xref: /freebsd/sys/net/if.c (revision 188bf8876fdc6e3edebeb76f80e764e419a65f1e)
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_private.h>
90 #include <net/if_vlan_var.h>
91 #include <net/radix.h>
92 #include <net/route.h>
93 #include <net/route/route_ctl.h>
94 #include <net/vnet.h>
95 
96 #if defined(INET) || defined(INET6)
97 #include <net/ethernet.h>
98 #include <netinet/in.h>
99 #include <netinet/in_var.h>
100 #include <netinet/ip.h>
101 #include <netinet/ip_carp.h>
102 #ifdef INET
103 #include <net/debugnet.h>
104 #include <netinet/if_ether.h>
105 #endif /* INET */
106 #ifdef INET6
107 #include <netinet6/in6_var.h>
108 #include <netinet6/in6_ifattach.h>
109 #endif /* INET6 */
110 #endif /* INET || INET6 */
111 
112 #include <security/mac/mac_framework.h>
113 
114 /*
115  * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name
116  * and ifr_ifru when it is used in SIOCGIFCONF.
117  */
118 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) ==
119     offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru");
120 
121 __read_mostly epoch_t net_epoch_preempt;
122 #ifdef COMPAT_FREEBSD32
123 #include <sys/mount.h>
124 #include <compat/freebsd32/freebsd32.h>
125 
126 struct ifreq_buffer32 {
127 	uint32_t	length;		/* (size_t) */
128 	uint32_t	buffer;		/* (void *) */
129 };
130 
131 /*
132  * Interface request structure used for socket
133  * ioctl's.  All interface ioctl's must have parameter
134  * definitions which begin with ifr_name.  The
135  * remainder may be interface specific.
136  */
137 struct ifreq32 {
138 	char	ifr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
139 	union {
140 		struct sockaddr	ifru_addr;
141 		struct sockaddr	ifru_dstaddr;
142 		struct sockaddr	ifru_broadaddr;
143 		struct ifreq_buffer32 ifru_buffer;
144 		short		ifru_flags[2];
145 		short		ifru_index;
146 		int		ifru_jid;
147 		int		ifru_metric;
148 		int		ifru_mtu;
149 		int		ifru_phys;
150 		int		ifru_media;
151 		uint32_t	ifru_data;
152 		int		ifru_cap[2];
153 		u_int		ifru_fib;
154 		u_char		ifru_vlan_pcp;
155 	} ifr_ifru;
156 };
157 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32));
158 CTASSERT(__offsetof(struct ifreq, ifr_ifru) ==
159     __offsetof(struct ifreq32, ifr_ifru));
160 
161 struct ifconf32 {
162 	int32_t	ifc_len;
163 	union {
164 		uint32_t	ifcu_buf;
165 		uint32_t	ifcu_req;
166 	} ifc_ifcu;
167 };
168 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
169 
170 struct ifdrv32 {
171 	char		ifd_name[IFNAMSIZ];
172 	uint32_t	ifd_cmd;
173 	uint32_t	ifd_len;
174 	uint32_t	ifd_data;
175 };
176 #define SIOCSDRVSPEC32	_IOC_NEWTYPE(SIOCSDRVSPEC, struct ifdrv32)
177 #define SIOCGDRVSPEC32	_IOC_NEWTYPE(SIOCGDRVSPEC, struct ifdrv32)
178 
179 struct ifgroupreq32 {
180 	char	ifgr_name[IFNAMSIZ];
181 	u_int	ifgr_len;
182 	union {
183 		char		ifgru_group[IFNAMSIZ];
184 		uint32_t	ifgru_groups;
185 	} ifgr_ifgru;
186 };
187 #define	SIOCAIFGROUP32	_IOC_NEWTYPE(SIOCAIFGROUP, struct ifgroupreq32)
188 #define	SIOCGIFGROUP32	_IOC_NEWTYPE(SIOCGIFGROUP, struct ifgroupreq32)
189 #define	SIOCDIFGROUP32	_IOC_NEWTYPE(SIOCDIFGROUP, struct ifgroupreq32)
190 #define	SIOCGIFGMEMB32	_IOC_NEWTYPE(SIOCGIFGMEMB, struct ifgroupreq32)
191 
192 struct ifmediareq32 {
193 	char		ifm_name[IFNAMSIZ];
194 	int		ifm_current;
195 	int		ifm_mask;
196 	int		ifm_status;
197 	int		ifm_active;
198 	int		ifm_count;
199 	uint32_t	ifm_ulist;	/* (int *) */
200 };
201 #define	SIOCGIFMEDIA32	_IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32)
202 #define	SIOCGIFXMEDIA32	_IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32)
203 #endif /* COMPAT_FREEBSD32 */
204 
205 union ifreq_union {
206 	struct ifreq	ifr;
207 #ifdef COMPAT_FREEBSD32
208 	struct ifreq32	ifr32;
209 #endif
210 };
211 
212 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
213     "Link layers");
214 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
215     "Generic link-management");
216 
217 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
218     &ifqmaxlen, 0, "max send queue size");
219 
220 /* Log link state change events */
221 static int log_link_state_change = 1;
222 
223 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
224 	&log_link_state_change, 0,
225 	"log interface link state change events");
226 
227 /* Log promiscuous mode change events */
228 static int log_promisc_mode_change = 1;
229 
230 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
231 	&log_promisc_mode_change, 1,
232 	"log promiscuous mode change events");
233 
234 /* Interface description */
235 static unsigned int ifdescr_maxlen = 1024;
236 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
237 	&ifdescr_maxlen, 0,
238 	"administrative maximum length for interface description");
239 
240 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
241 
242 /* global sx for non-critical path ifdescr */
243 static struct sx ifdescr_sx;
244 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
245 
246 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
247 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
248 /* These are external hooks for CARP. */
249 void	(*carp_linkstate_p)(struct ifnet *ifp);
250 void	(*carp_demote_adj_p)(int, char *);
251 int	(*carp_master_p)(struct ifaddr *);
252 #if defined(INET) || defined(INET6)
253 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
254 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
255     const struct sockaddr *sa);
256 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
257 int	(*carp_attach_p)(struct ifaddr *, int);
258 void	(*carp_detach_p)(struct ifaddr *, bool);
259 #endif
260 #ifdef INET
261 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
262 #endif
263 #ifdef INET6
264 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
265 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
266     const struct in6_addr *taddr);
267 #endif
268 
269 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
270 
271 /*
272  * XXX: Style; these should be sorted alphabetically, and unprototyped
273  * static functions should be prototyped. Currently they are sorted by
274  * declaration order.
275  */
276 static void	if_attachdomain(void *);
277 static void	if_attachdomain1(struct ifnet *);
278 static int	ifconf(u_long, caddr_t);
279 static void	if_input_default(struct ifnet *, struct mbuf *);
280 static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
281 static void	if_route(struct ifnet *, int flag, int fam);
282 static int	if_setflag(struct ifnet *, int, int, int *, int);
283 static int	if_transmit_default(struct ifnet *ifp, struct mbuf *m);
284 static void	if_unroute(struct ifnet *, int flag, int fam);
285 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
286 static void	do_link_state_change(void *, int);
287 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
288 static int	if_getgroupmembers(struct ifgroupreq *);
289 static void	if_delgroups(struct ifnet *);
290 static void	if_attach_internal(struct ifnet *, bool);
291 static int	if_detach_internal(struct ifnet *, bool);
292 static void	if_siocaddmulti(void *, int);
293 static void	if_link_ifnet(struct ifnet *);
294 static bool	if_unlink_ifnet(struct ifnet *, bool);
295 #ifdef VIMAGE
296 static int	if_vmove(struct ifnet *, struct vnet *);
297 #endif
298 
299 #ifdef INET6
300 /*
301  * XXX: declare here to avoid to include many inet6 related files..
302  * should be more generalized?
303  */
304 extern void	nd6_setmtu(struct ifnet *);
305 #endif
306 
307 /* ipsec helper hooks */
308 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
309 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
310 
311 int	ifqmaxlen = IFQ_MAXLEN;
312 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
313 VNET_DEFINE(struct ifgrouphead, ifg_head);
314 
315 /* Table of ifnet by index. */
316 static int if_index;
317 static int if_indexlim = 8;
318 static struct ifindex_entry {
319 	struct ifnet	*ife_ifnet;
320 	uint16_t	ife_gencnt;
321 } *ifindex_table;
322 
323 SYSCTL_NODE(_net_link_generic, IFMIB_SYSTEM, system,
324     CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
325     "Variables global to all interfaces");
326 static int
327 sysctl_ifcount(SYSCTL_HANDLER_ARGS)
328 {
329 	int rv = 0;
330 
331 	IFNET_RLOCK();
332 	for (int i = 1; i <= if_index; i++)
333 		if (ifindex_table[i].ife_ifnet != NULL &&
334 		    ifindex_table[i].ife_ifnet->if_vnet == curvnet)
335 			rv = i;
336 	IFNET_RUNLOCK();
337 
338 	return (sysctl_handle_int(oidp, &rv, 0, req));
339 }
340 SYSCTL_PROC(_net_link_generic_system, IFMIB_IFCOUNT, ifcount,
341     CTLTYPE_INT | CTLFLAG_VNET | CTLFLAG_RD, NULL, 0, sysctl_ifcount, "I",
342     "Maximum known interface index");
343 
344 /*
345  * The global network interface list (V_ifnet) and related state (such as
346  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock.
347  * This may be acquired to stabilise the list, or we may rely on NET_EPOCH.
348  */
349 struct sx ifnet_sxlock;
350 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
351 
352 struct sx ifnet_detach_sxlock;
353 SX_SYSINIT_FLAGS(ifnet_detach, &ifnet_detach_sxlock, "ifnet_detach_sx",
354     SX_RECURSE);
355 
356 #ifdef VIMAGE
357 #define	VNET_IS_SHUTTING_DOWN(_vnet)					\
358     ((_vnet)->vnet_shutdown && (_vnet)->vnet_state < SI_SUB_VNET_DONE)
359 #endif
360 
361 static	if_com_alloc_t *if_com_alloc[256];
362 static	if_com_free_t *if_com_free[256];
363 
364 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
365 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
366 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
367 
368 struct ifnet *
369 ifnet_byindex(u_int idx)
370 {
371 	struct ifnet *ifp;
372 
373 	NET_EPOCH_ASSERT();
374 
375 	if (__predict_false(idx > if_index))
376 		return (NULL);
377 
378 	ifp = ck_pr_load_ptr(&ifindex_table[idx].ife_ifnet);
379 
380 	if (curvnet != NULL && ifp != NULL && ifp->if_vnet != curvnet)
381 		ifp = NULL;
382 
383 	return (ifp);
384 }
385 
386 struct ifnet *
387 ifnet_byindex_ref(u_int idx)
388 {
389 	struct ifnet *ifp;
390 
391 	ifp = ifnet_byindex(idx);
392 	if (ifp == NULL || (ifp->if_flags & IFF_DYING))
393 		return (NULL);
394 	if (!if_try_ref(ifp))
395 		return (NULL);
396 	return (ifp);
397 }
398 
399 struct ifnet *
400 ifnet_byindexgen(uint16_t idx, uint16_t gen)
401 {
402 	struct ifnet *ifp;
403 
404 	NET_EPOCH_ASSERT();
405 
406 	if (__predict_false(idx > if_index))
407 		return (NULL);
408 
409 	ifp = ck_pr_load_ptr(&ifindex_table[idx].ife_ifnet);
410 
411 	if (ifindex_table[idx].ife_gencnt == gen)
412 		return (ifp);
413 	else
414 		return (NULL);
415 }
416 
417 /*
418  * Network interface utility routines.
419  *
420  * Routines with ifa_ifwith* names take sockaddr *'s as
421  * parameters.
422  */
423 
424 static void
425 if_init_idxtable(void *arg __unused)
426 {
427 
428 	ifindex_table = malloc(if_indexlim * sizeof(*ifindex_table),
429 	    M_IFNET, M_WAITOK | M_ZERO);
430 }
431 SYSINIT(if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, if_init_idxtable, NULL);
432 
433 static void
434 vnet_if_init(const void *unused __unused)
435 {
436 
437 	CK_STAILQ_INIT(&V_ifnet);
438 	CK_STAILQ_INIT(&V_ifg_head);
439 	vnet_if_clone_init();
440 }
441 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
442     NULL);
443 
444 static void
445 if_link_ifnet(struct ifnet *ifp)
446 {
447 
448 	IFNET_WLOCK();
449 	CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
450 #ifdef VIMAGE
451 	curvnet->vnet_ifcnt++;
452 #endif
453 	IFNET_WUNLOCK();
454 }
455 
456 static bool
457 if_unlink_ifnet(struct ifnet *ifp, bool vmove)
458 {
459 	struct ifnet *iter;
460 	int found = 0;
461 
462 	IFNET_WLOCK();
463 	CK_STAILQ_FOREACH(iter, &V_ifnet, if_link)
464 		if (iter == ifp) {
465 			CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link);
466 			if (!vmove)
467 				ifp->if_flags |= IFF_DYING;
468 			found = 1;
469 			break;
470 		}
471 #ifdef VIMAGE
472 	curvnet->vnet_ifcnt--;
473 #endif
474 	IFNET_WUNLOCK();
475 
476 	return (found);
477 }
478 
479 #ifdef VIMAGE
480 static void
481 vnet_if_return(const void *unused __unused)
482 {
483 	struct ifnet *ifp, *nifp;
484 	struct ifnet **pending;
485 	int found __diagused;
486 	int i;
487 
488 	i = 0;
489 
490 	/*
491 	 * We need to protect our access to the V_ifnet tailq. Ordinarily we'd
492 	 * enter NET_EPOCH, but that's not possible, because if_vmove() calls
493 	 * if_detach_internal(), which waits for NET_EPOCH callbacks to
494 	 * complete. We can't do that from within NET_EPOCH.
495 	 *
496 	 * However, we can also use the IFNET_xLOCK, which is the V_ifnet
497 	 * read/write lock. We cannot hold the lock as we call if_vmove()
498 	 * though, as that presents LOR w.r.t ifnet_sx, in_multi_sx and iflib
499 	 * ctx lock.
500 	 */
501 	IFNET_WLOCK();
502 
503 	pending = malloc(sizeof(struct ifnet *) * curvnet->vnet_ifcnt,
504 	    M_IFNET, M_WAITOK | M_ZERO);
505 
506 	/* Return all inherited interfaces to their parent vnets. */
507 	CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
508 		if (ifp->if_home_vnet != ifp->if_vnet) {
509 			found = if_unlink_ifnet(ifp, true);
510 			MPASS(found);
511 
512 			pending[i++] = ifp;
513 		}
514 	}
515 	IFNET_WUNLOCK();
516 
517 	for (int j = 0; j < i; j++) {
518 		sx_xlock(&ifnet_detach_sxlock);
519 		if_vmove(pending[j], pending[j]->if_home_vnet);
520 		sx_xunlock(&ifnet_detach_sxlock);
521 	}
522 
523 	free(pending, M_IFNET);
524 }
525 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
526     vnet_if_return, NULL);
527 #endif
528 
529 /*
530  * Allocate a struct ifnet and an index for an interface.  A layer 2
531  * common structure will also be allocated if an allocation routine is
532  * registered for the passed type.
533  */
534 static struct ifnet *
535 if_alloc_domain(u_char type, int numa_domain)
536 {
537 	struct ifnet *ifp;
538 	u_short idx;
539 
540 	KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large"));
541 	if (numa_domain == IF_NODOM)
542 		ifp = malloc(sizeof(struct ifnet), M_IFNET,
543 		    M_WAITOK | M_ZERO);
544 	else
545 		ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET,
546 		    DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO);
547 	ifp->if_type = type;
548 	ifp->if_alloctype = type;
549 	ifp->if_numa_domain = numa_domain;
550 #ifdef VIMAGE
551 	ifp->if_vnet = curvnet;
552 #endif
553 	if (if_com_alloc[type] != NULL) {
554 		ifp->if_l2com = if_com_alloc[type](type, ifp);
555 		KASSERT(ifp->if_l2com, ("%s: if_com_alloc[%u] failed", __func__,
556 		    type));
557 	}
558 
559 	IF_ADDR_LOCK_INIT(ifp);
560 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
561 	TASK_INIT(&ifp->if_addmultitask, 0, if_siocaddmulti, ifp);
562 	ifp->if_afdata_initialized = 0;
563 	IF_AFDATA_LOCK_INIT(ifp);
564 	CK_STAILQ_INIT(&ifp->if_addrhead);
565 	CK_STAILQ_INIT(&ifp->if_multiaddrs);
566 	CK_STAILQ_INIT(&ifp->if_groups);
567 #ifdef MAC
568 	mac_ifnet_init(ifp);
569 #endif
570 	ifq_init(&ifp->if_snd, ifp);
571 
572 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
573 	for (int i = 0; i < IFCOUNTERS; i++)
574 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
575 	ifp->if_get_counter = if_get_counter_default;
576 	ifp->if_pcp = IFNET_PCP_NONE;
577 
578 	/* Allocate an ifindex array entry. */
579 	IFNET_WLOCK();
580 	/*
581 	 * Try to find an empty slot below if_index.  If we fail, take the
582 	 * next slot.
583 	 */
584 	for (idx = 1; idx <= if_index; idx++) {
585 		if (ifindex_table[idx].ife_ifnet == NULL)
586 			break;
587 	}
588 
589 	/* Catch if_index overflow. */
590 	if (idx >= if_indexlim) {
591 		struct ifindex_entry *new, *old;
592 		int newlim;
593 
594 		newlim = if_indexlim * 2;
595 		new = malloc(newlim * sizeof(*new), M_IFNET, M_WAITOK | M_ZERO);
596 		memcpy(new, ifindex_table, if_indexlim * sizeof(*new));
597 		old = ifindex_table;
598 		ck_pr_store_ptr(&ifindex_table, new);
599 		if_indexlim = newlim;
600 		epoch_wait_preempt(net_epoch_preempt);
601 		free(old, M_IFNET);
602 	}
603 	if (idx > if_index)
604 		if_index = idx;
605 
606 	ifp->if_index = idx;
607 	ifp->if_idxgen = ifindex_table[idx].ife_gencnt;
608 	ck_pr_store_ptr(&ifindex_table[idx].ife_ifnet, ifp);
609 	IFNET_WUNLOCK();
610 
611 	return (ifp);
612 }
613 
614 struct ifnet *
615 if_alloc_dev(u_char type, device_t dev)
616 {
617 	int numa_domain;
618 
619 	if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0)
620 		return (if_alloc_domain(type, IF_NODOM));
621 	return (if_alloc_domain(type, numa_domain));
622 }
623 
624 struct ifnet *
625 if_alloc(u_char type)
626 {
627 
628 	return (if_alloc_domain(type, IF_NODOM));
629 }
630 /*
631  * Do the actual work of freeing a struct ifnet, and layer 2 common
632  * structure.  This call is made when the network epoch guarantees
633  * us that nobody holds a pointer to the interface.
634  */
635 static void
636 if_free_deferred(epoch_context_t ctx)
637 {
638 	struct ifnet *ifp = __containerof(ctx, struct ifnet, if_epoch_ctx);
639 
640 	KASSERT((ifp->if_flags & IFF_DYING),
641 	    ("%s: interface not dying", __func__));
642 
643 	if (if_com_free[ifp->if_alloctype] != NULL)
644 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
645 		    ifp->if_alloctype);
646 
647 #ifdef MAC
648 	mac_ifnet_destroy(ifp);
649 #endif /* MAC */
650 	IF_AFDATA_DESTROY(ifp);
651 	IF_ADDR_LOCK_DESTROY(ifp);
652 	ifq_delete(&ifp->if_snd);
653 
654 	for (int i = 0; i < IFCOUNTERS; i++)
655 		counter_u64_free(ifp->if_counters[i]);
656 
657 	if_freedescr(ifp->if_description);
658 	free(ifp->if_hw_addr, M_IFADDR);
659 	free(ifp, M_IFNET);
660 }
661 
662 /*
663  * Deregister an interface and free the associated storage.
664  */
665 void
666 if_free(struct ifnet *ifp)
667 {
668 
669 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
670 
671 	/*
672 	 * XXXGL: An interface index is really an alias to ifp pointer.
673 	 * Why would we clear the alias now, and not in the deferred
674 	 * context?  Indeed there is nothing wrong with some network
675 	 * thread obtaining ifp via ifnet_byindex() inside the network
676 	 * epoch and then dereferencing ifp while we perform if_free(),
677 	 * and after if_free() finished, too.
678 	 *
679 	 * This early index freeing was important back when ifindex was
680 	 * virtualized and interface would outlive the vnet.
681 	 */
682 	IFNET_WLOCK();
683 	MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
684 	ck_pr_store_ptr(&ifindex_table[ifp->if_index].ife_ifnet, NULL);
685 	ifindex_table[ifp->if_index].ife_gencnt++;
686 	while (if_index > 0 && ifindex_table[if_index].ife_ifnet == NULL)
687 		if_index--;
688 	IFNET_WUNLOCK();
689 
690 	if (refcount_release(&ifp->if_refcount))
691 		NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx);
692 }
693 
694 /*
695  * Interfaces to keep an ifnet type-stable despite the possibility of the
696  * driver calling if_free().  If there are additional references, we defer
697  * freeing the underlying data structure.
698  */
699 void
700 if_ref(struct ifnet *ifp)
701 {
702 	u_int old __diagused;
703 
704 	/* We don't assert the ifnet list lock here, but arguably should. */
705 	old = refcount_acquire(&ifp->if_refcount);
706 	KASSERT(old > 0, ("%s: ifp %p has 0 refs", __func__, ifp));
707 }
708 
709 bool
710 if_try_ref(struct ifnet *ifp)
711 {
712 	NET_EPOCH_ASSERT();
713 	return (refcount_acquire_if_not_zero(&ifp->if_refcount));
714 }
715 
716 void
717 if_rele(struct ifnet *ifp)
718 {
719 
720 	if (!refcount_release(&ifp->if_refcount))
721 		return;
722 	NET_EPOCH_CALL(if_free_deferred, &ifp->if_epoch_ctx);
723 }
724 
725 void
726 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
727 {
728 
729 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
730 
731 	if (ifq->ifq_maxlen == 0)
732 		ifq->ifq_maxlen = ifqmaxlen;
733 
734 	ifq->altq_type = 0;
735 	ifq->altq_disc = NULL;
736 	ifq->altq_flags &= ALTQF_CANTCHANGE;
737 	ifq->altq_tbr  = NULL;
738 	ifq->altq_ifp  = ifp;
739 }
740 
741 void
742 ifq_delete(struct ifaltq *ifq)
743 {
744 	mtx_destroy(&ifq->ifq_mtx);
745 }
746 
747 /*
748  * Perform generic interface initialization tasks and attach the interface
749  * to the list of "active" interfaces.  If vmove flag is set on entry
750  * to if_attach_internal(), perform only a limited subset of initialization
751  * tasks, given that we are moving from one vnet to another an ifnet which
752  * has already been fully initialized.
753  *
754  * Note that if_detach_internal() removes group membership unconditionally
755  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
756  * Thus, when if_vmove() is applied to a cloned interface, group membership
757  * is lost while a cloned one always joins a group whose name is
758  * ifc->ifc_name.  To recover this after if_detach_internal() and
759  * if_attach_internal(), the cloner should be specified to
760  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
761  * attempts to join a group whose name is ifc->ifc_name.
762  *
763  * XXX:
764  *  - The decision to return void and thus require this function to
765  *    succeed is questionable.
766  *  - We should probably do more sanity checking.  For instance we don't
767  *    do anything to insure if_xname is unique or non-empty.
768  */
769 void
770 if_attach(struct ifnet *ifp)
771 {
772 
773 	if_attach_internal(ifp, false);
774 }
775 
776 /*
777  * Compute the least common TSO limit.
778  */
779 void
780 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
781 {
782 	/*
783 	 * 1) If there is no limit currently, take the limit from
784 	 * the network adapter.
785 	 *
786 	 * 2) If the network adapter has a limit below the current
787 	 * limit, apply it.
788 	 */
789 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
790 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
791 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
792 	}
793 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
794 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
795 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
796 	}
797 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
798 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
799 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
800 	}
801 }
802 
803 /*
804  * Update TSO limit of a network adapter.
805  *
806  * Returns zero if no change. Else non-zero.
807  */
808 int
809 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
810 {
811 	int retval = 0;
812 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
813 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
814 		retval++;
815 	}
816 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
817 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
818 		retval++;
819 	}
820 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
821 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
822 		retval++;
823 	}
824 	return (retval);
825 }
826 
827 static void
828 if_attach_internal(struct ifnet *ifp, bool vmove)
829 {
830 	unsigned socksize, ifasize;
831 	int namelen, masklen;
832 	struct sockaddr_dl *sdl;
833 	struct ifaddr *ifa;
834 
835 	MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
836 
837 #ifdef VIMAGE
838 	ifp->if_vnet = curvnet;
839 	if (ifp->if_home_vnet == NULL)
840 		ifp->if_home_vnet = curvnet;
841 #endif
842 
843 	if_addgroup(ifp, IFG_ALL);
844 
845 #ifdef VIMAGE
846 	/* Restore group membership for cloned interface. */
847 	if (vmove)
848 		if_clone_restoregroup(ifp);
849 #endif
850 
851 	getmicrotime(&ifp->if_lastchange);
852 	ifp->if_epoch = time_uptime;
853 
854 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
855 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
856 	    ("transmit and qflush must both either be set or both be NULL"));
857 	if (ifp->if_transmit == NULL) {
858 		ifp->if_transmit = if_transmit_default;
859 		ifp->if_qflush = if_qflush;
860 	}
861 	if (ifp->if_input == NULL)
862 		ifp->if_input = if_input_default;
863 
864 	if (ifp->if_requestencap == NULL)
865 		ifp->if_requestencap = if_requestencap_default;
866 
867 	if (!vmove) {
868 #ifdef MAC
869 		mac_ifnet_create(ifp);
870 #endif
871 
872 		/*
873 		 * Create a Link Level name for this device.
874 		 */
875 		namelen = strlen(ifp->if_xname);
876 		/*
877 		 * Always save enough space for any possiable name so we
878 		 * can do a rename in place later.
879 		 */
880 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
881 		socksize = masklen + ifp->if_addrlen;
882 		if (socksize < sizeof(*sdl))
883 			socksize = sizeof(*sdl);
884 		socksize = roundup2(socksize, sizeof(long));
885 		ifasize = sizeof(*ifa) + 2 * socksize;
886 		ifa = ifa_alloc(ifasize, M_WAITOK);
887 		sdl = (struct sockaddr_dl *)(ifa + 1);
888 		sdl->sdl_len = socksize;
889 		sdl->sdl_family = AF_LINK;
890 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
891 		sdl->sdl_nlen = namelen;
892 		sdl->sdl_index = ifp->if_index;
893 		sdl->sdl_type = ifp->if_type;
894 		ifp->if_addr = ifa;
895 		ifa->ifa_ifp = ifp;
896 		ifa->ifa_addr = (struct sockaddr *)sdl;
897 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
898 		ifa->ifa_netmask = (struct sockaddr *)sdl;
899 		sdl->sdl_len = masklen;
900 		while (namelen != 0)
901 			sdl->sdl_data[--namelen] = 0xff;
902 		CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
903 		/* Reliably crash if used uninitialized. */
904 		ifp->if_broadcastaddr = NULL;
905 
906 		if (ifp->if_type == IFT_ETHER) {
907 			ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
908 			    M_WAITOK | M_ZERO);
909 		}
910 
911 #if defined(INET) || defined(INET6)
912 		/* Use defaults for TSO, if nothing is set */
913 		if (ifp->if_hw_tsomax == 0 &&
914 		    ifp->if_hw_tsomaxsegcount == 0 &&
915 		    ifp->if_hw_tsomaxsegsize == 0) {
916 			/*
917 			 * The TSO defaults needs to be such that an
918 			 * NFS mbuf list of 35 mbufs totalling just
919 			 * below 64K works and that a chain of mbufs
920 			 * can be defragged into at most 32 segments:
921 			 */
922 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
923 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
924 			ifp->if_hw_tsomaxsegcount = 35;
925 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
926 
927 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
928 			if (ifp->if_capabilities & IFCAP_TSO) {
929 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
930 				    ifp->if_hw_tsomax,
931 				    ifp->if_hw_tsomaxsegcount,
932 				    ifp->if_hw_tsomaxsegsize);
933 			}
934 		}
935 #endif
936 	}
937 #ifdef VIMAGE
938 	else {
939 		/*
940 		 * Update the interface index in the link layer address
941 		 * of the interface.
942 		 */
943 		for (ifa = ifp->if_addr; ifa != NULL;
944 		    ifa = CK_STAILQ_NEXT(ifa, ifa_link)) {
945 			if (ifa->ifa_addr->sa_family == AF_LINK) {
946 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
947 				sdl->sdl_index = ifp->if_index;
948 			}
949 		}
950 	}
951 #endif
952 
953 	if_link_ifnet(ifp);
954 
955 	if (domain_init_status >= 2)
956 		if_attachdomain1(ifp);
957 
958 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
959 	if (IS_DEFAULT_VNET(curvnet))
960 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
961 }
962 
963 static void
964 if_epochalloc(void *dummy __unused)
965 {
966 
967 	net_epoch_preempt = epoch_alloc("Net preemptible", EPOCH_PREEMPT);
968 }
969 SYSINIT(ifepochalloc, SI_SUB_EPOCH, SI_ORDER_ANY, if_epochalloc, NULL);
970 
971 static void
972 if_attachdomain(void *dummy)
973 {
974 	struct ifnet *ifp;
975 
976 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
977 		if_attachdomain1(ifp);
978 }
979 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
980     if_attachdomain, NULL);
981 
982 static void
983 if_attachdomain1(struct ifnet *ifp)
984 {
985 	struct domain *dp;
986 
987 	/*
988 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
989 	 * cannot lock ifp->if_afdata initialization, entirely.
990 	 */
991 	IF_AFDATA_LOCK(ifp);
992 	if (ifp->if_afdata_initialized >= domain_init_status) {
993 		IF_AFDATA_UNLOCK(ifp);
994 		log(LOG_WARNING, "%s called more than once on %s\n",
995 		    __func__, ifp->if_xname);
996 		return;
997 	}
998 	ifp->if_afdata_initialized = domain_init_status;
999 	IF_AFDATA_UNLOCK(ifp);
1000 
1001 	/* address family dependent data region */
1002 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
1003 	SLIST_FOREACH(dp, &domains, dom_next) {
1004 		if (dp->dom_ifattach)
1005 			ifp->if_afdata[dp->dom_family] =
1006 			    (*dp->dom_ifattach)(ifp);
1007 	}
1008 }
1009 
1010 /*
1011  * Remove any unicast or broadcast network addresses from an interface.
1012  */
1013 void
1014 if_purgeaddrs(struct ifnet *ifp)
1015 {
1016 	struct ifaddr *ifa;
1017 
1018 #ifdef INET6
1019 	/*
1020 	 * Need to leave multicast addresses of proxy NDP llentries
1021 	 * before in6_purgeifaddr() because the llentries are keys
1022 	 * for in6_multi objects of proxy NDP entries.
1023 	 * in6_purgeifaddr()s clean up llentries including proxy NDPs
1024 	 * then we would lose the keys if they are called earlier.
1025 	 */
1026 	in6_purge_proxy_ndp(ifp);
1027 #endif
1028 	while (1) {
1029 		struct epoch_tracker et;
1030 
1031 		NET_EPOCH_ENTER(et);
1032 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1033 			if (ifa->ifa_addr->sa_family != AF_LINK)
1034 				break;
1035 		}
1036 		NET_EPOCH_EXIT(et);
1037 
1038 		if (ifa == NULL)
1039 			break;
1040 #ifdef INET
1041 		/* XXX: Ugly!! ad hoc just for INET */
1042 		if (ifa->ifa_addr->sa_family == AF_INET) {
1043 			struct ifaliasreq ifr;
1044 
1045 			bzero(&ifr, sizeof(ifr));
1046 			ifr.ifra_addr = *ifa->ifa_addr;
1047 			if (ifa->ifa_dstaddr)
1048 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
1049 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
1050 			    NULL) == 0)
1051 				continue;
1052 		}
1053 #endif /* INET */
1054 #ifdef INET6
1055 		if (ifa->ifa_addr->sa_family == AF_INET6) {
1056 			in6_purgeifaddr((struct in6_ifaddr *)ifa);
1057 			/* ifp_addrhead is already updated */
1058 			continue;
1059 		}
1060 #endif /* INET6 */
1061 		IF_ADDR_WLOCK(ifp);
1062 		CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1063 		IF_ADDR_WUNLOCK(ifp);
1064 		ifa_free(ifa);
1065 	}
1066 }
1067 
1068 /*
1069  * Remove any multicast network addresses from an interface when an ifnet
1070  * is going away.
1071  */
1072 static void
1073 if_purgemaddrs(struct ifnet *ifp)
1074 {
1075 	struct ifmultiaddr *ifma;
1076 
1077 	IF_ADDR_WLOCK(ifp);
1078 	while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) {
1079 		ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs);
1080 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
1081 		if_delmulti_locked(ifp, ifma, 1);
1082 	}
1083 	IF_ADDR_WUNLOCK(ifp);
1084 }
1085 
1086 /*
1087  * Detach an interface, removing it from the list of "active" interfaces.
1088  * If vmove flag is set on entry to if_detach_internal(), perform only a
1089  * limited subset of cleanup tasks, given that we are moving an ifnet from
1090  * one vnet to another, where it must be fully operational.
1091  *
1092  * XXXRW: There are some significant questions about event ordering, and
1093  * how to prevent things from starting to use the interface during detach.
1094  */
1095 void
1096 if_detach(struct ifnet *ifp)
1097 {
1098 	bool found;
1099 
1100 	CURVNET_SET_QUIET(ifp->if_vnet);
1101 	found = if_unlink_ifnet(ifp, false);
1102 	if (found) {
1103 		sx_xlock(&ifnet_detach_sxlock);
1104 		if_detach_internal(ifp, false);
1105 		sx_xunlock(&ifnet_detach_sxlock);
1106 	}
1107 	CURVNET_RESTORE();
1108 }
1109 
1110 /*
1111  * The vmove flag, if set, indicates that we are called from a callpath
1112  * that is moving an interface to a different vnet instance.
1113  *
1114  * The shutdown flag, if set, indicates that we are called in the
1115  * process of shutting down a vnet instance.  Currently only the
1116  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
1117  * on a vnet instance shutdown without this flag being set, e.g., when
1118  * the cloned interfaces are destoyed as first thing of teardown.
1119  */
1120 static int
1121 if_detach_internal(struct ifnet *ifp, bool vmove)
1122 {
1123 	struct ifaddr *ifa;
1124 	int i;
1125 	struct domain *dp;
1126 #ifdef VIMAGE
1127 	bool shutdown;
1128 
1129 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1130 #endif
1131 
1132 	/*
1133 	 * At this point we know the interface still was on the ifnet list
1134 	 * and we removed it so we are in a stable state.
1135 	 */
1136 	epoch_wait_preempt(net_epoch_preempt);
1137 
1138 	/*
1139 	 * Ensure all pending EPOCH(9) callbacks have been executed. This
1140 	 * fixes issues about late destruction of multicast options
1141 	 * which lead to leave group calls, which in turn access the
1142 	 * belonging ifnet structure:
1143 	 */
1144 	NET_EPOCH_DRAIN_CALLBACKS();
1145 
1146 	/*
1147 	 * In any case (destroy or vmove) detach us from the groups
1148 	 * and remove/wait for pending events on the taskq.
1149 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
1150 	 */
1151 	if_delgroups(ifp);
1152 
1153 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1154 	taskqueue_drain(taskqueue_swi, &ifp->if_addmultitask);
1155 
1156 	if_down(ifp);
1157 
1158 #ifdef VIMAGE
1159 	/*
1160 	 * On VNET shutdown abort here as the stack teardown will do all
1161 	 * the work top-down for us.
1162 	 */
1163 	if (shutdown) {
1164 		/* Give interface users the chance to clean up. */
1165 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1166 
1167 		/*
1168 		 * In case of a vmove we are done here without error.
1169 		 * If we would signal an error it would lead to the same
1170 		 * abort as if we did not find the ifnet anymore.
1171 		 * if_detach() calls us in void context and does not care
1172 		 * about an early abort notification, so life is splendid :)
1173 		 */
1174 		goto finish_vnet_shutdown;
1175 	}
1176 #endif
1177 
1178 	/*
1179 	 * At this point we are not tearing down a VNET and are either
1180 	 * going to destroy or vmove the interface and have to cleanup
1181 	 * accordingly.
1182 	 */
1183 
1184 	/*
1185 	 * Remove routes and flush queues.
1186 	 */
1187 #ifdef ALTQ
1188 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1189 		altq_disable(&ifp->if_snd);
1190 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1191 		altq_detach(&ifp->if_snd);
1192 #endif
1193 
1194 	if_purgeaddrs(ifp);
1195 
1196 #ifdef INET
1197 	in_ifdetach(ifp);
1198 #endif
1199 
1200 #ifdef INET6
1201 	/*
1202 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
1203 	 * before removing routing entries below, since IPv6 interface direct
1204 	 * routes are expected to be removed by the IPv6-specific kernel API.
1205 	 * Otherwise, the kernel will detect some inconsistency and bark it.
1206 	 */
1207 	in6_ifdetach(ifp);
1208 #endif
1209 	if_purgemaddrs(ifp);
1210 
1211 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1212 	if (IS_DEFAULT_VNET(curvnet))
1213 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1214 
1215 	if (!vmove) {
1216 		/*
1217 		 * Prevent further calls into the device driver via ifnet.
1218 		 */
1219 		if_dead(ifp);
1220 
1221 		/*
1222 		 * Clean up all addresses.
1223 		 */
1224 		IF_ADDR_WLOCK(ifp);
1225 		if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) {
1226 			ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
1227 			CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1228 			IF_ADDR_WUNLOCK(ifp);
1229 			ifa_free(ifa);
1230 		} else
1231 			IF_ADDR_WUNLOCK(ifp);
1232 	}
1233 
1234 	rt_flushifroutes(ifp);
1235 
1236 #ifdef VIMAGE
1237 finish_vnet_shutdown:
1238 #endif
1239 	/*
1240 	 * We cannot hold the lock over dom_ifdetach calls as they might
1241 	 * sleep, for example trying to drain a callout, thus open up the
1242 	 * theoretical race with re-attaching.
1243 	 */
1244 	IF_AFDATA_LOCK(ifp);
1245 	i = ifp->if_afdata_initialized;
1246 	ifp->if_afdata_initialized = 0;
1247 	IF_AFDATA_UNLOCK(ifp);
1248 	if (i == 0)
1249 		return (0);
1250 	SLIST_FOREACH(dp, &domains, dom_next) {
1251 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1252 			(*dp->dom_ifdetach)(ifp,
1253 			    ifp->if_afdata[dp->dom_family]);
1254 			ifp->if_afdata[dp->dom_family] = NULL;
1255 		}
1256 	}
1257 
1258 	return (0);
1259 }
1260 
1261 #ifdef VIMAGE
1262 /*
1263  * if_vmove() performs a limited version of if_detach() in current
1264  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1265  */
1266 static int
1267 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1268 {
1269 #ifdef DEV_BPF
1270 	u_int bif_dlt, bif_hdrlen;
1271 #endif
1272 	int rc;
1273 
1274 #ifdef DEV_BPF
1275  	/*
1276 	 * if_detach_internal() will call the eventhandler to notify
1277 	 * interface departure.  That will detach if_bpf.  We need to
1278 	 * safe the dlt and hdrlen so we can re-attach it later.
1279 	 */
1280 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1281 #endif
1282 
1283 	/*
1284 	 * Detach from current vnet, but preserve LLADDR info, do not
1285 	 * mark as dead etc. so that the ifnet can be reattached later.
1286 	 * If we cannot find it, we lost the race to someone else.
1287 	 */
1288 	rc = if_detach_internal(ifp, true);
1289 	if (rc != 0)
1290 		return (rc);
1291 
1292 	/*
1293 	 * Perform interface-specific reassignment tasks, if provided by
1294 	 * the driver.
1295 	 */
1296 	if (ifp->if_reassign != NULL)
1297 		ifp->if_reassign(ifp, new_vnet, NULL);
1298 
1299 	/*
1300 	 * Switch to the context of the target vnet.
1301 	 */
1302 	CURVNET_SET_QUIET(new_vnet);
1303 	if_attach_internal(ifp, true);
1304 
1305 #ifdef DEV_BPF
1306 	if (ifp->if_bpf == NULL)
1307 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1308 #endif
1309 
1310 	CURVNET_RESTORE();
1311 	return (0);
1312 }
1313 
1314 /*
1315  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1316  */
1317 static int
1318 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1319 {
1320 	struct prison *pr;
1321 	struct ifnet *difp;
1322 	int error;
1323 	bool found __diagused;
1324 	bool shutdown;
1325 
1326 	MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
1327 
1328 	/* Try to find the prison within our visibility. */
1329 	sx_slock(&allprison_lock);
1330 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1331 	sx_sunlock(&allprison_lock);
1332 	if (pr == NULL)
1333 		return (ENXIO);
1334 	prison_hold_locked(pr);
1335 	mtx_unlock(&pr->pr_mtx);
1336 
1337 	/* Do not try to move the iface from and to the same prison. */
1338 	if (pr->pr_vnet == ifp->if_vnet) {
1339 		prison_free(pr);
1340 		return (EEXIST);
1341 	}
1342 
1343 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1344 	/* XXX Lock interfaces to avoid races. */
1345 	CURVNET_SET_QUIET(pr->pr_vnet);
1346 	difp = ifunit(ifname);
1347 	if (difp != NULL) {
1348 		CURVNET_RESTORE();
1349 		prison_free(pr);
1350 		return (EEXIST);
1351 	}
1352 	sx_xlock(&ifnet_detach_sxlock);
1353 
1354 	/* Make sure the VNET is stable. */
1355 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1356 	if (shutdown) {
1357 		sx_xunlock(&ifnet_detach_sxlock);
1358 		CURVNET_RESTORE();
1359 		prison_free(pr);
1360 		return (EBUSY);
1361 	}
1362 	CURVNET_RESTORE();
1363 
1364 	found = if_unlink_ifnet(ifp, true);
1365 	if (! found) {
1366 		sx_xunlock(&ifnet_detach_sxlock);
1367 		CURVNET_RESTORE();
1368 		prison_free(pr);
1369 		return (ENODEV);
1370 	}
1371 
1372 	/* Move the interface into the child jail/vnet. */
1373 	error = if_vmove(ifp, pr->pr_vnet);
1374 
1375 	/* Report the new if_xname back to the userland on success. */
1376 	if (error == 0)
1377 		sprintf(ifname, "%s", ifp->if_xname);
1378 
1379 	sx_xunlock(&ifnet_detach_sxlock);
1380 
1381 	prison_free(pr);
1382 	return (error);
1383 }
1384 
1385 static int
1386 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1387 {
1388 	struct prison *pr;
1389 	struct vnet *vnet_dst;
1390 	struct ifnet *ifp;
1391 	int error, found __diagused;
1392  	bool shutdown;
1393 
1394 	/* Try to find the prison within our visibility. */
1395 	sx_slock(&allprison_lock);
1396 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1397 	sx_sunlock(&allprison_lock);
1398 	if (pr == NULL)
1399 		return (ENXIO);
1400 	prison_hold_locked(pr);
1401 	mtx_unlock(&pr->pr_mtx);
1402 
1403 	/* Make sure the named iface exists in the source prison/vnet. */
1404 	CURVNET_SET(pr->pr_vnet);
1405 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1406 	if (ifp == NULL) {
1407 		CURVNET_RESTORE();
1408 		prison_free(pr);
1409 		return (ENXIO);
1410 	}
1411 
1412 	/* Do not try to move the iface from and to the same prison. */
1413 	vnet_dst = TD_TO_VNET(td);
1414 	if (vnet_dst == ifp->if_vnet) {
1415 		CURVNET_RESTORE();
1416 		prison_free(pr);
1417 		return (EEXIST);
1418 	}
1419 
1420 	/* Make sure the VNET is stable. */
1421 	shutdown = VNET_IS_SHUTTING_DOWN(ifp->if_vnet);
1422 	if (shutdown) {
1423 		CURVNET_RESTORE();
1424 		prison_free(pr);
1425 		return (EBUSY);
1426 	}
1427 
1428 	/* Get interface back from child jail/vnet. */
1429 	found = if_unlink_ifnet(ifp, true);
1430 	MPASS(found);
1431 	sx_xlock(&ifnet_detach_sxlock);
1432 	error = if_vmove(ifp, vnet_dst);
1433 	sx_xunlock(&ifnet_detach_sxlock);
1434 	CURVNET_RESTORE();
1435 
1436 	/* Report the new if_xname back to the userland on success. */
1437 	if (error == 0)
1438 		sprintf(ifname, "%s", ifp->if_xname);
1439 
1440 	prison_free(pr);
1441 	return (error);
1442 }
1443 #endif /* VIMAGE */
1444 
1445 /*
1446  * Add a group to an interface
1447  */
1448 int
1449 if_addgroup(struct ifnet *ifp, const char *groupname)
1450 {
1451 	struct ifg_list		*ifgl;
1452 	struct ifg_group	*ifg = NULL;
1453 	struct ifg_member	*ifgm;
1454 	int 			 new = 0;
1455 
1456 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1457 	    groupname[strlen(groupname) - 1] <= '9')
1458 		return (EINVAL);
1459 
1460 	IFNET_WLOCK();
1461 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1462 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1463 			IFNET_WUNLOCK();
1464 			return (EEXIST);
1465 		}
1466 
1467 	if ((ifgl = malloc(sizeof(*ifgl), M_TEMP, M_NOWAIT)) == NULL) {
1468 	    	IFNET_WUNLOCK();
1469 		return (ENOMEM);
1470 	}
1471 
1472 	if ((ifgm = malloc(sizeof(*ifgm), M_TEMP, M_NOWAIT)) == NULL) {
1473 		free(ifgl, M_TEMP);
1474 		IFNET_WUNLOCK();
1475 		return (ENOMEM);
1476 	}
1477 
1478 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1479 		if (!strcmp(ifg->ifg_group, groupname))
1480 			break;
1481 
1482 	if (ifg == NULL) {
1483 		if ((ifg = malloc(sizeof(*ifg), M_TEMP, M_NOWAIT)) == NULL) {
1484 			free(ifgl, M_TEMP);
1485 			free(ifgm, M_TEMP);
1486 			IFNET_WUNLOCK();
1487 			return (ENOMEM);
1488 		}
1489 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1490 		ifg->ifg_refcnt = 0;
1491 		CK_STAILQ_INIT(&ifg->ifg_members);
1492 		CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1493 		new = 1;
1494 	}
1495 
1496 	ifg->ifg_refcnt++;
1497 	ifgl->ifgl_group = ifg;
1498 	ifgm->ifgm_ifp = ifp;
1499 
1500 	IF_ADDR_WLOCK(ifp);
1501 	CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1502 	CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1503 	IF_ADDR_WUNLOCK(ifp);
1504 
1505 	IFNET_WUNLOCK();
1506 
1507 	if (new)
1508 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1509 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1510 
1511 	return (0);
1512 }
1513 
1514 /*
1515  * Helper function to remove a group out of an interface.  Expects the global
1516  * ifnet lock to be write-locked, and drops it before returning.
1517  */
1518 static void
1519 _if_delgroup_locked(struct ifnet *ifp, struct ifg_list *ifgl,
1520     const char *groupname)
1521 {
1522 	struct ifg_member *ifgm;
1523 	bool freeifgl;
1524 
1525 	IFNET_WLOCK_ASSERT();
1526 
1527 	IF_ADDR_WLOCK(ifp);
1528 	CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1529 	IF_ADDR_WUNLOCK(ifp);
1530 
1531 	CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) {
1532 		if (ifgm->ifgm_ifp == ifp) {
1533 			CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1534 			    ifg_member, ifgm_next);
1535 			break;
1536 		}
1537 	}
1538 
1539 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1540 		CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group,
1541 		    ifg_next);
1542 		freeifgl = true;
1543 	} else {
1544 		freeifgl = false;
1545 	}
1546 	IFNET_WUNLOCK();
1547 
1548 	epoch_wait_preempt(net_epoch_preempt);
1549 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1550 	if (freeifgl) {
1551 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1552 		free(ifgl->ifgl_group, M_TEMP);
1553 	}
1554 	free(ifgm, M_TEMP);
1555 	free(ifgl, M_TEMP);
1556 }
1557 
1558 /*
1559  * Remove a group from an interface
1560  */
1561 int
1562 if_delgroup(struct ifnet *ifp, const char *groupname)
1563 {
1564 	struct ifg_list *ifgl;
1565 
1566 	IFNET_WLOCK();
1567 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1568 		if (strcmp(ifgl->ifgl_group->ifg_group, groupname) == 0)
1569 			break;
1570 	if (ifgl == NULL) {
1571 		IFNET_WUNLOCK();
1572 		return (ENOENT);
1573 	}
1574 
1575 	_if_delgroup_locked(ifp, ifgl, groupname);
1576 
1577 	return (0);
1578 }
1579 
1580 /*
1581  * Remove an interface from all groups
1582  */
1583 static void
1584 if_delgroups(struct ifnet *ifp)
1585 {
1586 	struct ifg_list *ifgl;
1587 	char groupname[IFNAMSIZ];
1588 
1589 	IFNET_WLOCK();
1590 	while ((ifgl = CK_STAILQ_FIRST(&ifp->if_groups)) != NULL) {
1591 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1592 		_if_delgroup_locked(ifp, ifgl, groupname);
1593 		IFNET_WLOCK();
1594 	}
1595 	IFNET_WUNLOCK();
1596 }
1597 
1598 /*
1599  * Stores all groups from an interface in memory pointed to by ifgr.
1600  */
1601 static int
1602 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp)
1603 {
1604 	int			 len, error;
1605 	struct ifg_list		*ifgl;
1606 	struct ifg_req		 ifgrq, *ifgp;
1607 
1608 	NET_EPOCH_ASSERT();
1609 
1610 	if (ifgr->ifgr_len == 0) {
1611 		CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1612 			ifgr->ifgr_len += sizeof(struct ifg_req);
1613 		return (0);
1614 	}
1615 
1616 	len = ifgr->ifgr_len;
1617 	ifgp = ifgr->ifgr_groups;
1618 	/* XXX: wire */
1619 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1620 		if (len < sizeof(ifgrq))
1621 			return (EINVAL);
1622 		bzero(&ifgrq, sizeof ifgrq);
1623 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1624 		    sizeof(ifgrq.ifgrq_group));
1625 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req))))
1626 			return (error);
1627 		len -= sizeof(ifgrq);
1628 		ifgp++;
1629 	}
1630 
1631 	return (0);
1632 }
1633 
1634 /*
1635  * Stores all members of a group in memory pointed to by igfr
1636  */
1637 static int
1638 if_getgroupmembers(struct ifgroupreq *ifgr)
1639 {
1640 	struct ifg_group	*ifg;
1641 	struct ifg_member	*ifgm;
1642 	struct ifg_req		 ifgrq, *ifgp;
1643 	int			 len, error;
1644 
1645 	IFNET_RLOCK();
1646 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1647 		if (strcmp(ifg->ifg_group, ifgr->ifgr_name) == 0)
1648 			break;
1649 	if (ifg == NULL) {
1650 		IFNET_RUNLOCK();
1651 		return (ENOENT);
1652 	}
1653 
1654 	if (ifgr->ifgr_len == 0) {
1655 		CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1656 			ifgr->ifgr_len += sizeof(ifgrq);
1657 		IFNET_RUNLOCK();
1658 		return (0);
1659 	}
1660 
1661 	len = ifgr->ifgr_len;
1662 	ifgp = ifgr->ifgr_groups;
1663 	CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1664 		if (len < sizeof(ifgrq)) {
1665 			IFNET_RUNLOCK();
1666 			return (EINVAL);
1667 		}
1668 		bzero(&ifgrq, sizeof ifgrq);
1669 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1670 		    sizeof(ifgrq.ifgrq_member));
1671 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1672 			IFNET_RUNLOCK();
1673 			return (error);
1674 		}
1675 		len -= sizeof(ifgrq);
1676 		ifgp++;
1677 	}
1678 	IFNET_RUNLOCK();
1679 
1680 	return (0);
1681 }
1682 
1683 /*
1684  * Return counter values from counter(9)s stored in ifnet.
1685  */
1686 uint64_t
1687 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1688 {
1689 
1690 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1691 
1692 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1693 }
1694 
1695 /*
1696  * Increase an ifnet counter. Usually used for counters shared
1697  * between the stack and a driver, but function supports them all.
1698  */
1699 void
1700 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1701 {
1702 
1703 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1704 
1705 	counter_u64_add(ifp->if_counters[cnt], inc);
1706 }
1707 
1708 /*
1709  * Copy data from ifnet to userland API structure if_data.
1710  */
1711 void
1712 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1713 {
1714 
1715 	ifd->ifi_type = ifp->if_type;
1716 	ifd->ifi_physical = 0;
1717 	ifd->ifi_addrlen = ifp->if_addrlen;
1718 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1719 	ifd->ifi_link_state = ifp->if_link_state;
1720 	ifd->ifi_vhid = 0;
1721 	ifd->ifi_datalen = sizeof(struct if_data);
1722 	ifd->ifi_mtu = ifp->if_mtu;
1723 	ifd->ifi_metric = ifp->if_metric;
1724 	ifd->ifi_baudrate = ifp->if_baudrate;
1725 	ifd->ifi_hwassist = ifp->if_hwassist;
1726 	ifd->ifi_epoch = ifp->if_epoch;
1727 	ifd->ifi_lastchange = ifp->if_lastchange;
1728 
1729 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1730 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1731 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1732 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1733 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1734 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1735 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1736 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1737 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1738 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1739 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1740 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1741 }
1742 
1743 /*
1744  * Initialization, destruction and refcounting functions for ifaddrs.
1745  */
1746 struct ifaddr *
1747 ifa_alloc(size_t size, int flags)
1748 {
1749 	struct ifaddr *ifa;
1750 
1751 	KASSERT(size >= sizeof(struct ifaddr),
1752 	    ("%s: invalid size %zu", __func__, size));
1753 
1754 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1755 	if (ifa == NULL)
1756 		return (NULL);
1757 
1758 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1759 		goto fail;
1760 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1761 		goto fail;
1762 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1763 		goto fail;
1764 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1765 		goto fail;
1766 
1767 	refcount_init(&ifa->ifa_refcnt, 1);
1768 
1769 	return (ifa);
1770 
1771 fail:
1772 	/* free(NULL) is okay */
1773 	counter_u64_free(ifa->ifa_opackets);
1774 	counter_u64_free(ifa->ifa_ipackets);
1775 	counter_u64_free(ifa->ifa_obytes);
1776 	counter_u64_free(ifa->ifa_ibytes);
1777 	free(ifa, M_IFADDR);
1778 
1779 	return (NULL);
1780 }
1781 
1782 void
1783 ifa_ref(struct ifaddr *ifa)
1784 {
1785 	u_int old __diagused;
1786 
1787 	old = refcount_acquire(&ifa->ifa_refcnt);
1788 	KASSERT(old > 0, ("%s: ifa %p has 0 refs", __func__, ifa));
1789 }
1790 
1791 int
1792 ifa_try_ref(struct ifaddr *ifa)
1793 {
1794 
1795 	NET_EPOCH_ASSERT();
1796 	return (refcount_acquire_if_not_zero(&ifa->ifa_refcnt));
1797 }
1798 
1799 static void
1800 ifa_destroy(epoch_context_t ctx)
1801 {
1802 	struct ifaddr *ifa;
1803 
1804 	ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx);
1805 	counter_u64_free(ifa->ifa_opackets);
1806 	counter_u64_free(ifa->ifa_ipackets);
1807 	counter_u64_free(ifa->ifa_obytes);
1808 	counter_u64_free(ifa->ifa_ibytes);
1809 	free(ifa, M_IFADDR);
1810 }
1811 
1812 void
1813 ifa_free(struct ifaddr *ifa)
1814 {
1815 
1816 	if (refcount_release(&ifa->ifa_refcnt))
1817 		NET_EPOCH_CALL(ifa_destroy, &ifa->ifa_epoch_ctx);
1818 }
1819 
1820 /*
1821  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1822  * structs used to represent other address families, it is necessary
1823  * to perform a different comparison.
1824  */
1825 
1826 #define	sa_dl_equal(a1, a2)	\
1827 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1828 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1829 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1830 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1831 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1832 
1833 /*
1834  * Locate an interface based on a complete address.
1835  */
1836 /*ARGSUSED*/
1837 struct ifaddr *
1838 ifa_ifwithaddr(const struct sockaddr *addr)
1839 {
1840 	struct ifnet *ifp;
1841 	struct ifaddr *ifa;
1842 
1843 	NET_EPOCH_ASSERT();
1844 
1845 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1846 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1847 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1848 				continue;
1849 			if (sa_equal(addr, ifa->ifa_addr)) {
1850 				goto done;
1851 			}
1852 			/* IP6 doesn't have broadcast */
1853 			if ((ifp->if_flags & IFF_BROADCAST) &&
1854 			    ifa->ifa_broadaddr &&
1855 			    ifa->ifa_broadaddr->sa_len != 0 &&
1856 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1857 				goto done;
1858 			}
1859 		}
1860 	}
1861 	ifa = NULL;
1862 done:
1863 	return (ifa);
1864 }
1865 
1866 int
1867 ifa_ifwithaddr_check(const struct sockaddr *addr)
1868 {
1869 	struct epoch_tracker et;
1870 	int rc;
1871 
1872 	NET_EPOCH_ENTER(et);
1873 	rc = (ifa_ifwithaddr(addr) != NULL);
1874 	NET_EPOCH_EXIT(et);
1875 	return (rc);
1876 }
1877 
1878 /*
1879  * Locate an interface based on the broadcast address.
1880  */
1881 /* ARGSUSED */
1882 struct ifaddr *
1883 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
1884 {
1885 	struct ifnet *ifp;
1886 	struct ifaddr *ifa;
1887 
1888 	NET_EPOCH_ASSERT();
1889 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1890 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1891 			continue;
1892 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1893 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1894 				continue;
1895 			if ((ifp->if_flags & IFF_BROADCAST) &&
1896 			    ifa->ifa_broadaddr &&
1897 			    ifa->ifa_broadaddr->sa_len != 0 &&
1898 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1899 				goto done;
1900 			}
1901 		}
1902 	}
1903 	ifa = NULL;
1904 done:
1905 	return (ifa);
1906 }
1907 
1908 /*
1909  * Locate the point to point interface with a given destination address.
1910  */
1911 /*ARGSUSED*/
1912 struct ifaddr *
1913 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1914 {
1915 	struct ifnet *ifp;
1916 	struct ifaddr *ifa;
1917 
1918 	NET_EPOCH_ASSERT();
1919 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1920 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1921 			continue;
1922 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1923 			continue;
1924 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1925 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1926 				continue;
1927 			if (ifa->ifa_dstaddr != NULL &&
1928 			    sa_equal(addr, ifa->ifa_dstaddr)) {
1929 				goto done;
1930 			}
1931 		}
1932 	}
1933 	ifa = NULL;
1934 done:
1935 	return (ifa);
1936 }
1937 
1938 /*
1939  * Find an interface on a specific network.  If many, choice
1940  * is most specific found.
1941  */
1942 struct ifaddr *
1943 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
1944 {
1945 	struct ifnet *ifp;
1946 	struct ifaddr *ifa;
1947 	struct ifaddr *ifa_maybe = NULL;
1948 	u_int af = addr->sa_family;
1949 	const char *addr_data = addr->sa_data, *cplim;
1950 
1951 	NET_EPOCH_ASSERT();
1952 	/*
1953 	 * AF_LINK addresses can be looked up directly by their index number,
1954 	 * so do that if we can.
1955 	 */
1956 	if (af == AF_LINK) {
1957 		ifp = ifnet_byindex(
1958 		    ((const struct sockaddr_dl *)addr)->sdl_index);
1959 		return (ifp ? ifp->if_addr : NULL);
1960 	}
1961 
1962 	/*
1963 	 * Scan though each interface, looking for ones that have addresses
1964 	 * in this address family and the requested fib.
1965 	 */
1966 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1967 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
1968 			continue;
1969 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1970 			const char *cp, *cp2, *cp3;
1971 
1972 			if (ifa->ifa_addr->sa_family != af)
1973 next:				continue;
1974 			if (af == AF_INET &&
1975 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1976 				/*
1977 				 * This is a bit broken as it doesn't
1978 				 * take into account that the remote end may
1979 				 * be a single node in the network we are
1980 				 * looking for.
1981 				 * The trouble is that we don't know the
1982 				 * netmask for the remote end.
1983 				 */
1984 				if (ifa->ifa_dstaddr != NULL &&
1985 				    sa_equal(addr, ifa->ifa_dstaddr)) {
1986 					goto done;
1987 				}
1988 			} else {
1989 				/*
1990 				 * Scan all the bits in the ifa's address.
1991 				 * If a bit dissagrees with what we are
1992 				 * looking for, mask it with the netmask
1993 				 * to see if it really matters.
1994 				 * (A byte at a time)
1995 				 */
1996 				if (ifa->ifa_netmask == 0)
1997 					continue;
1998 				cp = addr_data;
1999 				cp2 = ifa->ifa_addr->sa_data;
2000 				cp3 = ifa->ifa_netmask->sa_data;
2001 				cplim = ifa->ifa_netmask->sa_len
2002 					+ (char *)ifa->ifa_netmask;
2003 				while (cp3 < cplim)
2004 					if ((*cp++ ^ *cp2++) & *cp3++)
2005 						goto next; /* next address! */
2006 				/*
2007 				 * If the netmask of what we just found
2008 				 * is more specific than what we had before
2009 				 * (if we had one), or if the virtual status
2010 				 * of new prefix is better than of the old one,
2011 				 * then remember the new one before continuing
2012 				 * to search for an even better one.
2013 				 */
2014 				if (ifa_maybe == NULL ||
2015 				    ifa_preferred(ifa_maybe, ifa) ||
2016 				    rn_refines((caddr_t)ifa->ifa_netmask,
2017 				    (caddr_t)ifa_maybe->ifa_netmask)) {
2018 					ifa_maybe = ifa;
2019 				}
2020 			}
2021 		}
2022 	}
2023 	ifa = ifa_maybe;
2024 	ifa_maybe = NULL;
2025 done:
2026 	return (ifa);
2027 }
2028 
2029 /*
2030  * Find an interface address specific to an interface best matching
2031  * a given address.
2032  */
2033 struct ifaddr *
2034 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
2035 {
2036 	struct ifaddr *ifa;
2037 	const char *cp, *cp2, *cp3;
2038 	char *cplim;
2039 	struct ifaddr *ifa_maybe = NULL;
2040 	u_int af = addr->sa_family;
2041 
2042 	if (af >= AF_MAX)
2043 		return (NULL);
2044 
2045 	NET_EPOCH_ASSERT();
2046 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2047 		if (ifa->ifa_addr->sa_family != af)
2048 			continue;
2049 		if (ifa_maybe == NULL)
2050 			ifa_maybe = ifa;
2051 		if (ifa->ifa_netmask == 0) {
2052 			if (sa_equal(addr, ifa->ifa_addr) ||
2053 			    (ifa->ifa_dstaddr &&
2054 			    sa_equal(addr, ifa->ifa_dstaddr)))
2055 				goto done;
2056 			continue;
2057 		}
2058 		if (ifp->if_flags & IFF_POINTOPOINT) {
2059 			if (ifa->ifa_dstaddr && sa_equal(addr, ifa->ifa_dstaddr))
2060 				goto done;
2061 		} else {
2062 			cp = addr->sa_data;
2063 			cp2 = ifa->ifa_addr->sa_data;
2064 			cp3 = ifa->ifa_netmask->sa_data;
2065 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2066 			for (; cp3 < cplim; cp3++)
2067 				if ((*cp++ ^ *cp2++) & *cp3)
2068 					break;
2069 			if (cp3 == cplim)
2070 				goto done;
2071 		}
2072 	}
2073 	ifa = ifa_maybe;
2074 done:
2075 	return (ifa);
2076 }
2077 
2078 /*
2079  * See whether new ifa is better than current one:
2080  * 1) A non-virtual one is preferred over virtual.
2081  * 2) A virtual in master state preferred over any other state.
2082  *
2083  * Used in several address selecting functions.
2084  */
2085 int
2086 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2087 {
2088 
2089 	return (cur->ifa_carp && (!next->ifa_carp ||
2090 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2091 }
2092 
2093 struct sockaddr_dl *
2094 link_alloc_sdl(size_t size, int flags)
2095 {
2096 
2097 	return (malloc(size, M_TEMP, flags));
2098 }
2099 
2100 void
2101 link_free_sdl(struct sockaddr *sa)
2102 {
2103 	free(sa, M_TEMP);
2104 }
2105 
2106 /*
2107  * Fills in given sdl with interface basic info.
2108  * Returns pointer to filled sdl.
2109  */
2110 struct sockaddr_dl *
2111 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2112 {
2113 	struct sockaddr_dl *sdl;
2114 
2115 	sdl = (struct sockaddr_dl *)paddr;
2116 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2117 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2118 	sdl->sdl_family = AF_LINK;
2119 	sdl->sdl_index = ifp->if_index;
2120 	sdl->sdl_type = iftype;
2121 
2122 	return (sdl);
2123 }
2124 
2125 /*
2126  * Mark an interface down and notify protocols of
2127  * the transition.
2128  */
2129 static void
2130 if_unroute(struct ifnet *ifp, int flag, int fam)
2131 {
2132 
2133 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2134 
2135 	ifp->if_flags &= ~flag;
2136 	getmicrotime(&ifp->if_lastchange);
2137 	ifp->if_qflush(ifp);
2138 
2139 	if (ifp->if_carp)
2140 		(*carp_linkstate_p)(ifp);
2141 	rt_ifmsg(ifp, IFF_UP);
2142 }
2143 
2144 /*
2145  * Mark an interface up and notify protocols of
2146  * the transition.
2147  */
2148 static void
2149 if_route(struct ifnet *ifp, int flag, int fam)
2150 {
2151 
2152 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2153 
2154 	ifp->if_flags |= flag;
2155 	getmicrotime(&ifp->if_lastchange);
2156 	if (ifp->if_carp)
2157 		(*carp_linkstate_p)(ifp);
2158 	rt_ifmsg(ifp, IFF_UP);
2159 #ifdef INET6
2160 	in6_if_up(ifp);
2161 #endif
2162 }
2163 
2164 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2165 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2166 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2167 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2168 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2169 int	(*vlan_pcp_p)(struct ifnet *, uint16_t *);
2170 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2171 void	*(*vlan_cookie_p)(struct ifnet *);
2172 
2173 /*
2174  * Handle a change in the interface link state. To avoid LORs
2175  * between driver lock and upper layer locks, as well as possible
2176  * recursions, we post event to taskqueue, and all job
2177  * is done in static do_link_state_change().
2178  */
2179 void
2180 if_link_state_change(struct ifnet *ifp, int link_state)
2181 {
2182 	/* Return if state hasn't changed. */
2183 	if (ifp->if_link_state == link_state)
2184 		return;
2185 
2186 	ifp->if_link_state = link_state;
2187 
2188 	/* XXXGL: reference ifp? */
2189 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2190 }
2191 
2192 static void
2193 do_link_state_change(void *arg, int pending)
2194 {
2195 	struct ifnet *ifp;
2196 	int link_state;
2197 
2198 	ifp = arg;
2199 	link_state = ifp->if_link_state;
2200 
2201 	CURVNET_SET(ifp->if_vnet);
2202 	rt_ifmsg(ifp, 0);
2203 	if (ifp->if_vlantrunk != NULL)
2204 		(*vlan_link_state_p)(ifp);
2205 
2206 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2207 	    ifp->if_l2com != NULL)
2208 		(*ng_ether_link_state_p)(ifp, link_state);
2209 	if (ifp->if_carp)
2210 		(*carp_linkstate_p)(ifp);
2211 	if (ifp->if_bridge)
2212 		ifp->if_bridge_linkstate(ifp);
2213 	if (ifp->if_lagg)
2214 		(*lagg_linkstate_p)(ifp, link_state);
2215 
2216 	if (IS_DEFAULT_VNET(curvnet))
2217 		devctl_notify("IFNET", ifp->if_xname,
2218 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2219 		    NULL);
2220 	if (pending > 1)
2221 		if_printf(ifp, "%d link states coalesced\n", pending);
2222 	if (log_link_state_change)
2223 		if_printf(ifp, "link state changed to %s\n",
2224 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2225 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2226 	CURVNET_RESTORE();
2227 }
2228 
2229 /*
2230  * Mark an interface down and notify protocols of
2231  * the transition.
2232  */
2233 void
2234 if_down(struct ifnet *ifp)
2235 {
2236 
2237 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2238 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2239 }
2240 
2241 /*
2242  * Mark an interface up and notify protocols of
2243  * the transition.
2244  */
2245 void
2246 if_up(struct ifnet *ifp)
2247 {
2248 
2249 	if_route(ifp, IFF_UP, AF_UNSPEC);
2250 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2251 }
2252 
2253 /*
2254  * Flush an interface queue.
2255  */
2256 void
2257 if_qflush(struct ifnet *ifp)
2258 {
2259 	struct mbuf *m, *n;
2260 	struct ifaltq *ifq;
2261 
2262 	ifq = &ifp->if_snd;
2263 	IFQ_LOCK(ifq);
2264 #ifdef ALTQ
2265 	if (ALTQ_IS_ENABLED(ifq))
2266 		ALTQ_PURGE(ifq);
2267 #endif
2268 	n = ifq->ifq_head;
2269 	while ((m = n) != NULL) {
2270 		n = m->m_nextpkt;
2271 		m_freem(m);
2272 	}
2273 	ifq->ifq_head = 0;
2274 	ifq->ifq_tail = 0;
2275 	ifq->ifq_len = 0;
2276 	IFQ_UNLOCK(ifq);
2277 }
2278 
2279 /*
2280  * Map interface name to interface structure pointer, with or without
2281  * returning a reference.
2282  */
2283 struct ifnet *
2284 ifunit_ref(const char *name)
2285 {
2286 	struct epoch_tracker et;
2287 	struct ifnet *ifp;
2288 
2289 	NET_EPOCH_ENTER(et);
2290 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2291 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2292 		    !(ifp->if_flags & IFF_DYING))
2293 			break;
2294 	}
2295 	if (ifp != NULL) {
2296 		if_ref(ifp);
2297 		MPASS(ifindex_table[ifp->if_index].ife_ifnet == ifp);
2298 	}
2299 
2300 	NET_EPOCH_EXIT(et);
2301 	return (ifp);
2302 }
2303 
2304 struct ifnet *
2305 ifunit(const char *name)
2306 {
2307 	struct epoch_tracker et;
2308 	struct ifnet *ifp;
2309 
2310 	NET_EPOCH_ENTER(et);
2311 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2312 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2313 			break;
2314 	}
2315 	NET_EPOCH_EXIT(et);
2316 	return (ifp);
2317 }
2318 
2319 void *
2320 ifr_buffer_get_buffer(void *data)
2321 {
2322 	union ifreq_union *ifrup;
2323 
2324 	ifrup = data;
2325 #ifdef COMPAT_FREEBSD32
2326 	if (SV_CURPROC_FLAG(SV_ILP32))
2327 		return ((void *)(uintptr_t)
2328 		    ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2329 #endif
2330 	return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2331 }
2332 
2333 static void
2334 ifr_buffer_set_buffer_null(void *data)
2335 {
2336 	union ifreq_union *ifrup;
2337 
2338 	ifrup = data;
2339 #ifdef COMPAT_FREEBSD32
2340 	if (SV_CURPROC_FLAG(SV_ILP32))
2341 		ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2342 	else
2343 #endif
2344 		ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2345 }
2346 
2347 size_t
2348 ifr_buffer_get_length(void *data)
2349 {
2350 	union ifreq_union *ifrup;
2351 
2352 	ifrup = data;
2353 #ifdef COMPAT_FREEBSD32
2354 	if (SV_CURPROC_FLAG(SV_ILP32))
2355 		return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2356 #endif
2357 	return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2358 }
2359 
2360 static void
2361 ifr_buffer_set_length(void *data, size_t len)
2362 {
2363 	union ifreq_union *ifrup;
2364 
2365 	ifrup = data;
2366 #ifdef COMPAT_FREEBSD32
2367 	if (SV_CURPROC_FLAG(SV_ILP32))
2368 		ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2369 	else
2370 #endif
2371 		ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2372 }
2373 
2374 void *
2375 ifr_data_get_ptr(void *ifrp)
2376 {
2377 	union ifreq_union *ifrup;
2378 
2379 	ifrup = ifrp;
2380 #ifdef COMPAT_FREEBSD32
2381 	if (SV_CURPROC_FLAG(SV_ILP32))
2382 		return ((void *)(uintptr_t)
2383 		    ifrup->ifr32.ifr_ifru.ifru_data);
2384 #endif
2385 		return (ifrup->ifr.ifr_ifru.ifru_data);
2386 }
2387 
2388 struct ifcap_nv_bit_name {
2389 	uint64_t cap_bit;
2390 	const char *cap_name;
2391 };
2392 #define CAPNV(x) {.cap_bit = IFCAP_##x, \
2393     .cap_name = __CONCAT(IFCAP_, __CONCAT(x, _NAME)) }
2394 const struct ifcap_nv_bit_name ifcap_nv_bit_names[] = {
2395 	CAPNV(RXCSUM),
2396 	CAPNV(TXCSUM),
2397 	CAPNV(NETCONS),
2398 	CAPNV(VLAN_MTU),
2399 	CAPNV(VLAN_HWTAGGING),
2400 	CAPNV(JUMBO_MTU),
2401 	CAPNV(POLLING),
2402 	CAPNV(VLAN_HWCSUM),
2403 	CAPNV(TSO4),
2404 	CAPNV(TSO6),
2405 	CAPNV(LRO),
2406 	CAPNV(WOL_UCAST),
2407 	CAPNV(WOL_MCAST),
2408 	CAPNV(WOL_MAGIC),
2409 	CAPNV(TOE4),
2410 	CAPNV(TOE6),
2411 	CAPNV(VLAN_HWFILTER),
2412 	CAPNV(VLAN_HWTSO),
2413 	CAPNV(LINKSTATE),
2414 	CAPNV(NETMAP),
2415 	CAPNV(RXCSUM_IPV6),
2416 	CAPNV(TXCSUM_IPV6),
2417 	CAPNV(HWSTATS),
2418 	CAPNV(TXRTLMT),
2419 	CAPNV(HWRXTSTMP),
2420 	CAPNV(MEXTPG),
2421 	CAPNV(TXTLS4),
2422 	CAPNV(TXTLS6),
2423 	CAPNV(VXLAN_HWCSUM),
2424 	CAPNV(VXLAN_HWTSO),
2425 	CAPNV(TXTLS_RTLMT),
2426 	{0, NULL}
2427 };
2428 #define CAP2NV(x) {.cap_bit = IFCAP2_BIT(IFCAP2_##x), \
2429     .cap_name = __CONCAT(IFCAP2_, __CONCAT(x, _NAME)) }
2430 const struct ifcap_nv_bit_name ifcap2_nv_bit_names[] = {
2431 	CAP2NV(RXTLS4),
2432 	CAP2NV(RXTLS6),
2433 	{0, NULL}
2434 };
2435 #undef CAPNV
2436 #undef CAP2NV
2437 
2438 int
2439 if_capnv_to_capint(const nvlist_t *nv, int *old_cap,
2440     const struct ifcap_nv_bit_name *nn, bool all)
2441 {
2442 	int i, res;
2443 
2444 	res = 0;
2445 	for (i = 0; nn[i].cap_name != NULL; i++) {
2446 		if (nvlist_exists_bool(nv, nn[i].cap_name)) {
2447 			if (all || nvlist_get_bool(nv, nn[i].cap_name))
2448 				res |= nn[i].cap_bit;
2449 		} else {
2450 			res |= *old_cap & nn[i].cap_bit;
2451 		}
2452 	}
2453 	return (res);
2454 }
2455 
2456 void
2457 if_capint_to_capnv(nvlist_t *nv, const struct ifcap_nv_bit_name *nn,
2458     int ifr_cap, int ifr_req)
2459 {
2460 	int i;
2461 
2462 	for (i = 0; nn[i].cap_name != NULL; i++) {
2463 		if ((nn[i].cap_bit & ifr_cap) != 0) {
2464 			nvlist_add_bool(nv, nn[i].cap_name,
2465 			    (nn[i].cap_bit & ifr_req) != 0);
2466 		}
2467 	}
2468 }
2469 
2470 /*
2471  * Hardware specific interface ioctls.
2472  */
2473 int
2474 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2475 {
2476 	struct ifreq *ifr;
2477 	int error = 0, do_ifup = 0;
2478 	int new_flags, temp_flags;
2479 	size_t descrlen, nvbuflen;
2480 	char *descrbuf;
2481 	char new_name[IFNAMSIZ];
2482 	void *buf;
2483 	nvlist_t *nvcap;
2484 	struct siocsifcapnv_driver_data drv_ioctl_data;
2485 
2486 	ifr = (struct ifreq *)data;
2487 	switch (cmd) {
2488 	case SIOCGIFINDEX:
2489 		ifr->ifr_index = ifp->if_index;
2490 		break;
2491 
2492 	case SIOCGIFFLAGS:
2493 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2494 		ifr->ifr_flags = temp_flags & 0xffff;
2495 		ifr->ifr_flagshigh = temp_flags >> 16;
2496 		break;
2497 
2498 	case SIOCGIFCAP:
2499 		ifr->ifr_reqcap = ifp->if_capabilities;
2500 		ifr->ifr_curcap = ifp->if_capenable;
2501 		break;
2502 
2503 	case SIOCGIFCAPNV:
2504 		if ((ifp->if_capabilities & IFCAP_NV) == 0) {
2505 			error = EINVAL;
2506 			break;
2507 		}
2508 		buf = NULL;
2509 		nvcap = nvlist_create(0);
2510 		for (;;) {
2511 			if_capint_to_capnv(nvcap, ifcap_nv_bit_names,
2512 			    ifp->if_capabilities, ifp->if_capenable);
2513 			if_capint_to_capnv(nvcap, ifcap2_nv_bit_names,
2514 			    ifp->if_capabilities2, ifp->if_capenable2);
2515 			error = (*ifp->if_ioctl)(ifp, SIOCGIFCAPNV,
2516 			    __DECONST(caddr_t, nvcap));
2517 			if (error != 0) {
2518 				if_printf(ifp,
2519 			    "SIOCGIFCAPNV driver mistake: nvlist error %d\n",
2520 				    error);
2521 				break;
2522 			}
2523 			buf = nvlist_pack(nvcap, &nvbuflen);
2524 			if (buf == NULL) {
2525 				error = nvlist_error(nvcap);
2526 				if (error == 0)
2527 					error = EDOOFUS;
2528 				break;
2529 			}
2530 			if (nvbuflen > ifr->ifr_cap_nv.buf_length) {
2531 				ifr->ifr_cap_nv.length = nvbuflen;
2532 				ifr->ifr_cap_nv.buffer = NULL;
2533 				error = EFBIG;
2534 				break;
2535 			}
2536 			ifr->ifr_cap_nv.length = nvbuflen;
2537 			error = copyout(buf, ifr->ifr_cap_nv.buffer, nvbuflen);
2538 			break;
2539 		}
2540 		free(buf, M_NVLIST);
2541 		nvlist_destroy(nvcap);
2542 		break;
2543 
2544 	case SIOCGIFDATA:
2545 	{
2546 		struct if_data ifd;
2547 
2548 		/* Ensure uninitialised padding is not leaked. */
2549 		memset(&ifd, 0, sizeof(ifd));
2550 
2551 		if_data_copy(ifp, &ifd);
2552 		error = copyout(&ifd, ifr_data_get_ptr(ifr), sizeof(ifd));
2553 		break;
2554 	}
2555 
2556 #ifdef MAC
2557 	case SIOCGIFMAC:
2558 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2559 		break;
2560 #endif
2561 
2562 	case SIOCGIFMETRIC:
2563 		ifr->ifr_metric = ifp->if_metric;
2564 		break;
2565 
2566 	case SIOCGIFMTU:
2567 		ifr->ifr_mtu = ifp->if_mtu;
2568 		break;
2569 
2570 	case SIOCGIFPHYS:
2571 		/* XXXGL: did this ever worked? */
2572 		ifr->ifr_phys = 0;
2573 		break;
2574 
2575 	case SIOCGIFDESCR:
2576 		error = 0;
2577 		sx_slock(&ifdescr_sx);
2578 		if (ifp->if_description == NULL)
2579 			error = ENOMSG;
2580 		else {
2581 			/* space for terminating nul */
2582 			descrlen = strlen(ifp->if_description) + 1;
2583 			if (ifr_buffer_get_length(ifr) < descrlen)
2584 				ifr_buffer_set_buffer_null(ifr);
2585 			else
2586 				error = copyout(ifp->if_description,
2587 				    ifr_buffer_get_buffer(ifr), descrlen);
2588 			ifr_buffer_set_length(ifr, descrlen);
2589 		}
2590 		sx_sunlock(&ifdescr_sx);
2591 		break;
2592 
2593 	case SIOCSIFDESCR:
2594 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2595 		if (error)
2596 			return (error);
2597 
2598 		/*
2599 		 * Copy only (length-1) bytes to make sure that
2600 		 * if_description is always nul terminated.  The
2601 		 * length parameter is supposed to count the
2602 		 * terminating nul in.
2603 		 */
2604 		if (ifr_buffer_get_length(ifr) > ifdescr_maxlen)
2605 			return (ENAMETOOLONG);
2606 		else if (ifr_buffer_get_length(ifr) == 0)
2607 			descrbuf = NULL;
2608 		else {
2609 			descrbuf = if_allocdescr(ifr_buffer_get_length(ifr), M_WAITOK);
2610 			error = copyin(ifr_buffer_get_buffer(ifr), descrbuf,
2611 			    ifr_buffer_get_length(ifr) - 1);
2612 			if (error) {
2613 				if_freedescr(descrbuf);
2614 				break;
2615 			}
2616 		}
2617 
2618 		if_setdescr(ifp, descrbuf);
2619 		getmicrotime(&ifp->if_lastchange);
2620 		break;
2621 
2622 	case SIOCGIFFIB:
2623 		ifr->ifr_fib = ifp->if_fib;
2624 		break;
2625 
2626 	case SIOCSIFFIB:
2627 		error = priv_check(td, PRIV_NET_SETIFFIB);
2628 		if (error)
2629 			return (error);
2630 		if (ifr->ifr_fib >= rt_numfibs)
2631 			return (EINVAL);
2632 
2633 		ifp->if_fib = ifr->ifr_fib;
2634 		break;
2635 
2636 	case SIOCSIFFLAGS:
2637 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2638 		if (error)
2639 			return (error);
2640 		/*
2641 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2642 		 * check, so we don't need special handling here yet.
2643 		 */
2644 		new_flags = (ifr->ifr_flags & 0xffff) |
2645 		    (ifr->ifr_flagshigh << 16);
2646 		if (ifp->if_flags & IFF_UP &&
2647 		    (new_flags & IFF_UP) == 0) {
2648 			if_down(ifp);
2649 		} else if (new_flags & IFF_UP &&
2650 		    (ifp->if_flags & IFF_UP) == 0) {
2651 			do_ifup = 1;
2652 		}
2653 		/* See if permanently promiscuous mode bit is about to flip */
2654 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2655 			if (new_flags & IFF_PPROMISC)
2656 				ifp->if_flags |= IFF_PROMISC;
2657 			else if (ifp->if_pcount == 0)
2658 				ifp->if_flags &= ~IFF_PROMISC;
2659 			if (log_promisc_mode_change)
2660                                 if_printf(ifp, "permanently promiscuous mode %s\n",
2661                                     ((new_flags & IFF_PPROMISC) ?
2662                                      "enabled" : "disabled"));
2663 		}
2664 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2665 			(new_flags &~ IFF_CANTCHANGE);
2666 		if (ifp->if_ioctl) {
2667 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2668 		}
2669 		if (do_ifup)
2670 			if_up(ifp);
2671 		getmicrotime(&ifp->if_lastchange);
2672 		break;
2673 
2674 	case SIOCSIFCAP:
2675 		error = priv_check(td, PRIV_NET_SETIFCAP);
2676 		if (error != 0)
2677 			return (error);
2678 		if (ifp->if_ioctl == NULL)
2679 			return (EOPNOTSUPP);
2680 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2681 			return (EINVAL);
2682 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2683 		if (error == 0)
2684 			getmicrotime(&ifp->if_lastchange);
2685 		break;
2686 
2687 	case SIOCSIFCAPNV:
2688 		error = priv_check(td, PRIV_NET_SETIFCAP);
2689 		if (error != 0)
2690 			return (error);
2691 		if (ifp->if_ioctl == NULL)
2692 			return (EOPNOTSUPP);
2693 		if ((ifp->if_capabilities & IFCAP_NV) == 0)
2694 			return (EINVAL);
2695 		if (ifr->ifr_cap_nv.length > IFR_CAP_NV_MAXBUFSIZE)
2696 			return (EINVAL);
2697 		nvcap = NULL;
2698 		buf = malloc(ifr->ifr_cap_nv.length, M_TEMP, M_WAITOK);
2699 		for (;;) {
2700 			error = copyin(ifr->ifr_cap_nv.buffer, buf,
2701 			    ifr->ifr_cap_nv.length);
2702 			if (error != 0)
2703 				break;
2704 			nvcap = nvlist_unpack(buf, ifr->ifr_cap_nv.length, 0);
2705 			if (nvcap == NULL) {
2706 				error = EINVAL;
2707 				break;
2708 			}
2709 			drv_ioctl_data.reqcap = if_capnv_to_capint(nvcap,
2710 			    &ifp->if_capenable, ifcap_nv_bit_names, false);
2711 			if ((drv_ioctl_data.reqcap &
2712 			    ~ifp->if_capabilities) != 0) {
2713 				error = EINVAL;
2714 				break;
2715 			}
2716 			drv_ioctl_data.reqcap2 = if_capnv_to_capint(nvcap,
2717 			    &ifp->if_capenable2, ifcap2_nv_bit_names, false);
2718 			if ((drv_ioctl_data.reqcap2 &
2719 			    ~ifp->if_capabilities2) != 0) {
2720 				error = EINVAL;
2721 				break;
2722 			}
2723 			drv_ioctl_data.nvcap = nvcap;
2724 			error = (*ifp->if_ioctl)(ifp, SIOCSIFCAPNV,
2725 			    (caddr_t)&drv_ioctl_data);
2726 			break;
2727 		}
2728 		nvlist_destroy(nvcap);
2729 		free(buf, M_TEMP);
2730 		if (error == 0)
2731 			getmicrotime(&ifp->if_lastchange);
2732 		break;
2733 
2734 #ifdef MAC
2735 	case SIOCSIFMAC:
2736 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2737 		break;
2738 #endif
2739 
2740 	case SIOCSIFNAME:
2741 		error = priv_check(td, PRIV_NET_SETIFNAME);
2742 		if (error)
2743 			return (error);
2744 		error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ,
2745 		    NULL);
2746 		if (error != 0)
2747 			return (error);
2748 		error = if_rename(ifp, new_name);
2749 		break;
2750 
2751 #ifdef VIMAGE
2752 	case SIOCSIFVNET:
2753 		error = priv_check(td, PRIV_NET_SETIFVNET);
2754 		if (error)
2755 			return (error);
2756 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2757 		break;
2758 #endif
2759 
2760 	case SIOCSIFMETRIC:
2761 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2762 		if (error)
2763 			return (error);
2764 		ifp->if_metric = ifr->ifr_metric;
2765 		getmicrotime(&ifp->if_lastchange);
2766 		break;
2767 
2768 	case SIOCSIFPHYS:
2769 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2770 		if (error)
2771 			return (error);
2772 		if (ifp->if_ioctl == NULL)
2773 			return (EOPNOTSUPP);
2774 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2775 		if (error == 0)
2776 			getmicrotime(&ifp->if_lastchange);
2777 		break;
2778 
2779 	case SIOCSIFMTU:
2780 	{
2781 		u_long oldmtu = ifp->if_mtu;
2782 
2783 		error = priv_check(td, PRIV_NET_SETIFMTU);
2784 		if (error)
2785 			return (error);
2786 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2787 			return (EINVAL);
2788 		if (ifp->if_ioctl == NULL)
2789 			return (EOPNOTSUPP);
2790 		/* Disallow MTU changes on bridge member interfaces. */
2791 		if (ifp->if_bridge)
2792 			return (EOPNOTSUPP);
2793 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2794 		if (error == 0) {
2795 			getmicrotime(&ifp->if_lastchange);
2796 			rt_ifmsg(ifp, 0);
2797 #ifdef INET
2798 			DEBUGNET_NOTIFY_MTU(ifp);
2799 #endif
2800 		}
2801 		/*
2802 		 * If the link MTU changed, do network layer specific procedure.
2803 		 */
2804 		if (ifp->if_mtu != oldmtu)
2805 			if_notifymtu(ifp);
2806 		break;
2807 	}
2808 
2809 	case SIOCADDMULTI:
2810 	case SIOCDELMULTI:
2811 		if (cmd == SIOCADDMULTI)
2812 			error = priv_check(td, PRIV_NET_ADDMULTI);
2813 		else
2814 			error = priv_check(td, PRIV_NET_DELMULTI);
2815 		if (error)
2816 			return (error);
2817 
2818 		/* Don't allow group membership on non-multicast interfaces. */
2819 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2820 			return (EOPNOTSUPP);
2821 
2822 		/* Don't let users screw up protocols' entries. */
2823 		if (ifr->ifr_addr.sa_family != AF_LINK)
2824 			return (EINVAL);
2825 
2826 		if (cmd == SIOCADDMULTI) {
2827 			struct epoch_tracker et;
2828 			struct ifmultiaddr *ifma;
2829 
2830 			/*
2831 			 * Userland is only permitted to join groups once
2832 			 * via the if_addmulti() KPI, because it cannot hold
2833 			 * struct ifmultiaddr * between calls. It may also
2834 			 * lose a race while we check if the membership
2835 			 * already exists.
2836 			 */
2837 			NET_EPOCH_ENTER(et);
2838 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2839 			NET_EPOCH_EXIT(et);
2840 			if (ifma != NULL)
2841 				error = EADDRINUSE;
2842 			else
2843 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2844 		} else {
2845 			error = if_delmulti(ifp, &ifr->ifr_addr);
2846 		}
2847 		if (error == 0)
2848 			getmicrotime(&ifp->if_lastchange);
2849 		break;
2850 
2851 	case SIOCSIFPHYADDR:
2852 	case SIOCDIFPHYADDR:
2853 #ifdef INET6
2854 	case SIOCSIFPHYADDR_IN6:
2855 #endif
2856 	case SIOCSIFMEDIA:
2857 	case SIOCSIFGENERIC:
2858 		error = priv_check(td, PRIV_NET_HWIOCTL);
2859 		if (error)
2860 			return (error);
2861 		if (ifp->if_ioctl == NULL)
2862 			return (EOPNOTSUPP);
2863 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2864 		if (error == 0)
2865 			getmicrotime(&ifp->if_lastchange);
2866 		break;
2867 
2868 	case SIOCGIFSTATUS:
2869 	case SIOCGIFPSRCADDR:
2870 	case SIOCGIFPDSTADDR:
2871 	case SIOCGIFMEDIA:
2872 	case SIOCGIFXMEDIA:
2873 	case SIOCGIFGENERIC:
2874 	case SIOCGIFRSSKEY:
2875 	case SIOCGIFRSSHASH:
2876 	case SIOCGIFDOWNREASON:
2877 		if (ifp->if_ioctl == NULL)
2878 			return (EOPNOTSUPP);
2879 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2880 		break;
2881 
2882 	case SIOCSIFLLADDR:
2883 		error = priv_check(td, PRIV_NET_SETLLADDR);
2884 		if (error)
2885 			return (error);
2886 		error = if_setlladdr(ifp,
2887 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2888 		break;
2889 
2890 	case SIOCGHWADDR:
2891 		error = if_gethwaddr(ifp, ifr);
2892 		break;
2893 
2894 	case SIOCAIFGROUP:
2895 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2896 		if (error)
2897 			return (error);
2898 		error = if_addgroup(ifp,
2899 		    ((struct ifgroupreq *)data)->ifgr_group);
2900 		if (error != 0)
2901 			return (error);
2902 		break;
2903 
2904 	case SIOCGIFGROUP:
2905 	{
2906 		struct epoch_tracker et;
2907 
2908 		NET_EPOCH_ENTER(et);
2909 		error = if_getgroup((struct ifgroupreq *)data, ifp);
2910 		NET_EPOCH_EXIT(et);
2911 		break;
2912 	}
2913 
2914 	case SIOCDIFGROUP:
2915 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2916 		if (error)
2917 			return (error);
2918 		error = if_delgroup(ifp,
2919 		    ((struct ifgroupreq *)data)->ifgr_group);
2920 		if (error != 0)
2921 			return (error);
2922 		break;
2923 
2924 	default:
2925 		error = ENOIOCTL;
2926 		break;
2927 	}
2928 	return (error);
2929 }
2930 
2931 /*
2932  * Interface ioctls.
2933  */
2934 int
2935 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2936 {
2937 #ifdef COMPAT_FREEBSD32
2938 	union {
2939 		struct ifconf ifc;
2940 		struct ifdrv ifd;
2941 		struct ifgroupreq ifgr;
2942 		struct ifmediareq ifmr;
2943 	} thunk;
2944 	u_long saved_cmd;
2945 	struct ifconf32 *ifc32;
2946 	struct ifdrv32 *ifd32;
2947 	struct ifgroupreq32 *ifgr32;
2948 	struct ifmediareq32 *ifmr32;
2949 #endif
2950 	struct ifnet *ifp;
2951 	struct ifreq *ifr;
2952 	int error;
2953 	int oif_flags;
2954 #ifdef VIMAGE
2955 	bool shutdown;
2956 #endif
2957 
2958 	CURVNET_SET(so->so_vnet);
2959 #ifdef VIMAGE
2960 	/* Make sure the VNET is stable. */
2961 	shutdown = VNET_IS_SHUTTING_DOWN(so->so_vnet);
2962 	if (shutdown) {
2963 		CURVNET_RESTORE();
2964 		return (EBUSY);
2965 	}
2966 #endif
2967 
2968 #ifdef COMPAT_FREEBSD32
2969 	saved_cmd = cmd;
2970 	switch (cmd) {
2971 	case SIOCGIFCONF32:
2972 		ifc32 = (struct ifconf32 *)data;
2973 		thunk.ifc.ifc_len = ifc32->ifc_len;
2974 		thunk.ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2975 		data = (caddr_t)&thunk.ifc;
2976 		cmd = SIOCGIFCONF;
2977 		break;
2978 	case SIOCGDRVSPEC32:
2979 	case SIOCSDRVSPEC32:
2980 		ifd32 = (struct ifdrv32 *)data;
2981 		memcpy(thunk.ifd.ifd_name, ifd32->ifd_name,
2982 		    sizeof(thunk.ifd.ifd_name));
2983 		thunk.ifd.ifd_cmd = ifd32->ifd_cmd;
2984 		thunk.ifd.ifd_len = ifd32->ifd_len;
2985 		thunk.ifd.ifd_data = PTRIN(ifd32->ifd_data);
2986 		data = (caddr_t)&thunk.ifd;
2987 		cmd = _IOC_NEWTYPE(cmd, struct ifdrv);
2988 		break;
2989 	case SIOCAIFGROUP32:
2990 	case SIOCGIFGROUP32:
2991 	case SIOCDIFGROUP32:
2992 	case SIOCGIFGMEMB32:
2993 		ifgr32 = (struct ifgroupreq32 *)data;
2994 		memcpy(thunk.ifgr.ifgr_name, ifgr32->ifgr_name,
2995 		    sizeof(thunk.ifgr.ifgr_name));
2996 		thunk.ifgr.ifgr_len = ifgr32->ifgr_len;
2997 		switch (cmd) {
2998 		case SIOCAIFGROUP32:
2999 		case SIOCDIFGROUP32:
3000 			memcpy(thunk.ifgr.ifgr_group, ifgr32->ifgr_group,
3001 			    sizeof(thunk.ifgr.ifgr_group));
3002 			break;
3003 		case SIOCGIFGROUP32:
3004 		case SIOCGIFGMEMB32:
3005 			thunk.ifgr.ifgr_groups = PTRIN(ifgr32->ifgr_groups);
3006 			break;
3007 		}
3008 		data = (caddr_t)&thunk.ifgr;
3009 		cmd = _IOC_NEWTYPE(cmd, struct ifgroupreq);
3010 		break;
3011 	case SIOCGIFMEDIA32:
3012 	case SIOCGIFXMEDIA32:
3013 		ifmr32 = (struct ifmediareq32 *)data;
3014 		memcpy(thunk.ifmr.ifm_name, ifmr32->ifm_name,
3015 		    sizeof(thunk.ifmr.ifm_name));
3016 		thunk.ifmr.ifm_current = ifmr32->ifm_current;
3017 		thunk.ifmr.ifm_mask = ifmr32->ifm_mask;
3018 		thunk.ifmr.ifm_status = ifmr32->ifm_status;
3019 		thunk.ifmr.ifm_active = ifmr32->ifm_active;
3020 		thunk.ifmr.ifm_count = ifmr32->ifm_count;
3021 		thunk.ifmr.ifm_ulist = PTRIN(ifmr32->ifm_ulist);
3022 		data = (caddr_t)&thunk.ifmr;
3023 		cmd = _IOC_NEWTYPE(cmd, struct ifmediareq);
3024 		break;
3025 	}
3026 #endif
3027 
3028 	switch (cmd) {
3029 	case SIOCGIFCONF:
3030 		error = ifconf(cmd, data);
3031 		goto out_noref;
3032 	}
3033 
3034 	ifr = (struct ifreq *)data;
3035 	switch (cmd) {
3036 #ifdef VIMAGE
3037 	case SIOCSIFRVNET:
3038 		error = priv_check(td, PRIV_NET_SETIFVNET);
3039 		if (error == 0)
3040 			error = if_vmove_reclaim(td, ifr->ifr_name,
3041 			    ifr->ifr_jid);
3042 		goto out_noref;
3043 #endif
3044 	case SIOCIFCREATE:
3045 	case SIOCIFCREATE2:
3046 		error = priv_check(td, PRIV_NET_IFCREATE);
3047 		if (error == 0)
3048 			error = if_clone_create(ifr->ifr_name,
3049 			    sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ?
3050 			    ifr_data_get_ptr(ifr) : NULL);
3051 		goto out_noref;
3052 	case SIOCIFDESTROY:
3053 		error = priv_check(td, PRIV_NET_IFDESTROY);
3054 
3055 		if (error == 0) {
3056 			sx_xlock(&ifnet_detach_sxlock);
3057 			error = if_clone_destroy(ifr->ifr_name);
3058 			sx_xunlock(&ifnet_detach_sxlock);
3059 		}
3060 		goto out_noref;
3061 
3062 	case SIOCIFGCLONERS:
3063 		error = if_clone_list((struct if_clonereq *)data);
3064 		goto out_noref;
3065 
3066 	case SIOCGIFGMEMB:
3067 		error = if_getgroupmembers((struct ifgroupreq *)data);
3068 		goto out_noref;
3069 
3070 #if defined(INET) || defined(INET6)
3071 	case SIOCSVH:
3072 	case SIOCGVH:
3073 		if (carp_ioctl_p == NULL)
3074 			error = EPROTONOSUPPORT;
3075 		else
3076 			error = (*carp_ioctl_p)(ifr, cmd, td);
3077 		goto out_noref;
3078 #endif
3079 	}
3080 
3081 	ifp = ifunit_ref(ifr->ifr_name);
3082 	if (ifp == NULL) {
3083 		error = ENXIO;
3084 		goto out_noref;
3085 	}
3086 
3087 	error = ifhwioctl(cmd, ifp, data, td);
3088 	if (error != ENOIOCTL)
3089 		goto out_ref;
3090 
3091 	oif_flags = ifp->if_flags;
3092 	if (so->so_proto == NULL) {
3093 		error = EOPNOTSUPP;
3094 		goto out_ref;
3095 	}
3096 
3097 	/*
3098 	 * Pass the request on to the socket control method, and if the
3099 	 * latter returns EOPNOTSUPP, directly to the interface.
3100 	 *
3101 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
3102 	 * trust SIOCSIFADDR et al to come from an already privileged
3103 	 * layer, and do not perform any credentials checks or input
3104 	 * validation.
3105 	 */
3106 	error = so->so_proto->pr_control(so, cmd, data, ifp, td);
3107 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
3108 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
3109 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
3110 		error = (*ifp->if_ioctl)(ifp, cmd, data);
3111 
3112 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
3113 #ifdef INET6
3114 		if (ifp->if_flags & IFF_UP)
3115 			in6_if_up(ifp);
3116 #endif
3117 	}
3118 
3119 out_ref:
3120 	if_rele(ifp);
3121 out_noref:
3122 	CURVNET_RESTORE();
3123 #ifdef COMPAT_FREEBSD32
3124 	if (error != 0)
3125 		return (error);
3126 	switch (saved_cmd) {
3127 	case SIOCGIFCONF32:
3128 		ifc32->ifc_len = thunk.ifc.ifc_len;
3129 		break;
3130 	case SIOCGDRVSPEC32:
3131 		/*
3132 		 * SIOCGDRVSPEC is IOWR, but nothing actually touches
3133 		 * the struct so just assert that ifd_len (the only
3134 		 * field it might make sense to update) hasn't
3135 		 * changed.
3136 		 */
3137 		KASSERT(thunk.ifd.ifd_len == ifd32->ifd_len,
3138 		    ("ifd_len was updated %u -> %zu", ifd32->ifd_len,
3139 			thunk.ifd.ifd_len));
3140 		break;
3141 	case SIOCGIFGROUP32:
3142 	case SIOCGIFGMEMB32:
3143 		ifgr32->ifgr_len = thunk.ifgr.ifgr_len;
3144 		break;
3145 	case SIOCGIFMEDIA32:
3146 	case SIOCGIFXMEDIA32:
3147 		ifmr32->ifm_current = thunk.ifmr.ifm_current;
3148 		ifmr32->ifm_mask = thunk.ifmr.ifm_mask;
3149 		ifmr32->ifm_status = thunk.ifmr.ifm_status;
3150 		ifmr32->ifm_active = thunk.ifmr.ifm_active;
3151 		ifmr32->ifm_count = thunk.ifmr.ifm_count;
3152 		break;
3153 	}
3154 #endif
3155 	return (error);
3156 }
3157 
3158 int
3159 if_rename(struct ifnet *ifp, char *new_name)
3160 {
3161 	struct ifaddr *ifa;
3162 	struct sockaddr_dl *sdl;
3163 	size_t namelen, onamelen;
3164 	char old_name[IFNAMSIZ];
3165 	char strbuf[IFNAMSIZ + 8];
3166 
3167 	if (new_name[0] == '\0')
3168 		return (EINVAL);
3169 	if (strcmp(new_name, ifp->if_xname) == 0)
3170 		return (0);
3171 	if (ifunit(new_name) != NULL)
3172 		return (EEXIST);
3173 
3174 	/*
3175 	 * XXX: Locking.  Nothing else seems to lock if_flags,
3176 	 * and there are numerous other races with the
3177 	 * ifunit() checks not being atomic with namespace
3178 	 * changes (renames, vmoves, if_attach, etc).
3179 	 */
3180 	ifp->if_flags |= IFF_RENAMING;
3181 
3182 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
3183 
3184 	if_printf(ifp, "changing name to '%s'\n", new_name);
3185 
3186 	IF_ADDR_WLOCK(ifp);
3187 	strlcpy(old_name, ifp->if_xname, sizeof(old_name));
3188 	strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
3189 	ifa = ifp->if_addr;
3190 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3191 	namelen = strlen(new_name);
3192 	onamelen = sdl->sdl_nlen;
3193 	/*
3194 	 * Move the address if needed.  This is safe because we
3195 	 * allocate space for a name of length IFNAMSIZ when we
3196 	 * create this in if_attach().
3197 	 */
3198 	if (namelen != onamelen) {
3199 		bcopy(sdl->sdl_data + onamelen,
3200 		    sdl->sdl_data + namelen, sdl->sdl_alen);
3201 	}
3202 	bcopy(new_name, sdl->sdl_data, namelen);
3203 	sdl->sdl_nlen = namelen;
3204 	sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
3205 	bzero(sdl->sdl_data, onamelen);
3206 	while (namelen != 0)
3207 		sdl->sdl_data[--namelen] = 0xff;
3208 	IF_ADDR_WUNLOCK(ifp);
3209 
3210 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
3211 
3212 	ifp->if_flags &= ~IFF_RENAMING;
3213 
3214 	snprintf(strbuf, sizeof(strbuf), "name=%s", new_name);
3215 	devctl_notify("IFNET", old_name, "RENAME", strbuf);
3216 
3217 	return (0);
3218 }
3219 
3220 /*
3221  * The code common to handling reference counted flags,
3222  * e.g., in ifpromisc() and if_allmulti().
3223  * The "pflag" argument can specify a permanent mode flag to check,
3224  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
3225  *
3226  * Only to be used on stack-owned flags, not driver-owned flags.
3227  */
3228 static int
3229 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
3230 {
3231 	struct ifreq ifr;
3232 	int error;
3233 	int oldflags, oldcount;
3234 
3235 	/* Sanity checks to catch programming errors */
3236 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
3237 	    ("%s: setting driver-owned flag %d", __func__, flag));
3238 
3239 	if (onswitch)
3240 		KASSERT(*refcount >= 0,
3241 		    ("%s: increment negative refcount %d for flag %d",
3242 		    __func__, *refcount, flag));
3243 	else
3244 		KASSERT(*refcount > 0,
3245 		    ("%s: decrement non-positive refcount %d for flag %d",
3246 		    __func__, *refcount, flag));
3247 
3248 	/* In case this mode is permanent, just touch refcount */
3249 	if (ifp->if_flags & pflag) {
3250 		*refcount += onswitch ? 1 : -1;
3251 		return (0);
3252 	}
3253 
3254 	/* Save ifnet parameters for if_ioctl() may fail */
3255 	oldcount = *refcount;
3256 	oldflags = ifp->if_flags;
3257 
3258 	/*
3259 	 * See if we aren't the only and touching refcount is enough.
3260 	 * Actually toggle interface flag if we are the first or last.
3261 	 */
3262 	if (onswitch) {
3263 		if ((*refcount)++)
3264 			return (0);
3265 		ifp->if_flags |= flag;
3266 	} else {
3267 		if (--(*refcount))
3268 			return (0);
3269 		ifp->if_flags &= ~flag;
3270 	}
3271 
3272 	/* Call down the driver since we've changed interface flags */
3273 	if (ifp->if_ioctl == NULL) {
3274 		error = EOPNOTSUPP;
3275 		goto recover;
3276 	}
3277 	ifr.ifr_flags = ifp->if_flags & 0xffff;
3278 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
3279 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3280 	if (error)
3281 		goto recover;
3282 	/* Notify userland that interface flags have changed */
3283 	rt_ifmsg(ifp, flag);
3284 	return (0);
3285 
3286 recover:
3287 	/* Recover after driver error */
3288 	*refcount = oldcount;
3289 	ifp->if_flags = oldflags;
3290 	return (error);
3291 }
3292 
3293 /*
3294  * Set/clear promiscuous mode on interface ifp based on the truth value
3295  * of pswitch.  The calls are reference counted so that only the first
3296  * "on" request actually has an effect, as does the final "off" request.
3297  * Results are undefined if the "off" and "on" requests are not matched.
3298  */
3299 int
3300 ifpromisc(struct ifnet *ifp, int pswitch)
3301 {
3302 	int error;
3303 	int oldflags = ifp->if_flags;
3304 
3305 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3306 			   &ifp->if_pcount, pswitch);
3307 	/* If promiscuous mode status has changed, log a message */
3308 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3309             log_promisc_mode_change)
3310 		if_printf(ifp, "promiscuous mode %s\n",
3311 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3312 	return (error);
3313 }
3314 
3315 /*
3316  * Return interface configuration
3317  * of system.  List may be used
3318  * in later ioctl's (above) to get
3319  * other information.
3320  */
3321 /*ARGSUSED*/
3322 static int
3323 ifconf(u_long cmd, caddr_t data)
3324 {
3325 	struct ifconf *ifc = (struct ifconf *)data;
3326 	struct ifnet *ifp;
3327 	struct ifaddr *ifa;
3328 	struct ifreq ifr;
3329 	struct sbuf *sb;
3330 	int error, full = 0, valid_len, max_len;
3331 
3332 	/* Limit initial buffer size to maxphys to avoid DoS from userspace. */
3333 	max_len = maxphys - 1;
3334 
3335 	/* Prevent hostile input from being able to crash the system */
3336 	if (ifc->ifc_len <= 0)
3337 		return (EINVAL);
3338 
3339 again:
3340 	if (ifc->ifc_len <= max_len) {
3341 		max_len = ifc->ifc_len;
3342 		full = 1;
3343 	}
3344 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3345 	max_len = 0;
3346 	valid_len = 0;
3347 
3348 	IFNET_RLOCK();
3349 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3350 		struct epoch_tracker et;
3351 		int addrs;
3352 
3353 		/*
3354 		 * Zero the ifr to make sure we don't disclose the contents
3355 		 * of the stack.
3356 		 */
3357 		memset(&ifr, 0, sizeof(ifr));
3358 
3359 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3360 		    >= sizeof(ifr.ifr_name)) {
3361 			sbuf_delete(sb);
3362 			IFNET_RUNLOCK();
3363 			return (ENAMETOOLONG);
3364 		}
3365 
3366 		addrs = 0;
3367 		NET_EPOCH_ENTER(et);
3368 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3369 			struct sockaddr *sa = ifa->ifa_addr;
3370 
3371 			if (prison_if(curthread->td_ucred, sa) != 0)
3372 				continue;
3373 			addrs++;
3374 			if (sa->sa_len <= sizeof(*sa)) {
3375 				if (sa->sa_len < sizeof(*sa)) {
3376 					memset(&ifr.ifr_ifru.ifru_addr, 0,
3377 					    sizeof(ifr.ifr_ifru.ifru_addr));
3378 					memcpy(&ifr.ifr_ifru.ifru_addr, sa,
3379 					    sa->sa_len);
3380 				} else
3381 					ifr.ifr_ifru.ifru_addr = *sa;
3382 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3383 				max_len += sizeof(ifr);
3384 			} else {
3385 				sbuf_bcat(sb, &ifr,
3386 				    offsetof(struct ifreq, ifr_addr));
3387 				max_len += offsetof(struct ifreq, ifr_addr);
3388 				sbuf_bcat(sb, sa, sa->sa_len);
3389 				max_len += sa->sa_len;
3390 			}
3391 
3392 			if (sbuf_error(sb) == 0)
3393 				valid_len = sbuf_len(sb);
3394 		}
3395 		NET_EPOCH_EXIT(et);
3396 		if (addrs == 0) {
3397 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3398 			max_len += sizeof(ifr);
3399 
3400 			if (sbuf_error(sb) == 0)
3401 				valid_len = sbuf_len(sb);
3402 		}
3403 	}
3404 	IFNET_RUNLOCK();
3405 
3406 	/*
3407 	 * If we didn't allocate enough space (uncommon), try again.  If
3408 	 * we have already allocated as much space as we are allowed,
3409 	 * return what we've got.
3410 	 */
3411 	if (valid_len != max_len && !full) {
3412 		sbuf_delete(sb);
3413 		goto again;
3414 	}
3415 
3416 	ifc->ifc_len = valid_len;
3417 	sbuf_finish(sb);
3418 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3419 	sbuf_delete(sb);
3420 	return (error);
3421 }
3422 
3423 /*
3424  * Just like ifpromisc(), but for all-multicast-reception mode.
3425  */
3426 int
3427 if_allmulti(struct ifnet *ifp, int onswitch)
3428 {
3429 
3430 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3431 }
3432 
3433 struct ifmultiaddr *
3434 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3435 {
3436 	struct ifmultiaddr *ifma;
3437 
3438 	IF_ADDR_LOCK_ASSERT(ifp);
3439 
3440 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3441 		if (sa->sa_family == AF_LINK) {
3442 			if (sa_dl_equal(ifma->ifma_addr, sa))
3443 				break;
3444 		} else {
3445 			if (sa_equal(ifma->ifma_addr, sa))
3446 				break;
3447 		}
3448 	}
3449 
3450 	return ifma;
3451 }
3452 
3453 /*
3454  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3455  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3456  * the ifnet multicast address list here, so the caller must do that and
3457  * other setup work (such as notifying the device driver).  The reference
3458  * count is initialized to 1.
3459  */
3460 static struct ifmultiaddr *
3461 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3462     int mflags)
3463 {
3464 	struct ifmultiaddr *ifma;
3465 	struct sockaddr *dupsa;
3466 
3467 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3468 	    M_ZERO);
3469 	if (ifma == NULL)
3470 		return (NULL);
3471 
3472 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3473 	if (dupsa == NULL) {
3474 		free(ifma, M_IFMADDR);
3475 		return (NULL);
3476 	}
3477 	bcopy(sa, dupsa, sa->sa_len);
3478 	ifma->ifma_addr = dupsa;
3479 
3480 	ifma->ifma_ifp = ifp;
3481 	ifma->ifma_refcount = 1;
3482 	ifma->ifma_protospec = NULL;
3483 
3484 	if (llsa == NULL) {
3485 		ifma->ifma_lladdr = NULL;
3486 		return (ifma);
3487 	}
3488 
3489 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3490 	if (dupsa == NULL) {
3491 		free(ifma->ifma_addr, M_IFMADDR);
3492 		free(ifma, M_IFMADDR);
3493 		return (NULL);
3494 	}
3495 	bcopy(llsa, dupsa, llsa->sa_len);
3496 	ifma->ifma_lladdr = dupsa;
3497 
3498 	return (ifma);
3499 }
3500 
3501 /*
3502  * if_freemulti: free ifmultiaddr structure and possibly attached related
3503  * addresses.  The caller is responsible for implementing reference
3504  * counting, notifying the driver, handling routing messages, and releasing
3505  * any dependent link layer state.
3506  */
3507 #ifdef MCAST_VERBOSE
3508 extern void kdb_backtrace(void);
3509 #endif
3510 static void
3511 if_freemulti_internal(struct ifmultiaddr *ifma)
3512 {
3513 
3514 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3515 	    ifma->ifma_refcount));
3516 
3517 	if (ifma->ifma_lladdr != NULL)
3518 		free(ifma->ifma_lladdr, M_IFMADDR);
3519 #ifdef MCAST_VERBOSE
3520 	kdb_backtrace();
3521 	printf("%s freeing ifma: %p\n", __func__, ifma);
3522 #endif
3523 	free(ifma->ifma_addr, M_IFMADDR);
3524 	free(ifma, M_IFMADDR);
3525 }
3526 
3527 static void
3528 if_destroymulti(epoch_context_t ctx)
3529 {
3530 	struct ifmultiaddr *ifma;
3531 
3532 	ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx);
3533 	if_freemulti_internal(ifma);
3534 }
3535 
3536 void
3537 if_freemulti(struct ifmultiaddr *ifma)
3538 {
3539 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d",
3540 	    ifma->ifma_refcount));
3541 
3542 	NET_EPOCH_CALL(if_destroymulti, &ifma->ifma_epoch_ctx);
3543 }
3544 
3545 /*
3546  * Register an additional multicast address with a network interface.
3547  *
3548  * - If the address is already present, bump the reference count on the
3549  *   address and return.
3550  * - If the address is not link-layer, look up a link layer address.
3551  * - Allocate address structures for one or both addresses, and attach to the
3552  *   multicast address list on the interface.  If automatically adding a link
3553  *   layer address, the protocol address will own a reference to the link
3554  *   layer address, to be freed when it is freed.
3555  * - Notify the network device driver of an addition to the multicast address
3556  *   list.
3557  *
3558  * 'sa' points to caller-owned memory with the desired multicast address.
3559  *
3560  * 'retifma' will be used to return a pointer to the resulting multicast
3561  * address reference, if desired.
3562  */
3563 int
3564 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3565     struct ifmultiaddr **retifma)
3566 {
3567 	struct ifmultiaddr *ifma, *ll_ifma;
3568 	struct sockaddr *llsa;
3569 	struct sockaddr_dl sdl;
3570 	int error;
3571 
3572 #ifdef INET
3573 	IN_MULTI_LIST_UNLOCK_ASSERT();
3574 #endif
3575 #ifdef INET6
3576 	IN6_MULTI_LIST_UNLOCK_ASSERT();
3577 #endif
3578 	/*
3579 	 * If the address is already present, return a new reference to it;
3580 	 * otherwise, allocate storage and set up a new address.
3581 	 */
3582 	IF_ADDR_WLOCK(ifp);
3583 	ifma = if_findmulti(ifp, sa);
3584 	if (ifma != NULL) {
3585 		ifma->ifma_refcount++;
3586 		if (retifma != NULL)
3587 			*retifma = ifma;
3588 		IF_ADDR_WUNLOCK(ifp);
3589 		return (0);
3590 	}
3591 
3592 	/*
3593 	 * The address isn't already present; resolve the protocol address
3594 	 * into a link layer address, and then look that up, bump its
3595 	 * refcount or allocate an ifma for that also.
3596 	 * Most link layer resolving functions returns address data which
3597 	 * fits inside default sockaddr_dl structure. However callback
3598 	 * can allocate another sockaddr structure, in that case we need to
3599 	 * free it later.
3600 	 */
3601 	llsa = NULL;
3602 	ll_ifma = NULL;
3603 	if (ifp->if_resolvemulti != NULL) {
3604 		/* Provide called function with buffer size information */
3605 		sdl.sdl_len = sizeof(sdl);
3606 		llsa = (struct sockaddr *)&sdl;
3607 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3608 		if (error)
3609 			goto unlock_out;
3610 	}
3611 
3612 	/*
3613 	 * Allocate the new address.  Don't hook it up yet, as we may also
3614 	 * need to allocate a link layer multicast address.
3615 	 */
3616 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3617 	if (ifma == NULL) {
3618 		error = ENOMEM;
3619 		goto free_llsa_out;
3620 	}
3621 
3622 	/*
3623 	 * If a link layer address is found, we'll need to see if it's
3624 	 * already present in the address list, or allocate is as well.
3625 	 * When this block finishes, the link layer address will be on the
3626 	 * list.
3627 	 */
3628 	if (llsa != NULL) {
3629 		ll_ifma = if_findmulti(ifp, llsa);
3630 		if (ll_ifma == NULL) {
3631 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3632 			if (ll_ifma == NULL) {
3633 				--ifma->ifma_refcount;
3634 				if_freemulti(ifma);
3635 				error = ENOMEM;
3636 				goto free_llsa_out;
3637 			}
3638 			ll_ifma->ifma_flags |= IFMA_F_ENQUEUED;
3639 			CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3640 			    ifma_link);
3641 		} else
3642 			ll_ifma->ifma_refcount++;
3643 		ifma->ifma_llifma = ll_ifma;
3644 	}
3645 
3646 	/*
3647 	 * We now have a new multicast address, ifma, and possibly a new or
3648 	 * referenced link layer address.  Add the primary address to the
3649 	 * ifnet address list.
3650 	 */
3651 	ifma->ifma_flags |= IFMA_F_ENQUEUED;
3652 	CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3653 
3654 	if (retifma != NULL)
3655 		*retifma = ifma;
3656 
3657 	/*
3658 	 * Must generate the message while holding the lock so that 'ifma'
3659 	 * pointer is still valid.
3660 	 */
3661 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3662 	IF_ADDR_WUNLOCK(ifp);
3663 
3664 	/*
3665 	 * We are certain we have added something, so call down to the
3666 	 * interface to let them know about it.
3667 	 */
3668 	if (ifp->if_ioctl != NULL) {
3669 		if (THREAD_CAN_SLEEP())
3670 			(void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3671 		else
3672 			taskqueue_enqueue(taskqueue_swi, &ifp->if_addmultitask);
3673 	}
3674 
3675 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3676 		link_free_sdl(llsa);
3677 
3678 	return (0);
3679 
3680 free_llsa_out:
3681 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3682 		link_free_sdl(llsa);
3683 
3684 unlock_out:
3685 	IF_ADDR_WUNLOCK(ifp);
3686 	return (error);
3687 }
3688 
3689 static void
3690 if_siocaddmulti(void *arg, int pending)
3691 {
3692 	struct ifnet *ifp;
3693 
3694 	ifp = arg;
3695 #ifdef DIAGNOSTIC
3696 	if (pending > 1)
3697 		if_printf(ifp, "%d SIOCADDMULTI coalesced\n", pending);
3698 #endif
3699 	CURVNET_SET(ifp->if_vnet);
3700 	(void )(*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3701 	CURVNET_RESTORE();
3702 }
3703 
3704 /*
3705  * Delete a multicast group membership by network-layer group address.
3706  *
3707  * Returns ENOENT if the entry could not be found. If ifp no longer
3708  * exists, results are undefined. This entry point should only be used
3709  * from subsystems which do appropriate locking to hold ifp for the
3710  * duration of the call.
3711  * Network-layer protocol domains must use if_delmulti_ifma().
3712  */
3713 int
3714 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3715 {
3716 	struct ifmultiaddr *ifma;
3717 	int lastref;
3718 
3719 	KASSERT(ifp, ("%s: NULL ifp", __func__));
3720 
3721 	IF_ADDR_WLOCK(ifp);
3722 	lastref = 0;
3723 	ifma = if_findmulti(ifp, sa);
3724 	if (ifma != NULL)
3725 		lastref = if_delmulti_locked(ifp, ifma, 0);
3726 	IF_ADDR_WUNLOCK(ifp);
3727 
3728 	if (ifma == NULL)
3729 		return (ENOENT);
3730 
3731 	if (lastref && ifp->if_ioctl != NULL) {
3732 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3733 	}
3734 
3735 	return (0);
3736 }
3737 
3738 /*
3739  * Delete all multicast group membership for an interface.
3740  * Should be used to quickly flush all multicast filters.
3741  */
3742 void
3743 if_delallmulti(struct ifnet *ifp)
3744 {
3745 	struct ifmultiaddr *ifma;
3746 	struct ifmultiaddr *next;
3747 
3748 	IF_ADDR_WLOCK(ifp);
3749 	CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3750 		if_delmulti_locked(ifp, ifma, 0);
3751 	IF_ADDR_WUNLOCK(ifp);
3752 }
3753 
3754 void
3755 if_delmulti_ifma(struct ifmultiaddr *ifma)
3756 {
3757 	if_delmulti_ifma_flags(ifma, 0);
3758 }
3759 
3760 /*
3761  * Delete a multicast group membership by group membership pointer.
3762  * Network-layer protocol domains must use this routine.
3763  *
3764  * It is safe to call this routine if the ifp disappeared.
3765  */
3766 void
3767 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags)
3768 {
3769 	struct ifnet *ifp;
3770 	int lastref;
3771 	MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma);
3772 #ifdef INET
3773 	IN_MULTI_LIST_UNLOCK_ASSERT();
3774 #endif
3775 	ifp = ifma->ifma_ifp;
3776 #ifdef DIAGNOSTIC
3777 	if (ifp == NULL) {
3778 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3779 	} else {
3780 		struct epoch_tracker et;
3781 		struct ifnet *oifp;
3782 
3783 		NET_EPOCH_ENTER(et);
3784 		CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link)
3785 			if (ifp == oifp)
3786 				break;
3787 		NET_EPOCH_EXIT(et);
3788 		if (ifp != oifp)
3789 			ifp = NULL;
3790 	}
3791 #endif
3792 	/*
3793 	 * If and only if the ifnet instance exists: Acquire the address lock.
3794 	 */
3795 	if (ifp != NULL)
3796 		IF_ADDR_WLOCK(ifp);
3797 
3798 	lastref = if_delmulti_locked(ifp, ifma, flags);
3799 
3800 	if (ifp != NULL) {
3801 		/*
3802 		 * If and only if the ifnet instance exists:
3803 		 *  Release the address lock.
3804 		 *  If the group was left: update the hardware hash filter.
3805 		 */
3806 		IF_ADDR_WUNLOCK(ifp);
3807 		if (lastref && ifp->if_ioctl != NULL) {
3808 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3809 		}
3810 	}
3811 }
3812 
3813 /*
3814  * Perform deletion of network-layer and/or link-layer multicast address.
3815  *
3816  * Return 0 if the reference count was decremented.
3817  * Return 1 if the final reference was released, indicating that the
3818  * hardware hash filter should be reprogrammed.
3819  */
3820 static int
3821 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3822 {
3823 	struct ifmultiaddr *ll_ifma;
3824 
3825 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3826 		KASSERT(ifma->ifma_ifp == ifp,
3827 		    ("%s: inconsistent ifp %p", __func__, ifp));
3828 		IF_ADDR_WLOCK_ASSERT(ifp);
3829 	}
3830 
3831 	ifp = ifma->ifma_ifp;
3832 	MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : "");
3833 
3834 	/*
3835 	 * If the ifnet is detaching, null out references to ifnet,
3836 	 * so that upper protocol layers will notice, and not attempt
3837 	 * to obtain locks for an ifnet which no longer exists. The
3838 	 * routing socket announcement must happen before the ifnet
3839 	 * instance is detached from the system.
3840 	 */
3841 	if (detaching) {
3842 #ifdef DIAGNOSTIC
3843 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3844 #endif
3845 		/*
3846 		 * ifp may already be nulled out if we are being reentered
3847 		 * to delete the ll_ifma.
3848 		 */
3849 		if (ifp != NULL) {
3850 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3851 			ifma->ifma_ifp = NULL;
3852 		}
3853 	}
3854 
3855 	if (--ifma->ifma_refcount > 0)
3856 		return 0;
3857 
3858 	if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) {
3859 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
3860 		ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3861 	}
3862 	/*
3863 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3864 	 * have been associated with it. Release it first if so.
3865 	 */
3866 	ll_ifma = ifma->ifma_llifma;
3867 	if (ll_ifma != NULL) {
3868 		KASSERT(ifma->ifma_lladdr != NULL,
3869 		    ("%s: llifma w/o lladdr", __func__));
3870 		if (detaching)
3871 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3872 		if (--ll_ifma->ifma_refcount == 0) {
3873 			if (ifp != NULL) {
3874 				if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) {
3875 					CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr,
3876 						ifma_link);
3877 					ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3878 				}
3879 			}
3880 			if_freemulti(ll_ifma);
3881 		}
3882 	}
3883 #ifdef INVARIANTS
3884 	if (ifp) {
3885 		struct ifmultiaddr *ifmatmp;
3886 
3887 		CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link)
3888 			MPASS(ifma != ifmatmp);
3889 	}
3890 #endif
3891 	if_freemulti(ifma);
3892 	/*
3893 	 * The last reference to this instance of struct ifmultiaddr
3894 	 * was released; the hardware should be notified of this change.
3895 	 */
3896 	return 1;
3897 }
3898 
3899 /*
3900  * Set the link layer address on an interface.
3901  *
3902  * At this time we only support certain types of interfaces,
3903  * and we don't allow the length of the address to change.
3904  *
3905  * Set noinline to be dtrace-friendly
3906  */
3907 __noinline int
3908 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3909 {
3910 	struct sockaddr_dl *sdl;
3911 	struct ifaddr *ifa;
3912 	struct ifreq ifr;
3913 
3914 	ifa = ifp->if_addr;
3915 	if (ifa == NULL)
3916 		return (EINVAL);
3917 
3918 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3919 	if (sdl == NULL)
3920 		return (EINVAL);
3921 
3922 	if (len != sdl->sdl_alen)	/* don't allow length to change */
3923 		return (EINVAL);
3924 
3925 	switch (ifp->if_type) {
3926 	case IFT_ETHER:
3927 	case IFT_XETHER:
3928 	case IFT_L2VLAN:
3929 	case IFT_BRIDGE:
3930 	case IFT_IEEE8023ADLAG:
3931 		bcopy(lladdr, LLADDR(sdl), len);
3932 		break;
3933 	default:
3934 		return (ENODEV);
3935 	}
3936 
3937 	/*
3938 	 * If the interface is already up, we need
3939 	 * to re-init it in order to reprogram its
3940 	 * address filter.
3941 	 */
3942 	if ((ifp->if_flags & IFF_UP) != 0) {
3943 		if (ifp->if_ioctl) {
3944 			ifp->if_flags &= ~IFF_UP;
3945 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3946 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3947 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3948 			ifp->if_flags |= IFF_UP;
3949 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3950 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3951 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3952 		}
3953 	}
3954 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3955 
3956 	return (0);
3957 }
3958 
3959 /*
3960  * Compat function for handling basic encapsulation requests.
3961  * Not converted stacks (FDDI, IB, ..) supports traditional
3962  * output model: ARP (and other similar L2 protocols) are handled
3963  * inside output routine, arpresolve/nd6_resolve() returns MAC
3964  * address instead of full prepend.
3965  *
3966  * This function creates calculated header==MAC for IPv4/IPv6 and
3967  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3968  * address families.
3969  */
3970 static int
3971 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3972 {
3973 	if (req->rtype != IFENCAP_LL)
3974 		return (EOPNOTSUPP);
3975 
3976 	if (req->bufsize < req->lladdr_len)
3977 		return (ENOMEM);
3978 
3979 	switch (req->family) {
3980 	case AF_INET:
3981 	case AF_INET6:
3982 		break;
3983 	default:
3984 		return (EAFNOSUPPORT);
3985 	}
3986 
3987 	/* Copy lladdr to storage as is */
3988 	memmove(req->buf, req->lladdr, req->lladdr_len);
3989 	req->bufsize = req->lladdr_len;
3990 	req->lladdr_off = 0;
3991 
3992 	return (0);
3993 }
3994 
3995 /*
3996  * Tunnel interfaces can nest, also they may cause infinite recursion
3997  * calls when misconfigured. We'll prevent this by detecting loops.
3998  * High nesting level may cause stack exhaustion. We'll prevent this
3999  * by introducing upper limit.
4000  *
4001  * Return 0, if tunnel nesting count is equal or less than limit.
4002  */
4003 int
4004 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie,
4005     int limit)
4006 {
4007 	struct m_tag *mtag;
4008 	int count;
4009 
4010 	count = 1;
4011 	mtag = NULL;
4012 	while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) {
4013 		if (*(struct ifnet **)(mtag + 1) == ifp) {
4014 			log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp));
4015 			return (EIO);
4016 		}
4017 		count++;
4018 	}
4019 	if (count > limit) {
4020 		log(LOG_NOTICE,
4021 		    "%s: if_output recursively called too many times(%d)\n",
4022 		    if_name(ifp), count);
4023 		return (EIO);
4024 	}
4025 	mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT);
4026 	if (mtag == NULL)
4027 		return (ENOMEM);
4028 	*(struct ifnet **)(mtag + 1) = ifp;
4029 	m_tag_prepend(m, mtag);
4030 	return (0);
4031 }
4032 
4033 /*
4034  * Get the link layer address that was read from the hardware at attach.
4035  *
4036  * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
4037  * their component interfaces as IFT_IEEE8023ADLAG.
4038  */
4039 int
4040 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
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_default(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 	m_freem(m);
4129 }
4130 
4131 int
4132 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
4133 {
4134 	int active = 0;
4135 
4136 	IF_LOCK(ifq);
4137 	if (_IF_QFULL(ifq)) {
4138 		IF_UNLOCK(ifq);
4139 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
4140 		m_freem(m);
4141 		return (0);
4142 	}
4143 	if (ifp != NULL) {
4144 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
4145 		if (m->m_flags & (M_BCAST|M_MCAST))
4146 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
4147 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
4148 	}
4149 	_IF_ENQUEUE(ifq, m);
4150 	IF_UNLOCK(ifq);
4151 	if (ifp != NULL && !active)
4152 		(*(ifp)->if_start)(ifp);
4153 	return (1);
4154 }
4155 
4156 void
4157 if_register_com_alloc(u_char type,
4158     if_com_alloc_t *a, if_com_free_t *f)
4159 {
4160 
4161 	KASSERT(if_com_alloc[type] == NULL,
4162 	    ("if_register_com_alloc: %d already registered", type));
4163 	KASSERT(if_com_free[type] == NULL,
4164 	    ("if_register_com_alloc: %d free already registered", type));
4165 
4166 	if_com_alloc[type] = a;
4167 	if_com_free[type] = f;
4168 }
4169 
4170 void
4171 if_deregister_com_alloc(u_char type)
4172 {
4173 
4174 	KASSERT(if_com_alloc[type] != NULL,
4175 	    ("if_deregister_com_alloc: %d not registered", type));
4176 	KASSERT(if_com_free[type] != NULL,
4177 	    ("if_deregister_com_alloc: %d free not registered", type));
4178 
4179 	/*
4180 	 * Ensure all pending EPOCH(9) callbacks have been executed. This
4181 	 * fixes issues about late invocation of if_destroy(), which leads
4182 	 * to memory leak from if_com_alloc[type] allocated if_l2com.
4183 	 */
4184 	NET_EPOCH_DRAIN_CALLBACKS();
4185 
4186 	if_com_alloc[type] = NULL;
4187 	if_com_free[type] = NULL;
4188 }
4189 
4190 /* API for driver access to network stack owned ifnet.*/
4191 uint64_t
4192 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
4193 {
4194 	uint64_t oldbrate;
4195 
4196 	oldbrate = ifp->if_baudrate;
4197 	ifp->if_baudrate = baudrate;
4198 	return (oldbrate);
4199 }
4200 
4201 uint64_t
4202 if_getbaudrate(const if_t ifp)
4203 {
4204 	return (ifp->if_baudrate);
4205 }
4206 
4207 int
4208 if_setcapabilities(if_t ifp, int capabilities)
4209 {
4210 	ifp->if_capabilities = capabilities;
4211 	return (0);
4212 }
4213 
4214 int
4215 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
4216 {
4217 	ifp->if_capabilities &= ~clearbit;
4218 	ifp->if_capabilities |= setbit;
4219 	return (0);
4220 }
4221 
4222 int
4223 if_getcapabilities(const if_t ifp)
4224 {
4225 	return (ifp->if_capabilities);
4226 }
4227 
4228 int
4229 if_setcapenable(if_t ifp, int capabilities)
4230 {
4231 	ifp->if_capenable = capabilities;
4232 	return (0);
4233 }
4234 
4235 int
4236 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
4237 {
4238 	ifp->if_capenable &= ~clearcap;
4239 	ifp->if_capenable |= setcap;
4240 	return (0);
4241 }
4242 
4243 int
4244 if_setcapabilities2(if_t ifp, int capabilities)
4245 {
4246 	ifp->if_capabilities2 = capabilities;
4247 	return (0);
4248 }
4249 
4250 int
4251 if_setcapabilities2bit(if_t ifp, int setbit, int clearbit)
4252 {
4253 	ifp->if_capabilities2 &= ~clearbit;
4254 	ifp->if_capabilities2 |= setbit;
4255 	return (0);
4256 }
4257 
4258 int
4259 if_getcapabilities2(const if_t ifp)
4260 {
4261 	return (ifp->if_capabilities2);
4262 }
4263 
4264 int
4265 if_setcapenable2(if_t ifp, int capabilities2)
4266 {
4267 	ifp->if_capenable2 = capabilities2;
4268 	return (0);
4269 }
4270 
4271 int
4272 if_setcapenable2bit(if_t ifp, int setcap, int clearcap)
4273 {
4274 	ifp->if_capenable2 &= ~clearcap;
4275 	ifp->if_capenable2 |= setcap;
4276 	return (0);
4277 }
4278 
4279 const char *
4280 if_getdname(const if_t ifp)
4281 {
4282 	return (ifp->if_dname);
4283 }
4284 
4285 void
4286 if_setdname(if_t ifp, const char *dname)
4287 {
4288 	ifp->if_dname = dname;
4289 }
4290 
4291 const char *
4292 if_name(if_t ifp)
4293 {
4294 	return (ifp->if_xname);
4295 }
4296 
4297 int
4298 if_setname(if_t ifp, const char *name)
4299 {
4300 	if (strlen(name) > sizeof(ifp->if_xname) - 1)
4301 		return (ENAMETOOLONG);
4302 	strcpy(ifp->if_xname, name);
4303 
4304 	return (0);
4305 }
4306 
4307 int
4308 if_togglecapenable(if_t ifp, int togglecap)
4309 {
4310 	ifp->if_capenable ^= togglecap;
4311 	return (0);
4312 }
4313 
4314 int
4315 if_getcapenable(const if_t ifp)
4316 {
4317 	return (ifp->if_capenable);
4318 }
4319 
4320 int
4321 if_togglecapenable2(if_t ifp, int togglecap)
4322 {
4323 	ifp->if_capenable2 ^= togglecap;
4324 	return (0);
4325 }
4326 
4327 int
4328 if_getcapenable2(const if_t ifp)
4329 {
4330 	return (ifp->if_capenable2);
4331 }
4332 
4333 int
4334 if_getdunit(const if_t ifp)
4335 {
4336 	return (ifp->if_dunit);
4337 }
4338 
4339 int
4340 if_getindex(const if_t ifp)
4341 {
4342 	return (ifp->if_index);
4343 }
4344 
4345 int
4346 if_getidxgen(const if_t ifp)
4347 {
4348 	return (ifp->if_idxgen);
4349 }
4350 
4351 void
4352 if_setdescr(if_t ifp, char *descrbuf)
4353 {
4354 	sx_xlock(&ifdescr_sx);
4355 	char *odescrbuf = ifp->if_description;
4356 	ifp->if_description = descrbuf;
4357 	sx_xunlock(&ifdescr_sx);
4358 
4359 	if_freedescr(odescrbuf);
4360 }
4361 
4362 char *
4363 if_allocdescr(size_t sz, int malloc_flag)
4364 {
4365 	malloc_flag &= (M_WAITOK | M_NOWAIT);
4366 	return (malloc(sz, M_IFDESCR, M_ZERO | malloc_flag));
4367 }
4368 
4369 void
4370 if_freedescr(char *descrbuf)
4371 {
4372 	free(descrbuf, M_IFDESCR);
4373 }
4374 
4375 int
4376 if_getalloctype(const if_t ifp)
4377 {
4378 	return (ifp->if_alloctype);
4379 }
4380 
4381 /*
4382  * This is largely undesirable because it ties ifnet to a device, but does
4383  * provide flexiblity for an embedded product vendor. Should be used with
4384  * the understanding that it violates the interface boundaries, and should be
4385  * a last resort only.
4386  */
4387 int
4388 if_setdev(if_t ifp, void *dev)
4389 {
4390 	return (0);
4391 }
4392 
4393 int
4394 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
4395 {
4396 	ifp->if_drv_flags &= ~clear_flags;
4397 	ifp->if_drv_flags |= set_flags;
4398 
4399 	return (0);
4400 }
4401 
4402 int
4403 if_getdrvflags(const if_t ifp)
4404 {
4405 	return (ifp->if_drv_flags);
4406 }
4407 
4408 int
4409 if_setdrvflags(if_t ifp, int flags)
4410 {
4411 	ifp->if_drv_flags = flags;
4412 	return (0);
4413 }
4414 
4415 int
4416 if_setflags(if_t ifp, int flags)
4417 {
4418 	ifp->if_flags = flags;
4419 	return (0);
4420 }
4421 
4422 int
4423 if_setflagbits(if_t ifp, int set, int clear)
4424 {
4425 	ifp->if_flags &= ~clear;
4426 	ifp->if_flags |= set;
4427 	return (0);
4428 }
4429 
4430 int
4431 if_getflags(const if_t ifp)
4432 {
4433 	return (ifp->if_flags);
4434 }
4435 
4436 int
4437 if_clearhwassist(if_t ifp)
4438 {
4439 	ifp->if_hwassist = 0;
4440 	return (0);
4441 }
4442 
4443 int
4444 if_sethwassistbits(if_t ifp, int toset, int toclear)
4445 {
4446 	ifp->if_hwassist &= ~toclear;
4447 	ifp->if_hwassist |= toset;
4448 
4449 	return (0);
4450 }
4451 
4452 int
4453 if_sethwassist(if_t ifp, int hwassist_bit)
4454 {
4455 	ifp->if_hwassist = hwassist_bit;
4456 	return (0);
4457 }
4458 
4459 int
4460 if_gethwassist(const if_t ifp)
4461 {
4462 	return (ifp->if_hwassist);
4463 }
4464 
4465 int
4466 if_togglehwassist(if_t ifp, int toggle_bits)
4467 {
4468 	ifp->if_hwassist ^= toggle_bits;
4469 	return (0);
4470 }
4471 
4472 int
4473 if_setmtu(if_t ifp, int mtu)
4474 {
4475 	ifp->if_mtu = mtu;
4476 	return (0);
4477 }
4478 
4479 void
4480 if_notifymtu(if_t ifp)
4481 {
4482 #ifdef INET6
4483 	nd6_setmtu(ifp);
4484 #endif
4485 	rt_updatemtu(ifp);
4486 }
4487 
4488 int
4489 if_getmtu(const if_t ifp)
4490 {
4491 	return (ifp->if_mtu);
4492 }
4493 
4494 int
4495 if_getmtu_family(const if_t ifp, int family)
4496 {
4497 	struct domain *dp;
4498 
4499 	SLIST_FOREACH(dp, &domains, dom_next) {
4500 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
4501 			return (dp->dom_ifmtu(ifp));
4502 	}
4503 
4504 	return (ifp->if_mtu);
4505 }
4506 
4507 /*
4508  * Methods for drivers to access interface unicast and multicast
4509  * link level addresses.  Driver shall not know 'struct ifaddr' neither
4510  * 'struct ifmultiaddr'.
4511  */
4512 u_int
4513 if_lladdr_count(if_t ifp)
4514 {
4515 	struct epoch_tracker et;
4516 	struct ifaddr *ifa;
4517 	u_int count;
4518 
4519 	count = 0;
4520 	NET_EPOCH_ENTER(et);
4521 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
4522 		if (ifa->ifa_addr->sa_family == AF_LINK)
4523 			count++;
4524 	NET_EPOCH_EXIT(et);
4525 
4526 	return (count);
4527 }
4528 
4529 int
4530 if_foreach(if_foreach_cb_t cb, void *cb_arg)
4531 {
4532 	if_t ifp;
4533 	int error;
4534 
4535 	NET_EPOCH_ASSERT();
4536 	MPASS(cb);
4537 
4538 	error = 0;
4539 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
4540 		error = cb(ifp, cb_arg);
4541 		if (error != 0)
4542 			break;
4543 	}
4544 
4545 	return (error);
4546 }
4547 
4548 /*
4549  * Iterates over the list of interfaces, permitting callback function @cb to sleep.
4550  * Stops iteration if @cb returns non-zero error code.
4551  * Returns the last error code from @cb.
4552  * @match_cb: optional match callback limiting the iteration to only matched interfaces
4553  * @match_arg: argument to pass to @match_cb
4554  * @cb: iteration callback
4555  * @cb_arg: argument to pass to @cb
4556  */
4557 int
4558 if_foreach_sleep(if_foreach_match_t match_cb, void *match_arg, if_foreach_cb_t cb,
4559     void *cb_arg)
4560 {
4561 	int match_count = 0, array_size = 16; /* 128 bytes for malloc */
4562 	struct ifnet **match_array = NULL;
4563 	int error = 0;
4564 
4565 	MPASS(cb);
4566 
4567 	while (true) {
4568 		struct ifnet **new_array;
4569 		int new_size = array_size;
4570 		struct epoch_tracker et;
4571 		struct ifnet *ifp;
4572 
4573 		while (new_size < match_count)
4574 			new_size *= 2;
4575 		new_array = malloc(new_size * sizeof(void *), M_TEMP, M_WAITOK);
4576 		if (match_array != NULL)
4577 			memcpy(new_array, match_array, array_size * sizeof(void *));
4578 		free(match_array, M_TEMP);
4579 		match_array = new_array;
4580 		array_size = new_size;
4581 
4582 		match_count = 0;
4583 		NET_EPOCH_ENTER(et);
4584 		CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
4585 			if (match_cb != NULL && !match_cb(ifp, match_arg))
4586 				continue;
4587 			if (match_count < array_size) {
4588 				if (if_try_ref(ifp))
4589 					match_array[match_count++] = ifp;
4590 			} else
4591 				match_count++;
4592 		}
4593 		NET_EPOCH_EXIT(et);
4594 
4595 		if (match_count > array_size) {
4596 			for (int i = 0; i < array_size; i++)
4597 				if_rele(match_array[i]);
4598 			continue;
4599 		} else {
4600 			for (int i = 0; i < match_count; i++) {
4601 				if (error == 0)
4602 					error = cb(match_array[i], cb_arg);
4603 				if_rele(match_array[i]);
4604 			}
4605 			free(match_array, M_TEMP);
4606 			break;
4607 		}
4608 	}
4609 
4610 	return (error);
4611 }
4612 
4613 
4614 /*
4615  * Uses just 1 pointer of the 4 available in the public struct.
4616  */
4617 if_t
4618 if_iter_start(struct if_iter *iter)
4619 {
4620 	if_t ifp;
4621 
4622 	NET_EPOCH_ASSERT();
4623 
4624 	bzero(iter, sizeof(*iter));
4625 	ifp = CK_STAILQ_FIRST(&V_ifnet);
4626 	if (ifp != NULL)
4627 		iter->context[0] = CK_STAILQ_NEXT(ifp, if_link);
4628 	else
4629 		iter->context[0] = NULL;
4630 	return (ifp);
4631 }
4632 
4633 if_t
4634 if_iter_next(struct if_iter *iter)
4635 {
4636 	if_t cur_ifp = iter->context[0];
4637 
4638 	if (cur_ifp != NULL)
4639 		iter->context[0] = CK_STAILQ_NEXT(cur_ifp, if_link);
4640 	return (cur_ifp);
4641 }
4642 
4643 void
4644 if_iter_finish(struct if_iter *iter)
4645 {
4646 	/* Nothing to do here for now. */
4647 }
4648 
4649 u_int
4650 if_foreach_lladdr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4651 {
4652 	struct epoch_tracker et;
4653 	struct ifaddr *ifa;
4654 	u_int count;
4655 
4656 	MPASS(cb);
4657 
4658 	count = 0;
4659 	NET_EPOCH_ENTER(et);
4660 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
4661 		if (ifa->ifa_addr->sa_family != AF_LINK)
4662 			continue;
4663 		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifa->ifa_addr,
4664 		    count);
4665 	}
4666 	NET_EPOCH_EXIT(et);
4667 
4668 	return (count);
4669 }
4670 
4671 u_int
4672 if_llmaddr_count(if_t ifp)
4673 {
4674 	struct epoch_tracker et;
4675 	struct ifmultiaddr *ifma;
4676 	int count;
4677 
4678 	count = 0;
4679 	NET_EPOCH_ENTER(et);
4680 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4681 		if (ifma->ifma_addr->sa_family == AF_LINK)
4682 			count++;
4683 	NET_EPOCH_EXIT(et);
4684 
4685 	return (count);
4686 }
4687 
4688 u_int
4689 if_foreach_llmaddr(if_t ifp, iflladdr_cb_t cb, void *cb_arg)
4690 {
4691 	struct epoch_tracker et;
4692 	struct ifmultiaddr *ifma;
4693 	u_int count;
4694 
4695 	MPASS(cb);
4696 
4697 	count = 0;
4698 	NET_EPOCH_ENTER(et);
4699 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
4700 		if (ifma->ifma_addr->sa_family != AF_LINK)
4701 			continue;
4702 		count += (*cb)(cb_arg, (struct sockaddr_dl *)ifma->ifma_addr,
4703 		    count);
4704 	}
4705 	NET_EPOCH_EXIT(et);
4706 
4707 	return (count);
4708 }
4709 
4710 u_int
4711 if_foreach_addr_type(if_t ifp, int type, if_addr_cb_t cb, void *cb_arg)
4712 {
4713 	struct epoch_tracker et;
4714 	struct ifaddr *ifa;
4715 	u_int count;
4716 
4717 	MPASS(cb);
4718 
4719 	count = 0;
4720 	NET_EPOCH_ENTER(et);
4721 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
4722 		if (ifa->ifa_addr->sa_family != type)
4723 			continue;
4724 		count += (*cb)(cb_arg, ifa, count);
4725 	}
4726 	NET_EPOCH_EXIT(et);
4727 
4728 	return (count);
4729 }
4730 
4731 int
4732 if_setsoftc(if_t ifp, void *softc)
4733 {
4734 	ifp->if_softc = softc;
4735 	return (0);
4736 }
4737 
4738 void *
4739 if_getsoftc(const if_t ifp)
4740 {
4741 	return (ifp->if_softc);
4742 }
4743 
4744 void
4745 if_setrcvif(struct mbuf *m, if_t ifp)
4746 {
4747 
4748 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
4749 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
4750 }
4751 
4752 void
4753 if_setvtag(struct mbuf *m, uint16_t tag)
4754 {
4755 	m->m_pkthdr.ether_vtag = tag;
4756 }
4757 
4758 uint16_t
4759 if_getvtag(struct mbuf *m)
4760 {
4761 	return (m->m_pkthdr.ether_vtag);
4762 }
4763 
4764 int
4765 if_sendq_empty(if_t ifp)
4766 {
4767 	return (IFQ_DRV_IS_EMPTY(&ifp->if_snd));
4768 }
4769 
4770 struct ifaddr *
4771 if_getifaddr(const if_t ifp)
4772 {
4773 	return (ifp->if_addr);
4774 }
4775 
4776 int
4777 if_getamcount(const if_t ifp)
4778 {
4779 	return (ifp->if_amcount);
4780 }
4781 
4782 int
4783 if_setsendqready(if_t ifp)
4784 {
4785 	IFQ_SET_READY(&ifp->if_snd);
4786 	return (0);
4787 }
4788 
4789 int
4790 if_setsendqlen(if_t ifp, int tx_desc_count)
4791 {
4792 	IFQ_SET_MAXLEN(&ifp->if_snd, tx_desc_count);
4793 	ifp->if_snd.ifq_drv_maxlen = tx_desc_count;
4794 	return (0);
4795 }
4796 
4797 void
4798 if_setnetmapadapter(if_t ifp, struct netmap_adapter *na)
4799 {
4800 	ifp->if_netmap = na;
4801 }
4802 
4803 struct netmap_adapter *
4804 if_getnetmapadapter(if_t ifp)
4805 {
4806 	return (ifp->if_netmap);
4807 }
4808 
4809 int
4810 if_vlantrunkinuse(if_t ifp)
4811 {
4812 	return (ifp->if_vlantrunk != NULL);
4813 }
4814 
4815 void
4816 if_init(if_t ifp, void *ctx)
4817 {
4818 	(*ifp->if_init)(ctx);
4819 }
4820 
4821 void
4822 if_input(if_t ifp, struct mbuf* sendmp)
4823 {
4824 	(*ifp->if_input)(ifp, sendmp);
4825 }
4826 
4827 int
4828 if_transmit(if_t ifp, struct mbuf *m)
4829 {
4830 	return ((*ifp->if_transmit)(ifp, m));
4831 }
4832 
4833 int
4834 if_resolvemulti(if_t ifp, struct sockaddr **srcs, struct sockaddr *dst)
4835 {
4836 	if (ifp->if_resolvemulti == NULL)
4837 		return (EOPNOTSUPP);
4838 
4839 	return (ifp->if_resolvemulti(ifp, srcs, dst));
4840 }
4841 
4842 struct mbuf *
4843 if_dequeue(if_t ifp)
4844 {
4845 	struct mbuf *m;
4846 
4847 	IFQ_DRV_DEQUEUE(&ifp->if_snd, m);
4848 	return (m);
4849 }
4850 
4851 int
4852 if_sendq_prepend(if_t ifp, struct mbuf *m)
4853 {
4854 	IFQ_DRV_PREPEND(&ifp->if_snd, m);
4855 	return (0);
4856 }
4857 
4858 int
4859 if_setifheaderlen(if_t ifp, int len)
4860 {
4861 	ifp->if_hdrlen = len;
4862 	return (0);
4863 }
4864 
4865 caddr_t
4866 if_getlladdr(const if_t ifp)
4867 {
4868 	return (IF_LLADDR(ifp));
4869 }
4870 
4871 void *
4872 if_gethandle(u_char type)
4873 {
4874 	return (if_alloc(type));
4875 }
4876 
4877 void
4878 if_bpfmtap(if_t ifp, struct mbuf *m)
4879 {
4880 	BPF_MTAP(ifp, m);
4881 }
4882 
4883 void
4884 if_etherbpfmtap(if_t ifp, struct mbuf *m)
4885 {
4886 	ETHER_BPF_MTAP(ifp, m);
4887 }
4888 
4889 void
4890 if_vlancap(if_t ifp)
4891 {
4892 	VLAN_CAPABILITIES(ifp);
4893 }
4894 
4895 int
4896 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4897 {
4898 	ifp->if_hw_tsomax = if_hw_tsomax;
4899         return (0);
4900 }
4901 
4902 int
4903 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4904 {
4905 	ifp->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4906         return (0);
4907 }
4908 
4909 int
4910 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4911 {
4912 	ifp->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4913         return (0);
4914 }
4915 
4916 u_int
4917 if_gethwtsomax(const if_t ifp)
4918 {
4919 	return (ifp->if_hw_tsomax);
4920 }
4921 
4922 u_int
4923 if_gethwtsomaxsegcount(const if_t ifp)
4924 {
4925 	return (ifp->if_hw_tsomaxsegcount);
4926 }
4927 
4928 u_int
4929 if_gethwtsomaxsegsize(const if_t ifp)
4930 {
4931 	return (ifp->if_hw_tsomaxsegsize);
4932 }
4933 
4934 void
4935 if_setinitfn(if_t ifp, if_init_fn_t init_fn)
4936 {
4937 	ifp->if_init = init_fn;
4938 }
4939 
4940 void
4941 if_setinputfn(if_t ifp, if_input_fn_t input_fn)
4942 {
4943 	ifp->if_input = input_fn;
4944 }
4945 
4946 if_input_fn_t
4947 if_getinputfn(if_t ifp)
4948 {
4949 	return (ifp->if_input);
4950 }
4951 
4952 void
4953 if_setioctlfn(if_t ifp, if_ioctl_fn_t ioctl_fn)
4954 {
4955 	ifp->if_ioctl = ioctl_fn;
4956 }
4957 
4958 void
4959 if_setoutputfn(if_t ifp, if_output_fn_t output_fn)
4960 {
4961 	ifp->if_output = output_fn;
4962 }
4963 
4964 void
4965 if_setstartfn(if_t ifp, if_start_fn_t start_fn)
4966 {
4967 	ifp->if_start = start_fn;
4968 }
4969 
4970 if_start_fn_t
4971 if_getstartfn(if_t ifp)
4972 {
4973 	return (ifp->if_start);
4974 }
4975 
4976 void
4977 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4978 {
4979 	ifp->if_transmit = start_fn;
4980 }
4981 
4982 if_transmit_fn_t
4983 if_gettransmitfn(if_t ifp)
4984 {
4985 	return (ifp->if_transmit);
4986 }
4987 
4988 void
4989 if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4990 {
4991 	ifp->if_qflush = flush_fn;
4992 }
4993 
4994 void
4995 if_setsndtagallocfn(if_t ifp, if_snd_tag_alloc_t alloc_fn)
4996 {
4997 	ifp->if_snd_tag_alloc = alloc_fn;
4998 }
4999 
5000 int
5001 if_snd_tag_alloc(if_t ifp, union if_snd_tag_alloc_params *params,
5002     struct m_snd_tag **mstp)
5003 {
5004 	if (ifp->if_snd_tag_alloc == NULL)
5005 		return (EOPNOTSUPP);
5006 	return (ifp->if_snd_tag_alloc(ifp, params, mstp));
5007 }
5008 
5009 void
5010 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
5011 {
5012 	ifp->if_get_counter = fn;
5013 }
5014 
5015 void
5016 if_setreassignfn(if_t ifp, if_reassign_fn_t fn)
5017 {
5018 	ifp->if_reassign = fn;
5019 }
5020 
5021 void
5022 if_setratelimitqueryfn(if_t ifp, if_ratelimit_query_t fn)
5023 {
5024 	ifp->if_ratelimit_query = fn;
5025 }
5026 
5027 void
5028 if_setdebugnet_methods(if_t ifp, struct debugnet_methods *m)
5029 {
5030 	ifp->if_debugnet_methods = m;
5031 }
5032 
5033 struct label *
5034 if_getmaclabel(if_t ifp)
5035 {
5036 	return (ifp->if_label);
5037 }
5038 
5039 void
5040 if_setmaclabel(if_t ifp, struct label *label)
5041 {
5042 	ifp->if_label = label;
5043 }
5044 
5045 int
5046 if_gettype(if_t ifp)
5047 {
5048 	return (ifp->if_type);
5049 }
5050 
5051 void *
5052 if_getllsoftc(if_t ifp)
5053 {
5054 	return (ifp->if_llsoftc);
5055 }
5056 
5057 void
5058 if_setllsoftc(if_t ifp, void *llsoftc)
5059 {
5060 	ifp->if_llsoftc = llsoftc;
5061 };
5062 
5063 int
5064 if_getlinkstate(if_t ifp)
5065 {
5066 	return (ifp->if_link_state);
5067 }
5068 
5069 const uint8_t *
5070 if_getbroadcastaddr(if_t ifp)
5071 {
5072 	return (ifp->if_broadcastaddr);
5073 }
5074 
5075 void
5076 if_setbroadcastaddr(if_t ifp, const uint8_t *addr)
5077 {
5078 	ifp->if_broadcastaddr = addr;
5079 }
5080 
5081 int
5082 if_getnumadomain(if_t ifp)
5083 {
5084 	return (ifp->if_numa_domain);
5085 }
5086 
5087 uint64_t
5088 if_getcounter(if_t ifp, ift_counter counter)
5089 {
5090 	return (ifp->if_get_counter(ifp, counter));
5091 }
5092 
5093 bool
5094 if_altq_is_enabled(if_t ifp)
5095 {
5096 	return (ALTQ_IS_ENABLED(&ifp->if_snd));
5097 }
5098 
5099 struct vnet *
5100 if_getvnet(if_t ifp)
5101 {
5102 	return (ifp->if_vnet);
5103 }
5104 
5105 void *
5106 if_getafdata(if_t ifp, int af)
5107 {
5108 	return (ifp->if_afdata[af]);
5109 }
5110 
5111 u_int
5112 if_getfib(if_t ifp)
5113 {
5114 	return (ifp->if_fib);
5115 }
5116 
5117 uint8_t
5118 if_getaddrlen(if_t ifp)
5119 {
5120 	return (ifp->if_addrlen);
5121 }
5122 
5123 struct bpf_if *
5124 if_getbpf(if_t ifp)
5125 {
5126 	return (ifp->if_bpf);
5127 }
5128 
5129 struct ifvlantrunk *
5130 if_getvlantrunk(if_t ifp)
5131 {
5132 	return (ifp->if_vlantrunk);
5133 }
5134 
5135 uint8_t
5136 if_getpcp(if_t ifp)
5137 {
5138 	return (ifp->if_pcp);
5139 }
5140 
5141 void *
5142 if_getl2com(if_t ifp)
5143 {
5144 	return (ifp->if_l2com);
5145 }
5146 
5147 #ifdef DDB
5148 static void
5149 if_show_ifnet(struct ifnet *ifp)
5150 {
5151 	if (ifp == NULL)
5152 		return;
5153 	db_printf("%s:\n", ifp->if_xname);
5154 #define	IF_DB_PRINTF(f, e)	db_printf("   %s = " f "\n", #e, ifp->e);
5155 	IF_DB_PRINTF("%s", if_dname);
5156 	IF_DB_PRINTF("%d", if_dunit);
5157 	IF_DB_PRINTF("%s", if_description);
5158 	IF_DB_PRINTF("%u", if_index);
5159 	IF_DB_PRINTF("%d", if_idxgen);
5160 	IF_DB_PRINTF("%u", if_refcount);
5161 	IF_DB_PRINTF("%p", if_softc);
5162 	IF_DB_PRINTF("%p", if_l2com);
5163 	IF_DB_PRINTF("%p", if_llsoftc);
5164 	IF_DB_PRINTF("%d", if_amcount);
5165 	IF_DB_PRINTF("%p", if_addr);
5166 	IF_DB_PRINTF("%p", if_broadcastaddr);
5167 	IF_DB_PRINTF("%p", if_afdata);
5168 	IF_DB_PRINTF("%d", if_afdata_initialized);
5169 	IF_DB_PRINTF("%u", if_fib);
5170 	IF_DB_PRINTF("%p", if_vnet);
5171 	IF_DB_PRINTF("%p", if_home_vnet);
5172 	IF_DB_PRINTF("%p", if_vlantrunk);
5173 	IF_DB_PRINTF("%p", if_bpf);
5174 	IF_DB_PRINTF("%u", if_pcount);
5175 	IF_DB_PRINTF("%p", if_bridge);
5176 	IF_DB_PRINTF("%p", if_lagg);
5177 	IF_DB_PRINTF("%p", if_pf_kif);
5178 	IF_DB_PRINTF("%p", if_carp);
5179 	IF_DB_PRINTF("%p", if_label);
5180 	IF_DB_PRINTF("%p", if_netmap);
5181 	IF_DB_PRINTF("0x%08x", if_flags);
5182 	IF_DB_PRINTF("0x%08x", if_drv_flags);
5183 	IF_DB_PRINTF("0x%08x", if_capabilities);
5184 	IF_DB_PRINTF("0x%08x", if_capenable);
5185 	IF_DB_PRINTF("%p", if_snd.ifq_head);
5186 	IF_DB_PRINTF("%p", if_snd.ifq_tail);
5187 	IF_DB_PRINTF("%d", if_snd.ifq_len);
5188 	IF_DB_PRINTF("%d", if_snd.ifq_maxlen);
5189 	IF_DB_PRINTF("%p", if_snd.ifq_drv_head);
5190 	IF_DB_PRINTF("%p", if_snd.ifq_drv_tail);
5191 	IF_DB_PRINTF("%d", if_snd.ifq_drv_len);
5192 	IF_DB_PRINTF("%d", if_snd.ifq_drv_maxlen);
5193 	IF_DB_PRINTF("%d", if_snd.altq_type);
5194 	IF_DB_PRINTF("%x", if_snd.altq_flags);
5195 #undef IF_DB_PRINTF
5196 }
5197 
5198 DB_SHOW_COMMAND(ifnet, db_show_ifnet)
5199 {
5200 	if (!have_addr) {
5201 		db_printf("usage: show ifnet <struct ifnet *>\n");
5202 		return;
5203 	}
5204 
5205 	if_show_ifnet((struct ifnet *)addr);
5206 }
5207 
5208 DB_SHOW_ALL_COMMAND(ifnets, db_show_all_ifnets)
5209 {
5210 	struct ifnet *ifp;
5211 	u_short idx;
5212 
5213 	for (idx = 1; idx <= if_index; idx++) {
5214 		ifp = ifindex_table[idx].ife_ifnet;
5215 		if (ifp == NULL)
5216 			continue;
5217 		db_printf( "%20s ifp=%p\n", ifp->if_xname, ifp);
5218 		if (db_pager_quit)
5219 			break;
5220 	}
5221 }
5222 #endif	/* DDB */
5223