1/*
2 * Copyright (c) 2000-2011 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/* Copyright (c) 1995, 1997 Apple Computer, Inc. All Rights Reserved */
29/*
30 * Copyright (c) 1982, 1986, 1989, 1991, 1993
31 *	The Regents of the University of California.  All rights reserved.
32 * (c) UNIX System Laboratories, Inc.
33 * All or some portions of this file are derived from material licensed
34 * to the University of California by American Telephone and Telegraph
35 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
36 * the permission of UNIX System Laboratories, Inc.
37 *
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
40 * are met:
41 * 1. Redistributions of source code must retain the above copyright
42 *    notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 *    notice, this list of conditions and the following disclaimer in the
45 *    documentation and/or other materials provided with the distribution.
46 * 3. All advertising materials mentioning features or use of this software
47 *    must display the following acknowledgement:
48 *	This product includes software developed by the University of
49 *	California, Berkeley and its contributors.
50 * 4. Neither the name of the University nor the names of its contributors
51 *    may be used to endorse or promote products derived from this software
52 *    without specific prior written permission.
53 *
54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64 * SUCH DAMAGE.
65 *
66 *	@(#)kern_exit.c	8.7 (Berkeley) 2/12/94
67 */
68/*
69 * NOTICE: This file was modified by SPARTA, Inc. in 2005 to introduce
70 * support for mandatory and extensible security protections.  This notice
71 * is included in support of clause 2.2 (b) of the Apple Public License,
72 * Version 2.0.
73 */
74
75#include <machine/reg.h>
76#include <machine/psl.h>
77
78#include "compat_43.h"
79
80#include <sys/param.h>
81#include <sys/systm.h>
82#include <sys/ioctl.h>
83#include <sys/proc_internal.h>
84#include <sys/proc.h>
85#include <sys/kauth.h>
86#include <sys/tty.h>
87#include <sys/time.h>
88#include <sys/resource.h>
89#include <sys/kernel.h>
90#include <sys/wait.h>
91#include <sys/file_internal.h>
92#include <sys/vnode_internal.h>
93#include <sys/syslog.h>
94#include <sys/malloc.h>
95#include <sys/resourcevar.h>
96#include <sys/ptrace.h>
97#include <sys/user.h>
98#include <sys/aio_kern.h>
99#include <sys/sysproto.h>
100#include <sys/signalvar.h>
101#include <sys/kdebug.h>
102#include <sys/filedesc.h>	/* fdfree */
103#if SYSV_SHM
104#include <sys/shm_internal.h>	/* shmexit */
105#endif
106#include <sys/acct.h>		/* acct_process */
107
108#include <security/audit/audit.h>
109#include <bsm/audit_kevents.h>
110
111#include <mach/mach_types.h>
112
113#include <kern/kern_types.h>
114#include <kern/kalloc.h>
115#include <kern/task.h>
116#include <kern/thread.h>
117#include <kern/thread_call.h>
118#include <kern/sched_prim.h>
119#include <kern/assert.h>
120#include <sys/codesign.h>
121
122#if VM_PRESSURE_EVENTS
123#include <kern/vm_pressure.h>
124#endif
125
126#if CONFIG_MEMORYSTATUS
127#include <sys/kern_memorystatus.h>
128#endif
129
130#if CONFIG_DTRACE
131/* Do not include dtrace.h, it redefines kmem_[alloc/free] */
132extern void (*dtrace_fasttrap_exit_ptr)(proc_t);
133extern void (*dtrace_helpers_cleanup)(proc_t);
134extern void dtrace_lazy_dofs_destroy(proc_t);
135
136#include <sys/dtrace_ptss.h>
137#endif
138
139#if CONFIG_MACF
140#include <security/mac.h>
141#include <sys/syscall.h>
142#endif
143
144#include <mach/mach_types.h>
145#include <mach/task.h>
146#include <mach/thread_act.h>
147
148#include <sys/sdt.h>
149
150extern char init_task_failure_data[];
151void proc_prepareexit(proc_t p, int rv, boolean_t perf_notify);
152void vfork_exit(proc_t p, int rv);
153void vproc_exit(proc_t p);
154__private_extern__ void munge_user64_rusage(struct rusage *a_rusage_p, struct user64_rusage *a_user_rusage_p);
155__private_extern__ void munge_user32_rusage(struct rusage *a_rusage_p, struct user32_rusage *a_user_rusage_p);
156static int reap_child_locked(proc_t parent, proc_t child, int deadparent, int reparentedtoinit, int locked, int droplock);
157
158/*
159 * Things which should have prototypes in headers, but don't
160 */
161void	*get_bsduthreadarg(thread_t);
162void	proc_exit(proc_t p);
163int	wait1continue(int result);
164int	waitidcontinue(int result);
165int	*get_bsduthreadrval(thread_t);
166kern_return_t sys_perf_notify(thread_t thread, int pid);
167kern_return_t task_exception_notify(exception_type_t exception,
168	mach_exception_data_type_t code, mach_exception_data_type_t subcode);
169void	delay(int);
170
171/*
172 * NOTE: Source and target may *NOT* overlap!
173 * XXX Should share code with bsd/dev/ppc/unix_signal.c
174 */
175void
176siginfo_user_to_user32(user_siginfo_t *in, user32_siginfo_t *out)
177{
178	out->si_signo	= in->si_signo;
179	out->si_errno	= in->si_errno;
180	out->si_code	= in->si_code;
181	out->si_pid	= in->si_pid;
182	out->si_uid	= in->si_uid;
183	out->si_status	= in->si_status;
184	out->si_addr	= CAST_DOWN_EXPLICIT(user32_addr_t,in->si_addr);
185	/* following cast works for sival_int because of padding */
186	out->si_value.sival_ptr	= CAST_DOWN_EXPLICIT(user32_addr_t,in->si_value.sival_ptr);
187	out->si_band	= in->si_band;			/* range reduction */
188}
189
190void
191siginfo_user_to_user64(user_siginfo_t *in, user64_siginfo_t *out)
192{
193	out->si_signo	= in->si_signo;
194	out->si_errno	= in->si_errno;
195	out->si_code	= in->si_code;
196	out->si_pid	= in->si_pid;
197	out->si_uid	= in->si_uid;
198	out->si_status	= in->si_status;
199	out->si_addr	= in->si_addr;
200	/* following cast works for sival_int because of padding */
201	out->si_value.sival_ptr	= in->si_value.sival_ptr;
202	out->si_band	= in->si_band;			/* range reduction */
203}
204
205static int
206copyoutsiginfo(user_siginfo_t *native, boolean_t is64, user_addr_t uaddr)
207{
208	if (is64) {
209		user64_siginfo_t sinfo64;
210
211		bzero(&sinfo64, sizeof (sinfo64));
212		siginfo_user_to_user64(native, &sinfo64);
213		return (copyout(&sinfo64, uaddr, sizeof (sinfo64)));
214	} else {
215		user32_siginfo_t sinfo32;
216
217		bzero(&sinfo32, sizeof (sinfo32));
218		siginfo_user_to_user32(native, &sinfo32);
219		return (copyout(&sinfo32, uaddr, sizeof (sinfo32)));
220	}
221}
222
223/*
224 * exit --
225 *	Death of process.
226 */
227void
228exit(proc_t p, struct exit_args *uap, int *retval)
229{
230	exit1(p, W_EXITCODE(uap->rval, 0), retval);
231
232	/* drop funnel before we return */
233	thread_exception_return();
234	/* NOTREACHED */
235	while (TRUE)
236		thread_block(THREAD_CONTINUE_NULL);
237	/* NOTREACHED */
238}
239
240/*
241 * Exit: deallocate address space and other resources, change proc state
242 * to zombie, and unlink proc from allproc and parent's lists.  Save exit
243 * status and rusage for wait().  Check for child processes and orphan them.
244 */
245int
246exit1(proc_t p, int rv, int *retval)
247{
248	return exit1_internal(p, rv, retval, TRUE, TRUE);
249}
250
251int
252exit1_internal(proc_t p, int rv, int *retval, boolean_t thread_can_terminate, boolean_t perf_notify)
253{
254	thread_t self = current_thread();
255	struct task *task = p->task;
256	struct uthread *ut;
257	int error = 0;
258
259	/*
260	 * If a thread in this task has already
261	 * called exit(), then halt any others
262	 * right here.
263	 */
264
265	 ut = get_bsdthread_info(self);
266	 if (ut->uu_flag & UT_VFORK) {
267		if (!thread_can_terminate) {
268			return EINVAL;
269		}
270
271		vfork_exit(p, rv);
272		vfork_return(p , retval, p->p_pid);
273		unix_syscall_return(0);
274		/* NOT REACHED */
275	 }
276
277	/*
278	 * The parameter list of audit_syscall_exit() was augmented to
279	 * take the Darwin syscall number as the first parameter,
280	 * which is currently required by mac_audit_postselect().
281	 */
282
283	/*
284	 * The BSM token contains two components: an exit status as passed
285	 * to exit(), and a return value to indicate what sort of exit it
286	 * was.  The exit status is WEXITSTATUS(rv), but it's not clear
287 	 * what the return value is.
288	 */
289	AUDIT_ARG(exit, WEXITSTATUS(rv), 0);
290	AUDIT_SYSCALL_EXIT(SYS_exit, p, ut, 0); /* Exit is always successfull */
291
292	DTRACE_PROC1(exit, int, CLD_EXITED);
293
294	/* mark process is going to exit and pull out of DBG/disk throttle */
295	proc_removethrottle(p);
296
297#if CONFIG_MEMORYSTATUS
298	memorystatus_list_remove(p->p_pid);
299#endif
300
301        proc_lock(p);
302	error = proc_transstart(p, 1);
303	if (error == EDEADLK) {
304		/* Temp: If deadlock error, then it implies multithreaded exec is
305		 * in progress. Instread of letting exit continue and
306		 * corrupting the freed memory, let the exit thread
307		 * return. This will save corruption in remote case.
308		 */
309		proc_unlock(p);
310		if (current_proc() == p){
311			thread_exception_return();
312		} else {
313			/* external termination like jetsam */
314			return(error);
315		}
316	}
317
318	while (p->exit_thread != self) {
319		if (sig_try_locked(p) <= 0) {
320			proc_transend(p, 1);
321			if (get_threadtask(self) != task) {
322				proc_unlock(p);
323				return(0);
324                        }
325			proc_unlock(p);
326
327			thread_terminate(self);
328			if (!thread_can_terminate) {
329				return 0;
330			}
331
332			thread_exception_return();
333			/* NOTREACHED */
334		}
335		sig_lock_to_exit(p);
336	}
337	if (p == initproc) {
338		proc_unlock(p);
339		printf("pid 1 exited (signal %d, exit %d)",
340		    WTERMSIG(rv), WEXITSTATUS(rv));
341		panic("%s died\nState at Last Exception:\n\n%s",
342							(p->p_comm[0] != '\0' ?
343								p->p_comm :
344								"launchd"),
345							init_task_failure_data);
346	}
347
348	p->p_lflag |= P_LEXIT;
349	p->p_xstat = rv;
350
351	proc_transend(p, 1);
352	proc_unlock(p);
353
354	proc_prepareexit(p, rv, perf_notify);
355
356	/* Last thread to terminate will call proc_exit() */
357	task_terminate_internal(task);
358
359	return(0);
360}
361
362void
363proc_prepareexit(proc_t p, int rv, boolean_t perf_notify)
364{
365	mach_exception_data_type_t code, subcode;
366	struct uthread *ut;
367	thread_t self = current_thread();
368	ut = get_bsdthread_info(self);
369
370 	/* If a core should be generated, notify crash reporter */
371	if (hassigprop(WTERMSIG(rv), SA_CORE) || ((p->p_csflags & CS_KILLED) != 0)) {
372		/*
373		 * Workaround for processes checking up on PT_DENY_ATTACH:
374		 * should be backed out post-Leopard (details in 5431025).
375		 */
376		if ((SIGSEGV == WTERMSIG(rv)) &&
377				(p->p_pptr->p_lflag & P_LNOATTACH)) {
378			goto skipcheck;
379		}
380
381		/*
382		 * Crash Reporter looks for the signal value, original exception
383		 * type, and low 20 bits of the original code in code[0]
384		 * (8, 4, and 20 bits respectively). code[1] is unmodified.
385		 */
386		code = ((WTERMSIG(rv) & 0xff) << 24) |
387			((ut->uu_exception & 0x0f) << 20) |
388			((int)ut->uu_code & 0xfffff);
389		subcode = ut->uu_subcode;
390		(void) task_exception_notify(EXC_CRASH, code, subcode);
391	}
392
393skipcheck:
394	/* Notify the perf server? */
395	if (perf_notify) {
396		(void)sys_perf_notify(self, p->p_pid);
397	}
398
399	/*
400	 * Remove proc from allproc queue and from pidhash chain.
401	 * Need to do this before we do anything that can block.
402	 * Not doing causes things like mount() find this on allproc
403	 * in partially cleaned state.
404	 */
405
406	proc_list_lock();
407
408	LIST_REMOVE(p, p_list);
409	LIST_INSERT_HEAD(&zombproc, p, p_list);	/* Place onto zombproc. */
410	/* will not be visible via proc_find */
411	p->p_listflag |= P_LIST_EXITED;
412
413	proc_list_unlock();
414
415
416#ifdef PGINPROF
417	vmsizmon();
418#endif
419	/*
420	 * If parent is waiting for us to exit or exec,
421	 * P_LPPWAIT is set; we will wakeup the parent below.
422	 */
423	proc_lock(p);
424	p->p_lflag &= ~(P_LTRACED | P_LPPWAIT);
425	p->p_sigignore = ~(sigcantmask);
426	ut->uu_siglist = 0;
427	proc_unlock(p);
428}
429
430void
431proc_exit(proc_t p)
432{
433	proc_t q;
434	proc_t pp;
435	struct task *task = p->task;
436	vnode_t tvp = NULLVP;
437	struct pgrp * pg;
438	struct session *sessp;
439	struct uthread * uth;
440	pid_t pid;
441	int exitval;
442	int knote_hint;
443
444	uth = (struct uthread *)get_bsdthread_info(current_thread());
445
446	proc_lock(p);
447	proc_transstart(p, 1);
448	if( !(p->p_lflag & P_LEXIT)) {
449		/*
450		 * This can happen if a thread_terminate() occurs
451		 * in a single-threaded process.
452		 */
453		p->p_lflag |= P_LEXIT;
454		proc_transend(p, 1);
455		proc_unlock(p);
456		proc_prepareexit(p, 0, TRUE);
457		(void) task_terminate_internal(task);
458		proc_lock(p);
459	} else {
460		proc_transend(p, 1);
461	}
462
463	p->p_lflag |= P_LPEXIT;
464
465	/*
466	 * Other kernel threads may be in the middle of signalling this process.
467	 * Wait for those threads to wrap it up before making the process
468	 * disappear on them.
469	 */
470	if ((p->p_lflag & P_LINSIGNAL) || (p->p_sigwaitcnt > 0)) {
471		p->p_sigwaitcnt++;
472		while ((p->p_lflag & P_LINSIGNAL) || (p->p_sigwaitcnt > 1))
473			msleep(&p->p_sigmask, &p->p_mlock, PWAIT, "proc_sigdrain", NULL);
474		p->p_sigwaitcnt--;
475	}
476
477	proc_unlock(p);
478	pid = p->p_pid;
479	exitval = p->p_xstat;
480	KERNEL_DEBUG_CONSTANT_IST(KDEBUG_COMMON,
481		BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXIT) | DBG_FUNC_START,
482		pid, exitval, 0, 0, 0);
483
484#if CONFIG_DTRACE
485	/*
486	 * Free any outstanding lazy dof entries. It is imperative we
487	 * always call dtrace_lazy_dofs_destroy, rather than null check
488	 * and call if !NULL. If we NULL test, during lazy dof faulting
489	 * we can race with the faulting code and proceed from here to
490	 * beyond the helpers cleanup. The lazy dof faulting will then
491	 * install new helpers which will never be cleaned up, and leak.
492	 */
493	dtrace_lazy_dofs_destroy(p);
494
495	/*
496	 * Clean up any DTrace helper actions or probes for the process.
497	 */
498	if (p->p_dtrace_helpers != NULL) {
499		(*dtrace_helpers_cleanup)(p);
500	}
501
502	/*
503	 * Clean up any DTrace probes associated with this process.
504	 */
505	/*
506	 * APPLE NOTE: We release ptss pages/entries in dtrace_fasttrap_exit_ptr(),
507	 * call this after dtrace_helpers_cleanup()
508	 */
509	proc_lock(p);
510	if (p->p_dtrace_probes && dtrace_fasttrap_exit_ptr) {
511		(*dtrace_fasttrap_exit_ptr)(p);
512	}
513	proc_unlock(p);
514#endif
515
516	/* XXX Zombie allocation may fail, in which case stats get lost */
517	MALLOC_ZONE(p->p_ru, struct rusage *,
518			sizeof (*p->p_ru), M_ZOMBIE, M_WAITOK);
519
520	nspace_proc_exit(p);
521
522#if VM_PRESSURE_EVENTS
523	vm_pressure_proc_cleanup(p);
524#endif
525
526	/*
527	 * need to cancel async IO requests that can be cancelled and wait for those
528	 * already active.  MAY BLOCK!
529	 */
530
531	proc_refdrain(p);
532
533	/* if any pending cpu limits action, clear it */
534	task_clear_cpuusage(p->task);
535
536	workqueue_mark_exiting(p);
537	workqueue_exit(p);
538
539	_aio_exit( p );
540
541	/*
542	 * Close open files and release open-file table.
543	 * This may block!
544	 */
545	fdfree(p);
546
547	if (uth->uu_lowpri_window) {
548	        /*
549		 * task is marked as a low priority I/O type
550		 * and the I/O we issued while in flushing files on close
551		 * collided with normal I/O operations...
552		 * no need to throttle this thread since its going away
553		 * but we do need to update our bookeeping w/r to throttled threads
554		 */
555		throttle_lowpri_io(FALSE);
556	}
557
558#if !CONFIG_EMBEDDED
559	if (p->p_legacy_behavior & PROC_LEGACY_BEHAVIOR_IOTHROTTLE) {
560		throttle_legacy_process_decr();
561	}
562#endif
563
564#if SYSV_SHM
565	/* Close ref SYSV Shared memory*/
566	if (p->vm_shm)
567		shmexit(p);
568#endif
569#if SYSV_SEM
570	/* Release SYSV semaphores */
571	semexit(p);
572#endif
573
574#if PSYNCH
575	pth_proc_hashdelete(p);
576#endif /* PSYNCH */
577
578	sessp = proc_session(p);
579	if (SESS_LEADER(p, sessp)) {
580
581		if (sessp->s_ttyvp != NULLVP) {
582			struct vnode *ttyvp;
583			int ttyvid;
584			struct vfs_context context;
585			struct tty *tp;
586
587			/*
588			 * Controlling process.
589			 * Signal foreground pgrp,
590			 * drain controlling terminal
591			 * and revoke access to controlling terminal.
592			 */
593			session_lock(sessp);
594			tp = SESSION_TP(sessp);
595			if ((tp != TTY_NULL) && (tp->t_session == sessp)) {
596				session_unlock(sessp);
597
598				tty_pgsignal(tp, SIGHUP, 1);
599
600				session_lock(sessp);
601				tp = SESSION_TP(sessp);
602			}
603			ttyvp = sessp->s_ttyvp;
604			ttyvid = sessp->s_ttyvid;
605			sessp->s_ttyvp = NULLVP;
606			sessp->s_ttyvid = 0;
607			sessp->s_ttyp = TTY_NULL;
608			sessp->s_ttypgrpid = NO_PID;
609			session_unlock(sessp);
610
611			if ((ttyvp != NULLVP) && (vnode_getwithvid(ttyvp, ttyvid) == 0)) {
612				if (tp != TTY_NULL) {
613					tty_lock(tp);
614					(void) ttywait(tp);
615					tty_unlock(tp);
616				}
617				context.vc_thread = proc_thread(p); /* XXX */
618				context.vc_ucred = kauth_cred_proc_ref(p);
619				vnode_rele(ttyvp);
620				VNOP_REVOKE(ttyvp, REVOKEALL, &context);
621				vnode_put(ttyvp);
622				kauth_cred_unref(&context.vc_ucred);
623				ttyvp = NULLVP;
624			}
625			if (ttyvp)
626				vnode_rele(ttyvp);
627			if (tp)
628				ttyfree(tp);
629		}
630		session_lock(sessp);
631		sessp->s_leader = NULL;
632		session_unlock(sessp);
633	}
634	session_rele(sessp);
635
636	pg = proc_pgrp(p);
637	fixjobc(p, pg, 0);
638	pg_rele(pg);
639
640	p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
641	(void)acct_process(p);
642
643	proc_list_lock();
644
645	if ((p->p_listflag & P_LIST_EXITCOUNT) == P_LIST_EXITCOUNT) {
646		p->p_listflag &= ~P_LIST_EXITCOUNT;
647		proc_shutdown_exitcount--;
648		if (proc_shutdown_exitcount == 0)
649			wakeup(&proc_shutdown_exitcount);
650	}
651
652	/* wait till parentrefs are dropped and grant no more */
653	proc_childdrainstart(p);
654	while ((q = p->p_children.lh_first) != NULL) {
655		int reparentedtoinit = (q->p_listflag & P_LIST_DEADPARENT) ? 1 : 0;
656		q->p_listflag |= P_LIST_DEADPARENT;
657		if (q->p_stat == SZOMB) {
658			if (p != q->p_pptr)
659				panic("parent child linkage broken");
660			/* check for sysctl zomb lookup */
661			while ((q->p_listflag & P_LIST_WAITING) == P_LIST_WAITING) {
662				msleep(&q->p_stat, proc_list_mlock, PWAIT, "waitcoll", 0);
663			}
664			q->p_listflag |= P_LIST_WAITING;
665			/*
666			 * This is a named reference and it is not granted
667			 * if the reap is already in progress. So we get
668			 * the reference here exclusively and their can be
669			 * no waiters. So there is no need for a wakeup
670			 * after we are done.  Also the reap frees the structure
671			 * and the proc struct cannot be used for wakeups as well.
672			 * It is safe to use q here as this is system reap
673			 */
674			(void)reap_child_locked(p, q, 1, reparentedtoinit, 1, 0);
675		} else {
676			proc_reparentlocked(q, initproc, 0, 1);
677			/*
678		 	* Traced processes are killed
679		 	* since their existence means someone is messing up.
680		 	*/
681			if (q->p_lflag & P_LTRACED) {
682				/*
683				 * Take a reference on the child process to
684				 * ensure it doesn't exit and disappear between
685				 * the time we drop the list_lock and attempt
686				 * to acquire its proc_lock.
687				 */
688				if (proc_ref_locked(q) != q)
689					continue;
690
691				proc_list_unlock();
692				proc_lock(q);
693				q->p_lflag &= ~P_LTRACED;
694				if (q->sigwait_thread) {
695					thread_t thread = q->sigwait_thread;
696
697					proc_unlock(q);
698					/*
699				 	* The sigwait_thread could be stopped at a
700				 	* breakpoint. Wake it up to kill.
701				 	* Need to do this as it could be a thread which is not
702				 	* the first thread in the task. So any attempts to kill
703				 	* the process would result into a deadlock on q->sigwait.
704				 	*/
705					thread_resume(thread);
706					clear_wait(thread, THREAD_INTERRUPTED);
707					threadsignal(thread, SIGKILL, 0);
708				} else {
709					proc_unlock(q);
710				}
711
712				psignal(q, SIGKILL);
713				proc_list_lock();
714				proc_rele_locked(q);
715			}
716		}
717	}
718
719	proc_childdrainend(p);
720	proc_list_unlock();
721
722	/*
723	 * Release reference to text vnode
724	 */
725	tvp = p->p_textvp;
726	p->p_textvp = NULL;
727	if (tvp != NULLVP) {
728		vnode_rele(tvp);
729	}
730
731	/*
732	 * Save exit status and final rusage info, adding in child rusage
733	 * info and self times.  If we were unable to allocate a zombie
734	 * structure, this information is lost.
735	 */
736	/* No need for locking here as no one than this thread can access this */
737	if (p->p_ru != NULL) {
738	    calcru(p, &p->p_stats->p_ru.ru_utime, &p->p_stats->p_ru.ru_stime, NULL);
739	    *p->p_ru = p->p_stats->p_ru;
740
741	    ruadd(p->p_ru, &p->p_stats->p_cru);
742	}
743
744	/*
745	 * Free up profiling buffers.
746	 */
747	{
748		struct uprof *p0 = &p->p_stats->p_prof, *p1, *pn;
749
750		p1 = p0->pr_next;
751		p0->pr_next = NULL;
752		p0->pr_scale = 0;
753
754		for (; p1 != NULL; p1 = pn) {
755			pn = p1->pr_next;
756			kfree(p1, sizeof *p1);
757		}
758	}
759
760	proc_spinlock(p);
761	if (thread_call_cancel(p->p_rcall))
762		p->p_ractive--;
763
764	while (p->p_ractive > 0) {
765		proc_spinunlock(p);
766
767		delay(1);
768
769		proc_spinlock(p);
770	}
771	proc_spinunlock(p);
772
773	thread_call_free(p->p_rcall);
774	p->p_rcall = NULL;
775
776	/*
777	 * Other substructures are freed from wait().
778	 */
779	FREE_ZONE(p->p_stats, sizeof *p->p_stats, M_PSTATS);
780	p->p_stats = NULL;
781
782	FREE_ZONE(p->p_sigacts, sizeof *p->p_sigacts, M_SIGACTS);
783	p->p_sigacts = NULL;
784
785	proc_limitdrop(p, 1);
786	p->p_limit = NULL;
787
788
789	/*
790	 * Finish up by terminating the task
791	 * and halt this thread (only if a
792	 * member of the task exiting).
793	 */
794	p->task = TASK_NULL;
795	set_bsdtask_info(task, NULL);
796
797	knote_hint = NOTE_EXIT | (p->p_xstat & 0xffff);
798	proc_knote(p, knote_hint);
799
800	/* mark the thread as the one that is doing proc_exit
801	 * no need to hold proc lock in uthread_free
802	 */
803	uth->uu_flag |= UT_PROCEXIT;
804	/*
805	 * Notify parent that we're gone.
806	 */
807	pp = proc_parent(p);
808	if (pp->p_flag & P_NOCLDWAIT) {
809
810#if 3839178
811		/*
812		 * If the parent is ignoring SIGCHLD, then POSIX requires
813		 * us to not add the resource usage to the parent process -
814		 * we are only going to hand it off to init to get reaped.
815		 * We should contest the standard in this case on the basis
816		 * of RLIMIT_CPU.
817		 */
818#else	/* !3839178 */
819		/*
820		 * Add child resource usage to parent before giving
821		 * zombie to init.  If we were unable to allocate a
822		 * zombie structure, this information is lost.
823		 */
824		if (p->p_ru != NULL) {
825			proc_lock(pp);
826			ruadd(&pp->p_stats->p_cru, p->p_ru);
827			proc_unlock(pp);
828		}
829#endif	/* !3839178 */
830
831		/* kernel can reap this one, no need to move it to launchd */
832		proc_list_lock();
833		p->p_listflag |= P_LIST_DEADPARENT;
834		proc_list_unlock();
835	}
836	if ((p->p_listflag & P_LIST_DEADPARENT) == 0 || p->p_oppid) {
837		if (pp != initproc) {
838			proc_lock(pp);
839			pp->si_pid = p->p_pid;
840			pp->si_status = p->p_xstat;
841			pp->si_code = CLD_EXITED;
842			/*
843			 * p_ucred usage is safe as it is an exiting process
844			 * and reference is dropped in reap
845			 */
846			pp->si_uid = kauth_cred_getruid(p->p_ucred);
847			proc_unlock(pp);
848		}
849		/* mark as a zombie */
850		/* No need to take proc lock as all refs are drained and
851		 * no one except parent (reaping ) can look at this.
852		 * The write is to an int and is coherent. Also parent is
853		 *  keyed off of list lock for reaping
854		 */
855		KERNEL_DEBUG_CONSTANT_IST(KDEBUG_COMMON,
856			BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXIT) | DBG_FUNC_END,
857			pid, exitval, 0, 0, 0);
858		p->p_stat = SZOMB;
859		/*
860		 * The current process can be reaped so, no one
861		 * can depend on this
862		 */
863
864		psignal(pp, SIGCHLD);
865
866		/* and now wakeup the parent */
867		proc_list_lock();
868		wakeup((caddr_t)pp);
869		proc_list_unlock();
870	} else {
871		/* should be fine as parent proc would be initproc */
872		/* mark as a zombie */
873		/* No need to take proc lock as all refs are drained and
874		 * no one except parent (reaping ) can look at this.
875		 * The write is to an int and is coherent. Also parent is
876		 *  keyed off of list lock for reaping
877		 */
878		proc_list_lock();
879		KERNEL_DEBUG_CONSTANT_IST(KDEBUG_COMMON,
880			BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXIT) | DBG_FUNC_END,
881			pid, exitval, 0, 0, 0);
882		/* check for sysctl zomb lookup */
883		while ((p->p_listflag & P_LIST_WAITING) == P_LIST_WAITING) {
884			msleep(&p->p_stat, proc_list_mlock, PWAIT, "waitcoll", 0);
885		}
886		/* safe to use p as this is a system reap */
887		p->p_stat = SZOMB;
888		p->p_listflag |= P_LIST_WAITING;
889
890		/*
891		 * This is a named reference and it is not granted
892		 * if the reap is already in progress. So we get
893		 * the reference here exclusively and their can be
894		 * no waiters. So there is no need for a wakeup
895		 * after we are done. AlsO  the reap frees the structure
896		 * and the proc struct cannot be used for wakeups as well.
897		 * It is safe to use p here as this is system reap
898		 */
899		(void)reap_child_locked(pp, p, 1, 0, 1, 1);
900		/* list lock dropped by reap_child_locked */
901	}
902	if (uth->uu_lowpri_window) {
903	        /*
904		 * task is marked as a low priority I/O type and we've
905		 * somehow picked up another throttle during exit processing...
906		 * no need to throttle this thread since its going away
907		 * but we do need to update our bookeeping w/r to throttled threads
908		 */
909		throttle_lowpri_io(FALSE);
910	}
911
912	proc_rele(pp);
913
914}
915
916
917/*
918 * reap_child_locked
919 *
920 * Description:	Given a process from which all status information needed
921 *		has already been extracted, if the process is a ptrace
922 *		attach process, detach it and give it back to its real
923 *		parent, else recover all resources remaining associated
924 *		with it.
925 *
926 * Parameters:	proc_t parent		Parent of process being reaped
927 *		proc_t child		Process to reap
928 *
929 * Returns:	0			Process was not reaped because it
930 *					came from an attach
931 *		1			Process was reaped
932 */
933static int
934reap_child_locked(proc_t parent, proc_t child, int deadparent, int reparentedtoinit, int locked, int droplock)
935{
936	proc_t trace_parent = PROC_NULL;	/* Traced parent process, if tracing */
937
938	if (locked == 1)
939		proc_list_unlock();
940
941	/*
942	 * If we got the child via a ptrace 'attach',
943	 * we need to give it back to the old parent.
944	 *
945	 * Exception: someone who has been reparented to launchd before being
946	 * ptraced can simply be reaped, refer to radar 5677288
947	 * 	p_oppid 		 -> ptraced
948	 * 	trace_parent == initproc -> away from launchd
949	 * 	reparentedtoinit	 -> came to launchd by reparenting
950	 */
951	if (child->p_oppid) {
952		int knote_hint;
953		pid_t oppid;
954
955		proc_lock(child);
956		oppid = child->p_oppid;
957		child->p_oppid = 0;
958		knote_hint = NOTE_EXIT | (child->p_xstat & 0xffff);
959		proc_unlock(child);
960
961		if ((trace_parent = proc_find(oppid))
962			&& !((trace_parent == initproc) && reparentedtoinit)) {
963
964			if (trace_parent != initproc) {
965				/*
966				 * proc internal fileds  and p_ucred usage safe
967				 * here as child is dead and is not reaped or
968				 * reparented yet
969				 */
970				proc_lock(trace_parent);
971				trace_parent->si_pid = child->p_pid;
972				trace_parent->si_status = child->p_xstat;
973				trace_parent->si_code = CLD_CONTINUED;
974				trace_parent->si_uid = kauth_cred_getruid(child->p_ucred);
975				proc_unlock(trace_parent);
976			}
977			proc_reparentlocked(child, trace_parent, 1, 0);
978
979			/* resend knote to original parent (and others) after reparenting */
980			proc_knote(child, knote_hint);
981
982			psignal(trace_parent, SIGCHLD);
983			proc_list_lock();
984			wakeup((caddr_t)trace_parent);
985			child->p_listflag &= ~P_LIST_WAITING;
986			wakeup(&child->p_stat);
987			proc_list_unlock();
988			proc_rele(trace_parent);
989			if ((locked == 1) && (droplock == 0))
990				proc_list_lock();
991			return (0);
992		}
993
994		/*
995		 * If we can't reparent (e.g. the original parent exited while child was being debugged, or
996		 * original parent is the same as the debugger currently exiting), we still need to satisfy
997		 * the knote lifecycle for other observers on the system. While the debugger was attached,
998		 * the NOTE_EXIT would not have been broadcast during initial child termination.
999		 */
1000		proc_knote(child, knote_hint);
1001
1002		if (trace_parent != PROC_NULL) {
1003			proc_rele(trace_parent);
1004		}
1005	}
1006
1007	proc_knote(child, NOTE_REAP);
1008	proc_knote_drain(child);
1009
1010	child->p_xstat = 0;
1011	if (child->p_ru) {
1012		proc_lock(parent);
1013#if 3839178
1014		/*
1015		 * If the parent is ignoring SIGCHLD, then POSIX requires
1016		 * us to not add the resource usage to the parent process -
1017		 * we are only going to hand it off to init to get reaped.
1018		 * We should contest the standard in this case on the basis
1019		 * of RLIMIT_CPU.
1020		 */
1021		if (!(parent->p_flag & P_NOCLDWAIT))
1022#endif	/* 3839178 */
1023			ruadd(&parent->p_stats->p_cru, child->p_ru);
1024		proc_unlock(parent);
1025		FREE_ZONE(child->p_ru, sizeof *child->p_ru, M_ZOMBIE);
1026		child->p_ru = NULL;
1027	} else {
1028		printf("Warning : lost p_ru for %s\n", child->p_comm);
1029	}
1030
1031	AUDIT_SESSION_PROCEXIT(child);
1032
1033	/*
1034	 * Decrement the count of procs running with this uid.
1035	 * p_ucred usage is safe here as it is an exited process.
1036	 * and refernce is dropped after these calls down below
1037	 * (locking protection is provided by list lock held in chgproccnt)
1038	 */
1039	(void)chgproccnt(kauth_cred_getruid(child->p_ucred), -1);
1040
1041#if CONFIG_LCTX
1042	ALLLCTX_LOCK;
1043	leavelctx(child);
1044	ALLLCTX_UNLOCK;
1045#endif
1046
1047	/*
1048	 * Free up credentials.
1049	 */
1050	if (IS_VALID_CRED(child->p_ucred)) {
1051		kauth_cred_unref(&child->p_ucred);
1052	}
1053
1054	/*  XXXX Note NOT SAFE TO USE p_ucred from this point onwards */
1055
1056	/*
1057	 * Finally finished with old proc entry.
1058	 * Unlink it from its process group and free it.
1059	 */
1060	leavepgrp(child);
1061
1062	proc_list_lock();
1063	LIST_REMOVE(child, p_list);	/* off zombproc */
1064	parent->p_childrencnt--;
1065	LIST_REMOVE(child, p_sibling);
1066	/* If there are no more children wakeup parent */
1067	if ((deadparent != 0) && (LIST_EMPTY(&parent->p_children)))
1068		wakeup((caddr_t)parent);	/* with list lock held */
1069	child->p_listflag &= ~P_LIST_WAITING;
1070	wakeup(&child->p_stat);
1071
1072	/* Take it out of process hash */
1073	LIST_REMOVE(child, p_hash);
1074	child->p_listflag &= ~P_LIST_INHASH;
1075	proc_checkdeadrefs(child);
1076	nprocs--;
1077
1078	proc_list_unlock();
1079
1080#if CONFIG_FINE_LOCK_GROUPS
1081	lck_mtx_destroy(&child->p_mlock, proc_mlock_grp);
1082	lck_mtx_destroy(&child->p_fdmlock, proc_fdmlock_grp);
1083#if CONFIG_DTRACE
1084	lck_mtx_destroy(&child->p_dtrace_sprlock, proc_lck_grp);
1085#endif
1086	lck_spin_destroy(&child->p_slock, proc_slock_grp);
1087#else /* CONFIG_FINE_LOCK_GROUPS */
1088	lck_mtx_destroy(&child->p_mlock, proc_lck_grp);
1089	lck_mtx_destroy(&child->p_fdmlock, proc_lck_grp);
1090#if CONFIG_DTRACE
1091	lck_mtx_destroy(&child->p_dtrace_sprlock, proc_lck_grp);
1092#endif
1093	lck_spin_destroy(&child->p_slock, proc_lck_grp);
1094#endif /* CONFIG_FINE_LOCK_GROUPS */
1095	workqueue_destroy_lock(child);
1096
1097	FREE_ZONE(child, sizeof *child, M_PROC);
1098	if ((locked == 1) && (droplock == 0))
1099		proc_list_lock();
1100
1101	return (1);
1102}
1103
1104
1105int
1106wait1continue(int result)
1107{
1108	void *vt;
1109	thread_t thread;
1110	int *retval;
1111	proc_t p;
1112
1113	if (result)
1114		return(result);
1115
1116	p = current_proc();
1117	thread = current_thread();
1118	vt = get_bsduthreadarg(thread);
1119	retval = get_bsduthreadrval(thread);
1120	return(wait4(p, (struct wait4_args *)vt, retval));
1121}
1122
1123int
1124wait4(proc_t q, struct wait4_args *uap, int32_t *retval)
1125{
1126	__pthread_testcancel(1);
1127	return(wait4_nocancel(q, (struct wait4_nocancel_args *)uap, retval));
1128}
1129
1130int
1131wait4_nocancel(proc_t q, struct wait4_nocancel_args *uap, int32_t *retval)
1132{
1133	int nfound;
1134	int sibling_count;
1135	proc_t p;
1136	int status, error;
1137
1138	AUDIT_ARG(pid, uap->pid);
1139
1140	if (uap->pid == 0)
1141		uap->pid = -q->p_pgrpid;
1142
1143loop:
1144	proc_list_lock();
1145loop1:
1146	nfound = 0;
1147	sibling_count = 0;
1148
1149	for (p = q->p_children.lh_first; p != 0; p = p->p_sibling.le_next) {
1150		if ( p->p_sibling.le_next != 0 )
1151			sibling_count++;
1152		if (uap->pid != WAIT_ANY &&
1153		    p->p_pid != uap->pid &&
1154		    p->p_pgrpid != -(uap->pid))
1155			continue;
1156
1157		nfound++;
1158
1159		/* XXX This is racy because we don't get the lock!!!! */
1160
1161		if (p->p_listflag & P_LIST_WAITING) {
1162			(void)msleep(&p->p_stat, proc_list_mlock, PWAIT, "waitcoll", 0);
1163			goto loop1;
1164		}
1165		p->p_listflag |= P_LIST_WAITING;   /* only allow single thread to wait() */
1166
1167
1168		if (p->p_stat == SZOMB) {
1169			int reparentedtoinit = (p->p_listflag & P_LIST_DEADPARENT) ? 1 : 0;
1170
1171			proc_list_unlock();
1172#if CONFIG_MACF
1173			if ((error = mac_proc_check_wait(q, p)) != 0)
1174				goto out;
1175#endif
1176			retval[0] = p->p_pid;
1177			if (uap->status) {
1178				/* Legacy apps expect only 8 bits of status */
1179				status = 0xffff & p->p_xstat;	/* convert to int */
1180				error = copyout((caddr_t)&status,
1181				   			uap->status,
1182						    sizeof(status));
1183				if (error)
1184					goto out;
1185			}
1186			if (uap->rusage) {
1187				if (p->p_ru == NULL) {
1188					error = ENOMEM;
1189				} else {
1190					if (IS_64BIT_PROCESS(q)) {
1191						struct user64_rusage	my_rusage;
1192						munge_user64_rusage(p->p_ru, &my_rusage);
1193						error = copyout((caddr_t)&my_rusage,
1194							uap->rusage,
1195							sizeof (my_rusage));
1196					}
1197					else {
1198						struct user32_rusage	my_rusage;
1199						munge_user32_rusage(p->p_ru, &my_rusage);
1200						error = copyout((caddr_t)&my_rusage,
1201							uap->rusage,
1202							sizeof (my_rusage));
1203					}
1204				}
1205				/* information unavailable? */
1206				if (error)
1207					goto out;
1208			}
1209
1210			/* Conformance change for 6577252.
1211			 * When SIGCHLD is blocked and wait() returns because the status
1212			 * of a child process is available and there are no other
1213			 * children processes, then any pending SIGCHLD signal is cleared.
1214			 */
1215			if ( sibling_count == 0 ) {
1216				int mask = sigmask(SIGCHLD);
1217				uthread_t uth = (struct uthread *)get_bsdthread_info(current_thread());
1218
1219				if ( (uth->uu_sigmask & mask) != 0 ) {
1220					/* we are blocking SIGCHLD signals.  clear any pending SIGCHLD.
1221					 * This locking looks funny but it is protecting access to the
1222					 * thread via p_uthlist.
1223					 */
1224					proc_lock(q);
1225					uth->uu_siglist &= ~mask;	/* clear pending signal */
1226					proc_unlock(q);
1227				}
1228			}
1229
1230			/* Clean up */
1231			(void)reap_child_locked(q, p, 0, reparentedtoinit, 0, 0);
1232
1233			return (0);
1234		}
1235		if (p->p_stat == SSTOP && (p->p_lflag & P_LWAITED) == 0 &&
1236		    (p->p_lflag & P_LTRACED || uap->options & WUNTRACED)) {
1237			proc_list_unlock();
1238#if CONFIG_MACF
1239			if ((error = mac_proc_check_wait(q, p)) != 0)
1240				goto out;
1241#endif
1242			proc_lock(p);
1243			p->p_lflag |= P_LWAITED;
1244			proc_unlock(p);
1245			retval[0] = p->p_pid;
1246			if (uap->status) {
1247				status = W_STOPCODE(p->p_xstat);
1248				error = copyout((caddr_t)&status,
1249					uap->status,
1250				    sizeof(status));
1251			} else
1252				error = 0;
1253			goto out;
1254		}
1255		/*
1256		 * If we are waiting for continued processses, and this
1257		 * process was continued
1258		 */
1259		if ((uap->options & WCONTINUED) &&
1260		    (p->p_flag & P_CONTINUED)) {
1261			proc_list_unlock();
1262#if CONFIG_MACF
1263			if ((error = mac_proc_check_wait(q, p)) != 0)
1264				goto out;
1265#endif
1266
1267			/* Prevent other process for waiting for this event */
1268			OSBitAndAtomic(~((uint32_t)P_CONTINUED), &p->p_flag);
1269			retval[0] = p->p_pid;
1270			if (uap->status) {
1271				status = W_STOPCODE(SIGCONT);
1272				error = copyout((caddr_t)&status,
1273					uap->status,
1274				    sizeof(status));
1275			} else
1276				error = 0;
1277			goto out;
1278		}
1279		p->p_listflag &= ~P_LIST_WAITING;
1280		wakeup(&p->p_stat);
1281	}
1282	/* list lock is held when we get here any which way */
1283	if (nfound == 0) {
1284		proc_list_unlock();
1285		return (ECHILD);
1286	}
1287
1288	if (uap->options & WNOHANG) {
1289		retval[0] = 0;
1290		proc_list_unlock();
1291		return (0);
1292	}
1293
1294	if ((error = msleep0((caddr_t)q, proc_list_mlock, PWAIT | PCATCH | PDROP, "wait", 0, wait1continue)))
1295		return (error);
1296
1297	goto loop;
1298out:
1299	proc_list_lock();
1300	p->p_listflag &= ~P_LIST_WAITING;
1301	wakeup(&p->p_stat);
1302	proc_list_unlock();
1303	return (error);
1304}
1305
1306#if DEBUG
1307#define ASSERT_LCK_MTX_OWNED(lock)	\
1308				lck_mtx_assert(lock, LCK_MTX_ASSERT_OWNED)
1309#else
1310#define ASSERT_LCK_MTX_OWNED(lock)	/* nothing */
1311#endif
1312
1313int
1314waitidcontinue(int result)
1315{
1316	void *vt;
1317	thread_t thread;
1318	int *retval;
1319
1320	if (result)
1321		return (result);
1322
1323	thread = current_thread();
1324	vt = get_bsduthreadarg(thread);
1325	retval = get_bsduthreadrval(thread);
1326	return (waitid(current_proc(), (struct waitid_args *)vt, retval));
1327}
1328
1329/*
1330 * Description:	Suspend the calling thread until one child of the process
1331 *		containing the calling thread changes state.
1332 *
1333 * Parameters:	uap->idtype		one of P_PID, P_PGID, P_ALL
1334 *		uap->id			pid_t or gid_t or ignored
1335 *		uap->infop		Address of siginfo_t struct in
1336 *					user space into which to return status
1337 *		uap->options		flag values
1338 *
1339 * Returns:	0			Success
1340 *		!0			Error returning status to user space
1341 */
1342int
1343waitid(proc_t q, struct waitid_args *uap, int32_t *retval)
1344{
1345	__pthread_testcancel(1);
1346	return (waitid_nocancel(q, (struct waitid_nocancel_args *)uap, retval));
1347}
1348
1349int
1350waitid_nocancel(proc_t q, struct waitid_nocancel_args *uap,
1351	__unused int32_t *retval)
1352{
1353	user_siginfo_t	siginfo;	/* siginfo data to return to caller */
1354	boolean_t caller64 = IS_64BIT_PROCESS(q);
1355	int nfound;
1356	proc_t p;
1357	int error;
1358
1359	if (uap->options == 0 ||
1360	    (uap->options & ~(WNOHANG|WNOWAIT|WCONTINUED|WSTOPPED|WEXITED)))
1361		return (EINVAL);	/* bits set that aren't recognized */
1362
1363	switch (uap->idtype) {
1364	case P_PID:	/* child with process ID equal to... */
1365	case P_PGID:	/* child with process group ID equal to... */
1366		if (((int)uap->id) < 0)
1367			return (EINVAL);
1368		break;
1369	case P_ALL:	/* any child */
1370		break;
1371	}
1372
1373loop:
1374	proc_list_lock();
1375loop1:
1376	nfound = 0;
1377	for (p = q->p_children.lh_first; p != 0; p = p->p_sibling.le_next) {
1378
1379		switch (uap->idtype) {
1380		case P_PID:	/* child with process ID equal to... */
1381			if (p->p_pid != (pid_t)uap->id)
1382				continue;
1383			break;
1384		case P_PGID:	/* child with process group ID equal to... */
1385			if (p->p_pgrpid != (pid_t)uap->id)
1386				continue;
1387			break;
1388		case P_ALL:	/* any child */
1389			break;
1390		}
1391
1392		/* XXX This is racy because we don't get the lock!!!! */
1393
1394		/*
1395		 * Wait collision; go to sleep and restart; used to maintain
1396		 * the single return for waited process guarantee.
1397		 */
1398		if (p->p_listflag & P_LIST_WAITING) {
1399			(void) msleep(&p->p_stat, proc_list_mlock,
1400				PWAIT, "waitidcoll", 0);
1401			goto loop1;
1402		}
1403		p->p_listflag |= P_LIST_WAITING;		/* mark busy */
1404
1405		nfound++;
1406
1407		bzero(&siginfo, sizeof (siginfo));
1408
1409		switch (p->p_stat) {
1410		case SZOMB:		/* Exited */
1411			if (!(uap->options & WEXITED))
1412				break;
1413			proc_list_unlock();
1414#if CONFIG_MACF
1415			if ((error = mac_proc_check_wait(q, p)) != 0)
1416				goto out;
1417#endif
1418			siginfo.si_signo = SIGCHLD;
1419			siginfo.si_pid = p->p_pid;
1420			siginfo.si_status = WEXITSTATUS(p->p_xstat);
1421			if (WIFSIGNALED(p->p_xstat)) {
1422				siginfo.si_code = WCOREDUMP(p->p_xstat) ?
1423					CLD_DUMPED : CLD_KILLED;
1424			} else
1425				siginfo.si_code = CLD_EXITED;
1426
1427			if ((error = copyoutsiginfo(&siginfo,
1428			    caller64, uap->infop)) != 0)
1429				goto out;
1430
1431			/* Prevent other process for waiting for this event? */
1432			if (!(uap->options & WNOWAIT)) {
1433				(void) reap_child_locked(q, p, 0, 0, 0, 0);
1434				return (0);
1435			}
1436			goto out;
1437
1438		case SSTOP:		/* Stopped */
1439			/*
1440			 * If we are not interested in stopped processes, then
1441			 * ignore this one.
1442			 */
1443			if (!(uap->options & WSTOPPED))
1444				break;
1445
1446			/*
1447			 * If someone has already waited it, we lost a race
1448			 * to be the one to return status.
1449			 */
1450			if ((p->p_lflag & P_LWAITED) != 0)
1451				break;
1452			proc_list_unlock();
1453#if CONFIG_MACF
1454			if ((error = mac_proc_check_wait(q, p)) != 0)
1455				goto out;
1456#endif
1457			siginfo.si_signo = SIGCHLD;
1458			siginfo.si_pid = p->p_pid;
1459			siginfo.si_status = p->p_xstat;	/* signal number */
1460			siginfo.si_code = CLD_STOPPED;
1461
1462			if ((error = copyoutsiginfo(&siginfo,
1463			    caller64, uap->infop)) != 0)
1464				goto out;
1465
1466			/* Prevent other process for waiting for this event? */
1467			if (!(uap->options & WNOWAIT)) {
1468				proc_lock(p);
1469				p->p_lflag |= P_LWAITED;
1470				proc_unlock(p);
1471			}
1472			goto out;
1473
1474		default:		/* All other states => Continued */
1475			if (!(uap->options & WCONTINUED))
1476				break;
1477
1478			/*
1479			 * If the flag isn't set, then this process has not
1480			 * been stopped and continued, or the status has
1481			 * already been reaped by another caller of waitid().
1482			 */
1483			if ((p->p_flag & P_CONTINUED) == 0)
1484				break;
1485			proc_list_unlock();
1486#if CONFIG_MACF
1487			if ((error = mac_proc_check_wait(q, p)) != 0)
1488				goto out;
1489#endif
1490			siginfo.si_signo = SIGCHLD;
1491			siginfo.si_code = CLD_CONTINUED;
1492			proc_lock(p);
1493			siginfo.si_pid = p->p_contproc;
1494			siginfo.si_status = p->p_xstat;
1495			proc_unlock(p);
1496
1497			if ((error = copyoutsiginfo(&siginfo,
1498			    caller64, uap->infop)) != 0)
1499				goto out;
1500
1501			/* Prevent other process for waiting for this event? */
1502			if (!(uap->options & WNOWAIT)) {
1503				OSBitAndAtomic(~((uint32_t)P_CONTINUED),
1504				    &p->p_flag);
1505			}
1506			goto out;
1507		}
1508		ASSERT_LCK_MTX_OWNED(proc_list_mlock);
1509
1510		/* Not a process we are interested in; go on to next child */
1511
1512		p->p_listflag &= ~P_LIST_WAITING;
1513		wakeup(&p->p_stat);
1514	}
1515	ASSERT_LCK_MTX_OWNED(proc_list_mlock);
1516
1517	/* No child processes that could possibly satisfy the request? */
1518
1519	if (nfound == 0) {
1520		proc_list_unlock();
1521		return (ECHILD);
1522	}
1523
1524	if (uap->options & WNOHANG) {
1525		proc_list_unlock();
1526#if CONFIG_MACF
1527		if ((error = mac_proc_check_wait(q, p)) != 0)
1528			return (error);
1529#endif
1530		/*
1531		 * The state of the siginfo structure in this case
1532		 * is undefined.  Some implementations bzero it, some
1533		 * (like here) leave it untouched for efficiency.
1534		 *
1535		 * Thus the most portable check for "no matching pid with
1536		 * WNOHANG" is to store a zero into si_pid before
1537		 * invocation, then check for a non-zero value afterwards.
1538		 */
1539		return (0);
1540	}
1541
1542	if ((error = msleep0(q, proc_list_mlock,
1543	    PWAIT | PCATCH | PDROP, "waitid", 0, waitidcontinue)) != 0)
1544		return (error);
1545
1546	goto loop;
1547out:
1548	proc_list_lock();
1549	p->p_listflag &= ~P_LIST_WAITING;
1550	wakeup(&p->p_stat);
1551	proc_list_unlock();
1552	return (error);
1553}
1554
1555/*
1556 * make process 'parent' the new parent of process 'child'.
1557 */
1558void
1559proc_reparentlocked(proc_t child, proc_t parent, int cansignal, int locked)
1560{
1561	proc_t oldparent = PROC_NULL;
1562
1563	if (child->p_pptr == parent)
1564		return;
1565
1566	if (locked == 0)
1567		proc_list_lock();
1568
1569	oldparent = child->p_pptr;
1570#if __PROC_INTERNAL_DEBUG
1571	if (oldparent == PROC_NULL)
1572		panic("proc_reparent: process %p does not have a parent\n", child);
1573#endif
1574
1575	LIST_REMOVE(child, p_sibling);
1576#if __PROC_INTERNAL_DEBUG
1577	if (oldparent->p_childrencnt == 0)
1578		panic("process children count already 0\n");
1579#endif
1580	oldparent->p_childrencnt--;
1581#if __PROC_INTERNAL_DEBUG1
1582	if (oldparent->p_childrencnt < 0)
1583		panic("process children count -ve\n");
1584#endif
1585	LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
1586	parent->p_childrencnt++;
1587	child->p_pptr = parent;
1588	child->p_ppid = parent->p_pid;
1589
1590	proc_list_unlock();
1591
1592	if ((cansignal != 0) && (initproc == parent) && (child->p_stat == SZOMB))
1593		psignal(initproc, SIGCHLD);
1594	if (locked == 1)
1595		proc_list_lock();
1596}
1597
1598/*
1599 * Exit: deallocate address space and other resources, change proc state
1600 * to zombie, and unlink proc from allproc and parent's lists.  Save exit
1601 * status and rusage for wait().  Check for child processes and orphan them.
1602 */
1603
1604void
1605vfork_exit(proc_t p, int rv)
1606{
1607	vfork_exit_internal(p, rv, 0);
1608}
1609
1610void
1611vfork_exit_internal(proc_t p, int rv, int forceexit)
1612{
1613	thread_t self = current_thread();
1614#ifdef FIXME
1615	struct task *task = p->task;
1616#endif
1617	struct uthread *ut;
1618
1619	/*
1620	 * If a thread in this task has already
1621	 * called exit(), then halt any others
1622	 * right here.
1623	 */
1624
1625	 ut = get_bsdthread_info(self);
1626
1627
1628	proc_lock(p);
1629	 if ((p->p_lflag & P_LPEXIT) == P_LPEXIT) {
1630		/*
1631	 	* This happens when a parent exits/killed and vfork is in progress
1632		* other threads. But shutdown code for ex has already called exit1()
1633	 	*/
1634		proc_unlock(p);
1635		return;
1636	}
1637	p->p_lflag |= (P_LEXIT | P_LPEXIT);
1638	proc_unlock(p);
1639
1640	if (forceexit == 0) {
1641		/*
1642		 * parent of a vfork child has already called exit() and the
1643		 * thread that has vfork in proress terminates. So there is no
1644		 * separate address space here and it has already been marked for
1645		 * termination. This was never covered before and could cause problems
1646		 * if we block here for outside code.
1647		 */
1648		/* Notify the perf server */
1649		(void)sys_perf_notify(self, p->p_pid);
1650	}
1651
1652	/*
1653	 * Remove proc from allproc queue and from pidhash chain.
1654	 * Need to do this before we do anything that can block.
1655	 * Not doing causes things like mount() find this on allproc
1656	 * in partially cleaned state.
1657	 */
1658
1659	proc_list_lock();
1660
1661	LIST_REMOVE(p, p_list);
1662	LIST_INSERT_HEAD(&zombproc, p, p_list);	/* Place onto zombproc. */
1663	/* will not be visible via proc_find */
1664	p->p_listflag |= P_LIST_EXITED;
1665
1666	proc_list_unlock();
1667
1668	proc_lock(p);
1669	p->p_xstat = rv;
1670	p->p_lflag &= ~(P_LTRACED | P_LPPWAIT);
1671	p->p_sigignore = ~0;
1672	proc_unlock(p);
1673
1674	proc_spinlock(p);
1675	if (thread_call_cancel(p->p_rcall))
1676		p->p_ractive--;
1677
1678	while (p->p_ractive > 0) {
1679		proc_spinunlock(p);
1680
1681		delay(1);
1682
1683		proc_spinlock(p);
1684	}
1685	proc_spinunlock(p);
1686
1687	thread_call_free(p->p_rcall);
1688	p->p_rcall = NULL;
1689
1690	ut->uu_siglist = 0;
1691
1692	vproc_exit(p);
1693}
1694
1695void
1696vproc_exit(proc_t p)
1697{
1698	proc_t q;
1699	proc_t pp;
1700
1701	vnode_t tvp;
1702#ifdef FIXME
1703	struct task *task = p->task;
1704#endif
1705	struct pgrp * pg;
1706	struct session *sessp;
1707
1708	/* XXX Zombie allocation may fail, in which case stats get lost */
1709	MALLOC_ZONE(p->p_ru, struct rusage *,
1710			sizeof (*p->p_ru), M_ZOMBIE, M_WAITOK);
1711
1712
1713	proc_refdrain(p);
1714
1715	/*
1716	 * Close open files and release open-file table.
1717	 * This may block!
1718	 */
1719	fdfree(p);
1720
1721#if !CONFIG_EMBEDDED
1722	if (p->p_legacy_behavior & PROC_LEGACY_BEHAVIOR_IOTHROTTLE) {
1723		throttle_legacy_process_decr();
1724	}
1725#endif
1726
1727	sessp = proc_session(p);
1728	if (SESS_LEADER(p, sessp)) {
1729
1730		if (sessp->s_ttyvp != NULLVP) {
1731			struct vnode *ttyvp;
1732			int ttyvid;
1733			struct vfs_context context;
1734			struct tty *tp;
1735
1736			/*
1737			 * Controlling process.
1738			 * Signal foreground pgrp,
1739			 * drain controlling terminal
1740			 * and revoke access to controlling terminal.
1741			 */
1742			session_lock(sessp);
1743			tp = SESSION_TP(sessp);
1744			if ((tp != TTY_NULL) && (tp->t_session == sessp)) {
1745				session_unlock(sessp);
1746
1747				tty_pgsignal(tp, SIGHUP, 1);
1748
1749				session_lock(sessp);
1750				tp = SESSION_TP(sessp);
1751			}
1752			ttyvp = sessp->s_ttyvp;
1753			ttyvid = sessp->s_ttyvid;
1754			sessp->s_ttyvp = NULL;
1755			sessp->s_ttyvid = 0;
1756			sessp->s_ttyp = TTY_NULL;
1757			sessp->s_ttypgrpid = NO_PID;
1758			session_unlock(sessp);
1759
1760		       if ((ttyvp != NULLVP) && (vnode_getwithvid(ttyvp, ttyvid) == 0)) {
1761			        if (tp != TTY_NULL) {
1762					tty_lock(tp);
1763					(void) ttywait(tp);
1764					tty_unlock(tp);
1765				}
1766				context.vc_thread = proc_thread(p); /* XXX */
1767				context.vc_ucred = kauth_cred_proc_ref(p);
1768				vnode_rele(ttyvp);
1769				VNOP_REVOKE(ttyvp, REVOKEALL, &context);
1770				vnode_put(ttyvp);
1771				kauth_cred_unref(&context.vc_ucred);
1772				ttyvp = NULLVP;
1773			}
1774			if (ttyvp)
1775				vnode_rele(ttyvp);
1776			if (tp)
1777				ttyfree(tp);
1778		}
1779		session_lock(sessp);
1780		sessp->s_leader = NULL;
1781		session_unlock(sessp);
1782	}
1783	session_rele(sessp);
1784
1785	pg = proc_pgrp(p);
1786	fixjobc(p, pg, 0);
1787	pg_rele(pg);
1788
1789	p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
1790
1791	proc_list_lock();
1792	proc_childdrainstart(p);
1793	while ((q = p->p_children.lh_first) != NULL) {
1794		q->p_listflag |= P_LIST_DEADPARENT;
1795		if (q->p_stat == SZOMB) {
1796			if (p != q->p_pptr)
1797				panic("parent child linkage broken");
1798			/* check for lookups by zomb sysctl */
1799			while ((q->p_listflag & P_LIST_WAITING) == P_LIST_WAITING) {
1800				msleep(&q->p_stat, proc_list_mlock, PWAIT, "waitcoll", 0);
1801			}
1802			q->p_listflag |= P_LIST_WAITING;
1803			/*
1804			 * This is a named reference and it is not granted
1805			 * if the reap is already in progress. So we get
1806			 * the reference here exclusively and their can be
1807			 * no waiters. So there is no need for a wakeup
1808			 * after we are done. AlsO  the reap frees the structure
1809			 * and the proc struct cannot be used for wakeups as well.
1810			 * It is safe to use q here as this is system reap
1811			 */
1812			(void)reap_child_locked(p, q, 1, 0, 1, 0);
1813		} else {
1814			proc_reparentlocked(q, initproc, 0, 1);
1815			/*
1816		 	* Traced processes are killed
1817		 	* since their existence means someone is messing up.
1818		 	*/
1819			if (q->p_lflag & P_LTRACED) {
1820				proc_list_unlock();
1821				proc_lock(q);
1822				q->p_lflag &= ~P_LTRACED;
1823				if (q->sigwait_thread) {
1824					thread_t thread = q->sigwait_thread;
1825
1826					proc_unlock(q);
1827					/*
1828				 	* The sigwait_thread could be stopped at a
1829				 	* breakpoint. Wake it up to kill.
1830				 	* Need to do this as it could be a thread which is not
1831				 	* the first thread in the task. So any attempts to kill
1832				 	* the process would result into a deadlock on q->sigwait.
1833				 	*/
1834					thread_resume(thread);
1835					clear_wait(thread, THREAD_INTERRUPTED);
1836					threadsignal(thread, SIGKILL, 0);
1837				} else {
1838					proc_unlock(q);
1839				}
1840
1841				psignal(q, SIGKILL);
1842				proc_list_lock();
1843			}
1844		}
1845	}
1846
1847	proc_childdrainend(p);
1848	proc_list_unlock();
1849
1850	/*
1851	 * Release reference to text vnode
1852	 */
1853	tvp = p->p_textvp;
1854	p->p_textvp = NULL;
1855	if (tvp != NULLVP) {
1856		vnode_rele(tvp);
1857	}
1858
1859	/*
1860	 * Save exit status and final rusage info, adding in child rusage
1861	 * info and self times.  If we were unable to allocate a zombie
1862	 * structure, this information is lost.
1863	 */
1864	/* No need for locking here as no one than this thread can access this */
1865	if (p->p_ru != NULL) {
1866	    *p->p_ru = p->p_stats->p_ru;
1867	    timerclear(&p->p_ru->ru_utime);
1868	    timerclear(&p->p_ru->ru_stime);
1869
1870#ifdef  FIXME
1871	    if (task) {
1872		mach_task_basic_info_data_t tinfo;
1873		task_thread_times_info_data_t ttimesinfo;
1874		int task_info_stuff, task_ttimes_stuff;
1875		struct timeval ut,st;
1876
1877		task_info_stuff	= MACH_TASK_BASIC_INFO_COUNT;
1878		task_info(task, MACH_TASK_BASIC_INFO,
1879			  &tinfo, &task_info_stuff);
1880		p->p_ru->ru_utime.tv_sec = tinfo.user_time.seconds;
1881		p->p_ru->ru_utime.tv_usec = tinfo.user_time.microseconds;
1882		p->p_ru->ru_stime.tv_sec = tinfo.system_time.seconds;
1883		p->p_ru->ru_stime.tv_usec = tinfo.system_time.microseconds;
1884
1885		task_ttimes_stuff = TASK_THREAD_TIMES_INFO_COUNT;
1886		task_info(task, TASK_THREAD_TIMES_INFO,
1887			  &ttimesinfo, &task_ttimes_stuff);
1888
1889		ut.tv_sec = ttimesinfo.user_time.seconds;
1890		ut.tv_usec = ttimesinfo.user_time.microseconds;
1891		st.tv_sec = ttimesinfo.system_time.seconds;
1892		st.tv_usec = ttimesinfo.system_time.microseconds;
1893		timeradd(&ut,&p->p_ru->ru_utime,&p->p_ru->ru_utime);
1894			timeradd(&st,&p->p_ru->ru_stime,&p->p_ru->ru_stime);
1895	    }
1896#endif /* FIXME */
1897
1898	    ruadd(p->p_ru, &p->p_stats->p_cru);
1899	}
1900
1901	/*
1902	 * Free up profiling buffers.
1903	 */
1904	{
1905		struct uprof *p0 = &p->p_stats->p_prof, *p1, *pn;
1906
1907		p1 = p0->pr_next;
1908		p0->pr_next = NULL;
1909		p0->pr_scale = 0;
1910
1911		for (; p1 != NULL; p1 = pn) {
1912			pn = p1->pr_next;
1913			kfree(p1, sizeof *p1);
1914		}
1915	}
1916
1917#if PSYNCH
1918	pth_proc_hashdelete(p);
1919#endif /* PSYNCH */
1920
1921	/*
1922	 * Other substructures are freed from wait().
1923	 */
1924	FREE_ZONE(p->p_stats, sizeof *p->p_stats, M_PSTATS);
1925	p->p_stats = NULL;
1926
1927	FREE_ZONE(p->p_sigacts, sizeof *p->p_sigacts, M_SIGACTS);
1928	p->p_sigacts = NULL;
1929
1930	proc_limitdrop(p, 1);
1931	p->p_limit = NULL;
1932
1933	/*
1934	 * Finish up by terminating the task
1935	 * and halt this thread (only if a
1936	 * member of the task exiting).
1937	 */
1938	p->task = TASK_NULL;
1939
1940	/*
1941	 * Notify parent that we're gone.
1942	 */
1943	pp = proc_parent(p);
1944	if ((p->p_listflag & P_LIST_DEADPARENT) == 0) {
1945		if (pp != initproc) {
1946			proc_lock(pp);
1947			pp->si_pid = p->p_pid;
1948			pp->si_status = p->p_xstat;
1949			pp->si_code = CLD_EXITED;
1950			/*
1951			 * p_ucred usage is safe as it is an exiting process
1952			 * and reference is dropped in reap
1953			 */
1954			pp->si_uid = kauth_cred_getruid(p->p_ucred);
1955			proc_unlock(pp);
1956		}
1957		/* mark as a zombie */
1958		/* mark as a zombie */
1959		/* No need to take proc lock as all refs are drained and
1960		 * no one except parent (reaping ) can look at this.
1961		 * The write is to an int and is coherent. Also parent is
1962		 *  keyed off of list lock for reaping
1963		 */
1964		p->p_stat = SZOMB;
1965
1966		psignal(pp, SIGCHLD);
1967
1968		/* and now wakeup the parent */
1969		proc_list_lock();
1970		wakeup((caddr_t)pp);
1971		proc_list_unlock();
1972	} else {
1973		proc_list_lock();
1974		/* check for lookups by zomb sysctl */
1975		while ((p->p_listflag & P_LIST_WAITING) == P_LIST_WAITING) {
1976			msleep(&p->p_stat, proc_list_mlock, PWAIT, "waitcoll", 0);
1977		}
1978		p->p_stat = SZOMB;
1979		p->p_listflag |= P_LIST_WAITING;
1980
1981		/*
1982		 * This is a named reference and it is not granted
1983		 * if the reap is already in progress. So we get
1984		 * the reference here exclusively and their can be
1985		 * no waiters. So there is no need for a wakeup
1986		 * after we are done. AlsO  the reap frees the structure
1987		 * and the proc struct cannot be used for wakeups as well.
1988		 * It is safe to use p here as this is system reap
1989		 */
1990		(void)reap_child_locked(pp, p, 0, 0, 1, 1);
1991		/* list lock dropped by reap_child_locked */
1992	}
1993	proc_rele(pp);
1994}
1995
1996
1997/*
1998 * munge_rusage
1999 *	LP64 support - long is 64 bits if we are dealing with a 64 bit user
2000 *	process.  We munge the kernel version of rusage into the
2001 *	64 bit version.
2002 */
2003__private_extern__  void
2004munge_user64_rusage(struct rusage *a_rusage_p, struct user64_rusage *a_user_rusage_p)
2005{
2006	/* timeval changes size, so utime and stime need special handling */
2007	a_user_rusage_p->ru_utime.tv_sec = a_rusage_p->ru_utime.tv_sec;
2008	a_user_rusage_p->ru_utime.tv_usec = a_rusage_p->ru_utime.tv_usec;
2009	a_user_rusage_p->ru_stime.tv_sec = a_rusage_p->ru_stime.tv_sec;
2010	a_user_rusage_p->ru_stime.tv_usec = a_rusage_p->ru_stime.tv_usec;
2011	/*
2012	 * everything else can be a direct assign, since there is no loss
2013	 * of precision implied boing 32->64.
2014	 */
2015	a_user_rusage_p->ru_maxrss = a_rusage_p->ru_maxrss;
2016	a_user_rusage_p->ru_ixrss = a_rusage_p->ru_ixrss;
2017	a_user_rusage_p->ru_idrss = a_rusage_p->ru_idrss;
2018	a_user_rusage_p->ru_isrss = a_rusage_p->ru_isrss;
2019	a_user_rusage_p->ru_minflt = a_rusage_p->ru_minflt;
2020	a_user_rusage_p->ru_majflt = a_rusage_p->ru_majflt;
2021	a_user_rusage_p->ru_nswap = a_rusage_p->ru_nswap;
2022	a_user_rusage_p->ru_inblock = a_rusage_p->ru_inblock;
2023	a_user_rusage_p->ru_oublock = a_rusage_p->ru_oublock;
2024	a_user_rusage_p->ru_msgsnd = a_rusage_p->ru_msgsnd;
2025	a_user_rusage_p->ru_msgrcv = a_rusage_p->ru_msgrcv;
2026	a_user_rusage_p->ru_nsignals = a_rusage_p->ru_nsignals;
2027	a_user_rusage_p->ru_nvcsw = a_rusage_p->ru_nvcsw;
2028	a_user_rusage_p->ru_nivcsw = a_rusage_p->ru_nivcsw;
2029}
2030
2031/* For a 64-bit kernel and 32-bit userspace, munging may be needed */
2032__private_extern__  void
2033munge_user32_rusage(struct rusage *a_rusage_p, struct user32_rusage *a_user_rusage_p)
2034{
2035	/* timeval changes size, so utime and stime need special handling */
2036	a_user_rusage_p->ru_utime.tv_sec = a_rusage_p->ru_utime.tv_sec;
2037	a_user_rusage_p->ru_utime.tv_usec = a_rusage_p->ru_utime.tv_usec;
2038	a_user_rusage_p->ru_stime.tv_sec = a_rusage_p->ru_stime.tv_sec;
2039	a_user_rusage_p->ru_stime.tv_usec = a_rusage_p->ru_stime.tv_usec;
2040	/*
2041	 * everything else can be a direct assign. We currently ignore
2042	 * the loss of precision
2043	 */
2044	a_user_rusage_p->ru_maxrss = a_rusage_p->ru_maxrss;
2045	a_user_rusage_p->ru_ixrss = a_rusage_p->ru_ixrss;
2046	a_user_rusage_p->ru_idrss = a_rusage_p->ru_idrss;
2047	a_user_rusage_p->ru_isrss = a_rusage_p->ru_isrss;
2048	a_user_rusage_p->ru_minflt = a_rusage_p->ru_minflt;
2049	a_user_rusage_p->ru_majflt = a_rusage_p->ru_majflt;
2050	a_user_rusage_p->ru_nswap = a_rusage_p->ru_nswap;
2051	a_user_rusage_p->ru_inblock = a_rusage_p->ru_inblock;
2052	a_user_rusage_p->ru_oublock = a_rusage_p->ru_oublock;
2053	a_user_rusage_p->ru_msgsnd = a_rusage_p->ru_msgsnd;
2054	a_user_rusage_p->ru_msgrcv = a_rusage_p->ru_msgrcv;
2055	a_user_rusage_p->ru_nsignals = a_rusage_p->ru_nsignals;
2056	a_user_rusage_p->ru_nvcsw = a_rusage_p->ru_nvcsw;
2057	a_user_rusage_p->ru_nivcsw = a_rusage_p->ru_nivcsw;
2058}
2059