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_kdtrace.h"
42 #include "opt_ktrace.h"
43 #include "opt_procdesc.h"
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/sysproto.h>
48 #include <sys/capability.h>
49 #include <sys/eventhandler.h>
50 #include <sys/kernel.h>
51 #include <sys/malloc.h>
53 #include <sys/mutex.h>
55 #include <sys/procdesc.h>
56 #include <sys/pioctl.h>
60 #include <sys/vmmeter.h>
61 #include <sys/vnode.h>
62 #include <sys/racct.h>
63 #include <sys/resourcevar.h>
65 #include <sys/signalvar.h>
66 #include <sys/sched.h>
68 #include <sys/syscallsubr.h>
69 #include <sys/syslog.h>
70 #include <sys/ptrace.h>
71 #include <sys/acct.h> /* for acct_process() function prototype */
72 #include <sys/filedesc.h>
77 #include <sys/ktrace.h>
80 #include <security/audit/audit.h>
81 #include <security/mac/mac_framework.h>
84 #include <vm/vm_extern.h>
85 #include <vm/vm_param.h>
87 #include <vm/vm_map.h>
88 #include <vm/vm_page.h>
92 #include <sys/dtrace_bsd.h>
93 dtrace_execexit_func_t dtrace_fasttrap_exit;
96 SDT_PROVIDER_DECLARE(proc);
97 SDT_PROBE_DEFINE1(proc, kernel, , exit, "int");
99 /* Hook for NFS teardown procedure. */
100 void (*nlminfo_release_p)(struct proc *p);
103 proc_realparent(struct proc *child)
105 struct proc *p, *parent;
107 sx_assert(&proctree_lock, SX_LOCKED);
108 if ((child->p_treeflag & P_TREE_ORPHANED) == 0) {
109 return (child->p_pptr->p_pid == child->p_oppid ?
110 child->p_pptr : initproc);
112 for (p = child; (p->p_treeflag & P_TREE_FIRST_ORPHAN) == 0;) {
113 /* Cannot use LIST_PREV(), since the list head is not known. */
114 p = __containerof(p->p_orphan.le_prev, struct proc,
116 KASSERT((p->p_treeflag & P_TREE_ORPHANED) != 0,
117 ("missing P_ORPHAN %p", p));
119 parent = __containerof(p->p_orphan.le_prev, struct proc,
125 clear_orphan(struct proc *p)
129 sx_assert(&proctree_lock, SA_XLOCKED);
130 if ((p->p_treeflag & P_TREE_ORPHANED) == 0)
132 if ((p->p_treeflag & P_TREE_FIRST_ORPHAN) != 0) {
133 p1 = LIST_NEXT(p, p_orphan);
135 p1->p_treeflag |= P_TREE_FIRST_ORPHAN;
136 p->p_treeflag &= ~P_TREE_FIRST_ORPHAN;
138 LIST_REMOVE(p, p_orphan);
139 p->p_treeflag &= ~P_TREE_ORPHANED;
143 * exit -- death of process.
146 sys_sys_exit(struct thread *td, struct sys_exit_args *uap)
149 exit1(td, W_EXITCODE(uap->rval, 0));
154 * Exit: deallocate address space and other resources, change proc state to
155 * zombie, and unlink proc from allproc and parent's lists. Save exit status
156 * and rusage for wait(). Check for child processes and orphan them.
159 exit1(struct thread *td, int rv)
161 struct proc *p, *nq, *q;
162 struct vnode *ttyvp = NULL;
164 mtx_assert(&Giant, MA_NOTOWNED);
168 * XXX in case we're rebooting we just let init die in order to
169 * work around an unsolved stack overflow seen very late during
170 * shutdown on sparc64 when the gmirror worker process exists.
172 if (p == initproc && rebooting == 0) {
173 printf("init died (signal %d, exit %d)\n",
174 WTERMSIG(rv), WEXITSTATUS(rv));
175 panic("Going nowhere without my init!");
179 * MUST abort all other threads before proceeding past here.
182 while (p->p_flag & P_HADTHREADS) {
184 * First check if some other thread got here before us.
185 * If so, act appropriately: exit or suspend.
187 thread_suspend_check(0);
190 * Kill off the other threads. This requires
191 * some co-operation from other parts of the kernel
192 * so it may not be instantaneous. With this state set
193 * any thread entering the kernel from userspace will
194 * thread_exit() in trap(). Any thread attempting to
195 * sleep will return immediately with EINTR or EWOULDBLOCK
196 * which will hopefully force them to back out to userland
197 * freeing resources as they go. Any thread attempting
198 * to return to userland will thread_exit() from userret().
199 * thread_exit() will unsuspend us when the last of the
200 * other threads exits.
201 * If there is already a thread singler after resumption,
202 * calling thread_single will fail; in that case, we just
203 * re-check all suspension request, the thread should
204 * either be suspended there or exit.
206 if (!thread_single(SINGLE_EXIT))
210 * All other activity in this process is now stopped.
211 * Threading support has been turned off.
214 KASSERT(p->p_numthreads == 1,
215 ("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
216 racct_sub(p, RACCT_NTHR, 1);
218 * Wakeup anyone in procfs' PIOCWAIT. They should have a hold
219 * on our vmspace, so we should block below until they have
220 * released their reference to us. Note that if they have
221 * requested S_EXIT stops we will block here until they ack
224 _STOPEVENT(p, S_EXIT, rv);
227 * Ignore any pending request to stop due to a stop signal.
228 * Once P_WEXIT is set, future requests will be ignored as
231 p->p_flag &= ~P_STOPPED_SIG;
232 KASSERT(!P_SHOULDSTOP(p), ("exiting process is stopped"));
235 * Note that we are exiting and do another wakeup of anyone in
236 * PIOCWAIT in case they aren't listening for S_EXIT stops or
237 * decided to wait again after we told them we are exiting.
239 p->p_flag |= P_WEXIT;
243 * Wait for any processes that have a hold on our vmspace to
244 * release their reference.
246 while (p->p_lock > 0)
247 msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0);
249 p->p_xstat = rv; /* Let event handler change exit status */
251 /* Drain the limit callout while we don't have the proc locked */
252 callout_drain(&p->p_limco);
256 * The Sun BSM exit token contains two components: an exit status as
257 * passed to exit(), and a return value to indicate what sort of exit
258 * it was. The exit status is WEXITSTATUS(rv), but it's not clear
259 * what the return value is.
261 AUDIT_ARG_EXIT(WEXITSTATUS(rv), 0);
262 AUDIT_SYSCALL_EXIT(0, td);
265 /* Are we a task leader? */
266 if (p == p->p_leader) {
267 mtx_lock(&ppeers_lock);
271 kern_psignal(q, SIGKILL);
275 while (p->p_peers != NULL)
276 msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
277 mtx_unlock(&ppeers_lock);
281 * Check if any loadable modules need anything done at process exit.
282 * E.g. SYSV IPC stuff
283 * XXX what if one of these generates an error?
285 EVENTHANDLER_INVOKE(process_exit, p);
288 * If parent is waiting for us to exit or exec,
289 * P_PPWAIT is set; we will wakeup the parent below.
292 rv = p->p_xstat; /* Event handler could change exit status */
294 p->p_flag &= ~(P_TRACED | P_PPWAIT | P_PPTRACE);
297 * Stop the real interval timer. If the handler is currently
298 * executing, prevent it from rearming itself and let it finish.
300 if (timevalisset(&p->p_realtimer.it_value) &&
301 callout_stop(&p->p_itcallout) == 0) {
302 timevalclear(&p->p_realtimer.it_interval);
303 msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
304 KASSERT(!timevalisset(&p->p_realtimer.it_value),
305 ("realtime timer is still armed"));
310 * Reset any sigio structures pointing to us as a result of
311 * F_SETOWN with our pid.
313 funsetownlst(&p->p_sigiolst);
316 * If this process has an nlminfo data area (for lockd), release it
318 if (nlminfo_release_p != NULL && p->p_nlminfo != NULL)
319 (*nlminfo_release_p)(p);
322 * Close open files and release open-file table.
328 * If this thread tickled GEOM, we need to wait for the giggling to
329 * stop before we return to userland
331 if (td->td_pflags & TDP_GEOM)
335 * Remove ourself from our leader's peer list and wake our leader.
337 mtx_lock(&ppeers_lock);
338 if (p->p_leader->p_peers) {
340 while (q->p_peers != p)
342 q->p_peers = p->p_peers;
345 mtx_unlock(&ppeers_lock);
349 sx_xlock(&proctree_lock);
350 if (SESS_LEADER(p)) {
351 struct session *sp = p->p_session;
355 * s_ttyp is not zero'd; we use this to indicate that
356 * the session once had a controlling terminal. (for
357 * logging and informational purposes)
368 * Signal foreground pgrp and revoke access to
369 * controlling terminal if it has not been revoked
372 * Because the TTY may have been revoked in the mean
373 * time and could already have a new session associated
374 * with it, make sure we don't send a SIGHUP to a
375 * foreground process group that does not belong to this
381 if (tp->t_session == sp)
382 tty_signal_pgrp(tp, SIGHUP);
387 sx_xunlock(&proctree_lock);
388 if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
389 VOP_REVOKE(ttyvp, REVOKEALL);
390 VOP_UNLOCK(ttyvp, 0);
392 sx_xlock(&proctree_lock);
395 fixjobc(p, p->p_pgrp, 0);
396 sx_xunlock(&proctree_lock);
397 (void)acct_process(td);
399 /* Release the TTY now we've unlocked everything. */
406 * Release reference to text vnode
408 if (p->p_textvp != NULL) {
414 * Release our limits structure.
416 lim_free(p->p_limit);
422 * Remove proc from allproc queue and pidhash chain.
423 * Place onto zombproc. Unlink from parent's child list.
425 sx_xlock(&allproc_lock);
426 LIST_REMOVE(p, p_list);
427 LIST_INSERT_HEAD(&zombproc, p, p_list);
428 LIST_REMOVE(p, p_hash);
429 sx_xunlock(&allproc_lock);
432 * Call machine-dependent code to release any
433 * machine-dependent resources other than the address space.
434 * The address space is released by "vmspace_exitfree(p)" in
439 WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
442 * Reparent all of our children to init.
444 sx_xlock(&proctree_lock);
445 q = LIST_FIRST(&p->p_children);
446 if (q != NULL) /* only need this if any child is S_ZOMB */
448 for (; q != NULL; q = nq) {
449 nq = LIST_NEXT(q, p_sibling);
451 proc_reparent(q, initproc);
452 q->p_sigparent = SIGCHLD;
454 * Traced processes are killed
455 * since their existence means someone is screwing up.
457 if (q->p_flag & P_TRACED) {
461 * Since q was found on our children list, the
462 * proc_reparent() call moved q to the orphan
463 * list due to present P_TRACED flag. Clear
464 * orphan link for q now while q is locked.
467 q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
468 FOREACH_THREAD_IN_PROC(q, temp)
469 temp->td_dbgflags &= ~TDB_SUSPEND;
470 kern_psignal(q, SIGKILL);
476 * Also get rid of our orphans.
478 while ((q = LIST_FIRST(&p->p_orphans)) != NULL) {
484 /* Save exit status. */
488 /* Tell the prison that we are gone. */
489 prison_proc_free(p->p_ucred->cr_prison);
493 * Tell the DTrace fasttrap provider about the exit if it
494 * has declared an interest.
496 if (dtrace_fasttrap_exit)
497 dtrace_fasttrap_exit(p);
501 * Notify interested parties of our demise.
503 KNOTE_LOCKED(&p->p_klist, NOTE_EXIT);
506 int reason = CLD_EXITED;
509 else if (WIFSIGNALED(rv))
511 SDT_PROBE(proc, kernel, , exit, reason, 0, 0, 0, 0);
515 * Just delete all entries in the p_klist. At this point we won't
516 * report any more events, and there are nasty race conditions that
517 * can beat us if we don't.
519 knlist_clear(&p->p_klist, 1);
522 * If this is a process with a descriptor, we may not need to deliver
523 * a signal to the parent. proctree_lock is held over
524 * procdesc_exit() to serialize concurrent calls to close() and
528 if (p->p_procdesc == NULL || procdesc_exit(p)) {
531 * Notify parent that we're gone. If parent has the
532 * PS_NOCLDWAIT flag set, or if the handler is set to SIG_IGN,
533 * notify process 1 instead (and hope it will handle this
536 PROC_LOCK(p->p_pptr);
537 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
538 if (p->p_pptr->p_sigacts->ps_flag &
539 (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
542 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
545 proc_reparent(p, initproc);
546 p->p_sigparent = SIGCHLD;
547 PROC_LOCK(p->p_pptr);
550 * Notify parent, so in case he was wait(2)ing or
551 * executing waitpid(2) with our pid, he will
556 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
558 if (p->p_pptr == initproc)
559 kern_psignal(p->p_pptr, SIGCHLD);
560 else if (p->p_sigparent != 0) {
561 if (p->p_sigparent == SIGCHLD)
563 else /* LINUX thread */
564 kern_psignal(p->p_pptr, p->p_sigparent);
568 PROC_LOCK(p->p_pptr);
570 sx_xunlock(&proctree_lock);
573 * The state PRS_ZOMBIE prevents other proesses from sending
574 * signal to the process, to avoid memory leak, we free memory
575 * for signal queue at the time when the state is set.
577 sigqueue_flush(&p->p_sigqueue);
578 sigqueue_flush(&td->td_sigqueue);
581 * We have to wait until after acquiring all locks before
582 * changing p_state. We need to avoid all possible context
583 * switches (including ones from blocking on a mutex) while
584 * marked as a zombie. We also have to set the zombie state
585 * before we release the parent process' proc lock to avoid
586 * a lost wakeup. So, we first call wakeup, then we grab the
587 * sched lock, update the state, and release the parent process'
591 cv_broadcast(&p->p_pwait);
592 sched_exit(p->p_pptr, td);
594 p->p_state = PRS_ZOMBIE;
595 PROC_UNLOCK(p->p_pptr);
598 * Hopefully no one will try to deliver a signal to the process this
601 knlist_destroy(&p->p_klist);
604 * Save our children's rusage information in our exit rusage.
606 ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
609 * Make sure the scheduler takes this thread out of its tables etc.
610 * This will also release this thread's reference to the ucred.
611 * Other thread parts to release include pcb bits and such.
617 #ifndef _SYS_SYSPROTO_H_
626 sys_abort2(struct thread *td, struct abort2_args *uap)
628 struct proc *p = td->td_proc;
634 * Do it right now so we can log either proper call of abort2(), or
635 * note, that invalid argument was passed. 512 is big enough to
636 * handle 16 arguments' descriptions with additional comments.
638 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN);
640 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ",
641 p->p_comm, p->p_pid, td->td_ucred->cr_uid);
643 * Since we can't return from abort2(), send SIGKILL in cases, where
644 * abort2() was called improperly
647 /* Prevent from DoSes from user-space. */
648 if (uap->nargs < 0 || uap->nargs > 16)
650 if (uap->nargs > 0) {
651 if (uap->args == NULL)
653 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *));
658 * Limit size of 'reason' string to 128. Will fit even when
659 * maximal number of arguments was chosen to be logged.
661 if (uap->why != NULL) {
662 error = sbuf_copyin(sb, uap->why, 128);
666 sbuf_printf(sb, "(null)");
668 if (uap->nargs > 0) {
669 sbuf_printf(sb, "(");
670 for (i = 0;i < uap->nargs; i++)
671 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]);
672 sbuf_printf(sb, ")");
675 * Final stage: arguments were proper, string has been
676 * successfully copied from userspace, and copying pointers
677 * from user-space succeed.
681 if (sig == SIGKILL) {
683 sbuf_printf(sb, " (Reason text inaccessible)");
687 log(LOG_INFO, "%s", sbuf_data(sb));
689 exit1(td, W_EXITCODE(0, sig));
696 * The dirty work is handled by kern_wait().
699 owait(struct thread *td, struct owait_args *uap __unused)
703 error = kern_wait(td, WAIT_ANY, &status, 0, NULL);
705 td->td_retval[1] = status;
708 #endif /* COMPAT_43 */
711 * The dirty work is handled by kern_wait().
714 sys_wait4(struct thread *td, struct wait4_args *uap)
716 struct rusage ru, *rup;
719 if (uap->rusage != NULL)
723 error = kern_wait(td, uap->pid, &status, uap->options, rup);
724 if (uap->status != NULL && error == 0)
725 error = copyout(&status, uap->status, sizeof(status));
726 if (uap->rusage != NULL && error == 0)
727 error = copyout(&ru, uap->rusage, sizeof(struct rusage));
732 sys_wait6(struct thread *td, struct wait6_args *uap)
734 struct __wrusage wru, *wrup;
740 idtype = uap->idtype;
743 if (uap->wrusage != NULL)
748 if (uap->info != NULL) {
750 bzero(sip, sizeof(*sip));
755 * We expect all callers of wait6() to know about WEXITED and
758 error = kern_wait6(td, idtype, id, &status, uap->options, wrup, sip);
760 if (uap->status != NULL && error == 0)
761 error = copyout(&status, uap->status, sizeof(status));
762 if (uap->wrusage != NULL && error == 0)
763 error = copyout(&wru, uap->wrusage, sizeof(wru));
764 if (uap->info != NULL && error == 0)
765 error = copyout(&si, uap->info, sizeof(si));
770 * Reap the remains of a zombie process and optionally return status and
771 * rusage. Asserts and will release both the proctree_lock and the process
772 * lock as part of its work.
775 proc_reap(struct thread *td, struct proc *p, int *status, int options)
779 sx_assert(&proctree_lock, SA_XLOCKED);
780 PROC_LOCK_ASSERT(p, MA_OWNED);
781 PROC_SLOCK_ASSERT(p, MA_OWNED);
782 KASSERT(p->p_state == PRS_ZOMBIE, ("proc_reap: !PRS_ZOMBIE"));
787 td->td_retval[0] = p->p_pid;
789 *status = p->p_xstat; /* convert to int */
790 if (options & WNOWAIT) {
792 * Only poll, returning the status. Caller does not wish to
793 * release the proc struct just yet.
796 sx_xunlock(&proctree_lock);
801 sigqueue_take(p->p_ksi);
806 * If we got the child via a ptrace 'attach', we need to give it back
809 if (p->p_oppid != 0) {
810 t = proc_realparent(p);
816 pksignal(t, SIGCHLD, p->p_ksi);
818 cv_broadcast(&p->p_pwait);
820 sx_xunlock(&proctree_lock);
825 * Remove other references to this process to ensure we have an
826 * exclusive reference.
828 sx_xlock(&allproc_lock);
829 LIST_REMOVE(p, p_list); /* off zombproc */
830 sx_xunlock(&allproc_lock);
831 LIST_REMOVE(p, p_sibling);
837 if (p->p_procdesc != NULL)
840 sx_xunlock(&proctree_lock);
843 * As a side effect of this lock, we know that all other writes to
844 * this proc are visible now, so no more locking is needed for p.
847 p->p_xstat = 0; /* XXX: why? */
850 ruadd(&q->p_stats->p_cru, &q->p_crux, &p->p_ru, &p->p_rux);
854 * Decrement the count of procs running with this uid.
856 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
859 * Destroy resource accounting information associated with the process.
863 racct_sub(p, RACCT_NPROC, 1);
869 * Free credentials, arguments, and sigacts.
873 pargs_drop(p->p_args);
875 sigacts_free(p->p_sigacts);
879 * Do any thread-system specific cleanups.
884 * Give vm and machine-dependent layer a chance to free anything that
885 * cpu_exit couldn't release while still running in process context.
891 KASSERT(FIRST_THREAD_IN_PROC(p),
892 ("proc_reap: no residual thread!"));
893 uma_zfree(proc_zone, p);
894 sx_xlock(&allproc_lock);
896 sx_xunlock(&allproc_lock);
900 proc_to_reap(struct thread *td, struct proc *p, idtype_t idtype, id_t id,
901 int *status, int options, struct __wrusage *wrusage, siginfo_t *siginfo)
906 sx_assert(&proctree_lock, SA_XLOCKED);
915 if (p->p_pid != (pid_t)id) {
921 if (p->p_pgid != (pid_t)id) {
927 if (p->p_session->s_sid != (pid_t)id) {
933 if (p->p_ucred->cr_uid != (uid_t)id) {
939 if (p->p_ucred->cr_gid != (gid_t)id) {
945 if (p->p_ucred->cr_prison->pr_id != (int)id) {
951 * It seems that the thread structures get zeroed out
952 * at process exit. This makes it impossible to
953 * support P_SETID, P_CID or P_CPUID.
960 if (p_canwait(td, p)) {
965 if (((options & WEXITED) == 0) && (p->p_state == PRS_ZOMBIE)) {
971 * This special case handles a kthread spawned by linux_clone
972 * (see linux_misc.c). The linux_wait4 and linux_waitpid
973 * functions need to be able to distinguish between waiting
974 * on a process and waiting on a thread. It is a thread if
975 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
976 * signifies we want to wait for threads and not processes.
978 if ((p->p_sigparent != SIGCHLD) ^
979 ((options & WLINUXCLONE) != 0)) {
986 if (siginfo != NULL) {
987 bzero(siginfo, sizeof(*siginfo));
988 siginfo->si_errno = 0;
991 * SUSv4 requires that the si_signo value is always
992 * SIGCHLD. Obey it despite the rfork(2) interface
993 * allows to request other signal for child exit
996 siginfo->si_signo = SIGCHLD;
999 * This is still a rough estimate. We will fix the
1000 * cases TRAPPED, STOPPED, and CONTINUED later.
1002 if (WCOREDUMP(p->p_xstat)) {
1003 siginfo->si_code = CLD_DUMPED;
1004 siginfo->si_status = WTERMSIG(p->p_xstat);
1005 } else if (WIFSIGNALED(p->p_xstat)) {
1006 siginfo->si_code = CLD_KILLED;
1007 siginfo->si_status = WTERMSIG(p->p_xstat);
1009 siginfo->si_code = CLD_EXITED;
1010 siginfo->si_status = WEXITSTATUS(p->p_xstat);
1013 siginfo->si_pid = p->p_pid;
1014 siginfo->si_uid = p->p_ucred->cr_uid;
1017 * The si_addr field would be useful additional
1018 * detail, but apparently the PC value may be lost
1019 * when we reach this point. bzero() above sets
1020 * siginfo->si_addr to NULL.
1025 * There should be no reason to limit resources usage info to
1026 * exited processes only. A snapshot about any resources used
1027 * by a stopped process may be exactly what is needed.
1029 if (wrusage != NULL) {
1030 rup = &wrusage->wru_self;
1032 calcru(p, &rup->ru_utime, &rup->ru_stime);
1034 rup = &wrusage->wru_children;
1035 *rup = p->p_stats->p_cru;
1036 calccru(p, &rup->ru_utime, &rup->ru_stime);
1039 if (p->p_state == PRS_ZOMBIE) {
1040 proc_reap(td, p, status, options);
1049 kern_wait(struct thread *td, pid_t pid, int *status, int options,
1050 struct rusage *rusage)
1052 struct __wrusage wru, *wrup;
1058 * Translate the special pid values into the (idtype, pid)
1059 * pair for kern_wait6. The WAIT_MYPGRP case is handled by
1060 * kern_wait6() on its own.
1062 if (pid == WAIT_ANY) {
1065 } else if (pid < 0) {
1079 * For backward compatibility we implicitly add flags WEXITED
1080 * and WTRAPPED here.
1082 options |= WEXITED | WTRAPPED;
1083 ret = kern_wait6(td, idtype, id, status, options, wrup, NULL);
1085 *rusage = wru.wru_self;
1090 kern_wait6(struct thread *td, idtype_t idtype, id_t id, int *status,
1091 int options, struct __wrusage *wrusage, siginfo_t *siginfo)
1094 int error, nfound, ret;
1096 AUDIT_ARG_VALUE((int)idtype); /* XXX - This is likely wrong! */
1097 AUDIT_ARG_PID((pid_t)id); /* XXX - This may be wrong! */
1098 AUDIT_ARG_VALUE(options);
1102 if ((pid_t)id == WAIT_MYPGRP && (idtype == P_PID || idtype == P_PGID)) {
1104 id = (id_t)q->p_pgid;
1109 /* If we don't know the option, just return. */
1110 if ((options & ~(WUNTRACED | WNOHANG | WCONTINUED | WNOWAIT |
1111 WEXITED | WTRAPPED | WLINUXCLONE)) != 0)
1113 if ((options & (WEXITED | WUNTRACED | WCONTINUED | WTRAPPED)) == 0) {
1115 * We will be unable to find any matching processes,
1116 * because there are no known events to look for.
1117 * Prefer to return error instead of blocking
1124 if (q->p_flag & P_STATCHILD) {
1126 q->p_flag &= ~P_STATCHILD;
1130 sx_xlock(&proctree_lock);
1131 LIST_FOREACH(p, &q->p_children, p_sibling) {
1132 ret = proc_to_reap(td, p, idtype, id, status, options,
1144 if ((options & WTRAPPED) != 0 &&
1145 (p->p_flag & P_TRACED) != 0 &&
1146 (p->p_flag & (P_STOPPED_TRACE | P_STOPPED_SIG)) != 0 &&
1147 (p->p_suspcount == p->p_numthreads) &&
1148 ((p->p_flag & P_WAITED) == 0)) {
1150 if ((options & WNOWAIT) == 0)
1151 p->p_flag |= P_WAITED;
1152 sx_xunlock(&proctree_lock);
1153 td->td_retval[0] = p->p_pid;
1156 *status = W_STOPCODE(p->p_xstat);
1157 if (siginfo != NULL) {
1158 siginfo->si_status = p->p_xstat;
1159 siginfo->si_code = CLD_TRAPPED;
1161 if ((options & WNOWAIT) == 0) {
1163 sigqueue_take(p->p_ksi);
1170 if ((options & WUNTRACED) != 0 &&
1171 (p->p_flag & P_STOPPED_SIG) != 0 &&
1172 (p->p_suspcount == p->p_numthreads) &&
1173 ((p->p_flag & P_WAITED) == 0)) {
1175 if ((options & WNOWAIT) == 0)
1176 p->p_flag |= P_WAITED;
1177 sx_xunlock(&proctree_lock);
1178 td->td_retval[0] = p->p_pid;
1181 *status = W_STOPCODE(p->p_xstat);
1182 if (siginfo != NULL) {
1183 siginfo->si_status = p->p_xstat;
1184 siginfo->si_code = CLD_STOPPED;
1186 if ((options & WNOWAIT) == 0) {
1188 sigqueue_take(p->p_ksi);
1196 if ((options & WCONTINUED) != 0 &&
1197 (p->p_flag & P_CONTINUED) != 0) {
1198 sx_xunlock(&proctree_lock);
1199 td->td_retval[0] = p->p_pid;
1200 if ((options & WNOWAIT) == 0) {
1201 p->p_flag &= ~P_CONTINUED;
1203 sigqueue_take(p->p_ksi);
1210 if (siginfo != NULL) {
1211 siginfo->si_status = SIGCONT;
1212 siginfo->si_code = CLD_CONTINUED;
1220 * Look in the orphans list too, to allow the parent to
1221 * collect it's child exit status even if child is being
1224 * Debugger detaches from the parent upon successful
1225 * switch-over from parent to child. At this point due to
1226 * re-parenting the parent loses the child to debugger and a
1227 * wait4(2) call would report that it has no children to wait
1228 * for. By maintaining a list of orphans we allow the parent
1229 * to successfully wait until the child becomes a zombie.
1231 LIST_FOREACH(p, &q->p_orphans, p_orphan) {
1232 ret = proc_to_reap(td, p, idtype, id, status, options,
1242 sx_xunlock(&proctree_lock);
1245 if (options & WNOHANG) {
1246 sx_xunlock(&proctree_lock);
1247 td->td_retval[0] = 0;
1251 sx_xunlock(&proctree_lock);
1252 if (q->p_flag & P_STATCHILD) {
1253 q->p_flag &= ~P_STATCHILD;
1256 error = msleep(q, &q->p_mtx, PWAIT | PCATCH, "wait", 0);
1264 * Make process 'parent' the new parent of process 'child'.
1265 * Must be called with an exclusive hold of proctree lock.
1268 proc_reparent(struct proc *child, struct proc *parent)
1271 sx_assert(&proctree_lock, SX_XLOCKED);
1272 PROC_LOCK_ASSERT(child, MA_OWNED);
1273 if (child->p_pptr == parent)
1276 PROC_LOCK(child->p_pptr);
1277 sigqueue_take(child->p_ksi);
1278 PROC_UNLOCK(child->p_pptr);
1279 LIST_REMOVE(child, p_sibling);
1280 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
1282 clear_orphan(child);
1283 if (child->p_flag & P_TRACED) {
1284 if (LIST_EMPTY(&child->p_pptr->p_orphans)) {
1285 child->p_treeflag |= P_TREE_FIRST_ORPHAN;
1286 LIST_INSERT_HEAD(&child->p_pptr->p_orphans, child,
1289 LIST_INSERT_AFTER(LIST_FIRST(&child->p_pptr->p_orphans),
1292 child->p_treeflag |= P_TREE_ORPHANED;
1295 child->p_pptr = parent;