1/*
2 * Copyright (c) 1999-2007 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 *	Data Link Inteface Layer
30 *	Author: Ted Walker
31 */
32/*
33 * NOTICE: This file was modified by SPARTA, Inc. in 2005 to introduce
34 * support for mandatory and extensible security protections.  This notice
35 * is included in support of clause 2.2 (b) of the Apple Public License,
36 * Version 2.0.
37 */
38
39#include <sys/param.h>
40#include <sys/systm.h>
41#include <sys/kernel.h>
42#include <sys/malloc.h>
43#include <sys/mbuf.h>
44#include <sys/socket.h>
45#include <sys/domain.h>
46#include <sys/user.h>
47#include <sys/random.h>
48#include <net/if_dl.h>
49#include <net/if.h>
50#include <net/route.h>
51#include <net/if_var.h>
52#include <net/dlil.h>
53#include <net/if_arp.h>
54#include <sys/kern_event.h>
55#include <sys/kdebug.h>
56
57#include <kern/assert.h>
58#include <kern/task.h>
59#include <kern/thread.h>
60#include <kern/sched_prim.h>
61#include <kern/locks.h>
62#include <net/kpi_protocol.h>
63
64#include <net/if_types.h>
65#include <net/kpi_interfacefilter.h>
66
67#include <libkern/OSAtomic.h>
68
69#include <machine/machine_routines.h>
70
71#include <mach/thread_act.h>
72
73#if CONFIG_MACF_NET
74#include <security/mac_framework.h>
75#endif /* MAC_NET */
76
77#define DBG_LAYER_BEG			DLILDBG_CODE(DBG_DLIL_STATIC, 0)
78#define DBG_LAYER_END			DLILDBG_CODE(DBG_DLIL_STATIC, 2)
79#define DBG_FNC_DLIL_INPUT      DLILDBG_CODE(DBG_DLIL_STATIC, (1 << 8))
80#define DBG_FNC_DLIL_OUTPUT     DLILDBG_CODE(DBG_DLIL_STATIC, (2 << 8))
81#define DBG_FNC_DLIL_IFOUT      DLILDBG_CODE(DBG_DLIL_STATIC, (3 << 8))
82
83
84#define MAX_FRAME_TYPE_SIZE 4 /* LONGWORDS */
85#define MAX_LINKADDR	    4 /* LONGWORDS */
86#define M_NKE M_IFADDR
87
88#if 1
89#define DLIL_PRINTF	printf
90#else
91#define DLIL_PRINTF	kprintf
92#endif
93
94
95enum {
96	kProtoKPI_v1	= 1,
97	kProtoKPI_v2	= 2
98};
99
100struct if_proto {
101    SLIST_ENTRY(if_proto)	 next_hash;
102    int						 refcount;
103    int						 detaching;
104    struct ifnet			 *ifp;
105    struct domain			 *dl_domain;
106    protocol_family_t		protocol_family;
107    int						proto_kpi;
108    union {
109		struct {
110			proto_media_input			input;
111			proto_media_preout			pre_output;
112			proto_media_event			event;
113			proto_media_ioctl			ioctl;
114			proto_media_detached		detached;
115			proto_media_resolve_multi	resolve_multi;
116			proto_media_send_arp		send_arp;
117		} v1;
118		struct {
119			proto_media_input_v2		input;
120			proto_media_preout			pre_output;
121			proto_media_event			event;
122			proto_media_ioctl			ioctl;
123			proto_media_detached		detached;
124			proto_media_resolve_multi	resolve_multi;
125			proto_media_send_arp		send_arp;
126		} v2;
127	} kpi;
128};
129
130SLIST_HEAD(proto_hash_entry, if_proto);
131
132
133struct dlil_ifnet {
134    /* ifnet and drvr_ext are used by the stack and drivers
135    drvr_ext extends the public ifnet and must follow dl_if */
136    struct ifnet	dl_if;			/* public ifnet */
137
138    /* dlil private fields */
139    TAILQ_ENTRY(dlil_ifnet) dl_if_link;	/* dlil_ifnet are link together */
140    								/* it is not the ifnet list */
141    void		*if_uniqueid;	/* unique id identifying the interface */
142    size_t		if_uniqueid_len;/* length of the unique id */
143    char		if_namestorage[IFNAMSIZ]; /* interface name storage */
144};
145
146struct ifnet_filter {
147	TAILQ_ENTRY(ifnet_filter)	filt_next;
148    ifnet_t						filt_ifp;
149    int							filt_detaching;
150
151	const char					*filt_name;
152	void						*filt_cookie;
153    protocol_family_t			filt_protocol;
154    iff_input_func				filt_input;
155    iff_output_func				filt_output;
156    iff_event_func				filt_event;
157    iff_ioctl_func				filt_ioctl;
158    iff_detached_func			filt_detached;
159};
160
161struct proto_input_entry;
162
163static TAILQ_HEAD(, dlil_ifnet) dlil_ifnet_head;
164static lck_grp_t *dlil_lock_group;
165static lck_grp_t *ifnet_lock_group;
166static lck_grp_t *ifnet_head_lock_group;
167static lck_attr_t *ifnet_lock_attr;
168static lck_rw_t *ifnet_head_mutex;
169static lck_mtx_t *dlil_ifnet_mutex;
170static lck_mtx_t *dlil_mutex;
171static unsigned long dlil_read_count = 0;
172static unsigned long dlil_detach_waiting = 0;
173extern u_int32_t	ipv4_ll_arp_aware;
174
175static struct dlil_threading_info dlil_lo_thread;
176__private_extern__  struct dlil_threading_info *dlil_lo_thread_ptr = &dlil_lo_thread;
177
178static struct mbuf *dlil_lo_input_mbuf_head = NULL;
179static struct mbuf *dlil_lo_input_mbuf_tail = NULL;
180
181#if IFNET_INPUT_SANITY_CHK
182static int dlil_lo_input_mbuf_count = 0;
183int dlil_input_sanity_check = 0;	/* sanity checking of input packet lists received */
184#endif
185int dlil_multithreaded_input = 1;
186static int cur_dlil_input_threads = 0;
187
188static int dlil_event_internal(struct ifnet *ifp, struct kev_msg *msg);
189static int dlil_detach_filter_internal(interface_filter_t filter, int detached);
190static void dlil_call_delayed_detach_thread(void);
191
192static void	dlil_read_begin(void);
193static __inline__ void	dlil_read_end(void);
194static int	dlil_write_begin(void);
195static void	dlil_write_end(void);
196
197unsigned int net_affinity = 1;
198static kern_return_t dlil_affinity_set(struct thread *, u_int32_t);
199
200extern void bpfdetach(struct ifnet*);
201extern void proto_input_run(void); // new run_netisr
202
203void dlil_input_packet_list(struct ifnet  *ifp, struct mbuf *m);
204static void dlil_input_thread_func(struct dlil_threading_info *inpthread);
205__private_extern__ int dlil_create_input_thread(
206		ifnet_t, struct dlil_threading_info *);
207__private_extern__ void dlil_terminate_input_thread(
208		struct dlil_threading_info *);
209
210__private_extern__ void link_rtrequest(int, struct rtentry *, struct sockaddr *);
211
212int dlil_expand_mcl;
213
214extern u_int32_t	inject_buckets;
215
216static const u_int32_t dlil_writer_waiting = 0x80000000;
217static	lck_grp_attr_t	*dlil_grp_attributes = NULL;
218static	lck_attr_t	*dlil_lck_attributes = NULL;
219static	lck_grp_t	*dlil_input_lock_grp = NULL;
220
221static inline void*
222_cast_non_const(const void * ptr) {
223	union {
224		const void*		cval;
225		void*			val;
226	} ret;
227
228	ret.cval = ptr;
229	return (ret.val);
230}
231
232/* Should these be inline? */
233static void
234dlil_read_begin(void)
235{
236	unsigned long new_value;
237	unsigned long old_value;
238	struct uthread *uth = get_bsdthread_info(current_thread());
239
240	if (uth->dlil_incremented_read == dlil_writer_waiting)
241		panic("dlil_read_begin - thread is already a writer");
242
243	do {
244again:
245		old_value = dlil_read_count;
246
247		if ((old_value & dlil_writer_waiting) != 0 && uth->dlil_incremented_read == 0)
248		{
249			tsleep(&dlil_read_count, PRIBIO, "dlil_read_count", 1);
250			goto again;
251		}
252
253		new_value = old_value + 1;
254	} while (!OSCompareAndSwap((UInt32)old_value, (UInt32)new_value, (UInt32*)&dlil_read_count));
255
256	uth->dlil_incremented_read++;
257}
258
259static void
260dlil_read_end(void)
261{
262	struct uthread *uth = get_bsdthread_info(current_thread());
263
264	OSDecrementAtomic((SInt32*)&dlil_read_count);
265	uth->dlil_incremented_read--;
266	if (dlil_read_count == dlil_writer_waiting)
267		wakeup(_cast_non_const(&dlil_writer_waiting));
268}
269
270static int
271dlil_write_begin(void)
272{
273	struct uthread *uth = get_bsdthread_info(current_thread());
274
275	if (uth->dlil_incremented_read != 0) {
276		return EDEADLK;
277	}
278	lck_mtx_lock(dlil_mutex);
279	OSBitOrAtomic((UInt32)dlil_writer_waiting, (UInt32*)&dlil_read_count);
280again:
281	if (dlil_read_count == dlil_writer_waiting) {
282		uth->dlil_incremented_read = dlil_writer_waiting;
283		return 0;
284	}
285	else {
286		tsleep(_cast_non_const(&dlil_writer_waiting), PRIBIO, "dlil_writer_waiting", 1);
287		goto again;
288	}
289}
290
291static void
292dlil_write_end(void)
293{
294	struct uthread *uth = get_bsdthread_info(current_thread());
295
296	if (uth->dlil_incremented_read != dlil_writer_waiting)
297		panic("dlil_write_end - thread is not a writer");
298	OSBitAndAtomic((UInt32)~dlil_writer_waiting, (UInt32*)&dlil_read_count);
299	lck_mtx_unlock(dlil_mutex);
300	uth->dlil_incremented_read = 0;
301	wakeup(&dlil_read_count);
302}
303
304#define PROTO_HASH_SLOTS	0x5
305
306/*
307 * Internal functions.
308 */
309
310static int
311proto_hash_value(u_long protocol_family)
312{
313	/*
314	 * dlil_proto_unplumb_all() depends on the mapping between
315	 * the hash bucket index and the protocol family defined
316	 * here; future changes must be applied there as well.
317	 */
318	switch(protocol_family) {
319		case PF_INET:
320			return 0;
321		case PF_INET6:
322			return 1;
323		case PF_APPLETALK:
324			return 2;
325		case PF_VLAN:
326			return 3;
327		default:
328			return 4;
329	}
330}
331
332static struct if_proto*
333find_attached_proto(struct ifnet *ifp, u_long protocol_family)
334{
335	struct if_proto *proto = NULL;
336	u_long i = proto_hash_value(protocol_family);
337	if (ifp->if_proto_hash) {
338		proto = SLIST_FIRST(&ifp->if_proto_hash[i]);
339	}
340
341	while(proto && proto->protocol_family != protocol_family) {
342		proto = SLIST_NEXT(proto, next_hash);
343	}
344
345	return proto;
346}
347
348static void
349if_proto_ref(struct if_proto *proto)
350{
351	OSAddAtomic(1, (SInt32*)&proto->refcount);
352}
353
354static void
355if_proto_free(struct if_proto *proto)
356{
357	int oldval = OSAddAtomic(-1, (SInt32*)&proto->refcount);
358
359	if (oldval == 1) { /* This was the last reference */
360		FREE(proto, M_IFADDR);
361	}
362}
363
364__private_extern__ void
365ifnet_lock_assert(
366	__unused struct ifnet *ifp,
367	__unused int what)
368{
369#if IFNET_RW_LOCK
370	/*
371	 * Not implemented for rw locks.
372	 *
373	 * Function exists so when/if we use mutex we can
374	 * enable this check.
375	 */
376#else
377	lck_mtx_assert(ifp->if_lock, what);
378#endif
379}
380
381__private_extern__ void
382ifnet_lock_shared(
383	struct ifnet *ifp)
384{
385#if IFNET_RW_LOCK
386	lck_rw_lock_shared(ifp->if_lock);
387#else
388	lck_mtx_assert(ifp->if_lock, LCK_MTX_ASSERT_NOTOWNED);
389	lck_mtx_lock(ifp->if_lock);
390#endif
391}
392
393__private_extern__ void
394ifnet_lock_exclusive(
395	struct ifnet *ifp)
396{
397#if IFNET_RW_LOCK
398	lck_rw_lock_exclusive(ifp->if_lock);
399#else
400	lck_mtx_assert(ifp->if_lock, LCK_MTX_ASSERT_NOTOWNED);
401	lck_mtx_lock(ifp->if_lock);
402#endif
403}
404
405__private_extern__ void
406ifnet_lock_done(
407	struct ifnet *ifp)
408{
409#if IFNET_RW_LOCK
410	lck_rw_done(ifp->if_lock);
411#else
412	lck_mtx_assert(ifp->if_lock, LCK_MTX_ASSERT_OWNED);
413	lck_mtx_unlock(ifp->if_lock);
414#endif
415}
416
417__private_extern__ void
418ifnet_head_lock_shared(void)
419{
420	lck_rw_lock_shared(ifnet_head_mutex);
421}
422
423__private_extern__ void
424ifnet_head_lock_exclusive(void)
425{
426	lck_rw_lock_exclusive(ifnet_head_mutex);
427}
428
429__private_extern__ void
430ifnet_head_done(void)
431{
432	lck_rw_done(ifnet_head_mutex);
433}
434
435static int dlil_ifp_proto_count(struct ifnet * ifp)
436{
437	int				count = 0;
438	int				i;
439
440	if (ifp->if_proto_hash != NULL) {
441		for (i = 0; i < PROTO_HASH_SLOTS; i++) {
442			struct if_proto *proto;
443			SLIST_FOREACH(proto, &ifp->if_proto_hash[i], next_hash) {
444				count++;
445			}
446		}
447	}
448
449	return count;
450}
451
452__private_extern__ void
453dlil_post_msg(struct ifnet *ifp, u_long event_subclass, u_long event_code,
454		   struct net_event_data *event_data, u_long event_data_len)
455{
456	struct net_event_data  	ev_data;
457	struct kev_msg  		ev_msg;
458
459	/*
460	 * a net event always starts with a net_event_data structure
461	 * but the caller can generate a simple net event or
462	 * provide a longer event structure to post
463	 */
464
465	ev_msg.vendor_code    = KEV_VENDOR_APPLE;
466	ev_msg.kev_class      = KEV_NETWORK_CLASS;
467	ev_msg.kev_subclass   = event_subclass;
468	ev_msg.event_code 	  = event_code;
469
470	if (event_data == 0) {
471		event_data = &ev_data;
472		event_data_len = sizeof(struct net_event_data);
473	}
474
475	strncpy(&event_data->if_name[0], ifp->if_name, IFNAMSIZ);
476	event_data->if_family = ifp->if_family;
477	event_data->if_unit   = (unsigned long) ifp->if_unit;
478
479	ev_msg.dv[0].data_length = event_data_len;
480	ev_msg.dv[0].data_ptr    = event_data;
481	ev_msg.dv[1].data_length = 0;
482
483	dlil_event_internal(ifp, &ev_msg);
484}
485
486__private_extern__ int
487dlil_create_input_thread(
488	ifnet_t ifp, struct dlil_threading_info *inputthread)
489{
490	int error;
491
492	bzero(inputthread, sizeof(*inputthread));
493	// loopback ifp may not be configured at dlil_init time.
494	if (ifp == lo_ifp)
495		strlcat(inputthread->input_name, "dlil_input_main_thread_mtx", 32);
496	else
497		snprintf(inputthread->input_name, 32, "dlil_input_%s%d_mtx", ifp->if_name, ifp->if_unit);
498
499	inputthread->lck_grp = lck_grp_alloc_init(inputthread->input_name, dlil_grp_attributes);
500	inputthread->input_lck  = lck_mtx_alloc_init(inputthread->lck_grp, dlil_lck_attributes);
501
502	error= kernel_thread_start((thread_continue_t)dlil_input_thread_func, inputthread, &inputthread->input_thread);
503	if (error == 0) {
504       		ml_thread_policy(inputthread->input_thread, MACHINE_GROUP,
505				 (MACHINE_NETWORK_GROUP|MACHINE_NETWORK_NETISR));
506		/*
507		 * Except for the loopback dlil input thread, we create
508		 * an affinity set so that the matching workloop thread
509		 * can be scheduled on the same processor set.
510		 */
511		if (net_affinity && inputthread != dlil_lo_thread_ptr) {
512			struct thread *tp = inputthread->input_thread;
513			u_int32_t tag;
514			/*
515			 * Randomize to reduce the probability
516			 * of affinity tag namespace collision.
517			 */
518			read_random(&tag, sizeof (tag));
519			if (dlil_affinity_set(tp, tag) == KERN_SUCCESS) {
520				thread_reference(tp);
521				inputthread->tag = tag;
522				inputthread->net_affinity = TRUE;
523			}
524		}
525	} else {
526		panic("dlil_create_input_thread: couldn't create thread\n");
527	}
528	OSAddAtomic(1, (SInt32*)&cur_dlil_input_threads);
529#if DLIL_DEBUG
530	printf("dlil_create_input_thread: threadinfo: %p input_thread=%p threads: cur=%d max=%d\n",
531		inputthread, inputthread->input_thread, dlil_multithreaded_input, cur_dlil_input_threads);
532#endif
533	return error;
534}
535__private_extern__ void
536dlil_terminate_input_thread(
537	struct dlil_threading_info *inputthread)
538{
539	OSAddAtomic(-1, (SInt32*)&cur_dlil_input_threads);
540
541	lck_mtx_unlock(inputthread->input_lck);
542	lck_mtx_free(inputthread->input_lck, inputthread->lck_grp);
543	lck_grp_free(inputthread->lck_grp);
544
545	FREE(inputthread, M_NKE);
546
547	/* For the extra reference count from kernel_thread_start() */
548	thread_deallocate(current_thread());
549
550	thread_terminate(current_thread());
551}
552
553static kern_return_t
554dlil_affinity_set(struct thread *tp, u_int32_t tag)
555{
556	thread_affinity_policy_data_t policy;
557
558	bzero(&policy, sizeof (policy));
559	policy.affinity_tag = tag;
560	return (thread_policy_set(tp, THREAD_AFFINITY_POLICY,
561	    (thread_policy_t)&policy, THREAD_AFFINITY_POLICY_COUNT));
562}
563
564void
565dlil_init(void)
566{
567	PE_parse_boot_argn("net_affinity", &net_affinity, sizeof (net_affinity));
568
569	TAILQ_INIT(&dlil_ifnet_head);
570	TAILQ_INIT(&ifnet_head);
571
572	/* Setup the lock groups we will use */
573	dlil_grp_attributes = lck_grp_attr_alloc_init();
574
575	dlil_lock_group = lck_grp_alloc_init("dlil internal locks", dlil_grp_attributes);
576	ifnet_lock_group = lck_grp_alloc_init("ifnet locks", dlil_grp_attributes);
577	ifnet_head_lock_group = lck_grp_alloc_init("ifnet head lock", dlil_grp_attributes);
578	dlil_input_lock_grp = lck_grp_alloc_init("dlil input lock", dlil_grp_attributes);
579
580	/* Setup the lock attributes we will use */
581	dlil_lck_attributes = lck_attr_alloc_init();
582
583	ifnet_lock_attr = lck_attr_alloc_init();
584
585
586	ifnet_head_mutex = lck_rw_alloc_init(ifnet_head_lock_group, dlil_lck_attributes);
587	dlil_ifnet_mutex = lck_mtx_alloc_init(dlil_lock_group, dlil_lck_attributes);
588	dlil_mutex = lck_mtx_alloc_init(dlil_lock_group, dlil_lck_attributes);
589
590	lck_attr_free(dlil_lck_attributes);
591	dlil_lck_attributes = NULL;
592
593	/*
594	 * Create and start up the first dlil input thread once everything is initialized
595	 */
596	dlil_create_input_thread(0, dlil_lo_thread_ptr);
597
598	(void) kernel_thread(kernel_task, dlil_call_delayed_detach_thread);
599}
600
601__private_extern__ int
602dlil_attach_filter(
603	struct ifnet			*ifp,
604	const struct iff_filter	*if_filter,
605	interface_filter_t		*filter_ref)
606{
607    int retval = 0;
608    struct ifnet_filter	*filter;
609
610	MALLOC(filter, struct ifnet_filter *, sizeof(*filter), M_NKE, M_WAITOK);
611	if (filter == NULL)
612		return ENOMEM;
613	bzero(filter, sizeof(*filter));
614
615
616	filter->filt_ifp = ifp;
617	filter->filt_cookie = if_filter->iff_cookie;
618	filter->filt_name = if_filter->iff_name;
619	filter->filt_protocol = if_filter->iff_protocol;
620	filter->filt_input = if_filter->iff_input;
621	filter->filt_output = if_filter->iff_output;
622	filter->filt_event = if_filter->iff_event;
623	filter->filt_ioctl = if_filter->iff_ioctl;
624	filter->filt_detached = if_filter->iff_detached;
625
626	if ((retval = dlil_write_begin()) != 0) {
627		/* Failed to acquire the write lock */
628		FREE(filter, M_NKE);
629		return retval;
630	}
631	TAILQ_INSERT_TAIL(&ifp->if_flt_head, filter, filt_next);
632	dlil_write_end();
633	*filter_ref = filter;
634	return retval;
635}
636
637static int
638dlil_detach_filter_internal(
639	interface_filter_t	filter,
640	int					detached)
641{
642	int retval = 0;
643
644	if (detached == 0) {
645		ifnet_t				ifp = NULL;
646		interface_filter_t	entry = NULL;
647
648		/* Take the write lock */
649	 	retval = dlil_write_begin();
650	 	if (retval != 0 && retval != EDEADLK)
651	 		return retval;
652
653	 	/*
654	 	 * At this point either we have the write lock (retval == 0)
655	 	 * or we couldn't get it (retval == EDEADLK) because someone
656	 	 * else up the stack is holding the read lock. It is safe to
657	 	 * read, either the read or write is held. Verify the filter
658	 	 * parameter before proceeding.
659	 	 */
660		ifnet_head_lock_shared();
661		TAILQ_FOREACH(ifp, &ifnet_head, if_link) {
662			TAILQ_FOREACH(entry, &ifp->if_flt_head, filt_next) {
663				if (entry == filter)
664					break;
665			}
666			if (entry == filter)
667				break;
668		}
669		ifnet_head_done();
670
671		if (entry != filter) {
672			/* filter parameter is not a valid filter ref */
673			if (retval == 0) {
674				dlil_write_end();
675			}
676			return EINVAL;
677		}
678
679		if (retval == EDEADLK) {
680			/* Perform a delayed detach */
681			filter->filt_detaching = 1;
682			dlil_detach_waiting = 1;
683			wakeup(&dlil_detach_waiting);
684			return 0;
685		}
686
687		/* Remove the filter from the list */
688		TAILQ_REMOVE(&ifp->if_flt_head, filter, filt_next);
689		dlil_write_end();
690	}
691
692	/* Call the detached funciton if there is one */
693	if (filter->filt_detached)
694		filter->filt_detached(filter->filt_cookie, filter->filt_ifp);
695
696	/* Free the filter */
697	FREE(filter, M_NKE);
698
699	return retval;
700}
701
702__private_extern__ void
703dlil_detach_filter(interface_filter_t filter)
704{
705	if (filter == NULL)
706		return;
707	dlil_detach_filter_internal(filter, 0);
708}
709
710static void
711dlil_input_thread_func(
712	struct dlil_threading_info *inputthread)
713{
714	while (1) {
715		struct mbuf *m = NULL, *m_loop = NULL;
716#if IFNET_INPUT_SANITY_CHK
717		int		loop_cnt = 0, mbuf_cnt;
718		int		count;
719		struct mbuf *m1;
720#endif /* IFNET_INPUT_SANITY_CHK */
721
722		lck_mtx_lock(inputthread->input_lck);
723
724		/* Wait until there is work to be done */
725		while ((inputthread->input_waiting & ~DLIL_INPUT_RUNNING) == 0) {
726			inputthread->input_waiting &= ~DLIL_INPUT_RUNNING;
727			msleep(&inputthread->input_waiting, inputthread->input_lck, 0, inputthread->input_name, 0);
728		}
729
730
731		lck_mtx_assert(inputthread->input_lck, LCK_MTX_ASSERT_OWNED);
732
733		m = inputthread->mbuf_head;
734		inputthread->mbuf_head = NULL;
735		inputthread->mbuf_tail = NULL;
736
737		if (inputthread->input_waiting & DLIL_INPUT_TERMINATE) {
738				if (m)
739					mbuf_freem_list(m);
740				/* this is the end */
741				dlil_terminate_input_thread(inputthread);
742				return;
743		}
744
745		inputthread->input_waiting |= DLIL_INPUT_RUNNING;
746		inputthread->input_waiting &= ~DLIL_INPUT_WAITING;
747
748		if (inputthread == dlil_lo_thread_ptr) {
749			m_loop = dlil_lo_input_mbuf_head;
750			dlil_lo_input_mbuf_head = NULL;
751			dlil_lo_input_mbuf_tail = NULL;
752		}
753
754#if IFNET_INPUT_SANITY_CHK
755		if (dlil_input_sanity_check != 0) {
756			mbuf_cnt = inputthread->mbuf_count;
757			inputthread->mbuf_count = 0;
758			if (inputthread == dlil_lo_thread_ptr) {
759				loop_cnt = dlil_lo_input_mbuf_count;
760				dlil_lo_input_mbuf_count = 0;
761			}
762
763			lck_mtx_unlock(inputthread->input_lck);
764
765			for (m1 = m, count = 0; m1; m1 = mbuf_nextpkt(m1)) {
766				count++;
767			}
768			if (count != mbuf_cnt) {
769				panic("dlil_input_func - thread=%p reg. loop queue has %d packets, should have %d\n",
770					  inputthread, count, mbuf_cnt);
771			}
772
773			if (inputthread == dlil_lo_thread_ptr) {
774				for (m1 = m_loop, count = 0; m1; m1 = mbuf_nextpkt(m1)) {
775					count++;
776				}
777				if (count != loop_cnt) {
778					panic("dlil_input_func - thread=%p loop queue has %d packets, should have %d\n",
779					  inputthread, count, loop_cnt);
780				}
781			}
782		} else
783#endif /* IFNET_INPUT_SANITY_CHK */
784		{
785			lck_mtx_unlock(inputthread->input_lck);
786		}
787
788
789		/*
790		* NOTE warning %%% attention !!!!
791		* We should think about putting some thread starvation safeguards if
792		* we deal with long chains of packets.
793		*/
794		if (m_loop) {
795			if (inputthread == dlil_lo_thread_ptr)
796				dlil_input_packet_list(lo_ifp, m_loop);
797#if IFNET_INPUT_SANITY_CHK
798			else
799				panic("dlil_input_func - thread=%p loop queue has %d packets, should have none!\n",
800				  inputthread, loop_cnt);
801#endif /* IFNET_INPUT_SANITY_CHK */
802		}
803
804
805		if (m)
806			dlil_input_packet_list(0, m);
807
808
809		lck_mtx_lock(inputthread->input_lck);
810
811		if ((inputthread->input_waiting & (DLIL_PROTO_WAITING | DLIL_PROTO_REGISTER)) != 0)  {
812			lck_mtx_unlock(inputthread->input_lck);
813			proto_input_run();
814		}
815		else
816			lck_mtx_unlock(inputthread->input_lck);
817	}
818}
819
820errno_t
821ifnet_input(
822	ifnet_t									ifp,
823	mbuf_t									m_head,
824	const struct ifnet_stat_increment_param	*stats)
825{
826	struct thread *tp = current_thread();
827	mbuf_t		m_tail;
828	struct dlil_threading_info *inp;
829#if IFNET_INPUT_SANITY_CHK
830	u_int32_t	pkt_count = 0;
831#endif /* IFNET_INPUT_SANITY_CHK */
832
833	if (ifp == NULL || m_head == NULL) {
834		if (m_head)
835			mbuf_freem_list(m_head);
836		return EINVAL;
837	}
838
839	m_tail = m_head;
840	while (1) {
841#if IFNET_INPUT_SANITY_CHK
842		if (dlil_input_sanity_check != 0) {
843			ifnet_t	rcvif;
844
845			rcvif = mbuf_pkthdr_rcvif(m_tail);
846			pkt_count++;
847
848			if (rcvif == NULL ||
849				(ifp->if_type != IFT_LOOP && rcvif != ifp) ||
850				(mbuf_flags(m_head) & MBUF_PKTHDR) == 0) {
851				panic("ifnet_input - invalid mbuf %p\n", m_tail);
852			}
853		}
854#endif /* IFNET_INPUT_SANITY_CHK */
855		if (mbuf_nextpkt(m_tail) == NULL)
856			break;
857		m_tail = mbuf_nextpkt(m_tail);
858	}
859
860	inp = ifp->if_input_thread;
861
862	if (dlil_multithreaded_input == 0 || inp == NULL)
863		inp = dlil_lo_thread_ptr;
864
865	/*
866	 * If there is a matching dlil input thread associated with an
867	 * affinity set, associate this workloop thread with the same set.
868	 * We will only do this once.
869	 */
870	lck_mtx_lock(inp->input_lck);
871	if (inp->net_affinity && inp->workloop_thread == NULL) {
872		u_int32_t tag = inp->tag;
873		inp->workloop_thread = tp;
874		lck_mtx_unlock(inp->input_lck);
875
876		/* Associated the current thread with the new affinity tag */
877		(void) dlil_affinity_set(tp, tag);
878
879		/*
880		 * Take a reference on the workloop (current) thread; during
881		 * detach, we will need to refer to it in order ot tear down
882		 * its affinity.
883		 */
884		thread_reference(tp);
885		lck_mtx_lock(inp->input_lck);
886	}
887
888        /* WARNING
889	 * Because of loopbacked multicast we cannot stuff the ifp in
890	 * the rcvif of the packet header: loopback has its own dlil
891	 * input queue
892	 */
893
894	if (inp == dlil_lo_thread_ptr && ifp->if_type == IFT_LOOP) {
895		if (dlil_lo_input_mbuf_head == NULL)
896			dlil_lo_input_mbuf_head = m_head;
897		else if (dlil_lo_input_mbuf_tail != NULL)
898			dlil_lo_input_mbuf_tail->m_nextpkt = m_head;
899		dlil_lo_input_mbuf_tail = m_tail;
900#if IFNET_INPUT_SANITY_CHK
901		if (dlil_input_sanity_check != 0) {
902			dlil_lo_input_mbuf_count += pkt_count;
903			inp->input_mbuf_cnt += pkt_count;
904			inp->input_wake_cnt++;
905
906			lck_mtx_assert(inp->input_lck, LCK_MTX_ASSERT_OWNED);
907		}
908#endif
909	}
910	else {
911		if (inp->mbuf_head == NULL)
912			inp->mbuf_head = m_head;
913		else if (inp->mbuf_tail != NULL)
914			inp->mbuf_tail->m_nextpkt = m_head;
915		inp->mbuf_tail = m_tail;
916#if IFNET_INPUT_SANITY_CHK
917		if (dlil_input_sanity_check != 0) {
918			inp->mbuf_count += pkt_count;
919			inp->input_mbuf_cnt += pkt_count;
920			inp->input_wake_cnt++;
921
922			lck_mtx_assert(inp->input_lck, LCK_MTX_ASSERT_OWNED);
923		}
924#endif
925	}
926
927
928	inp->input_waiting |= DLIL_INPUT_WAITING;
929	if ((inp->input_waiting & DLIL_INPUT_RUNNING) == 0) {
930		wakeup((caddr_t)&inp->input_waiting);
931	}
932	if (stats) {
933		ifp->if_data.ifi_ipackets += stats->packets_in;
934		ifp->if_data.ifi_ibytes += stats->bytes_in;
935		ifp->if_data.ifi_ierrors += stats->errors_in;
936
937		ifp->if_data.ifi_opackets += stats->packets_out;
938		ifp->if_data.ifi_obytes += stats->bytes_out;
939		ifp->if_data.ifi_oerrors += stats->errors_out;
940
941		ifp->if_data.ifi_collisions += stats->collisions;
942		ifp->if_data.ifi_iqdrops += stats->dropped;
943	}
944
945	lck_mtx_unlock(inp->input_lck);
946
947	return 0;
948}
949
950static int
951dlil_interface_filters_input(struct ifnet * ifp, struct mbuf * * m_p,
952			     char * * frame_header_p,
953			     protocol_family_t protocol_family)
954{
955	struct ifnet_filter * 		filter;
956
957	TAILQ_FOREACH(filter, &ifp->if_flt_head, filt_next) {
958		int result;
959
960		if (filter->filt_input
961		    && (filter->filt_protocol == 0
962			|| filter->filt_protocol == protocol_family)) {
963			result = (*filter->filt_input)(filter->filt_cookie,
964						       ifp, protocol_family,
965						       m_p, frame_header_p);
966			if (result != 0) {
967				return (result);
968			}
969		}
970	}
971	return (0);
972}
973
974static void
975dlil_ifproto_input(struct if_proto * ifproto, mbuf_t m)
976{
977	int error;
978
979	if (ifproto->proto_kpi == kProtoKPI_v1) {
980		/* Version 1 protocols get one packet at a time */
981		while (m != NULL) {
982			char *	frame_header;
983			mbuf_t	next_packet;
984
985			next_packet = m->m_nextpkt;
986			m->m_nextpkt = NULL;
987			frame_header = m->m_pkthdr.header;
988			m->m_pkthdr.header = NULL;
989			error = (*ifproto->kpi.v1.input)(ifproto->ifp,
990							 ifproto->protocol_family,
991							 m, frame_header);
992			if (error != 0 && error != EJUSTRETURN)
993				m_freem(m);
994			m = next_packet;
995		}
996	}
997	else if (ifproto->proto_kpi == kProtoKPI_v2) {
998		/* Version 2 protocols support packet lists */
999		error = (*ifproto->kpi.v2.input)(ifproto->ifp,
1000						 ifproto->protocol_family,
1001						 m);
1002		if (error != 0 && error != EJUSTRETURN)
1003			m_freem_list(m);
1004	}
1005	return;
1006}
1007
1008__private_extern__ void
1009dlil_input_packet_list(struct ifnet * ifp_param, struct mbuf *m)
1010{
1011	int				error = 0;
1012	int				locked = 0;
1013	protocol_family_t		protocol_family;
1014	mbuf_t				next_packet;
1015	ifnet_t				ifp = ifp_param;
1016	char *				frame_header;
1017	struct if_proto	*		last_ifproto = NULL;
1018	mbuf_t				pkt_first = NULL;
1019	mbuf_t *			pkt_next = NULL;
1020
1021	KERNEL_DEBUG(DBG_FNC_DLIL_INPUT | DBG_FUNC_START,0,0,0,0,0);
1022
1023	while (m != NULL) {
1024		struct if_proto *	ifproto = NULL;
1025
1026		next_packet = m->m_nextpkt;
1027		m->m_nextpkt = NULL;
1028		if (ifp_param == NULL)
1029			ifp = m->m_pkthdr.rcvif;
1030		frame_header = m->m_pkthdr.header;
1031		m->m_pkthdr.header = NULL;
1032
1033		if (locked == 0) {
1034			/* dlil lock protects the demux and interface filters */
1035			locked = 1;
1036			dlil_read_begin();
1037		}
1038		/* find which protocol family this packet is for */
1039		error = (*ifp->if_demux)(ifp, m, frame_header,
1040					 &protocol_family);
1041		if (error != 0) {
1042			if (error == EJUSTRETURN) {
1043				goto next;
1044			}
1045			protocol_family = 0;
1046		}
1047
1048		/* DANGER!!! */
1049		if (m->m_flags & (M_BCAST|M_MCAST))
1050			ifp->if_imcasts++;
1051
1052		/* run interface filters, exclude VLAN packets PR-3586856 */
1053		if ((m->m_pkthdr.csum_flags & CSUM_VLAN_TAG_VALID) == 0) {
1054			int	filter_result;
1055
1056			filter_result = dlil_interface_filters_input(ifp, &m,
1057							  &frame_header,
1058							  protocol_family);
1059			if (filter_result != 0) {
1060				if (filter_result != EJUSTRETURN) {
1061					m_freem(m);
1062				}
1063				goto next;
1064			}
1065		}
1066		if (error != 0 || ((m->m_flags & M_PROMISC) != 0) ) {
1067			m_freem(m);
1068			goto next;
1069		}
1070
1071		/* Lookup the protocol attachment to this interface */
1072		if (protocol_family == 0) {
1073			ifproto = NULL;
1074		}
1075		else if (last_ifproto != NULL
1076			 && last_ifproto->ifp == ifp
1077			 && (last_ifproto->protocol_family
1078			     == protocol_family)) {
1079			ifproto = last_ifproto;
1080		}
1081		else {
1082			ifproto	= find_attached_proto(ifp, protocol_family);
1083		}
1084		if (ifproto == NULL) {
1085			/* no protocol for this packet, discard */
1086			m_freem(m);
1087			goto next;
1088		}
1089		if (ifproto != last_ifproto) {
1090			/* make sure ifproto can't go away during input */
1091			if_proto_ref(ifproto);
1092			if (last_ifproto != NULL) {
1093				/* pass up the list for the previous protocol */
1094				dlil_read_end();
1095
1096				dlil_ifproto_input(last_ifproto, pkt_first);
1097				pkt_first = NULL;
1098				if_proto_free(last_ifproto);
1099				dlil_read_begin();
1100			}
1101			last_ifproto = ifproto;
1102		}
1103		/* extend the list */
1104		m->m_pkthdr.header = frame_header;
1105		if (pkt_first == NULL) {
1106			pkt_first = m;
1107		} else {
1108			*pkt_next = m;
1109		}
1110		pkt_next = &m->m_nextpkt;
1111
1112	next:
1113		if (next_packet == NULL && last_ifproto != NULL) {
1114			/* pass up the last list of packets */
1115			dlil_read_end();
1116
1117			dlil_ifproto_input(last_ifproto, pkt_first);
1118			if_proto_free(last_ifproto);
1119			locked = 0;
1120		}
1121		m = next_packet;
1122
1123	}
1124	if (locked != 0) {
1125		dlil_read_end();
1126	}
1127	KERNEL_DEBUG(DBG_FNC_DLIL_INPUT | DBG_FUNC_END,0,0,0,0,0);
1128	return;
1129}
1130
1131static int
1132dlil_event_internal(struct ifnet *ifp, struct kev_msg *event)
1133{
1134	struct ifnet_filter *filter;
1135
1136	if (ifp_use(ifp, kIfNetUseCount_MustNotBeZero) == 0) {
1137		dlil_read_begin();
1138
1139		/* Pass the event to the interface filters */
1140		TAILQ_FOREACH(filter, &ifp->if_flt_head, filt_next) {
1141			if (filter->filt_event)
1142				filter->filt_event(filter->filt_cookie, ifp, filter->filt_protocol, event);
1143		}
1144
1145		if (ifp->if_proto_hash) {
1146			int i;
1147
1148			for (i = 0; i < PROTO_HASH_SLOTS; i++) {
1149				struct if_proto *proto;
1150
1151				SLIST_FOREACH(proto, &ifp->if_proto_hash[i], next_hash) {
1152					proto_media_event eventp = proto->proto_kpi == kProtoKPI_v1
1153						 ? proto->kpi.v1.event : proto->kpi.v2.event;
1154
1155					if (eventp)
1156						eventp(ifp, proto->protocol_family, event);
1157				}
1158			}
1159		}
1160
1161		dlil_read_end();
1162
1163		/* Pass the event to the interface */
1164		if (ifp->if_event)
1165			ifp->if_event(ifp, event);
1166
1167		if (ifp_unuse(ifp))
1168			ifp_use_reached_zero(ifp);
1169	}
1170
1171	return kev_post_msg(event);
1172}
1173
1174errno_t
1175ifnet_event(
1176	ifnet_t					ifp,
1177	struct kern_event_msg	*event)
1178{
1179	struct kev_msg               kev_msg;
1180	int result = 0;
1181
1182	if (ifp == NULL || event == NULL) return EINVAL;
1183
1184	kev_msg.vendor_code    = event->vendor_code;
1185	kev_msg.kev_class      = event->kev_class;
1186	kev_msg.kev_subclass   = event->kev_subclass;
1187	kev_msg.event_code     = event->event_code;
1188	kev_msg.dv[0].data_ptr = &event->event_data[0];
1189	kev_msg.dv[0].data_length = event->total_size - KEV_MSG_HEADER_SIZE;
1190	kev_msg.dv[1].data_length = 0;
1191
1192	result = dlil_event_internal(ifp, &kev_msg);
1193
1194	return result;
1195}
1196
1197#if CONFIG_MACF_NET
1198#include <netinet/ip6.h>
1199#include <netinet/ip.h>
1200static int dlil_get_socket_type(struct mbuf **mp, int family, int raw)
1201{
1202	struct mbuf *m;
1203	struct ip *ip;
1204	struct ip6_hdr *ip6;
1205	int type = SOCK_RAW;
1206
1207	if (!raw) {
1208		switch (family) {
1209		case PF_INET:
1210			m = m_pullup(*mp, sizeof(struct ip));
1211			if (m == NULL)
1212				break;
1213			*mp = m;
1214			ip = mtod(m, struct ip *);
1215			if (ip->ip_p == IPPROTO_TCP)
1216				type = SOCK_STREAM;
1217			else if (ip->ip_p == IPPROTO_UDP)
1218				type = SOCK_DGRAM;
1219			break;
1220		case PF_INET6:
1221			m = m_pullup(*mp, sizeof(struct ip6_hdr));
1222			if (m == NULL)
1223				break;
1224			*mp = m;
1225			ip6 = mtod(m, struct ip6_hdr *);
1226			if (ip6->ip6_nxt == IPPROTO_TCP)
1227				type = SOCK_STREAM;
1228			else if (ip6->ip6_nxt == IPPROTO_UDP)
1229				type = SOCK_DGRAM;
1230			break;
1231		}
1232	}
1233
1234	return (type);
1235}
1236#endif
1237
1238#if 0
1239int
1240dlil_output_list(
1241	struct ifnet* ifp,
1242	u_long proto_family,
1243	struct mbuf		*packetlist,
1244	caddr_t		route,
1245	const struct sockaddr	*dest,
1246	int						raw)
1247{
1248	char			*frame_type = NULL;
1249	char			*dst_linkaddr = NULL;
1250	int				retval = 0;
1251	char			frame_type_buffer[MAX_FRAME_TYPE_SIZE * 4];
1252	char			dst_linkaddr_buffer[MAX_LINKADDR * 4];
1253	struct ifnet_filter *filter;
1254	struct if_proto	*proto = 0;
1255	mbuf_t	m;
1256	mbuf_t	send_head = NULL;
1257	mbuf_t	*send_tail = &send_head;
1258
1259	KERNEL_DEBUG(DBG_FNC_DLIL_OUTPUT | DBG_FUNC_START,0,0,0,0,0);
1260
1261	dlil_read_begin();
1262
1263	frame_type	   = frame_type_buffer;
1264	dst_linkaddr   = dst_linkaddr_buffer;
1265
1266	if (raw == 0) {
1267		proto = find_attached_proto(ifp, proto_family);
1268		if (proto == NULL) {
1269			retval = ENXIO;
1270			goto cleanup;
1271		}
1272	}
1273
1274preout_again:
1275	if (packetlist == NULL)
1276		goto cleanup;
1277	m = packetlist;
1278	packetlist = packetlist->m_nextpkt;
1279	m->m_nextpkt = NULL;
1280
1281	if (raw == 0) {
1282		proto_media_preout preoutp = proto->proto_kpi == kProtoKPI_v1
1283			 ? proto->kpi.v1.pre_output : proto->kpi.v2.pre_output;
1284		retval = 0;
1285		if (preoutp)
1286			retval = preoutp(ifp, proto_family, &m, dest, route, frame_type, dst_linkaddr);
1287
1288		if (retval) {
1289			if (retval == EJUSTRETURN) {
1290				goto preout_again;
1291			}
1292
1293			m_freem(m);
1294			goto cleanup;
1295		}
1296	}
1297
1298	do {
1299#if CONFIG_MACF_NET
1300		retval = mac_ifnet_check_transmit(ifp, m, proto_family,
1301		    dlil_get_socket_type(&m, proto_family, raw));
1302		if (retval) {
1303			m_freem(m);
1304			goto cleanup;
1305		}
1306#endif
1307
1308		if (raw == 0 && ifp->if_framer) {
1309			retval = ifp->if_framer(ifp, &m, dest, dst_linkaddr, frame_type);
1310			if (retval) {
1311				if (retval != EJUSTRETURN) {
1312					m_freem(m);
1313				}
1314				goto next;
1315			}
1316		}
1317
1318#if BRIDGE
1319		/* !!!LOCKING!!!
1320		 *
1321		 * Need to consider how to handle this.
1322		 * Also note that return should be a goto cleanup
1323		 */
1324		broken-locking
1325		if (do_bridge) {
1326			struct mbuf *m0 = m;
1327			struct ether_header *eh = mtod(m, struct ether_header *);
1328
1329			if (m->m_pkthdr.rcvif)
1330				m->m_pkthdr.rcvif = NULL;
1331			ifp = bridge_dst_lookup(eh);
1332			bdg_forward(&m0, ifp);
1333			if (m0)
1334				m_freem(m0);
1335
1336			return 0 - should be goto cleanup?
1337		}
1338#endif
1339
1340		/*
1341		 * Let interface filters (if any) do their thing ...
1342		 */
1343		/* Do not pass VLAN tagged packets to filters PR-3586856 */
1344		if ((m->m_pkthdr.csum_flags & CSUM_VLAN_TAG_VALID) == 0) {
1345			TAILQ_FOREACH(filter, &ifp->if_flt_head, filt_next) {
1346				if ((filter->filt_protocol == 0 || (filter->filt_protocol == proto_family)) &&
1347					filter->filt_output) {
1348					retval = filter->filt_output(filter->filt_cookie, ifp, proto_family, &m);
1349					if (retval) {
1350						if (retval != EJUSTRETURN)
1351							m_freem(m);
1352						goto next;
1353					}
1354				}
1355			}
1356		}
1357
1358		/*
1359		 * Finally, call the driver.
1360		 */
1361
1362		if ((ifp->if_eflags & IFEF_SENDLIST) != 0) {
1363			*send_tail = m;
1364			send_tail = &m->m_nextpkt;
1365		}
1366		else {
1367			KERNEL_DEBUG(DBG_FNC_DLIL_IFOUT | DBG_FUNC_START, 0,0,0,0,0);
1368			retval = ifp->if_output(ifp, m);
1369			if (retval) {
1370				printf("dlil_output: output error retval = %x\n", retval);
1371			}
1372			KERNEL_DEBUG(DBG_FNC_DLIL_IFOUT | DBG_FUNC_END, 0,0,0,0,0);
1373		}
1374		KERNEL_DEBUG(DBG_FNC_DLIL_IFOUT | DBG_FUNC_END, 0,0,0,0,0);
1375
1376next:
1377		m = packetlist;
1378		if (m) {
1379			packetlist = packetlist->m_nextpkt;
1380			m->m_nextpkt = NULL;
1381		}
1382	} while (m);
1383
1384	if (send_head) {
1385		KERNEL_DEBUG(DBG_FNC_DLIL_IFOUT | DBG_FUNC_START, 0,0,0,0,0);
1386		retval = ifp->if_output(ifp, send_head);
1387		if (retval) {
1388			printf("dlil_output: output error retval = %x\n", retval);
1389		}
1390		KERNEL_DEBUG(DBG_FNC_DLIL_IFOUT | DBG_FUNC_END, 0,0,0,0,0);
1391	}
1392
1393	KERNEL_DEBUG(DBG_FNC_DLIL_OUTPUT | DBG_FUNC_END,0,0,0,0,0);
1394
1395cleanup:
1396	dlil_read_end();
1397	if (packetlist) /* if any packet left, clean up */
1398		mbuf_freem_list(packetlist);
1399	if (retval == EJUSTRETURN)
1400		retval = 0;
1401	return retval;
1402}
1403#endif
1404
1405/*
1406 * dlil_output
1407 *
1408 * Caller should have a lock on the protocol domain if the protocol
1409 * doesn't support finer grained locking. In most cases, the lock
1410 * will be held from the socket layer and won't be released until
1411 * we return back to the socket layer.
1412 *
1413 * This does mean that we must take a protocol lock before we take
1414 * an interface lock if we're going to take both. This makes sense
1415 * because a protocol is likely to interact with an ifp while it
1416 * is under the protocol lock.
1417 */
1418__private_extern__ errno_t
1419dlil_output(
1420	ifnet_t					ifp,
1421	protocol_family_t		proto_family,
1422	mbuf_t					packetlist,
1423	void					*route,
1424	const struct sockaddr	*dest,
1425	int						raw)
1426{
1427	char			*frame_type = NULL;
1428	char			*dst_linkaddr = NULL;
1429	int				retval = 0;
1430	char			frame_type_buffer[MAX_FRAME_TYPE_SIZE * 4];
1431	char			dst_linkaddr_buffer[MAX_LINKADDR * 4];
1432	struct ifnet_filter *filter;
1433	struct if_proto	*proto = 0;
1434	mbuf_t	m;
1435	mbuf_t	send_head = NULL;
1436	mbuf_t	*send_tail = &send_head;
1437
1438	KERNEL_DEBUG(DBG_FNC_DLIL_OUTPUT | DBG_FUNC_START,0,0,0,0,0);
1439
1440	dlil_read_begin();
1441
1442	frame_type	   = frame_type_buffer;
1443	dst_linkaddr   = dst_linkaddr_buffer;
1444
1445	if (raw == 0) {
1446		proto = find_attached_proto(ifp, proto_family);
1447		if (proto == NULL) {
1448			retval = ENXIO;
1449			goto cleanup;
1450		}
1451	}
1452
1453preout_again:
1454	if (packetlist == NULL)
1455		goto cleanup;
1456	m = packetlist;
1457	packetlist = packetlist->m_nextpkt;
1458	m->m_nextpkt = NULL;
1459
1460	if (raw == 0) {
1461		proto_media_preout preoutp = proto->proto_kpi == kProtoKPI_v1
1462			 ? proto->kpi.v1.pre_output : proto->kpi.v2.pre_output;
1463		retval = 0;
1464		if (preoutp)
1465			retval = preoutp(ifp, proto_family, &m, dest, route, frame_type, dst_linkaddr);
1466
1467		if (retval) {
1468			if (retval == EJUSTRETURN) {
1469				goto preout_again;
1470			}
1471
1472			m_freem(m);
1473			goto cleanup;
1474		}
1475	}
1476
1477#if CONFIG_MACF_NET
1478	retval = mac_ifnet_check_transmit(ifp, m, proto_family,
1479	    dlil_get_socket_type(&m, proto_family, raw));
1480	if (retval) {
1481		m_freem(m);
1482		goto cleanup;
1483	}
1484#endif
1485
1486	do {
1487		if (raw == 0 && ifp->if_framer) {
1488			retval = ifp->if_framer(ifp, &m, dest, dst_linkaddr, frame_type);
1489			if (retval) {
1490				if (retval != EJUSTRETURN) {
1491					m_freem(m);
1492				}
1493				goto next;
1494			}
1495		}
1496
1497#if BRIDGE
1498		/* !!!LOCKING!!!
1499		 *
1500		 * Need to consider how to handle this.
1501		 * Also note that return should be a goto cleanup
1502		 */
1503		broken-locking
1504		if (do_bridge) {
1505			struct mbuf *m0 = m;
1506			struct ether_header *eh = mtod(m, struct ether_header *);
1507
1508			if (m->m_pkthdr.rcvif)
1509				m->m_pkthdr.rcvif = NULL;
1510			ifp = bridge_dst_lookup(eh);
1511			bdg_forward(&m0, ifp);
1512			if (m0)
1513				m_freem(m0);
1514
1515			return 0 - should be goto cleanup?
1516		}
1517#endif
1518
1519		/*
1520		 * Let interface filters (if any) do their thing ...
1521		 */
1522		/* Do not pass VLAN tagged packets to filters PR-3586856 */
1523		if ((m->m_pkthdr.csum_flags & CSUM_VLAN_TAG_VALID) == 0) {
1524			TAILQ_FOREACH(filter, &ifp->if_flt_head, filt_next) {
1525				if ((filter->filt_protocol == 0 || (filter->filt_protocol == proto_family)) &&
1526					filter->filt_output) {
1527					retval = filter->filt_output(filter->filt_cookie, ifp, proto_family, &m);
1528					if (retval) {
1529						if (retval != EJUSTRETURN)
1530							m_freem(m);
1531						goto next;
1532					}
1533				}
1534			}
1535		}
1536
1537		/*
1538		 * If the underlying interface is not capable of handling a
1539		 * packet whose data portion spans across physically disjoint
1540		 * pages, we need to "normalize" the packet so that we pass
1541		 * down a chain of mbufs where each mbuf points to a span that
1542		 * resides in the system page boundary.  If the packet does
1543		 * not cross page(s), the following is a no-op.
1544		 */
1545		if (!(ifp->if_hwassist & IFNET_MULTIPAGES)) {
1546			if ((m = m_normalize(m)) == NULL)
1547				goto next;
1548		}
1549
1550		/*
1551		 * Finally, call the driver.
1552		 */
1553
1554		if ((ifp->if_eflags & IFEF_SENDLIST) != 0) {
1555			*send_tail = m;
1556			send_tail = &m->m_nextpkt;
1557		}
1558		else {
1559			KERNEL_DEBUG(DBG_FNC_DLIL_IFOUT | DBG_FUNC_START, 0,0,0,0,0);
1560			retval = ifp->if_output(ifp, m);
1561			if (retval) {
1562				printf("dlil_output: output error retval = %x\n", retval);
1563			}
1564			KERNEL_DEBUG(DBG_FNC_DLIL_IFOUT | DBG_FUNC_END, 0,0,0,0,0);
1565		}
1566		KERNEL_DEBUG(DBG_FNC_DLIL_IFOUT | DBG_FUNC_END, 0,0,0,0,0);
1567
1568next:
1569		m = packetlist;
1570		if (m) {
1571			packetlist = packetlist->m_nextpkt;
1572			m->m_nextpkt = NULL;
1573		}
1574	} while (m);
1575
1576	if (send_head) {
1577		KERNEL_DEBUG(DBG_FNC_DLIL_IFOUT | DBG_FUNC_START, 0,0,0,0,0);
1578		retval = ifp->if_output(ifp, send_head);
1579		if (retval) {
1580			printf("dlil_output: output error retval = %x\n", retval);
1581		}
1582		KERNEL_DEBUG(DBG_FNC_DLIL_IFOUT | DBG_FUNC_END, 0,0,0,0,0);
1583	}
1584
1585	KERNEL_DEBUG(DBG_FNC_DLIL_OUTPUT | DBG_FUNC_END,0,0,0,0,0);
1586
1587cleanup:
1588	dlil_read_end();
1589	if (packetlist) /* if any packet left, clean up */
1590		mbuf_freem_list(packetlist);
1591	if (retval == EJUSTRETURN)
1592		retval = 0;
1593	return retval;
1594}
1595
1596errno_t
1597ifnet_ioctl(
1598	ifnet_t				ifp,
1599	protocol_family_t	proto_fam,
1600	u_int32_t			ioctl_code,
1601	void				*ioctl_arg)
1602{
1603	struct ifnet_filter		*filter;
1604	int						retval = EOPNOTSUPP;
1605	int						result = 0;
1606	int						holding_read = 0;
1607
1608	if (ifp == NULL || ioctl_code == 0)
1609		return EINVAL;
1610
1611	/* Attempt to increment the use count. If it's zero, bail out, the ifp is invalid */
1612	result = ifp_use(ifp, kIfNetUseCount_MustNotBeZero);
1613	if (result != 0)
1614		return EOPNOTSUPP;
1615
1616	dlil_read_begin();
1617	holding_read = 1;
1618
1619	/* Run the interface filters first.
1620	 * We want to run all filters before calling the protocol,
1621	 * interface family, or interface.
1622	 */
1623	TAILQ_FOREACH(filter, &ifp->if_flt_head, filt_next) {
1624		if ((filter->filt_protocol == 0 || (filter->filt_protocol == proto_fam)) &&
1625			filter->filt_ioctl != NULL) {
1626			result = filter->filt_ioctl(filter->filt_cookie, ifp, proto_fam, ioctl_code, ioctl_arg);
1627			/* Only update retval if no one has handled the ioctl */
1628			if (retval == EOPNOTSUPP || result == EJUSTRETURN) {
1629				if (result == ENOTSUP)
1630					result = EOPNOTSUPP;
1631				retval = result;
1632				if (retval && retval != EOPNOTSUPP) {
1633					goto cleanup;
1634				}
1635			}
1636		}
1637	}
1638
1639	/* Allow the protocol to handle the ioctl */
1640	if (proto_fam) {
1641		struct if_proto	*proto = find_attached_proto(ifp, proto_fam);
1642
1643		if (proto != 0) {
1644			proto_media_ioctl ioctlp = proto->proto_kpi == kProtoKPI_v1
1645				 ? proto->kpi.v1.ioctl : proto->kpi.v2.ioctl;
1646			result = EOPNOTSUPP;
1647			if (ioctlp)
1648				result = ioctlp(ifp, proto_fam, ioctl_code, ioctl_arg);
1649
1650			/* Only update retval if no one has handled the ioctl */
1651			if (retval == EOPNOTSUPP || result == EJUSTRETURN) {
1652				if (result == ENOTSUP)
1653					result = EOPNOTSUPP;
1654				retval = result;
1655				if (retval && retval != EOPNOTSUPP) {
1656					goto cleanup;
1657				}
1658			}
1659		}
1660	}
1661
1662	/*
1663	 * Since we have incremented the use count on the ifp, we are guaranteed
1664	 * that the ifp will not go away (the function pointers may not be changed).
1665	 * We release the dlil read lock so the interface ioctl may trigger a
1666	 * protocol attach. This happens with vlan and may occur with other virtual
1667	 * interfaces.
1668	 */
1669	dlil_read_end();
1670	holding_read = 0;
1671
1672	/* retval is either 0 or EOPNOTSUPP */
1673
1674	/*
1675	 * Let the interface handle this ioctl.
1676	 * If it returns EOPNOTSUPP, ignore that, we may have
1677	 * already handled this in the protocol or family.
1678	 */
1679	if (ifp->if_ioctl)
1680		result = (*ifp->if_ioctl)(ifp, ioctl_code, ioctl_arg);
1681
1682	/* Only update retval if no one has handled the ioctl */
1683	if (retval == EOPNOTSUPP || result == EJUSTRETURN) {
1684		if (result == ENOTSUP)
1685			result = EOPNOTSUPP;
1686		retval = result;
1687		if (retval && retval != EOPNOTSUPP) {
1688			goto cleanup;
1689		}
1690	}
1691
1692cleanup:
1693	if (holding_read)
1694		dlil_read_end();
1695	if (ifp_unuse(ifp))
1696		ifp_use_reached_zero(ifp);
1697
1698	if (retval == EJUSTRETURN)
1699		retval = 0;
1700	return retval;
1701}
1702
1703__private_extern__ errno_t
1704dlil_set_bpf_tap(
1705	ifnet_t			ifp,
1706	bpf_tap_mode	mode,
1707	bpf_packet_func	callback)
1708{
1709	errno_t	error = 0;
1710
1711	dlil_read_begin();
1712	if (ifp->if_set_bpf_tap)
1713		error = ifp->if_set_bpf_tap(ifp, mode, callback);
1714	dlil_read_end();
1715
1716	return error;
1717}
1718
1719errno_t
1720dlil_resolve_multi(
1721	struct ifnet *ifp,
1722	const struct sockaddr *proto_addr,
1723	struct sockaddr *ll_addr,
1724	size_t ll_len)
1725{
1726	errno_t	result = EOPNOTSUPP;
1727	struct if_proto *proto;
1728	const struct sockaddr *verify;
1729	proto_media_resolve_multi resolvep;
1730
1731	dlil_read_begin();
1732
1733	bzero(ll_addr, ll_len);
1734
1735	/* Call the protocol first */
1736	proto = find_attached_proto(ifp, proto_addr->sa_family);
1737	if (proto != NULL) {
1738		resolvep = proto->proto_kpi == kProtoKPI_v1
1739			 ? proto->kpi.v1.resolve_multi : proto->kpi.v2.resolve_multi;
1740		if (resolvep != NULL)
1741			result = resolvep(ifp, proto_addr,(struct sockaddr_dl*)ll_addr,
1742							  ll_len);
1743	}
1744
1745	/* Let the interface verify the multicast address */
1746	if ((result == EOPNOTSUPP || result == 0) && ifp->if_check_multi) {
1747		if (result == 0)
1748			verify = ll_addr;
1749		else
1750			verify = proto_addr;
1751		result = ifp->if_check_multi(ifp, verify);
1752	}
1753
1754	dlil_read_end();
1755
1756	return result;
1757}
1758
1759__private_extern__ errno_t
1760dlil_send_arp_internal(
1761	ifnet_t	ifp,
1762	u_short arpop,
1763	const struct sockaddr_dl* sender_hw,
1764	const struct sockaddr* sender_proto,
1765	const struct sockaddr_dl* target_hw,
1766	const struct sockaddr* target_proto)
1767{
1768	struct if_proto *proto;
1769	errno_t	result = 0;
1770
1771	dlil_read_begin();
1772
1773	proto = find_attached_proto(ifp, target_proto->sa_family);
1774	if (proto == NULL) {
1775		result = ENOTSUP;
1776	}
1777	else {
1778		proto_media_send_arp	arpp;
1779		arpp = proto->proto_kpi == kProtoKPI_v1
1780			 ? proto->kpi.v1.send_arp : proto->kpi.v2.send_arp;
1781		if (arpp == NULL)
1782			result = ENOTSUP;
1783		else
1784			result = arpp(ifp, arpop, sender_hw, sender_proto, target_hw,
1785						  target_proto);
1786	}
1787
1788	dlil_read_end();
1789
1790	return result;
1791}
1792
1793static __inline__ int
1794_is_announcement(const struct sockaddr_in * sender_sin,
1795		     const struct sockaddr_in * target_sin)
1796{
1797	if (sender_sin == NULL) {
1798		return FALSE;
1799	}
1800	return (sender_sin->sin_addr.s_addr == target_sin->sin_addr.s_addr);
1801}
1802
1803__private_extern__ errno_t
1804dlil_send_arp(
1805	ifnet_t	ifp,
1806	u_short arpop,
1807	const struct sockaddr_dl* sender_hw,
1808	const struct sockaddr* sender_proto,
1809	const struct sockaddr_dl* target_hw,
1810	const struct sockaddr* target_proto)
1811{
1812	errno_t	result = 0;
1813	const struct sockaddr_in * sender_sin;
1814	const struct sockaddr_in * target_sin;
1815
1816	if (target_proto == NULL || (sender_proto &&
1817		sender_proto->sa_family != target_proto->sa_family))
1818		return EINVAL;
1819
1820	/*
1821	 * If this is an ARP request and the target IP is IPv4LL,
1822	 * send the request on all interfaces.  The exception is
1823	 * an announcement, which must only appear on the specific
1824	 * interface.
1825	 */
1826	sender_sin = (const struct sockaddr_in *)sender_proto;
1827	target_sin = (const struct sockaddr_in *)target_proto;
1828	if (target_proto->sa_family == AF_INET
1829	    && IN_LINKLOCAL(ntohl(target_sin->sin_addr.s_addr))
1830	    && ipv4_ll_arp_aware != 0
1831	    && arpop == ARPOP_REQUEST
1832	    && !_is_announcement(target_sin, sender_sin)) {
1833		ifnet_t		*ifp_list;
1834		u_int32_t	count;
1835		u_int32_t	ifp_on;
1836
1837		result = ENOTSUP;
1838
1839		if (ifnet_list_get(IFNET_FAMILY_ANY, &ifp_list, &count) == 0) {
1840			for (ifp_on = 0; ifp_on < count; ifp_on++) {
1841				errno_t				new_result;
1842				ifaddr_t			source_hw = NULL;
1843				ifaddr_t			source_ip = NULL;
1844				struct sockaddr_in	source_ip_copy;
1845
1846				/*
1847				 * Only arp on interfaces marked for IPv4LL ARPing. This may
1848				 * mean that we don't ARP on the interface the subnet route
1849				 * points to.
1850				 */
1851				if ((ifp_list[ifp_on]->if_eflags & IFEF_ARPLL) == 0) {
1852					continue;
1853				}
1854
1855				source_hw = TAILQ_FIRST(&ifp_list[ifp_on]->if_addrhead);
1856
1857				/* Find the source IP address */
1858				ifnet_lock_shared(ifp_list[ifp_on]);
1859				TAILQ_FOREACH(source_ip, &ifp_list[ifp_on]->if_addrhead,
1860							  ifa_link) {
1861					if (source_ip->ifa_addr &&
1862						source_ip->ifa_addr->sa_family == AF_INET) {
1863						break;
1864					}
1865				}
1866
1867				/* No IP Source, don't arp */
1868				if (source_ip == NULL) {
1869					ifnet_lock_done(ifp_list[ifp_on]);
1870					continue;
1871				}
1872
1873				/* Copy the source IP address */
1874				source_ip_copy = *(struct sockaddr_in*)source_ip->ifa_addr;
1875
1876				ifnet_lock_done(ifp_list[ifp_on]);
1877
1878				/* Send the ARP */
1879				new_result = dlil_send_arp_internal(ifp_list[ifp_on], arpop,
1880									(struct sockaddr_dl*)source_hw->ifa_addr,
1881									(struct sockaddr*)&source_ip_copy, NULL,
1882									target_proto);
1883
1884				if (result == ENOTSUP) {
1885					result = new_result;
1886				}
1887			}
1888		}
1889
1890		ifnet_list_free(ifp_list);
1891	}
1892	else {
1893		result = dlil_send_arp_internal(ifp, arpop, sender_hw, sender_proto,
1894										target_hw, target_proto);
1895	}
1896
1897	return result;
1898}
1899
1900__private_extern__ int
1901ifp_use(
1902	struct ifnet *ifp,
1903	int	handle_zero)
1904{
1905	int old_value;
1906	int retval = 0;
1907
1908	do {
1909		old_value = ifp->if_usecnt;
1910		if (old_value == 0 && handle_zero == kIfNetUseCount_MustNotBeZero) {
1911			retval = ENXIO; // ifp is invalid
1912			break;
1913		}
1914	} while (!OSCompareAndSwap((UInt32)old_value, (UInt32)old_value + 1, (UInt32*)&ifp->if_usecnt));
1915
1916	return retval;
1917}
1918
1919/* ifp_unuse is broken into two pieces.
1920 *
1921 * ifp_use and ifp_unuse must be called between when the caller calls
1922 * dlil_write_begin and dlil_write_end. ifp_unuse needs to perform some
1923 * operations after dlil_write_end has been called. For this reason,
1924 * anyone calling ifp_unuse must call ifp_use_reached_zero if ifp_unuse
1925 * returns a non-zero value. The caller must call ifp_use_reached_zero
1926 * after the caller has called dlil_write_end.
1927 */
1928__private_extern__ void
1929ifp_use_reached_zero(
1930	struct ifnet *ifp)
1931{
1932	ifnet_detached_func	free_func;
1933
1934	dlil_read_begin();
1935
1936	if (ifp->if_usecnt != 0)
1937		panic("ifp_use_reached_zero: ifp->if_usecnt != 0");
1938
1939	ifnet_head_lock_exclusive();
1940	ifnet_lock_exclusive(ifp);
1941
1942	/* Remove ourselves from the list */
1943	TAILQ_REMOVE(&ifnet_head, ifp, if_link);
1944	ifnet_addrs[ifp->if_index - 1] = NULL;
1945
1946	/* ifp should be removed from the interface list */
1947	while (ifp->if_multiaddrs.lh_first) {
1948		struct ifmultiaddr *ifma = ifp->if_multiaddrs.lh_first;
1949
1950		/*
1951		 * When the interface is gone, we will no longer
1952		 * be listening on these multicasts. Various bits
1953		 * of the stack may be referencing these multicasts,
1954		 * release only our reference.
1955		 */
1956		LIST_REMOVE(ifma, ifma_link);
1957		ifma->ifma_ifp = NULL;
1958		ifma_release(ifma);
1959	}
1960	ifnet_head_done();
1961
1962	ifp->if_eflags &= ~IFEF_DETACHING; // clear the detaching flag
1963	ifnet_lock_done(ifp);
1964
1965	free_func = ifp->if_free;
1966	dlil_read_end();
1967	dlil_post_msg(ifp, KEV_DL_SUBCLASS, KEV_DL_IF_DETACHED, NULL, 0);
1968
1969	if (free_func)
1970		free_func(ifp);
1971}
1972
1973__private_extern__ int
1974ifp_unuse(
1975	struct ifnet *ifp)
1976{
1977	int	oldval;
1978	oldval = OSDecrementAtomic((SInt32*)&ifp->if_usecnt);
1979	if (oldval == 0)
1980		panic("ifp_unuse: ifp(%s%d)->if_usecnt was zero\n", ifp->if_name, ifp->if_unit);
1981
1982	if (oldval > 1)
1983		return 0;
1984
1985	if ((ifp->if_eflags & IFEF_DETACHING) == 0)
1986		panic("ifp_unuse: use count reached zero but detching flag is not set!");
1987
1988 	return 1; /* caller must call ifp_use_reached_zero */
1989}
1990
1991extern lck_mtx_t 	*domain_proto_mtx;
1992
1993static errno_t
1994dlil_attach_protocol_internal(
1995	struct if_proto	*proto,
1996	const struct ifnet_demux_desc *demux_list,
1997	u_int32_t	demux_count)
1998{
1999	struct kev_dl_proto_data	ev_pr_data;
2000	struct ifnet *ifp = proto->ifp;
2001	int retval = 0;
2002	u_long hash_value = proto_hash_value(proto->protocol_family);
2003
2004    /* setup some of the common values */
2005	{
2006		struct domain *dp;
2007		lck_mtx_lock(domain_proto_mtx);
2008		dp = domains;
2009		while (dp && (protocol_family_t)dp->dom_family != proto->protocol_family)
2010			dp = dp->dom_next;
2011		proto->dl_domain = dp;
2012		lck_mtx_unlock(domain_proto_mtx);
2013	}
2014
2015	/*
2016	 * Take the write lock to protect readers and exclude other writers.
2017	 */
2018	if ((retval = dlil_write_begin()) != 0) {
2019		printf("dlil_attach_protocol_internal - dlil_write_begin returned %d\n", retval);
2020		return retval;
2021	}
2022
2023	/* Check that the interface isn't currently detaching */
2024	ifnet_lock_shared(ifp);
2025	if ((ifp->if_eflags & IFEF_DETACHING) != 0) {
2026		ifnet_lock_done(ifp);
2027		dlil_write_end();
2028		return ENXIO;
2029	}
2030	ifnet_lock_done(ifp);
2031
2032	if (find_attached_proto(ifp, proto->protocol_family) != NULL) {
2033		dlil_write_end();
2034		return EEXIST;
2035	}
2036
2037	/*
2038	 * Call family module add_proto routine so it can refine the
2039	 * demux descriptors as it wishes.
2040	 */
2041	retval = ifp->if_add_proto(ifp, proto->protocol_family, demux_list, demux_count);
2042	if (retval) {
2043		dlil_write_end();
2044		return retval;
2045	}
2046
2047	/*
2048	 * We can't fail from this point on.
2049	 * Increment the number of uses (protocol attachments + interface attached).
2050	 */
2051	ifp_use(ifp, kIfNetUseCount_MustNotBeZero);
2052
2053	/*
2054	 * Insert the protocol in the hash
2055	 */
2056	{
2057		struct if_proto*	prev_proto = SLIST_FIRST(&ifp->if_proto_hash[hash_value]);
2058		while (prev_proto && SLIST_NEXT(prev_proto, next_hash) != NULL)
2059			prev_proto = SLIST_NEXT(prev_proto, next_hash);
2060		if (prev_proto)
2061			SLIST_INSERT_AFTER(prev_proto, proto, next_hash);
2062		else
2063			SLIST_INSERT_HEAD(&ifp->if_proto_hash[hash_value], proto, next_hash);
2064	}
2065
2066	/*
2067	 * Add to if_proto list for this interface
2068	 */
2069	if_proto_ref(proto);
2070	dlil_write_end();
2071
2072	/* the reserved field carries the number of protocol still attached (subject to change) */
2073	ev_pr_data.proto_family = proto->protocol_family;
2074	ev_pr_data.proto_remaining_count = dlil_ifp_proto_count(ifp);
2075	dlil_post_msg(ifp, KEV_DL_SUBCLASS, KEV_DL_PROTO_ATTACHED,
2076				  (struct net_event_data *)&ev_pr_data,
2077				  sizeof(struct kev_dl_proto_data));
2078#if 0
2079	DLIL_PRINTF("dlil. Attached protocol %d to %s%d - %d\n", proto->protocol_family,
2080			 ifp->if_name, ifp->if_unit, retval);
2081#endif
2082	return retval;
2083}
2084
2085errno_t
2086ifnet_attach_protocol(ifnet_t ifp, protocol_family_t protocol,
2087	const struct ifnet_attach_proto_param *proto_details)
2088{
2089	int retval = 0;
2090	struct if_proto  *ifproto = NULL;
2091
2092	if (ifp == NULL || protocol == 0 || proto_details == NULL)
2093		return EINVAL;
2094
2095	ifproto = _MALLOC(sizeof(struct if_proto), M_IFADDR, M_WAITOK);
2096	if (ifproto == 0) {
2097		DLIL_PRINTF("ERROR - dlil failed if_proto allocation\n");
2098		retval = ENOMEM;
2099		goto end;
2100	}
2101	bzero(ifproto, sizeof(*ifproto));
2102
2103	ifproto->ifp = ifp;
2104	ifproto->protocol_family = protocol;
2105	ifproto->proto_kpi = kProtoKPI_v1;
2106	ifproto->kpi.v1.input = proto_details->input;
2107	ifproto->kpi.v1.pre_output = proto_details->pre_output;
2108	ifproto->kpi.v1.event = proto_details->event;
2109	ifproto->kpi.v1.ioctl = proto_details->ioctl;
2110	ifproto->kpi.v1.detached = proto_details->detached;
2111	ifproto->kpi.v1.resolve_multi = proto_details->resolve;
2112	ifproto->kpi.v1.send_arp = proto_details->send_arp;
2113
2114	retval = dlil_attach_protocol_internal(ifproto,
2115				proto_details->demux_list, proto_details->demux_count);
2116
2117end:
2118	if (retval && ifproto)
2119		FREE(ifproto, M_IFADDR);
2120	return retval;
2121}
2122
2123errno_t
2124ifnet_attach_protocol_v2(ifnet_t ifp, protocol_family_t protocol,
2125	const struct ifnet_attach_proto_param_v2 *proto_details)
2126{
2127	int retval = 0;
2128	struct if_proto  *ifproto = NULL;
2129
2130	if (ifp == NULL || protocol == 0 || proto_details == NULL)
2131		return EINVAL;
2132
2133	ifproto = _MALLOC(sizeof(struct if_proto), M_IFADDR, M_WAITOK);
2134	if (ifproto == 0) {
2135		DLIL_PRINTF("ERROR - dlil failed if_proto allocation\n");
2136		retval = ENOMEM;
2137		goto end;
2138	}
2139	bzero(ifproto, sizeof(*ifproto));
2140
2141	ifproto->ifp = ifp;
2142	ifproto->protocol_family = protocol;
2143	ifproto->proto_kpi = kProtoKPI_v2;
2144	ifproto->kpi.v2.input = proto_details->input;
2145	ifproto->kpi.v2.pre_output = proto_details->pre_output;
2146	ifproto->kpi.v2.event = proto_details->event;
2147	ifproto->kpi.v2.ioctl = proto_details->ioctl;
2148	ifproto->kpi.v2.detached = proto_details->detached;
2149	ifproto->kpi.v2.resolve_multi = proto_details->resolve;
2150	ifproto->kpi.v2.send_arp = proto_details->send_arp;
2151
2152	retval = dlil_attach_protocol_internal(ifproto,
2153				proto_details->demux_list, proto_details->demux_count);
2154
2155end:
2156	if (retval && ifproto)
2157		FREE(ifproto, M_IFADDR);
2158	return retval;
2159}
2160
2161extern void if_rtproto_del(struct ifnet *ifp, int protocol);
2162
2163static int
2164dlil_detach_protocol_internal(
2165	struct if_proto *proto)
2166{
2167	struct ifnet *ifp = proto->ifp;
2168	u_long proto_family = proto->protocol_family;
2169	struct kev_dl_proto_data	ev_pr_data;
2170
2171	if (proto->proto_kpi == kProtoKPI_v1) {
2172		if (proto->kpi.v1.detached)
2173			proto->kpi.v1.detached(ifp, proto->protocol_family);
2174	}
2175	if (proto->proto_kpi == kProtoKPI_v2) {
2176		if (proto->kpi.v2.detached)
2177			proto->kpi.v2.detached(ifp, proto->protocol_family);
2178	}
2179	if_proto_free(proto);
2180
2181	/*
2182	 * Cleanup routes that may still be in the routing table for that interface/protocol pair.
2183	 */
2184
2185	if_rtproto_del(ifp, proto_family);
2186
2187	/* the reserved field carries the number of protocol still attached (subject to change) */
2188	ev_pr_data.proto_family   = proto_family;
2189	ev_pr_data.proto_remaining_count = dlil_ifp_proto_count(ifp);
2190	dlil_post_msg(ifp, KEV_DL_SUBCLASS, KEV_DL_PROTO_DETACHED,
2191				  (struct net_event_data *)&ev_pr_data,
2192				  sizeof(struct kev_dl_proto_data));
2193	return 0;
2194}
2195
2196errno_t
2197ifnet_detach_protocol(ifnet_t ifp, protocol_family_t proto_family)
2198{
2199	struct if_proto *proto = NULL;
2200	int	retval = 0;
2201	int use_reached_zero = 0;
2202
2203	if (ifp == NULL || proto_family == 0) return EINVAL;
2204
2205	if ((retval = dlil_write_begin()) != 0) {
2206		if (retval == EDEADLK) {
2207			retval = 0;
2208			dlil_read_begin();
2209			proto = find_attached_proto(ifp, proto_family);
2210			if (proto == 0) {
2211				retval = ENXIO;
2212			}
2213			else {
2214				proto->detaching = 1;
2215				dlil_detach_waiting = 1;
2216				wakeup(&dlil_detach_waiting);
2217			}
2218			dlil_read_end();
2219		}
2220		goto end;
2221	}
2222
2223	proto = find_attached_proto(ifp, proto_family);
2224
2225	if (proto == NULL) {
2226		retval = ENXIO;
2227		dlil_write_end();
2228		goto end;
2229	}
2230
2231	/*
2232	 * Call family module del_proto
2233	 */
2234
2235	if (ifp->if_del_proto)
2236		ifp->if_del_proto(ifp, proto->protocol_family);
2237
2238	SLIST_REMOVE(&ifp->if_proto_hash[proto_hash_value(proto_family)], proto, if_proto, next_hash);
2239
2240	/*
2241	 * We can do the rest of the work outside of the write lock.
2242	 */
2243	use_reached_zero = ifp_unuse(ifp);
2244	dlil_write_end();
2245
2246	dlil_detach_protocol_internal(proto);
2247
2248	/*
2249	 * Only handle the case where the interface will go away after
2250	 * we've sent the message. This way post message can send the
2251	 * message to the interface safely.
2252	 */
2253
2254	if (use_reached_zero)
2255		ifp_use_reached_zero(ifp);
2256
2257end:
2258    return retval;
2259}
2260
2261/*
2262 * dlil_delayed_detach_thread is responsible for detaching
2263 * protocols, protocol filters, and interface filters after
2264 * an attempt was made to detach one of those items while
2265 * it was not safe to do so (i.e. called dlil_read_begin).
2266 *
2267 * This function will take the dlil write lock and walk
2268 * through each of the interfaces looking for items with
2269 * the detaching flag set. When an item is found, it is
2270 * detached from the interface and placed on a local list.
2271 * After all of the items have been collected, we drop the
2272 * write lock and performed the post detach. This is done
2273 * so we only have to take the write lock once.
2274 *
2275 * When detaching a protocol filter, if we find that we
2276 * have detached the very last protocol and we need to call
2277 * ifp_use_reached_zero, we have to break out of our work
2278 * to drop the write lock so we can call ifp_use_reached_zero.
2279 */
2280
2281static void
2282dlil_delayed_detach_thread(__unused void* foo, __unused wait_result_t wait)
2283{
2284	thread_t self = current_thread();
2285	int asserted = 0;
2286
2287	ml_thread_policy(self, MACHINE_GROUP,
2288					 (MACHINE_NETWORK_GROUP|MACHINE_NETWORK_NETISR));
2289
2290
2291	while (1) {
2292		if (dlil_detach_waiting != 0 && dlil_write_begin() == 0) {
2293			struct ifnet *ifp;
2294			struct proto_hash_entry detached_protos;
2295			struct ifnet_filter_head detached_filters;
2296			struct if_proto	*proto;
2297			struct if_proto *next_proto;
2298			struct ifnet_filter *filt;
2299			struct ifnet_filter *next_filt;
2300			int reached_zero;
2301
2302			reached_zero = 0;
2303
2304			/* Clear the detach waiting flag */
2305			dlil_detach_waiting = 0;
2306			TAILQ_INIT(&detached_filters);
2307			SLIST_INIT(&detached_protos);
2308
2309			ifnet_head_lock_shared();
2310			TAILQ_FOREACH(ifp, &ifnet_head, if_link) {
2311				int i;
2312
2313				// Look for protocols and protocol filters
2314				for (i = 0; i < PROTO_HASH_SLOTS && !reached_zero; i++) {
2315					struct if_proto **prev_nextptr = &SLIST_FIRST(&ifp->if_proto_hash[i]);
2316					for (proto = *prev_nextptr; proto; proto = *prev_nextptr) {
2317
2318						// Detach this protocol
2319						if (proto->detaching) {
2320							if (ifp->if_del_proto)
2321								ifp->if_del_proto(ifp, proto->protocol_family);
2322							*prev_nextptr = SLIST_NEXT(proto, next_hash);
2323							SLIST_INSERT_HEAD(&detached_protos, proto, next_hash);
2324							reached_zero = ifp_unuse(ifp);
2325							if (reached_zero) {
2326								break;
2327							}
2328						}
2329						else {
2330							// Update prev_nextptr to point to our next ptr
2331							prev_nextptr = &SLIST_NEXT(proto, next_hash);
2332						}
2333					}
2334				}
2335
2336				// look for interface filters that need to be detached
2337				for (filt = TAILQ_FIRST(&ifp->if_flt_head); filt; filt = next_filt) {
2338					next_filt = TAILQ_NEXT(filt, filt_next);
2339					if (filt->filt_detaching != 0) {
2340						// take this interface filter off the interface filter list
2341						TAILQ_REMOVE(&ifp->if_flt_head, filt, filt_next);
2342
2343						// put this interface filter on the detached filters list
2344						TAILQ_INSERT_TAIL(&detached_filters, filt, filt_next);
2345					}
2346				}
2347
2348				if (ifp->if_delayed_detach) {
2349					ifp->if_delayed_detach = 0;
2350					reached_zero = ifp_unuse(ifp);
2351				}
2352
2353				if (reached_zero)
2354					break;
2355			}
2356			ifnet_head_done();
2357			dlil_write_end();
2358
2359			for (filt = TAILQ_FIRST(&detached_filters); filt; filt = next_filt) {
2360				next_filt = TAILQ_NEXT(filt, filt_next);
2361				/*
2362				 * dlil_detach_filter_internal won't remove an item from
2363				 * the list if it is already detached (second parameter).
2364				 * The item will be freed though.
2365				 */
2366				dlil_detach_filter_internal(filt, 1);
2367			}
2368
2369			for (proto = SLIST_FIRST(&detached_protos); proto; proto = next_proto) {
2370				next_proto = SLIST_NEXT(proto, next_hash);
2371				dlil_detach_protocol_internal(proto);
2372			}
2373
2374			if (reached_zero) {
2375				ifp_use_reached_zero(ifp);
2376				dlil_detach_waiting = 1; // we may have missed something
2377			}
2378		}
2379
2380		if (!asserted && dlil_detach_waiting == 0) {
2381			asserted = 1;
2382			assert_wait(&dlil_detach_waiting, THREAD_UNINT);
2383		}
2384
2385		if (dlil_detach_waiting == 0) {
2386			asserted = 0;
2387			thread_block(dlil_delayed_detach_thread);
2388		}
2389	}
2390}
2391
2392static void
2393dlil_call_delayed_detach_thread(void) {
2394	dlil_delayed_detach_thread(NULL, THREAD_RESTART);
2395}
2396
2397extern int if_next_index(void);
2398
2399errno_t
2400ifnet_attach(
2401	ifnet_t						ifp,
2402	const struct sockaddr_dl	*ll_addr)
2403{
2404	u_long		    interface_family;
2405	struct ifnet *tmp_if;
2406	struct proto_hash_entry *new_proto_list = NULL;
2407	int locked = 0;
2408
2409	if (ifp == NULL) return EINVAL;
2410	if (ll_addr && ifp->if_addrlen == 0) {
2411		ifp->if_addrlen = ll_addr->sdl_alen;
2412	}
2413	else if (ll_addr && ll_addr->sdl_alen != ifp->if_addrlen) {
2414		return EINVAL;
2415	}
2416
2417	interface_family = ifp->if_family;
2418
2419	ifnet_head_lock_shared();
2420
2421	/* Verify we aren't already on the list */
2422	TAILQ_FOREACH(tmp_if, &ifnet_head, if_link) {
2423		if (tmp_if == ifp) {
2424			ifnet_head_done();
2425			return EEXIST;
2426		}
2427	}
2428
2429	ifnet_head_done();
2430
2431	if ((ifp->if_eflags & IFEF_REUSE) == 0 || ifp->if_lock == 0)
2432#if IFNET_RW_LOCK
2433		ifp->if_lock = lck_rw_alloc_init(ifnet_lock_group, ifnet_lock_attr);
2434#else
2435		ifp->if_lock = lck_mtx_alloc_init(ifnet_lock_group, ifnet_lock_attr);
2436#endif
2437
2438	if (ifp->if_lock == 0) {
2439		return ENOMEM;
2440	}
2441
2442	/*
2443	 * Allow interfaces withouth protocol families to attach
2444	 * only if they have the necessary fields filled out.
2445	 */
2446
2447	if (ifp->if_add_proto == 0 || ifp->if_del_proto == 0) {
2448		DLIL_PRINTF("dlil Attempt to attach interface without family module - %ld\n",
2449				interface_family);
2450		return ENODEV;
2451	}
2452
2453	if ((ifp->if_eflags & IFEF_REUSE) == 0 || ifp->if_proto_hash == NULL) {
2454		MALLOC(new_proto_list, struct proto_hash_entry*, sizeof(struct proto_hash_entry) * PROTO_HASH_SLOTS,
2455			   M_NKE, M_WAITOK);
2456
2457		if (new_proto_list == 0) {
2458			return ENOBUFS;
2459		}
2460	}
2461
2462	dlil_write_begin();
2463	locked = 1;
2464
2465	TAILQ_INIT(&ifp->if_flt_head);
2466
2467
2468	if (new_proto_list) {
2469		bzero(new_proto_list, (PROTO_HASH_SLOTS * sizeof(struct proto_hash_entry)));
2470		ifp->if_proto_hash = new_proto_list;
2471		new_proto_list = NULL;
2472	}
2473
2474	/* old_if_attach */
2475	{
2476		char workbuf[64];
2477		int namelen, masklen, socksize, ifasize;
2478		struct ifaddr *ifa = NULL;
2479
2480		if (ifp->if_snd.ifq_maxlen == 0)
2481			ifp->if_snd.ifq_maxlen = ifqmaxlen;
2482		TAILQ_INIT(&ifp->if_prefixhead);
2483		LIST_INIT(&ifp->if_multiaddrs);
2484		ifnet_touch_lastchange(ifp);
2485
2486		/* usecount to track attachment to the ifnet list */
2487		ifp_use(ifp, kIfNetUseCount_MayBeZero);
2488
2489		/* Lock the list of interfaces */
2490		ifnet_head_lock_exclusive();
2491		ifnet_lock_exclusive(ifp);
2492
2493		if ((ifp->if_eflags & IFEF_REUSE) == 0 || ifp->if_index == 0)
2494			ifp->if_index = if_next_index();
2495		else
2496			ifa = TAILQ_FIRST(&ifp->if_addrhead);
2497
2498		namelen = snprintf(workbuf, sizeof(workbuf), "%s%d", ifp->if_name, ifp->if_unit);
2499#define _offsetof(t, m) ((int)((caddr_t)&((t *)0)->m))
2500		masklen = _offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
2501		socksize = masklen + ifp->if_addrlen;
2502#define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
2503		if ((u_long)socksize < sizeof(struct sockaddr_dl))
2504			socksize = sizeof(struct sockaddr_dl);
2505		socksize = ROUNDUP(socksize);
2506		ifasize = sizeof(struct ifaddr) + 2 * socksize;
2507
2508		/*
2509		 * Allocate a new ifa if we don't have one
2510		 * or the old one is too small.
2511		 */
2512		if (ifa == NULL || socksize > ifa->ifa_addr->sa_len) {
2513			if (ifa)
2514				if_detach_ifa(ifp, ifa);
2515			ifa = (struct ifaddr*)_MALLOC(ifasize, M_IFADDR, M_WAITOK);
2516		}
2517
2518		if (ifa) {
2519			struct sockaddr_dl *sdl = (struct sockaddr_dl *)(ifa + 1);
2520			ifnet_addrs[ifp->if_index - 1] = ifa;
2521			bzero(ifa, ifasize);
2522			sdl->sdl_len = socksize;
2523			sdl->sdl_family = AF_LINK;
2524			bcopy(workbuf, sdl->sdl_data, namelen);
2525			sdl->sdl_nlen = namelen;
2526			sdl->sdl_index = ifp->if_index;
2527			sdl->sdl_type = ifp->if_type;
2528			if (ll_addr) {
2529				sdl->sdl_alen = ll_addr->sdl_alen;
2530				if (ll_addr->sdl_alen != ifp->if_addrlen)
2531					panic("ifnet_attach - ll_addr->sdl_alen != ifp->if_addrlen");
2532				bcopy(CONST_LLADDR(ll_addr), LLADDR(sdl), sdl->sdl_alen);
2533			}
2534			ifa->ifa_ifp = ifp;
2535			ifa->ifa_rtrequest = link_rtrequest;
2536			ifa->ifa_addr = (struct sockaddr*)sdl;
2537			sdl = (struct sockaddr_dl*)(socksize + (caddr_t)sdl);
2538			ifa->ifa_netmask = (struct sockaddr*)sdl;
2539			sdl->sdl_len = masklen;
2540			while (namelen != 0)
2541				sdl->sdl_data[--namelen] = 0xff;
2542		}
2543
2544		TAILQ_INIT(&ifp->if_addrhead);
2545		ifa = ifnet_addrs[ifp->if_index - 1];
2546
2547		if (ifa) {
2548			/*
2549			 * We don't use if_attach_ifa because we want
2550			 * this address to be first on the list.
2551			 */
2552			ifaref(ifa);
2553			ifa->ifa_debug |= IFA_ATTACHED;
2554			TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
2555		}
2556#if CONFIG_MACF_NET
2557		mac_ifnet_label_associate(ifp);
2558#endif
2559
2560		TAILQ_INSERT_TAIL(&ifnet_head, ifp, if_link);
2561		ifindex2ifnet[ifp->if_index] = ifp;
2562
2563		ifnet_head_done();
2564	}
2565
2566	/*
2567	 * A specific dlil input thread is created per Ethernet interface.
2568	 * pseudo interfaces or other types of interfaces use the main ("loopback") thread.
2569	 * If the sysctl "net.link.generic.system.multi_threaded_input" is set to zero, all packets will
2570	 * be handled by the main loopback thread, reverting to 10.4.x behaviour.
2571	 *
2572	 */
2573
2574	if (ifp->if_type == IFT_ETHER) {
2575		int err;
2576
2577		if (dlil_multithreaded_input > 0) {
2578			ifp->if_input_thread = _MALLOC(sizeof(struct dlil_threading_info), M_NKE, M_WAITOK);
2579			if (ifp->if_input_thread == NULL)
2580				panic("ifnet_attach ifp=%p couldn't alloc threading\n", ifp);
2581			if ((err = dlil_create_input_thread(ifp, ifp->if_input_thread)) != 0)
2582				panic("ifnet_attach ifp=%p couldn't get a thread. err=%x\n", ifp, err);
2583#ifdef DLIL_DEBUG
2584			printf("ifnet_attach: dlil thread for ifp=%p if_index=%x\n", ifp, ifp->if_index);
2585#endif
2586		}
2587	}
2588    	dlil_write_end();
2589	ifnet_lock_done(ifp);
2590
2591	dlil_post_msg(ifp, KEV_DL_SUBCLASS, KEV_DL_IF_ATTACHED, NULL, 0);
2592
2593    return 0;
2594}
2595
2596errno_t
2597ifnet_detach(
2598	ifnet_t	ifp)
2599{
2600	struct ifnet_filter *filter;
2601	struct ifnet_filter	*filter_next;
2602	int zeroed = 0;
2603	int retval = 0;
2604	struct ifnet_filter_head fhead;
2605	struct dlil_threading_info *inputthread;
2606
2607	if (ifp == NULL) return EINVAL;
2608
2609	ifnet_lock_exclusive(ifp);
2610
2611	if ((ifp->if_eflags & IFEF_DETACHING) != 0) {
2612		/* Interface has already been detached */
2613		ifnet_lock_done(ifp);
2614		return ENXIO;
2615	}
2616
2617	/*
2618	 * Indicate this interface is being detached.
2619	 *
2620	 * This should prevent protocols from attaching
2621	 * from this point on. Interface will remain on
2622	 * the list until all of the protocols are detached.
2623	 */
2624	ifp->if_eflags |= IFEF_DETACHING;
2625	ifnet_lock_done(ifp);
2626
2627	dlil_post_msg(ifp, KEV_DL_SUBCLASS, KEV_DL_IF_DETACHING, NULL, 0);
2628
2629	/* Let BPF know we're detaching */
2630	bpfdetach(ifp);
2631
2632	if ((retval = dlil_write_begin()) != 0) {
2633		if (retval == EDEADLK) {
2634			retval = 0;
2635
2636			/* We need to perform a delayed detach */
2637			ifp->if_delayed_detach = 1;
2638			dlil_detach_waiting = 1;
2639			wakeup(&dlil_detach_waiting);
2640		}
2641		return retval;
2642	}
2643
2644	/* Steal the list of interface filters */
2645	fhead = ifp->if_flt_head;
2646	TAILQ_INIT(&ifp->if_flt_head);
2647
2648	/* unuse the interface */
2649	zeroed = ifp_unuse(ifp);
2650
2651	/*
2652	 * If thread affinity was set for the workloop thread, we will need
2653	 * to tear down the affinity and release the extra reference count
2654	 * taken at attach time;
2655	 */
2656	if ((inputthread = ifp->if_input_thread) != NULL) {
2657		if (inputthread->net_affinity) {
2658			struct thread *tp;
2659
2660			if (inputthread == dlil_lo_thread_ptr)
2661				panic("Thread affinity should not be enabled "
2662				    "on the loopback dlil input thread\n");
2663
2664			lck_mtx_lock(inputthread->input_lck);
2665			tp = inputthread->workloop_thread;
2666			inputthread->workloop_thread = NULL;
2667			inputthread->tag = 0;
2668			inputthread->net_affinity = FALSE;
2669			lck_mtx_unlock(inputthread->input_lck);
2670
2671			/* Tear down workloop thread affinity */
2672			if (tp != NULL) {
2673				(void) dlil_affinity_set(tp,
2674				    THREAD_AFFINITY_TAG_NULL);
2675				thread_deallocate(tp);
2676			}
2677
2678			/* Tear down dlil input thread affinity */
2679			tp = inputthread->input_thread;
2680			(void) dlil_affinity_set(tp, THREAD_AFFINITY_TAG_NULL);
2681			thread_deallocate(tp);
2682		}
2683
2684		/* cleanup ifp dlil input thread, if any */
2685		ifp->if_input_thread = NULL;
2686
2687		if (inputthread != dlil_lo_thread_ptr) {
2688#ifdef DLIL_DEBUG
2689			printf("ifnet_detach: wakeup thread threadinfo: %p "
2690			    "input_thread=%p threads: cur=%d max=%d\n",
2691			    inputthread, inputthread->input_thread,
2692			    dlil_multithreaded_input, cur_dlil_input_threads);
2693#endif
2694			lck_mtx_lock(inputthread->input_lck);
2695
2696			inputthread->input_waiting |= DLIL_INPUT_TERMINATE;
2697			if ((inputthread->input_waiting & DLIL_INPUT_RUNNING) == 0) {
2698				wakeup((caddr_t)&inputthread->input_waiting);
2699			}
2700			lck_mtx_unlock(inputthread->input_lck);
2701		}
2702	}
2703	dlil_write_end();
2704
2705	for (filter = TAILQ_FIRST(&fhead); filter; filter = filter_next) {
2706		filter_next = TAILQ_NEXT(filter, filt_next);
2707		dlil_detach_filter_internal(filter, 1);
2708	}
2709
2710	if (zeroed != 0) {
2711		ifp_use_reached_zero(ifp);
2712	}
2713
2714	return retval;
2715}
2716
2717static errno_t
2718dlil_recycle_ioctl(
2719	__unused ifnet_t ifnet_ptr,
2720	__unused u_int32_t ioctl_code,
2721	__unused void *ioctl_arg)
2722{
2723    return EOPNOTSUPP;
2724}
2725
2726static int
2727dlil_recycle_output(
2728	__unused struct ifnet *ifnet_ptr,
2729	struct mbuf *m)
2730{
2731    m_freem(m);
2732    return 0;
2733}
2734
2735static void
2736dlil_recycle_free(
2737	__unused ifnet_t ifnet_ptr)
2738{
2739}
2740
2741static errno_t
2742dlil_recycle_set_bpf_tap(
2743	__unused ifnet_t ifp,
2744	__unused bpf_tap_mode mode,
2745	__unused bpf_packet_func callback)
2746{
2747    /* XXX not sure what to do here */
2748    return 0;
2749}
2750
2751__private_extern__
2752int dlil_if_acquire(
2753	u_long family,
2754	const void *uniqueid,
2755	size_t uniqueid_len,
2756	struct ifnet **ifp)
2757{
2758    struct ifnet	*ifp1 = NULL;
2759    struct dlil_ifnet	*dlifp1 = NULL;
2760    int	ret = 0;
2761
2762    lck_mtx_lock(dlil_ifnet_mutex);
2763    TAILQ_FOREACH(dlifp1, &dlil_ifnet_head, dl_if_link) {
2764
2765        ifp1 = (struct ifnet *)dlifp1;
2766
2767		if (ifp1->if_family == family)  {
2768
2769            /* same uniqueid and same len or no unique id specified */
2770            if ((uniqueid_len == dlifp1->if_uniqueid_len)
2771                && !bcmp(uniqueid, dlifp1->if_uniqueid, uniqueid_len)) {
2772
2773				/* check for matching interface in use */
2774				if (ifp1->if_eflags & IFEF_INUSE) {
2775					if (uniqueid_len) {
2776						ret = EBUSY;
2777						goto end;
2778					}
2779				}
2780				else {
2781					if (!ifp1->if_lock)
2782						panic("ifp's lock is gone\n");
2783					ifnet_lock_exclusive(ifp1);
2784					ifp1->if_eflags |= (IFEF_INUSE | IFEF_REUSE);
2785					ifnet_lock_done(ifp1);
2786					*ifp = ifp1;
2787					goto end;
2788            	}
2789            }
2790        }
2791    }
2792
2793    /* no interface found, allocate a new one */
2794    MALLOC(dlifp1, struct dlil_ifnet *, sizeof(*dlifp1), M_NKE, M_WAITOK);
2795    if (dlifp1 == 0) {
2796        ret = ENOMEM;
2797        goto end;
2798    }
2799
2800    bzero(dlifp1, sizeof(*dlifp1));
2801
2802    if (uniqueid_len) {
2803        MALLOC(dlifp1->if_uniqueid, void *, uniqueid_len, M_NKE, M_WAITOK);
2804        if (dlifp1->if_uniqueid == 0) {
2805            FREE(dlifp1, M_NKE);
2806            ret = ENOMEM;
2807           goto end;
2808        }
2809        bcopy(uniqueid, dlifp1->if_uniqueid, uniqueid_len);
2810        dlifp1->if_uniqueid_len = uniqueid_len;
2811    }
2812
2813    ifp1 = (struct ifnet *)dlifp1;
2814    ifp1->if_eflags |= IFEF_INUSE;
2815    ifp1->if_name = dlifp1->if_namestorage;
2816#if CONFIG_MACF_NET
2817    mac_ifnet_label_init(ifp1);
2818#endif
2819
2820    TAILQ_INSERT_TAIL(&dlil_ifnet_head, dlifp1, dl_if_link);
2821
2822     *ifp = ifp1;
2823
2824end:
2825	lck_mtx_unlock(dlil_ifnet_mutex);
2826
2827    return ret;
2828}
2829
2830__private_extern__ void
2831dlil_if_release(
2832	ifnet_t	ifp)
2833{
2834    struct dlil_ifnet	*dlifp = (struct dlil_ifnet *)ifp;
2835
2836    /* Interface does not have a lock until it is attached - radar 3713951 */
2837    if (ifp->if_lock)
2838		ifnet_lock_exclusive(ifp);
2839    ifp->if_eflags &= ~IFEF_INUSE;
2840    ifp->if_ioctl = dlil_recycle_ioctl;
2841    ifp->if_output = dlil_recycle_output;
2842    ifp->if_free = dlil_recycle_free;
2843    ifp->if_set_bpf_tap = dlil_recycle_set_bpf_tap;
2844
2845    strncpy(dlifp->if_namestorage, ifp->if_name, IFNAMSIZ);
2846    ifp->if_name = dlifp->if_namestorage;
2847#if CONFIG_MACF_NET
2848    /*
2849     * We can either recycle the MAC label here or in dlil_if_acquire().
2850     * It seems logical to do it here but this means that anything that
2851     * still has a handle on ifp will now see it as unlabeled.
2852     * Since the interface is "dead" that may be OK.  Revisit later.
2853     */
2854    mac_ifnet_label_recycle(ifp);
2855#endif
2856    if (ifp->if_lock)
2857		ifnet_lock_done(ifp);
2858
2859}
2860
2861__private_extern__ void
2862dlil_proto_unplumb_all(struct ifnet *ifp)
2863{
2864	/*
2865	 * if_proto_hash[0-3] are for PF_INET, PF_INET6, PF_APPLETALK
2866	 * and PF_VLAN, where each bucket contains exactly one entry;
2867	 * PF_VLAN does not need an explicit unplumb.
2868	 *
2869	 * if_proto_hash[4] is for other protocols; we expect anything
2870	 * in this bucket to respond to the DETACHING event (which would
2871	 * have happened by now) and do the unplumb then.
2872	 */
2873	(void) proto_unplumb(PF_INET, ifp);
2874#if INET6
2875	(void) proto_unplumb(PF_INET6, ifp);
2876#endif /* INET6 */
2877#if NETAT
2878	(void) proto_unplumb(PF_APPLETALK, ifp);
2879#endif /* NETAT */
2880}
2881