xref: /freebsd/sys/net/if.c (revision b9f654b163bce26de79705e77b872427c9f2afa1)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1980, 1986, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)if.c	8.5 (Berkeley) 1/9/95
32  * $FreeBSD$
33  */
34 
35 #include "opt_inet6.h"
36 #include "opt_inet.h"
37 
38 #include <sys/param.h>
39 #include <sys/conf.h>
40 #include <sys/eventhandler.h>
41 #include <sys/malloc.h>
42 #include <sys/domainset.h>
43 #include <sys/sbuf.h>
44 #include <sys/bus.h>
45 #include <sys/epoch.h>
46 #include <sys/mbuf.h>
47 #include <sys/systm.h>
48 #include <sys/priv.h>
49 #include <sys/proc.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/protosw.h>
53 #include <sys/kernel.h>
54 #include <sys/lock.h>
55 #include <sys/refcount.h>
56 #include <sys/module.h>
57 #include <sys/rwlock.h>
58 #include <sys/sockio.h>
59 #include <sys/syslog.h>
60 #include <sys/sysctl.h>
61 #include <sys/sysent.h>
62 #include <sys/taskqueue.h>
63 #include <sys/domain.h>
64 #include <sys/jail.h>
65 #include <sys/priv.h>
66 
67 #include <machine/stdarg.h>
68 #include <vm/uma.h>
69 
70 #include <net/bpf.h>
71 #include <net/ethernet.h>
72 #include <net/if.h>
73 #include <net/if_arp.h>
74 #include <net/if_clone.h>
75 #include <net/if_dl.h>
76 #include <net/if_types.h>
77 #include <net/if_var.h>
78 #include <net/if_media.h>
79 #include <net/if_vlan_var.h>
80 #include <net/radix.h>
81 #include <net/route.h>
82 #include <net/vnet.h>
83 
84 #if defined(INET) || defined(INET6)
85 #include <net/ethernet.h>
86 #include <netinet/in.h>
87 #include <netinet/in_var.h>
88 #include <netinet/ip.h>
89 #include <netinet/ip_carp.h>
90 #ifdef INET
91 #include <netinet/if_ether.h>
92 #include <netinet/netdump/netdump.h>
93 #endif /* INET */
94 #ifdef INET6
95 #include <netinet6/in6_var.h>
96 #include <netinet6/in6_ifattach.h>
97 #endif /* INET6 */
98 #endif /* INET || INET6 */
99 
100 #include <security/mac/mac_framework.h>
101 
102 /*
103  * Consumers of struct ifreq such as tcpdump assume no pad between ifr_name
104  * and ifr_ifru when it is used in SIOCGIFCONF.
105  */
106 _Static_assert(sizeof(((struct ifreq *)0)->ifr_name) ==
107     offsetof(struct ifreq, ifr_ifru), "gap between ifr_name and ifr_ifru");
108 
109 __read_mostly epoch_t net_epoch_preempt;
110 __read_mostly epoch_t net_epoch;
111 #ifdef COMPAT_FREEBSD32
112 #include <sys/mount.h>
113 #include <compat/freebsd32/freebsd32.h>
114 
115 struct ifreq_buffer32 {
116 	uint32_t	length;		/* (size_t) */
117 	uint32_t	buffer;		/* (void *) */
118 };
119 
120 /*
121  * Interface request structure used for socket
122  * ioctl's.  All interface ioctl's must have parameter
123  * definitions which begin with ifr_name.  The
124  * remainder may be interface specific.
125  */
126 struct ifreq32 {
127 	char	ifr_name[IFNAMSIZ];		/* if name, e.g. "en0" */
128 	union {
129 		struct sockaddr	ifru_addr;
130 		struct sockaddr	ifru_dstaddr;
131 		struct sockaddr	ifru_broadaddr;
132 		struct ifreq_buffer32 ifru_buffer;
133 		short		ifru_flags[2];
134 		short		ifru_index;
135 		int		ifru_jid;
136 		int		ifru_metric;
137 		int		ifru_mtu;
138 		int		ifru_phys;
139 		int		ifru_media;
140 		uint32_t	ifru_data;
141 		int		ifru_cap[2];
142 		u_int		ifru_fib;
143 		u_char		ifru_vlan_pcp;
144 	} ifr_ifru;
145 };
146 CTASSERT(sizeof(struct ifreq) == sizeof(struct ifreq32));
147 CTASSERT(__offsetof(struct ifreq, ifr_ifru) ==
148     __offsetof(struct ifreq32, ifr_ifru));
149 
150 struct ifgroupreq32 {
151 	char	ifgr_name[IFNAMSIZ];
152 	u_int	ifgr_len;
153 	union {
154 		char		ifgru_group[IFNAMSIZ];
155 		uint32_t	ifgru_groups;
156 	} ifgr_ifgru;
157 };
158 
159 struct ifmediareq32 {
160 	char		ifm_name[IFNAMSIZ];
161 	int		ifm_current;
162 	int		ifm_mask;
163 	int		ifm_status;
164 	int		ifm_active;
165 	int		ifm_count;
166 	uint32_t	ifm_ulist;	/* (int *) */
167 };
168 #define	SIOCGIFMEDIA32	_IOC_NEWTYPE(SIOCGIFMEDIA, struct ifmediareq32)
169 #define	SIOCGIFXMEDIA32	_IOC_NEWTYPE(SIOCGIFXMEDIA, struct ifmediareq32)
170 
171 #define	_CASE_IOC_IFGROUPREQ_32(cmd)				\
172     _IOC_NEWTYPE((cmd), struct ifgroupreq32): case
173 #else /* !COMPAT_FREEBSD32 */
174 #define _CASE_IOC_IFGROUPREQ_32(cmd)
175 #endif /* !COMPAT_FREEBSD32 */
176 
177 #define CASE_IOC_IFGROUPREQ(cmd)	\
178     _CASE_IOC_IFGROUPREQ_32(cmd)	\
179     (cmd)
180 
181 union ifreq_union {
182 	struct ifreq	ifr;
183 #ifdef COMPAT_FREEBSD32
184 	struct ifreq32	ifr32;
185 #endif
186 };
187 
188 union ifgroupreq_union {
189 	struct ifgroupreq ifgr;
190 #ifdef COMPAT_FREEBSD32
191 	struct ifgroupreq32 ifgr32;
192 #endif
193 };
194 
195 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
196 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
197 
198 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
199     &ifqmaxlen, 0, "max send queue size");
200 
201 /* Log link state change events */
202 static int log_link_state_change = 1;
203 
204 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
205 	&log_link_state_change, 0,
206 	"log interface link state change events");
207 
208 /* Log promiscuous mode change events */
209 static int log_promisc_mode_change = 1;
210 
211 SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
212 	&log_promisc_mode_change, 1,
213 	"log promiscuous mode change events");
214 
215 /* Interface description */
216 static unsigned int ifdescr_maxlen = 1024;
217 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
218 	&ifdescr_maxlen, 0,
219 	"administrative maximum length for interface description");
220 
221 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
222 
223 /* global sx for non-critical path ifdescr */
224 static struct sx ifdescr_sx;
225 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
226 
227 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
228 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
229 /* These are external hooks for CARP. */
230 void	(*carp_linkstate_p)(struct ifnet *ifp);
231 void	(*carp_demote_adj_p)(int, char *);
232 int	(*carp_master_p)(struct ifaddr *);
233 #if defined(INET) || defined(INET6)
234 int	(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
235 int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
236     const struct sockaddr *sa);
237 int	(*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
238 int	(*carp_attach_p)(struct ifaddr *, int);
239 void	(*carp_detach_p)(struct ifaddr *, bool);
240 #endif
241 #ifdef INET
242 int	(*carp_iamatch_p)(struct ifaddr *, uint8_t **);
243 #endif
244 #ifdef INET6
245 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
246 caddr_t	(*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
247     const struct in6_addr *taddr);
248 #endif
249 
250 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
251 
252 /*
253  * XXX: Style; these should be sorted alphabetically, and unprototyped
254  * static functions should be prototyped. Currently they are sorted by
255  * declaration order.
256  */
257 static void	if_attachdomain(void *);
258 static void	if_attachdomain1(struct ifnet *);
259 static int	ifconf(u_long, caddr_t);
260 static void	*if_grow(void);
261 static void	if_input_default(struct ifnet *, struct mbuf *);
262 static int	if_requestencap_default(struct ifnet *, struct if_encap_req *);
263 static void	if_route(struct ifnet *, int flag, int fam);
264 static int	if_setflag(struct ifnet *, int, int, int *, int);
265 static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
266 static void	if_unroute(struct ifnet *, int flag, int fam);
267 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
268 static void	do_link_state_change(void *, int);
269 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
270 static int	if_getgroupmembers(struct ifgroupreq *);
271 static void	if_delgroups(struct ifnet *);
272 static void	if_attach_internal(struct ifnet *, int, struct if_clone *);
273 static int	if_detach_internal(struct ifnet *, int, struct if_clone **);
274 #ifdef VIMAGE
275 static void	if_vmove(struct ifnet *, struct vnet *);
276 #endif
277 
278 #ifdef INET6
279 /*
280  * XXX: declare here to avoid to include many inet6 related files..
281  * should be more generalized?
282  */
283 extern void	nd6_setmtu(struct ifnet *);
284 #endif
285 
286 /* ipsec helper hooks */
287 VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
288 VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
289 
290 VNET_DEFINE(int, if_index);
291 int	ifqmaxlen = IFQ_MAXLEN;
292 VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
293 VNET_DEFINE(struct ifgrouphead, ifg_head);
294 
295 VNET_DEFINE_STATIC(int, if_indexlim) = 8;
296 
297 /* Table of ifnet by index. */
298 VNET_DEFINE(struct ifnet **, ifindex_table);
299 
300 #define	V_if_indexlim		VNET(if_indexlim)
301 #define	V_ifindex_table		VNET(ifindex_table)
302 
303 /*
304  * The global network interface list (V_ifnet) and related state (such as
305  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
306  * an rwlock.  Either may be acquired shared to stablize the list, but both
307  * must be acquired writable to modify the list.  This model allows us to
308  * both stablize the interface list during interrupt thread processing, but
309  * also to stablize it over long-running ioctls, without introducing priority
310  * inversions and deadlocks.
311  */
312 struct rwlock ifnet_rwlock;
313 RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
314 struct sx ifnet_sxlock;
315 SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
316 
317 /*
318  * The allocation of network interfaces is a rather non-atomic affair; we
319  * need to select an index before we are ready to expose the interface for
320  * use, so will use this pointer value to indicate reservation.
321  */
322 #define	IFNET_HOLD	(void *)(uintptr_t)(-1)
323 
324 static	if_com_alloc_t *if_com_alloc[256];
325 static	if_com_free_t *if_com_free[256];
326 
327 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
328 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
329 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
330 
331 struct ifnet *
332 ifnet_byindex_locked(u_short idx)
333 {
334 
335 	if (idx > V_if_index)
336 		return (NULL);
337 	if (V_ifindex_table[idx] == IFNET_HOLD)
338 		return (NULL);
339 	return (V_ifindex_table[idx]);
340 }
341 
342 struct ifnet *
343 ifnet_byindex(u_short idx)
344 {
345 	struct ifnet *ifp;
346 
347 	ifp = ifnet_byindex_locked(idx);
348 	return (ifp);
349 }
350 
351 struct ifnet *
352 ifnet_byindex_ref(u_short idx)
353 {
354 	struct epoch_tracker et;
355 	struct ifnet *ifp;
356 
357 	NET_EPOCH_ENTER(et);
358 	ifp = ifnet_byindex_locked(idx);
359 	if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
360 		NET_EPOCH_EXIT(et);
361 		return (NULL);
362 	}
363 	if_ref(ifp);
364 	NET_EPOCH_EXIT(et);
365 	return (ifp);
366 }
367 
368 /*
369  * Allocate an ifindex array entry; return 0 on success or an error on
370  * failure.
371  */
372 static u_short
373 ifindex_alloc(void **old)
374 {
375 	u_short idx;
376 
377 	IFNET_WLOCK_ASSERT();
378 	/*
379 	 * Try to find an empty slot below V_if_index.  If we fail, take the
380 	 * next slot.
381 	 */
382 	for (idx = 1; idx <= V_if_index; idx++) {
383 		if (V_ifindex_table[idx] == NULL)
384 			break;
385 	}
386 
387 	/* Catch if_index overflow. */
388 	if (idx >= V_if_indexlim) {
389 		*old = if_grow();
390 		return (USHRT_MAX);
391 	}
392 	if (idx > V_if_index)
393 		V_if_index = idx;
394 	return (idx);
395 }
396 
397 static void
398 ifindex_free_locked(u_short idx)
399 {
400 
401 	IFNET_WLOCK_ASSERT();
402 
403 	V_ifindex_table[idx] = NULL;
404 	while (V_if_index > 0 &&
405 	    V_ifindex_table[V_if_index] == NULL)
406 		V_if_index--;
407 }
408 
409 static void
410 ifindex_free(u_short idx)
411 {
412 
413 	IFNET_WLOCK();
414 	ifindex_free_locked(idx);
415 	IFNET_WUNLOCK();
416 }
417 
418 static void
419 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
420 {
421 
422 	V_ifindex_table[idx] = ifp;
423 }
424 
425 struct ifaddr *
426 ifaddr_byindex(u_short idx)
427 {
428 	struct epoch_tracker et;
429 	struct ifnet *ifp;
430 	struct ifaddr *ifa = NULL;
431 
432 	NET_EPOCH_ENTER(et);
433 	ifp = ifnet_byindex_locked(idx);
434 	if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
435 		ifa_ref(ifa);
436 	NET_EPOCH_EXIT(et);
437 	return (ifa);
438 }
439 
440 /*
441  * Network interface utility routines.
442  *
443  * Routines with ifa_ifwith* names take sockaddr *'s as
444  * parameters.
445  */
446 
447 static void
448 vnet_if_init(const void *unused __unused)
449 {
450 	void *old;
451 
452 	CK_STAILQ_INIT(&V_ifnet);
453 	CK_STAILQ_INIT(&V_ifg_head);
454 	IFNET_WLOCK();
455 	old = if_grow();				/* create initial table */
456 	IFNET_WUNLOCK();
457 	epoch_wait_preempt(net_epoch_preempt);
458 	free(old, M_IFNET);
459 	vnet_if_clone_init();
460 }
461 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
462     NULL);
463 
464 #ifdef VIMAGE
465 static void
466 vnet_if_uninit(const void *unused __unused)
467 {
468 
469 	VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
470 	    "not empty", __func__, __LINE__, &V_ifnet));
471 	VNET_ASSERT(CK_STAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
472 	    "not empty", __func__, __LINE__, &V_ifg_head));
473 
474 	free((caddr_t)V_ifindex_table, M_IFNET);
475 }
476 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
477     vnet_if_uninit, NULL);
478 
479 static void
480 vnet_if_return(const void *unused __unused)
481 {
482 	struct ifnet *ifp, *nifp;
483 
484 	/* Return all inherited interfaces to their parent vnets. */
485 	CK_STAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
486 		if (ifp->if_home_vnet != ifp->if_vnet)
487 			if_vmove(ifp, ifp->if_home_vnet);
488 	}
489 }
490 VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
491     vnet_if_return, NULL);
492 #endif
493 
494 
495 static void *
496 if_grow(void)
497 {
498 	int oldlim;
499 	u_int n;
500 	struct ifnet **e;
501 	void *old;
502 
503 	old = NULL;
504 	IFNET_WLOCK_ASSERT();
505 	oldlim = V_if_indexlim;
506 	IFNET_WUNLOCK();
507 	n = (oldlim << 1) * sizeof(*e);
508 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
509 	IFNET_WLOCK();
510 	if (V_if_indexlim != oldlim) {
511 		free(e, M_IFNET);
512 		return (NULL);
513 	}
514 	if (V_ifindex_table != NULL) {
515 		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
516 		old = V_ifindex_table;
517 	}
518 	V_if_indexlim <<= 1;
519 	V_ifindex_table = e;
520 	return (old);
521 }
522 
523 /*
524  * Allocate a struct ifnet and an index for an interface.  A layer 2
525  * common structure will also be allocated if an allocation routine is
526  * registered for the passed type.
527  */
528 struct ifnet *
529 if_alloc_domain(u_char type, int numa_domain)
530 {
531 	struct ifnet *ifp;
532 	u_short idx;
533 	void *old;
534 
535 	KASSERT(numa_domain <= IF_NODOM, ("numa_domain too large"));
536 	if (numa_domain == IF_NODOM)
537 		ifp = malloc(sizeof(struct ifnet), M_IFNET,
538 		    M_WAITOK | M_ZERO);
539 	else
540 		ifp = malloc_domainset(sizeof(struct ifnet), M_IFNET,
541 		    DOMAINSET_PREF(numa_domain), M_WAITOK | M_ZERO);
542  restart:
543 	IFNET_WLOCK();
544 	idx = ifindex_alloc(&old);
545 	if (__predict_false(idx == USHRT_MAX)) {
546 		IFNET_WUNLOCK();
547 		epoch_wait_preempt(net_epoch_preempt);
548 		free(old, M_IFNET);
549 		goto restart;
550 	}
551 	ifnet_setbyindex(idx, IFNET_HOLD);
552 	IFNET_WUNLOCK();
553 	ifp->if_index = idx;
554 	ifp->if_type = type;
555 	ifp->if_alloctype = type;
556 	ifp->if_numa_domain = numa_domain;
557 #ifdef VIMAGE
558 	ifp->if_vnet = curvnet;
559 #endif
560 	if (if_com_alloc[type] != NULL) {
561 		ifp->if_l2com = if_com_alloc[type](type, ifp);
562 		if (ifp->if_l2com == NULL) {
563 			free(ifp, M_IFNET);
564 			ifindex_free(idx);
565 			return (NULL);
566 		}
567 	}
568 
569 	IF_ADDR_LOCK_INIT(ifp);
570 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
571 	ifp->if_afdata_initialized = 0;
572 	IF_AFDATA_LOCK_INIT(ifp);
573 	CK_STAILQ_INIT(&ifp->if_addrhead);
574 	CK_STAILQ_INIT(&ifp->if_multiaddrs);
575 	CK_STAILQ_INIT(&ifp->if_groups);
576 #ifdef MAC
577 	mac_ifnet_init(ifp);
578 #endif
579 	ifq_init(&ifp->if_snd, ifp);
580 
581 	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
582 	for (int i = 0; i < IFCOUNTERS; i++)
583 		ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
584 	ifp->if_get_counter = if_get_counter_default;
585 	ifp->if_pcp = IFNET_PCP_NONE;
586 	ifnet_setbyindex(ifp->if_index, ifp);
587 	return (ifp);
588 }
589 
590 struct ifnet *
591 if_alloc_dev(u_char type, device_t dev)
592 {
593 	int numa_domain;
594 
595 	if (dev == NULL || bus_get_domain(dev, &numa_domain) != 0)
596 		return (if_alloc_domain(type, IF_NODOM));
597 	return (if_alloc_domain(type, numa_domain));
598 }
599 
600 struct ifnet *
601 if_alloc(u_char type)
602 {
603 
604 	return (if_alloc_domain(type, IF_NODOM));
605 }
606 /*
607  * Do the actual work of freeing a struct ifnet, and layer 2 common
608  * structure.  This call is made when the last reference to an
609  * interface is released.
610  */
611 static void
612 if_free_internal(struct ifnet *ifp)
613 {
614 
615 	KASSERT((ifp->if_flags & IFF_DYING),
616 	    ("if_free_internal: interface not dying"));
617 
618 	if (if_com_free[ifp->if_alloctype] != NULL)
619 		if_com_free[ifp->if_alloctype](ifp->if_l2com,
620 		    ifp->if_alloctype);
621 
622 #ifdef MAC
623 	mac_ifnet_destroy(ifp);
624 #endif /* MAC */
625 	IF_AFDATA_DESTROY(ifp);
626 	IF_ADDR_LOCK_DESTROY(ifp);
627 	ifq_delete(&ifp->if_snd);
628 
629 	for (int i = 0; i < IFCOUNTERS; i++)
630 		counter_u64_free(ifp->if_counters[i]);
631 
632 	free(ifp->if_description, M_IFDESCR);
633 	free(ifp->if_hw_addr, M_IFADDR);
634 	if (ifp->if_numa_domain == IF_NODOM)
635 		free(ifp, M_IFNET);
636 	else
637 		free_domain(ifp, M_IFNET);
638 }
639 
640 static void
641 if_destroy(epoch_context_t ctx)
642 {
643 	struct ifnet *ifp;
644 
645 	ifp = __containerof(ctx, struct ifnet, if_epoch_ctx);
646 	if_free_internal(ifp);
647 }
648 
649 /*
650  * Deregister an interface and free the associated storage.
651  */
652 void
653 if_free(struct ifnet *ifp)
654 {
655 
656 	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
657 
658 	CURVNET_SET_QUIET(ifp->if_vnet);
659 	IFNET_WLOCK();
660 	KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
661 	    ("%s: freeing unallocated ifnet", ifp->if_xname));
662 
663 	ifindex_free_locked(ifp->if_index);
664 	IFNET_WUNLOCK();
665 
666 	if (refcount_release(&ifp->if_refcount))
667 		epoch_call(net_epoch_preempt, &ifp->if_epoch_ctx, if_destroy);
668 	CURVNET_RESTORE();
669 }
670 
671 /*
672  * Interfaces to keep an ifnet type-stable despite the possibility of the
673  * driver calling if_free().  If there are additional references, we defer
674  * freeing the underlying data structure.
675  */
676 void
677 if_ref(struct ifnet *ifp)
678 {
679 
680 	/* We don't assert the ifnet list lock here, but arguably should. */
681 	refcount_acquire(&ifp->if_refcount);
682 }
683 
684 void
685 if_rele(struct ifnet *ifp)
686 {
687 
688 	if (!refcount_release(&ifp->if_refcount))
689 		return;
690 	epoch_call(net_epoch_preempt, &ifp->if_epoch_ctx, if_destroy);
691 }
692 
693 void
694 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
695 {
696 
697 	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
698 
699 	if (ifq->ifq_maxlen == 0)
700 		ifq->ifq_maxlen = ifqmaxlen;
701 
702 	ifq->altq_type = 0;
703 	ifq->altq_disc = NULL;
704 	ifq->altq_flags &= ALTQF_CANTCHANGE;
705 	ifq->altq_tbr  = NULL;
706 	ifq->altq_ifp  = ifp;
707 }
708 
709 void
710 ifq_delete(struct ifaltq *ifq)
711 {
712 	mtx_destroy(&ifq->ifq_mtx);
713 }
714 
715 /*
716  * Perform generic interface initialization tasks and attach the interface
717  * to the list of "active" interfaces.  If vmove flag is set on entry
718  * to if_attach_internal(), perform only a limited subset of initialization
719  * tasks, given that we are moving from one vnet to another an ifnet which
720  * has already been fully initialized.
721  *
722  * Note that if_detach_internal() removes group membership unconditionally
723  * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
724  * Thus, when if_vmove() is applied to a cloned interface, group membership
725  * is lost while a cloned one always joins a group whose name is
726  * ifc->ifc_name.  To recover this after if_detach_internal() and
727  * if_attach_internal(), the cloner should be specified to
728  * if_attach_internal() via ifc.  If it is non-NULL, if_attach_internal()
729  * attempts to join a group whose name is ifc->ifc_name.
730  *
731  * XXX:
732  *  - The decision to return void and thus require this function to
733  *    succeed is questionable.
734  *  - We should probably do more sanity checking.  For instance we don't
735  *    do anything to insure if_xname is unique or non-empty.
736  */
737 void
738 if_attach(struct ifnet *ifp)
739 {
740 
741 	if_attach_internal(ifp, 0, NULL);
742 }
743 
744 /*
745  * Compute the least common TSO limit.
746  */
747 void
748 if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
749 {
750 	/*
751 	 * 1) If there is no limit currently, take the limit from
752 	 * the network adapter.
753 	 *
754 	 * 2) If the network adapter has a limit below the current
755 	 * limit, apply it.
756 	 */
757 	if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
758 	    ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
759 		pmax->tsomaxbytes = ifp->if_hw_tsomax;
760 	}
761 	if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
762 	    ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
763 		pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
764 	}
765 	if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
766 	    ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
767 		pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
768 	}
769 }
770 
771 /*
772  * Update TSO limit of a network adapter.
773  *
774  * Returns zero if no change. Else non-zero.
775  */
776 int
777 if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
778 {
779 	int retval = 0;
780 	if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
781 		ifp->if_hw_tsomax = pmax->tsomaxbytes;
782 		retval++;
783 	}
784 	if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
785 		ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
786 		retval++;
787 	}
788 	if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
789 		ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
790 		retval++;
791 	}
792 	return (retval);
793 }
794 
795 static void
796 if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
797 {
798 	unsigned socksize, ifasize;
799 	int namelen, masklen;
800 	struct sockaddr_dl *sdl;
801 	struct ifaddr *ifa;
802 
803 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
804 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
805 		    ifp->if_xname);
806 
807 #ifdef VIMAGE
808 	ifp->if_vnet = curvnet;
809 	if (ifp->if_home_vnet == NULL)
810 		ifp->if_home_vnet = curvnet;
811 #endif
812 
813 	if_addgroup(ifp, IFG_ALL);
814 
815 	/* Restore group membership for cloned interfaces. */
816 	if (vmove && ifc != NULL)
817 		if_clone_addgroup(ifp, ifc);
818 
819 	getmicrotime(&ifp->if_lastchange);
820 	ifp->if_epoch = time_uptime;
821 
822 	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
823 	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
824 	    ("transmit and qflush must both either be set or both be NULL"));
825 	if (ifp->if_transmit == NULL) {
826 		ifp->if_transmit = if_transmit;
827 		ifp->if_qflush = if_qflush;
828 	}
829 	if (ifp->if_input == NULL)
830 		ifp->if_input = if_input_default;
831 
832 	if (ifp->if_requestencap == NULL)
833 		ifp->if_requestencap = if_requestencap_default;
834 
835 	if (!vmove) {
836 #ifdef MAC
837 		mac_ifnet_create(ifp);
838 #endif
839 
840 		/*
841 		 * Create a Link Level name for this device.
842 		 */
843 		namelen = strlen(ifp->if_xname);
844 		/*
845 		 * Always save enough space for any possiable name so we
846 		 * can do a rename in place later.
847 		 */
848 		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
849 		socksize = masklen + ifp->if_addrlen;
850 		if (socksize < sizeof(*sdl))
851 			socksize = sizeof(*sdl);
852 		socksize = roundup2(socksize, sizeof(long));
853 		ifasize = sizeof(*ifa) + 2 * socksize;
854 		ifa = ifa_alloc(ifasize, M_WAITOK);
855 		sdl = (struct sockaddr_dl *)(ifa + 1);
856 		sdl->sdl_len = socksize;
857 		sdl->sdl_family = AF_LINK;
858 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
859 		sdl->sdl_nlen = namelen;
860 		sdl->sdl_index = ifp->if_index;
861 		sdl->sdl_type = ifp->if_type;
862 		ifp->if_addr = ifa;
863 		ifa->ifa_ifp = ifp;
864 		ifa->ifa_addr = (struct sockaddr *)sdl;
865 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
866 		ifa->ifa_netmask = (struct sockaddr *)sdl;
867 		sdl->sdl_len = masklen;
868 		while (namelen != 0)
869 			sdl->sdl_data[--namelen] = 0xff;
870 		CK_STAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
871 		/* Reliably crash if used uninitialized. */
872 		ifp->if_broadcastaddr = NULL;
873 
874 		if (ifp->if_type == IFT_ETHER) {
875 			ifp->if_hw_addr = malloc(ifp->if_addrlen, M_IFADDR,
876 			    M_WAITOK | M_ZERO);
877 		}
878 
879 #if defined(INET) || defined(INET6)
880 		/* Use defaults for TSO, if nothing is set */
881 		if (ifp->if_hw_tsomax == 0 &&
882 		    ifp->if_hw_tsomaxsegcount == 0 &&
883 		    ifp->if_hw_tsomaxsegsize == 0) {
884 			/*
885 			 * The TSO defaults needs to be such that an
886 			 * NFS mbuf list of 35 mbufs totalling just
887 			 * below 64K works and that a chain of mbufs
888 			 * can be defragged into at most 32 segments:
889 			 */
890 			ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
891 			    (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
892 			ifp->if_hw_tsomaxsegcount = 35;
893 			ifp->if_hw_tsomaxsegsize = 2048;	/* 2K */
894 
895 			/* XXX some drivers set IFCAP_TSO after ethernet attach */
896 			if (ifp->if_capabilities & IFCAP_TSO) {
897 				if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
898 				    ifp->if_hw_tsomax,
899 				    ifp->if_hw_tsomaxsegcount,
900 				    ifp->if_hw_tsomaxsegsize);
901 			}
902 		}
903 #endif
904 	}
905 #ifdef VIMAGE
906 	else {
907 		/*
908 		 * Update the interface index in the link layer address
909 		 * of the interface.
910 		 */
911 		for (ifa = ifp->if_addr; ifa != NULL;
912 		    ifa = CK_STAILQ_NEXT(ifa, ifa_link)) {
913 			if (ifa->ifa_addr->sa_family == AF_LINK) {
914 				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
915 				sdl->sdl_index = ifp->if_index;
916 			}
917 		}
918 	}
919 #endif
920 
921 	IFNET_WLOCK();
922 	CK_STAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
923 #ifdef VIMAGE
924 	curvnet->vnet_ifcnt++;
925 #endif
926 	IFNET_WUNLOCK();
927 
928 	if (domain_init_status >= 2)
929 		if_attachdomain1(ifp);
930 
931 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
932 	if (IS_DEFAULT_VNET(curvnet))
933 		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
934 
935 	/* Announce the interface. */
936 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
937 }
938 
939 static void
940 if_epochalloc(void *dummy __unused)
941 {
942 
943 	net_epoch_preempt = epoch_alloc(EPOCH_PREEMPT);
944 	net_epoch = epoch_alloc(0);
945 }
946 SYSINIT(ifepochalloc, SI_SUB_TASKQ + 1, SI_ORDER_ANY,
947     if_epochalloc, NULL);
948 
949 static void
950 if_attachdomain(void *dummy)
951 {
952 	struct ifnet *ifp;
953 
954 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
955 		if_attachdomain1(ifp);
956 }
957 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
958     if_attachdomain, NULL);
959 
960 static void
961 if_attachdomain1(struct ifnet *ifp)
962 {
963 	struct domain *dp;
964 
965 	/*
966 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
967 	 * cannot lock ifp->if_afdata initialization, entirely.
968 	 */
969 	IF_AFDATA_LOCK(ifp);
970 	if (ifp->if_afdata_initialized >= domain_init_status) {
971 		IF_AFDATA_UNLOCK(ifp);
972 		log(LOG_WARNING, "%s called more than once on %s\n",
973 		    __func__, ifp->if_xname);
974 		return;
975 	}
976 	ifp->if_afdata_initialized = domain_init_status;
977 	IF_AFDATA_UNLOCK(ifp);
978 
979 	/* address family dependent data region */
980 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
981 	for (dp = domains; dp; dp = dp->dom_next) {
982 		if (dp->dom_ifattach)
983 			ifp->if_afdata[dp->dom_family] =
984 			    (*dp->dom_ifattach)(ifp);
985 	}
986 }
987 
988 /*
989  * Remove any unicast or broadcast network addresses from an interface.
990  */
991 void
992 if_purgeaddrs(struct ifnet *ifp)
993 {
994 	struct ifaddr *ifa;
995 
996 	while (1) {
997 		struct epoch_tracker et;
998 
999 		NET_EPOCH_ENTER(et);
1000 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1001 			if (ifa->ifa_addr->sa_family != AF_LINK)
1002 				break;
1003 		}
1004 		NET_EPOCH_EXIT(et);
1005 
1006 		if (ifa == NULL)
1007 			break;
1008 #ifdef INET
1009 		/* XXX: Ugly!! ad hoc just for INET */
1010 		if (ifa->ifa_addr->sa_family == AF_INET) {
1011 			struct ifaliasreq ifr;
1012 
1013 			bzero(&ifr, sizeof(ifr));
1014 			ifr.ifra_addr = *ifa->ifa_addr;
1015 			if (ifa->ifa_dstaddr)
1016 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
1017 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
1018 			    NULL) == 0)
1019 				continue;
1020 		}
1021 #endif /* INET */
1022 #ifdef INET6
1023 		if (ifa->ifa_addr->sa_family == AF_INET6) {
1024 			in6_purgeaddr(ifa);
1025 			/* ifp_addrhead is already updated */
1026 			continue;
1027 		}
1028 #endif /* INET6 */
1029 		IF_ADDR_WLOCK(ifp);
1030 		CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1031 		IF_ADDR_WUNLOCK(ifp);
1032 		ifa_free(ifa);
1033 	}
1034 }
1035 
1036 /*
1037  * Remove any multicast network addresses from an interface when an ifnet
1038  * is going away.
1039  */
1040 static void
1041 if_purgemaddrs(struct ifnet *ifp)
1042 {
1043 	struct ifmultiaddr *ifma;
1044 
1045 	IF_ADDR_WLOCK(ifp);
1046 	while (!CK_STAILQ_EMPTY(&ifp->if_multiaddrs)) {
1047 		ifma = CK_STAILQ_FIRST(&ifp->if_multiaddrs);
1048 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
1049 		if_delmulti_locked(ifp, ifma, 1);
1050 	}
1051 	IF_ADDR_WUNLOCK(ifp);
1052 }
1053 
1054 /*
1055  * Detach an interface, removing it from the list of "active" interfaces.
1056  * If vmove flag is set on entry to if_detach_internal(), perform only a
1057  * limited subset of cleanup tasks, given that we are moving an ifnet from
1058  * one vnet to another, where it must be fully operational.
1059  *
1060  * XXXRW: There are some significant questions about event ordering, and
1061  * how to prevent things from starting to use the interface during detach.
1062  */
1063 void
1064 if_detach(struct ifnet *ifp)
1065 {
1066 
1067 	CURVNET_SET_QUIET(ifp->if_vnet);
1068 	if_detach_internal(ifp, 0, NULL);
1069 	CURVNET_RESTORE();
1070 }
1071 
1072 /*
1073  * The vmove flag, if set, indicates that we are called from a callpath
1074  * that is moving an interface to a different vnet instance.
1075  *
1076  * The shutdown flag, if set, indicates that we are called in the
1077  * process of shutting down a vnet instance.  Currently only the
1078  * vnet_if_return SYSUNINIT function sets it.  Note: we can be called
1079  * on a vnet instance shutdown without this flag being set, e.g., when
1080  * the cloned interfaces are destoyed as first thing of teardown.
1081  */
1082 static int
1083 if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
1084 {
1085 	struct ifaddr *ifa;
1086 	int i;
1087 	struct domain *dp;
1088  	struct ifnet *iter;
1089  	int found = 0;
1090 #ifdef VIMAGE
1091 	int shutdown;
1092 
1093 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1094 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1095 #endif
1096 	IFNET_WLOCK();
1097 	CK_STAILQ_FOREACH(iter, &V_ifnet, if_link)
1098 		if (iter == ifp) {
1099 			CK_STAILQ_REMOVE(&V_ifnet, ifp, ifnet, if_link);
1100 			if (!vmove)
1101 				ifp->if_flags |= IFF_DYING;
1102 			found = 1;
1103 			break;
1104 		}
1105 	IFNET_WUNLOCK();
1106 	if (!found) {
1107 		/*
1108 		 * While we would want to panic here, we cannot
1109 		 * guarantee that the interface is indeed still on
1110 		 * the list given we don't hold locks all the way.
1111 		 */
1112 		return (ENOENT);
1113 #if 0
1114 		if (vmove)
1115 			panic("%s: ifp=%p not on the ifnet tailq %p",
1116 			    __func__, ifp, &V_ifnet);
1117 		else
1118 			return; /* XXX this should panic as well? */
1119 #endif
1120 	}
1121 
1122 	/*
1123 	 * At this point we know the interface still was on the ifnet list
1124 	 * and we removed it so we are in a stable state.
1125 	 */
1126 #ifdef VIMAGE
1127 	curvnet->vnet_ifcnt--;
1128 #endif
1129 	epoch_wait_preempt(net_epoch_preempt);
1130 	/*
1131 	 * In any case (destroy or vmove) detach us from the groups
1132 	 * and remove/wait for pending events on the taskq.
1133 	 * XXX-BZ in theory an interface could still enqueue a taskq change?
1134 	 */
1135 	if_delgroups(ifp);
1136 
1137 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
1138 
1139 	/*
1140 	 * Check if this is a cloned interface or not. Must do even if
1141 	 * shutting down as a if_vmove_reclaim() would move the ifp and
1142 	 * the if_clone_addgroup() will have a corrupted string overwise
1143 	 * from a gibberish pointer.
1144 	 */
1145 	if (vmove && ifcp != NULL)
1146 		*ifcp = if_clone_findifc(ifp);
1147 
1148 	if_down(ifp);
1149 
1150 #ifdef VIMAGE
1151 	/*
1152 	 * On VNET shutdown abort here as the stack teardown will do all
1153 	 * the work top-down for us.
1154 	 */
1155 	if (shutdown) {
1156 		/* Give interface users the chance to clean up. */
1157 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1158 
1159 		/*
1160 		 * In case of a vmove we are done here without error.
1161 		 * If we would signal an error it would lead to the same
1162 		 * abort as if we did not find the ifnet anymore.
1163 		 * if_detach() calls us in void context and does not care
1164 		 * about an early abort notification, so life is splendid :)
1165 		 */
1166 		goto finish_vnet_shutdown;
1167 	}
1168 #endif
1169 
1170 	/*
1171 	 * At this point we are not tearing down a VNET and are either
1172 	 * going to destroy or vmove the interface and have to cleanup
1173 	 * accordingly.
1174 	 */
1175 
1176 	/*
1177 	 * Remove routes and flush queues.
1178 	 */
1179 #ifdef ALTQ
1180 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
1181 		altq_disable(&ifp->if_snd);
1182 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
1183 		altq_detach(&ifp->if_snd);
1184 #endif
1185 
1186 	if_purgeaddrs(ifp);
1187 
1188 #ifdef INET
1189 	in_ifdetach(ifp);
1190 #endif
1191 
1192 #ifdef INET6
1193 	/*
1194 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
1195 	 * before removing routing entries below, since IPv6 interface direct
1196 	 * routes are expected to be removed by the IPv6-specific kernel API.
1197 	 * Otherwise, the kernel will detect some inconsistency and bark it.
1198 	 */
1199 	in6_ifdetach(ifp);
1200 #endif
1201 	if_purgemaddrs(ifp);
1202 
1203 	/* Announce that the interface is gone. */
1204 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1205 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1206 	if (IS_DEFAULT_VNET(curvnet))
1207 		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
1208 
1209 	if (!vmove) {
1210 		/*
1211 		 * Prevent further calls into the device driver via ifnet.
1212 		 */
1213 		if_dead(ifp);
1214 
1215 		/*
1216 		 * Clean up all addresses.
1217 		 */
1218 		IF_ADDR_WLOCK(ifp);
1219 		if (!CK_STAILQ_EMPTY(&ifp->if_addrhead)) {
1220 			ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
1221 			CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
1222 			IF_ADDR_WUNLOCK(ifp);
1223 			ifa_free(ifa);
1224 		} else
1225 			IF_ADDR_WUNLOCK(ifp);
1226 	}
1227 
1228 	rt_flushifroutes(ifp);
1229 
1230 #ifdef VIMAGE
1231 finish_vnet_shutdown:
1232 #endif
1233 	/*
1234 	 * We cannot hold the lock over dom_ifdetach calls as they might
1235 	 * sleep, for example trying to drain a callout, thus open up the
1236 	 * theoretical race with re-attaching.
1237 	 */
1238 	IF_AFDATA_LOCK(ifp);
1239 	i = ifp->if_afdata_initialized;
1240 	ifp->if_afdata_initialized = 0;
1241 	IF_AFDATA_UNLOCK(ifp);
1242 	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
1243 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
1244 			(*dp->dom_ifdetach)(ifp,
1245 			    ifp->if_afdata[dp->dom_family]);
1246 			ifp->if_afdata[dp->dom_family] = NULL;
1247 		}
1248 	}
1249 
1250 	return (0);
1251 }
1252 
1253 #ifdef VIMAGE
1254 /*
1255  * if_vmove() performs a limited version of if_detach() in current
1256  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
1257  * An attempt is made to shrink if_index in current vnet, find an
1258  * unused if_index in target vnet and calls if_grow() if necessary,
1259  * and finally find an unused if_xname for the target vnet.
1260  */
1261 static void
1262 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
1263 {
1264 	struct if_clone *ifc;
1265 	u_int bif_dlt, bif_hdrlen;
1266 	void *old;
1267 	int rc;
1268 
1269  	/*
1270 	 * if_detach_internal() will call the eventhandler to notify
1271 	 * interface departure.  That will detach if_bpf.  We need to
1272 	 * safe the dlt and hdrlen so we can re-attach it later.
1273 	 */
1274 	bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
1275 
1276 	/*
1277 	 * Detach from current vnet, but preserve LLADDR info, do not
1278 	 * mark as dead etc. so that the ifnet can be reattached later.
1279 	 * If we cannot find it, we lost the race to someone else.
1280 	 */
1281 	rc = if_detach_internal(ifp, 1, &ifc);
1282 	if (rc != 0)
1283 		return;
1284 
1285 	/*
1286 	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
1287 	 * the if_index for that vnet if possible.
1288 	 *
1289 	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
1290 	 * or we'd lock on one vnet and unlock on another.
1291 	 */
1292 	IFNET_WLOCK();
1293 	ifindex_free_locked(ifp->if_index);
1294 	IFNET_WUNLOCK();
1295 
1296 	/*
1297 	 * Perform interface-specific reassignment tasks, if provided by
1298 	 * the driver.
1299 	 */
1300 	if (ifp->if_reassign != NULL)
1301 		ifp->if_reassign(ifp, new_vnet, NULL);
1302 
1303 	/*
1304 	 * Switch to the context of the target vnet.
1305 	 */
1306 	CURVNET_SET_QUIET(new_vnet);
1307  restart:
1308 	IFNET_WLOCK();
1309 	ifp->if_index = ifindex_alloc(&old);
1310 	if (__predict_false(ifp->if_index == USHRT_MAX)) {
1311 		IFNET_WUNLOCK();
1312 		epoch_wait_preempt(net_epoch_preempt);
1313 		free(old, M_IFNET);
1314 		goto restart;
1315 	}
1316 	ifnet_setbyindex(ifp->if_index, ifp);
1317 	IFNET_WUNLOCK();
1318 
1319 	if_attach_internal(ifp, 1, ifc);
1320 
1321 	if (ifp->if_bpf == NULL)
1322 		bpfattach(ifp, bif_dlt, bif_hdrlen);
1323 
1324 	CURVNET_RESTORE();
1325 }
1326 
1327 /*
1328  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1329  */
1330 static int
1331 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1332 {
1333 	struct prison *pr;
1334 	struct ifnet *difp;
1335 	int shutdown;
1336 
1337 	/* Try to find the prison within our visibility. */
1338 	sx_slock(&allprison_lock);
1339 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1340 	sx_sunlock(&allprison_lock);
1341 	if (pr == NULL)
1342 		return (ENXIO);
1343 	prison_hold_locked(pr);
1344 	mtx_unlock(&pr->pr_mtx);
1345 
1346 	/* Do not try to move the iface from and to the same prison. */
1347 	if (pr->pr_vnet == ifp->if_vnet) {
1348 		prison_free(pr);
1349 		return (EEXIST);
1350 	}
1351 
1352 	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1353 	/* XXX Lock interfaces to avoid races. */
1354 	CURVNET_SET_QUIET(pr->pr_vnet);
1355 	difp = ifunit(ifname);
1356 	if (difp != NULL) {
1357 		CURVNET_RESTORE();
1358 		prison_free(pr);
1359 		return (EEXIST);
1360 	}
1361 
1362 	/* Make sure the VNET is stable. */
1363 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1364 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1365 	if (shutdown) {
1366 		CURVNET_RESTORE();
1367 		prison_free(pr);
1368 		return (EBUSY);
1369 	}
1370 	CURVNET_RESTORE();
1371 
1372 	/* Move the interface into the child jail/vnet. */
1373 	if_vmove(ifp, pr->pr_vnet);
1374 
1375 	/* Report the new if_xname back to the userland. */
1376 	sprintf(ifname, "%s", ifp->if_xname);
1377 
1378 	prison_free(pr);
1379 	return (0);
1380 }
1381 
1382 static int
1383 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1384 {
1385 	struct prison *pr;
1386 	struct vnet *vnet_dst;
1387 	struct ifnet *ifp;
1388  	int shutdown;
1389 
1390 	/* Try to find the prison within our visibility. */
1391 	sx_slock(&allprison_lock);
1392 	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1393 	sx_sunlock(&allprison_lock);
1394 	if (pr == NULL)
1395 		return (ENXIO);
1396 	prison_hold_locked(pr);
1397 	mtx_unlock(&pr->pr_mtx);
1398 
1399 	/* Make sure the named iface exists in the source prison/vnet. */
1400 	CURVNET_SET(pr->pr_vnet);
1401 	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1402 	if (ifp == NULL) {
1403 		CURVNET_RESTORE();
1404 		prison_free(pr);
1405 		return (ENXIO);
1406 	}
1407 
1408 	/* Do not try to move the iface from and to the same prison. */
1409 	vnet_dst = TD_TO_VNET(td);
1410 	if (vnet_dst == ifp->if_vnet) {
1411 		CURVNET_RESTORE();
1412 		prison_free(pr);
1413 		return (EEXIST);
1414 	}
1415 
1416 	/* Make sure the VNET is stable. */
1417 	shutdown = (ifp->if_vnet->vnet_state > SI_SUB_VNET &&
1418 		 ifp->if_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
1419 	if (shutdown) {
1420 		CURVNET_RESTORE();
1421 		prison_free(pr);
1422 		return (EBUSY);
1423 	}
1424 
1425 	/* Get interface back from child jail/vnet. */
1426 	if_vmove(ifp, vnet_dst);
1427 	CURVNET_RESTORE();
1428 
1429 	/* Report the new if_xname back to the userland. */
1430 	sprintf(ifname, "%s", ifp->if_xname);
1431 
1432 	prison_free(pr);
1433 	return (0);
1434 }
1435 #endif /* VIMAGE */
1436 
1437 /*
1438  * Add a group to an interface
1439  */
1440 int
1441 if_addgroup(struct ifnet *ifp, const char *groupname)
1442 {
1443 	struct ifg_list		*ifgl;
1444 	struct ifg_group	*ifg = NULL;
1445 	struct ifg_member	*ifgm;
1446 	int 			 new = 0;
1447 
1448 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1449 	    groupname[strlen(groupname) - 1] <= '9')
1450 		return (EINVAL);
1451 
1452 	IFNET_WLOCK();
1453 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1454 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1455 			IFNET_WUNLOCK();
1456 			return (EEXIST);
1457 		}
1458 
1459 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1460 	    M_NOWAIT)) == NULL) {
1461 	    	IFNET_WUNLOCK();
1462 		return (ENOMEM);
1463 	}
1464 
1465 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1466 	    M_TEMP, M_NOWAIT)) == NULL) {
1467 		free(ifgl, M_TEMP);
1468 		IFNET_WUNLOCK();
1469 		return (ENOMEM);
1470 	}
1471 
1472 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1473 		if (!strcmp(ifg->ifg_group, groupname))
1474 			break;
1475 
1476 	if (ifg == NULL) {
1477 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1478 		    M_TEMP, M_NOWAIT)) == NULL) {
1479 			free(ifgl, M_TEMP);
1480 			free(ifgm, M_TEMP);
1481 			IFNET_WUNLOCK();
1482 			return (ENOMEM);
1483 		}
1484 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1485 		ifg->ifg_refcnt = 0;
1486 		CK_STAILQ_INIT(&ifg->ifg_members);
1487 		CK_STAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1488 		new = 1;
1489 	}
1490 
1491 	ifg->ifg_refcnt++;
1492 	ifgl->ifgl_group = ifg;
1493 	ifgm->ifgm_ifp = ifp;
1494 
1495 	IF_ADDR_WLOCK(ifp);
1496 	CK_STAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1497 	CK_STAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1498 	IF_ADDR_WUNLOCK(ifp);
1499 
1500 	IFNET_WUNLOCK();
1501 
1502 	if (new)
1503 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1504 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1505 
1506 	return (0);
1507 }
1508 
1509 /*
1510  * Remove a group from an interface
1511  */
1512 int
1513 if_delgroup(struct ifnet *ifp, const char *groupname)
1514 {
1515 	struct ifg_list		*ifgl;
1516 	struct ifg_member	*ifgm;
1517 	int freeifgl;
1518 
1519 	IFNET_WLOCK();
1520 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1521 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1522 			break;
1523 	if (ifgl == NULL) {
1524 		IFNET_WUNLOCK();
1525 		return (ENOENT);
1526 	}
1527 
1528 	freeifgl = 0;
1529 	IF_ADDR_WLOCK(ifp);
1530 	CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1531 	IF_ADDR_WUNLOCK(ifp);
1532 
1533 	CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1534 		if (ifgm->ifgm_ifp == ifp)
1535 			break;
1536 
1537 	if (ifgm != NULL)
1538 		CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifg_member, ifgm_next);
1539 
1540 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1541 		CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, ifg_next);
1542 		freeifgl = 1;
1543 	}
1544 	IFNET_WUNLOCK();
1545 
1546 	epoch_wait_preempt(net_epoch_preempt);
1547 	if (freeifgl) {
1548 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1549 		free(ifgl->ifgl_group, M_TEMP);
1550 	}
1551 	free(ifgm, M_TEMP);
1552 	free(ifgl, M_TEMP);
1553 
1554 	EVENTHANDLER_INVOKE(group_change_event, groupname);
1555 
1556 	return (0);
1557 }
1558 
1559 /*
1560  * Remove an interface from all groups
1561  */
1562 static void
1563 if_delgroups(struct ifnet *ifp)
1564 {
1565 	struct ifg_list		*ifgl;
1566 	struct ifg_member	*ifgm;
1567 	char groupname[IFNAMSIZ];
1568 	int ifglfree;
1569 
1570 	IFNET_WLOCK();
1571 	while (!CK_STAILQ_EMPTY(&ifp->if_groups)) {
1572 		ifgl = CK_STAILQ_FIRST(&ifp->if_groups);
1573 
1574 		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1575 
1576 		IF_ADDR_WLOCK(ifp);
1577 		CK_STAILQ_REMOVE(&ifp->if_groups, ifgl, ifg_list, ifgl_next);
1578 		IF_ADDR_WUNLOCK(ifp);
1579 
1580 		CK_STAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1581 			if (ifgm->ifgm_ifp == ifp)
1582 				break;
1583 
1584 		if (ifgm != NULL)
1585 			CK_STAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifg_member,
1586 			    ifgm_next);
1587 		ifglfree = 0;
1588 		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1589 			CK_STAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_group, ifg_next);
1590 			ifglfree = 1;
1591 		}
1592 
1593 		IFNET_WUNLOCK();
1594 		epoch_wait_preempt(net_epoch_preempt);
1595 		free(ifgm, M_TEMP);
1596 		if (ifglfree) {
1597 			EVENTHANDLER_INVOKE(group_detach_event,
1598 								ifgl->ifgl_group);
1599 			free(ifgl->ifgl_group, M_TEMP);
1600 		}
1601 		EVENTHANDLER_INVOKE(group_change_event, groupname);
1602 
1603 		IFNET_WLOCK();
1604 	}
1605 	IFNET_WUNLOCK();
1606 }
1607 
1608 static char *
1609 ifgr_group_get(void *ifgrp)
1610 {
1611 	union ifgroupreq_union *ifgrup;
1612 
1613 	ifgrup = ifgrp;
1614 #ifdef COMPAT_FREEBSD32
1615 	if (SV_CURPROC_FLAG(SV_ILP32))
1616 		return (&ifgrup->ifgr32.ifgr_ifgru.ifgru_group[0]);
1617 #endif
1618 	return (&ifgrup->ifgr.ifgr_ifgru.ifgru_group[0]);
1619 }
1620 
1621 static struct ifg_req *
1622 ifgr_groups_get(void *ifgrp)
1623 {
1624 	union ifgroupreq_union *ifgrup;
1625 
1626 	ifgrup = ifgrp;
1627 #ifdef COMPAT_FREEBSD32
1628 	if (SV_CURPROC_FLAG(SV_ILP32))
1629 		return ((struct ifg_req *)(uintptr_t)
1630 		    ifgrup->ifgr32.ifgr_ifgru.ifgru_groups);
1631 #endif
1632 	return (ifgrup->ifgr.ifgr_ifgru.ifgru_groups);
1633 }
1634 
1635 /*
1636  * Stores all groups from an interface in memory pointed to by ifgr.
1637  */
1638 static int
1639 if_getgroup(struct ifgroupreq *ifgr, struct ifnet *ifp)
1640 {
1641 	struct epoch_tracker	 et;
1642 	int			 len, error;
1643 	struct ifg_list		*ifgl;
1644 	struct ifg_req		 ifgrq, *ifgp;
1645 
1646 	if (ifgr->ifgr_len == 0) {
1647 		NET_EPOCH_ENTER(et);
1648 		CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1649 			ifgr->ifgr_len += sizeof(struct ifg_req);
1650 		NET_EPOCH_EXIT(et);
1651 		return (0);
1652 	}
1653 
1654 	len = ifgr->ifgr_len;
1655 	ifgp = ifgr_groups_get(ifgr);
1656 	/* XXX: wire */
1657 	NET_EPOCH_ENTER(et);
1658 	CK_STAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1659 		if (len < sizeof(ifgrq)) {
1660 			NET_EPOCH_EXIT(et);
1661 			return (EINVAL);
1662 		}
1663 		bzero(&ifgrq, sizeof ifgrq);
1664 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1665 		    sizeof(ifgrq.ifgrq_group));
1666 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1667 		    	NET_EPOCH_EXIT(et);
1668 			return (error);
1669 		}
1670 		len -= sizeof(ifgrq);
1671 		ifgp++;
1672 	}
1673 	NET_EPOCH_EXIT(et);
1674 
1675 	return (0);
1676 }
1677 
1678 /*
1679  * Stores all members of a group in memory pointed to by igfr
1680  */
1681 static int
1682 if_getgroupmembers(struct ifgroupreq *ifgr)
1683 {
1684 	struct ifg_group	*ifg;
1685 	struct ifg_member	*ifgm;
1686 	struct ifg_req		 ifgrq, *ifgp;
1687 	int			 len, error;
1688 
1689 	IFNET_RLOCK();
1690 	CK_STAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1691 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1692 			break;
1693 	if (ifg == NULL) {
1694 		IFNET_RUNLOCK();
1695 		return (ENOENT);
1696 	}
1697 
1698 	if (ifgr->ifgr_len == 0) {
1699 		CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1700 			ifgr->ifgr_len += sizeof(ifgrq);
1701 		IFNET_RUNLOCK();
1702 		return (0);
1703 	}
1704 
1705 	len = ifgr->ifgr_len;
1706 	ifgp = ifgr_groups_get(ifgr);
1707 	CK_STAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1708 		if (len < sizeof(ifgrq)) {
1709 			IFNET_RUNLOCK();
1710 			return (EINVAL);
1711 		}
1712 		bzero(&ifgrq, sizeof ifgrq);
1713 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1714 		    sizeof(ifgrq.ifgrq_member));
1715 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1716 			IFNET_RUNLOCK();
1717 			return (error);
1718 		}
1719 		len -= sizeof(ifgrq);
1720 		ifgp++;
1721 	}
1722 	IFNET_RUNLOCK();
1723 
1724 	return (0);
1725 }
1726 
1727 /*
1728  * Return counter values from counter(9)s stored in ifnet.
1729  */
1730 uint64_t
1731 if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
1732 {
1733 
1734 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1735 
1736 	return (counter_u64_fetch(ifp->if_counters[cnt]));
1737 }
1738 
1739 /*
1740  * Increase an ifnet counter. Usually used for counters shared
1741  * between the stack and a driver, but function supports them all.
1742  */
1743 void
1744 if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
1745 {
1746 
1747 	KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
1748 
1749 	counter_u64_add(ifp->if_counters[cnt], inc);
1750 }
1751 
1752 /*
1753  * Copy data from ifnet to userland API structure if_data.
1754  */
1755 void
1756 if_data_copy(struct ifnet *ifp, struct if_data *ifd)
1757 {
1758 
1759 	ifd->ifi_type = ifp->if_type;
1760 	ifd->ifi_physical = 0;
1761 	ifd->ifi_addrlen = ifp->if_addrlen;
1762 	ifd->ifi_hdrlen = ifp->if_hdrlen;
1763 	ifd->ifi_link_state = ifp->if_link_state;
1764 	ifd->ifi_vhid = 0;
1765 	ifd->ifi_datalen = sizeof(struct if_data);
1766 	ifd->ifi_mtu = ifp->if_mtu;
1767 	ifd->ifi_metric = ifp->if_metric;
1768 	ifd->ifi_baudrate = ifp->if_baudrate;
1769 	ifd->ifi_hwassist = ifp->if_hwassist;
1770 	ifd->ifi_epoch = ifp->if_epoch;
1771 	ifd->ifi_lastchange = ifp->if_lastchange;
1772 
1773 	ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
1774 	ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
1775 	ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
1776 	ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
1777 	ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
1778 	ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
1779 	ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
1780 	ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
1781 	ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
1782 	ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
1783 	ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
1784 	ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
1785 }
1786 
1787 /*
1788  * Wrapper functions for struct ifnet address list locking macros.  These are
1789  * used by kernel modules to avoid encoding programming interface or binary
1790  * interface assumptions that may be violated when kernel-internal locking
1791  * approaches change.
1792  */
1793 void
1794 if_addr_rlock(struct ifnet *ifp)
1795 {
1796 
1797 	epoch_enter_preempt(net_epoch_preempt, curthread->td_et);
1798 }
1799 
1800 void
1801 if_addr_runlock(struct ifnet *ifp)
1802 {
1803 
1804 	epoch_exit_preempt(net_epoch_preempt, curthread->td_et);
1805 }
1806 
1807 void
1808 if_maddr_rlock(if_t ifp)
1809 {
1810 
1811 	epoch_enter_preempt(net_epoch_preempt, curthread->td_et);
1812 }
1813 
1814 void
1815 if_maddr_runlock(if_t ifp)
1816 {
1817 
1818 	epoch_exit_preempt(net_epoch_preempt, curthread->td_et);
1819 }
1820 
1821 /*
1822  * Initialization, destruction and refcounting functions for ifaddrs.
1823  */
1824 struct ifaddr *
1825 ifa_alloc(size_t size, int flags)
1826 {
1827 	struct ifaddr *ifa;
1828 
1829 	KASSERT(size >= sizeof(struct ifaddr),
1830 	    ("%s: invalid size %zu", __func__, size));
1831 
1832 	ifa = malloc(size, M_IFADDR, M_ZERO | flags);
1833 	if (ifa == NULL)
1834 		return (NULL);
1835 
1836 	if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
1837 		goto fail;
1838 	if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
1839 		goto fail;
1840 	if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
1841 		goto fail;
1842 	if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
1843 		goto fail;
1844 
1845 	refcount_init(&ifa->ifa_refcnt, 1);
1846 
1847 	return (ifa);
1848 
1849 fail:
1850 	/* free(NULL) is okay */
1851 	counter_u64_free(ifa->ifa_opackets);
1852 	counter_u64_free(ifa->ifa_ipackets);
1853 	counter_u64_free(ifa->ifa_obytes);
1854 	counter_u64_free(ifa->ifa_ibytes);
1855 	free(ifa, M_IFADDR);
1856 
1857 	return (NULL);
1858 }
1859 
1860 void
1861 ifa_ref(struct ifaddr *ifa)
1862 {
1863 
1864 	refcount_acquire(&ifa->ifa_refcnt);
1865 }
1866 
1867 static void
1868 ifa_destroy(epoch_context_t ctx)
1869 {
1870 	struct ifaddr *ifa;
1871 
1872 	ifa = __containerof(ctx, struct ifaddr, ifa_epoch_ctx);
1873 	counter_u64_free(ifa->ifa_opackets);
1874 	counter_u64_free(ifa->ifa_ipackets);
1875 	counter_u64_free(ifa->ifa_obytes);
1876 	counter_u64_free(ifa->ifa_ibytes);
1877 	free(ifa, M_IFADDR);
1878 }
1879 
1880 void
1881 ifa_free(struct ifaddr *ifa)
1882 {
1883 
1884 	if (refcount_release(&ifa->ifa_refcnt))
1885 		epoch_call(net_epoch_preempt, &ifa->ifa_epoch_ctx, ifa_destroy);
1886 }
1887 
1888 
1889 static int
1890 ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
1891     struct sockaddr *ia)
1892 {
1893 	struct epoch_tracker et;
1894 	int error;
1895 	struct rt_addrinfo info;
1896 	struct sockaddr_dl null_sdl;
1897 	struct ifnet *ifp;
1898 
1899 	ifp = ifa->ifa_ifp;
1900 
1901 	bzero(&info, sizeof(info));
1902 	if (cmd != RTM_DELETE)
1903 		info.rti_ifp = V_loif;
1904 	if (cmd == RTM_ADD) {
1905 		/* explicitly specify (loopback) ifa */
1906 		if (info.rti_ifp != NULL) {
1907 			NET_EPOCH_ENTER(et);
1908 			info.rti_ifa = ifaof_ifpforaddr(ifa->ifa_addr, info.rti_ifp);
1909 			if (info.rti_ifa != NULL)
1910 				ifa_ref(info.rti_ifa);
1911 			NET_EPOCH_EXIT(et);
1912 		}
1913 	}
1914 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
1915 	info.rti_info[RTAX_DST] = ia;
1916 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1917 	link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
1918 
1919 	error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib);
1920 
1921 	if (error != 0 &&
1922 	    !(cmd == RTM_ADD && error == EEXIST) &&
1923 	    !(cmd == RTM_DELETE && error == ENOENT))
1924 		if_printf(ifp, "%s failed: %d\n", otype, error);
1925 
1926 	return (error);
1927 }
1928 
1929 int
1930 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1931 {
1932 
1933 	return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
1934 }
1935 
1936 int
1937 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1938 {
1939 
1940 	return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
1941 }
1942 
1943 int
1944 ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1945 {
1946 
1947 	return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
1948 }
1949 
1950 /*
1951  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1952  * structs used to represent other address families, it is necessary
1953  * to perform a different comparison.
1954  */
1955 
1956 #define	sa_dl_equal(a1, a2)	\
1957 	((((const struct sockaddr_dl *)(a1))->sdl_len ==		\
1958 	 ((const struct sockaddr_dl *)(a2))->sdl_len) &&		\
1959 	 (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)),		\
1960 	       CLLADDR((const struct sockaddr_dl *)(a2)),		\
1961 	       ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1962 
1963 /*
1964  * Locate an interface based on a complete address.
1965  */
1966 /*ARGSUSED*/
1967 struct ifaddr *
1968 ifa_ifwithaddr(const struct sockaddr *addr)
1969 {
1970 	struct ifnet *ifp;
1971 	struct ifaddr *ifa;
1972 
1973 	MPASS(in_epoch(net_epoch_preempt));
1974 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1975 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1976 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1977 				continue;
1978 			if (sa_equal(addr, ifa->ifa_addr)) {
1979 				goto done;
1980 			}
1981 			/* IP6 doesn't have broadcast */
1982 			if ((ifp->if_flags & IFF_BROADCAST) &&
1983 			    ifa->ifa_broadaddr &&
1984 			    ifa->ifa_broadaddr->sa_len != 0 &&
1985 			    sa_equal(ifa->ifa_broadaddr, addr)) {
1986 				goto done;
1987 			}
1988 		}
1989 	}
1990 	ifa = NULL;
1991 done:
1992 	return (ifa);
1993 }
1994 
1995 int
1996 ifa_ifwithaddr_check(const struct sockaddr *addr)
1997 {
1998 	struct epoch_tracker et;
1999 	int rc;
2000 
2001 	NET_EPOCH_ENTER(et);
2002 	rc = (ifa_ifwithaddr(addr) != NULL);
2003 	NET_EPOCH_EXIT(et);
2004 	return (rc);
2005 }
2006 
2007 /*
2008  * Locate an interface based on the broadcast address.
2009  */
2010 /* ARGSUSED */
2011 struct ifaddr *
2012 ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
2013 {
2014 	struct ifnet *ifp;
2015 	struct ifaddr *ifa;
2016 
2017 	MPASS(in_epoch(net_epoch_preempt));
2018 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2019 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2020 			continue;
2021 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2022 			if (ifa->ifa_addr->sa_family != addr->sa_family)
2023 				continue;
2024 			if ((ifp->if_flags & IFF_BROADCAST) &&
2025 			    ifa->ifa_broadaddr &&
2026 			    ifa->ifa_broadaddr->sa_len != 0 &&
2027 			    sa_equal(ifa->ifa_broadaddr, addr)) {
2028 				goto done;
2029 			}
2030 		}
2031 	}
2032 	ifa = NULL;
2033 done:
2034 	return (ifa);
2035 }
2036 
2037 /*
2038  * Locate the point to point interface with a given destination address.
2039  */
2040 /*ARGSUSED*/
2041 struct ifaddr *
2042 ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
2043 {
2044 	struct ifnet *ifp;
2045 	struct ifaddr *ifa;
2046 
2047 	MPASS(in_epoch(net_epoch_preempt));
2048 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2049 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
2050 			continue;
2051 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2052 			continue;
2053 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2054 			if (ifa->ifa_addr->sa_family != addr->sa_family)
2055 				continue;
2056 			if (ifa->ifa_dstaddr != NULL &&
2057 			    sa_equal(addr, ifa->ifa_dstaddr)) {
2058 				goto done;
2059 			}
2060 		}
2061 	}
2062 	ifa = NULL;
2063 done:
2064 	return (ifa);
2065 }
2066 
2067 /*
2068  * Find an interface on a specific network.  If many, choice
2069  * is most specific found.
2070  */
2071 struct ifaddr *
2072 ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
2073 {
2074 	struct ifnet *ifp;
2075 	struct ifaddr *ifa;
2076 	struct ifaddr *ifa_maybe = NULL;
2077 	u_int af = addr->sa_family;
2078 	const char *addr_data = addr->sa_data, *cplim;
2079 
2080 	MPASS(in_epoch(net_epoch_preempt));
2081 	/*
2082 	 * AF_LINK addresses can be looked up directly by their index number,
2083 	 * so do that if we can.
2084 	 */
2085 	if (af == AF_LINK) {
2086 	    const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
2087 	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
2088 		return (ifaddr_byindex(sdl->sdl_index));
2089 	}
2090 
2091 	/*
2092 	 * Scan though each interface, looking for ones that have addresses
2093 	 * in this address family and the requested fib.
2094 	 */
2095 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2096 		if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
2097 			continue;
2098 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2099 			const char *cp, *cp2, *cp3;
2100 
2101 			if (ifa->ifa_addr->sa_family != af)
2102 next:				continue;
2103 			if (af == AF_INET &&
2104 			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
2105 				/*
2106 				 * This is a bit broken as it doesn't
2107 				 * take into account that the remote end may
2108 				 * be a single node in the network we are
2109 				 * looking for.
2110 				 * The trouble is that we don't know the
2111 				 * netmask for the remote end.
2112 				 */
2113 				if (ifa->ifa_dstaddr != NULL &&
2114 				    sa_equal(addr, ifa->ifa_dstaddr)) {
2115 					goto done;
2116 				}
2117 			} else {
2118 				/*
2119 				 * Scan all the bits in the ifa's address.
2120 				 * If a bit dissagrees with what we are
2121 				 * looking for, mask it with the netmask
2122 				 * to see if it really matters.
2123 				 * (A byte at a time)
2124 				 */
2125 				if (ifa->ifa_netmask == 0)
2126 					continue;
2127 				cp = addr_data;
2128 				cp2 = ifa->ifa_addr->sa_data;
2129 				cp3 = ifa->ifa_netmask->sa_data;
2130 				cplim = ifa->ifa_netmask->sa_len
2131 					+ (char *)ifa->ifa_netmask;
2132 				while (cp3 < cplim)
2133 					if ((*cp++ ^ *cp2++) & *cp3++)
2134 						goto next; /* next address! */
2135 				/*
2136 				 * If the netmask of what we just found
2137 				 * is more specific than what we had before
2138 				 * (if we had one), or if the virtual status
2139 				 * of new prefix is better than of the old one,
2140 				 * then remember the new one before continuing
2141 				 * to search for an even better one.
2142 				 */
2143 				if (ifa_maybe == NULL ||
2144 				    ifa_preferred(ifa_maybe, ifa) ||
2145 				    rn_refines((caddr_t)ifa->ifa_netmask,
2146 				    (caddr_t)ifa_maybe->ifa_netmask)) {
2147 					ifa_maybe = ifa;
2148 				}
2149 			}
2150 		}
2151 	}
2152 	ifa = ifa_maybe;
2153 	ifa_maybe = NULL;
2154 done:
2155 	return (ifa);
2156 }
2157 
2158 /*
2159  * Find an interface address specific to an interface best matching
2160  * a given address.
2161  */
2162 struct ifaddr *
2163 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
2164 {
2165 	struct ifaddr *ifa;
2166 	const char *cp, *cp2, *cp3;
2167 	char *cplim;
2168 	struct ifaddr *ifa_maybe = NULL;
2169 	u_int af = addr->sa_family;
2170 
2171 	if (af >= AF_MAX)
2172 		return (NULL);
2173 
2174 	MPASS(in_epoch(net_epoch_preempt));
2175 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2176 		if (ifa->ifa_addr->sa_family != af)
2177 			continue;
2178 		if (ifa_maybe == NULL)
2179 			ifa_maybe = ifa;
2180 		if (ifa->ifa_netmask == 0) {
2181 			if (sa_equal(addr, ifa->ifa_addr) ||
2182 			    (ifa->ifa_dstaddr &&
2183 			    sa_equal(addr, ifa->ifa_dstaddr)))
2184 				goto done;
2185 			continue;
2186 		}
2187 		if (ifp->if_flags & IFF_POINTOPOINT) {
2188 			if (sa_equal(addr, ifa->ifa_dstaddr))
2189 				goto done;
2190 		} else {
2191 			cp = addr->sa_data;
2192 			cp2 = ifa->ifa_addr->sa_data;
2193 			cp3 = ifa->ifa_netmask->sa_data;
2194 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
2195 			for (; cp3 < cplim; cp3++)
2196 				if ((*cp++ ^ *cp2++) & *cp3)
2197 					break;
2198 			if (cp3 == cplim)
2199 				goto done;
2200 		}
2201 	}
2202 	ifa = ifa_maybe;
2203 done:
2204 	return (ifa);
2205 }
2206 
2207 /*
2208  * See whether new ifa is better than current one:
2209  * 1) A non-virtual one is preferred over virtual.
2210  * 2) A virtual in master state preferred over any other state.
2211  *
2212  * Used in several address selecting functions.
2213  */
2214 int
2215 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
2216 {
2217 
2218 	return (cur->ifa_carp && (!next->ifa_carp ||
2219 	    ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
2220 }
2221 
2222 struct sockaddr_dl *
2223 link_alloc_sdl(size_t size, int flags)
2224 {
2225 
2226 	return (malloc(size, M_TEMP, flags));
2227 }
2228 
2229 void
2230 link_free_sdl(struct sockaddr *sa)
2231 {
2232 	free(sa, M_TEMP);
2233 }
2234 
2235 /*
2236  * Fills in given sdl with interface basic info.
2237  * Returns pointer to filled sdl.
2238  */
2239 struct sockaddr_dl *
2240 link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
2241 {
2242 	struct sockaddr_dl *sdl;
2243 
2244 	sdl = (struct sockaddr_dl *)paddr;
2245 	memset(sdl, 0, sizeof(struct sockaddr_dl));
2246 	sdl->sdl_len = sizeof(struct sockaddr_dl);
2247 	sdl->sdl_family = AF_LINK;
2248 	sdl->sdl_index = ifp->if_index;
2249 	sdl->sdl_type = iftype;
2250 
2251 	return (sdl);
2252 }
2253 
2254 /*
2255  * Mark an interface down and notify protocols of
2256  * the transition.
2257  */
2258 static void
2259 if_unroute(struct ifnet *ifp, int flag, int fam)
2260 {
2261 	struct ifaddr *ifa;
2262 
2263 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
2264 
2265 	ifp->if_flags &= ~flag;
2266 	getmicrotime(&ifp->if_lastchange);
2267 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2268 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2269 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
2270 	ifp->if_qflush(ifp);
2271 
2272 	if (ifp->if_carp)
2273 		(*carp_linkstate_p)(ifp);
2274 	rt_ifmsg(ifp);
2275 }
2276 
2277 /*
2278  * Mark an interface up and notify protocols of
2279  * the transition.
2280  */
2281 static void
2282 if_route(struct ifnet *ifp, int flag, int fam)
2283 {
2284 	struct ifaddr *ifa;
2285 
2286 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
2287 
2288 	ifp->if_flags |= flag;
2289 	getmicrotime(&ifp->if_lastchange);
2290 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
2291 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
2292 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
2293 	if (ifp->if_carp)
2294 		(*carp_linkstate_p)(ifp);
2295 	rt_ifmsg(ifp);
2296 #ifdef INET6
2297 	in6_if_up(ifp);
2298 #endif
2299 }
2300 
2301 void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
2302 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
2303 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
2304 struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
2305 int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
2306 int	(*vlan_pcp_p)(struct ifnet *, uint16_t *);
2307 int	(*vlan_setcookie_p)(struct ifnet *, void *);
2308 void	*(*vlan_cookie_p)(struct ifnet *);
2309 
2310 /*
2311  * Handle a change in the interface link state. To avoid LORs
2312  * between driver lock and upper layer locks, as well as possible
2313  * recursions, we post event to taskqueue, and all job
2314  * is done in static do_link_state_change().
2315  */
2316 void
2317 if_link_state_change(struct ifnet *ifp, int link_state)
2318 {
2319 	/* Return if state hasn't changed. */
2320 	if (ifp->if_link_state == link_state)
2321 		return;
2322 
2323 	ifp->if_link_state = link_state;
2324 
2325 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
2326 }
2327 
2328 static void
2329 do_link_state_change(void *arg, int pending)
2330 {
2331 	struct ifnet *ifp = (struct ifnet *)arg;
2332 	int link_state = ifp->if_link_state;
2333 	CURVNET_SET(ifp->if_vnet);
2334 
2335 	/* Notify that the link state has changed. */
2336 	rt_ifmsg(ifp);
2337 	if (ifp->if_vlantrunk != NULL)
2338 		(*vlan_link_state_p)(ifp);
2339 
2340 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
2341 	    ifp->if_l2com != NULL)
2342 		(*ng_ether_link_state_p)(ifp, link_state);
2343 	if (ifp->if_carp)
2344 		(*carp_linkstate_p)(ifp);
2345 	if (ifp->if_bridge)
2346 		ifp->if_bridge_linkstate(ifp);
2347 	if (ifp->if_lagg)
2348 		(*lagg_linkstate_p)(ifp, link_state);
2349 
2350 	if (IS_DEFAULT_VNET(curvnet))
2351 		devctl_notify("IFNET", ifp->if_xname,
2352 		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
2353 		    NULL);
2354 	if (pending > 1)
2355 		if_printf(ifp, "%d link states coalesced\n", pending);
2356 	if (log_link_state_change)
2357 		if_printf(ifp, "link state changed to %s\n",
2358 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
2359 	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, link_state);
2360 	CURVNET_RESTORE();
2361 }
2362 
2363 /*
2364  * Mark an interface down and notify protocols of
2365  * the transition.
2366  */
2367 void
2368 if_down(struct ifnet *ifp)
2369 {
2370 
2371 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_DOWN);
2372 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
2373 }
2374 
2375 /*
2376  * Mark an interface up and notify protocols of
2377  * the transition.
2378  */
2379 void
2380 if_up(struct ifnet *ifp)
2381 {
2382 
2383 	if_route(ifp, IFF_UP, AF_UNSPEC);
2384 	EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_UP);
2385 }
2386 
2387 /*
2388  * Flush an interface queue.
2389  */
2390 void
2391 if_qflush(struct ifnet *ifp)
2392 {
2393 	struct mbuf *m, *n;
2394 	struct ifaltq *ifq;
2395 
2396 	ifq = &ifp->if_snd;
2397 	IFQ_LOCK(ifq);
2398 #ifdef ALTQ
2399 	if (ALTQ_IS_ENABLED(ifq))
2400 		ALTQ_PURGE(ifq);
2401 #endif
2402 	n = ifq->ifq_head;
2403 	while ((m = n) != NULL) {
2404 		n = m->m_nextpkt;
2405 		m_freem(m);
2406 	}
2407 	ifq->ifq_head = 0;
2408 	ifq->ifq_tail = 0;
2409 	ifq->ifq_len = 0;
2410 	IFQ_UNLOCK(ifq);
2411 }
2412 
2413 /*
2414  * Map interface name to interface structure pointer, with or without
2415  * returning a reference.
2416  */
2417 struct ifnet *
2418 ifunit_ref(const char *name)
2419 {
2420 	struct epoch_tracker et;
2421 	struct ifnet *ifp;
2422 
2423 	NET_EPOCH_ENTER(et);
2424 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2425 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2426 		    !(ifp->if_flags & IFF_DYING))
2427 			break;
2428 	}
2429 	if (ifp != NULL)
2430 		if_ref(ifp);
2431 	NET_EPOCH_EXIT(et);
2432 	return (ifp);
2433 }
2434 
2435 struct ifnet *
2436 ifunit(const char *name)
2437 {
2438 	struct epoch_tracker et;
2439 	struct ifnet *ifp;
2440 
2441 	NET_EPOCH_ENTER(et);
2442 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2443 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2444 			break;
2445 	}
2446 	NET_EPOCH_EXIT(et);
2447 	return (ifp);
2448 }
2449 
2450 static void *
2451 ifr_buffer_get_buffer(void *data)
2452 {
2453 	union ifreq_union *ifrup;
2454 
2455 	ifrup = data;
2456 #ifdef COMPAT_FREEBSD32
2457 	if (SV_CURPROC_FLAG(SV_ILP32))
2458 		return ((void *)(uintptr_t)
2459 		    ifrup->ifr32.ifr_ifru.ifru_buffer.buffer);
2460 #endif
2461 	return (ifrup->ifr.ifr_ifru.ifru_buffer.buffer);
2462 }
2463 
2464 static void
2465 ifr_buffer_set_buffer_null(void *data)
2466 {
2467 	union ifreq_union *ifrup;
2468 
2469 	ifrup = data;
2470 #ifdef COMPAT_FREEBSD32
2471 	if (SV_CURPROC_FLAG(SV_ILP32))
2472 		ifrup->ifr32.ifr_ifru.ifru_buffer.buffer = 0;
2473 	else
2474 #endif
2475 		ifrup->ifr.ifr_ifru.ifru_buffer.buffer = NULL;
2476 }
2477 
2478 static size_t
2479 ifr_buffer_get_length(void *data)
2480 {
2481 	union ifreq_union *ifrup;
2482 
2483 	ifrup = data;
2484 #ifdef COMPAT_FREEBSD32
2485 	if (SV_CURPROC_FLAG(SV_ILP32))
2486 		return (ifrup->ifr32.ifr_ifru.ifru_buffer.length);
2487 #endif
2488 	return (ifrup->ifr.ifr_ifru.ifru_buffer.length);
2489 }
2490 
2491 static void
2492 ifr_buffer_set_length(void *data, size_t len)
2493 {
2494 	union ifreq_union *ifrup;
2495 
2496 	ifrup = data;
2497 #ifdef COMPAT_FREEBSD32
2498 	if (SV_CURPROC_FLAG(SV_ILP32))
2499 		ifrup->ifr32.ifr_ifru.ifru_buffer.length = len;
2500 	else
2501 #endif
2502 		ifrup->ifr.ifr_ifru.ifru_buffer.length = len;
2503 }
2504 
2505 void *
2506 ifr_data_get_ptr(void *ifrp)
2507 {
2508 	union ifreq_union *ifrup;
2509 
2510 	ifrup = ifrp;
2511 #ifdef COMPAT_FREEBSD32
2512 	if (SV_CURPROC_FLAG(SV_ILP32))
2513 		return ((void *)(uintptr_t)
2514 		    ifrup->ifr32.ifr_ifru.ifru_data);
2515 #endif
2516 		return (ifrup->ifr.ifr_ifru.ifru_data);
2517 }
2518 
2519 /*
2520  * Hardware specific interface ioctls.
2521  */
2522 int
2523 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2524 {
2525 	struct ifreq *ifr;
2526 	int error = 0, do_ifup = 0;
2527 	int new_flags, temp_flags;
2528 	size_t namelen, onamelen;
2529 	size_t descrlen;
2530 	char *descrbuf, *odescrbuf;
2531 	char new_name[IFNAMSIZ];
2532 	struct ifaddr *ifa;
2533 	struct sockaddr_dl *sdl;
2534 
2535 	ifr = (struct ifreq *)data;
2536 	switch (cmd) {
2537 	case SIOCGIFINDEX:
2538 		ifr->ifr_index = ifp->if_index;
2539 		break;
2540 
2541 	case SIOCGIFFLAGS:
2542 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2543 		ifr->ifr_flags = temp_flags & 0xffff;
2544 		ifr->ifr_flagshigh = temp_flags >> 16;
2545 		break;
2546 
2547 	case SIOCGIFCAP:
2548 		ifr->ifr_reqcap = ifp->if_capabilities;
2549 		ifr->ifr_curcap = ifp->if_capenable;
2550 		break;
2551 
2552 #ifdef MAC
2553 	case SIOCGIFMAC:
2554 		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2555 		break;
2556 #endif
2557 
2558 	case SIOCGIFMETRIC:
2559 		ifr->ifr_metric = ifp->if_metric;
2560 		break;
2561 
2562 	case SIOCGIFMTU:
2563 		ifr->ifr_mtu = ifp->if_mtu;
2564 		break;
2565 
2566 	case SIOCGIFPHYS:
2567 		/* XXXGL: did this ever worked? */
2568 		ifr->ifr_phys = 0;
2569 		break;
2570 
2571 	case SIOCGIFDESCR:
2572 		error = 0;
2573 		sx_slock(&ifdescr_sx);
2574 		if (ifp->if_description == NULL)
2575 			error = ENOMSG;
2576 		else {
2577 			/* space for terminating nul */
2578 			descrlen = strlen(ifp->if_description) + 1;
2579 			if (ifr_buffer_get_length(ifr) < descrlen)
2580 				ifr_buffer_set_buffer_null(ifr);
2581 			else
2582 				error = copyout(ifp->if_description,
2583 				    ifr_buffer_get_buffer(ifr), descrlen);
2584 			ifr_buffer_set_length(ifr, descrlen);
2585 		}
2586 		sx_sunlock(&ifdescr_sx);
2587 		break;
2588 
2589 	case SIOCSIFDESCR:
2590 		error = priv_check(td, PRIV_NET_SETIFDESCR);
2591 		if (error)
2592 			return (error);
2593 
2594 		/*
2595 		 * Copy only (length-1) bytes to make sure that
2596 		 * if_description is always nul terminated.  The
2597 		 * length parameter is supposed to count the
2598 		 * terminating nul in.
2599 		 */
2600 		if (ifr_buffer_get_length(ifr) > ifdescr_maxlen)
2601 			return (ENAMETOOLONG);
2602 		else if (ifr_buffer_get_length(ifr) == 0)
2603 			descrbuf = NULL;
2604 		else {
2605 			descrbuf = malloc(ifr_buffer_get_length(ifr),
2606 			    M_IFDESCR, M_WAITOK | M_ZERO);
2607 			error = copyin(ifr_buffer_get_buffer(ifr), descrbuf,
2608 			    ifr_buffer_get_length(ifr) - 1);
2609 			if (error) {
2610 				free(descrbuf, M_IFDESCR);
2611 				break;
2612 			}
2613 		}
2614 
2615 		sx_xlock(&ifdescr_sx);
2616 		odescrbuf = ifp->if_description;
2617 		ifp->if_description = descrbuf;
2618 		sx_xunlock(&ifdescr_sx);
2619 
2620 		getmicrotime(&ifp->if_lastchange);
2621 		free(odescrbuf, M_IFDESCR);
2622 		break;
2623 
2624 	case SIOCGIFFIB:
2625 		ifr->ifr_fib = ifp->if_fib;
2626 		break;
2627 
2628 	case SIOCSIFFIB:
2629 		error = priv_check(td, PRIV_NET_SETIFFIB);
2630 		if (error)
2631 			return (error);
2632 		if (ifr->ifr_fib >= rt_numfibs)
2633 			return (EINVAL);
2634 
2635 		ifp->if_fib = ifr->ifr_fib;
2636 		break;
2637 
2638 	case SIOCSIFFLAGS:
2639 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2640 		if (error)
2641 			return (error);
2642 		/*
2643 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2644 		 * check, so we don't need special handling here yet.
2645 		 */
2646 		new_flags = (ifr->ifr_flags & 0xffff) |
2647 		    (ifr->ifr_flagshigh << 16);
2648 		if (ifp->if_flags & IFF_UP &&
2649 		    (new_flags & IFF_UP) == 0) {
2650 			if_down(ifp);
2651 		} else if (new_flags & IFF_UP &&
2652 		    (ifp->if_flags & IFF_UP) == 0) {
2653 			do_ifup = 1;
2654 		}
2655 		/* See if permanently promiscuous mode bit is about to flip */
2656 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2657 			if (new_flags & IFF_PPROMISC)
2658 				ifp->if_flags |= IFF_PROMISC;
2659 			else if (ifp->if_pcount == 0)
2660 				ifp->if_flags &= ~IFF_PROMISC;
2661 			if (log_promisc_mode_change)
2662                                 if_printf(ifp, "permanently promiscuous mode %s\n",
2663                                     ((new_flags & IFF_PPROMISC) ?
2664                                      "enabled" : "disabled"));
2665 		}
2666 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2667 			(new_flags &~ IFF_CANTCHANGE);
2668 		if (ifp->if_ioctl) {
2669 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2670 		}
2671 		if (do_ifup)
2672 			if_up(ifp);
2673 		getmicrotime(&ifp->if_lastchange);
2674 		break;
2675 
2676 	case SIOCSIFCAP:
2677 		error = priv_check(td, PRIV_NET_SETIFCAP);
2678 		if (error)
2679 			return (error);
2680 		if (ifp->if_ioctl == NULL)
2681 			return (EOPNOTSUPP);
2682 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2683 			return (EINVAL);
2684 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2685 		if (error == 0)
2686 			getmicrotime(&ifp->if_lastchange);
2687 		break;
2688 
2689 #ifdef MAC
2690 	case SIOCSIFMAC:
2691 		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2692 		break;
2693 #endif
2694 
2695 	case SIOCSIFNAME:
2696 		error = priv_check(td, PRIV_NET_SETIFNAME);
2697 		if (error)
2698 			return (error);
2699 		error = copyinstr(ifr_data_get_ptr(ifr), new_name, IFNAMSIZ,
2700 		    NULL);
2701 		if (error != 0)
2702 			return (error);
2703 		if (new_name[0] == '\0')
2704 			return (EINVAL);
2705 		if (new_name[IFNAMSIZ-1] != '\0') {
2706 			new_name[IFNAMSIZ-1] = '\0';
2707 			if (strlen(new_name) == IFNAMSIZ-1)
2708 				return (EINVAL);
2709 		}
2710 		if (ifunit(new_name) != NULL)
2711 			return (EEXIST);
2712 
2713 		/*
2714 		 * XXX: Locking.  Nothing else seems to lock if_flags,
2715 		 * and there are numerous other races with the
2716 		 * ifunit() checks not being atomic with namespace
2717 		 * changes (renames, vmoves, if_attach, etc).
2718 		 */
2719 		ifp->if_flags |= IFF_RENAMING;
2720 
2721 		/* Announce the departure of the interface. */
2722 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2723 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2724 
2725 		if_printf(ifp, "changing name to '%s'\n", new_name);
2726 
2727 		IF_ADDR_WLOCK(ifp);
2728 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2729 		ifa = ifp->if_addr;
2730 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2731 		namelen = strlen(new_name);
2732 		onamelen = sdl->sdl_nlen;
2733 		/*
2734 		 * Move the address if needed.  This is safe because we
2735 		 * allocate space for a name of length IFNAMSIZ when we
2736 		 * create this in if_attach().
2737 		 */
2738 		if (namelen != onamelen) {
2739 			bcopy(sdl->sdl_data + onamelen,
2740 			    sdl->sdl_data + namelen, sdl->sdl_alen);
2741 		}
2742 		bcopy(new_name, sdl->sdl_data, namelen);
2743 		sdl->sdl_nlen = namelen;
2744 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2745 		bzero(sdl->sdl_data, onamelen);
2746 		while (namelen != 0)
2747 			sdl->sdl_data[--namelen] = 0xff;
2748 		IF_ADDR_WUNLOCK(ifp);
2749 
2750 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2751 		/* Announce the return of the interface. */
2752 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2753 
2754 		ifp->if_flags &= ~IFF_RENAMING;
2755 		break;
2756 
2757 #ifdef VIMAGE
2758 	case SIOCSIFVNET:
2759 		error = priv_check(td, PRIV_NET_SETIFVNET);
2760 		if (error)
2761 			return (error);
2762 		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2763 		break;
2764 #endif
2765 
2766 	case SIOCSIFMETRIC:
2767 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2768 		if (error)
2769 			return (error);
2770 		ifp->if_metric = ifr->ifr_metric;
2771 		getmicrotime(&ifp->if_lastchange);
2772 		break;
2773 
2774 	case SIOCSIFPHYS:
2775 		error = priv_check(td, PRIV_NET_SETIFPHYS);
2776 		if (error)
2777 			return (error);
2778 		if (ifp->if_ioctl == NULL)
2779 			return (EOPNOTSUPP);
2780 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2781 		if (error == 0)
2782 			getmicrotime(&ifp->if_lastchange);
2783 		break;
2784 
2785 	case SIOCSIFMTU:
2786 	{
2787 		u_long oldmtu = ifp->if_mtu;
2788 
2789 		error = priv_check(td, PRIV_NET_SETIFMTU);
2790 		if (error)
2791 			return (error);
2792 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2793 			return (EINVAL);
2794 		if (ifp->if_ioctl == NULL)
2795 			return (EOPNOTSUPP);
2796 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2797 		if (error == 0) {
2798 			getmicrotime(&ifp->if_lastchange);
2799 			rt_ifmsg(ifp);
2800 #ifdef INET
2801 			NETDUMP_REINIT(ifp);
2802 #endif
2803 		}
2804 		/*
2805 		 * If the link MTU changed, do network layer specific procedure.
2806 		 */
2807 		if (ifp->if_mtu != oldmtu) {
2808 #ifdef INET6
2809 			nd6_setmtu(ifp);
2810 #endif
2811 			rt_updatemtu(ifp);
2812 		}
2813 		break;
2814 	}
2815 
2816 	case SIOCADDMULTI:
2817 	case SIOCDELMULTI:
2818 		if (cmd == SIOCADDMULTI)
2819 			error = priv_check(td, PRIV_NET_ADDMULTI);
2820 		else
2821 			error = priv_check(td, PRIV_NET_DELMULTI);
2822 		if (error)
2823 			return (error);
2824 
2825 		/* Don't allow group membership on non-multicast interfaces. */
2826 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2827 			return (EOPNOTSUPP);
2828 
2829 		/* Don't let users screw up protocols' entries. */
2830 		if (ifr->ifr_addr.sa_family != AF_LINK)
2831 			return (EINVAL);
2832 
2833 		if (cmd == SIOCADDMULTI) {
2834 			struct epoch_tracker et;
2835 			struct ifmultiaddr *ifma;
2836 
2837 			/*
2838 			 * Userland is only permitted to join groups once
2839 			 * via the if_addmulti() KPI, because it cannot hold
2840 			 * struct ifmultiaddr * between calls. It may also
2841 			 * lose a race while we check if the membership
2842 			 * already exists.
2843 			 */
2844 			NET_EPOCH_ENTER(et);
2845 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2846 			NET_EPOCH_EXIT(et);
2847 			if (ifma != NULL)
2848 				error = EADDRINUSE;
2849 			else
2850 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2851 		} else {
2852 			error = if_delmulti(ifp, &ifr->ifr_addr);
2853 		}
2854 		if (error == 0)
2855 			getmicrotime(&ifp->if_lastchange);
2856 		break;
2857 
2858 	case SIOCSIFPHYADDR:
2859 	case SIOCDIFPHYADDR:
2860 #ifdef INET6
2861 	case SIOCSIFPHYADDR_IN6:
2862 #endif
2863 	case SIOCSIFMEDIA:
2864 	case SIOCSIFGENERIC:
2865 		error = priv_check(td, PRIV_NET_HWIOCTL);
2866 		if (error)
2867 			return (error);
2868 		if (ifp->if_ioctl == NULL)
2869 			return (EOPNOTSUPP);
2870 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2871 		if (error == 0)
2872 			getmicrotime(&ifp->if_lastchange);
2873 		break;
2874 
2875 	case SIOCGIFSTATUS:
2876 	case SIOCGIFPSRCADDR:
2877 	case SIOCGIFPDSTADDR:
2878 	case SIOCGIFMEDIA:
2879 	case SIOCGIFXMEDIA:
2880 	case SIOCGIFGENERIC:
2881 	case SIOCGIFRSSKEY:
2882 	case SIOCGIFRSSHASH:
2883 		if (ifp->if_ioctl == NULL)
2884 			return (EOPNOTSUPP);
2885 		error = (*ifp->if_ioctl)(ifp, cmd, data);
2886 		break;
2887 
2888 	case SIOCSIFLLADDR:
2889 		error = priv_check(td, PRIV_NET_SETLLADDR);
2890 		if (error)
2891 			return (error);
2892 		error = if_setlladdr(ifp,
2893 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2894 		break;
2895 
2896 	case SIOCGHWADDR:
2897 		error = if_gethwaddr(ifp, ifr);
2898 		break;
2899 
2900 	case CASE_IOC_IFGROUPREQ(SIOCAIFGROUP):
2901 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2902 		if (error)
2903 			return (error);
2904 		if ((error = if_addgroup(ifp,
2905 		    ifgr_group_get((struct ifgroupreq *)data))))
2906 			return (error);
2907 		break;
2908 
2909 	case CASE_IOC_IFGROUPREQ(SIOCGIFGROUP):
2910 		if ((error = if_getgroup((struct ifgroupreq *)data, ifp)))
2911 			return (error);
2912 		break;
2913 
2914 	case CASE_IOC_IFGROUPREQ(SIOCDIFGROUP):
2915 		error = priv_check(td, PRIV_NET_DELIFGROUP);
2916 		if (error)
2917 			return (error);
2918 		if ((error = if_delgroup(ifp,
2919 		    ifgr_group_get((struct ifgroupreq *)data))))
2920 			return (error);
2921 		break;
2922 
2923 	default:
2924 		error = ENOIOCTL;
2925 		break;
2926 	}
2927 	return (error);
2928 }
2929 
2930 #ifdef COMPAT_FREEBSD32
2931 struct ifconf32 {
2932 	int32_t	ifc_len;
2933 	union {
2934 		uint32_t	ifcu_buf;
2935 		uint32_t	ifcu_req;
2936 	} ifc_ifcu;
2937 };
2938 #define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2939 #endif
2940 
2941 #ifdef COMPAT_FREEBSD32
2942 static void
2943 ifmr_init(struct ifmediareq *ifmr, caddr_t data)
2944 {
2945 	struct ifmediareq32 *ifmr32;
2946 
2947 	ifmr32 = (struct ifmediareq32 *)data;
2948 	memcpy(ifmr->ifm_name, ifmr32->ifm_name,
2949 	    sizeof(ifmr->ifm_name));
2950 	ifmr->ifm_current = ifmr32->ifm_current;
2951 	ifmr->ifm_mask = ifmr32->ifm_mask;
2952 	ifmr->ifm_status = ifmr32->ifm_status;
2953 	ifmr->ifm_active = ifmr32->ifm_active;
2954 	ifmr->ifm_count = ifmr32->ifm_count;
2955 	ifmr->ifm_ulist = (int *)(uintptr_t)ifmr32->ifm_ulist;
2956 }
2957 
2958 static void
2959 ifmr_update(const struct ifmediareq *ifmr, caddr_t data)
2960 {
2961 	struct ifmediareq32 *ifmr32;
2962 
2963 	ifmr32 = (struct ifmediareq32 *)data;
2964 	ifmr32->ifm_current = ifmr->ifm_current;
2965 	ifmr32->ifm_mask = ifmr->ifm_mask;
2966 	ifmr32->ifm_status = ifmr->ifm_status;
2967 	ifmr32->ifm_active = ifmr->ifm_active;
2968 	ifmr32->ifm_count = ifmr->ifm_count;
2969 }
2970 #endif
2971 
2972 /*
2973  * Interface ioctls.
2974  */
2975 int
2976 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2977 {
2978 #ifdef COMPAT_FREEBSD32
2979 	caddr_t saved_data = NULL;
2980 	struct ifmediareq ifmr;
2981 	struct ifmediareq *ifmrp;
2982 #endif
2983 	struct ifnet *ifp;
2984 	struct ifreq *ifr;
2985 	int error;
2986 	int oif_flags;
2987 #ifdef VIMAGE
2988 	int shutdown;
2989 #endif
2990 
2991 	CURVNET_SET(so->so_vnet);
2992 #ifdef VIMAGE
2993 	/* Make sure the VNET is stable. */
2994 	shutdown = (so->so_vnet->vnet_state > SI_SUB_VNET &&
2995 		 so->so_vnet->vnet_state < SI_SUB_VNET_DONE) ? 1 : 0;
2996 	if (shutdown) {
2997 		CURVNET_RESTORE();
2998 		return (EBUSY);
2999 	}
3000 #endif
3001 
3002 
3003 	switch (cmd) {
3004 	case SIOCGIFCONF:
3005 		error = ifconf(cmd, data);
3006 		CURVNET_RESTORE();
3007 		return (error);
3008 
3009 #ifdef COMPAT_FREEBSD32
3010 	case SIOCGIFCONF32:
3011 		{
3012 			struct ifconf32 *ifc32;
3013 			struct ifconf ifc;
3014 
3015 			ifc32 = (struct ifconf32 *)data;
3016 			ifc.ifc_len = ifc32->ifc_len;
3017 			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
3018 
3019 			error = ifconf(SIOCGIFCONF, (void *)&ifc);
3020 			CURVNET_RESTORE();
3021 			if (error == 0)
3022 				ifc32->ifc_len = ifc.ifc_len;
3023 			return (error);
3024 		}
3025 #endif
3026 	}
3027 
3028 #ifdef COMPAT_FREEBSD32
3029 	ifmrp = NULL;
3030 	switch (cmd) {
3031 	case SIOCGIFMEDIA32:
3032 	case SIOCGIFXMEDIA32:
3033 		ifmrp = &ifmr;
3034 		ifmr_init(ifmrp, data);
3035 		cmd = _IOC_NEWTYPE(cmd, struct ifmediareq);
3036 		saved_data = data;
3037 		data = (caddr_t)ifmrp;
3038 	}
3039 #endif
3040 
3041 	ifr = (struct ifreq *)data;
3042 	switch (cmd) {
3043 #ifdef VIMAGE
3044 	case SIOCSIFRVNET:
3045 		error = priv_check(td, PRIV_NET_SETIFVNET);
3046 		if (error == 0)
3047 			error = if_vmove_reclaim(td, ifr->ifr_name,
3048 			    ifr->ifr_jid);
3049 		goto out_noref;
3050 #endif
3051 	case SIOCIFCREATE:
3052 	case SIOCIFCREATE2:
3053 		error = priv_check(td, PRIV_NET_IFCREATE);
3054 		if (error == 0)
3055 			error = if_clone_create(ifr->ifr_name,
3056 			    sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ?
3057 			    ifr_data_get_ptr(ifr) : NULL);
3058 		goto out_noref;
3059 	case SIOCIFDESTROY:
3060 		error = priv_check(td, PRIV_NET_IFDESTROY);
3061 		if (error == 0)
3062 			error = if_clone_destroy(ifr->ifr_name);
3063 		goto out_noref;
3064 
3065 	case SIOCIFGCLONERS:
3066 		error = if_clone_list((struct if_clonereq *)data);
3067 		goto out_noref;
3068 
3069 	case CASE_IOC_IFGROUPREQ(SIOCGIFGMEMB):
3070 		error = if_getgroupmembers((struct ifgroupreq *)data);
3071 		goto out_noref;
3072 
3073 #if defined(INET) || defined(INET6)
3074 	case SIOCSVH:
3075 	case SIOCGVH:
3076 		if (carp_ioctl_p == NULL)
3077 			error = EPROTONOSUPPORT;
3078 		else
3079 			error = (*carp_ioctl_p)(ifr, cmd, td);
3080 		goto out_noref;
3081 #endif
3082 	}
3083 
3084 	ifp = ifunit_ref(ifr->ifr_name);
3085 	if (ifp == NULL) {
3086 		error = ENXIO;
3087 		goto out_noref;
3088 	}
3089 
3090 	error = ifhwioctl(cmd, ifp, data, td);
3091 	if (error != ENOIOCTL)
3092 		goto out_ref;
3093 
3094 	oif_flags = ifp->if_flags;
3095 	if (so->so_proto == NULL) {
3096 		error = EOPNOTSUPP;
3097 		goto out_ref;
3098 	}
3099 
3100 	/*
3101 	 * Pass the request on to the socket control method, and if the
3102 	 * latter returns EOPNOTSUPP, directly to the interface.
3103 	 *
3104 	 * Make an exception for the legacy SIOCSIF* requests.  Drivers
3105 	 * trust SIOCSIFADDR et al to come from an already privileged
3106 	 * layer, and do not perform any credentials checks or input
3107 	 * validation.
3108 	 */
3109 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
3110 	    ifp, td));
3111 	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
3112 	    cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
3113 	    cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
3114 		error = (*ifp->if_ioctl)(ifp, cmd, data);
3115 
3116 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
3117 #ifdef INET6
3118 		if (ifp->if_flags & IFF_UP)
3119 			in6_if_up(ifp);
3120 #endif
3121 	}
3122 
3123 out_ref:
3124 	if_rele(ifp);
3125 out_noref:
3126 #ifdef COMPAT_FREEBSD32
3127 	if (ifmrp != NULL) {
3128 		KASSERT((cmd == SIOCGIFMEDIA || cmd == SIOCGIFXMEDIA),
3129 		    ("ifmrp non-NULL, but cmd is not an ifmedia req 0x%lx",
3130 		     cmd));
3131 		data = saved_data;
3132 		ifmr_update(ifmrp, data);
3133 	}
3134 #endif
3135 	CURVNET_RESTORE();
3136 	return (error);
3137 }
3138 
3139 /*
3140  * The code common to handling reference counted flags,
3141  * e.g., in ifpromisc() and if_allmulti().
3142  * The "pflag" argument can specify a permanent mode flag to check,
3143  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
3144  *
3145  * Only to be used on stack-owned flags, not driver-owned flags.
3146  */
3147 static int
3148 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
3149 {
3150 	struct ifreq ifr;
3151 	int error;
3152 	int oldflags, oldcount;
3153 
3154 	/* Sanity checks to catch programming errors */
3155 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
3156 	    ("%s: setting driver-owned flag %d", __func__, flag));
3157 
3158 	if (onswitch)
3159 		KASSERT(*refcount >= 0,
3160 		    ("%s: increment negative refcount %d for flag %d",
3161 		    __func__, *refcount, flag));
3162 	else
3163 		KASSERT(*refcount > 0,
3164 		    ("%s: decrement non-positive refcount %d for flag %d",
3165 		    __func__, *refcount, flag));
3166 
3167 	/* In case this mode is permanent, just touch refcount */
3168 	if (ifp->if_flags & pflag) {
3169 		*refcount += onswitch ? 1 : -1;
3170 		return (0);
3171 	}
3172 
3173 	/* Save ifnet parameters for if_ioctl() may fail */
3174 	oldcount = *refcount;
3175 	oldflags = ifp->if_flags;
3176 
3177 	/*
3178 	 * See if we aren't the only and touching refcount is enough.
3179 	 * Actually toggle interface flag if we are the first or last.
3180 	 */
3181 	if (onswitch) {
3182 		if ((*refcount)++)
3183 			return (0);
3184 		ifp->if_flags |= flag;
3185 	} else {
3186 		if (--(*refcount))
3187 			return (0);
3188 		ifp->if_flags &= ~flag;
3189 	}
3190 
3191 	/* Call down the driver since we've changed interface flags */
3192 	if (ifp->if_ioctl == NULL) {
3193 		error = EOPNOTSUPP;
3194 		goto recover;
3195 	}
3196 	ifr.ifr_flags = ifp->if_flags & 0xffff;
3197 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
3198 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3199 	if (error)
3200 		goto recover;
3201 	/* Notify userland that interface flags have changed */
3202 	rt_ifmsg(ifp);
3203 	return (0);
3204 
3205 recover:
3206 	/* Recover after driver error */
3207 	*refcount = oldcount;
3208 	ifp->if_flags = oldflags;
3209 	return (error);
3210 }
3211 
3212 /*
3213  * Set/clear promiscuous mode on interface ifp based on the truth value
3214  * of pswitch.  The calls are reference counted so that only the first
3215  * "on" request actually has an effect, as does the final "off" request.
3216  * Results are undefined if the "off" and "on" requests are not matched.
3217  */
3218 int
3219 ifpromisc(struct ifnet *ifp, int pswitch)
3220 {
3221 	int error;
3222 	int oldflags = ifp->if_flags;
3223 
3224 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
3225 			   &ifp->if_pcount, pswitch);
3226 	/* If promiscuous mode status has changed, log a message */
3227 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC) &&
3228             log_promisc_mode_change)
3229 		if_printf(ifp, "promiscuous mode %s\n",
3230 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
3231 	return (error);
3232 }
3233 
3234 /*
3235  * Return interface configuration
3236  * of system.  List may be used
3237  * in later ioctl's (above) to get
3238  * other information.
3239  */
3240 /*ARGSUSED*/
3241 static int
3242 ifconf(u_long cmd, caddr_t data)
3243 {
3244 	struct ifconf *ifc = (struct ifconf *)data;
3245 	struct ifnet *ifp;
3246 	struct ifaddr *ifa;
3247 	struct ifreq ifr;
3248 	struct sbuf *sb;
3249 	int error, full = 0, valid_len, max_len;
3250 
3251 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
3252 	max_len = MAXPHYS - 1;
3253 
3254 	/* Prevent hostile input from being able to crash the system */
3255 	if (ifc->ifc_len <= 0)
3256 		return (EINVAL);
3257 
3258 again:
3259 	if (ifc->ifc_len <= max_len) {
3260 		max_len = ifc->ifc_len;
3261 		full = 1;
3262 	}
3263 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
3264 	max_len = 0;
3265 	valid_len = 0;
3266 
3267 	IFNET_RLOCK();
3268 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
3269 		struct epoch_tracker et;
3270 		int addrs;
3271 
3272 		/*
3273 		 * Zero the ifr to make sure we don't disclose the contents
3274 		 * of the stack.
3275 		 */
3276 		memset(&ifr, 0, sizeof(ifr));
3277 
3278 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
3279 		    >= sizeof(ifr.ifr_name)) {
3280 			sbuf_delete(sb);
3281 			IFNET_RUNLOCK();
3282 			return (ENAMETOOLONG);
3283 		}
3284 
3285 		addrs = 0;
3286 		NET_EPOCH_ENTER(et);
3287 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3288 			struct sockaddr *sa = ifa->ifa_addr;
3289 
3290 			if (prison_if(curthread->td_ucred, sa) != 0)
3291 				continue;
3292 			addrs++;
3293 			if (sa->sa_len <= sizeof(*sa)) {
3294 				if (sa->sa_len < sizeof(*sa)) {
3295 					memset(&ifr.ifr_ifru.ifru_addr, 0,
3296 					    sizeof(ifr.ifr_ifru.ifru_addr));
3297 					memcpy(&ifr.ifr_ifru.ifru_addr, sa,
3298 					    sa->sa_len);
3299 				} else
3300 					ifr.ifr_ifru.ifru_addr = *sa;
3301 				sbuf_bcat(sb, &ifr, sizeof(ifr));
3302 				max_len += sizeof(ifr);
3303 			} else {
3304 				sbuf_bcat(sb, &ifr,
3305 				    offsetof(struct ifreq, ifr_addr));
3306 				max_len += offsetof(struct ifreq, ifr_addr);
3307 				sbuf_bcat(sb, sa, sa->sa_len);
3308 				max_len += sa->sa_len;
3309 			}
3310 
3311 			if (sbuf_error(sb) == 0)
3312 				valid_len = sbuf_len(sb);
3313 		}
3314 		NET_EPOCH_EXIT(et);
3315 		if (addrs == 0) {
3316 			sbuf_bcat(sb, &ifr, sizeof(ifr));
3317 			max_len += sizeof(ifr);
3318 
3319 			if (sbuf_error(sb) == 0)
3320 				valid_len = sbuf_len(sb);
3321 		}
3322 	}
3323 	IFNET_RUNLOCK();
3324 
3325 	/*
3326 	 * If we didn't allocate enough space (uncommon), try again.  If
3327 	 * we have already allocated as much space as we are allowed,
3328 	 * return what we've got.
3329 	 */
3330 	if (valid_len != max_len && !full) {
3331 		sbuf_delete(sb);
3332 		goto again;
3333 	}
3334 
3335 	ifc->ifc_len = valid_len;
3336 	sbuf_finish(sb);
3337 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
3338 	sbuf_delete(sb);
3339 	return (error);
3340 }
3341 
3342 /*
3343  * Just like ifpromisc(), but for all-multicast-reception mode.
3344  */
3345 int
3346 if_allmulti(struct ifnet *ifp, int onswitch)
3347 {
3348 
3349 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
3350 }
3351 
3352 struct ifmultiaddr *
3353 if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
3354 {
3355 	struct ifmultiaddr *ifma;
3356 
3357 	IF_ADDR_LOCK_ASSERT(ifp);
3358 
3359 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3360 		if (sa->sa_family == AF_LINK) {
3361 			if (sa_dl_equal(ifma->ifma_addr, sa))
3362 				break;
3363 		} else {
3364 			if (sa_equal(ifma->ifma_addr, sa))
3365 				break;
3366 		}
3367 	}
3368 
3369 	return ifma;
3370 }
3371 
3372 /*
3373  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
3374  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
3375  * the ifnet multicast address list here, so the caller must do that and
3376  * other setup work (such as notifying the device driver).  The reference
3377  * count is initialized to 1.
3378  */
3379 static struct ifmultiaddr *
3380 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
3381     int mflags)
3382 {
3383 	struct ifmultiaddr *ifma;
3384 	struct sockaddr *dupsa;
3385 
3386 	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
3387 	    M_ZERO);
3388 	if (ifma == NULL)
3389 		return (NULL);
3390 
3391 	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
3392 	if (dupsa == NULL) {
3393 		free(ifma, M_IFMADDR);
3394 		return (NULL);
3395 	}
3396 	bcopy(sa, dupsa, sa->sa_len);
3397 	ifma->ifma_addr = dupsa;
3398 
3399 	ifma->ifma_ifp = ifp;
3400 	ifma->ifma_refcount = 1;
3401 	ifma->ifma_protospec = NULL;
3402 
3403 	if (llsa == NULL) {
3404 		ifma->ifma_lladdr = NULL;
3405 		return (ifma);
3406 	}
3407 
3408 	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
3409 	if (dupsa == NULL) {
3410 		free(ifma->ifma_addr, M_IFMADDR);
3411 		free(ifma, M_IFMADDR);
3412 		return (NULL);
3413 	}
3414 	bcopy(llsa, dupsa, llsa->sa_len);
3415 	ifma->ifma_lladdr = dupsa;
3416 
3417 	return (ifma);
3418 }
3419 
3420 /*
3421  * if_freemulti: free ifmultiaddr structure and possibly attached related
3422  * addresses.  The caller is responsible for implementing reference
3423  * counting, notifying the driver, handling routing messages, and releasing
3424  * any dependent link layer state.
3425  */
3426 #ifdef MCAST_VERBOSE
3427 extern void kdb_backtrace(void);
3428 #endif
3429 static void
3430 if_freemulti_internal(struct ifmultiaddr *ifma)
3431 {
3432 
3433 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
3434 	    ifma->ifma_refcount));
3435 
3436 	if (ifma->ifma_lladdr != NULL)
3437 		free(ifma->ifma_lladdr, M_IFMADDR);
3438 #ifdef MCAST_VERBOSE
3439 	kdb_backtrace();
3440 	printf("%s freeing ifma: %p\n", __func__, ifma);
3441 #endif
3442 	free(ifma->ifma_addr, M_IFMADDR);
3443 	free(ifma, M_IFMADDR);
3444 }
3445 
3446 static void
3447 if_destroymulti(epoch_context_t ctx)
3448 {
3449 	struct ifmultiaddr *ifma;
3450 
3451 	ifma = __containerof(ctx, struct ifmultiaddr, ifma_epoch_ctx);
3452 	if_freemulti_internal(ifma);
3453 }
3454 
3455 void
3456 if_freemulti(struct ifmultiaddr *ifma)
3457 {
3458 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti_epoch: refcount %d",
3459 	    ifma->ifma_refcount));
3460 
3461 	epoch_call(net_epoch_preempt, &ifma->ifma_epoch_ctx, if_destroymulti);
3462 }
3463 
3464 
3465 /*
3466  * Register an additional multicast address with a network interface.
3467  *
3468  * - If the address is already present, bump the reference count on the
3469  *   address and return.
3470  * - If the address is not link-layer, look up a link layer address.
3471  * - Allocate address structures for one or both addresses, and attach to the
3472  *   multicast address list on the interface.  If automatically adding a link
3473  *   layer address, the protocol address will own a reference to the link
3474  *   layer address, to be freed when it is freed.
3475  * - Notify the network device driver of an addition to the multicast address
3476  *   list.
3477  *
3478  * 'sa' points to caller-owned memory with the desired multicast address.
3479  *
3480  * 'retifma' will be used to return a pointer to the resulting multicast
3481  * address reference, if desired.
3482  */
3483 int
3484 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
3485     struct ifmultiaddr **retifma)
3486 {
3487 	struct ifmultiaddr *ifma, *ll_ifma;
3488 	struct sockaddr *llsa;
3489 	struct sockaddr_dl sdl;
3490 	int error;
3491 
3492 #ifdef INET
3493 	IN_MULTI_LIST_UNLOCK_ASSERT();
3494 #endif
3495 #ifdef INET6
3496 	IN6_MULTI_LIST_UNLOCK_ASSERT();
3497 #endif
3498 	/*
3499 	 * If the address is already present, return a new reference to it;
3500 	 * otherwise, allocate storage and set up a new address.
3501 	 */
3502 	IF_ADDR_WLOCK(ifp);
3503 	ifma = if_findmulti(ifp, sa);
3504 	if (ifma != NULL) {
3505 		ifma->ifma_refcount++;
3506 		if (retifma != NULL)
3507 			*retifma = ifma;
3508 		IF_ADDR_WUNLOCK(ifp);
3509 		return (0);
3510 	}
3511 
3512 	/*
3513 	 * The address isn't already present; resolve the protocol address
3514 	 * into a link layer address, and then look that up, bump its
3515 	 * refcount or allocate an ifma for that also.
3516 	 * Most link layer resolving functions returns address data which
3517 	 * fits inside default sockaddr_dl structure. However callback
3518 	 * can allocate another sockaddr structure, in that case we need to
3519 	 * free it later.
3520 	 */
3521 	llsa = NULL;
3522 	ll_ifma = NULL;
3523 	if (ifp->if_resolvemulti != NULL) {
3524 		/* Provide called function with buffer size information */
3525 		sdl.sdl_len = sizeof(sdl);
3526 		llsa = (struct sockaddr *)&sdl;
3527 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
3528 		if (error)
3529 			goto unlock_out;
3530 	}
3531 
3532 	/*
3533 	 * Allocate the new address.  Don't hook it up yet, as we may also
3534 	 * need to allocate a link layer multicast address.
3535 	 */
3536 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
3537 	if (ifma == NULL) {
3538 		error = ENOMEM;
3539 		goto free_llsa_out;
3540 	}
3541 
3542 	/*
3543 	 * If a link layer address is found, we'll need to see if it's
3544 	 * already present in the address list, or allocate is as well.
3545 	 * When this block finishes, the link layer address will be on the
3546 	 * list.
3547 	 */
3548 	if (llsa != NULL) {
3549 		ll_ifma = if_findmulti(ifp, llsa);
3550 		if (ll_ifma == NULL) {
3551 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3552 			if (ll_ifma == NULL) {
3553 				--ifma->ifma_refcount;
3554 				if_freemulti(ifma);
3555 				error = ENOMEM;
3556 				goto free_llsa_out;
3557 			}
3558 			ll_ifma->ifma_flags |= IFMA_F_ENQUEUED;
3559 			CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3560 			    ifma_link);
3561 		} else
3562 			ll_ifma->ifma_refcount++;
3563 		ifma->ifma_llifma = ll_ifma;
3564 	}
3565 
3566 	/*
3567 	 * We now have a new multicast address, ifma, and possibly a new or
3568 	 * referenced link layer address.  Add the primary address to the
3569 	 * ifnet address list.
3570 	 */
3571 	ifma->ifma_flags |= IFMA_F_ENQUEUED;
3572 	CK_STAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3573 
3574 	if (retifma != NULL)
3575 		*retifma = ifma;
3576 
3577 	/*
3578 	 * Must generate the message while holding the lock so that 'ifma'
3579 	 * pointer is still valid.
3580 	 */
3581 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3582 	IF_ADDR_WUNLOCK(ifp);
3583 
3584 	/*
3585 	 * We are certain we have added something, so call down to the
3586 	 * interface to let them know about it.
3587 	 */
3588 	if (ifp->if_ioctl != NULL) {
3589 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3590 	}
3591 
3592 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3593 		link_free_sdl(llsa);
3594 
3595 	return (0);
3596 
3597 free_llsa_out:
3598 	if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
3599 		link_free_sdl(llsa);
3600 
3601 unlock_out:
3602 	IF_ADDR_WUNLOCK(ifp);
3603 	return (error);
3604 }
3605 
3606 /*
3607  * Delete a multicast group membership by network-layer group address.
3608  *
3609  * Returns ENOENT if the entry could not be found. If ifp no longer
3610  * exists, results are undefined. This entry point should only be used
3611  * from subsystems which do appropriate locking to hold ifp for the
3612  * duration of the call.
3613  * Network-layer protocol domains must use if_delmulti_ifma().
3614  */
3615 int
3616 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3617 {
3618 	struct ifmultiaddr *ifma;
3619 	int lastref;
3620 #ifdef INVARIANTS
3621 	struct epoch_tracker et;
3622 	struct ifnet *oifp;
3623 
3624 	NET_EPOCH_ENTER(et);
3625 	CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link)
3626 		if (ifp == oifp)
3627 			break;
3628 	if (ifp != oifp)
3629 		ifp = NULL;
3630 	NET_EPOCH_EXIT(et);
3631 
3632 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3633 #endif
3634 	if (ifp == NULL)
3635 		return (ENOENT);
3636 
3637 	IF_ADDR_WLOCK(ifp);
3638 	lastref = 0;
3639 	ifma = if_findmulti(ifp, sa);
3640 	if (ifma != NULL)
3641 		lastref = if_delmulti_locked(ifp, ifma, 0);
3642 	IF_ADDR_WUNLOCK(ifp);
3643 
3644 	if (ifma == NULL)
3645 		return (ENOENT);
3646 
3647 	if (lastref && ifp->if_ioctl != NULL) {
3648 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3649 	}
3650 
3651 	return (0);
3652 }
3653 
3654 /*
3655  * Delete all multicast group membership for an interface.
3656  * Should be used to quickly flush all multicast filters.
3657  */
3658 void
3659 if_delallmulti(struct ifnet *ifp)
3660 {
3661 	struct ifmultiaddr *ifma;
3662 	struct ifmultiaddr *next;
3663 
3664 	IF_ADDR_WLOCK(ifp);
3665 	CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3666 		if_delmulti_locked(ifp, ifma, 0);
3667 	IF_ADDR_WUNLOCK(ifp);
3668 }
3669 
3670 void
3671 if_delmulti_ifma(struct ifmultiaddr *ifma)
3672 {
3673 	if_delmulti_ifma_flags(ifma, 0);
3674 }
3675 
3676 /*
3677  * Delete a multicast group membership by group membership pointer.
3678  * Network-layer protocol domains must use this routine.
3679  *
3680  * It is safe to call this routine if the ifp disappeared.
3681  */
3682 void
3683 if_delmulti_ifma_flags(struct ifmultiaddr *ifma, int flags)
3684 {
3685 	struct ifnet *ifp;
3686 	int lastref;
3687 	MCDPRINTF("%s freeing ifma: %p\n", __func__, ifma);
3688 #ifdef INET
3689 	IN_MULTI_LIST_UNLOCK_ASSERT();
3690 #endif
3691 	ifp = ifma->ifma_ifp;
3692 #ifdef DIAGNOSTIC
3693 	if (ifp == NULL) {
3694 		printf("%s: ifma_ifp seems to be detached\n", __func__);
3695 	} else {
3696 		struct epoch_tracker et;
3697 		struct ifnet *oifp;
3698 
3699 		NET_EPOCH_ENTER(et);
3700 		CK_STAILQ_FOREACH(oifp, &V_ifnet, if_link)
3701 			if (ifp == oifp)
3702 				break;
3703 		if (ifp != oifp)
3704 			ifp = NULL;
3705 		NET_EPOCH_EXIT(et);
3706 	}
3707 #endif
3708 	/*
3709 	 * If and only if the ifnet instance exists: Acquire the address lock.
3710 	 */
3711 	if (ifp != NULL)
3712 		IF_ADDR_WLOCK(ifp);
3713 
3714 	lastref = if_delmulti_locked(ifp, ifma, flags);
3715 
3716 	if (ifp != NULL) {
3717 		/*
3718 		 * If and only if the ifnet instance exists:
3719 		 *  Release the address lock.
3720 		 *  If the group was left: update the hardware hash filter.
3721 		 */
3722 		IF_ADDR_WUNLOCK(ifp);
3723 		if (lastref && ifp->if_ioctl != NULL) {
3724 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3725 		}
3726 	}
3727 }
3728 
3729 /*
3730  * Perform deletion of network-layer and/or link-layer multicast address.
3731  *
3732  * Return 0 if the reference count was decremented.
3733  * Return 1 if the final reference was released, indicating that the
3734  * hardware hash filter should be reprogrammed.
3735  */
3736 static int
3737 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3738 {
3739 	struct ifmultiaddr *ll_ifma;
3740 
3741 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3742 		KASSERT(ifma->ifma_ifp == ifp,
3743 		    ("%s: inconsistent ifp %p", __func__, ifp));
3744 		IF_ADDR_WLOCK_ASSERT(ifp);
3745 	}
3746 
3747 	ifp = ifma->ifma_ifp;
3748 	MCDPRINTF("%s freeing %p from %s \n", __func__, ifma, ifp ? ifp->if_xname : "");
3749 
3750 	/*
3751 	 * If the ifnet is detaching, null out references to ifnet,
3752 	 * so that upper protocol layers will notice, and not attempt
3753 	 * to obtain locks for an ifnet which no longer exists. The
3754 	 * routing socket announcement must happen before the ifnet
3755 	 * instance is detached from the system.
3756 	 */
3757 	if (detaching) {
3758 #ifdef DIAGNOSTIC
3759 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3760 #endif
3761 		/*
3762 		 * ifp may already be nulled out if we are being reentered
3763 		 * to delete the ll_ifma.
3764 		 */
3765 		if (ifp != NULL) {
3766 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3767 			ifma->ifma_ifp = NULL;
3768 		}
3769 	}
3770 
3771 	if (--ifma->ifma_refcount > 0)
3772 		return 0;
3773 
3774 	if (ifp != NULL && detaching == 0 && (ifma->ifma_flags & IFMA_F_ENQUEUED)) {
3775 		CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifmultiaddr, ifma_link);
3776 		ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3777 	}
3778 	/*
3779 	 * If this ifma is a network-layer ifma, a link-layer ifma may
3780 	 * have been associated with it. Release it first if so.
3781 	 */
3782 	ll_ifma = ifma->ifma_llifma;
3783 	if (ll_ifma != NULL) {
3784 		KASSERT(ifma->ifma_lladdr != NULL,
3785 		    ("%s: llifma w/o lladdr", __func__));
3786 		if (detaching)
3787 			ll_ifma->ifma_ifp = NULL;	/* XXX */
3788 		if (--ll_ifma->ifma_refcount == 0) {
3789 			if (ifp != NULL) {
3790 				if (ll_ifma->ifma_flags & IFMA_F_ENQUEUED) {
3791 					CK_STAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifmultiaddr,
3792 						ifma_link);
3793 					ll_ifma->ifma_flags &= ~IFMA_F_ENQUEUED;
3794 				}
3795 			}
3796 			if_freemulti(ll_ifma);
3797 		}
3798 	}
3799 #ifdef INVARIANTS
3800 	if (ifp) {
3801 		struct ifmultiaddr *ifmatmp;
3802 
3803 		CK_STAILQ_FOREACH(ifmatmp, &ifp->if_multiaddrs, ifma_link)
3804 			MPASS(ifma != ifmatmp);
3805 	}
3806 #endif
3807 	if_freemulti(ifma);
3808 	/*
3809 	 * The last reference to this instance of struct ifmultiaddr
3810 	 * was released; the hardware should be notified of this change.
3811 	 */
3812 	return 1;
3813 }
3814 
3815 /*
3816  * Set the link layer address on an interface.
3817  *
3818  * At this time we only support certain types of interfaces,
3819  * and we don't allow the length of the address to change.
3820  *
3821  * Set noinline to be dtrace-friendly
3822  */
3823 __noinline int
3824 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3825 {
3826 	struct sockaddr_dl *sdl;
3827 	struct ifaddr *ifa;
3828 	struct ifreq ifr;
3829 	struct epoch_tracker et;
3830 	int rc;
3831 
3832 	rc = 0;
3833 	NET_EPOCH_ENTER(et);
3834 	ifa = ifp->if_addr;
3835 	if (ifa == NULL) {
3836 		rc = EINVAL;
3837 		goto out;
3838 	}
3839 
3840 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3841 	if (sdl == NULL) {
3842 		rc = EINVAL;
3843 		goto out;
3844 	}
3845 	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3846 		rc = EINVAL;
3847 		goto out;
3848 	}
3849 	switch (ifp->if_type) {
3850 	case IFT_ETHER:
3851 	case IFT_XETHER:
3852 	case IFT_L2VLAN:
3853 	case IFT_BRIDGE:
3854 	case IFT_IEEE8023ADLAG:
3855 		bcopy(lladdr, LLADDR(sdl), len);
3856 		break;
3857 	default:
3858 		rc = ENODEV;
3859 		goto out;
3860 	}
3861 
3862 	/*
3863 	 * If the interface is already up, we need
3864 	 * to re-init it in order to reprogram its
3865 	 * address filter.
3866 	 */
3867 	NET_EPOCH_EXIT(et);
3868 	if ((ifp->if_flags & IFF_UP) != 0) {
3869 		if (ifp->if_ioctl) {
3870 			ifp->if_flags &= ~IFF_UP;
3871 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3872 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3873 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3874 			ifp->if_flags |= IFF_UP;
3875 			ifr.ifr_flags = ifp->if_flags & 0xffff;
3876 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3877 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3878 		}
3879 	}
3880 	EVENTHANDLER_INVOKE(iflladdr_event, ifp);
3881 	return (0);
3882  out:
3883 	NET_EPOCH_EXIT(et);
3884 	return (rc);
3885 }
3886 
3887 /*
3888  * Compat function for handling basic encapsulation requests.
3889  * Not converted stacks (FDDI, IB, ..) supports traditional
3890  * output model: ARP (and other similar L2 protocols) are handled
3891  * inside output routine, arpresolve/nd6_resolve() returns MAC
3892  * address instead of full prepend.
3893  *
3894  * This function creates calculated header==MAC for IPv4/IPv6 and
3895  * returns EAFNOSUPPORT (which is then handled in ARP code) for other
3896  * address families.
3897  */
3898 static int
3899 if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
3900 {
3901 
3902 	if (req->rtype != IFENCAP_LL)
3903 		return (EOPNOTSUPP);
3904 
3905 	if (req->bufsize < req->lladdr_len)
3906 		return (ENOMEM);
3907 
3908 	switch (req->family) {
3909 	case AF_INET:
3910 	case AF_INET6:
3911 		break;
3912 	default:
3913 		return (EAFNOSUPPORT);
3914 	}
3915 
3916 	/* Copy lladdr to storage as is */
3917 	memmove(req->buf, req->lladdr, req->lladdr_len);
3918 	req->bufsize = req->lladdr_len;
3919 	req->lladdr_off = 0;
3920 
3921 	return (0);
3922 }
3923 
3924 /*
3925  * Tunnel interfaces can nest, also they may cause infinite recursion
3926  * calls when misconfigured. We'll prevent this by detecting loops.
3927  * High nesting level may cause stack exhaustion. We'll prevent this
3928  * by introducing upper limit.
3929  *
3930  * Return 0, if tunnel nesting count is equal or less than limit.
3931  */
3932 int
3933 if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, uint32_t cookie,
3934     int limit)
3935 {
3936 	struct m_tag *mtag;
3937 	int count;
3938 
3939 	count = 1;
3940 	mtag = NULL;
3941 	while ((mtag = m_tag_locate(m, cookie, 0, mtag)) != NULL) {
3942 		if (*(struct ifnet **)(mtag + 1) == ifp) {
3943 			log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp));
3944 			return (EIO);
3945 		}
3946 		count++;
3947 	}
3948 	if (count > limit) {
3949 		log(LOG_NOTICE,
3950 		    "%s: if_output recursively called too many times(%d)\n",
3951 		    if_name(ifp), count);
3952 		return (EIO);
3953 	}
3954 	mtag = m_tag_alloc(cookie, 0, sizeof(struct ifnet *), M_NOWAIT);
3955 	if (mtag == NULL)
3956 		return (ENOMEM);
3957 	*(struct ifnet **)(mtag + 1) = ifp;
3958 	m_tag_prepend(m, mtag);
3959 	return (0);
3960 }
3961 
3962 /*
3963  * Get the link layer address that was read from the hardware at attach.
3964  *
3965  * This is only set by Ethernet NICs (IFT_ETHER), but laggX interfaces re-type
3966  * their component interfaces as IFT_IEEE8023ADLAG.
3967  */
3968 int
3969 if_gethwaddr(struct ifnet *ifp, struct ifreq *ifr)
3970 {
3971 
3972 	if (ifp->if_hw_addr == NULL)
3973 		return (ENODEV);
3974 
3975 	switch (ifp->if_type) {
3976 	case IFT_ETHER:
3977 	case IFT_IEEE8023ADLAG:
3978 		bcopy(ifp->if_hw_addr, ifr->ifr_addr.sa_data, ifp->if_addrlen);
3979 		return (0);
3980 	default:
3981 		return (ENODEV);
3982 	}
3983 }
3984 
3985 /*
3986  * The name argument must be a pointer to storage which will last as
3987  * long as the interface does.  For physical devices, the result of
3988  * device_get_name(dev) is a good choice and for pseudo-devices a
3989  * static string works well.
3990  */
3991 void
3992 if_initname(struct ifnet *ifp, const char *name, int unit)
3993 {
3994 	ifp->if_dname = name;
3995 	ifp->if_dunit = unit;
3996 	if (unit != IF_DUNIT_NONE)
3997 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3998 	else
3999 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
4000 }
4001 
4002 int
4003 if_printf(struct ifnet *ifp, const char *fmt, ...)
4004 {
4005 	char if_fmt[256];
4006 	va_list ap;
4007 
4008 	snprintf(if_fmt, sizeof(if_fmt), "%s: %s", ifp->if_xname, fmt);
4009 	va_start(ap, fmt);
4010 	vlog(LOG_INFO, if_fmt, ap);
4011 	va_end(ap);
4012 	return (0);
4013 }
4014 
4015 void
4016 if_start(struct ifnet *ifp)
4017 {
4018 
4019 	(*(ifp)->if_start)(ifp);
4020 }
4021 
4022 /*
4023  * Backwards compatibility interface for drivers
4024  * that have not implemented it
4025  */
4026 static int
4027 if_transmit(struct ifnet *ifp, struct mbuf *m)
4028 {
4029 	int error;
4030 
4031 	IFQ_HANDOFF(ifp, m, error);
4032 	return (error);
4033 }
4034 
4035 static void
4036 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
4037 {
4038 
4039 	m_freem(m);
4040 }
4041 
4042 int
4043 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
4044 {
4045 	int active = 0;
4046 
4047 	IF_LOCK(ifq);
4048 	if (_IF_QFULL(ifq)) {
4049 		IF_UNLOCK(ifq);
4050 		if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
4051 		m_freem(m);
4052 		return (0);
4053 	}
4054 	if (ifp != NULL) {
4055 		if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
4056 		if (m->m_flags & (M_BCAST|M_MCAST))
4057 			if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
4058 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
4059 	}
4060 	_IF_ENQUEUE(ifq, m);
4061 	IF_UNLOCK(ifq);
4062 	if (ifp != NULL && !active)
4063 		(*(ifp)->if_start)(ifp);
4064 	return (1);
4065 }
4066 
4067 void
4068 if_register_com_alloc(u_char type,
4069     if_com_alloc_t *a, if_com_free_t *f)
4070 {
4071 
4072 	KASSERT(if_com_alloc[type] == NULL,
4073 	    ("if_register_com_alloc: %d already registered", type));
4074 	KASSERT(if_com_free[type] == NULL,
4075 	    ("if_register_com_alloc: %d free already registered", type));
4076 
4077 	if_com_alloc[type] = a;
4078 	if_com_free[type] = f;
4079 }
4080 
4081 void
4082 if_deregister_com_alloc(u_char type)
4083 {
4084 
4085 	KASSERT(if_com_alloc[type] != NULL,
4086 	    ("if_deregister_com_alloc: %d not registered", type));
4087 	KASSERT(if_com_free[type] != NULL,
4088 	    ("if_deregister_com_alloc: %d free not registered", type));
4089 	if_com_alloc[type] = NULL;
4090 	if_com_free[type] = NULL;
4091 }
4092 
4093 /* API for driver access to network stack owned ifnet.*/
4094 uint64_t
4095 if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
4096 {
4097 	uint64_t oldbrate;
4098 
4099 	oldbrate = ifp->if_baudrate;
4100 	ifp->if_baudrate = baudrate;
4101 	return (oldbrate);
4102 }
4103 
4104 uint64_t
4105 if_getbaudrate(if_t ifp)
4106 {
4107 
4108 	return (((struct ifnet *)ifp)->if_baudrate);
4109 }
4110 
4111 int
4112 if_setcapabilities(if_t ifp, int capabilities)
4113 {
4114 	((struct ifnet *)ifp)->if_capabilities = capabilities;
4115 	return (0);
4116 }
4117 
4118 int
4119 if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
4120 {
4121 	((struct ifnet *)ifp)->if_capabilities |= setbit;
4122 	((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
4123 
4124 	return (0);
4125 }
4126 
4127 int
4128 if_getcapabilities(if_t ifp)
4129 {
4130 	return ((struct ifnet *)ifp)->if_capabilities;
4131 }
4132 
4133 int
4134 if_setcapenable(if_t ifp, int capabilities)
4135 {
4136 	((struct ifnet *)ifp)->if_capenable = capabilities;
4137 	return (0);
4138 }
4139 
4140 int
4141 if_setcapenablebit(if_t ifp, int setcap, int clearcap)
4142 {
4143 	if(setcap)
4144 		((struct ifnet *)ifp)->if_capenable |= setcap;
4145 	if(clearcap)
4146 		((struct ifnet *)ifp)->if_capenable &= ~clearcap;
4147 
4148 	return (0);
4149 }
4150 
4151 const char *
4152 if_getdname(if_t ifp)
4153 {
4154 	return ((struct ifnet *)ifp)->if_dname;
4155 }
4156 
4157 int
4158 if_togglecapenable(if_t ifp, int togglecap)
4159 {
4160 	((struct ifnet *)ifp)->if_capenable ^= togglecap;
4161 	return (0);
4162 }
4163 
4164 int
4165 if_getcapenable(if_t ifp)
4166 {
4167 	return ((struct ifnet *)ifp)->if_capenable;
4168 }
4169 
4170 /*
4171  * This is largely undesirable because it ties ifnet to a device, but does
4172  * provide flexiblity for an embedded product vendor. Should be used with
4173  * the understanding that it violates the interface boundaries, and should be
4174  * a last resort only.
4175  */
4176 int
4177 if_setdev(if_t ifp, void *dev)
4178 {
4179 	return (0);
4180 }
4181 
4182 int
4183 if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
4184 {
4185 	((struct ifnet *)ifp)->if_drv_flags |= set_flags;
4186 	((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
4187 
4188 	return (0);
4189 }
4190 
4191 int
4192 if_getdrvflags(if_t ifp)
4193 {
4194 	return ((struct ifnet *)ifp)->if_drv_flags;
4195 }
4196 
4197 int
4198 if_setdrvflags(if_t ifp, int flags)
4199 {
4200 	((struct ifnet *)ifp)->if_drv_flags = flags;
4201 	return (0);
4202 }
4203 
4204 
4205 int
4206 if_setflags(if_t ifp, int flags)
4207 {
4208 	((struct ifnet *)ifp)->if_flags = flags;
4209 	return (0);
4210 }
4211 
4212 int
4213 if_setflagbits(if_t ifp, int set, int clear)
4214 {
4215 	((struct ifnet *)ifp)->if_flags |= set;
4216 	((struct ifnet *)ifp)->if_flags &= ~clear;
4217 
4218 	return (0);
4219 }
4220 
4221 int
4222 if_getflags(if_t ifp)
4223 {
4224 	return ((struct ifnet *)ifp)->if_flags;
4225 }
4226 
4227 int
4228 if_clearhwassist(if_t ifp)
4229 {
4230 	((struct ifnet *)ifp)->if_hwassist = 0;
4231 	return (0);
4232 }
4233 
4234 int
4235 if_sethwassistbits(if_t ifp, int toset, int toclear)
4236 {
4237 	((struct ifnet *)ifp)->if_hwassist |= toset;
4238 	((struct ifnet *)ifp)->if_hwassist &= ~toclear;
4239 
4240 	return (0);
4241 }
4242 
4243 int
4244 if_sethwassist(if_t ifp, int hwassist_bit)
4245 {
4246 	((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
4247 	return (0);
4248 }
4249 
4250 int
4251 if_gethwassist(if_t ifp)
4252 {
4253 	return ((struct ifnet *)ifp)->if_hwassist;
4254 }
4255 
4256 int
4257 if_setmtu(if_t ifp, int mtu)
4258 {
4259 	((struct ifnet *)ifp)->if_mtu = mtu;
4260 	return (0);
4261 }
4262 
4263 int
4264 if_getmtu(if_t ifp)
4265 {
4266 	return ((struct ifnet *)ifp)->if_mtu;
4267 }
4268 
4269 int
4270 if_getmtu_family(if_t ifp, int family)
4271 {
4272 	struct domain *dp;
4273 
4274 	for (dp = domains; dp; dp = dp->dom_next) {
4275 		if (dp->dom_family == family && dp->dom_ifmtu != NULL)
4276 			return (dp->dom_ifmtu((struct ifnet *)ifp));
4277 	}
4278 
4279 	return (((struct ifnet *)ifp)->if_mtu);
4280 }
4281 
4282 int
4283 if_setsoftc(if_t ifp, void *softc)
4284 {
4285 	((struct ifnet *)ifp)->if_softc = softc;
4286 	return (0);
4287 }
4288 
4289 void *
4290 if_getsoftc(if_t ifp)
4291 {
4292 	return ((struct ifnet *)ifp)->if_softc;
4293 }
4294 
4295 void
4296 if_setrcvif(struct mbuf *m, if_t ifp)
4297 {
4298 
4299 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
4300 	m->m_pkthdr.rcvif = (struct ifnet *)ifp;
4301 }
4302 
4303 void
4304 if_setvtag(struct mbuf *m, uint16_t tag)
4305 {
4306 	m->m_pkthdr.ether_vtag = tag;
4307 }
4308 
4309 uint16_t
4310 if_getvtag(struct mbuf *m)
4311 {
4312 
4313 	return (m->m_pkthdr.ether_vtag);
4314 }
4315 
4316 int
4317 if_sendq_empty(if_t ifp)
4318 {
4319 	return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
4320 }
4321 
4322 struct ifaddr *
4323 if_getifaddr(if_t ifp)
4324 {
4325 	return ((struct ifnet *)ifp)->if_addr;
4326 }
4327 
4328 int
4329 if_getamcount(if_t ifp)
4330 {
4331 	return ((struct ifnet *)ifp)->if_amcount;
4332 }
4333 
4334 
4335 int
4336 if_setsendqready(if_t ifp)
4337 {
4338 	IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
4339 	return (0);
4340 }
4341 
4342 int
4343 if_setsendqlen(if_t ifp, int tx_desc_count)
4344 {
4345 	IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
4346 	((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
4347 
4348 	return (0);
4349 }
4350 
4351 int
4352 if_vlantrunkinuse(if_t ifp)
4353 {
4354 	return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
4355 }
4356 
4357 int
4358 if_input(if_t ifp, struct mbuf* sendmp)
4359 {
4360 	(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
4361 	return (0);
4362 
4363 }
4364 
4365 /* XXX */
4366 #ifndef ETH_ADDR_LEN
4367 #define ETH_ADDR_LEN 6
4368 #endif
4369 
4370 int
4371 if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
4372 {
4373 	struct ifmultiaddr *ifma;
4374 	uint8_t *lmta = (uint8_t *)mta;
4375 	int mcnt = 0;
4376 
4377 	CK_STAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4378 		if (ifma->ifma_addr->sa_family != AF_LINK)
4379 			continue;
4380 
4381 		if (mcnt == max)
4382 			break;
4383 
4384 		bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
4385 		    &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
4386 		mcnt++;
4387 	}
4388 	*cnt = mcnt;
4389 
4390 	return (0);
4391 }
4392 
4393 int
4394 if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
4395 {
4396 	int error;
4397 
4398 	if_maddr_rlock(ifp);
4399 	error = if_setupmultiaddr(ifp, mta, cnt, max);
4400 	if_maddr_runlock(ifp);
4401 	return (error);
4402 }
4403 
4404 int
4405 if_multiaddr_count(if_t ifp, int max)
4406 {
4407 	struct ifmultiaddr *ifma;
4408 	int count;
4409 
4410 	count = 0;
4411 	if_maddr_rlock(ifp);
4412 	CK_STAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
4413 		if (ifma->ifma_addr->sa_family != AF_LINK)
4414 			continue;
4415 		count++;
4416 		if (count == max)
4417 			break;
4418 	}
4419 	if_maddr_runlock(ifp);
4420 	return (count);
4421 }
4422 
4423 int
4424 if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg)
4425 {
4426 	struct ifmultiaddr *ifma;
4427 	int cnt = 0;
4428 
4429 	if_maddr_rlock(ifp);
4430 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
4431 		cnt += filter(arg, ifma, cnt);
4432 	if_maddr_runlock(ifp);
4433 	return (cnt);
4434 }
4435 
4436 struct mbuf *
4437 if_dequeue(if_t ifp)
4438 {
4439 	struct mbuf *m;
4440 	IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
4441 
4442 	return (m);
4443 }
4444 
4445 int
4446 if_sendq_prepend(if_t ifp, struct mbuf *m)
4447 {
4448 	IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
4449 	return (0);
4450 }
4451 
4452 int
4453 if_setifheaderlen(if_t ifp, int len)
4454 {
4455 	((struct ifnet *)ifp)->if_hdrlen = len;
4456 	return (0);
4457 }
4458 
4459 caddr_t
4460 if_getlladdr(if_t ifp)
4461 {
4462 	return (IF_LLADDR((struct ifnet *)ifp));
4463 }
4464 
4465 void *
4466 if_gethandle(u_char type)
4467 {
4468 	return (if_alloc(type));
4469 }
4470 
4471 void
4472 if_bpfmtap(if_t ifh, struct mbuf *m)
4473 {
4474 	struct ifnet *ifp = (struct ifnet *)ifh;
4475 
4476 	BPF_MTAP(ifp, m);
4477 }
4478 
4479 void
4480 if_etherbpfmtap(if_t ifh, struct mbuf *m)
4481 {
4482 	struct ifnet *ifp = (struct ifnet *)ifh;
4483 
4484 	ETHER_BPF_MTAP(ifp, m);
4485 }
4486 
4487 void
4488 if_vlancap(if_t ifh)
4489 {
4490 	struct ifnet *ifp = (struct ifnet *)ifh;
4491 	VLAN_CAPABILITIES(ifp);
4492 }
4493 
4494 int
4495 if_sethwtsomax(if_t ifp, u_int if_hw_tsomax)
4496 {
4497 
4498 	((struct ifnet *)ifp)->if_hw_tsomax = if_hw_tsomax;
4499         return (0);
4500 }
4501 
4502 int
4503 if_sethwtsomaxsegcount(if_t ifp, u_int if_hw_tsomaxsegcount)
4504 {
4505 
4506 	((struct ifnet *)ifp)->if_hw_tsomaxsegcount = if_hw_tsomaxsegcount;
4507         return (0);
4508 }
4509 
4510 int
4511 if_sethwtsomaxsegsize(if_t ifp, u_int if_hw_tsomaxsegsize)
4512 {
4513 
4514 	((struct ifnet *)ifp)->if_hw_tsomaxsegsize = if_hw_tsomaxsegsize;
4515         return (0);
4516 }
4517 
4518 u_int
4519 if_gethwtsomax(if_t ifp)
4520 {
4521 
4522 	return (((struct ifnet *)ifp)->if_hw_tsomax);
4523 }
4524 
4525 u_int
4526 if_gethwtsomaxsegcount(if_t ifp)
4527 {
4528 
4529 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegcount);
4530 }
4531 
4532 u_int
4533 if_gethwtsomaxsegsize(if_t ifp)
4534 {
4535 
4536 	return (((struct ifnet *)ifp)->if_hw_tsomaxsegsize);
4537 }
4538 
4539 void
4540 if_setinitfn(if_t ifp, void (*init_fn)(void *))
4541 {
4542 	((struct ifnet *)ifp)->if_init = init_fn;
4543 }
4544 
4545 void
4546 if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
4547 {
4548 	((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
4549 }
4550 
4551 void
4552 if_setstartfn(if_t ifp, void (*start_fn)(if_t))
4553 {
4554 	((struct ifnet *)ifp)->if_start = (void *)start_fn;
4555 }
4556 
4557 void
4558 if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
4559 {
4560 	((struct ifnet *)ifp)->if_transmit = start_fn;
4561 }
4562 
4563 void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
4564 {
4565 	((struct ifnet *)ifp)->if_qflush = flush_fn;
4566 
4567 }
4568 
4569 void
4570 if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
4571 {
4572 
4573 	ifp->if_get_counter = fn;
4574 }
4575 
4576 /* Revisit these - These are inline functions originally. */
4577 int
4578 drbr_inuse_drv(if_t ifh, struct buf_ring *br)
4579 {
4580 	return drbr_inuse(ifh, br);
4581 }
4582 
4583 struct mbuf*
4584 drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
4585 {
4586 	return drbr_dequeue(ifh, br);
4587 }
4588 
4589 int
4590 drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
4591 {
4592 	return drbr_needs_enqueue(ifh, br);
4593 }
4594 
4595 int
4596 drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
4597 {
4598 	return drbr_enqueue(ifh, br, m);
4599 
4600 }
4601