latter returns EOPNOTSUPP, directly to the interface.
3160 	 *
3161 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
3162 	 * trust SIOCSIFADDR et al to come from an already privileged
3163 	 * layer, and do not perform any credentials checks or input
3164 	 * validation.
3165 	 */
3166 	error = so->so_proto->pr_control(so, cmd, data, ifp, td);
3167 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
3168 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
3169 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
3170 		error = (*ifp->if_ioctl)(ifp, cmd, data);
3171 
3172 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
3173 #ifdef INET6
3174 		if (ifp->if_flags & IFF_UP)
3175 			in6_if_up(ifp);
3176 #endif
3177 	}
3178 
3179 out_ref:
3180 	if_rele(ifp);
3181 out_noref:
3182 	CURVNET_RESTORE();
3183 #ifdef COMPAT_FREEBSD32
3184 	if (error != 0)
3185 		return (error);
3186 	switch (saved_cmd) {
3187 	case SIOCGIFCONF32:
3188 		ifc32->ifc_len = thunk.ifc.ifc_len;
3189 		break;
3190 	case SIOCGDRVSPEC32:
3191 		/*
3192 		 * SIOCGDRVSPEC is IOWR, but nothing actually touches
3193 		 * the struct so just assert that ifd_len (the only
3194 		 * field it might make sense to update) hasn't
3195 		 * changed.
3196 		 */
3197 		KASSERT(thunk.ifd.ifd_len == ifd32->ifd_len,
3198 		    ("ifd_len was updated %u -> %zu", ifd32->ifd_len,
3199 			thunk.ifd.ifd_len));
3200 		break;
3201 	case SIOCGIFGROUP32:
3202 	case SIOCGIFGMEMB32:
3203 		ifgr32->ifgr_len = thunk.ifgr.ifgr_len;
3204 		break;
3205 	case SIOCGIFMEDIA32:
3206 	case SIOCGIFXMEDIA32:
3207 		ifmr32->ifm_current = thunk.ifmr.ifm_current;
3208 		ifmr32->ifm_mask = thunk.ifmr.ifm_mask;
3209 		ifmr32->ifm_status = thunk.ifmr.ifm_status;
3210 		ifmr32->ifm_active = thunk.ifmr.ifm_active;
3211 		ifmr32->ifm_count = thunk.ifmr.ifm_count;
3212 		break;
3213 	}
3214 #endif
3215 	return (error);
3216 }
3217 
3218 /*
3219  * The code common to handling reference counted flags,
3220  * e.g., in ifpromisc() and if_allmulti().
3221  * The "pflag" argument can specify a permanent mode flag to check,
3222  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
3223  *
3224  * Only to be used on stack-owned flags, not driver-owned flags.
3225  */
3226 static int
3227 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
3228 {
3229 	struct ifreq ifr;
3230 	int error;
3231 	int oldflags, oldcount;
3232 
3233 	/* Sanity checks to catch programming errors */
3234 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
3235 	    ("%s: setting driver-owned flag %d", __func__, flag));
3236 
3237 	if (onswitch)
3238 		KASSERT(*refcount >= 0,
3239 		    ("%s: increment negative refcount %d for flag %d",
3240 		    __func__, *refcount, flag));
3241 	else
3242 		KASSERT(*refcount > 0,
3243 		    ("%s: decrement non-positive refcount %d for flag %d",
3244 		    __func__, *refcount, flag));
3245 
3246 	/* In case this mode is permanent, just touch refcount */
3247 	if (ifp->if_flags & pflag) {
3248 		*refcount += onswitch ? 1 : -1;
3249 		return (0);
3250 	}
3251 
3252 	/* Save ifnet parameters for if_ioctl() may fail */
3253 	oldcount = *refcount;
3254 	oldflags = ifp->if_flags;
3255 
3256 	/*
3257 	 * See if we aren't the only and touching refcount is enough.
3258 	 * Actually toggle interface flag if we are the first or last.
3259 	 */
3260 	if (onswitch) {
3261 		if ((*refcount)++)
3262 			return (0);
3263 		ifp->if_flags |= flag;
3264 	} else {
3265 		if (--(*refcount))
3266 			return (0);
3267 		ifp->if_flags &= ~flag;
3268 	}
3269 
3270 	/* Call down the driver since we've changed interface flags */
3271 	if (ifp->if_ioctl == NULL) {
3272 		error = EOPNOTSUPP;
3273 		goto recover;
3274 	}
3275 	ifr.ifr_flags = ifp->if_flags & 0xffff;
3276 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
3277 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3278 	if (error)
3279 		goto recover;
3280 	/* Notify userland that interface flags have changed */
3281 	rt_ifmsg(ifp);
3282 	return (0);
3283 
3284 recover:
3285 	/* Recover after driver error */
3286 	*refcount = oldcount;
3287 	ifp->if_flags = oldflags;
3288 	return (error);
3289 }
3290 
3291 /*
3292  * Set/clear promiscuous mode on interface ifp based on the truth value
3293  * of pswitch.  The calls are reference counted so that only the first
3294  * "on" request actually has an effect, as does the final "off" request.
3295  * Results are undefined if the "off" and "on" requests are not matched.
3296  */
3297 int
3298 ifpromisc(struct ifnet *ifp, int pswitch)
3299 {
3300 	int error;
3301 	int oldflags = ifp->if_flags;
3302 
3303 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3304 			   &ifp->if_pcount, pswitch);
3305 	/* If promiscuous mode status has changed, log a message */
3306 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3307             log_promisc_mode_change)
3308 		if_printf(ifp, "promiscuous mode %s\n",
3309 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3310 	return (error);
3311 }
3312 
3313 /*
3314  * Return interface configuration
3315  * of system.  List may be used
3316  * in later ioctl's (above) to get
3317  * other information.
3318  */
3319 /*ARGSUSED*/
3320 static int
3321 ifconf(u_long cmd, caddr_t data)
3322 {
3323 	struct ifconf *ifc = (struct ifconf *)data;
3324 	struct ifnet *ifp;
3325 	struct ifaddr *ifa;
3326 	struct ifreq ifr;
3327 	struct sbuf *sb;
3328 	int error, full = 0, valid_len, max_len;
3329 
3330 	/* Limit initial buffer size to maxphys to avoid DoS from userspace. */
3331 	max_len = maxphys - 1;
3332 
3333 	/* Prevent hostile input from being able to crash the system */
3334 	if (ifc->ifc_len <= 0)
3335 		return (EINVAL);
3336 
3337 again:
3338 	if (ifc->ifc_len <= max_len) {
3339 		max_len = ifc->ifc_len;
3340 		full = 1;
3341 	}
3342 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3343 	max_len = 0;
3344 	valid_len = 0;
3345 
3346 	IFNET_RLOCK();
3347 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3348 		struct epoch_tracker et;
3349 		int addrs;
3350 
3351 		/*
3352 		 * Zero the ifr to make sure we don't disclose the contents
3353 		 * of the stack.
3354 		 */
3355 		memset(&ifr, 0, sizeof(ifr));
3356 
3357 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3358 		    >= sizeof(ifr.ifr_name)) {
3359 			sbuf_delete(sb);
3360 			IFNET_RUNLOCK();
3361 			return (ENAMETOOLONG);
3362 		}
3363 
3364 		addrs = 0;
3365 		NET_EPOCH_ENTER(et);
3366 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3367 			struct sockaddr *sa = ifa->ifa_addr;
3368 
3369 			if (prison_if(curthread->td_ucred, sa) != 0)
3370 				continue;
3371 			addrs++;
3372 			if (sa->sa_len <= sizeof(*sa)) {
3373 				if (sa->sa_len < sizeof(*sa)) {
3374 					memset(&ifr.ifr_ifru.ifru_addr, 0,
3375 					    sizeof(ifr.ifr_ifru.ifru_addr));
3376 					memcpy(&ifr.ifr_ifru.ifru_addr, sa,
3377 					    sa->sa_len);
3378 				} else
3379 					ifr.ifr_ifru.ifru_addr = *sa;
3380 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3381 				max_len += sizeof(ifr);
3382 			} else {
3383 				sbuf_bcat(sb, &ifr,
3384 				    offsetof(struct ifreq, ifr_addr));
3385 				max_len += offsetof(struct ifreq, ifr_addr);
3386 				sbuf_bcat(sb, sa, sa->sa_len);
3387 				max_len += sa->sa_len;
3388 			}
3389 
3390 			if (sbuf_error(sb) == 0)
3391 				valid_len = sbuf_len(sb);
3392 		}
3393 		NET_EPOCH_EXIT(et);
3394 		if (addrs == 0) {
3395 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3396 			max_len += sizeof(ifr);
3397 
3398 			if (sbuf_error(sb) == 0)
3399 				valid_len = sbuf_len(sb);
3400 		}
3401 	}
3402 	IFNET_RUNLOCK();
3403 
3404 	/*
3405 	 * If we didn't allocate enough space (uncommon), try again.  If
3406 	 * we have already allocated as much space as we are allowed,
3407 	 * return what we've got.
3408 	 */
3409 	if (valid_len != max_len && !full) {
3410 		sbuf_delete(sb);
3411 		goto again;
3412 	}
3413 
3414 	ifc->ifc_len = valid_len;
3415 	sbuf_finish(sb);
3416 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3417 	sbuf_delete(sb);
3418 	return (error);
3419 }
3420 
3421 /*
3422  * Just like ifpromisc(), but for all-multicast-reception mode.
3423  */
3424 int
3425 if_allmulti(struct ifnet *ifp, int onswitch)
3426 {
3427 
3428 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3429 }
3430 
3431 struct ifmultiaddr *
3432 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3433 {
3434 	struct ifmultiaddr *ifma;
3435 
3436 	IF_ADDR_LOCK_ASSERT(ifp);
3437 
3438 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3439 		if (sa->sa_family == AF_LINK) {
3440 			if (sa_dl_equal(ifma->ifma_addr, sa))
3441 				break;
3442 		} else {
3443 			if (sa_equal(ifma->ifma_addr, sa))
3444 				break;
3445 		}
3446 	}
3447 
3448 	return ifma;
3449 }
3450 
3451 /*
3452  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3453  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3454  * the ifnet multicast address list here, so the caller must do that and
3455  * other setup work (such as notifying the device driver).  The reference
3456  * count is initialized to 1.
3457  */
3458 static struct ifmultiaddr *
3459 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3460     int mflags)
3461 {
3462 	struct ifmultiaddr *ifma;
3463 	struct sockaddr *dupsa;
3464 
3465 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3466 	    M_ZERO);
3467 	if (ifma == NULL)
3468 		return (NULL);
3469 
3470 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3471 	if (dupsa == NULL) {
3472 		free(ifma, M_IFMADDR);
3473 		return (NULL);
3474 	}
3475 	bcopy(sa, dupsa, sa->sa_len);
3476 	ifma->ifma_addr = dupsa;
3477 
3478 	ifma->ifma_ifp = ifp;
3479 	ifma->ifma_refcount = 1;
3480 	ifma->ifma_protospec = NULL;
3481 
3482 	if (llsa == NULL) {
3483 		ifma->ifma_lladdr = NULL;
3484 		return (ifma);
3485 	}
3486 
3487 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3488 	if (dupsa == NULL) {
3489 		free(ifma->ifma_addr, M_IFMADDR);
3490 		free(ifma, M_IFMADDR);
3491 		return (NULL);
3492 	}
3493 	bcopy(llsa, dupsa, llsa->sa_len);
3494 	ifma->ifma_lladdr = dupsa;
3495 
3496 	return (ifma);
3497 }
3498 
3499 /*
3500  * if_freemulti: free ifmultiaddr structure and possibly attached related
3501  * addresses.  The caller is responsible for implementing reference
3502  * counting, notifying the driver, handling routing messages, and releasing
3503  * any dependent link layer state.
3504  */
3505 #ifdef MCAST_VERBOSE
3506 extern void kdb_backtrace(void);
3507 #endif
3508 static void
3509 if_freemulti_internal(struct ifmultiaddr *ifma)
3510 {
3511 
3512 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3513 	    ifma->ifma_refcount));
3514 
3515 	if (ifma->ifma_lladdr != NULL)
3516 		free(ifma->ifma_lladdr, M_IFMADDR);
3517 #ifdef MCAST_VERBOSE
3518 	kdb_backtrace();
3519 	printf("%s freeing ifma: %p\n", __func__, ifma);
3520 #endif
3521 	free(ifma->ifma_addr, M_IFMADDR);
3522 	free(ifma, M_IFMADDR);
3523 }
3524 
3525 static void
3526 if_destroymulti(epoch_context_t ctx)
3527 {
3528 	struct ifmultiaddr *ifma;
3529 
3530 	ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx);
3531 	if_freemulti_internal(ifma);
3532 }
3533 
3534 void
3535 if_freemulti(struct ifmultiaddr *ifma)
3536 {
3537 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d",
3538 	    ifma->ifma_refcount));
3539 
3540 	NET_EPOCH_CALL(if_destroymulti, &ifma->ifma_epoch_ctx);
3541 }
3542 
3543 /*
3544  * Register an additional multicast address with a network interface.
3545  *
3546  * - If the address is already present, bump the reference count on the
3547  *   address and return.
3548  * - If the address is not link-layer, look up a link layer address.
3549  * - Allocate address structures for one or both addresses, and attach to the
3550  *   multicast address list on the interface.  If automatically adding a link
3551  *   layer address, the protocol address will own a reference to the link
3552  *   layer address, to be freed when it is freed.
3553  * - Notify the network device driver of an addition to the multicast address
3554  *   list.
3555  *
3556  * 'sa' points to caller-owned memory with the desired multicast address.
3557  *
3558  * 'retifma' will be used to return a pointer to the resulting multicast
3559  * address reference, if desired.
3560  */
3561 int
3562 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3563     struct ifmultiaddr **retifma)
3564 {
3565 	struct ifmultiaddr *ifma, *ll_ifma;
3566 	struct sockaddr *llsa;
3567 	struct sockaddr_dl sdl;
3568 	int error;
3569 
3570 #ifdef INET
3571 	IN_MULTI_LIST_UNLOCK_ASSERT();
3572 #endif
3573 #ifdef INET6
3574 	IN6_MULTI_LIST_UNLOCK_ASSERT();
3575 #endif
3576 	/*
3577 	 * If the address is already present, return a new reference to it;
3578 	 * otherwise, allocate storage and set up a new address.
3579 	 */
3580 	IF_ADDR_WLOCK(ifp);
3581 	ifma = if_findmulti(ifp, sa);
3582 	if (ifma != NULL) {
3583 		ifma->ifma_refcount++;
3584 		if (retifma != NULL)
3585 			*retifma = ifma;
3586 		IF_ADDR_WUNLOCK(ifp);
3587 		return (0);
3588 	}
3589 
3590 	/*
3591 	 * The address isn't already present; resolve the protocol address
3592 	 * into a link layer address, and then look that up, bump its
3593 	 * refcount or allocate an ifma for that also.
3594 	 * Most link layer resolving functions returns address data which
3595 	 * fits inside default sockaddr_dl structure. However callback
3596 	 * can allocate another sockaddr structure, in that case we need to
3597 	 * free it later.
3598 	 */
3599 	llsa = NULL;
3600 	ll_ifma = NULL;
3601 	if (ifp->if_resolvemulti != NULL) {
3602 		/* Provide called function with buffer size information */
3603 		sdl.sdl_len = sizeof(sdl);
3604 		llsa = (struct sockaddr *)&sdl;
3605 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3606 		if (error)
3607 			goto unlock_out;
3608 	}
3609 
3610 	/*
3611 	 * Allocate the new address.  Don't hook it up yet, as we may also
3612 	 * need to allocate a link layer multicast address.
3613 	 */
3614 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3615 	if (ifma == NULL) {
3616 		error = ENOMEM;
3617 		goto free_llsa_out;
3618 	}
3619 
3620 	/*
3621 	 * If a link layer address is found, we'll need to see if it's
3622 	 * already present in the address list, or allocate is as well.
3623 	 * When this block finishes, the link layer address will be on the
3624 	 * list.
3625 	 */
3626 	if (llsa != NULL) {
3627 		ll_ifma = if_findmulti(ifp, llsa);
3628 		if (ll_ifma == NULL) {
3629 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3630 			if (ll_ifma == NULL) {
3631 				--ifma->ifma_refcount;
3632 				if_freemulti(ifma);
3633 				error = ENOMEM;
3634 				goto free_llsa_out;
3635 			}
3636 			ll_ifma->ifma_flags |= IFMA_F_ENQUEUED;
3637 			CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3638 			    ifma_link);
3639 		} else
3640 			ll_ifma->ifma_refcount++;
3641 		ifma->ifma_llifma = ll_ifma;
3642 	}
3643 
3644 	/*
3645 	 * We now have a new multicast address, ifma, and possibly a new or
3646 	 * referenced link layer address.  Add the primary address to the
3647 	 * ifnet address list.
3648 	 */
3649 	ifma->ifma_flags |= IFMA_F_ENQUEUED;
3650 	CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3651 
3652 	if (retifma != NULL)
3653 		*retifma = ifma;
3654 
3655 	/*
3656 	 * Must generate the message while holding the lock so that 'ifma'
3657 	 * pointer is still valid.
3658 	 */
3659 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3660 	IF_ADDR_WUNLOCK(ifp);
3661 
3662 	/*
3663 	 * We are certain we have added something, so call down to the
3664 	 * interface to let them know about it.
3665 	 */
3666 	if (ifp->if_ioctl != NULL) {
3667 		if (THREAD_CAN_SLEEP())
3668 			(void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3669 		else
3670 			taskqueue_enqueue(taskqueue_swi, &ifp->if_addmultitask);
3671 	}
3672 
3673 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3674 		link_free_sdl(llsa);
3675 
3676 	return (0);
3677 
3678 free_llsa_out:
3679 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3680 		link_free_sdl(llsa);
3681 
3682 unlock_out:
3683 	IF_ADDR_WUNLOCK(ifp);
3684 	return (error);
3685 }
3686 
3687 static void
3688 if_siocaddmulti(void *arg, int pending)
3689 {
3690 	struct ifnet *ifp;
3691 
3692 	ifp = arg;
3693 #ifdef DIAGNOSTIC
3694 	if (pending > 1)
3695 		if_printf(ifp, "%d SIOCADDMULTI coalesced\n", pending);
3696 #endif
3697 	CURVNET_SET(ifp->if_vnet);
3698 	(void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3699 	CURVNET_RESTORE();
3700 }
3701 
3702 /*
3703  * Delete a multicast group membership by network-layer group address.
3704  *
3705  * Returns ENOENT if the entry could not be found. If ifp no longer
3706  * exists, results are undefined. This entry point should only be used
3707  * from subsystems which do appropriate locking to hold ifp for the
3708  * duration of the call.
3709  * Network-layer protocol domains must use if_delmulti_ifma().
3710  */
3711 int
3712 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3713 {
3714 	struct ifmultiaddr *ifma;
3715 	int lastref;
3716 
3717 	KASSERT(ifp, ("%s: NULL ifp", __func__));
3718 
3719 	IF_ADDR_WLOCK(ifp);
3720 	lastref = 0;
3721 	ifma = if_findmulti(ifp, sa);
3722 	if (ifma != NULL)
3723 		lastref = if_delmulti_locked(ifp, ifma, 0);
3724 	IF_ADDR_WUNLOCK(ifp);
3725 
3726 	if (ifma == NULL)
3727 		return (ENOENT);
3728 
3729 	if (lastref && ifp->if_ioctl != NULL) {
3730 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3731 	}
3732 
3733 	return (0);
3734 }
3735 
3736 /*
3737  * Delete all multicast group membership for an interface.
3738  * Should be used to quickly flush all multicast filters.
3739  */
3740 void
3741 if_delallmulti(struct ifnet *ifp)
3742 {
3743 	struct ifmultiaddr *ifma;
3744 	struct ifmultiaddr *next;
3745 
3746 	IF_ADDR_WLOCK(ifp);
3747 	CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3748 		if_delmulti_locked(ifp, ifma, 0);
3749 	IF_ADDR_WUNLOCK(ifp);
3750 }
3751 
3752 void
3753 if_delmulti_ifma(struct ifmultiaddr *ifma)
3754 {
3755 	if_delmulti_ifma_flags(ifma, 0);
3756 }
3757 
3758 /*
3759  * Delete a multicast group membership by group membership pointer.
3760  * Network-layer protocol domains must use this routine.
3761  *
3762  * It is safe to call this routine if the ifp disappeared.
3763  */
3764 void
3765 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags)
3766 {
3767 	struct ifnet *ifp;
3768 	int lastref;
3769 	MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma);
3770 #ifdef INET
3771 	IN_MULTI_LIST_UNLOCK_ASSERT();
3772 #endif
3773 	ifp = ifma->ifma_ifp;
3774 #ifdef DIAGNOSTIC
3775 	if (ifp == NULL) {
3776 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3777 	} else {
3778 		struct epoch_tracker et;
3779 		struct ifnet *oifp;
3780 
3781 		NET_EPOCH_ENTER(et);
3782 		CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link)
3783 			if (ifp == oifp)
3784 				break;
3785 		NET_EPOCH_EXIT(et);
3786 		if (ifp != oifp)
3787 			ifp = NULL;
3788 	}
3789 #endif
3790 	/*
3791 	 * If and only if the ifnet instance exists: Acquire the address lock.
3792 	 */
3793 	if (ifp != NULL)
3794 		IF_ADDR_WLOCK(ifp);
3795 
3796 	lastref = if_delmulti_locked(ifp, ifma, flags);
3797 
3798 	if (ifp != NULL) {
3799 		/*
3800 		 * If and only if the ifnet instance exists:
3801 		 *  Release the address lock.
3802 		 *  If the group was left: update the hardware hash filter.
3803 		 */
3804 		IF_ADDR_WUNLOCK(ifp);
3805 		if (lastref && ifp->if_ioctl != NULL) {
3806 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3807 		}
3808 	}
3809 }
3810 
3811 /*
3812  * Perform deletion of network-layer and/or link-layer multicast address.
3813  *
3814  * Return 0 if the reference count was decremented.
3815  * Return 1 if the final reference was released, indicating that the
3816  * hardware hash filter should be reprogrammed.
3817  */
3818 static int
3819 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3820 {
3821 	struct ifmultiaddr *ll_ifma;
3822 
3823 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3824 		KASSERT(ifma->ifma_ifp == ifp,
3825 		    ("%s: inconsistent ifp %p", __func__, ifp));
3826 		IF_ADDR_WLOCK_ASSERT(ifp);
3827 	}
3828 
3829 	ifp = ifma->ifma_ifp;
3830 	MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : "");
3831 
3832 	/*
3833 	 * If the ifnet is detaching, null out references to ifnet,
3834 	 * so that upper protocol layers will notice, and not attempt
3835 	 * to obtain locks for an ifnet which no longer exists. The
3836 	 * routing socket announcement must happen before the ifnet
3837 	 * instance is detached from the system.
3838 	 */
3839 	if (detaching) {
3840 #ifdef DIAGNOSTIC
3841 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3842 #endif
3843 		/*
3844 		 * ifp may already be nulled out if we are being reentered
3845 		 * to delete the ll_ifma.
3846 		 */
3847 		if (ifp != NULL) {
3848 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3849 			ifma->ifma_ifp = NULL;
3850 		}
3851 	}
3852 
3853 	if (--ifma->ifma_refcount > 0)
3854 		return 0;
3855 
3856 	if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) {
3857 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
3858 		ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3859 	}
3860 	/*
3861 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3862 	 * have been associated with it. Release it first if so.
3863 	 */
3864 	ll_ifma = ifma->ifma_llifma;
3865 	if (ll_ifma != NULL) {
3866 		KASSERT(ifma->ifma_lladdr != NULL,
3867 		    ("%s: llifma w/o lladdr", __func__));
3868 		if (detaching)
3869 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3870 		if (--ll_ifma->ifma_refcount == 0) {
3871 			if (ifp != NULL) {
3872 				if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) {
3873 					CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr,
3874 						ifma_link);
3875 					ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3876 				}
3877 			}
3878 			if_freemulti(ll_ifma);
3879 		}
3880 	}
3881 #ifdef INVARIANTS
3882 	if (ifp) {
3883 		struct ifmultiaddr *ifmatmp;
3884 
3885 		CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link)
3886 			MPASS(ifma != ifmatmp);
3887 	}
3888 #endif
3889 	if_freemulti(ifma);
3890 	/*
3891 	 * The last reference to this instance of struct ifmultiaddr
3892 	 * was released; the hardware should be notified of this change.
3893 	 */
3894 	return 1;
3895 }
3896 
3897 /*
3898  * Set the link layer address on an interface.
3899  *
3900  * At this time we only support certain types of interfaces,
3901  * and we don't allow the length of the address to change.
3902  *
3903  * Set noinline to be dtrace-friendly
3904  */
3905 __noinline int
3906 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3907 {
3908 	struct sockaddr_dl *sdl;
3909 	struct ifaddr *ifa;
3910 	struct ifreq ifr;
3911 
3912 	ifa = ifp->if_addr;
3913 	if (ifa == NULL)
3914 		return (EINVAL);
3915 
3916 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3917 	if (sdl == NULL)
3918 		return (EINVAL);
3919 
3920 	if (len != sdl->sdl_alen)	/* don't allow length to change */
3921 		return (EINVAL);
3922 
3923 	switch (ifp->if_type) {
3924 	case IFT_ETHER:
3925 	case IFT_XETHER:
3926 	case IFT_L2VLAN:
3927 	case IFT_BRIDGE:
3928 	case IFT_IEEE8023ADLAG:
3929 		bcopy(lladdr, LLADDR(sdl), len);
3930 		break;
3931 	default:
3932 		return (ENODEV);
3933 	}
3934 
3935 	/*
3936 	 * If the interface is already up, we need
3937 	 * to re-init it in order to reprogram its
3938 	 * address filter.
3939 	 */
3940 	if ((ifp->if_flags & IFF_UP) != 0) {
3941 		if (ifp->if_ioctl) {
3942 			ifp->if_flags &= ~IFF_UP;
3943 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3944 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3945 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3946 			ifp->if_flags |= IFF_UP;
3947 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3948 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3949 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3950 		}
3951 	}
3952 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3953 
3954 	return (0);
3955 }
3956 
3957 /*
3958  * Compat function for handling basic encapsulation requests.
3959  * Not converted stacks (FDDI, IB, ..) supports traditional
3960  * output model: ARP (and other similar L2 protocols) are handled
3961  * inside output routine, arpresolve/nd6_resolve() returns MAC
3962  * address instead of full prepend.
3963  *
3964  * This function creates calculated header==MAC for IPv4/IPv6 and
3965  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3966  * address families.
3967  */
3968 static int
3969 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3970 {
3971 
3972 	if (req->rtype != IFENCAP_LL)
3973 		return (EOPNOTSUPP);
3974 
3975 	if (req->bufsize < req->lladdr_len)
3976 		return (ENOMEM);
3977 
3978 	switch (req->family) {
3979 	case AF_INET:
3980 	case AF_INET6:
3981 		break;
3982 	default:
3983 		return (EAFNOSUPPORT);
3984 	}
3985 
3986 	/* Copy lladdr to storage as is */
3987 	memmove(req->buf, req->lladdr, req->lladdr_len);
3988 	req->bufsize = req->lladdr_len;
3989 	req->lladdr_off = 0;
3990 
3991 	return (0);
3992 }
3993 
3994 /*
3995  * Tunnel interfaces can nest, also they may cause infinite recursion
3996  * calls when misconfigured. We'll prevent this by detecting loops.
3997  * High nesting level may cause stack exhaustion. We'll prevent this
3998  * by introducing upper limit.
3999  *
4000  * Return 0, if tunnel nesting count is equal or less than limit.
4001  */
4002 int
4003 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie,
4004     int limit)
4005 {
4006 	struct m_tag *mtag;
4007 	int count;
4008 
4009 	count = 1;
4010 	mtag = NULL;
4011 	while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) {
4012 		if (*(struct ifnet **)(mtag + 1) == ifp) {
4013 			log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp));
4014 			return (EIO);
4015 		}
4016 		count++;
4017 	}
4018 	if (count > limit) {
4019 		log(LOG_NOTICE,
4020 		    "%s: if_output recursively called too many times(%d)\n",
4021 		    if_name(ifp), count);
4022 		return (EIO);
4023 	}
4024 	mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT);
4025 	if (mtag == NULL)
4026 		return (ENOMEM);
4027 	*(struct ifnet **)(mtag + 1) = ifp;
4028 	m_tag_prepend(m, mtag);
4029 	return (0);
4030 }
4031 
4032 /*
4033  * Get the link layer address that was read from the hardware at attach.
4034  *
4035  * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
4036  * their component interfaces as IFT_IEEE8023ADLAG.
4037  */
4038 int
4039 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
4040 {
4041 
4042 	if (ifp->if_hw_addr == NULL)
4043 		return (ENODEV);
4044 
4045 	switch (ifp->if_type) {
4046 	case IFT_ETHER:
4047 	case IFT_IEEE8023ADLAG:
4048 		bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
4049 		return (0);
4050 	default:
4051 		return (ENODEV);
4052 	}
4053 }
4054 
4055 /*
4056  * The name argument must be a pointer to storage which will last as
4057  * long as the interface does.  For physical devices, the result of
4058  * device_get_name(dev) is a good choice and for pseudo-devices a
4059  * static string works well.
4060  */
4061 void
4062 if_initname(struct ifnet *ifp, const char *name, int unit)
4063 {
4064 	ifp->if_dname = name;
4065 	ifp->if_dunit = unit;
4066 	if (unit != IF_DUNIT_NONE)
4067 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
4068 	else
4069 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
4070 }
4071 
4072 static int
4073 if_vlog(struct ifnet *ifp, int pri, const char *fmt, va_list ap)
4074 {
4075 	char if_fmt[256];
4076 
4077 	snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt);
4078 	vlog(pri, if_fmt, ap);
4079 	return (0);
4080 }
4081 
4082 
4083 int
4084 if_printf(struct ifnet *ifp, const char *fmt, ...)
4085 {
4086 	va_list ap;
4087 
4088 	va_start(ap, fmt);
4089 	if_vlog(ifp, LOG_INFO, fmt, ap);
4090 	va_end(ap);
4091 	return (0);
4092 }
4093 
4094 int
4095 if_log(struct ifnet *ifp, int pri, const char *fmt, ...)
4096 {
4097 	va_list ap;
4098 
4099 	va_start(ap, fmt);
4100 	if_vlog(ifp, pri, fmt, ap);
4101 	va_end(ap);
4102 	return (0);
4103 }
4104 
4105 void
4106 if_start(struct ifnet *ifp)
4107 {
4108 
4109 	(*(ifp)->if_start)(ifp);
4110 }
4111 
4112 /*
4113  * Backwards compatibility interface for drivers
4114  * that have not implemented it
4115  */
4116 static int
4117 if_transmit(struct ifnet *ifp, struct mbuf *m)
4118 {
4119 	int error;
4120 
4121 	IFQ_HANDOFF(ifp, m, error);
4122 	return (error);
4123 }
4124 
4125 static void
4126 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
4127 {
4128 
4129 	m_freem(m);
4130 }
4131 
4132 int
4133 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
4134 {
4135 	int active = 0;
4136 
4137 	IF_LOCK(ifq);
4138 	if (_IF_QFULL(ifq)) {
4139 		IF_UNLOCK(ifq);
4140 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
4141 		m_freem(m);
4142 		return (0);
4143 	}
4144 	if (ifp != NULL) {
4145 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
4146 		if (m->m_flags & (M_BCAST|M_MCAST))
4147 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
4148 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
4149 	}
4150 	_IF_ENQUEUE(ifq, m);
4151 	IF_UNLOCK(ifq);
4152 	if (ifp != NULL && !active)
4153 		(*(ifp)->if_start)(ifp);
4154 	return (1);
4155 }
4156 
4157 void
4158 if_register_com_alloc(u_char type,
4159     if_com_alloc_t *a, if_com_free_t *f)
4160 {
4161 
4162 	KASSERT(if_com_alloc[type] == NULL,
4163 	    ("if_register_com_alloc: %d already registered", type));
4164 	KASSERT(if_com_free[type] == NULL,
4165 	    ("if_register_com_alloc: %d free already registered", type));
4166 
4167 	if_com_alloc[type] = a;
4168 	if_com_free[type] = f;
4169 }
4170 
4171 void
4172 if_deregister_com_alloc(u_char type)
4173 {
4174 
4175 	KASSERT(if_com_alloc[type] != NULL,
4176 	    ("if_deregister_com_alloc: %d not registered", type));
4177 	KASSERT(if_com_free[type] != NULL,
4178 	    ("if_deregister_com_alloc: %d free not registered", type));
4179 
4180 	/*
4181 	 * Ensure all pending EPOCH(9) callbacks have been executed. This
4182 	 * fixes issues about late invocation of if_destroy(), which leads
4183 	 * to memory leak from if_com_alloc[type] allocated if_l2com.
4184 	 */
4185 	NET_EPOCH_DRAIN_CALLBACKS();
4186 
4187 	if_com_alloc[type] = NULL;
4188 	if_com_free[type] = NULL;
4189 }
4190 
4191 /* API for driver access to network stack owned ifnet.*/
4192 uint64_t
4193 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
4194 {
4195 	uint64_t oldbrate;
4196 
4197 	oldbrate = ifp->if_baudrate;
4198 	ifp->if_baudrate = baudrate;
4199 	return (oldbrate);
4200 }
4201 
4202 uint64_t
4203 if_getbaudrate(if_t ifp)
4204 {
4205 
4206 	return (((struct ifnet *)ifp)->if_baudrate);
4207 }
4208 
4209 int
4210 if_setcapabilities(if_t ifp, int capabilities)
4211 {
4212 	((struct ifnet *)ifp)->if_capabilities = capabilities;
4213 	return (0);
4214 }
4215 
4216 int
4217 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
4218 {
4219 	((struct ifnet *)ifp)->if_capabilities |= setbit;
4220 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
4221 
4222 	return (0);
4223 }
4224 
4225 int
4226 if_getcapabilities(if_t ifp)
4227 {
4228 	return ((struct ifnet *)ifp)->if_capabilities;
4229 }
4230 
4231 int
4232 if_setcapenable(if_t ifp, int capabilities)
4233 {
4234 	((struct ifnet *)ifp)->if_capenable = capabilities;
4235 	return (0);
4236 }
4237 
4238 int
4239 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
4240 {
4241 	if(setcap)
4242 		((struct ifnet *)ifp)->if_capenable |= setcap;
4243 	if(clearcap)
4244 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
4245 
4246 	return (0);
4247 }
4248 
4249 const char *
4250 if_getdname(if_t ifp)
4251 {
4252 	return ((struct ifnet *)ifp)->if_dname;
4253 }
4254 
4255 int
4256 if_togglecapenable(if_t ifp, int togglecap)
4257 {
4258 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
4259 	return (0);
4260 }
4261 
4262 int
4263 if_getcapenable(if_t ifp)
4264 {
4265 	return ((struct ifnet *)ifp)->if_capenable;
4266 }
4267 
4268 /*
4269  * This is largely undesirable because it ties ifnet to a device, but does
4270  * provide flexiblity for an embedded product vendor. Should be used with
4271  * the understanding that it violates the interface boundaries, and should be
4272  * a last resort only.
4273  */
4274 int
4275 if_setdev(if_t ifp, void *dev)
4276 {
4277 	return (0);
4278 }
4279 
4280 int
4281 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
4282 {
4283 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
4284 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
4285 
4286 	return (0);
4287 }
4288 
4289 int
4290 if_getdrvflags(if_t ifp)
4291 {
4292 	return ((struct ifnet *)ifp)->if_drv_flags;
4293 }
4294 
4295 int
4296 if_setdrvflags(if_t ifp, int flags)
4297 {
4298 	((struct ifnet *)ifp)->if_drv_flags = flags;
4299 	return (0);
4300 }
4301 
4302 int
4303 if_setflags(if_t ifp, int flags)
4304 {
4305 
4306 	ifp->if_flags = flags;
4307 	return (0);
4308 }
4309 
4310 int
4311 if_setflagbits(if_t ifp, int set, int clear)
4312 {
4313 	((struct ifnet *)ifp)->if_flags |= set;
4314 	((struct ifnet *)ifp)->if_flags &= ~clear;
4315 
4316 	return (0);
4317 }
4318 
4319 int
4320 if_getflags(if_t ifp)
4321 {
4322 	return ((struct ifnet *)ifp)->if_flags;
4323 }
4324 
4325 int
4326 if_clearhwassist(if_t ifp)
4327 {
4328 	((struct ifnet *)ifp)->if_hwassist = 0;
4329 	return (0);
4330 }
4331 
4332 int
4333 if_sethwassistbits(if_t ifp, int toset, int toclear)
4334 {
4335 	((struct ifnet *)ifp)->if_hwassist |= toset;
4336 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
4337 
4338 	return (0);
4339 }
4340 
4341 int
4342 if_sethwassist(if_t ifp, int hwassist_bit)
4343 {
4344 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
4345 	return (0);
4346 }
4347 
4348 int
4349 if_gethwassist(if_t ifp)
4350 {
4351 	return ((struct ifnet *)ifp)->if_hwassist;
4352 }
4353 
4354 int
4355 if_setmtu(if_t ifp, int mtu)
4356 {
4357 	((struct ifnet *)ifp)->if_mtu = mtu;
4358 	return (0);
4359 }
4360 
4361 int
4362 if_getmtu(if_t ifp)
4363 {
4364 	return ((struct ifnet *)ifp)->if_mtu;
4365 }
4366 
4367 int
4368 if_getmtu_family(if_t ifp, int family)
4369 {
4370 	struct domain *dp;
4371 
4372 	for (dp = domains; dp; dp = dp->dom_next) {
4373 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
4374 			return (dp->dom_ifmtu((struct ifnet *)ifp));
4375 	}
4376 
4377 	return (((struct ifnet *)ifp)->if_mtu);
4378 }
4379 
4380 /*
4381  * Methods for drivers to access interface unicast and multicast
4382  * link level addresses.  Driver shall not know 'struct ifaddr' neither
4383  * 'struct ifmultiaddr'.
4384  */
4385 u_int
4386 if_lladdr_count(if_t ifp)
4387 {
4388 	struct epoch_tracker et;
4389 	struct ifaddr *ifa;
4390 	u_int count;
4391 
4392 	count = 0;
4393 	NET_EPOCH_ENTER(et);
4394 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
4395 		if (ifa->ifa_addr->sa_family == AF_LINK)
4396 			count++;
4397 	NET_EPOCH_EXIT(et);
4398 
4399 	return (count);
4400 }
4401 
4402 u_int
4403 if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4404 {
4405 	struct epoch_tracker et;
4406 	struct ifaddr *ifa;
4407 	u_int count;
4408 
4409 	MPASS(cb);
4410 
4411 	count = 0;
4412 	NET_EPOCH_ENTER(et);
4413 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
4414 		if (ifa->ifa_addr->sa_family != AF_LINK)
4415 			continue;
4416 		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr,
4417 		    count);
4418 	}
4419 	NET_EPOCH_EXIT(et);
4420 
4421 	return (count);
4422 }
4423 
4424 u_int
4425 if_llmaddr_count(if_t ifp)
4426 {
4427 	struct epoch_tracker et;
4428 	struct ifmultiaddr *ifma;
4429 	int count;
4430 
4431 	count = 0;
4432 	NET_EPOCH_ENTER(et);
4433 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4434 		if (ifma->ifma_addr->sa_family == AF_LINK)
4435 			count++;
4436 	NET_EPOCH_EXIT(et);
4437 
4438 	return (count);
4439 }
4440 
4441 u_int
4442 if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4443 {
4444 	struct epoch_tracker et;
4445 	struct ifmultiaddr *ifma;
4446 	u_int count;
4447 
4448 	MPASS(cb);
4449 
4450 	count = 0;
4451 	NET_EPOCH_ENTER(et);
4452 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
4453 		if (ifma->ifma_addr->sa_family != AF_LINK)
4454 			continue;
4455 		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr,
4456 		    count);
4457 	}
4458 	NET_EPOCH_EXIT(et);
4459 
4460 	return (count);
4461 }
4462 
4463 int
4464 if_setsoftc(if_t ifp, void *softc)
4465 {
4466 	((struct ifnet *)ifp)->if_softc = softc;
4467 	return (0);
4468 }
4469 
4470 void *
4471 if_getsoftc(if_t ifp)
4472 {
4473 	return ((struct ifnet *)ifp)->if_softc;
4474 }
4475 
4476 void
4477 if_setrcvif(struct mbuf *m, if_t ifp)
4478 {
4479 
4480 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
4481 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
4482 }
4483 
4484 void
4485 if_setvtag(struct mbuf *m, uint16_t tag)
4486 {
4487 	m->m_pkthdr.ether_vtag = tag;
4488 }
4489 
4490 uint16_t
4491 if_getvtag(struct mbuf *m)
4492 {
4493 
4494 	return (m->m_pkthdr.ether_vtag);
4495 }
4496 
4497 int
4498 if_sendq_empty(if_t ifp)
4499 {
4500 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
4501 }
4502 
4503 struct ifaddr *
4504 if_getifaddr(if_t ifp)
4505 {
4506 	return ((struct ifnet *)ifp)->if_addr;
4507 }
4508 
4509 int
4510 if_getamcount(if_t ifp)
4511 {
4512 	return ((struct ifnet *)ifp)->if_amcount;
4513 }
4514 
4515 int
4516 if_setsendqready(if_t ifp)
4517 {
4518 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
4519 	return (0);
4520 }
4521 
4522 int
4523 if_setsendqlen(if_t ifp, int tx_desc_count)
4524 {
4525 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
4526 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
4527 
4528 	return (0);
4529 }
4530 
4531 int
4532 if_vlantrunkinuse(if_t ifp)
4533 {
4534 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
4535 }
4536 
4537 int
4538 if_input(if_t ifp, struct mbuf* sendmp)
4539 {
4540 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
4541 	return (0);
4542 
4543 }
4544 
4545 struct mbuf *
4546 if_dequeue(if_t ifp)
4547 {
4548 	struct mbuf *m;
4549 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
4550 
4551 	return (m);
4552 }
4553 
4554 int
4555 if_sendq_prepend(if_t ifp, struct mbuf *m)
4556 {
4557 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
4558 	return (0);
4559 }
4560 
4561 int
4562 if_setifheaderlen(if_t ifp, int len)
4563 {
4564 	((struct ifnet *)ifp)->if_hdrlen = len;
4565 	return (0);
4566 }
4567 
4568 caddr_t
4569 if_getlladdr(if_t ifp)
4570 {
4571 	return (IF_LLADDR((struct ifnet *)ifp));
4572 }
4573 
4574 void *
4575 if_gethandle(u_char type)
4576 {
4577 	return (if_alloc(type));
4578 }
4579 
4580 void
4581 if_bpfmtap(if_t ifh, struct mbuf *m)
4582 {
4583 	struct ifnet *ifp = (struct ifnet *)ifh;
4584 
4585 	BPF_MTAP(ifp, m);
4586 }
4587 
4588 void
4589 if_etherbpfmtap(if_t ifh, struct mbuf *m)
4590 {
4591 	struct ifnet *ifp = (struct ifnet *)ifh;
4592 
4593 	ETHER_BPF_MTAP(ifp, m);
4594 }
4595 
4596 void
4597 if_vlancap(if_t ifh)
4598 {
4599 	struct ifnet *ifp = (struct ifnet *)ifh;
4600 	VLAN_CAPABILITIES(ifp);
4601 }
4602 
4603 int
4604 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4605 {
4606 
4607 	((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax;
4608         return (0);
4609 }
4610 
4611 int
4612 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4613 {
4614 
4615 	((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4616         return (0);
4617 }
4618 
4619 int
4620 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4621 {
4622 
4623 	((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4624         return (0);
4625 }
4626 
4627 u_int
4628 if_gethwtsomax(if_t ifp)
4629 {
4630 
4631 	return (((struct ifnet *)ifp)->if_hw_tsomax);
4632 }
4633 
4634 u_int
4635 if_gethwtsomaxsegcount(if_t ifp)
4636 {
4637 
4638 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount);
4639 }
4640 
4641 u_int
4642 if_gethwtsomaxsegsize(if_t ifp)
4643 {
4644 
4645 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize);
4646 }
4647 
4648 void
4649 if_setinitfn(if_t ifp, void (*init_fn)(void *))
4650 {
4651 	((struct ifnet *)ifp)->if_init = init_fn;
4652 }
4653 
4654 void
4655 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
4656 {
4657 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4658 }
4659 
4660 void
4661 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4662 {
4663 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
4664 }
4665 
4666 void
4667 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4668 {
4669 	((struct ifnet *)ifp)->if_transmit = start_fn;
4670 }
4671 
4672 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4673 {
4674 	((struct ifnet *)ifp)->if_qflush = flush_fn;
4675 
4676 }
4677 
4678 void
4679 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4680 {
4681 
4682 	ifp->if_get_counter = fn;
4683 }
4684 
4685 #ifdef DDB
4686 static void
4687 if_show_ifnet(struct ifnet *ifp)
4688 {
4689 
4690 	if (ifp == NULL)
4691 		return;
4692 	db_printf("%s:\n", ifp->if_xname);
4693 #define	IF_DB_PRINTF(f, e)	db_printf("   %s = " f "\n", #e, ifp->e);
4694 	IF_DB_PRINTF("%s", if_dname);
4695 	IF_DB_PRINTF("%d", if_dunit);
4696 	IF_DB_PRINTF("%s", if_description);
4697 	IF_DB_PRINTF("%u", if_index);
4698 	IF_DB_PRINTF("%d", if_idxgen);
4699 	IF_DB_PRINTF("%u", if_refcount);
4700 	IF_DB_PRINTF("%p", if_softc);
4701 	IF_DB_PRINTF("%p", if_l2com);
4702 	IF_DB_PRINTF("%p", if_llsoftc);
4703 	IF_DB_PRINTF("%d", if_amcount);
4704 	IF_DB_PRINTF("%p", if_addr);
4705 	IF_DB_PRINTF("%p", if_broadcastaddr);
4706 	IF_DB_PRINTF("%p", if_afdata);
4707 	IF_DB_PRINTF("%d", if_afdata_initialized);
4708 	IF_DB_PRINTF("%u", if_fib);
4709 	IF_DB_PRINTF("%p", if_vnet);
4710 	IF_DB_PRINTF("%p", if_home_vnet);
4711 	IF_DB_PRINTF("%p", if_vlantrunk);
4712 	IF_DB_PRINTF("%p", if_bpf);
4713 	IF_DB_PRINTF("%u", if_pcount);
4714 	IF_DB_PRINTF("%p", if_bridge);
4715 	IF_DB_PRINTF("%p", if_lagg);
4716 	IF_DB_PRINTF("%p", if_pf_kif);
4717 	IF_DB_PRINTF("%p", if_carp);
4718 	IF_DB_PRINTF("%p", if_label);
4719 	IF_DB_PRINTF("%p", if_netmap);
4720 	IF_DB_PRINTF("0x%08x", if_flags);
4721 	IF_DB_PRINTF("0x%08x", if_drv_flags);
4722 	IF_DB_PRINTF("0x%08x", if_capabilities);
4723 	IF_DB_PRINTF("0x%08x", if_capenable);
4724 	IF_DB_PRINTF("%p", if_snd.ifq_head);
4725 	IF_DB_PRINTF("%p", if_snd.ifq_tail);
4726 	IF_DB_PRINTF("%d", if_snd.ifq_len);
4727 	IF_DB_PRINTF("%d", if_snd.ifq_maxlen);
4728 	IF_DB_PRINTF("%p", if_snd.ifq_drv_head);
4729 	IF_DB_PRINTF("%p", if_snd.ifq_drv_tail);
4730 	IF_DB_PRINTF("%d", if_snd.ifq_drv_len);
4731 	IF_DB_PRINTF("%d", if_snd.ifq_drv_maxlen);
4732 	IF_DB_PRINTF("%d", if_snd.altq_type);
4733 	IF_DB_PRINTF("%x", if_snd.altq_flags);
4734 #undef IF_DB_PRINTF
4735 }
4736 
4737 DB_SHOW_COMMAND(ifnet, db_show_ifnet)
4738 {
4739 
4740 	if (!have_addr) {
4741 		db_printf("usage: show ifnet <struct ifnet *>\n");
4742 		return;
4743 	}
4744 
4745 	if_show_ifnet((struct ifnet *)addr);
4746 }
4747 
4748 DB_SHOW_ALL_COMMAND(ifnets, db_show_all_ifnets)
4749 {
4750 	struct ifnet *ifp;
4751 	u_short idx;
4752 
4753 	for (idx = 1; idx <= if_index; idx++) {
4754 		ifp = ifindex_table[idx].ife_ifnet;
4755 		if (ifp == NULL)
4756 			continue;
4757 		db_printf( "%20s ifp=%p\n", ifp->if_xname, ifp);
4758 		if (db_pager_quit)
4759 			break;
4760 	}
4761 }
4762 #endif	/* DDB */
4763