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 <vm/vm_protos.h>
149
150#include <sys/sdt.h>
151
152extern char init_task_failure_data[];
153void proc_prepareexit(proc_t p, int rv, boolean_t perf_notify);
154void vfork_exit(proc_t p, int rv);
155void vproc_exit(proc_t p);
156__private_extern__ void munge_user64_rusage(struct rusage *a_rusage_p, struct user64_rusage *a_user_rusage_p);
157__private_extern__ void munge_user32_rusage(struct rusage *a_rusage_p, struct user32_rusage *a_user_rusage_p);
158static int reap_child_locked(proc_t parent, proc_t child, int deadparent, int reparentedtoinit, int locked, int droplock);
159
160/*
161 * Things which should have prototypes in headers, but don't
162 */
163void	proc_exit(proc_t p);
164int	wait1continue(int result);
165int	waitidcontinue(int result);
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);
170void gather_rusage_info(proc_t p, rusage_info_current *ru, int flavor);
171
172/*
173 * NOTE: Source and target may *NOT* overlap!
174 * XXX Should share code with bsd/dev/ppc/unix_signal.c
175 */
176void
177siginfo_user_to_user32(user_siginfo_t *in, user32_siginfo_t *out)
178{
179	out->si_signo	= in->si_signo;
180	out->si_errno	= in->si_errno;
181	out->si_code	= in->si_code;
182	out->si_pid	= in->si_pid;
183	out->si_uid	= in->si_uid;
184	out->si_status	= in->si_status;
185	out->si_addr	= CAST_DOWN_EXPLICIT(user32_addr_t,in->si_addr);
186	/* following cast works for sival_int because of padding */
187	out->si_value.sival_ptr	= CAST_DOWN_EXPLICIT(user32_addr_t,in->si_value.sival_ptr);
188	out->si_band	= in->si_band;			/* range reduction */
189}
190
191void
192siginfo_user_to_user64(user_siginfo_t *in, user64_siginfo_t *out)
193{
194	out->si_signo	= in->si_signo;
195	out->si_errno	= in->si_errno;
196	out->si_code	= in->si_code;
197	out->si_pid	= in->si_pid;
198	out->si_uid	= in->si_uid;
199	out->si_status	= in->si_status;
200	out->si_addr	= in->si_addr;
201	/* following cast works for sival_int because of padding */
202	out->si_value.sival_ptr	= in->si_value.sival_ptr;
203	out->si_band	= in->si_band;			/* range reduction */
204}
205
206static int
207copyoutsiginfo(user_siginfo_t *native, boolean_t is64, user_addr_t uaddr)
208{
209	if (is64) {
210		user64_siginfo_t sinfo64;
211
212		bzero(&sinfo64, sizeof (sinfo64));
213		siginfo_user_to_user64(native, &sinfo64);
214		return (copyout(&sinfo64, uaddr, sizeof (sinfo64)));
215	} else {
216		user32_siginfo_t sinfo32;
217
218		bzero(&sinfo32, sizeof (sinfo32));
219		siginfo_user_to_user32(native, &sinfo32);
220		return (copyout(&sinfo32, uaddr, sizeof (sinfo32)));
221	}
222}
223
224/*
225 * exit --
226 *	Death of process.
227 */
228void
229exit(proc_t p, struct exit_args *uap, int *retval)
230{
231	exit1(p, W_EXITCODE(uap->rval, 0), retval);
232
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, 0);
249}
250
251int
252exit1_internal(proc_t p, int rv, int *retval, boolean_t thread_can_terminate, boolean_t perf_notify,
253	       int jetsam_flags)
254{
255	thread_t self = current_thread();
256	struct task *task = p->task;
257	struct uthread *ut;
258	int error = 0;
259
260	/*
261	 * If a thread in this task has already
262	 * called exit(), then halt any others
263	 * right here.
264	 */
265
266	 ut = get_bsdthread_info(self);
267	 if (ut->uu_flag & UT_VFORK) {
268		if (!thread_can_terminate) {
269			return EINVAL;
270		}
271
272		vfork_exit(p, rv);
273		vfork_return(p , retval, p->p_pid);
274		unix_syscall_return(0);
275		/* NOT REACHED */
276	 }
277
278	/*
279	 * The parameter list of audit_syscall_exit() was augmented to
280	 * take the Darwin syscall number as the first parameter,
281	 * which is currently required by mac_audit_postselect().
282	 */
283
284	/*
285	 * The BSM token contains two components: an exit status as passed
286	 * to exit(), and a return value to indicate what sort of exit it
287	 * was.  The exit status is WEXITSTATUS(rv), but it's not clear
288 	 * what the return value is.
289	 */
290	AUDIT_ARG(exit, WEXITSTATUS(rv), 0);
291	AUDIT_SYSCALL_EXIT(SYS_exit, p, ut, 0); /* Exit is always successfull */
292
293	DTRACE_PROC1(exit, int, CLD_EXITED);
294
295	/* mark process is going to exit and pull out of DBG/disk throttle */
296	/* TODO: This should be done after becoming exit thread */
297	proc_set_task_policy(p->task, THREAD_NULL, TASK_POLICY_ATTRIBUTE,
298	                     TASK_POLICY_TERMINATED, TASK_POLICY_ENABLE);
299
300        proc_lock(p);
301	error = proc_transstart(p, 1, ((jetsam_flags & P_JETSAM_VNODE) ? 1 : 0));
302	if (error == EDEADLK) {
303		/* Temp: If deadlock error, then it implies multithreaded exec is
304		 * in progress. Instread of letting exit continue and
305		 * corrupting the freed memory, let the exit thread
306		 * return. This will save corruption in remote case.
307		 */
308		proc_unlock(p);
309		if (current_proc() == p){
310			if (p->exit_thread == self)
311				printf("exit_thread failed to exit, leaving process %s[%d] in unkillable limbo\n",
312				       p->p_comm, p->p_pid);
313			thread_exception_return();
314		} else {
315			/* external termination like jetsam */
316			return(error);
317		}
318	}
319
320	while (p->exit_thread != self) {
321		if (sig_try_locked(p) <= 0) {
322			proc_transend(p, 1);
323			if (get_threadtask(self) != task) {
324				proc_unlock(p);
325				return(0);
326                        }
327			proc_unlock(p);
328
329			thread_terminate(self);
330			if (!thread_can_terminate) {
331				return 0;
332			}
333
334			thread_exception_return();
335			/* NOTREACHED */
336		}
337		sig_lock_to_exit(p);
338	}
339	if (p == initproc && current_proc() == p) {
340		proc_unlock(p);
341		printf("pid 1 exited (signal %d, exit %d)",
342		    WTERMSIG(rv), WEXITSTATUS(rv));
343#if (DEVELOPMENT || DEBUG)
344		int err;
345		/*
346		 * For debugging purposes, generate a core file of initproc before
347		 * panicking. Leave at least 300 MB free on the root volume, and ignore
348		 * the process's corefile ulimit.
349		 */
350		if ((err = coredump(p, 300, 1)) != 0) {
351			printf("Failed to generate initproc core file: error %d", err);
352		} else {
353			printf("Generated initproc core file");
354			sync(p, (void *)NULL, (int *)NULL);
355		}
356#endif
357		panic("%s died\nState at Last Exception:\n\n%s",
358							(p->p_comm[0] != '\0' ?
359								p->p_comm :
360								"launchd"),
361							init_task_failure_data);
362	}
363
364	p->p_lflag |= P_LEXIT;
365	p->p_xstat = rv;
366	p->p_lflag |= jetsam_flags;
367
368	proc_transend(p, 1);
369	proc_unlock(p);
370
371	proc_prepareexit(p, rv, perf_notify);
372
373	/* Last thread to terminate will call proc_exit() */
374	task_terminate_internal(task);
375
376	return(0);
377}
378
379void
380proc_prepareexit(proc_t p, int rv, boolean_t perf_notify)
381{
382	mach_exception_data_type_t code, subcode;
383	struct uthread *ut;
384	thread_t self = current_thread();
385	ut = get_bsdthread_info(self);
386	struct rusage_superset *rup;
387
388 	/* If a core should be generated, notify crash reporter */
389	if (hassigprop(WTERMSIG(rv), SA_CORE) || ((p->p_csflags & CS_KILLED) != 0)) {
390		/*
391		 * Workaround for processes checking up on PT_DENY_ATTACH:
392		 * should be backed out post-Leopard (details in 5431025).
393		 */
394		if ((SIGSEGV == WTERMSIG(rv)) &&
395				(p->p_pptr->p_lflag & P_LNOATTACH)) {
396			goto skipcheck;
397		}
398
399		/*
400		 * Crash Reporter looks for the signal value, original exception
401		 * type, and low 20 bits of the original code in code[0]
402		 * (8, 4, and 20 bits respectively). code[1] is unmodified.
403		 */
404		code = ((WTERMSIG(rv) & 0xff) << 24) |
405			((ut->uu_exception & 0x0f) << 20) |
406			((int)ut->uu_code & 0xfffff);
407		subcode = ut->uu_subcode;
408		(void) task_exception_notify(EXC_CRASH, code, subcode);
409	}
410
411skipcheck:
412	/* Notify the perf server? */
413	if (perf_notify) {
414		(void)sys_perf_notify(self, p->p_pid);
415	}
416
417	/*
418	 * Before this process becomes a zombie, stash resource usage
419	 * stats in the proc for external observers to query
420	 * via proc_pid_rusage().
421	 *
422	 * If the zombie allocation fails, just punt the stats.
423	 */
424	MALLOC_ZONE(rup, struct rusage_superset *,
425			sizeof (*rup), M_ZOMBIE, M_WAITOK);
426	if (rup != NULL) {
427		gather_rusage_info(p, &rup->ri, RUSAGE_INFO_CURRENT);
428		rup->ri.ri_phys_footprint = 0;
429		rup->ri.ri_proc_exit_abstime = mach_absolute_time();
430
431		/*
432		 * Make the rusage_info visible to external observers
433		 * only after it has been completely filled in.
434		 */
435		p->p_ru = rup;
436	}
437
438	/*
439	 * Remove proc from allproc queue and from pidhash chain.
440	 * Need to do this before we do anything that can block.
441	 * Not doing causes things like mount() find this on allproc
442	 * in partially cleaned state.
443	 */
444
445	proc_list_lock();
446
447#if CONFIG_MEMORYSTATUS
448	memorystatus_remove(p, TRUE);
449#endif
450
451	LIST_REMOVE(p, p_list);
452	LIST_INSERT_HEAD(&zombproc, p, p_list);	/* Place onto zombproc. */
453	/* will not be visible via proc_find */
454	p->p_listflag |= P_LIST_EXITED;
455
456	proc_list_unlock();
457
458
459#ifdef PGINPROF
460	vmsizmon();
461#endif
462	/*
463	 * If parent is waiting for us to exit or exec,
464	 * P_LPPWAIT is set; we will wakeup the parent below.
465	 */
466	proc_lock(p);
467	p->p_lflag &= ~(P_LTRACED | P_LPPWAIT);
468	p->p_sigignore = ~(sigcantmask);
469	ut->uu_siglist = 0;
470	proc_unlock(p);
471}
472
473void
474proc_exit(proc_t p)
475{
476	proc_t q;
477	proc_t pp;
478	struct task *task = p->task;
479	vnode_t tvp = NULLVP;
480	struct pgrp * pg;
481	struct session *sessp;
482	struct uthread * uth;
483	pid_t pid;
484	int exitval;
485	int knote_hint;
486
487	uth = current_uthread();
488
489	proc_lock(p);
490	proc_transstart(p, 1, 0);
491	if( !(p->p_lflag & P_LEXIT)) {
492		/*
493		 * This can happen if a thread_terminate() occurs
494		 * in a single-threaded process.
495		 */
496		p->p_lflag |= P_LEXIT;
497		proc_transend(p, 1);
498		proc_unlock(p);
499		proc_prepareexit(p, 0, TRUE);
500		(void) task_terminate_internal(task);
501		proc_lock(p);
502	} else {
503		proc_transend(p, 1);
504	}
505
506	p->p_lflag |= P_LPEXIT;
507
508	/*
509	 * Other kernel threads may be in the middle of signalling this process.
510	 * Wait for those threads to wrap it up before making the process
511	 * disappear on them.
512	 */
513	if ((p->p_lflag & P_LINSIGNAL) || (p->p_sigwaitcnt > 0)) {
514		p->p_sigwaitcnt++;
515		while ((p->p_lflag & P_LINSIGNAL) || (p->p_sigwaitcnt > 1))
516			msleep(&p->p_sigmask, &p->p_mlock, PWAIT, "proc_sigdrain", NULL);
517		p->p_sigwaitcnt--;
518	}
519
520	proc_unlock(p);
521	pid = p->p_pid;
522	exitval = p->p_xstat;
523	KERNEL_DEBUG_CONSTANT_IST(KDEBUG_COMMON,
524		BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXIT) | DBG_FUNC_START,
525		pid, exitval, 0, 0, 0);
526
527#if CONFIG_DTRACE
528	/*
529	 * Free any outstanding lazy dof entries. It is imperative we
530	 * always call dtrace_lazy_dofs_destroy, rather than null check
531	 * and call if !NULL. If we NULL test, during lazy dof faulting
532	 * we can race with the faulting code and proceed from here to
533	 * beyond the helpers cleanup. The lazy dof faulting will then
534	 * install new helpers which will never be cleaned up, and leak.
535	 */
536	dtrace_lazy_dofs_destroy(p);
537
538	/*
539	 * Clean up any DTrace helper actions or probes for the process.
540	 */
541	if (p->p_dtrace_helpers != NULL) {
542		(*dtrace_helpers_cleanup)(p);
543	}
544
545	/*
546	 * Clean up any DTrace probes associated with this process.
547	 */
548	/*
549	 * APPLE NOTE: We release ptss pages/entries in dtrace_fasttrap_exit_ptr(),
550	 * call this after dtrace_helpers_cleanup()
551	 */
552	proc_lock(p);
553	if (p->p_dtrace_probes && dtrace_fasttrap_exit_ptr) {
554		(*dtrace_fasttrap_exit_ptr)(p);
555	}
556	proc_unlock(p);
557#endif
558
559	nspace_proc_exit(p);
560
561#if VM_PRESSURE_EVENTS
562	vm_pressure_proc_cleanup(p);
563#endif
564
565	/*
566	 * need to cancel async IO requests that can be cancelled and wait for those
567	 * already active.  MAY BLOCK!
568	 */
569
570	proc_refdrain(p);
571
572	/* if any pending cpu limits action, clear it */
573	task_clear_cpuusage(p->task, TRUE);
574
575	workqueue_mark_exiting(p);
576	workqueue_exit(p);
577
578	_aio_exit( p );
579
580	/*
581	 * Close open files and release open-file table.
582	 * This may block!
583	 */
584	fdfree(p);
585
586	if (uth->uu_lowpri_window) {
587	        /*
588		 * task is marked as a low priority I/O type
589		 * and the I/O we issued while in flushing files on close
590		 * collided with normal I/O operations...
591		 * no need to throttle this thread since its going away
592		 * but we do need to update our bookeeping w/r to throttled threads
593		 */
594		throttle_lowpri_io(0);
595	}
596
597#if SYSV_SHM
598	/* Close ref SYSV Shared memory*/
599	if (p->vm_shm)
600		shmexit(p);
601#endif
602#if SYSV_SEM
603	/* Release SYSV semaphores */
604	semexit(p);
605#endif
606
607#if PSYNCH
608	pth_proc_hashdelete(p);
609#endif /* PSYNCH */
610
611	sessp = proc_session(p);
612	if (SESS_LEADER(p, sessp)) {
613
614		if (sessp->s_ttyvp != NULLVP) {
615			struct vnode *ttyvp;
616			int ttyvid;
617			int cttyflag = 0;
618			struct vfs_context context;
619			struct tty *tp;
620
621			/*
622			 * Controlling process.
623			 * Signal foreground pgrp,
624			 * drain controlling terminal
625			 * and revoke access to controlling terminal.
626			 */
627			session_lock(sessp);
628			tp = SESSION_TP(sessp);
629			if ((tp != TTY_NULL) && (tp->t_session == sessp)) {
630				session_unlock(sessp);
631
632				/*
633				 * We're going to SIGHUP the foreground process
634				 * group. It can't change from this point on
635				 * until the revoke is complete.
636				 * The process group changes under both the tty
637				 * lock and proc_list_lock but we need only one
638				 */
639				tty_lock(tp);
640				ttysetpgrphup(tp);
641				tty_unlock(tp);
642
643				tty_pgsignal(tp, SIGHUP, 1);
644
645				session_lock(sessp);
646				tp = SESSION_TP(sessp);
647			}
648			cttyflag = sessp->s_flags & S_CTTYREF;
649			sessp->s_flags &= ~S_CTTYREF;
650			ttyvp = sessp->s_ttyvp;
651			ttyvid = sessp->s_ttyvid;
652			sessp->s_ttyvp = NULLVP;
653			sessp->s_ttyvid = 0;
654			sessp->s_ttyp = TTY_NULL;
655			sessp->s_ttypgrpid = NO_PID;
656			session_unlock(sessp);
657
658			if ((ttyvp != NULLVP) && (vnode_getwithvid(ttyvp, ttyvid) == 0)) {
659				if (tp != TTY_NULL) {
660					tty_lock(tp);
661					(void) ttywait(tp);
662					tty_unlock(tp);
663				}
664				context.vc_thread = proc_thread(p); /* XXX */
665				context.vc_ucred = kauth_cred_proc_ref(p);
666				VNOP_REVOKE(ttyvp, REVOKEALL, &context);
667				if (cttyflag) {
668					/*
669					 * Release the extra usecount taken in cttyopen.
670					 * usecount should be released after VNOP_REVOKE is called.
671					 * This usecount was taken to ensure that
672					 * the VNOP_REVOKE results in a close to
673					 * the tty since cttyclose is a no-op.
674					 */
675					vnode_rele(ttyvp);
676				}
677				vnode_put(ttyvp);
678				kauth_cred_unref(&context.vc_ucred);
679				ttyvp = NULLVP;
680			}
681			if (tp) {
682				/*
683				 * This is cleared even if not set. This is also done in
684				 * spec_close to ensure that the flag is cleared.
685				 */
686				tty_lock(tp);
687				ttyclrpgrphup(tp);
688				tty_unlock(tp);
689
690				ttyfree(tp);
691			}
692		}
693		session_lock(sessp);
694		sessp->s_leader = NULL;
695		session_unlock(sessp);
696	}
697	session_rele(sessp);
698
699	pg = proc_pgrp(p);
700	fixjobc(p, pg, 0);
701	pg_rele(pg);
702
703	p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
704	(void)acct_process(p);
705
706	proc_list_lock();
707
708	if ((p->p_listflag & P_LIST_EXITCOUNT) == P_LIST_EXITCOUNT) {
709		p->p_listflag &= ~P_LIST_EXITCOUNT;
710		proc_shutdown_exitcount--;
711		if (proc_shutdown_exitcount == 0)
712			wakeup(&proc_shutdown_exitcount);
713	}
714
715	/* wait till parentrefs are dropped and grant no more */
716	proc_childdrainstart(p);
717	while ((q = p->p_children.lh_first) != NULL) {
718		int reparentedtoinit = (q->p_listflag & P_LIST_DEADPARENT) ? 1 : 0;
719		if (q->p_stat == SZOMB) {
720			if (p != q->p_pptr)
721				panic("parent child linkage broken");
722			/* check for sysctl zomb lookup */
723			while ((q->p_listflag & P_LIST_WAITING) == P_LIST_WAITING) {
724				msleep(&q->p_stat, proc_list_mlock, PWAIT, "waitcoll", 0);
725			}
726			q->p_listflag |= P_LIST_WAITING;
727			/*
728			 * This is a named reference and it is not granted
729			 * if the reap is already in progress. So we get
730			 * the reference here exclusively and their can be
731			 * no waiters. So there is no need for a wakeup
732			 * after we are done.  Also the reap frees the structure
733			 * and the proc struct cannot be used for wakeups as well.
734			 * It is safe to use q here as this is system reap
735			 */
736			(void)reap_child_locked(p, q, 1, reparentedtoinit, 1, 0);
737		} else {
738			/*
739		 	* Traced processes are killed
740		 	* since their existence means someone is messing up.
741		 	*/
742			if (q->p_lflag & P_LTRACED) {
743				struct proc *opp;
744
745				/*
746				 * Take a reference on the child process to
747				 * ensure it doesn't exit and disappear between
748				 * the time we drop the list_lock and attempt
749				 * to acquire its proc_lock.
750				 */
751				if (proc_ref_locked(q) != q)
752					continue;
753
754				proc_list_unlock();
755
756				opp = proc_find(q->p_oppid);
757				if (opp != PROC_NULL) {
758					proc_list_lock();
759					q->p_oppid = 0;
760					proc_list_unlock();
761					proc_reparentlocked(q, opp, 0, 0);
762					proc_rele(opp);
763				} else {
764					/* original parent exited while traced */
765					proc_list_lock();
766					q->p_listflag |= P_LIST_DEADPARENT;
767					q->p_oppid = 0;
768					proc_list_unlock();
769					proc_reparentlocked(q, initproc, 0, 0);
770				}
771
772				proc_lock(q);
773				q->p_lflag &= ~P_LTRACED;
774
775				if (q->sigwait_thread) {
776					thread_t thread = q->sigwait_thread;
777
778					proc_unlock(q);
779					/*
780				 	* The sigwait_thread could be stopped at a
781				 	* breakpoint. Wake it up to kill.
782				 	* Need to do this as it could be a thread which is not
783				 	* the first thread in the task. So any attempts to kill
784				 	* the process would result into a deadlock on q->sigwait.
785				 	*/
786					thread_resume(thread);
787					clear_wait(thread, THREAD_INTERRUPTED);
788					threadsignal(thread, SIGKILL, 0);
789				} else {
790					proc_unlock(q);
791				}
792
793				psignal(q, SIGKILL);
794				proc_list_lock();
795				proc_rele_locked(q);
796			} else {
797				q->p_listflag |= P_LIST_DEADPARENT;
798				proc_reparentlocked(q, initproc, 0, 1);
799			}
800		}
801	}
802
803	proc_childdrainend(p);
804	proc_list_unlock();
805
806	/*
807	 * Release reference to text vnode
808	 */
809	tvp = p->p_textvp;
810	p->p_textvp = NULL;
811	if (tvp != NULLVP) {
812		vnode_rele(tvp);
813	}
814
815	/*
816	 * Save exit status and final rusage info, adding in child rusage
817	 * info and self times.  If we were unable to allocate a zombie
818	 * structure, this information is lost.
819	 */
820	if (p->p_ru != NULL) {
821	    calcru(p, &p->p_stats->p_ru.ru_utime, &p->p_stats->p_ru.ru_stime, NULL);
822	    p->p_ru->ru = p->p_stats->p_ru;
823
824	    ruadd(&(p->p_ru->ru), &p->p_stats->p_cru);
825	}
826
827	/*
828	 * Free up profiling buffers.
829	 */
830	{
831		struct uprof *p0 = &p->p_stats->p_prof, *p1, *pn;
832
833		p1 = p0->pr_next;
834		p0->pr_next = NULL;
835		p0->pr_scale = 0;
836
837		for (; p1 != NULL; p1 = pn) {
838			pn = p1->pr_next;
839			kfree(p1, sizeof *p1);
840		}
841	}
842
843	proc_spinlock(p);
844	if (thread_call_cancel(p->p_rcall))
845		p->p_ractive--;
846
847	while (p->p_ractive > 0) {
848		proc_spinunlock(p);
849
850		delay(1);
851
852		proc_spinlock(p);
853	}
854	proc_spinunlock(p);
855
856	thread_call_free(p->p_rcall);
857	p->p_rcall = NULL;
858
859	/*
860	 * Other substructures are freed from wait().
861	 */
862	FREE_ZONE(p->p_stats, sizeof *p->p_stats, M_PSTATS);
863	p->p_stats = NULL;
864
865	FREE_ZONE(p->p_sigacts, sizeof *p->p_sigacts, M_SIGACTS);
866	p->p_sigacts = NULL;
867
868	proc_limitdrop(p, 1);
869	p->p_limit = NULL;
870
871	vm_purgeable_disown(p->task);
872
873	/*
874	 * Finish up by terminating the task
875	 * and halt this thread (only if a
876	 * member of the task exiting).
877	 */
878	p->task = TASK_NULL;
879	set_bsdtask_info(task, NULL);
880
881	knote_hint = NOTE_EXIT | (p->p_xstat & 0xffff);
882	proc_knote(p, knote_hint);
883
884	/* mark the thread as the one that is doing proc_exit
885	 * no need to hold proc lock in uthread_free
886	 */
887	uth->uu_flag |= UT_PROCEXIT;
888	/*
889	 * Notify parent that we're gone.
890	 */
891	pp = proc_parent(p);
892	if (pp->p_flag & P_NOCLDWAIT) {
893
894		if (p->p_ru != NULL) {
895			proc_lock(pp);
896#if 3839178
897		/*
898		 * If the parent is ignoring SIGCHLD, then POSIX requires
899		 * us to not add the resource usage to the parent process -
900		 * we are only going to hand it off to init to get reaped.
901		 * We should contest the standard in this case on the basis
902		 * of RLIMIT_CPU.
903		 */
904#else	/* !3839178 */
905		/*
906		 * Add child resource usage to parent before giving
907		 * zombie to init.  If we were unable to allocate a
908		 * zombie structure, this information is lost.
909		 */
910			ruadd(&pp->p_stats->p_cru, &p->p_ru->ru);
911#endif	/* !3839178 */
912			update_rusage_info_child(&pp->p_stats->ri_child, &p->p_ru->ri);
913			proc_unlock(pp);
914		}
915
916		/* kernel can reap this one, no need to move it to launchd */
917		proc_list_lock();
918		p->p_listflag |= P_LIST_DEADPARENT;
919		proc_list_unlock();
920	}
921	if ((p->p_listflag & P_LIST_DEADPARENT) == 0 || p->p_oppid) {
922		if (pp != initproc) {
923			proc_lock(pp);
924			pp->si_pid = p->p_pid;
925			pp->si_status = p->p_xstat;
926			pp->si_code = CLD_EXITED;
927			/*
928			 * p_ucred usage is safe as it is an exiting process
929			 * and reference is dropped in reap
930			 */
931			pp->si_uid = kauth_cred_getruid(p->p_ucred);
932			proc_unlock(pp);
933		}
934		/* mark as a zombie */
935		/* No need to take proc lock as all refs are drained and
936		 * no one except parent (reaping ) can look at this.
937		 * The write is to an int and is coherent. Also parent is
938		 *  keyed off of list lock for reaping
939		 */
940		KERNEL_DEBUG_CONSTANT_IST(KDEBUG_COMMON,
941			BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXIT) | DBG_FUNC_END,
942			pid, exitval, 0, 0, 0);
943		p->p_stat = SZOMB;
944		/*
945		 * The current process can be reaped so, no one
946		 * can depend on this
947		 */
948
949		psignal(pp, SIGCHLD);
950
951		/* and now wakeup the parent */
952		proc_list_lock();
953		wakeup((caddr_t)pp);
954		proc_list_unlock();
955	} else {
956		/* should be fine as parent proc would be initproc */
957		/* mark as a zombie */
958		/* No need to take proc lock as all refs are drained and
959		 * no one except parent (reaping ) can look at this.
960		 * The write is to an int and is coherent. Also parent is
961		 *  keyed off of list lock for reaping
962		 */
963		proc_list_lock();
964		KERNEL_DEBUG_CONSTANT_IST(KDEBUG_COMMON,
965			BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXIT) | DBG_FUNC_END,
966			pid, exitval, 0, 0, 0);
967		/* check for sysctl zomb lookup */
968		while ((p->p_listflag & P_LIST_WAITING) == P_LIST_WAITING) {
969			msleep(&p->p_stat, proc_list_mlock, PWAIT, "waitcoll", 0);
970		}
971		/* safe to use p as this is a system reap */
972		p->p_stat = SZOMB;
973		p->p_listflag |= P_LIST_WAITING;
974
975		/*
976		 * This is a named reference and it is not granted
977		 * if the reap is already in progress. So we get
978		 * the reference here exclusively and their can be
979		 * no waiters. So there is no need for a wakeup
980		 * after we are done. AlsO  the reap frees the structure
981		 * and the proc struct cannot be used for wakeups as well.
982		 * It is safe to use p here as this is system reap
983		 */
984		(void)reap_child_locked(pp, p, 1, 0, 1, 1);
985		/* list lock dropped by reap_child_locked */
986	}
987	if (uth->uu_lowpri_window) {
988	        /*
989		 * task is marked as a low priority I/O type and we've
990		 * somehow picked up another throttle during exit processing...
991		 * no need to throttle this thread since its going away
992		 * but we do need to update our bookeeping w/r to throttled threads
993		 */
994		throttle_lowpri_io(0);
995	}
996
997	proc_rele(pp);
998
999}
1000
1001
1002/*
1003 * reap_child_locked
1004 *
1005 * Description:	Given a process from which all status information needed
1006 *		has already been extracted, if the process is a ptrace
1007 *		attach process, detach it and give it back to its real
1008 *		parent, else recover all resources remaining associated
1009 *		with it.
1010 *
1011 * Parameters:	proc_t parent		Parent of process being reaped
1012 *		proc_t child		Process to reap
1013 *
1014 * Returns:	0			Process was not reaped because it
1015 *					came from an attach
1016 *		1			Process was reaped
1017 */
1018static int
1019reap_child_locked(proc_t parent, proc_t child, int deadparent, int reparentedtoinit, int locked, int droplock)
1020{
1021	proc_t trace_parent = PROC_NULL;	/* Traced parent process, if tracing */
1022
1023	if (locked == 1)
1024		proc_list_unlock();
1025
1026	/*
1027	 * If we got the child via a ptrace 'attach',
1028	 * we need to give it back to the old parent.
1029	 *
1030	 * Exception: someone who has been reparented to launchd before being
1031	 * ptraced can simply be reaped, refer to radar 5677288
1032	 * 	p_oppid 		 -> ptraced
1033	 * 	trace_parent == initproc -> away from launchd
1034	 * 	reparentedtoinit	 -> came to launchd by reparenting
1035	 */
1036	if (child->p_oppid) {
1037		int knote_hint;
1038		pid_t oppid;
1039
1040		proc_lock(child);
1041		oppid = child->p_oppid;
1042		child->p_oppid = 0;
1043		knote_hint = NOTE_EXIT | (child->p_xstat & 0xffff);
1044		proc_unlock(child);
1045
1046		if ((trace_parent = proc_find(oppid))
1047			&& !((trace_parent == initproc) && reparentedtoinit)) {
1048
1049			if (trace_parent != initproc) {
1050				/*
1051				 * proc internal fileds  and p_ucred usage safe
1052				 * here as child is dead and is not reaped or
1053				 * reparented yet
1054				 */
1055				proc_lock(trace_parent);
1056				trace_parent->si_pid = child->p_pid;
1057				trace_parent->si_status = child->p_xstat;
1058				trace_parent->si_code = CLD_CONTINUED;
1059				trace_parent->si_uid = kauth_cred_getruid(child->p_ucred);
1060				proc_unlock(trace_parent);
1061			}
1062			proc_reparentlocked(child, trace_parent, 1, 0);
1063
1064			/* resend knote to original parent (and others) after reparenting */
1065			proc_knote(child, knote_hint);
1066
1067			psignal(trace_parent, SIGCHLD);
1068			proc_list_lock();
1069			wakeup((caddr_t)trace_parent);
1070			child->p_listflag &= ~P_LIST_WAITING;
1071			wakeup(&child->p_stat);
1072			proc_list_unlock();
1073			proc_rele(trace_parent);
1074			if ((locked == 1) && (droplock == 0))
1075				proc_list_lock();
1076			return (0);
1077		}
1078
1079		/*
1080		 * If we can't reparent (e.g. the original parent exited while child was being debugged, or
1081		 * original parent is the same as the debugger currently exiting), we still need to satisfy
1082		 * the knote lifecycle for other observers on the system. While the debugger was attached,
1083		 * the NOTE_EXIT would not have been broadcast during initial child termination.
1084		 */
1085		proc_knote(child, knote_hint);
1086
1087		if (trace_parent != PROC_NULL) {
1088			proc_rele(trace_parent);
1089		}
1090	}
1091
1092#pragma clang diagnostic push
1093#pragma clang diagnostic ignored "-Wdeprecated-declarations"
1094	proc_knote(child, NOTE_REAP);
1095#pragma clang diagnostic pop
1096
1097	proc_knote_drain(child);
1098
1099	child->p_xstat = 0;
1100	if (child->p_ru) {
1101		proc_lock(parent);
1102#if 3839178
1103		/*
1104		 * If the parent is ignoring SIGCHLD, then POSIX requires
1105		 * us to not add the resource usage to the parent process -
1106		 * we are only going to hand it off to init to get reaped.
1107		 * We should contest the standard in this case on the basis
1108		 * of RLIMIT_CPU.
1109		 */
1110		if (!(parent->p_flag & P_NOCLDWAIT))
1111#endif	/* 3839178 */
1112			ruadd(&parent->p_stats->p_cru, &child->p_ru->ru);
1113		update_rusage_info_child(&parent->p_stats->ri_child, &child->p_ru->ri);
1114		proc_unlock(parent);
1115		FREE_ZONE(child->p_ru, sizeof *child->p_ru, M_ZOMBIE);
1116		child->p_ru = NULL;
1117	} else {
1118		printf("Warning : lost p_ru for %s\n", child->p_comm);
1119	}
1120
1121	AUDIT_SESSION_PROCEXIT(child);
1122
1123	/*
1124	 * Decrement the count of procs running with this uid.
1125	 * p_ucred usage is safe here as it is an exited process.
1126	 * and refernce is dropped after these calls down below
1127	 * (locking protection is provided by list lock held in chgproccnt)
1128	 */
1129	(void)chgproccnt(kauth_cred_getruid(child->p_ucred), -1);
1130
1131#if CONFIG_LCTX
1132	ALLLCTX_LOCK;
1133	leavelctx(child);
1134	ALLLCTX_UNLOCK;
1135#endif
1136
1137	/*
1138	 * Free up credentials.
1139	 */
1140	if (IS_VALID_CRED(child->p_ucred)) {
1141		kauth_cred_unref(&child->p_ucred);
1142	}
1143
1144	/*  XXXX Note NOT SAFE TO USE p_ucred from this point onwards */
1145
1146	/*
1147	 * Finally finished with old proc entry.
1148	 * Unlink it from its process group and free it.
1149	 */
1150	leavepgrp(child);
1151
1152	proc_list_lock();
1153	LIST_REMOVE(child, p_list);	/* off zombproc */
1154	parent->p_childrencnt--;
1155	LIST_REMOVE(child, p_sibling);
1156	/* If there are no more children wakeup parent */
1157	if ((deadparent != 0) && (LIST_EMPTY(&parent->p_children)))
1158		wakeup((caddr_t)parent);	/* with list lock held */
1159	child->p_listflag &= ~P_LIST_WAITING;
1160	wakeup(&child->p_stat);
1161
1162	/* Take it out of process hash */
1163	LIST_REMOVE(child, p_hash);
1164	child->p_listflag &= ~P_LIST_INHASH;
1165	proc_checkdeadrefs(child);
1166	nprocs--;
1167
1168	if (deadparent) {
1169		/*
1170		 * If a child zombie is being reaped because its parent
1171		 * is exiting, make sure we update the list flag
1172		 */
1173		child->p_listflag |= P_LIST_DEADPARENT;
1174	}
1175
1176	proc_list_unlock();
1177
1178#if CONFIG_FINE_LOCK_GROUPS
1179	lck_mtx_destroy(&child->p_mlock, proc_mlock_grp);
1180	lck_mtx_destroy(&child->p_fdmlock, proc_fdmlock_grp);
1181#if CONFIG_DTRACE
1182	lck_mtx_destroy(&child->p_dtrace_sprlock, proc_lck_grp);
1183#endif
1184	lck_spin_destroy(&child->p_slock, proc_slock_grp);
1185#else /* CONFIG_FINE_LOCK_GROUPS */
1186	lck_mtx_destroy(&child->p_mlock, proc_lck_grp);
1187	lck_mtx_destroy(&child->p_fdmlock, proc_lck_grp);
1188#if CONFIG_DTRACE
1189	lck_mtx_destroy(&child->p_dtrace_sprlock, proc_lck_grp);
1190#endif
1191	lck_spin_destroy(&child->p_slock, proc_lck_grp);
1192#endif /* CONFIG_FINE_LOCK_GROUPS */
1193	workqueue_destroy_lock(child);
1194
1195	FREE_ZONE(child, sizeof *child, M_PROC);
1196	if ((locked == 1) && (droplock == 0))
1197		proc_list_lock();
1198
1199	return (1);
1200}
1201
1202
1203int
1204wait1continue(int result)
1205{
1206	proc_t p;
1207	thread_t thread;
1208	uthread_t uth;
1209	struct _wait4_data *wait4_data;
1210	struct wait4_nocancel_args *uap;
1211	int *retval;
1212
1213	if (result)
1214		return(result);
1215
1216	p = current_proc();
1217	thread = current_thread();
1218	uth = (struct uthread *)get_bsdthread_info(thread);
1219
1220	wait4_data = &uth->uu_kevent.uu_wait4_data;
1221	uap = wait4_data->args;
1222	retval = wait4_data->retval;
1223	return(wait4_nocancel(p, uap, retval));
1224}
1225
1226int
1227wait4(proc_t q, struct wait4_args *uap, int32_t *retval)
1228{
1229	__pthread_testcancel(1);
1230	return(wait4_nocancel(q, (struct wait4_nocancel_args *)uap, retval));
1231}
1232
1233int
1234wait4_nocancel(proc_t q, struct wait4_nocancel_args *uap, int32_t *retval)
1235{
1236	int nfound;
1237	int sibling_count;
1238	proc_t p;
1239	int status, error;
1240	uthread_t uth;
1241	struct _wait4_data *wait4_data;
1242
1243	AUDIT_ARG(pid, uap->pid);
1244
1245	if (uap->pid == 0)
1246		uap->pid = -q->p_pgrpid;
1247
1248loop:
1249	proc_list_lock();
1250loop1:
1251	nfound = 0;
1252	sibling_count = 0;
1253
1254	for (p = q->p_children.lh_first; p != 0; p = p->p_sibling.le_next) {
1255		if ( p->p_sibling.le_next != 0 )
1256			sibling_count++;
1257		if (uap->pid != WAIT_ANY &&
1258		    p->p_pid != uap->pid &&
1259		    p->p_pgrpid != -(uap->pid))
1260			continue;
1261
1262		nfound++;
1263
1264		/* XXX This is racy because we don't get the lock!!!! */
1265
1266		if (p->p_listflag & P_LIST_WAITING) {
1267			(void)msleep(&p->p_stat, proc_list_mlock, PWAIT, "waitcoll", 0);
1268			goto loop1;
1269		}
1270		p->p_listflag |= P_LIST_WAITING;   /* only allow single thread to wait() */
1271
1272
1273		if (p->p_stat == SZOMB) {
1274			int reparentedtoinit = (p->p_listflag & P_LIST_DEADPARENT) ? 1 : 0;
1275
1276			proc_list_unlock();
1277#if CONFIG_MACF
1278			if ((error = mac_proc_check_wait(q, p)) != 0)
1279				goto out;
1280#endif
1281			retval[0] = p->p_pid;
1282			if (uap->status) {
1283				/* Legacy apps expect only 8 bits of status */
1284				status = 0xffff & p->p_xstat;	/* convert to int */
1285				error = copyout((caddr_t)&status,
1286				   			uap->status,
1287						    sizeof(status));
1288				if (error)
1289					goto out;
1290			}
1291			if (uap->rusage) {
1292				if (p->p_ru == NULL) {
1293					error = ENOMEM;
1294				} else {
1295					if (IS_64BIT_PROCESS(q)) {
1296						struct user64_rusage	my_rusage;
1297						munge_user64_rusage(&p->p_ru->ru, &my_rusage);
1298						error = copyout((caddr_t)&my_rusage,
1299							uap->rusage,
1300							sizeof (my_rusage));
1301					}
1302					else {
1303						struct user32_rusage	my_rusage;
1304						munge_user32_rusage(&p->p_ru->ru, &my_rusage);
1305						error = copyout((caddr_t)&my_rusage,
1306							uap->rusage,
1307							sizeof (my_rusage));
1308					}
1309				}
1310				/* information unavailable? */
1311				if (error)
1312					goto out;
1313			}
1314
1315			/* Conformance change for 6577252.
1316			 * When SIGCHLD is blocked and wait() returns because the status
1317			 * of a child process is available and there are no other
1318			 * children processes, then any pending SIGCHLD signal is cleared.
1319			 */
1320			if ( sibling_count == 0 ) {
1321				int mask = sigmask(SIGCHLD);
1322				uth = current_uthread();
1323
1324				if ( (uth->uu_sigmask & mask) != 0 ) {
1325					/* we are blocking SIGCHLD signals.  clear any pending SIGCHLD.
1326					 * This locking looks funny but it is protecting access to the
1327					 * thread via p_uthlist.
1328					 */
1329					proc_lock(q);
1330					uth->uu_siglist &= ~mask;	/* clear pending signal */
1331					proc_unlock(q);
1332				}
1333			}
1334
1335			/* Clean up */
1336			(void)reap_child_locked(q, p, 0, reparentedtoinit, 0, 0);
1337
1338			return (0);
1339		}
1340		if (p->p_stat == SSTOP && (p->p_lflag & P_LWAITED) == 0 &&
1341		    (p->p_lflag & P_LTRACED || uap->options & WUNTRACED)) {
1342			proc_list_unlock();
1343#if CONFIG_MACF
1344			if ((error = mac_proc_check_wait(q, p)) != 0)
1345				goto out;
1346#endif
1347			proc_lock(p);
1348			p->p_lflag |= P_LWAITED;
1349			proc_unlock(p);
1350			retval[0] = p->p_pid;
1351			if (uap->status) {
1352				status = W_STOPCODE(p->p_xstat);
1353				error = copyout((caddr_t)&status,
1354					uap->status,
1355				    sizeof(status));
1356			} else
1357				error = 0;
1358			goto out;
1359		}
1360		/*
1361		 * If we are waiting for continued processses, and this
1362		 * process was continued
1363		 */
1364		if ((uap->options & WCONTINUED) &&
1365		    (p->p_flag & P_CONTINUED)) {
1366			proc_list_unlock();
1367#if CONFIG_MACF
1368			if ((error = mac_proc_check_wait(q, p)) != 0)
1369				goto out;
1370#endif
1371
1372			/* Prevent other process for waiting for this event */
1373			OSBitAndAtomic(~((uint32_t)P_CONTINUED), &p->p_flag);
1374			retval[0] = p->p_pid;
1375			if (uap->status) {
1376				status = W_STOPCODE(SIGCONT);
1377				error = copyout((caddr_t)&status,
1378					uap->status,
1379				    sizeof(status));
1380			} else
1381				error = 0;
1382			goto out;
1383		}
1384		p->p_listflag &= ~P_LIST_WAITING;
1385		wakeup(&p->p_stat);
1386	}
1387	/* list lock is held when we get here any which way */
1388	if (nfound == 0) {
1389		proc_list_unlock();
1390		return (ECHILD);
1391	}
1392
1393	if (uap->options & WNOHANG) {
1394		retval[0] = 0;
1395		proc_list_unlock();
1396		return (0);
1397	}
1398
1399	/* Save arguments for continuation. Backing storage is in uthread->uu_arg, and will not be deallocated */
1400	uth = current_uthread();
1401	wait4_data = &uth->uu_kevent.uu_wait4_data;
1402	wait4_data->args = uap;
1403	wait4_data->retval = retval;
1404
1405	if ((error = msleep0((caddr_t)q, proc_list_mlock, PWAIT | PCATCH | PDROP, "wait", 0, wait1continue)))
1406		return (error);
1407
1408	goto loop;
1409out:
1410	proc_list_lock();
1411	p->p_listflag &= ~P_LIST_WAITING;
1412	wakeup(&p->p_stat);
1413	proc_list_unlock();
1414	return (error);
1415}
1416
1417#if DEBUG
1418#define ASSERT_LCK_MTX_OWNED(lock)	\
1419				lck_mtx_assert(lock, LCK_MTX_ASSERT_OWNED)
1420#else
1421#define ASSERT_LCK_MTX_OWNED(lock)	/* nothing */
1422#endif
1423
1424int
1425waitidcontinue(int result)
1426{
1427	proc_t p;
1428	thread_t thread;
1429	uthread_t uth;
1430	struct _waitid_data *waitid_data;
1431	struct waitid_nocancel_args *uap;
1432	int *retval;
1433
1434	if (result)
1435		return (result);
1436
1437	p = current_proc();
1438	thread = current_thread();
1439	uth = (struct uthread *)get_bsdthread_info(thread);
1440
1441	waitid_data = &uth->uu_kevent.uu_waitid_data;
1442	uap = waitid_data->args;
1443	retval = waitid_data->retval;
1444	return(waitid_nocancel(p, uap, retval));
1445}
1446
1447/*
1448 * Description:	Suspend the calling thread until one child of the process
1449 *		containing the calling thread changes state.
1450 *
1451 * Parameters:	uap->idtype		one of P_PID, P_PGID, P_ALL
1452 *		uap->id			pid_t or gid_t or ignored
1453 *		uap->infop		Address of siginfo_t struct in
1454 *					user space into which to return status
1455 *		uap->options		flag values
1456 *
1457 * Returns:	0			Success
1458 *		!0			Error returning status to user space
1459 */
1460int
1461waitid(proc_t q, struct waitid_args *uap, int32_t *retval)
1462{
1463	__pthread_testcancel(1);
1464	return (waitid_nocancel(q, (struct waitid_nocancel_args *)uap, retval));
1465}
1466
1467int
1468waitid_nocancel(proc_t q, struct waitid_nocancel_args *uap,
1469	__unused int32_t *retval)
1470{
1471	user_siginfo_t	siginfo;	/* siginfo data to return to caller */
1472	boolean_t caller64 = IS_64BIT_PROCESS(q);
1473	int nfound;
1474	proc_t p;
1475	int error;
1476	uthread_t uth;
1477	struct _waitid_data *waitid_data;
1478
1479	if (uap->options == 0 ||
1480	    (uap->options & ~(WNOHANG|WNOWAIT|WCONTINUED|WSTOPPED|WEXITED)))
1481		return (EINVAL);	/* bits set that aren't recognized */
1482
1483	switch (uap->idtype) {
1484	case P_PID:	/* child with process ID equal to... */
1485	case P_PGID:	/* child with process group ID equal to... */
1486		if (((int)uap->id) < 0)
1487			return (EINVAL);
1488		break;
1489	case P_ALL:	/* any child */
1490		break;
1491	}
1492
1493loop:
1494	proc_list_lock();
1495loop1:
1496	nfound = 0;
1497	for (p = q->p_children.lh_first; p != 0; p = p->p_sibling.le_next) {
1498
1499		switch (uap->idtype) {
1500		case P_PID:	/* child with process ID equal to... */
1501			if (p->p_pid != (pid_t)uap->id)
1502				continue;
1503			break;
1504		case P_PGID:	/* child with process group ID equal to... */
1505			if (p->p_pgrpid != (pid_t)uap->id)
1506				continue;
1507			break;
1508		case P_ALL:	/* any child */
1509			break;
1510		}
1511
1512		/* XXX This is racy because we don't get the lock!!!! */
1513
1514		/*
1515		 * Wait collision; go to sleep and restart; used to maintain
1516		 * the single return for waited process guarantee.
1517		 */
1518		if (p->p_listflag & P_LIST_WAITING) {
1519			(void) msleep(&p->p_stat, proc_list_mlock,
1520				PWAIT, "waitidcoll", 0);
1521			goto loop1;
1522		}
1523		p->p_listflag |= P_LIST_WAITING;		/* mark busy */
1524
1525		nfound++;
1526
1527		bzero(&siginfo, sizeof (siginfo));
1528
1529		switch (p->p_stat) {
1530		case SZOMB:		/* Exited */
1531			if (!(uap->options & WEXITED))
1532				break;
1533			proc_list_unlock();
1534#if CONFIG_MACF
1535			if ((error = mac_proc_check_wait(q, p)) != 0)
1536				goto out;
1537#endif
1538			siginfo.si_signo = SIGCHLD;
1539			siginfo.si_pid = p->p_pid;
1540			siginfo.si_status = WEXITSTATUS(p->p_xstat);
1541			if (WIFSIGNALED(p->p_xstat)) {
1542				siginfo.si_code = WCOREDUMP(p->p_xstat) ?
1543					CLD_DUMPED : CLD_KILLED;
1544			} else
1545				siginfo.si_code = CLD_EXITED;
1546
1547			if ((error = copyoutsiginfo(&siginfo,
1548			    caller64, uap->infop)) != 0)
1549				goto out;
1550
1551			/* Prevent other process for waiting for this event? */
1552			if (!(uap->options & WNOWAIT)) {
1553				(void) reap_child_locked(q, p, 0, 0, 0, 0);
1554				return (0);
1555			}
1556			goto out;
1557
1558		case SSTOP:		/* Stopped */
1559			/*
1560			 * If we are not interested in stopped processes, then
1561			 * ignore this one.
1562			 */
1563			if (!(uap->options & WSTOPPED))
1564				break;
1565
1566			/*
1567			 * If someone has already waited it, we lost a race
1568			 * to be the one to return status.
1569			 */
1570			if ((p->p_lflag & P_LWAITED) != 0)
1571				break;
1572			proc_list_unlock();
1573#if CONFIG_MACF
1574			if ((error = mac_proc_check_wait(q, p)) != 0)
1575				goto out;
1576#endif
1577			siginfo.si_signo = SIGCHLD;
1578			siginfo.si_pid = p->p_pid;
1579			siginfo.si_status = p->p_xstat;	/* signal number */
1580			siginfo.si_code = CLD_STOPPED;
1581
1582			if ((error = copyoutsiginfo(&siginfo,
1583			    caller64, uap->infop)) != 0)
1584				goto out;
1585
1586			/* Prevent other process for waiting for this event? */
1587			if (!(uap->options & WNOWAIT)) {
1588				proc_lock(p);
1589				p->p_lflag |= P_LWAITED;
1590				proc_unlock(p);
1591			}
1592			goto out;
1593
1594		default:		/* All other states => Continued */
1595			if (!(uap->options & WCONTINUED))
1596				break;
1597
1598			/*
1599			 * If the flag isn't set, then this process has not
1600			 * been stopped and continued, or the status has
1601			 * already been reaped by another caller of waitid().
1602			 */
1603			if ((p->p_flag & P_CONTINUED) == 0)
1604				break;
1605			proc_list_unlock();
1606#if CONFIG_MACF
1607			if ((error = mac_proc_check_wait(q, p)) != 0)
1608				goto out;
1609#endif
1610			siginfo.si_signo = SIGCHLD;
1611			siginfo.si_code = CLD_CONTINUED;
1612			proc_lock(p);
1613			siginfo.si_pid = p->p_contproc;
1614			siginfo.si_status = p->p_xstat;
1615			proc_unlock(p);
1616
1617			if ((error = copyoutsiginfo(&siginfo,
1618			    caller64, uap->infop)) != 0)
1619				goto out;
1620
1621			/* Prevent other process for waiting for this event? */
1622			if (!(uap->options & WNOWAIT)) {
1623				OSBitAndAtomic(~((uint32_t)P_CONTINUED),
1624				    &p->p_flag);
1625			}
1626			goto out;
1627		}
1628		ASSERT_LCK_MTX_OWNED(proc_list_mlock);
1629
1630		/* Not a process we are interested in; go on to next child */
1631
1632		p->p_listflag &= ~P_LIST_WAITING;
1633		wakeup(&p->p_stat);
1634	}
1635	ASSERT_LCK_MTX_OWNED(proc_list_mlock);
1636
1637	/* No child processes that could possibly satisfy the request? */
1638
1639	if (nfound == 0) {
1640		proc_list_unlock();
1641		return (ECHILD);
1642	}
1643
1644	if (uap->options & WNOHANG) {
1645		proc_list_unlock();
1646#if CONFIG_MACF
1647		if ((error = mac_proc_check_wait(q, p)) != 0)
1648			return (error);
1649#endif
1650		/*
1651		 * The state of the siginfo structure in this case
1652		 * is undefined.  Some implementations bzero it, some
1653		 * (like here) leave it untouched for efficiency.
1654		 *
1655		 * Thus the most portable check for "no matching pid with
1656		 * WNOHANG" is to store a zero into si_pid before
1657		 * invocation, then check for a non-zero value afterwards.
1658		 */
1659		return (0);
1660	}
1661
1662	/* Save arguments for continuation. Backing storage is in uthread->uu_arg, and will not be deallocated */
1663	uth = current_uthread();
1664	waitid_data = &uth->uu_kevent.uu_waitid_data;
1665	waitid_data->args = uap;
1666	waitid_data->retval = retval;
1667
1668	if ((error = msleep0(q, proc_list_mlock,
1669	    PWAIT | PCATCH | PDROP, "waitid", 0, waitidcontinue)) != 0)
1670		return (error);
1671
1672	goto loop;
1673out:
1674	proc_list_lock();
1675	p->p_listflag &= ~P_LIST_WAITING;
1676	wakeup(&p->p_stat);
1677	proc_list_unlock();
1678	return (error);
1679}
1680
1681/*
1682 * make process 'parent' the new parent of process 'child'.
1683 */
1684void
1685proc_reparentlocked(proc_t child, proc_t parent, int cansignal, int locked)
1686{
1687	proc_t oldparent = PROC_NULL;
1688
1689	if (child->p_pptr == parent)
1690		return;
1691
1692	if (locked == 0)
1693		proc_list_lock();
1694
1695	oldparent = child->p_pptr;
1696#if __PROC_INTERNAL_DEBUG
1697	if (oldparent == PROC_NULL)
1698		panic("proc_reparent: process %p does not have a parent\n", child);
1699#endif
1700
1701	LIST_REMOVE(child, p_sibling);
1702#if __PROC_INTERNAL_DEBUG
1703	if (oldparent->p_childrencnt == 0)
1704		panic("process children count already 0\n");
1705#endif
1706	oldparent->p_childrencnt--;
1707#if __PROC_INTERNAL_DEBUG1
1708	if (oldparent->p_childrencnt < 0)
1709		panic("process children count -ve\n");
1710#endif
1711	LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
1712	parent->p_childrencnt++;
1713	child->p_pptr = parent;
1714	child->p_ppid = parent->p_pid;
1715
1716	proc_list_unlock();
1717
1718	if ((cansignal != 0) && (initproc == parent) && (child->p_stat == SZOMB))
1719		psignal(initproc, SIGCHLD);
1720	if (locked == 1)
1721		proc_list_lock();
1722}
1723
1724/*
1725 * Exit: deallocate address space and other resources, change proc state
1726 * to zombie, and unlink proc from allproc and parent's lists.  Save exit
1727 * status and rusage for wait().  Check for child processes and orphan them.
1728 */
1729
1730void
1731vfork_exit(proc_t p, int rv)
1732{
1733	vfork_exit_internal(p, rv, 0);
1734}
1735
1736void
1737vfork_exit_internal(proc_t p, int rv, int forceexit)
1738{
1739	thread_t self = current_thread();
1740#ifdef FIXME
1741	struct task *task = p->task;
1742#endif
1743	struct uthread *ut;
1744
1745	/*
1746	 * If a thread in this task has already
1747	 * called exit(), then halt any others
1748	 * right here.
1749	 */
1750
1751	 ut = get_bsdthread_info(self);
1752
1753
1754	proc_lock(p);
1755	 if ((p->p_lflag & P_LPEXIT) == P_LPEXIT) {
1756		/*
1757	 	* This happens when a parent exits/killed and vfork is in progress
1758		* other threads. But shutdown code for ex has already called exit1()
1759	 	*/
1760		proc_unlock(p);
1761		return;
1762	}
1763	p->p_lflag |= (P_LEXIT | P_LPEXIT);
1764	proc_unlock(p);
1765
1766	if (forceexit == 0) {
1767		/*
1768		 * parent of a vfork child has already called exit() and the
1769		 * thread that has vfork in proress terminates. So there is no
1770		 * separate address space here and it has already been marked for
1771		 * termination. This was never covered before and could cause problems
1772		 * if we block here for outside code.
1773		 */
1774		/* Notify the perf server */
1775		(void)sys_perf_notify(self, p->p_pid);
1776	}
1777
1778	/*
1779	 * Remove proc from allproc queue and from pidhash chain.
1780	 * Need to do this before we do anything that can block.
1781	 * Not doing causes things like mount() find this on allproc
1782	 * in partially cleaned state.
1783	 */
1784
1785	proc_list_lock();
1786
1787#if CONFIG_MEMORYSTATUS
1788	memorystatus_remove(p, TRUE);
1789#endif
1790
1791	LIST_REMOVE(p, p_list);
1792	LIST_INSERT_HEAD(&zombproc, p, p_list);	/* Place onto zombproc. */
1793	/* will not be visible via proc_find */
1794	p->p_listflag |= P_LIST_EXITED;
1795
1796	proc_list_unlock();
1797
1798	proc_lock(p);
1799	p->p_xstat = rv;
1800	p->p_lflag &= ~(P_LTRACED | P_LPPWAIT);
1801	p->p_sigignore = ~0;
1802	proc_unlock(p);
1803
1804	proc_spinlock(p);
1805	if (thread_call_cancel(p->p_rcall))
1806		p->p_ractive--;
1807
1808	while (p->p_ractive > 0) {
1809		proc_spinunlock(p);
1810
1811		delay(1);
1812
1813		proc_spinlock(p);
1814	}
1815	proc_spinunlock(p);
1816
1817	thread_call_free(p->p_rcall);
1818	p->p_rcall = NULL;
1819
1820	ut->uu_siglist = 0;
1821
1822	vproc_exit(p);
1823}
1824
1825void
1826vproc_exit(proc_t p)
1827{
1828	proc_t q;
1829	proc_t pp;
1830
1831	vnode_t tvp;
1832#ifdef FIXME
1833	struct task *task = p->task;
1834#endif
1835	struct pgrp * pg;
1836	struct session *sessp;
1837	struct rusage_superset *rup;
1838
1839	/* XXX Zombie allocation may fail, in which case stats get lost */
1840	MALLOC_ZONE(rup, struct rusage_superset *,
1841			sizeof (*rup), M_ZOMBIE, M_WAITOK);
1842
1843	proc_refdrain(p);
1844
1845	/*
1846	 * Close open files and release open-file table.
1847	 * This may block!
1848	 */
1849	fdfree(p);
1850
1851	sessp = proc_session(p);
1852	if (SESS_LEADER(p, sessp)) {
1853
1854		if (sessp->s_ttyvp != NULLVP) {
1855			struct vnode *ttyvp;
1856			int ttyvid;
1857			int cttyflag = 0;
1858			struct vfs_context context;
1859			struct tty *tp;
1860
1861			/*
1862			 * Controlling process.
1863			 * Signal foreground pgrp,
1864			 * drain controlling terminal
1865			 * and revoke access to controlling terminal.
1866			 */
1867			session_lock(sessp);
1868			tp = SESSION_TP(sessp);
1869			if ((tp != TTY_NULL) && (tp->t_session == sessp)) {
1870				session_unlock(sessp);
1871
1872				/*
1873				 * We're going to SIGHUP the foreground process
1874				 * group. It can't change from this point on
1875				 * until the revoke is complete.
1876				 * The process group changes under both the tty
1877				 * lock and proc_list_lock but we need only one
1878				 */
1879				tty_lock(tp);
1880				ttysetpgrphup(tp);
1881				tty_unlock(tp);
1882
1883				tty_pgsignal(tp, SIGHUP, 1);
1884
1885				session_lock(sessp);
1886				tp = SESSION_TP(sessp);
1887			}
1888			cttyflag = sessp->s_flags & S_CTTYREF;
1889			sessp->s_flags &= ~S_CTTYREF;
1890			ttyvp = sessp->s_ttyvp;
1891			ttyvid = sessp->s_ttyvid;
1892			sessp->s_ttyvp = NULL;
1893			sessp->s_ttyvid = 0;
1894			sessp->s_ttyp = TTY_NULL;
1895			sessp->s_ttypgrpid = NO_PID;
1896			session_unlock(sessp);
1897
1898		       if ((ttyvp != NULLVP) && (vnode_getwithvid(ttyvp, ttyvid) == 0)) {
1899			        if (tp != TTY_NULL) {
1900					tty_lock(tp);
1901					(void) ttywait(tp);
1902					tty_unlock(tp);
1903				}
1904				context.vc_thread = proc_thread(p); /* XXX */
1905				context.vc_ucred = kauth_cred_proc_ref(p);
1906				VNOP_REVOKE(ttyvp, REVOKEALL, &context);
1907				if (cttyflag) {
1908					/*
1909					 * Release the extra usecount taken in cttyopen.
1910					 * usecount should be released after VNOP_REVOKE is called.
1911					 * This usecount was taken to ensure that
1912					 * the VNOP_REVOKE results in a close to
1913					 * the tty since cttyclose is a no-op.
1914					 */
1915					vnode_rele(ttyvp);
1916				}
1917				vnode_put(ttyvp);
1918				kauth_cred_unref(&context.vc_ucred);
1919				ttyvp = NULLVP;
1920			}
1921			if (tp) {
1922				/*
1923				 * This is cleared even if not set. This is also done in
1924				 * spec_close to ensure that the flag is cleared.
1925				 */
1926				tty_lock(tp);
1927				ttyclrpgrphup(tp);
1928				tty_unlock(tp);
1929
1930				ttyfree(tp);
1931			}
1932		}
1933		session_lock(sessp);
1934		sessp->s_leader = NULL;
1935		session_unlock(sessp);
1936	}
1937	session_rele(sessp);
1938
1939	pg = proc_pgrp(p);
1940	fixjobc(p, pg, 0);
1941	pg_rele(pg);
1942
1943	p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
1944
1945	proc_list_lock();
1946	proc_childdrainstart(p);
1947	while ((q = p->p_children.lh_first) != NULL) {
1948		if (q->p_stat == SZOMB) {
1949			if (p != q->p_pptr)
1950				panic("parent child linkage broken");
1951			/* check for lookups by zomb sysctl */
1952			while ((q->p_listflag & P_LIST_WAITING) == P_LIST_WAITING) {
1953				msleep(&q->p_stat, proc_list_mlock, PWAIT, "waitcoll", 0);
1954			}
1955			q->p_listflag |= P_LIST_WAITING;
1956			/*
1957			 * This is a named reference and it is not granted
1958			 * if the reap is already in progress. So we get
1959			 * the reference here exclusively and their can be
1960			 * no waiters. So there is no need for a wakeup
1961			 * after we are done. AlsO  the reap frees the structure
1962			 * and the proc struct cannot be used for wakeups as well.
1963			 * It is safe to use q here as this is system reap
1964			 */
1965			(void)reap_child_locked(p, q, 1, 0, 1, 0);
1966		} else {
1967			/*
1968		 	* Traced processes are killed
1969		 	* since their existence means someone is messing up.
1970		 	*/
1971			if (q->p_lflag & P_LTRACED) {
1972				struct proc *opp;
1973
1974				proc_list_unlock();
1975
1976				opp = proc_find(q->p_oppid);
1977				if (opp != PROC_NULL) {
1978					proc_list_lock();
1979					q->p_oppid = 0;
1980					proc_list_unlock();
1981					proc_reparentlocked(q, opp, 0, 0);
1982					proc_rele(opp);
1983				} else {
1984					/* original parent exited while traced */
1985					proc_list_lock();
1986					q->p_listflag |= P_LIST_DEADPARENT;
1987					q->p_oppid = 0;
1988					proc_list_unlock();
1989					proc_reparentlocked(q, initproc, 0, 0);
1990				}
1991
1992				proc_lock(q);
1993				q->p_lflag &= ~P_LTRACED;
1994
1995				if (q->sigwait_thread) {
1996					thread_t thread = q->sigwait_thread;
1997
1998					proc_unlock(q);
1999					/*
2000				 	* The sigwait_thread could be stopped at a
2001				 	* breakpoint. Wake it up to kill.
2002				 	* Need to do this as it could be a thread which is not
2003				 	* the first thread in the task. So any attempts to kill
2004				 	* the process would result into a deadlock on q->sigwait.
2005				 	*/
2006					thread_resume(thread);
2007					clear_wait(thread, THREAD_INTERRUPTED);
2008					threadsignal(thread, SIGKILL, 0);
2009				} else {
2010					proc_unlock(q);
2011				}
2012
2013				psignal(q, SIGKILL);
2014				proc_list_lock();
2015			} else {
2016				q->p_listflag |= P_LIST_DEADPARENT;
2017				proc_reparentlocked(q, initproc, 0, 1);
2018			}
2019		}
2020	}
2021
2022	proc_childdrainend(p);
2023	proc_list_unlock();
2024
2025	/*
2026	 * Release reference to text vnode
2027	 */
2028	tvp = p->p_textvp;
2029	p->p_textvp = NULL;
2030	if (tvp != NULLVP) {
2031		vnode_rele(tvp);
2032	}
2033
2034	/*
2035	 * Save exit status and final rusage info, adding in child rusage
2036	 * info and self times.  If we were unable to allocate a zombie
2037	 * structure, this information is lost.
2038	 */
2039	if (rup != NULL) {
2040	    rup->ru = p->p_stats->p_ru;
2041	    timerclear(&rup->ru.ru_utime);
2042	    timerclear(&rup->ru.ru_stime);
2043
2044#ifdef  FIXME
2045	    if (task) {
2046		mach_task_basic_info_data_t tinfo;
2047		task_thread_times_info_data_t ttimesinfo;
2048		int task_info_stuff, task_ttimes_stuff;
2049		struct timeval ut,st;
2050
2051		task_info_stuff	= MACH_TASK_BASIC_INFO_COUNT;
2052		task_info(task, MACH_TASK_BASIC_INFO,
2053			  &tinfo, &task_info_stuff);
2054		p->p_ru->ru.ru_utime.tv_sec = tinfo.user_time.seconds;
2055		p->p_ru->ru.ru_utime.tv_usec = tinfo.user_time.microseconds;
2056		p->p_ru->ru.ru_stime.tv_sec = tinfo.system_time.seconds;
2057		p->p_ru->ru.ru_stime.tv_usec = tinfo.system_time.microseconds;
2058
2059		task_ttimes_stuff = TASK_THREAD_TIMES_INFO_COUNT;
2060		task_info(task, TASK_THREAD_TIMES_INFO,
2061			  &ttimesinfo, &task_ttimes_stuff);
2062
2063		ut.tv_sec = ttimesinfo.user_time.seconds;
2064		ut.tv_usec = ttimesinfo.user_time.microseconds;
2065		st.tv_sec = ttimesinfo.system_time.seconds;
2066		st.tv_usec = ttimesinfo.system_time.microseconds;
2067		timeradd(&ut,&p->p_ru->ru.ru_utime,&p->p_ru->ru.ru_utime);
2068			timeradd(&st,&p->p_ru->ru.ru_stime,&p->p_ru->ru.ru_stime);
2069	    }
2070#endif /* FIXME */
2071
2072	    ruadd(&rup->ru, &p->p_stats->p_cru);
2073
2074		gather_rusage_info(p, &rup->ri, RUSAGE_INFO_CURRENT);
2075		rup->ri.ri_phys_footprint = 0;
2076		rup->ri.ri_proc_exit_abstime = mach_absolute_time();
2077
2078		/*
2079		 * Now that we have filled in the rusage info, make it
2080		 * visible to an external observer via proc_pid_rusage().
2081		 */
2082		p->p_ru = rup;
2083	}
2084
2085	/*
2086	 * Free up profiling buffers.
2087	 */
2088	{
2089		struct uprof *p0 = &p->p_stats->p_prof, *p1, *pn;
2090
2091		p1 = p0->pr_next;
2092		p0->pr_next = NULL;
2093		p0->pr_scale = 0;
2094
2095		for (; p1 != NULL; p1 = pn) {
2096			pn = p1->pr_next;
2097			kfree(p1, sizeof *p1);
2098		}
2099	}
2100
2101#if PSYNCH
2102	pth_proc_hashdelete(p);
2103#endif /* PSYNCH */
2104
2105	/*
2106	 * Other substructures are freed from wait().
2107	 */
2108	FREE_ZONE(p->p_stats, sizeof *p->p_stats, M_PSTATS);
2109	p->p_stats = NULL;
2110
2111	FREE_ZONE(p->p_sigacts, sizeof *p->p_sigacts, M_SIGACTS);
2112	p->p_sigacts = NULL;
2113
2114	proc_limitdrop(p, 1);
2115	p->p_limit = NULL;
2116
2117	/*
2118	 * Finish up by terminating the task
2119	 * and halt this thread (only if a
2120	 * member of the task exiting).
2121	 */
2122	p->task = TASK_NULL;
2123
2124	/*
2125	 * Notify parent that we're gone.
2126	 */
2127	pp = proc_parent(p);
2128	if ((p->p_listflag & P_LIST_DEADPARENT) == 0) {
2129		if (pp != initproc) {
2130			proc_lock(pp);
2131			pp->si_pid = p->p_pid;
2132			pp->si_status = p->p_xstat;
2133			pp->si_code = CLD_EXITED;
2134			/*
2135			 * p_ucred usage is safe as it is an exiting process
2136			 * and reference is dropped in reap
2137			 */
2138			pp->si_uid = kauth_cred_getruid(p->p_ucred);
2139			proc_unlock(pp);
2140		}
2141		/* mark as a zombie */
2142		/* mark as a zombie */
2143		/* No need to take proc lock as all refs are drained and
2144		 * no one except parent (reaping ) can look at this.
2145		 * The write is to an int and is coherent. Also parent is
2146		 *  keyed off of list lock for reaping
2147		 */
2148		p->p_stat = SZOMB;
2149
2150		psignal(pp, SIGCHLD);
2151
2152		/* and now wakeup the parent */
2153		proc_list_lock();
2154		wakeup((caddr_t)pp);
2155		proc_list_unlock();
2156	} else {
2157		proc_list_lock();
2158		/* check for lookups by zomb sysctl */
2159		while ((p->p_listflag & P_LIST_WAITING) == P_LIST_WAITING) {
2160			msleep(&p->p_stat, proc_list_mlock, PWAIT, "waitcoll", 0);
2161		}
2162		p->p_stat = SZOMB;
2163		p->p_listflag |= P_LIST_WAITING;
2164
2165		/*
2166		 * This is a named reference and it is not granted
2167		 * if the reap is already in progress. So we get
2168		 * the reference here exclusively and their can be
2169		 * no waiters. So there is no need for a wakeup
2170		 * after we are done. AlsO  the reap frees the structure
2171		 * and the proc struct cannot be used for wakeups as well.
2172		 * It is safe to use p here as this is system reap
2173		 */
2174		(void)reap_child_locked(pp, p, 0, 0, 1, 1);
2175		/* list lock dropped by reap_child_locked */
2176	}
2177	proc_rele(pp);
2178}
2179
2180
2181/*
2182 * munge_rusage
2183 *	LP64 support - long is 64 bits if we are dealing with a 64 bit user
2184 *	process.  We munge the kernel version of rusage into the
2185 *	64 bit version.
2186 */
2187__private_extern__  void
2188munge_user64_rusage(struct rusage *a_rusage_p, struct user64_rusage *a_user_rusage_p)
2189{
2190	/* timeval changes size, so utime and stime need special handling */
2191	a_user_rusage_p->ru_utime.tv_sec = a_rusage_p->ru_utime.tv_sec;
2192	a_user_rusage_p->ru_utime.tv_usec = a_rusage_p->ru_utime.tv_usec;
2193	a_user_rusage_p->ru_stime.tv_sec = a_rusage_p->ru_stime.tv_sec;
2194	a_user_rusage_p->ru_stime.tv_usec = a_rusage_p->ru_stime.tv_usec;
2195	/*
2196	 * everything else can be a direct assign, since there is no loss
2197	 * of precision implied boing 32->64.
2198	 */
2199	a_user_rusage_p->ru_maxrss = a_rusage_p->ru_maxrss;
2200	a_user_rusage_p->ru_ixrss = a_rusage_p->ru_ixrss;
2201	a_user_rusage_p->ru_idrss = a_rusage_p->ru_idrss;
2202	a_user_rusage_p->ru_isrss = a_rusage_p->ru_isrss;
2203	a_user_rusage_p->ru_minflt = a_rusage_p->ru_minflt;
2204	a_user_rusage_p->ru_majflt = a_rusage_p->ru_majflt;
2205	a_user_rusage_p->ru_nswap = a_rusage_p->ru_nswap;
2206	a_user_rusage_p->ru_inblock = a_rusage_p->ru_inblock;
2207	a_user_rusage_p->ru_oublock = a_rusage_p->ru_oublock;
2208	a_user_rusage_p->ru_msgsnd = a_rusage_p->ru_msgsnd;
2209	a_user_rusage_p->ru_msgrcv = a_rusage_p->ru_msgrcv;
2210	a_user_rusage_p->ru_nsignals = a_rusage_p->ru_nsignals;
2211	a_user_rusage_p->ru_nvcsw = a_rusage_p->ru_nvcsw;
2212	a_user_rusage_p->ru_nivcsw = a_rusage_p->ru_nivcsw;
2213}
2214
2215/* For a 64-bit kernel and 32-bit userspace, munging may be needed */
2216__private_extern__  void
2217munge_user32_rusage(struct rusage *a_rusage_p, struct user32_rusage *a_user_rusage_p)
2218{
2219	/* timeval changes size, so utime and stime need special handling */
2220	a_user_rusage_p->ru_utime.tv_sec = a_rusage_p->ru_utime.tv_sec;
2221	a_user_rusage_p->ru_utime.tv_usec = a_rusage_p->ru_utime.tv_usec;
2222	a_user_rusage_p->ru_stime.tv_sec = a_rusage_p->ru_stime.tv_sec;
2223	a_user_rusage_p->ru_stime.tv_usec = a_rusage_p->ru_stime.tv_usec;
2224	/*
2225	 * everything else can be a direct assign. We currently ignore
2226	 * the loss of precision
2227	 */
2228	a_user_rusage_p->ru_maxrss = a_rusage_p->ru_maxrss;
2229	a_user_rusage_p->ru_ixrss = a_rusage_p->ru_ixrss;
2230	a_user_rusage_p->ru_idrss = a_rusage_p->ru_idrss;
2231	a_user_rusage_p->ru_isrss = a_rusage_p->ru_isrss;
2232	a_user_rusage_p->ru_minflt = a_rusage_p->ru_minflt;
2233	a_user_rusage_p->ru_majflt = a_rusage_p->ru_majflt;
2234	a_user_rusage_p->ru_nswap = a_rusage_p->ru_nswap;
2235	a_user_rusage_p->ru_inblock = a_rusage_p->ru_inblock;
2236	a_user_rusage_p->ru_oublock = a_rusage_p->ru_oublock;
2237	a_user_rusage_p->ru_msgsnd = a_rusage_p->ru_msgsnd;
2238	a_user_rusage_p->ru_msgrcv = a_rusage_p->ru_msgrcv;
2239	a_user_rusage_p->ru_nsignals = a_rusage_p->ru_nsignals;
2240	a_user_rusage_p->ru_nvcsw = a_rusage_p->ru_nvcsw;
2241	a_user_rusage_p->ru_nivcsw = a_rusage_p->ru_nivcsw;
2242}
2243