1/*
2 * Copyright (c) 2000-2012 Apple Inc. All rights reserved.
3 *
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5 *
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
14 *
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
17 *
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
25 *
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27 */
28/*
29 * Copyright (c) 1980, 1986, 1993
30 *	The Regents of the University of California.  All rights reserved.
31 *
32 * Redistribution and use in source and binary forms, with or without
33 * modification, are permitted provided that the following conditions
34 * are met:
35 * 1. Redistributions of source code must retain the above copyright
36 *    notice, this list of conditions and the following disclaimer.
37 * 2. Redistributions in binary form must reproduce the above copyright
38 *    notice, this list of conditions and the following disclaimer in the
39 *    documentation and/or other materials provided with the distribution.
40 * 3. All advertising materials mentioning features or use of this software
41 *    must display the following acknowledgement:
42 *	This product includes software developed by the University of
43 *	California, Berkeley and its contributors.
44 * 4. Neither the name of the University nor the names of its contributors
45 *    may be used to endorse or promote products derived from this software
46 *    without specific prior written permission.
47 *
48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58 * SUCH DAMAGE.
59 *
60 *	@(#)if.c	8.3 (Berkeley) 1/4/94
61 * $FreeBSD: src/sys/net/if.c,v 1.85.2.9 2001/07/24 19:10:17 brooks Exp $
62 */
63/*
64 * NOTICE: This file was modified by SPARTA, Inc. in 2006 to introduce
65 * support for mandatory and extensible security protections.  This notice
66 * is included in support of clause 2.2 (b) of the Apple Public License,
67 * Version 2.0.
68 */
69
70#include <kern/locks.h>
71
72#include <sys/param.h>
73#include <sys/malloc.h>
74#include <sys/mbuf.h>
75#include <sys/systm.h>
76#include <sys/proc.h>
77#include <sys/socket.h>
78#include <sys/socketvar.h>
79#include <sys/protosw.h>
80#include <sys/kernel.h>
81#include <sys/sockio.h>
82#include <sys/syslog.h>
83#include <sys/sysctl.h>
84#include <sys/mcache.h>
85#include <kern/zalloc.h>
86
87#include <machine/endian.h>
88
89#include <pexpert/pexpert.h>
90
91#include <net/if.h>
92#include <net/if_arp.h>
93#include <net/if_dl.h>
94#include <net/if_types.h>
95#include <net/if_var.h>
96#include <net/net_osdep.h>
97#include <net/ethernet.h>
98
99#include <net/radix.h>
100#include <net/route.h>
101#ifdef __APPLE__
102#include <net/dlil.h>
103//#include <string.h>
104#include <sys/domain.h>
105#include <libkern/OSAtomic.h>
106#endif
107
108#if INET || INET6
109/*XXX*/
110#include <netinet/in.h>
111#include <netinet/in_var.h>
112#include <netinet/ip_var.h>
113#include <netinet/ip6.h>
114#include <netinet/ip_var.h>
115#include <netinet/tcp.h>
116#include <netinet/tcp_var.h>
117#include <netinet/udp.h>
118#include <netinet/udp_var.h>
119#if INET6
120#include <netinet6/in6_var.h>
121#include <netinet6/in6_ifattach.h>
122#include <netinet6/ip6_var.h>
123#endif
124#endif
125
126#if CONFIG_MACF_NET
127#include <security/mac_framework.h>
128#endif
129
130#if PF_ALTQ
131#include <net/altq/if_altq.h>
132#endif /* !PF_ALTQ */
133
134/*
135 * System initialization
136 */
137
138/* Lock group and attribute for ifaddr lock */
139lck_attr_t	*ifa_mtx_attr;
140lck_grp_t	*ifa_mtx_grp;
141static lck_grp_attr_t	*ifa_mtx_grp_attr;
142
143static int ifioctl_ifreq(struct socket *, u_long, struct ifreq *,
144    struct proc *);
145static int ifconf(u_long cmd, user_addr_t ifrp, int * ret_space);
146__private_extern__ void link_rtrequest(int, struct rtentry *, struct sockaddr *);
147void if_rtproto_del(struct ifnet *ifp, int protocol);
148
149static int if_addmulti_common(struct ifnet *, const struct sockaddr *,
150    struct ifmultiaddr **, int);
151static int if_delmulti_common(struct ifmultiaddr *, struct ifnet *,
152    const struct sockaddr *, int);
153
154static int if_rtmtu(struct radix_node *, void *);
155static void if_rtmtu_update(struct ifnet *);
156
157#if IF_CLONE_LIST
158static int	if_clone_list(int count, int * total, user_addr_t dst);
159#endif /* IF_CLONE_LIST */
160
161MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
162
163struct	ifnethead ifnet_head = TAILQ_HEAD_INITIALIZER(ifnet_head);
164
165static int	if_cloners_count;
166LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
167
168static struct ifaddr *ifa_ifwithnet_common(const struct sockaddr *,
169    unsigned int);
170static void if_attach_ifa_common(struct ifnet *, struct ifaddr *, int);
171static void if_detach_ifa_common(struct ifnet *, struct ifaddr *, int);
172
173static void if_attach_ifma(struct ifnet *, struct ifmultiaddr *, int);
174static int if_detach_ifma(struct ifnet *, struct ifmultiaddr *, int);
175
176static struct ifmultiaddr *ifma_alloc(int);
177static void ifma_free(struct ifmultiaddr *);
178static void ifma_trace(struct ifmultiaddr *, int);
179
180#if DEBUG
181static unsigned int ifma_debug = 1;	/* debugging (enabled) */
182#else
183static unsigned int ifma_debug;		/* debugging (disabled) */
184#endif /* !DEBUG */
185static unsigned int ifma_size;		/* size of zone element */
186static struct zone *ifma_zone;		/* zone for ifmultiaddr */
187
188#define	IFMA_TRACE_HIST_SIZE	32	/* size of trace history */
189
190/* For gdb */
191__private_extern__ unsigned int ifma_trace_hist_size = IFMA_TRACE_HIST_SIZE;
192
193struct ifmultiaddr_dbg {
194	struct ifmultiaddr	ifma;			/* ifmultiaddr */
195	u_int16_t		ifma_refhold_cnt;	/* # of ref */
196	u_int16_t		ifma_refrele_cnt;	/* # of rele */
197	/*
198	 * Circular lists of IFA_ADDREF and IFA_REMREF callers.
199	 */
200	ctrace_t		ifma_refhold[IFMA_TRACE_HIST_SIZE];
201	ctrace_t		ifma_refrele[IFMA_TRACE_HIST_SIZE];
202	/*
203	 * Trash list linkage
204	 */
205	TAILQ_ENTRY(ifmultiaddr_dbg) ifma_trash_link;
206};
207
208/* List of trash ifmultiaddr entries protected by ifma_trash_lock */
209static TAILQ_HEAD(, ifmultiaddr_dbg) ifma_trash_head;
210static decl_lck_mtx_data(, ifma_trash_lock);
211
212#define	IFMA_ZONE_MAX		64		/* maximum elements in zone */
213#define	IFMA_ZONE_NAME		"ifmultiaddr"	/* zone name */
214
215#if INET6
216/*
217 * XXX: declare here to avoid to include many inet6 related files..
218 * should be more generalized?
219 */
220extern void	nd6_setmtu(struct ifnet *);
221extern lck_mtx_t *nd6_mutex;
222#endif
223
224
225void
226ifa_init(void)
227{
228	/* Setup lock group and attribute for ifaddr */
229	ifa_mtx_grp_attr = lck_grp_attr_alloc_init();
230	ifa_mtx_grp = lck_grp_alloc_init("ifaddr", ifa_mtx_grp_attr);
231	ifa_mtx_attr = lck_attr_alloc_init();
232
233	PE_parse_boot_argn("ifa_debug", &ifma_debug, sizeof (ifma_debug));
234
235	ifma_size = (ifma_debug == 0) ? sizeof (struct ifmultiaddr) :
236	    sizeof (struct ifmultiaddr_dbg);
237
238	ifma_zone = zinit(ifma_size, IFMA_ZONE_MAX * ifma_size, 0,
239	    IFMA_ZONE_NAME);
240	if (ifma_zone == NULL) {
241		panic("%s: failed allocating %s", __func__, IFMA_ZONE_NAME);
242		/* NOTREACHED */
243	}
244	zone_change(ifma_zone, Z_EXPAND, TRUE);
245	zone_change(ifma_zone, Z_CALLERACCT, FALSE);
246
247	lck_mtx_init(&ifma_trash_lock, ifa_mtx_grp, ifa_mtx_attr);
248	TAILQ_INIT(&ifma_trash_head);
249}
250
251/*
252 * Network interface utility routines.
253 *
254 * Routines with ifa_ifwith* names take sockaddr *'s as
255 * parameters.
256 */
257
258int if_index;
259struct ifaddr **ifnet_addrs;
260struct ifnet **ifindex2ifnet;
261
262__private_extern__ void
263if_attach_ifa(struct ifnet *ifp, struct ifaddr *ifa)
264{
265	if_attach_ifa_common(ifp, ifa, 0);
266}
267
268__private_extern__ void
269if_attach_link_ifa(struct ifnet *ifp, struct ifaddr *ifa)
270{
271	if_attach_ifa_common(ifp, ifa, 1);
272}
273
274static void
275if_attach_ifa_common(struct ifnet *ifp, struct ifaddr *ifa, int link)
276{
277	ifnet_lock_assert(ifp, IFNET_LCK_ASSERT_EXCLUSIVE);
278	IFA_LOCK_ASSERT_HELD(ifa);
279
280	if (ifa->ifa_ifp != ifp) {
281		panic("%s: Mismatch ifa_ifp=%p != ifp=%p", __func__,
282		    ifa->ifa_ifp, ifp);
283		/* NOTREACHED */
284	} else if (ifa->ifa_debug & IFD_ATTACHED) {
285		panic("%s: Attempt to attach an already attached ifa=%p",
286		    __func__, ifa);
287		/* NOTREACHED */
288	} else if (link && !(ifa->ifa_debug & IFD_LINK)) {
289		panic("%s: Unexpected non-link address ifa=%p", __func__, ifa);
290		/* NOTREACHED */
291	} else if (!link && (ifa->ifa_debug & IFD_LINK)) {
292		panic("%s: Unexpected link address ifa=%p", __func__, ifa);
293		/* NOTREACHED */
294	}
295	IFA_ADDREF_LOCKED(ifa);
296	ifa->ifa_debug |= IFD_ATTACHED;
297	if (link)
298		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
299	else
300		TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
301
302	if (ifa->ifa_attached != NULL)
303		(*ifa->ifa_attached)(ifa);
304}
305
306__private_extern__ void
307if_detach_ifa(struct ifnet *ifp, struct ifaddr *ifa)
308{
309	if_detach_ifa_common(ifp, ifa, 0);
310}
311
312__private_extern__ void
313if_detach_link_ifa(struct ifnet *ifp, struct ifaddr *ifa)
314{
315	if_detach_ifa_common(ifp, ifa, 1);
316}
317
318static void
319if_detach_ifa_common(struct ifnet *ifp, struct ifaddr *ifa, int link)
320{
321	ifnet_lock_assert(ifp, IFNET_LCK_ASSERT_EXCLUSIVE);
322	IFA_LOCK_ASSERT_HELD(ifa);
323
324	if (link && !(ifa->ifa_debug & IFD_LINK)) {
325		panic("%s: Unexpected non-link address ifa=%p", __func__, ifa);
326		/* NOTREACHED */
327	} else if (link && ifa != TAILQ_FIRST(&ifp->if_addrhead)) {
328		panic("%s: Link address ifa=%p not first", __func__, ifa);
329		/* NOTREACHED */
330	} else if (!link && (ifa->ifa_debug & IFD_LINK)) {
331		panic("%s: Unexpected link address ifa=%p", __func__, ifa);
332		/* NOTREACHED */
333	} else if (!(ifa->ifa_debug & IFD_ATTACHED)) {
334		panic("%s: Attempt to detach an unattached address ifa=%p",
335		    __func__, ifa);
336		/* NOTREACHED */
337	} else if (ifa->ifa_ifp != ifp) {
338		panic("%s: Mismatch ifa_ifp=%p, ifp=%p", __func__,
339		    ifa->ifa_ifp, ifp);
340		/* NOTREACHED */
341	} else if (ifa->ifa_debug & IFD_DEBUG) {
342		struct ifaddr *ifa2;
343		TAILQ_FOREACH(ifa2, &ifp->if_addrhead, ifa_link) {
344			if (ifa2 == ifa)
345				break;
346		}
347		if (ifa2 != ifa) {
348			panic("%s: Attempt to detach a stray address ifa=%p",
349			    __func__, ifa);
350			/* NOTREACHED */
351		}
352	}
353	TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
354	/* This must not be the last reference to the ifaddr */
355	if (IFA_REMREF_LOCKED(ifa) == NULL) {
356		panic("%s: unexpected (missing) refcnt ifa=%p", __func__, ifa);
357		/* NOTREACHED */
358	}
359	ifa->ifa_debug &= ~IFD_ATTACHED;
360
361	if (ifa->ifa_detached != NULL)
362		(*ifa->ifa_detached)(ifa);
363}
364
365#define INITIAL_IF_INDEXLIM	8
366
367/*
368 * Function: if_next_index
369 * Purpose:
370 *   Return the next available interface index.
371 *   Grow the ifnet_addrs[] and ifindex2ifnet[] arrays to accomodate the
372 *   added entry when necessary.
373 *
374 * Note:
375 *   ifnet_addrs[] is indexed by (if_index - 1), whereas
376 *   ifindex2ifnet[] is indexed by ifp->if_index.  That requires us to
377 *   always allocate one extra element to hold ifindex2ifnet[0], which
378 *   is unused.
379 */
380int if_next_index(void);
381
382__private_extern__ int
383if_next_index(void)
384{
385	static int 	if_indexlim = 0;
386	int		new_index;
387
388	new_index = ++if_index;
389	if (if_index > if_indexlim) {
390		unsigned 	n;
391		int		new_if_indexlim;
392		caddr_t		new_ifnet_addrs;
393		caddr_t		new_ifindex2ifnet;
394		caddr_t		old_ifnet_addrs;
395
396		old_ifnet_addrs = (caddr_t)ifnet_addrs;
397		if (ifnet_addrs == NULL) {
398			new_if_indexlim = INITIAL_IF_INDEXLIM;
399		} else {
400			new_if_indexlim = if_indexlim << 1;
401		}
402
403		/* allocate space for the larger arrays */
404		n = (2 * new_if_indexlim + 1) * sizeof(caddr_t);
405		new_ifnet_addrs = _MALLOC(n, M_IFADDR, M_WAITOK);
406		if (new_ifnet_addrs == NULL) {
407			--if_index;
408			return -1;
409		}
410
411		new_ifindex2ifnet = new_ifnet_addrs
412			+ new_if_indexlim * sizeof(caddr_t);
413		bzero(new_ifnet_addrs, n);
414		if (ifnet_addrs != NULL) {
415			/* copy the existing data */
416			bcopy((caddr_t)ifnet_addrs, new_ifnet_addrs,
417			      if_indexlim * sizeof(caddr_t));
418			bcopy((caddr_t)ifindex2ifnet,
419			      new_ifindex2ifnet,
420			      (if_indexlim + 1) * sizeof(caddr_t));
421		}
422
423		/* switch to the new tables and size */
424		ifnet_addrs = (struct ifaddr **)(void *)new_ifnet_addrs;
425		ifindex2ifnet = (struct ifnet **)(void *)new_ifindex2ifnet;
426		if_indexlim = new_if_indexlim;
427
428		/* release the old data */
429		if (old_ifnet_addrs != NULL) {
430			_FREE((caddr_t)old_ifnet_addrs, M_IFADDR);
431		}
432	}
433	return (new_index);
434}
435
436/*
437 * Create a clone network interface.
438 */
439static int
440if_clone_create(char *name, int len, void *params)
441{
442	struct if_clone *ifc;
443	char *dp;
444	int wildcard;
445	u_int32_t bytoff, bitoff;
446	u_int32_t unit;
447	int err;
448
449	ifc = if_clone_lookup(name, &unit);
450	if (ifc == NULL)
451		return (EINVAL);
452
453	if (ifunit(name) != NULL)
454		return (EEXIST);
455
456	bytoff = bitoff = 0;
457	wildcard = (unit == UINT32_MAX);
458	/*
459	 * Find a free unit if none was given.
460	 */
461	if (wildcard) {
462		while ((bytoff < ifc->ifc_bmlen)
463		    && (ifc->ifc_units[bytoff] == 0xff))
464			bytoff++;
465		if (bytoff >= ifc->ifc_bmlen)
466			return (ENOSPC);
467		while ((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0)
468			bitoff++;
469		unit = (bytoff << 3) + bitoff;
470	}
471
472	if (unit > ifc->ifc_maxunit)
473		return (ENXIO);
474
475	err = (*ifc->ifc_create)(ifc, unit, params);
476	if (err != 0)
477		return (err);
478
479	if (!wildcard) {
480		bytoff = unit >> 3;
481		bitoff = unit - (bytoff << 3);
482	}
483
484	/*
485	 * Allocate the unit in the bitmap.
486	 */
487	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) == 0,
488	    ("%s: bit is already set", __func__));
489	ifc->ifc_units[bytoff] |= (1 << bitoff);
490
491	/* In the wildcard case, we need to update the name. */
492	if (wildcard) {
493		for (dp = name; *dp != '\0'; dp++);
494		if (snprintf(dp, len - (dp-name), "%d", unit) >
495		    len - (dp-name) - 1) {
496			/*
497			 * This can only be a programmer error and
498			 * there's no straightforward way to recover if
499			 * it happens.
500			 */
501			panic("%s: interface name too long", __func__);
502			/* NOTREACHED */
503		}
504
505	}
506
507	return (0);
508}
509
510/*
511 * Destroy a clone network interface.
512 */
513static int
514if_clone_destroy(const char *name)
515{
516	struct if_clone *ifc;
517	struct ifnet *ifp;
518	int bytoff, bitoff;
519	u_int32_t unit;
520
521	ifc = if_clone_lookup(name, &unit);
522	if (ifc == NULL)
523		return (EINVAL);
524
525	if (unit < ifc->ifc_minifs)
526		return (EINVAL);
527
528	ifp = ifunit(name);
529	if (ifp == NULL)
530		return (ENXIO);
531
532	if (ifc->ifc_destroy == NULL)
533		return (EOPNOTSUPP);
534
535	(*ifc->ifc_destroy)(ifp);
536
537	/*
538	 * Compute offset in the bitmap and deallocate the unit.
539	 */
540	bytoff = unit >> 3;
541	bitoff = unit - (bytoff << 3);
542	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0,
543	    ("%s: bit is already cleared", __func__));
544	ifc->ifc_units[bytoff] &= ~(1 << bitoff);
545	return (0);
546}
547
548/*
549 * Look up a network interface cloner.
550 */
551
552__private_extern__ struct if_clone *
553if_clone_lookup(const char *name, u_int32_t *unitp)
554{
555	struct if_clone *ifc;
556	const char *cp;
557	size_t i;
558
559	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
560		for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
561			if (ifc->ifc_name[i] != *cp)
562				goto next_ifc;
563		}
564		goto found_name;
565 next_ifc:
566		ifc = LIST_NEXT(ifc, ifc_list);
567	}
568
569	/* No match. */
570	return ((struct if_clone *)NULL);
571
572 found_name:
573	if (*cp == '\0') {
574		i = 0xffff;
575	} else {
576		for (i = 0; *cp != '\0'; cp++) {
577			if (*cp < '0' || *cp > '9') {
578				/* Bogus unit number. */
579				return (NULL);
580			}
581			i = (i * 10) + (*cp - '0');
582		}
583	}
584
585	if (unitp != NULL)
586		*unitp = i;
587	return (ifc);
588}
589
590/*
591 * Register a network interface cloner.
592 */
593int
594if_clone_attach(struct if_clone *ifc)
595{
596	int bytoff, bitoff;
597	int err;
598	int len, maxclone;
599	u_int32_t unit;
600
601	KASSERT(ifc->ifc_minifs - 1 <= ifc->ifc_maxunit,
602	    ("%s: %s requested more units then allowed (%d > %d)",
603	    __func__, ifc->ifc_name, ifc->ifc_minifs,
604	    ifc->ifc_maxunit + 1));
605	/*
606	 * Compute bitmap size and allocate it.
607	 */
608	maxclone = ifc->ifc_maxunit + 1;
609	len = maxclone >> 3;
610	if ((len << 3) < maxclone)
611		len++;
612	ifc->ifc_units = _MALLOC(len, M_CLONE, M_WAITOK | M_ZERO);
613	if (ifc->ifc_units == NULL)
614		return ENOBUFS;
615	bzero(ifc->ifc_units, len);
616	ifc->ifc_bmlen = len;
617
618	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
619	if_cloners_count++;
620
621	for (unit = 0; unit < ifc->ifc_minifs; unit++) {
622		err = (*ifc->ifc_create)(ifc, unit, NULL);
623		KASSERT(err == 0,
624		    ("%s: failed to create required interface %s%d",
625		    __func__, ifc->ifc_name, unit));
626
627		/* Allocate the unit in the bitmap. */
628		bytoff = unit >> 3;
629		bitoff = unit - (bytoff << 3);
630		ifc->ifc_units[bytoff] |= (1 << bitoff);
631	}
632
633	return 0;
634}
635
636/*
637 * Unregister a network interface cloner.
638 */
639void
640if_clone_detach(struct if_clone *ifc)
641{
642
643	LIST_REMOVE(ifc, ifc_list);
644	FREE(ifc->ifc_units, M_CLONE);
645	if_cloners_count--;
646}
647
648#if IF_CLONE_LIST
649/*
650 * Provide list of interface cloners to userspace.
651 */
652static int
653if_clone_list(int count, int * total, user_addr_t dst)
654{
655	char outbuf[IFNAMSIZ];
656	struct if_clone *ifc;
657	int error = 0;
658
659	*total = if_cloners_count;
660	if (dst == USER_ADDR_NULL) {
661		/* Just asking how many there are. */
662		return (0);
663	}
664
665	if (count < 0)
666		return (EINVAL);
667
668	count = (if_cloners_count < count) ? if_cloners_count : count;
669
670	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
671	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
672		strlcpy(outbuf, ifc->ifc_name, IFNAMSIZ);
673		error = copyout(outbuf, dst, IFNAMSIZ);
674		if (error)
675			break;
676	}
677
678	return (error);
679}
680#endif /* IF_CLONE_LIST */
681
682/*
683 * Similar to ifa_ifwithaddr, except that this is IPv4 specific
684 * and that it matches only the local (not broadcast) address.
685 */
686__private_extern__ struct in_ifaddr *
687ifa_foraddr(unsigned int addr)
688{
689	return (ifa_foraddr_scoped(addr, IFSCOPE_NONE));
690}
691
692/*
693 * Similar to ifa_foraddr, except with the added interface scope
694 * constraint (unless the caller passes in IFSCOPE_NONE in which
695 * case there is no scope restriction).
696 */
697__private_extern__ struct in_ifaddr *
698ifa_foraddr_scoped(unsigned int addr, unsigned int scope)
699{
700	struct in_ifaddr *ia = NULL;
701
702	lck_rw_lock_shared(in_ifaddr_rwlock);
703	TAILQ_FOREACH(ia, INADDR_HASH(addr), ia_hash) {
704		IFA_LOCK_SPIN(&ia->ia_ifa);
705		if (ia->ia_addr.sin_addr.s_addr == addr &&
706		    (scope == IFSCOPE_NONE || ia->ia_ifp->if_index == scope)) {
707			IFA_ADDREF_LOCKED(&ia->ia_ifa);	/* for caller */
708			IFA_UNLOCK(&ia->ia_ifa);
709			break;
710		}
711		IFA_UNLOCK(&ia->ia_ifa);
712	}
713	lck_rw_done(in_ifaddr_rwlock);
714	return (ia);
715}
716
717#if INET6
718/*
719 * Similar to ifa_foraddr, except that this for IPv6.
720 */
721__private_extern__ struct in6_ifaddr *
722ifa_foraddr6(struct in6_addr *addr6)
723{
724	return (ifa_foraddr6_scoped(addr6, IFSCOPE_NONE));
725}
726
727__private_extern__ struct in6_ifaddr *
728ifa_foraddr6_scoped(struct in6_addr *addr6, unsigned int scope)
729{
730	struct in6_ifaddr *ia = NULL;
731
732	lck_rw_lock_shared(&in6_ifaddr_rwlock);
733	for (ia = in6_ifaddrs; ia; ia = ia->ia_next) {
734		IFA_LOCK(&ia->ia_ifa);
735		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, addr6) &&
736		    (scope == IFSCOPE_NONE || ia->ia_ifp->if_index == scope)) {
737			IFA_ADDREF_LOCKED(&ia->ia_ifa); /* for caller */
738			IFA_UNLOCK(&ia->ia_ifa);
739			break;
740		}
741		IFA_UNLOCK(&ia->ia_ifa);
742	}
743	lck_rw_done(&in6_ifaddr_rwlock);
744
745	return (ia);
746}
747#endif /* INET6 */
748
749/*
750 * Return the first (primary) address of a given family on an interface.
751 */
752__private_extern__ struct ifaddr *
753ifa_ifpgetprimary(struct ifnet *ifp, int family)
754{
755	struct ifaddr *ifa;
756
757	ifnet_lock_shared(ifp);
758	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
759		IFA_LOCK_SPIN(ifa);
760		if (ifa->ifa_addr->sa_family == family) {
761			IFA_ADDREF_LOCKED(ifa);	/* for caller */
762			IFA_UNLOCK(ifa);
763			break;
764		}
765		IFA_UNLOCK(ifa);
766	}
767	ifnet_lock_done(ifp);
768
769	return (ifa);
770}
771
772/*
773 * Locate an interface based on a complete address.
774 */
775/*ARGSUSED*/
776struct ifaddr *
777ifa_ifwithaddr(const struct sockaddr *addr)
778{
779	struct ifnet *ifp;
780	struct ifaddr *ifa;
781	struct ifaddr *result = NULL;
782
783#define	equal(a1, a2)							\
784	(bcmp((const void*)(a1), (const void*)(a2),			\
785	    ((const struct sockaddr *)(a1))->sa_len) == 0)
786
787	ifnet_head_lock_shared();
788	for (ifp = ifnet_head.tqh_first; ifp && !result;
789	    ifp = ifp->if_link.tqe_next) {
790		ifnet_lock_shared(ifp);
791		for (ifa = ifp->if_addrhead.tqh_first; ifa;
792		    ifa = ifa->ifa_link.tqe_next) {
793			IFA_LOCK_SPIN(ifa);
794			if (ifa->ifa_addr->sa_family != addr->sa_family) {
795				IFA_UNLOCK(ifa);
796				continue;
797			}
798			if (equal(addr, ifa->ifa_addr)) {
799				result = ifa;
800				IFA_ADDREF_LOCKED(ifa);	/* for caller */
801				IFA_UNLOCK(ifa);
802				break;
803			}
804			if ((ifp->if_flags & IFF_BROADCAST) &&
805			    ifa->ifa_broadaddr != NULL &&
806			    /* IP6 doesn't have broadcast */
807			    ifa->ifa_broadaddr->sa_len != 0 &&
808			    equal(ifa->ifa_broadaddr, addr)) {
809				result = ifa;
810				IFA_ADDREF_LOCKED(ifa);	/* for caller */
811				IFA_UNLOCK(ifa);
812				break;
813			}
814			IFA_UNLOCK(ifa);
815		}
816		ifnet_lock_done(ifp);
817	}
818	ifnet_head_done();
819
820	return (result);
821}
822/*
823 * Locate the point to point interface with a given destination address.
824 */
825/*ARGSUSED*/
826struct ifaddr *
827ifa_ifwithdstaddr(const struct sockaddr *addr)
828{
829	struct ifnet *ifp;
830	struct ifaddr *ifa;
831	struct ifaddr *result = NULL;
832
833	ifnet_head_lock_shared();
834	for (ifp = ifnet_head.tqh_first; ifp && !result;
835	    ifp = ifp->if_link.tqe_next) {
836	    if ((ifp->if_flags & IFF_POINTOPOINT)) {
837			ifnet_lock_shared(ifp);
838			for (ifa = ifp->if_addrhead.tqh_first; ifa;
839			    ifa = ifa->ifa_link.tqe_next) {
840				IFA_LOCK_SPIN(ifa);
841				if (ifa->ifa_addr->sa_family !=
842				    addr->sa_family) {
843					IFA_UNLOCK(ifa);
844					continue;
845				}
846				if (ifa->ifa_dstaddr &&
847				    equal(addr, ifa->ifa_dstaddr)) {
848					result = ifa;
849					IFA_ADDREF_LOCKED(ifa);	/* for caller */
850					IFA_UNLOCK(ifa);
851					break;
852				}
853				IFA_UNLOCK(ifa);
854			}
855			ifnet_lock_done(ifp);
856		}
857	}
858	ifnet_head_done();
859	return (result);
860}
861
862/*
863 * Locate the source address of an interface based on a complete address.
864 */
865struct ifaddr *
866ifa_ifwithaddr_scoped(const struct sockaddr *addr, unsigned int ifscope)
867{
868	struct ifaddr *result = NULL;
869	struct ifnet *ifp;
870
871	if (ifscope == IFSCOPE_NONE)
872		return (ifa_ifwithaddr(addr));
873
874	ifnet_head_lock_shared();
875	if (ifscope > (unsigned int)if_index) {
876		ifnet_head_done();
877		return (NULL);
878	}
879
880	ifp = ifindex2ifnet[ifscope];
881	if (ifp != NULL) {
882		struct ifaddr *ifa = NULL;
883
884		/*
885		 * This is suboptimal; there should be a better way
886		 * to search for a given address of an interface
887		 * for any given address family.
888		 */
889		ifnet_lock_shared(ifp);
890		for (ifa = ifp->if_addrhead.tqh_first; ifa != NULL;
891		    ifa = ifa->ifa_link.tqe_next) {
892			IFA_LOCK_SPIN(ifa);
893			if (ifa->ifa_addr->sa_family != addr->sa_family) {
894				IFA_UNLOCK(ifa);
895				continue;
896			}
897			if (equal(addr, ifa->ifa_addr)) {
898				result = ifa;
899				IFA_ADDREF_LOCKED(ifa);	/* for caller */
900				IFA_UNLOCK(ifa);
901				break;
902			}
903			if ((ifp->if_flags & IFF_BROADCAST) &&
904			    ifa->ifa_broadaddr != NULL &&
905			    /* IP6 doesn't have broadcast */
906			    ifa->ifa_broadaddr->sa_len != 0 &&
907			    equal(ifa->ifa_broadaddr, addr)) {
908				result = ifa;
909				IFA_ADDREF_LOCKED(ifa);	/* for caller */
910				IFA_UNLOCK(ifa);
911				break;
912			}
913			IFA_UNLOCK(ifa);
914		}
915		ifnet_lock_done(ifp);
916	}
917	ifnet_head_done();
918
919	return (result);
920}
921
922struct ifaddr *
923ifa_ifwithnet(const struct sockaddr *addr)
924{
925	return (ifa_ifwithnet_common(addr, IFSCOPE_NONE));
926}
927
928struct ifaddr *
929ifa_ifwithnet_scoped(const struct sockaddr *addr, unsigned int ifscope)
930{
931	return (ifa_ifwithnet_common(addr, ifscope));
932}
933
934/*
935 * Find an interface on a specific network.  If many, choice
936 * is most specific found.
937 */
938static struct ifaddr *
939ifa_ifwithnet_common(const struct sockaddr *addr, unsigned int ifscope)
940{
941	struct ifnet *ifp;
942	struct ifaddr *ifa = NULL;
943	struct ifaddr *ifa_maybe = NULL;
944	u_int af = addr->sa_family;
945	const char *addr_data = addr->sa_data, *cplim;
946
947#if INET6
948	if ((af != AF_INET && af != AF_INET6) ||
949	    (af == AF_INET && !ip_doscopedroute) ||
950	    (af == AF_INET6 && !ip6_doscopedroute))
951#else
952	if (af != AF_INET || !ip_doscopedroute)
953#endif /* !INET6 */
954		ifscope = IFSCOPE_NONE;
955
956	ifnet_head_lock_shared();
957	/*
958	 * AF_LINK addresses can be looked up directly by their index number,
959	 * so do that if we can.
960	 */
961	if (af == AF_LINK) {
962		const struct sockaddr_dl *sdl =
963		    (const struct sockaddr_dl *)(uintptr_t)(size_t)addr;
964		if (sdl->sdl_index && sdl->sdl_index <= if_index) {
965			ifa = ifnet_addrs[sdl->sdl_index - 1];
966			if (ifa != NULL)
967				IFA_ADDREF(ifa);
968
969			ifnet_head_done();
970			return (ifa);
971		}
972	}
973
974	/*
975	 * Scan though each interface, looking for ones that have
976	 * addresses in this address family.
977	 */
978	for (ifp = ifnet_head.tqh_first; ifp; ifp = ifp->if_link.tqe_next) {
979		ifnet_lock_shared(ifp);
980		for (ifa = ifp->if_addrhead.tqh_first; ifa;
981		     ifa = ifa->ifa_link.tqe_next) {
982			const char *cp, *cp2, *cp3;
983
984			IFA_LOCK(ifa);
985			if (ifa->ifa_addr == NULL ||
986			    ifa->ifa_addr->sa_family != af) {
987next:
988				IFA_UNLOCK(ifa);
989				continue;
990			}
991#ifndef __APPLE__
992/* This breaks tunneling application trying to install a route with
993 * a specific subnet and the local address as the destination
994 * It's breaks binary compatibility with previous version of MacOS X
995 */
996			if (
997#if INET6 /* XXX: for maching gif tunnel dst as routing entry gateway */
998			    addr->sa_family != AF_INET6 &&
999#endif
1000			    ifp->if_flags & IFF_POINTOPOINT) {
1001				/*
1002				 * This is a bit broken as it doesn't
1003				 * take into account that the remote end may
1004				 * be a single node in the network we are
1005				 * looking for.
1006				 * The trouble is that we don't know the
1007				 * netmask for the remote end.
1008				 */
1009				if (ifa->ifa_dstaddr != 0 &&
1010				    equal(addr, ifa->ifa_dstaddr)) {
1011					IFA_ADDREF_LOCKED(ifa);
1012					IFA_UNLOCK(ifa);
1013					break;
1014				}
1015				IFA_UNLOCK(ifa);
1016			} else
1017#endif /* __APPLE__*/
1018			{
1019				/*
1020				 * If we're looking up with a scope,
1021				 * find using a matching interface.
1022				 */
1023				if (ifscope != IFSCOPE_NONE &&
1024				    ifp->if_index != ifscope) {
1025					IFA_UNLOCK(ifa);
1026					continue;
1027				}
1028
1029				/*
1030				 * Scan all the bits in the ifa's address.
1031				 * If a bit dissagrees with what we are
1032				 * looking for, mask it with the netmask
1033				 * to see if it really matters.
1034				 * (A byte at a time)
1035				 */
1036				if (ifa->ifa_netmask == 0) {
1037					IFA_UNLOCK(ifa);
1038					continue;
1039				}
1040				cp = addr_data;
1041				cp2 = ifa->ifa_addr->sa_data;
1042				cp3 = ifa->ifa_netmask->sa_data;
1043				cplim = ifa->ifa_netmask->sa_len
1044					+ (char *)ifa->ifa_netmask;
1045				while (cp3 < cplim)
1046					if ((*cp++ ^ *cp2++) & *cp3++)
1047						goto next; /* next address! */
1048				/*
1049				 * If the netmask of what we just found
1050				 * is more specific than what we had before
1051				 * (if we had one) then remember the new one
1052				 * before continuing to search
1053				 * for an even better one.
1054				 */
1055				if (ifa_maybe == NULL ||
1056				    rn_refines((caddr_t)ifa->ifa_netmask,
1057				    (caddr_t)ifa_maybe->ifa_netmask)) {
1058					IFA_ADDREF_LOCKED(ifa);	/* ifa_maybe */
1059					IFA_UNLOCK(ifa);
1060					if (ifa_maybe != NULL)
1061						IFA_REMREF(ifa_maybe);
1062					ifa_maybe = ifa;
1063				} else {
1064					IFA_UNLOCK(ifa);
1065				}
1066			}
1067			IFA_LOCK_ASSERT_NOTHELD(ifa);
1068		}
1069		ifnet_lock_done(ifp);
1070
1071		if (ifa != NULL)
1072			break;
1073	}
1074	ifnet_head_done();
1075
1076	if (ifa == NULL)
1077		ifa = ifa_maybe;
1078	else if (ifa_maybe != NULL)
1079		IFA_REMREF(ifa_maybe);
1080
1081	return (ifa);
1082}
1083
1084/*
1085 * Find an interface address specific to an interface best matching
1086 * a given address.
1087 */
1088struct ifaddr *
1089ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1090{
1091	struct ifaddr *ifa = NULL;
1092	const char *cp, *cp2, *cp3;
1093	char *cplim;
1094	struct ifaddr *ifa_maybe = NULL;
1095	struct ifaddr *better_ifa_maybe = NULL;
1096	u_int af = addr->sa_family;
1097
1098	if (af >= AF_MAX)
1099		return (NULL);
1100
1101	ifnet_lock_shared(ifp);
1102	for (ifa = ifp->if_addrhead.tqh_first; ifa;
1103	     ifa = ifa->ifa_link.tqe_next) {
1104		IFA_LOCK(ifa);
1105		if (ifa->ifa_addr->sa_family != af) {
1106			IFA_UNLOCK(ifa);
1107			continue;
1108		}
1109		if (ifa_maybe == NULL) {
1110			IFA_ADDREF_LOCKED(ifa);	/* for ifa_maybe */
1111			ifa_maybe = ifa;
1112		}
1113		if (ifa->ifa_netmask == 0) {
1114			if (equal(addr, ifa->ifa_addr) || (ifa->ifa_dstaddr &&
1115			    equal(addr, ifa->ifa_dstaddr))) {
1116				IFA_ADDREF_LOCKED(ifa);	/* for caller */
1117				IFA_UNLOCK(ifa);
1118				break;
1119			}
1120			IFA_UNLOCK(ifa);
1121			continue;
1122		}
1123		if (ifp->if_flags & IFF_POINTOPOINT) {
1124			if (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr)) {
1125				IFA_ADDREF_LOCKED(ifa);	/* for caller */
1126				IFA_UNLOCK(ifa);
1127				break;
1128			}
1129		} else {
1130			if (equal(addr, ifa->ifa_addr)) {
1131				/* exact match */
1132				IFA_ADDREF_LOCKED(ifa);	/* for caller */
1133				IFA_UNLOCK(ifa);
1134				break;
1135			}
1136			cp = addr->sa_data;
1137			cp2 = ifa->ifa_addr->sa_data;
1138			cp3 = ifa->ifa_netmask->sa_data;
1139			cplim = ifa->ifa_netmask->sa_len +
1140			    (char *)ifa->ifa_netmask;
1141			for (; cp3 < cplim; cp3++)
1142				if ((*cp++ ^ *cp2++) & *cp3)
1143					break;
1144			if (cp3 == cplim) {
1145				/* subnet match */
1146				if (better_ifa_maybe == NULL) {
1147					/* for better_ifa_maybe */
1148					IFA_ADDREF_LOCKED(ifa);
1149					better_ifa_maybe = ifa;
1150				}
1151			}
1152		}
1153		IFA_UNLOCK(ifa);
1154	}
1155
1156	if (ifa == NULL) {
1157		if (better_ifa_maybe != NULL) {
1158			ifa = better_ifa_maybe;
1159			better_ifa_maybe = NULL;
1160		} else {
1161			ifa = ifa_maybe;
1162			ifa_maybe = NULL;
1163		}
1164	}
1165
1166	ifnet_lock_done(ifp);
1167
1168	if (better_ifa_maybe != NULL)
1169		IFA_REMREF(better_ifa_maybe);
1170	if (ifa_maybe != NULL)
1171		IFA_REMREF(ifa_maybe);
1172
1173	return (ifa);
1174}
1175
1176#include <net/route.h>
1177
1178/*
1179 * Default action when installing a route with a Link Level gateway.
1180 * Lookup an appropriate real ifa to point to.
1181 * This should be moved to /sys/net/link.c eventually.
1182 */
1183void
1184link_rtrequest(int cmd, struct rtentry *rt, struct sockaddr *sa)
1185{
1186	struct ifaddr *ifa;
1187	struct sockaddr *dst;
1188	struct ifnet *ifp;
1189	void (*ifa_rtrequest)(int, struct rtentry *, struct sockaddr *);
1190
1191	lck_mtx_assert(rnh_lock, LCK_MTX_ASSERT_OWNED);
1192	RT_LOCK_ASSERT_HELD(rt);
1193
1194	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1195	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1196		return;
1197
1198	/* Become a regular mutex, just in case */
1199	RT_CONVERT_LOCK(rt);
1200
1201	ifa = ifaof_ifpforaddr(dst, ifp);
1202	if (ifa) {
1203		rtsetifa(rt, ifa);
1204		IFA_LOCK_SPIN(ifa);
1205		ifa_rtrequest = ifa->ifa_rtrequest;
1206		IFA_UNLOCK(ifa);
1207		if (ifa_rtrequest != NULL && ifa_rtrequest != link_rtrequest)
1208			ifa_rtrequest(cmd, rt, sa);
1209		IFA_REMREF(ifa);
1210	}
1211}
1212
1213/*
1214 * if_updown will set the interface up or down. It will
1215 * prevent other up/down events from occurring until this
1216 * up/down event has completed.
1217 *
1218 * Caller must lock ifnet. This function will drop the
1219 * lock. This allows ifnet_set_flags to set the rest of
1220 * the flags after we change the up/down state without
1221 * dropping the interface lock between setting the
1222 * up/down state and updating the rest of the flags.
1223 */
1224__private_extern__ void
1225if_updown(
1226	struct ifnet	*ifp,
1227	int				up)
1228{
1229	int i;
1230	struct ifaddr **ifa;
1231	struct timespec	tv;
1232	struct ifclassq *ifq = &ifp->if_snd;
1233
1234	/* Wait until no one else is changing the up/down state */
1235	while ((ifp->if_eflags & IFEF_UPDOWNCHANGE) != 0) {
1236		tv.tv_sec = 0;
1237		tv.tv_nsec = NSEC_PER_SEC / 10;
1238		ifnet_lock_done(ifp);
1239		msleep(&ifp->if_eflags, NULL, 0, "if_updown", &tv);
1240		ifnet_lock_exclusive(ifp);
1241	}
1242
1243	/* Verify that the interface isn't already in the right state */
1244	if ((!up && (ifp->if_flags & IFF_UP) == 0) ||
1245		(up && (ifp->if_flags & IFF_UP) == IFF_UP)) {
1246		return;
1247	}
1248
1249	/* Indicate that the up/down state is changing */
1250	ifp->if_eflags |= IFEF_UPDOWNCHANGE;
1251
1252	/* Mark interface up or down */
1253	if (up) {
1254		ifp->if_flags |= IFF_UP;
1255	}
1256	else {
1257		ifp->if_flags &= ~IFF_UP;
1258	}
1259
1260	ifnet_touch_lastchange(ifp);
1261
1262	/* Drop the lock to notify addresses and route */
1263	ifnet_lock_done(ifp);
1264	if (ifnet_get_address_list(ifp, &ifa) == 0) {
1265		for (i = 0; ifa[i] != 0; i++) {
1266			pfctlinput(up ? PRC_IFUP : PRC_IFDOWN, ifa[i]->ifa_addr);
1267		}
1268		ifnet_free_address_list(ifa);
1269	}
1270	rt_ifmsg(ifp);
1271
1272	if (!up)
1273		if_qflush(ifp, 0);
1274
1275	/* Inform all transmit queues about the new link state */
1276	IFCQ_LOCK(ifq);
1277	ifnet_update_sndq(ifq, up ? CLASSQ_EV_LINK_UP : CLASSQ_EV_LINK_DOWN);
1278	IFCQ_UNLOCK(ifq);
1279
1280	/* Aquire the lock to clear the changing flag */
1281	ifnet_lock_exclusive(ifp);
1282	ifp->if_eflags &= ~IFEF_UPDOWNCHANGE;
1283	wakeup(&ifp->if_eflags);
1284}
1285
1286/*
1287 * Mark an interface down and notify protocols of
1288 * the transition.
1289 */
1290void
1291if_down(
1292	struct ifnet *ifp)
1293{
1294	ifnet_lock_exclusive(ifp);
1295	if_updown(ifp, 0);
1296	ifnet_lock_done(ifp);
1297}
1298
1299/*
1300 * Mark an interface up and notify protocols of
1301 * the transition.
1302 */
1303void
1304if_up(
1305	struct ifnet *ifp)
1306{
1307	ifnet_lock_exclusive(ifp);
1308	if_updown(ifp, 1);
1309	ifnet_lock_done(ifp);
1310}
1311
1312/*
1313 * Flush an interface queue.
1314 */
1315void
1316if_qflush(struct ifnet *ifp, int ifq_locked)
1317{
1318	struct ifclassq *ifq = &ifp->if_snd;
1319
1320	if (!ifq_locked)
1321		IFCQ_LOCK(ifq);
1322
1323	if (IFCQ_IS_ENABLED(ifq))
1324		IFCQ_PURGE(ifq);
1325#if PF_ALTQ
1326	if (IFCQ_IS_DRAINING(ifq))
1327		ifq->ifcq_drain = 0;
1328	if (ALTQ_IS_ENABLED(IFCQ_ALTQ(ifq)))
1329		ALTQ_PURGE(IFCQ_ALTQ(ifq));
1330#endif /* PF_ALTQ */
1331
1332	VERIFY(IFCQ_IS_EMPTY(ifq));
1333
1334	if (!ifq_locked)
1335		IFCQ_UNLOCK(ifq);
1336}
1337
1338void
1339if_qflush_sc(struct ifnet *ifp, mbuf_svc_class_t sc, u_int32_t flow,
1340    u_int32_t *packets, u_int32_t *bytes, int ifq_locked)
1341{
1342	struct ifclassq *ifq = &ifp->if_snd;
1343	u_int32_t cnt = 0, len = 0;
1344	u_int32_t a_cnt = 0, a_len = 0;
1345
1346	VERIFY(sc == MBUF_SC_UNSPEC || MBUF_VALID_SC(sc));
1347	VERIFY(flow != 0);
1348
1349	if (!ifq_locked)
1350		IFCQ_LOCK(ifq);
1351
1352	if (IFCQ_IS_ENABLED(ifq))
1353		IFCQ_PURGE_SC(ifq, sc, flow, cnt, len);
1354#if PF_ALTQ
1355	if (IFCQ_IS_DRAINING(ifq)) {
1356		VERIFY((signed)(ifq->ifcq_drain - cnt) >= 0);
1357		ifq->ifcq_drain -= cnt;
1358	}
1359	if (ALTQ_IS_ENABLED(IFCQ_ALTQ(ifq)))
1360		ALTQ_PURGE_SC(IFCQ_ALTQ(ifq), sc, flow, a_cnt, a_len);
1361#endif /* PF_ALTQ */
1362
1363	if (!ifq_locked)
1364		IFCQ_UNLOCK(ifq);
1365
1366	if (packets != NULL)
1367		*packets = cnt + a_cnt;
1368	if (bytes != NULL)
1369		*bytes = len + a_len;
1370}
1371
1372/*
1373 * Map interface name to
1374 * interface structure pointer.
1375 */
1376struct ifnet *
1377ifunit(const char *name)
1378{
1379	char namebuf[IFNAMSIZ + 1];
1380	const char *cp;
1381	struct ifnet *ifp;
1382	int unit;
1383	unsigned len, m;
1384	char c;
1385
1386	len = strlen(name);
1387	if (len < 2 || len > IFNAMSIZ)
1388		return (NULL);
1389	cp = name + len - 1;
1390	c = *cp;
1391	if (c < '0' || c > '9')
1392		return (NULL);		/* trailing garbage */
1393	unit = 0;
1394	m = 1;
1395	do {
1396		if (cp == name)
1397			return (NULL);	/* no interface name */
1398		unit += (c - '0') * m;
1399		if (unit > 1000000)
1400			return (NULL);	/* number is unreasonable */
1401		m *= 10;
1402		c = *--cp;
1403	} while (c >= '0' && c <= '9');
1404	len = cp - name + 1;
1405	bcopy(name, namebuf, len);
1406	namebuf[len] = '\0';
1407	/*
1408	 * Now search all the interfaces for this name/number
1409	 */
1410	ifnet_head_lock_shared();
1411	TAILQ_FOREACH(ifp, &ifnet_head, if_link) {
1412		if (strncmp(ifp->if_name, namebuf, len))
1413			continue;
1414		if (unit == ifp->if_unit)
1415			break;
1416	}
1417	ifnet_head_done();
1418	return (ifp);
1419}
1420
1421
1422/*
1423 * Map interface name in a sockaddr_dl to
1424 * interface structure pointer.
1425 */
1426struct ifnet *
1427if_withname(struct sockaddr *sa)
1428{
1429	char ifname[IFNAMSIZ+1];
1430	struct sockaddr_dl *sdl = (struct sockaddr_dl *)(void *)sa;
1431
1432	if ( (sa->sa_family != AF_LINK) || (sdl->sdl_nlen == 0) ||
1433	     (sdl->sdl_nlen > IFNAMSIZ) )
1434		return (NULL);
1435
1436	/*
1437	 * ifunit wants a null-terminated name.  It may not be null-terminated
1438	 * in the sockaddr.  We don't want to change the caller's sockaddr,
1439	 * and there might not be room to put the trailing null anyway, so we
1440	 * make a local copy that we know we can null terminate safely.
1441	 */
1442
1443	bcopy(sdl->sdl_data, ifname, sdl->sdl_nlen);
1444	ifname[sdl->sdl_nlen] = '\0';
1445	return (ifunit(ifname));
1446}
1447
1448
1449/*
1450 * Interface ioctls.
1451 */
1452int
1453ifioctl(struct socket *so, u_long cmd, caddr_t data, struct proc *p)
1454{
1455	char ifname[IFNAMSIZ + 1];
1456	struct ifnet *ifp = NULL;
1457	struct ifstat *ifs = NULL;
1458	int error = 0;
1459
1460	bzero(ifname, sizeof (ifname));
1461
1462	/*
1463	 * ioctls which don't require ifp, or ifreq ioctls
1464	 */
1465	switch (cmd) {
1466	case OSIOCGIFCONF32:			/* struct ifconf32 */
1467	case SIOCGIFCONF32: {			/* struct ifconf32 */
1468		struct ifconf32 ifc;
1469		bcopy(data, &ifc, sizeof (ifc));
1470		error = ifconf(cmd, CAST_USER_ADDR_T(ifc.ifc_req),
1471		    &ifc.ifc_len);
1472		bcopy(&ifc, data, sizeof (ifc));
1473		goto done;
1474	}
1475
1476	case SIOCGIFCONF64:			/* struct ifconf64 */
1477	case OSIOCGIFCONF64: {			/* struct ifconf64 */
1478		struct ifconf64 ifc;
1479		bcopy(data, &ifc, sizeof (ifc));
1480		error = ifconf(cmd, ifc.ifc_req, &ifc.ifc_len);
1481		bcopy(&ifc, data, sizeof (ifc));
1482		goto done;
1483	}
1484
1485#if IF_CLONE_LIST
1486	case SIOCIFGCLONERS32: {		/* struct if_clonereq32 */
1487		struct if_clonereq32 ifcr;
1488		bcopy(data, &ifcr, sizeof (ifcr));
1489		error = if_clone_list(ifcr.ifcr_count, &ifcr.ifcr_total,
1490		    CAST_USER_ADDR_T(ifcr.ifcru_buffer));
1491		bcopy(&ifcr, data, sizeof (ifcr));
1492		goto done;
1493	}
1494
1495	case SIOCIFGCLONERS64: {		/* struct if_clonereq64 */
1496		struct if_clonereq64 ifcr;
1497		bcopy(data, &ifcr, sizeof (ifcr));
1498		error = if_clone_list(ifcr.ifcr_count, &ifcr.ifcr_total,
1499		    ifcr.ifcru_buffer);
1500		bcopy(&ifcr, data, sizeof (ifcr));
1501		goto done;
1502	}
1503#endif /* IF_CLONE_LIST */
1504
1505	case SIOCSIFDSTADDR:			/* struct ifreq */
1506	case SIOCSIFADDR:			/* struct ifreq */
1507	case SIOCSIFBRDADDR:			/* struct ifreq */
1508	case SIOCSIFNETMASK:			/* struct ifreq */
1509	case OSIOCGIFADDR:			/* struct ifreq */
1510	case OSIOCGIFDSTADDR:			/* struct ifreq */
1511	case OSIOCGIFBRDADDR:			/* struct ifreq */
1512	case OSIOCGIFNETMASK:			/* struct ifreq */
1513	case SIOCSIFKPI:			/* struct ifreq */
1514		if (so->so_proto == NULL) {
1515			error = EOPNOTSUPP;
1516			goto done;
1517		}
1518		/* FALLTHRU */
1519	case SIOCIFCREATE:			/* struct ifreq */
1520	case SIOCIFCREATE2:			/* struct ifreq */
1521	case SIOCIFDESTROY:			/* struct ifreq */
1522	case SIOCGIFFLAGS:			/* struct ifreq */
1523	case SIOCGIFEFLAGS:			/* struct ifreq */
1524	case SIOCGIFCAP:			/* struct ifreq */
1525	case SIOCGIFMAC:			/* struct ifreq */
1526	case SIOCGIFMETRIC:			/* struct ifreq */
1527	case SIOCGIFMTU:			/* struct ifreq */
1528	case SIOCGIFPHYS:			/* struct ifreq */
1529	case SIOCSIFFLAGS:			/* struct ifreq */
1530	case SIOCSIFCAP:			/* struct ifreq */
1531	case SIOCSIFPHYS:			/* struct ifreq */
1532	case SIOCSIFMTU:			/* struct ifreq */
1533	case SIOCADDMULTI:			/* struct ifreq */
1534	case SIOCDELMULTI:			/* struct ifreq */
1535	case SIOCDIFPHYADDR:			/* struct ifreq */
1536	case SIOCSIFMEDIA:			/* struct ifreq */
1537	case SIOCSIFGENERIC:			/* struct ifreq */
1538	case SIOCSIFLLADDR:			/* struct ifreq */
1539	case SIOCSIFALTMTU:			/* struct ifreq */
1540	case SIOCSIFVLAN:			/* struct ifreq */
1541	case SIOCSIFBOND:			/* struct ifreq */
1542	case SIOCGIFPSRCADDR:			/* struct ifreq */
1543	case SIOCGIFPDSTADDR:			/* struct ifreq */
1544	case SIOCGIFGENERIC:			/* struct ifreq */
1545	case SIOCGIFDEVMTU:			/* struct ifreq */
1546	case SIOCGIFVLAN:			/* struct ifreq */
1547	case SIOCGIFBOND:			/* struct ifreq */
1548	case SIOCGIFWAKEFLAGS:			/* struct ifreq */
1549	case SIOCGIFGETRTREFCNT:		/* struct ifreq */
1550	case SIOCSIFOPPORTUNISTIC:		/* struct ifreq */
1551	case SIOCGIFOPPORTUNISTIC:		/* struct ifreq */
1552	case SIOCGIFLINKQUALITYMETRIC: {	/* struct ifreq */
1553		struct ifreq ifr;
1554		bcopy(data, &ifr, sizeof (ifr));
1555		error = ifioctl_ifreq(so, cmd, &ifr, p);
1556		bcopy(&ifr, data, sizeof (ifr));
1557		goto done;
1558	}
1559	}
1560
1561	/*
1562	 * ioctls which require ifp.  Note that we acquire dlil_ifnet_lock
1563	 * here to ensure that the ifnet, if found, has been fully attached.
1564	 */
1565	dlil_if_lock();
1566	switch (cmd) {
1567	case SIOCSIFPHYADDR: {			/* struct ifaliasreq */
1568		bcopy(((struct ifaliasreq *)(void *)data)->ifra_name,
1569		    ifname, IFNAMSIZ);
1570		ifp = ifunit(ifname);
1571		break;
1572	}
1573
1574#if INET6
1575	case SIOCSIFPHYADDR_IN6_32: {		/* struct in6_aliasreq_32 */
1576		bcopy(((struct in6_aliasreq_32 *)(void *)data)->ifra_name,
1577		    ifname, IFNAMSIZ);
1578		ifp = ifunit(ifname);
1579		break;
1580	}
1581
1582	case SIOCSIFPHYADDR_IN6_64: {		/* struct in6_aliasreq_64 */
1583		bcopy(((struct in6_aliasreq_64 *)(void *)data)->ifra_name,
1584		    ifname, IFNAMSIZ);
1585		ifp = ifunit(ifname);
1586		break;
1587	}
1588#endif
1589
1590	case SIOCSLIFPHYADDR:			/* struct if_laddrreq */
1591	case SIOCGLIFPHYADDR: {			/* struct if_laddrreq */
1592		bcopy(((struct if_laddrreq *)(void *)data)->iflr_name,
1593		    ifname, IFNAMSIZ);
1594		ifp = ifunit(ifname);
1595		break;
1596	}
1597
1598	case SIOCGIFSTATUS: {			/* struct ifstat */
1599		ifs = _MALLOC(sizeof (*ifs), M_DEVBUF, M_WAITOK);
1600		if (ifs == NULL) {
1601			error = ENOMEM;
1602			dlil_if_unlock();
1603			goto done;
1604		}
1605		bcopy(data, ifs, sizeof (*ifs));
1606		ifs->ifs_name[IFNAMSIZ - 1] = '\0';
1607		ifp = ifunit(ifs->ifs_name);
1608		break;
1609	}
1610
1611	case SIOCGIFMEDIA32: {			/* struct ifmediareq32 */
1612		bcopy(((struct ifmediareq32 *)(void *)data)->ifm_name,
1613		    ifname, IFNAMSIZ);
1614		ifp = ifunit(ifname);
1615		break;
1616	}
1617
1618	case SIOCGIFMEDIA64: {			/* struct ifmediareq64 */
1619		bcopy(((struct ifmediareq64 *)(void *)data)->ifm_name,
1620		    ifname, IFNAMSIZ);
1621		ifp = ifunit(ifname);
1622		break;
1623	}
1624
1625	case SIOCSIFDESC:			/* struct if_descreq */
1626	case SIOCGIFDESC: {			/* struct if_descreq */
1627		bcopy(((struct if_descreq *)(void *)data)->ifdr_name,
1628		    ifname, IFNAMSIZ);
1629		ifp = ifunit(ifname);
1630		break;
1631	}
1632
1633	case SIOCSIFLINKPARAMS:			/* struct if_linkparamsreq */
1634	case SIOCGIFLINKPARAMS: {		/* struct if_linkparamsreq */
1635		bcopy(((struct if_linkparamsreq *)(void *)data)->iflpr_name,
1636		    ifname, IFNAMSIZ);
1637		ifp = ifunit(ifname);
1638		break;
1639	}
1640
1641	case SIOCGIFQUEUESTATS: {		/* struct if_qstatsreq */
1642		bcopy(((struct if_qstatsreq *)(void *)data)->ifqr_name,
1643		    ifname, IFNAMSIZ);
1644		ifp = ifunit(ifname);
1645		break;
1646	}
1647
1648	case SIOCSIFTHROTTLE:			/* struct if_throttlereq */
1649	case SIOCGIFTHROTTLE: {			/* struct if_throttlereq */
1650		bcopy(((struct if_throttlereq *)(void *)data)->ifthr_name,
1651		    ifname, IFNAMSIZ);
1652		ifp = ifunit(ifname);
1653		break;
1654	}
1655
1656	default: {
1657		/*
1658		 * This is a bad assumption, but the code seems to
1659		 * have been doing this in the past; caveat emptor.
1660		 */
1661		bcopy(((struct ifreq *)(void *)data)->ifr_name,
1662		    ifname, IFNAMSIZ);
1663		ifp = ifunit(ifname);
1664		break;
1665	}
1666	}
1667	dlil_if_unlock();
1668
1669	if (ifp == NULL) {
1670		error = ENXIO;
1671		goto done;
1672	}
1673
1674	switch (cmd) {
1675	case SIOCSIFPHYADDR:			/* struct ifaliasreq */
1676#if INET6
1677	case SIOCSIFPHYADDR_IN6_32:		/* struct in6_aliasreq_32 */
1678	case SIOCSIFPHYADDR_IN6_64:		/* struct in6_aliasreq_64 */
1679#endif
1680	case SIOCSLIFPHYADDR:			/* struct if_laddrreq */
1681		error = proc_suser(p);
1682		if (error != 0)
1683			break;
1684
1685		error = ifnet_ioctl(ifp, so->so_proto->pr_domain->dom_family,
1686		    cmd, data);
1687		if (error != 0)
1688			break;
1689
1690		ifnet_touch_lastchange(ifp);
1691		break;
1692
1693	case SIOCGIFSTATUS:			/* struct ifstat */
1694		VERIFY(ifs != NULL);
1695		ifs->ascii[0] = '\0';
1696
1697		error = ifnet_ioctl(ifp, so->so_proto->pr_domain->dom_family,
1698		    cmd, (caddr_t)ifs);
1699
1700		bcopy(ifs, data, sizeof (*ifs));
1701		break;
1702
1703	case SIOCGLIFPHYADDR:			/* struct if_laddrreq */
1704	case SIOCGIFMEDIA32:			/* struct ifmediareq32 */
1705	case SIOCGIFMEDIA64:			/* struct ifmediareq64 */
1706		error = ifnet_ioctl(ifp, so->so_proto->pr_domain->dom_family,
1707		    cmd, data);
1708		break;
1709
1710	case SIOCSIFDESC: {			/* struct if_descreq */
1711		struct if_descreq *ifdr = (struct if_descreq *)(void *)data;
1712		u_int32_t ifdr_len;
1713
1714		if ((error = proc_suser(p)) != 0)
1715                        break;
1716
1717		ifnet_lock_exclusive(ifp);
1718		bcopy(&ifdr->ifdr_len, &ifdr_len, sizeof (ifdr_len));
1719		if (ifdr_len > sizeof (ifdr->ifdr_desc) ||
1720		    ifdr_len > ifp->if_desc.ifd_maxlen) {
1721			error = EINVAL;
1722			ifnet_lock_done(ifp);
1723			break;
1724		}
1725
1726		bzero(ifp->if_desc.ifd_desc, ifp->if_desc.ifd_maxlen);
1727		if ((ifp->if_desc.ifd_len = ifdr_len) > 0) {
1728			bcopy(ifdr->ifdr_desc, ifp->if_desc.ifd_desc,
1729			    MIN(ifdr_len, ifp->if_desc.ifd_maxlen));
1730		}
1731		ifnet_lock_done(ifp);
1732		break;
1733	}
1734
1735	case SIOCGIFDESC: {			/* struct if_descreq */
1736		struct if_descreq *ifdr = (struct if_descreq *)(void *)data;
1737		u_int32_t ifdr_len;
1738
1739		ifnet_lock_shared(ifp);
1740		ifdr_len = MIN(ifp->if_desc.ifd_len, sizeof (ifdr->ifdr_desc));
1741		bcopy(&ifdr_len, &ifdr->ifdr_len, sizeof (ifdr_len));
1742		bzero(&ifdr->ifdr_desc, sizeof (ifdr->ifdr_desc));
1743		if (ifdr_len > 0) {
1744			bcopy(ifp->if_desc.ifd_desc, ifdr->ifdr_desc, ifdr_len);
1745		}
1746		ifnet_lock_done(ifp);
1747		break;
1748	}
1749
1750	case SIOCSIFLINKPARAMS: {		/* struct if_linkparamsreq */
1751		struct if_linkparamsreq *iflpr =
1752		    (struct if_linkparamsreq *)(void *)data;
1753		struct ifclassq *ifq = &ifp->if_snd;
1754		struct tb_profile tb = { 0, 0, 0 };
1755
1756		if ((error = proc_suser(p)) != 0)
1757                        break;
1758
1759		IFCQ_LOCK(ifq);
1760		if (!IFCQ_IS_READY(ifq)) {
1761			error = ENXIO;
1762			IFCQ_UNLOCK(ifq);
1763			break;
1764		}
1765		bcopy(&iflpr->iflpr_output_tbr_rate, &tb.rate,
1766		    sizeof (tb.rate));
1767		bcopy(&iflpr->iflpr_output_tbr_percent, &tb.percent,
1768		    sizeof (tb.percent));
1769		error = ifclassq_tbr_set(ifq, &tb, TRUE);
1770		IFCQ_UNLOCK(ifq);
1771		break;
1772	}
1773
1774	case SIOCGIFLINKPARAMS: {		/* struct if_linkparamsreq */
1775		struct if_linkparamsreq *iflpr =
1776		    (struct if_linkparamsreq *)(void *)data;
1777		struct ifclassq *ifq = &ifp->if_snd;
1778		u_int32_t sched_type = PKTSCHEDT_NONE, flags = 0;
1779		u_int64_t tbr_bw = 0, tbr_pct = 0;
1780
1781		IFCQ_LOCK(ifq);
1782#if PF_ALTQ
1783		if (ALTQ_IS_ENABLED(IFCQ_ALTQ(ifq))) {
1784			sched_type = IFCQ_ALTQ(ifq)->altq_type;
1785			flags |= IFLPRF_ALTQ;
1786		} else
1787#endif /* PF_ALTQ */
1788		{
1789			if (IFCQ_IS_ENABLED(ifq))
1790				sched_type = ifq->ifcq_type;
1791		}
1792		bcopy(&sched_type, &iflpr->iflpr_output_sched,
1793		    sizeof (iflpr->iflpr_output_sched));
1794
1795		if (IFCQ_TBR_IS_ENABLED(ifq)) {
1796			tbr_bw = ifq->ifcq_tbr.tbr_rate_raw;
1797			tbr_pct = ifq->ifcq_tbr.tbr_percent;
1798		}
1799		bcopy(&tbr_bw, &iflpr->iflpr_output_tbr_rate,
1800		    sizeof (iflpr->iflpr_output_tbr_rate));
1801		bcopy(&tbr_pct, &iflpr->iflpr_output_tbr_percent,
1802		    sizeof (iflpr->iflpr_output_tbr_percent));
1803		IFCQ_UNLOCK(ifq);
1804
1805		if (ifp->if_output_sched_model ==
1806		    IFNET_SCHED_MODEL_DRIVER_MANAGED)
1807			flags |= IFLPRF_DRVMANAGED;
1808		bcopy(&flags, &iflpr->iflpr_flags, sizeof (iflpr->iflpr_flags));
1809		bcopy(&ifp->if_output_bw, &iflpr->iflpr_output_bw,
1810		    sizeof (iflpr->iflpr_output_bw));
1811		bcopy(&ifp->if_input_bw, &iflpr->iflpr_input_bw,
1812		    sizeof (iflpr->iflpr_input_bw));
1813		break;
1814	}
1815
1816	case SIOCGIFQUEUESTATS: {		/* struct if_qstatsreq */
1817		struct if_qstatsreq *ifqr = (struct if_qstatsreq *)(void *)data;
1818		u_int32_t ifqr_len, ifqr_slot;
1819
1820		bcopy(&ifqr->ifqr_slot, &ifqr_slot, sizeof (ifqr_slot));
1821		bcopy(&ifqr->ifqr_len, &ifqr_len, sizeof (ifqr_len));
1822		error = ifclassq_getqstats(&ifp->if_snd, ifqr_slot,
1823		    ifqr->ifqr_buf, &ifqr_len);
1824		if (error != 0)
1825			ifqr_len = 0;
1826		bcopy(&ifqr_len, &ifqr->ifqr_len, sizeof (ifqr_len));
1827		break;
1828	}
1829
1830	case SIOCSIFTHROTTLE: {			/* struct if_throttlereq */
1831		struct if_throttlereq *ifthr =
1832		    (struct if_throttlereq *)(void *)data;
1833		u_int32_t ifthr_level;
1834
1835		/*
1836		 * XXX: Use priv_check_cred() instead of root check?
1837		 */
1838		if ((error = proc_suser(p)) != 0)
1839                        break;
1840
1841		bcopy(&ifthr->ifthr_level, &ifthr_level, sizeof (ifthr_level));
1842		error = ifnet_set_throttle(ifp, ifthr_level);
1843		if (error == EALREADY)
1844			error = 0;
1845		break;
1846	}
1847
1848	case SIOCGIFTHROTTLE: {			/* struct if_throttlereq */
1849		struct if_throttlereq *ifthr =
1850		    (struct if_throttlereq *)(void *)data;
1851		u_int32_t ifthr_level;
1852
1853		if ((error = ifnet_get_throttle(ifp, &ifthr_level)) == 0) {
1854			bcopy(&ifthr_level, &ifthr->ifthr_level,
1855			    sizeof (ifthr_level));
1856		}
1857		break;
1858	}
1859
1860	default:
1861		if (so->so_proto == NULL) {
1862			error = EOPNOTSUPP;
1863			break;
1864		}
1865
1866		socket_lock(so, 1);
1867		error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
1868		    data, ifp, p));
1869		socket_unlock(so, 1);
1870
1871		if (error == EOPNOTSUPP || error == ENOTSUP) {
1872			error = ifnet_ioctl(ifp,
1873			    so->so_proto->pr_domain->dom_family, cmd, data);
1874		}
1875		break;
1876	}
1877
1878done:
1879	if (ifs != NULL)
1880		_FREE(ifs, M_DEVBUF);
1881
1882	return (error);
1883}
1884
1885static int
1886ifioctl_ifreq(struct socket *so, u_long cmd, struct ifreq *ifr, struct proc *p)
1887{
1888	struct ifnet *ifp;
1889	u_long ocmd = cmd;
1890	int error = 0;
1891	struct kev_msg ev_msg;
1892	struct net_event_data ev_data;
1893
1894	bzero(&ev_data, sizeof (struct net_event_data));
1895	bzero(&ev_msg, sizeof (struct kev_msg));
1896
1897	ifr->ifr_name[IFNAMSIZ - 1] = '\0';
1898
1899	switch (cmd) {
1900	case SIOCIFCREATE:
1901	case SIOCIFCREATE2:
1902                error = proc_suser(p);
1903                if (error)
1904                        return (error);
1905                return (if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name),
1906		    cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL));
1907	case SIOCIFDESTROY:
1908		error = proc_suser(p);
1909		if (error)
1910			return (error);
1911		return (if_clone_destroy(ifr->ifr_name));
1912	}
1913
1914	ifp = ifunit(ifr->ifr_name);
1915	if (ifp == NULL)
1916		return (ENXIO);
1917
1918	switch (cmd) {
1919	case SIOCGIFFLAGS:
1920		ifnet_lock_shared(ifp);
1921		ifr->ifr_flags = ifp->if_flags;
1922		ifnet_lock_done(ifp);
1923		break;
1924
1925	case SIOCGIFEFLAGS:
1926		ifnet_lock_shared(ifp);
1927		ifr->ifr_eflags = ifp->if_eflags;
1928		ifnet_lock_done(ifp);
1929		break;
1930
1931	case SIOCGIFCAP:
1932		ifnet_lock_shared(ifp);
1933		ifr->ifr_reqcap = ifp->if_capabilities;
1934		ifr->ifr_curcap = ifp->if_capenable;
1935		ifnet_lock_done(ifp);
1936		break;
1937
1938#if CONFIG_MACF_NET
1939	case SIOCGIFMAC:
1940		error = mac_ifnet_label_get(kauth_cred_get(), ifr, ifp);
1941		break;
1942#endif
1943	case SIOCGIFMETRIC:
1944		ifnet_lock_shared(ifp);
1945		ifr->ifr_metric = ifp->if_metric;
1946		ifnet_lock_done(ifp);
1947		break;
1948
1949	case SIOCGIFMTU:
1950		ifnet_lock_shared(ifp);
1951		ifr->ifr_mtu = ifp->if_mtu;
1952		ifnet_lock_done(ifp);
1953		break;
1954
1955	case SIOCGIFPHYS:
1956		ifnet_lock_shared(ifp);
1957		ifr->ifr_phys = ifp->if_physical;
1958		ifnet_lock_done(ifp);
1959		break;
1960
1961	case SIOCGIFWAKEFLAGS:
1962		ifnet_lock_shared(ifp);
1963		ifr->ifr_wake_flags = ifnet_get_wake_flags(ifp);
1964		ifnet_lock_done(ifp);
1965		break;
1966
1967	case SIOCGIFGETRTREFCNT:
1968		ifnet_lock_shared(ifp);
1969		ifr->ifr_route_refcnt = ifp->if_route_refcnt;
1970		ifnet_lock_done(ifp);
1971		break;
1972
1973	case SIOCGIFLINKQUALITYMETRIC:
1974		ifnet_lock_shared(ifp);
1975		ifr->ifr_link_quality_metric = ifp->if_lqm;
1976		ifnet_lock_done(ifp);
1977		break;
1978
1979	case SIOCSIFFLAGS:
1980		error = proc_suser(p);
1981		if (error != 0)
1982			break;
1983
1984		(void) ifnet_set_flags(ifp, ifr->ifr_flags,
1985		    (u_int16_t)~IFF_CANTCHANGE);
1986
1987		/*
1988		 * Note that we intentionally ignore any error from below
1989		 * for the SIOCSIFFLAGS case.
1990		 */
1991		(void) ifnet_ioctl(ifp, so->so_proto->pr_domain->dom_family,
1992		    cmd, (caddr_t)ifr);
1993
1994		/*
1995		 * Send the event even upon error from the driver because
1996		 * we changed the flags.
1997		 */
1998		ev_msg.vendor_code    = KEV_VENDOR_APPLE;
1999		ev_msg.kev_class      = KEV_NETWORK_CLASS;
2000		ev_msg.kev_subclass   = KEV_DL_SUBCLASS;
2001
2002		ev_msg.event_code = KEV_DL_SIFFLAGS;
2003		strlcpy(&ev_data.if_name[0], ifp->if_name, IFNAMSIZ);
2004		ev_data.if_family = ifp->if_family;
2005		ev_data.if_unit   = (u_int32_t) ifp->if_unit;
2006		ev_msg.dv[0].data_length = sizeof(struct net_event_data);
2007		ev_msg.dv[0].data_ptr    = &ev_data;
2008		ev_msg.dv[1].data_length = 0;
2009		kev_post_msg(&ev_msg);
2010
2011		ifnet_touch_lastchange(ifp);
2012		break;
2013
2014	case SIOCSIFCAP:
2015		error = proc_suser(p);
2016		if (error != 0)
2017			break;
2018
2019		if ((ifr->ifr_reqcap & ~ifp->if_capabilities)) {
2020			error = EINVAL;
2021			break;
2022		}
2023		error = ifnet_ioctl(ifp, so->so_proto->pr_domain->dom_family,
2024		    cmd, (caddr_t)ifr);
2025
2026		ifnet_touch_lastchange(ifp);
2027		break;
2028
2029#if CONFIG_MACF_NET
2030	case SIOCSIFMAC:
2031		error = mac_ifnet_label_set(kauth_cred_get(), ifr, ifp);
2032		break;
2033#endif
2034	case SIOCSIFMETRIC:
2035		error = proc_suser(p);
2036		if (error != 0)
2037			break;
2038
2039		ifp->if_metric = ifr->ifr_metric;
2040
2041		ev_msg.vendor_code    = KEV_VENDOR_APPLE;
2042		ev_msg.kev_class      = KEV_NETWORK_CLASS;
2043		ev_msg.kev_subclass   = KEV_DL_SUBCLASS;
2044
2045		ev_msg.event_code = KEV_DL_SIFMETRICS;
2046		strlcpy(&ev_data.if_name[0], ifp->if_name, IFNAMSIZ);
2047		ev_data.if_family = ifp->if_family;
2048		ev_data.if_unit   = (u_int32_t) ifp->if_unit;
2049		ev_msg.dv[0].data_length = sizeof(struct net_event_data);
2050		ev_msg.dv[0].data_ptr    = &ev_data;
2051
2052		ev_msg.dv[1].data_length = 0;
2053		kev_post_msg(&ev_msg);
2054
2055		ifnet_touch_lastchange(ifp);
2056		break;
2057
2058	case SIOCSIFPHYS:
2059		error = proc_suser(p);
2060		if (error != 0)
2061			break;
2062
2063		error = ifnet_ioctl(ifp, so->so_proto->pr_domain->dom_family,
2064		    cmd, (caddr_t)ifr);
2065		if (error != 0)
2066			break;
2067
2068		ev_msg.vendor_code    = KEV_VENDOR_APPLE;
2069		ev_msg.kev_class      = KEV_NETWORK_CLASS;
2070		ev_msg.kev_subclass   = KEV_DL_SUBCLASS;
2071
2072		ev_msg.event_code = KEV_DL_SIFPHYS;
2073		strlcpy(&ev_data.if_name[0], ifp->if_name, IFNAMSIZ);
2074		ev_data.if_family = ifp->if_family;
2075		ev_data.if_unit   = (u_int32_t) ifp->if_unit;
2076		ev_msg.dv[0].data_length = sizeof(struct net_event_data);
2077		ev_msg.dv[0].data_ptr    = &ev_data;
2078		ev_msg.dv[1].data_length = 0;
2079		kev_post_msg(&ev_msg);
2080
2081		ifnet_touch_lastchange(ifp);
2082		break;
2083
2084	case SIOCSIFMTU: {
2085		u_int32_t oldmtu = ifp->if_mtu;
2086		struct ifclassq *ifq = &ifp->if_snd;
2087
2088		error = proc_suser(p);
2089		if (error != 0)
2090			break;
2091
2092		if (ifp->if_ioctl == NULL) {
2093			error = EOPNOTSUPP;
2094			break;
2095		}
2096		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) {
2097			error = EINVAL;
2098			break;
2099		}
2100		error = ifnet_ioctl(ifp, so->so_proto->pr_domain->dom_family,
2101		    cmd, (caddr_t)ifr);
2102		if (error != 0)
2103			break;
2104
2105		ev_msg.vendor_code    = KEV_VENDOR_APPLE;
2106		ev_msg.kev_class      = KEV_NETWORK_CLASS;
2107		ev_msg.kev_subclass   = KEV_DL_SUBCLASS;
2108
2109		ev_msg.event_code = KEV_DL_SIFMTU;
2110		strlcpy(&ev_data.if_name[0], ifp->if_name, IFNAMSIZ);
2111		ev_data.if_family = ifp->if_family;
2112		ev_data.if_unit   = (u_int32_t) ifp->if_unit;
2113		ev_msg.dv[0].data_length = sizeof(struct net_event_data);
2114		ev_msg.dv[0].data_ptr    = &ev_data;
2115		ev_msg.dv[1].data_length = 0;
2116		kev_post_msg(&ev_msg);
2117
2118		ifnet_touch_lastchange(ifp);
2119		rt_ifmsg(ifp);
2120
2121		/*
2122		 * If the link MTU changed, do network layer specific procedure
2123		 * and update all route entries associated with the interface,
2124		 * so that their MTU metric gets updated.
2125		 */
2126		if (ifp->if_mtu != oldmtu) {
2127			if_rtmtu_update(ifp);
2128#if INET6
2129			nd6_setmtu(ifp);
2130#endif
2131			/* Inform all transmit queues about the new MTU */
2132			IFCQ_LOCK(ifq);
2133			ifnet_update_sndq(ifq, CLASSQ_EV_LINK_MTU);
2134			IFCQ_UNLOCK(ifq);
2135		}
2136		break;
2137	}
2138
2139	case SIOCADDMULTI:
2140	case SIOCDELMULTI:
2141		error = proc_suser(p);
2142		if (error != 0)
2143			break;
2144
2145		/* Don't allow group membership on non-multicast interfaces. */
2146		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
2147			error = EOPNOTSUPP;
2148			break;
2149		}
2150
2151		/* Don't let users screw up protocols' entries. */
2152		if (ifr->ifr_addr.sa_family != AF_UNSPEC &&
2153		    ifr->ifr_addr.sa_family != AF_LINK) {
2154			error = EINVAL;
2155			break;
2156		}
2157
2158		/*
2159		 * User is permitted to anonymously join a particular link
2160		 * multicast group via SIOCADDMULTI.  Subsequent join requested
2161		 * for the same record which has an outstanding refcnt from a
2162		 * past if_addmulti_anon() will not result in EADDRINUSE error
2163		 * (unlike other BSDs.)  Anonymously leaving a group is also
2164		 * allowed only as long as there is an outstanding refcnt held
2165		 * by a previous anonymous request, or else ENOENT (even if the
2166		 * link-layer multicast membership exists for a network-layer
2167		 * membership.)
2168		 */
2169		if (cmd == SIOCADDMULTI) {
2170			error = if_addmulti_anon(ifp, &ifr->ifr_addr, NULL);
2171			ev_msg.event_code = KEV_DL_ADDMULTI;
2172		} else {
2173			error = if_delmulti_anon(ifp, &ifr->ifr_addr);
2174			ev_msg.event_code = KEV_DL_DELMULTI;
2175		}
2176		if (error != 0)
2177			break;
2178
2179		ev_msg.vendor_code    = KEV_VENDOR_APPLE;
2180		ev_msg.kev_class      = KEV_NETWORK_CLASS;
2181		ev_msg.kev_subclass   = KEV_DL_SUBCLASS;
2182		strlcpy(&ev_data.if_name[0], ifp->if_name, IFNAMSIZ);
2183
2184		ev_data.if_family = ifp->if_family;
2185		ev_data.if_unit   = (u_int32_t) ifp->if_unit;
2186		ev_msg.dv[0].data_length = sizeof(struct net_event_data);
2187		ev_msg.dv[0].data_ptr    = &ev_data;
2188		ev_msg.dv[1].data_length = 0;
2189		kev_post_msg(&ev_msg);
2190
2191		ifnet_touch_lastchange(ifp);
2192		break;
2193
2194	case SIOCDIFPHYADDR:
2195	case SIOCSIFMEDIA:
2196	case SIOCSIFGENERIC:
2197	case SIOCSIFLLADDR:
2198	case SIOCSIFALTMTU:
2199	case SIOCSIFVLAN:
2200	case SIOCSIFBOND:
2201		error = proc_suser(p);
2202		if (error != 0)
2203			break;
2204
2205		error = ifnet_ioctl(ifp, so->so_proto->pr_domain->dom_family,
2206		    cmd, (caddr_t)ifr);
2207		if (error != 0)
2208			break;
2209
2210		ifnet_touch_lastchange(ifp);
2211		break;
2212
2213	case SIOCGIFPSRCADDR:
2214	case SIOCGIFPDSTADDR:
2215	case SIOCGIFGENERIC:
2216	case SIOCGIFDEVMTU:
2217	case SIOCGIFVLAN:
2218	case SIOCGIFBOND:
2219		error = ifnet_ioctl(ifp, so->so_proto->pr_domain->dom_family,
2220		    cmd, (caddr_t)ifr);
2221		break;
2222
2223	case SIOCSIFOPPORTUNISTIC:
2224	case SIOCGIFOPPORTUNISTIC:
2225		error = ifnet_getset_opportunistic(ifp, cmd, ifr, p);
2226		break;
2227
2228	case SIOCSIFDSTADDR:
2229	case SIOCSIFADDR:
2230	case SIOCSIFBRDADDR:
2231	case SIOCSIFNETMASK:
2232	case OSIOCGIFADDR:
2233	case OSIOCGIFDSTADDR:
2234	case OSIOCGIFBRDADDR:
2235	case OSIOCGIFNETMASK:
2236	case SIOCSIFKPI:
2237		VERIFY(so->so_proto != NULL);
2238
2239		if (cmd == SIOCSIFDSTADDR || cmd == SIOCSIFADDR ||
2240		    cmd == SIOCSIFBRDADDR || cmd == SIOCSIFNETMASK) {
2241#if BYTE_ORDER != BIG_ENDIAN
2242			if (ifr->ifr_addr.sa_family == 0 &&
2243			    ifr->ifr_addr.sa_len < 16) {
2244				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
2245				ifr->ifr_addr.sa_len = 16;
2246			}
2247#else
2248			if (ifr->ifr_addr.sa_len == 0)
2249				ifr->ifr_addr.sa_len = 16;
2250#endif
2251		} else if (cmd == OSIOCGIFADDR) {
2252			cmd = SIOCGIFADDR;	/* struct ifreq */
2253		} else if (cmd == OSIOCGIFDSTADDR) {
2254			cmd = SIOCGIFDSTADDR;	/* struct ifreq */
2255		} else if (cmd == OSIOCGIFBRDADDR) {
2256			cmd = SIOCGIFBRDADDR;	/* struct ifreq */
2257		} else if (cmd == OSIOCGIFNETMASK) {
2258			cmd = SIOCGIFNETMASK;	/* struct ifreq */
2259		}
2260
2261		socket_lock(so, 1);
2262		error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
2263		    (caddr_t)ifr, ifp, p));
2264		socket_unlock(so, 1);
2265
2266		switch (ocmd) {
2267		case OSIOCGIFADDR:
2268		case OSIOCGIFDSTADDR:
2269		case OSIOCGIFBRDADDR:
2270		case OSIOCGIFNETMASK:
2271			bcopy(&ifr->ifr_addr.sa_family, &ifr->ifr_addr,
2272			    sizeof (u_short));
2273		}
2274
2275		if (cmd == SIOCSIFKPI) {
2276			int temperr = proc_suser(p);
2277			if (temperr != 0)
2278				error = temperr;
2279		}
2280
2281		if (error == EOPNOTSUPP || error == ENOTSUP) {
2282			error = ifnet_ioctl(ifp,
2283			    so->so_proto->pr_domain->dom_family, cmd,
2284			    (caddr_t)ifr);
2285		}
2286		break;
2287
2288	default:
2289		VERIFY(0);
2290		/* NOTREACHED */
2291	}
2292
2293	return (error);
2294}
2295
2296int
2297ifioctllocked(struct socket *so, u_long cmd, caddr_t data, struct proc *p)
2298{
2299	int error;
2300
2301	socket_unlock(so, 0);
2302	error = ifioctl(so, cmd, data, p);
2303	socket_lock(so, 0);
2304	return(error);
2305}
2306
2307/*
2308 * Set/clear promiscuous mode on interface ifp based on the truth value
2309 * of pswitch.  The calls are reference counted so that only the first
2310 * "on" request actually has an effect, as does the final "off" request.
2311 * Results are undefined if the "off" and "on" requests are not matched.
2312 */
2313errno_t
2314ifnet_set_promiscuous(
2315	ifnet_t	ifp,
2316	int pswitch)
2317{
2318	int error = 0;
2319	int oldflags = 0;
2320	int newflags = 0;
2321
2322	ifnet_lock_exclusive(ifp);
2323	oldflags = ifp->if_flags;
2324	ifp->if_pcount += pswitch ? 1 : -1;
2325
2326	if (ifp->if_pcount > 0)
2327		ifp->if_flags |= IFF_PROMISC;
2328	else
2329		ifp->if_flags &= ~IFF_PROMISC;
2330
2331	newflags = ifp->if_flags;
2332	ifnet_lock_done(ifp);
2333
2334	if (newflags != oldflags && (newflags & IFF_UP) != 0) {
2335		error = ifnet_ioctl(ifp, 0, SIOCSIFFLAGS, NULL);
2336		if (error == 0) {
2337			rt_ifmsg(ifp);
2338		} else {
2339			ifnet_lock_exclusive(ifp);
2340			// revert the flags
2341			ifp->if_pcount -= pswitch ? 1 : -1;
2342			if (ifp->if_pcount > 0)
2343			    ifp->if_flags |= IFF_PROMISC;
2344			else
2345			    ifp->if_flags &= ~IFF_PROMISC;
2346			ifnet_lock_done(ifp);
2347		}
2348	}
2349
2350	if (newflags != oldflags) {
2351		log(LOG_INFO, "%s%d: promiscuous mode %s%s\n",
2352		    ifp->if_name, ifp->if_unit,
2353		    (newflags & IFF_PROMISC) != 0 ? "enable" : "disable",
2354		    error != 0 ? " failed" : " succeeded");
2355	}
2356	return error;
2357}
2358
2359/*
2360 * Return interface configuration
2361 * of system.  List may be used
2362 * in later ioctl's (above) to get
2363 * other information.
2364 */
2365/*ARGSUSED*/
2366static int
2367ifconf(u_long cmd, user_addr_t ifrp, int * ret_space)
2368{
2369	struct ifnet *ifp = NULL;
2370	struct ifaddr *ifa;
2371	struct ifreq ifr;
2372	int error = 0;
2373	size_t space;
2374
2375	/*
2376	 * Zero the ifr buffer to make sure we don't
2377	 * disclose the contents of the stack.
2378	 */
2379	bzero(&ifr, sizeof(struct ifreq));
2380
2381	space = *ret_space;
2382	ifnet_head_lock_shared();
2383	for (ifp = ifnet_head.tqh_first; space > sizeof(ifr) &&
2384	    ifp; ifp = ifp->if_link.tqe_next) {
2385		char workbuf[64];
2386		size_t ifnlen, addrs;
2387
2388		ifnlen = snprintf(workbuf, sizeof(workbuf),
2389		    "%s%d", ifp->if_name, ifp->if_unit);
2390		if(ifnlen + 1 > sizeof ifr.ifr_name) {
2391			error = ENAMETOOLONG;
2392			break;
2393		} else {
2394			strlcpy(ifr.ifr_name, workbuf, IFNAMSIZ);
2395		}
2396
2397		ifnet_lock_shared(ifp);
2398
2399		addrs = 0;
2400		ifa = ifp->if_addrhead.tqh_first;
2401		for ( ; space > sizeof (ifr) && ifa;
2402		    ifa = ifa->ifa_link.tqe_next) {
2403			struct sockaddr *sa;
2404
2405			IFA_LOCK(ifa);
2406			sa = ifa->ifa_addr;
2407#ifndef __APPLE__
2408			if (curproc->p_prison && prison_if(curproc, sa)) {
2409				IFA_UNLOCK(ifa);
2410				continue;
2411			}
2412#endif
2413			addrs++;
2414			if (cmd == OSIOCGIFCONF32 || cmd == OSIOCGIFCONF64) {
2415				struct osockaddr *osa =
2416				    (struct osockaddr *)(void *)&ifr.ifr_addr;
2417				ifr.ifr_addr = *sa;
2418				osa->sa_family = sa->sa_family;
2419				error = copyout((caddr_t)&ifr, ifrp,
2420				    sizeof (ifr));
2421				ifrp += sizeof(struct ifreq);
2422			} else if (sa->sa_len <= sizeof(*sa)) {
2423				ifr.ifr_addr = *sa;
2424				error = copyout((caddr_t)&ifr, ifrp,
2425				    sizeof (ifr));
2426				ifrp += sizeof(struct ifreq);
2427			} else {
2428				if (space <
2429				    sizeof (ifr) + sa->sa_len - sizeof(*sa)) {
2430					IFA_UNLOCK(ifa);
2431					break;
2432				}
2433				space -= sa->sa_len - sizeof(*sa);
2434				error = copyout((caddr_t)&ifr, ifrp,
2435				    sizeof (ifr.ifr_name));
2436				if (error == 0) {
2437				    error = copyout((caddr_t)sa, (ifrp +
2438				        offsetof(struct ifreq, ifr_addr)),
2439					sa->sa_len);
2440				}
2441				ifrp += (sa->sa_len + offsetof(struct ifreq,
2442				    ifr_addr));
2443			}
2444			IFA_UNLOCK(ifa);
2445			if (error)
2446				break;
2447			space -= sizeof (ifr);
2448		}
2449		ifnet_lock_done(ifp);
2450
2451		if (error)
2452			break;
2453		if (!addrs) {
2454			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
2455			error = copyout((caddr_t)&ifr, ifrp, sizeof (ifr));
2456			if (error)
2457				break;
2458			space -= sizeof (ifr);
2459			ifrp += sizeof(struct ifreq);
2460		}
2461	}
2462	ifnet_head_done();
2463	*ret_space -= space;
2464	return (error);
2465}
2466
2467/*
2468 * Just like if_promisc(), but for all-multicast-reception mode.
2469 */
2470int
2471if_allmulti(struct ifnet *ifp, int onswitch)
2472{
2473	int error = 0;
2474	int	modified = 0;
2475
2476	ifnet_lock_exclusive(ifp);
2477
2478	if (onswitch) {
2479		if (ifp->if_amcount++ == 0) {
2480			ifp->if_flags |= IFF_ALLMULTI;
2481			modified = 1;
2482		}
2483	} else {
2484		if (ifp->if_amcount > 1) {
2485			ifp->if_amcount--;
2486		} else {
2487			ifp->if_amcount = 0;
2488			ifp->if_flags &= ~IFF_ALLMULTI;
2489			modified = 1;
2490		}
2491	}
2492	ifnet_lock_done(ifp);
2493
2494	if (modified)
2495		error = ifnet_ioctl(ifp, 0, SIOCSIFFLAGS, NULL);
2496
2497	if (error == 0)
2498		rt_ifmsg(ifp);
2499	return error;
2500}
2501
2502static struct ifmultiaddr *
2503ifma_alloc(int how)
2504{
2505	struct ifmultiaddr *ifma;
2506
2507	ifma = (how == M_WAITOK) ? zalloc(ifma_zone) :
2508	    zalloc_noblock(ifma_zone);
2509
2510	if (ifma != NULL) {
2511		bzero(ifma, ifma_size);
2512		lck_mtx_init(&ifma->ifma_lock, ifa_mtx_grp, ifa_mtx_attr);
2513		ifma->ifma_debug |= IFD_ALLOC;
2514		if (ifma_debug != 0) {
2515			ifma->ifma_debug |= IFD_DEBUG;
2516			ifma->ifma_trace = ifma_trace;
2517		}
2518	}
2519	return (ifma);
2520}
2521
2522static void
2523ifma_free(struct ifmultiaddr *ifma)
2524{
2525	IFMA_LOCK(ifma);
2526
2527	if (ifma->ifma_protospec != NULL) {
2528		panic("%s: Protospec not NULL for ifma=%p", __func__, ifma);
2529		/* NOTREACHED */
2530	} else if ((ifma->ifma_flags & IFMAF_ANONYMOUS) ||
2531	    ifma->ifma_anoncnt != 0) {
2532		panic("%s: Freeing ifma=%p with outstanding anon req",
2533		    __func__, ifma);
2534		/* NOTREACHED */
2535	} else if (ifma->ifma_debug & IFD_ATTACHED) {
2536		panic("%s: ifma=%p attached to ifma_ifp=%p is being freed",
2537		    __func__, ifma, ifma->ifma_ifp);
2538		/* NOTREACHED */
2539	} else if (!(ifma->ifma_debug & IFD_ALLOC)) {
2540		panic("%s: ifma %p cannot be freed", __func__, ifma);
2541		/* NOTREACHED */
2542	} else if (ifma->ifma_refcount != 0) {
2543		panic("%s: non-zero refcount ifma=%p", __func__, ifma);
2544		/* NOTREACHED */
2545	} else if (ifma->ifma_reqcnt != 0) {
2546		panic("%s: non-zero reqcnt ifma=%p", __func__, ifma);
2547		/* NOTREACHED */
2548	} else if (ifma->ifma_ifp != NULL) {
2549		panic("%s: non-NULL ifma_ifp=%p for ifma=%p", __func__,
2550		    ifma->ifma_ifp, ifma);
2551		/* NOTREACHED */
2552	} else if (ifma->ifma_ll != NULL) {
2553		panic("%s: non-NULL ifma_ll=%p for ifma=%p", __func__,
2554		    ifma->ifma_ll, ifma);
2555		/* NOTREACHED */
2556	}
2557	ifma->ifma_debug &= ~IFD_ALLOC;
2558	if ((ifma->ifma_debug & (IFD_DEBUG | IFD_TRASHED)) ==
2559	    (IFD_DEBUG | IFD_TRASHED)) {
2560		lck_mtx_lock(&ifma_trash_lock);
2561		TAILQ_REMOVE(&ifma_trash_head, (struct ifmultiaddr_dbg *)ifma,
2562		    ifma_trash_link);
2563		lck_mtx_unlock(&ifma_trash_lock);
2564		ifma->ifma_debug &= ~IFD_TRASHED;
2565	}
2566	IFMA_UNLOCK(ifma);
2567
2568	if (ifma->ifma_addr != NULL) {
2569		FREE(ifma->ifma_addr, M_IFADDR);
2570		ifma->ifma_addr = NULL;
2571	}
2572	lck_mtx_destroy(&ifma->ifma_lock, ifa_mtx_grp);
2573	zfree(ifma_zone, ifma);
2574}
2575
2576static void
2577ifma_trace(struct ifmultiaddr *ifma, int refhold)
2578{
2579	struct ifmultiaddr_dbg *ifma_dbg = (struct ifmultiaddr_dbg *)ifma;
2580	ctrace_t *tr;
2581	u_int32_t idx;
2582	u_int16_t *cnt;
2583
2584	if (!(ifma->ifma_debug & IFD_DEBUG)) {
2585		panic("%s: ifma %p has no debug structure", __func__, ifma);
2586		/* NOTREACHED */
2587	}
2588	if (refhold) {
2589		cnt = &ifma_dbg->ifma_refhold_cnt;
2590		tr = ifma_dbg->ifma_refhold;
2591	} else {
2592		cnt = &ifma_dbg->ifma_refrele_cnt;
2593		tr = ifma_dbg->ifma_refrele;
2594	}
2595
2596	idx = atomic_add_16_ov(cnt, 1) % IFMA_TRACE_HIST_SIZE;
2597	ctrace_record(&tr[idx]);
2598}
2599
2600void
2601ifma_addref(struct ifmultiaddr *ifma, int locked)
2602{
2603	if (!locked)
2604		IFMA_LOCK(ifma);
2605	else
2606		IFMA_LOCK_ASSERT_HELD(ifma);
2607
2608	if (++ifma->ifma_refcount == 0) {
2609		panic("%s: ifma=%p wraparound refcnt", __func__, ifma);
2610		/* NOTREACHED */
2611	} else if (ifma->ifma_trace != NULL) {
2612		(*ifma->ifma_trace)(ifma, TRUE);
2613	}
2614	if (!locked)
2615		IFMA_UNLOCK(ifma);
2616}
2617
2618void
2619ifma_remref(struct ifmultiaddr *ifma)
2620{
2621	struct ifmultiaddr *ll;
2622
2623	IFMA_LOCK(ifma);
2624
2625	if (ifma->ifma_refcount == 0) {
2626		panic("%s: ifma=%p negative refcnt", __func__, ifma);
2627		/* NOTREACHED */
2628	} else if (ifma->ifma_trace != NULL) {
2629		(*ifma->ifma_trace)(ifma, FALSE);
2630	}
2631
2632	--ifma->ifma_refcount;
2633	if (ifma->ifma_refcount > 0) {
2634		IFMA_UNLOCK(ifma);
2635		return;
2636	}
2637
2638	ll = ifma->ifma_ll;
2639	ifma->ifma_ifp = NULL;
2640	ifma->ifma_ll = NULL;
2641	IFMA_UNLOCK(ifma);
2642	ifma_free(ifma);	/* deallocate it */
2643
2644	if (ll != NULL)
2645		IFMA_REMREF(ll);
2646}
2647
2648static void
2649if_attach_ifma(struct ifnet *ifp, struct ifmultiaddr *ifma, int anon)
2650{
2651	ifnet_lock_assert(ifp, IFNET_LCK_ASSERT_EXCLUSIVE);
2652	IFMA_LOCK_ASSERT_HELD(ifma);
2653
2654	if (ifma->ifma_ifp != ifp) {
2655		panic("%s: Mismatch ifma_ifp=%p != ifp=%p", __func__,
2656		    ifma->ifma_ifp, ifp);
2657		/* NOTREACHED */
2658	} else if (ifma->ifma_debug & IFD_ATTACHED) {
2659		panic("%s: Attempt to attach an already attached ifma=%p",
2660		    __func__, ifma);
2661		/* NOTREACHED */
2662	} else if (anon && (ifma->ifma_flags & IFMAF_ANONYMOUS)) {
2663		panic("%s: ifma=%p unexpected IFMAF_ANONYMOUS", __func__, ifma);
2664		/* NOTREACHED */
2665	} else if (ifma->ifma_debug & IFD_TRASHED) {
2666		panic("%s: Attempt to reattach a detached ifma=%p",
2667		    __func__, ifma);
2668		/* NOTREACHED */
2669	}
2670
2671	ifma->ifma_reqcnt++;
2672	VERIFY(ifma->ifma_reqcnt == 1);
2673	IFMA_ADDREF_LOCKED(ifma);
2674	ifma->ifma_debug |= IFD_ATTACHED;
2675	if (anon) {
2676		ifma->ifma_anoncnt++;
2677		VERIFY(ifma->ifma_anoncnt == 1);
2678		ifma->ifma_flags |= IFMAF_ANONYMOUS;
2679	}
2680
2681	LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
2682}
2683
2684static int
2685if_detach_ifma(struct ifnet *ifp, struct ifmultiaddr *ifma, int anon)
2686{
2687	ifnet_lock_assert(ifp, IFNET_LCK_ASSERT_EXCLUSIVE);
2688	IFMA_LOCK_ASSERT_HELD(ifma);
2689
2690	if (ifma->ifma_reqcnt == 0) {
2691		panic("%s: ifma=%p negative reqcnt", __func__, ifma);
2692		/* NOTREACHED */
2693	} else if (anon && !(ifma->ifma_flags & IFMAF_ANONYMOUS)) {
2694		panic("%s: ifma=%p missing IFMAF_ANONYMOUS", __func__, ifma);
2695		/* NOTREACHED */
2696	} else if (anon && ifma->ifma_anoncnt == 0) {
2697		panic("%s: ifma=%p negative anonreqcnt", __func__, ifma);
2698		/* NOTREACHED */
2699	} else if (ifma->ifma_ifp != ifp) {
2700		panic("%s: Mismatch ifma_ifp=%p, ifp=%p", __func__,
2701		    ifma->ifma_ifp, ifp);
2702		/* NOTREACHED */
2703	}
2704
2705	if (anon) {
2706		--ifma->ifma_anoncnt;
2707		if (ifma->ifma_anoncnt > 0)
2708			return (0);
2709		ifma->ifma_flags &= ~IFMAF_ANONYMOUS;
2710	}
2711
2712	--ifma->ifma_reqcnt;
2713	if (ifma->ifma_reqcnt > 0)
2714		return (0);
2715
2716	if (ifma->ifma_protospec != NULL) {
2717		panic("%s: Protospec not NULL for ifma=%p", __func__, ifma);
2718		/* NOTREACHED */
2719	} else if ((ifma->ifma_flags & IFMAF_ANONYMOUS) ||
2720	    ifma->ifma_anoncnt != 0) {
2721		panic("%s: Detaching ifma=%p with outstanding anon req",
2722		    __func__, ifma);
2723		/* NOTREACHED */
2724	} else if (!(ifma->ifma_debug & IFD_ATTACHED)) {
2725		panic("%s: Attempt to detach an unattached address ifma=%p",
2726		    __func__, ifma);
2727		/* NOTREACHED */
2728	} else if (ifma->ifma_debug & IFD_TRASHED) {
2729		panic("%s: ifma %p is already in trash list", __func__, ifma);
2730		/* NOTREACHED */
2731	}
2732
2733	/*
2734	 * NOTE: Caller calls IFMA_REMREF
2735	 */
2736	ifma->ifma_debug &= ~IFD_ATTACHED;
2737	LIST_REMOVE(ifma, ifma_link);
2738	if (LIST_EMPTY(&ifp->if_multiaddrs))
2739		ifp->if_updatemcasts = 0;
2740
2741	if (ifma->ifma_debug & IFD_DEBUG) {
2742		/* Become a regular mutex, just in case */
2743		IFMA_CONVERT_LOCK(ifma);
2744		lck_mtx_lock(&ifma_trash_lock);
2745		TAILQ_INSERT_TAIL(&ifma_trash_head,
2746		    (struct ifmultiaddr_dbg *)ifma, ifma_trash_link);
2747		lck_mtx_unlock(&ifma_trash_lock);
2748		ifma->ifma_debug |= IFD_TRASHED;
2749	}
2750
2751	return (1);
2752}
2753
2754/*
2755 * Find an ifmultiaddr that matches a socket address on an interface.
2756 *
2757 * Caller is responsible for holding the ifnet_lock while calling
2758 * this function.
2759 */
2760static int
2761if_addmulti_doesexist(struct ifnet *ifp, const struct sockaddr *sa,
2762    struct ifmultiaddr **retifma, int anon)
2763{
2764	struct ifmultiaddr *ifma;
2765
2766	for (ifma = LIST_FIRST(&ifp->if_multiaddrs); ifma != NULL;
2767	     ifma = LIST_NEXT(ifma, ifma_link)) {
2768		IFMA_LOCK_SPIN(ifma);
2769		if (!equal(sa, ifma->ifma_addr)) {
2770			IFMA_UNLOCK(ifma);
2771			continue;
2772		}
2773		if (anon) {
2774			VERIFY(!(ifma->ifma_flags & IFMAF_ANONYMOUS) ||
2775			    ifma->ifma_anoncnt != 0);
2776			VERIFY((ifma->ifma_flags & IFMAF_ANONYMOUS) ||
2777			    ifma->ifma_anoncnt == 0);
2778			ifma->ifma_anoncnt++;
2779			if (!(ifma->ifma_flags & IFMAF_ANONYMOUS)) {
2780				VERIFY(ifma->ifma_anoncnt == 1);
2781				ifma->ifma_flags |= IFMAF_ANONYMOUS;
2782			}
2783		}
2784		if (!anon || ifma->ifma_anoncnt == 1) {
2785			ifma->ifma_reqcnt++;
2786			VERIFY(ifma->ifma_reqcnt > 1);
2787		}
2788		if (retifma != NULL) {
2789			*retifma = ifma;
2790			IFMA_ADDREF_LOCKED(ifma);
2791		}
2792		IFMA_UNLOCK(ifma);
2793		return (0);
2794	}
2795	return (ENOENT);
2796}
2797
2798/*
2799 * Radar 3642395, make sure all multicasts are in a standard format.
2800 */
2801static struct sockaddr*
2802copy_and_normalize(const struct sockaddr *original)
2803{
2804	int			alen = 0;
2805	const u_char		*aptr = NULL;
2806	struct sockaddr		*copy = NULL;
2807	struct sockaddr_dl	*sdl_new = NULL;
2808	int			len = 0;
2809
2810	if (original->sa_family != AF_LINK &&
2811	    original->sa_family != AF_UNSPEC) {
2812		/* Just make a copy */
2813		MALLOC(copy, struct sockaddr*, original->sa_len,
2814		    M_IFADDR, M_WAITOK);
2815		if (copy != NULL)
2816			bcopy(original, copy, original->sa_len);
2817		return (copy);
2818	}
2819
2820	switch (original->sa_family) {
2821		case AF_LINK: {
2822			const struct sockaddr_dl *sdl_original =
2823			    (struct sockaddr_dl*)(uintptr_t)(size_t)original;
2824
2825			if (sdl_original->sdl_nlen + sdl_original->sdl_alen +
2826			    sdl_original->sdl_slen +
2827			    offsetof(struct sockaddr_dl, sdl_data) >
2828			    sdl_original->sdl_len)
2829				return (NULL);
2830
2831			alen = sdl_original->sdl_alen;
2832			aptr = CONST_LLADDR(sdl_original);
2833		}
2834		break;
2835
2836		case AF_UNSPEC: {
2837			if (original->sa_len < ETHER_ADDR_LEN +
2838			    offsetof(struct sockaddr, sa_data)) {
2839				return (NULL);
2840			}
2841
2842			alen = ETHER_ADDR_LEN;
2843			aptr = (const u_char*)original->sa_data;
2844		}
2845		break;
2846	}
2847
2848	if (alen == 0 || aptr == NULL)
2849		return (NULL);
2850
2851	len = alen + offsetof(struct sockaddr_dl, sdl_data);
2852	MALLOC(sdl_new, struct sockaddr_dl*, len, M_IFADDR, M_WAITOK);
2853
2854	if (sdl_new != NULL) {
2855		bzero(sdl_new, len);
2856		sdl_new->sdl_len = len;
2857		sdl_new->sdl_family = AF_LINK;
2858		sdl_new->sdl_alen = alen;
2859		bcopy(aptr, LLADDR(sdl_new), alen);
2860	}
2861
2862	return ((struct sockaddr*)sdl_new);
2863}
2864
2865/*
2866 * Network-layer protocol domains which hold references to the underlying
2867 * link-layer record must use this routine.
2868 */
2869int
2870if_addmulti(struct ifnet *ifp, const struct sockaddr *sa,
2871    struct ifmultiaddr **retifma)
2872{
2873	return (if_addmulti_common(ifp, sa, retifma, 0));
2874}
2875
2876/*
2877 * Anything other than network-layer protocol domains which hold references
2878 * to the underlying link-layer record must use this routine: SIOCADDMULTI
2879 * ioctl, ifnet_add_multicast(), AppleTalk, if_bond.
2880 */
2881int
2882if_addmulti_anon(struct ifnet *ifp, const struct sockaddr *sa,
2883    struct ifmultiaddr **retifma)
2884{
2885	return (if_addmulti_common(ifp, sa, retifma, 1));
2886}
2887
2888/*
2889 * Register an additional multicast address with a network interface.
2890 *
2891 * - If the address is already present, bump the reference count on the
2892 *   address and return.
2893 * - If the address is not link-layer, look up a link layer address.
2894 * - Allocate address structures for one or both addresses, and attach to the
2895 *   multicast address list on the interface.  If automatically adding a link
2896 *   layer address, the protocol address will own a reference to the link
2897 *   layer address, to be freed when it is freed.
2898 * - Notify the network device driver of an addition to the multicast address
2899 *   list.
2900 *
2901 * 'sa' points to caller-owned memory with the desired multicast address.
2902 *
2903 * 'retifma' will be used to return a pointer to the resulting multicast
2904 * address reference, if desired.
2905 *
2906 * 'anon' indicates a link-layer address with no protocol address reference
2907 * made to it.  Anything other than network-layer protocol domain requests
2908 * are considered as anonymous.
2909 */
2910static int
2911if_addmulti_common(struct ifnet *ifp, const struct sockaddr *sa,
2912    struct ifmultiaddr **retifma, int anon)
2913{
2914	struct sockaddr_storage storage;
2915	struct sockaddr *llsa = NULL;
2916	struct sockaddr *dupsa = NULL;
2917	int error = 0, ll_firstref = 0, lladdr;
2918	struct ifmultiaddr *ifma = NULL;
2919	struct ifmultiaddr *llifma = NULL;
2920
2921	/* Only AF_UNSPEC/AF_LINK is allowed for an "anonymous" address */
2922	VERIFY(!anon || sa->sa_family == AF_UNSPEC ||
2923	    sa->sa_family == AF_LINK);
2924
2925	/* If sa is a AF_LINK or AF_UNSPEC, duplicate and normalize it */
2926	if (sa->sa_family == AF_LINK || sa->sa_family == AF_UNSPEC) {
2927		dupsa = copy_and_normalize(sa);
2928		if (dupsa == NULL) {
2929			error = ENOMEM;
2930			goto cleanup;
2931		}
2932		sa = dupsa;
2933	}
2934
2935	ifnet_lock_exclusive(ifp);
2936	if (!(ifp->if_flags & IFF_MULTICAST)) {
2937		error = EADDRNOTAVAIL;
2938		ifnet_lock_done(ifp);
2939		goto cleanup;
2940	}
2941
2942	/* If the address is already present, return a new reference to it */
2943	error = if_addmulti_doesexist(ifp, sa, retifma, anon);
2944	ifnet_lock_done(ifp);
2945	if (error == 0)
2946		goto cleanup;
2947
2948	/*
2949	 * The address isn't already present; give the link layer a chance
2950	 * to accept/reject it, and also find out which AF_LINK address this
2951	 * maps to, if it isn't one already.
2952	 */
2953	error = dlil_resolve_multi(ifp, sa, (struct sockaddr *)&storage,
2954	    sizeof (storage));
2955	if (error == 0 && storage.ss_len != 0) {
2956		llsa = copy_and_normalize((struct sockaddr *)&storage);
2957		if (llsa == NULL) {
2958			error = ENOMEM;
2959			goto cleanup;
2960		}
2961
2962		llifma = ifma_alloc(M_WAITOK);
2963		if (llifma == NULL) {
2964			error = ENOMEM;
2965			goto cleanup;
2966		}
2967	}
2968
2969	/* to be similar to FreeBSD */
2970	if (error == EOPNOTSUPP)
2971		error = 0;
2972	else if (error != 0)
2973		goto cleanup;
2974
2975	/* Allocate while we aren't holding any locks */
2976	if (dupsa == NULL) {
2977		dupsa = copy_and_normalize(sa);
2978		if (dupsa == NULL) {
2979			error = ENOMEM;
2980			goto cleanup;
2981		}
2982	}
2983	ifma = ifma_alloc(M_WAITOK);
2984	if (ifma == NULL) {
2985		error = ENOMEM;
2986		goto cleanup;
2987	}
2988
2989	ifnet_lock_exclusive(ifp);
2990	/*
2991	 * Check again for the matching multicast.
2992	 */
2993	error = if_addmulti_doesexist(ifp, sa, retifma, anon);
2994	if (error == 0) {
2995		ifnet_lock_done(ifp);
2996		goto cleanup;
2997	}
2998
2999	if (llifma != NULL) {
3000		VERIFY(!anon);	/* must not get here if "anonymous" */
3001		if (if_addmulti_doesexist(ifp, llsa, &ifma->ifma_ll, 0) == 0) {
3002			FREE(llsa, M_IFADDR);
3003			llsa = NULL;
3004			ifma_free(llifma);
3005			llifma = NULL;
3006			VERIFY(ifma->ifma_ll->ifma_ifp == ifp);
3007		} else {
3008			ll_firstref = 1;
3009			llifma->ifma_addr = llsa;
3010			llifma->ifma_ifp = ifp;
3011			IFMA_LOCK(llifma);
3012			if_attach_ifma(ifp, llifma, 0);
3013			/* add extra refcnt for ifma */
3014			IFMA_ADDREF_LOCKED(llifma);
3015			IFMA_UNLOCK(llifma);
3016			ifma->ifma_ll = llifma;
3017		}
3018	}
3019
3020	/* "anonymous" request should not result in network address */
3021	VERIFY(!anon || ifma->ifma_ll == NULL);
3022
3023	ifma->ifma_addr = dupsa;
3024	ifma->ifma_ifp = ifp;
3025	IFMA_LOCK(ifma);
3026	if_attach_ifma(ifp, ifma, anon);
3027	IFMA_ADDREF_LOCKED(ifma);		/* for this routine */
3028	if (retifma != NULL) {
3029		*retifma = ifma;
3030		IFMA_ADDREF_LOCKED(*retifma);	/* for caller */
3031	}
3032	lladdr = (ifma->ifma_addr->sa_family == AF_UNSPEC ||
3033	    ifma->ifma_addr->sa_family == AF_LINK);
3034	IFMA_UNLOCK(ifma);
3035	ifnet_lock_done(ifp);
3036
3037	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3038	IFMA_REMREF(ifma);			/* for this routine */
3039
3040	/*
3041	 * We are certain we have added something, so call down to the
3042	 * interface to let them know about it.  Do this only for newly-
3043	 * added AF_LINK/AF_UNSPEC address in the if_multiaddrs set.
3044	 */
3045	if (lladdr || ll_firstref)
3046		(void) ifnet_ioctl(ifp, 0, SIOCADDMULTI, NULL);
3047
3048	if (ifp->if_updatemcasts > 0)
3049		ifp->if_updatemcasts = 0;
3050
3051	return (0);
3052
3053cleanup:
3054	if (ifma != NULL)
3055		ifma_free(ifma);
3056	if (dupsa != NULL)
3057		FREE(dupsa, M_IFADDR);
3058	if (llifma != NULL)
3059		ifma_free(llifma);
3060	if (llsa != NULL)
3061		FREE(llsa, M_IFADDR);
3062
3063	return (error);
3064}
3065
3066/*
3067 * Delete a multicast group membership by network-layer group address.
3068 * This routine is deprecated.
3069 */
3070int
3071if_delmulti(struct ifnet *ifp, const struct sockaddr *sa)
3072{
3073	return (if_delmulti_common(NULL, ifp, sa, 0));
3074}
3075
3076/*
3077 * Delete a multicast group membership by group membership pointer.
3078 * Network-layer protocol domains must use this routine.
3079 */
3080int
3081if_delmulti_ifma(struct ifmultiaddr *ifma)
3082{
3083	return (if_delmulti_common(ifma, NULL, NULL, 0));
3084}
3085
3086/*
3087 * Anything other than network-layer protocol domains which hold references
3088 * to the underlying link-layer record must use this routine: SIOCDELMULTI
3089 * ioctl, ifnet_remove_multicast(), AppleTalk, if_bond.
3090 */
3091int
3092if_delmulti_anon(struct ifnet *ifp, const struct sockaddr *sa)
3093{
3094	return (if_delmulti_common(NULL, ifp, sa, 1));
3095}
3096
3097/*
3098 * Delete a multicast group membership by network-layer group address.
3099 *
3100 * Returns ENOENT if the entry could not be found.
3101 */
3102static int
3103if_delmulti_common(struct ifmultiaddr *ifma, struct ifnet *ifp,
3104    const struct sockaddr *sa, int anon)
3105{
3106	struct sockaddr		*dupsa = NULL;
3107	int			lastref, ll_lastref = 0, lladdr;
3108	struct ifmultiaddr	*ll = NULL;
3109
3110	/* sanity check for callers */
3111	VERIFY(ifma != NULL || (ifp != NULL && sa != NULL));
3112
3113	if (ifma != NULL)
3114		ifp = ifma->ifma_ifp;
3115
3116	if (sa != NULL &&
3117	    (sa->sa_family == AF_LINK || sa->sa_family == AF_UNSPEC)) {
3118		dupsa = copy_and_normalize(sa);
3119		if (dupsa == NULL)
3120			return (ENOMEM);
3121		sa = dupsa;
3122	}
3123
3124	ifnet_lock_exclusive(ifp);
3125	if (ifma == NULL) {
3126		for (ifma = LIST_FIRST(&ifp->if_multiaddrs); ifma != NULL;
3127		     ifma = LIST_NEXT(ifma, ifma_link)) {
3128			IFMA_LOCK(ifma);
3129			if (!equal(sa, ifma->ifma_addr) ||
3130			    (anon && !(ifma->ifma_flags & IFMAF_ANONYMOUS))) {
3131				VERIFY(!(ifma->ifma_flags & IFMAF_ANONYMOUS) ||
3132				    ifma->ifma_anoncnt != 0);
3133				IFMA_UNLOCK(ifma);
3134				continue;
3135			}
3136			/* found; keep it locked */
3137			break;
3138		}
3139		if (ifma == NULL) {
3140			if (dupsa != NULL)
3141				FREE(dupsa, M_IFADDR);
3142			ifnet_lock_done(ifp);
3143			return (ENOENT);
3144		}
3145	} else {
3146		IFMA_LOCK(ifma);
3147	}
3148	IFMA_LOCK_ASSERT_HELD(ifma);
3149	IFMA_ADDREF_LOCKED(ifma);	/* for this routine */
3150	lastref = if_detach_ifma(ifp, ifma, anon);
3151	VERIFY(!lastref || (!(ifma->ifma_debug & IFD_ATTACHED) &&
3152	    ifma->ifma_reqcnt == 0));
3153	VERIFY(!anon || ifma->ifma_ll == NULL);
3154	ll = ifma->ifma_ll;
3155	lladdr = (ifma->ifma_addr->sa_family == AF_UNSPEC ||
3156	    ifma->ifma_addr->sa_family == AF_LINK);
3157	IFMA_UNLOCK(ifma);
3158	if (lastref && ll != NULL) {
3159		IFMA_LOCK(ll);
3160		ll_lastref = if_detach_ifma(ifp, ll, 0);
3161		IFMA_UNLOCK(ll);
3162	}
3163	ifnet_lock_done(ifp);
3164
3165	if (lastref)
3166		rt_newmaddrmsg(RTM_DELMADDR, ifma);
3167
3168	if ((ll == NULL && lastref && lladdr) || ll_lastref) {
3169		/*
3170		 * Make sure the interface driver is notified in the
3171		 * case of a link layer mcast group being left.  Do
3172		 * this only for a AF_LINK/AF_UNSPEC address that has
3173		 * been removed from the if_multiaddrs set.
3174		 */
3175		ifnet_ioctl(ifp, 0, SIOCDELMULTI, NULL);
3176	}
3177
3178	if (lastref)
3179		IFMA_REMREF(ifma);	/* for if_multiaddrs list */
3180	if (ll_lastref)
3181		IFMA_REMREF(ll);	/* for if_multiaddrs list */
3182
3183	IFMA_REMREF(ifma);		/* for this routine */
3184	if (dupsa != NULL)
3185		FREE(dupsa, M_IFADDR);
3186
3187	return (0);
3188}
3189
3190/*
3191 * We don't use if_setlladdr, our interfaces are responsible for
3192 * handling the SIOCSIFLLADDR ioctl.
3193 */
3194#ifndef __APPLE__
3195int
3196if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3197{
3198	...
3199}
3200#endif
3201
3202SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW|CTLFLAG_LOCKED, 0, "Link layers");
3203SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW|CTLFLAG_LOCKED, 0, "Generic link-management");
3204
3205
3206/*
3207 * Shutdown all network activity.  Used boot() when halting
3208 * system.
3209 */
3210int
3211if_down_all(void)
3212{
3213	struct ifnet **ifp;
3214	u_int32_t	count;
3215	u_int32_t	i;
3216
3217	if (ifnet_list_get_all(IFNET_FAMILY_ANY, &ifp, &count) == 0) {
3218		for (i = 0; i < count; i++) {
3219			if_down(ifp[i]);
3220			dlil_proto_unplumb_all(ifp[i]);
3221		}
3222		ifnet_list_free(ifp);
3223	}
3224
3225	return 0;
3226}
3227
3228/*
3229 * Delete Routes for a Network Interface
3230 *
3231 * Called for each routing entry via the rnh->rnh_walktree() call above
3232 * to delete all route entries referencing a detaching network interface.
3233 *
3234 * Arguments:
3235 *	rn	pointer to node in the routing table
3236 *	arg	argument passed to rnh->rnh_walktree() - detaching interface
3237 *
3238 * Returns:
3239 *	0	successful
3240 *	errno	failed - reason indicated
3241 *
3242 */
3243static int
3244if_rtdel(struct radix_node *rn, void *arg)
3245{
3246	struct rtentry	*rt = (struct rtentry *)rn;
3247	struct ifnet	*ifp = arg;
3248	int		err;
3249
3250	if (rt == NULL)
3251		return (0);
3252	/*
3253	 * Checking against RTF_UP protects against walktree
3254	 * recursion problems with cloned routes.
3255	 */
3256	RT_LOCK(rt);
3257	if (rt->rt_ifp == ifp && (rt->rt_flags & RTF_UP)) {
3258		/*
3259		 * Safe to drop rt_lock and use rt_key, rt_gateway,
3260		 * since holding rnh_lock here prevents another thread
3261		 * from calling rt_setgate() on this route.
3262		 */
3263		RT_UNLOCK(rt);
3264		err = rtrequest_locked(RTM_DELETE, rt_key(rt), rt->rt_gateway,
3265		    rt_mask(rt), rt->rt_flags, NULL);
3266		if (err) {
3267			log(LOG_WARNING, "if_rtdel: error %d\n", err);
3268		}
3269	} else {
3270		RT_UNLOCK(rt);
3271	}
3272	return (0);
3273}
3274
3275/*
3276 * Removes routing table reference to a given interface
3277 * for a given protocol family
3278 */
3279void
3280if_rtproto_del(struct ifnet *ifp, int protocol)
3281{
3282	struct radix_node_head  *rnh;
3283
3284	if (use_routegenid)
3285		routegenid_update();
3286	if ((protocol <= AF_MAX) && (protocol >= 0) &&
3287		((rnh = rt_tables[protocol]) != NULL) && (ifp != NULL)) {
3288		lck_mtx_lock(rnh_lock);
3289		(void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
3290		lck_mtx_unlock(rnh_lock);
3291	}
3292}
3293
3294static int
3295if_rtmtu(struct radix_node *rn, void *arg)
3296{
3297	struct rtentry *rt = (struct rtentry *)rn;
3298	struct ifnet *ifp = arg;
3299
3300	RT_LOCK(rt);
3301	if (rt->rt_ifp == ifp) {
3302		/*
3303		 * Update the MTU of this entry only if the MTU
3304		 * has not been locked (RTV_MTU is not set) and
3305		 * if it was non-zero to begin with.
3306		 */
3307		if (!(rt->rt_rmx.rmx_locks & RTV_MTU) && rt->rt_rmx.rmx_mtu)
3308			rt->rt_rmx.rmx_mtu = ifp->if_mtu;
3309	}
3310	RT_UNLOCK(rt);
3311
3312	return (0);
3313}
3314
3315/*
3316 * Update the MTU metric of all route entries in all protocol tables
3317 * associated with a particular interface; this is called when the
3318 * MTU of that interface has changed.
3319 */
3320static
3321void if_rtmtu_update(struct ifnet *ifp)
3322{
3323	struct radix_node_head *rnh;
3324	int p;
3325
3326	for (p = 0; p < AF_MAX + 1; p++) {
3327		if ((rnh = rt_tables[p]) == NULL)
3328			continue;
3329
3330		lck_mtx_lock(rnh_lock);
3331		(void) rnh->rnh_walktree(rnh, if_rtmtu, ifp);
3332		lck_mtx_unlock(rnh_lock);
3333	}
3334
3335	if (use_routegenid)
3336		routegenid_update();
3337}
3338
3339__private_extern__ void
3340if_data_internal_to_if_data(struct ifnet *ifp,
3341    const struct if_data_internal *if_data_int, struct if_data *if_data)
3342{
3343#pragma unused(ifp)
3344#define COPYFIELD(fld)		if_data->fld = if_data_int->fld
3345#define COPYFIELD32(fld)	if_data->fld = (u_int32_t)(if_data_int->fld)
3346/* compiler will cast down to 32-bit */
3347#define	COPYFIELD32_ATOMIC(fld) do {					\
3348	atomic_get_64(if_data->fld,					\
3349	    (u_int64_t *)(void *)(uintptr_t)&if_data_int->fld);		\
3350} while (0)
3351
3352	COPYFIELD(ifi_type);
3353	COPYFIELD(ifi_typelen);
3354	COPYFIELD(ifi_physical);
3355	COPYFIELD(ifi_addrlen);
3356	COPYFIELD(ifi_hdrlen);
3357	COPYFIELD(ifi_recvquota);
3358	COPYFIELD(ifi_xmitquota);
3359	if_data->ifi_unused1 = 0;
3360	COPYFIELD(ifi_mtu);
3361	COPYFIELD(ifi_metric);
3362	if (if_data_int->ifi_baudrate & 0xFFFFFFFF00000000LL) {
3363		if_data->ifi_baudrate = 0xFFFFFFFF;
3364	} else {
3365		COPYFIELD32(ifi_baudrate);
3366	}
3367
3368	COPYFIELD32_ATOMIC(ifi_ipackets);
3369	COPYFIELD32_ATOMIC(ifi_ierrors);
3370	COPYFIELD32_ATOMIC(ifi_opackets);
3371	COPYFIELD32_ATOMIC(ifi_oerrors);
3372	COPYFIELD32_ATOMIC(ifi_collisions);
3373	COPYFIELD32_ATOMIC(ifi_ibytes);
3374	COPYFIELD32_ATOMIC(ifi_obytes);
3375	COPYFIELD32_ATOMIC(ifi_imcasts);
3376	COPYFIELD32_ATOMIC(ifi_omcasts);
3377	COPYFIELD32_ATOMIC(ifi_iqdrops);
3378	COPYFIELD32_ATOMIC(ifi_noproto);
3379
3380	COPYFIELD(ifi_recvtiming);
3381	COPYFIELD(ifi_xmittiming);
3382
3383	if_data->ifi_lastchange.tv_sec = if_data_int->ifi_lastchange.tv_sec;
3384	if_data->ifi_lastchange.tv_usec = if_data_int->ifi_lastchange.tv_usec;
3385
3386#if IF_LASTCHANGEUPTIME
3387	if_data->ifi_lastchange.tv_sec += boottime_sec();
3388#endif
3389
3390	if_data->ifi_unused2 = 0;
3391	COPYFIELD(ifi_hwassist);
3392	if_data->ifi_reserved1 = 0;
3393	if_data->ifi_reserved2 = 0;
3394#undef COPYFIELD32_ATOMIC
3395#undef COPYFIELD32
3396#undef COPYFIELD
3397}
3398
3399__private_extern__ void
3400if_data_internal_to_if_data64(struct ifnet *ifp,
3401    const struct if_data_internal *if_data_int,
3402    struct if_data64 *if_data64)
3403{
3404#pragma unused(ifp)
3405#define COPYFIELD64(fld)	if_data64->fld = if_data_int->fld
3406#define COPYFIELD64_ATOMIC(fld) do {					\
3407	atomic_get_64(if_data64->fld,					\
3408	    (u_int64_t *)(void *)(uintptr_t)&if_data_int->fld);		\
3409} while (0)
3410
3411	COPYFIELD64(ifi_type);
3412	COPYFIELD64(ifi_typelen);
3413	COPYFIELD64(ifi_physical);
3414	COPYFIELD64(ifi_addrlen);
3415	COPYFIELD64(ifi_hdrlen);
3416	COPYFIELD64(ifi_recvquota);
3417	COPYFIELD64(ifi_xmitquota);
3418	if_data64->ifi_unused1 = 0;
3419	COPYFIELD64(ifi_mtu);
3420	COPYFIELD64(ifi_metric);
3421	COPYFIELD64(ifi_baudrate);
3422
3423	COPYFIELD64_ATOMIC(ifi_ipackets);
3424	COPYFIELD64_ATOMIC(ifi_ierrors);
3425	COPYFIELD64_ATOMIC(ifi_opackets);
3426	COPYFIELD64_ATOMIC(ifi_oerrors);
3427	COPYFIELD64_ATOMIC(ifi_collisions);
3428	COPYFIELD64_ATOMIC(ifi_ibytes);
3429	COPYFIELD64_ATOMIC(ifi_obytes);
3430	COPYFIELD64_ATOMIC(ifi_imcasts);
3431	COPYFIELD64_ATOMIC(ifi_omcasts);
3432	COPYFIELD64_ATOMIC(ifi_iqdrops);
3433	COPYFIELD64_ATOMIC(ifi_noproto);
3434
3435	/* Note these two fields are actually 32 bit, so doing COPYFIELD64_ATOMIC will
3436	 * cause them to be misaligned
3437	 */
3438	COPYFIELD64(ifi_recvtiming);
3439	COPYFIELD64(ifi_xmittiming);
3440
3441	if_data64->ifi_lastchange.tv_sec = if_data_int->ifi_lastchange.tv_sec;
3442	if_data64->ifi_lastchange.tv_usec = if_data_int->ifi_lastchange.tv_usec;
3443
3444#if IF_LASTCHANGEUPTIME
3445	if_data64->ifi_lastchange.tv_sec += boottime_sec();
3446#endif
3447
3448#undef COPYFIELD64
3449}
3450
3451__private_extern__ void
3452if_copy_traffic_class(struct ifnet *ifp,
3453    struct if_traffic_class *if_tc)
3454{
3455#define COPY_IF_TC_FIELD64_ATOMIC(fld) do {			\
3456	atomic_get_64(if_tc->fld,				\
3457	    (u_int64_t *)(void *)(uintptr_t)&ifp->if_tc.fld);	\
3458} while (0)
3459
3460	bzero(if_tc, sizeof (*if_tc));
3461	COPY_IF_TC_FIELD64_ATOMIC(ifi_ibepackets);
3462	COPY_IF_TC_FIELD64_ATOMIC(ifi_ibebytes);
3463	COPY_IF_TC_FIELD64_ATOMIC(ifi_obepackets);
3464	COPY_IF_TC_FIELD64_ATOMIC(ifi_obebytes);
3465	COPY_IF_TC_FIELD64_ATOMIC(ifi_ibkpackets);
3466	COPY_IF_TC_FIELD64_ATOMIC(ifi_ibkbytes);
3467	COPY_IF_TC_FIELD64_ATOMIC(ifi_obkpackets);
3468	COPY_IF_TC_FIELD64_ATOMIC(ifi_obkbytes);
3469	COPY_IF_TC_FIELD64_ATOMIC(ifi_ivipackets);
3470	COPY_IF_TC_FIELD64_ATOMIC(ifi_ivibytes);
3471	COPY_IF_TC_FIELD64_ATOMIC(ifi_ovipackets);
3472	COPY_IF_TC_FIELD64_ATOMIC(ifi_ovibytes);
3473	COPY_IF_TC_FIELD64_ATOMIC(ifi_ivopackets);
3474	COPY_IF_TC_FIELD64_ATOMIC(ifi_ivobytes);
3475	COPY_IF_TC_FIELD64_ATOMIC(ifi_ovopackets);
3476	COPY_IF_TC_FIELD64_ATOMIC(ifi_ovobytes);
3477	COPY_IF_TC_FIELD64_ATOMIC(ifi_ipvpackets);
3478	COPY_IF_TC_FIELD64_ATOMIC(ifi_ipvbytes);
3479	COPY_IF_TC_FIELD64_ATOMIC(ifi_opvpackets);
3480	COPY_IF_TC_FIELD64_ATOMIC(ifi_opvbytes);
3481
3482#undef COPY_IF_TC_FIELD64_ATOMIC
3483}
3484
3485void
3486if_copy_data_extended(struct ifnet *ifp, struct if_data_extended *if_de)
3487{
3488#define COPY_IF_DE_FIELD64_ATOMIC(fld) do {			\
3489	atomic_get_64(if_de->fld,				\
3490	    (u_int64_t *)(void *)(uintptr_t)&ifp->if_data.fld);	\
3491} while (0)
3492
3493	bzero(if_de, sizeof (*if_de));
3494	COPY_IF_DE_FIELD64_ATOMIC(ifi_alignerrs);
3495
3496#undef COPY_IF_DE_FIELD64_ATOMIC
3497}
3498
3499void
3500if_copy_packet_stats(struct ifnet *ifp, struct if_packet_stats *if_ps)
3501{
3502#define COPY_IF_PS_TCP_FIELD64_ATOMIC(fld) do {				\
3503	atomic_get_64(if_ps->ifi_tcp_##fld,				\
3504	    (u_int64_t *)(void *)(uintptr_t)&ifp->if_tcp_stat->fld);	\
3505} while (0)
3506
3507#define COPY_IF_PS_UDP_FIELD64_ATOMIC(fld) do {				\
3508	atomic_get_64(if_ps->ifi_udp_##fld,				\
3509	    (u_int64_t *)(void *)(uintptr_t)&ifp->if_udp_stat->fld);	\
3510} while (0)
3511
3512	COPY_IF_PS_TCP_FIELD64_ATOMIC(badformat);
3513	COPY_IF_PS_TCP_FIELD64_ATOMIC(unspecv6);
3514	COPY_IF_PS_TCP_FIELD64_ATOMIC(synfin);
3515	COPY_IF_PS_TCP_FIELD64_ATOMIC(badformatipsec);
3516	COPY_IF_PS_TCP_FIELD64_ATOMIC(noconnnolist);
3517	COPY_IF_PS_TCP_FIELD64_ATOMIC(noconnlist);
3518	COPY_IF_PS_TCP_FIELD64_ATOMIC(listbadsyn);
3519	COPY_IF_PS_TCP_FIELD64_ATOMIC(icmp6unreach);
3520	COPY_IF_PS_TCP_FIELD64_ATOMIC(deprecate6);
3521	COPY_IF_PS_TCP_FIELD64_ATOMIC(ooopacket);
3522	COPY_IF_PS_TCP_FIELD64_ATOMIC(rstinsynrcv);
3523	COPY_IF_PS_TCP_FIELD64_ATOMIC(dospacket);
3524	COPY_IF_PS_TCP_FIELD64_ATOMIC(cleanup);
3525	COPY_IF_PS_TCP_FIELD64_ATOMIC(synwindow);
3526
3527	COPY_IF_PS_UDP_FIELD64_ATOMIC(port_unreach);
3528	COPY_IF_PS_UDP_FIELD64_ATOMIC(faithprefix);
3529	COPY_IF_PS_UDP_FIELD64_ATOMIC(port0);
3530	COPY_IF_PS_UDP_FIELD64_ATOMIC(badlength);
3531	COPY_IF_PS_UDP_FIELD64_ATOMIC(badchksum);
3532	COPY_IF_PS_UDP_FIELD64_ATOMIC(badmcast);
3533	COPY_IF_PS_UDP_FIELD64_ATOMIC(cleanup);
3534	COPY_IF_PS_UDP_FIELD64_ATOMIC(badipsec);
3535
3536#undef COPY_IF_PS_TCP_FIELD64_ATOMIC
3537#undef COPY_IF_PS_UDP_FIELD64_ATOMIC
3538}
3539
3540void
3541if_copy_rxpoll_stats(struct ifnet *ifp, struct if_rxpoll_stats *if_rs)
3542{
3543	bzero(if_rs, sizeof (*if_rs));
3544	if (!(ifp->if_eflags & IFEF_RXPOLL) || !ifnet_is_attached(ifp, 1))
3545		return;
3546
3547	/* by now, ifnet will stay attached so if_inp must be valid */
3548	VERIFY(ifp->if_inp != NULL);
3549	bcopy(&ifp->if_inp->pstats, if_rs, sizeof (*if_rs));
3550
3551	/* Release the IO refcnt */
3552	ifnet_decr_iorefcnt(ifp);
3553}
3554
3555struct ifaddr *
3556ifa_remref(struct ifaddr *ifa, int locked)
3557{
3558	if (!locked)
3559		IFA_LOCK_SPIN(ifa);
3560	else
3561		IFA_LOCK_ASSERT_HELD(ifa);
3562
3563	if (ifa->ifa_refcnt == 0)
3564		panic("%s: ifa %p negative refcnt\n", __func__, ifa);
3565	else if (ifa->ifa_trace != NULL)
3566		(*ifa->ifa_trace)(ifa, FALSE);
3567	if (--ifa->ifa_refcnt == 0) {
3568		if (ifa->ifa_debug & IFD_ATTACHED)
3569			panic("ifa %p attached to ifp is being freed\n", ifa);
3570		/*
3571		 * Some interface addresses are allocated either statically
3572		 * or carved out of a larger block; e.g. AppleTalk addresses.
3573		 * Only free it if it was allocated via MALLOC or via the
3574		 * corresponding per-address family allocator.  Otherwise,
3575		 * leave it alone.
3576		 */
3577		if (ifa->ifa_debug & IFD_ALLOC) {
3578			if (ifa->ifa_free == NULL) {
3579				IFA_UNLOCK(ifa);
3580				FREE(ifa, M_IFADDR);
3581			} else {
3582				/* Become a regular mutex */
3583				IFA_CONVERT_LOCK(ifa);
3584				/* callee will unlock */
3585				(*ifa->ifa_free)(ifa);
3586			}
3587		} else {
3588			IFA_UNLOCK(ifa);
3589		}
3590		ifa = NULL;
3591	}
3592
3593	if (!locked && ifa != NULL)
3594		IFA_UNLOCK(ifa);
3595
3596	return (ifa);
3597}
3598
3599void
3600ifa_addref(struct ifaddr *ifa, int locked)
3601{
3602	if (!locked)
3603		IFA_LOCK_SPIN(ifa);
3604	else
3605		IFA_LOCK_ASSERT_HELD(ifa);
3606
3607	if (++ifa->ifa_refcnt == 0) {
3608		panic("%s: ifa %p wraparound refcnt\n", __func__, ifa);
3609		/* NOTREACHED */
3610	} else if (ifa->ifa_trace != NULL) {
3611		(*ifa->ifa_trace)(ifa, TRUE);
3612	}
3613	if (!locked)
3614		IFA_UNLOCK(ifa);
3615}
3616
3617void
3618ifa_lock_init(struct ifaddr *ifa)
3619{
3620	lck_mtx_init(&ifa->ifa_lock, ifa_mtx_grp, ifa_mtx_attr);
3621}
3622
3623void
3624ifa_lock_destroy(struct ifaddr *ifa)
3625{
3626	IFA_LOCK_ASSERT_NOTHELD(ifa);
3627	lck_mtx_destroy(&ifa->ifa_lock, ifa_mtx_grp);
3628}
3629