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 return (child->p_pptr->p_pid == child->p_oppid ?
108 child->p_pptr : initproc);
110 for (p = child; (p->p_treeflag & P_TREE_FIRST_ORPHAN) == 0;) {
111 /* Cannot use LIST_PREV(), since the list head is not known. */
112 p = __containerof(p->p_orphan.le_prev, struct proc,
114 KASSERT((p->p_treeflag & P_TREE_ORPHANED) != 0,
115 ("missing P_ORPHAN %p", p));
117 parent = __containerof(p->p_orphan.le_prev, struct proc,
123 clear_orphan(struct proc *p)
127 sx_assert(&proctree_lock, SA_XLOCKED);
128 if ((p->p_treeflag & P_TREE_ORPHANED) == 0)
130 if ((p->p_treeflag & P_TREE_FIRST_ORPHAN) != 0) {
131 p1 = LIST_NEXT(p, p_orphan);
133 p1->p_treeflag |= P_TREE_FIRST_ORPHAN;
134 p->p_treeflag &= ~P_TREE_FIRST_ORPHAN;
136 LIST_REMOVE(p, p_orphan);
137 p->p_treeflag &= ~P_TREE_ORPHANED;
141 * exit -- death of process.
144 sys_sys_exit(struct thread *td, struct sys_exit_args *uap)
147 exit1(td, W_EXITCODE(uap->rval, 0));
152 * Exit: deallocate address space and other resources, change proc state to
153 * zombie, and unlink proc from allproc and parent's lists. Save exit status
154 * and rusage for wait(). Check for child processes and orphan them.
157 exit1(struct thread *td, int rv)
159 struct proc *p, *nq, *q;
160 struct vnode *ttyvp = NULL;
162 mtx_assert(&Giant, MA_NOTOWNED);
166 * XXX in case we're rebooting we just let init die in order to
167 * work around an unsolved stack overflow seen very late during
168 * shutdown on sparc64 when the gmirror worker process exists.
170 if (p == initproc && rebooting == 0) {
171 printf("init died (signal %d, exit %d)\n",
172 WTERMSIG(rv), WEXITSTATUS(rv));
173 panic("Going nowhere without my init!");
177 * MUST abort all other threads before proceeding past here.
180 while (p->p_flag & P_HADTHREADS) {
182 * First check if some other thread got here before us.
183 * If so, act appropriately: exit or suspend.
185 thread_suspend_check(0);
188 * Kill off the other threads. This requires
189 * some co-operation from other parts of the kernel
190 * so it may not be instantaneous. With this state set
191 * any thread entering the kernel from userspace will
192 * thread_exit() in trap(). Any thread attempting to
193 * sleep will return immediately with EINTR or EWOULDBLOCK
194 * which will hopefully force them to back out to userland
195 * freeing resources as they go. Any thread attempting
196 * to return to userland will thread_exit() from userret().
197 * thread_exit() will unsuspend us when the last of the
198 * other threads exits.
199 * If there is already a thread singler after resumption,
200 * calling thread_single will fail; in that case, we just
201 * re-check all suspension request, the thread should
202 * either be suspended there or exit.
204 if (!thread_single(SINGLE_EXIT))
208 * All other activity in this process is now stopped.
209 * Threading support has been turned off.
212 KASSERT(p->p_numthreads == 1,
213 ("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
214 racct_sub(p, RACCT_NTHR, 1);
216 * Wakeup anyone in procfs' PIOCWAIT. They should have a hold
217 * on our vmspace, so we should block below until they have
218 * released their reference to us. Note that if they have
219 * requested S_EXIT stops we will block here until they ack
222 _STOPEVENT(p, S_EXIT, rv);
225 * Ignore any pending request to stop due to a stop signal.
226 * Once P_WEXIT is set, future requests will be ignored as
229 p->p_flag &= ~P_STOPPED_SIG;
230 KASSERT(!P_SHOULDSTOP(p), ("exiting process is stopped"));
233 * Note that we are exiting and do another wakeup of anyone in
234 * PIOCWAIT in case they aren't listening for S_EXIT stops or
235 * decided to wait again after we told them we are exiting.
237 p->p_flag |= P_WEXIT;
241 * Wait for any processes that have a hold on our vmspace to
242 * release their reference.
244 while (p->p_lock > 0)
245 msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0);
247 p->p_xstat = rv; /* Let event handler change exit status */
249 /* Drain the limit callout while we don't have the proc locked */
250 callout_drain(&p->p_limco);
254 * The Sun BSM exit token contains two components: an exit status as
255 * passed to exit(), and a return value to indicate what sort of exit
256 * it was. The exit status is WEXITSTATUS(rv), but it's not clear
257 * what the return value is.
259 AUDIT_ARG_EXIT(WEXITSTATUS(rv), 0);
260 AUDIT_SYSCALL_EXIT(0, td);
263 /* Are we a task leader? */
264 if (p == p->p_leader) {
265 mtx_lock(&ppeers_lock);
269 kern_psignal(q, SIGKILL);
273 while (p->p_peers != NULL)
274 msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
275 mtx_unlock(&ppeers_lock);
279 * Check if any loadable modules need anything done at process exit.
280 * E.g. SYSV IPC stuff
281 * XXX what if one of these generates an error?
283 EVENTHANDLER_INVOKE(process_exit, p);
286 * If parent is waiting for us to exit or exec,
287 * P_PPWAIT is set; we will wakeup the parent below.
290 rv = p->p_xstat; /* Event handler could change exit status */
292 p->p_flag &= ~(P_TRACED | P_PPWAIT | P_PPTRACE);
295 * Stop the real interval timer. If the handler is currently
296 * executing, prevent it from rearming itself and let it finish.
298 if (timevalisset(&p->p_realtimer.it_value) &&
299 callout_stop(&p->p_itcallout) == 0) {
300 timevalclear(&p->p_realtimer.it_interval);
301 msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
302 KASSERT(!timevalisset(&p->p_realtimer.it_value),
303 ("realtime timer is still armed"));
308 * Reset any sigio structures pointing to us as a result of
309 * F_SETOWN with our pid.
311 funsetownlst(&p->p_sigiolst);
314 * If this process has an nlminfo data area (for lockd), release it
316 if (nlminfo_release_p != NULL && p->p_nlminfo != NULL)
317 (*nlminfo_release_p)(p);
320 * Close open files and release open-file table.
326 * If this thread tickled GEOM, we need to wait for the giggling to
327 * stop before we return to userland
329 if (td->td_pflags & TDP_GEOM)
333 * Remove ourself from our leader's peer list and wake our leader.
335 mtx_lock(&ppeers_lock);
336 if (p->p_leader->p_peers) {
338 while (q->p_peers != p)
340 q->p_peers = p->p_peers;
343 mtx_unlock(&ppeers_lock);
347 sx_xlock(&proctree_lock);
348 if (SESS_LEADER(p)) {
349 struct session *sp = p->p_session;
353 * s_ttyp is not zero'd; we use this to indicate that
354 * the session once had a controlling terminal. (for
355 * logging and informational purposes)
366 * Signal foreground pgrp and revoke access to
367 * controlling terminal if it has not been revoked
370 * Because the TTY may have been revoked in the mean
371 * time and could already have a new session associated
372 * with it, make sure we don't send a SIGHUP to a
373 * foreground process group that does not belong to this
379 if (tp->t_session == sp)
380 tty_signal_pgrp(tp, SIGHUP);
385 sx_xunlock(&proctree_lock);
386 if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
387 VOP_REVOKE(ttyvp, REVOKEALL);
388 VOP_UNLOCK(ttyvp, 0);
390 sx_xlock(&proctree_lock);
393 fixjobc(p, p->p_pgrp, 0);
394 sx_xunlock(&proctree_lock);
395 (void)acct_process(td);
397 /* Release the TTY now we've unlocked everything. */
404 * Release reference to text vnode
406 if (p->p_textvp != NULL) {
412 * Release our limits structure.
414 lim_free(p->p_limit);
420 * Remove proc from allproc queue and pidhash chain.
421 * Place onto zombproc. Unlink from parent's child list.
423 sx_xlock(&allproc_lock);
424 LIST_REMOVE(p, p_list);
425 LIST_INSERT_HEAD(&zombproc, p, p_list);
426 LIST_REMOVE(p, p_hash);
427 sx_xunlock(&allproc_lock);
430 * Call machine-dependent code to release any
431 * machine-dependent resources other than the address space.
432 * The address space is released by "vmspace_exitfree(p)" in
437 WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
440 * Reparent all of our children to init.
442 sx_xlock(&proctree_lock);
443 q = LIST_FIRST(&p->p_children);
444 if (q != NULL) /* only need this if any child is S_ZOMB */
446 for (; q != NULL; q = nq) {
447 nq = LIST_NEXT(q, p_sibling);
449 proc_reparent(q, initproc);
450 q->p_sigparent = SIGCHLD;
452 * Traced processes are killed
453 * since their existence means someone is screwing up.
455 if (q->p_flag & P_TRACED) {
459 * Since q was found on our children list, the
460 * proc_reparent() call moved q to the orphan
461 * list due to present P_TRACED flag. Clear
462 * orphan link for q now while q is locked.
465 q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
466 FOREACH_THREAD_IN_PROC(q, temp)
467 temp->td_dbgflags &= ~TDB_SUSPEND;
468 kern_psignal(q, SIGKILL);
474 * Also get rid of our orphans.
476 while ((q = LIST_FIRST(&p->p_orphans)) != NULL) {
482 /* Save exit status. */
486 /* Tell the prison that we are gone. */
487 prison_proc_free(p->p_ucred->cr_prison);
491 * Tell the DTrace fasttrap provider about the exit if it
492 * has declared an interest.
494 if (dtrace_fasttrap_exit)
495 dtrace_fasttrap_exit(p);
499 * Notify interested parties of our demise.
501 KNOTE_LOCKED(&p->p_klist, NOTE_EXIT);
504 int reason = CLD_EXITED;
507 else if (WIFSIGNALED(rv))
509 SDT_PROBE(proc, kernel, , exit, reason, 0, 0, 0, 0);
513 * Just delete all entries in the p_klist. At this point we won't
514 * report any more events, and there are nasty race conditions that
515 * can beat us if we don't.
517 knlist_clear(&p->p_klist, 1);
520 * If this is a process with a descriptor, we may not need to deliver
521 * a signal to the parent. proctree_lock is held over
522 * procdesc_exit() to serialize concurrent calls to close() and
525 if (p->p_procdesc == NULL || procdesc_exit(p)) {
527 * Notify parent that we're gone. If parent has the
528 * PS_NOCLDWAIT flag set, or if the handler is set to SIG_IGN,
529 * notify process 1 instead (and hope it will handle this
532 PROC_LOCK(p->p_pptr);
533 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
534 if (p->p_pptr->p_sigacts->ps_flag &
535 (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
538 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
541 proc_reparent(p, initproc);
542 p->p_sigparent = SIGCHLD;
543 PROC_LOCK(p->p_pptr);
546 * Notify parent, so in case he was wait(2)ing or
547 * executing waitpid(2) with our pid, he will
552 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
554 if (p->p_pptr == initproc)
555 kern_psignal(p->p_pptr, SIGCHLD);
556 else if (p->p_sigparent != 0) {
557 if (p->p_sigparent == SIGCHLD)
559 else /* LINUX thread */
560 kern_psignal(p->p_pptr, p->p_sigparent);
563 PROC_LOCK(p->p_pptr);
564 sx_xunlock(&proctree_lock);
567 * The state PRS_ZOMBIE prevents other proesses from sending
568 * signal to the process, to avoid memory leak, we free memory
569 * for signal queue at the time when the state is set.
571 sigqueue_flush(&p->p_sigqueue);
572 sigqueue_flush(&td->td_sigqueue);
575 * We have to wait until after acquiring all locks before
576 * changing p_state. We need to avoid all possible context
577 * switches (including ones from blocking on a mutex) while
578 * marked as a zombie. We also have to set the zombie state
579 * before we release the parent process' proc lock to avoid
580 * a lost wakeup. So, we first call wakeup, then we grab the
581 * sched lock, update the state, and release the parent process'
585 cv_broadcast(&p->p_pwait);
586 sched_exit(p->p_pptr, td);
588 p->p_state = PRS_ZOMBIE;
589 PROC_UNLOCK(p->p_pptr);
592 * Hopefully no one will try to deliver a signal to the process this
595 knlist_destroy(&p->p_klist);
598 * Save our children's rusage information in our exit rusage.
600 ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
603 * Make sure the scheduler takes this thread out of its tables etc.
604 * This will also release this thread's reference to the ucred.
605 * Other thread parts to release include pcb bits and such.
611 #ifndef _SYS_SYSPROTO_H_
620 sys_abort2(struct thread *td, struct abort2_args *uap)
622 struct proc *p = td->td_proc;
628 * Do it right now so we can log either proper call of abort2(), or
629 * note, that invalid argument was passed. 512 is big enough to
630 * handle 16 arguments' descriptions with additional comments.
632 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN);
634 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ",
635 p->p_comm, p->p_pid, td->td_ucred->cr_uid);
637 * Since we can't return from abort2(), send SIGKILL in cases, where
638 * abort2() was called improperly
641 /* Prevent from DoSes from user-space. */
642 if (uap->nargs < 0 || uap->nargs > 16)
644 if (uap->nargs > 0) {
645 if (uap->args == NULL)
647 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *));
652 * Limit size of 'reason' string to 128. Will fit even when
653 * maximal number of arguments was chosen to be logged.
655 if (uap->why != NULL) {
656 error = sbuf_copyin(sb, uap->why, 128);
660 sbuf_printf(sb, "(null)");
662 if (uap->nargs > 0) {
663 sbuf_printf(sb, "(");
664 for (i = 0;i < uap->nargs; i++)
665 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]);
666 sbuf_printf(sb, ")");
669 * Final stage: arguments were proper, string has been
670 * successfully copied from userspace, and copying pointers
671 * from user-space succeed.
675 if (sig == SIGKILL) {
677 sbuf_printf(sb, " (Reason text inaccessible)");
681 log(LOG_INFO, "%s", sbuf_data(sb));
683 exit1(td, W_EXITCODE(0, sig));
690 * The dirty work is handled by kern_wait().
693 owait(struct thread *td, struct owait_args *uap __unused)
697 error = kern_wait(td, WAIT_ANY, &status, 0, NULL);
699 td->td_retval[1] = status;
702 #endif /* COMPAT_43 */
705 * The dirty work is handled by kern_wait().
708 sys_wait4(struct thread *td, struct wait4_args *uap)
710 struct rusage ru, *rup;
713 if (uap->rusage != NULL)
717 error = kern_wait(td, uap->pid, &status, uap->options, rup);
718 if (uap->status != NULL && error == 0)
719 error = copyout(&status, uap->status, sizeof(status));
720 if (uap->rusage != NULL && error == 0)
721 error = copyout(&ru, uap->rusage, sizeof(struct rusage));
726 sys_wait6(struct thread *td, struct wait6_args *uap)
728 struct __wrusage wru, *wrup;
734 idtype = uap->idtype;
737 if (uap->wrusage != NULL)
742 if (uap->info != NULL) {
744 bzero(sip, sizeof(*sip));
749 * We expect all callers of wait6() to know about WEXITED and
752 error = kern_wait6(td, idtype, id, &status, uap->options, wrup, sip);
754 if (uap->status != NULL && error == 0)
755 error = copyout(&status, uap->status, sizeof(status));
756 if (uap->wrusage != NULL && error == 0)
757 error = copyout(&wru, uap->wrusage, sizeof(wru));
758 if (uap->info != NULL && error == 0)
759 error = copyout(&si, uap->info, sizeof(si));
764 * Reap the remains of a zombie process and optionally return status and
765 * rusage. Asserts and will release both the proctree_lock and the process
766 * lock as part of its work.
769 proc_reap(struct thread *td, struct proc *p, int *status, int options)
773 sx_assert(&proctree_lock, SA_XLOCKED);
774 PROC_LOCK_ASSERT(p, MA_OWNED);
775 PROC_SLOCK_ASSERT(p, MA_OWNED);
776 KASSERT(p->p_state == PRS_ZOMBIE, ("proc_reap: !PRS_ZOMBIE"));
781 td->td_retval[0] = p->p_pid;
783 *status = p->p_xstat; /* convert to int */
784 if (options & WNOWAIT) {
786 * Only poll, returning the status. Caller does not wish to
787 * release the proc struct just yet.
790 sx_xunlock(&proctree_lock);
795 sigqueue_take(p->p_ksi);
800 * If we got the child via a ptrace 'attach', we need to give it back
803 if (p->p_oppid != 0) {
804 t = proc_realparent(p);
810 pksignal(t, SIGCHLD, p->p_ksi);
812 cv_broadcast(&p->p_pwait);
814 sx_xunlock(&proctree_lock);
819 * Remove other references to this process to ensure we have an
820 * exclusive reference.
822 sx_xlock(&allproc_lock);
823 LIST_REMOVE(p, p_list); /* off zombproc */
824 sx_xunlock(&allproc_lock);
825 LIST_REMOVE(p, p_sibling);
830 if (p->p_procdesc != NULL)
832 sx_xunlock(&proctree_lock);
835 * As a side effect of this lock, we know that all other writes to
836 * this proc are visible now, so no more locking is needed for p.
839 p->p_xstat = 0; /* XXX: why? */
842 ruadd(&q->p_stats->p_cru, &q->p_crux, &p->p_ru, &p->p_rux);
846 * Decrement the count of procs running with this uid.
848 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
851 * Destroy resource accounting information associated with the process.
855 racct_sub(p, RACCT_NPROC, 1);
861 * Free credentials, arguments, and sigacts.
865 pargs_drop(p->p_args);
867 sigacts_free(p->p_sigacts);
871 * Do any thread-system specific cleanups.
876 * Give vm and machine-dependent layer a chance to free anything that
877 * cpu_exit couldn't release while still running in process context.
883 KASSERT(FIRST_THREAD_IN_PROC(p),
884 ("proc_reap: no residual thread!"));
885 uma_zfree(proc_zone, p);
886 sx_xlock(&allproc_lock);
888 sx_xunlock(&allproc_lock);
892 proc_to_reap(struct thread *td, struct proc *p, idtype_t idtype, id_t id,
893 int *status, int options, struct __wrusage *wrusage, siginfo_t *siginfo)
898 sx_assert(&proctree_lock, SA_XLOCKED);
907 if (p->p_pid != (pid_t)id) {
913 if (p->p_pgid != (pid_t)id) {
919 if (p->p_session->s_sid != (pid_t)id) {
925 if (p->p_ucred->cr_uid != (uid_t)id) {
931 if (p->p_ucred->cr_gid != (gid_t)id) {
937 if (p->p_ucred->cr_prison->pr_id != (int)id) {
943 * It seems that the thread structures get zeroed out
944 * at process exit. This makes it impossible to
945 * support P_SETID, P_CID or P_CPUID.
952 if (p_canwait(td, p)) {
957 if (((options & WEXITED) == 0) && (p->p_state == PRS_ZOMBIE)) {
963 * This special case handles a kthread spawned by linux_clone
964 * (see linux_misc.c). The linux_wait4 and linux_waitpid
965 * functions need to be able to distinguish between waiting
966 * on a process and waiting on a thread. It is a thread if
967 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
968 * signifies we want to wait for threads and not processes.
970 if ((p->p_sigparent != SIGCHLD) ^
971 ((options & WLINUXCLONE) != 0)) {
978 if (siginfo != NULL) {
979 bzero(siginfo, sizeof(*siginfo));
980 siginfo->si_errno = 0;
983 * SUSv4 requires that the si_signo value is always
984 * SIGCHLD. Obey it despite the rfork(2) interface
985 * allows to request other signal for child exit
988 siginfo->si_signo = SIGCHLD;
991 * This is still a rough estimate. We will fix the
992 * cases TRAPPED, STOPPED, and CONTINUED later.
994 if (WCOREDUMP(p->p_xstat)) {
995 siginfo->si_code = CLD_DUMPED;
996 siginfo->si_status = WTERMSIG(p->p_xstat);
997 } else if (WIFSIGNALED(p->p_xstat)) {
998 siginfo->si_code = CLD_KILLED;
999 siginfo->si_status = WTERMSIG(p->p_xstat);
1001 siginfo->si_code = CLD_EXITED;
1002 siginfo->si_status = WEXITSTATUS(p->p_xstat);
1005 siginfo->si_pid = p->p_pid;
1006 siginfo->si_uid = p->p_ucred->cr_uid;
1009 * The si_addr field would be useful additional
1010 * detail, but apparently the PC value may be lost
1011 * when we reach this point. bzero() above sets
1012 * siginfo->si_addr to NULL.
1017 * There should be no reason to limit resources usage info to
1018 * exited processes only. A snapshot about any resources used
1019 * by a stopped process may be exactly what is needed.
1021 if (wrusage != NULL) {
1022 rup = &wrusage->wru_self;
1024 calcru(p, &rup->ru_utime, &rup->ru_stime);
1026 rup = &wrusage->wru_children;
1027 *rup = p->p_stats->p_cru;
1028 calccru(p, &rup->ru_utime, &rup->ru_stime);
1031 if (p->p_state == PRS_ZOMBIE) {
1032 proc_reap(td, p, status, options);
1041 kern_wait(struct thread *td, pid_t pid, int *status, int options,
1042 struct rusage *rusage)
1044 struct __wrusage wru, *wrup;
1050 * Translate the special pid values into the (idtype, pid)
1051 * pair for kern_wait6. The WAIT_MYPGRP case is handled by
1052 * kern_wait6() on its own.
1054 if (pid == WAIT_ANY) {
1057 } else if (pid < 0) {
1071 * For backward compatibility we implicitly add flags WEXITED
1072 * and WTRAPPED here.
1074 options |= WEXITED | WTRAPPED;
1075 ret = kern_wait6(td, idtype, id, status, options, wrup, NULL);
1077 *rusage = wru.wru_self;
1082 kern_wait6(struct thread *td, idtype_t idtype, id_t id, int *status,
1083 int options, struct __wrusage *wrusage, siginfo_t *siginfo)
1086 int error, nfound, ret;
1088 AUDIT_ARG_VALUE((int)idtype); /* XXX - This is likely wrong! */
1089 AUDIT_ARG_PID((pid_t)id); /* XXX - This may be wrong! */
1090 AUDIT_ARG_VALUE(options);
1094 if ((pid_t)id == WAIT_MYPGRP && (idtype == P_PID || idtype == P_PGID)) {
1096 id = (id_t)q->p_pgid;
1101 /* If we don't know the option, just return. */
1102 if ((options & ~(WUNTRACED | WNOHANG | WCONTINUED | WNOWAIT |
1103 WEXITED | WTRAPPED | WLINUXCLONE)) != 0)
1105 if ((options & (WEXITED | WUNTRACED | WCONTINUED | WTRAPPED)) == 0) {
1107 * We will be unable to find any matching processes,
1108 * because there are no known events to look for.
1109 * Prefer to return error instead of blocking
1116 if (q->p_flag & P_STATCHILD) {
1118 q->p_flag &= ~P_STATCHILD;
1122 sx_xlock(&proctree_lock);
1123 LIST_FOREACH(p, &q->p_children, p_sibling) {
1124 ret = proc_to_reap(td, p, idtype, id, status, options,
1136 if ((options & WTRAPPED) != 0 &&
1137 (p->p_flag & P_TRACED) != 0 &&
1138 (p->p_flag & (P_STOPPED_TRACE | P_STOPPED_SIG)) != 0 &&
1139 (p->p_suspcount == p->p_numthreads) &&
1140 ((p->p_flag & P_WAITED) == 0)) {
1142 if ((options & WNOWAIT) == 0)
1143 p->p_flag |= P_WAITED;
1144 sx_xunlock(&proctree_lock);
1145 td->td_retval[0] = p->p_pid;
1148 *status = W_STOPCODE(p->p_xstat);
1149 if (siginfo != NULL) {
1150 siginfo->si_status = p->p_xstat;
1151 siginfo->si_code = CLD_TRAPPED;
1153 if ((options & WNOWAIT) == 0) {
1155 sigqueue_take(p->p_ksi);
1162 if ((options & WUNTRACED) != 0 &&
1163 (p->p_flag & P_STOPPED_SIG) != 0 &&
1164 (p->p_suspcount == p->p_numthreads) &&
1165 ((p->p_flag & P_WAITED) == 0)) {
1167 if ((options & WNOWAIT) == 0)
1168 p->p_flag |= P_WAITED;
1169 sx_xunlock(&proctree_lock);
1170 td->td_retval[0] = p->p_pid;
1173 *status = W_STOPCODE(p->p_xstat);
1174 if (siginfo != NULL) {
1175 siginfo->si_status = p->p_xstat;
1176 siginfo->si_code = CLD_STOPPED;
1178 if ((options & WNOWAIT) == 0) {
1180 sigqueue_take(p->p_ksi);
1188 if ((options & WCONTINUED) != 0 &&
1189 (p->p_flag & P_CONTINUED) != 0) {
1190 sx_xunlock(&proctree_lock);
1191 td->td_retval[0] = p->p_pid;
1192 if ((options & WNOWAIT) == 0) {
1193 p->p_flag &= ~P_CONTINUED;
1195 sigqueue_take(p->p_ksi);
1202 if (siginfo != NULL) {
1203 siginfo->si_status = SIGCONT;
1204 siginfo->si_code = CLD_CONTINUED;
1212 * Look in the orphans list too, to allow the parent to
1213 * collect it's child exit status even if child is being
1216 * Debugger detaches from the parent upon successful
1217 * switch-over from parent to child. At this point due to
1218 * re-parenting the parent loses the child to debugger and a
1219 * wait4(2) call would report that it has no children to wait
1220 * for. By maintaining a list of orphans we allow the parent
1221 * to successfully wait until the child becomes a zombie.
1223 LIST_FOREACH(p, &q->p_orphans, p_orphan) {
1224 ret = proc_to_reap(td, p, idtype, id, status, options,
1234 sx_xunlock(&proctree_lock);
1237 if (options & WNOHANG) {
1238 sx_xunlock(&proctree_lock);
1239 td->td_retval[0] = 0;
1243 sx_xunlock(&proctree_lock);
1244 if (q->p_flag & P_STATCHILD) {
1245 q->p_flag &= ~P_STATCHILD;
1248 error = msleep(q, &q->p_mtx, PWAIT | PCATCH, "wait", 0);
1256 * Make process 'parent' the new parent of process 'child'.
1257 * Must be called with an exclusive hold of proctree lock.
1260 proc_reparent(struct proc *child, struct proc *parent)
1263 sx_assert(&proctree_lock, SX_XLOCKED);
1264 PROC_LOCK_ASSERT(child, MA_OWNED);
1265 if (child->p_pptr == parent)
1268 PROC_LOCK(child->p_pptr);
1269 sigqueue_take(child->p_ksi);
1270 PROC_UNLOCK(child->p_pptr);
1271 LIST_REMOVE(child, p_sibling);
1272 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
1274 clear_orphan(child);
1275 if (child->p_flag & P_TRACED) {
1276 if (LIST_EMPTY(&child->p_pptr->p_orphans)) {
1277 child->p_treeflag |= P_TREE_FIRST_ORPHAN;
1278 LIST_INSERT_HEAD(&child->p_pptr->p_orphans, child,
1281 LIST_INSERT_AFTER(LIST_FIRST(&child->p_pptr->p_orphans),
1284 child->p_treeflag |= P_TREE_ORPHANED;
1287 child->p_pptr = parent;