2 * Copyright (c) 1982, 1986, 1989, 1991, 1993
3 * The Regents of the University of California. All rights reserved.
4 * (c) UNIX System Laboratories, Inc.
5 * All or some portions of this file are derived from material licensed
6 * to the University of California by American Telephone and Telegraph
7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8 * the permission of UNIX System Laboratories, Inc.
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 4. Neither the name of the University nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * @(#)kern_exit.c 8.7 (Berkeley) 2/12/94
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
40 #include "opt_compat.h"
41 #include "opt_ktrace.h"
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/sysproto.h>
46 #include <sys/capsicum.h>
47 #include <sys/eventhandler.h>
48 #include <sys/kernel.h>
49 #include <sys/malloc.h>
51 #include <sys/mutex.h>
53 #include <sys/procdesc.h>
54 #include <sys/pioctl.h>
58 #include <sys/vmmeter.h>
59 #include <sys/vnode.h>
60 #include <sys/racct.h>
61 #include <sys/resourcevar.h>
63 #include <sys/signalvar.h>
64 #include <sys/sched.h>
66 #include <sys/syscallsubr.h>
67 #include <sys/syslog.h>
68 #include <sys/ptrace.h>
69 #include <sys/acct.h> /* for acct_process() function prototype */
70 #include <sys/filedesc.h>
75 #include <sys/ktrace.h>
78 #include <security/audit/audit.h>
79 #include <security/mac/mac_framework.h>
82 #include <vm/vm_extern.h>
83 #include <vm/vm_param.h>
85 #include <vm/vm_map.h>
86 #include <vm/vm_page.h>
90 #include <sys/dtrace_bsd.h>
91 dtrace_execexit_func_t dtrace_fasttrap_exit;
94 SDT_PROVIDER_DECLARE(proc);
95 SDT_PROBE_DEFINE1(proc, kernel, , exit, "int");
97 /* Hook for NFS teardown procedure. */
98 void (*nlminfo_release_p)(struct proc *p);
101 proc_realparent(struct proc *child)
103 struct proc *p, *parent;
105 sx_assert(&proctree_lock, SX_LOCKED);
106 if ((child->p_treeflag & P_TREE_ORPHANED) == 0) {
107 if (child->p_oppid == 0 ||
108 child->p_pptr->p_pid == child->p_oppid)
109 parent = child->p_pptr;
114 for (p = child; (p->p_treeflag & P_TREE_FIRST_ORPHAN) == 0;) {
115 /* Cannot use LIST_PREV(), since the list head is not known. */
116 p = __containerof(p->p_orphan.le_prev, struct proc,
118 KASSERT((p->p_treeflag & P_TREE_ORPHANED) != 0,
119 ("missing P_ORPHAN %p", p));
121 parent = __containerof(p->p_orphan.le_prev, struct proc,
127 clear_orphan(struct proc *p)
131 sx_assert(&proctree_lock, SA_XLOCKED);
132 if ((p->p_treeflag & P_TREE_ORPHANED) == 0)
134 if ((p->p_treeflag & P_TREE_FIRST_ORPHAN) != 0) {
135 p1 = LIST_NEXT(p, p_orphan);
137 p1->p_treeflag |= P_TREE_FIRST_ORPHAN;
138 p->p_treeflag &= ~P_TREE_FIRST_ORPHAN;
140 LIST_REMOVE(p, p_orphan);
141 p->p_treeflag &= ~P_TREE_ORPHANED;
145 * exit -- death of process.
148 sys_sys_exit(struct thread *td, struct sys_exit_args *uap)
151 exit1(td, W_EXITCODE(uap->rval, 0));
156 * Exit: deallocate address space and other resources, change proc state to
157 * zombie, and unlink proc from allproc and parent's lists. Save exit status
158 * and rusage for wait(). Check for child processes and orphan them.
161 exit1(struct thread *td, int rv)
163 struct proc *p, *nq, *q, *t;
165 struct vnode *ttyvp = NULL;
167 mtx_assert(&Giant, MA_NOTOWNED);
171 * XXX in case we're rebooting we just let init die in order to
172 * work around an unsolved stack overflow seen very late during
173 * shutdown on sparc64 when the gmirror worker process exists.
175 if (p == initproc && rebooting == 0) {
176 printf("init died (signal %d, exit %d)\n",
177 WTERMSIG(rv), WEXITSTATUS(rv));
178 panic("Going nowhere without my init!");
182 * MUST abort all other threads before proceeding past here.
185 while (p->p_flag & P_HADTHREADS) {
187 * First check if some other thread got here before us.
188 * If so, act appropriately: exit or suspend.
190 thread_suspend_check(0);
193 * Kill off the other threads. This requires
194 * some co-operation from other parts of the kernel
195 * so it may not be instantaneous. With this state set
196 * any thread entering the kernel from userspace will
197 * thread_exit() in trap(). Any thread attempting to
198 * sleep will return immediately with EINTR or EWOULDBLOCK
199 * which will hopefully force them to back out to userland
200 * freeing resources as they go. Any thread attempting
201 * to return to userland will thread_exit() from userret().
202 * thread_exit() will unsuspend us when the last of the
203 * other threads exits.
204 * If there is already a thread singler after resumption,
205 * calling thread_single will fail; in that case, we just
206 * re-check all suspension request, the thread should
207 * either be suspended there or exit.
209 if (!thread_single(SINGLE_EXIT))
213 * All other activity in this process is now stopped.
214 * Threading support has been turned off.
217 KASSERT(p->p_numthreads == 1,
218 ("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
219 racct_sub(p, RACCT_NTHR, 1);
221 * Wakeup anyone in procfs' PIOCWAIT. They should have a hold
222 * on our vmspace, so we should block below until they have
223 * released their reference to us. Note that if they have
224 * requested S_EXIT stops we will block here until they ack
227 _STOPEVENT(p, S_EXIT, rv);
230 * Ignore any pending request to stop due to a stop signal.
231 * Once P_WEXIT is set, future requests will be ignored as
234 p->p_flag &= ~P_STOPPED_SIG;
235 KASSERT(!P_SHOULDSTOP(p), ("exiting process is stopped"));
238 * Note that we are exiting and do another wakeup of anyone in
239 * PIOCWAIT in case they aren't listening for S_EXIT stops or
240 * decided to wait again after we told them we are exiting.
242 p->p_flag |= P_WEXIT;
246 * Wait for any processes that have a hold on our vmspace to
247 * release their reference.
249 while (p->p_lock > 0)
250 msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0);
252 p->p_xstat = rv; /* Let event handler change exit status */
254 /* Drain the limit callout while we don't have the proc locked */
255 callout_drain(&p->p_limco);
259 * The Sun BSM exit token contains two components: an exit status as
260 * passed to exit(), and a return value to indicate what sort of exit
261 * it was. The exit status is WEXITSTATUS(rv), but it's not clear
262 * what the return value is.
264 AUDIT_ARG_EXIT(WEXITSTATUS(rv), 0);
265 AUDIT_SYSCALL_EXIT(0, td);
268 /* Are we a task leader? */
269 if (p == p->p_leader) {
270 mtx_lock(&ppeers_lock);
274 kern_psignal(q, SIGKILL);
278 while (p->p_peers != NULL)
279 msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
280 mtx_unlock(&ppeers_lock);
284 * Check if any loadable modules need anything done at process exit.
285 * E.g. SYSV IPC stuff
286 * XXX what if one of these generates an error?
288 EVENTHANDLER_INVOKE(process_exit, p);
291 * If parent is waiting for us to exit or exec,
292 * P_PPWAIT is set; we will wakeup the parent below.
295 rv = p->p_xstat; /* Event handler could change exit status */
297 p->p_flag &= ~(P_TRACED | P_PPWAIT | P_PPTRACE);
300 * Stop the real interval timer. If the handler is currently
301 * executing, prevent it from rearming itself and let it finish.
303 if (timevalisset(&p->p_realtimer.it_value) &&
304 callout_stop(&p->p_itcallout) == 0) {
305 timevalclear(&p->p_realtimer.it_interval);
306 msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
307 KASSERT(!timevalisset(&p->p_realtimer.it_value),
308 ("realtime timer is still armed"));
313 * Reset any sigio structures pointing to us as a result of
314 * F_SETOWN with our pid.
316 funsetownlst(&p->p_sigiolst);
319 * If this process has an nlminfo data area (for lockd), release it
321 if (nlminfo_release_p != NULL && p->p_nlminfo != NULL)
322 (*nlminfo_release_p)(p);
325 * Close open files and release open-file table.
331 * If this thread tickled GEOM, we need to wait for the giggling to
332 * stop before we return to userland
334 if (td->td_pflags & TDP_GEOM)
338 * Remove ourself from our leader's peer list and wake our leader.
340 mtx_lock(&ppeers_lock);
341 if (p->p_leader->p_peers) {
343 while (q->p_peers != p)
345 q->p_peers = p->p_peers;
348 mtx_unlock(&ppeers_lock);
352 sx_xlock(&proctree_lock);
353 if (SESS_LEADER(p)) {
354 struct session *sp = p->p_session;
358 * s_ttyp is not zero'd; we use this to indicate that
359 * the session once had a controlling terminal. (for
360 * logging and informational purposes)
371 * Signal foreground pgrp and revoke access to
372 * controlling terminal if it has not been revoked
375 * Because the TTY may have been revoked in the mean
376 * time and could already have a new session associated
377 * with it, make sure we don't send a SIGHUP to a
378 * foreground process group that does not belong to this
384 if (tp->t_session == sp)
385 tty_signal_pgrp(tp, SIGHUP);
390 sx_xunlock(&proctree_lock);
391 if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
392 VOP_REVOKE(ttyvp, REVOKEALL);
393 VOP_UNLOCK(ttyvp, 0);
395 sx_xlock(&proctree_lock);
398 fixjobc(p, p->p_pgrp, 0);
399 sx_xunlock(&proctree_lock);
400 (void)acct_process(td);
402 /* Release the TTY now we've unlocked everything. */
409 * Release reference to text vnode
411 if (p->p_textvp != NULL) {
417 * Release our limits structure.
419 lim_free(p->p_limit);
425 * Remove proc from allproc queue and pidhash chain.
426 * Place onto zombproc. Unlink from parent's child list.
428 sx_xlock(&allproc_lock);
429 LIST_REMOVE(p, p_list);
430 LIST_INSERT_HEAD(&zombproc, p, p_list);
431 LIST_REMOVE(p, p_hash);
432 sx_xunlock(&allproc_lock);
435 * Call machine-dependent code to release any
436 * machine-dependent resources other than the address space.
437 * The address space is released by "vmspace_exitfree(p)" in
442 WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
445 * Reparent all children processes:
446 * - traced ones to the original parent (or init if we are that parent)
449 sx_xlock(&proctree_lock);
450 q = LIST_FIRST(&p->p_children);
451 if (q != NULL) /* only need this if any child is S_ZOMB */
453 for (; q != NULL; q = nq) {
454 nq = LIST_NEXT(q, p_sibling);
456 q->p_sigparent = SIGCHLD;
458 if (!(q->p_flag & P_TRACED)) {
459 proc_reparent(q, initproc);
462 * Traced processes are killed since their existence
463 * means someone is screwing up.
465 t = proc_realparent(q);
467 proc_reparent(q, initproc);
474 * Since q was found on our children list, the
475 * proc_reparent() call moved q to the orphan
476 * list due to present P_TRACED flag. Clear
477 * orphan link for q now while q is locked.
480 q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
481 FOREACH_THREAD_IN_PROC(q, tdt)
482 tdt->td_dbgflags &= ~TDB_SUSPEND;
483 kern_psignal(q, SIGKILL);
489 * Also get rid of our orphans.
491 while ((q = LIST_FIRST(&p->p_orphans)) != NULL) {
497 /* Save exit status. */
501 /* Tell the prison that we are gone. */
502 prison_proc_free(p->p_ucred->cr_prison);
506 * Tell the DTrace fasttrap provider about the exit if it
507 * has declared an interest.
509 if (dtrace_fasttrap_exit)
510 dtrace_fasttrap_exit(p);
514 * Notify interested parties of our demise.
516 KNOTE_LOCKED(&p->p_klist, NOTE_EXIT);
519 int reason = CLD_EXITED;
522 else if (WIFSIGNALED(rv))
524 SDT_PROBE(proc, kernel, , exit, reason, 0, 0, 0, 0);
528 * Just delete all entries in the p_klist. At this point we won't
529 * report any more events, and there are nasty race conditions that
530 * can beat us if we don't.
532 knlist_clear(&p->p_klist, 1);
535 * If this is a process with a descriptor, we may not need to deliver
536 * a signal to the parent. proctree_lock is held over
537 * procdesc_exit() to serialize concurrent calls to close() and
540 if (p->p_procdesc == NULL || procdesc_exit(p)) {
542 * Notify parent that we're gone. If parent has the
543 * PS_NOCLDWAIT flag set, or if the handler is set to SIG_IGN,
544 * notify process 1 instead (and hope it will handle this
547 PROC_LOCK(p->p_pptr);
548 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
549 if (p->p_pptr->p_sigacts->ps_flag &
550 (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
553 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
556 proc_reparent(p, initproc);
557 p->p_sigparent = SIGCHLD;
558 PROC_LOCK(p->p_pptr);
561 * Notify parent, so in case he was wait(2)ing or
562 * executing waitpid(2) with our pid, he will
567 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
569 if (p->p_pptr == initproc)
570 kern_psignal(p->p_pptr, SIGCHLD);
571 else if (p->p_sigparent != 0) {
572 if (p->p_sigparent == SIGCHLD)
574 else /* LINUX thread */
575 kern_psignal(p->p_pptr, p->p_sigparent);
578 PROC_LOCK(p->p_pptr);
579 sx_xunlock(&proctree_lock);
582 * The state PRS_ZOMBIE prevents other proesses from sending
583 * signal to the process, to avoid memory leak, we free memory
584 * for signal queue at the time when the state is set.
586 sigqueue_flush(&p->p_sigqueue);
587 sigqueue_flush(&td->td_sigqueue);
590 * We have to wait until after acquiring all locks before
591 * changing p_state. We need to avoid all possible context
592 * switches (including ones from blocking on a mutex) while
593 * marked as a zombie. We also have to set the zombie state
594 * before we release the parent process' proc lock to avoid
595 * a lost wakeup. So, we first call wakeup, then we grab the
596 * sched lock, update the state, and release the parent process'
600 cv_broadcast(&p->p_pwait);
601 sched_exit(p->p_pptr, td);
603 p->p_state = PRS_ZOMBIE;
604 PROC_UNLOCK(p->p_pptr);
607 * Hopefully no one will try to deliver a signal to the process this
610 knlist_destroy(&p->p_klist);
613 * Save our children's rusage information in our exit rusage.
615 ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
618 * Make sure the scheduler takes this thread out of its tables etc.
619 * This will also release this thread's reference to the ucred.
620 * Other thread parts to release include pcb bits and such.
626 #ifndef _SYS_SYSPROTO_H_
635 sys_abort2(struct thread *td, struct abort2_args *uap)
637 struct proc *p = td->td_proc;
643 * Do it right now so we can log either proper call of abort2(), or
644 * note, that invalid argument was passed. 512 is big enough to
645 * handle 16 arguments' descriptions with additional comments.
647 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN);
649 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ",
650 p->p_comm, p->p_pid, td->td_ucred->cr_uid);
652 * Since we can't return from abort2(), send SIGKILL in cases, where
653 * abort2() was called improperly
656 /* Prevent from DoSes from user-space. */
657 if (uap->nargs < 0 || uap->nargs > 16)
659 if (uap->nargs > 0) {
660 if (uap->args == NULL)
662 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *));
667 * Limit size of 'reason' string to 128. Will fit even when
668 * maximal number of arguments was chosen to be logged.
670 if (uap->why != NULL) {
671 error = sbuf_copyin(sb, uap->why, 128);
675 sbuf_printf(sb, "(null)");
677 if (uap->nargs > 0) {
678 sbuf_printf(sb, "(");
679 for (i = 0;i < uap->nargs; i++)
680 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]);
681 sbuf_printf(sb, ")");
684 * Final stage: arguments were proper, string has been
685 * successfully copied from userspace, and copying pointers
686 * from user-space succeed.
690 if (sig == SIGKILL) {
692 sbuf_printf(sb, " (Reason text inaccessible)");
696 log(LOG_INFO, "%s", sbuf_data(sb));
698 exit1(td, W_EXITCODE(0, sig));
705 * The dirty work is handled by kern_wait().
708 owait(struct thread *td, struct owait_args *uap __unused)
712 error = kern_wait(td, WAIT_ANY, &status, 0, NULL);
714 td->td_retval[1] = status;
717 #endif /* COMPAT_43 */
720 * The dirty work is handled by kern_wait().
723 sys_wait4(struct thread *td, struct wait4_args *uap)
725 struct rusage ru, *rup;
728 if (uap->rusage != NULL)
732 error = kern_wait(td, uap->pid, &status, uap->options, rup);
733 if (uap->status != NULL && error == 0)
734 error = copyout(&status, uap->status, sizeof(status));
735 if (uap->rusage != NULL && error == 0)
736 error = copyout(&ru, uap->rusage, sizeof(struct rusage));
741 sys_wait6(struct thread *td, struct wait6_args *uap)
743 struct __wrusage wru, *wrup;
749 idtype = uap->idtype;
752 if (uap->wrusage != NULL)
757 if (uap->info != NULL) {
759 bzero(sip, sizeof(*sip));
764 * We expect all callers of wait6() to know about WEXITED and
767 error = kern_wait6(td, idtype, id, &status, uap->options, wrup, sip);
769 if (uap->status != NULL && error == 0)
770 error = copyout(&status, uap->status, sizeof(status));
771 if (uap->wrusage != NULL && error == 0)
772 error = copyout(&wru, uap->wrusage, sizeof(wru));
773 if (uap->info != NULL && error == 0)
774 error = copyout(&si, uap->info, sizeof(si));
779 * Reap the remains of a zombie process and optionally return status and
780 * rusage. Asserts and will release both the proctree_lock and the process
781 * lock as part of its work.
784 proc_reap(struct thread *td, struct proc *p, int *status, int options)
788 sx_assert(&proctree_lock, SA_XLOCKED);
789 PROC_LOCK_ASSERT(p, MA_OWNED);
790 PROC_SLOCK_ASSERT(p, MA_OWNED);
791 KASSERT(p->p_state == PRS_ZOMBIE, ("proc_reap: !PRS_ZOMBIE"));
796 td->td_retval[0] = p->p_pid;
798 *status = p->p_xstat; /* convert to int */
799 if (options & WNOWAIT) {
801 * Only poll, returning the status. Caller does not wish to
802 * release the proc struct just yet.
805 sx_xunlock(&proctree_lock);
810 sigqueue_take(p->p_ksi);
815 * If we got the child via a ptrace 'attach', we need to give it back
818 if (p->p_oppid != 0) {
819 t = proc_realparent(p);
825 pksignal(t, SIGCHLD, p->p_ksi);
827 cv_broadcast(&p->p_pwait);
829 sx_xunlock(&proctree_lock);
834 * Remove other references to this process to ensure we have an
835 * exclusive reference.
837 sx_xlock(&allproc_lock);
838 LIST_REMOVE(p, p_list); /* off zombproc */
839 sx_xunlock(&allproc_lock);
840 LIST_REMOVE(p, p_sibling);
845 if (p->p_procdesc != NULL)
847 sx_xunlock(&proctree_lock);
850 * As a side effect of this lock, we know that all other writes to
851 * this proc are visible now, so no more locking is needed for p.
854 p->p_xstat = 0; /* XXX: why? */
857 ruadd(&q->p_stats->p_cru, &q->p_crux, &p->p_ru, &p->p_rux);
861 * Decrement the count of procs running with this uid.
863 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
866 * Destroy resource accounting information associated with the process.
870 racct_sub(p, RACCT_NPROC, 1);
876 * Free credentials, arguments, and sigacts.
880 pargs_drop(p->p_args);
882 sigacts_free(p->p_sigacts);
886 * Do any thread-system specific cleanups.
891 * Give vm and machine-dependent layer a chance to free anything that
892 * cpu_exit couldn't release while still running in process context.
898 KASSERT(FIRST_THREAD_IN_PROC(p),
899 ("proc_reap: no residual thread!"));
900 uma_zfree(proc_zone, p);
901 sx_xlock(&allproc_lock);
903 sx_xunlock(&allproc_lock);
907 proc_to_reap(struct thread *td, struct proc *p, idtype_t idtype, id_t id,
908 int *status, int options, struct __wrusage *wrusage, siginfo_t *siginfo)
913 sx_assert(&proctree_lock, SA_XLOCKED);
922 if (p->p_pid != (pid_t)id) {
928 if (p->p_pgid != (pid_t)id) {
934 if (p->p_session->s_sid != (pid_t)id) {
940 if (p->p_ucred->cr_uid != (uid_t)id) {
946 if (p->p_ucred->cr_gid != (gid_t)id) {
952 if (p->p_ucred->cr_prison->pr_id != (int)id) {
958 * It seems that the thread structures get zeroed out
959 * at process exit. This makes it impossible to
960 * support P_SETID, P_CID or P_CPUID.
967 if (p_canwait(td, p)) {
972 if (((options & WEXITED) == 0) && (p->p_state == PRS_ZOMBIE)) {
978 * This special case handles a kthread spawned by linux_clone
979 * (see linux_misc.c). The linux_wait4 and linux_waitpid
980 * functions need to be able to distinguish between waiting
981 * on a process and waiting on a thread. It is a thread if
982 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
983 * signifies we want to wait for threads and not processes.
985 if ((p->p_sigparent != SIGCHLD) ^
986 ((options & WLINUXCLONE) != 0)) {
993 if (siginfo != NULL) {
994 bzero(siginfo, sizeof(*siginfo));
995 siginfo->si_errno = 0;
998 * SUSv4 requires that the si_signo value is always
999 * SIGCHLD. Obey it despite the rfork(2) interface
1000 * allows to request other signal for child exit
1003 siginfo->si_signo = SIGCHLD;
1006 * This is still a rough estimate. We will fix the
1007 * cases TRAPPED, STOPPED, and CONTINUED later.
1009 if (WCOREDUMP(p->p_xstat)) {
1010 siginfo->si_code = CLD_DUMPED;
1011 siginfo->si_status = WTERMSIG(p->p_xstat);
1012 } else if (WIFSIGNALED(p->p_xstat)) {
1013 siginfo->si_code = CLD_KILLED;
1014 siginfo->si_status = WTERMSIG(p->p_xstat);
1016 siginfo->si_code = CLD_EXITED;
1017 siginfo->si_status = WEXITSTATUS(p->p_xstat);
1020 siginfo->si_pid = p->p_pid;
1021 siginfo->si_uid = p->p_ucred->cr_uid;
1024 * The si_addr field would be useful additional
1025 * detail, but apparently the PC value may be lost
1026 * when we reach this point. bzero() above sets
1027 * siginfo->si_addr to NULL.
1032 * There should be no reason to limit resources usage info to
1033 * exited processes only. A snapshot about any resources used
1034 * by a stopped process may be exactly what is needed.
1036 if (wrusage != NULL) {
1037 rup = &wrusage->wru_self;
1039 calcru(p, &rup->ru_utime, &rup->ru_stime);
1041 rup = &wrusage->wru_children;
1042 *rup = p->p_stats->p_cru;
1043 calccru(p, &rup->ru_utime, &rup->ru_stime);
1046 if (p->p_state == PRS_ZOMBIE) {
1047 proc_reap(td, p, status, options);
1056 kern_wait(struct thread *td, pid_t pid, int *status, int options,
1057 struct rusage *rusage)
1059 struct __wrusage wru, *wrup;
1065 * Translate the special pid values into the (idtype, pid)
1066 * pair for kern_wait6. The WAIT_MYPGRP case is handled by
1067 * kern_wait6() on its own.
1069 if (pid == WAIT_ANY) {
1072 } else if (pid < 0) {
1086 * For backward compatibility we implicitly add flags WEXITED
1087 * and WTRAPPED here.
1089 options |= WEXITED | WTRAPPED;
1090 ret = kern_wait6(td, idtype, id, status, options, wrup, NULL);
1092 *rusage = wru.wru_self;
1097 kern_wait6(struct thread *td, idtype_t idtype, id_t id, int *status,
1098 int options, struct __wrusage *wrusage, siginfo_t *siginfo)
1101 int error, nfound, ret;
1103 AUDIT_ARG_VALUE((int)idtype); /* XXX - This is likely wrong! */
1104 AUDIT_ARG_PID((pid_t)id); /* XXX - This may be wrong! */
1105 AUDIT_ARG_VALUE(options);
1109 if ((pid_t)id == WAIT_MYPGRP && (idtype == P_PID || idtype == P_PGID)) {
1111 id = (id_t)q->p_pgid;
1116 /* If we don't know the option, just return. */
1117 if ((options & ~(WUNTRACED | WNOHANG | WCONTINUED | WNOWAIT |
1118 WEXITED | WTRAPPED | WLINUXCLONE)) != 0)
1120 if ((options & (WEXITED | WUNTRACED | WCONTINUED | WTRAPPED)) == 0) {
1122 * We will be unable to find any matching processes,
1123 * because there are no known events to look for.
1124 * Prefer to return error instead of blocking
1131 if (q->p_flag & P_STATCHILD) {
1133 q->p_flag &= ~P_STATCHILD;
1137 sx_xlock(&proctree_lock);
1138 LIST_FOREACH(p, &q->p_children, p_sibling) {
1139 ret = proc_to_reap(td, p, idtype, id, status, options,
1151 if ((options & WTRAPPED) != 0 &&
1152 (p->p_flag & P_TRACED) != 0 &&
1153 (p->p_flag & (P_STOPPED_TRACE | P_STOPPED_SIG)) != 0 &&
1154 (p->p_suspcount == p->p_numthreads) &&
1155 ((p->p_flag & P_WAITED) == 0)) {
1157 if ((options & WNOWAIT) == 0)
1158 p->p_flag |= P_WAITED;
1159 sx_xunlock(&proctree_lock);
1160 td->td_retval[0] = p->p_pid;
1163 *status = W_STOPCODE(p->p_xstat);
1164 if (siginfo != NULL) {
1165 siginfo->si_status = p->p_xstat;
1166 siginfo->si_code = CLD_TRAPPED;
1168 if ((options & WNOWAIT) == 0) {
1170 sigqueue_take(p->p_ksi);
1177 if ((options & WUNTRACED) != 0 &&
1178 (p->p_flag & P_STOPPED_SIG) != 0 &&
1179 (p->p_suspcount == p->p_numthreads) &&
1180 ((p->p_flag & P_WAITED) == 0)) {
1182 if ((options & WNOWAIT) == 0)
1183 p->p_flag |= P_WAITED;
1184 sx_xunlock(&proctree_lock);
1185 td->td_retval[0] = p->p_pid;
1188 *status = W_STOPCODE(p->p_xstat);
1189 if (siginfo != NULL) {
1190 siginfo->si_status = p->p_xstat;
1191 siginfo->si_code = CLD_STOPPED;
1193 if ((options & WNOWAIT) == 0) {
1195 sigqueue_take(p->p_ksi);
1203 if ((options & WCONTINUED) != 0 &&
1204 (p->p_flag & P_CONTINUED) != 0) {
1205 sx_xunlock(&proctree_lock);
1206 td->td_retval[0] = p->p_pid;
1207 if ((options & WNOWAIT) == 0) {
1208 p->p_flag &= ~P_CONTINUED;
1210 sigqueue_take(p->p_ksi);
1217 if (siginfo != NULL) {
1218 siginfo->si_status = SIGCONT;
1219 siginfo->si_code = CLD_CONTINUED;
1227 * Look in the orphans list too, to allow the parent to
1228 * collect it's child exit status even if child is being
1231 * Debugger detaches from the parent upon successful
1232 * switch-over from parent to child. At this point due to
1233 * re-parenting the parent loses the child to debugger and a
1234 * wait4(2) call would report that it has no children to wait
1235 * for. By maintaining a list of orphans we allow the parent
1236 * to successfully wait until the child becomes a zombie.
1238 LIST_FOREACH(p, &q->p_orphans, p_orphan) {
1239 ret = proc_to_reap(td, p, idtype, id, status, options,
1249 sx_xunlock(&proctree_lock);
1252 if (options & WNOHANG) {
1253 sx_xunlock(&proctree_lock);
1254 td->td_retval[0] = 0;
1258 sx_xunlock(&proctree_lock);
1259 if (q->p_flag & P_STATCHILD) {
1260 q->p_flag &= ~P_STATCHILD;
1263 error = msleep(q, &q->p_mtx, PWAIT | PCATCH, "wait", 0);
1271 * Make process 'parent' the new parent of process 'child'.
1272 * Must be called with an exclusive hold of proctree lock.
1275 proc_reparent(struct proc *child, struct proc *parent)
1278 sx_assert(&proctree_lock, SX_XLOCKED);
1279 PROC_LOCK_ASSERT(child, MA_OWNED);
1280 if (child->p_pptr == parent)
1283 PROC_LOCK(child->p_pptr);
1284 sigqueue_take(child->p_ksi);
1285 PROC_UNLOCK(child->p_pptr);
1286 LIST_REMOVE(child, p_sibling);
1287 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
1289 clear_orphan(child);
1290 if (child->p_flag & P_TRACED) {
1291 if (LIST_EMPTY(&child->p_pptr->p_orphans)) {
1292 child->p_treeflag |= P_TREE_FIRST_ORPHAN;
1293 LIST_INSERT_HEAD(&child->p_pptr->p_orphans, child,
1296 LIST_INSERT_AFTER(LIST_FIRST(&child->p_pptr->p_orphans),
1299 child->p_treeflag |= P_TREE_ORPHANED;
1302 child->p_pptr = parent;