in_pcb.c source code [xnu/bsd/netinet/in_pcb.c]

1	/*
2	* Copyright (c) 2000-2021 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) 1982, 1986, 1991, 1993, 1995
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	* @(#)in_pcb.c 8.4 (Berkeley) 5/24/95
61	* $FreeBSD: src/sys/netinet/in_pcb.c,v 1.59.2.17 2001/08/13 16:26:17 ume Exp $
62	*/
63
64	#include <sys/param.h>
65	#include <sys/systm.h>
66	#include <sys/malloc.h>
67	#include <sys/mbuf.h>
68	#include <sys/domain.h>
69	#include <sys/protosw.h>
70	#include <sys/socket.h>
71	#include <sys/socketvar.h>
72	#include <sys/proc.h>
73	#include <sys/kernel.h>
74	#include <sys/sysctl.h>
75	#include <sys/mcache.h>
76	#include <sys/kauth.h>
77	#include <sys/priv.h>
78	#include <sys/proc_uuid_policy.h>
79	#include <sys/syslog.h>
80	#include <sys/priv.h>
81	#include <sys/file_internal.h>
82	#include <net/dlil.h>
83
84	#include <libkern/OSAtomic.h>
85	#include <kern/locks.h>
86
87	#include <machine/limits.h>
88
89	#include <kern/zalloc.h>
90
91	#include <net/if.h>
92	#include <net/if_types.h>
93	#include <net/route.h>
94	#include <net/flowhash.h>
95	#include <net/flowadv.h>
96	#include <net/nat464_utils.h>
97	#include <net/ntstat.h>
98	#include <net/nwk_wq.h>
99	#include <net/restricted_in_port.h>
100
101	#include <netinet/in.h>
102	#include <netinet/in_pcb.h>
103	#include <netinet/inp_log.h>
104	#include <netinet/in_var.h>
105	#include <netinet/ip_var.h>
106
107	#include <netinet/ip6.h>
108	#include <netinet6/ip6_var.h>
109
110	#include <sys/kdebug.h>
111	#include <sys/random.h>
112
113	#include <dev/random/randomdev.h>
114	#include <mach/boolean.h>
115
116	#include <atm/atm_internal.h>
117	#include <pexpert/pexpert.h>
118
119	#if NECP
120	#include <net/necp.h>
121	#endif
122
123	#include <sys/stat.h>
124	#include <sys/ubc.h>
125	#include <sys/vnode.h>
126
127	#include <os/log.h>
128
129	#if SKYWALK
130	#include <skywalk/namespace/flowidns.h>
131	#endif /* SKYWALK */
132
133	#include <IOKit/IOBSD.h>
134
135	#include <net/sockaddr_utils.h>
136
137	extern const char proc_name_address(struct* proc *);
138
139	static LCK_GRP_DECLARE(inpcb_lock_grp, "inpcb");
140	static LCK_ATTR_DECLARE(inpcb_lock_attr, `0`, `0`);
141	static LCK_MTX_DECLARE_ATTR(inpcb_lock, &inpcb_lock_grp, &inpcb_lock_attr);
142	static LCK_MTX_DECLARE_ATTR(inpcb_timeout_lock, &inpcb_lock_grp, &inpcb_lock_attr);
143
144	static TAILQ_HEAD(, inpcbinfo) inpcb_head = TAILQ_HEAD_INITIALIZER(inpcb_head);
145
146	static u_int16_t inpcb_timeout_run = `0`; / INPCB timer is scheduled to run /
147	static boolean_t inpcb_garbage_collecting = FALSE; / gc timer is scheduled /
148	static boolean_t inpcb_ticking = FALSE; / "slow" timer is scheduled /
149	static boolean_t inpcb_fast_timer_on = FALSE;
150
151	#define INPCB_GCREQ_THRESHOLD 50000
152
153	static thread_call_t inpcb_thread_call, inpcb_fast_thread_call;
154	static void inpcb_sched_timeout(void);
155	static void inpcb_sched_lazy_timeout(void);
156	static void _inpcb_sched_timeout(unsigned int);
157	static void inpcb_timeout(void , void* *);
158	const int inpcb_timeout_lazy = `10`; / 10 seconds leeway for lazy timers /
159	extern int tvtohz(struct timeval *);
160
161	#if CONFIG_PROC_UUID_POLICY
162	static void inp_update_cellular_policy(struct inpcb *, boolean_t);
163	#if NECP
164	static void inp_update_necp_want_app_policy(struct inpcb *, boolean_t);
165	#endif /* NECP */
166	#endif /* !CONFIG_PROC_UUID_POLICY */
167
168	#define DBG_FNC_PCB_LOOKUP NETDBG_CODE(DBG_NETTCP, (6 << 8))
169	#define DBG_FNC_PCB_HLOOKUP NETDBG_CODE(DBG_NETTCP, ((6 << 8) \| 1))
170
171	int allow_udp_port_exhaustion = `0`;
172
173	/*
174	* These configure the range of local port addresses assigned to
175	* "unspecified" outgoing connections/packets/whatever.
176	*/
177	int ipport_lowfirstauto = IPPORT_RESERVED - `1`; / 1023 /
178	int ipport_lowlastauto = IPPORT_RESERVEDSTART; / 600 /
179	int ipport_firstauto = IPPORT_HIFIRSTAUTO; / 49152 /
180	int ipport_lastauto = IPPORT_HILASTAUTO; / 65535 /
181	int ipport_hifirstauto = IPPORT_HIFIRSTAUTO; / 49152 /
182	int ipport_hilastauto = IPPORT_HILASTAUTO; / 65535 /
183
184	#define RANGECHK(var, min, max) \
185	if ((var) < (min)) { (var) = (min); } \
186	else if ((var) > (max)) { (var) = (max); }
187
188	static int
189	sysctl_net_ipport_check SYSCTL_HANDLER_ARGS
190	{
191	#pragma unused(arg1, arg2)
192	int error;
193	int new_value = (int* *)oidp->oid_arg1;
194	#if (DEBUG \| DEVELOPMENT)
195	int old_value = (int* *)oidp->oid_arg1;
196	/*
197	* For unit testing allow a non-superuser process with the
198	* proper entitlement to modify the variables
199	*/
200	if (req->newptr) {
201	if (proc_suser(current_proc()) != `0` &&
202	(error = priv_check_cred(kauth_cred_get(),
203	PRIV_NETINET_RESERVEDPORT, `0`))) {
204	return EPERM;
205	}
206	}
207	#endif /* (DEBUG \| DEVELOPMENT) */
208
209	error = sysctl_handle_int(oidp, arg1: &new_value, arg2: `0`, req);
210	if (!error) {
211	if (oidp->oid_arg1 == &ipport_lowfirstauto \|\| oidp->oid_arg1 == &ipport_lowlastauto) {
212	RANGECHK(new_value, `1`, IPPORT_RESERVED - `1`);
213	} else {
214	RANGECHK(new_value, IPPORT_RESERVED, USHRT_MAX);
215	}
216	(int* *)oidp->oid_arg1 = new_value;
217	}
218
219	#if (DEBUG \| DEVELOPMENT)
220	os_log(OS_LOG_DEFAULT,
221	"%s:%u sysctl net.restricted_port.verbose: %d -> %d)",
222	proc_best_name(current_proc()), proc_selfpid(),
223	old_value, (int* *)oidp->oid_arg1);
224	#endif /* (DEBUG \| DEVELOPMENT) */
225
226	return error;
227	}
228
229	#undef RANGECHK
230
231	SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange,
232	CTLFLAG_RW \| CTLFLAG_LOCKED, `0`, "IP Ports");
233
234	#if (DEBUG \| DEVELOPMENT)
235	#define CTLFAGS_IP_PORTRANGE (CTLTYPE_INT \| CTLFLAG_RW \| CTLFLAG_LOCKED \| CTLFLAG_ANYBODY)
236	#else
237	#define CTLFAGS_IP_PORTRANGE (CTLTYPE_INT \| CTLFLAG_RW \| CTLFLAG_LOCKED)
238	#endif /* (DEBUG \| DEVELOPMENT) */
239
240	SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst,
241	CTLFAGS_IP_PORTRANGE,
242	&ipport_lowfirstauto, `0`, &sysctl_net_ipport_check, "I", "");
243	SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast,
244	CTLFAGS_IP_PORTRANGE,
245	&ipport_lowlastauto, `0`, &sysctl_net_ipport_check, "I", "");
246	SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first,
247	CTLFAGS_IP_PORTRANGE,
248	&ipport_firstauto, `0`, &sysctl_net_ipport_check, "I", "");
249	SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last,
250	CTLFAGS_IP_PORTRANGE,
251	&ipport_lastauto, `0`, &sysctl_net_ipport_check, "I", "");
252	SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst,
253	CTLFAGS_IP_PORTRANGE,
254	&ipport_hifirstauto, `0`, &sysctl_net_ipport_check, "I", "");
255	SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast,
256	CTLFAGS_IP_PORTRANGE,
257	&ipport_hilastauto, `0`, &sysctl_net_ipport_check, "I", "");
258	SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, ipport_allow_udp_port_exhaustion,
259	CTLFLAG_LOCKED \| CTLFLAG_RW, &allow_udp_port_exhaustion, `0`, "");
260
261	static uint32_t apn_fallbk_debug = `0`;
262	#define apn_fallbk_log(x) do { if (apn_fallbk_debug >= 1) log x; } while (0)
263
264	#if !XNU_TARGET_OS_OSX
265	static boolean_t apn_fallbk_enabled = TRUE;
266
267	SYSCTL_DECL(_net_inet);
268	SYSCTL_NODE(_net_inet, OID_AUTO, apn_fallback, CTLFLAG_RW \| CTLFLAG_LOCKED, `0`, "APN Fallback");
269	SYSCTL_UINT(_net_inet_apn_fallback, OID_AUTO, enable, CTLFLAG_RW \| CTLFLAG_LOCKED,
270	&apn_fallbk_enabled, `0`, "APN fallback enable");
271	SYSCTL_UINT(_net_inet_apn_fallback, OID_AUTO, debug, CTLFLAG_RW \| CTLFLAG_LOCKED,
272	&apn_fallbk_debug, `0`, "APN fallback debug enable");
273	#else /* XNU_TARGET_OS_OSX */
274	static boolean_t apn_fallbk_enabled = FALSE;
275	#endif /* XNU_TARGET_OS_OSX */
276
277	extern int udp_use_randomport;
278	extern int tcp_use_randomport;
279
280	/ Structs used for flowhash computation /
281	struct inp_flowhash_key_addr {
282	union {
283	struct in_addr v4;
284	struct in6_addr v6;
285	u_int8_t addr8[`16`];
286	u_int16_t addr16[`8`];
287	u_int32_t addr32[`4`];
288	} infha;
289	};
290
291	struct inp_flowhash_key {
292	struct inp_flowhash_key_addr infh_laddr;
293	struct inp_flowhash_key_addr infh_faddr;
294	u_int32_t infh_lport;
295	u_int32_t infh_fport;
296	u_int32_t infh_af;
297	u_int32_t infh_proto;
298	u_int32_t infh_rand1;
299	u_int32_t infh_rand2;
300	};
301
302	#if !SKYWALK
303	static u_int32_t inp_hash_seed = `0`;
304	#endif /* !SKYWALK */
305
306	static int infc_cmp(const struct inpcb , const* struct inpcb *);
307
308	/ Flags used by inp_fc_getinp /
309	#define INPFC_SOLOCKED 0x1
310	#define INPFC_REMOVE 0x2
311	static struct inpcb *inp_fc_getinp(u_int32_t, u_int32_t);
312
313	static void inp_fc_feedback(struct inpcb *);
314	extern void tcp_remove_from_time_wait(struct inpcb *inp);
315
316	static LCK_MTX_DECLARE_ATTR(inp_fc_lck, &inpcb_lock_grp, &inpcb_lock_attr);
317
318	RB_HEAD(inp_fc_tree, inpcb) inp_fc_tree;
319	RB_PROTOTYPE(inp_fc_tree, inpcb, infc_link, infc_cmp);
320	RB_GENERATE(inp_fc_tree, inpcb, infc_link, infc_cmp);
321
322	/*
323	* Use this inp as a key to find an inp in the flowhash tree.
324	* Accesses to it are protected by inp_fc_lck.
325	*/
326	struct inpcb key_inp;
327
328	/*
329	* in_pcb.c: manage the Protocol Control Blocks.
330	*/
331
332	void
333	in_pcbinit(void)
334	{
335	static int inpcb_initialized = `0`;
336	uint32_t logging_config;
337
338	VERIFY(!inpcb_initialized);
339	inpcb_initialized = `1`;
340
341	logging_config = atm_get_diagnostic_config();
342	if (logging_config & `0x80000000`) {
343	inp_log_privacy = `1`;
344	}
345
346	inpcb_thread_call = thread_call_allocate_with_priority(func: inpcb_timeout,
347	NULL, pri: THREAD_CALL_PRIORITY_KERNEL);
348	/ Give it an arg so that we know that this is the fast timer /
349	inpcb_fast_thread_call = thread_call_allocate_with_priority(
350	func: inpcb_timeout, param0: &inpcb_timeout, pri: THREAD_CALL_PRIORITY_KERNEL);
351	if (inpcb_thread_call == NULL \|\| inpcb_fast_thread_call == NULL) {
352	panic("unable to alloc the inpcb thread call");
353	}
354
355	/*
356	* Initialize data structures required to deliver
357	* flow advisories.
358	*/
359	lck_mtx_lock(lck: &inp_fc_lck);
360	RB_INIT(&inp_fc_tree);
361	bzero(s: &key_inp, n: sizeof(key_inp));
362	lck_mtx_unlock(lck: &inp_fc_lck);
363	}
364
365	#define INPCB_HAVE_TIMER_REQ(req) (((req).intimer_lazy > 0) \|\| \
366	((req).intimer_fast > 0) \|\| ((req).intimer_nodelay > 0))
367	static void
368	inpcb_timeout(void arg0, void* *arg1)
369	{
370	#pragma unused(arg1)
371	struct inpcbinfo *ipi;
372	boolean_t t, gc;
373	struct intimercount gccnt, tmcnt;
374
375	/*
376	* Update coarse-grained networking timestamp (in sec.); the idea
377	* is to piggy-back on the timeout callout to update the counter
378	* returnable via net_uptime().
379	*/
380	net_update_uptime();
381
382	bzero(s: &gccnt, n: sizeof(gccnt));
383	bzero(s: &tmcnt, n: sizeof(tmcnt));
384
385	lck_mtx_lock_spin(lck: &inpcb_timeout_lock);
386	gc = inpcb_garbage_collecting;
387	inpcb_garbage_collecting = FALSE;
388
389	t = inpcb_ticking;
390	inpcb_ticking = FALSE;
391
392	if (gc \|\| t) {
393	lck_mtx_unlock(lck: &inpcb_timeout_lock);
394
395	lck_mtx_lock(lck: &inpcb_lock);
396	TAILQ_FOREACH(ipi, &inpcb_head, ipi_entry) {
397	if (INPCB_HAVE_TIMER_REQ(ipi->ipi_gc_req)) {
398	bzero(s: &ipi->ipi_gc_req,
399	n: sizeof(ipi->ipi_gc_req));
400	if (gc && ipi->ipi_gc != NULL) {
401	ipi->ipi_gc(ipi);
402	gccnt.intimer_lazy +=
403	ipi->ipi_gc_req.intimer_lazy;
404	gccnt.intimer_fast +=
405	ipi->ipi_gc_req.intimer_fast;
406	gccnt.intimer_nodelay +=
407	ipi->ipi_gc_req.intimer_nodelay;
408	}
409	}
410	if (INPCB_HAVE_TIMER_REQ(ipi->ipi_timer_req)) {
411	bzero(s: &ipi->ipi_timer_req,
412	n: sizeof(ipi->ipi_timer_req));
413	if (t && ipi->ipi_timer != NULL) {
414	ipi->ipi_timer(ipi);
415	tmcnt.intimer_lazy +=
416	ipi->ipi_timer_req.intimer_lazy;
417	tmcnt.intimer_fast +=
418	ipi->ipi_timer_req.intimer_fast;
419	tmcnt.intimer_nodelay +=
420	ipi->ipi_timer_req.intimer_nodelay;
421	}
422	}
423	}
424	lck_mtx_unlock(lck: &inpcb_lock);
425	lck_mtx_lock_spin(lck: &inpcb_timeout_lock);
426	}
427
428	/ lock was dropped above, so check first before overriding /
429	if (!inpcb_garbage_collecting) {
430	inpcb_garbage_collecting = INPCB_HAVE_TIMER_REQ(gccnt);
431	}
432	if (!inpcb_ticking) {
433	inpcb_ticking = INPCB_HAVE_TIMER_REQ(tmcnt);
434	}
435
436	/ arg0 will be set if we are the fast timer /
437	if (arg0 != NULL) {
438	inpcb_fast_timer_on = FALSE;
439	}
440	inpcb_timeout_run--;
441	VERIFY(inpcb_timeout_run >= `0` && inpcb_timeout_run < `2`);
442
443	/ re-arm the timer if there's work to do /
444	if (gccnt.intimer_nodelay > `0` \|\| tmcnt.intimer_nodelay > `0`) {
445	inpcb_sched_timeout();
446	} else if ((gccnt.intimer_fast + tmcnt.intimer_fast) <= `5`) {
447	/ be lazy when idle with little activity /
448	inpcb_sched_lazy_timeout();
449	} else {
450	inpcb_sched_timeout();
451	}
452
453	lck_mtx_unlock(lck: &inpcb_timeout_lock);
454	}
455
456	static void
457	inpcb_sched_timeout(void)
458	{
459	_inpcb_sched_timeout(`0`);
460	}
461
462	static void
463	inpcb_sched_lazy_timeout(void)
464	{
465	_inpcb_sched_timeout(inpcb_timeout_lazy);
466	}
467
468	static void
469	_inpcb_sched_timeout(unsigned int offset)
470	{
471	uint64_t deadline, leeway;
472
473	clock_interval_to_deadline(interval: `1`, NSEC_PER_SEC, result: &deadline);
474	LCK_MTX_ASSERT(&inpcb_timeout_lock, LCK_MTX_ASSERT_OWNED);
475	if (inpcb_timeout_run == `0` &&
476	(inpcb_garbage_collecting \|\| inpcb_ticking)) {
477	lck_mtx_convert_spin(lck: &inpcb_timeout_lock);
478	inpcb_timeout_run++;
479	if (offset == `0`) {
480	inpcb_fast_timer_on = TRUE;
481	thread_call_enter_delayed(call: inpcb_fast_thread_call,
482	deadline);
483	} else {
484	inpcb_fast_timer_on = FALSE;
485	clock_interval_to_absolutetime_interval(interval: offset,
486	NSEC_PER_SEC, result: &leeway);
487	thread_call_enter_delayed_with_leeway(
488	call: inpcb_thread_call, NULL, deadline, leeway,
489	THREAD_CALL_DELAY_LEEWAY);
490	}
491	} else if (inpcb_timeout_run == `1` &&
492	offset == `0` && !inpcb_fast_timer_on) {
493	/*
494	* Since the request was for a fast timer but the
495	* scheduled timer is a lazy timer, try to schedule
496	* another instance of fast timer also.
497	*/
498	lck_mtx_convert_spin(lck: &inpcb_timeout_lock);
499	inpcb_timeout_run++;
500	inpcb_fast_timer_on = TRUE;
501	thread_call_enter_delayed(call: inpcb_fast_thread_call, deadline);
502	}
503	}
504
505	void
506	inpcb_gc_sched(struct inpcbinfo *ipi, u_int32_t type)
507	{
508	u_int32_t gccnt;
509
510	lck_mtx_lock_spin(lck: &inpcb_timeout_lock);
511	inpcb_garbage_collecting = TRUE;
512	gccnt = ipi->ipi_gc_req.intimer_nodelay +
513	ipi->ipi_gc_req.intimer_fast;
514
515	if (gccnt > INPCB_GCREQ_THRESHOLD) {
516	type = INPCB_TIMER_FAST;
517	}
518
519	switch (type) {
520	case INPCB_TIMER_NODELAY:
521	os_atomic_inc(&ipi->ipi_gc_req.intimer_nodelay, relaxed);
522	inpcb_sched_timeout();
523	break;
524	case INPCB_TIMER_FAST:
525	os_atomic_inc(&ipi->ipi_gc_req.intimer_fast, relaxed);
526	inpcb_sched_timeout();
527	break;
528	default:
529	os_atomic_inc(&ipi->ipi_gc_req.intimer_lazy, relaxed);
530	inpcb_sched_lazy_timeout();
531	break;
532	}
533	lck_mtx_unlock(lck: &inpcb_timeout_lock);
534	}
535
536	void
537	inpcb_timer_sched(struct inpcbinfo *ipi, u_int32_t type)
538	{
539	lck_mtx_lock_spin(lck: &inpcb_timeout_lock);
540	inpcb_ticking = TRUE;
541	switch (type) {
542	case INPCB_TIMER_NODELAY:
543	os_atomic_inc(&ipi->ipi_timer_req.intimer_nodelay, relaxed);
544	inpcb_sched_timeout();
545	break;
546	case INPCB_TIMER_FAST:
547	os_atomic_inc(&ipi->ipi_timer_req.intimer_fast, relaxed);
548	inpcb_sched_timeout();
549	break;
550	default:
551	os_atomic_inc(&ipi->ipi_timer_req.intimer_lazy, relaxed);
552	inpcb_sched_lazy_timeout();
553	break;
554	}
555	lck_mtx_unlock(lck: &inpcb_timeout_lock);
556	}
557
558	void
559	in_pcbinfo_attach(struct inpcbinfo *ipi)
560	{
561	struct inpcbinfo *ipi0;
562
563	lck_mtx_lock(lck: &inpcb_lock);
564	TAILQ_FOREACH(ipi0, &inpcb_head, ipi_entry) {
565	if (ipi0 == ipi) {
566	panic("%s: ipi %p already in the list",
567	__func__, ipi);
568	/ NOTREACHED /
569	}
570	}
571	TAILQ_INSERT_TAIL(&inpcb_head, ipi, ipi_entry);
572	lck_mtx_unlock(lck: &inpcb_lock);
573	}
574
575	int
576	in_pcbinfo_detach(struct inpcbinfo *ipi)
577	{
578	struct inpcbinfo *ipi0;
579	int error = `0`;
580
581	lck_mtx_lock(lck: &inpcb_lock);
582	TAILQ_FOREACH(ipi0, &inpcb_head, ipi_entry) {
583	if (ipi0 == ipi) {
584	break;
585	}
586	}
587	if (ipi0 != NULL) {
588	TAILQ_REMOVE(&inpcb_head, ipi0, ipi_entry);
589	} else {
590	error = ENXIO;
591	}
592	lck_mtx_unlock(lck: &inpcb_lock);
593
594	return error;
595	}
596
597	__attribute__((noinline))
598	char *
599	inp_snprintf_tuple(struct inpcb inp, char* *buf, size_t buflen)
600	{
601	char laddrstr[MAX_IPv6_STR_LEN];
602	char faddrstr[MAX_IPv6_STR_LEN];
603	uint16_t lport = `0`;
604	uint16_t fport = `0`;
605	uint16_t proto = IPPROTO_IP;
606
607	if (inp->inp_socket != NULL) {
608	proto = SOCK_PROTO(inp->inp_socket);
609
610	if (proto == IPPROTO_TCP \|\| proto == IPPROTO_UDP) {
611	lport = inp->inp_lport;
612	fport = inp->inp_fport;
613	}
614	}
615	if (inp->inp_vflag & INP_IPV4) {
616	inet_ntop(AF_INET, (void )&inp->inp_laddr.s_addr, laddrstr, sizeof*(laddrstr));
617	inet_ntop(AF_INET, (void )&inp->inp_faddr.s_addr, faddrstr, sizeof*(faddrstr));
618	} else if (inp->inp_vflag & INP_IPV6) {
619	inet_ntop(AF_INET6, (void )&inp->in6p_faddr, laddrstr, sizeof*(laddrstr));
620	inet_ntop(AF_INET6, (void )&inp->in6p_faddr, faddrstr, sizeof*(faddrstr));
621	}
622	snprintf(buf, count: buflen, "[%u %s:%u %s:%u]",
623	proto, laddrstr, ntohs(lport), faddrstr, ntohs(fport));
624
625	return buf;
626	}
627
628	__attribute__((noinline))
629	void
630	in_pcb_check_management_entitled(struct inpcb *inp)
631	{
632	if (inp->inp_flags2 & INP2_MANAGEMENT_CHECKED) {
633	return;
634	}
635
636	if (management_data_unrestricted) {
637	inp->inp_flags2 \|= INP2_MANAGEMENT_ALLOWED;
638	inp->inp_flags2 \|= INP2_MANAGEMENT_CHECKED;
639	} else if (if_management_interface_check_needed == true) {
640	inp->inp_flags2 \|= INP2_MANAGEMENT_CHECKED;
641	/*
642	* Note that soopt_cred_check check both intcoproc entitlements
643	* We check MANAGEMENT_DATA_ENTITLEMENT as there is no corresponding PRIV value
644	*/
645	if (soopt_cred_check(so: inp->inp_socket, PRIV_NET_RESTRICTED_INTCOPROC, false, false) == `0`
646	\|\| IOCurrentTaskHasEntitlement(MANAGEMENT_DATA_ENTITLEMENT) == true
647	#if DEBUG \|\| DEVELOPMENT
648	\|\| IOCurrentTaskHasEntitlement(MANAGEMENT_DATA_ENTITLEMENT_DEVELOPMENT) == true
649	#endif /* DEBUG \|\| DEVELOPMENT */
650	) {
651	inp->inp_flags2 \|= INP2_MANAGEMENT_ALLOWED;
652	} else {
653	if (__improbable(if_management_verbose > `1`)) {
654	char buf[`128`];
655
656	os_log(OS_LOG_DEFAULT, "in_pcb_check_management_entitled %s:%d not management entitled %s",
657	proc_best_name(current_proc()),
658	proc_selfpid(),
659	inp_snprintf_tuple(inp, buf, sizeof(buf)));
660	}
661	}
662	}
663	}
664
665	/*
666	* Allocate a PCB and associate it with the socket.
667	*
668	* Returns: 0 Success
669	* ENOBUFS
670	* ENOMEM
671	*/
672	int
673	in_pcballoc(struct socket so, struct* inpcbinfo pcbinfo, struct* proc *p)
674	{
675	#pragma unused(p)
676	struct inpcb *inp;
677	caddr_t temp;
678
679	if ((so->so_flags1 & SOF1_CACHED_IN_SOCK_LAYER) == `0`) {
680	inp = zalloc_flags(pcbinfo->ipi_zone,
681	Z_WAITOK \| Z_ZERO \| Z_NOFAIL);
682	} else {
683	inp = (struct inpcb )(void* *)so->so_saved_pcb;
684	temp = inp->inp_saved_ppcb;
685	bzero(s: (caddr_t)inp, n: sizeof(*inp));
686	inp->inp_saved_ppcb = temp;
687	}
688
689	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
690	inp->inp_pcbinfo = pcbinfo;
691	inp->inp_socket = so;
692	/ make sure inp_stat is always 64-bit aligned /
693	inp->inp_stat = (struct inp_stat *)P2ROUNDUP(inp->inp_stat_store,
694	sizeof(u_int64_t));
695	if (((uintptr_t)inp->inp_stat - (uintptr_t)inp->inp_stat_store) +
696	sizeof(inp->inp_stat) > sizeof*(inp->inp_stat_store)) {
697	panic("%s: insufficient space to align inp_stat", __func__);
698	/ NOTREACHED /
699	}
700
701	/ make sure inp_cstat is always 64-bit aligned /
702	inp->inp_cstat = (struct inp_stat *)P2ROUNDUP(inp->inp_cstat_store,
703	sizeof(u_int64_t));
704	if (((uintptr_t)inp->inp_cstat - (uintptr_t)inp->inp_cstat_store) +
705	sizeof(inp->inp_cstat) > sizeof*(inp->inp_cstat_store)) {
706	panic("%s: insufficient space to align inp_cstat", __func__);
707	/ NOTREACHED /
708	}
709
710	/ make sure inp_wstat is always 64-bit aligned /
711	inp->inp_wstat = (struct inp_stat *)P2ROUNDUP(inp->inp_wstat_store,
712	sizeof(u_int64_t));
713	if (((uintptr_t)inp->inp_wstat - (uintptr_t)inp->inp_wstat_store) +
714	sizeof(inp->inp_wstat) > sizeof*(inp->inp_wstat_store)) {
715	panic("%s: insufficient space to align inp_wstat", __func__);
716	/ NOTREACHED /
717	}
718
719	/ make sure inp_Wstat is always 64-bit aligned /
720	inp->inp_Wstat = (struct inp_stat *)P2ROUNDUP(inp->inp_Wstat_store,
721	sizeof(u_int64_t));
722	if (((uintptr_t)inp->inp_Wstat - (uintptr_t)inp->inp_Wstat_store) +
723	sizeof(inp->inp_Wstat) > sizeof*(inp->inp_Wstat_store)) {
724	panic("%s: insufficient space to align inp_Wstat", __func__);
725	/ NOTREACHED /
726	}
727
728	so->so_pcb = (caddr_t)inp;
729
730	if (so->so_proto->pr_flags & PR_PCBLOCK) {
731	lck_mtx_init(lck: &inp->inpcb_mtx, grp: pcbinfo->ipi_lock_grp,
732	attr: &pcbinfo->ipi_lock_attr);
733	}
734
735	if (SOCK_DOM(so) == PF_INET6 && !ip6_mapped_addr_on) {
736	inp->inp_flags \|= IN6P_IPV6_V6ONLY;
737	}
738
739	if (ip6_auto_flowlabel) {
740	inp->inp_flags \|= IN6P_AUTOFLOWLABEL;
741	}
742	if (intcoproc_unrestricted) {
743	inp->inp_flags2 \|= INP2_INTCOPROC_ALLOWED;
744	}
745
746	(void) inp_update_policy(inp);
747
748	lck_rw_lock_exclusive(lck: &pcbinfo->ipi_lock);
749	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
750	LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list);
751	pcbinfo->ipi_count++;
752	lck_rw_done(lck: &pcbinfo->ipi_lock);
753	return `0`;
754	}
755
756	/*
757	* in_pcblookup_local_and_cleanup does everything
758	* in_pcblookup_local does but it checks for a socket
759	* that's going away. Since we know that the lock is
760	* held read+write when this function is called, we
761	* can safely dispose of this socket like the slow
762	* timer would usually do and return NULL. This is
763	* great for bind.
764	*/
765	struct inpcb *
766	in_pcblookup_local_and_cleanup(struct inpcbinfo pcbinfo, struct* in_addr laddr,
767	u_int lport_arg, int wild_okay)
768	{
769	struct inpcb *inp;
770
771	/ Perform normal lookup /
772	inp = in_pcblookup_local(pcbinfo, laddr, lport_arg, wild_okay);
773
774	/ Check if we found a match but it's waiting to be disposed /
775	if (inp != NULL && inp->inp_wantcnt == WNT_STOPUSING) {
776	struct socket *so = inp->inp_socket;
777
778	socket_lock(so, refcount: `0`);
779
780	if (so->so_usecount == `0`) {
781	if (inp->inp_state != INPCB_STATE_DEAD) {
782	in_pcbdetach(inp);
783	}
784	in_pcbdispose(inp); / will unlock & destroy /
785	inp = NULL;
786	} else {
787	socket_unlock(so, refcount: `0`);
788	}
789	}
790
791	return inp;
792	}
793
794	static void
795	in_pcb_conflict_post_msg(u_int16_t port)
796	{
797	/*
798	* Radar 5523020 send a kernel event notification if a
799	* non-participating socket tries to bind the port a socket
800	* who has set SOF_NOTIFYCONFLICT owns.
801	*/
802	struct kev_msg ev_msg;
803	struct kev_in_portinuse in_portinuse;
804
805	bzero(s: &in_portinuse, n: sizeof(struct kev_in_portinuse));
806	bzero(s: &ev_msg, n: sizeof(struct kev_msg));
807	in_portinuse.port = ntohs(port); / port in host order /
808	in_portinuse.req_pid = proc_selfpid();
809	ev_msg.vendor_code = KEV_VENDOR_APPLE;
810	ev_msg.kev_class = KEV_NETWORK_CLASS;
811	ev_msg.kev_subclass = KEV_INET_SUBCLASS;
812	ev_msg.event_code = KEV_INET_PORTINUSE;
813	ev_msg.dv[`0`].data_ptr = &in_portinuse;
814	ev_msg.dv[`0`].data_length = sizeof(struct kev_in_portinuse);
815	ev_msg.dv[`1`].data_length = `0`;
816	dlil_post_complete_msg(NULL, &ev_msg);
817	}
818
819	/*
820	* Bind an INPCB to an address and/or port. This routine should not alter
821	* the caller-supplied local address "nam".
822	*
823	* Returns: 0 Success
824	* EADDRNOTAVAIL Address not available.
825	* EINVAL Invalid argument
826	* EAFNOSUPPORT Address family not supported [notdef]
827	* EACCES Permission denied
828	* EADDRINUSE Address in use
829	* EAGAIN Resource unavailable, try again
830	* priv_check_cred:EPERM Operation not permitted
831	*/
832	int
833	in_pcbbind(struct inpcb inp, struct* sockaddr nam, struct* proc *p)
834	{
835	struct socket *so = inp->inp_socket;
836	unsigned short *lastport;
837	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
838	u_short lport = `0`, rand_port = `0`;
839	int wild = `0`;
840	int reuseport = (so->so_options & SO_REUSEPORT);
841	int error = `0`;
842	int randomport;
843	int conflict = `0`;
844	boolean_t anonport = FALSE;
845	kauth_cred_t cred;
846	struct in_addr laddr;
847	struct ifnet *outif = NULL;
848
849	if (inp->inp_flags2 & INP2_BIND_IN_PROGRESS) {
850	return EINVAL;
851	}
852	inp->inp_flags2 \|= INP2_BIND_IN_PROGRESS;
853
854	if (TAILQ_EMPTY(&in_ifaddrhead)) { / XXX broken! /
855	error = EADDRNOTAVAIL;
856	goto done;
857	}
858	if (!(so->so_options & (SO_REUSEADDR \| SO_REUSEPORT))) {
859	wild = `1`;
860	}
861
862	bzero(s: &laddr, n: sizeof(laddr));
863
864	socket_unlock(so, refcount: `0`); / keep reference on socket /
865	lck_rw_lock_exclusive(lck: &pcbinfo->ipi_lock);
866	if (inp->inp_lport != `0` \|\| inp->inp_laddr.s_addr != INADDR_ANY) {
867	/ another thread completed the bind /
868	lck_rw_done(lck: &pcbinfo->ipi_lock);
869	socket_lock(so, refcount: `0`);
870	error = EINVAL;
871	goto done;
872	}
873
874	if (nam != NULL) {
875	if (nam->sa_len != sizeof(struct sockaddr_in)) {
876	lck_rw_done(lck: &pcbinfo->ipi_lock);
877	socket_lock(so, refcount: `0`);
878	error = EINVAL;
879	goto done;
880	}
881	#if 0
882	/*
883	* We should check the family, but old programs
884	* incorrectly fail to initialize it.
885	*/
886	if (nam->sa_family != AF_INET) {
887	lck_rw_done(&pcbinfo->ipi_lock);
888	socket_lock(so, `0`);
889	error = EAFNOSUPPORT;
890	goto done;
891	}
892	#endif /* 0 */
893	lport = SIN(nam)->sin_port;
894
895	if (IN_MULTICAST(ntohl(SIN(nam)->sin_addr.s_addr))) {
896	/*
897	* Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
898	* allow complete duplication of binding if
899	* SO_REUSEPORT is set, or if SO_REUSEADDR is set
900	* and a multicast address is bound on both
901	* new and duplicated sockets.
902	*/
903	if (so->so_options & SO_REUSEADDR) {
904	reuseport = SO_REUSEADDR \| SO_REUSEPORT;
905	}
906	} else if (SIN(nam)->sin_addr.s_addr != INADDR_ANY) {
907	struct sockaddr_in sin;
908	struct ifaddr *ifa;
909
910	/ Sanitized for interface address searches /
911	SOCKADDR_ZERO(&sin, sizeof(sin));
912	sin.sin_family = AF_INET;
913	sin.sin_len = sizeof(struct sockaddr_in);
914	sin.sin_addr.s_addr = SIN(nam)->sin_addr.s_addr;
915
916	ifa = ifa_ifwithaddr(SA(&sin));
917	if (ifa == NULL) {
918	lck_rw_done(lck: &pcbinfo->ipi_lock);
919	socket_lock(so, refcount: `0`);
920	error = EADDRNOTAVAIL;
921	goto done;
922	} else {
923	/*
924	* Opportunistically determine the outbound
925	* interface that may be used; this may not
926	* hold true if we end up using a route
927	* going over a different interface, e.g.
928	* when sending to a local address. This
929	* will get updated again after sending.
930	*/
931	IFA_LOCK(ifa);
932	outif = ifa->ifa_ifp;
933	IFA_UNLOCK(ifa);
934	ifa_remref(ifa);
935	}
936	}
937
938	#if SKYWALK
939	if (inp->inp_flags2 & INP2_EXTERNAL_PORT) {
940	// Extract the external flow info
941	struct ns_flow_info nfi = {};
942	error = necp_client_get_netns_flow_info(client_id: inp->necp_client_uuid,
943	flow_info: &nfi);
944	if (error != `0`) {
945	lck_rw_done(lck: &pcbinfo->ipi_lock);
946	socket_lock(so, refcount: `0`);
947	goto done;
948	}
949
950	// Extract the reserved port
951	u_int16_t reserved_lport = `0`;
952	if (nfi.nfi_laddr.sa.sa_family == AF_INET) {
953	reserved_lport = nfi.nfi_laddr.sin.sin_port;
954	} else if (nfi.nfi_laddr.sa.sa_family == AF_INET6) {
955	reserved_lport = nfi.nfi_laddr.sin6.sin6_port;
956	} else {
957	lck_rw_done(lck: &pcbinfo->ipi_lock);
958	socket_lock(so, refcount: `0`);
959	error = EINVAL;
960	goto done;
961	}
962
963	// Validate or use the reserved port
964	if (lport == `0`) {
965	lport = reserved_lport;
966	} else if (lport != reserved_lport) {
967	lck_rw_done(lck: &pcbinfo->ipi_lock);
968	socket_lock(so, refcount: `0`);
969	error = EINVAL;
970	goto done;
971	}
972	}
973
974	/ Do not allow reserving a UDP port if remaining UDP port count is below 4096 /
975	if (SOCK_PROTO(so) == IPPROTO_UDP && !allow_udp_port_exhaustion) {
976	uint32_t current_reservations = `0`;
977	if (inp->inp_vflag & INP_IPV6) {
978	current_reservations = netns_lookup_reservations_count_in6(addr: inp->in6p_laddr, IPPROTO_UDP);
979	} else {
980	current_reservations = netns_lookup_reservations_count_in(addr: inp->inp_laddr, IPPROTO_UDP);
981	}
982	if (USHRT_MAX - UDP_RANDOM_PORT_RESERVE < current_reservations) {
983	log(LOG_ERR, "UDP port not available, less than 4096 UDP ports left");
984	lck_rw_done(lck: &pcbinfo->ipi_lock);
985	socket_lock(so, refcount: `0`);
986	error = EADDRNOTAVAIL;
987	goto done;
988	}
989	}
990
991	#endif /* SKYWALK */
992
993	if (lport != `0`) {
994	struct inpcb *t;
995	uid_t u;
996
997	#if XNU_TARGET_OS_OSX
998	if (ntohs(lport) < IPPORT_RESERVED &&
999	SIN(nam)->sin_addr.s_addr != `0` &&
1000	!(inp->inp_flags2 & INP2_EXTERNAL_PORT)) {
1001	cred = kauth_cred_proc_ref(procp: p);
1002	error = priv_check_cred(cred,
1003	PRIV_NETINET_RESERVEDPORT, flags: `0`);
1004	kauth_cred_unref(&cred);
1005	if (error != `0`) {
1006	lck_rw_done(lck: &pcbinfo->ipi_lock);
1007	socket_lock(so, refcount: `0`);
1008	error = EACCES;
1009	goto done;
1010	}
1011	}
1012	#endif /* XNU_TARGET_OS_OSX */
1013	/*
1014	* Check wether the process is allowed to bind to a restricted port
1015	*/
1016	if (!current_task_can_use_restricted_in_port(port: lport,
1017	protocol: (uint8_t)SOCK_PROTO(so), PORT_FLAGS_BSD)) {
1018	lck_rw_done(lck: &pcbinfo->ipi_lock);
1019	socket_lock(so, refcount: `0`);
1020	error = EADDRINUSE;
1021	goto done;
1022	}
1023
1024	if (!IN_MULTICAST(ntohl(SIN(nam)->sin_addr.s_addr)) &&
1025	(u = kauth_cred_getuid(cred: so->so_cred)) != `0` &&
1026	(t = in_pcblookup_local_and_cleanup(
1027	pcbinfo: inp->inp_pcbinfo, SIN(nam)->sin_addr, lport_arg: lport,
1028	INPLOOKUP_WILDCARD)) != NULL &&
1029	(SIN(nam)->sin_addr.s_addr != INADDR_ANY \|\|
1030	t->inp_laddr.s_addr != INADDR_ANY \|\|
1031	!(t->inp_socket->so_options & SO_REUSEPORT)) &&
1032	(u != kauth_cred_getuid(cred: t->inp_socket->so_cred)) &&
1033	!(t->inp_socket->so_flags & SOF_REUSESHAREUID) &&
1034	(SIN(nam)->sin_addr.s_addr != INADDR_ANY \|\|
1035	t->inp_laddr.s_addr != INADDR_ANY) &&
1036	(!(t->inp_flags2 & INP2_EXTERNAL_PORT) \|\|
1037	!(inp->inp_flags2 & INP2_EXTERNAL_PORT) \|\|
1038	uuid_compare(uu1: t->necp_client_uuid, uu2: inp->necp_client_uuid) != `0`)) {
1039	if ((t->inp_socket->so_flags &
1040	SOF_NOTIFYCONFLICT) &&
1041	!(so->so_flags & SOF_NOTIFYCONFLICT)) {
1042	conflict = `1`;
1043	}
1044
1045	lck_rw_done(lck: &pcbinfo->ipi_lock);
1046
1047	if (conflict) {
1048	in_pcb_conflict_post_msg(port: lport);
1049	}
1050
1051	socket_lock(so, refcount: `0`);
1052	error = EADDRINUSE;
1053	goto done;
1054	}
1055	t = in_pcblookup_local_and_cleanup(pcbinfo,
1056	SIN(nam)->sin_addr, lport_arg: lport, wild_okay: wild);
1057	if (t != NULL &&
1058	(reuseport & t->inp_socket->so_options) == `0` &&
1059	(!(t->inp_flags2 & INP2_EXTERNAL_PORT) \|\|
1060	!(inp->inp_flags2 & INP2_EXTERNAL_PORT) \|\|
1061	uuid_compare(uu1: t->necp_client_uuid, uu2: inp->necp_client_uuid) != `0`)) {
1062	if (SIN(nam)->sin_addr.s_addr != INADDR_ANY \|\|
1063	t->inp_laddr.s_addr != INADDR_ANY \|\|
1064	SOCK_DOM(so) != PF_INET6 \|\|
1065	SOCK_DOM(t->inp_socket) != PF_INET6) {
1066	if ((t->inp_socket->so_flags &
1067	SOF_NOTIFYCONFLICT) &&
1068	!(so->so_flags & SOF_NOTIFYCONFLICT)) {
1069	conflict = `1`;
1070	}
1071
1072	lck_rw_done(lck: &pcbinfo->ipi_lock);
1073
1074	if (conflict) {
1075	in_pcb_conflict_post_msg(port: lport);
1076	}
1077	socket_lock(so, refcount: `0`);
1078	error = EADDRINUSE;
1079	goto done;
1080	}
1081	}
1082	#if SKYWALK
1083	if ((SOCK_PROTO(so) == IPPROTO_TCP \|\|
1084	SOCK_PROTO(so) == IPPROTO_UDP) &&
1085	!(inp->inp_flags2 & INP2_EXTERNAL_PORT)) {
1086	int res_err = `0`;
1087	if (inp->inp_vflag & INP_IPV6) {
1088	res_err = netns_reserve_in6(
1089	token: &inp->inp_netns_token,
1090	SIN6(nam)->sin6_addr,
1091	proto: (uint8_t)SOCK_PROTO(so), port: lport, NETNS_BSD,
1092	NULL);
1093	} else {
1094	res_err = netns_reserve_in(
1095	token: &inp->inp_netns_token,
1096	SIN(nam)->sin_addr, proto: (uint8_t)SOCK_PROTO(so),
1097	port: lport, NETNS_BSD, NULL);
1098	}
1099	if (res_err != `0`) {
1100	lck_rw_done(lck: &pcbinfo->ipi_lock);
1101	socket_lock(so, refcount: `0`);
1102	error = EADDRINUSE;
1103	goto done;
1104	}
1105	}
1106	#endif /* SKYWALK */
1107	}
1108	laddr = SIN(nam)->sin_addr;
1109	}
1110	if (lport == `0`) {
1111	u_short first, last;
1112	int count;
1113	bool found;
1114
1115	/*
1116	* Override wild = 1 for implicit bind (mainly used by connect)
1117	* For implicit bind (lport == 0), we always use an unused port,
1118	* so REUSEADDR\|REUSEPORT don't apply
1119	*/
1120	wild = `1`;
1121
1122	randomport = (so->so_flags & SOF_BINDRANDOMPORT) \|\|
1123	(so->so_type == SOCK_STREAM ? tcp_use_randomport :
1124	udp_use_randomport);
1125
1126	/*
1127	* Even though this looks similar to the code in
1128	* in6_pcbsetport, the v6 vs v4 checks are different.
1129	*/
1130	anonport = TRUE;
1131	if (inp->inp_flags & INP_HIGHPORT) {
1132	first = (u_short)ipport_hifirstauto; / sysctl /
1133	last = (u_short)ipport_hilastauto;
1134	lastport = &pcbinfo->ipi_lasthi;
1135	} else if (inp->inp_flags & INP_LOWPORT) {
1136	cred = kauth_cred_proc_ref(procp: p);
1137	error = priv_check_cred(cred,
1138	PRIV_NETINET_RESERVEDPORT, flags: `0`);
1139	kauth_cred_unref(&cred);
1140	if (error != `0`) {
1141	lck_rw_done(lck: &pcbinfo->ipi_lock);
1142	socket_lock(so, refcount: `0`);
1143	goto done;
1144	}
1145	first = (u_short)ipport_lowfirstauto; / 1023 /
1146	last = (u_short)ipport_lowlastauto; / 600 /
1147	lastport = &pcbinfo->ipi_lastlow;
1148	} else {
1149	first = (u_short)ipport_firstauto; / sysctl /
1150	last = (u_short)ipport_lastauto;
1151	lastport = &pcbinfo->ipi_lastport;
1152	}
1153	/ No point in randomizing if only one port is available /
1154
1155	if (first == last) {
1156	randomport = `0`;
1157	}
1158	/*
1159	* Simple check to ensure all ports are not used up causing
1160	* a deadlock here.
1161	*
1162	* We split the two cases (up and down) so that the direction
1163	* is not being tested on each round of the loop.
1164	*/
1165	if (first > last) {
1166	struct in_addr lookup_addr;
1167
1168	/*
1169	* counting down
1170	*/
1171	if (randomport) {
1172	read_frandom(buffer: &rand_port, numBytes: sizeof(rand_port));
1173	*lastport =
1174	first - (rand_port % (first - last));
1175	}
1176	count = first - last;
1177
1178	lookup_addr = (laddr.s_addr != INADDR_ANY) ? laddr :
1179	inp->inp_laddr;
1180
1181	found = false;
1182	do {
1183	if (count-- < `0`) { / completely used? /
1184	lck_rw_done(lck: &pcbinfo->ipi_lock);
1185	socket_lock(so, refcount: `0`);
1186	error = EADDRNOTAVAIL;
1187	goto done;
1188	}
1189	--*lastport;
1190	if (lastport > first \|\| lastport < last) {
1191	*lastport = first;
1192	}
1193	lport = htons(*lastport);
1194
1195	/*
1196	* Skip if this is a restricted port as we do not want to
1197	* restricted ports as ephemeral
1198	*/
1199	if (IS_RESTRICTED_IN_PORT(lport)) {
1200	continue;
1201	}
1202
1203	found = in_pcblookup_local_and_cleanup(pcbinfo,
1204	laddr: lookup_addr, lport_arg: lport, wild_okay: wild) == NULL;
1205	#if SKYWALK
1206	if (found &&
1207	(SOCK_PROTO(so) == IPPROTO_TCP \|\|
1208	SOCK_PROTO(so) == IPPROTO_UDP) &&
1209	!(inp->inp_flags2 & INP2_EXTERNAL_PORT)) {
1210	int res_err;
1211	if (inp->inp_vflag & INP_IPV6) {
1212	res_err = netns_reserve_in6(
1213	token: &inp->inp_netns_token,
1214	addr: inp->in6p_laddr,
1215	proto: (uint8_t)SOCK_PROTO(so), port: lport,
1216	NETNS_BSD, NULL);
1217	} else {
1218	res_err = netns_reserve_in(
1219	token: &inp->inp_netns_token,
1220	addr: lookup_addr, proto: (uint8_t)SOCK_PROTO(so),
1221	port: lport, NETNS_BSD, NULL);
1222	}
1223	found = res_err == `0`;
1224	}
1225	#endif /* SKYWALK */
1226	} while (!found);
1227	} else {
1228	struct in_addr lookup_addr;
1229
1230	/*
1231	* counting up
1232	*/
1233	if (randomport) {
1234	read_frandom(buffer: &rand_port, numBytes: sizeof(rand_port));
1235	*lastport =
1236	first + (rand_port % (first - last));
1237	}
1238	count = last - first;
1239
1240	lookup_addr = (laddr.s_addr != INADDR_ANY) ? laddr :
1241	inp->inp_laddr;
1242
1243	found = false;
1244	do {
1245	if (count-- < `0`) { / completely used? /
1246	lck_rw_done(lck: &pcbinfo->ipi_lock);
1247	socket_lock(so, refcount: `0`);
1248	error = EADDRNOTAVAIL;
1249	goto done;
1250	}
1251	++*lastport;
1252	if (lastport < first \|\| lastport > last) {
1253	*lastport = first;
1254	}
1255	lport = htons(*lastport);
1256
1257	/*
1258	* Skip if this is a restricted port as we do not want to
1259	* restricted ports as ephemeral
1260	*/
1261	if (IS_RESTRICTED_IN_PORT(lport)) {
1262	continue;
1263	}
1264
1265	found = in_pcblookup_local_and_cleanup(pcbinfo,
1266	laddr: lookup_addr, lport_arg: lport, wild_okay: wild) == NULL;
1267	#if SKYWALK
1268	if (found &&
1269	(SOCK_PROTO(so) == IPPROTO_TCP \|\|
1270	SOCK_PROTO(so) == IPPROTO_UDP) &&
1271	!(inp->inp_flags2 & INP2_EXTERNAL_PORT)) {
1272	int res_err;
1273	if (inp->inp_vflag & INP_IPV6) {
1274	res_err = netns_reserve_in6(
1275	token: &inp->inp_netns_token,
1276	addr: inp->in6p_laddr,
1277	proto: (uint8_t)SOCK_PROTO(so), port: lport,
1278	NETNS_BSD, NULL);
1279	} else {
1280	res_err = netns_reserve_in(
1281	token: &inp->inp_netns_token,
1282	addr: lookup_addr, proto: (uint8_t)SOCK_PROTO(so),
1283	port: lport, NETNS_BSD, NULL);
1284	}
1285	found = res_err == `0`;
1286	}
1287	#endif /* SKYWALK */
1288	} while (!found);
1289	}
1290	}
1291	socket_lock(so, refcount: `0`);
1292
1293	/*
1294	* We unlocked socket's protocol lock for a long time.
1295	* The socket might have been dropped/defuncted.
1296	* Checking if world has changed since.
1297	*/
1298	if (inp->inp_state == INPCB_STATE_DEAD) {
1299	#if SKYWALK
1300	netns_release(token: &inp->inp_netns_token);
1301	#endif /* SKYWALK */
1302	lck_rw_done(lck: &pcbinfo->ipi_lock);
1303	error = ECONNABORTED;
1304	goto done;
1305	}
1306
1307	if (inp->inp_lport != `0` \|\| inp->inp_laddr.s_addr != INADDR_ANY) {
1308	#if SKYWALK
1309	netns_release(token: &inp->inp_netns_token);
1310	#endif /* SKYWALK */
1311	lck_rw_done(lck: &pcbinfo->ipi_lock);
1312	error = EINVAL;
1313	goto done;
1314	}
1315
1316	if (laddr.s_addr != INADDR_ANY) {
1317	inp->inp_laddr = laddr;
1318	inp->inp_last_outifp = outif;
1319	#if SKYWALK
1320	if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
1321	netns_set_ifnet(token: &inp->inp_netns_token, ifp: outif);
1322	}
1323	#endif /* SKYWALK */
1324	}
1325	inp->inp_lport = lport;
1326	if (anonport) {
1327	inp->inp_flags \|= INP_ANONPORT;
1328	}
1329
1330	if (in_pcbinshash(inp, `1`) != `0`) {
1331	inp->inp_laddr.s_addr = INADDR_ANY;
1332	inp->inp_last_outifp = NULL;
1333
1334	#if SKYWALK
1335	netns_release(token: &inp->inp_netns_token);
1336	#endif /* SKYWALK */
1337	inp->inp_lport = `0`;
1338	if (anonport) {
1339	inp->inp_flags &= ~INP_ANONPORT;
1340	}
1341	lck_rw_done(lck: &pcbinfo->ipi_lock);
1342	error = EAGAIN;
1343	goto done;
1344	}
1345	lck_rw_done(lck: &pcbinfo->ipi_lock);
1346	sflt_notify(so, event: sock_evt_bound, NULL);
1347
1348	in_pcb_check_management_entitled(inp);
1349	done:
1350	inp->inp_flags2 &= ~INP2_BIND_IN_PROGRESS;
1351	return error;
1352	}
1353
1354	#define APN_FALLBACK_IP_FILTER(a) \
1355	(IN_LINKLOCAL(ntohl((a)->sin_addr.s_addr)) \|\| \
1356	IN_LOOPBACK(ntohl((a)->sin_addr.s_addr)) \|\| \
1357	IN_ZERONET(ntohl((a)->sin_addr.s_addr)) \|\| \
1358	IN_MULTICAST(ntohl((a)->sin_addr.s_addr)) \|\| \
1359	IN_PRIVATE(ntohl((a)->sin_addr.s_addr)))
1360
1361	#define APN_FALLBACK_NOTIF_INTERVAL 2 /* Magic Number */
1362	static uint64_t last_apn_fallback = `0`;
1363
1364	static boolean_t
1365	apn_fallback_required(proc_t proc, struct socket so, struct* sockaddr_in *p_dstv4)
1366	{
1367	uint64_t timenow;
1368	struct sockaddr_storage lookup_default_addr;
1369	struct rtentry *rt = NULL;
1370
1371	VERIFY(proc != NULL);
1372
1373	if (apn_fallbk_enabled == FALSE) {
1374	return FALSE;
1375	}
1376
1377	if (proc == kernproc) {
1378	return FALSE;
1379	}
1380
1381	if (so && (so->so_options & SO_NOAPNFALLBK)) {
1382	return FALSE;
1383	}
1384
1385	timenow = net_uptime();
1386	if ((timenow - last_apn_fallback) < APN_FALLBACK_NOTIF_INTERVAL) {
1387	apn_fallbk_log((LOG_INFO, "APN fallback notification throttled.\n"));
1388	return FALSE;
1389	}
1390
1391	if (p_dstv4 && APN_FALLBACK_IP_FILTER(p_dstv4)) {
1392	return FALSE;
1393	}
1394
1395	/ Check if we have unscoped IPv6 default route through cellular /
1396	bzero(s: &lookup_default_addr, n: sizeof(lookup_default_addr));
1397	lookup_default_addr.ss_family = AF_INET6;
1398	lookup_default_addr.ss_len = sizeof(struct sockaddr_in6);
1399
1400	rt = rtalloc1(SA(&lookup_default_addr), `0`, `0`);
1401	if (NULL == rt) {
1402	apn_fallbk_log((LOG_INFO, "APN fallback notification could not find "
1403	"unscoped default IPv6 route.\n"));
1404	return FALSE;
1405	}
1406
1407	if (!IFNET_IS_CELLULAR(rt->rt_ifp)) {
1408	rtfree(rt);
1409	apn_fallbk_log((LOG_INFO, "APN fallback notification could not find "
1410	"unscoped default IPv6 route through cellular interface.\n"));
1411	return FALSE;
1412	}
1413
1414	/*
1415	* We have a default IPv6 route, ensure that
1416	* we do not have IPv4 default route before triggering
1417	* the event
1418	*/
1419	rtfree(rt);
1420	rt = NULL;
1421
1422	bzero(s: &lookup_default_addr, n: sizeof(lookup_default_addr));
1423	lookup_default_addr.ss_family = AF_INET;
1424	lookup_default_addr.ss_len = sizeof(struct sockaddr_in);
1425
1426	rt = rtalloc1(SA(&lookup_default_addr), `0`, `0`);
1427
1428	if (rt) {
1429	rtfree(rt);
1430	rt = NULL;
1431	apn_fallbk_log((LOG_INFO, "APN fallback notification found unscoped "
1432	"IPv4 default route!\n"));
1433	return FALSE;
1434	}
1435
1436	{
1437	/*
1438	* We disable APN fallback if the binary is not a third-party app.
1439	* Note that platform daemons use their process name as a
1440	* bundle ID so we filter out bundle IDs without dots.
1441	*/
1442	const char *bundle_id = cs_identity_get(proc);
1443	if (bundle_id == NULL \|\|
1444	bundle_id[`0`] == `'\0'` \|\|
1445	strchr(s: bundle_id, c: `'.'`) == NULL \|\|
1446	strncmp(s1: bundle_id, s2: "com.apple.", n: sizeof("com.apple.") - `1`) == `0`) {
1447	apn_fallbk_log((LOG_INFO, "Abort: APN fallback notification found first-"
1448	"party bundle ID \"%s\"!\n", (bundle_id ? bundle_id : "NULL")));
1449	return FALSE;
1450	}
1451	}
1452
1453	{
1454	/*
1455	* The Apple App Store IPv6 requirement started on
1456	* June 1st, 2016 at 12:00:00 AM PDT.
1457	* We disable APN fallback if the binary is more recent than that.
1458	* We check both atime and birthtime since birthtime is not always supported.
1459	*/
1460	static const long ipv6_start_date = `1464764400L`;
1461	vfs_context_t context;
1462	struct stat64 sb;
1463	int vn_stat_error;
1464
1465	bzero(s: &sb, n: sizeof(struct stat64));
1466	context = vfs_context_create(NULL);
1467	vn_stat_error = vn_stat(vp: proc->p_textvp, sb: &sb, NULL, isstat64: `1`, needsrealdev: `0`, ctx: context);
1468	(void)vfs_context_rele(ctx: context);
1469
1470	if (vn_stat_error != `0` \|\|
1471	sb.st_atimespec.tv_sec >= ipv6_start_date \|\|
1472	sb.st_birthtimespec.tv_sec >= ipv6_start_date) {
1473	apn_fallbk_log((LOG_INFO, "Abort: APN fallback notification found binary "
1474	"too recent! (err %d atime %ld mtime %ld ctime %ld birthtime %ld)\n",
1475	vn_stat_error, sb.st_atimespec.tv_sec, sb.st_mtimespec.tv_sec,
1476	sb.st_ctimespec.tv_sec, sb.st_birthtimespec.tv_sec));
1477	return FALSE;
1478	}
1479	}
1480	return TRUE;
1481	}
1482
1483	static void
1484	apn_fallback_trigger(proc_t proc, struct socket *so)
1485	{
1486	pid_t pid = `0`;
1487	struct kev_msg ev_msg;
1488	struct kev_netevent_apnfallbk_data apnfallbk_data;
1489
1490	last_apn_fallback = net_uptime();
1491	pid = proc_pid(proc);
1492	uuid_t application_uuid;
1493	uuid_clear(uu: application_uuid);
1494	proc_getexecutableuuid(proc, application_uuid,
1495	sizeof(application_uuid));
1496
1497	bzero(s: &ev_msg, n: sizeof(struct kev_msg));
1498	ev_msg.vendor_code = KEV_VENDOR_APPLE;
1499	ev_msg.kev_class = KEV_NETWORK_CLASS;
1500	ev_msg.kev_subclass = KEV_NETEVENT_SUBCLASS;
1501	ev_msg.event_code = KEV_NETEVENT_APNFALLBACK;
1502
1503	bzero(s: &apnfallbk_data, n: sizeof(apnfallbk_data));
1504
1505	if (so->so_flags & SOF_DELEGATED) {
1506	apnfallbk_data.epid = so->e_pid;
1507	uuid_copy(dst: apnfallbk_data.euuid, src: so->e_uuid);
1508	} else {
1509	apnfallbk_data.epid = so->last_pid;
1510	uuid_copy(dst: apnfallbk_data.euuid, src: so->last_uuid);
1511	}
1512
1513	ev_msg.dv[`0`].data_ptr = &apnfallbk_data;
1514	ev_msg.dv[`0`].data_length = sizeof(apnfallbk_data);
1515	kev_post_msg(event: &ev_msg);
1516	apn_fallbk_log((LOG_INFO, "APN fallback notification issued.\n"));
1517	}
1518
1519	/*
1520	* Transform old in_pcbconnect() into an inner subroutine for new
1521	* in_pcbconnect(); do some validity-checking on the remote address
1522	* (in "nam") and then determine local host address (i.e., which
1523	* interface) to use to access that remote host.
1524	*
1525	* This routine may alter the caller-supplied remote address "nam".
1526	*
1527	* The caller may override the bound-to-interface setting of the socket
1528	* by specifying the ifscope parameter (e.g. from IP_PKTINFO.)
1529	*
1530	* This routine might return an ifp with a reference held if the caller
1531	* provides a non-NULL outif, even in the error case. The caller is
1532	* responsible for releasing its reference.
1533	*
1534	* Returns: 0 Success
1535	* EINVAL Invalid argument
1536	* EAFNOSUPPORT Address family not supported
1537	* EADDRNOTAVAIL Address not available
1538	*/
1539	int
1540	in_pcbladdr(struct inpcb inp, struct* sockaddr nam, struct* in_addr *laddr,
1541	unsigned int ifscope, struct ifnet *outif, int* raw)
1542	{
1543	struct route *ro = &inp->inp_route;
1544	struct in_ifaddr *ia = NULL;
1545	struct sockaddr_in sin;
1546	int error = `0`;
1547	boolean_t restricted = FALSE;
1548
1549	if (outif != NULL) {
1550	*outif = NULL;
1551	}
1552	if (nam->sa_len != sizeof(struct sockaddr_in)) {
1553	return EINVAL;
1554	}
1555	if (SIN(nam)->sin_family != AF_INET) {
1556	return EAFNOSUPPORT;
1557	}
1558	if (raw == `0` && SIN(nam)->sin_port == `0`) {
1559	return EADDRNOTAVAIL;
1560	}
1561
1562	in_pcb_check_management_entitled(inp);
1563
1564	/*
1565	* If the destination address is INADDR_ANY,
1566	* use the primary local address.
1567	* If the supplied address is INADDR_BROADCAST,
1568	* and the primary interface supports broadcast,
1569	* choose the broadcast address for that interface.
1570	*/
1571	if (raw == `0` && (SIN(nam)->sin_addr.s_addr == INADDR_ANY \|\|
1572	SIN(nam)->sin_addr.s_addr == (u_int32_t)INADDR_BROADCAST)) {
1573	lck_rw_lock_shared(lck: &in_ifaddr_rwlock);
1574	if (!TAILQ_EMPTY(&in_ifaddrhead)) {
1575	ia = TAILQ_FIRST(&in_ifaddrhead);
1576	IFA_LOCK_SPIN(&ia->ia_ifa);
1577	if (SIN(nam)->sin_addr.s_addr == INADDR_ANY) {
1578	SIN(nam)->sin_addr = IA_SIN(ia)->sin_addr;
1579	} else if (ia->ia_ifp->if_flags & IFF_BROADCAST) {
1580	SIN(nam)->sin_addr =
1581	SIN(&ia->ia_broadaddr)->sin_addr;
1582	}
1583	IFA_UNLOCK(&ia->ia_ifa);
1584	ia = NULL;
1585	}
1586	lck_rw_done(lck: &in_ifaddr_rwlock);
1587	}
1588	/*
1589	* Otherwise, if the socket has already bound the source, just use it.
1590	*/
1591	if (inp->inp_laddr.s_addr != INADDR_ANY) {
1592	VERIFY(ia == NULL);
1593	*laddr = inp->inp_laddr;
1594	return `0`;
1595	}
1596
1597	/*
1598	* If the ifscope is specified by the caller (e.g. IP_PKTINFO)
1599	* then it overrides the sticky ifscope set for the socket.
1600	*/
1601	if (ifscope == IFSCOPE_NONE && (inp->inp_flags & INP_BOUND_IF)) {
1602	ifscope = inp->inp_boundifp->if_index;
1603	}
1604
1605	/*
1606	* If route is known or can be allocated now,
1607	* our src addr is taken from the i/f, else punt.
1608	* Note that we should check the address family of the cached
1609	* destination, in case of sharing the cache with IPv6.
1610	*/
1611	if (ro->ro_rt != NULL) {
1612	RT_LOCK_SPIN(ro->ro_rt);
1613	}
1614	if (ROUTE_UNUSABLE(ro) \|\| ro->ro_dst.sa_family != AF_INET \|\|
1615	SIN(&ro->ro_dst)->sin_addr.s_addr != SIN(nam)->sin_addr.s_addr \|\|
1616	(inp->inp_socket->so_options & SO_DONTROUTE)) {
1617	if (ro->ro_rt != NULL) {
1618	RT_UNLOCK(ro->ro_rt);
1619	}
1620	ROUTE_RELEASE(ro);
1621	}
1622	if (!(inp->inp_socket->so_options & SO_DONTROUTE) &&
1623	(ro->ro_rt == NULL \|\| ro->ro_rt->rt_ifp == NULL)) {
1624	if (ro->ro_rt != NULL) {
1625	RT_UNLOCK(ro->ro_rt);
1626	}
1627	ROUTE_RELEASE(ro);
1628	/ No route yet, so try to acquire one /
1629	SOCKADDR_ZERO(&ro->ro_dst, sizeof(struct sockaddr_in));
1630	ro->ro_dst.sa_family = AF_INET;
1631	ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
1632	SIN(&ro->ro_dst)->sin_addr = SIN(nam)->sin_addr;
1633	rtalloc_scoped(ro, ifscope);
1634	if (ro->ro_rt != NULL) {
1635	RT_LOCK_SPIN(ro->ro_rt);
1636	}
1637	}
1638	/ Sanitized local copy for interface address searches /
1639	SOCKADDR_ZERO(&sin, sizeof(sin));
1640	sin.sin_family = AF_INET;
1641	sin.sin_len = sizeof(struct sockaddr_in);
1642	sin.sin_addr.s_addr = SIN(nam)->sin_addr.s_addr;
1643	/*
1644	* If we did not find (or use) a route, assume dest is reachable
1645	* on a directly connected network and try to find a corresponding
1646	* interface to take the source address from.
1647	*/
1648	if (ro->ro_rt == NULL) {
1649	proc_t proc = current_proc();
1650
1651	VERIFY(ia == NULL);
1652	ia = ifatoia(ifa_ifwithdstaddr(SA(&sin)));
1653	if (ia == NULL) {
1654	ia = ifatoia(ifa_ifwithnet_scoped(SA(&sin), ifscope));
1655	}
1656	error = ((ia == NULL) ? ENETUNREACH : `0`);
1657
1658	if (apn_fallback_required(proc, so: inp->inp_socket,
1659	p_dstv4: (void *)nam)) {
1660	apn_fallback_trigger(proc, so: inp->inp_socket);
1661	}
1662
1663	goto done;
1664	}
1665	RT_LOCK_ASSERT_HELD(ro->ro_rt);
1666	/*
1667	* If the outgoing interface on the route found is not
1668	* a loopback interface, use the address from that interface.
1669	*/
1670	if (!(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK)) {
1671	VERIFY(ia == NULL);
1672	/*
1673	* If the route points to a cellular interface and the
1674	* caller forbids our using interfaces of such type,
1675	* pretend that there is no route.
1676	* Apply the same logic for expensive interfaces.
1677	*/
1678	if (inp_restricted_send(inp, ro->ro_rt->rt_ifp)) {
1679	RT_UNLOCK(ro->ro_rt);
1680	ROUTE_RELEASE(ro);
1681	error = EHOSTUNREACH;
1682	restricted = TRUE;
1683	} else {
1684	/ Become a regular mutex /
1685	RT_CONVERT_LOCK(ro->ro_rt);
1686	ia = ifatoia(ro->ro_rt->rt_ifa);
1687	ifa_addref(ifa: &ia->ia_ifa);
1688
1689	/*
1690	* Mark the control block for notification of
1691	* a possible flow that might undergo clat46
1692	* translation.
1693	*
1694	* We defer the decision to a later point when
1695	* inpcb is being disposed off.
1696	* The reason is that we only want to send notification
1697	* if the flow was ever used to send data.
1698	*/
1699	if (IS_INTF_CLAT46(ro->ro_rt->rt_ifp)) {
1700	inp->inp_flags2 \|= INP2_CLAT46_FLOW;
1701	}
1702
1703	RT_UNLOCK(ro->ro_rt);
1704	error = `0`;
1705	}
1706	goto done;
1707	}
1708	VERIFY(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK);
1709	RT_UNLOCK(ro->ro_rt);
1710	/*
1711	* The outgoing interface is marked with 'loopback net', so a route
1712	* to ourselves is here.
1713	* Try to find the interface of the destination address and then
1714	* take the address from there. That interface is not necessarily
1715	* a loopback interface.
1716	*/
1717	VERIFY(ia == NULL);
1718	ia = ifatoia(ifa_ifwithdstaddr(SA(&sin)));
1719	if (ia == NULL) {
1720	ia = ifatoia(ifa_ifwithaddr_scoped(SA(&sin), ifscope));
1721	}
1722	if (ia == NULL) {
1723	ia = ifatoia(ifa_ifwithnet_scoped(SA(&sin), ifscope));
1724	}
1725	if (ia == NULL) {
1726	RT_LOCK(ro->ro_rt);
1727	ia = ifatoia(ro->ro_rt->rt_ifa);
1728	if (ia != NULL) {
1729	ifa_addref(ifa: &ia->ia_ifa);
1730	}
1731	RT_UNLOCK(ro->ro_rt);
1732	}
1733	error = ((ia == NULL) ? ENETUNREACH : `0`);
1734
1735	done:
1736	/*
1737	* If the destination address is multicast and an outgoing
1738	* interface has been set as a multicast option, use the
1739	* address of that interface as our source address.
1740	*/
1741	if (IN_MULTICAST(ntohl(SIN(nam)->sin_addr.s_addr)) &&
1742	inp->inp_moptions != NULL) {
1743	struct ip_moptions *imo;
1744	struct ifnet *ifp;
1745
1746	imo = inp->inp_moptions;
1747	IMO_LOCK(imo);
1748	if (imo->imo_multicast_ifp != NULL && (ia == NULL \|\|
1749	ia->ia_ifp != imo->imo_multicast_ifp)) {
1750	ifp = imo->imo_multicast_ifp;
1751	if (ia != NULL) {
1752	ifa_remref(ifa: &ia->ia_ifa);
1753	}
1754	lck_rw_lock_shared(lck: &in_ifaddr_rwlock);
1755	TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
1756	if (ia->ia_ifp == ifp) {
1757	break;
1758	}
1759	}
1760	if (ia != NULL) {
1761	ifa_addref(ifa: &ia->ia_ifa);
1762	}
1763	lck_rw_done(lck: &in_ifaddr_rwlock);
1764	if (ia == NULL) {
1765	error = EADDRNOTAVAIL;
1766	} else {
1767	error = `0`;
1768	}
1769	}
1770	IMO_UNLOCK(imo);
1771	}
1772	/*
1773	* Don't do pcblookup call here; return interface in laddr
1774	* and exit to caller, that will do the lookup.
1775	*/
1776	if (ia != NULL) {
1777	/*
1778	* If the source address belongs to a cellular interface
1779	* and the socket forbids our using interfaces of such
1780	* type, pretend that there is no source address.
1781	* Apply the same logic for expensive interfaces.
1782	*/
1783	IFA_LOCK_SPIN(&ia->ia_ifa);
1784	if (inp_restricted_send(inp, ia->ia_ifa.ifa_ifp)) {
1785	IFA_UNLOCK(&ia->ia_ifa);
1786	error = EHOSTUNREACH;
1787	restricted = TRUE;
1788	} else if (error == `0`) {
1789	*laddr = ia->ia_addr.sin_addr;
1790	if (outif != NULL) {
1791	struct ifnet *ifp;
1792
1793	if (ro->ro_rt != NULL) {
1794	ifp = ro->ro_rt->rt_ifp;
1795	} else {
1796	ifp = ia->ia_ifp;
1797	}
1798
1799	VERIFY(ifp != NULL);
1800	IFA_CONVERT_LOCK(&ia->ia_ifa);
1801	ifnet_reference(interface: ifp); / for caller /
1802	if (*outif != NULL) {
1803	ifnet_release(interface: *outif);
1804	}
1805	*outif = ifp;
1806	}
1807	IFA_UNLOCK(&ia->ia_ifa);
1808	} else {
1809	IFA_UNLOCK(&ia->ia_ifa);
1810	}
1811	ifa_remref(ifa: &ia->ia_ifa);
1812	ia = NULL;
1813	}
1814
1815	if (restricted && error == EHOSTUNREACH) {
1816	soevent(so: inp->inp_socket, hint: (SO_FILT_HINT_LOCKED \|
1817	SO_FILT_HINT_IFDENIED));
1818	}
1819
1820	return error;
1821	}
1822
1823	/*
1824	* Outer subroutine:
1825	* Connect from a socket to a specified address.
1826	* Both address and port must be specified in argument sin.
1827	* If don't have a local address for this socket yet,
1828	* then pick one.
1829	*
1830	* The caller may override the bound-to-interface setting of the socket
1831	* by specifying the ifscope parameter (e.g. from IP_PKTINFO.)
1832	*/
1833	int
1834	in_pcbconnect(struct inpcb inp, struct* sockaddr nam, struct* proc *p,
1835	unsigned int ifscope, struct ifnet **outif)
1836	{
1837	struct in_addr laddr;
1838	struct sockaddr_in *sin = SIN(nam);
1839	struct inpcb *pcb;
1840	int error;
1841	struct socket *so = inp->inp_socket;
1842
1843	#if CONTENT_FILTER
1844	if (so) {
1845	so->so_state_change_cnt++;
1846	}
1847	#endif
1848
1849	/*
1850	* Call inner routine, to assign local interface address.
1851	*/
1852	if ((error = in_pcbladdr(inp, nam, laddr: &laddr, ifscope, outif, raw: `0`)) != `0`) {
1853	return error;
1854	}
1855
1856	socket_unlock(so, refcount: `0`);
1857	pcb = in_pcblookup_hash(inp->inp_pcbinfo, sin->sin_addr, sin->sin_port,
1858	inp->inp_laddr.s_addr ? inp->inp_laddr : laddr,
1859	inp->inp_lport, `0`, NULL);
1860	socket_lock(so, refcount: `0`);
1861
1862	/*
1863	* Check if the socket is still in a valid state. When we unlock this
1864	* embryonic socket, it can get aborted if another thread is closing
1865	* the listener (radar 7947600).
1866	*/
1867	if ((so->so_flags & SOF_ABORTED) != `0`) {
1868	return ECONNREFUSED;
1869	}
1870
1871	if (pcb != NULL) {
1872	in_pcb_checkstate(pcb, WNT_RELEASE, pcb == inp ? `1` : `0`);
1873	return EADDRINUSE;
1874	}
1875	if (inp->inp_laddr.s_addr == INADDR_ANY) {
1876	if (inp->inp_lport == `0`) {
1877	error = in_pcbbind(inp, NULL, p);
1878	if (error) {
1879	return error;
1880	}
1881	}
1882	if (!lck_rw_try_lock_exclusive(lck: &inp->inp_pcbinfo->ipi_lock)) {
1883	/*
1884	* Lock inversion issue, mostly with udp
1885	* multicast packets.
1886	*/
1887	socket_unlock(so, refcount: `0`);
1888	lck_rw_lock_exclusive(lck: &inp->inp_pcbinfo->ipi_lock);
1889	socket_lock(so, refcount: `0`);
1890	}
1891	inp->inp_laddr = laddr;
1892	/ no reference needed /
1893	inp->inp_last_outifp = (outif != NULL) ? *outif : NULL;
1894	#if SKYWALK
1895	if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
1896	netns_set_ifnet(token: &inp->inp_netns_token,
1897	ifp: inp->inp_last_outifp);
1898	}
1899	#endif /* SKYWALK */
1900	inp->inp_flags \|= INP_INADDR_ANY;
1901	} else {
1902	/*
1903	* Usage of IP_PKTINFO, without local port already
1904	* speficified will cause kernel to panic,
1905	* see rdar://problem/18508185.
1906	* For now returning error to avoid a kernel panic
1907	* This routines can be refactored and handle this better
1908	* in future.
1909	*/
1910	if (inp->inp_lport == `0`) {
1911	return EINVAL;
1912	}
1913	if (!lck_rw_try_lock_exclusive(lck: &inp->inp_pcbinfo->ipi_lock)) {
1914	/*
1915	* Lock inversion issue, mostly with udp
1916	* multicast packets.
1917	*/
1918	socket_unlock(so, refcount: `0`);
1919	lck_rw_lock_exclusive(lck: &inp->inp_pcbinfo->ipi_lock);
1920	socket_lock(so, refcount: `0`);
1921	}
1922	}
1923	inp->inp_faddr = sin->sin_addr;
1924	inp->inp_fport = sin->sin_port;
1925	if (nstat_collect && SOCK_PROTO(so) == IPPROTO_UDP) {
1926	nstat_pcb_invalidate_cache(inp);
1927	}
1928	in_pcbrehash(inp);
1929	lck_rw_done(lck: &inp->inp_pcbinfo->ipi_lock);
1930	return `0`;
1931	}
1932
1933	void
1934	in_pcbdisconnect(struct inpcb *inp)
1935	{
1936	struct socket *so = inp->inp_socket;
1937
1938	if (nstat_collect && SOCK_PROTO(so) == IPPROTO_UDP) {
1939	nstat_pcb_cache(inp);
1940	}
1941
1942	inp->inp_faddr.s_addr = INADDR_ANY;
1943	inp->inp_fport = `0`;
1944
1945	#if CONTENT_FILTER
1946	if (so) {
1947	so->so_state_change_cnt++;
1948	}
1949	#endif
1950
1951	if (!lck_rw_try_lock_exclusive(lck: &inp->inp_pcbinfo->ipi_lock)) {
1952	/ lock inversion issue, mostly with udp multicast packets /
1953	socket_unlock(so, refcount: `0`);
1954	lck_rw_lock_exclusive(lck: &inp->inp_pcbinfo->ipi_lock);
1955	socket_lock(so, refcount: `0`);
1956	}
1957
1958	in_pcbrehash(inp);
1959	lck_rw_done(lck: &inp->inp_pcbinfo->ipi_lock);
1960	/*
1961	* A multipath subflow socket would have its SS_NOFDREF set by default,
1962	* so check for SOF_MP_SUBFLOW socket flag before detaching the PCB;
1963	* when the socket is closed for real, SOF_MP_SUBFLOW would be cleared.
1964	*/
1965	if (!(so->so_flags & SOF_MP_SUBFLOW) && (so->so_state & SS_NOFDREF)) {
1966	in_pcbdetach(inp);
1967	}
1968	}
1969
1970	void
1971	in_pcbdetach(struct inpcb *inp)
1972	{
1973	struct socket *so = inp->inp_socket;
1974
1975	if (so->so_pcb == NULL) {
1976	/ PCB has been disposed /
1977	panic("%s: inp=%p so=%p proto=%d so_pcb is null!", __func__,
1978	inp, so, SOCK_PROTO(so));
1979	/ NOTREACHED /
1980	}
1981
1982	#if IPSEC
1983	if (inp->inp_sp != NULL) {
1984	(void) ipsec4_delete_pcbpolicy(inp);
1985	}
1986	#endif /* IPSEC */
1987
1988	if (inp->inp_stat != NULL && SOCK_PROTO(so) == IPPROTO_UDP) {
1989	if (inp->inp_stat->rxpackets == `0` && inp->inp_stat->txpackets == `0`) {
1990	INC_ATOMIC_INT64_LIM(net_api_stats.nas_socket_inet_dgram_no_data);
1991	}
1992	}
1993
1994	/*
1995	* Let NetworkStatistics know this PCB is going away
1996	* before we detach it.
1997	*/
1998	if (nstat_collect &&
1999	(SOCK_PROTO(so) == IPPROTO_TCP \|\| SOCK_PROTO(so) == IPPROTO_UDP)) {
2000	nstat_pcb_detach(inp);
2001	}
2002
2003	/ Free memory buffer held for generating keep alives /
2004	if (inp->inp_keepalive_data != NULL) {
2005	kfree_data(inp->inp_keepalive_data, inp->inp_keepalive_datalen);
2006	inp->inp_keepalive_data = NULL;
2007	}
2008
2009	/ mark socket state as dead /
2010	if (in_pcb_checkstate(inp, WNT_STOPUSING, `1`) != WNT_STOPUSING) {
2011	panic("%s: so=%p proto=%d couldn't set to STOPUSING",
2012	__func__, so, SOCK_PROTO(so));
2013	/ NOTREACHED /
2014	}
2015
2016	#if SKYWALK
2017	/ Free up the port in the namespace registrar if not in TIME_WAIT /
2018	if (!(inp->inp_flags2 & INP2_TIMEWAIT)) {
2019	netns_release(token: &inp->inp_netns_token);
2020	netns_release(token: &inp->inp_wildcard_netns_token);
2021	}
2022	#endif /* SKYWALK */
2023
2024	if (!(so->so_flags & SOF_PCBCLEARING)) {
2025	struct ip_moptions *imo;
2026
2027	inp->inp_vflag = `0`;
2028	if (inp->inp_options != NULL) {
2029	(void) m_free(inp->inp_options);
2030	inp->inp_options = NULL;
2031	}
2032	ROUTE_RELEASE(&inp->inp_route);
2033	imo = inp->inp_moptions;
2034	if (imo != NULL) {
2035	IMO_REMREF(imo);
2036	}
2037	inp->inp_moptions = NULL;
2038	sofreelastref(so, `0`);
2039	inp->inp_state = INPCB_STATE_DEAD;
2040
2041	/*
2042	* Enqueue an event to send kernel event notification
2043	* if the flow has to CLAT46 for data packets
2044	*/
2045	if (inp->inp_flags2 & INP2_CLAT46_FLOW) {
2046	/*
2047	* If there has been any exchange of data bytes
2048	* over this flow.
2049	* Schedule a notification to report that flow is
2050	* using client side translation.
2051	*/
2052	if (inp->inp_stat != NULL &&
2053	(inp->inp_stat->txbytes != `0` \|\|
2054	inp->inp_stat->rxbytes != `0`)) {
2055	if (so->so_flags & SOF_DELEGATED) {
2056	in6_clat46_event_enqueue_nwk_wq_entry(
2057	IN6_CLAT46_EVENT_V4_FLOW,
2058	so->e_pid,
2059	so->e_uuid);
2060	} else {
2061	in6_clat46_event_enqueue_nwk_wq_entry(
2062	IN6_CLAT46_EVENT_V4_FLOW,
2063	so->last_pid,
2064	so->last_uuid);
2065	}
2066	}
2067	}
2068
2069	/ makes sure we're not called twice from so_close /
2070	so->so_flags \|= SOF_PCBCLEARING;
2071
2072	inpcb_gc_sched(ipi: inp->inp_pcbinfo, type: INPCB_TIMER_FAST);
2073	}
2074	}
2075
2076
2077	void
2078	in_pcbdispose(struct inpcb *inp)
2079	{
2080	struct socket *so = inp->inp_socket;
2081	struct inpcbinfo *ipi = inp->inp_pcbinfo;
2082
2083	if (so != NULL && so->so_usecount != `0`) {
2084	panic("%s: so %p [%d,%d] usecount %d lockhistory %s",
2085	__func__, so, SOCK_DOM(so), SOCK_TYPE(so), so->so_usecount,
2086	solockhistory_nr(so));
2087	/ NOTREACHED /
2088	} else if (inp->inp_wantcnt != WNT_STOPUSING) {
2089	if (so != NULL) {
2090	panic_plain("%s: inp %p invalid wantcnt %d, so %p "
2091	"[%d,%d] usecount %d retaincnt %d state 0x%x "
2092	"flags 0x%x lockhistory %s\n", __func__, inp,
2093	inp->inp_wantcnt, so, SOCK_DOM(so), SOCK_TYPE(so),
2094	so->so_usecount, so->so_retaincnt, so->so_state,
2095	so->so_flags, solockhistory_nr(so));
2096	/ NOTREACHED /
2097	} else {
2098	panic("%s: inp %p invalid wantcnt %d no socket",
2099	__func__, inp, inp->inp_wantcnt);
2100	/ NOTREACHED /
2101	}
2102	}
2103
2104	LCK_RW_ASSERT(&ipi->ipi_lock, LCK_RW_ASSERT_EXCLUSIVE);
2105
2106	inp->inp_gencnt = ++ipi->ipi_gencnt;
2107	/ access ipi in in_pcbremlists /
2108	in_pcbremlists(inp);
2109
2110	if (so != NULL) {
2111	if (so->so_proto->pr_flags & PR_PCBLOCK) {
2112	sofreelastref(so, `0`);
2113	if (so->so_rcv.sb_cc > `0` \|\| so->so_snd.sb_cc > `0`) {
2114	/*
2115	* selthreadclear() already called
2116	* during sofreelastref() above.
2117	*/
2118	sbrelease(sb: &so->so_rcv);
2119	sbrelease(sb: &so->so_snd);
2120	}
2121	if (so->so_head != NULL) {
2122	panic("%s: so=%p head still exist",
2123	__func__, so);
2124	/ NOTREACHED /
2125	}
2126	lck_mtx_unlock(lck: &inp->inpcb_mtx);
2127
2128	#if NECP
2129	necp_inpcb_remove_cb(inp);
2130	#endif /* NECP */
2131
2132	lck_mtx_destroy(lck: &inp->inpcb_mtx, grp: ipi->ipi_lock_grp);
2133	}
2134	/ makes sure we're not called twice from so_close /
2135	so->so_flags \|= SOF_PCBCLEARING;
2136	so->so_saved_pcb = (caddr_t)inp;
2137	so->so_pcb = NULL;
2138	inp->inp_socket = NULL;
2139	#if NECP
2140	necp_inpcb_dispose(inp);
2141	#endif /* NECP */
2142	/*
2143	* In case there a route cached after a detach (possible
2144	* in the tcp case), make sure that it is freed before
2145	* we deallocate the structure.
2146	*/
2147	ROUTE_RELEASE(&inp->inp_route);
2148	if ((so->so_flags1 & SOF1_CACHED_IN_SOCK_LAYER) == `0`) {
2149	zfree(ipi->ipi_zone, inp);
2150	}
2151	sodealloc(so);
2152	}
2153	}
2154
2155	/*
2156	* The calling convention of in_getsockaddr() and in_getpeeraddr() was
2157	* modified to match the pru_sockaddr() and pru_peeraddr() entry points
2158	* in struct pr_usrreqs, so that protocols can just reference then directly
2159	* without the need for a wrapper function.
2160	*/
2161	int
2162	in_getsockaddr(struct socket so, struct* sockaddr **nam)
2163	{
2164	struct inpcb *inp;
2165	struct sockaddr_in *sin;
2166
2167	/*
2168	* Do the malloc first in case it blocks.
2169	*/
2170	sin = SIN(alloc_sockaddr(sizeof(*sin),
2171	Z_WAITOK \| Z_NOFAIL));
2172
2173	sin->sin_family = AF_INET;
2174
2175	if ((inp = sotoinpcb(so)) == NULL) {
2176	free_sockaddr(sin);
2177	return EINVAL;
2178	}
2179	sin->sin_port = inp->inp_lport;
2180	sin->sin_addr = inp->inp_laddr;
2181
2182	*nam = SA(sin);
2183	return `0`;
2184	}
2185
2186	int
2187	in_getsockaddr_s(struct socket so, struct* sockaddr_in *ss)
2188	{
2189	struct sockaddr_in *sin = ss;
2190	struct inpcb *inp;
2191
2192	VERIFY(ss != NULL);
2193	SOCKADDR_ZERO(ss, sizeof(*ss));
2194
2195	sin->sin_family = AF_INET;
2196	sin->sin_len = sizeof(*sin);
2197
2198	if ((inp = sotoinpcb(so)) == NULL) {
2199	return EINVAL;
2200	}
2201
2202	sin->sin_port = inp->inp_lport;
2203	sin->sin_addr = inp->inp_laddr;
2204	return `0`;
2205	}
2206
2207	int
2208	in_getpeeraddr(struct socket so, struct* sockaddr **nam)
2209	{
2210	struct inpcb *inp;
2211	struct sockaddr_in *sin;
2212
2213	/*
2214	* Do the malloc first in case it blocks.
2215	*/
2216	sin = SIN(alloc_sockaddr(sizeof(*sin),
2217	Z_WAITOK \| Z_NOFAIL));
2218
2219	sin->sin_family = AF_INET;
2220
2221	if ((inp = sotoinpcb(so)) == NULL) {
2222	free_sockaddr(sin);
2223	return EINVAL;
2224	}
2225	sin->sin_port = inp->inp_fport;
2226	sin->sin_addr = inp->inp_faddr;
2227
2228	*nam = SA(sin);
2229	return `0`;
2230	}
2231
2232	void
2233	in_pcbnotifyall(struct inpcbinfo pcbinfo, struct* in_addr faddr,
2234	int errno, void (notify)(struct* inpcb , int*))
2235	{
2236	struct inpcb *inp;
2237
2238	lck_rw_lock_shared(lck: &pcbinfo->ipi_lock);
2239
2240	LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) {
2241	if (!(inp->inp_vflag & INP_IPV4)) {
2242	continue;
2243	}
2244	if (inp->inp_faddr.s_addr != faddr.s_addr \|\|
2245	inp->inp_socket == NULL) {
2246	continue;
2247	}
2248	if (in_pcb_checkstate(inp, WNT_ACQUIRE, `0`) == WNT_STOPUSING) {
2249	continue;
2250	}
2251	socket_lock(so: inp->inp_socket, refcount: `1`);
2252	(*notify)(inp, errno);
2253	(void) in_pcb_checkstate(inp, WNT_RELEASE, `1`);
2254	socket_unlock(so: inp->inp_socket, refcount: `1`);
2255	}
2256	lck_rw_done(lck: &pcbinfo->ipi_lock);
2257	}
2258
2259	/*
2260	* Check for alternatives when higher level complains
2261	* about service problems. For now, invalidate cached
2262	* routing information. If the route was created dynamically
2263	* (by a redirect), time to try a default gateway again.
2264	*/
2265	void
2266	in_losing(struct inpcb *inp)
2267	{
2268	boolean_t release = FALSE;
2269	struct rtentry *rt;
2270
2271	if ((rt = inp->inp_route.ro_rt) != NULL) {
2272	struct in_ifaddr *ia = NULL;
2273
2274	RT_LOCK(rt);
2275	if (rt->rt_flags & RTF_DYNAMIC) {
2276	/*
2277	* Prevent another thread from modifying rt_key,
2278	* rt_gateway via rt_setgate() after rt_lock is
2279	* dropped by marking the route as defunct.
2280	*/
2281	rt->rt_flags \|= RTF_CONDEMNED;
2282	RT_UNLOCK(rt);
2283	(void) rtrequest(RTM_DELETE, rt_key(rt),
2284	rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
2285	} else {
2286	RT_UNLOCK(rt);
2287	}
2288	/ if the address is gone keep the old route in the pcb /
2289	if (inp->inp_laddr.s_addr != INADDR_ANY &&
2290	(ia = ifa_foraddr(inp->inp_laddr.s_addr)) != NULL) {
2291	/*
2292	* Address is around; ditch the route. A new route
2293	* can be allocated the next time output is attempted.
2294	*/
2295	release = TRUE;
2296	}
2297	if (ia != NULL) {
2298	ifa_remref(ifa: &ia->ia_ifa);
2299	}
2300	}
2301	if (rt == NULL \|\| release) {
2302	ROUTE_RELEASE(&inp->inp_route);
2303	}
2304	}
2305
2306	/*
2307	* After a routing change, flush old routing
2308	* and allocate a (hopefully) better one.
2309	*/
2310	void
2311	in_rtchange(struct inpcb inp, int* errno)
2312	{
2313	#pragma unused(errno)
2314	boolean_t release = FALSE;
2315	struct rtentry *rt;
2316
2317	if ((rt = inp->inp_route.ro_rt) != NULL) {
2318	struct in_ifaddr *ia = NULL;
2319
2320	/ if address is gone, keep the old route /
2321	if (inp->inp_laddr.s_addr != INADDR_ANY &&
2322	(ia = ifa_foraddr(inp->inp_laddr.s_addr)) != NULL) {
2323	/*
2324	* Address is around; ditch the route. A new route
2325	* can be allocated the next time output is attempted.
2326	*/
2327	release = TRUE;
2328	}
2329	if (ia != NULL) {
2330	ifa_remref(ifa: &ia->ia_ifa);
2331	}
2332	}
2333	if (rt == NULL \|\| release) {
2334	ROUTE_RELEASE(&inp->inp_route);
2335	}
2336	}
2337
2338	/*
2339	* Lookup a PCB based on the local address and port.
2340	*/
2341	struct inpcb *
2342	in_pcblookup_local(struct inpcbinfo pcbinfo, struct* in_addr laddr,
2343	unsigned int lport_arg, int wild_okay)
2344	{
2345	struct inpcb *inp;
2346	int matchwild = `3`, wildcard;
2347	u_short lport = (u_short)lport_arg;
2348
2349	KERNEL_DEBUG(DBG_FNC_PCB_LOOKUP \| DBG_FUNC_START, `0`, `0`, `0`, `0`, `0`);
2350
2351	if (!wild_okay) {
2352	struct inpcbhead *head;
2353	/*
2354	* Look for an unconnected (wildcard foreign addr) PCB that
2355	* matches the local address and port we're looking for.
2356	*/
2357	head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, `0`,
2358	pcbinfo->ipi_hashmask)];
2359	LIST_FOREACH(inp, head, inp_hash) {
2360	if (!(inp->inp_vflag & INP_IPV4)) {
2361	continue;
2362	}
2363	if (inp->inp_faddr.s_addr == INADDR_ANY &&
2364	inp->inp_laddr.s_addr == laddr.s_addr &&
2365	inp->inp_lport == lport) {
2366	/*
2367	* Found.
2368	*/
2369	return inp;
2370	}
2371	}
2372	/*
2373	* Not found.
2374	*/
2375	KERNEL_DEBUG(DBG_FNC_PCB_LOOKUP \| DBG_FUNC_END, `0`, `0`, `0`, `0`, `0`);
2376	return NULL;
2377	} else {
2378	struct inpcbporthead *porthash;
2379	struct inpcbport *phd;
2380	struct inpcb *match = NULL;
2381	/*
2382	* Best fit PCB lookup.
2383	*
2384	* First see if this local port is in use by looking on the
2385	* port hash list.
2386	*/
2387	porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
2388	pcbinfo->ipi_porthashmask)];
2389	LIST_FOREACH(phd, porthash, phd_hash) {
2390	if (phd->phd_port == lport) {
2391	break;
2392	}
2393	}
2394	if (phd != NULL) {
2395	/*
2396	* Port is in use by one or more PCBs. Look for best
2397	* fit.
2398	*/
2399	LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
2400	wildcard = `0`;
2401	if (!(inp->inp_vflag & INP_IPV4)) {
2402	continue;
2403	}
2404	if (inp->inp_faddr.s_addr != INADDR_ANY) {
2405	wildcard++;
2406	}
2407	if (inp->inp_laddr.s_addr != INADDR_ANY) {
2408	if (laddr.s_addr == INADDR_ANY) {
2409	wildcard++;
2410	} else if (inp->inp_laddr.s_addr !=
2411	laddr.s_addr) {
2412	continue;
2413	}
2414	} else {
2415	if (laddr.s_addr != INADDR_ANY) {
2416	wildcard++;
2417	}
2418	}
2419	if (wildcard < matchwild) {
2420	match = inp;
2421	matchwild = wildcard;
2422	if (matchwild == `0`) {
2423	break;
2424	}
2425	}
2426	}
2427	}
2428	KERNEL_DEBUG(DBG_FNC_PCB_LOOKUP \| DBG_FUNC_END, match,
2429	`0`, `0`, `0`, `0`);
2430	return match;
2431	}
2432	}
2433
2434	/*
2435	* Check if PCB exists in hash list.
2436	*/
2437	int
2438	in_pcblookup_hash_exists(struct inpcbinfo pcbinfo, struct* in_addr faddr,
2439	u_int fport_arg, struct in_addr laddr, u_int lport_arg, int wildcard,
2440	uid_t uid, gid_t gid, struct ifnet *ifp)
2441	{
2442	struct inpcbhead *head;
2443	struct inpcb *inp;
2444	u_short fport = (u_short)fport_arg, lport = (u_short)lport_arg;
2445	int found = `0`;
2446	struct inpcb *local_wild = NULL;
2447	struct inpcb *local_wild_mapped = NULL;
2448
2449	*uid = UID_MAX;
2450	*gid = GID_MAX;
2451
2452	/*
2453	* We may have found the pcb in the last lookup - check this first.
2454	*/
2455
2456	lck_rw_lock_shared(lck: &pcbinfo->ipi_lock);
2457
2458	/*
2459	* First look for an exact match.
2460	*/
2461	head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
2462	pcbinfo->ipi_hashmask)];
2463	LIST_FOREACH(inp, head, inp_hash) {
2464	if (!(inp->inp_vflag & INP_IPV4)) {
2465	continue;
2466	}
2467	if (inp_restricted_recv(inp, ifp)) {
2468	continue;
2469	}
2470
2471	#if NECP
2472	if (!necp_socket_is_allowed_to_recv_on_interface(inp, interface: ifp)) {
2473	continue;
2474	}
2475	#endif /* NECP */
2476
2477	if (inp->inp_faddr.s_addr == faddr.s_addr &&
2478	inp->inp_laddr.s_addr == laddr.s_addr &&
2479	inp->inp_fport == fport &&
2480	inp->inp_lport == lport) {
2481	if ((found = (inp->inp_socket != NULL))) {
2482	/*
2483	* Found.
2484	*/
2485	*uid = kauth_cred_getuid(
2486	cred: inp->inp_socket->so_cred);
2487	*gid = kauth_cred_getgid(
2488	cred: inp->inp_socket->so_cred);
2489	}
2490	lck_rw_done(lck: &pcbinfo->ipi_lock);
2491	return found;
2492	}
2493	}
2494
2495	if (!wildcard) {
2496	/*
2497	* Not found.
2498	*/
2499	lck_rw_done(lck: &pcbinfo->ipi_lock);
2500	return `0`;
2501	}
2502
2503	head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, `0`,
2504	pcbinfo->ipi_hashmask)];
2505	LIST_FOREACH(inp, head, inp_hash) {
2506	if (!(inp->inp_vflag & INP_IPV4)) {
2507	continue;
2508	}
2509	if (inp_restricted_recv(inp, ifp)) {
2510	continue;
2511	}
2512
2513	#if NECP
2514	if (!necp_socket_is_allowed_to_recv_on_interface(inp, interface: ifp)) {
2515	continue;
2516	}
2517	#endif /* NECP */
2518
2519	if (inp->inp_faddr.s_addr == INADDR_ANY &&
2520	inp->inp_lport == lport) {
2521	if (inp->inp_laddr.s_addr == laddr.s_addr) {
2522	if ((found = (inp->inp_socket != NULL))) {
2523	*uid = kauth_cred_getuid(
2524	cred: inp->inp_socket->so_cred);
2525	*gid = kauth_cred_getgid(
2526	cred: inp->inp_socket->so_cred);
2527	}
2528	lck_rw_done(lck: &pcbinfo->ipi_lock);
2529	return found;
2530	} else if (inp->inp_laddr.s_addr == INADDR_ANY) {
2531	if (inp->inp_socket &&
2532	SOCK_CHECK_DOM(inp->inp_socket, PF_INET6)) {
2533	local_wild_mapped = inp;
2534	} else {
2535	local_wild = inp;
2536	}
2537	}
2538	}
2539	}
2540	if (local_wild == NULL) {
2541	if (local_wild_mapped != NULL) {
2542	if ((found = (local_wild_mapped->inp_socket != NULL))) {
2543	*uid = kauth_cred_getuid(
2544	cred: local_wild_mapped->inp_socket->so_cred);
2545	*gid = kauth_cred_getgid(
2546	cred: local_wild_mapped->inp_socket->so_cred);
2547	}
2548	lck_rw_done(lck: &pcbinfo->ipi_lock);
2549	return found;
2550	}
2551	lck_rw_done(lck: &pcbinfo->ipi_lock);
2552	return `0`;
2553	}
2554	if ((found = (local_wild->inp_socket != NULL))) {
2555	*uid = kauth_cred_getuid(
2556	cred: local_wild->inp_socket->so_cred);
2557	*gid = kauth_cred_getgid(
2558	cred: local_wild->inp_socket->so_cred);
2559	}
2560	lck_rw_done(lck: &pcbinfo->ipi_lock);
2561	return found;
2562	}
2563
2564	/*
2565	* Lookup PCB in hash list.
2566	*/
2567	struct inpcb *
2568	in_pcblookup_hash(struct inpcbinfo pcbinfo, struct* in_addr faddr,
2569	u_int fport_arg, struct in_addr laddr, u_int lport_arg, int wildcard,
2570	struct ifnet *ifp)
2571	{
2572	struct inpcbhead *head;
2573	struct inpcb *inp;
2574	u_short fport = (u_short)fport_arg, lport = (u_short)lport_arg;
2575	struct inpcb *local_wild = NULL;
2576	struct inpcb *local_wild_mapped = NULL;
2577
2578	/*
2579	* We may have found the pcb in the last lookup - check this first.
2580	*/
2581
2582	lck_rw_lock_shared(lck: &pcbinfo->ipi_lock);
2583
2584	/*
2585	* First look for an exact match.
2586	*/
2587	head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
2588	pcbinfo->ipi_hashmask)];
2589	LIST_FOREACH(inp, head, inp_hash) {
2590	if (!(inp->inp_vflag & INP_IPV4)) {
2591	continue;
2592	}
2593	if (inp_restricted_recv(inp, ifp)) {
2594	continue;
2595	}
2596
2597	#if NECP
2598	if (!necp_socket_is_allowed_to_recv_on_interface(inp, interface: ifp)) {
2599	continue;
2600	}
2601	#endif /* NECP */
2602
2603	if (inp->inp_faddr.s_addr == faddr.s_addr &&
2604	inp->inp_laddr.s_addr == laddr.s_addr &&
2605	inp->inp_fport == fport &&
2606	inp->inp_lport == lport) {
2607	/*
2608	* Found.
2609	*/
2610	if (in_pcb_checkstate(inp, WNT_ACQUIRE, `0`) !=
2611	WNT_STOPUSING) {
2612	lck_rw_done(lck: &pcbinfo->ipi_lock);
2613	return inp;
2614	} else {
2615	/ it's there but dead, say it isn't found /
2616	lck_rw_done(lck: &pcbinfo->ipi_lock);
2617	return NULL;
2618	}
2619	}
2620	}
2621
2622	if (!wildcard) {
2623	/*
2624	* Not found.
2625	*/
2626	lck_rw_done(lck: &pcbinfo->ipi_lock);
2627	return NULL;
2628	}
2629
2630	head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, `0`,
2631	pcbinfo->ipi_hashmask)];
2632	LIST_FOREACH(inp, head, inp_hash) {
2633	if (!(inp->inp_vflag & INP_IPV4)) {
2634	continue;
2635	}
2636	if (inp_restricted_recv(inp, ifp)) {
2637	continue;
2638	}
2639
2640	#if NECP
2641	if (!necp_socket_is_allowed_to_recv_on_interface(inp, interface: ifp)) {
2642	continue;
2643	}
2644	#endif /* NECP */
2645
2646	if (inp->inp_faddr.s_addr == INADDR_ANY &&
2647	inp->inp_lport == lport) {
2648	if (inp->inp_laddr.s_addr == laddr.s_addr) {
2649	if (in_pcb_checkstate(inp, WNT_ACQUIRE, `0`) !=
2650	WNT_STOPUSING) {
2651	lck_rw_done(lck: &pcbinfo->ipi_lock);
2652	return inp;
2653	} else {
2654	/ it's dead; say it isn't found /
2655	lck_rw_done(lck: &pcbinfo->ipi_lock);
2656	return NULL;
2657	}
2658	} else if (inp->inp_laddr.s_addr == INADDR_ANY) {
2659	if (SOCK_CHECK_DOM(inp->inp_socket, PF_INET6)) {
2660	local_wild_mapped = inp;
2661	} else {
2662	local_wild = inp;
2663	}
2664	}
2665	}
2666	}
2667	if (local_wild == NULL) {
2668	if (local_wild_mapped != NULL) {
2669	if (in_pcb_checkstate(local_wild_mapped,
2670	WNT_ACQUIRE, `0`) != WNT_STOPUSING) {
2671	lck_rw_done(lck: &pcbinfo->ipi_lock);
2672	return local_wild_mapped;
2673	} else {
2674	/ it's dead; say it isn't found /
2675	lck_rw_done(lck: &pcbinfo->ipi_lock);
2676	return NULL;
2677	}
2678	}
2679	lck_rw_done(lck: &pcbinfo->ipi_lock);
2680	return NULL;
2681	}
2682	if (in_pcb_checkstate(local_wild, WNT_ACQUIRE, `0`) != WNT_STOPUSING) {
2683	lck_rw_done(lck: &pcbinfo->ipi_lock);
2684	return local_wild;
2685	}
2686	/*
2687	* It's either not found or is already dead.
2688	*/
2689	lck_rw_done(lck: &pcbinfo->ipi_lock);
2690	return NULL;
2691	}
2692
2693	/*
2694	* @brief Insert PCB onto various hash lists.
2695	*
2696	* @param inp Pointer to internet protocol control block
2697	* @param locked Implies if ipi_lock (protecting pcb list)
2698	* is already locked or not.
2699	*
2700	* @return int error on failure and 0 on success
2701	*/
2702	int
2703	in_pcbinshash(struct inpcb inp, int* locked)
2704	{
2705	struct inpcbhead *pcbhash;
2706	struct inpcbporthead *pcbporthash;
2707	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
2708	struct inpcbport *phd;
2709	u_int32_t hashkey_faddr;
2710
2711	if (!locked) {
2712	if (!lck_rw_try_lock_exclusive(lck: &pcbinfo->ipi_lock)) {
2713	/*
2714	* Lock inversion issue, mostly with udp
2715	* multicast packets
2716	*/
2717	socket_unlock(so: inp->inp_socket, refcount: `0`);
2718	lck_rw_lock_exclusive(lck: &pcbinfo->ipi_lock);
2719	socket_lock(so: inp->inp_socket, refcount: `0`);
2720	}
2721	}
2722
2723	/*
2724	* This routine or its caller may have given up
2725	* socket's protocol lock briefly.
2726	* During that time the socket may have been dropped.
2727	* Safe-guarding against that.
2728	*/
2729	if (inp->inp_state == INPCB_STATE_DEAD) {
2730	if (!locked) {
2731	lck_rw_done(lck: &pcbinfo->ipi_lock);
2732	}
2733	return ECONNABORTED;
2734	}
2735
2736
2737	if (inp->inp_vflag & INP_IPV6) {
2738	hashkey_faddr = inp->in6p_faddr.s6_addr32[`3`] / XXX /;
2739	} else {
2740	hashkey_faddr = inp->inp_faddr.s_addr;
2741	}
2742
2743	inp->inp_hash_element = INP_PCBHASH(hashkey_faddr, inp->inp_lport,
2744	inp->inp_fport, pcbinfo->ipi_hashmask);
2745
2746	pcbhash = &pcbinfo->ipi_hashbase[inp->inp_hash_element];
2747
2748	pcbporthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(inp->inp_lport,
2749	pcbinfo->ipi_porthashmask)];
2750
2751	/*
2752	* Go through port list and look for a head for this lport.
2753	*/
2754	LIST_FOREACH(phd, pcbporthash, phd_hash) {
2755	if (phd->phd_port == inp->inp_lport) {
2756	break;
2757	}
2758	}
2759
2760	/*
2761	* If none exists, malloc one and tack it on.
2762	*/
2763	if (phd == NULL) {
2764	phd = kalloc_type(struct inpcbport, Z_WAITOK \| Z_NOFAIL);
2765	phd->phd_port = inp->inp_lport;
2766	LIST_INIT(&phd->phd_pcblist);
2767	LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
2768	}
2769
2770	VERIFY(!(inp->inp_flags2 & INP2_INHASHLIST));
2771
2772	#if SKYWALK
2773	int err;
2774	struct socket *so = inp->inp_socket;
2775	if ((SOCK_PROTO(so) == IPPROTO_TCP \|\| SOCK_PROTO(so) == IPPROTO_UDP) &&
2776	!(inp->inp_flags2 & INP2_EXTERNAL_PORT)) {
2777	if (inp->inp_vflag & INP_IPV6) {
2778	err = netns_reserve_in6(token: &inp->inp_netns_token,
2779	addr: inp->in6p_laddr, proto: (uint8_t)SOCK_PROTO(so), port: inp->inp_lport,
2780	NETNS_BSD \| NETNS_PRERESERVED, NULL);
2781	} else {
2782	err = netns_reserve_in(token: &inp->inp_netns_token,
2783	addr: inp->inp_laddr, proto: (uint8_t)SOCK_PROTO(so), port: inp->inp_lport,
2784	NETNS_BSD \| NETNS_PRERESERVED, NULL);
2785	}
2786	if (err) {
2787	if (!locked) {
2788	lck_rw_done(lck: &pcbinfo->ipi_lock);
2789	}
2790	return err;
2791	}
2792	netns_set_ifnet(token: &inp->inp_netns_token, ifp: inp->inp_last_outifp);
2793	inp_update_netns_flags(so);
2794	}
2795	#endif /* SKYWALK */
2796
2797	inp->inp_phd = phd;
2798	LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
2799	LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
2800	inp->inp_flags2 \|= INP2_INHASHLIST;
2801
2802	if (!locked) {
2803	lck_rw_done(lck: &pcbinfo->ipi_lock);
2804	}
2805
2806	#if NECP
2807	// This call catches the original setting of the local address
2808	inp_update_necp_policy(inp, NULL, NULL, `0`);
2809	#endif /* NECP */
2810
2811	return `0`;
2812	}
2813
2814	/*
2815	* Move PCB to the proper hash bucket when { faddr, fport } have been
2816	* changed. NOTE: This does not handle the case of the lport changing (the
2817	* hashed port list would have to be updated as well), so the lport must
2818	* not change after in_pcbinshash() has been called.
2819	*/
2820	void
2821	in_pcbrehash(struct inpcb *inp)
2822	{
2823	struct inpcbhead *head;
2824	u_int32_t hashkey_faddr;
2825
2826	#if SKYWALK
2827	struct socket *so = inp->inp_socket;
2828	if ((SOCK_PROTO(so) == IPPROTO_TCP \|\| SOCK_PROTO(so) == IPPROTO_UDP) &&
2829	!(inp->inp_flags2 & INP2_EXTERNAL_PORT)) {
2830	int err;
2831	if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
2832	if (inp->inp_vflag & INP_IPV6) {
2833	err = netns_change_addr_in6(
2834	token: &inp->inp_netns_token, addr: inp->in6p_laddr);
2835	} else {
2836	err = netns_change_addr_in(
2837	token: &inp->inp_netns_token, addr: inp->inp_laddr);
2838	}
2839	} else {
2840	if (inp->inp_vflag & INP_IPV6) {
2841	err = netns_reserve_in6(token: &inp->inp_netns_token,
2842	addr: inp->in6p_laddr, proto: (uint8_t)SOCK_PROTO(so),
2843	port: inp->inp_lport, NETNS_BSD, NULL);
2844	} else {
2845	err = netns_reserve_in(token: &inp->inp_netns_token,
2846	addr: inp->inp_laddr, proto: (uint8_t)SOCK_PROTO(so),
2847	port: inp->inp_lport, NETNS_BSD, NULL);
2848	}
2849	}
2850	/ We are assuming that whatever code paths result in a rehash*
2851	* did their due diligence and ensured that the given
2852	* <proto, laddr, lport> tuple was free ahead of time. Just
2853	* reserving the lport on INADDR_ANY should be enough, since
2854	* that will block Skywalk from trying to reserve that same
2855	* port. Given this assumption, the above netns calls should
2856	* never fail*/
2857	VERIFY(err == `0`);
2858
2859	netns_set_ifnet(token: &inp->inp_netns_token, ifp: inp->inp_last_outifp);
2860	inp_update_netns_flags(so);
2861	}
2862	#endif /* SKYWALK */
2863	if (inp->inp_vflag & INP_IPV6) {
2864	hashkey_faddr = inp->in6p_faddr.s6_addr32[`3`] / XXX /;
2865	} else {
2866	hashkey_faddr = inp->inp_faddr.s_addr;
2867	}
2868
2869	inp->inp_hash_element = INP_PCBHASH(hashkey_faddr, inp->inp_lport,
2870	inp->inp_fport, inp->inp_pcbinfo->ipi_hashmask);
2871	head = &inp->inp_pcbinfo->ipi_hashbase[inp->inp_hash_element];
2872
2873	if (inp->inp_flags2 & INP2_INHASHLIST) {
2874	LIST_REMOVE(inp, inp_hash);
2875	inp->inp_flags2 &= ~INP2_INHASHLIST;
2876	}
2877
2878	VERIFY(!(inp->inp_flags2 & INP2_INHASHLIST));
2879	LIST_INSERT_HEAD(head, inp, inp_hash);
2880	inp->inp_flags2 \|= INP2_INHASHLIST;
2881
2882	#if NECP
2883	// This call catches updates to the remote addresses
2884	inp_update_necp_policy(inp, NULL, NULL, `0`);
2885	#endif /* NECP */
2886	}
2887
2888	/*
2889	* Remove PCB from various lists.
2890	* Must be called pcbinfo lock is held in exclusive mode.
2891	*/
2892	void
2893	in_pcbremlists(struct inpcb *inp)
2894	{
2895	inp->inp_gencnt = ++inp->inp_pcbinfo->ipi_gencnt;
2896
2897	/*
2898	* Check if it's in hashlist -- an inp is placed in hashlist when
2899	* it's local port gets assigned. So it should also be present
2900	* in the port list.
2901	*/
2902	if (inp->inp_flags2 & INP2_INHASHLIST) {
2903	struct inpcbport *phd = inp->inp_phd;
2904
2905	VERIFY(phd != NULL && inp->inp_lport > `0`);
2906
2907	LIST_REMOVE(inp, inp_hash);
2908	inp->inp_hash.le_next = NULL;
2909	inp->inp_hash.le_prev = NULL;
2910
2911	LIST_REMOVE(inp, inp_portlist);
2912	inp->inp_portlist.le_next = NULL;
2913	inp->inp_portlist.le_prev = NULL;
2914	if (LIST_EMPTY(&phd->phd_pcblist)) {
2915	LIST_REMOVE(phd, phd_hash);
2916	kfree_type(struct inpcbport, phd);
2917	}
2918	inp->inp_phd = NULL;
2919	inp->inp_flags2 &= ~INP2_INHASHLIST;
2920	#if SKYWALK
2921	/ Free up the port in the namespace registrar /
2922	netns_release(token: &inp->inp_netns_token);
2923	netns_release(token: &inp->inp_wildcard_netns_token);
2924	#endif /* SKYWALK */
2925	}
2926	VERIFY(!(inp->inp_flags2 & INP2_INHASHLIST));
2927
2928	if (inp->inp_flags2 & INP2_TIMEWAIT) {
2929	/ Remove from time-wait queue /
2930	tcp_remove_from_time_wait(inp);
2931	inp->inp_flags2 &= ~INP2_TIMEWAIT;
2932	VERIFY(inp->inp_pcbinfo->ipi_twcount != `0`);
2933	inp->inp_pcbinfo->ipi_twcount--;
2934	} else {
2935	/ Remove from global inp list if it is not time-wait /
2936	LIST_REMOVE(inp, inp_list);
2937	}
2938
2939	if (inp->inp_flags2 & INP2_IN_FCTREE) {
2940	inp_fc_getinp(inp->inp_flowhash, (INPFC_SOLOCKED \| INPFC_REMOVE));
2941	VERIFY(!(inp->inp_flags2 & INP2_IN_FCTREE));
2942	}
2943
2944	inp->inp_pcbinfo->ipi_count--;
2945	}
2946
2947	/*
2948	* Mechanism used to defer the memory release of PCBs
2949	* The pcb list will contain the pcb until the reaper can clean it up if
2950	* the following conditions are met:
2951	* 1) state "DEAD",
2952	* 2) wantcnt is STOPUSING
2953	* 3) usecount is 0
2954	* This function will be called to either mark the pcb as
2955	*/
2956	int
2957	in_pcb_checkstate(struct inpcb pcb, int* mode, int locked)
2958	{
2959	volatile UInt32 wantcnt = (volatile* UInt32 *)&pcb->inp_wantcnt;
2960	UInt32 origwant;
2961	UInt32 newwant;
2962
2963	switch (mode) {
2964	case WNT_STOPUSING:
2965	/*
2966	* Try to mark the pcb as ready for recycling. CAS with
2967	* STOPUSING, if success we're good, if it's in use, will
2968	* be marked later
2969	*/
2970	if (locked == `0`) {
2971	socket_lock(so: pcb->inp_socket, refcount: `1`);
2972	}
2973	pcb->inp_state = INPCB_STATE_DEAD;
2974
2975	stopusing:
2976	if (pcb->inp_socket->so_usecount < `0`) {
2977	panic("%s: pcb=%p so=%p usecount is negative",
2978	__func__, pcb, pcb->inp_socket);
2979	/ NOTREACHED /
2980	}
2981	if (locked == `0`) {
2982	socket_unlock(so: pcb->inp_socket, refcount: `1`);
2983	}
2984
2985	inpcb_gc_sched(ipi: pcb->inp_pcbinfo, type: INPCB_TIMER_FAST);
2986
2987	origwant = *wantcnt;
2988	if ((UInt16) origwant == `0xffff`) { / should stop using /
2989	return WNT_STOPUSING;
2990	}
2991	newwant = `0xffff`;
2992	if ((UInt16) origwant == `0`) {
2993	/ try to mark it as unsuable now /
2994	OSCompareAndSwap(origwant, newwant, wantcnt);
2995	}
2996	return WNT_STOPUSING;
2997
2998	case WNT_ACQUIRE:
2999	/*
3000	* Try to increase reference to pcb. If WNT_STOPUSING
3001	* should bail out. If socket state DEAD, try to set count
3002	* to STOPUSING, return failed otherwise increase cnt.
3003	*/
3004	do {
3005	origwant = *wantcnt;
3006	if ((UInt16) origwant == `0xffff`) {
3007	/ should stop using /
3008	return WNT_STOPUSING;
3009	}
3010	newwant = origwant + `1`;
3011	} while (!OSCompareAndSwap(origwant, newwant, wantcnt));
3012	return WNT_ACQUIRE;
3013
3014	case WNT_RELEASE:
3015	/*
3016	* Release reference. If result is null and pcb state
3017	* is DEAD, set wanted bit to STOPUSING
3018	*/
3019	if (locked == `0`) {
3020	socket_lock(so: pcb->inp_socket, refcount: `1`);
3021	}
3022
3023	do {
3024	origwant = *wantcnt;
3025	if ((UInt16) origwant == `0x0`) {
3026	panic("%s: pcb=%p release with zero count",
3027	__func__, pcb);
3028	/ NOTREACHED /
3029	}
3030	if ((UInt16) origwant == `0xffff`) {
3031	/ should stop using /
3032	if (locked == `0`) {
3033	socket_unlock(so: pcb->inp_socket, refcount: `1`);
3034	}
3035	return WNT_STOPUSING;
3036	}
3037	newwant = origwant - `1`;
3038	} while (!OSCompareAndSwap(origwant, newwant, wantcnt));
3039
3040	if (pcb->inp_state == INPCB_STATE_DEAD) {
3041	goto stopusing;
3042	}
3043	if (pcb->inp_socket->so_usecount < `0`) {
3044	panic("%s: RELEASE pcb=%p so=%p usecount is negative",
3045	__func__, pcb, pcb->inp_socket);
3046	/ NOTREACHED /
3047	}
3048
3049	if (locked == `0`) {
3050	socket_unlock(so: pcb->inp_socket, refcount: `1`);
3051	}
3052	return WNT_RELEASE;
3053
3054	default:
3055	panic("%s: so=%p not a valid state =%x", __func__,
3056	pcb->inp_socket, mode);
3057	/ NOTREACHED /
3058	}
3059
3060	/ NOTREACHED /
3061	return mode;
3062	}
3063
3064	/*
3065	* inpcb_to_compat copies specific bits of an inpcb to a inpcb_compat.
3066	* The inpcb_compat data structure is passed to user space and must
3067	* not change. We intentionally avoid copying pointers.
3068	*/
3069	void
3070	inpcb_to_compat(struct inpcb inp, struct* inpcb_compat *inp_compat)
3071	{
3072	bzero(s: inp_compat, n: sizeof(*inp_compat));
3073	inp_compat->inp_fport = inp->inp_fport;
3074	inp_compat->inp_lport = inp->inp_lport;
3075	inp_compat->nat_owner = `0`;
3076	inp_compat->nat_cookie = `0`;
3077	inp_compat->inp_gencnt = inp->inp_gencnt;
3078	inp_compat->inp_flags = inp->inp_flags;
3079	inp_compat->inp_flow = inp->inp_flow;
3080	inp_compat->inp_vflag = inp->inp_vflag;
3081	inp_compat->inp_ip_ttl = inp->inp_ip_ttl;
3082	inp_compat->inp_ip_p = inp->inp_ip_p;
3083	inp_compat->inp_dependfaddr.inp6_foreign =
3084	inp->inp_dependfaddr.inp6_foreign;
3085	inp_compat->inp_dependladdr.inp6_local =
3086	inp->inp_dependladdr.inp6_local;
3087	inp_compat->inp_depend4.inp4_ip_tos = inp->inp_depend4.inp4_ip_tos;
3088	inp_compat->inp_depend6.inp6_hlim = `0`;
3089	inp_compat->inp_depend6.inp6_cksum = inp->inp_depend6.inp6_cksum;
3090	inp_compat->inp_depend6.inp6_ifindex = `0`;
3091	inp_compat->inp_depend6.inp6_hops = inp->inp_depend6.inp6_hops;
3092	}
3093
3094	#if XNU_TARGET_OS_OSX
3095	void
3096	inpcb_to_xinpcb64(struct inpcb inp, struct* xinpcb64 *xinp)
3097	{
3098	xinp->inp_fport = inp->inp_fport;
3099	xinp->inp_lport = inp->inp_lport;
3100	xinp->inp_gencnt = inp->inp_gencnt;
3101	xinp->inp_flags = inp->inp_flags;
3102	xinp->inp_flow = inp->inp_flow;
3103	xinp->inp_vflag = inp->inp_vflag;
3104	xinp->inp_ip_ttl = inp->inp_ip_ttl;
3105	xinp->inp_ip_p = inp->inp_ip_p;
3106	xinp->inp_dependfaddr.inp6_foreign = inp->inp_dependfaddr.inp6_foreign;
3107	xinp->inp_dependladdr.inp6_local = inp->inp_dependladdr.inp6_local;
3108	xinp->inp_depend4.inp4_ip_tos = inp->inp_depend4.inp4_ip_tos;
3109	xinp->inp_depend6.inp6_hlim = `0`;
3110	xinp->inp_depend6.inp6_cksum = inp->inp_depend6.inp6_cksum;
3111	xinp->inp_depend6.inp6_ifindex = `0`;
3112	xinp->inp_depend6.inp6_hops = inp->inp_depend6.inp6_hops;
3113	}
3114	#endif /* XNU_TARGET_OS_OSX */
3115
3116	/*
3117	* The following routines implement this scheme:
3118	*
3119	* Callers of ip_output() that intend to cache the route in the inpcb pass
3120	* a local copy of the struct route to ip_output(). Using a local copy of
3121	* the cached route significantly simplifies things as IP no longer has to
3122	* worry about having exclusive access to the passed in struct route, since
3123	* it's defined in the caller's stack; in essence, this allows for a lock-
3124	* less operation when updating the struct route at the IP level and below,
3125	* whenever necessary. The scheme works as follows:
3126	*
3127	* Prior to dropping the socket's lock and calling ip_output(), the caller
3128	* copies the struct route from the inpcb into its stack, and adds a reference
3129	* to the cached route entry, if there was any. The socket's lock is then
3130	* dropped and ip_output() is called with a pointer to the copy of struct
3131	* route defined on the stack (not to the one in the inpcb.)
3132	*
3133	* Upon returning from ip_output(), the caller then acquires the socket's
3134	* lock and synchronizes the cache; if there is no route cached in the inpcb,
3135	* it copies the local copy of struct route (which may or may not contain any
3136	* route) back into the cache; otherwise, if the inpcb has a route cached in
3137	* it, the one in the local copy will be freed, if there's any. Trashing the
3138	* cached route in the inpcb can be avoided because ip_output() is single-
3139	* threaded per-PCB (i.e. multiple transmits on a PCB are always serialized
3140	* by the socket/transport layer.)
3141	*/
3142	void
3143	inp_route_copyout(struct inpcb inp, struct* route *dst)
3144	{
3145	struct route *src = &inp->inp_route;
3146
3147	socket_lock_assert_owned(so: inp->inp_socket);
3148
3149	/*
3150	* If the route in the PCB is stale or not for IPv4, blow it away;
3151	* this is possible in the case of IPv4-mapped address case.
3152	*/
3153	if (ROUTE_UNUSABLE(src) \|\| rt_key(src->ro_rt)->sa_family != AF_INET) {
3154	ROUTE_RELEASE(src);
3155	}
3156
3157	route_copyout(dst, src, sizeof(*dst));
3158	}
3159
3160	void
3161	inp_route_copyin(struct inpcb inp, struct* route *src)
3162	{
3163	struct route *dst = &inp->inp_route;
3164
3165	socket_lock_assert_owned(so: inp->inp_socket);
3166
3167	/ Minor sanity check /
3168	if (src->ro_rt != NULL && rt_key(src->ro_rt)->sa_family != AF_INET) {
3169	panic("%s: wrong or corrupted route: %p", __func__, src);
3170	}
3171
3172	route_copyin(src, dst, sizeof(*src));
3173	}
3174
3175	/*
3176	* Handler for setting IP_BOUND_IF/IPV6_BOUND_IF socket option.
3177	*/
3178	int
3179	inp_bindif(struct inpcb inp, unsigned* int ifscope, struct ifnet **pifp)
3180	{
3181	struct ifnet *ifp = NULL;
3182
3183	ifnet_head_lock_shared();
3184	if ((ifscope > (unsigned)if_index) \|\| (ifscope != IFSCOPE_NONE &&
3185	(ifp = ifindex2ifnet[ifscope]) == NULL)) {
3186	ifnet_head_done();
3187	return ENXIO;
3188	}
3189	ifnet_head_done();
3190
3191	VERIFY(ifp != NULL \|\| ifscope == IFSCOPE_NONE);
3192
3193	/*
3194	* A zero interface scope value indicates an "unbind".
3195	* Otherwise, take in whatever value the app desires;
3196	* the app may already know the scope (or force itself
3197	* to such a scope) ahead of time before the interface
3198	* gets attached. It doesn't matter either way; any
3199	* route lookup from this point on will require an
3200	* exact match for the embedded interface scope.
3201	*/
3202	inp->inp_boundifp = ifp;
3203	if (inp->inp_boundifp == NULL) {
3204	inp->inp_flags &= ~INP_BOUND_IF;
3205	} else {
3206	inp->inp_flags \|= INP_BOUND_IF;
3207	}
3208
3209	/ Blow away any cached route in the PCB /
3210	ROUTE_RELEASE(&inp->inp_route);
3211
3212	if (pifp != NULL) {
3213	*pifp = ifp;
3214	}
3215
3216	return `0`;
3217	}
3218
3219	/*
3220	* Handler for setting IP_NO_IFT_CELLULAR/IPV6_NO_IFT_CELLULAR socket option,
3221	* as well as for setting PROC_UUID_NO_CELLULAR policy.
3222	*/
3223	void
3224	inp_set_nocellular(struct inpcb *inp)
3225	{
3226	inp->inp_flags \|= INP_NO_IFT_CELLULAR;
3227
3228	/ Blow away any cached route in the PCB /
3229	ROUTE_RELEASE(&inp->inp_route);
3230	}
3231
3232	/*
3233	* Handler for clearing IP_NO_IFT_CELLULAR/IPV6_NO_IFT_CELLULAR socket option,
3234	* as well as for clearing PROC_UUID_NO_CELLULAR policy.
3235	*/
3236	void
3237	inp_clear_nocellular(struct inpcb *inp)
3238	{
3239	struct socket *so = inp->inp_socket;
3240
3241	/*
3242	* SO_RESTRICT_DENY_CELLULAR socket restriction issued on the socket
3243	* has a higher precendence than INP_NO_IFT_CELLULAR. Clear the flag
3244	* if and only if the socket is unrestricted.
3245	*/
3246	if (so != NULL && !(so->so_restrictions & SO_RESTRICT_DENY_CELLULAR)) {
3247	inp->inp_flags &= ~INP_NO_IFT_CELLULAR;
3248
3249	/ Blow away any cached route in the PCB /
3250	ROUTE_RELEASE(&inp->inp_route);
3251	}
3252	}
3253
3254	void
3255	inp_set_noexpensive(struct inpcb *inp)
3256	{
3257	inp->inp_flags2 \|= INP2_NO_IFF_EXPENSIVE;
3258
3259	/ Blow away any cached route in the PCB /
3260	ROUTE_RELEASE(&inp->inp_route);
3261	}
3262
3263	void
3264	inp_set_noconstrained(struct inpcb *inp)
3265	{
3266	inp->inp_flags2 \|= INP2_NO_IFF_CONSTRAINED;
3267
3268	/ Blow away any cached route in the PCB /
3269	ROUTE_RELEASE(&inp->inp_route);
3270	}
3271
3272	void
3273	inp_set_awdl_unrestricted(struct inpcb *inp)
3274	{
3275	inp->inp_flags2 \|= INP2_AWDL_UNRESTRICTED;
3276
3277	/ Blow away any cached route in the PCB /
3278	ROUTE_RELEASE(&inp->inp_route);
3279	}
3280
3281	boolean_t
3282	inp_get_awdl_unrestricted(struct inpcb *inp)
3283	{
3284	return (inp->inp_flags2 & INP2_AWDL_UNRESTRICTED) ? TRUE : FALSE;
3285	}
3286
3287	void
3288	inp_clear_awdl_unrestricted(struct inpcb *inp)
3289	{
3290	inp->inp_flags2 &= ~INP2_AWDL_UNRESTRICTED;
3291
3292	/ Blow away any cached route in the PCB /
3293	ROUTE_RELEASE(&inp->inp_route);
3294	}
3295
3296	void
3297	inp_set_intcoproc_allowed(struct inpcb *inp)
3298	{
3299	inp->inp_flags2 \|= INP2_INTCOPROC_ALLOWED;
3300
3301	/ Blow away any cached route in the PCB /
3302	ROUTE_RELEASE(&inp->inp_route);
3303	}
3304
3305	boolean_t
3306	inp_get_intcoproc_allowed(struct inpcb *inp)
3307	{
3308	return (inp->inp_flags2 & INP2_INTCOPROC_ALLOWED) ? TRUE : FALSE;
3309	}
3310
3311	void
3312	inp_clear_intcoproc_allowed(struct inpcb *inp)
3313	{
3314	inp->inp_flags2 &= ~INP2_INTCOPROC_ALLOWED;
3315
3316	/ Blow away any cached route in the PCB /
3317	ROUTE_RELEASE(&inp->inp_route);
3318	}
3319
3320	void
3321	inp_set_management_allowed(struct inpcb *inp)
3322	{
3323	inp->inp_flags2 \|= INP2_MANAGEMENT_ALLOWED;
3324	inp->inp_flags2 \|= INP2_MANAGEMENT_CHECKED;
3325
3326	/ Blow away any cached route in the PCB /
3327	ROUTE_RELEASE(&inp->inp_route);
3328	}
3329
3330	boolean_t
3331	inp_get_management_allowed(struct inpcb *inp)
3332	{
3333	return (inp->inp_flags2 & INP2_MANAGEMENT_ALLOWED) ? TRUE : FALSE;
3334	}
3335
3336	void
3337	inp_clear_management_allowed(struct inpcb *inp)
3338	{
3339	inp->inp_flags2 &= ~INP2_MANAGEMENT_ALLOWED;
3340
3341	/ Blow away any cached route in the PCB /
3342	ROUTE_RELEASE(&inp->inp_route);
3343	}
3344
3345	#if NECP
3346	/*
3347	* Called when PROC_UUID_NECP_APP_POLICY is set.
3348	*/
3349	void
3350	inp_set_want_app_policy(struct inpcb *inp)
3351	{
3352	inp->inp_flags2 \|= INP2_WANT_APP_POLICY;
3353	}
3354
3355	/*
3356	* Called when PROC_UUID_NECP_APP_POLICY is cleared.
3357	*/
3358	void
3359	inp_clear_want_app_policy(struct inpcb *inp)
3360	{
3361	inp->inp_flags2 &= ~INP2_WANT_APP_POLICY;
3362	}
3363	#endif /* NECP */
3364
3365	/*
3366	* Calculate flow hash for an inp, used by an interface to identify a
3367	* flow. When an interface provides flow control advisory, this flow
3368	* hash is used as an identifier.
3369	*/
3370	u_int32_t
3371	inp_calc_flowhash(struct inpcb *inp)
3372	{
3373	#if SKYWALK
3374
3375	uint32_t flowid;
3376	struct flowidns_flow_key fk;
3377
3378	bzero(s: &fk, n: sizeof(fk));
3379
3380	if (inp->inp_vflag & INP_IPV4) {
3381	fk.ffk_af = AF_INET;
3382	fk.ffk_laddr_v4 = inp->inp_laddr;
3383	fk.ffk_raddr_v4 = inp->inp_faddr;
3384	} else {
3385	fk.ffk_af = AF_INET6;
3386	fk.ffk_laddr_v6 = inp->in6p_laddr;
3387	fk.ffk_raddr_v6 = inp->in6p_faddr;
3388	/ clear embedded scope ID /
3389	if (IN6_IS_SCOPE_EMBED(&fk.ffk_laddr_v6)) {
3390	fk.ffk_laddr_v6.s6_addr16[`1`] = `0`;
3391	}
3392	if (IN6_IS_SCOPE_EMBED(&fk.ffk_raddr_v6)) {
3393	fk.ffk_raddr_v6.s6_addr16[`1`] = `0`;
3394	}
3395	}
3396
3397	fk.ffk_lport = inp->inp_lport;
3398	fk.ffk_rport = inp->inp_fport;
3399	fk.ffk_proto = (inp->inp_ip_p != `0`) ? inp->inp_ip_p :
3400	(uint8_t)SOCK_PROTO(inp->inp_socket);
3401	flowidns_allocate_flowid(domain: FLOWIDNS_DOMAIN_INPCB, flow_key: &fk, flowid: &flowid);
3402	/ Insert the inp into inp_fc_tree /
3403	lck_mtx_lock_spin(lck: &inp_fc_lck);
3404	ASSERT(inp->inp_flowhash == `0`);
3405	ASSERT((inp->inp_flags2 & INP2_IN_FCTREE) == `0`);
3406	inp->inp_flowhash = flowid;
3407	VERIFY(RB_INSERT(inp_fc_tree, &inp_fc_tree, inp) == NULL);
3408	inp->inp_flags2 \|= INP2_IN_FCTREE;
3409	lck_mtx_unlock(lck: &inp_fc_lck);
3410
3411	return flowid;
3412
3413	#else /* !SKYWALK */
3414
3415	struct inp_flowhash_key fh __attribute__((aligned(`8`)));
3416	u_int32_t flowhash = `0`;
3417	struct inpcb *tmp_inp = NULL;
3418
3419	if (inp_hash_seed == `0`) {
3420	inp_hash_seed = RandomULong();
3421	}
3422
3423	bzero(&fh, sizeof(fh));
3424
3425	bcopy(&inp->inp_dependladdr, &fh.infh_laddr, sizeof(fh.infh_laddr));
3426	bcopy(&inp->inp_dependfaddr, &fh.infh_faddr, sizeof(fh.infh_faddr));
3427
3428	fh.infh_lport = inp->inp_lport;
3429	fh.infh_fport = inp->inp_fport;
3430	fh.infh_af = (inp->inp_vflag & INP_IPV6) ? AF_INET6 : AF_INET;
3431	fh.infh_proto = inp->inp_ip_p;
3432	fh.infh_rand1 = RandomULong();
3433	fh.infh_rand2 = RandomULong();
3434
3435	try_again:
3436	flowhash = net_flowhash(&fh, sizeof(fh), inp_hash_seed);
3437	if (flowhash == `0`) {
3438	/ try to get a non-zero flowhash /
3439	inp_hash_seed = RandomULong();
3440	goto try_again;
3441	}
3442
3443	inp->inp_flowhash = flowhash;
3444
3445	/ Insert the inp into inp_fc_tree /
3446	lck_mtx_lock_spin(&inp_fc_lck);
3447	tmp_inp = RB_FIND(inp_fc_tree, &inp_fc_tree, inp);
3448	if (tmp_inp != NULL) {
3449	/*
3450	* There is a different inp with the same flowhash.
3451	* There can be a collision on flow hash but the
3452	* probability is low. Let's recompute the
3453	* flowhash.
3454	*/
3455	lck_mtx_unlock(&inp_fc_lck);
3456	/ recompute hash seed /
3457	inp_hash_seed = RandomULong();
3458	goto try_again;
3459	}
3460
3461	RB_INSERT(inp_fc_tree, &inp_fc_tree, inp);
3462	inp->inp_flags2 \|= INP2_IN_FCTREE;
3463	lck_mtx_unlock(&inp_fc_lck);
3464
3465	return flowhash;
3466
3467	#endif /* !SKYWALK */
3468	}
3469
3470	void
3471	inp_flowadv(uint32_t flowhash)
3472	{
3473	struct inpcb *inp;
3474
3475	inp = inp_fc_getinp(flowhash, `0`);
3476
3477	if (inp == NULL) {
3478	return;
3479	}
3480	inp_fc_feedback(inp);
3481	}
3482
3483	/*
3484	* Function to compare inp_fc_entries in inp flow control tree
3485	*/
3486	static inline int
3487	infc_cmp(const struct inpcb inp1, const* struct inpcb *inp2)
3488	{
3489	return memcmp(s1: &(inp1->inp_flowhash), s2: &(inp2->inp_flowhash),
3490	n: sizeof(inp1->inp_flowhash));
3491	}
3492
3493	static struct inpcb *
3494	inp_fc_getinp(u_int32_t flowhash, u_int32_t flags)
3495	{
3496	struct inpcb *inp = NULL;
3497	int locked = (flags & INPFC_SOLOCKED) ? `1` : `0`;
3498
3499	lck_mtx_lock_spin(lck: &inp_fc_lck);
3500	key_inp.inp_flowhash = flowhash;
3501	inp = RB_FIND(inp_fc_tree, &inp_fc_tree, &key_inp);
3502	if (inp == NULL) {
3503	/ inp is not present, return /
3504	lck_mtx_unlock(lck: &inp_fc_lck);
3505	return NULL;
3506	}
3507
3508	if (flags & INPFC_REMOVE) {
3509	ASSERT((inp->inp_flags2 & INP2_IN_FCTREE) != `0`);
3510	lck_mtx_convert_spin(lck: &inp_fc_lck);
3511	RB_REMOVE(inp_fc_tree, &inp_fc_tree, inp);
3512	bzero(s: &(inp->infc_link), n: sizeof(inp->infc_link));
3513	#if SKYWALK
3514	VERIFY(inp->inp_flowhash != `0`);
3515	flowidns_release_flowid(flowid: inp->inp_flowhash);
3516	inp->inp_flowhash = `0`;
3517	#endif /* !SKYWALK */
3518	inp->inp_flags2 &= ~INP2_IN_FCTREE;
3519	lck_mtx_unlock(lck: &inp_fc_lck);
3520	return NULL;
3521	}
3522
3523	if (in_pcb_checkstate(pcb: inp, WNT_ACQUIRE, locked) == WNT_STOPUSING) {
3524	inp = NULL;
3525	}
3526	lck_mtx_unlock(lck: &inp_fc_lck);
3527
3528	return inp;
3529	}
3530
3531	static void
3532	inp_fc_feedback(struct inpcb *inp)
3533	{
3534	struct socket *so = inp->inp_socket;
3535
3536	/ we already hold a want_cnt on this inp, socket can't be null /
3537	VERIFY(so != NULL);
3538	socket_lock(so, refcount: `1`);
3539
3540	if (in_pcb_checkstate(pcb: inp, WNT_RELEASE, locked: `1`) == WNT_STOPUSING) {
3541	socket_unlock(so, refcount: `1`);
3542	return;
3543	}
3544
3545	if (inp->inp_sndinprog_cnt > `0`) {
3546	inp->inp_flags \|= INP_FC_FEEDBACK;
3547	}
3548
3549	/*
3550	* Return if the connection is not in flow-controlled state.
3551	* This can happen if the connection experienced
3552	* loss while it was in flow controlled state
3553	*/
3554	if (!INP_WAIT_FOR_IF_FEEDBACK(inp)) {
3555	socket_unlock(so, refcount: `1`);
3556	return;
3557	}
3558	inp_reset_fc_state(inp);
3559
3560	if (SOCK_TYPE(so) == SOCK_STREAM) {
3561	inp_fc_unthrottle_tcp(inp);
3562	}
3563
3564	socket_unlock(so, refcount: `1`);
3565	}
3566
3567	static void
3568	inp_reset_fc_timerstat(struct inpcb *inp)
3569	{
3570	uint64_t now;
3571
3572	if (inp->inp_fadv_start_time == `0`) {
3573	return;
3574	}
3575
3576	now = net_uptime_us();
3577	ASSERT(now >= inp->inp_fadv_start_time);
3578
3579	inp->inp_fadv_total_time += (now - inp->inp_fadv_start_time);
3580	inp->inp_fadv_cnt++;
3581
3582	inp->inp_fadv_start_time = `0`;
3583	}
3584
3585	static void
3586	inp_set_fc_timerstat(struct inpcb *inp)
3587	{
3588	if (inp->inp_fadv_start_time != `0`) {
3589	return;
3590	}
3591
3592	inp->inp_fadv_start_time = net_uptime_us();
3593	}
3594
3595	void
3596	inp_reset_fc_state(struct inpcb *inp)
3597	{
3598	struct socket *so = inp->inp_socket;
3599	int suspended = (INP_IS_FLOW_SUSPENDED(inp)) ? `1` : `0`;
3600	int needwakeup = (INP_WAIT_FOR_IF_FEEDBACK(inp)) ? `1` : `0`;
3601
3602	inp->inp_flags &= ~(INP_FLOW_CONTROLLED \| INP_FLOW_SUSPENDED);
3603
3604	inp_reset_fc_timerstat(inp);
3605
3606	if (suspended) {
3607	so->so_flags &= ~(SOF_SUSPENDED);
3608	soevent(so, hint: (SO_FILT_HINT_LOCKED \| SO_FILT_HINT_RESUME));
3609	}
3610
3611	/ Give a write wakeup to unblock the socket /
3612	if (needwakeup) {
3613	sowwakeup(so);
3614	}
3615	}
3616
3617	int
3618	inp_set_fc_state(struct inpcb inp, int* advcode)
3619	{
3620	boolean_t is_flow_controlled = INP_WAIT_FOR_IF_FEEDBACK(inp);
3621	struct inpcb *tmp_inp = NULL;
3622	/*
3623	* If there was a feedback from the interface when
3624	* send operation was in progress, we should ignore
3625	* this flow advisory to avoid a race between setting
3626	* flow controlled state and receiving feedback from
3627	* the interface
3628	*/
3629	if (inp->inp_flags & INP_FC_FEEDBACK) {
3630	return `0`;
3631	}
3632
3633	inp->inp_flags &= ~(INP_FLOW_CONTROLLED \| INP_FLOW_SUSPENDED);
3634	if ((tmp_inp = inp_fc_getinp(flowhash: inp->inp_flowhash,
3635	INPFC_SOLOCKED)) != NULL) {
3636	if (in_pcb_checkstate(pcb: tmp_inp, WNT_RELEASE, locked: `1`) == WNT_STOPUSING) {
3637	goto exit_reset;
3638	}
3639	VERIFY(tmp_inp == inp);
3640	switch (advcode) {
3641	case FADV_FLOW_CONTROLLED:
3642	inp->inp_flags \|= INP_FLOW_CONTROLLED;
3643	inp_set_fc_timerstat(inp);
3644	break;
3645	case FADV_SUSPENDED:
3646	inp->inp_flags \|= INP_FLOW_SUSPENDED;
3647	inp_set_fc_timerstat(inp);
3648
3649	soevent(so: inp->inp_socket,
3650	hint: (SO_FILT_HINT_LOCKED \| SO_FILT_HINT_SUSPEND));
3651
3652	/ Record the fact that suspend event was sent /
3653	inp->inp_socket->so_flags \|= SOF_SUSPENDED;
3654	break;
3655	}
3656
3657	if (!is_flow_controlled && SOCK_TYPE(inp->inp_socket) == SOCK_STREAM) {
3658	inp_fc_throttle_tcp(inp);
3659	}
3660	return `1`;
3661	}
3662
3663	exit_reset:
3664	inp_reset_fc_timerstat(inp);
3665
3666	return `0`;
3667	}
3668
3669	/*
3670	* Handler for SO_FLUSH socket option.
3671	*/
3672	int
3673	inp_flush(struct inpcb inp, int* optval)
3674	{
3675	u_int32_t flowhash = inp->inp_flowhash;
3676	struct ifnet rtifp, oifp;
3677
3678	/ Either all classes or one of the valid ones /
3679	if (optval != SO_TC_ALL && !SO_VALID_TC(optval)) {
3680	return EINVAL;
3681	}
3682
3683	/ We need a flow hash for identification /
3684	if (flowhash == `0`) {
3685	return `0`;
3686	}
3687
3688	/ Grab the interfaces from the route and pcb /
3689	rtifp = ((inp->inp_route.ro_rt != NULL) ?
3690	inp->inp_route.ro_rt->rt_ifp : NULL);
3691	oifp = inp->inp_last_outifp;
3692
3693	if (rtifp != NULL) {
3694	if_qflush_sc(rtifp, so_tc2msc(optval), flowhash, NULL, NULL, `0`);
3695	}
3696	if (oifp != NULL && oifp != rtifp) {
3697	if_qflush_sc(oifp, so_tc2msc(optval), flowhash, NULL, NULL, `0`);
3698	}
3699
3700	return `0`;
3701	}
3702
3703	/*
3704	* Clear the INP_INADDR_ANY flag (special case for PPP only)
3705	*/
3706	void
3707	inp_clear_INP_INADDR_ANY(struct socket *so)
3708	{
3709	struct inpcb *inp = NULL;
3710
3711	socket_lock(so, refcount: `1`);
3712	inp = sotoinpcb(so);
3713	if (inp) {
3714	inp->inp_flags &= ~INP_INADDR_ANY;
3715	}
3716	socket_unlock(so, refcount: `1`);
3717	}
3718
3719	void
3720	inp_get_soprocinfo(struct inpcb inp, struct* so_procinfo *soprocinfo)
3721	{
3722	struct socket *so = inp->inp_socket;
3723
3724	soprocinfo->spi_pid = so->last_pid;
3725	strlcpy(dst: &soprocinfo->spi_proc_name[`0`], src: &inp->inp_last_proc_name[`0`],
3726	n: sizeof(soprocinfo->spi_proc_name));
3727	if (so->last_pid != `0`) {
3728	uuid_copy(dst: soprocinfo->spi_uuid, src: so->last_uuid);
3729	}
3730	/*
3731	* When not delegated, the effective pid is the same as the real pid
3732	*/
3733	if (so->so_flags & SOF_DELEGATED) {
3734	soprocinfo->spi_delegated = `1`;
3735	soprocinfo->spi_epid = so->e_pid;
3736	uuid_copy(dst: soprocinfo->spi_euuid, src: so->e_uuid);
3737	} else {
3738	soprocinfo->spi_delegated = `0`;
3739	soprocinfo->spi_epid = so->last_pid;
3740	}
3741	strlcpy(dst: &soprocinfo->spi_e_proc_name[`0`], src: &inp->inp_e_proc_name[`0`],
3742	n: sizeof(soprocinfo->spi_e_proc_name));
3743	}
3744
3745	int
3746	inp_findinpcb_procinfo(struct inpcbinfo *pcbinfo, uint32_t flowhash,
3747	struct so_procinfo *soprocinfo)
3748	{
3749	struct inpcb *inp = NULL;
3750	int found = `0`;
3751
3752	bzero(s: soprocinfo, n: sizeof(struct so_procinfo));
3753
3754	if (!flowhash) {
3755	return -`1`;
3756	}
3757
3758	lck_rw_lock_shared(lck: &pcbinfo->ipi_lock);
3759	LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) {
3760	if (inp->inp_state != INPCB_STATE_DEAD &&
3761	inp->inp_socket != NULL &&
3762	inp->inp_flowhash == flowhash) {
3763	found = `1`;
3764	inp_get_soprocinfo(inp, soprocinfo);
3765	break;
3766	}
3767	}
3768	lck_rw_done(lck: &pcbinfo->ipi_lock);
3769
3770	return found;
3771	}
3772
3773	#if CONFIG_PROC_UUID_POLICY
3774	static void
3775	inp_update_cellular_policy(struct inpcb *inp, boolean_t set)
3776	{
3777	struct socket *so = inp->inp_socket;
3778	int before, after;
3779
3780	VERIFY(so != NULL);
3781	VERIFY(inp->inp_state != INPCB_STATE_DEAD);
3782
3783	before = INP_NO_CELLULAR(inp);
3784	if (set) {
3785	inp_set_nocellular(inp);
3786	} else {
3787	inp_clear_nocellular(inp);
3788	}
3789	after = INP_NO_CELLULAR(inp);
3790	if (net_io_policy_log && (before != after)) {
3791	static const char *ok = "OK";
3792	static const char *nok = "NOACCESS";
3793	uuid_string_t euuid_buf;
3794	pid_t epid;
3795
3796	if (so->so_flags & SOF_DELEGATED) {
3797	uuid_unparse(uu: so->e_uuid, out: euuid_buf);
3798	epid = so->e_pid;
3799	} else {
3800	uuid_unparse(uu: so->last_uuid, out: euuid_buf);
3801	epid = so->last_pid;
3802	}
3803
3804	/ allow this socket to generate another notification event /
3805	so->so_ifdenied_notifies = `0`;
3806
3807	log(LOG_DEBUG, "%s: so %llu [%d,%d] epid %d "
3808	"euuid %s%s %s->%s\n", __func__,
3809	so->so_gencnt, SOCK_DOM(so),
3810	SOCK_TYPE(so), epid, euuid_buf,
3811	(so->so_flags & SOF_DELEGATED) ?
3812	" [delegated]" : "",
3813	((before < after) ? ok : nok),
3814	((before < after) ? nok : ok));
3815	}
3816	}
3817
3818	#if NECP
3819	static void
3820	inp_update_necp_want_app_policy(struct inpcb *inp, boolean_t set)
3821	{
3822	struct socket *so = inp->inp_socket;
3823	int before, after;
3824
3825	VERIFY(so != NULL);
3826	VERIFY(inp->inp_state != INPCB_STATE_DEAD);
3827
3828	before = (inp->inp_flags2 & INP2_WANT_APP_POLICY);
3829	if (set) {
3830	inp_set_want_app_policy(inp);
3831	} else {
3832	inp_clear_want_app_policy(inp);
3833	}
3834	after = (inp->inp_flags2 & INP2_WANT_APP_POLICY);
3835	if (net_io_policy_log && (before != after)) {
3836	static const char *wanted = "WANTED";
3837	static const char *unwanted = "UNWANTED";
3838	uuid_string_t euuid_buf;
3839	pid_t epid;
3840
3841	if (so->so_flags & SOF_DELEGATED) {
3842	uuid_unparse(uu: so->e_uuid, out: euuid_buf);
3843	epid = so->e_pid;
3844	} else {
3845	uuid_unparse(uu: so->last_uuid, out: euuid_buf);
3846	epid = so->last_pid;
3847	}
3848
3849	log(LOG_DEBUG, "%s: so %llu [%d,%d] epid %d "
3850	"euuid %s%s %s->%s\n", __func__,
3851	so->so_gencnt, SOCK_DOM(so),
3852	SOCK_TYPE(so), epid, euuid_buf,
3853	(so->so_flags & SOF_DELEGATED) ?
3854	" [delegated]" : "",
3855	((before < after) ? unwanted : wanted),
3856	((before < after) ? wanted : unwanted));
3857	}
3858	}
3859	#endif /* NECP */
3860	#endif /* !CONFIG_PROC_UUID_POLICY */
3861
3862	#if NECP
3863	void
3864	inp_update_necp_policy(struct inpcb inp, struct* sockaddr override_local_addr, struct* sockaddr *override_remote_addr, u_int override_bound_interface)
3865	{
3866	necp_socket_find_policy_match(inp, override_local_addr, override_remote_addr, override_bound_interface);
3867	if (necp_socket_should_rescope(inp) &&
3868	inp->inp_lport == `0` &&
3869	inp->inp_laddr.s_addr == INADDR_ANY &&
3870	IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
3871	// If we should rescope, and the socket is not yet bound
3872	inp_bindif(inp, ifscope: necp_socket_get_rescope_if_index(inp), NULL);
3873	inp->inp_flags2 \|= INP2_SCOPED_BY_NECP;
3874	}
3875	}
3876	#endif /* NECP */
3877
3878	int
3879	inp_update_policy(struct inpcb *inp)
3880	{
3881	#if CONFIG_PROC_UUID_POLICY
3882	struct socket *so = inp->inp_socket;
3883	uint32_t pflags = `0`;
3884	int32_t ogencnt;
3885	int err = `0`;
3886	uint8_t *lookup_uuid = NULL;
3887
3888	if (!net_io_policy_uuid \|\|
3889	so == NULL \|\| inp->inp_state == INPCB_STATE_DEAD) {
3890	return `0`;
3891	}
3892
3893	/*
3894	* Kernel-created sockets that aren't delegating other sockets
3895	* are currently exempted from UUID policy checks.
3896	*/
3897	if (so->last_pid == `0` && !(so->so_flags & SOF_DELEGATED)) {
3898	return `0`;
3899	}
3900
3901	#if defined(XNU_TARGET_OS_OSX)
3902	if (so->so_rpid > `0`) {
3903	lookup_uuid = so->so_ruuid;
3904	ogencnt = so->so_policy_gencnt;
3905	err = proc_uuid_policy_lookup(uuid: lookup_uuid, flags: &pflags, gencount: &so->so_policy_gencnt);
3906	}
3907	#endif
3908	if (lookup_uuid == NULL \|\| err == ENOENT) {
3909	lookup_uuid = ((so->so_flags & SOF_DELEGATED) ? so->e_uuid : so->last_uuid);
3910	ogencnt = so->so_policy_gencnt;
3911	err = proc_uuid_policy_lookup(uuid: lookup_uuid, flags: &pflags, gencount: &so->so_policy_gencnt);
3912	}
3913
3914	/*
3915	* Discard cached generation count if the entry is gone (ENOENT),
3916	* so that we go thru the checks below.
3917	*/
3918	if (err == ENOENT && ogencnt != `0`) {
3919	so->so_policy_gencnt = `0`;
3920	}
3921
3922	/*
3923	* If the generation count has changed, inspect the policy flags
3924	* and act accordingly. If a policy flag was previously set and
3925	* the UUID is no longer present in the table (ENOENT), treat it
3926	* as if the flag has been cleared.
3927	*/
3928	if ((err == `0` \|\| err == ENOENT) && ogencnt != so->so_policy_gencnt) {
3929	/ update cellular policy for this socket /
3930	if (err == `0` && (pflags & PROC_UUID_NO_CELLULAR)) {
3931	inp_update_cellular_policy(inp, TRUE);
3932	} else if (!(pflags & PROC_UUID_NO_CELLULAR)) {
3933	inp_update_cellular_policy(inp, FALSE);
3934	}
3935	#if NECP
3936	/ update necp want app policy for this socket /
3937	if (err == `0` && (pflags & PROC_UUID_NECP_APP_POLICY)) {
3938	inp_update_necp_want_app_policy(inp, TRUE);
3939	} else if (!(pflags & PROC_UUID_NECP_APP_POLICY)) {
3940	inp_update_necp_want_app_policy(inp, FALSE);
3941	}
3942	#endif /* NECP */
3943	}
3944
3945	return (err == ENOENT) ? `0` : err;
3946	#else /* !CONFIG_PROC_UUID_POLICY */
3947	#pragma unused(inp)
3948	return `0`;
3949	#endif /* !CONFIG_PROC_UUID_POLICY */
3950	}
3951
3952	unsigned int log_restricted;
3953	SYSCTL_DECL(_net_inet);
3954	SYSCTL_INT(_net_inet, OID_AUTO, log_restricted,
3955	CTLFLAG_RW \| CTLFLAG_LOCKED, &log_restricted, `0`,
3956	"Log network restrictions");
3957
3958
3959	/*
3960	* Called when we need to enforce policy restrictions in the input path.
3961	*
3962	* Returns TRUE if we're not allowed to receive data, otherwise FALSE.
3963	*/
3964	static boolean_t
3965	_inp_restricted_recv(struct inpcb inp, struct* ifnet *ifp)
3966	{
3967	VERIFY(inp != NULL);
3968
3969	/*
3970	* Inbound restrictions.
3971	*/
3972	if (!sorestrictrecv) {
3973	return FALSE;
3974	}
3975
3976	if (ifp == NULL) {
3977	return FALSE;
3978	}
3979
3980	if (IFNET_IS_CELLULAR(ifp) && INP_NO_CELLULAR(inp)) {
3981	return TRUE;
3982	}
3983
3984	if (IFNET_IS_EXPENSIVE(ifp) && INP_NO_EXPENSIVE(inp)) {
3985	return TRUE;
3986	}
3987
3988	if (IFNET_IS_CONSTRAINED(ifp) && INP_NO_CONSTRAINED(inp)) {
3989	return TRUE;
3990	}
3991
3992	if (IFNET_IS_AWDL_RESTRICTED(ifp) && !INP_AWDL_UNRESTRICTED(inp)) {
3993	return TRUE;
3994	}
3995
3996	if (!(ifp->if_eflags & IFEF_RESTRICTED_RECV)) {
3997	return FALSE;
3998	}
3999
4000	if (inp->inp_flags & INP_RECV_ANYIF) {
4001	return FALSE;
4002	}
4003
4004	/*
4005	* An entitled process can use the management interface without being bound
4006	* to the interface
4007	*/
4008	if (IFNET_IS_MANAGEMENT(ifp)) {
4009	if (INP_MANAGEMENT_ALLOWED(inp)) {
4010	return FALSE;
4011	}
4012	if (if_management_verbose > `1`) {
4013	os_log(OS_LOG_DEFAULT, "_inp_restricted_recv %s:%d not allowed on management interface %s",
4014	proc_best_name(current_proc()), proc_getpid(current_proc()),
4015	ifp->if_xname);
4016	}
4017	return TRUE;
4018	}
4019
4020	if ((inp->inp_flags & INP_BOUND_IF) && inp->inp_boundifp == ifp) {
4021	return FALSE;
4022	}
4023
4024	if (IFNET_IS_INTCOPROC(ifp) && !INP_INTCOPROC_ALLOWED(inp)) {
4025	return TRUE;
4026	}
4027
4028
4029	return TRUE;
4030	}
4031
4032	boolean_t
4033	inp_restricted_recv(struct inpcb inp, struct* ifnet *ifp)
4034	{
4035	boolean_t ret;
4036
4037	ret = _inp_restricted_recv(inp, ifp);
4038	if (ret == TRUE && log_restricted) {
4039	printf("pid %d (%s) is unable to receive packets on %s\n",
4040	proc_getpid(current_proc()), proc_best_name(p: current_proc()),
4041	ifp->if_xname);
4042	}
4043	return ret;
4044	}
4045
4046	/*
4047	* Called when we need to enforce policy restrictions in the output path.
4048	*
4049	* Returns TRUE if we're not allowed to send data out, otherwise FALSE.
4050	*/
4051	static boolean_t
4052	_inp_restricted_send(struct inpcb inp, struct* ifnet *ifp)
4053	{
4054	VERIFY(inp != NULL);
4055
4056	/*
4057	* Outbound restrictions.
4058	*/
4059	if (!sorestrictsend) {
4060	return FALSE;
4061	}
4062
4063	if (ifp == NULL) {
4064	return FALSE;
4065	}
4066
4067	if (IFNET_IS_CELLULAR(ifp) && INP_NO_CELLULAR(inp)) {
4068	return TRUE;
4069	}
4070
4071	if (IFNET_IS_EXPENSIVE(ifp) && INP_NO_EXPENSIVE(inp)) {
4072	return TRUE;
4073	}
4074
4075	if (IFNET_IS_CONSTRAINED(ifp) && INP_NO_CONSTRAINED(inp)) {
4076	return TRUE;
4077	}
4078
4079	if (IFNET_IS_AWDL_RESTRICTED(ifp) && !INP_AWDL_UNRESTRICTED(inp)) {
4080	return TRUE;
4081	}
4082
4083	if (IFNET_IS_MANAGEMENT(ifp)) {
4084	if (!INP_MANAGEMENT_ALLOWED(inp)) {
4085	if (if_management_verbose > `1`) {
4086	os_log(OS_LOG_DEFAULT, "_inp_restricted_send %s:%d not allowed on management interface %s",
4087	proc_best_name(current_proc()), proc_getpid(current_proc()),
4088	ifp->if_xname);
4089	}
4090	return TRUE;
4091	}
4092	}
4093
4094	if (IFNET_IS_INTCOPROC(ifp) && !INP_INTCOPROC_ALLOWED(inp)) {
4095	return TRUE;
4096	}
4097
4098	return FALSE;
4099	}
4100
4101	boolean_t
4102	inp_restricted_send(struct inpcb inp, struct* ifnet *ifp)
4103	{
4104	boolean_t ret;
4105
4106	ret = _inp_restricted_send(inp, ifp);
4107	if (ret == TRUE && log_restricted) {
4108	printf("pid %d (%s) is unable to transmit packets on %s\n",
4109	proc_getpid(current_proc()), proc_best_name(p: current_proc()),
4110	ifp->if_xname);
4111	}
4112	return ret;
4113	}
4114
4115	inline void
4116	inp_count_sndbytes(struct inpcb *inp, u_int32_t th_ack)
4117	{
4118	struct ifnet *ifp = inp->inp_last_outifp;
4119	struct socket *so = inp->inp_socket;
4120	if (ifp != NULL && !(so->so_flags & SOF_MP_SUBFLOW) &&
4121	(ifp->if_type == IFT_CELLULAR \|\| IFNET_IS_WIFI(ifp))) {
4122	int32_t unsent;
4123
4124	so->so_snd.sb_flags \|= SB_SNDBYTE_CNT;
4125
4126	/*
4127	* There can be data outstanding before the connection
4128	* becomes established -- TFO case
4129	*/
4130	if (so->so_snd.sb_cc > `0`) {
4131	inp_incr_sndbytes_total(so, so->so_snd.sb_cc);
4132	}
4133
4134	unsent = inp_get_sndbytes_allunsent(so, th_ack);
4135	if (unsent > `0`) {
4136	inp_incr_sndbytes_unsent(so, unsent);
4137	}
4138	}
4139	}
4140
4141	inline void
4142	inp_incr_sndbytes_total(struct socket *so, int32_t len)
4143	{
4144	struct inpcb inp = (struct* inpcb *)so->so_pcb;
4145	struct ifnet *ifp = inp->inp_last_outifp;
4146
4147	if (ifp != NULL) {
4148	VERIFY(ifp->if_sndbyte_total >= `0`);
4149	OSAddAtomic64(len, &ifp->if_sndbyte_total);
4150	}
4151	}
4152
4153	inline void
4154	inp_decr_sndbytes_total(struct socket *so, int32_t len)
4155	{
4156	struct inpcb inp = (struct* inpcb *)so->so_pcb;
4157	struct ifnet *ifp = inp->inp_last_outifp;
4158
4159	if (ifp != NULL) {
4160	if (ifp->if_sndbyte_total >= len) {
4161	OSAddAtomic64(-len, &ifp->if_sndbyte_total);
4162	} else {
4163	ifp->if_sndbyte_total = `0`;
4164	}
4165	}
4166	}
4167
4168	inline void
4169	inp_incr_sndbytes_unsent(struct socket *so, int32_t len)
4170	{
4171	struct inpcb inp = (struct* inpcb *)so->so_pcb;
4172	struct ifnet *ifp = inp->inp_last_outifp;
4173
4174	if (ifp != NULL) {
4175	VERIFY(ifp->if_sndbyte_unsent >= `0`);
4176	OSAddAtomic64(len, &ifp->if_sndbyte_unsent);
4177	}
4178	}
4179
4180	inline void
4181	inp_decr_sndbytes_unsent(struct socket *so, int32_t len)
4182	{
4183	if (so == NULL \|\| !(so->so_snd.sb_flags & SB_SNDBYTE_CNT)) {
4184	return;
4185	}
4186
4187	struct inpcb inp = (struct* inpcb *)so->so_pcb;
4188	struct ifnet *ifp = inp->inp_last_outifp;
4189
4190	if (ifp != NULL) {
4191	if (ifp->if_sndbyte_unsent >= len) {
4192	OSAddAtomic64(-len, &ifp->if_sndbyte_unsent);
4193	} else {
4194	ifp->if_sndbyte_unsent = `0`;
4195	}
4196	}
4197	}
4198
4199	inline void
4200	inp_decr_sndbytes_allunsent(struct socket *so, u_int32_t th_ack)
4201	{
4202	int32_t len;
4203
4204	if (so == NULL \|\| !(so->so_snd.sb_flags & SB_SNDBYTE_CNT)) {
4205	return;
4206	}
4207
4208	len = inp_get_sndbytes_allunsent(so, th_ack);
4209	inp_decr_sndbytes_unsent(so, len);
4210	}
4211
4212	#if SKYWALK
4213	inline void
4214	inp_update_netns_flags(struct socket *so)
4215	{
4216	struct inpcb *inp;
4217	uint32_t set_flags = `0`;
4218	uint32_t clear_flags = `0`;
4219
4220	if (!(SOCK_CHECK_DOM(so, AF_INET) \|\| SOCK_CHECK_DOM(so, AF_INET6))) {
4221	return;
4222	}
4223
4224	inp = sotoinpcb(so);
4225
4226	if (inp == NULL) {
4227	return;
4228	}
4229
4230	if (!NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
4231	return;
4232	}
4233
4234	if (so->so_options & SO_NOWAKEFROMSLEEP) {
4235	set_flags \|= NETNS_NOWAKEFROMSLEEP;
4236	} else {
4237	clear_flags \|= NETNS_NOWAKEFROMSLEEP;
4238	}
4239
4240	if (inp->inp_flags & INP_RECV_ANYIF) {
4241	set_flags \|= NETNS_RECVANYIF;
4242	} else {
4243	clear_flags \|= NETNS_RECVANYIF;
4244	}
4245
4246	if (so->so_flags1 & SOF1_EXTEND_BK_IDLE_WANTED) {
4247	set_flags \|= NETNS_EXTBGIDLE;
4248	} else {
4249	clear_flags \|= NETNS_EXTBGIDLE;
4250	}
4251
4252	netns_change_flags(token: &inp->inp_netns_token, set_flags, clear_flags);
4253	}
4254	#endif /* SKYWALK */
4255
4256	inline void
4257	inp_set_activity_bitmap(struct inpcb *inp)
4258	{
4259	in_stat_set_activity_bitmap(activity: &inp->inp_nw_activity, now: net_uptime());
4260	}
4261
4262	inline void
4263	inp_get_activity_bitmap(struct inpcb inp, activity_bitmap_t ab)
4264	{
4265	bcopy(src: &inp->inp_nw_activity, dst: ab, n: sizeof(*ab));
4266	}
4267
4268	inline void
4269	inp_clear_activity_bitmap(struct inpcb *inp)
4270	{
4271	in_stat_clear_activity_bitmap(activity: &inp->inp_nw_activity);
4272	}
4273
4274	void
4275	inp_update_last_owner(struct socket so, struct* proc p, struct* proc *ep)
4276	{
4277	struct inpcb inp = (struct* inpcb *)so->so_pcb;
4278
4279	if (inp == NULL) {
4280	return;
4281	}
4282
4283	if (p != NULL) {
4284	strlcpy(dst: &inp->inp_last_proc_name[`0`], src: proc_name_address(p), n: sizeof(inp->inp_last_proc_name));
4285	}
4286	if (so->so_flags & SOF_DELEGATED) {
4287	if (ep != NULL) {
4288	strlcpy(dst: &inp->inp_e_proc_name[`0`], src: proc_name_address(ep), n: sizeof(inp->inp_e_proc_name));
4289	} else {
4290	inp->inp_e_proc_name[`0`] = `0`;
4291	}
4292	} else {
4293	inp->inp_e_proc_name[`0`] = `0`;
4294	}
4295	}
4296
4297	void
4298	inp_copy_last_owner(struct socket so, struct* socket *head)
4299	{
4300	struct inpcb inp = (struct* inpcb *)so->so_pcb;
4301	struct inpcb head_inp = (struct* inpcb *)head->so_pcb;
4302
4303	if (inp == NULL \|\| head_inp == NULL) {
4304	return;
4305	}
4306
4307	strlcpy(dst: &inp->inp_last_proc_name[`0`], src: &head_inp->inp_last_proc_name[`0`], n: sizeof(inp->inp_last_proc_name));
4308	strlcpy(dst: &inp->inp_e_proc_name[`0`], src: &head_inp->inp_e_proc_name[`0`], n: sizeof(inp->inp_e_proc_name));
4309	}
4310
4311	static int
4312	in_check_management_interface_proc_callout(proc_t proc, void *arg __unused)
4313	{
4314	struct fileproc *fp = NULL;
4315	task_t task = proc_task(proc);
4316	bool allowed = false;
4317
4318	if (IOTaskHasEntitlement(task, INTCOPROC_RESTRICTED_ENTITLEMENT) == true
4319	\|\| IOTaskHasEntitlement(task, MANAGEMENT_DATA_ENTITLEMENT) == true
4320	#if DEBUG \|\| DEVELOPMENT
4321	\|\| IOTaskHasEntitlement(task, INTCOPROC_RESTRICTED_ENTITLEMENT_DEVELOPMENT) == true
4322	\|\| IOTaskHasEntitlement(task, MANAGEMENT_DATA_ENTITLEMENT_DEVELOPMENT) == true
4323	#endif /* DEBUG \|\| DEVELOPMENT */
4324	) {
4325	allowed = true;
4326	}
4327	if (allowed == false && management_data_unrestricted == false) {
4328	return PROC_RETURNED;
4329	}
4330
4331	proc_fdlock(proc);
4332	fdt_foreach(fp, proc) {
4333	struct fileglob *fg = fp->fp_glob;
4334	struct socket *so;
4335	struct inpcb *inp;
4336
4337	if (FILEGLOB_DTYPE(fg) != DTYPE_SOCKET) {
4338	continue;
4339	}
4340
4341	so = (struct socket *)fp_get_data(fp);
4342	if (SOCK_DOM(so) != PF_INET && SOCK_DOM(so) != PF_INET6) {
4343	continue;
4344	}
4345
4346	inp = (struct inpcb *)so->so_pcb;
4347
4348	if (in_pcb_checkstate(pcb: inp, WNT_ACQUIRE, locked: `0`) == WNT_STOPUSING) {
4349	continue;
4350	}
4351
4352	socket_lock(so, refcount: `1`);
4353
4354	if (in_pcb_checkstate(pcb: inp, WNT_RELEASE, locked: `1`) == WNT_STOPUSING) {
4355	socket_unlock(so, refcount: `1`);
4356	continue;
4357	}
4358	inp->inp_flags2 \|= INP2_MANAGEMENT_ALLOWED;
4359	inp->inp_flags2 \|= INP2_MANAGEMENT_CHECKED;
4360
4361	socket_unlock(so, refcount: `1`);
4362	}
4363	proc_fdunlock(proc);
4364
4365	return PROC_RETURNED;
4366	}
4367
4368	static bool in_management_interface_checked = false;
4369
4370	static void
4371	in_management_interface_event_callback(struct nwk_wq_entry *nwk_item)
4372	{
4373	kfree_type(struct nwk_wq_entry, nwk_item);
4374
4375	if (in_management_interface_checked == true) {
4376	return;
4377	}
4378	in_management_interface_checked = true;
4379
4380	proc_iterate(PROC_ALLPROCLIST,
4381	callout: in_check_management_interface_proc_callout,
4382	NULL, NULL, NULL);
4383	}
4384
4385	void
4386	in_management_interface_check(void)
4387	{
4388	struct nwk_wq_entry *nwk_item;
4389
4390	if (if_management_interface_check_needed == false \|\|
4391	in_management_interface_checked == true) {
4392	return;
4393	}
4394
4395	nwk_item = kalloc_type(struct nwk_wq_entry,
4396	Z_WAITOK \| Z_ZERO \| Z_NOFAIL);
4397
4398	nwk_item->func = in_management_interface_event_callback;
4399
4400	nwk_wq_enqueue(nwk_item);
4401	}
4402

Browse the source code of xnu/bsd/netinet/in_pcb.c