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 if (child->p_oppid == 0 ||
110 child->p_pptr->p_pid == child->p_oppid)
111 parent = child->p_pptr;
116 for (p = child; (p->p_treeflag & P_TREE_FIRST_ORPHAN) == 0;) {
117 /* Cannot use LIST_PREV(), since the list head is not known. */
118 p = __containerof(p->p_orphan.le_prev, struct proc,
120 KASSERT((p->p_treeflag & P_TREE_ORPHANED) != 0,
121 ("missing P_ORPHAN %p", p));
123 parent = __containerof(p->p_orphan.le_prev, struct proc,
129 clear_orphan(struct proc *p)
133 sx_assert(&proctree_lock, SA_XLOCKED);
134 if ((p->p_treeflag & P_TREE_ORPHANED) == 0)
136 if ((p->p_treeflag & P_TREE_FIRST_ORPHAN) != 0) {
137 p1 = LIST_NEXT(p, p_orphan);
139 p1->p_treeflag |= P_TREE_FIRST_ORPHAN;
140 p->p_treeflag &= ~P_TREE_FIRST_ORPHAN;
142 LIST_REMOVE(p, p_orphan);
143 p->p_treeflag &= ~P_TREE_ORPHANED;
147 * exit -- death of process.
150 sys_sys_exit(struct thread *td, struct sys_exit_args *uap)
153 exit1(td, W_EXITCODE(uap->rval, 0));
158 * Exit: deallocate address space and other resources, change proc state to
159 * zombie, and unlink proc from allproc and parent's lists. Save exit status
160 * and rusage for wait(). Check for child processes and orphan them.
163 exit1(struct thread *td, int rv)
165 struct proc *p, *nq, *q, *t;
168 struct vnode *ttyvp = NULL;
172 mtx_assert(&Giant, MA_NOTOWNED);
176 * XXX in case we're rebooting we just let init die in order to
177 * work around an unsolved stack overflow seen very late during
178 * shutdown on sparc64 when the gmirror worker process exists.
180 if (p == initproc && rebooting == 0) {
181 printf("init died (signal %d, exit %d)\n",
182 WTERMSIG(rv), WEXITSTATUS(rv));
183 panic("Going nowhere without my init!");
187 * MUST abort all other threads before proceeding past here.
190 while (p->p_flag & P_HADTHREADS) {
192 * First check if some other thread got here before us..
193 * if so, act apropriatly, (exit or suspend);
195 thread_suspend_check(0);
198 * Kill off the other threads. This requires
199 * some co-operation from other parts of the kernel
200 * so it may not be instantaneous. With this state set
201 * any thread entering the kernel from userspace will
202 * thread_exit() in trap(). Any thread attempting to
203 * sleep will return immediately with EINTR or EWOULDBLOCK
204 * which will hopefully force them to back out to userland
205 * freeing resources as they go. Any thread attempting
206 * to return to userland will thread_exit() from userret().
207 * thread_exit() will unsuspend us when the last of the
208 * other threads exits.
209 * If there is already a thread singler after resumption,
210 * calling thread_single will fail; in that case, we just
211 * re-check all suspension request, the thread should
212 * either be suspended there or exit.
214 if (! thread_single(SINGLE_EXIT))
218 * All other activity in this process is now stopped.
219 * Threading support has been turned off.
222 KASSERT(p->p_numthreads == 1,
223 ("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
224 racct_sub(p, RACCT_NTHR, 1);
226 * Wakeup anyone in procfs' PIOCWAIT. They should have a hold
227 * on our vmspace, so we should block below until they have
228 * released their reference to us. Note that if they have
229 * requested S_EXIT stops we will block here until they ack
232 _STOPEVENT(p, S_EXIT, rv);
235 * Ignore any pending request to stop due to a stop signal.
236 * Once P_WEXIT is set, future requests will be ignored as
239 p->p_flag &= ~P_STOPPED_SIG;
240 KASSERT(!P_SHOULDSTOP(p), ("exiting process is stopped"));
243 * Note that we are exiting and do another wakeup of anyone in
244 * PIOCWAIT in case they aren't listening for S_EXIT stops or
245 * decided to wait again after we told them we are exiting.
247 p->p_flag |= P_WEXIT;
251 * Wait for any processes that have a hold on our vmspace to
252 * release their reference.
254 while (p->p_lock > 0)
255 msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0);
257 p->p_xstat = rv; /* Let event handler change exit status */
259 /* Drain the limit callout while we don't have the proc locked */
260 callout_drain(&p->p_limco);
264 * The Sun BSM exit token contains two components: an exit status as
265 * passed to exit(), and a return value to indicate what sort of exit
266 * it was. The exit status is WEXITSTATUS(rv), but it's not clear
267 * what the return value is.
269 AUDIT_ARG_EXIT(WEXITSTATUS(rv), 0);
270 AUDIT_SYSCALL_EXIT(0, td);
273 /* Are we a task leader? */
274 if (p == p->p_leader) {
275 mtx_lock(&ppeers_lock);
279 kern_psignal(q, SIGKILL);
283 while (p->p_peers != NULL)
284 msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
285 mtx_unlock(&ppeers_lock);
289 * Check if any loadable modules need anything done at process exit.
290 * E.g. SYSV IPC stuff
291 * XXX what if one of these generates an error?
293 EVENTHANDLER_INVOKE(process_exit, p);
296 * If parent is waiting for us to exit or exec,
297 * P_PPWAIT is set; we will wakeup the parent below.
300 rv = p->p_xstat; /* Event handler could change exit status */
302 p->p_flag &= ~(P_TRACED | P_PPWAIT | P_PPTRACE);
305 * Stop the real interval timer. If the handler is currently
306 * executing, prevent it from rearming itself and let it finish.
308 if (timevalisset(&p->p_realtimer.it_value) &&
309 callout_stop(&p->p_itcallout) == 0) {
310 timevalclear(&p->p_realtimer.it_interval);
311 msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
312 KASSERT(!timevalisset(&p->p_realtimer.it_value),
313 ("realtime timer is still armed"));
318 * Reset any sigio structures pointing to us as a result of
319 * F_SETOWN with our pid.
321 funsetownlst(&p->p_sigiolst);
324 * If this process has an nlminfo data area (for lockd), release it
326 if (nlminfo_release_p != NULL && p->p_nlminfo != NULL)
327 (*nlminfo_release_p)(p);
330 * Close open files and release open-file table.
336 * If this thread tickled GEOM, we need to wait for the giggling to
337 * stop before we return to userland
339 if (td->td_pflags & TDP_GEOM)
343 * Remove ourself from our leader's peer list and wake our leader.
345 mtx_lock(&ppeers_lock);
346 if (p->p_leader->p_peers) {
348 while (q->p_peers != p)
350 q->p_peers = p->p_peers;
353 mtx_unlock(&ppeers_lock);
357 sx_xlock(&proctree_lock);
358 if (SESS_LEADER(p)) {
359 struct session *sp = p->p_session;
363 * s_ttyp is not zero'd; we use this to indicate that
364 * the session once had a controlling terminal. (for
365 * logging and informational purposes)
376 * Signal foreground pgrp and revoke access to
377 * controlling terminal if it has not been revoked
380 * Because the TTY may have been revoked in the mean
381 * time and could already have a new session associated
382 * with it, make sure we don't send a SIGHUP to a
383 * foreground process group that does not belong to this
389 if (tp->t_session == sp)
390 tty_signal_pgrp(tp, SIGHUP);
395 sx_xunlock(&proctree_lock);
396 if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
397 VOP_REVOKE(ttyvp, REVOKEALL);
398 VOP_UNLOCK(ttyvp, 0);
400 sx_xlock(&proctree_lock);
403 fixjobc(p, p->p_pgrp, 0);
404 sx_xunlock(&proctree_lock);
405 (void)acct_process(td);
407 /* Release the TTY now we've unlocked everything. */
414 * Release reference to text vnode
416 if ((vtmp = p->p_textvp) != NULL) {
418 locked = VFS_LOCK_GIANT(vtmp->v_mount);
420 VFS_UNLOCK_GIANT(locked);
424 * Release our limits structure.
435 * Remove proc from allproc queue and pidhash chain.
436 * Place onto zombproc. Unlink from parent's child list.
438 sx_xlock(&allproc_lock);
439 LIST_REMOVE(p, p_list);
440 LIST_INSERT_HEAD(&zombproc, p, p_list);
441 LIST_REMOVE(p, p_hash);
442 sx_xunlock(&allproc_lock);
445 * Call machine-dependent code to release any
446 * machine-dependent resources other than the address space.
447 * The address space is released by "vmspace_exitfree(p)" in
452 WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
455 * Reparent all children processes:
456 * - traced ones to the original parent (or init if we are that parent)
459 sx_xlock(&proctree_lock);
460 q = LIST_FIRST(&p->p_children);
461 if (q != NULL) /* only need this if any child is S_ZOMB */
463 for (; q != NULL; q = nq) {
464 nq = LIST_NEXT(q, p_sibling);
466 q->p_sigparent = SIGCHLD;
468 if (!(q->p_flag & P_TRACED)) {
469 proc_reparent(q, initproc);
472 * Traced processes are killed since their existence
473 * means someone is screwing up.
475 t = proc_realparent(q);
477 proc_reparent(q, initproc);
484 * Since q was found on our children list, the
485 * proc_reparent() call moved q to the orphan
486 * list due to present P_TRACED flag. Clear
487 * orphan link for q now while q is locked.
490 q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
491 FOREACH_THREAD_IN_PROC(q, tdt)
492 tdt->td_dbgflags &= ~TDB_SUSPEND;
493 kern_psignal(q, SIGKILL);
499 * Also get rid of our orphans.
501 while ((q = LIST_FIRST(&p->p_orphans)) != NULL) {
503 CTR2(KTR_PTRACE, "exit: pid %d, clearing orphan %d", p->p_pid,
509 /* Save exit status. */
513 /* Tell the prison that we are gone. */
514 prison_proc_free(p->p_ucred->cr_prison);
518 * Tell the DTrace fasttrap provider about the exit if it
519 * has declared an interest.
521 if (dtrace_fasttrap_exit)
522 dtrace_fasttrap_exit(p);
526 * Notify interested parties of our demise.
528 KNOTE_LOCKED(&p->p_klist, NOTE_EXIT);
531 int reason = CLD_EXITED;
534 else if (WIFSIGNALED(rv))
536 SDT_PROBE1(proc, kernel, , exit, reason);
540 * Just delete all entries in the p_klist. At this point we won't
541 * report any more events, and there are nasty race conditions that
542 * can beat us if we don't.
544 knlist_clear(&p->p_klist, 1);
547 * If this is a process with a descriptor, we may not need to deliver
548 * a signal to the parent. proctree_lock is held over
549 * procdesc_exit() to serialize concurrent calls to close() and
553 if (p->p_procdesc == NULL || procdesc_exit(p)) {
556 * Notify parent that we're gone. If parent has the
557 * PS_NOCLDWAIT flag set, or if the handler is set to SIG_IGN,
558 * notify process 1 instead (and hope it will handle this
561 PROC_LOCK(p->p_pptr);
562 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
563 if (p->p_pptr->p_sigacts->ps_flag &
564 (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
567 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
570 proc_reparent(p, initproc);
571 p->p_sigparent = SIGCHLD;
572 PROC_LOCK(p->p_pptr);
575 * Notify parent, so in case he was wait(2)ing or
576 * executing waitpid(2) with our pid, he will
581 mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
583 if (p->p_pptr == initproc)
584 kern_psignal(p->p_pptr, SIGCHLD);
585 else if (p->p_sigparent != 0) {
586 if (p->p_sigparent == SIGCHLD)
588 else /* LINUX thread */
589 kern_psignal(p->p_pptr, p->p_sigparent);
593 PROC_LOCK(p->p_pptr);
595 sx_xunlock(&proctree_lock);
598 * The state PRS_ZOMBIE prevents other proesses from sending
599 * signal to the process, to avoid memory leak, we free memory
600 * for signal queue at the time when the state is set.
602 sigqueue_flush(&p->p_sigqueue);
603 sigqueue_flush(&td->td_sigqueue);
606 * We have to wait until after acquiring all locks before
607 * changing p_state. We need to avoid all possible context
608 * switches (including ones from blocking on a mutex) while
609 * marked as a zombie. We also have to set the zombie state
610 * before we release the parent process' proc lock to avoid
611 * a lost wakeup. So, we first call wakeup, then we grab the
612 * sched lock, update the state, and release the parent process'
616 cv_broadcast(&p->p_pwait);
617 sched_exit(p->p_pptr, td);
619 p->p_state = PRS_ZOMBIE;
620 PROC_UNLOCK(p->p_pptr);
623 * Hopefully no one will try to deliver a signal to the process this
626 knlist_destroy(&p->p_klist);
629 * Save our children's rusage information in our exit rusage.
631 ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
634 * Make sure the scheduler takes this thread out of its tables etc.
635 * This will also release this thread's reference to the ucred.
636 * Other thread parts to release include pcb bits and such.
642 #ifndef _SYS_SYSPROTO_H_
651 sys_abort2(struct thread *td, struct abort2_args *uap)
653 struct proc *p = td->td_proc;
659 * Do it right now so we can log either proper call of abort2(), or
660 * note, that invalid argument was passed. 512 is big enough to
661 * handle 16 arguments' descriptions with additional comments.
663 sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN);
665 sbuf_printf(sb, "%s(pid %d uid %d) aborted: ",
666 p->p_comm, p->p_pid, td->td_ucred->cr_uid);
668 * Since we can't return from abort2(), send SIGKILL in cases, where
669 * abort2() was called improperly
672 /* Prevent from DoSes from user-space. */
673 if (uap->nargs < 0 || uap->nargs > 16)
675 if (uap->nargs > 0) {
676 if (uap->args == NULL)
678 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *));
683 * Limit size of 'reason' string to 128. Will fit even when
684 * maximal number of arguments was chosen to be logged.
686 if (uap->why != NULL) {
687 error = sbuf_copyin(sb, uap->why, 128);
691 sbuf_printf(sb, "(null)");
693 if (uap->nargs > 0) {
694 sbuf_printf(sb, "(");
695 for (i = 0;i < uap->nargs; i++)
696 sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]);
697 sbuf_printf(sb, ")");
700 * Final stage: arguments were proper, string has been
701 * successfully copied from userspace, and copying pointers
702 * from user-space succeed.
706 if (sig == SIGKILL) {
708 sbuf_printf(sb, " (Reason text inaccessible)");
712 log(LOG_INFO, "%s", sbuf_data(sb));
714 exit1(td, W_EXITCODE(0, sig));
721 * The dirty work is handled by kern_wait().
724 owait(struct thread *td, struct owait_args *uap __unused)
728 error = kern_wait(td, WAIT_ANY, &status, 0, NULL);
730 td->td_retval[1] = status;
733 #endif /* COMPAT_43 */
736 * The dirty work is handled by kern_wait().
739 sys_wait4(struct thread *td, struct wait4_args *uap)
741 struct rusage ru, *rup;
744 if (uap->rusage != NULL)
748 error = kern_wait(td, uap->pid, &status, uap->options, rup);
749 if (uap->status != NULL && error == 0)
750 error = copyout(&status, uap->status, sizeof(status));
751 if (uap->rusage != NULL && error == 0)
752 error = copyout(&ru, uap->rusage, sizeof(struct rusage));
757 sys_wait6(struct thread *td, struct wait6_args *uap)
759 struct __wrusage wru, *wrup;
765 idtype = uap->idtype;
768 if (uap->wrusage != NULL)
773 if (uap->info != NULL) {
775 bzero(sip, sizeof(*sip));
780 * We expect all callers of wait6() to know about WEXITED and
783 error = kern_wait6(td, idtype, id, &status, uap->options, wrup, sip);
785 if (uap->status != NULL && error == 0)
786 error = copyout(&status, uap->status, sizeof(status));
787 if (uap->wrusage != NULL && error == 0)
788 error = copyout(&wru, uap->wrusage, sizeof(wru));
789 if (uap->info != NULL && error == 0)
790 error = copyout(&si, uap->info, sizeof(si));
795 * Reap the remains of a zombie process and optionally return status and
796 * rusage. Asserts and will release both the proctree_lock and the process
797 * lock as part of its work.
800 proc_reap(struct thread *td, struct proc *p, int *status, int options)
804 sx_assert(&proctree_lock, SA_XLOCKED);
805 PROC_LOCK_ASSERT(p, MA_OWNED);
806 PROC_SLOCK_ASSERT(p, MA_OWNED);
807 KASSERT(p->p_state == PRS_ZOMBIE, ("proc_reap: !PRS_ZOMBIE"));
812 td->td_retval[0] = p->p_pid;
814 *status = p->p_xstat; /* convert to int */
815 if (options & WNOWAIT) {
817 * Only poll, returning the status. Caller does not wish to
818 * release the proc struct just yet.
821 sx_xunlock(&proctree_lock);
826 sigqueue_take(p->p_ksi);
830 * If we got the child via a ptrace 'attach', we need to give it back
833 if (p->p_oppid != 0 && p->p_oppid != p->p_pptr->p_pid) {
835 t = proc_realparent(p);
839 "wait: traced child %d moved back to parent %d", p->p_pid,
844 pksignal(t, SIGCHLD, p->p_ksi);
846 cv_broadcast(&p->p_pwait);
848 sx_xunlock(&proctree_lock);
855 * Remove other references to this process to ensure we have an
856 * exclusive reference.
858 sx_xlock(&allproc_lock);
859 LIST_REMOVE(p, p_list); /* off zombproc */
860 sx_xunlock(&allproc_lock);
861 LIST_REMOVE(p, p_sibling);
867 if (p->p_procdesc != NULL)
870 sx_xunlock(&proctree_lock);
873 * As a side effect of this lock, we know that all other writes to
874 * this proc are visible now, so no more locking is needed for p.
877 p->p_xstat = 0; /* XXX: why? */
880 ruadd(&q->p_stats->p_cru, &q->p_crux, &p->p_ru, &p->p_rux);
884 * Decrement the count of procs running with this uid.
886 (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
889 * Destroy resource accounting information associated with the process.
893 racct_sub(p, RACCT_NPROC, 1);
899 * Free credentials, arguments, and sigacts.
903 pargs_drop(p->p_args);
905 sigacts_free(p->p_sigacts);
909 * Do any thread-system specific cleanups.
914 * Give vm and machine-dependent layer a chance to free anything that
915 * cpu_exit couldn't release while still running in process context.
921 KASSERT(FIRST_THREAD_IN_PROC(p),
922 ("proc_reap: no residual thread!"));
923 uma_zfree(proc_zone, p);
924 sx_xlock(&allproc_lock);
926 sx_xunlock(&allproc_lock);
930 proc_to_reap(struct thread *td, struct proc *p, idtype_t idtype, id_t id,
931 int *status, int options, struct __wrusage *wrusage, siginfo_t *siginfo,
937 sx_assert(&proctree_lock, SA_XLOCKED);
946 if (p->p_pid != (pid_t)id) {
952 if (p->p_pgid != (pid_t)id) {
958 if (p->p_session->s_sid != (pid_t)id) {
964 if (p->p_ucred->cr_uid != (uid_t)id) {
970 if (p->p_ucred->cr_gid != (gid_t)id) {
976 if (p->p_ucred->cr_prison == NULL ||
977 (p->p_ucred->cr_prison->pr_id != (int)id)) {
983 * It seems that the thread structures get zeroed out
984 * at process exit. This makes it impossible to
985 * support P_SETID, P_CID or P_CPUID.
992 if (p_canwait(td, p)) {
997 if (((options & WEXITED) == 0) && (p->p_state == PRS_ZOMBIE)) {
1003 * This special case handles a kthread spawned by linux_clone
1004 * (see linux_misc.c). The linux_wait4 and linux_waitpid
1005 * functions need to be able to distinguish between waiting
1006 * on a process and waiting on a thread. It is a thread if
1007 * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
1008 * signifies we want to wait for threads and not processes.
1010 if ((p->p_sigparent != SIGCHLD) ^
1011 ((options & WLINUXCLONE) != 0)) {
1018 if (siginfo != NULL) {
1019 bzero(siginfo, sizeof(*siginfo));
1020 siginfo->si_errno = 0;
1023 * SUSv4 requires that the si_signo value is always
1024 * SIGCHLD. Obey it despite the rfork(2) interface
1025 * allows to request other signal for child exit
1028 siginfo->si_signo = SIGCHLD;
1031 * This is still a rough estimate. We will fix the
1032 * cases TRAPPED, STOPPED, and CONTINUED later.
1034 if (WCOREDUMP(p->p_xstat)) {
1035 siginfo->si_code = CLD_DUMPED;
1036 siginfo->si_status = WTERMSIG(p->p_xstat);
1037 } else if (WIFSIGNALED(p->p_xstat)) {
1038 siginfo->si_code = CLD_KILLED;
1039 siginfo->si_status = WTERMSIG(p->p_xstat);
1041 siginfo->si_code = CLD_EXITED;
1042 siginfo->si_status = WEXITSTATUS(p->p_xstat);
1045 siginfo->si_pid = p->p_pid;
1046 siginfo->si_uid = p->p_ucred->cr_uid;
1049 * The si_addr field would be useful additional
1050 * detail, but apparently the PC value may be lost
1051 * when we reach this point. bzero() above sets
1052 * siginfo->si_addr to NULL.
1057 * There should be no reason to limit resources usage info to
1058 * exited processes only. A snapshot about any resources used
1059 * by a stopped process may be exactly what is needed.
1061 if (wrusage != NULL) {
1062 rup = &wrusage->wru_self;
1064 calcru(p, &rup->ru_utime, &rup->ru_stime);
1066 rup = &wrusage->wru_children;
1067 *rup = p->p_stats->p_cru;
1068 calccru(p, &rup->ru_utime, &rup->ru_stime);
1071 if (p->p_state == PRS_ZOMBIE && !check_only) {
1072 proc_reap(td, p, status, options);
1081 kern_wait(struct thread *td, pid_t pid, int *status, int options,
1082 struct rusage *rusage)
1084 struct __wrusage wru, *wrup;
1090 * Translate the special pid values into the (idtype, pid)
1091 * pair for kern_wait6. The WAIT_MYPGRP case is handled by
1092 * kern_wait6() on its own.
1094 if (pid == WAIT_ANY) {
1097 } else if (pid < 0) {
1111 * For backward compatibility we implicitly add flags WEXITED
1112 * and WTRAPPED here.
1114 options |= WEXITED | WTRAPPED;
1115 ret = kern_wait6(td, idtype, id, status, options, wrup, NULL);
1117 *rusage = wru.wru_self;
1122 kern_wait6(struct thread *td, idtype_t idtype, id_t id, int *status,
1123 int options, struct __wrusage *wrusage, siginfo_t *siginfo)
1126 int error, nfound, ret;
1128 AUDIT_ARG_VALUE((int)idtype); /* XXX - This is likely wrong! */
1129 AUDIT_ARG_PID((pid_t)id); /* XXX - This may be wrong! */
1130 AUDIT_ARG_VALUE(options);
1134 if ((pid_t)id == WAIT_MYPGRP && (idtype == P_PID || idtype == P_PGID)) {
1136 id = (id_t)q->p_pgid;
1141 /* If we don't know the option, just return. */
1142 if ((options & ~(WUNTRACED | WNOHANG | WCONTINUED | WNOWAIT |
1143 WEXITED | WTRAPPED | WLINUXCLONE)) != 0)
1145 if ((options & (WEXITED | WUNTRACED | WCONTINUED | WTRAPPED)) == 0) {
1147 * We will be unable to find any matching processes,
1148 * because there are no known events to look for.
1149 * Prefer to return error instead of blocking
1156 if (q->p_flag & P_STATCHILD) {
1158 q->p_flag &= ~P_STATCHILD;
1162 sx_xlock(&proctree_lock);
1163 LIST_FOREACH(p, &q->p_children, p_sibling) {
1164 ret = proc_to_reap(td, p, idtype, id, status, options,
1165 wrusage, siginfo, 0);
1176 if ((options & WTRAPPED) != 0 &&
1177 (p->p_flag & P_TRACED) != 0 &&
1178 (p->p_flag & (P_STOPPED_TRACE | 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_TRAPPED;
1193 if ((options & WNOWAIT) == 0) {
1195 sigqueue_take(p->p_ksi);
1200 "wait: returning trapped pid %d status %#x (xstat %d) xthread %d",
1201 p->p_pid, W_STOPCODE(p->p_xstat), p->p_xstat,
1202 p->p_xthread != NULL ? p->p_xthread->td_tid : -1);
1206 if ((options & WUNTRACED) != 0 &&
1207 (p->p_flag & P_STOPPED_SIG) != 0 &&
1208 (p->p_suspcount == p->p_numthreads) &&
1209 ((p->p_flag & P_WAITED) == 0)) {
1211 if ((options & WNOWAIT) == 0)
1212 p->p_flag |= P_WAITED;
1213 sx_xunlock(&proctree_lock);
1214 td->td_retval[0] = p->p_pid;
1217 *status = W_STOPCODE(p->p_xstat);
1218 if (siginfo != NULL) {
1219 siginfo->si_status = p->p_xstat;
1220 siginfo->si_code = CLD_STOPPED;
1222 if ((options & WNOWAIT) == 0) {
1224 sigqueue_take(p->p_ksi);
1232 if ((options & WCONTINUED) != 0 &&
1233 (p->p_flag & P_CONTINUED) != 0) {
1234 sx_xunlock(&proctree_lock);
1235 td->td_retval[0] = p->p_pid;
1236 if ((options & WNOWAIT) == 0) {
1237 p->p_flag &= ~P_CONTINUED;
1239 sigqueue_take(p->p_ksi);
1246 if (siginfo != NULL) {
1247 siginfo->si_status = SIGCONT;
1248 siginfo->si_code = CLD_CONTINUED;
1256 * Look in the orphans list too, to allow the parent to
1257 * collect it's child exit status even if child is being
1260 * Debugger detaches from the parent upon successful
1261 * switch-over from parent to child. At this point due to
1262 * re-parenting the parent loses the child to debugger and a
1263 * wait4(2) call would report that it has no children to wait
1264 * for. By maintaining a list of orphans we allow the parent
1265 * to successfully wait until the child becomes a zombie.
1268 LIST_FOREACH(p, &q->p_orphans, p_orphan) {
1269 ret = proc_to_reap(td, p, idtype, id, NULL, options,
1272 KASSERT(ret != -1, ("reaped an orphan (pid %d)",
1273 (int)td->td_retval[0]));
1280 sx_xunlock(&proctree_lock);
1283 if (options & WNOHANG) {
1284 sx_xunlock(&proctree_lock);
1285 td->td_retval[0] = 0;
1289 sx_xunlock(&proctree_lock);
1290 if (q->p_flag & P_STATCHILD) {
1291 q->p_flag &= ~P_STATCHILD;
1294 error = msleep(q, &q->p_mtx, PWAIT | PCATCH, "wait", 0);
1302 * Make process 'parent' the new parent of process 'child'.
1303 * Must be called with an exclusive hold of proctree lock.
1306 proc_reparent(struct proc *child, struct proc *parent)
1309 sx_assert(&proctree_lock, SX_XLOCKED);
1310 PROC_LOCK_ASSERT(child, MA_OWNED);
1311 if (child->p_pptr == parent)
1314 PROC_LOCK(child->p_pptr);
1315 sigqueue_take(child->p_ksi);
1316 PROC_UNLOCK(child->p_pptr);
1317 LIST_REMOVE(child, p_sibling);
1318 LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
1320 clear_orphan(child);
1321 if (child->p_flag & P_TRACED) {
1322 if (LIST_EMPTY(&child->p_pptr->p_orphans)) {
1323 child->p_treeflag |= P_TREE_FIRST_ORPHAN;
1324 LIST_INSERT_HEAD(&child->p_pptr->p_orphans, child,
1327 LIST_INSERT_AFTER(LIST_FIRST(&child->p_pptr->p_orphans),
1330 child->p_treeflag |= P_TREE_ORPHANED;
1333 child->p_pptr = parent;